Daniel Sargent

The American Joint Committee on Cancer (AJCC) has increasingly recognized the need for more personalized probabilistic predictions than those delivered by ordinal staging systems, particularly through the use of accurate risk models or calculators. However, judging the quality and acceptability of a risk model is complex. The AJCC Precision Medicine Core conducted a 2-day meeting to discuss characteristics necessary for a quality risk model in cancer patients. More specifically, the committee established inclusion and exclusion criteria necessary for a risk model to potentially be endorsed by the AJCC. This committee reviewed and discussed relevant literature before creating a checklist unique to this need of AJCC risk model endorsement. The committee identified 13 inclusion and 3 exclusion criteria for AJCC risk model endorsement in cancer. The emphasis centered on performance metrics, implementation clarity, and clinical relevance. The facilitation of personalized probabilistic pr...

Development of oncologic therapies has traditionally been performed in a sequence of clinical trials intended to assess safety (phase I), preliminary efficacy (phase II), and improvement over the standard of care (phase III) in homogeneous (in terms of tumor type and disease stage) patient populations. As cancer has become increasingly understood on the molecular level, newer "targeted" drugs that inhibit specific cancer cell growth and survival mechanisms have increased the need for new clinical trial designs, wherein pertinent questions on the relationship between patient biomarkers and response to treatment can be answered. Herein, we review the clinical trial design literature from initial to more recently proposed designs for targeted agents or those treatments hypothesized to have enhanced effectiveness within patient subgroups (e.g., those with a certain biomarker value or who harbor a certain genetic tumor mutation). We also describe a number of real clinical trial...

Factors associated with early mortality after surgery and treatment with adjuvant chemotherapy in colon cancer are poorly understood. We aimed to characterize the determinants of early mortality in a large cohort of colon cancer trial participants. A pooled analysis of 37,568 patients in 25 randomized trials of adjuvant systemic therapy was conducted. Multivariable logistic regression models with several definitions of early mortality (30, 60, and 90 days, and 6 months) were constructed, adjusting for clinically and statistically significant variables. A nomogram for 6-month mortality was developed and validated. Median age among patients was 61 years, patient demographics included 54% men and 90% White, 29% and 71% had stage II and III disease, respectively, and 79%, 20%, and 1% had an Eastern Cooperative Oncology Group performance status (PS) of 0, 1, and ≥ 2, respectively. Early mortality was low: 0.3% at 30 days, 0.6% at 60 days, 0.8% at 90 days, and 1.4% at 6 months. Of those p...

Fluorouracil plus leucovorin (FU + LV) adjuvant chemotherapy reduced the risk of recurrence and death across all time points in a pooled analysis of 20,898 patients with colon cancer from 18 randomized studies. The impact of oxaliplatin added to FU + LV on the time course of recurrence and survival remains unknown. A total of 12,233 patients enrolled to the randomized trials C-07, C-08, N0147, MOSAIC (Adjuvant Treatment of Colon Cancer), and XELOXA (Adjuvant XELOX) were pooled to examine the impact of oxaliplatin and tumor-specific factors on the time course of recurrence and death. For each end point, continuous-time risk was modeled over 6 years post treatment in all oxaliplatin-treated patients and patients concurrently randomized to FU + LV with or without oxaliplatin; the latter analyses supported time-dependent treatment comparisons. Addition of oxaliplatin significantly reduced the risk of recurrence within the first 14 months post treatment for patients with stage II disease...

Phase II clinical trials inform go/no-go decisions for proceeding to phase III trials, and appropriate end points in phase II trials are critical for facilitating this decision. Phase II solid tumor trials have traditionally used end points such as tumor response defined by Response Evaluation Criteria for Solid Tumors (RECIST). We previously reported that absolute and relative changes in tumor measurements demonstrated potential, but not convincing, improvement over RECIST to predict overall survival (OS). We have evaluated the metrics by using additional measures of clinical utility and data from phase III trials. Resampling methods were used to assess the clinical utility of metrics to predict phase III outcomes from simulated phase II trials. In all, 2,000 phase II trials were simulated from four actual phase III trials (two positive for OS and two negative for OS). Cox models for three metrics landmarked at 12 weeks and adjusted for baseline tumor burden were fit for each phase...

