CN1688244A - Method, system, and computer program product for the processing of self-monitoring blood glucose(smbg)data to enhance diabetic self-management - Google Patents
Method, system, and computer program product for the processing of self-monitoring blood glucose(smbg)data to enhance diabetic self-management Download PDFInfo
- Publication number
- CN1688244A CN1688244A CNA038240092A CN03824009A CN1688244A CN 1688244 A CN1688244 A CN 1688244A CN A038240092 A CNA038240092 A CN A038240092A CN 03824009 A CN03824009 A CN 03824009A CN 1688244 A CN1688244 A CN 1688244A
- Authority
- CN
- China
- Prior art keywords
- lbgi
- hba
- data
- risk
- predetermined duration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 152
- 239000008280 blood Substances 0.000 title claims abstract description 64
- 210000004369 blood Anatomy 0.000 title claims abstract description 64
- 238000004590 computer program Methods 0.000 title claims abstract description 41
- 206010012601 diabetes mellitus Diseases 0.000 title abstract description 63
- 239000008103 glucose Substances 0.000 title abstract description 58
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 title abstract description 57
- 238000012544 monitoring process Methods 0.000 title abstract description 32
- 238000012545 processing Methods 0.000 title description 11
- 230000002218 hypoglycaemic effect Effects 0.000 claims abstract description 175
- 208000013016 Hypoglycemia Diseases 0.000 claims abstract description 161
- 230000007774 longterm Effects 0.000 claims abstract description 31
- 238000012360 testing method Methods 0.000 claims description 113
- 101100121625 Caenorhabditis elegans rhi-1 gene Proteins 0.000 claims description 75
- 238000004364 calculation method Methods 0.000 claims description 39
- 101000655352 Homo sapiens Telomerase reverse transcriptase Proteins 0.000 claims description 36
- 102100032938 Telomerase reverse transcriptase Human genes 0.000 claims description 36
- 230000002641 glycemic effect Effects 0.000 claims description 35
- 238000007781 pre-processing Methods 0.000 claims description 17
- 230000007246 mechanism Effects 0.000 claims description 6
- 201000001421 hyperglycemia Diseases 0.000 abstract description 9
- 201000010099 disease Diseases 0.000 abstract description 5
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 abstract description 5
- 238000004519 manufacturing process Methods 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 198
- NOESYZHRGYRDHS-UHFFFAOYSA-N insulin Chemical compound N1C(=O)C(NC(=O)C(CCC(N)=O)NC(=O)C(CCC(O)=O)NC(=O)C(C(C)C)NC(=O)C(NC(=O)CN)C(C)CC)CSSCC(C(NC(CO)C(=O)NC(CC(C)C)C(=O)NC(CC=2C=CC(O)=CC=2)C(=O)NC(CCC(N)=O)C(=O)NC(CC(C)C)C(=O)NC(CCC(O)=O)C(=O)NC(CC(N)=O)C(=O)NC(CC=2C=CC(O)=CC=2)C(=O)NC(CSSCC(NC(=O)C(C(C)C)NC(=O)C(CC(C)C)NC(=O)C(CC=2C=CC(O)=CC=2)NC(=O)C(CC(C)C)NC(=O)C(C)NC(=O)C(CCC(O)=O)NC(=O)C(C(C)C)NC(=O)C(CC(C)C)NC(=O)C(CC=2NC=NC=2)NC(=O)C(CO)NC(=O)CNC2=O)C(=O)NCC(=O)NC(CCC(O)=O)C(=O)NC(CCCNC(N)=N)C(=O)NCC(=O)NC(CC=3C=CC=CC=3)C(=O)NC(CC=3C=CC=CC=3)C(=O)NC(CC=3C=CC(O)=CC=3)C(=O)NC(C(C)O)C(=O)N3C(CCC3)C(=O)NC(CCCCN)C(=O)NC(C)C(O)=O)C(=O)NC(CC(N)=O)C(O)=O)=O)NC(=O)C(C(C)CC)NC(=O)C(CO)NC(=O)C(C(C)O)NC(=O)C1CSSCC2NC(=O)C(CC(C)C)NC(=O)C(NC(=O)C(CCC(N)=O)NC(=O)C(CC(N)=O)NC(=O)C(NC(=O)C(N)CC=1C=CC=CC=1)C(C)C)CC1=CN=CN1 NOESYZHRGYRDHS-UHFFFAOYSA-N 0.000 description 42
- 238000012549 training Methods 0.000 description 39
- 238000005259 measurement Methods 0.000 description 32
- 230000006870 function Effects 0.000 description 25
- 102000004877 Insulin Human genes 0.000 description 22
- 108090001061 Insulin Proteins 0.000 description 22
- 238000009826 distribution Methods 0.000 description 22
- 238000004458 analytical method Methods 0.000 description 21
- 229940125396 insulin Drugs 0.000 description 21
- 238000005457 optimization Methods 0.000 description 21
- 238000003556 assay Methods 0.000 description 20
- 238000004891 communication Methods 0.000 description 20
- 230000008859 change Effects 0.000 description 19
- 238000003860 storage Methods 0.000 description 17
- 230000000694 effects Effects 0.000 description 16
- 239000000243 solution Substances 0.000 description 16
- 230000015654 memory Effects 0.000 description 15
- 206010067584 Type 1 diabetes mellitus Diseases 0.000 description 13
- 238000013480 data collection Methods 0.000 description 13
- 230000002354 daily effect Effects 0.000 description 11
- 238000000540 analysis of variance Methods 0.000 description 10
- 238000011160 research Methods 0.000 description 10
- 230000008901 benefit Effects 0.000 description 9
- 238000011156 evaluation Methods 0.000 description 9
- 230000007717 exclusion Effects 0.000 description 9
- 238000007726 management method Methods 0.000 description 8
- 238000012806 monitoring device Methods 0.000 description 7
- 230000000422 nocturnal effect Effects 0.000 description 7
- 208000001072 type 2 diabetes mellitus Diseases 0.000 description 7
- 102000017011 Glycated Hemoglobin A Human genes 0.000 description 6
- 230000000875 corresponding effect Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 239000003550 marker Substances 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 206010010071 Coma Diseases 0.000 description 5
- 108010014663 Glycated Hemoglobin A Proteins 0.000 description 5
- 108010041872 Islet Amyloid Polypeptide Proteins 0.000 description 5
- 102000036770 Islet Amyloid Polypeptide Human genes 0.000 description 5
- 238000012937 correction Methods 0.000 description 5
- 238000007405 data analysis Methods 0.000 description 5
- 229940079593 drug Drugs 0.000 description 5
- 239000003814 drug Substances 0.000 description 5
- 238000012417 linear regression Methods 0.000 description 5
- 230000009466 transformation Effects 0.000 description 5
- 208000003443 Unconsciousness Diseases 0.000 description 4
- 239000012491 analyte Substances 0.000 description 4
- 235000021152 breakfast Nutrition 0.000 description 4
- 230000002596 correlated effect Effects 0.000 description 4
- 230000007423 decrease Effects 0.000 description 4
- 230000001934 delay Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 230000003203 everyday effect Effects 0.000 description 4
- 238000009472 formulation Methods 0.000 description 4
- 238000011545 laboratory measurement Methods 0.000 description 4
- 238000012886 linear function Methods 0.000 description 4
- 239000000203 mixture Substances 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 238000000926 separation method Methods 0.000 description 4
- 208000024891 symptom Diseases 0.000 description 4
- 238000010200 validation analysis Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 3
- 230000001517 counterregulatory effect Effects 0.000 description 3
- 230000003247 decreasing effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000003345 hyperglycaemic effect Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000000291 postprandial effect Effects 0.000 description 3
- 230000001105 regulatory effect Effects 0.000 description 3
- 210000004243 sweat Anatomy 0.000 description 3
- 238000002560 therapeutic procedure Methods 0.000 description 3
- GLDQAMYCGOIJDV-UHFFFAOYSA-N 2,3-dihydroxybenzoic acid Chemical compound OC(=O)C1=CC=CC(O)=C1O GLDQAMYCGOIJDV-UHFFFAOYSA-N 0.000 description 2
- 230000001154 acute effect Effects 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 230000001413 cellular effect Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 238000005183 dynamical system Methods 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- XZWYZXLIPXDOLR-UHFFFAOYSA-N metformin Chemical compound CN(C)C(=N)NC(N)=N XZWYZXLIPXDOLR-UHFFFAOYSA-N 0.000 description 2
- 229960003105 metformin Drugs 0.000 description 2
- 230000000926 neurological effect Effects 0.000 description 2
- 108010029667 pramlintide Proteins 0.000 description 2
- 229960003611 pramlintide Drugs 0.000 description 2
- NRKVKVQDUCJPIZ-MKAGXXMWSA-N pramlintide acetate Chemical compound C([C@@H](C(=O)NCC(=O)N1CCC[C@H]1C(=O)N[C@@H]([C@@H](C)CC)C(=O)N[C@@H](CC(C)C)C(=O)N1[C@@H](CCC1)C(=O)N1[C@@H](CCC1)C(=O)N[C@@H]([C@@H](C)O)C(=O)N[C@@H](CC(N)=O)C(=O)N[C@@H](C(C)C)C(=O)NCC(=O)N[C@@H](CO)C(=O)N[C@@H](CC(N)=O)C(=O)N[C@@H]([C@@H](C)O)C(=O)N[C@@H](CC=1C=CC(O)=CC=1)C(N)=O)NC(=O)[C@H](CC(N)=O)NC(=O)[C@H](CC(N)=O)NC(=O)[C@H](CO)NC(=O)[C@H](CO)NC(=O)[C@H](CC=1NC=NC=1)NC(=O)[C@@H](NC(=O)[C@H](CC(C)C)NC(=O)[C@H](CC=1C=CC=CC=1)NC(=O)[C@H](CC(N)=O)NC(=O)[C@H](C)NC(=O)[C@H](CC(C)C)NC(=O)[C@H](CCCNC(N)=N)NC(=O)[C@H](CCC(N)=O)NC(=O)[C@@H](NC(=O)[C@H](C)NC(=O)[C@H](CS)NC(=O)[C@@H](NC(=O)[C@H](C)NC(=O)[C@@H](NC(=O)[C@H](CC(N)=O)NC(=O)[C@H](CS)NC(=O)[C@@H](N)CCCCN)[C@@H](C)O)[C@@H](C)O)[C@@H](C)O)C(C)C)C1=CC=CC=C1 NRKVKVQDUCJPIZ-MKAGXXMWSA-N 0.000 description 2
- 238000012552 review Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 208000035408 type 1 diabetes mellitus 1 Diseases 0.000 description 2
- 230000003442 weekly effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 208000028698 Cognitive impairment Diseases 0.000 description 1
- 208000002249 Diabetes Complications Diseases 0.000 description 1
- 101001051478 Equus caballus Beta-lactoglobulin-1 Proteins 0.000 description 1
- 102100030373 HSPB1-associated protein 1 Human genes 0.000 description 1
- 102100027685 Hemoglobin subunit alpha Human genes 0.000 description 1
- 102100021467 Histone acetyltransferase type B catalytic subunit Human genes 0.000 description 1
- 101000843045 Homo sapiens HSPB1-associated protein 1 Proteins 0.000 description 1
- 101001009007 Homo sapiens Hemoglobin subunit alpha Proteins 0.000 description 1
- 101000898976 Homo sapiens Histone acetyltransferase type B catalytic subunit Proteins 0.000 description 1
- 101000851696 Homo sapiens Steroid hormone receptor ERR2 Proteins 0.000 description 1
- 238000004566 IR spectroscopy Methods 0.000 description 1
- 206010022489 Insulin Resistance Diseases 0.000 description 1
- 102100036831 Steroid hormone receptor ERR2 Human genes 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000009098 adjuvant therapy Methods 0.000 description 1
- 230000002547 anomalous effect Effects 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003833 bile salt Substances 0.000 description 1
- 229940093761 bile salts Drugs 0.000 description 1
- 150000001720 carbohydrates Chemical class 0.000 description 1
- 235000014633 carbohydrates Nutrition 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 230000001713 cholinergic effect Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 208000010877 cognitive disease Diseases 0.000 description 1
- 238000013479 data entry Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013501 data transformation Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 239000007933 dermal patch Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000003292 diminished effect Effects 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000003826 endocrine responses Effects 0.000 description 1
- 210000000750 endocrine system Anatomy 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000003623 enhancer Substances 0.000 description 1
- 210000003722 extracellular fluid Anatomy 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 108091005995 glycated hemoglobin Proteins 0.000 description 1
- 231100001261 hazardous Toxicity 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 108010013766 hemoglobin A(0) Proteins 0.000 description 1
- 230000003054 hormonal effect Effects 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 239000004026 insulin derivative Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000001307 laser spectroscopy Methods 0.000 description 1
- 235000012054 meals Nutrition 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 230000002503 metabolic effect Effects 0.000 description 1
- 230000004060 metabolic process Effects 0.000 description 1
- 230000001019 normoglycemic effect Effects 0.000 description 1
- 229940126701 oral medication Drugs 0.000 description 1
- 206010033675 panniculitis Diseases 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 230000037081 physical activity Effects 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000007391 self-medication Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000012109 statistical procedure Methods 0.000 description 1
- 210000004304 subcutaneous tissue Anatomy 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000000153 supplemental effect Effects 0.000 description 1
- 230000009469 supplementation Effects 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 230000002459 sustained effect Effects 0.000 description 1
- 210000000106 sweat gland Anatomy 0.000 description 1
- 239000013076 target substance Substances 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000002023 wood Substances 0.000 description 1
Images
Landscapes
- Investigating Or Analysing Biological Materials (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
一种方法、系统和计算机程序产品涉及根据由自我监测血糖设备收集的血糖数据保持对糖尿病的最佳控制,并致力于预测对高血糖症的长期暴露,以及糖尿病患者罹患严重或者中度低血糖症的长期和短期危险。本方法、系统和计算机程序产品适合于通过引入一种能够同时预测HbA1c和低血糖症高危期的智能数据判读组件直接提高现有家用血糖监测设备的性能,并通过相同的特征提高将来生产的连续监测设备的性能。利用这些预测,糖尿病患者能够采取措施防止与高血糖症和低血糖症有关的负面后果。
A method, system and computer program product related to maintaining optimal control of diabetes based on blood glucose data collected by a self-monitoring blood glucose device, and aimed at predicting long-term exposure to hyperglycemia, and severe or moderate hypoglycemia in diabetic patients long-term and short-term risks of the disease. The method, system and computer program product are suitable for directly improving the performance of existing household blood glucose monitoring equipment by introducing an intelligent data interpretation component capable of simultaneously predicting HbA 1c and high-risk periods of hypoglycemia, and improving future production of blood glucose monitoring equipment through the same features. Continuously monitor device performance. Using these predictions, diabetics are able to take steps to prevent the negative outcomes associated with hyperglycemia and hypoglycemia.
Description
相关专利申请Related Patent Applications
本发明要求获得如下专利申请的优先权,即美国临时专利申请系列No.60/402,976,其于2002年8月13日提出申请,标题为“用于处理自我监测血糖(SMBG)数据从而提高糖尿病患者自我管理的方法、系统和计算机程序产品”(Method,System,and Computer ProgramProduct for Processing of Self-monitoring Blood Glucose(SMBG)Data to Enhance Diabetic Self-managemnet),和No.60/478,377,其于2003年6月13日提出申请,标题为“用于处理自我监测血糖(SMBG)数据从而提高糖尿病患者自我管理的方法、系统和计算机程序产品”(Method,System,and Computer Program Product for Processingof Self-monitoring Blood Glucose(SMBG)Data to Enhance DiabeticSelf-managemnet),本文引用上述公开的全部内容作为参考。This application claims priority to U.S. Provisional Patent Application Serial No. 60/402,976, filed August 13, 2002, and entitled "Use for Processing Self-Monitoring Blood Glucose (SMBG) Data to Improve Diabetes Patient self-management method, system and computer program product" (Method, System, and Computer Program Product for Processing of Self-monitoring Blood Glucose (SMBG) Data to Enhance Diabetic Self-managemnet), and No.60/478,377, published in 2003 The application was filed on June 13, 2009, entitled "Method, System, and Computer Program Product for Processing of Self-monitoring Blood Glucose (SMBG) Data to Improve Self-management of Diabetes Patients" (Method, System, and Computer Program Product for Processing of Self-monitoring Blood Glucose (SMBG) Data to Enhance Diabetic Self-managemnet), this article quotes the entire content of the above disclosure as a reference.
本发明涉及国际专利申请No.PCT/US01/09884,其于2001年3月29日提出申请(专利申请Nos.WO 01/72208 A2,WO 01/72208 A3),标题为“由自我监测数据评估糖尿病血糖控制的方法、系统和计算机程序产品”(Method,System,and Computer Program Product for theEvaluation of Glycemic Control in Diabetes from Self-monitoringData),和美国专利申请系列No.:10/240,228,其于2002年9月26日提出申请,标题为“由自我监测数据评估糖尿病血糖控制的方法、系统和计算机程序产品”(Method,System,and Computer ProgramProduct for the Evaluation of Glycemic Control in Diabetes fromSelf-monitoring Data),本文引用其全部的公开内容作为参考。This invention relates to International Patent Application No. PCT/US01/09884, filed on March 29, 2001 (Patent Application Nos. WO 01/72208 A2, WO 01/72208 A3), entitled "Assessment by self-monitoring data Method, System, and Computer Program Product for the Evaluation of Glycemic Control in Diabetes from Self-monitoring Data, and US Patent Application Serial No.: 10/240,228, filed in 2002 An application was filed on September 26, entitled "Method, System, and Computer Program Product for the Evaluation of Glycemic Control in Diabetes from Self-monitoring Data" (Method, System, and Computer Program Product for the Evaluation of Glycemic Control in Diabetes from Self-monitoring Data), this article The entire disclosure content thereof is incorporated by reference.
技术领域technical field
本系统总体上涉及糖尿病个体的血糖控制,更特别地,涉及一种基于计算机的系统和方法,用于评估预测糖基化血红蛋白(HbA1c和HbA1)和发生低血糖症的危险。The present system relates generally to glycemic control in individuals with diabetes and, more particularly, to a computer-based system and method for assessing the risk of predicting glycated hemoglobin (HbA 1c and HbA 1 ) and developing hypoglycemia.
背景技术Background technique
大量的研究反复证实,防止糖尿病长期并发症的最有效方法是采取胰岛素强化治疗将血糖(BG)水平严格控制在正常范围之内,这些研究包括糖尿病控制和并发症实验(DCCT)(见DCCT研究会:糖尿病强化治疗对胰岛素依赖型糖尿病长期并发症的发生和发展的影响(The Effect Of Intensive Treatment Of Diabetes On The DevelopmentAnd Progresion Of Long-Term Complications Of Insulin-DependentDiabetes Mellitus).New England Journal ofMedicine,329:978-986,1993)),斯德哥尔摩糖尿病干预研究(见Reichard P,Phil M:斯德哥尔摩糖尿病干预研究中在传统胰岛素长期强化治疗期间的死亡率和治疗副作用(Mortality and TreatmentSide Effects Druing Long-term Intensified Conventional InsulinTreatment in the Stockholm Diabetes Intervention Study).Diabetes,3:313-317,1994)),和英国前瞻性糖尿病研究(见英国前瞻性糖尿病研究会:用二甲双胍进行强化血糖控制对于2型糖尿病患者并发症的影响(Effect of Intensive Blood Glucose Control With Metformin OnComplications In Patients With Type 2 Diabetes)(UKPDS 34).
Lancet,352:837-853,1998)。A large number of studies have repeatedly confirmed that the most effective way to prevent long-term complications of diabetes is to take intensive insulin therapy to strictly control blood glucose (BG) levels within the normal range. These studies include the Diabetes Control and Complications Trial (DCCT) (see DCCT Research Meeting: The Effect Of Intensive Treatment Of Diabetes On The Development And Progress Of Long-Term Complications Of Insulin-Dependent Diabetes Mellitus. New England Journal of Medicine, 329: 978-986, 1993)), Stockholm Diabetes Intervention Study (see Reichard P, Phil M: Mortality and Treatment Side Effects Druing Long-term Intensified Conventional Insulin Treatment in the Stockholm Diabetes Intervention Study in the Stockholm Diabetes Intervention Study). Diabetes, 3:313-317, 1994)), and the UK Prospective Diabetes Study (see UK Prospective Diabetes Study: Effect of Intensive Glycemic Control with Metformin on Complications in Patients with
然而,相同的研究也证实了胰岛素强化治疗的一些副作用,最严重的是会增加频繁严重低血糖症(SH)的危险,这是一种无法进行自我治疗和需要外界帮助才能恢复的神经低血糖症事件(见DCCT研究会:糖尿病控制和并发症实验中严重低血糖症的流行病学(Epidemiology of Severe Hypoglycemia In The Diabetes Control andComplications Trial).American Journal of Medicine,90:450-459,1991,和DCCT研究会:糖尿病控制和并发症实验中的低血糖症(Hypoglycemia in the Diabetes Control and ComplicationsTrial).Diabetes,46:271-286,1997)。因为SH会导致意外事故、昏迷甚至死亡,所以患者和康护人员对继续进行强化治疗感到灰心。结果,低血糖症被认定为是提高血糖控制的一个主要障碍(Cryer PE:低血糖症是糖尿病管理的限制因素(Hypoglycemia is the Limiting Factor inthe Management Of Diabetes).Diabetes Metab ResRev,15:42-46,1999)。However, the same study also demonstrated some side effects of intensive insulin therapy, the most serious being an increased risk of frequent severe hypoglycemia (SH), a neurological hypoglycemia that cannot be self-treated and requires outside assistance to recover (see DCCT Research Council: Epidemiology of Severe Hypoglycemia In The Diabetes Control and Complications Trial). American Journal of Medicine, 90: 450-459, 1991, and DCCT Research Council: Hypoglycemia in the Diabetes Control and Complications Trial. Diabetes, 46:271-286, 1997). Because SH can lead to accidents, coma, and even death, patients and health care providers are discouraged from continuing intensive treatment. As a result, hypoglycemia was identified as a major barrier to improved glycemic control (Cryer PE: Hypoglycemia is the Limiting Factor in the Management Of Diabetes). Diabetes Metab ResRev, 15:42-46 , 1999).
因此,糖尿病患者面临着一个终生的优化问题,即在保持对血糖严格控制的同时又不增加低血糖症的危险。与这个问题有关的主要挑战是产生一种能够同时评估患者的血糖控制及其低血糖症危险,并且能够在日常环境中使用的简单而可靠的方法。Diabetics thus face a lifelong optimization problem of maintaining tight glycemic control without increasing the risk of hypoglycemia. The main challenge related to this problem is to generate a simple and reliable method that can simultaneously assess a patient's glycemic control and his risk of hypoglycemia and that can be used in an everyday setting.
二十年来人们已经熟知,糖基化血红蛋白是糖尿病(1型或2型)个体血糖控制的标志。大量的研究人员对这一关系进行了研究并且发现,糖基化血红蛋白基本上反映了患者过去两个月内的平均BG水平。因为在大多数糖尿病患者体内,BG水平在一段时间内会有相当大的波动,所以建议整体血糖控制与HbA1c之间的实际关联只能在已知患者在一个较长时期内处于稳定的血糖控制的情况下进行观察。It has been well known for two decades that glycosylated hemoglobin is a marker of glycemic control in individuals with diabetes (
对该类患者的早期研究建立了前5周内平均BG水平与HbA1c之间的几乎确定的关系,并且该曲线关系产生了大小为0.98的相关系数(见Aaby Svendsen P,Lauritzen T,Soegard U,Nerup J(1982).1型(胰岛素依赖型)糖尿病中的糖基化血红蛋白与稳态平均血糖浓度(Glycosylated Hemoglobin and Steady-State Mean Blood GlucoseConcentration in Type 1(Insulin-Dependent)Diabetes). Diabetologia,23,403-405)。1993年,DCCT得出结论,HbA1c是黄金标准糖基化血红蛋白化验(gold-standard glycosylated hemoglobin assay)的“合理推荐”,并且DCCT确定了先前平均BG与HbA1c之间的线性关系(见Santiago JV(1993).来自糖尿病控制与并发症实验的教训(Lessonsfrom the Diabetes Control and Complications Trial), Diabetes,42,1549-1554)。Early studies in this category of patients established an almost certain relationship between mean BG levels over the previous 5 weeks and HbA 1c , and this curvilinear relationship yielded a correlation coefficient of 0.98 (see Aaby Svendsen P, Lauritzen T, Soegard U , Nerup J(1982). Glycosylated Hemoglobin and Steady-State Mean Blood Glucose Concentration in Type 1 (Insulin-Dependent) Diabetes. Diabetologia , 23 , 403-405). In 1993, the DCCT concluded that HbA 1c was a "reasonable recommendation" for the gold-standard glycosylated hemoglobin assay, and the DCCT established a linear relationship between prior mean BG and HbA 1c (see Santiago JV (1993). Lessons from the Diabetes Control and Complications Trial, Diabetes , 42 , 1549-1554).
已提出的指导方针表明,7%的HbA1c相应于8.3mM(150mg/dl)的平均BG,9%的HbA1c相应于11.7mM(210mg/dl)的平均BG,并且HbA1c增加1%相应于平均BG增加1.7mM(30mg/dl,2)。DCCT还建议,因为直接测量平均BG并不现实,所以可以用单一的简单测试,即HbA1c,评估患者的血糖控制。然而,研究清晰地表明,HbA1c对低血糖症并不敏感。Proposed guidelines indicate that an HbA 1c of 7% corresponds to a mean BG of 8.3 mM (150 mg/dl), an HbA 1c of 9% corresponds to a mean BG of 11.7 mM (210 mg/dl), and a 1% increase in HbA 1c corresponds to The average BG increased by 1.7mM (30mg/dl, 2). The DCCT also recommends that, since direct measurement of mean BG is impractical, a single simple test, HbA 1c , can be used to assess glycemic control in patients. However, research clearly shows that HbA 1c is not sensitive to hypoglycemia.
确实,由任何数据都不能得到患者SH直接危险的可靠预测值。DCCT得出结论,只有大约8%的将来SH能够由已知的参数,例如SH历史、低HbA1c和低血糖症昏迷(unawareness),预报出来。一篇近期的综述详细介绍了该问题的当前临床状况,并且为预防SH提供了患者及其康护人员能够获得的选择(见Bolli,GB:如何在1型糖尿病强化及非强化治疗中改善低血糖症问题(How To Ameliorate ThePreblem of Hypoglycemia In Intensive As Well As NonintensiveTreatment Of Type I Diabetes).Diabetes Care,22,Supplement2:B43-B52,1999)。Indeed, no reliable predictor of a patient's immediate risk for SH can be derived from any data. The DCCT concluded that only about 8% of future SH could be predicted by known parameters such as SH history, low HbA 1c and hypoglycemic unawareness. A recent review details the current clinical state of the issue and provides options for preventing SH that are available to patients and their caregivers (see Bolli, GB: How to improve low Glycemia problem (How To Ameliorate The Preblem of Hypoglycemia In Intensive As Well As Nonintensive Treatment Of Type I Diabetes). Diabetes Care, 22, Supplement 2: B43-B52, 1999).
现代家用BG监测器通过自我监测BG(SMBG)提供了进行频繁BG测量的装置。然而,SMBG的问题在于,通过BG监测器收集的数据与HbA1c和低血糖症之间缺乏联系。换言之,目前还没有可靠的方法能够根据SMBG读数估计HbA1c和识别即将发生的低血糖症(见Bremer T和Gough DA:血糖能够由先前值预测出来吗?数据的引发(Is blood glucose predictable from previous values?A solicitationfor data).Diabetes 48:445-451,1999)。Modern home BG monitors provide means for frequent BG measurements via self-monitoring BG (SMBG). The problem with SMBG, however, is the lack of association between data collected by BG monitors and HbA 1c and hypoglycemia. In other words, there is currently no reliable method for estimating HbA 1c and identifying impending hypoglycemia from SMBG readings (see Bremer T and Gough DA: Can blood glucose be predicted from previous values? Data elicited (Is blood glucose predictable from previous values? A solicitation for data). Diabetes 48:445-451, 1999).
因此,本发明的一个目的是通过提出三种不同但相互兼容的算法给出这一缺失的联系,三种算法用于由SMBG数据估计HbA1c和低血糖症危险,以预测低血糖症的长期和短期危险以及高血糖症的长期危险。It is therefore an object of the present invention to give this missing link by proposing three different but mutually compatible algorithms for estimating HbA 1c and hypoglycemia risk from SMBG data to predict long-term hypoglycemia and short-term risks as well as long-term risks of hyperglycemia.
发明人在先前曾经报道过,SMBG数据的日常获得与估计HbA1c和低血糖症危险之间缺乏联系的原因在于,在糖尿病研究中使用的数据收集与临床评估的精密方法很少获得糖尿病特异的和数学上精密的统计处理的支持。The inventors have previously reported that the reason for the lack of association between the routine acquisition of SMBG data and the estimation of HbA 1c and risk of hypoglycemia is that the sophisticated methods of data collection and clinical assessment used in diabetes research rarely yield diabetes-specific data. and support for mathematically sophisticated statistical processing.
出于对能够顾及BG特殊分布的统计分析的需要,发明人提出了一种对血糖测量值范围的对称变换(见Kovatchev BP,Cox DJ,Gonder-Frederick LA和WL Clarke(1997).血糖测量值范围的对称化及其应用(Symmetization of the Blood Glucose Measurement Scaleand Its Applications), Diabetes Care, 20,1655-1658),其操作如下。BG水平在美国以mg/dl为单位进行测量,在其他大多数国家中以mmol/L(或mM)为单位。两种标度的直接关系为18mg/dl=1mM。在大多数参考文献中给出的整体BG值范围是1.1-33.3mM,并且认为它实际地覆盖了全部的观察值。根据DCCT的推荐(见DCCT研究会(1993),糖尿病强化治疗对胰岛素依赖型糖尿病并发症的发生和发展的影响(TheEffect Of Intensive Treatment Of Diabetes On The Development AndProgresion Of Long-Term Complications Of Insulin-DependentDiabetes Mellitus).New England Journal ofMedicine,329:978-986,1993)),糖尿病人的目标BG值范围——也称作正常血糖范围——是3.9-10mM,当BG降到低于3.9mM时则出现低血糖症,当BG升高到10mM以上时则出现高血糖症。不幸的是,该范围在数值上并不对称——高血糖范围(10-33.3mM)宽于低血糖范围(1.1-3.9Mm),且正常血糖范围(3.9-10mM)不在该范围的中心。发明人通过引入一种变换f(BG)修正了该对称性,其中f(BG)是定义域为BG范围[1.1,33.3]上的连续函数,具有双参数解析形式:Out of the need for a statistical analysis capable of taking into account the particular distribution of BG, the inventors proposed a symmetric transformation of the range of blood glucose measurements (see Kovatchev BP, Cox DJ, Gonder-Frederick LA, and WL Clarke (1997). Blood glucose measurements Symmetization of the Blood Glucose Measurement Scale and Its Applications (Symmetization of the Blood Glucose Measurement Scale and Its Applications), Diabetes Care, 20, 1655-1658), which operates as follows. BG levels are measured in mg/dl in the US and mmol/L (or mM) in most other countries. The direct relationship between the two scales is 18 mg/dl = 1 mM. The overall BG value range given in most references is 1.1-33.3 mM and is considered to cover practically all observed values. According to the recommendation of DCCT (see DCCT Research Association (1993), The Effect Of Intensive Treatment Of Diabetes On The Development And Progression Of Long-Term Complications Of Insulin-Dependent Diabetes Mellitus ). New England Journal of Medicine, 329:978-986, 1993)), the target BG value range for diabetics - also known as normal blood sugar range - is 3.9-10mM, when BG drops below 3.9mM, it will appear Hypoglycemia, hyperglycemia occurs when BG rises above 10mM. Unfortunately, the range is not numerically symmetrical - the hyperglycemic range (10-33.3mM) is wider than the hypoglycemic range (1.1-3.9Mm), and the normoglycemic range (3.9-10mM) is not in the center of the range. The inventors have corrected this symmetry by introducing a transformation f(BG), where f(BG) is a continuous function with domain BG in the range [1.1, 33.3], with a two-parameter analytic form:
f(BG,α,β)=[(ln(BG))α-β],α,β>0f(BG, α, β)=[(ln(BG)) α -β], α, β>0
它满足假设:It satisfies the assumptions:
A1:f(33.3,α,β)=-f(1.1,α,β)和A1: f(33.3, α, β) = -f(1.1, α, β) and
A2:f(10.0,α,β)=-f(3.9,α,β)。A2: f(10.0, α, β)=-f(3.9, α, β).
接着,f(.)乘以第三比例参数从而将经过变换的BG范围的最小和最大值分别固定在 和 这些数值很方便,因为一个具有标准正态分布的随意变量在区间[ ]内具有99.8%的数值。如果BG以mmol/L为单位测量,则根据假设A1和A2进行数值解时,函数f(BG,α,β)的参数为α=1.026,β=1.861,比例参数γ=1.794。如果BG以mg/dl为单位测量,则计算出的参数为α=1.084,β=5.381,γ=1.509。Next, f(.) is multiplied by the third scale parameter to fix the minimum and maximum values of the transformed BG range at and These values are convenient because a random variable with a standard normal distribution is in the interval [ ] has a value of 99.8%. If BG is measured in mmol/L, the parameters of the function f(BG, α, β) are α = 1.026, β = 1.861, and the proportional parameter γ = 1.794 when numerically solved according to assumptions A1 and A2. If BG is measured in mg/dl, the calculated parameters are α = 1.084, β = 5.381, and γ = 1.509.
因此,当以mmol/L为单位测量BG时,对称变换是f(BG)=1.794[(ln(BG))1.026-1.861],当以mg/dl为单位测量BG时,对称变换为f(BG)=1.509[(ln(BG))1.084-5.381]。Therefore, when BG is measured in mmol/L, the symmetric transformation is f(BG)=1.794[(ln(BG)) 1.026 -1.861], and when BG is measured in mg/dl, the symmetric transformation is f( BG) = 1.509 [(ln(BG)) 1.084 - 5.381].
根据对称变换f(.),发明人引入了低BG指数——一个用于由SMBG读数估计低血糖症危险的新测量值(见Cox DJ,Kovatchev BP,Julian DM,Gonder-Frederick LA,Polonsky WH,Schlundt DG,Clarke WL:IDDM中的严重低血糖症频率能够由自我监测血糖数据进行预测(Frequency of Severe Hypoglycemia In IDDM Can BePredicted From Self-Monitoring Blood Glucose Data).Journal ofClinical Endocrinology and Metabolism,79:1659-1662,1994,和Kovatchev BP,Cox DJ,Gonder-Frederick LA,Young-Hyman D,Schlundt D,Clarke WL.IDDM成年人中严重低血糖症危险估计:低血糖指数的有效性(Assessment of Risk for Severe HypoglycemiaAmong Adults With IDDM:Validation of the Low Blood GlucoseIndex),Diabetes Care 21:1870-1875,1998)。给定一系列SMBG数据,当f(BG)<0时,低BG指数计算为10.f(BG)2的平均值,否则为0。还提出了高BG指数,按照和低BG指数对称的方式计算而得,但是该指数还没有发现其实际用途。From the symmetric transformation f(.), the inventors introduced the low BG index - a new measure for estimating the risk of hypoglycemia from SMBG readings (see Cox DJ, Kovatchev BP, Julian DM, Gonder-Frederick LA, Polonsky WH , Schlundt DG, Clarke WL: Frequency of Severe Hypoglycemia In IDDM Can Be Predicted From Self-Monitoring Blood Glucose Data in IDDM. Journal of Clinical Endocrinology and Metabolism, 79: 1659 -1662, 1994, and Kovatchev BP, Cox DJ, Gonder-Frederick LA, Young-Hyman D, Schlundt D, Clarke WL. Risk estimation of severe hypoglycemia in adults with IDDM: the validity of the Assessment of Risk for Severe Hypoglycemia Among Adults With IDDM: Validation of the Low Blood Glucose Index), Diabetes Care 21: 1870-1875, 1998). Given a series of SMBG data, the low BG index is calculated as the mean of 10.f(BG) 2 when f(BG)<0, and 0 otherwise. A high BG index has also been proposed, calculated in a symmetrical manner to the low BG index, but this index has not yet found its practical use.
