Diseases and Disorders – Diabetes The effect of having the condition diabetes on road safety 1 Summary Goldenbeld, C. & van Schagen, I., SWOV Institute for Road Safety Research, The Hague, Netherlands, August 1.1 COLOUR CODE: YELLOW Studies generally show a (small) elevated crash risk. However, effects are not always statistically significant. Many studies have low quality, e.g. because they did not adjust for exposure or mileage. Furthermore, the results are possibly compromised by national countermeasures, e.g., some countries impose driving restrictions on drivers with insulin-treated diabetes. When the higher risk diabetes drivers are not allowed to participate in traffic, this will affect the overall risk of diabetes identified in that country. 1.2 KEYWORDS diseases and disorders, diabetes, crash risk, car drivers 1.3 ABSTRACT This chapter discusses the effect of diabetes on road safety. Diabetes mellitus is a group of metabolic diseases characterised by defects in insulin secretion, insulin action, or both. Studies generally show a (slightly) higher risk for drivers with diabetes, although differences are often not statistically significant. Two main approaches have been used to study the relationship between diabetes and crash risk. The most common approach compares crash rates of individuals with diabetes with crash rates of individuals without diabetes. The less common approach first distinguishes between drivers who have and who have not been involved in a crash, and then compares the prevalence of diabetes in these two groups. Most research has been done in the USA, Canada, and Europe. Most of the research is on private drivers; very few studies are on commercial drivers. 1.4 BACKGROUND What is diabetes? Diabetes mellitus (or diabetes) is a chronic condition that affects the body's ability to use glucose in food for energy. For this the hormone insulin is needed. With diabetes the body either does not make sufficient insulin or it cannot use the insulin, or both. The two main types of diabetes are type 1 and type 2. Type 1 diabetes, often called insulin-dependent diabetes or juvenile-onset diabetes, accounts for 5 to 10 percent of all diagnosed cases of diabetes; type 2 diabetes, often called noninsulin-dependent diabetes or adult-onset diabetes, accounts for 90 to 95 percent (Bieber-Tregear et al., 2011). Risk factors for developing type 2 diabetes include older age, obesity, family history of diabetes, history of gestational diabetes, impaired glucose tolerance, physical inactivity, and race/ethnicity (Bieber-Tregear et al., 2011). Treatments for diabetes aim to maintain blood glucose levels near normal (euglycemia) at all times. Exact treatment differs for type 1 and type 2 diabetes but generally includes diet control, physical activity, home blood glucose testing several times a day, and regular insulin injections or oral medication. How does diabetes affect road safety? The most important acute threat of diabetes for road safety is hypoglycaemia (Bieber-Tregear et al., 2011; Graveling & Frier, 2015). Hypoglycemia is a clinical syndrome that results from abnormally low levels of blood glucose which can arise as a result of treatments for diabetes. The body’s biochemical response to hypoglycemia usually starts when blood sugar levels fall below 65 to 70 mg/dl (3.6 to 3.9 mmol/L). If the blood glucose level falls below 60 mg/dl (3.3 mmol/L), physical symptoms begin to become apparent: sweating, tremor, hunger, anxiety, and palpitations. Experimental laboratory studies have demonstrated that cognitive functions critical to driving (such as attention, reaction times and hand-eye coordination) are impaired during hypoglycaemia (Graveling & Frier, 2015). Hypoglycemia also affects the visual information processing and visual perception, and hence driving performance. This is most apparent under conditions of limited perceptual time and low visual contrast (poor light). Driving simulator studies have shown that driving performance is already affected adversely by moderate hypoglycemia, causing problems such as inappropriate speeding or braking, ignoring road signs and traffic lights and not keeping to traffic lanes (Graveling & Frier, 2015). Furthermore, there are a number of medical complications associated with diabetes that could affect driving competency, including cardiovascular disease, diabetic neuropathy, and diabetic retinopathy (Bieber-Tregear et al, 2011; Graveling & Frier, 2015). How many people have diabetes? In the European Region, there are about 60 million people with diabetes; i.e. about 10.3% of men and 9.6% of women aged 25 years and over (WHO, site accessed 2 May 2016). The prevalence of diabetes is increasing among all ages in the European region, mostly due to increases in overweight and obesity, unhealthy diet and physical inactivity (WHO site accessed 2 May 2016). The prevalence of diabetes varies widely in the 56 diverse countries in the European region, from 2.4% in Moldova to 14.9% in Turkey (Tamayo et al., 2014). In the USA around 9.3% of the population has diabetes (American Diabetes Association site accessed 2 May 2016). Which factors influence the effect of diabetes on road safety? In theory, the diabetes-risk relationship could be affected by personal factors (e.g. gender, age, type of driving), specific treatment factors, and national conditions (e.g. national screening and countermeasures for diabetes). However systematic comparable evidence about the influence of these factors is scarce. Diabetes risk studies that have included gender and age often use them as covariates rather than independent variables. There is no systematic evidence of the effect of specific treatment on crash risk. A meta-analysis found that insulin-treatment of diabetes was associated with non-significant 21% risk increase