Current standard evaluation of Peripheral Neuropathy (PN) is based on an investigator-reported classification system that is commonly unable to correctly reflect the subjective symptoms for patients. Thus more reliable methods to assess PN are needed. This study assessed alternative methods of assessing patient-reported PN in 5 North Central Cancer Treatment Group (NCCTG) clinical trials. Two single-item assessments relating to numbness and tingling were used to measure PN. Patients' Quality Of Life (QOL) was also assessed using the Uniscale, Symptom Distress Scale (SDS), Profile of Mood States (POMS), Brief Pain Inventory (BPI) and Subject Global Impression of Change (SGIC). Wilcoxon tests compared QOL scores between patients with PN (score > 50) vs. no PN (score ≤ 50). Changes from baseline in QOL were compared by Wilcoxon rank sum test with a 20-point change in PN defined as clinically meaningful. Both distribution-based and anchor-based approaches were used to derive estimates of Minimal Important Differences (MID). Standardized Response Means (SRM), Effect Sizes (ES) and Guyatt's responsiveness statistic were used to measure responsiveness. The proportion of patients reporting numbness (tingling) at baseline was 10.7% (10.0%) and 18.4% (17.8%) at last assessment. The correlation between numbness and tingling at baseline was 0.81, and at last assessment was 0.83. Patients with substantial PN reported an average of 10 points lower overall QOL, mood and worse symptom distress and 20 points lower in the BPI interference items. Patients having a ≤ 20 point worsening in PN score reported significantly worse in symptom distress and BPI worst pain, but not in POMS or overall QOL. The MID estimates were similar between numbness and tingling items but varied depending on the approach used. Responsiveness statistics indicated that the two PN assessments are sensitive and responsive instruments for cancer patients with PN. The two PN items for numbness and tingling were redundant. Evidence of criterion validity and responsiveness indicates that these simple measures of PN can be used successfully in cancer clinical trials.

To compare operative morbidity, mortality, quality of life, and survival after pancreatoduodenectomy (PD) versus pancreatoduodenectomy with extended lymphadenectomy (PD/ELND) in patients with resectable pancreatic cancer. From May 1997 to July 2003 there were 132 patients with biopsy examination-proven or suspected adenocarcinoma of the pancreatic head who agreed to participate in a single-institution, prospective, randomized trial. If resectable at operation, patients then were randomized to standard PD (40 patients) or PD/ELND (39 patients). Quality of life was assessed by using the Functional Assessment of Response to Cancer Therapy specific to the pancreas. Morbidity, mortality, and survival were analyzed. Demographics and pathologic characteristics for both groups were similar. When comparing PD/ELND with standard PD, the median operating time was greater for the PD/ELND group (7.6 h vs 6.2 h, P < .01), blood transfusion more likely (44% vs 22%, P < .05), and the median number of lymph nodes resected was greater (36 vs 15 nodes, P < .01). Morbidity and mortality rates were comparable. Median durations of stay were 11 and 10.5 days (P = NS), respectively. There were no significant differences in 1-year (71% vs 82%), 3-year (25% vs 41%), 5-year (16.5% vs 16.4%), and median (19 vs 26 mo) survival (P = .32). At 4 months postoperatively, diarrhea, body appearance, and bowel control scored lower on the Functional Assessment of Response to Cancer Therapy specific to the pancreas after PD/ELND (P < .05). Although a much larger study would have more power to compare statistically the survival between groups, both the decrement in quality of life and similar studies showing no survival difference make PD/ELND unattractive for further prospective investigation.

... Leonard Saltz. Memorial Sloan-Kettering Cancer Center, New York, NY. Alan Venook. University of California San Francisco, San Francisco, CA. Greg Yothers. University of Pittsburgh, Pittsburgh, PA. Daniel Sargent. Mayo Clinic, Rochester, MN. ...

In this article, we provide a high-level overview of statistical concepts related to study design and data analysis in oncology research. These concepts are discussed for 2 main types of clinical research: (1) observational studies, which focus on biomarker discovery in order to predict disease risk and prognosis, and (2) prospectively designed, well-controlled clinical trials, which are critical for the development of new cancer treatments. Throughout the article, we emphasize the importance of appropriate design and prospectively determined analysis plans. We also hope to promote effective collaboration between oncology investigators and statisticians who center their research on the development of cancer treatments.

Prognostic models have been proposed to predict survival for non-small-cell lung cancer (NSCLC). It is important to evaluate whether these models perform better than performance status (PS) alone in stage- and age-specific subgroups. The validation cohort included 2060 stage I and 1611 stage IV NSCLC patients from 23CALGB studies. For stage I, Blanchon (B), Chansky (C) and Gail (G) models were evaluated along with the PS only model. For stage IV, Blanchon (B) and Mandrekar (M) models were compared with the PS only model. The c-index was used to assess the concordance between survival and risk scores. The c-index difference (c-difference) and the integrated discrimination improvement (IDI) were used to determine the improvement of these models over the PS only model. For stage I, B and PS have better survival separation. The c-index for B, PS, C and G are 0.61, 0.58, 0.57 and 0.52, respectively, and B performs significantly better than PS with c-difference=0.034. For stage IV, B, M a...

... 12. Gennaro Galizia, Eva Lieto, Francesca Ferraraccio, Ferdinando Vita, Paolo Castellano, Michele Orditura, Vincenzo Imperatore, Anna La Mura, Giovanni La Manna, MargheritaPinto, Giuseppe Catalano, Carlo Pignatelli, Fortunato Ciardiello. ...