在回归模型中使用低BG指数,发明人能够根据SH历史和SMBG数据解释随后6个月中SH事件40%的变异,后来将该预测提高到了46%(见Kovatchev BP,Straume M,Farhi LS,Cox DJ:估计血糖转变速度及其与严重低血糖的关系(Estimating the Speed ofBlood Glncose Transitions and its Relationship With SevereHypoglycemia),Diabetes,48:Supplement 1,A363,1999)。Using a low BG index in a regression model, the inventors were able to explain 40% of the variance in SH events over the subsequent 6 months based on SH history and SMBG data, later improving this prediction to 46% (see Kovatchev BP, Straume M, Farhi LS, Cox DJ: Estimating the Speed of Blood Glncose Transitions and its Relationship With Severe Hypoglycemia, Diabetes, 48:
此外,发明人还提出了一些关于HbA1c和SMBG的数据(见Kovatchev BP,Cox DJ,Straume M,Farhi LS.自我监测血糖曲线与糖基化血红蛋白的关系(Association of Self-monitoring Blood GlucoseProfiles with Glycosylated Hemoglobin), In:Methods in Enzymology, vol.321:Numerical Computer Methods,Part C,Mechael Johnson和Ludvig Brand,Eds.,Academic Press,NY;2000)。In addition, the inventors present some data on HbA 1c and SMBG (see Kovatchev BP, Cox DJ, Straume M, Farhi LS. Association of Self-monitoring Blood Glucose Profiles with Glycosylated Hemoglobin Hemoglobin), In: Methods in Enzymology, vol. 321: Numerical Computer Methods, Part C , Michael Johnson and Ludvig Brand, Eds., Academic Press, NY; 2000).
这些研究成果成为本发明理论背景的一部分。为了将该理论付诸于实践,添加了几个关键的理论参量,在下面的部分中将进行说明。特别地,提出了三种方法用于估计HbA1c、低血糖症的长期和短期危险。这些方法的提出是基于,但不仅限于,对867个糖尿病个体超过300,000个SMBG读数、严重低血糖症记录和HbA1c结果的详细分析。These research results form part of the theoretical background of the present invention. To put this theory into practice, several key theoretical parameters are added, which are explained in the following sections. In particular, three methods are proposed for estimating HbA 1c , long-term and short-term risks of hypoglycemia. These methods are based on, but not limited to, a detailed analysis of more than 300,000 SMBG readings, severe hypoglycemia records, and HbA 1c results from 867 diabetic individuals.
因此,发明人试图改进与传统方法有关的前述限制,借此提供简单而可靠的方法,从而能够用于同时评估患者的血糖控制及其低血糖症的危险,并能够在其日常环境中使用。Therefore, the inventors attempted to improve the aforementioned limitations related to traditional methods, thereby providing a simple and reliable method that can be used to simultaneously assess a patient's glycemic control and his risk of hypoglycemia and can be used in their daily environment.
发明内容Contents of the invention
本发明包括一种数据分析方法和基于计算机的系统,用于由日常收集的SMBG数据同时估计糖尿病血糖控制中两种最重要的分量:HbA1c和低血糖症危险。出于本文献的目的,BG自我监测(SMBG)被定义为用于在糖尿病患者的自然条件下确定血糖的方法,并且包括当前通常存储200-250个BG读数的SMBG设备使用的方法,以及将来生产的连续检测技术使用的方法。通过给出SMBG的这一广泛定义,本发明直接致力于通过引入一种能够同时预测HbA1c和低血糖症高危期的智能数据判读组件提高(但不仅限于)现有家用血糖监测设备的性能,以及通过相同的部件提高将来生产的连续监测设备的性能。The present invention includes a data analysis method and computer-based system for simultaneously estimating the two most important components in diabetic glycemic control: HbA 1c and hypoglycemia risk, from routinely collected SMBG data. For the purposes of this document, self-monitoring of BG (SMBG) is defined as the method used to determine blood glucose under natural conditions in diabetic patients, and includes methods currently used by SMBG devices that typically store 200-250 BG readings, as well as future The method used in the continuous detection technique of production. By giving this broad definition of SMBG, the present invention is directed towards improving (but not limited to) the performance of existing home blood glucose monitoring devices by introducing an intelligent data interpretation component capable of simultaneously predicting HbA 1c and high-risk periods of hypoglycemia, As well as improving the performance of continuous monitoring equipment produced in the future by the same components.
本发明的一个方面包括一种方法、系统和计算机程序产品,用于由在预定时期内,例如大约4-6周,收集的SMBG数据估计HbA1c。在一个实施例中,本发明提供了一种计算机化的方法和系统,用于根据在预定时期内收集的BG数据估计患者的HbA1c。本方法(或系统或计算机可用介质)包括根据在第一预定时期内收集的BG数据估计患者的HbA1c。本方法包括:使用如下定义的预定序列数学公式准备用于估计HbA1c的数据:数据预处理;使用四个预定公式中的至少一个估计HbA1c;和通过样本选择标准验证估计的有效性。One aspect of the invention includes a method, system and computer program product for estimating HbA 1c from SMBG data collected over a predetermined period of time, eg, about 4-6 weeks. In one embodiment, the present invention provides a computerized method and system for estimating a patient's HbA 1c from BG data collected over a predetermined period of time. The method (or system or computer usable medium) includes estimating the patient's HbA 1c based on BG data collected over a first predetermined period of time. The method comprises: preparing data for estimating HbA 1c using a predetermined sequence of mathematical formulas defined as follows: data preprocessing; estimating HbA 1c using at least one of four predetermined formulas; and validating the estimates by sample selection criteria.
本发明的另一个方面包括一种方法、系统和计算机程序产品,用于估计严重低血糖症(SH)的长期概率。本方法使用预定时期内,例如4-6周,的SMBG读数,并预测随后大约6个月内的SH危险。在一个实施例中,本发明提供了一种计算机化的方法和系统,用于根据在预定时期内收集的BG数据估计严重低血糖症(SH)的长期概率。本方法(或系统或计算机可用介质)包括根据在预定持续时间内收集的BG数据估计患者严重低血糖症(SH)或者中度低血糖症(MH)的长期概率。本方法包括:根据所收集的BG数据计算LBGI;和根据计算出的LBGI利用预定的数学公式估计将来SH事件的数目。Another aspect of the invention includes a method, system and computer program product for estimating the long-term probability of severe hypoglycemia (SH). The method uses SMBG readings over a predetermined period of time, eg, 4-6 weeks, and predicts the risk of SH over a subsequent period of approximately 6 months. In one embodiment, the present invention provides a computerized method and system for estimating the long-term probability of severe hypoglycemia (SH) from BG data collected over a predetermined period of time. The method (or system or computer usable medium) includes estimating a patient's long-term probability of severe hypoglycemia (SH) or moderate hypoglycemia (MH) from BG data collected over a predetermined duration. The method includes: calculating LBGI based on the collected BG data; and estimating the number of future SH events based on the calculated LBGI using a predetermined mathematical formula.
本发明的再一个方面包括一种方法、系统和计算机程序产品,用于识别24小时内(或者其他选择时期)的低血糖症高危期。其是通过使用在先前24小时收集的SMBG读数计算低血糖症的短期危险而实现的。在一个实施例中,本发明提供了一种计算机化的方法和系统,用于根据在预定时期内收集的BG数据估计严重低血糖症(SH)的短期危险。本方法(或系统或计算机可用介质)包括根据在预定持续时间内收集的BG数据估计患者严重低血糖症(SH)的短期概率。本方法包括:根据所收集的BG数据计算scale值;并为每个BG数据计算低BG危险值(RLO)。Yet another aspect of the invention includes a method, system and computer program product for identifying periods of high risk for hypoglycemia within 24 hours (or other selected periods). This was achieved by calculating the short-term risk of hypoglycemia using SMBG readings collected over the previous 24 hours. In one embodiment, the present invention provides a computerized method and system for estimating the short-term risk of severe hypoglycemia (SH) from BG data collected over a predetermined period of time. The method (or system or computer usable medium) includes estimating a short-term probability of severe hypoglycemia (SH) in a patient based on BG data collected over a predetermined duration. The method includes: calculating a scale value according to the collected BG data; and calculating a low BG risk value (RLO) for each BG data.
本发明的这三个方面以及本文献全文讨论的其他方面,能够组合在一起提供有关糖尿病个体血糖控制的连续信息,并提高对低血糖症危险的监控。These three aspects of the invention, as well as others discussed throughout this document, can be combined to provide continuous information on glycemic control in diabetic individuals and improve monitoring of the risk of hypoglycemia.
本文中,本发明的这些及其他目标以及其优点和特点从下面的说明、附图和权利要求中将变得更加显而易见。These and other objects of the present invention, together with its advantages and features, will herein become more apparent from the following description, drawings and claims.
附图说明Description of drawings
通过联系附图一起阅读如下的优选实施例说明将对本发明的前述及其他目的、特点和优点以及本发明自身具有更全面的理解,其中:The aforementioned and other objects, features and advantages of the present invention and the present invention itself will have a more comprehensive understanding by reading the following description of preferred embodiments in conjunction with the accompanying drawings, wherein:
图1图示地提供了在对15个由实例No.1的低BG指数定义的危险水平范围的每一个进行SMBG估计之后1个月内发生中度(虚线)和严重(实线)低血糖症的经验和理论概率。Figure 1 graphically presents the occurrence of moderate (dashed line) and severe (solid line) hypoglycemia within 1 month after SMBG estimation for each of the 15 risk level ranges defined by the low BG index of Example No. 1 Empirical and theoretical probabilities of symptoms.
图2图示地提供了在对15个由实例No.1的低BG指数定义的危险水平范围的每一个进行SMBG估计之后3个月内发生中度(虚线)和严重(实线)低血糖症的经验与理论概率。Figure 2 graphically presents the occurrence of moderate (dotted line) and severe (solid line) hypoglycemia within 3 months after SMBG estimation for each of the 15 risk level ranges defined by the low BG index of Example No. 1 Empirical and theoretical probabilities of the disease.
图3图示地提供了在对15个由实例No.1的低BG指数定义的危险水平范围的每一个进行SMBG估计之后6个月内发生中度(虚线)和严重(实线)低血糖症的经验与理论概率。Figure 3 graphically presents the occurrence of moderate (dotted line) and severe (solid line) hypoglycemia within 6 months after SMBG estimation for each of the 15 risk level ranges defined by the low BG index of Example No. 1 Empirical and theoretical probabilities of the disease.
图4图示地提供了在对15个由实例No.1的低BG指数定义的危险水平范围的每一个进行SMBG估计之后3个月内发生2次或更多次中度(虚线)和严重(实线)低血糖症的经验与理论概率。Figure 4 graphically provides 2 or more occurrences of moderate (dotted line) and severe (Solid line) Empirical versus theoretical probability of hypoglycemia.
图5图示地提供了在对15个由实例No.1的低BG指数定义的危险水平范围的每一个进行SMBG估计之后6个月内发生2次或更多次中度(虚线)和严重(实线)低血糖症的经验与理论概率。Figure 5 graphically provides 2 or more occurrences of moderate (dotted line) and severe (Solid line) Empirical versus theoretical probability of hypoglycemia.
图6是用于实现本发明的计算机系统的功能框图。Fig. 6 is a functional block diagram of a computer system for implementing the present invention.
图7-9是本发明相关处理器、通信连接和系统的可替代变型的示意性框图。7-9 are schematic block diagrams of alternative variations of the associated processors, communication connections and systems of the present invention.
图10图示地提供了在对15个由实例No.1的低BG指数定义的危险水平范围的每一个进行SMBG估计之后6个月内发生3次或更多次中度(虚线)和严重(实线)低血糖症的经验与理论概率。Figure 10 graphically provides 3 or more occurrences of moderate (dotted line) and severe (Solid line) Empirical versus theoretical probability of hypoglycemia.
图11图示地显示了本模型的残差分析,显示出与实例No.1训练数据组1残差的正态分布相近。Figure 11 graphically shows the residual analysis of this model, showing that it is close to the normal distribution of the residuals of the
图12图示地显示了本模型的残差分析,显示出与实例No.1残差的正态分布相近。Figure 12 graphically shows the analysis of the residuals of this model, showing a close to normal distribution of the residuals of Example No.1.
图13图示地显示了由实例No.1的正态概率图1给出的统计证据。Figure 13 graphically shows the statistical evidence given by the normal probability figure 1 for Example No. 1.
图14图示地提供了实例No.1中用百分率表示的命中率和比值Rud之间的平滑依赖性。Fig. 14 graphically provides the smoothed dependence between the hit rate expressed in percentage and the ratio Rud in Example No. 1.
图15图示地提供了实例No.1中预测时期与相应命中率之间的依赖性。Fig. 15 graphically provides the dependence between the prediction epoch and the corresponding hit rate in Example No. 1.
图16(A)-(B)图示地提供了由LBGI预测的T1DM显著低血糖症1个月内的危险,在实例No.2中各危险组严重低血糖症事件数目的ANOVA(F=7.2,p<0.001)和各危险组中度低血糖症事件数目的ANOVA(F=13.9,p<0.001)。Figure 16 (A)-(B) graphically provides the risk within 1 month of T1DM significant hypoglycemia predicted by LBGI, the ANOVA (F= 7.2, p<0.001) and ANOVA of the number of moderate hypoglycemia events in each risk group (F=13.9, p<0.001).
图17(A)-(B)图示地提供了由LBGI预测的T1DM显著低血糖症3个月内的危险,在实例No.2中各危险组严重低血糖症事件数目的ANOVA(F=9.2,p<0.001)和各危险组中度低血糖症事件数目的ANOVA(F=14.7,p<0.001)。Figure 17 (A)-(B) graphically provides the risk within 3 months of T1DM significant hypoglycemia predicted by LBGI, the ANOVA (F= 9.2, p<0.001) and ANOVA of the number of moderate hypoglycemia events in each risk group (F=14.7, p<0.001).
图18(A)-(B)图示地提供了由LBGI预测的T2DM显著低血糖症1个月内的危险,在实例No.2中各危险组严重低血糖症事件数目的ANOVA(F=6.0,p<0.001)和各危险组中度低血糖症事件数目的ANOVA(F=25.1,p<0.001)。Figure 18 (A)-(B) graphically provides the risk within 1 month of significant hypoglycemia in T2DM predicted by LBGI, the ANOVA (F= 6.0, p<0.001) and ANOVA of the number of moderate hypoglycemia events in each risk group (F=25.1, p<0.001).
图19(A)-(B)图示地提供了由LBGI预测的T2DM显著低血糖症3个月内的危险,在实例No.2中各危险组严重低血糖症事件数目的ANOVA(F=5.3,p<0.01)和各危险组中度低血糖症事件数目的ANOVA(F=20.1,p<0.001)。Figure 19 (A)-(B) graphically provides the risk of significant hypoglycemia in T2DM predicted by LBGI within 3 months, ANOVA (F= 5.3, p<0.01) and ANOVA of the number of moderate hypoglycemia events in each risk group (F=20.1, p<0.001).
具体实施方式Detailed ways
本发明使得,但不仅限于,有可能产生用于估计糖尿病患者血糖控制的精确方法,并且包括在计算该方法的关键分量时使用的固件和软件编码。用于估计HbA1c、SH长期概率和低血糖短期危险的发明方法也能够根据所收集的大量数据加以验证,并且在本文的后面将进行讨论。最后,这些方法的方案能够被组合成结构化显示或者矩阵。The present invention enables, but is not limited to, the creation of an accurate method for estimating glycemic control in diabetic patients, and includes firmware and software codes used in calculating the key components of the method. The inventive method for estimating HbA 1c , SH long-term probability, and hypoglycemia short-term risk can also be validated on the large amount of data collected and will be discussed later in this paper. Finally, the schemes of these methods can be combined into structured displays or matrices.
I.估计HbA1c I. Estimated HbA 1c
本发明的一个方面包括一种方法、系统和计算机程序软件,用于由在预定时期内,例如4-6周,收集的SMBG数据估计HbA1c。在一个实施例中,本发明提供了一种计算机化(或者其他类型)的方法和系统,用于根据在预定持续时间内收集的BG数据估计患者的HbA1c。本方法包括根据在第一预定持续时间内收集的BG数据估计患者的HbA1c,本方法包括:利用预定序列的数学公式准备用于估计HbA1c的数据。该数学公式被定义为:数据的预处理;使用四个预定公式中的至少一个估计HbA1c;和通过样本选择标准验证估计的有效性。第一预定持续时间能够为大约60天,或者选择地,第一预定持续时间范围是大约45天到大约75天,或者大约45天到大约90天,或者根据期望。每个患者的数据预处理包括:将血浆BG转换成全血BGmg/dl;将以mg/dl为单位测量的BG转换成mmol/l的单位;和计算低血糖指数(RLO1)和高血糖指数(RHI1)。对每个患者的数据预处理使用如下定义的预定数学公式:通过BG=PLASBG(mg/dl)/1.12将血浆BG转换成全血BG mg/dl;通过BGMM=BG/18将以mg/dl测量的BG转换成mmol/l单位;和计算低血糖指数(RLO1)和高血糖指数(RHI1)。数据的预处理进一步使用如下定义的预定数学公式:Scale=[In(BG)]1.0845-5.381,其中BG是以mg/dl为单位测量的;Risk1=22.765(Scale)2,其中RiskLO=Risk1,如果(BG小于大约112.5),因此存在LBGI的危险,否则RiskLO=0;RiskHI=Risk1,如果(BG大于大约112.5),因此存在HGBI的危险,否则RiskHI=0;BGMM1=每个患者的平均BGMM;RLO1=每个患者的平均RsikLO;RHI1=每个患者的平均RiskHI;L06=只对夜间读数计算的平均RiskLO,如果没有夜间读数则缺省;N06,N12,N24是各时间间隔中SMBG读数的百分率;NC1=第一预定持续时间内SMBG读数的总数;NDAYS=第一预定持续时间内具有SMBG读数的天数。N06、N12、N24分别如下时间间隔中SMBG读书的百分率,即大约0-6:59、大约7-12:59和大约18-23:59,或者其他的期望百分率和间隔数目。One aspect of the invention includes a method, system and computer program software for estimating HbA 1c from SMBG data collected over a predetermined period of time, eg, 4-6 weeks. In one embodiment, the present invention provides a computerized (or other type) method and system for estimating a patient's HbA 1c from BG data collected over a predetermined duration. The method includes estimating the patient's HbA 1c from BG data collected over a first predetermined duration, the method includes preparing the data for estimating the HbA 1c using a predetermined sequence of mathematical formulas. The mathematical formula is defined as: preprocessing of data; estimating HbA 1c using at least one of four predetermined formulas; and validating the estimate by sample selection criteria. The first predetermined duration can be about 60 days, or alternatively, the first predetermined duration ranges from about 45 days to about 75 days, or from about 45 days to about 90 days, or as desired. Data preprocessing for each patient included: conversion of plasma BG to whole blood BG mg/dl; conversion of BG measured in mg/dl to mmol/l; and calculation of low glycemic index (RLO1) and high glycemic index (RHI1). Data preprocessing for each patient uses a predetermined mathematical formula defined as follows: convert plasma BG to whole blood BG mg/dl by BG=PLASBG(mg/dl)/1.12; convert BG in mg/dl by BGMM=BG/18 The measured BG was converted into mmol/l units; and the low glycemic index (RLO1) and high glycemic index (RHI1) were calculated. The preprocessing of the data further uses a predetermined mathematical formula defined as follows: Scale=[In(BG)] 1.0845-5.381 , where BG is measured in mg/dl; Risk1=22.765(Scale) 2 , where RiskLO=Risk1, If (BG is less than about 112.5) then there is a risk of LBGI, otherwise RiskLO = 0; RiskHI = Risk1, if (BG is greater than about 112.5) then there is a risk of HGBI, otherwise RiskHI = 0; BGMM1 = mean BGMM per patient ; RLO1 = average RsikLO per patient; RHI1 = average RiskHI per patient; L06 = average RiskLO calculated for nighttime readings only, default if no nighttime readings; N06, N12, N24 are SMBG readings in each time interval NC1 = total number of SMBG readings within the first predetermined duration; NDAYS = number of days with SMBG readings within the first predetermined duration. N06, N12, N24 are respectively the percentages of SMBG readings in the following time intervals, namely about 0-6:59, about 7-12:59 and about 18-23:59, or other desired percentages and interval numbers.
本方法进一步包括根据用预定数学公式计算的患者高BG指数为组赋值。该公式可以定义为:如果(RHI1≤大约5.25或者如果RHI1≥大约16),则赋值group=0;如果(RHI1>大约5.25并且如果RHI1<大约7.0),则赋值group=1;如果(RHI1≥大约7.0并且如果RHI1<大约8.5),则赋值group=2;和如果(RHI1≥大约8.5并且如果RHI1<大约16),则赋值group=3。The method further includes assigning a group value based on the patient's high BG index calculated using a predetermined mathematical formula. The formula can be defined as: if (RHI1≤about 5.25 or if RHI1≥about 16), then assign group=0; if (RHI1>about 5.25 and if RHI1<about 7.0), then assign group=1; if (RHI1≥ About 7.0 and if RHI1<about 8.5), assign group=2; and if (RHI1≥about 8.5 and if RHI1<about 16), assign group=3.
接着,本方法可以进一步包括使用如下定义的预定数学公式给出估计:Then, the method may further include giving an estimate using a predetermined mathematical formula defined as follows:
E0=0.55555*BGMM1+2.95;E1=0.50567*BGMM1+0.074*L06+2.69;E0=0.55555*BGMM1+2.95; E1=0.50567*BGMM1+0.074*L06+2.69;
E2=0.5555*BGMM1-0.074*L06+2.96;E2=0.5555*BGMM1-0.074*L06+2.96;
E3=0.44000*BGMM1+0.035*L06+3.65;并且如果(group=1),则E3=0.44000*BGMM1+0.035*L06+3.65; and if (group=1), then
EST2=E1,或者如果(group=2)则EST2=E2,或者如果(group=3)EST2=E1, or if (group=2) then EST2=E2, or if (group=3)
则EST2=E3,否则EST2=E0。Then EST2=E3, otherwise EST2=E0.
本方法包括使用如下定义的预定数学公式对估计进行进一步的修正:如果(缺省(L06)),EST2=E0,如果(RLO1≤大约0.5并且RHI1≤大约2.0),则EST2=E0-0.25;如果(RLO1≤大约2.5并且RHI1>大约26),则EST2=E0-1.5*RLO1;并且如果((RLO1/RHI1)≤大约0.25并且L06>大约1.3)则EST2=EST2-0.08。The method includes further correction of the estimate using a predetermined mathematical formula defined as follows: if (default (L06)), EST2 = E0, if (RLO1 < about 0.5 and RHI1 < about 2.0), then EST2 = E0 - 0.25; If (RLO1≦about 2.5 and RHI1>about 26), then EST2=E0-1.5*RLO1; and if ((RLO1/RHI1)≦about 0.25 and L06>about 1.3) then EST2=EST2-0.08.
根据在第一预定持续时间内收集的BG数据估计患者的HbA1c能够通过使用四个预定数学公式中的至少一个估计HbA1c而实现,四个公式的定义如下:Estimating the patient's HbA 1c from BG data collected during the first predetermined duration can be accomplished by estimating the HbA 1c using at least one of four predetermined mathematical formulas, defined as follows:
a)HbA1c=如上面定义的或者如上面修正的EST2;a) HbA 1c = EST2 as defined above or as amended above;
b)HbA1c=0.809098*BGMM1+0.064540*RLO1-0.151673*RHI1+1.873325,其中BGMM1是BG的平均值(mmol/l),RLO1是低BG指数,RHI1是高BG指数;b) HbA 1c =0.809098*BGMM1+0.064540*RLO1-0.151673*RHI1+1.873325, where BGMM1 is the average value of BG (mmol/l), RLO1 is the low BG index, and RHI1 is the high BG index;
c)HbA1c=0.682742*HBA0+0.054377*RHI1+1.553277,其中HBA0是在估计之前的大约第二预定时期内采用的先前参考HbA1c读数,其中RHI1是高BG指数;或者 or _
d)HbA1c=0.41046*BGMM+4.0775,其中BGMM1是BG的平均值(mmol/l)。第二预定持续时间能够为大约3个月;大约2.5个月到大约3.5个月;或者大约2.5个月到6个月;或者根据期望。d) HbA 1c =0.41046*BGMM+4.0775, where BGMM1 is the mean value of BG (mmol/l). The second predetermined duration can be about 3 months; about 2.5 months to about 3.5 months; or about 2.5 months to 6 months; or as desired.
只有当第一预定持续时间样本满足如下四个标准中的至少一个时,才用HbA1c估计的样本选择标准验证估计的有效性:The sample selection criteria for HbA 1c estimation were used to verify the validity of the estimate only if the first predetermined duration sample met at least one of the following four criteria:
a)测试频率标准,其中第一预定持续时间样本平均每天至少测试大a) Test frequency criteria, where the first predetermined duration sample averages at least one test per day
约1.5到大约2.5次;b)可选择测试频率标准,其中预定持续时间样本在第三预定采样时期内读数的平均频率为大约1.8个读数/天(或者其他期望的平均频率);about 1.5 to about 2.5 times; b) an optional test frequency criterion, wherein the average frequency of readings for samples of the predetermined duration over a third predetermined sampling period is about 1.8 readings/day (or other desired average frequency);
c)数据随机化标准-1,其中只有当比值(RLO1/RHI1)>=大约0.005时c) Data Randomization Criterion-1, where only if the ratio (RLO1/RHI1) >= about 0.005
才验证和显示HbA1c估计,其中RLO1是低BG指数,RHI1是高BG指数;或者to validate and display the HbA 1c estimate, where RLO1 is the low BG index and RHI1 is the high BG index; or
d)数据随机化标准,其中只有当比值(NO6>=大约3%)时才验证和显示HbA1c估计,且其中NO6是夜间读数的平均值。第三预定持续时间能够为至少35天,范围从大约35天到大约40天,或者从大约35天到大约和第一预定持续时间一样长,或者根据期望。d) Data randomization criteria, where HbA 1c estimates are validated and displayed only if ratio (NO6 >= about 3%), and where NO6 is the mean of nighttime readings. The third predetermined duration can be at least 35 days, ranging from about 35 days to about 40 days, or from about 35 days to about as long as the first predetermined duration, or as desired.
II.严重低血糖症(SH)的长期概率II. Long-term probability of severe hypoglycemia (SH)
本发明的另一个方面包括一种方法、系统和计算机程序产品,用于估计严重低血糖症(SH)的长期概率。该方法使用预定时期,例如大约4-6周,的SMBG读数并预测在随后大约6个月内SH的危险。在一个实施例中,本发明提供了一种计算机化的(或者其他类型)方法和系统,用于根据在预定持续时间内收集的BG数据估计患者严重低血糖症(SH)的长期概率。根据在预定持续时间内收集的BG数据估计患者严重低血糖症(SH)或者中度低血糖症(MH)的长期概率的方法包括:根据收集的BG数据计算LBGI;和根据计算的LBGI利用预定的数学公式估计将来SH事件的数目。LBGI的计算由在时间点t1,t2,…,tn采集一系列BG读数x1,x2,…,xn数学地定义:Another aspect of the invention includes a method, system and computer program product for estimating the long-term probability of severe hypoglycemia (SH). The method uses SMBG readings for a predetermined period of time, eg, about 4-6 weeks, and predicts the risk of SH in the following about 6 months. In one embodiment, the present invention provides a computerized (or other type) method and system for estimating a patient's long-term probability of severe hypoglycemia (SH) from BG data collected over a predetermined duration. A method for estimating a patient's long-term probability of severe hypoglycemia (SH) or moderate hypoglycemia (MH) from BG data collected over a predetermined duration includes: calculating LBGI from the collected BG data; The mathematical formula estimates the number of future SH events. The calculation of LBGI is defined mathematically by taking a series of BG readings x1 , x2 , ..., xn at time points t1 , t2 , ..., tn :
我们定义了预定危险范围(risk category)(RCAT),借此每个危险范围(RCAT)表示LBGI的一个数值范围;并且将LBGI赋值给所述危险范围(RCAT)中的至少一个。危险范围(RCAT)的定义如下:We define predetermined risk categories (RCAT), whereby each risk category (RCAT) represents a numerical range of LBGI; and assign LBGI to at least one of said risk categories (RCAT). The Risk Range (RCAT) is defined as follows:
范围1,其中所述LBGI小于大约0.25;
范围2,其中所述LBGI介于大约0.25-大约0.50;
范围3,其中所述LBGI介于大约0.50-大约0.75;
范围4,其中所述LBGI介于大约0.75-大约1.0;
范围5,其中所述LBGI介于大约1.0-大约1.25;
范围6,其中所述LBGI介于大约1.25-大约1.50;
范围7,其中所述LBGI介于大约1.50-大约1.75;
范围8,其中所述LBGI介于大约1.75-大约2.0;
范围9,其中所述LBGI介于大约2.0-大约2.5;
范围10,其中所述LBGI介于大约2.5-大约3.0;
范围11,其中所述LBGI介于大约3.0-大约3.5;
范围12,其中所述LBGI介于大约3.5-大约4.25;
范围13,其中所述LBGI介于大约4.25-大约5.0;
范围14,其中所述LBGI介于大约5.0-大约6.5;和
范围15,其中所述LBGI大于大约6.5。Range 15, wherein the LBGI is greater than about 6.5.
接着,分别为每个所述指定危险范围(RCAT)限定发生所选数目SH事件的概率。利用如下的公式分别为每个所述指定危险范围(RCAT)限定在下一个第一预定持续时间内发生所选数目SH事件的概率:F(x)=1-exp(-a.xb),x>0,否则为0,其中:a≈-4.19,b≈1.75(a和/或b可以是其他的期望值)。第一预定持续时间能够是大约1个月,范围从0.5个月到大约1.5个月,或者范围从大约0.5个月到大约3个月,或者根据期望。Next, the probability of occurrence of the selected number of SH events is defined separately for each of said specified risk ranges (RCAT). The probability of the selected number of SH events occurring within the next first predetermined duration is defined separately for each of said specified risk ranges (RCAT) using the following formula: F(x)=1-exp(-ax b ), x> 0, otherwise 0, wherein: a≈-4.19, b≈1.75 (a and/or b can be other expected values). The first predetermined duration can be about 1 month, range from about 0.5 months to about 1.5 months, or range from about 0.5 months to about 3 months, or as desired.
此外,利用如下的公式分别为每个所述指定危险范围(RCAT)限定在下一个第二预定持续时间内发生所选数目SH事件的概率:F(x)=1-exp(-a.xb),x>0,否则为0,其中:a≈-3.28,b≈1.50(a和/或b可以是其他的期望值)。第二预定持续时间能够为大约3个月,范围从大约2个月到大约4个月,或者从大约3个月到大约6个月,或者根据期望。Furthermore, the probability of the selected number of SH events occurring within the next second predetermined duration is defined separately for each of said specified risk ranges (RCAT) using the formula: F(x)=1-exp(-ax b ), x>0, otherwise 0, wherein: a≈-3.28, b≈1.50 (a and/or b can be other desired values). The second predetermined duration can be about 3 months, ranging from about 2 months to about 4 months, or from about 3 months to about 6 months, or as desired.
进一步,利用如下的公式分别为每个所述指定危险范围(RCAT)限定在下一个第三预定持续时间内发生所选数目SH事件的概率:F(x)=1-exp(-a.xb),x>0,否则为0,其中:a≈-3.06,b≈1.45(a和/或b可以是其他的期值)。第三预定持续时间能够为大约6个月,范围从大约5个月到大约7个月,或者从大约3个月到大约9个月,或者根据期望。Further, the probability of a selected number of SH events occurring within the next third predetermined duration is defined for each of the specified risk ranges (RCAT) using the following formula: F(x)=1-exp(-ax b ), x>0, otherwise it is 0, wherein: a≈-3.06, b≈1.45 (a and/or b can be other expected values). The third predetermined duration can be about 6 months, ranging from about 5 months to about 7 months, or from about 3 months to about 9 months, or as desired.
选择地,利用如下的公式分别为每个所述指定危险范围(RCAT)限定在下一个第一预定时期内(范围是大约1个月,大约0.5-1.5个月,大约0.5-3个月,或者根据期望)发生所选数目MH事件的概率:F(x)=1-exp(-a.xb),x>0,否则为0,其中:a≈-1.58,b≈1.05(a和/或b可以是其他的期望值)。Optionally, each of said specified risk ranges (RCAT) is defined within the next first predetermined period (range is about 1 month, about 0.5-1.5 months, about 0.5-3 months, or Probability of occurrence of a selected number of MH events according to expectations: F(x)=1-exp(-ax b ), x>0, otherwise 0, where: a≈-1.58, b≈1.05 (a and/or b can be other expected values).
选择地,利用如下的公式分别为每个所述指定危险范围(RCAT)限定在下一个第二预定持续时间内(范围是大约3个月,大约2-4个月,大约3-6个月,或者根据期望)发生所选数目MH事件的概率:F(x)=1-exp(-a.xb),x>0,否则为0,其中:a≈-1.37,b≈1.14(a和/或b可以是其他的期望值)。Optionally, each of the specified risk ranges (RCAT) is limited to the next second predetermined duration (range is about 3 months, about 2-4 months, about 3-6 months, or according to expectation) probability of occurrence of selected number of MH events: F(x)=1-exp(-ax b ), x>0, otherwise 0, where: a≈-1.37, b≈1.14 (a and/or b can be other desired value).
选择地,利用如下的公式分别为每个所述指定危险范围(RCAT)限定在下一个第三预定持续时间内(范围是大约6个月,大约5-7个月,大约3-9个月,或者根据期望)发生所选数目MH事件的概率:F(x)=1-exp(-a.xb),x>0,否则为0,其中:a≈-1.37,b≈1.35(a和/或b可以是其他的期数值)。Optionally, each of said specified risk ranges (RCAT) is limited to the next third predetermined duration (range is about 6 months, about 5-7 months, about 3-9 months, or according to expectation) probability of occurrence of selected number of MH events: F(x)=1-exp(-ax b ), x>0, otherwise 0, where: a≈-1.37, b≈1.35 (a and/or b can be other period values).
而且,指定了患者在将来发生显著低血糖症的危险分类。该分类的定义如下:最小危险,其中所述LBGI小于大约1.25;低危险,其中所述LBGI为大约1.25-大约2.50;中度危险,其中所述LBGI为大约2.50-大约5之间;和高危险,其中所述LBGI大于大约5(也能够根据期望实现其他的分类范围)。Furthermore, a risk category for developing significant hypoglycemia in the future was assigned to the patient. The categories are defined as follows: minimal risk, wherein the LBGI is less than about 1.25; low risk, wherein the LBGI is between about 1.25 and about 2.50; moderate risk, wherein the LBGI is between about 2.50 and about 5; and high Hazardous, wherein the LBGI is greater than about 5 (other classification ranges can also be implemented as desired).
III.严重低血糖症(SH)短期概率III. Short-term probability of severe hypoglycemia (SH)
本发明的再一个方面包括一种用于识别24小时内(或者其他选择时期)低血糖症高危期的方法、系统和计算机程序产品。这通过使用在先前24小时收集的SMBG读数计算低血糖症的短期危险而实现。在一个实施例中,本发明提供了一种计算机化方法和系统,用于根据在预定持续时间内收集的BG数据估计患者严重低血糖症(SH)的短期危险。根据在预定持续时间内收集的BG数据估计患者严重低血糖症(SH)短期危险的方法包括:根据所述收集的BG数据计算scale值;和为每个BG数据计算低BG危险值(RLO)。RLO(BG)的计算被数学定义为:Scale=[In(BG)]1.0845-5.381,其中BG是以mg/dl为单位测量的;Risk=22.765(Scale)2,如果(BG小于大约112.5),则:RLO(BG)=Risk,否则RLO(BG)=0。选择地,RLO(BG)的计算被数学地定义为:Scale=[In(BG)]1.026-1.861,其中BG是以mmol/l为单位测量的;Risk=32.184(Scale)2,如果(BG小于大约112.5),则:RLO(BG)=Rsik,否则RLO(BG)=0。Yet another aspect of the invention includes a method, system and computer program product for identifying periods of high risk for hypoglycemia within 24 hours (or other selected periods). This was achieved by calculating the short-term risk of hypoglycemia using SMBG readings collected over the previous 24 hours. In one embodiment, the present invention provides a computerized method and system for estimating a patient's short-term risk of severe hypoglycemia (SH) from BG data collected over a predetermined duration. A method for estimating a patient's short-term risk of severe hypoglycemia (SH) based on BG data collected within a predetermined duration includes: calculating a scale value based on said collected BG data; and calculating a risk of low BG (RLO) for each BG data . Calculation of RLO(BG) is defined mathematically as: Scale = [In(BG)] 1.0845 -5.381, where BG is measured in mg/dl; Risk = 22.765(Scale) 2 if (BG is less than approximately 112.5) , then: RLO(BG)=Risk, otherwise RLO(BG)=0. Optionally, the calculation of RLO(BG) is defined mathematically as: Scale=[In(BG)] 1.026-1.861 , where BG is measured in mmol/l; Risk=32.184(Scale) 2 , if (BG is less than approximately 112.5), then: RLO(BG)=Rsik, otherwise RLO(BG)=0.