compared to non-treatment with insulin. Concerning national conditions, a meta-analysis found that the increased risk of drivers with diabetes was significant in the USA, but not in other countries (Canada, Norway, Northern Ireland, Scotland, Sweden). This difference was attributed to stricter diabetes checks and regulations for drivers in Europe and Canada. How is the effect of diabetes on road safety measured? Two main approaches have been used to study the relationship between diabetes and crash risk. The most common approach is the case-control study. Such a study compares the crash rates of individuals with diabetes (cases) with crash rates of individuals without diabetes (controls). The less common approach first distinguishes between drivers who have and who have not been involved in a crash, and then compares the prevalence of diabetes in these two groups. Most research has been done in the USA, Canada, and Europe. Most of the research is on private drivers; very few studies are on commercial drivers. Most studies included both type 1 and type 2 diabetes. The studies generally look at one of the following measures of road safety:  Actual/police-registered crash involvement,  Actual/police-registered at fault (culpable) crash involvement,  Self-reported crash involvement,  Self-reported at fault crash involvement. 1.5 OVERVIEW RESULTS Studies generally show that drivers with diabetes have a slightly increased crash risk compared to drivers without diabetes. However, effects are often statistically non-significant. Hence we cannot exclude the possibility that the observed difference in crash risk is not a real difference but based on chance and accidental fluctuations. In theory, factors that could influence the effect of diabetes on crash risk are personal factors (e.g. gender, age), medical treatment factors, or national conditions (e.g. screening and counter-measures for diabetes). However, the evidence concerning these types of modifying conditions is scarce and indirect. Most of the research is on private drivers; very few studies are on commercial drivers. There has been no systematic research on differences between type 1 and type 2 diabetes on crash risk. 1.6 NOTES ON RESEARCH AND ANALYSIS METHOD Most studies that aim to assess the risk of diabetes compare the accident risk of diabetes patients with the accident risk of people without diabetes. It is also possible to compare the prevalence of diabetes in people who have and who have not been involved in an accident. This is a less common method. 2 Scientific Details 2.1 DESCRIPTION OF CODED STUDIES Table 1 provides further description of the background characteristics of the coded studies on diabetes and crash risk. Table 1: Characteristics of coded studies on diabetes and driving risk Author, Year, Country Method and analysis Risk group/ Cases Control group/ Controls Modifying conditions/ control variables BieberTregear et al. 2011 International Meta-analysis. Random effects. 15 studies comparing crash involvement between diabetic and non-diabetic drivers Diabetic drivers (crash/no crash) Non-diabetic drivers (crash/no crash) Comparison US and nonUS studies BieberTregear et al. 2011 International Meta-analysis. Random effects. 6 studies comparing crash involvement between insulin treated diabetic drivers and otherwise treated diabetic drivers Insulin treated diabetic drivers Oral medication or diet treated diabetic drivers Comparison US and nonUS studies BieberTregear et al. 2011 International Meta-analysis. Random effects. 4 studies comparing prevalence of diabetes between crashinvolved and non-crash involved drivers Crash-involved drivers (diabetes/no diabetes) Non-crash involved drivers (diabetes/no diabetes) 4 studies also reported on conditions of diabetes treatment (insulin, pharma-cotherapy, controlled diet alone). BieberTregear et al. 2011 International Meta-analysis. Fixed effects. 4 studies comparing prevalence of insulin-treated diabetic drivers among crash-involved and noncrash-involved drivers Crash-involved drivers (insulin/ no insulin) Non-crash involved drivers (insulin/no insulin) Comparison US and nonUS studies Sagberg, 2006, Norway Self-report questionnaires from 4448 crash-involved drivers. Odds ratio calculated. Cases: At fault (n = 2226) Controls: Not at fault (n = 1840) Analysis adjusted for age and annual driving distance. Redelmeier et al. 2009, Canada In 2-y study interval 795 diabetic patients who had HbA1c values documented were reported to licensing authorities. Logistic regression. Cases: 57 patients were involved in a crash Controls: 738 were not involved in a crash Analyses controlled for age, gender, medical complication, history severe hypoglycemia, age diabetes diagnosed Signorovitch et al. 2012 USA Diabetes-2 people (not insulin treated) identified from a claims database (1998–2010). Crash occurrence leading to hospital visits was com-pared between people with, and without claims for hypoglycaemia. Analysis by multivariate Cox proportional hazard models. n=5.582 people with claims for hypoglycaemia n=27.910 with no such claims were Analysis adjusted for demo-graphics, comorbidities, prior treatments and prior medical service use Vingilis & Wilk 2012 Canada Population-based large-scale panel research (N = 12.387). 