... Daniel F. Hayes. REFERENCES. 1. Ellis M, Hayes DF, Lippman ME: Treatment of metastatic breast cancer, in Harris J, Lippman M, Morrow M, et al (eds): Diseases of the Breast (ed 3). Philadelphia, PA, Lippincott Williams & Wilkins, 2004, pp 1101-1162. 2. Sloan JA, Dueck A ...

A general Gibbs sampling algorithm for analyzing a broad class of linear models under a Bayesian framework is presented using Markov Chain Monte Carlo (MCMC) methodology in the SAS system. The analysis of a North Central Cancer Treatment Group (NCCTG) oncology clinical trial involving a two-period two-treatment crossover design is presented as an example. Results for the Bayesian model are compared to standard linear models analysis of variance procedures.

This article discusses end points and novel clinical trial design in the era of targeted therapies, and the need to accelerate the drug development process so that the right therapies can quickly be delivered to the right patients with gastrointestinal cancers. The increasing number of novel targeted therapies available for testing for cancers, including those of the gastrointestinal tract, requires new approaches to clinical trials to identify promising agents for Phase 3 testing—particularly with respect to end points and trial designs.

Developing biomarkers that can predict whether patients are likely to benefit from an intervention is a pressing objective in many areas of medicine. Recent guidance documents have recommended that the accuracy of predictive biomarkers, ie, sensitivity, specificity, and positive and negative predictive values, should be assessed. We clarify the meanings of these entities for predictive markers and demonstrate that generally they cannot be estimated from data without making strong untestable assumptions. Language suggesting that predictive biomarkers can identify patients who benefit from an intervention is also widespread. We show that in general one cannot estimate the chance that a patient will benefit from treatment. We recommend instead that predictive biomarkers be evaluated with respect to their ability to predict clinical outcomes among patients treated and among patients receiving standard of care, and the population impact of treatment rules based on those predictions. Ideally these entities are estimated from a randomized trial comparing the experimental intervention with standard of care.

Consensus practice guidelines and the implementation of clinical therapeutic advances are usually based on the results of large, randomized clinical trials (RCTs). However, RCTs generally inform us on an average treatment effect for a presumably homogeneous population, but therapeutic interventions rarely benefit the entire population targeted. Indeed, multiple RCTs have demonstrated that interindividual variability exists both in drug response and in the development of adverse effects. The field of pharmacogenomics promises to deliver the right drug to the right patient. Substantial progress has been made in this field, with advances in technology, statistical and computational methods, and the use of cell and animal model systems. However, clinical implementation of pharmacogenetic principles has been difficult because RCTs demonstrating benefit are lacking. For patients, the potential benefits of performing such trials include the individualization of therapy to maximize efficacy and minimize adverse effects. These trials would also enable investigators to reduce sample size and hence contain costs for trial sponsors. Multiple ethical, legal, and practical issues need to be considered for the conduct of genotype-based RCTs. Whether pre-emptive genotyping embedded in electronic health records will preclude the need for performing genotype-based RCTs remains to be seen.

The database of the Adjuvant Colon Cancer End Points (ACCENT) Group was assembled to address questions in early stage colon cancer that could be best answered by information pooled across many similar trials. Today, the ACCENT database contains individual patient-level data from over 33,000 patients enrolled onto 25 adjuvant colon cancer trials conducted between 1977 and 2008. Since its flagship analysis of 3-year disease-free survival as a surrogate endpoint for 5-year overall survival in 2005, the ACCENT group has produced many noteworthy scientific findings addressing a variety of clinical questions, which we describe here. Additionally, we provide an overview of the history, collaboration, construction, principles, and future of the ACCENT database, as it has set a precedent for multi-trial database creation in other types of cancer.

Inferences from tumor marker studies are complicated by a variety of statistical concerns, which can make proper interpretation of results difficult. This article focuses on important issues that should be addressed when designing, conducting, and analyzing tumor marker studies. To highlight the importance of considering statistical significance, risk ratios, statistical power, reproducibility, multiple testing, confirmatory studies, and missing data in the design of marker studies used for prognosis. Suggestions are provided for more effectively conducting marker studies. These include more careful attention to adequacy of the number of subjects for a marker study and improved documentation and standardization of assay methods. The importance of complete reporting of study results and description of the statistical analysis methods used is also emphasized. Cooperation among clinicians, laboratory scientists, and statisticians will be required to conduct statistically sound tumor ma...

Research on methods for studying time-to-event data (survival analysis) has been extensive in recent years. The basic model in use today represents the hazard function for an individual through a proportional hazards model (Cox, 1972). Typically, it is assumed that a covariate's effect on the hazard function is constant throughout the course of the study. In this paper we propose a method to allow for possible deviations from the standard Cox model, by allowing the effect of a covariate to vary over time. This method is based on a dynamic linear model. We present our method in terms of a Bayesian hierarchical model. We fit the model to the data using Markov chain Monte Carlo methods. Finally, we illustrate the approach with several examples.