根据所收集的BG数据能够计算LBGI。LBGI的计算由在时间点t1,t2,…,tn采集一系列BG读数x1,x2,…,xn数学地定义:LBGI can be calculated from the collected BG data. The calculation of LBGI is defined mathematically by taking a series of BG readings x1 , x2 , ..., xn at time points t1 , t2 , ..., tn :
根据已收集的BG数据能够计算临时LBGI。临时LBGI的计算被数学地定义为:Temporary LBGI can be calculated from the collected BG data. The calculation of temporary LBGI is defined mathematically as:
LBGI(1)=RLO(x1);RLO2(1)=0;LBGI(j)=((j-1)/j)*LBGI(j-1)+(1/j)*RLO(xj);和RLO2(j)=((j-1)/j)*RLO2(j-1)+(1/j)*RLO(xj)-LBGI(j))2。LBGI(1)=RLO(x1); RLO2(1)=0; LBGI(j)=((j-1)/j)*LBGI(j-1)+(1/j)*RLO(xj); and RLO2(j)=((j−1)/j)*RLO2(j−1)+(1/j)*RLO(xj)−LBGI(j)) 2 .
SBGI能够用如下定义的数学公式加以计算:SBGI can be calculated using the mathematical formula defined below:
接着,本发明提供了对即将发生的短期SH的认证(qualification)和报警。在下述情况下将进行认证和报警,如果:(LBGI(150)≥2.5且(LBGI(50)≥(1.5*LBGI(150)且SBGI(50)≥SBGI(150)),则确认或者发出所述报警,或者RLO≥(LBGI(150)+1.5*(SBGI(150)),则确认或者发出所述报警,否则不需要认证或者提供报警。Next, the present invention provides qualification and alert of impending short-term SH. Authentication and alarm will be carried out in the following cases, if: (LBGI(150)≥2.5 and (LBGI(50)≥(1.5*LBGI(150) and SBGI(50)≥SBGI(150)), then confirm or issue the If the above alarm, or RLO≥(LBGI(150)+1.5*(SBGI(150)), then confirm or issue the above alarm, otherwise no authentication or alarm is required.
接着选择地,本发明提供了对即将发生的短期SH的认证或报警。在下述情况下将进行认证和报警,如果:(LBGI(n)≥α且LBGI(n)ge(β)),则确认或者发出所述报警,和/或(RLO(n)≥(LBGI(n)+γ*SBGI(n))),则确认或者发出所述报警,否则不需要认证或者提供报警,其中α、β和γ是阈值参数。Optionally then, the present invention provides authentication or warning of impending short-term SH. Authentication and alarm will be performed in the following cases, if: (LBGI(n)≥α and LBGI(n)ge(β)), then confirm or issue the alarm, and/or (RLO(n)≥(LBGI( n)+γ*SBGI(n))), then confirm or issue the alarm, otherwise no authentication or alarm is provided, where α, β and γ are threshold parameters.
阈值参数α、β和γ定义为α≈5,β≈7.5,γ≈1.5。在下面的表格中给出了其他可能的参数组合。该数值可以与下面给出的值相近,也可以是下面表格中数值的任意中间组合。
IV.示例系统IV. Example System
本发明的方法可以用硬件、软件或者其组合加以实现,并且能够在一个或多个计算机系统或者其他处理系统中实现,例如个人数字助理(PDA),或者直接在具有足够存储和处理能力的血糖自我监测设备(SMBG存储仪表)中实现。在一个示例实施例中,本发明是在如图6所示的通用计算机900上运行的软件。计算机系统600包括一个或多个处理器,例如处理器604。处理器604连接通信基础设备606(例如,通信总线,交叉棒(cross-over bar)或者网络)。计算机系统600可以包括显示接口602,其从通信基础设备606(或者从未显示的帧缓冲器)传送图表、文本和其他数据用于在显示单元630上进行显示。The method of the present invention can be implemented in hardware, software or a combination thereof, and can be implemented in one or more computer systems or other processing systems, such as a Personal Digital Assistant (PDA), or directly on a blood glucose control system with sufficient storage and processing capabilities. Implemented in Self-Monitoring Device (SMBG Storage Meter). In an example embodiment, the present invention is software running on a general purpose computer 900 as shown in FIG. 6 .
计算机系统600还包括主存储器608,优选地随机存取存储器(RAM),并还可以包括一个二级存储器610。二级存储器610可以包括,例如,硬盘驱动612和/或可移动存储驱动器614,其代表软盘驱动、磁带驱动、光盘驱动、闪存等。可移动存储驱动614以众所周知的方式从和/或向可移动存储单元618读和/或写。可移动存储单元618代表软盘、磁带、光盘等,其由可移动存储驱动614读出和写入。可以意识到,可移动存储单元618包括计算机可用的存储介质,其中存储有计算机软件和/或数据。
在可选择实施例中,二级存储器610可以包括其他的装置,用于允许计算机程序或者其他指令装载到计算机系统600中。这类装置可以包括,例如,可移动存储单元622和接口620。这种可移动存储单元/接口的实例包括一个程序匣(cartridge)和匣接口(例如在视频游戏设备中见到的),可移动存储芯片(例如ROM、PROM、EPROM或者EEPROM)及相关插槽,和其他可移动存储单元622和接口,其允许插槽和数据从可移动存储单元62转移到计算机系统600。In alternative embodiments, secondary storage 610 may include other means for allowing computer programs or other instructions to be loaded into
计算机系统600还可以包括通信接口624。通信接口624允许软件和数据在计算机系统600和外部设备之间进行传递。通信接口624的实例包括一个调制解调器,一个网络接口(例如以太网卡),一个通信端口(例如串联或者并联等),PCMCIA插槽和卡、一个调制解调器等。通过通信接口624传递的软件和数据呈信号628的形式,其可以是电子、电磁、光或者其他能够被通信接口624接收的信号。信号628通过通信路径(也就是通道)626提供给通信接口624。通道626携带信号628,并且可以用导线或者电缆、光纤、电话线、便携式电话连接、RF连接、红外连接和其他通信通道加以实现。
在本文献中,术语“计算机程序介质”和“计算机可用介质”用于一般地指如下的介质,例如可移动存储驱动614,安装在硬盘驱动612中的硬盘,和信号628。这些计算机程序产品是用于提供软件到计算机系统600的装置。本发明包括这种计算机程序产品。In this document, the terms “computer program medium” and “computer usable medium” are used to refer generally to media such as removable storage drive 614 , hard disk installed in
计算机程序(也称作计算机控制逻辑)被存储在主存储器608和/或二级存储器610中。计算机程序还可以通过通信接口624被接收。这种计算机程序在执行时能够使计算机系统600执行如下文中讨论的本发明的特征。特别地,计算机程序在被执行时使处理器604能够执行本发明的功能。因此,这种计算机程序代表了计算机系统600的控制器。Computer programs (also referred to as computer control logic) are stored in
在用软件实现本发明的一个实施例中,软件可以存储在计算机程序产品中,并用可移动存储驱动器614、硬盘驱动612或者通信接口624装载到计算机系统600中。控制逻辑(软件)在被处理器604执行时使处理器604执行如下文所述的本发明的功能。In one embodiment where the invention is implemented in software, the software may be stored in a computer program product and loaded into
在另一个实施例中,本发明主要以硬件形式实现,其使用例如硬件组件如设备专用集成电路(application specific integrated circuit)(ASIC)。执行硬件状态机器从而执行这里说明的功能,对于相关领域的技术人员而言是显而易见的。In another embodiment, the invention is implemented primarily in hardware using, for example, hardware components such as application specific integrated circuits (ASICs). It will be apparent to those skilled in the relevant arts to implement a hardware state machine to perform the functions described herein.
在另一个实施例中,本发明使用硬件与软件的组合加以实现。In another embodiment, the invention is implemented using a combination of hardware and software.
在本发明的一个示例软件实施例中,上述的方法用SPSS控制语言实现,但也能够用其他的程序实现,例如,但不仅限于,C++程序语言或者其他本领域技术人员能够获得的程序。In an example software embodiment of the present invention, above-mentioned method is realized with SPSS control language, but also can realize with other programs, for example, but not limited to, C++ programming language or other programs that those skilled in the art can obtain.
图7-9显示了代表本发明可选择实施例的框图。参考图7,显示了代表系统710的框图,其主要包括由患者712使用的葡萄糖计728用于记录,尤其是,胰岛素剂量读数和测量的血糖(“BG”)水平。由葡萄糖计728获得的数据优选地通过合适的通信连接714或者数据调制解调器732转移到处理站或者芯片,例如个人计算机740、PDA,或者通过便携式电话或通过合适的互连网端口。例如,所存储的数据可以存储在葡萄糖计728中并通过合适的接口电缆直接下载到个人计算机,然后通过互连网发送到处理站。一个实例是由LifeScan有限公司生产的ONE TOUCH监测系统或测量计,其与IN TOUCH软件兼容,并包括接口电缆以便下载到个人计算机。7-9 show block diagrams representing alternative embodiments of the present invention. Referring to FIG. 7 , there is shown a block diagram representative of a system 710 that primarily includes a glucose meter 728 used by a patient 712 for recording, inter alia, insulin dose readings and measured blood glucose ("BG") levels. Data obtained by the glucose meter 728 is preferably transferred via a suitable communication link 714 or data modem 732 to a processing station or chip such as a personal computer 740, PDA, or via a cellular phone or via a suitable internet port. For example, the stored data can be stored in the glucose meter 728 and downloaded directly to a personal computer via a suitable interface cable, and then sent via the Internet to a processing station. An example is the ONE TOUCH monitoring system or meter produced by LifeScan Ltd, which is compatible with INTOUCH software and includes an interface cable for downloading to a personal computer.
血糖计是行业中通用的,并且主要包括任何能够起BG获取机构功能的设备。BG计或者获取机构、设备、工具或者系统包括各种用于为每次测试提取血样(例如通过刺手指)的传统方法,并包括使用通过电机或者claorimetric方法读出葡萄糖浓度的装置确定葡萄糖的水平。最近,开发出了多种无需取血便可确定血液分析物浓度的方法。例如授予Yang等的美国专利No.5,267,152(本文引用作为参考),说明了一种利用近IR辐射漫反射激光光谱测量血糖浓度的非侵入技术。类似的近IR光谱设备在授予Rosenthal等的美国专利No.5,086,229,和授予Robinson等的美国专利No.4,975,581中有所说明(本文引用作为参考)。Glucose meters are common in the industry and generally include any device capable of functioning as a BG acquisition mechanism. BG meters or acquisition mechanisms, devices, tools, or systems include various traditional methods for drawing blood samples for each test (such as by finger pricks) and include determining glucose levels using devices that read out glucose concentrations by electromechanical or claorimetric methods . Recently, various methods have been developed to determine blood analyte concentrations without drawing blood. For example, US Patent No. 5,267,152 to Yang et al., incorporated herein by reference, describes a non-invasive technique for measuring blood glucose concentration using diffuse reflectance laser spectroscopy of near IR radiation. Similar near-IR spectroscopy devices are described in US Patent No. 5,086,229 to Rosenthal et al., and US Patent No. 4,975,581 to Robinson et al. (incorporated herein by reference).
授予Stanley的美国专利No.5,139,023(本文引用作为参考)说明了一种经皮血糖监测装置,其依靠渗透性增强器(例如胆汁盐)方便葡萄糖沿着在间隙液和接收介质之间建立的浓度梯度经皮运动。授予Sembrowich的美国专利No.5,036,861(本文引用作为参考)说明了一种被动葡萄糖监测器,其通过皮肤贴收集汗液,其中使用类胆碱能的制剂刺激汗腺分泌汗液。类似的汗液收集设备在授予Schoendorfer的美国专利No.5,076,273和授予Schroeder的美国专利No.5,140,985中有说明(本文引用作为参考)。U.S. Patent No. 5,139,023 to Stanley (incorporated herein by reference) describes a transcutaneous glucose monitoring device that relies on permeability enhancers (such as bile salts) to facilitate glucose along the concentration established between the interstitial fluid and the receiving medium. Gradient percutaneous motion. US Patent No. 5,036,861 to Sembrowich (incorporated herein by reference) describes a passive glucose monitor that collects sweat via a skin patch in which the sweat glands are stimulated to secrete sweat using a cholinergic agent. Similar sweat collection devices are described in US Patent No. 5,076,273 to Schoendorfer and US Patent No. 5,140,985 to Schroeder (incorporated herein by reference).
此外,授予Glikfeld的美国专利No.5,279,543(本文引用作为参考)说明了使用离子电渗疗法非侵入地将物质通过皮肤采样到皮肤表面的容器中。Glikfeld讲解了,该采样过程能够与葡萄糖特异的生物传感器或者葡萄糖特异的电极耦合从而监测血糖。而且,授予Tamada的国际公开No.WO 96/00110(本文引用作为参考)说明了一种用于经皮监测目标物质的离子电渗疗装置,其中离子电渗电极用于将分析物移动进入收集器中,并使用生物传感器探测容器中的目标分析物。最后,授予Berner的美国专利No.6,144,869(本文引用作为参考)说明了一种用于测量所存在分析物的浓度的采样系统。Additionally, US Patent No. 5,279,543 to Glikfeld, incorporated herein by reference, describes the use of iontophoresis to non-invasively sample substances through the skin into a reservoir on the skin surface. Glikfeld explained that the sampling process can be coupled with glucose-specific biosensors or glucose-specific electrodes to monitor blood glucose. Also, International Publication No. WO 96/00110 to Tamada (incorporated herein by reference) describes an iontophoretic device for transdermal monitoring of a target substance in which iontophoretic electrodes are used to move the analyte into a collection container, and use the biosensor to detect the target analyte in the container. Finally, US Patent No. 6,144,869 to Berner, incorporated herein by reference, describes a sampling system for measuring the concentration of an analyte present.
进一步,BG计或者获取结构还可以包括内在导管和皮下组织液采样。Further, the BG meter or acquisition structure may also include an internal catheter and sampling of subcutaneous tissue fluid.
计算机或者PDA 740包括软件和硬件,它们是根据预定的流程顺序(如上面详细说明的)处理、分析和判读自我记录的糖尿病患者数据并产生合适的数据判读输出所必需的。优选地,根据由计算机740存储的患者数据执行的数据分析和判读的结果通过与个人计算机740相连的打印机生成报表加以显示。选择地,数据判读程序的结果可以直接显示在与计算机740相连的视频显示单元上。The computer or PDA 740 includes the software and hardware necessary to process, analyze and interpret the self-recorded diabetic patient data according to a predetermined sequence of procedures (as detailed above) and to produce an appropriate data interpretation output. Preferably, the results of the data analysis and interpretation performed based on the patient data stored by the computer 740 are displayed through a printer connected to the personal computer 740 to generate a report. Alternatively, the results of the data interpretation program can be displayed directly on a video display unit connected to the computer 740 .
图8显示的框图代表了具有糖尿病管理系统的可选择实施例,其是由患者操作的装置810,其外壳(housing)优选地足够紧凑从而使装置810能够用手持并由患者携带。在外壳816的表面上装载有一个用于接收血糖测试条(未显示)的条导向装置。测试条用于接收来自患者的血液样本。该装置包括一个微处理器822和与微处理器相连的存储器824。微处理器822被设计成执行存储在存储器824中的计算机程序从而执行各种计算和控制功能,如上面详细说明的。袖珍键盘816通过标准袖珍键盘解码器826与微处理器822相连。显示器814通过显示驱动器830与微处理器822相连。微处理器822通过接口与显示驱动器830通信,并且显示驱动器830在微处理器822的控制下修改和刷新显示器814。扬声器854和时钟856也连接微处理器822。扬声器854在微处理器822的控制下工作,从而发出可听的语音,使患者对将来可能的低血糖症警觉。时钟856向微处理器822提供当前的日期和时间。The block diagram shown in Figure 8 represents an alternative embodiment having a diabetes management system with a device 810 operated by the patient, the housing of which is preferably compact enough that the device 810 can be hand-held and carried by the patient. On the surface of
存储器824也存储患者812的血糖值、胰岛素剂量值、胰岛素类型和微处理器822使用的参数值,用以计算将来血糖值、补充胰岛素剂量和碳水化合物补充。每个血糖值和胰岛素剂量值都与相应的日期和时间一起存储在存储器824中,存储器824优选地是非易失性存储器,例如电擦除只读存储器(EEPROM)。Memory 824 also stores
装置810还包括一个与微处理器822相连的血糖计828。葡萄糖计828被设计用于测量被接收在血糖测试条上的血液样品,并产生血液样品测量的血糖值。如前所述,这种葡萄糖计在本领域是熟知的。葡萄糖计828的类型优选地是产生直接输出到微处理器822的数字值。选择地,血糖计828的类型可以是产生模拟值。在该选择实施例中,血糖计828通过模拟-数字转换器(未显示)与微处理器822相连。Device 810 also includes a blood glucose meter 828 coupled to microprocessor 822 . Glucose meter 828 is designed to measure a blood sample received on a blood glucose test strip and generate a blood glucose value measured by the blood sample. As previously mentioned, such glucose meters are well known in the art. Glucose meter 828 is preferably of the type that produces a digital value that is output directly to microprocessor 822 . Alternatively, blood glucose meter 828 may be of the type that generates analog values. In the alternative embodiment, blood glucose meter 828 is connected to microprocessor 822 through an analog-to-digital converter (not shown).
装置810进一步包括一个输入/输出端口834,优选地,包括一系列连接微处理器822的端口。端口834通过接口,优选地,通过标准RS232接口连接调制解调器832。调制解调器832用于通过通信网络836在装置810与个人计算机840,或者康护人员的计算机838,之间建立通信。通过连接电缆连接电子设备的专用技术在本领域是熟知的。另一种可选择实例是“蓝牙”技术通信。Device 810 further includes an input/output port 834 , preferably comprising a series of ports connected to microprocessor 822 . Port 834 interfaces with modem 832, preferably via a standard RS232 interface. A modem 832 is used to establish communication between the device 810 and a personal computer 840 , or a healthcare provider's
选择地,图9显示的框图代表了具有糖尿病管理系统的可选择实施例,其是由患者操作的装置910,与图8所示类似,其外壳(housing)优选地足够紧凑从而使装置910能够用手持并由患者携带。例如,分离式或者可差卸式葡萄糖计或者BG获取机构/模块928。早已经有了自我监测设备,其能够直接计算算法1、2、3并向患者显示结果而无需将数据发送给其他事物。这种设备的实例是Lifescan有限公司生产的ULTRA SMART,Therasense,Alameda,CA生产的Milpitas,CA和FREESTYLE TRACKER。Optionally, the block diagram shown in FIG. 9 represents an alternative embodiment having a diabetes management system, which is a device 910 operated by a patient, similar to that shown in FIG. Hand-held and carried by the patient. For example, a separate or differential glucose meter or BG acquisition mechanism/
因此,这里说明的实施例能够在数据通信网络例如互连网上实现,使得任何遥远位置的任何处理器和计算机都能够获得该评价、估计和信息,如图6-9和/或授予Wood的美国专利No.5,851,186中描述的,本文引用其内容作为参考。选择地,遥远位置处的患者可以将BG数据发送到中心康护人员或者医疗所或者不同的遥远位置。Accordingly, the embodiments described herein can be implemented over a data communications network, such as the Internet, so that any processor and computer at any remote location can obtain the evaluations, estimates, and information, as shown in FIGS. 6-9 and/or in U.S. Patents to Wood. No. 5,851,186, the contents of which are incorporated herein by reference. Optionally, a patient at a remote location can send BG data to a central healthcare provider or clinic or a different remote location.
总之,本发明提出了一种计算机化的(或者非计算机化的)数据分析方法和系统,用于同时估计糖尿病个体血糖控制中的两个最重要的分量:HbA1c和低血糖症危险。本方法尽管只使用日常的SMBG数据,但是除了其他的之外,提供了三组输出。In summary, the present invention proposes a computerized (or non-computerized) data analysis method and system for simultaneously estimating the two most important components in glycemic control in diabetic individuals: HbA 1c and risk of hypoglycemia. This method, although only using daily SMBG data, provides three sets of outputs, among others.
本发明方法、系统和计算机程序产品的潜力在于提供了如下的优点,但不仅限于此。首先,本发明通过执行和显示如下的内容提高了现有家用BG监测设备的性能:1)估计HbA1c的范围,2)估计在随后6个月内SH的概率,和3)估计低血糖症的短期危险(也就是,今后24小时)。后者可以包括警告,例如报警,表明即将发生低血糖症事件。这三个部分还能够组合在一起,从而提供有关糖尿病个体血糖控制的连续信息,进而提高对低血糖症危险的监测。The method, system and computer program product of the present invention have the potential to provide the following advantages, but are not limited thereto. First, the present invention improves the performance of existing home BG monitoring devices by performing and displaying: 1) estimating the range of HbA 1c , 2) estimating the probability of SH in the following 6 months, and 3) estimating hypoglycemia short-term risk (that is, the next 24 hours). The latter may include warnings, such as alarms, that a hypoglycemic event is imminent. These three components can also be combined to provide continuous information on the glycemic control of diabetic individuals, thereby improving the monitoring of the risk of hypoglycemia.
作为附加的优点,本发明提高了现有SMBG数据检索软件或者硬件的性能。几乎每个家用BG监测设备的制造商都生产这种软件或者硬件,并且患者和康护人员通常使用它判读SMBG数据。本发明的方法和系统能够直接合并到现有家用血糖监测器中,或者通过引入能够同时预测HbA1c和低血糖症高危期的数据判读组件提高SMBG数据检索软件的性能。As an added advantage, the present invention improves the performance of existing SMBG data retrieval software or hardware. Almost every manufacturer of home BG monitoring equipment produces this software or hardware, and it is commonly used by patients and healthcare professionals to interpret SMBG data. The method and system of the present invention can be directly incorporated into existing household blood glucose monitors, or the performance of SMBG data retrieval software can be improved by introducing data interpretation components capable of simultaneously predicting HbA 1c and high-risk periods of hypoglycemia.
另外一个优点是,本发明能够同时在低和高BG范围内以及在BG的整个数值范围内对家用BG监测设备的精确性进行评估。An additional advantage is that the present invention enables the assessment of the accuracy of home BG monitoring devices simultaneously in the low and high BG ranges and across the entire range of BG values.
而且,另一个优点,本发明能够评估各种糖尿病疗法的有效性。Also, as another advantage, the present invention enables the assessment of the effectiveness of various diabetes therapies.
再进一步,因为糖尿病患者终生面临着在保持严格血糖控制的同时而不使其低血糖症危险增加的优化问题,所以本发明通过使用其简单而可靠的方法缓和了这一相关问题,也就是,本发明能够同时评估患者的血糖控制及其低血糖症的危险,同时能够在患者的日常条件下加以应用。Still further, since diabetics face lifelong optimization problems of maintaining tight glycemic control without increasing their risk of hypoglycemia, the present invention alleviates this related problem by using its simple and reliable method, namely, The present invention enables simultaneous assessment of a patient's glycemic control and his risk of hypoglycemia, while being applicable in the patient's day-to-day conditions.
另外,本发明通过提出三种截然不同但相互兼容的算法给出了缺失的联系,该三种算法均用于由SMBG数据估计HbA1c和低血糖症的危险,从而用于预测低血糖症的短期和长期危险以及高血糖症的长期危险。Additionally, the present invention provides the missing link by proposing three distinct but mutually compatible algorithms for estimating HbA 1c and the risk of hypoglycemia from SMBG data, and thus for predicting the risk of hypoglycemia. Short-term and long-term dangers and long-term dangers of hyperglycemia.
另一个优点,本发明能够评估新型胰岛素或者胰岛素投放(delivery)设备的有效性。任何胰岛素或胰岛素投放设备的制造商或者研究人员都能够利用本发明的实施例测试其提出或者检测的胰岛素类型或者设备投放设计的相对成功性。As another advantage, the present invention enables the evaluation of the effectiveness of new insulins or insulin delivery devices. Any manufacturer or researcher of insulin or insulin delivery devices can utilize embodiments of the present invention to test the relative success of their proposed or tested insulin types or device delivery designs.
最后,另一个优点,本发明能够评估胰岛素辅助治疗药物的有效性。Finally, as another advantage, the present invention enables the assessment of the effectiveness of insulin adjuvant therapy drugs.
发明实例Invention example
I.实例No.1I. Example No.1
实例No.1包括三种算法,用于由日常SMBG数据同时估计糖尿病血糖控制中的两个最重要的分量,HbA1c和低血糖症危险。该方法直接适合于通过引入能够同时预测HbA1c和低血糖症高危期的智能数据判读组件提高现有家用BG监测设备的性能。该数据分析方法具有三个部分(算法):Example No. 1 includes three algorithms for simultaneously estimating the two most important components in diabetic glycemic control, HbA 1c and hypoglycemia risk, from daily SMBG data. This approach is directly suitable for improving the performance of existing home BG monitoring devices by introducing an intelligent data interpretation component capable of simultaneously predicting HbA 1c and high-risk periods of hypoglycemia. This data analysis method has three parts (algorithms):
●算法1:估计HbA1c ●Algorithm 1: Estimation of HbA 1c
●算法2:估计严重低血糖症(SH)的长期危险,和Algorithm 2: Estimation of the long-term risk of severe hypoglycemia (SH), and
●算法3:估计低血糖症的短期(24-48小时内)危险。• Algorithm 3: Estimate short-term (within 24-48 hours) risk of hypoglycemia.
算法1和2提供了有关1型或者2型糖尿病(T1DM,T2DM)个体整体血糖控制的不间断监测和信息,同时覆盖BG数值范围的上限和下限。当算法2指示低血糖症的长期危险增加时,算法3被激活。一旦激活,算法3要求更加频繁的监测(每天4次),并提供对中度/严重低血糖症的24-48小时预报。
实例1的另一个重要目标是用现有数据检验大量的假说和观点,这些假说和观点有可能产生其他其他的算法,它们以概念上不同于本发明公开所提供的方式估计HbA1c和计算低血糖症的危险。目标是找到潜在的更好的解决方法,或者简单地证实某些观点不能够产生更好的结果,其主要用于优化和改进对目前在本研究实例No.2中收集的数据的分析。Another important goal of Example 1 was to use existing data to test a large number of hypotheses and ideas that could potentially lead to other algorithms that estimate HbA 1c and calculate low The danger of blood sugar. The goal is to find potentially better solutions, or simply to prove that some ideas are not able to produce better results, which is mainly used to optimize and improve the analysis of the data currently collected in this research example No.2.
数据组(data sets) data sets
为了保证我们的最优化结果能够推广到更广泛的水平,算法1和2首先用训练数据组加以优化,然后用不相关的测试数据组检验其精度。对于算法3,我们目前只有一组含有平行SMBG和SH记录的数据。患者人口调查(population)的详细说明如下:In order to ensure that our optimization results can be generalized to a wider level,
(1)
训练数据组1:96名患者至少在本研究2年之前被诊断患有TIDM。其中有43名患者报告在过去一年里有过至少2次严重低血糖症事件,53名患者报告在相同的时间里没有这种事件。其中有38名男性和58名妇女。平均年龄为35±8岁,糖尿病的平均持续时间为16±10年,每天的平均胰岛素为0.58±0.19单位/kg,平均HbA1c为8.6±1.8%。这些对象在40-45天的时间内收集了大约13,000个SMBG读数。SMBG的频率为大约3个读数/天。这些数据收集继续进行了6个月,按月记录中度和严重低血糖症事件。该数据组用作算法1(无先前HbA1c)和算法2的训练数据组。(1) Training data set 1 : 96 patients were diagnosed with TIDM at least 2 years before this study. Of these, 43 patients reported at least 2 severe hypoglycemic events in the past year, and 53 patients reported no such events during the same time period. Among them were 38 men and 58 women. The mean age was 35±8 years, the mean duration of diabetes was 16±10 years, the mean daily insulin was 0.58±0.19 units/kg, and the mean HbA 1c was 8.6±1.8%. Approximately 13,000 SMBG readings were collected from these subjects over a period of 40-45 days. The frequency of SMBG was approximately 3 readings/day. These data collections continued for 6 months, with moderate and severe hypoglycemia events recorded monthly. This data set was used as the training data set for Algorithm 1 (without prior HbA 1c ) and
(2)
训练数据组:85名患者至少在本研究2年之前被诊断患有TIDM,所有患者均报告在过去一年里有SH事件。其中有44名男性和41名妇女。平均年龄为44±10岁,糖尿病的平均持续时间为26±11年,每天的平均胰岛素为0.6±0.2单位/kg,平均基准HbA1c为7.7±1.1%,且平均6个月HbA1c为7.4±1%(60名对象具有6个月HbA1c)。这些对象在两次HbA1c化验之间的6个月内收集了大约75,500个SMBG读数。数据组2中的SMBG频率更高,为4-5个读数/天。此外,在SMBG的6个月期间,这些对象保持记录中度和严重低血糖症事件及其发生的日期和时间,结果有399次SH事件。该数据组用作算法1(有先前HbA1c)和所有与算法3有关的分析的训练数据组。(2) Training data set : 85 patients were diagnosed with TIDM at least 2 years before this study, and all patients reported SH events in the past year. Among them were 44 men and 41 women. Mean age was 44±10 years, mean duration of diabetes was 26±11 years, mean daily insulin was 0.6±0.2 units/kg, mean baseline HbA 1c was 7.7±1.1%, and mean 6-month HbA 1c was 7.4 ±1% (60 subjects with 6-month HbA 1c ). These subjects had approximately 75,500 SMBG readings collected in the 6-month period between two HbA 1c assays. The frequency of SMBG in
(3) 测试数据组:我们使用N=600个对象的数据,其中277名具有T1DM,323名具有T2DM,他们全部使用胰岛素治疗糖尿病。这些数据由美国中部圣地亚哥Amylin公司的药剂师收集,包括6-8个月的SMBG数据(大约300,000个读数),同时具有基准和6个月HbA1c结果以及一些人口统计数据。这些对象参加了pramlintide(剂量为60-120微克)对代谢控制的临床实验调查。T1DM和T2DM组中的对象使用pramlintide是随机的(表1)。(3) Test data set : We used the data of N=600 subjects, among which 277 had T1DM and 323 had T2DM, all of them were treated with insulin for diabetes. The data were collected by Amylin's pharmacists in San Diego, Central USA, and consisted of 6-8 months of SMBG data (approximately 300,000 readings), with both baseline and 6-month HbA 1c results and some demographic data. These subjects participated in a clinical trial investigating the metabolic control of pramlintide (dose 60-120 micrograms). Subjects in the T1DM and T2DM groups were randomized to pramlintide (Table 1).
表1:测试数据组中对象的人口统计特征Table 1: Demographic characteristics of subjects in the test data set
表1给出了人口统计特征以及T1DM与T2DM对象的比较。在研究最初的6个月,T1DM和T2DM组的平均HbA1c都显著下降,这可能是由于药物的使用,这超出了本报告书(presentation)(表1)的范围。HbA1c相对快速的变化允许更好地评估算法1的预测能力。在所有数据组中,SMBG用Lifescan公司的ONE TOUCH II或ONETOUCH PROFILE测量计执行。Table 1 presents demographic characteristics and a comparison of T1DM versus T2DM subjects. Mean HbA 1c decreased significantly in both T1DM and T2DM groups during the first 6 months of the study, possibly due to medication use, which is beyond the scope of this presentation (Table 1). The relatively rapid change in HbA 1c allowed for a better assessment of the predictive power of
算法1:估计HbA1c Algorithm 1: Estimating HbA 1c
实例No.1给出了,但不仅限于,HbA1c预测(算法1)的优化,其通过:(1)SMBG越接近中心,权重越高;(2)高BG事件时间越长,权重越高;(3)用较早前的HbA1c校正高BG指数;和(4)合并其他的患者变量,例如年龄、性别和疾病持续时间。Example No.1 presents, but is not limited to, the optimization of HbA 1c prediction (Algorithm 1) by: (1) the closer the SMBG is to the center, the higher the weight; (2) the longer the high BG event, the higher the weight ; (3) correcting for a high BG index with an earlier HbA 1c ; and (4) incorporating other patient variables such as age, sex, and disease duration.
算法1包括SMBG数据的优化函数,其估计随后的HbA1c,并推荐数据收集周期的最佳持续时间和该周期内自我监测的最佳频率。然而要非常注意,算法1的更广泛目的是评估患者血糖控制的状态。尽管HbA1c被接受作为评估血糖控制的“黄金标准”,但是目前并不清楚其他的测量量,例如平均SMBG或者高BG指数是不是比HbA1c更好的糖尿病长期并发症的预报因子。在该问题被澄清之前,算法1的目的都是估计HbA1c。为了尽可能地接近将来算法1的实际应用,我们如下进行:
(1)首先,由我们在先前对T1DM患者的研究中收集的两个训练数据组1和2推导出多个使用不同独立变量的优化函数,最佳持续时间和最佳SMBG频率;(1) First, multiple optimization functions using different independent variables, optimal duration and optimal SMBG frequency, were derived from two
(2)然后,固定所有的系数,将算法1应用于大得多的测试数据组,该测试数据组同时包括在由Amylin药剂师指导的临床实验中在非常不同的条件下收集的T1DM和T2DM对象的数据;(2) Then, with all coefficients fixed,
(3)仅用测试数据组详细评估算法1对各种优化函数的精度。(3) The accuracy of
训练和测试数据组分离允许我们声明,算法1的估计精度能够一般化到T1DM或者T2DM对象任何其他的数据。而且,因为Amylin数据(测试数据组)是从正接受治疗以降低其HbA1c的对象中收集的,因此他们的HbA1c在6个月的观察周期内显示出了不正常的大变异,我们能够声明,算法1不仅能够预测相对恒定的HbA1c,而且能够预测大而不正常快速变化的HbA1c。在相同的行中,算法1对于希望优化其HbA1c的患者是最有用的,可以推测,该患者群体很可能对于寻求具有先进特征的测量计,例如连续估计HbA1c,感兴趣。The separation of training and test data sets allows us to state that the estimated accuracy of
结果摘要Summary of Results
● 最佳SMBG数据收集周期为45天;● The optimal SMBG data collection cycle is 45 days;
● 最佳SMBG频率是3次读数/天;● Optimum SMBG frequency is 3 readings/day;
● 提出了两个最佳HbA1c估计函数:F1-只使用SMBG数据,和F2-使用SMBG数据加上在所预测HbA1c之前大约6个月获得的HbA1c读数;● Two optimal HbA 1c estimation functions are proposed: F1 - using only SMBG data, and F2 - using SMBG data plus HbA 1c readings obtained approximately 6 months before the predicted HbA 1c ;
● 通过在下页(表2)中详细说明的多个标准对测试数据组(N=573个对象)中HbA1c预测的精度进行评估。我们在这里说明,在T1DM中,F1的总精度(HbA1c测量值的20%之内)为96.5%,而F2的总精度为95.7%。对于T2DM,F1的总精度为95.9%,F2的总精度为98.4%。因此,F1和F2的精度均与HbA1c的直接测量结果相当;• The accuracy of HbA 1c prediction in the test data set (N=573 subjects) was assessed by a number of criteria detailed on the next page (Table 2). We show here that in T1DM, the overall accuracy (within 20% of the HbA 1c measurement) was 96.5% for F1 and 95.7% for F2. For T2DM, the overall accuracy is 95.9% for F1 and 98.4% for F2. Therefore, the accuracy of both F1 and F2 is comparable to the direct measurement of HbA 1c ;
● 最重要的,对于HbA1c从其基准读数改变2个或者更多个单位的患者(N=68),F1无论对于T1DM还是T2DM,预测该变化的精度均为100%,而F2对于T1DM和T2DM的精度分别为71%和85%;● Most importantly, for patients whose HbA 1c changed by 2 or more units from their baseline reading (N=68), F1 predicted this change with 100% accuracy for both T1DM and T2DM, while F2 for T1DM and The accuracy of T2DM is 71% and 85%, respectively;
● F1和F2对于6个月HbA1c的预测都比0月时的最初HbA1c估计要精确得多。但使用平均BG作为HbA1c的直接估计并不准确;● Both F1 and F2 were much more accurate in predicting 6-month HbA 1c than the original HbA 1c estimate at 0-month. But using mean BG as a direct estimate of HbA 1c is not accurate;
● 试验了大量可选择的方法,例如选择每天的特殊时间点(餐后读数)估计HbA1c,根据每次SMBG读数和HbA1c确定之间间隔的时间给SMBG读数不同的权重,对具有不同平均血糖/HbA1c比值的对象分别进行评估等。尽管这些可选择方法中有一些获得了比上面提出的两个函数好一些的结果,但在整体上没有一个更好。我们能够得出结论:在算法1的将来应用中将使用优化函数F1和F2。● Experimented with a number of alternative methods, such as choosing specific time points of the day (postprandial readings) to estimate HbA 1c , giving different weights to SMBG readings according to the time interval between each SMBG reading and HbA 1c determination, and to The blood glucose/HbA 1c ratio was assessed separately among the subjects. Although some of these alternatives achieve somewhat better results than the two functions proposed above, none is better overall. We can conclude that in future applications of
详细结果-测试数据组Detailed Results - Test Dataset
算法1评估最重要的部分是评估它对于如下数据的性能,即与用于提出并优化它的数据不相关的数据。由测试数据组,573个对象的数据,其中T1DM的N=254,T2DM的N=319,足以用于评估算法1。The most important part of the evaluation of
最佳算法1:对于每个对象,选择了他/她SMBG读数45天的子集。该子集的开始日期大约在对象6个月HbA1c化验之前的75天,结束日期大约为该化验前的30天。因为在该数据组中,HbA1c化验的时间只是大概地知道,所以在上一次用于分析的SMBG读数与HbA1c之间间隔的时间并不确切。该时间周期是通过连续优化其持续时间和其结束点(HbA1c之前的时间)选择出来的。最佳持续时间为45天。最佳结束时间为HbA1c之前1个月。换言之,45天的SMBG能够提前大约1个月预测HbA1c的数值。然而,预测45-75天之间任何一个HbA1c数值都几乎一样好——差别只是数字而非临床的显著度。类似地,45天的监测周期与60天的监测周期之间的差异并不大。但是,小于45天的监测周期会使得预测能力快速下降。 Optimal Algorithm 1 : For each subject, a 45-day subset of his/her SMBG readings was selected. The start date of the subset is approximately 75 days prior to the subject's 6-month HbA 1c assay and the end date is approximately 30 days prior to the assay. Because the time of HbA 1c assay was only known approximately in this data set, the time interval between the last SMBG reading used for analysis and HbA 1c was not exact. This time period was chosen by continuously optimizing its duration and its end point (time before HbA 1c ). The optimal duration is 45 days. The best end time is 1 month before HbA 1c . In other words, 45-day SMBG was able to predict HbA 1c values about 1 month in advance. However, predicting any HbA 1c value between 45 and 75 days was almost as good - the difference was numerical and not clinically significant. Similarly, the difference between the 45-day monitoring period and the 60-day monitoring period was not large. However, the monitoring period of less than 45 days will make the prediction ability decline rapidly.