524 (4.2%) reporting an motor vehicle injury MVI 1996- 2007. Path analyses examined the odds of subsequent MVI. Diabetes reporting MVI, n =14 Diabetic drivers not reporting MVI, n = 346 Analysis controlled for age, gender and independent effects of medication use. Orriols et al. 2014 France 69.630 drivers involved in an injurious crash in France 20052008. Logistic regression analysis ; outcome = odds of being responsible for crash Cases: drivers who were deemed responsible for the crash (n =33.200) Controls: drivers who were not responsible for crash (n = 36.450). Analysis adjusted for age, gender, socioeconomic category, month, time of day, vehicle type, alcohol level, injury severity, exposure to medicines and other long-term diseases. Description of main research methods In the coded studies two main approaches were used for investigating crash risk in individuals with diabetes (Figure 1, taken from Bieber-Tregear et al. 2011). On the one hand, cohorts can be identified based on whether or not they have diabetes. In this scenario, crash rates among a group of individuals with diabetes (i.e., cases) are compared with crash rates among a group of individuals without diabetes. An alternative less used approach is to identify cohorts on the basis of whether or not they have had a crash, and then compare the prevalence of diabetes in the two groups. Figure 1: Scenarios for investigating risk of crash in diabetes (from: Bieber-Tregear et al., 2011) According to Bieber-Tregear et al. (2011) 15 studies on diabetes-crash risk relationship can be classified under scenario 1, and 4 studies under scenario 2. Most research has been done in the USA, Canada, and Europe. All of the coded studies were on private drivers. Most of the coded studies included both type 1 and type 2 diabetes. The coded studies generally looked at one of the following measures of road safety:  Actual/police-registered crash involvement,  Actual/police-registered at fault (culpable) crash involvement,  Self-reported crash involvement,  Self-reported at fault crash involvement. Besides these basic approaches the coded studies on diabetes and risk varied on:  Matching of cases and controls,     2.2 Recorded crashes versus self-reported crashes, Inclusion of independent variables (e.g. several treatment conditions), Inclusion of statistical control variables (e.g. gender, age, mileage, treatment, other diseases). RESULTS Literature review Diabetes mellitus (or diabetes) is a chronic condition that affects the body's ability to use the glucose in food for energy. For this the hormone insulin is needed. With diabetes the body either does not make sufficient insulin or it cannot use the insulin, or both. The two main types of diabetes are type 1 and type 2. Type 1 diabetes, often called insulin-dependent diabetes or juvenile-onset diabetes, accounts for 5 to 10 percent of all diagnosed cases of diabetes; type 2 diabetes, often called noninsulin-dependent diabetes or adult-onset diabetes, accounts for 90 to 95 percent (Bieber-Tregear et al., 2011). Risk factors for developing type 2 diabetes include older age, obesity, family history of diabetes, history of gestational diabetes, impaired glucose tolerance, physical inactivity, and race/ethnicity (Bieber-Tregear et al., 2011). Treatments for diabetes aim to maintain blood glucose levels near normal (euglycemia) at all times. Exact treatment differs for type 1 and type 2 diabetes but generally includes diet control, physical activity, home blood glucose testing several times a day, and regular insulin injections or oral medication. In the European Region, there are about 60 million people with diabetes; i.e. about 10.3% of men and 9.6% of women aged 25 years and over (WHO, site accessed 2 May 2016). The prevalence of diabetes is increasing among all ages in the European region, mostly due to increases in overweight and obesity, unhealthy diet and physical inactivity (WHO site accessed 2 May 2016). The prevalence of diabetes varies widely in the 56 diverse countries in the European region from 2.4% in Moldova to 14.9% in Turkey (Tamayo et al., 2014). In the USA around 9.3% of the population have diabetes (American Diabetes Association site accessed 2 May 2016). The most important acute threat of diabetes for road safety is hypoglycaemia (Bieber-Tregear et al., 2011; Graveling & Frier, 2015). Hypoglycemia is a clinical syndrome that results from abnormally low levels of blood glucose which can arise as a result of treatments for diabetes. The body’s biochemical response to hypo-glycemia usually starts when blood sugar levels fall below 65 to 70 mg/dl (3.6 to 3.9 mmol/L). If the blood glucose level falls below 60 mg/dl (3.3 mmol/L), physical symptoms begin to become apparent: sweating, tremor, hunger, anxiety, and palpitations. Experimental laboratory studies have demonstrated that cognitive functions critical to driving (such as attention, reaction times and hand-eye coordination) are impaired during hypoglycaemia (Graveling & Frier, 2015). Hypoglycemia also affects the visual information processing and visual perception, and hence driving performance. This is most apparent under conditions of limited perceptual time and low visual contrast (poor light). Driving simulator studies have shown that driving performance is already affected adversely by moderate hypoglycemia, causing problems such as inappropriate speeding or braking, ignoring road signs and traffic lights and not keeping to traffic lanes (Graveling & Frier, 2015). Furthermore, there are a number of medical complications associated with diabetes that could affect driving competency, including cardiovascular disease, diabetic neuropathy, and diabetic retinopathy (Bieber-Tregear et al, 2011; Graveling & Frier, 2015). In theory, the diabetes-risk relationship could be affected by personal factors (e.g. gender, age, type of driving), specific treatment factors, and national conditions (e.g. national screening and countermeasures for diabetes). However systematic comparable evidence about the influence of these factors is scarce. Diabetes risk studies that have included gender and age often use them as covariates rather than independent variables. There is no systematic evidence of the effect of specific treatment on crash risk. A meta-analysis found that insulin-treatment of diabetes was associated with non-significant slight 21% risk increase compared to non-treatment with insulin. Concerning national conditions, a meta-analysis found that the increased risk of drivers with diabetes was significant in the USA, but not in other countries (Canada, Sweden, Norway, Northern Ireland, Scotland, Sweden). This difference was attributed to stricter diabetes checks and regulations for drivers in Europe and Canada. Meta-analysis A 2011 meta-analysis of 15 case-control studies indicated that the magnitude of increased crash risk was small and not statistically significant (Risk Ratio=1.126; 95% CI: 0.847–1.497; p=0.415). These case-control studies used a scenario 1 design (as described in Figure 1). A subgroup meta-analysis on 6 studies compared the crash risk of insulin-treated drivers with diabetes to that of drivers with diabetes who control their condition with pharmacotherapy or diet alone. The result of this analysis was not significant: OR = 1.537 (95% CI: 0.603–3.915, p=0.368). A random-effects meta-analysis on 4 studies (with a design according to scenario 2 in Figure 1) found that drivers with diabetes are not over-represented among samples of drivers who have experienced a crash (OR = 1.052, 95% CI: 0.970–1.141; p=0.220). A fixed effects analysis on the same 4 studies found that drivers with diabetes who controlled diabetes using insulin had a non-significant higher crash rate when compared with those who do not use insulin to control their diabetes (OR = 1.212; 95% CI: 0.939–1.563, p=0.139). Additional studies The studies not included in the meta-analysis and/or appearing after the meta-analysis also showed mixed results (Sagberg, 2006; Redelmeier et al., 2009; Orriols et al., 2014; Vingilis & Wilk, 2012; Signorovitch et al., 2012). In a Norwegian study Sagberg analysed questionnaire data from 4448 crash-involved drivers. For drivers with untreated diabetes (diabetes type 2), he found a significant odds ratio indicating increased risk for being at fault for a crash (OR=3.08, p = 0.05). No effect was found for treated or medicated diabetes. A strong point of the study was that the researcher corrected for age and mileage; at the same time the questionnaire response was low and the method (induced-exposure method) does not allow one to determine crash risk of diabetics when compared with rest of population. A large, nationally representative, longitudinal, self-report only study of Canadians indicated a non-significant crash risk of diabetes after controlling for age and gender (OR = 1.479, 95% CI: 0.743 - 2.944; p = 0.266; Vingilis & Wilk, 2012). A large scale longitudinal French study (combining information from the national healthcare insurance database, police reports and the national police database of injury crashes) found a significant effect of type 1 diabetes on being responsible for a crash (OR = 1.47, CL=1.12–1.92; p = 0.0047; Orriols et al., 2014). This estimate was corrected for various covariates (age, gender, socioeconomic category, month, time of day, vehicle type, alcohol level, injury severity, exposure to medicines affecting driving abilities and other long-term diseases). In large scale prospective database study, Signorovitch et al. (2012) compared the occurrence of accidents resulting in hospital visits between people with, and without, claims for hypoglycaemia after the initiation of a non-insulin antidiabetic drug. These researchers adjusted the risk estimates for demographics, comorbidities, prior treatments and prior medical service use, and they also conducted analyses stratified by age (< 65; 65 years or older). After adjusting for baseline characteristics, hypoglycaemia was associated with significantly increased risks for motor vehicle accidents (Hazard Ratio = 1.82, 95% CI 1.18–2.80, p=0.007). Modifying conditions In theory, conditions that might modify the diabetes-risk relationship could be personal factors (e.g. gender, age), specific treatment factors, or national conditions (e.g. screening and countermeasures for diabetes).  Age and gender Several studies have used age and/or gender as statistical control variables (Sagberg, 2006; Orriols et al., 2014; Vingilis & Wilk, 2012) but very few have used them as independent variables. It seems that diabetes driving risk does not increase with age. Skurtveit (2009; included in the Bieber-Tregear et al. meta-analysis) found that highest crash involvement was among 18-34 yrs. Signorovitch et al. (2012) found that hypoglycemia was associated with greater hazards of driving-related accidents in people younger than 65. Hypoglycaemia was associated with greater hazards of driving-related accidents in people younger than 65. Among the younger people, hypoglycemia was significantly associated with a greater than 130% increase in the risk of motor vehicle accidents (adjusted HR: 2.31; 95% CI: 1.44–3.70).  National differences Subgroup analysis indicated that the relative risk effect was significant in the USA (RR = 1.284; 95% CI=1.124-1.466; p <0.0001), but not in non-USA countries (1.035; 95%, CI: 0.720-1.487; p=0.854) (Bieber-Tregear et al., 2011). This difference in findings has been attributed to stricter diabetes checks and regulations for EU-drivers (Bieber-Tregear et al., 2011).  