最佳估计函数是线性的,并且由下给出:The best estimate function is linear and is given by:
估计1——不知道先前的HbA1c: Estimate 1 - Unknown previous HbA 1c :
F1=0.809098*BGMM1+0.064540*LBGI1-0.151673*RHI1+1.873325F1=0.809098*BGMM1+0.064540*LBGI1-0.151673*RHI1+1.873325
估计2——已知先前的HbA1c(大约6个月之前) Estimate 2 - known prior HbA 1c (approximately 6 months ago)
F2=0.682742*HBA0+0.054377*RHI1+1.553277F2=0.682742*HBA0+0.054377*RHI1+1.553277
在这些公式中,BGMM1是由45天的SMBG读数计算出的平均血糖;LBGI1和RHI1是由相同读数计算出的低和高BG指数,HBA0是只用于估计2的基准HbA1c读数。系数值用训练数据组进行优化,并且在详细结果——训练数据组部分中给出了相关的统计和绘图。In these formulas, BGMM1 is the average blood glucose calculated from 45 days of SMBG readings; LBGI1 and RHI1 are the low and high BG indices calculated from the same readings, and HBA0 is the baseline HbA 1c reading used only to estimate 2. Coefficient values were optimized using the training data set, and relevant statistics and plots are given in the Detailed Results - Training Data Set section.
函数F1和F2产生了HbA1c的点估计,也就是说,每个函数产生了一个HbA1c估计。利用详细结果——训练数据组部分中提供的回归误差估计能够获得区间估计。然而,对于测试数据组,这些区间估计并不是HbA1c真正的90%或者95%置信区间,因为它们最初得自于训练数据组并且只是应用于测试数据(见下一部分的统计注意)。Functions F1 and F2 produced point estimates of HbA 1c , that is, each function produced an estimate of HbA 1c . Interval estimates can be obtained using the regression error estimates provided in the Detailed Results - Training Dataset section. However, for the test data set, these interval estimates are not true 90% or 95% confidence intervals for HbA 1c because they were originally derived from the training data set and were only applied to the test data (see Statistical Notes in the next section).
算法1精度评估:表2A和2B给出了分别用T1DM和T2DM对象的测试数据组数据评估最佳算法1的结果。使用了多个标准:
(1)从HbA1c测量值估计的绝对偏差(AERR);(1) Estimated absolute deviation (AERR) from HbA 1c measurements;
(2)从HbA1c测量值估计的绝对百分比偏差(PERR);(2) Estimated absolute percentage deviation (PERR) from HbA 1c measurements;
(3)HbA1c测量值20%内的百分比估计(HIT20);(3) Percentage estimate within 20% of HbA 1c measurement (HIT20);
(4)HbA1c测量值10%内的百分比读数(HIT10);和(4) Percent reading within 10% of HbA 1c measurement (HIT10); and
(5)HbA1c测量值周围25%区间之外的百分比读数(MISS 25)。(5) Percent readings outside the 25% interval around the HbA 1c measurement (MISS 25).
表2A:算法1在T1DM中的精度(N=254个对象)Table 2A: Accuracy of
表2B:算法1在T2DM中的精度(N=319个对象)Table 2B: Accuracy of
表2A和2B的前两列分别给出了优化函数F1和F2的结果。第三行给出了当采用平均BG(mmol/L)作为HbA1c估计时估计的精度。第四行给出了用0时的HbA1c化验作为6个月HbA1c估计计算的相同精度测量结果。显然,对于T1DM和T2DM,F2对于估计HbA1c在整体上都比F1略好。最重要的是,F1和F2对于HbA1c的估计都比用其较早数值或者平均BG进行估计要好得多。对于落在25%精度区间之外的%估计,这特别地真。执行F1和F2与由先前HbA1c化验进行估计的差异是高度显著的(第4列)。The first two columns of Tables 2A and 2B give the results of optimizing functions F1 and F2, respectively. The third row gives the estimated precision when using mean BG (mmol/L) as the HbA 1c estimate. The fourth row gives the same precision measure calculated using the HbA 1c assay at
统计注意:要重点注意到,使用传统的回归型标准,例如从ANOVA表中获得的R2或F和p值,评估算法1的精度是不合适的。这是因为,参数估计得自于另一个不相关的数据组(训练数据),而只是应用于该测试数据组。因此,基本(underlying)模型的统计假定被违反了(例如在测试数据组中,残差的总和不是零),因此R2或F和p值失去了它们的统计学意义。 Statistical Note : It is important to note that it is not appropriate to assess the precision of
通过检查SMBG基准读数与随后6个月的读数有显著变化的T1DM和T2DM对象,对算法1在测试数据组中的精度进行了进一步的评估。表3A和3B给出了HbA1c的绝对改变等于或者大于2个单位的T1DM和T2DM对象的列表。在每个对象组中,有34个对象的HbA1c具有这种变化。算法1,函数F1,在T1DM和T2DM中都100%地预测出了这种改变。由于公式中含有基准HbA1c(其部分地将估计值拉回到HbA1c的基准值),所以F2的预测力变小,在T1DM中为71%,在T2DM中为85%。除两个对象之外,基准HbA1c落在6个月HbA1c值20%区间的外部:The accuracy of
表3A:HbA1c变化>=2单位的T1DM对象Table 3A: T1DM subjects with HbA 1c change >= 2 units
ID HBA0 HBA6 DHBA F1 F2 HIT F1 HIT F2 HITID HBA0 HBA6 DHBA F1 F2 HIT F1 HIT F2 HIT
HBA0HBA0
6504 12.0 7.0 5.00 6.82 9.90 100.00 .00 .006504 12.0 7.0 5.00 6.82 9.90 100.00 .00 .00
6613 10.5 6.8 3.70 8.02 9.37 100.00 .00 .006613 10.5 6.8 3.70 8.02 9.37 100.00 .00 .00
4003 12.4 8.9 3.50 8.45 10.73 100.00 .00 .004003 12.4 8.9 3.50 8.45 10.73 100.00 .00 .00
6204 11.0 7.5 3.50 7.29 9.45 100.00 .00 .006204 11.0 7.5 3.50 7.29 9.45 100.00 .00 .00
3709 13.0 9.7 3.30 8.99 11.54 100.00 100.00 .003709 13.0 9.7 3.30 8.99 11.54 100.00 100.00 .00
4701 12.8 9.5 3.30 9.50 11.61 100.00 .00 .004701 12.8 9.5 3.30 9.50 11.61 100.00 .00 .00
3614 11.9 8.7 3.20 8.24 10.30 100.00 100.00 .003614 11.9 8.7 3.20 8.24 10.30 100.00 100.00 .00
3602 11.5 8.3 3.20 7.93 9.94 100.00 100.00 .003602 11.5 8.3 3.20 7.93 9.94 100.00 100.00 .00
6008 11.3 8.3 3.00 9.30 10.53 100.00 .00 .006008 11.3 8.3 3.00 9.30 10.53 100.00 .00 .00
3723 13.0 10.1 2.90 8.80 11.46 100.00 100.00 .003723 13.0 10.1 2.90 8.80 11.46 100.00 100.00 .00
7010 12.7 9.8 2.90 8.09 10.89 100.00 100.00 .007010 12.7 9.8 2.90 8.09 10.89 100.00 100.00 .00
6208 11.5 8.7 2.80 8.42 10.09 100.00 100.00 .006208 11.5 8.7 2.80 8.42 10.09 100.00 100.00 .00
6202 10.6 7.8 2.80 7.91 9.37 100.00 .00 .006202 10.6 7.8 2.80 7.91 9.37 100.00 .00 .00
3924 9.9 7.2 2.70 7.71 8.72 100.00 .00 .003924 9.9 7.2 2.70 7.71 8.72 100.00 .00 .00
8211 11.0 8.3 2.70 8.76 10.32 100.00 .00 .008211 11.0 8.3 2.70 8.76 10.32 100.00 .00 .00
6012 9.3 6.7 2.60 7.82 8.35 100.00 .00 .006012 9.3 6.7 2.60 7.82 8.35 100.00 .00 .00
3913 11.0 8.4 2.60 7.88 9.54 100.00 100.00 .003913 11.0 8.4 2.60 7.88 9.54 100.00 100.00 .00
6701 11.2 8.6 2.60 8.75 10.07 100.00 100.00 .006701 11.2 8.6 2.60 8.75 10.07 100.00 100.00 .00
2307 10.6 8.1 2.50 7.95 9.27 100.00 100.00 .002307 10.6 8.1 2.50 7.95 9.27 100.00 100.00 .00
3516 11.8 9.3 2.50 7.76 10.03 100.00 100.00 .003516 11.8 9.3 2.50 7.76 10.03 100.00 100.00 .00
5808 9.6 7.2 2.40 7.61 8.52 100.00 100.00 .005808 9.6 7.2 2.40 7.61 8.52 100.00 100.00 .00
2201 11.8 9.5 2.30 8.90 10.71 100.00 100.00 .002201 11.8 9.5 2.30 8.90 10.71 100.00 100.00 .00
4010 12.4 10.1 2.30 8.57 11.15 100.00 100.00 .004010 12.4 10.1 2.30 8.57 11.15 100.00 100.00 .00
6210 11.9 9.6 2.30 8.33 10.40 100.00 100.00 .006210 11.9 9.6 2.30 8.33 10.40 100.00 100.00 .00
4904 11.3 9.1 2.20 8.63 10.29 100.00 100.00 .004904 11.3 9.1 2.20 8.63 10.29 100.00 100.00 .00
6709 10.3 8.1 2.20 7.83 9.04 100.00 100.00 .006709 10.3 8.1 2.20 7.83 9.04 100.00 100.00 .00
6619 9.5 7.3 2.20 7.64 8.57 100.00 100.00 .006619 9.5 7.3 2.20 7.64 8.57 100.00 100.00 .00
3921 10.9 8.8 2.10 7.20 9.19 100.00 100.00 .003921 10.9 8.8 2.10 7.20 9.19 100.00 100.00 .00
6603 11.0 8.9 2.10 8.18 9.89 100.00 100.00 .006603 11.0 8.9 2.10 8.18 9.89 100.00 100.00 .00
7415 10.6 8.5 2.10 7.94 9.27 100.00 100.00 .007415 10.6 8.5 2.10 7.94 9.27 100.00 100.00 .00
6515 9.8 7.8 2.00 7.13 9.54 100.00 100.00 .006515 9.8 7.8 2.00 7.13 9.54 100.00 100.00 .00
3611 10.3 8.3 2.00 8.36 9.23 100.00 100.00 .003611 10.3 8.3 2.00 8.36 9.23 100.00 100.00 .00
3732 13.2 11.2 2.00 9.30 l1.99 100.00 100.00 100.003732 13.2 11.2 2.00 9.30 l1.99 100.00 100.00 100.00
7409 10.0 8.0 2.00 7.99 9.04 100.00 100.00 .007409 10.0 8.0 2.00 7.99 9.04 100.00 100.00 .00
表3B:HbA1c变化>=2单位的T1DM对象Table 3B: T1DM subjects with HbA 1c change >= 2 units
ID HBA0 HBA6 DHBA F1 F2 HIT F1 HIT F2 HITID HBA0 HBA6 DHBA F1 F2 HIT F1 HIT F2 HIT
HBA0HBA0
6754 10.8 7.0 3.80 6.90 9.03 100.00 .00 .006754 10.8 7.0 3.80 6.90 9.03 100.00 .00 .00
6361 11.3 7.6 3.70 8.51 10.20 100.00 .00 .006361 11.3 7.6 3.70 8.51 10.20 100.00 .00 .00
6270 12.0 8.6 3.40 7.85 10.03 100.00 100.00 .006270 12.0 8.6 3.40 7.85 10.03 100.00 100.00 .00
6264 11.1 7.8 3.30 8.31 9.70 100.00 .00 .006264 11.1 7.8 3.30 8.31 9.70 100.00 .00 .00
6355 11.8 8.6 3.20 7.99 9.90 100.00 100.00 .006355 11.8 8.6 3.20 7.99 9.90 100.00 100.00 .00
3961 10.8 8.0 2.80 9.13 9.73 100.00 .00 .003961 10.8 8.0 2.80 9.13 9.73 100.00 .00 .00
6555 11.1 8.3 2.80 8.11 9.55 100.00 100.00 .006555 11.1 8.3 2.80 8.11 9.55 100.00 100.00 .00
8052 11.7 8.9 2.80 7.68 9.80 100.00 100.00 .008052 11.7 8.9 2.80 7.68 9.80 100.00 100.00 .00
5356 9.7 7.0 2.70 6.75 8.20 100.00 100.00 .005356 9.7 7.0 2.70 6.75 8.20 100.00 100.00 .00
3966 10.3 7.7 2.60 8.08 9.07 100.00 100.00 .003966 10.3 7.7 2.60 8.08 9.07 100.00 100.00 .00
908 9.5 6.9 2.60 7.47 8.23 100.00 100.00 .00908 9.5 6.9 2.60 7.47 8.23 100.00 100.00 .00
6554 10.7 8.1 2.60 8.16 9.42 100.00 100.00 .006554 10.7 8.1 2.60 8.16 9.42 100.00 100.00 .00
2353 11.1 8.7 2.40 8.99 9.90 100.00 100.00 .002353 11.1 8.7 2.40 8.99 9.90 100.00 100.00 .00
4064 11.3 8.9 2.40 7.89 9.88 100.00 100.00 .004064 11.3 8.9 2.40 7.89 9.88 100.00 100.00 .00
6351 10.1 7.7 2.40 7.92 8.63 100.00 100.00 .006351 10.1 7.7 2.40 7.92 8.63 100.00 100.00 .00
7551 12.2 9.8 2.40 9.17 11.02 100.00 100.00 .007551 12.2 9.8 2.40 9.17 11.02 100.00 100.00 .00
6358 8.4 6.1 2.30 7.00 7.32 100.00 .00 .006358 8.4 6.1 2.30 7.00 7.32 100.00 .00 .00
3965 10.1 7.8 2.30 7.83 8.64 100.00 100.00 .003965 10.1 7.8 2.30 7.83 8.64 100.00 100.00 .00
914 11.1 8.8 2.30 9.57 10.33 100.00 100.00 .00914 11.1 8.8 2.30 9.57 10.33 100.00 100.00 .00
1603 10.2 7.9 2.30 8.02 8.88 100.00 100.00 .001603 10.2 7.9 2.30 8.02 8.88 100.00 100.00 .00
1708 10.8 8.6 2.20 7.62 9.24 100.00 100.00 .001708 10.8 8.6 2.20 7.62 9.24 100.00 100.00 .00
3761 12.4 10.2 2.20 9.13 10.86 100.00 100.00 .003761 12.4 10.2 2.20 9.13 10.86 100.00 100.00 .00
3768 11.2 9.0 2.20 8.29 9.74 100.00 100.00 .003768 11.2 9.0 2.20 8.29 9.74 100.00 100.00 .00
326 10.3 8.2 2.10 7.45 8.78 100.00 100.00 .00326 10.3 8.2 2.10 7.45 8.78 100.00 100.00 .00
109 9.3 7.2 2.10 7.70 8.18 100.00 100.00 .00109 9.3 7.2 2.10 7.70 8.18 100.00 100.00 .00
1501 11.9 9.8 2.10 8.52 10.18 100.00 100.00 .001501 11.9 9.8 2.10 8.52 10.18 100.00 100.00 .00
3964 13.7 11.6 2.10 10.08 12.65 100.00 100.00 100.003964 13.7 11.6 2.10 10.08 12.65 100.00 100.00 100.00
4352 12.2 10.1 2.10 9.51 11.14 100.00 100.00 .004352 12.2 10.1 2.10 9.51 11.14 100.00 100.00 .00
7858 12.1 10.0 2.10 9.53 l1.01 100.00 100.00 .007858 12.1 10.0 2.10 9.53 l1.01 100.00 100.00 .00
4256 10.6 8.6 2.00 8.76 9.69 100.00 100.00 .004256 10.6 8.6 2.00 8.76 9.69 100.00 100.00 .00
4752 10.1 8.1 2.00 8.51 8.87 100.00 100.00 .004752 10.1 8.1 2.00 8.51 8.87 100.00 100.00 .00
6556 11.1 9.1 2.00 8.72 9.68 100.00 100.00 .006556 11.1 9.1 2.00 8.72 9.68 100.00 100.00 .00
6562 7.9 5.9 2.00 7.07 7.04 100.00 100.00 .006562 7.9 5.9 2.00 7.07 7.04 100.00 100.00 .00
8255 10.9 8.9 2.00 8.90 9.87 100.00 100.00 .008255 10.9 8.9 2.00 8.90 9.87 100.00 100.00 .00
在表3A和3B中:In Tables 3A and 3B:
ID——对象的ID数字;ID - the ID number of the object;
HBA0——基准HbA1c;HBA0 - baseline HbA 1c ;
HBA6——HbA1c的6个月测量值;HBA6 - 6-month measurement of HbA 1c ;
DHBA——HbA1c基准值与6个月值之间的绝对差值;DHBA—the absolute difference between the HbA 1c baseline value and the 6-month value;
F1——由函数F1估计的HbA1c,只使用SMBG数据;F1 - HbA 1c estimated by function F1, using only SMBG data;
F2——由函数F2估计的HbA1c,使用先前的HbA1c化验结果;F2——HbA 1c estimated by function F2, using previous HbA 1c assay results;
HitF1=100,如果F1在6个月HbA1c值的20%之内,否则为0;HitF1 = 100, if F1 is within 20% of the 6-month HbA 1c value, otherwise 0;
HitF2=100,如果F2在6个月HbA1c值的20%之内,否则为0;和HitF2 = 100 if F2 is within 20% of the 6-month HbA 1c value, 0 otherwise; and
Hit HbA0=100,如果基准HbA1c在6个月HbA1c读数的20%之内,否则为0。Hit HbA0 = 100 if baseline HbA 1c is within 20% of the 6-month HbA 1c reading, 0 otherwise.
详细结果——训练数据组Detailed Results - Training Dataset
这一部分说明了优化算法1的步骤。该优化包括两个部分:(1)假定不能获得先前的HbA1c读数,和(2)假定能够使用先前的HbA1c预测HbA1c。This section illustrates the steps to optimize
我们考虑了多个不同的函数用于描述SMBG数据与HbA1c的关系。根据计算的精度和简单性,如果不使用先前的HbA1c读数,最佳函数似乎是SMBG读数的平均值、低和高BG指数的线性函数,另一个是先前HbA1c与高B指数的线性函数。非线性关系并没有提高该模型的匹配良好度,因此不考虑用于实际应用。We considered several different functions for describing the relationship between SMBG data and HbA 1c . Based on the accuracy and simplicity of the calculations, if previous HbA 1c readings are not used, the best function appears to be a linear function of the mean of the SMBG readings, low and high BG index, and another linear function of previous HbA 1c versus high B index . Non-linear relationships do not improve the goodness of fit of the model and are therefore not considered for practical use.
训练数据组1-无先前的HbA1c 使用线性回归模型优化函数F1的系数。最佳系数在前面的部分中已经给出。这里,我们给出有关模型匹配良好度的数据:Training data set 1 - no prior HbA 1c The coefficients of function F1 were optimized using a linear regression model. The optimal coefficients are given in the previous section. Here, we give data on how well the model fits:
倍数R(Multiple R) .71461Multiple R (Multiple R) .71461
R方 .51067R Square .51067
方差分析variance analysis
DF 平方和 均方DF DF Sum of Squares Mean Square
回归 3 154.57097 51.52366return 3 154.57097 51.52366
残差 90 148.10903 1.64566residual 90 148.10903 1.64566
F=31.30889 显著度F=.0000F=31.30889 Significant degree F=.0000
该模型的残差分析显示出接近于残差的正态分布(见图11)。残差的SD为1.2(根据定义其平均值为0)。因此,我们能够接受,该模型良好地描述该数据。Analysis of the residuals from this model revealed a close to normal distribution of the residuals (see Figure 11). The SD of the residuals is 1.2 (which by definition has a mean of 0). Therefore, we can accept that the model describes the data well.
训练数据组2-先前的HbA1c 再一次,使用线性回归模型优化函数F3的系数。最佳系数在前面的部分中已经给出。这里,我们给出有关模型匹配良好度的数据:Training Data Set 2 - Previous HbA 1c Again, the coefficients of function F3 were optimized using a linear regression model. The optimal coefficients are given in the previous section. Here, we give data on how well the model fits:
倍数R .86907Multiple R .86907
R方 .75528R Square .75528
方差分析variance analysis
DF 平方和 均方DF DF Sum of Squares Mean Square
回归 4 38.70237 9.67559return 4 38.70237 9.67559
残差 54 12.54000 .23222residual 54 12.54000 .23222
F=41.66522 显著度F=.0000F=41.66522 Significant degree F=.0000
该模型的残差分析显示出接近残差的正态分布(见图12)。残差的SD为0.47(根据定义其平均值为0)。因此,我们能够接受,该模型良好地描述了该数据。Analysis of the residuals from this model revealed a nearly normal distribution of the residuals (see Figure 12). The SD of the residuals is 0.47 (which by definition has a mean of 0). Therefore, we can accept that the model describes the data well.
此外,比较没有和有先前HbA1c的模型,我们能够得出结论,如果在计算中包括先前的HbA1c,则最终的模型比根据R2和根据残差误差都要好得多。Furthermore, comparing models without and with prior HbA 1c , we were able to conclude that if prior HbA 1c was included in the calculation, the final model was much better both in terms of R 2 and in terms of residual error.
然而,正如我们在先前部分中见到的,在不相关数据组中,先前的HbA1c对于预测的总精度没有贡献,在某些情况下,HbA1c变化显著甚至由于变化快速而妨碍了算法的能力。因此,我们能够得出结论,即使先前的HbA1c可能从统计的观点看更好,但是不可能有足够的实际效用,用于校正将来的测量计的读数输入。我们还不知道HbA1c化验与SMBG检测(profile)之间间隔的时间,但仍然使HbA1c的输入有用。或许,这取决于该时间周期中HbA1c的改变——正如我们在前一部分中看到的,2个HbA1c单位的变化就使先前的HbA1c读数完全无用。However, as we saw in the previous section, in uncorrelated data sets, previous HbA 1c did not contribute to the overall accuracy of the prediction, and in some cases HbA 1c changed significantly and even hampered the algorithm's performance due to rapid changes. ability. Therefore, we can conclude that even though previous HbA 1c may be better from a statistical point of view, it is unlikely to have sufficient practical utility for correcting future meter reading inputs. We do not yet know the time interval between the HbA 1c assay and the SMBG profile, but still makes the HbA 1c entry useful. Perhaps, it depends on the change in HbA 1c over that time period - as we saw in the previous section, a change of 2 HbA 1c units renders the previous HbA 1c reading completely useless.
SMBG/HbA1c比值SMBG/HbA 1c ratio
我们现在提供了一种可选择的方法,以改进模型匹配的统计精度并保持相当高的临床应用性。可见,45天SMBG读数的平均值与HbA1c的比值是一个具有最完全正态分布的测量量(由Kolmogorov-Simov实验可以证实),并且最重要地,区分了三组对象,其比值为<1.0、1.0-1.2和>1.2。前两组的每一个都能解释大约40%的对象,第三组能够解释大约20%的对象。这对于T1DM和T2DM都是有效的,并且在训练数据组以及测试数据组中都能观察到。此外,该比值对于时间似乎特别稳定,因此可能是反映患者SMBG习惯的测量量(例如,如果SMBG主要在BG较低时进行,则最终的平均值将低估HbA1c,从而相应的比值将小于1.0)。注意这只是一个假说,并不能够用可获得的数据加以证实,我们进行了一些分析,似乎证实能够在一定程度上知道每个人在某个时间点的比例。这似乎相当于知道先前的HbA1c,并且有可能等价于数据输入,但是该比值的应用与先前HbA1c的使用非常不同。除了直接用于该预测公式之外,该比值还用于将患者进行分类,对它们应用三个不同预测公式中的某一个。这些新的公式不直接包括HbA1c,因此不会受包含HbA1c导致的惯性(inertia)的影响。另外,由该比值限定的三个组之间平均HbA1c差异并不大,并且与该比值不相关,所以不同人中比例不同的原因一定与HbA1c不相关。We now provide an alternative method to improve the statistical accuracy of model matching while maintaining considerable clinical applicability. It can be seen that the ratio of the mean of the 45-day SMBG readings to HbA 1c is a measure with the most complete normal distribution (confirmed by the Kolmogorov-Simov experiment), and most importantly, distinguishes three groups of subjects whose ratio is < 1.0, 1.0-1.2 and >1.2. Each of the first two groups was able to explain about 40% of the objects, and the third group was able to explain about 20% of the objects. This is valid for both T1DM and T2DM and is observed in the training dataset as well as the test dataset. Furthermore, the ratio appears to be particularly stable over time, and thus may be a measure reflecting the patient's SMBG habit (eg, if SMBG is performed primarily when BG is low, the final mean will underestimate HbA 1c and the corresponding ratio will be less than 1.0 ). Note that this is only a hypothesis and cannot be confirmed with the available data, we have performed some analysis and it seems to confirm that the proportion of each individual at a certain point in time can be known to some extent. This appears to be equivalent to knowing the prior HbA 1c , and possibly equivalent to data entry, but the application of this ratio is very different from the use of prior HbA 1c . In addition to being used directly in the predictive formula, the ratio is used to classify patients, applying one of three different predictive formulas to them. These new formulas do not directly include HbA 1c and therefore are not affected by the inertia caused by including HbA 1c . Also, the difference in mean HbA 1c between the three groups defined by this ratio is not large and is not related to this ratio, so the reason for the different proportions in different people must not be related to HbA 1c .
如果我们首先根据他们的比值将对象分成3组并在训练数据组中分别进行回归,则回归模型的配合良好度显著增加:(1)在组1(比值<1.0)中,我们得到倍数R=0.86,R2=0.73;(2)在组2(比值为1.0-1.2)中,匹配度最佳,R=0.97,R2=0.94,和(3)在组3(比值>1.2)中,匹配度最差,R=0.69,R2=0.47。因为所有三种回归模型都不包括先前的HbA1c,所以我们得出结论,对于大约80%的对象,匹配良好度显著增加,剩余20%的对象,匹配良好度保持相同,匹配度将变差的对象能够预先被识别出来。If we first divide the subjects into 3 groups according to their ratio and regress them separately in the training data set, the goodness of fit of the regression model increases significantly: (1) In group 1 (ratio < 1.0), we get the multiple R = 0.86, R 2 =0.73; (2) in group 2 (ratio 1.0-1.2), the best match, R=0.97, R 2 =0.94, and (3) in group 3 (ratio>1.2), The matching degree is the worst, R=0.69, R 2 =0.47. Because all three regression models do not include prior HbA 1c , we conclude that for about 80% of subjects, goodness of match increases significantly, and for the remaining 20% of subjects, goodness of match remains the same, and goodness of match becomes worse objects can be identified in advance.
进一步,根据对象的比值将测试数据组分成3组。我们获得的预测精度与先前获得的精度相似(表4A和4B):Further, the test data group is divided into 3 groups according to the ratio of the objects. We obtained prediction accuracies similar to those obtained previously (Tables 4A and 4B):
表4A:算法1在T1DM(N=254个对象)中的精度Table 4A: Accuracy of
表4B:算法1在T2DM(N=319个对象)中的精度Table 4B: Accuracy of
简而言之,知道每个对象的SMBG/HBA1c比值并相应地分别进行估计,似乎能够提高该模型的统计性能而不损失临床精度。In short, knowing the SMBG/HBA 1c ratio for each subject and estimating it separately accordingly appeared to improve the statistical performance of the model without loss of clinical precision.
被测试的其他假说和思路Other hypotheses and ideas tested
我们测试了大量其他的假说和思路,它们被证实至少对于促进和更集中地分析由实例No.2收集的数据是有用的。简要的结果说明如下:We tested a number of other hypotheses and ideas that proved useful at least for facilitating and more focused analysis of the data collected by Example No. 2. A brief description of the results follows:
(1)HbA1c与下午获取的SMBG读数——中午12点至下午6点——最关联(相关),与空腹SMBG读数(4a.m.-8a.m.)最少关联。然而,并不是只采集餐后SMBG读数就会提高HbA1c的预测,相反,如果忽略全天所有小时(相对较小但重要)的贡献,则预测会变得更糟。如果给全天其它小时的读数进行不同的加权,则有可能提高HbA1c的预测,但是该提高并不足以弥补模型附加的复杂性;(1) HbA 1c was most associated (correlated) with SMBG readings taken in the afternoon - 12 noon to 6 pm - and least correlated with fasting SMBG readings (4a.m.-8a.m.). However, it is not that taking postprandial SMBG readings alone improves the prediction of HbA 1c , rather the prediction gets worse if the (relatively small but important) contribution of all hours throughout the day is ignored. It is possible to improve the prediction of HbA 1c if readings at other hours of the day are weighted differently, but this improvement is not enough to compensate for the added complexity of the model;
(2)在T2DM中,HbA1c与平均SMBG的关系比在T1DM中更强,即使两组的HbA1c相互匹配。根据直接相关性,在整个研究中,T1DM中的系数为大约0.6,而T2DM中的系数大约为0.75;(2) The relationship between HbA 1c and mean SMBG was stronger in T2DM than in T1DM, even though HbA 1c was matched between the two groups. According to the direct correlation, the coefficient in T1DM was about 0.6 and that in T2DM was about 0.75 throughout the study;
(3)根据SMBG与HbA1c化验之间间隔的时间对SMBG读数进行不同的加权(例如结果越接近中间,加权越高)并不能获得对HbA1c更好的预测;(3) Weighting SMBG readings differently according to the time interval between SMBG and HbA 1c assays (eg, the closer the result is to the middle, the higher the weighting) does not yield a better prediction of HbA 1c ;
(4)包含人口统计变量,例如年龄、糖尿病持续时间、性别等,并不能提高HbA1c的预测;(4) Including demographic variables, such as age, duration of diabetes, gender, etc., does not improve the prediction of HbA 1c ;
(5)HbA1c与平均SMBG(以mmol/l为单位测量)之间可能最简单的线性关系由如下公式给出:HbA1c=0.41046*BGMM+4.0775。尽管与F1和F2相比在统计上较差,但是该公式给出的HbA1c估计在T1DM和T2DM中具有大约95%的精度(根据与HbA1c化验的偏差小于20%),并且如果通过计算低和高BG指数将其合并在一个测量计中,则可能是有用的(但是,不计算低BG指数是不能够实现低血糖症的预测的,因此该公式可能只对包括算法1但不包括算法2和3的测量计有用)。(5) The simplest possible linear relationship between HbA 1c and mean SMBG (measured in mmol/l) is given by the following formula: HbA 1c =0.41046*BGMM+4.0775. Although statistically inferior compared to F1 and F2, the formula gives an estimate of HbA 1c with approximately 95% accuracy in T1DM and T2DM (less than 20% deviation from HbA 1c assays), and if calculated by It may be useful to combine both low and high BG indices in one meter (however, prediction of hypoglycemia cannot be achieved without calculating the low BG index, so this formula may only be useful for including
算法2:SH长期危险的评估Algorithm 2: Assessment of the long-term risk of SH
实例No.1提供了,但不仅限于,算法2的扩展,包括估计个体生化显著低血糖症(biochemical significant hyoglycemia)(BSH,定义为BG读数<=39mg/dl)或者生化中度低血糖症(BMH,定义为39mg/dl<BG<=55mg/dl)的概率。另外,我们计划评估,与日间SH相比,算法2预测夜间(午夜至7:00am)SH的发生率是否更好。Example No. 1 provides, but is not limited to, an extension of
算法2是一个分类算法。也就是说,根据对象的SMBG数据,它将对象的将来BSH或者MSH分类成特定的危险范围。为了尽可能接近算法2将来真实的应用,我们如下地进行:
(4)首先,从训练数据组1得出多个最佳分类变量和最佳分类范围、最佳持续时间和最佳SMBG频率;(4) at first, draw a plurality of optimal classification variables and optimal classification scope, optimal duration and optimal SMBG frequency from training data set 1;
(5)然后,将测试数据组分成两个部分:最初45天和其余的数据部分。将算法2的最佳参数应用于数据的最初45天部分,从而使用将来BSH或者MSH的概率估计值预测数据第二部分的BSH和MSH;(5) Then, the test data set is divided into two parts: the first 45 days and the remaining data part. Apply the optimal parameters of
(6)只用测试数据对算法2精度进行详细评估。(6) Perform a detailed evaluation of
训练与测试数据组分离允许我们声明,算法2的估计精度能够一般化到T1DM或者T2DM患者任何其他的数据。而且,因为Amylin数据是从经过强化治疗的对象中收集的,所以我们能够推测,算法2在低血糖症危险不断变化且不断增加的对象中进行试验并且证实是有效的。The separation of the training and test datasets allows us to state that the estimated accuracy of
结果摘要Summary of Results
●估计将来BSH或者BMH概率所需的最佳SMBG数据收集周期为40-45天。SMBG的最佳频率为3-4个读数/天。大量的读数并不能够导致算法2预测能力的显著增加。如果每天的读数小于3个,则预测能力下降。然而,该要求是参考45天观察周期内每天的平均读数,并非意味着,每天都需要执行3-4次读数;● The optimal SMBG data collection period needed to estimate the probability of future BSH or BMH is 40-45 days. The optimal frequency of SMBG is 3-4 readings/day. A large number of reads did not lead to a significant increase in the predictive power of
●预测子变量与将来SH和MH之间的关系严格地非线性。因此,线性方法不能用于优化预测,尽管通过直接线性模型能够获得R2=50%(相比之下,DCCT的最佳结果为预测8%的将来SH);• The relationship between the predictor variables and future SH and MH is strictly non-linear. Therefore, linear methods cannot be used to optimize predictions, although R2 = 50% can be obtained by direct linear models (compared to the best result of DCCT for predicting 8% of future SH);
●单独预测夜间SH一般比预测日间SH更弱;Prediction of nocturnal SH alone is generally weaker than prediction of daytime SH;
●确定了15个将来BSH和BMH的危险范围。范围的最佳分离只是根据低BG指数获得的,尽管低BG指数与其它变量的组合能够发挥类似良好的作用;●Identified 15 future hazard areas for BSH and BMH. The best separation of ranges was obtained only for low BG index, although combinations of low BG index and other variables can work similarly well;
●尽管BSH和BMH的频率在T1DM和T2DM中不同(见表5),但是当给定一个危险范围时,条件频率在T1DM和T2DM之间并无差异。这允许获得统一的预测SH和MH危险的方法;●Although the frequency of BSH and BMH differed in T1DM and T2DM (see Table 5), the conditional frequency did not differ between T1DM and T2DM when given a risk range. This allows to obtain a unified method for predicting SH and MH hazards;
●为15个危险范围计算了将来的各经验概率并进行了比较。所有的比较都具有高的显著度,p’s<0.0005。• The individual future empirical probabilities were calculated and compared for the 15 hazard ranges. All comparisons were highly significant, p's < 0.0005.