Specific treatment factors Although insulin treatment is a risk factor for hypoglycemia and hypoglycemia is considered to be a vital mechanism explaining the increased risk for diabetes, there is no evidence that insulin-treated persons are over-represented in crashes. A fixed effects meta-analysis on 4 studies (method 2, Figure 1) found that drivers with insulin controlled diabetes tend to be over-represented among samples of drivers who have experienced a crash: this result was not statistically significant (odds ratio =1.212; 95% CI: 0.939–1.563, p=0.139) (Bieber-Tregear et al., 2011). Conclusions There is some evidence that drivers with diabetes have a slight increased crash risk compared to drivers without diabetes. Most studies indicate a slightly elevated risk estimate, but results are not always significant (see Table 2 below). Table 2: Overview of studies and their (simplified) main outcomes (↑ = statistical significant increase in crash involvement; ↑ n.s. = statistical not significant increase; = = no effect). Study Simplified summary of main outcomes Bieber-Tregear et al. 2011, intern. Crash involvement ↑ n.s. Bieber-Tregear et al. 2011, intern. Crash involvement insulin-treated vs. oral or diet treatment ↑ n.s. Bieber-Tregear et al. 2011, intern. Crash involvement = Bieber-Tregear et al. 2011, intern. Crash involvement insulin-treated vs. oral or diet treatment ↑ n.s. Sagberg, 2006, Norway Self-reported crash culpability non-treated ↑ Sagberg, 2006, Norway Self-reported crash culpability treated = Redelmeier et al. 2009, Canada Crash involvement ↑ Signorovitch et al., 2012, USA Crash involvement (resulting in hospital visit) ↑ Vingilis & Wilk, 2012, Canada Motor vehicle injury ↑ n.s. Orriols et al., 2014, France Crash culpability ↑ The 2011 meta-analysis was based on a comprehensive literature search, review guiding decision rules and clearly defined quality assessment criteria. Most of the 15 case-control studies included in the meta-analysis were rated as low in quality; for example 9 of 15 case-control studies did not adjust for exposure (mileage). The five studies after the meta-analysis or not included in the meta-analyses have used different designs and different outcome measures, and thus are not homogeneous, and their results cannot be pooled. Other complications are that over time studies are difficult to compare since the effects of treatments/medicine for diabetes may change over time. Comparison between countries is difficult since countries may differ in regulations concerning diabetes and driving. One reason why studies may fail to show a significant difference in crash rates at a population level between people at risk of hypoglycemia (mainly those with insulin-treated diabetes) and the general population with driving licenses is that countries impose restrictions on drivers with insulin-treated diabetes and remove those who are at high risk of having an accident. Studies generally show that drivers with diabetes have a slightly increased crash risk compared to drivers without diabetes. However, effects are often statistically non-significant. Hence we cannot exclude the possibility that the observed difference in crash risk is not a real difference but based on chance and accidental fluctuations. The main studies and their outcomes are:  A 2011 meta-analysis of 15 case-control studies indicated a non-significant increase of the crash risk (actual crashes) of 13% (Bieber-Tregear et al., 2011).  A large-scale Canadian longitudinal study indicated a non-significant increase in the crash risk (self-reports) of 48% (Vingilis & Wilk, 2012).  A large-scale longitudinal French study, based on information from the national health insurance database and the national police injury crash database, reported a significant increase in crash risk (actual at fault crashes) of 47% (Orriols et al., 2014). In theory, factors that could influence the effect of diabetes on crash risk are personal factors (e.g. gender, age), medical treatment factors, or national conditions (e.g. screening and countermeasures for diabetes). However, the evidence concerning these types of modifying conditions is scarce and indirect. Most of the research is on private drivers; very few studies are on commercial drivers. There has been no systematic research on differences between type 1 and type 2 diabetes on crash risk. 3 Supporting documents 3.1 LITERATURE SEARCH STRATEGY The literature on diabetes and traffic risk was searched for in the international database Scopus on 23 March 2016. Scopus is the largest international peer-reviewed database. The literature was searched over the period 1999-2016; the search terms (Table 3) were searched in title, abstract and keywords. Also references were looked at in very recent review like texts. This search produced 164 hits. Database: Scopus, Date: 23 March 2016 Table 3: Used search terms and logical operators Search terms/logical operators/combined queries hits History Search Terms ( TITLE-ABS-KEY ( diabetes OR hypoglycemia OR hypoglycaemia OR hyperglycemia OR hyperglyceamia ) AND TITLE-ABS-KEY ( "road accident" OR "traffic accident" OR "accident risk" OR "crash risk" OR "road risk" OR "risky driving" OR "road safety" OR crash OR collision ) AND TITLE-ABS-KEY ( driving OR driver ) ) AND SUBJAREA ( mult OR agri OR bioc OR immu OR neur OR phar OR mult OR medi OR nurs OR vete OR dent OR heal OR mult OR arts OR busi OR deci OR econ OR psyc OR soci ) AND PUBYEAR > 1999 164 In a first screening round these 164 references were screened on potential relevance for coding based on title and abstract information. Criteria A to E describe reasons for not selecting publications in the first round (Table 4). A= B= C= D= E= Not selected because paper refers to measure/intervention. Not selected because diabetes is side subject (and not itself investigated in relationship to traffic risk). Not selected because written in non-English. Not selected because better or more complete results were published earlier or later in another publication (duplication). Not selected because general review-like text. Table 4: Initial screening of studies on diabetes and crash risk Exclusion criterion Not selected first round A. Measure/intervention 21 B. Side subject (not itself directly investigated) 59 C .Non-English 20 D. Duplication 4 E. General review-like text 33 Selected first round Initially selected 27 The 27 studies selected in this initial screening were further screened on relevance for coding in a second round. In the second round the same criteria were used but were checked on full-text copies of the papers. Table 5 presents the results of the second screening round and describes the final decisions concerning coding of the studies. In the end, 12 studies were coded. Of these 12 studies 6 were eventually analysed. 5 had been included in a recent meta-analysis, and consequently, were not analysed separately. One was an older meta-analysis which was replaced by a more recent meta-analysis. Table 5: Selection of studies after the second screening round Reference Relevant Coded Analysed 1 Abu Dabrh A.M. , Firwana B., Cowl C.T., Steinkraus L.W. , Prokop L.J. , & Murad M.H.(2014). Health assessment of commercial drivers: A meta-narrative systematic review (2014) BMJ Open, 4 (3), art. no. e003434 Review can point to specific studies, not relevant for coding but potentially relevant for review chapter No 2 Avalos, M. et al., 2014. Variable selection on large case-crossover data: application to a registry-based study of prescription drugs and road traffic crashes. Pharmacoepidemiology and drug safety, 23, 140–151. Likely not relevant for coding: this article is about the problem how to screen hypotheses using probabilistic reasoning, selecting drug classes or individual drugs that most warrant further hypothesis testing. No 3 Bieber-Tregear, M., Funmilayo, D., Amana, A., Connor, D., & Tregear, S. (2011). Diabetes and commercial motor vehicle safety. Washington: Department of Transportation’s Federal Motor Carrier Safety Administration. Yes. Meta-analysis Yes 4 Bloomfield, H.E., Greer, N., Newman, D., MacDonald, R., Carlyle, M., Fitzgerald, P., & Rutks, I. (2012). Predictors and Consequences of Severe Hypoglycemia in Adults with Diabetes - A Systematic Review of the Evidence. Department of Veterans Affairs, Washington. Review not relevant for coding No 5 Burki, T.K.( 2013). Diabetes and driving. The lancet. Diabetes & endocrinology, 1, e7–8 Not relevant for coding since this article discusses new European legislation. No 6 Campbell, L.K. et al. (2010). Neurocognitive Differences Between Drivers with Type 1 Diabetes with and without a Recent History of Recurrent Driving Mishaps. International journal of diabetes mellitus, 2, 73–77. Likely not relevant for coding since the study deals with neurocognitive differences No 7 Cox, D.J., Gonder- Frederick, L.A., Kovatchev, B. P., Julian, D.M., Clarke, W.L. (2000). Progressive hypoglycemia ’s impact on driving simulation performance: occurrence, awareness, and correction. Diabetes Care 23, 163–170. Yes Yes No 8 Cox D.J., Ford D., Gonder-Frederick L., Clarke W., Mazze R., & Weinger K. (2009). Driving mishaps among individuals with type 1 diabetes. Diabetes Care, 32, 2177-2180. Yes Yes No Yes 9 Cox, D.J., Singh, H. & Lorber, D.(2013). Diabetes and driving safety: science, ethics, legality and practice. The American journal of the medical sciences, 345, 263–5 Review, not relevant for coding, but relevant for final review chapter No 10 Harsch, I.A., Stocker, S., Radespiel-Tröger, M., Hahn, E.G., Konturek, P.C., Ficker, J.H., & Lohmann, T. (2002). Traffic hypoglycaemias and accidents in patients with diabetes mellitus treated with different antidiabetic regimens. Journal of Internal Medicine, 252, 352-360. Yes Yes 11 Hassoun, A.A.K. et al., 2015. Driving and diabetes mellitus in the Gulf Cooperation Council countries: Call for action. Diabetes research and clinical practice, 110(1), 91–4. Likely not relevant for coding since this paper seems more about extent of problem in Gulf countries and possible countermeasures. No 12 Hemmelgarn, B., Levesque, L.E., Suissa, S. (2006). Anti-diabetic drug use and the risk of motor vehicle crash in the elderly. Canadian Journal Clinical Pharmacology, 13, e112–20. Yes Yes 13 Hitosugi, M. et al., 2015. Main factors causing healthrelated vehicle collisions and incidents in Japanese taxi drivers. Romanian Journal of Legal Medicine, 23, 83–86 Likely not relevant since study concentrates on prevalence of several diseases (diabetes) among taxi drivers) No 14 Kagan A. , Hashemi G. , Korner-Bitensky N. (2010). Diabetes and fitness to drive: A systematic review of the evidence with a focus on older drivers. Canadian Journal of Diabetes, 34, 233-242. Review can point to specific studies; at least relevant for review chapter but not for coding. No 15 Laberge-Nadeau, C., Dionne, G., Ekoe, J.M., Hamet, P., Desjardins, D., Messier S., & Maag, U (2000). Impact of diabetes on crash risks of truck-permit holders and commercial drivers. Diabetes Care, 23, 612-617. Yes Yes No 16 Lonnen, K.F., Powell, R.J., Taylor, D., et al. (2008). Road traffic accidents and diabetes: insulin use does not determine risk. Diabetes Medicine,25, 578–84. Yes Yes No 17 Marrero, D. & Edelman, S.(2000). Hypoglycemia and driving performance: A flashing yellow light? Diabetes Care, 23, 146–147. Not suited for coding since this article mainly reviews some of earlier research in particular also a study by Cox et al. 2000 No 18 Matsumura, M. et al. (2014). Hypoglycemic attacks in diabetic patients while driving an automobile. Journal of the Japan Diabetes Society, 57(5), 329–336. In Japanese language No 19 Orriols L. , Avalos-Fernandez M., Moore N. , Philip P. , Delorme B. , Laumon B. , Gadegbeku B. , Salmi L.-R., Lagarde E. (2014). Long-term chronic diseases and crash responsibility: A record linkage study. Accident Analysis & Prevention, 71, 137-143 Yes Yes 20 Parmentier, G. et al., (2005). Road mobility and the risk of road traffic accident as a driver. The impact of medical conditions and life events. Accident Analysis & Prevention, 37, 1121–1134 Not relevant, this study only mentions diabetes as a factor for road mobility mot as a risk factor No 21 Raubenheimer, P. (2012). Diabetes mellitus and Not relevant for coding, review-like No No No Yes driving. Journal of Endocrinology, Metabolism and Diabetes of South Africa, 17(2 SUPPL. 