●这些经验概率用双参数Weibull分布(two-parameter Weibulldistribution)加以近似,产生了每个危险范围中将来BSH和BMH的理论概率。• These empirical probabilities were approximated by a two-parameter Weibull distribution, yielding theoretical probabilities of future BSH and BMH in each hazard range.
●这些近似的匹配良好度非常好——所有的确定系数都大于85%,一些高达98%(见图1-5和9-10)。• The goodness of fit of these approximations is very good - all have coefficients of determination greater than 85%, some as high as 98% (see Figures 1-5 and 9-10).
详细结果——测试数据组Detailed Results - Test Dataset
确定SH/MH的个人危险范围:总共600个对象的数据用于该分析。由他/她前45天的SMBG数据收集为每个对象计算低BG指数(LBGI)。然后,将LBGI分类到15个最佳危险范围中的其中一个(变量RCAT的范围是0-14),如在训练数据组1中得出的。这些危险范围通过如下的不等式定义: Determination of Individual Risk Ranges for SH/MH : Data from a total of 600 subjects were used for this analysis. A Low BG Index (LBGI) was calculated for each subject from his/her SMBG data collection for the previous 45 days. Then, LBGI is classified into one of 15 optimal risk ranges (variable RCAT ranges from 0-14), as derived in
if(LBGI≤0.25),RCAT=0If (LBGI≤0.25), RCAT=0
if(0.25<LBGI≤0.5),RCAT=1if(0.25<LBGI≤0.5), RCAT=1
if(0.50<LBGI≤0.75),RCAT=2if(0.50<LBGI≤0.75), RCAT=2
if(0.75<LBGI≤1.00),RCAT=3if(0.75<LBGI≤1.00), RCAT=3
if(1.00<LBGI≤1.25),RCAT=4if(1.00<LBGI≤1.25), RCAT=4
if(1.25<LBGI≤1.50),RCAT=5if(1.25<LBGI≤1.50), RCAT=5
if(1.50<LBGI≤1.75),RCAT=6if(1.50<LBGI≤1.75), RCAT=6
if(1.75<LBGI≤2.00),RCAT=7if(1.75<LBGI≤2.00), RCAT=7
if(2.00<LBGI≤2.50),RCAT=8if(2.00<LBGI≤2.50), RCAT=8
if(3.00<LBGI≤3.50),RCAT=9if(3.00<LBGI≤3.50), RCAT=9
if(3.50<LBGI≤4.00),RCAT=10if(3.50<LBGI≤4.00), RCAT=10
if(4.00<LBGI≤4.50),RCAT=11if(4.00<LBGI≤4.50), RCAT=11
if(4.50<LBGI≤5.25),RCAT=12if(4.50<LBGI≤5.25), RCAT=12
if(5.25<LBGI≤6.50),RCAT=13if(5.25<LBGI≤6.50), RCAT=13
if(LBGI>6.50),RCAT=14if(LBGI>6.50), RCAT=14
BSH和BMH的观察频率:对于每个对象,在最初45天数据收集之后的1个月、3个月和6个月内计数由SMBG指示的BSH和BMH的发生。表5A给出了在T1DM中观察到0、>=1、>=2、>=3次BSH和BMH的频率,表5B给出了在T2DM中观察到的相同数据: Frequency of BSH and BMH observations : For each subject, the occurrence of BSH and BMH as indicated by SMBG was counted at 1 month, 3 months and 6 months after the initial 45 days of data collection. Table 5A presents the frequency of 0, >=1, >=2, >=3 BSH and BMH observed in T1DM, and Table 5B presents the same data observed in T2DM:
表5A:在T1DM中观察到的BSH和BMH频率Table 5A: BSH and BMH frequencies observed in T1DM
表5B:在T2DM中观察到的BSH和BMH频率Table 5B: BSH and BMH frequencies observed in T2DM
夜间BSH和BMH占由SMBG指示的全部事件的大约15%。在训练数据组中,夜间事件与全部预测子变量之间的相关性较弱。我们得出结论,夜间事件的目标预测是无效的。 Nocturnal BSH and BMH accounted for approximately 15% of all events indicated by SMBG. In the training data set, nocturnal events are less correlated with all predictor variables. We conclude that target prediction of nighttime events is not valid.
将来BSH和BMH的经验概率:我们计算了15个危险范围中每一个的将来BSH和BMH的特定经验概率。这些概率包括:(1)在之后1个月、3个月和6个月内发生至少一次BSH或者BMH的概率;(2)在之后3个月和6个月内发生至少二次BSH或者BMH的概率;和(3)在之后6个月内发生至少三次BSH或者BMH的概率。当然,有可能根据要求计算任何其他的组合概率。 Empirical probabilities of future BSH and BMH : We calculated specific empirical probabilities of future BSH and BMH for each of the 15 hazard ranges. These probabilities include: (1) the probability of at least one BSH or BMH occurring within the next 1, 3, and 6 months; (2) the probability of at least two BSH or BMH occurring within the following 3 and 6 months and (3) the probability of at least three occurrences of BSH or BMH within the following 6 months. Of course, it is possible to calculate any other combined probabilities on request.
从该分析得出的最重要的结论是,给定一个危险范围,将来BSH和BMH的概率在T1DM和T2DM之间没有显著差异。这允许获得统一的经验性和理论性预测T1DM和T2DM中这些概率的方法。结果,T1DM和T2DM患者的数据被组合用于如下的分析。The most important conclusion to draw from this analysis is that, given a risk spectrum, the probability of future BSH and BMH does not differ significantly between T1DM and T2DM. This allows obtaining a unified empirical and theoretical approach to predicting these probabilities in T1DM and T2DM. As a result, the data of T1DM and T2DM patients were combined for the following analysis.
图1-5和9-10给出了按照15个危险范围绘制的6个计算经验概率的散点图。BSH的经验概率用黑三角表示,而BMH的经验概率用红方形表示。Figures 1-5 and 9-10 show scatterplots of six calculated empirical probabilities plotted against 15 hazard ranges. Empirical probabilities for BSH are indicated by black triangles, while empirical probabilities for BMH are indicated by red squares.
用单变量ANOVA在15个危险范围内对所有组的经验概率进行比较,所有的p水平都小于0.0005。因此,我们观察到,在不同的危险范围内BSH和BMH事件之间的差异高度显著。Empirical probabilities for all groups were compared using univariate ANOVA at 15 hazards, and all p levels were less than 0.0005. Thus, we observed highly significant differences between BSH and BMH events across different risk spectrums.
将来BSH和BMH的理论概率:为了能够使用直接的公式估计将来BSH和BMH的概率,我们用双参数Weibull概率分布对经验概率进行了近似。Weibull概率函数由如下公式给出: Theoretical probabilities of future BSH and BMH : To be able to estimate the probabilities of future BSH and BMH using straightforward formulas, we approximated the empirical probabilities with a two-parameter Weibull probability distribution. The Weibull probability function is given by the following formula:
F(x)=1-exp(-a,xb),x>0;否则F(x)=0F(x)=1-exp(-a, x b ), x>0; otherwise F(x)=0
统计注意:参数a和b大于0,并且分别称作尺度参数和形状参数。在b=1的特殊实例中,Weibull分布变成指数型。该分布经常在工程问题中使用,因为随机发生技术失败的分布彼此并不完全无关(如果失败完全无关,则它们将形成Piosson处理,这将在指数分布中加以说明,例如b=1)。这里的情况很不相同——我们需要描述不完全独立并且趋向于群集发生的事件(失效)的分布,例如我们先前的研究所证实的。 Statistical Note : Parameters a and b are greater than 0 and are called scale and shape parameters, respectively. In the special case of b=1, the Weibull distribution becomes exponential. This distribution is often used in engineering problems because the distributions of randomly occurring technical failures are not completely independent of each other (if the failures were completely independent, they would form a Piosson process, which would be accounted for in an exponential distribution, eg b=1). The situation here is quite different - we need to describe the distribution of events (failures) that are not completely independent and tend to occur in clusters, such as demonstrated in our previous work.
每组经验概率都用上面给出的理论公式加以近似。用非线性最小平方法对参数进行估计(用线性双对数模型给出的初始参数估计)。每种模型的匹配良好度用其确定系数(coefficient of determination)(D2)进行评估。该统计的意义与线性回归中的R2相似,但是R2不能应用于非线性模型。Each set of empirical probabilities is approximated using the theoretical formula given above. Parameters were estimated using a nonlinear least squares method (initial parameter estimates given by a linear log-log model). The goodness of fit of each model was assessed by its coefficient of determination (D 2 ). The significance of this statistic is similar to R2 in linear regression, but R2 cannot be applied to nonlinear models.
图1-6中给出了模型匹配,黑线用于BSH的概率,虚线用于BMH的概率。在每个图上,我们给出了对相应模型的参数估计,因此我们给出了直接的公式,用于计算初始SMBG之后1个月、3个月和6个月内发生0、>=1、>=2、>=3次BSH和BMH的频率。在监测设备或者软件中能够包括其中一些公式或其变型作为SH和MH危险的指示子。The model fit is given in Figures 1-6, with the black line for the probability of BSH and the dashed line for the probability of BMH. On each plot, we give parameter estimates for the corresponding model, so we give straightforward formulas for calculating 0, >= 1 occurring 1, 3 and 6 months after the initial SMBG , >=2, >=3 times the frequency of BSH and BMH. Some of these formulas or variations thereof can be included in monitoring equipment or software as indicators of SH and MH hazard.
每个图的下方给出了D2值,作为近似精度的指示子。所有的数值都大于85%,一些达到了98%,这证实该近似度非常好,并且证实在将来的研究/应用中能够使用理论概率,而不是经验概率。The D2 value is given below each figure as an indicator of approximation accuracy. All values are greater than 85%, some reach 98%, which confirms that the approximation is very good and that theoretical probabilities can be used instead of empirical probabilities in future studies/applications.
发生一次或多次中度或严重低血糖症事件的理论概率由图1所示的公式给出:The theoretical probability of one or more events of moderate or severe hypoglycemia is given by the formula shown in Figure 1:
P(MH>=1)=1-exp(-exp(-1.5839)*Risk**1.0483)P(MH>=1)=1-exp(-exp(-1.5839)*Risk**1.0483)
P(SH>=1)=1-exp(-exp(-4.1947)*Risk**1.7472)P(SH>=1)=1-exp(-exp(-4.1947)*Risk**1.7472)
图1给出了在15个由低BG指数定义的危险水平的每一个内在SMBG估计之后1个月内中度(虚线)和严重(黑线)低血糖症的经验和理论概率。因为模型是非线性的,所以用它们的确定系数D2估计匹配良好度,D2是线性模型中R2的类似物。确定系数及其平方根如下:Figure 1 presents the empirical and theoretical probabilities of moderate (dashed line) and severe (black line) hypoglycemia within 1 month following intrinsic SMBG estimates for each of the 15 risk levels defined by the low BG index. Because the models are non - linear, the goodness of fit was estimated with their coefficient of determination D2 , which is the analogue of R2 in linear models. The coefficient of determination and its square root are as follows:
SH模型:D2=96%,D=98%SH model: D2 = 96%, D = 98%
MH模型:D2=87%,D=93%MH model: D2 = 87%, D = 93%
发生一次或多次中度或严重低血糖症事件的理论概率由图2所示的公式给出:The theoretical probability of one or more events of moderate or severe hypoglycemia is given by the formula shown in Figure 2:
P(MH>=1)=1-exp(-exp(-1.3731)*Risk**1.1351)P(MH>=1)=1-exp(-exp(-1.3731)*Risk**1.1351)
P(SH>=1)=1-exp(-exp(-3.2802)*Risk**1.5050)P(SH>=1)=1-exp(-exp(-3.2802)*Risk**1.5050)
图2给出了在15个由低BG指数定义的危险水平的每一个内在SMBG估计之后3个月内中度(虚线)和严重(黑线)低血糖症的经验和理论概率。Figure 2 presents the empirical and theoretical probabilities of moderate (dashed line) and severe (black line) hypoglycemia within 3 months following intrinsic SMBG estimates for each of the 15 risk levels defined by the low BG index.
确定系数及其平方根如下:The coefficient of determination and its square root are as follows:
SH模型:D2=93%,D=97%SH model: D2 = 93%, D = 97%
MH模型:D2=87%,D=93%MH model: D2 = 87%, D = 93%
发生一次或多次中度或严重低血糖症事件的理论概率由图3所示的公式给出:The theoretical probability of one or more events of moderate or severe hypoglycemia is given by the formula shown in Figure 3:
P(MH>=1)=1-exp(-exp(-1.3721)*Risk**1.3511)P(MH>=1)=1-exp(-exp(-1.3721)*Risk**1.3511)
P(SH>=1)=1-exp(-exp(-3.0591)*Risk**1.4549)P(SH>=1)=1-exp(-exp(-3.0591)*Risk**1.4549)
图3给出了在15个由低BG指数定义的危险水平的每一个内在SMBG估计之后6个月内中度(虚线)和严重(黑线)低血糖症的经验和理论概率。Figure 3 presents the empirical and theoretical probabilities of moderate (dashed line) and severe (black line) hypoglycemia within 6 months following intrinsic SMBG estimates for each of the 15 risk levels defined by the low BG index.
确定系数及其平方根如下:The coefficient of determination and its square root are as follows:
SH模型:D2=86%,D=93%SH model: D2 = 86%, D = 93%
MH模型:D2=89%,D=95%MH model: D2 = 89%, D = 95%
发生两次或多次中度或严重低血糖症事件的理论概率由图4所示的公式给出:The theoretical probability of two or more events of moderate or severe hypoglycemia is given by the formula shown in Figure 4:
P(MH>=2)=1-exp(-exp(-1.6209)*Risk**1.0515)P(MH>=2)=1-exp(-exp(-1.6209)*Risk**1.0515)
P(SH>=2)=1-exp(-exp(-4.6862)*Risk**1.8580)P(SH>=2)=1-exp(-exp(-4.6862)*Risk**1.8580)
图4给出了在15个由低BG指数定义的危险水平的每一个内在SMBG估计之后3个月内发生2次或多次中度(虚线)和严重(黑线)低血糖症的经验和理论概率。Figure 4 presents the empirical and theoretical probability.
确定系数及其平方根如下:The coefficient of determination and its square root are as follows:
SH模型:D2=98%,D=99%SH model: D2 = 98%, D = 99%
MH模型:D2=90%,D=95%MH model: D2 = 90%, D = 95%
发生两次或多次中度或严重低血糖症事件的理论概率由图5所示的公式给出:The theoretical probability of two or more events of moderate or severe hypoglycemia is given by the formula shown in Figure 5:
P(MH>=2)=1-exp(-exp(-1.7081)*Risk**1.19555)P(MH>=2)=1-exp(-exp(-1.7081)*Risk**1.19555)
P(SH>=2)=1-exp(-exp(-4.5241)*Risk**1.9402)P(SH>=2)=1-exp(-exp(-4.5241)*Risk**1.9402)
图5给出了在15个由低BG指数定义的危险水平的每一个内在SMBG估计之后6个月内发生2次或多次中度(虚线)和严重(黑线)低血糖症的经验和理论概率。Figure 5 presents the empirical and theoretical probability.
确定系数及其平方根如下:The coefficient of determination and its square root are as follows:
SH模型:D2=98%,D=99%SH model: D2 = 98%, D = 99%
MH模型:D2=89%,D=95%MH model: D2 = 89%, D = 95%
发生三次或多次中度或严重低血糖症事件的理论概率由图9所示的公式给出:The theoretical probability of three or more events of moderate or severe hypoglycemia is given by the formula shown in Figure 9:
P(MH>=3)=1-exp(-exp(-2.0222)*Risk**1.2091)P(MH>=3)=1-exp(-exp(-2.0222)*Risk**1.2091)
P(SH>=3)=1-exp(-exp(-5.5777)*Risk**2.2467)P(SH>=3)=1-exp(-exp(-5.5777)*Risk**2.2467)
图10给出了在15个由低BG指数定义的危险水平的每一个内在SMBG估计之后6个月内发生3次或多次中度(虚线)和严重(黑线)低血糖症的经验和理论概率。Figure 10 presents the empirical and theoretical probability.
确定系数及其平方根如下:The coefficient of determination and its square root are as follows:
SH模型:D2=97%,D=99%SH model: D2 = 97%, D = 99%
MH模型:D2=90%,D=95%。MH model: D 2 =90%, D=95%.
详细结果——训练数据组Detailed Results - Training Dataset
训练数据组包括SMBG数据及之后按月记录的严重低血糖症记录。与用截止(cutoff)BG数值确定BSH和BMH的测试数据组相对,按月的记录所包括的严重症状的报告被定义为由低血糖症导致的无意识、昏迷、无法自我治疗或者显著认知损伤。在SMBG之后的6个月内,研究对象报告称每个人有平均2.24次该事件,其中67%的对象报告没有该事件。从统计的观点看,这只会使SH事件的分布显著偏颇,并且不适合于应用线性方法。尽管线性回归能够用于估计各种变量的相对分布从而预测SH,但是并不能用于构建最终的模型。我们进行了如下的三种分析:The training data set consisted of SMBG data followed by severe hypoglycemia records recorded on a monthly basis. In contrast to the test data set using cutoff BG values to determine BSH and BMH, monthly records included reports of severe symptoms defined as unconsciousness, coma, inability to self-medicate, or significant cognitive impairment due to hypoglycemia . In the 6 months following SMBG, subjects reported an average of 2.24 events per individual, with 67% of subjects reporting no events. From a statistical point of view, this only significantly skews the distribution of SH events and is not suitable for applying linear methods. Although linear regression can be used to estimate the relative distribution of various variables to predict SH, it cannot be used to construct the final model. We performed the following three analyses:
(1) 不知道SH历史:忽略SH的任何历史知识,我们由基准HbA1c和SMBG特征,例如平均BG、低BG指数和变化的估计BG危险率,通过回归预测将来的SH(所有的变量在本发明最初的公开中已经说明)。和以前发现的一样,HbA1c和平均BG对预测SH没有任何贡献。最终的回归模型包括低BG指数和BG危险率的变化,并具有如下的匹配良好度:(1) Unknown SH history : Ignoring any historical knowledge of SH, we predicted future SH by regression from baseline HbA 1c and SMBG characteristics such as mean BG, low BG index, and changing estimated BG hazard rates (all variables in described in the original publication of the present invention). As previously found, HbA 1c and mean BG did not contribute to the prediction of SH. The final regression model included low BG index and change in BG hazard rate with goodness of fit as follows:
倍数R .61548Multiple R .61548
S平方 .37882S squared .37882
方差分析variance analysis
F=27.74772 显著度F=.0000F=27.74772 Significant degree F=.0000
————————等式中的变量——————————
(1) 知道先前的SH:当我们包括了前一年SH事件的数目时,如筛选调查表中报告的,这一变量能够解释另外11%的将来SH变异:(1) Knowing previous SH : When we included the number of SH events in the previous year, as reported in the screening questionnaire, this variable was able to explain an additional 11% of future SH variation:
倍数R .70328Multiple R .70328
S平方 .49461S squared .49461
方差分析variance analysis
F=29.35999 显著度F=.0000F=29.35999 Significant degree F=.0000
——————等式中的变量——————————
(3)不知道先前SH的数目,只知道某人先前是有还是没有过SH,我们只用SMBG变量能够解释45%的将来SH变异;(3) We don't know the number of previous SH, but only know whether someone has SH or not before, we can explain 45% of the future SH variation by using only the SMBG variable;
(4)最后,两个单独的线性模型能够解释55%的日间SH变异和25%的夜间SH变异。所有预测子变量与夜间SH的直接相关性也微弱。夜间事件占全部SH的30%。(4) Finally, two separate linear models were able to explain 55% of the diurnal SH variation and 25% of the nighttime SH variation. Direct associations of all predictor variables with nocturnal SH were also weak. Nocturnal events accounted for 30% of all SH.
我们得出结论,线性预测模型能够直接解释大约40-50%的将来SH变异。然而,根据其残余误差,该模型并未良好平衡(这是由于在糖尿病患者人口统计中SH事件数目的高度偏颇分布)。图13的正态概率图给出了其统计证据,该图显示出了标准残差与其预测值的显著偏差。We conclude that linear predictive models are able to directly explain approximately 40–50% of future SH variation. However, the model was not well balanced in terms of its residual error (due to the highly skewed distribution of the number of SH events in the diabetic population). The statistical evidence for this is given by the normal probability plot in Figure 13, which shows a significant deviation of the standard residuals from their predicted values.
因此,我们采用了另一种预测SH的方法,利用他们的SMBG数据将对象分成各种危险范围,并估计这些范围中随后SH的概率。我们尝试了各种分类模型,使危险范围之间的差异最大并试图获得最大的危险估计分辨率(根据范围的最大数目)。We therefore took an alternative approach to predicting SH by using their SMBG data to classify subjects into various risk ranges and estimate the probability of subsequent SH in these ranges. We tried various classification models, maximizing the difference between hazard ranges and trying to obtain maximum risk estimate resolution (in terms of maximum number of ranges).
通过只用低BG指数进行分类得到15个危险范围(在先一部分的开始部分给出)获得了最佳的结果。The best results were obtained by classifying only the low BG index for the 15 hazard ranges (given at the beginning of the previous section).
除了范围之间最佳分离之外,该结果还具有其他的优点:(1)不需要先前的SH历史知识;(2)计算相对简单且不需要跟踪时间变量,例如BG的变化速度,和(3)分类似乎能够同等地应用于T1DM和T2DM患者(不需要任何先前SH的知识)。In addition to the best separation between ranges, this result has other advantages: (1) no prior knowledge of SH history is required; (2) the calculation is relatively simple and does not require tracking temporal variables such as the rate of change of BG, and ( 3) The classification appears to be equally applicable to T1DM and T2DM patients (no prior knowledge of SH is required).
算法3:低血糖短期危险的评估Algorithm 3: Assessment of short-term risk of hypoglycemia
实例No.1给出了,但不仅限于,算法3的优化:Example No.1 gives, but is not limited to, the optimization of Algorithm 3:
(1)利用基准长期危险(来自算法2)和HbA1c(来自算法1);(1) Using baseline long-term risk (from Algorithm 2) and HbA 1c (from Algorithm 1);
(2)低血糖症报警的危险标准/阈值;(2) Dangerous standard/threshold for hypoglycemia alarm;
(3)SMBG频率;(3) SMBG frequency;
(4)如果检测到低血糖症危险增加并且已经有一段时间没有SMBG,则是否应当发出低血糖症报警,和(4) whether a hypoglycemia alarm should be issued if an increased risk of hypoglycemia is detected and there has been no SMBG for a period of time, and
(5)人口统计学变量的贡献,例如严重低血糖症历史。(5) Contribution of demographic variables, such as history of severe hypoglycemia.
引言introduction
与具有长期发展历史的算法1和算法2不同,算法3处理的问题直到最近都被认为是不可能的。实际上,人们仍然普遍认为根据先前已知的数值不可能预测将来的BG值(特别是低血糖症)(Bremer T和Gough DA.能够从先前的数值预测血糖吗?数据引发.(Is bloodglucose predictable from previous values?A solicitation for data)Diabetes,1999,
48:445-451)。我们先前的工作,其在能够从Lifescan有限公司获得的一个原稿中有过报道并且在本发明公开中被详细地提出,质疑了这个普遍的看法。为了解释该质疑的基础以及阐明算法3背后的原理,我们包含了下面的段落。Unlike
我们量化糖尿病特征的“哲学”:Our "philosophy" for quantifying diabetes characteristics:
根据所研究的内分泌系统,激素间的相互作用是由动态调控生化网络控制的,该网络具有由主要节点和管道组成的复杂或者简单的结构。糖尿病能够在各种水平上扰乱调控胰岛素-葡萄糖动力学的网络。例如,在T1DM中,胰岛素的自然产生被完全消除,而在T2DM中,胰岛素在细胞内的使用受到更强大的胰岛素抵抗的阻碍。在T1DM中(在T2DM中也常见),需要某种形式的外部胰岛素替代品,这使得调控系统对于有缺陷的外部因素变得脆弱,这些外部因素包括丸药和胰岛素注射的时限和剂量,所吃的食物,体育活动等。这通常会导致BG极度偏差出现低血糖症和高血糖症。在许多但不是所有病例中,低血糖症激发内分泌反应,称之为反向调节。因此,在数学领域,BG在一段时间内的波动是受大量内部和外部因素影响的复杂动力学系统活动的可测量结果。然而,根据众所周知的动力学系统理论,当调控的复杂性增加时,一个纯粹的确定性系统会发展成显示随机的宏观行为。因此,在短时期内(分钟),在人的水平上观察到的BG波动将是接近确定的,然而从长期看,波动将是接近随机的,这包括极端过渡,例如SH事件。因此,随机建模和统计推论最适于在较长的时期内分析系统——算法1和算法2采用的范例,其在特定的观察周期之后利用我们最初提出的手段,例如LBGI和HBGI,预测数值范围和一个事件的概率。BG在短期内的波动能够用确定性网络加以建模和预测,这将在能够进行连续检测的将来智能胰岛素投递设备中实现。Depending on the endocrine system studied, hormonal interactions are governed by dynamically regulated biochemical networks with complex or simple structures consisting of major nodes and conduits. Diabetes can perturb the network that regulates insulin-glucose dynamics at various levels. For example, in T1DM, the natural production of insulin is completely abolished, whereas in T2DM, insulin use within cells is hampered by more robust insulin resistance. In T1DM (and also common in T2DM), some form of external insulin replacement is required, making the regulatory system vulnerable to defective external factors such as the timing and dose of bolus and insulin injections, the food, physical activity, etc. This often results in hypoglycemia and hyperglycemia with extreme BG deviations. In many but not all cases, hypoglycemia triggers an endocrine response known as counterregulation. Therefore, in the field of mathematics, fluctuations of BG over time are measurable outcomes of the activity of complex dynamical systems influenced by a large number of internal and external factors. However, according to the well-known theory of dynamical systems, a purely deterministic system develops to display stochastic macroscopic behavior when the regulatory complexity increases. Thus, over short periods of time (minutes), BG fluctuations observed at the human level will be close to deterministic, whereas over the long term, fluctuations will be close to random, including extreme transitions such as SH events. Therefore, stochastic modeling and statistical inference are best suited for analyzing systems over longer periods of time—the paradigms adopted by
算法3在数小时到数天的中间时间范围内工作,所以要求结合统计推论和确定性建模。前者用于估计个体SH的基准危险,而后者用于动态跟踪个体参数和在SH事件出现之前进行预报。当在一个装置中执行时,算法3将如下地工作:
(1)装置收集研究对象的某些基准信息,建立个人基准参数;(1) The device collects certain benchmark information of the research object and establishes personal benchmark parameters;
(2)然后,装置开始追踪SMBG数据的某组特征;(2) Then, the device starts to track a certain set of characteristics of the SMBG data;
(3)装置具有判定规则,确定何时建立即将发生SH的标志,以及当数据表明危险减弱时何时降低该标志;(3) The device has a decision rule that determines when to establish a sign of impending SH and when to lower the sign when the data indicate that the hazard has diminished;
(4)当建立了标志时,我们认为对象在随后24小时内(预测时间)将接受到SH报警。(4) When the flag is established, we assume that the subject will receive a SH alert within the next 24 hours (predicted time).
这种动态预测同时在模型参数优化水平上和最佳方法的精度评估水平上产生了理论上的问题。我们开始阐述第二个问题,因为它对于理解算法3的作用是最重要的。This dynamic prediction creates theoretical problems both at the level of model parameter optimization and at the level of accuracy assessment of the best methods. We start by addressing the second question, since it is the most important for understanding what
算法3精度评估:虽然算法1和算法2使用静态预测,并且评估这些算法的标准在理论上是显然的——预测值更好,但是对于算法3,优化的标准不再是直接。这是因为在提高被预测SH事件的百分率的同时,我们不可避免地增加了“已建立标志”的数目,这反过来增加了潜在的“错误报警”的数目。因为“错误报警”并没有被清晰地定义,所以使问题进一步复杂化。在其纯粹意义上,错误报警是指建立了标志但随后并未发生SH事件。然而,如果人察觉到了症状并且采取了合适的行动,则SH是可以避免的。因此,即使可能存在SH的生化潜力,但事件可能并不出现。为了解决这个问题,我们采取了如下的优化标准:
(1)使24小时内将发生SH的预测最大化;(1) Maximize the prediction that SH will occur within 24 hours;
(2)“标记升高”和“标记降低”的持续时期的比值Rud最小化。(2) The ratio R ud of the duration of "mark up" and "mark down" is minimized.
尽管这两点的第一点是清晰的,但是第二点可能需要额外的解释。从在测量计中执行算法3的前景来看,在每次确定SMBG时,测量计都确定是否建立或者不建立即将发生SH的标记。当标记被建立时,它可能会持续一段时间(伴随着随后的数个SMBG读数)直到做出降低标记的决定为止。因此,我们将具有交替的“标志升高”和“标志降低”过程,其在SMBG点发生变化。参考上面点(2)的比值Rud是一个人标志升高时的平均时间除以标志降低时的平均时间。While the first of these two points is clear, the second may require additional explanation. From the perspective of implementing
在发明公开中给出的我们先前的最好结果是预测了44%的24小时内SH事件,且Rud=1∶7,即1天高危险警报与7天没有警报相交替。因为当时我们假定报警周期至少为24小时,所以算法被优化为建立标志的频率不超过每周1次。假如该分析是用具有高比率SH事件的对象的数据实现的,则这一比值是可以接受的。Our previous best result given in the Invention Disclosure predicted 44% of SH events within 24 hours with Rud = 1:7,
在本研究的实例No.1中,我们不得不使用相同的数据组完善算法3,因为没有其他可用的同时包括SMBG记录和SH记录的数据。我们也使用了相似的标准评估算法3的精度。然而,我们基本上改变了其他所有的事情。数据的跟踪、参数估计、全部阈值和判定规则都不再一样。这些变化是由于一个新的思路导致的,即在SH之前机体对反向调节的储备(reserve)会有一定的“损耗”,并且该损耗可以用SMBG数据进行跟踪。这一思路的确切执行在“判定规则”部分中有描述。因为判定规则包括一个连续的标准和一个在某种程度上的人为中止,所以存在多个解,我们选择了最佳的一个用于进一步的研究。然而,根据这些结果的表述,我们可以决定选择另一个解,从而在算法3的将来应用中加以执行。In case No. 1 of this study, we had to refine
结果摘要Summary of results
首先,要重点注意到下面给出的全部结果恰好超过统计的显著性。正如我们将在下一部分的几个实例中见到的,所观察到的差异一直非常显著(p值低于任何可能的显著水平)。算法3的观点是在个人的基础上预测SH事件的发生。结果是:First, it is important to note that all of the results presented below just exceed statistical significance. As we will see in several instances in the next section, the observed differences are consistently significant (p-values below any possible significance level). The idea of
(1)最小的基准观察周期是以每天3-4个读数的频率在大约两周的时间内获得50个SMBG读数。之后将每个对象分类到两个危险组中的一个,后者使用不同的判定规则;(1) The minimum baseline observation period is 50 SMBG readings taken over a period of approximately two weeks at a frequency of 3-4 readings per day. Each object is then classified into one of two risk groups, the latter using different decision rules;
(2)从我们具有的六个月数据中我们发现,在观察开始时足以进行该组的分配。因此,我们假定大约每六个月测量计将使用50个读数重新评价它自己的组分配;(2) From the six month data we have we find that at the beginning of the observations it is sufficient to do the group assignment. We therefore assume that approximately every six months the meter will re-evaluate its own group assignments using 50 readings;
(3)SMBG跟踪的最佳延迟是以每天3-4个读数的频率采集100-150个读数。换言之,最佳的判定标准是根据使用测量计存储器中全部150个读数进行的计算。这通过模拟ONE TOUCH ULTRA的存储容量而实现。一般地,使用在一周内获得的只有20个读数的延迟就能够获得良好的结果,但更长的延迟能够产生更好的预测;(3) The optimal latency for SMBG tracking is to acquire 100-150 readings at a frequency of 3-4 readings per day. In other words, the best decision criterion is based on calculations using all 150 readings in the meter's memory. This is achieved by simulating the storage capacity of ONE TOUCH ULTRA. In general, good results were obtained with a delay of only 20 readings taken within a week, but longer delays yielded better predictions;
(4)判定规则是根据新的计算程序,其使用“临时平均值”计算跟踪对象的低BG指数和其他相关参数。我们设计了专用软件用以执行该程序并处理我们能够得到的数据。从编程的观点看,执行该程序所需的编码只有大约20行,包括LBGI的计算;(4) The judgment rule is based on a new calculation program, which uses "temporary average" to calculate the low BG index and other related parameters of the tracked object. We have designed special software to carry out this procedure and process the data available to us. From a programming point of view, the code required to execute the program is only about 20 lines, including the calculation of LBGI;
(5)我们调查了多个判定规则(使用各种参数)。忽略SMBG频率,这些规则获得的24小时内SH预测从43.4%,Rud=1∶25到53.4%,Rud=1∶7。因此,与我们先前的结果相比,24小时内SH的预测提高了10%;(5) We investigated multiple decision rules (using various parameters). Neglecting SMBG frequency, these rules obtained SH predictions within 24 hours ranging from 43.4%, Rud = 1:25 to 53.4%, Rud = 1:7. Therefore, compared to our previous results, the prediction of SH within 24 hours is improved by 10%;
(6)作为进一步研究的最佳解,我们选择的判定规则可预测50%的24小时内SH,且Rud=1∶10。下面的结果参考了该最佳解在不同条件下的结果:(6) As the best solution for further research, the decision rule we choose can predict 50% of SH within 24 hours, and R ud =1:10. The following results refer to the results of this optimal solution under different conditions:
(7)SMBG的最佳频率是每天4个读数。如果达到了该频率,则24小时内SH的预测会增加到57.2%,并具有相同的Rud=1∶10。其他的SMBG频率也有调查和报道;(7) The optimal frequency of SMBG is 4 readings per day. If this frequency is achieved, the prediction of SH within 24 hours increases to 57.2%, with the same Rud = 1:10. Other SMBG frequencies have also been investigated and reported;
(8)如果我们将预测周期延长到36或48小时,则SH的预测分别增加到57%和63%,且具有同样的Rud=1∶10;(8) If we extend the prediction period to 36 or 48 hours, the prediction of SH increases to 57% and 63% respectively, and has the same R ud =1:10;
(9)利用基准信息能够显著增加SH的预测。实际上,比算法3先前的版本增加10%完全是由于使用了基准跟踪。但是,该基准跟踪现在被建模作为不使用患者任何额外输入的测量计在两周时期内的自我校正;(9) Utilizing benchmark information can significantly increase the prediction of SH. In fact, the 10% increase over the previous version of
(10)个人/人口统计信息,例如SH的历史或先前的HbA1c,对于更好的SH短期预测没有贡献;(10) Personal/demographic information, such as history of SH or previous HbA 1c , did not contribute to better short-term prediction of SH;
(11)任何时候当长时间没有SMBG活动时建立标志是不恰当的。只有测量计发出即将发生SH警报的次数才是使用次数。这是因为SH预测的主要部分是根据非常低BGs的重现(群集)。对该重现的估计在为2002年6月ADA会议准备的摘要中给出(Kovatchev等.具有大量SH历史的T1DM患者中周期性发生的低血糖症和严重低血糖症(SH))(见附录)。(11) Anytime there is no SMBG activity for an extended period of time it is inappropriate to establish a flag. Only the number of times the meter issued an impending SH alarm is counted as usage. This is because a major part of SH prediction is based on recurrences (clusters) of very low BGs. An estimate of this recurrence is given in an abstract prepared for the June 2002 ADA meeting (Kovatchev et al. Periodic hypoglycemia and severe hypoglycemia (SH) in T1DM patients with a history of extensive SH) (see appendix).
数据处理的详细说明Detailed description of data processing
测量计将SMBG读数与每个读数的日期和准确时间(小时,分,秒)存储在一起。因此,在训练数据组中,我们具有每个对象一定时间顺序的SMBG记录。在研究期间,从参与者的存储测量计中总共下载了75,495个SMBG读数(每人每天平均4.0±1.5个)。由对象的月记中,我们获得了所发生SH事件的日期和时间。对象报告了399次SH事件(每人4.7±6.0次)。有68位参与者(占80%)曾经历一次或者多次SH事件。根据他们的人口统计学特征,这些对象与那些没有经历过SH的人(剩余20%的对象)没有不同。The gauge stores SMBG readings with the date and exact time (hours, minutes, seconds) of each reading. Therefore, in the training data set, we have a certain chronological sequence of SMBG records for each object. A total of 75,495 SMBG readings were downloaded from participants' stored meters during the study period (average 4.0 ± 1.5 per person per day). From the subject's monthly diary, we obtained the date and time of the SH event that occurred. Subjects reported 399 SH events (4.7 ± 6.0 per subject). 68 participants (80%) experienced one or more SH events. Based on their demographic characteristics, these subjects were no different from those who had not experienced SH (the remaining 20% of subjects).
数据预处理: Data preprocessing :
我们开发了专用的软件用于数据预处理。这包括:(1)将每个对象的存储测量计数据与BG读数的连续6-8个月序列组合在一起,和(2)按照日期和时间使每位对象的SH记录与该序列相匹配。后者的执行如下:为每个SMBG读数计算到最近一次SH事件的时间(小时/分)和从上一次SH事件经历的时间。因此有可能:(1)每次SH事件之前或之后的24小时或48小时等时间周期,(2)SMBG读数之间的时间周期。由于SH(昏迷,无意识)的本质,在SH时不会有SMBG,因此出于算法3的目的,SH事件不包括用于算法2的生化显著低血糖症。平均每个SH事件在SH与之前最近SMBG读数之间的最小时间间隔为5.2±4.1小时;有29次SH事件(占7%)在之前15分钟内具有SMBG读数。对每个SH事件,我们计数了在该事件之前24h、36h、48h和72h内执行SMBG读数的次数。We developed dedicated software for data preprocessing. This included: (1) combining each subject's stored gauge data with a continuous 6-8 month sequence of BG readings, and (2) matching each subject's SH record to that sequence by date and time . The latter was performed as follows: time to last SH event (hours/minutes) and elapsed time since last SH event were calculated for each SMBG reading. It is thus possible: (1) time periods such as 24 or 48 hours before or after each SH event, (2) time periods between SMBG readings. Due to the nature of SH (comatose, unconscious), there will be no SMBG during SH, so for the purposes of
基准危险值的计算和自我校正: Calculation and self-correction of baseline risk values :
根据他/她的第一SMBG读数为每个对象计算低BG指数。已经确定,计算基准LBGI所需的最少读数是在大约2周内采集50个。因此对于每个新的测量计,我们需要期望一段最初两周的自我校正时期,在此期间测量计将扫描其拥有者的SH总体危险。初始期之后,人们被分到两个危险组中的一个:低中度危险(LBGI≤3.5,LM组)或者中高度危险(LBGI>3.5,MH组)。我们的测试数据显示,更精确的分组是不必要的。这种分组允许在LM和MH组中使用不同的判定规则,从而与本发明公开中给出的其最初命中率相比,算法的命中率提高了大约10%。A low BG index was calculated for each subject based on his/her first SMBG reading. It has been determined that the minimum number of readings required to calculate a baseline LBGI is 50 readings taken over approximately 2 weeks. So with each new meter we need to expect an initial two week self-calibration period during which the meter will scan its owner's SH general hazard. After the initial period, people were assigned to one of two risk groups: low-intermediate risk (LBGI < 3.5, LM group) or intermediate-high risk (LBGI > 3.5, MH group). Our test data shows that more precise grouping is unnecessary. This grouping allows the use of different decision rules in the LM and MH groups, thereby increasing the hit rate of the algorithm by about 10% compared to its original hit rate given in the present disclosure.
使用测试数据,不需要重新校正基准危险。因此我们能够假定,如果人在治疗中没有经历变化,则大约每六个月进行一次重新校正。这与算法2的结果是一致的,其显示在初始观察周期之后的6个月,SH的长期预测是非常准确的。Using test data, there is no need to recalibrate the baseline hazard. We can therefore assume that recalibration occurs approximately every six months if the person experiences no changes during treatment. This is consistent with the results of
然而,如果人在其血糖控制中经历了迅速的变化,则可能需要更加频繁地重新校正。重新校正的判定可能是自动的,并且根据所观察到的运行危险值(见下段)与基准LBGI之间的差异不断增加。然而,可获得的数据并不能允许我们阐明这个问题,因为我们所观察对象的低血糖症危险并没有显著变化。However, more frequent recalibration may be required if a person experiences rapid changes in their blood sugar control. The decision to recalibrate may be automatic and incrementally based on the observed difference between the operational risk value (see next paragraph) and the baseline LBGI. However, the available data did not allow us to shed light on this issue, as the risk of hypoglycemia did not change significantly among our subjects.
计算SMBG参数:在预处理步骤之后,我们设计了另一个软件来计算用于预测即将发生的SH的SMBG参数。该软件包括: Calculation of SMBG parameters : After the preprocessing step, we designed another software to calculate the SMBG parameters for predicting the upcoming SH. The software includes:
(1)为每个BG读数计算低BG危险值(RLO),这通过如下的编码实现(这里BG以mg/dl为单位测量,如果单位是mmol/l则系数会不同):(1) Calculation of risk of low BG (RLO) for each BG reading, which is achieved by the following code (here BG is measured in mg/dl, if the unit is mmol/l the coefficient will be different):
scale=(In(bg))**1.08405-5.381scale=(In(bg))**1.08405-5.381
risk=22.765*scale*scalerisk=22.765*scale*scale
if(bg_1≤112.5)thenif(bg_1≤112.5)then
RLO=riskRLO = risk
elseelse
RLO=0RLO=0
endifendif
(2)为每个具有序列数n的SMBG读数,BG(n),计算LBGI(n)的运行值,和另一个统计量,SBGI(n),其是低BG危险值的标准偏差。这两个参数用每个SMBG读数之后的某个延迟(k)加以计算,例如包括该读数,BG(n),和BG(n)之前采集的(k-1)个读数。(2) For each SMBG read with sequence number n, BG(n), compute a running value of LBGI(n), and another statistic, SBGI(n), which is the standard deviation of low BG risk values. These two parameters are calculated with some delay (k) after each SMBG reading, eg including that reading, BG(n), and (k-1) readings taken before BG(n).
(3)LBGI(n)和SBGI(n)的计算采用一个新临时方法程序(newprovisional means procedure),其基于如下的递归式编码:(3) The calculation of LBGI(n) and SBGI(n) adopts a new provisional means procedure, which is based on the following recursive coding:
初值在n-k(或者精确地在最大值(1,n-k),以便解释序数小于k的测量计读数):Initial value at n-k (or precisely at maximum value (1,n-k) to account for meter readings with ordinals less than k):
LBGI(n-k)=rlo(n-k)LBGI(n-k)=rlo(n-k)
rlo2(n-k)=0rlo2(n-k)=0
对于n-k与n之间任何连续迭代j的数值:For the value of any successive iteration j between n-k and n:
LBGI(j)=((j-1)/j)*LBGI(j-1)+(1/j)*RLO(j)LBGI(j)=((j-1)/j)*LBGI(j-1)+(1/j)*RLO(j)
rlo2(j)=((j-1)/j)*rlo2(j-1)+(1/j)*(RLO(j)-LBGI(j))**2rlo2(j)=((j-1)/j)*rlo2(j-1)+(1/j)*(RLO(j)-LBGI(j))**2
在该循环完成之后,我们得到了LBGI(n)的值,并计算After this loop completes, we get the value of LBGI(n) and calculate
SBGI(n)=sqrt(rlo2(n))SBGI(n)=sqrt(rlo2(n))
因为对于ONE TOUCH ULTRA测量计,最大值n是150,所以在k=10到k=150范围内寻找最佳的延迟k。虽然性能上的差异并不显著,但是最佳的延迟确定为k=150(见例如下一部分)。Since the maximum value n is 150 for the ONE TOUCH ULTRA meter, the optimum delay k is found in the range k=10 to k=150. Although the difference in performance is not significant, the optimal delay was determined to be k=150 (see eg next section).
判定规则:在每次SMBG读数时,程序判定是否建立警告即将发生SH的标志。如果已经建立了标志,则程序确定是否使其降低。这些判定是根据三个阈值参数:α、β、γ,其运行如下: Decision rule : At each SMBG reading, the program determines whether to set a flag warning that SH is about to occur. If the flag is already set, the program determines whether to lower it. These decisions are based on three threshold parameters: α, β, γ, which operate as follows:
对于对低中度危险(LM组)的对象:For subjects at low to moderate risk (LM group):
FLAG=0FLAG=0
if(LBGI(n)≥α and SBGI(n)≥β)FLAG=1if(LBGI(n)≥α and SBGI(n)≥β)FLAG=1
if(RLO(n)≥(LBGI(n)+γ*SBGI(n)))FLAG=1if(RLO(n)≥(LBGI(n)+γ*SBGI(n)))FLAG=1
对于中高度危险组的对象,只有第二if叙述起作用。换言之,如果LBGI(n)的运行值和它的标准偏差SBGI(n)都超过了某个阈值,则建立标志(即等于1),而且如果低BG危险RLO(n)的当前值超过LBGI(n)加γ标准偏差的值则标志也建立。For subjects in the medium to high risk group, only the second if statement worked. In other words, if both the operating value of LBGI(n) and its standard deviation SBGI(n) exceed a certain threshold, a flag is established (i.e. equal to 1), and if the current value of low BG hazard RLO(n) exceeds LBGI( n) Add the value of gamma standard deviation and the flag is also established.
一个启发性的解释:LBGI(n)知SBGI(n)的数值反映了低血糖症危险的更缓慢变化——要经过数天的SMBG才能显著改变这些数值。因为LBGI(n)平均值越高,近期低血糖症越频繁和极端,所以我们可得出结论,LBGI(n)和SBGI(n)反映了反向调节储备在数天的时期内持续的损耗(或者缺乏补给)。此外,SBGI(n)是系统稳定性的一个标志——较大的SBGI(n)表明对象的BG波动增加,因此控制系统变得不稳定和容易受到极端异常的影响。因此,第一逻辑表达式反映的观点是,无论何时当反向调节防御被耗尽且控制(外部或内部)变得不稳定时就会发生SH。第二逻辑表达式说明了低BG危险的急剧改变,无论何时当目前低BG危险值突然大于其运行平均值时则会发出标志。实际上对于中高度危险组的对象,只有第二逻辑表达式与这些对象的最终“永久耗尽”和“永久不稳定”状态相对一致。因为这些对象连续运行低BG值,并且他们的BG是不稳定的,所以任何急剧的低血糖事件都可能激发SH。大体上,严重低血糖症标志或者低不稳定BG周期之后建立,或者在急性低血糖事件之后建立,急性低血糖事件严重偏离了(在危险空间内)最近的运行危险平均值(可能早经很高)。这些之前没有任何报警信号的SH事件仍然不能用本算法加以解释。在下面的表5C中,我们给出了解释算法3对多个对象的作用的样本输出: A heuristic explanation : LBGI(n) and SBGI(n) values reflect more gradual changes in the risk of hypoglycemia—it takes days of SMBG to change these values significantly. Because the higher the LBGI(n) mean, the more frequent and extreme the recent hypoglycemia, we can conclude that LBGI(n) and SBGI(n) reflect a sustained depletion of counter-regulatory reserves over a period of several days (or lack of supplies). Furthermore, SBGI(n) is an indicator of system stability - a larger SBGI(n) indicates that the BG fluctuations of the subject increase and thus the control system becomes unstable and susceptible to extreme anomalies. Thus, the first logical expression reflects the idea that SH occurs whenever counterregulatory defenses are exhausted and controls (external or internal) become unstable. The second logical expression accounts for a sharp change in low BG hazard, and a flag is raised whenever the current low BG hazard value is suddenly greater than its running average. In fact, only the second logical expression is relatively consistent with the final "permanently exhausted" and "permanently unstable" states of these subjects for the subjects in the medium-high risk group. Because these subjects run continuously with low BG values, and their BG is unstable, any acute hypoglycemic event may trigger SH. In general, hallmarks of severe hypoglycemia are established after periods of low unstable BG, or after acute hypoglycemic events that deviate significantly (within the hazard space) from the most recent running hazard average (possibly long ago high). These SH events without any previous warning signal still cannot be explained by this algorithm. In Table 5C below, we present sample output explaining the effect of
表5C:解释算法3对多个对象的作用的样本输出:
该输出的每一行都给出了一个SMBG读数,或一个SH事件(没有读数)。ID是对象的ID号码,BG是以mg/dl计的BG水平,当发生SH事件时SH=1。如果算法3决定建立flag则FLAG=1;TIME是以小时计的最近一次SH事件的时间。Each line of this output gives an SMBG reading, or a SH event (no reading). ID is the ID number of the subject, BG is the BG level in mg/dl, and SH=1 when a SH event occurs. If
临时方法程序延迟的优化:在先前的出版物中,我们报道了在SH之前的48到24小时周期内,平均BG水平降低而BG方差增加。在SH之前的24小时周期内,平均值BG水平进一步降低,BG的方差继续增加,LBGI急剧增加。在SH之后的24小时周期内,平均BG水平正常化,而BG方差仍大幅增加。在SH之后的48小时内,BG的平均值和方差返回到基准水平(见Kovatchev et,al.1型糖尿病中严重低血糖症事件之前和之后48小时内可测得的血糖扰动(Episodes ofSevere Hypoglycemia in Type 1 Diabetes are Preceded,and Followed,within 48 Hours by Measurable Disturbances in Blood Glucose).
J of Clinical Endocrinology and Metabolism,85:4287-4292,2000)。我们现在使用这些观察值根据在SH之前24小时内观察到的LBGI(n)和SBGI(n)的平均值优化算法3所采用的临时方法程序的延迟,k。简言之,用于计算LBGI(n)和SBGI(n))的延迟被选择使SH之前24小时内的这些测量显示与其余的研究相比最大化,但不包括SH之后系统仍处于不平衡的时期。发现最佳延迟是k=150。表6A和6B给出了LBGI(n)和SBGI(n)对于参数k的多个值和对于低中度危险组和中高度危险组两个对象组的平均值。显然,k的各个值之间的差异不大,因此实际使用任何k≥10的数值都是合适的。然而,根据目前的数据,我们推荐k=150,并且所有进一步的计算使用该延迟。该推荐还是根据LBGI(n)和SBGI(n)的方差岁延迟值的增大而减小,这能够由下面更大的t-值加以反映: Optimization of interim method procedural delay : In a previous publication, we reported that mean BG levels decreased while BG variance increased during the 48 to 24 h period preceding SH. During the 24-h period preceding SH, the mean BG level decreased further, the variance of BG continued to increase, and the LBGI increased sharply. During the 24-h period following SH, mean BG levels normalized, while BG variance remained substantially increased. The mean and variance of BG returned to baseline levels within 48 hours after SH (see Kovatchev et al. in
表6A:不同延迟下SH之前24小时内与其余时间的LBGI(n)Table 6A: LBGI(n) in the 24 hours before SH and the rest of the time at different delays
表6B:不同延迟下SH之前24小时内与其余时间的SBGI(n)Table 6B: SBGI(n) in the 24 hours before SH and the rest of the time at different delays
*最佳解 * best solution
从表6A和6B可以看出,LBGI和SBGI在SH之前的24小时内都高度显著地增加。因此,人们试图运行一种直接的判别式或者对数模型来预测即将发生的SH。不幸的是,这种标准统计的效果并不很好,尽管两种模型在统计上都非常显著。判别式模型(其效果比对数回归要好)能够正确预测52.6%的即将发生的SH事件。然而,其标志升高(flag-up)与标志降低(flag-down)的比非常差——Rud=1∶4。因此,该模型倾向于更大数量的数据点,这是在任何统计程序中都会预见到的倾向。因此,我们不得不必须采用上面给出的判定规则。As can be seen from Tables 6A and 6B, both LBGI and SBGI were highly significantly increased in the 24 hours before SH. Therefore, people try to run a direct discriminant or logarithmic model to predict the upcoming SH. Unfortunately, this standard statistic doesn't work very well, although both models are statistically very significant. The discriminative model (which performed better than log regression) was able to correctly predict 52.6% of upcoming SH events. However, the ratio of flag-up to flag-down is very poor - Rud = 1:4. Therefore, the model favors a larger number of data points, a tendency to be expected in any statistical procedure. Therefore, we have to adopt the decision rules given above.
严重低血糖症的预测精度Prediction Accuracy of Severe Hypoglycemia
阈值参数α
、β
知γ
的优化 下面我们详细说明了使用阈值参数α、β和γ的各种组合的算法3的预测能力。因为这些参数与所期望结果(高度预测SH且比值Rud最小)之间的关系非常复杂,因此我们所采用的优化程序不能够获得单一的解。此外,似乎也不需要获得单一的解。它似乎是一个商业的而非数学的判定,即在给定的“标志升高”与“标志降低”的比值之下,获得可以接受的SH预测百分率。因此,我们并不声明下面给出的任何解是最优解。但是,为了进一步探索这个主题,我们接受能预测50%的将来SH且Rud=1∶10作为基准,用于研究除24小时之外的预测周期以及为了获得更好的危险分布(profile)所需的每天SMBG读数个数的各种要求。 Optimization of Threshold Parameters α , β and γ Below we detail the predictive power of
表7给出了算法3在α、β和γ的各种数值组合下的性能,这些组合代表了SH的预测百分率(命中率)与我们称之为“烦扰指数”(annoyance index)的比值Rud之间的关系。表7还包括在研究期间每个对象在被警告与无警告状态下经历的平均总时间(以天为单位),也就是说,用阐明了比值Rud意义的算法对研究对象经历的被警告-无警告周期交替过程的摘要结果。Table 7 shows the performance of
表7:SH预测:命中率、烦扰指数和平均时间Table 7: SH Prediction: Hit Rate, Annoyance Index and Average Time
最好的解用于进一步的分析。假定本研究中的参与者平均经历4.7次SH事件,那么如果报警可以预防50%的SH,则19天的高报警期似乎是可以接受的。此外,高报警期趋向于群集地(in clusters)来临。因此,我们能够假设,在实践中,长而且相对平静的时期会与少数几天高危险报警期交替。表7中的最后一行给出了Rud=1∶7的解,其与本发明公开中给出的解相当。然而,当前的解具有大约高10%的命中率:53.4%对我们先前算法的44%。当命中率与我们先前的算法相当时,烦扰比小于1∶20,也就是说,好了3倍。The best solution is used for further analysis. Assuming that participants in this study experienced an average of 4.7 SH episodes, a high alarm period of 19 days would seem acceptable if alarming would prevent 50% of SH. Furthermore, periods of high alarm tend to come in clusters. We can therefore assume that, in practice, long periods of relative calm alternate with periods of high-risk alarm of a few days. The last row in Table 7 gives the solution for Rud = 1:7, which is comparable to the solution given in the present disclosure. However, the current solution has about 10% higher hit rate: 53.4% vs. 44% of our previous algorithm. When the hit rate is comparable to our previous algorithm, the annoyance ratio is less than 1:20, that is, 3 times better.
图14给出了以百分数表示的命中率与比值Rud之间的平滑依赖关系。显然,当算法3的命中率增加时,“标志升高”与“标志降低”之间的比值迅速增加。因此,给定这些数据,寻找能够产生高于50%命中率的参数组合是不恰当的:Figure 14 presents the smoothed dependence of the hit rate expressed as a percentage on the ratio R ud . Obviously, when the hit rate of
可选择的预测周期:在开始说明算法3时,我们制定了基本假设,如果在SH事件之前的24小时内建立了标志,则认为SH事件被预测到了。该假设产生了在前一部分报告的命中率。现在我们根据范围从12小时到72小时的其他预测周期计算命中率。在整个实验中,参数α、β和γ被分别固定为5.0、7.5和1.5,也就是,固定在它们在表7中最佳解的数值。因此,标志升高比值与该解保持相同,Rud=1∶10,只是命中率发生了变化,因为我们改变了命中的定义。图15给出了预测周期与相应命中率之间的依赖关系。 Alternative prediction periods : At the beginning of the illustration of
显然,随着预测周期增加到大约24小时,命中率迅速增加,然后命中率的增加缓慢下降。因此,我们能够得出结论,提前24小时是最佳的且合理的预报周期。Clearly, as the prediction period increases to about 24 hours, the hit rate increases rapidly, and then the increase in hit rate decreases slowly. Therefore, we can conclude that 24 hours ahead is the optimal and reasonable forecast period.
每天SMBG读数的最佳数目最后,我们进行了实验,研究为了产生最佳的SH预报每天需要多少个读数。 Optimal number of SMBG readings per day Finally, we conducted experiments to investigate how many readings per day are required to produce optimal SH forecasts.
正如我们在开始时说的,总共报告了399个SH事件。这些事件中,343个具有24小时之前的SMBG读数(还有3个事件具有之前48小时内的读数,另有4个事件具有之前72小时的读数)。另外有超过50个SH事件(14%)没有任何有助于预测的合理预先SMBG读数。在先前24小时内至少具有一个SMBG读数的343个事件用于计算前一部分中的命中率。其余事件自然被排斥在计算之外。As we said at the beginning, a total of 399 SH events were reported. Of these events, 343 had SMBG readings 24 hours earlier (3 more events had readings within the previous 48 hours, and 4 events had readings within the previous 72 hours). There were also more than 50 SH events (14%) without any plausible prior SMBG readings to aid in prediction. 343 events with at least one SMBG reading in the previous 24 hours were used to calculate the hit rate in the previous part. The remaining events are naturally excluded from the calculation.
进一步的分析显示,随着SH事件之前采集的读数的增加,命中率快速增加。然而,如果我们强行严格要求一定数目的可获得读数以便考虑一个SH事件,则我们会发现,满足这一要求的SH事件数目会迅速下降(表8)。这是由于对象不遵守研究的要求,并且可能是在将来的测量计中出现某种警报的良好理由,即如果没有以合适的速度采集SMBG读数,则算法3将不再有用并且测量计会被关闭。Further analysis revealed a rapid increase in hit rate as the number of reads acquired prior to the SH event increased. However, if we imposed a strict requirement for a certain number of available reads in order to consider an SH event, we found that the number of SH events satisfying this requirement dropped rapidly (Table 8). This is due to subjects not adhering to the requirements of the study, and may be a good reason for some kind of alarm in future meters that if SMBG readings are not taken at an appropriate rate,
表8给出了具有一定数目先前SMBG读数的SH事件的数目以及算法3对这些事件的命中率。表中最佳的行包括了表7的最优解,其作为后面所有计算的基础。所有的命中率都是以24小时预测周期给出的,也就是,SH之前24小时内的标志。我们能够得出结论,随着对象顺从性的增加,算法3预测SH的精度也显著增加。每天进行5次SMBG读数,则精度从基准的50%命中率提高10%:Table 8 gives the number of SH events with a certain number of previous SMBG readings and the hit ratio of
表8:当给出一定数目的先前SMBG读数时,算法3的性能Table 8: Performance of
被测试的其他可能的改善Other possible improvements tested
通过包含外部参数,例如前一年SH的数目或者HbA1c的基准值,提高算法3预测能力的尝试并不成功。显然,SH的短期预测主要取决于当前或近期的事件。然而,该研究的限制条件是,所有的参与者在前一年经历≥2次SH。Attempts to improve the predictive power of
最后,我们检验了当探测到低血糖症的危险增加时并且当时已经有一段时间没有SMBG,则是否应当发出SH的报警。这样做是想预测至少一部分之前没有SMBG读数的SH事件。但并不成功,所产生的主要是错误的报警。这一结果更加证实了遵守SMBG协议的重要性,该协议要有足够频繁的SMBG读数。Finally, we examined whether an SH alarm should be issued when an increased risk of hypoglycemia is detected and there has been no SMBG for a period of time. This was done in an attempt to predict at least a portion of SH events for which there were no previous SMBG readings. But it was not successful, and the result was mainly false alarms. This result further confirms the importance of following the SMBG protocol with sufficiently frequent SMBG readings.
附录:摘要Appendix: Summary
实例No.1估计了低血糖症(BG<3.9mmol/l)事件之后再发生低血糖症和SH(定义为无法进行自我治疗的昏迷或者无意识)的频率。Example No. 1 estimates the frequency of reoccurrence of hypoglycemia and SH (defined as comatose or unconscious without self-medication) following episodes of hypoglycemia (BG < 3.9 mmol/l).
85名在上一年经历>2次SH事件的T1DM患者(41名女性),每天进行3-5次SMBG达6-8个月,并根据日期和时间记录SH事件。对象的平均年龄为44±10岁,糖尿病的持续时间为26±11年,HbA1c为7.7±1.1%。Eighty-five T1DM patients (41 women) who experienced >2 SH events in the previous year underwent SMBG 3-5 times a day for 6-8 months, and recorded SH events according to date and time. The mean age of the subjects was 44±10 years, the duration of diabetes was 26±11 years, and the HbA 1c was 7.7±1.1%.
所有的SMBG读数(75,495)按照日期和时间与对象的SH事件结合(n=399;SH事件一般没有相应的SMBG读数)。为每个SMBG读数或者SH事件计算从最近的上一次低BG(<3.9mmol/l)之后经历的时间。下面的表9给出了3个低血糖范围:BG<1.9mmol/l、1.9-2.8mmol/l和2.8-3.9mmol/l中读数的百分比,以及在之前24小时、24-48小时、48-72小时以及超过72小时具有低BG读数(BG<3.9mmol/l)的SH事件的百分比。最后一列给出了运行检验,其拒绝了如下的假设,即包含低BG读数(或者SH事件)的日期在整个时间范围上随机分布。检验的负Z值显示出具有和没有低血糖读数或者SH事件的日期“群集出现”。All SMBG readings (75,495) were combined by date and time with the subject's SH events (n=399; SH events generally had no corresponding SMBG readings). The time elapsed since the last low BG (<3.9 mmol/l) was calculated for each SMBG reading or SH event. Table 9 below gives the percentage of readings in the 3 hypoglycemic ranges: BG<1.9mmol/l, 1.9-2.8mmol/l and 2.8-3.9mmol/l, and the percentage of readings in the previous 24 hours, 24-48 hours, 48 - Percentage of SH events with low BG readings (BG < 3.9 mmol/l) at 72 hours and beyond. The last column presents a running test that rejects the hypothesis that dates containing low BG readings (or SH events) are randomly distributed over the entire time range. Negative Z values for the test show "clustering" of days with and without hypoglycemia readings or SH events.
表9.之前具有低RG的低血糖症/SH的百分比:Table 9. Percentage of hypoglycemia/SH with previous low RG:
我们得出结论,有超过一半的低血糖SMBG读数和大约2/3的SH事件在其之前24小时内有至少一个低血糖读数。此外,低血糖事件趋向于群集地出现。因此,最初的低血糖事件可能是即将再次出现低血糖症的报警信号。We concluded that more than half of the hypoglycemic SMBG readings and about two-thirds of the SH events had at least one hypoglycemic reading within 24 hours before them. In addition, hypoglycemic events tend to occur in clusters. Thus, an initial hypoglycemic episode may be a warning sign of impending recurrent hypoglycemia.
II.实例No.2II. Example No.2
本发明采用日常的自我监测血糖(SMBG)数据,并且直接适合于通过引入能够预测HbA1c和显著低血糖症高危期的智能数据判读逻辑提高家用SMBG设备的性能。该方法包括两个部分:(1)算法1估计HbA1c,和(2)算法2&3分别预测长期和短期(24小时内)的显著低血糖症。在该报告中,我们描述了提出、优化和验证HbA1c估计算法1的步骤以及其在估计实验室获得的HbA1c的精度。The present invention employs daily self-monitoring blood glucose (SMBG) data and is directly suitable for improving the performance of home SMBG devices by introducing intelligent data interpretation logic capable of predicting HbA 1c and high-risk periods of significant hypoglycemia. The method consists of two parts: (1)
目标:Target:
我们的主要目的是达到有95%的测量结果位于实验室参考值±1HbA1c单位内的精度,这是HbA1c化验精度的国家糖基化血红蛋白标准化大纲(NGSP)准则(National Glycohemoglobin StandardizationProgram Criterion)。Our primary objective was to achieve a precision of 95% of measurements within ±1 HbA 1c unit of the laboratory reference value, which is the National Glycohemoglobin Standardization Program Criterion (NGSP) for HbA 1c assay precision.
方法: method :
对象:SMBG数据是由100名1型糖尿病对象和100名2型糖尿病对象(T1DM,T2DM)分别经过6个月和4个月获得的,并且在T1DM中于第0、3和6个月进行HbA1c检测,而在T2DM中于第0、2和4个月进行检测。 Subjects : SMBG data were obtained from 100 subjects with
算法1的提出和优化:训练数据组包括由T1DM收集了3个月的SMBG和HbA1c数据,和由T2DM收集了2个月的SMBG和HbA1c数据。这些训练数据用于优化算法1,和用于评估大量的用于保证更高精度的样本选择标准。样本选择标准是对通过测量计收集的SMBG样本的要求,如果能够满足则能够保证由样本精确地估计HbA1c。因此,测量计会扫描每一个SMBG样本,如果满足样本选择标准,则计算和显示HbA1c估计。在经过分析大量的切点(cut point)之后,我们选择了如下的标准: Proposal and optimization of Algorithm 1 : The training data set included SMBG and HbA 1c data collected by T1DM for 3 months, and SMBG and HbA 1c data collected by T2DM for 2 months. These training data are used to optimize
1.测试频率:为了产生HbA1c的估计,测量计需要在过去的60天内平均每天进行2.5次或者更多次测试,也就是,在过去的两个月内进行总共要有150个SMBG读数。注意这是每天的平均数,并未要求每天进行测试,这一点是重要的。1. Testing Frequency: To generate an estimate of HbA 1c , the meter needs to take an average of 2.5 or more tests per day over the past 60 days, ie, a total of 150 SMBG readings over the past two months. It is important to note that this is a daily average and does not require daily testing.
2.数据随机化:60天样本中只具有餐后测试或者夜间测试不够(<3%样本)的样本将被排除。此外还包括避免高度集中在每天的一个最常用的时间内进行的测试。这些标准本报告中有详细说明。2. Data randomization: 60-day samples with only post-meal tests or insufficient overnight tests (<3% of samples) will be excluded. It also includes avoiding tests that are highly concentrated during one of the most commonly used times of the day. These criteria are detailed in this report.
结果:算法的前瞻性验证和精度: Results: Prospective Validation and Accuracy of Algorithms :
然后,将算法,包括样本选择标准,应用于测试数据组1和独立测试数据组2,其中测试数据组1包括T1DM和T2DM对象在上一次HbA1c测试之前2个月内的SMBG和HbA1c数据,而独立测试数据组2由参加了先前NIH研究的60名T1DM对象的数据构成。出于验证的目的,将由算法1获得的估计与参考HbA1c水平进行比较。在测试数据组1中,算法达到了NGSP标准,位于实验室参考值±1HbA1c单位内的精度为95.1%。在测试数据组2中,算法达到了NGSP标准,同样位于实验室参考值±1HbA1c单位内的精度为95.1%。样本选择标准的研究显示,有72.5%的对象能够每天产生一个这样精确的估计,有94%的对象能够大约每5天产生1个这样精确的估计。Then, the algorithm, including sample selection criteria, was applied to test data set 1 and independent test data set 2, where test data set 1 included SMBG and HbA 1c data for T1DM and T2DM subjects within 2 months before the last HbA 1c test , while the independent test data set 2 consisted of data from 60 T1DM subjects who participated in a previous NIH study. For validation purposes, the estimates obtained by
结论:日常SMBG数据允许精确地估计HbA1c,并且满足直接HbA1c化验精度的NGSP标准。 Conclusions: Routine SMBG data allow accurate estimation of HbA 1c and meet NGSP criteria for direct HbA 1c assay accuracy.
对象&取舍标准Objects & selection criteria
我们得到了100名患有1型糖尿病(T1DM)的对象和100名患有2型糖尿病(T2DM)的对象的允许。179名对象完成了SMBG数据收集的主要部分,其中90名患有T1DM,89名患有T2DM。这179名对象的数据用于检验算法2和3。然而,检验算法1要求对象不仅具有SMBG数据,而且要有在SMBG之前60天内采集的HbA1c数据和SMBG记录。在本研究的第3个月(对于T2DM是死2个月),153名对象(78患有T1DM)完成了满足上述标准的HbA1c数据和SMBG记录。此外,我们使用N=60名患有T1DM的对象的数据检验算法1,这些对象参与了我们先前的NIH研究(NIH)。表10给出了所有对象的人口统计学特征。We obtained consent for 100 subjects with
表10:对象的人口统计学特征Table 10: Demographic Characteristics of Subjects
观察到的测量计误差Observed meter error
我们的调查显示,HbA1c化验或者其他化验之前60天内没有完成数据的主要原因不是对象不顺从,而是测量计失效。显然如果患者按下“M”按钮太长时间,则ONE TOUCH ULTRA测量计的时间和日期会“跳转”到随机的日期/时间(例如2017年11月)。一返回我们就检查每个测量计的日期/时间,我们发现在整个研究过程中有60个测量计发生了这种事件。时间/日期的偏移影响了15,280个读数,或者将近全部读数的10%。我们单独保存这些读数并让一名学生检查它们。在许多情况下,但并非在所有的情况下,他都能够恢复读数的日期/时间序列。该误差以及在邮寄过程中丢失了少数测量计,导致具有用于分析算法1的良好数据的对象数目从179减少到了140。恢复了12名对象的数据,最终采集到了153名对象的数据,其中78名患有T1DM,75名患有T2DM,他们在HbA1c之前的数据的时间序列没有被扰乱,因此适合于检验算法1。Our investigations indicated that the main reason for HbA 1c assays or other assays not completed within 60 days prior to data was not subject non-compliance but meter failure. Apparently if the patient presses the "M" button for too long, the time and date of the ONE TOUCH ULTRA meter will "jump" to a random date/time (e.g. November 2017). Upon return we checked the date/time of each meter and we found that 60 meters had this event over the course of the study. The time/date offset affected 15,280 readings, or nearly 10% of all readings. We keep these readings individually and have a student review them. In many cases, but not all, he was able to recover the date/time sequence of the readings. This error, along with the loss of a few gauges in the mail, reduced the number of subjects with good data for
程序program
所有的对象都填写了IRB推荐的准许表格并参加了指导会(orientation meeting),在会上给他们介绍了ONE TOUCH ULTRA测量计并让他们完成了屏幕问卷。指导会之后,所有的对象立即参观UVA实验室并被采血获得了基准HbA1c。T1DM对象进行了6个月,于第3和第6月进行实验室HbA1c化验;T2DM对象进行了4个月,于第2和第4月进行实验室HbA1c化验。自我监测(SMBG)数据被规律地从测量计下载下来并保存在数据库中。每两周通过自动e-mail/电话跟踪系统对显著低血糖和高血糖事件进行平行记录。All subjects completed the IRB recommended clearance form and attended an orientation meeting where they were introduced to the ONE TOUCH ULTRA meter and completed an on-screen questionnaire. Immediately after the orientation meeting, all subjects visited the UVA laboratory and were blood drawn to obtain baseline HbA 1c . T1DM subjects were treated for 6 months, and laboratory HbA 1c tests were performed on the 3rd and 6th months; T2DM subjects were treated for 4 months, and laboratory HbA 1c tests were performed on the 2nd and 4th months. Self-monitoring (SMBG) data is regularly downloaded from the gauges and stored in a database. Parallel records of significant hypoglycemic and hyperglycemic events were performed every two weeks through an automated e-mail/telephone tracking system.