1). paper, giving guidelines 22 Raubenheimer, P.(2012). Diabetes mellitus and driving. Journal of Endocrinology, Metabolism and Diabetes of South Africa, 17(1). Not relevant for coding, review-like paper, giving guidelines No 23 Redelmeier, D.A., Kenshole A.B., Ray J.G. (2009)., Motor vehicle crashes in diabetic patients with tight glycemic control: A population-based case control analysis. PLoS Med, 6, e1000192 Yes Yes Yes 24 Sagberg F. (2006). Driver Health and Crash Involvement: A Case-Control Study. Accident. Analysis & Prevention, 38, 28-34. Yes Yes Yes 25 Signorovitch, J.E., Macaulay, D., Diener, M., Yan, Y., Wu, E.Q., Gruenberger, J.-B.& Frier,B.M. (2012). Hypoglycaemia and accident risk in people with type 2 diabetes mellitus treated with non-insulin antidiabetes drugs. Diabetes, Obesity and Metabolism, 15, 335– 341. Yes Yes Yes 26 Vaa, T. (2003) Impairment, Diseases, Age and Their Relative Risks of Accident Involvement: Results from Meta-Analysis. Oslo: TØI Report 690 for the Institute of Transport Economics.. Yes Yes No 27 Vingilis, E., & Wilk, P. (2012). Medical conditions, medication use, and their relationship with subsequent motor vehicle injuries: examination of the Canadian National Population Health Survey. Traffic Injury Prevention, 13, 327-36. Yes Yes Yes 3.2 BACKGROUND CHARACTERISTICS OF THE ANALYSED STUDIES Table 6 provides a detailed description of the background characteristics of the analysed studies. It should be noted that the meta-analysis of Bieber-Tregear et al. (2011) consisted of 4 separate analyses each of which is included in Table 6. Table 6: Characteristics of analysed studies Author, Year, Country Sample, method/design and analysis Risk group/ Cases Control group/ Controls Research conditions/ control variables BieberTregear et al. 2011 International Meta-analysis. Random effects. 15 studies comparing crash involvement between diabetic and non-diabetic drivers Diabetic drivers (crash/no crash) Non-diabetic drivers (crash/no crash) Comparison US and non-US studies BieberTregear et al. 2011 International Meta-analysis. Random effects. 6 studies comparing crash involvement between insulin treated diabetic drivers and otherwise treated diabetic drivers Insulin treated diabetic drivers Oral medication or diet treated diabetic drivers Comparison US and non-US studies BieberTregear et al. 2011 International Meta-analysis. Random effects. 4 studies comparing prevalence of diabetes between crash-involved and not crash-involved drivers Crash-involved drivers (diabetes/no diabetes) Non-crash involved drivers (diabetes/no diabetes) 4 studies also reported on conditions of diabetes treatment (insulin, pharmacotherapy, controlled diet alone). BieberTregear et al. 2011 International Meta-analysis. Fixed effects. 4 studies comparing prevalence of insulin-treated diabetic drivers among crash-involved and noncrash-involved drivers Crash-involved drivers (insulin/ no insulin) Non-crash involved drivers (insulin/no insulin) Comparison US and non-US studies Sagberg, 2006, Norway Self-report questionnaires from 4448 crash-involved drivers. Odds ratio calculated. Cases: At fault (n = 2226) Controls: Not at fault (n = 1840) Analysis adjusted for age and annual driving distance. Redelmeier et al. 2009, Canada In 2-y study interval 795 diabetic patients who had HbA1c values documented were reported to licensing authorities. Logistic regression. Cases: 57 patients were involved in a crash Controls: 738 were not involved in a crash Analyses controlled for age, gender, medical complication, history severe hypoglycemia, age diabetes diagnosed Signorovitch et al. 2012 USA Diabetes-2 people (not insulin treated) identified from a claims database (1998–2010). Crash occurrence leading to hospital visits was com-pared between people with, and without claims for hypo-glycaemia. Analysis by multivariate Cox proportional hazard models. n=5.582 people with claims for hypoglycaemia n=27.910 with no such claims were Analysis adjusted for demographics, comorbidities, prior treatments and prior medical service use Vingilis & Wilk 2012 Canada Population-based large-scale panel research (N = 12.387). 524 (4.2%) reporting an motor vehicle injury MVI 1996- 2007. Path analyses examined the odds of subsequent MVI. Diabetes reporting MVI, n =14 Diabetic drivers not reporting MVI, n = 346 Analysis controlled for age, gender and independent effects of medication use. Orriols et al. 2014 France 69.630 drivers involved in an injurious crash in France 20052008. Logistic regression analysis ; outcome = odds of being responsible for crash Cases: drivers who were deemed responsible for the crash (n =33.200) Controls: drivers who were not responsible for crash (n = 36.450). Analysis adjusted for age, gender, socio-economic category, month, time of day, vehicle type, alcohol level, injury severity, exposure to medicines and other longterm diseases. The meta-analysis was based on a comprehensive literature search, review guiding decision rules and clearly defined quality assessment criteria. Most of the 15 case-control studies included in the meta-analysis were rated as low in quality; for example 9 of 15 case-control studies did not adjust for exposure (mileage). The five studies after meta-analysis or not included in the meta-analysis used different designs and different outcome measures, and thus are not homogeneous, and their results cannot be pooled. Other complications are that over time studies are difficult to compare since the effects of treatments/medicine for diabetes may change over time. Comparison between countries is difficult since countries may differ in regulations concerning diabetes and driving. One reason why studies may fail to show a significant difference in crash rates at a population level between people at risk of hypoglycemia (mainly those with insulin-treated diabetes) and the general population with driving licenses is that countries impose restrictions on drivers with insulin-treated diabetes and remove those who are at high risk of having an accident. 