数据存储和清理Data Storage and Cleaning
T1DM和T2DM对象的ONE TOUCH ULTRA粗数据被分别存储在InTouch数据库中。使用定制开发的软件清理这些粗数据的对象和测量计误差,在某些情况下进行手动数据清理(见上面的测量计误差)。当不可能进行修正时,则放弃该数据。ONE TOUCH ULTRA raw data for T1DM and T2DM objects are stored separately in the InTouch database. These crude data were cleaned of object and gauge errors using custom-developed software, and in some cases manual data cleaning (see gauge errors above). When correction is not possible, the data is discarded.
为了保证我们的优化结果能够一般化到全部人口的水平,首先用训练数据组优化算法,然后用测试数据组进行检验。In order to ensure that our optimization results can be generalized to the level of the entire population, the algorithm is first optimized with the training data set, and then tested with the test data set.
训练数据组包括T1DM对象在确定第3个月HbA1c之前采集的60天的SMBG数据。该数据组用于优化算法1的公式。T2DM对象在确定他们第2个月HbA1c之前收集的数据用于鉴定样本选择标准,该标准在T1DM数据中不明显。然而,T2DM对象的数据不用于优化算法1的公式。含有这些数据的文件是PASS01.DAT。 The training data set consisted of 60 days of SMBG data collected from T1DM subjects prior to determination of HbA 1c at 3 months. This data set is used to optimize the formula of
测试数据组1包括T1DM对象在确定第6个月HbA1c之前,T2DM对象在确定第4个月HbA1c之前采集的60天SMBG数据。下面我们将这些数据称作数据组1。含有这些数据的文件是PASS02.DAT。 Test data set 1 includes 60-day SMBG data collected for T1DM subjects before determining HbA 1c at
测试数据组2包括先前NIH研究中N=60名T1DM对象的数据。这些数据用ONE TOUCH PROFILE测量计收集。下面我们将这些数据称作数据组2。含有这些数据的文件是HAT0.XLS。 Test data set 2 included data from N=60 T1DM subjects from a previous NIH study. These data were collected with the ONE TOUCH PROFILE meter. We refer to these data as
PASS01.DAT、PASS02.DAT和HAT0.XLS中的变量如下:The variables in PASS01.DAT, PASS02.DAT, and HAT0.XLS are as follows:
ID、月、日、时、年——自我解释的ID数字和读数时间。ID, Month, Day, Hour, Year - Self explanatory ID number and readout time.
PLASBG——由One Touch Ultra记录的BG(HAT0.DAT中的N/A,因为使用了One Touch Profile)。PLASBG - BG recorded by One Touch Ultra (N/A in HAT0.DAT because One Touch Profile is used).
RISKLO、RISKHI——表示数据转换结果的控制变量(见下文)。RISKLO, RISKHI - control variables representing the results of data transformations (see below).
BG和BGMM——将BG转换成全血BG,然后以mmol/l表达(见下文)。BG and BGMM - BG was converted to whole blood BG and then expressed in mmol/l (see below).
(每个对象的)合计数据、HbA1c、其估计值以及估计误差被存储在Excel文件PASS01.XLS和PASS02.XLS中。Aggregated data (for each subject), HbA 1c , its estimated value and estimated error are stored in Excel files PASS01.XLS and PASS02.XLS.
PASS01.XLS、PASS002.XLS知HAT1.XLS中的变量如下:The variables in PASS01.XLS, PASS002.XLS and HAT1.XLS are as follows:
ID、 (糖尿病)类型ID, (diabetes) type
HBA1——参考基准HbA1c值HBA1——reference base HbA 1c value
HBA2——第3个月的参考HbA1c(T2DM为2个月)——这需要加以预测;HBA2 - reference HbA 1c at 3 months (2 months for T2DM) - this needs to be predicted;
EST2和ERR2——HbA1c及其误差的估计;EST2 and ERR2 - estimates of HbA 1c and their errors;
控制变量(所有被算法1使用的变量):Control variables (all variables used by Algorithm 1):
BGMM1——以mmol/l为单位的平均BG(见下面的第2部分);BGMM1 - mean BG in mmol/l (see
RLO1、RHI1——低和高BG指数(见下面的第2部分);RLO1, RHI1 - low and high BG indices (see
L06-夜间低BG指数——根据午夜-6:59a.m的读数加以计算(也就是,if(0.1e.HOUR.1e.6));L06 - Nighttime low BG index - calculated from midnight - 6:59a.m readings (ie, if(0.1e.HOUR.1e.6));
NC1=过去60天中SMBG读数的个数;NC1 = number of SMBG readings in the past 60 days;
NDAYS=过去60天中具有SMBG读数的天数。NDAYS = number of days with SMBG readings in the past 60 days.
N06-时间间隔0-6:59;7-12:59中SMBG读数的百分数;N06 - Percentage of SMBG readings in time interval 0-6:59; 7-12:59;
ECLUDE=0,1——如果ECLUDE=1,则建议算法排除该样本。ECLUDE=0,1 - If ECLUDE=1, the suggestion algorithm excludes this sample.
文件PASS01.DAT和PASS1.XLS能够通过对象的ID数字加以匹配。类似地,文件PASS02.DAT和PASS2.XLS,HAT0.XLS和AHAT1.XLS也能够通过对象的ID数字加以匹配。粗数据和全部第二代数据文件被发送给LifeScan有限公司。The files PASS01.DAT and PASS1.XLS can be matched by the ID numbers of the objects. Similarly, the files PASS02.DAT and PASS2.XLS, HAT0.XLS and AHAT1.XLS can also be matched by the ID numbers of the objects. Raw data and all second generation data files were sent to LifeScan Ltd.
算法1的提出Proposition of
公式推导:Formula Derivation:
在本项目的实例No.1中给出了算法1大部分的解释和提出。实例No.1不包括数据收集。相反,我们使用由Amylin药剂师在临床试验中收集的数据组。实例No.1提出了三种可能的公式用于由SMBG数据估计HbA1c:(1)使用平均SMBG、低和高BG指数的公式;(2)使用平均SMBG和先前参考HbA1c读数的公式;和(3)只使用平均SMBG的简单线性公式(见实例No.1)。Most of the explanation and proposal of
还提出了另一个用于评估HbA1c精度的客观标准(在实例No.1中,我们使用最小平方估计、%误差和绝对误差估计每个公式的精度)。这一新的要求被转变成算法1的一个不同的优化标准,即公式的优化不再是产生最小的误差平方和(最小平方估计),而是使估计值固定在HbA1c参考值±1的均匀范围内。Another objective criterion for assessing the precision of HbA 1c was also proposed (in Example No. 1 we estimated the precision of each formula using least squares estimation, % error and absolute error). This new requirement is translated into a different optimization criterion for
为了这样做,我们只用T1DM对象的训练数据分析我们最初的线性模型(实例No.1的公式)相对于该均匀匹配的误差。我们发现,这些误差与对象的高BG指数正相关(r=0.3),并且我们使用这一关系修正我们最初的线性模型。我们发现,最好使用高BG指数作为分组变量,将对象样本分成高BG指数渐增的组,并且在每个组中将修正引入该线性模型。我们的想法是将使用低BG指数的修正引入到每个特殊的组内,而不是象实例No.1中建议的那样引入到所有的样本。这一变化用基于NBSP标准的不同优化方案加以指示。To do so, we analyzed the error of our original linear model (formula from Example No. 1) with respect to this uniform fit using only the training data of T1DM subjects. We found that these errors were positively correlated with a subject's high BG index (r=0.3), and we used this relationship to refine our original linear model. We have found that it is best to use high BG index as a grouping variable, divide the sample of subjects into groups of increasing high BG index, and introduce corrections into the linear model within each group. The idea is to introduce a correction using a low BG index within each particular group, rather than all samples as suggested in Example No. 1. This variation is indicated by different optimization schemes based on NBSP criteria.
因此,根据T1DM对象的训练数据,我们完成了如下的算法1:Therefore, according to the training data of T1DM subjects, we completed
部分1-数据预处理Part 1 - Data Preprocessing
BG=PLASBG/1.12(将血浆BG转换成全血BG,其是通用的)。BG = PLASBG/1.12 (converts plasma BG to whole blood BG, which is common).
BGMM=BG/18(将BG转换成mmol/l)。BGMM = BG/18 (convert BG to mmol/l).
下面各行计算每个SMBG读数的低和高BG指数:The following lines calculate the low and high BG indices for each SMBG reading:
COM SCALE=(in(G))**1.08405-5.381.COM SCALE=(in(G))**1.08405-5.381.
COM RISK1=22.765*SCALE*SCALE.COM RISK1=22.765*SCALE*SCALE.
COM RISKLO=0.COM RISKLO = 0.
IF(BG≤112.5)RISKLO=RISK1.IF(BG≤112.5)RISKLO=RISK1.
COM RISKHI=0.COM RISKHI=0.
IF(BG>112.5)RISKHI=RISK1.IF(BG>112.5)RISKHI=RISK1.
下面各行合计了每个对象的数据:The following rows total the data for each object:
BGMM1=每个对象的平均(BGMM);BGMM1 = mean(BGMM) per subject;
RLO1=每个对象的平均(RISKLO);RLO1 = average (RISKLO) per subject;
RHI1=每个对象的平均(RISKHI);RHI1 = mean (RISKHI) per subject;
L06=只为夜间读数计算的平均(RISKLO),如果没有夜间读数则缺省。L06 = average calculated for night readings only (RISKLO), default if no night readings.
N06,N12,N24——分别是时间间隔0-6:59;7-12:59和18-23:59中SMBG读数的百分数,例如if(0≤HOUR≤6);if(7≤HOUR≤12)和if(18≤HOUR≤24)。N06, N12, N24—the percentage of SMBG readings in time intervals 0-6:59; 7-12:59 and 18-23:59, respectively, for example if(0≤HOUR≤6); if(7≤HOUR≤ 12) and if (18≤HOUR≤24).
NC1=过去60天中SMBG读数的总数;NC1 = total number of SMBG readings in the past 60 days;
NDAYS=过去60天中具有SMBG读数的天数。NDAYS = number of days with SMBG readings in the past 60 days.
部分2-估计程序: Part 2 - Estimation procedure :
该估计程序是基于我们实例No.1的线性模型:The estimation procedure is based on the linear model of our example No.1:
HbA1c=0.41046*BGMM+4.0775.HbA 1c =0.41046*BGMM+4.0775.
分析该公式的误差我们发现,误差取决于高BG指数。因此,我们根据对象的高BG指数对全部对象进行分类,然后在每个组中对线性模型进行修正,如下:Analyzing the error of this formula, we found that the error depends on the high BG index. Therefore, we classify all objects according to their high BG index, and then modify the linear model in each group as follows:
A.每个对象根据他/她的高BG指数指派一个组:A. Each object is assigned a group based on his/her high BG index:
if(RHI1≤5.25 or RHI1≥16) GRP=0.if(RHI1≤5.25 or RHI1≥16) GRP=0.
if(RHI1>5.25 and RHI1<7.0)GRP=1.if(RHI1>5.25 and RHI1<7.0)GRP=1.
if(RHI1≥7.0 or RHI1<8.5) GRP=2.if(RHI1≥7.0 or RHI1<8.5) GRP=2.
if(RHI1≥8.5 or RHI1<16) GRP=3.if(RHI1≥8.5 or RHI1<16) GRP=3.
B.对于每个组,我们进行如下估计:B. For each group, we estimate as follows:
E0=0.55555*BGMM1+2.95.E0=0.55555*BGMM1+2.95.
E1=0.50567*BGMM1+0.074*L06+2.69.E1=0.50567*BGMM1+0.074*L06+2.69.
E2=0.55555*BGMM1-0.074*L06+2.96E2=0.55555*BGMM1-0.074*L06+2.96
E3=0.44000*BGMM1+0.035*L06+3.65.E3=0.44000*BGMM1+0.035*L06+3.65.
EST2=E0EST2=E0
if(GRP=1) EST2=E1.if(GRP=1) EST2=E1.
if(GRP=2) EST2=E2.if(GRP=2) EST2=E2.
if(GRP=3) EST2=E3.if(GRP=3) EST2=E3.
C.对极少数局外值(outlier)进行修正:C. Correct a very small number of outliers:
if(missing(L06)) EST2=E0.if(missing(L06)) EST2=E0.
if(RL01≤0.5 and RHI1≤2.0) EST2=E0-0.25.if(RL01≤0.5 and RHI1≤2.0) EST2=E0-0.25.
if(RL01≤2.5 and RHI1>26) EST2=E0-1.5*RLO1.if(RL01≤2.5 and RHI1>26) EST2=E0-1.5*RLO1.
if((RLO1/RHI1)≤0.25 and L06>1.3) EST2=EST2-0.08.if((RLO1/RHI1)≤0.25 and L06>1.3) EST2=EST2-0.08.
精度标准Accuracy standard
为了评估算法1的精度,我们使用了多个标准判据:To evaluate the accuracy of
1)NGSP精度标准要求全部估计中有至少95%位于HbA1c参考值±1HbA1c单位内。1) The NGSP precision criteria require at least 95% of all estimates to be within ±1 HbA 1c units of the HbA 1c reference value.
2)HbA1c估计值与测量值的平均绝对偏差;2) The mean absolute deviation between the estimated value of HbA 1c and the measured value;
3)HbA1c估计值与测量值的平均百分比偏差。3) The average percentage deviation of estimated and measured HbA 1c values.
重点注意:NGSP精度标准是为检验直接测量HbA1c的设备而设计的。这里,我们应用该标准到由SMBG数据对HbA1c的估计。但是,这种估计的目的不是取代HbA1c实验室测量,而是帮助患者和医生进行糖尿病的日常管理。与实验室测量相对,该估计采用能够以任何方法获得的并且能够在日常获得的数据,而不要求特殊的设备或者去医生诊所。 IMPORTANT NOTE : The NGSP precision standard is designed for testing devices that directly measure HbA 1c . Here, we apply this criterion to estimates of HbA 1c from SMBG data. However, the purpose of this estimate is not to replace HbA 1c laboratory measurements, but to assist patients and physicians in the daily management of diabetes. In contrast to laboratory measurements, this estimation uses data that can be obtained by any means and can be obtained on a daily basis without requiring special equipment or a visit to a doctor's office.
为了证实HbA1c的其他直接测量是与传统的实验室测量是否一致,我们检测了21名IDDM患者的血样并同时用DCA2000和临床化验分析HbA1c。这21个检测结果中,有一个大的误差,为2.5个HbA1c单位。表11给出了该FDA推荐的办公设备的精度结果:To confirm that other direct measurements of HbA 1c are consistent with traditional laboratory measurements, we tested blood samples from 21 IDDM patients and analyzed HbA 1c with both DCA2000 and clinical assays. Among the 21 test results, there was a large error of 2.5 HbA 1c units. Table 11 gives the accuracy results for this FDA-recommended office equipment:
表11:DCA2000在T1DM中的精度Table 11: Accuracy of DCA2000 in T1DM
样本选择标准Sample Selection Criteria
公式推导:Formula Derivation:
HbA1c的估计使用连续60天的SMBG。我们以这连续60天的SMBG数据作为样本。在他/她的SMBG期间,每个人会产生大量的样本。实际上,每次新的测量都会产生一个与先前一个略微不同的新样本。因此,自然会假定测量计具有一些控制点,以保证用于估计HbA1c的SMBG样本数据的质量。HbA 1c was estimated using SMBG for 60 consecutive days. We take the 60 consecutive days of SMBG data as a sample. During his/her SMBG, each person generates a large number of samples. In effect, each new measurement produces a new sample that is slightly different from the previous one. Therefore, it is natural to assume that the meter has some control points to assure the quality of the SMBG sample data used to estimate HbA 1c .
因此,一般化算法公式被优化之后,便被应用于整个训练数据组(T1DM和T2DM对象数据),以调查SMBG样本会产生不准确HbA1c估计的条件。Therefore, the generalized algorithm formulation, after being optimized, was applied to the entire training data set (T1DM and T2DM subject data) to investigate the conditions under which SMBG samples would yield inaccurate HbA 1c estimates.
该调查集中于如下的在SMBG中出现的导致不准确估计的模式:The investigation focuses on the following patterns that occur in SMBG leading to inaccurate estimates:
1)SMBG不频繁——在两个月中需要一定数目的读数以便估计HbA1c,如果该数目没有达到,则估计可能不准确;1) SMBG is infrequent - a certain number of readings are required in two months in order to estimate HbA 1c , if this number is not achieved, the estimate may be inaccurate;
2)当对象主要在餐后进行测试或者服用口服药物而主要关注于高BG时,会出现SMBG偏向于高血糖症的模式;2) A SMBG-biased pattern of hyperglycemia occurs when subjects are primarily concerned with high BG when they are tested primarily after meals or taking oral medications;
3)SMBG的时间偏颇模式,即主要在每天的固定时间进行测试,使得对象的BG波动没有良好的日内分布。3) The time-biased mode of SMBG, that is, the test is mainly performed at a fixed time every day, so that the subject's BG fluctuations do not have a good intraday distribution.
调查完这些模式之后,我们根据最精确的和最少排除的切点选择了最佳的样本选择标准。关于该程序逻辑的详细说明和编码目的的叙述请参考附录A。After investigating these patterns, we selected the best sample selection criteria based on the most precise and least excluded cut points. Please refer to Appendix A for a detailed description of the program logic and a description of the coding purpose.
最终样本选择标准:Final Sample Selection Criteria:
标准1.测试频率:算法要求60天样本包括平均每天至少2.5次测试,即在过去的60天内至少有150个SMBG读数以产生HbA1c估计值(NC1>=150)。
标准2.数据随机化:
2a)口服治疗/餐后检测:(RLO1/RHI1>=0.005)。在一些SMBG样本中,SMBG的分布似乎非常偏向于高血糖症。这主要在T2DM对象中发生,他们似乎只在夜间测量BG。我们假定这些样本不包括低血糖范围内的测试。我们的调查显示,这些样本有大约1/3会对HbA1c产生过高的估计(2/3仍然会产生准确的估计)。据此我们推荐,如果遇到偏颇的样本,则测量计不显示结果,该计算被公式化为LBGI至少为HBGI的1/2%。 2a) Oral treatment/postprandial detection : (RLO1/RHI1>=0.005). In some SMBG samples, the distribution of SMBG appeared to be very skewed towards hyperglycemia. This mainly occurs in T2DM subjects, who seem to measure BG only at night. We assumed that these samples did not include tests in the hypoglycemic range. Our survey showed that approximately 1/3 of these samples would have yielded an overestimation of HbA 1c (2/3 would still have yielded an accurate estimate). Accordingly we recommend that the meter not display results if a biased sample is encountered, the calculation being formulated so that the LBGI is at least 1/2% of the HBGI.
2b)夜间测试:(NO6>=3%)。该标准保证至少有一部分夜间低血糖症能够得到解释。该标准要求全部读数的3%在夜间进行(午夜-7:00am)。换言之,如果2个月内采集的150个读数中有至少5个在夜间,则该样本是可以接受的。注意,患者通常被建议在夜间进行测试,因此该标准能够促进良好的管理。 2b) Night test : (NO6>=3%). This criterion guarantees that at least some nocturnal hypoglycemia can be explained. The standard requires that 3% of all readings be taken at night (midnight - 7:00am). In other words, a sample is acceptable if at least 5 of the 150 readings taken over 2 months are at night. Note that patients are often advised to be tested at night, so this standard can promote good management.
2c)防止高度异常的测试模式:如果超过3/4的读数在每天的任意6小时间隔内进行,则该样本不能够产生估计值。例如,如果样本中有80%的测试是恰在早饭之后进行的,则不进行估计。该标准是LifeScan有限公司所要求的,以防止人们试图“困惑算法”(beat thealgorithm),借此允许我们确保有效,特别是对临床医生。 2c) Prevents highly outlier test patterns : If more than 3/4 of the readings are taken in any 6-hour interval per day, the sample cannot produce an estimate. For example, if 80% of the tests in the sample were taken just after breakfast, no estimate is made. This standard is required by LifeScan Ltd to prevent people from trying to "beat the algorithm", thereby allowing us to ensure effectiveness, especially for clinicians.
顺序采用选择标准在训练数据组中的精度Sequentially adopt the accuracy of the selection criteria in the training data set
下面的表说明了所选样本选择标准对在训练数据组中的精度和排除数目的影响。注意提出作为本研究(最终算法)一部分的算法1最终版本的精度和在实例No.1中提出的并包含在实例No.1中的最简线性函数的精度(见第一线性模型)。The table below illustrates the effect of selected sample selection criteria on the accuracy and number of exclusions in the training data set. Note the accuracy of the final version of
我们给出了每个模型在没有任何样本选择标准时和顺序应用了样本选择标准1——检测频率,#读数NR≥150,和标准2——数据随机化,时的精度,如上所述。We give the accuracy of each model without any sample selection criteria and when
如在所有表中见到的,算法1的精度随着顺序采用样本选择标准而提高,并且在应用了所有标准之后达到了NGSP要求的95%。后一结果在表中是突出的。As seen in all tables, the accuracy of
表12A:训练数据组中的最终样本选择标准——所有对象:Table 12A: Final Sample Selection Criteria in Training Dataset - All Subjects:
表12B:训练数据组中的最终样本选择标准——T1DM:在该样本中优化了算法1的系数,其解释了即使没有样本选择时的高精度。
表12C:训练数据组中的最终样本选择标准——T2DM:样本选择标准2(数据随机化)主要用这一样本提出,其解释了当应用该标准时增加的5%精度。
训练数据中的样本排除频率: Sample exclusion frequency in training data :
在每次新读数时测量计都有机会估计HbA1c。如果样本没有满足选择标准,则测量计不会显示HbA1c的估计,并且会:The meter has the opportunity to estimate HbA 1c at each new reading. If the sample does not meet the selection criteria, the meter will not display an estimate of HbA 1c and will:
(a)等待直到收集到合适的样本,或者(a) wait until a suitable sample is collected, or
(b)如果没有收集到合适的样本,例如某人具有永久的偏颇测量方式,则测量计会发出修正SMBG模式的提示。(b) If no suitable sample is collected, eg someone has a permanently biased measurement pattern, the meter will prompt to correct the SMBG pattern.
我们的调查显示,大部分的对象(>95%)在60天内会获得至少10个HbA1c估计(只要他们的测量频率足够),而只有2%的对象会由于偏颇的测量方式而没有获得估计。需要提示这2%的对象修正他们的测量方式。该调查的最终结果在下文给出:Our survey showed that most subjects (>95%) will get at least 10 HbA 1c estimates within 60 days (provided they are measured frequently enough), while only 2% of subjects will get no estimates due to biased measurement methods . The 2% of subjects need to be prompted to revise their measurements. The final results of the survey are given below:
我们计算了(在60天中)有多少天测量计会由于样本不满足选择标准而不能向用户显示HbA1c结果:We calculated (out of 60 days) how many days the meter would not be able to display the HbA 1c result to the user because the sample did not meet the selection criteria:
1)对于全部对象的72.5%,测量计能够每天报告HbA1c;1) The meter was able to report HbA 1c daily for 72.5% of all subjects;
2)对于全部对象的另外7.5%,测量计能够报告(60天中)45-59天的HbA1c;2) For another 7.5% of all subjects, the meter was able to report (out of 60 days) HbA 1c for 45-59 days;
3)对于全部对象的另外10%,测量计能够报告(60天中)12-44天的HbA1c;3) For another 10% of all subjects, the meter is capable of reporting HbA 1c for 12-44 days (out of 60 days);
4)对于9名对象(5.9%),测量计不能够报告HbA1c,除非他们改变SMBG方式。4) For 9 subjects (5.9%), the meter was unable to report HbA 1c unless they changed the SMBG modality.
重点注意:这些对象中的大多数都不会获得估计,因为他们没有满足测试频率标准1,即他们的样本总是小于150个读数。因此,全部对象中至少有94%能够大约每5天获得至少一个HbA1c,而无需改变他们的测量方式(这包括T1DM和T2DM)。 Important note : Most of these subjects will not get an estimate because they do not meet the
如果我们要求在60天内至少有150个读数,则只有3名对象不能获得HbA1c估计:If we ask for at least 150 readings within 60 days, there are only 3 subjects who cannot obtain HbA 1c estimates:
1)95.6%在60天内得到至少10个HbA1c估计;1) 95.6% get at least 10 HbA 1c estimates within 60 days;
2)2.2%不会获得任何估计。2) 2.2% won't get any estimates.
因此,在60天内有大约98%的平均每天测量2.5次的对象会得到HbA1c估计,有>95%的将每周至少获得1个估计。我们得出结论,样本选择标准2——数据随机化在一段时期内对显示HbA1c估计的影响很微小。只有大约2%的对象需要被提示改进他们的SMBG方式。Thus, approximately 98% of subjects with an average of 2.5 measurements per day will have an HbA 1c estimate over 60 days and >95% will have at least 1 estimate per week. We concluded that
应当注意,样本选择标准能够用于提高任何估计HbA1c的公式的精度。该选择标准独立于任何特殊的算法/公式,并且在估计开始之前应用。例如,当被使用时,样本选择标准将提高最新提出的作为本研究一部分的算法1的精度,以及在实例No.1中提出的我们最初的线性模型的精度。It should be noted that sample selection criteria can be used to improve the precision of any formula for estimating HbA 1c . This selection criterion is independent of any special algorithm/formula and is applied before the estimation begins. For example, when used, sample selection criteria will improve the accuracy of the newly proposed
此外,检查其他一些样本选择标准的效果显示,我们能够进一步提高精度,这是被期望的。例如,当将其中一个原始测试频率标准应用于该数据时,能够证实具有更多的效力。该标准在附录E中有进一步的说明。Furthermore, examining the effect of some other sample selection criteria shows that we are able to further improve the accuracy, which is expected. For example, more power can be demonstrated when one of the original test frequency criteria is applied to the data. The standard is further described in Appendix E.
算法1的前瞻性验证:Prospective verification of Algorithm 1:
在测试数据组1中的精度: Accuracy on test data set 1 :
随后将算法,包括最终的样本选择标准,应用于测试数据组1(T1DM1和T2DM对象最后一次HbA1c之前2个月的SMBG)以产生HbA1c估计。然后将这些估计值与HbA1c参考值进行比较,从而前瞻性验证算法1。表13给出了该验证的概要。关于每个样本选择标准对算法的影响的更详细说明请见附录C。The algorithm, including final sample selection criteria, was then applied to test data set 1 (
表13:算法前瞻性应用的精度:
测试数据组2中的精度: Accuracy in test data set 2 :
另一个独立的NIH数据组(N=60名患有T1DM的对象)证实了结果具有类似的精度,即有95.5%处于实验室参考值±1HbA1c单位之内(表14):Another independent NIH data set (N = 60 subjects with T1DM) confirmed the results with similar precision, namely 95.5% within ± 1 HbA 1c unit of the laboratory reference value (Table 14):
表14:算法1在独立NIH数据组中的精度:
算法1的精度与FDA推荐的办公设备精度的比较: Comparison of the accuracy of
如下面的表15中显示的,算法1的精度与医生诊所中使用的HbA1c化验精度相当。如在精度标准部分中说明的,DCA2000数据用于证实HbA1c的其他直接测量结果是否与实验室测量结果一致。我们同时用DCA2000和临床化验分析了21名T1DM患者的HbA1c血液样本。在这21次测试中有较大的误差,为2.5HbA1c单位:As shown in Table 15 below, the accuracy of
表15:DCA2000在T1DM中的精度与算法1比较:
测试数据组中样本排除的频率: Frequency of sample exclusion in the test data set :
正如我们在提出算法1的部分中讨论的,在每次新读数的时候,测量计都有机会估计HbA1c。如果样本不满足选择标准,那么测量计就不会显示HbA1c。As we discussed in the
我们使用测试数据组1和2前瞻性估计样本排除的频率。为此,我们的计算是根据测量计能够有多少天(60天中)向人显示HbA1c,也就是人们能够有多少天具有满足样本选择标准的样本。表16A和16B给出了测试数据组1和2中这些结果的摘要。我们包括了所有对象的数据,这些数据分成平均测量1.5次/天的对象(在60天中有90个SMBG读数)和2.5次/天的对象的数据:We prospectively estimated the frequency of sample exclusions using
表16A:测试数据组1中样本排除的频率Table 16A: Frequency of Sample Exclusions in
表16B:测试数据组2中样本排除的频率Table 16B: Frequency of Sample Exclusions in
结论: Conclusion :
表13-16证实,测量计能够平均一周产生一个满足95%NGSP精度标准的HbA1c精确估计,对于平均每天测量2.5次的人则为>96%。Tables 13-16 demonstrate that the meters were able to produce an accurate estimate of HbA 1c meeting the 95% NGSP accuracy criteria for a week on average, and >96% for a person averaging 2.5 measurements per day.
附录-样本选择标准的软件逻辑APPENDIX - SOFTWARE LOGIC FOR SAMPLE SELECTION CRITERIA
向对象发送样本选择标准——建议通过算法排除一些SMBG样本,或者消息。该样本选择标准被编程如下:Send Sample Selection Criteria to Objects - It is suggested to algorithmically exclude some SMBG samples, or messages. The sample selection criteria were programmed as follows:
标准1.测试频率:本算法要求60天样本要包括平均每天至少测量2.5次,即在过去的60天中要有至少150个SMBG读数以产生HbA1c估计:
EXCLUDE=0EXCLUDE=0
if(NC1>=150) EXCLUDE=1if(NC1>=150) EXCLUDE=1
标准2.数据随机化:
2a) 口服治疗/餐后测试:在一些SMBG样本中,SMBG的分布似乎非常偏向于高血糖症。这主要在T2MD对象中发生,他们似乎只在高BG时进行测量。我们假定,这些样本不包括在低血糖范围测量的结果。我们的调查显示,这些对象中大约有1/3会过高地估计HbA1c(2/3的对象仍然产生精确的估计)。据此,我们建议如果样本出现了偏颇,则测量计不显示结果,该计算被公式化为LBGI至少为HBGI的1/2%。2a) Oral treatment/fed test : In some SMBG samples, the distribution of SMBG appeared to be very skewed towards hyperglycemia. This mainly occurs in T2MD subjects, who only seem to measure at high BG. We assume that these samples do not include results measured in the hypoglycemic range. Our survey showed that approximately 1/3 of these subjects overestimated HbA 1c (2/3 subjects still produced accurate estimates). Accordingly, we recommend that the meter not display results if the sample is biased, the calculation being formulated so that the LBGI is at least 1/2% of the HBGI.
if(RLO1-RHI1<0.005)EXCLUDE=1。if(RLO1−RHI1<0.005) EXCLUDE=1.
2b) 夜间测试:(NO6>=3%)。该标准保证至少能够解释一部分的夜间低血糖症。该标准要求全部读数的3%在夜间进行(午夜-7:00am)。换言之,如果在两个月中获得的150个读数中至少有5个在夜间,则该样本可以接受。注意,通常建议患者在夜间进行测试,所以该标准能够促进良好的管理。2b) Night test : (NO6>=3%). This criterion is guaranteed to account for at least some of the nocturnal hypoglycemia. The standard requires that 3% of all readings be taken at night (midnight - 7:00am). In other words, a sample is acceptable if at least 5 of the 150 readings taken over two months are at night. Note that it is often recommended that patients be tested at night, so this criterion can facilitate good management.
if(NO6≤3.0) EXCLUDE=1。if(NO6≤3.0) EXCLUDE=1.
2c) 确保防止高度异常的测试模式:如果有超过3/4的读数一天中的任意6小时间隔内进行,则样本不会产生估计值。例如,如果样本中有80%的测试恰好在早饭之后进行,则不会产生估计值。该标准是LifeScan公司要求的以便保证人们试图“困惑算法”,借此允许我们确保有效性,特别是对于医师。在我们的数据中,没有样本如此高度地异常以致于触发该标准(详细信息见附录B——标准2c)。依靠软件执行,需要从SMBG数据中计算如下的频率:2c) Ensure protection against highly anomalous test patterns : If more than 3/4 of the readings are taken in any 6-hour interval of the day, the sample will not yield an estimate. For example, if 80% of the tests in the sample were taken just after breakfast, no estimate would be produced. This standard is required by LifeScan in order to ensure that people try to "confuse the algorithm", thereby allowing us to ensure validity, especially for physicians. In our data, no samples were so highly abnormal as to trigger this criterion (see Appendix B - Criterion 2c for details). Relying on software implementation, the following frequencies need to be calculated from the SMBG data:
M12-6:00am-中午的SMBG读数%(早餐)M12 - 6:00am - % of SMBG reading at noon (breakfast)
M18-中午-6:00pm的SMBG读数%(午餐)M18 - % of SMBG reading at noon - 6:00pm (lunch)
M24-6:00pm-12:00的SMBG读数%(晚餐)M24-6:00pm-12:00 SMBG reading % (Dinner)
M06-12:00-6:00am的SMBG读数%(夜间)M06 - % of SMBG readings from 12:00-6:00am (nighttime)
M15-9:00am-3:00pm的SMBG读数%M15 - SMBG reading % from 9:00am-3:00pm
M21-3:00pm-9:00pm的SMBG读数%M21 - SMBG reading % from 3:00pm-9:00pm
M03-9:00pm-3:00am的SMBG读数%M03 - % of SMBG readings from 9:00pm-3:00am
M09-3:00am-9:00am的SMBG读数%M09 - SMBG reading % from 3:00am-9:00am
然后对于上述的任何组合(i,j):Then for any combination (i, j) of the above:
if(Mij>75.0)EXCLUDE=1。if (Mij > 75.0) EXCLUDE = 1.
附录B-样本选择标准2CAppendix B - Sample Selection Criteria 2C
本标准是LifeScan公司要求的,以确保防止高度异常的测试模式。该标准的目的是防止人们“困惑算法”。This standard is required by LifeScan Corporation to ensure protection against highly abnormal test patterns. The purpose of the standard is to prevent people from "confusing algorithms".
基本上,本标准规定:如果你的读数的3/4(或者其他期望的数目)在一天的任意6小时间隔或者其他期望的间隔内进行,则你将不能够得到估计值。Basically, this standard states: If 3/4 (or other desired number) of your readings are taken at any 6-hour interval of the day or other desired interval, then you will not be able to get an estimate.
因此,例如,如果有超过3/4测试在晚饭后进行,则不会得到估计值。这将更加支持我们的一般声明:本计算中不包括测试不随机的人。我认为,对此的特殊计算和编码可能看上去复杂,但关键的是我们可以只输入“你在一天中必须随机进行测试”或者其他类似的话作为面板声明,就可以覆盖我们全部的排除标准(排除测试频率)。如果我们需要,则我们能够以精练的语句(fine print)简单地输入更加精确的定义“一天中任何6小时间隔内的读数都不可以超过全部读数的75%”。这保证,遵循我们的标准,可以提高算法的临床接受度。So, for example, if more than 3/4 of the tests were done after dinner, you won't get an estimate. This would more support our general statement: people whose tests were not randomized were not included in this calculation. I think the special calculations and coding for this may seem complicated, but the point is that we can just enter "you have to take the test randomly throughout the day" or something like that as a panel statement and override our entire exclusion criteria ( Exclude test frequency). If we wanted, we could simply enter a more precise definition in fine print that "the readings in any 6-hour interval during the day may not exceed 75% of the total readings". This guarantees that following our criteria improves the clinical acceptance of the algorithm.
更详细内容:More details:
4个6小时间隔定义如下:The four 6-hour intervals are defined as follows:
6:00am-中午(早饭)6:00am-noon (breakfast)
中午-6:00pm(午饭)Noon-6:00pm(Lunch)
6:00pm-12:00(晚饭)6:00pm-12:00(dinner)
12:00-6:00am(夜间)12:00-6:00am (Night)
该标准能够以不同的时间间隔运行两次,从而防止人们集中在6小时间隔的交界点附近进行测试但仍然不正确地满足该第一标准。例如,如果人们在11:50pm有40%,在12:10pm有40%,则依然是集中的测试,虽然满足间隔的第一关(pass),但不满足间隔的第二关。This criterion can be run twice at different time intervals, preventing people from concentrating on testing near the junction of the 6 hour intervals and still incorrectly meeting this first criterion. For example, if a person has 40% at 11:50pm and 40% at 12:10pm, it is still a focused test, satisfying the first pass of the interval but not the second pass of the interval.
第二组间隔:Second set of intervals:
9:00am-3:00pm9:00am-3:00pm
3:00pm-9:00pm3:00pm-9:00pm
9:00pm-3:00am9:00pm-3:00am
3:00am-9:00pm3:00am-9:00pm
注意选择地,从编码的观点来看,人们可以如下地运行而得到相同的结果:Note that alternatively, from a coding point of view, one can get the same result by running:
任意18小时周期的读数不得少于总读数的25%。The readings for any 18-hour period shall not be less than 25% of the total readings.
你必须在相互重叠3个小时的18小时周期内运行它:You have to run it in 18-hour periods that overlap each other by 3 hours:
9:00am-3:00am9:00am-3:00am
12:00中午-6:00am12:00 noon-6:00am
3:00pm-9:00am3:00pm-9:00am
6:00pm-12:00中午6:00pm-12:00 noon
9:00pm-3:00pm9:00pm-3:00pm
12:00午夜-6:00pm12:00 midnight-6:00pm
3:00am-9:00pm3:00am-9:00pm
6:00am-12:00午夜6:00am-12:00midnight
附录C-测试数据组1中样本选择标准对算法精度的增加效应Appendix C - Incremental Effect of Sample Selection Criteria in
如在算法提出部分中说明的,如下的表格涉及了被提出作为本研究一部分的算法1(最终算法)和在实例No.1中提出的并且包含在实例No.1中的最简线性函数的精度(见最初线性模型)。该表格给出了算法1在测试数据组1中无样本排除并依次应用两个样本选择标准的精度:As explained in the Algorithm Proposal section, the following table relates to Algorithm 1 (Final Algorithm) proposed as part of this study and the simplest linear function proposed and included in Example No. 1 Accuracy (see Initial Linear Model). This table gives the accuracy of
标准1——测试频率,#读数NR≥150,和
标准2——数据随机化,如在样本选择标准中说明的:Criterion 2 - Randomization of data, as stated in the sample selection criteria:
表17A:算法1的精度——全部对象:
表17B:算法1在T1DM中的精度:
表17C:算法1在T2DM中的精度:
附录D-可选择测试频率标准Appendix D - Optional Test Frequency Criteria
高级的测试频率标准有利于更显著地提高算法1的精度。这是因为,采用测试频率标准1并不仅是基于数据分析,而且还基于其他的考虑。如果发现要求2个月内有150个读数的标准1太严格,则可以采用可替换的解决方法。就是原始的测试频率标准,其要求有35天(60天中)的SMBG读数频率为1.8个读数/天,也就是60天中有35天的总读数为63个。表18用该原始的宽松测试频率标准加上标准2(数据随机化)证实这一点,算法1的精度超过95%:Advanced test frequency criteria are beneficial to improve the accuracy of
表18:使用选择测试频率标准(35天的读数为1.8个读数/天) 和数据随机化标准的算法1的精度:
重点注意:除此之外,该可选择标准可以筛选掉具有大量缺失数据的样本,例如,如果SMBG间断达4周之后再恢复,则不会显示HbA1c估计。这种模式的一个明显实例出现在测试数据组2中——他/她的HbA1c估计值具有最大误差的对象只在60天的30天中收集了159个读数。因此,通过在少数几天内快速收集读数,对象仍可以满足收集150个读数的要求,但会导致不准确的HbA1c估计。 IMPORTANT NOTE : Among other things, this optional criterion can filter out samples with a large number of missing data, e.g. HbA 1c estimates will not be shown if SMBG is resumed after a 4-week break. A clear instance of this pattern occurs in test data set 2 - the subject with the largest error in his/her HbA 1c estimate had only 159 readings collected in 30 out of 60 days. Therefore, by rapidly collecting readings over a few days, subjects can still meet the requirement of collecting 150 readings, but result in inaccurate HbA 1c estimates.
实例No.2的典型定义(但对本文没有限制)Typical definition for instance No.2 (but not limiting to this article)
1)严重低血糖症(SH)定义为导致无法自我治疗的昏迷、疾病发作或者无意识的低血糖(BG);1) Severe hypoglycemia (SH) is defined as coma, disease attack or unconscious hypoglycemia (BG) that cannot be self-treated;
2)中度低血糖症(MH)定为扰乱对象的活动但不妨碍自我治疗的严重神经低血糖症;2) Moderate hypoglycemia (MH) is defined as severe neurological hypoglycemia that disturbs the subject's activities but does not hinder self-treatment;
3)生化严重低血糖症(BSH)定义为血浆BG读数<=39mg/dl;3) Biochemical severe hypoglycemia (BSH) is defined as plasma BG reading <= 39mg/dl;
4)生化中度低血糖症(BMH)定义为血浆BG读数为39-55mg/dl;4) Biochemical moderate hypoglycemia (BMH) is defined as a plasma BG reading of 39-55 mg/dl;
5)上面全部病症均称作 显著低血糖症。5) All of the above symptoms are referred to as significant hypoglycemia .
附加目的additional purpose
本实例的数据用于前瞻性验证如下的算法:The data of this example is used to prospectively verify the following algorithms:
算法2——使用某一对象30-45天SMBG数据的分类算法,将对象分成发生将来显著低血糖症的一定危险范围。该分类是临时的,例如,当对象的SMBG方式改变时,分类也改变。
算法3——数据跟踪/判定算法,其使用一定序列的SMBG数据判断是否为即将发生的(24小时)显著低血糖症建立标志。现在我们详细说明算法1&2及其测试结果。Algorithm 3 - Data Tracking/Decision Algorithm which uses a sequence of SMBG data to determine whether to establish a flag for imminent (24 hour) significant hypoglycemia. Now we detail
对象object
我们获得了100名患有1型糖尿病(T1DM)的对象和100名患有2形糖尿病(T2DM)的对象的许可。170名对象,其中90名患有T1DM,89名患有T2DM,完成了SMBG数据收集的主要部分。We obtained permission for 100 subjects with
程序program
全部对象都分配了一个IRB推荐的许可表格并参加了指导会,在会上向他们介绍了ONE TOUCH ULTRA测量计并完成了屏幕问卷调查。指导会之后,全部对象立即参观了UVA实验室并采血测量了HbA1c基准值。T1DM对象在随后的6个月里于第3和6月进行实验室HbA1c化验;T2DM对象在随后的4个月里于第2和4月进行实验室HbA1c化验。自我监测(SMBG)数据被规则地从测量计下载下来并保存在数据库中。通过定制设计的自动e-mail/电话跟踪系统平行地记录显著低血糖症和高血糖症事件,其每2周与全部参加者进行联系。表19给出了SMBG和严重低血糖症/中度低血糖症[SH/MH]数据收集的摘要。All subjects were assigned an IRB recommended clearance form and attended an orientation session where they were introduced to the ONE TOUCH ULTRA meter and completed an on-screen questionnaire. After the guidance meeting, all subjects visited the UVA laboratory immediately and took blood to measure the HbA 1c baseline value. Subjects with T1DM had laboratory HbA 1c assays at
表19:数据收集摘要Table 19: Summary of data collection
算法2和3的公式没有显著改变。这些公式与2002年3月实例No.1的报告中给出的公式特别相似。只有两个变化:(a)修正了SH/MH(实例No.1)危险范围列表中的类型和(b)算法3中有一行改变。后者的原因在下面进行解释。The formulations of
因为算法1和2保持未变,所以我们能够将整个实例No.2的数据收集看作这些算法的前瞻性测试。Because
算法2的公式Formula of
算法2的执行如下:
1)根据一个月的SMBG数据,根据每个对象的低血糖指数如下地将他/她分类成15个危险范围(RCAT)中的一个:1) Based on one month of SMBG data, classify each subject into one of 15 risk areas (RCAT) according to his/her hypoglycemic index as follows:
if(LBGI≤0.25),RCAT=0if(LBGI≤0.25), RCAT=0
if(0.25<LBGI≤0.5),RCAT=1if(0.25<LBGI≤0.5), RCAT=1
if(0.50<LBGI≤0.75),RCAT=2if(0.50<LBGI≤0.75), RCAT=2
if(0.75<LBGI≤1.00),RCAT=3if(0.75<LBGI≤1.00), RCAT=3
if(1.00<LBGI≤1.25),RCAT=4if(1.00<LBGI≤1.25), RCAT=4
if(1.25<LBGI≤1.50),RCAT=5if(1.25<LBGI≤1.50), RCAT=5
if(1.50<LBGI≤1.75),RCAT=6if(1.50<LBGI≤1.75), RCAT=6
if(1.75<LBGI≤2.00),RCAT=7if(1.75<LBGI≤2.00), RCAT=7
if(2.00<LBGI≤2.50),RCAT=8if(2.00<LBGI≤2.50), RCAT=8
if(2.50<LBGI≤3.00),RCAT=9if(2.50<LBGI≤3.00), RCAT=9
if(3.00<LBGI≤3.50),RCAT=10if(3.00<LBGI≤3.50), RCAT=10
if(3.50<LBGI≤4.25),RCAT=11if(3.50<LBGI≤4.25), RCAT=11
if(4.25<LBGI≤5.00),RCAT=12if(4.25<LBGI≤5.00), RCAT=12
if(5.00<LBGI≤6.50),RCAT=13if(5.00<LBGI≤6.50), RCAT=13
if(LBGI>6.50),RCAT=14if(LBGI>6.50), RCAT=14
2)通过双参数Weibull概率分布并利用如下给出的分布函数计算将来显著低血糖症的理论概率:2) Calculate the theoretical probability of significant hypoglycemia in the future by using the two-parameter Weibull probability distribution and using the distribution function given below:
F(x)=1-exp(-a,xb)对于任何x>0;否则0。该分布的参数取决于期望的预测持续时间,并且在实例No.1的报告中有说明。如果用测量计执行,该步骤将提供显著低血糖症危险的连续型估计,例如“在下个月中50%”。F(x)=1-exp(-a, xb ) for any x>0; otherwise 0. The parameters of this distribution depend on the desired forecast duration and are reported in Example No. 1. If performed with a meter, this step will provide a continuous type estimate of the risk of significant hypoglycemia, eg "50% in the next month".
3)将每个对象分类到将来显著低血糖症的最小、低、中或者高危险组中:这些范围的定义如下:最小危险(LBGI≤1.25);低危险(1.25<LBGI≤2.5);中度危险(2.5<LBGI≤5)和高危险(LBGI>5)。如果用测量计执行,则该步骤将提供显著低血糖症危险的离散型估计,例如“在下个月为高危险”。3) Classify each subject into minimal, low, medium, or high risk groups for future significant hypoglycemia: these ranges are defined as follows: minimal risk (LBGI≤1.25); low risk (1.25<LBGI≤2.5); medium Moderate risk (2.5<LBGI≤5) and high risk (LBGI>5). If performed on a meter, this step would provide a discrete estimate of the risk of significant hypoglycemia, eg "high risk in the next month".
算法3的公式Formula of
首先,为了避免计算在算法3实例No.1报告说明书中给出的基准危险值,我们修改了编码中的一行。现在,算法3转而使用算法2的结果。我们引入该变化用于介绍2002年10月28日两个对象的样本结果。此时,显然用简单的Excel电子数据表例证算法3的行动是方便的,并且如果避免了基准值的计算则是可能的。该步骤没有改变算法3的精度,所以保留作为方便算法3编程的永久改变。2002年10月28日之后不再给算法3引入改变。这里,我们给出了与实例No.1报告中相同的算法3公式,被改变的行已经标记出来。First, we modified one line in the code to avoid calculating the baseline risk value given in the specification for the report of
1)通过如下的编码为每个BG读数计算低BG危险值(RLO)(这里BG以mg/dl进行测量,如果单位是mmol/l则系数会不同):1) Calculate the risk of low BG (RLO) for each BG reading by the following code (here BG is measured in mg/dl, the coefficient will be different if the unit is mmol/l):
scale=(In(bg))**1.08405-5.381scale=(In(bg))**1.08405-5.381
risk=22.765*scale*scalerisk=22.765*scale*scale
if(bg_1≤112.5)thenif(bg_1≤112.5)then
RLO=riskRLO = risk
elseelse
RLO=0RLO=0
endifendif
2)对于每个SMBG读数,我们计算了运行值LBGI(n),和另一个统计量SBGI(n),其是低BG危险值的标准差。这两个参数的计算是用特定的标记(n)从每个SMBG读数逆推算的,也就是包括该读数及该读数之前的(n-1)个读数。2) For each SMBG reading, we calculated the running value LBGI(n), and another statistic SBGI(n), which is the standard deviation of the low BG risk values. The calculation of these two parameters was back-extrapolated from each SMBG reading with a specific marker (n), ie including that reading and (n-1) readings preceding it.
3)LBGI(n)和SBGI(n)的计算使用临时平均程序(provisionalmeans procedure),其使根据如下的递归编码:3) The calculation of LBGI(n) and SBGI(n) uses a provisional means procedure (provisional means procedure), which makes it according to the following recursive coding:
初值在n(或者精确地在最大值(1,n-k),以便解释序数小于k的测量计读数):Initially at n (or precisely at max(1,n-k) to account for meter readings with ordinals less than k):
LBGI(n)=rlo(n)LBGI(n)=rlo(n)
Rlo2(n)=0Rlo2(n)=0
倒着计数n与1之间的任何连续迭代j值:Count backwards for any successive iteration j values between n and 1:
LBGI(j)=((j-1)/j)*BLGI(j-1)+(1/j)*RLO(j)LBGI(j)=((j-1)/j)*BLGI(j-1)+(1/j)*RLO(j)
rlo2(j)=((j-1)/j)*rlo2(j-1)+(1/j)*(RLO(j)-LBGI(j))**2rlo2(j)=((j-1)/j)*rlo2(j-1)+(1/j)*(RLO(j)-LBGI(j))**2
完成该循环之后,我们获得了LBGI(n)的数值,接着计算After completing the loop, we obtain the value of LBGI(n), and then calculate
SBGI(n)=sqrt(rlo2(n))SBGI(n)=sqrt(rlo2(n))
由该计算,我们保存了两组数据:n=150和n=50(例如前150个和前50个观察值)。From this calculation, we save two sets of data: n=150 and n=50 (eg first 150 and first 50 observations).
4) 判定规则:在每次SMBG读数,程序都判定是否建立即将发生SH的警告标志。如果标志已经建立,则程序判定是否降低该标志。这些判定取决于三个阈值参数α,β,γ,运行如下:4) Judgment rule : at each SMBG reading, the program judges whether to establish a warning sign of impending SH. If the flag is already set, the program decides whether to lower the flag. These decisions depend on three threshold parameters α, β, γ, which operate as follows:
对于低中危险的对象(LM组):For subjects at low and intermediate risk (LM group):
FLAG=0.FLAG=0.
if(LBGI(150)≥2.5 and LBGI(50)≥(1.5*LGI(150)and SBGI(50)≥SBGI(150)) FLAG=1.if(LBGI(150)≥2.5 and LBGI(50)≥(1.5*LGI(150)and SBGI(50)≥SBGI(150)) FLAG=1.
if(RLO≥(LBGI(150)+1.5*SBGI(150)) FLAG=1.if(RLO≥(LBGI(150)+1.5*SBGI(150)) FLAG=1.
换言之,在每次SMBG读数,如果满足两个条件中的一条则建立标志:In other words, at each SMBG reading, a flag is established if one of two conditions is met:
1)根据算法2由前150次试验进行的分类对象处于中度高危险SH,且在前50次试验中LBGI和LBGI的SD增加;1) Subjects classified according to
2)或者,通过第二个不等式确定出低BG指数突增。2) Alternatively, a low BG index burst is determined by the second inequality.
这些叙述的启发性解释在实例No.1的报告中给出。如上面描述的,第一个“if”的陈述已经改变了其最初的形式,以便避免使用基准LBGI,从而使用算法2的输出。A heuristic explanation of these narratives is given in the report of Example No. 1. As described above, the first "if" statement has been changed from its original form in order to avoid using the baseline LBGI and thus use the output of
如实例No.1的报告中说明的,一旦建立起标志,则它将保持24小时。为了评估算法3的精度,我们使用以前提出的技术——计算两个测量值:As stated in the report of Example No. 1, once the flag is established, it remains for 24 hours. To evaluate the accuracy of
1)24小时内即将发生SH/MH事件的预测%,和1) Predicted % of upcoming SH/MH events within 24 hours, and
2)“标志升高”与“标志降低”持续周期的比值Rud(烦扰指数)。2) Ratio R ud (disturbance index) of the continuous period of "mark up" and "mark down".
SH事件的预测%需要比较高,而比值Rud需要比较低。这是因为通过增加预测SH事件的百分率,我们会不可避免地增加“建立标志”的数目,其反过来增加潜在“错误报警”的数目。因为“错误报警”没有被清晰地定义(见实例No.1的报告),所以我们使用Rud作为算法3有效性的指标。The predicted % of SH events needs to be relatively high, while the ratio Rud needs to be relatively low. This is because by increasing the percentage of predicted SH events we would inevitably increase the number of "establishment flags", which in turn increases the number of potential "false alarms". Because "false alarm" is not clearly defined (see report in Example No. 1), we use Rud as an indicator of the effectiveness of
在实例No.1的报告中给出的我们先前最佳的结果是,预测出50%的24小时内SH/MH事件,Rud=1∶10,即在高危险报警一天之后,紧接着10天没有报警。这里我们将保持相同的标志升高/标志降低比值,并分别为T1DM和T2DM对象计算24小时内SH和MH事件的%预测。对于该预测,我们不使用BSH和BMH事件,因为这是通过测量计记录的,因此是预测函数的一部分。Our previous best result given in the report of Example No. 1 was to predict 50% of SH/MH events within 24 hours, Rud = 1:10, i.e. after one day of high-risk alarm, followed by 10 Day did not call the police. Here we will keep the same flag-increased/flag-decreased ratio and calculate % predictions of SH and MH events within 24 hours for T1DM and T2DM subjects, respectively. For this prediction we do not use BSH and BMH events as this is recorded by the gauge and thus part of the prediction function.
估计1-3个月内显著低血糖症的危险:算法2的精度Estimating the risk of significant hypoglycemia in 1-3 months: the precision of
我们如下地估计了算法2的预测能力:We estimated the predictive power of
1)首先,我们由一个月的SMBG数据计算了LBGI,并且如上所述地将每个对象分类到显著低血糖症的最小、低、中和高危险组。1) First, we calculated LBGI from one month of SMBG data and classified each subject into minimal, low, medium and high risk groups for significant hypoglycemia as described above.
2)然后,在随后的1-3个月中,我们计数每个对象前瞻性记录的SH、BSH、MH和BMH事件的数目。2) We then counted the number of prospectively recorded SH, BSH, MH and BMH events per subject during the subsequent 1-3 months.
下面的图16-19分别为T1DM和T2DM给出了在一个月的SMBG之后每个对象在未来1个月或者3个月内观察到的SH、BSH、MH和BMH事件的数目。同时还包括了统计比较。Figures 16-19 below give the number of SH, BSH, MH and BMH events observed for each subject in the next 1 month or 3 months after one month of SMBG for T1DM and T2DM, respectively. Statistical comparisons are also included.
此外,直接线性回归使用LBGI、在屏幕问卷中根据过去一年中SH事件数目报告的SH历史,和基准HbA1c,显著地预测了(R2=0.62,f=48,p<0.0001)在之后3个月内即将发生显著低血糖症事件的总数(SH+MH+BSH+BMH)。预测变量按照其重要性的顺序为:1)LBGI(t=8.2,p<0.0001),可单独解释将来显著低血糖症55%的变异(例如R2=0.55);2)SH历史(t=3.6,p<0.0005),可解释另外5%的变异,和HbA1c(t=2.2,p<0.03),可解释另外2%的变异。这证实了先前的结果,即LBGI是将来低血糖症最重要的预测指标,而HbA1c对该预测的贡献是中等的。Furthermore, direct linear regression using LBGI, SH history reported in a screen questionnaire based on the number of SH events in the past year, and baseline HbA 1c , significantly predicted (R 2 =0.62, f=48, p<0.0001) after Total number of imminent significant hypoglycemic events (SH+MH+BSH+BMH) within 3 months. The predictors, in order of their importance, were: 1) LBGI (t=8.2, p<0.0001), which alone explained 55% of the variation in future significant hypoglycemia (eg, R2 =0.55); 2) SH history (t= 3.6, p<0.0005), which explained another 5% of the variation, and HbA 1c (t=2.2, p<0.03), which explained another 2% of the variation. This confirms previous results that LBGI was the most important predictor of future hypoglycemia, whereas the contribution of HbA 1c to this prediction was moderate.
通过Weibull模型计算的将来显著低血糖症的理论概率与未来观察到的显著低血糖症事件非常一致——对于严重和中度事件,确定系数均超过90%。Theoretical probabilities of future significant hypoglycemia calculated by the Weibull model were in good agreement with future observed significant hypoglycemia events—the coefficients of determination exceeded 90% for both severe and moderate events.
即将发生的(24小时内)显著低血糖症的预测:算法3的精度Prediction of imminent (within 24 hours) significant hypoglycemia: the accuracy of
下面的表分别给出了T1DM和T2DM对象SH和MH事件短期预测(24小时内)的精度。如果在24小时周期内可以获得一定数目的SMBG读数用于预测,则表20和21的每一行给出了预测到的事件的百分率。例如,每个表的第一行给出了被预测到的事件的百分率,而不管在某个事件之前24小时内是否有SMBG读数。可见,预测的精度随着事件之前读数数目的增加而增加。因此,如果某人每天测量3次或者更多次,则测量计能够报警且有可能帮助避免超过一半的显著低血糖症事件。The table below presents the accuracies for short-term prediction (within 24 hours) of SH and MH events in T1DM and T2DM subjects, respectively. Each row of Tables 20 and 21 gives the percentage of events predicted if a certain number of SMBG readings were available for prediction during the 24-hour period. For example, the first row of each table gives the percentage of events that were predicted, regardless of whether there were SMBG readings within 24 hours before an event. It can be seen that the accuracy of the prediction increases with the number of readings preceding the event. Thus, if someone measures 3 or more times per day, the meter can alert and potentially help avoid more than half of significant hypoglycemic events.
重点注意:出于评估算法3精度的目的,我们只使用了通过e-mail/电话系统报告的独立于SMBG的SH和MH事件,该系统要求参加者每两周报告SH和MH的日期和时间。如我们的调查所显示的,有时参加者在他们的报告中使用某一次事件之前的上一个SMBG读数的时间和日期,而不是该事件的精确时间和日期,因为查询测量计有助于帮助他们进行回忆。结果,有大量的事件,其之前最近一次SMBG读数的时刻与该事件时刻之间的时间间隔接近于零。为了解释这种可疑的时间记录,每个表的第3列给出了只严格限制于如下事件的算法3的精度,即前导报警时间至少为15分钟。假定平均前导报警时间为11小时,我们得出结论,在大多数情况下,报警的出现足够早从而有利于充分地自我治疗。 Important note : For the purpose of assessing the accuracy of
在表20和21中,烦扰指数被设定为Rud>=10以便和实例No.1的报告相匹配。In Tables 20 and 21, the nuisance index was set to Ru ud >= 10 to match the report of Example No.1.
表20:算法3在T1DM中的精度Table 20: Accuracy of
表20:算法3在T2DM中的精度Table 20: Accuracy of
本发明可以用其他的特殊形式加以实现,而不背离其精神或者基本特征。因此从任何方面讲,都应认为前述的实施例只是例证,而非对本文说明的发明具有限制。因此本发明的范围由附属权利要求指定,而不是由前述的说明,因此在与权利要求等价的意义和范围内的任何变化都应当包含在本文的范围之内。The present invention may be embodied in other specific forms without departing from its spirit or essential characteristics. Therefore, in any respect, the foregoing embodiments should be regarded as illustrations rather than limitations on the invention described herein. Therefore, the scope of the present invention is specified by the appended claims rather than the foregoing description, and any changes within the meaning and scope equivalent to the claims should be embraced in the scope of this document.
Claims (111)
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US40297602P | 2002-08-13 | 2002-08-13 | |
US60/402,976 | 2002-08-13 | ||
US60/478,377 | 2003-06-13 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA2007101032423A Division CN101079082A (en) | 2002-08-13 | 2003-08-08 | Method, system, and computer program product for the processing of self-monitoring blood glucose(AMBG)data to enhance diabetic self-management |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1688244A true CN1688244A (en) | 2005-10-26 |
CN100466965C CN100466965C (en) | 2009-03-11 |
Family
ID=35306237
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA2007101032423A Pending CN101079082A (en) | 2002-08-13 | 2003-08-08 | Method, system, and computer program product for the processing of self-monitoring blood glucose(AMBG)data to enhance diabetic self-management |
CNB038240092A Expired - Lifetime CN100466965C (en) | 2002-08-13 | 2003-08-08 | Method, system and computer program product for processing self-monitoring blood glucose (SMBG) data to improve self-management of diabetic patients |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA2007101032423A Pending CN101079082A (en) | 2002-08-13 | 2003-08-08 | Method, system, and computer program product for the processing of self-monitoring blood glucose(AMBG)data to enhance diabetic self-management |
Country Status (1)
Country | Link |
---|---|
CN (2) | CN101079082A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101785701A (en) * | 2009-01-23 | 2010-07-28 | 李秀 | Device and method for processing information, equipment and reagent carrier |
CN102824180A (en) * | 2011-04-25 | 2012-12-19 | 爱科来株式会社 | Information processing apparatus and user terminal |
CN103889317A (en) * | 2011-06-23 | 2014-06-25 | 弗吉尼亚大学专利基金会 | Unified platform for monitoring and control of blood glucose levels in diabetic patients |
CN104183104A (en) * | 2014-08-13 | 2014-12-03 | 上海电机学院 | Blood sugar test result communication method and system |
CN105190316A (en) * | 2013-02-21 | 2015-12-23 | 弗吉尼亚大学专利基金会 | Tracking changes in average glycemia in diabetics |
CN106955391A (en) * | 2009-02-04 | 2017-07-18 | 赛诺菲-安万特德国有限公司 | Medical device and method for providing information for glycemic control |
CN107003315A (en) * | 2014-12-25 | 2017-08-01 | 株式会社日立制作所 | Insulin secreting ability analytical equipment, the insulin secreting ability analysis system and insulin secreting ability analysis method for possessing the device |
CN107077525A (en) * | 2014-10-22 | 2017-08-18 | 德克斯康公司 | The user interface monitored for continuous glucose |
CN111816294A (en) * | 2012-07-09 | 2020-10-23 | 德克斯康公司 | System and method for utilizing smartphone features in continuous glucose monitoring |
CN113450906A (en) * | 2020-03-26 | 2021-09-28 | 苏州佳世达光电有限公司 | Clinical detection and diagnosis system and clinical detection and diagnosis method |
CN113505925A (en) * | 2021-07-09 | 2021-10-15 | 重庆邮电大学 | ANFIS-based laboratory dangerous chemical abnormal information early warning method |
CN114343580A (en) * | 2015-07-06 | 2022-04-15 | 雅培糖尿病护理公司 | Systems, devices and methods for symptom onset detection and assessment |
CN114689877A (en) * | 2020-12-30 | 2022-07-01 | 北京大学人民医院 | Data processing device for individualized prediction and calibration of glycated hemoglobin value and assist in judging the risk of elevated blood sugar |
CN115762782A (en) * | 2022-11-22 | 2023-03-07 | 南京晶捷生物科技有限公司 | Blood glucose analysis method based on CGM dynamic graph and edge calculation device |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100330598A1 (en) * | 2009-06-26 | 2010-12-30 | Roche Diagnostics Operations, Inc. | METHOD, SYSTEM, AND COMPUTER PROGRAM PRODUCT FOR PROVIDING BOTH AN ESTIMATED TRUE MEAN BLOOD GLUCOSE VALUE AND ESTIMATED GLYCATED HEMOGLOBIN (HbA1C) VALUE FROM STRUCTURED SPOT MEASUREMENTS OF BLOOD GLUCOSE |
EP2465060A1 (en) | 2009-08-10 | 2012-06-20 | Diabetes Tools Sweden Ab | Apparatus and method for processing a set of data values |
US8532933B2 (en) * | 2010-06-18 | 2013-09-10 | Roche Diagnostics Operations, Inc. | Insulin optimization systems and testing methods with adjusted exit criterion accounting for system noise associated with biomarkers |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5206144A (en) * | 1985-03-29 | 1993-04-27 | Novo Industri A/S | Determination of glycated (glycosylated) hemoglobin in blood |
JP3150857B2 (en) * | 1994-10-19 | 2001-03-26 | 富士写真フイルム株式会社 | Analysis element and method for measuring glycohemoglobin content ratio |
US5695949A (en) * | 1995-04-07 | 1997-12-09 | Lxn Corp. | Combined assay for current glucose level and intermediate or long-term glycemic control |
JPH11296598A (en) * | 1998-04-07 | 1999-10-29 | Seizaburo Arita | System and method for predicting blood-sugar level and record medium where same method is recorded |
-
2003
- 2003-08-08 CN CNA2007101032423A patent/CN101079082A/en active Pending
- 2003-08-08 CN CNB038240092A patent/CN100466965C/en not_active Expired - Lifetime
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101785701A (en) * | 2009-01-23 | 2010-07-28 | 李秀 | Device and method for processing information, equipment and reagent carrier |
CN106955391A (en) * | 2009-02-04 | 2017-07-18 | 赛诺菲-安万特德国有限公司 | Medical device and method for providing information for glycemic control |
CN106955391B (en) * | 2009-02-04 | 2020-11-20 | 赛诺菲-安万特德国有限公司 | Medical device and method for providing information for glycemic control |
US11135367B2 (en) | 2009-02-04 | 2021-10-05 | Sanofi-Aventis Deutschland Gmbh | Medical device and method for providing information for glycemic control |
CN102824180A (en) * | 2011-04-25 | 2012-12-19 | 爱科来株式会社 | Information processing apparatus and user terminal |
CN102824180B (en) * | 2011-04-25 | 2015-03-18 | 爱科来株式会社 | Information processing apparatus and user terminal |
CN103889317B (en) * | 2011-06-23 | 2017-12-29 | 弗吉尼亚大学专利基金会 | For the unified platform for the blood sugar level that diabetic is monitored and controlled |
CN103889317A (en) * | 2011-06-23 | 2014-06-25 | 弗吉尼亚大学专利基金会 | Unified platform for monitoring and control of blood glucose levels in diabetic patients |
CN111816294A (en) * | 2012-07-09 | 2020-10-23 | 德克斯康公司 | System and method for utilizing smartphone features in continuous glucose monitoring |
CN105190316A (en) * | 2013-02-21 | 2015-12-23 | 弗吉尼亚大学专利基金会 | Tracking changes in average glycemia in diabetics |
CN104183104A (en) * | 2014-08-13 | 2014-12-03 | 上海电机学院 | Blood sugar test result communication method and system |
US11670418B2 (en) | 2014-10-22 | 2023-06-06 | Dexcom, Inc. | User interfaces for continuous glucose monitoring |
CN107077525A (en) * | 2014-10-22 | 2017-08-18 | 德克斯康公司 | The user interface monitored for continuous glucose |
CN107077525B (en) * | 2014-10-22 | 2022-03-22 | 德克斯康公司 | User interface for continuous glucose monitoring |
CN107003315B (en) * | 2014-12-25 | 2018-09-25 | 株式会社日立制作所 | Insulin secreting ability analytical equipment, the insulin secreting ability analysis system and insulin secreting ability analysis method for having the device |
CN107003315A (en) * | 2014-12-25 | 2017-08-01 | 株式会社日立制作所 | Insulin secreting ability analytical equipment, the insulin secreting ability analysis system and insulin secreting ability analysis method for possessing the device |
CN114343580A (en) * | 2015-07-06 | 2022-04-15 | 雅培糖尿病护理公司 | Systems, devices and methods for symptom onset detection and assessment |
CN113450906A (en) * | 2020-03-26 | 2021-09-28 | 苏州佳世达光电有限公司 | Clinical detection and diagnosis system and clinical detection and diagnosis method |
CN114689877A (en) * | 2020-12-30 | 2022-07-01 | 北京大学人民医院 | Data processing device for individualized prediction and calibration of glycated hemoglobin value and assist in judging the risk of elevated blood sugar |
CN113505925A (en) * | 2021-07-09 | 2021-10-15 | 重庆邮电大学 | ANFIS-based laboratory dangerous chemical abnormal information early warning method |
CN113505925B (en) * | 2021-07-09 | 2022-07-15 | 重庆邮电大学 | A kind of abnormal information early warning method of laboratory hazardous chemicals based on ANFIS |
CN115762782A (en) * | 2022-11-22 | 2023-03-07 | 南京晶捷生物科技有限公司 | Blood glucose analysis method based on CGM dynamic graph and edge calculation device |
Also Published As
Publication number | Publication date |
---|---|
CN100466965C (en) | 2009-03-11 |
CN101079082A (en) | 2007-11-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN1688244A (en) | Method, system, and computer program product for the processing of self-monitoring blood glucose(smbg)data to enhance diabetic self-management | |
Husser et al. | Prevalence of clinically apparent hypertrophic cardiomyopathy in Germany—An analysis of over 5 million patients | |
Tao et al. | Estimating the cost of type 1 diabetes in the US: a propensity score matching method | |
JP5072182B2 (en) | Method, system and computer program product for processing self-monitoring blood glucose (SMBG) data to promote diabetes self-management | |
Cheetham et al. | Association of testosterone replacement with cardiovascular outcomes among men with androgen deficiency | |
Woolhandler et al. | The relationship of health insurance and mortality: is lack of insurance deadly? | |
Fancourt et al. | Trajectories of anxiety and depressive symptoms during enforced isolation due to COVID-19: longitudinal analyses of 36,520 adults in England | |
Peasgood et al. | The impact of diabetes-related complications on preference-based measures of health-related quality of life in adults with type I diabetes | |
Basu et al. | Issues for the next generation of health care cost analyses | |
CN1230266A (en) | Computerized medical diagnosis and treatment consultation system with network access | |
Lameijer et al. | Determinants of HbA1c reduction with FreeStyle Libre flash glucose monitoring (FLARE-NL 5) | |
CN1674030A (en) | Vital sign processing apparatus and vital sign processing system | |
CN1902650A (en) | Electronic medical information system, electronic medical information programs, and recording media | |
CN1604111A (en) | Disease management system and method including correlation assessment | |
Braithwaite et al. | Influence of alternative thresholds for initiating HIV treatment on quality-adjusted life expectancy: a decision model | |
CN1701343A (en) | Computer program product, system and method for information discovery and association analysis | |
Flores et al. | The economic burden of uncontrolled gout: how controlling gout reduces cost | |
Wan et al. | Clinical inertia in type 2 diabetes management in a middle-income country: A retrospective cohort study | |
Liu et al. | Regulation and competition in the Taiwanese pharmaceutical market under national health insurance | |
US20150379229A1 (en) | Computer-Implemented System And Method For Improving Glucose Management Through Modeling Of Circadian Profiles | |
Schmitz et al. | The impact of late-career job loss and genetic risk on body mass index: Evidence from variance polygenic scores | |
Ruiz-Casas et al. | Clinical and sociodemographic determinants of disease progression in patients with nonalcoholic steatohepatitis in the United States | |
Smith et al. | Improving the reliability of physician “report cards” | |
Lee et al. | Association of Sasang constitutional types with incident hypertension: a 12-year follow-up study | |
Jensen et al. | Long-term glucose-lowering effect of intermittently scanned continuous glucose monitoring for type 1 diabetes patients in poor glycaemic control from Region North Denmark: An observational real-world cohort study |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1084845 Country of ref document: HK |
|
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
REG | Reference to a national code |
Ref country code: HK Ref legal event code: GR Ref document number: 1084845 Country of ref document: HK |
|
CX01 | Expiry of patent term | ||
CX01 | Expiry of patent term |
Granted publication date: 20090311 |