3.3 OVERVIEW OF THE RESULTS OF THE ANALYSED STUDIES An overview of the main results of the analysed studies is presented in Table 7. Author, Year, Country Risk factor Study type Outcome variable Effects for Road Safety Main outcome -description Bieber-Tregear et al. 2011 International Diabetes 1 and 2 Meta-analysis Random effects 15 studies Crash involvement RR=1.126; 95% CI: 0.847–1.497; p=0.415 Increased crash risk was small and not statistically significant Bieber-Tregear et al. 2011 International Insulintreated diabetes Meta-analysis. Random effects. 6 studies Crash involvement OR = 1.537; 95% CI: 0.603–3.915, p=0.368). Non-significant increase in crash risk for insulin-treated drivers when compared with drivers treated with oral medication and/or diet alone. Bieber-Tregear et al.2011 International Diabetes 1 and 2 Meta-analysis Random effect 4 studies Crash involvement OR = 1.052; 95%, CI: 0.970–1.141; p=0.220 Drivers with diabetes are not over-represented among samples of drivers who have experienced a crash Bieber-Tregear et al.2011 International Insulintreated diabetes Meta-analysis Fixed effects 4 studies Crash involvement OR=1.212; 95% CI: 0.939–1.563, p=0.139 Drivers with insulin controlled diabetes tend to be overrepresented among samples of drivers who have experienced a crash; not statistically significant Sagberg 2006 Norway Diabetes 1 and 2 Questionnaire study. Induced exposure: at fault crash-involved drivers compared not at fault. Selfreported crash culpability Non-medicated diabetic drivers: (Diabetes Type II) OR=3.08, p = 0.05 The adjusted odds ratio was significant for non-medicated diabetic drivers For diabetic drivers on medication ( Diabetes 1) the OR was nonsignificant. Redelmeier et al. 2009 Canada Glycemic control A populationbased case control analysis Crash involvement OR= 1.26, 95% Cl:1.03–1.54) Crash risk increases 26% for each 1% reduction in HbA1c (finding robust after control for confounders) Signorovitch et al 2012 USA Hypoglycemia (Diabete s 2) Case-control comparing diabetes 2 patients with and without evidence hypoglycemia Crash involvement (resulting in hospital visit) Hazard ratio (HR) = 1.82 (95% CL: 1.18, 2.80) People < 65 years; HR = 2.31 (95% CL: 1.44, 3.70) After adjusting for baseline characteristics, hypoglycaemia significantly increased hazard Vingilis & Wilk 2012 Canada Diabetes 1 and 2 Population-based large-scale panel research Motor vehicle injury OR = 1.479, 95% CI: 0.743 - 2.944; p = 0.266 (NS). No significantly increased odds of subsequent MVI was found for diabetes Orriols et al. 2014 France Diabetes 1 and 2 Case-control analysis comparing responsible vs. non responsible crashinvolved drivers. Crash culpability (estimated by standard method) Diabetes type 1: OR = 1.47; 95% Cl 1.12–1.92; p = 0.0047 Significantly increased risk of being responsible for a crash found for drivers with type 1 diabetes. Type 2 diabetes not selected in final risk model. Table 7: Main results of analysed studies 3.4 REFERENCES Analysed studies Bieber-Tregear, M., Funmilayo, D., Amana, A., Connor, D., & Tregear, S. (2011). Diabetes and commercial motor vehicle safety. Washington DC: Department of Transportation’s Federal Motor Carrier Safety Administration. Orriols, L., Avalos-Fernandez M., Moore N., Philip P., Delorme B., Laumon B., Gadegbeku B., Salmi L.R., Lagarde E.. (2014). Long-term chronic diseases and crash responsibility: a record linkage study. Accident Analysis & Prevention, 71,137–143. Redelmeier, D.A., Kenshole, A.B. & Ray, J.G. (2009). Motor vehicle crashes in diabetic patients with tight glycemic control: a population-based case control analysis. PLoS medicine, 6, p.e1000192. Sagberg, F. (2006). Driver health and crash involvement: a case-control study. Accident Analysis & Prevention, 38, 28–34. Signorovitch, J.E., Macaulay, D., Diener, M., Yan, Y., Wu, E.Q., Gruenberger, J.-B.& Frier,B.M. (2012). Hypoglycaemia and accident risk in people with type 2 diabetes mellitus treated with non-insulin antidiabetes drugs. Diabetes, Obesity and Metabolism, 15, 335–341. Vingilis, E., & Wilk, P. (2012). Medical conditions, medication use, and their relationship with subsequent motor vehicle injuries: examination of the Canadian National Population Health Survey. Traffic Injury Prevention, 13, 327-36. Other references Cox, D.J., Singh, H. & Lorber, D. (2013). Diabetes and Driving Safety: Science, Ethics, Legality & Practice. American Journal of the Medical Sciences, 345, 263–265. European Road Observatory (2012). Fitness to drive. Brussels, EC. Retrieved May 2, 2016, from: http://www.roadsafetyobservatory.com/Summary/drivers/fitness-to-drive Graveling, A.J. & Frier, B.M. (2015). Driving and diabetes: problems, licensing restrictions and recommendations for safe driving. Clinical Diabetes and Endocrinology. DOI 10.1186/s40842015-0007-3 Inkster, B. & Frier, B.M. (2013). Diabetes and driving. Diabetes Obesity and Metabolism, 15,. 775-783. Skurtveit, S., Strom, H., Skrivarhaug, T., et al. (2009). Road Traffic Accident Risk in Patients with Diabetes Mellitus Receiving Blood Glucose-lowering Drugs. Prospective Follow-up Study. Diabetic Medicine, 26, 404–408. Tamayo, T., Rosenbauer,J., Wild, S.H., Spijkerman, A.M., Baan, C., Forouhi,N.G., Herder, C., & Rathmann, W. (2014). Diabetes in Europe: an update. Diabetes Research and Clinical Practice, 103, 206–217.