A structured framework for adaptively incorporating external evidence in sequentially monitored clinical trials.

We present a Bayesian framework for sequential monitoring that allows for use of external data, and that can be applied in a wide range of clinical trial applications. The basis for this framework is the idea that, in many cases, specification of priors used for sequential monitoring and the stopping criteria can be semi-algorithmic byproducts of the trial hypotheses and relevant external data, simplifying the process of prior elicitation. Monitoring priors are defined using the family of generalized normal distributions, which comprise a flexible class of priors, naturally allowing one to construct a prior that is peaked or flat about the parameter values thought to be most likely. External data are incorporated into the monitoring process through mixing an a priori skeptical prior with an enthusiastic prior using a weight that can be fixed or adaptively estimated. In particular, we introduce an adaptive monitoring prior for efficacy evaluation that dynamically weighs skeptical and enthusiastic prior components based on the degree to which observed data are consistent with an enthusiastic perspective. The proposed approach allows for prospective and pre-specified use of external data in the monitoring procedure. We illustrate the method for both single-arm and two-arm randomized controlled trials. For the latter case, we also include a retrospective analysis of actual trial data using the proposed adaptive sequential monitoring procedure. Both examples are motivated by completed pediatric trials, and the designs incorporate information from adult trials to varying degrees. Preposterior analysis and frequentist operating characteristics of each trial design are discussed.

A practical Bayesian adaptive design incorporating data from historical controls.

In this paper, we develop the fixed-borrowing adaptive design, a Bayesian adaptive design which facilitates information borrowing from a historical trial using subject-level control data while assuring a reasonable upper bound on the maximum type I error rate and lower bound on the minimum power. First, one constructs an informative power prior from the historical data to be used for design and analysis of the new trial. At an interim analysis opportunity, one evaluates the degree of prior-data conflict. If there is too much conflict between the new trial data and the historical control data, the prior information is discarded and the study proceeds to the final analysis opportunity at which time a noninformative prior is used for analysis. Otherwise, the trial is stopped early and the informative power prior is used for analysis. Simulation studies are used to calibrate the early stopping rule. The proposed design methodology seamlessly accommodates covariates in the statistical model, which the authors argue is necessary to justify borrowing information from historical controls. Implementation of the proposed methodology is straightforward for many common data models, including linear regression models, generalized linear regression models, and proportional hazards models. We demonstrate the methodology to design a cardiovascular outcomes trial for a hypothetical new therapy for treatment of type 2 diabetes mellitus and borrow information from the SAVOR trial, one of the earliest cardiovascular outcomes trials designed to assess cardiovascular risk in antidiabetic therapies.

The role of quantitative safety evaluation in regulatory decision making of drugs.

Evaluation of safety is a critical component of drug review at the US Food and Drug Administration (FDA). Statisticians are playing an increasingly visible role in quantitative safety evaluation and regulatory decision-making. This article reviews the history and the recent events relating to quantitative drug safety evaluation at the FDA. The article then focuses on five active areas of quantitative drug safety evaluation and the role Division of Biometrics VII (DBVII) plays in these areas, namely meta-analysis for safety evaluation, large safety outcome trials, post-marketing requirements (PMRs), the Sentinel Initiative, and the evaluation of risk from extended/long-acting opioids. This article will focus chiefly on developments related to quantitative drug safety evaluation and not on the many additional developments in drug safety in general.

Seeing is believing: good graphic design principles for medical research.

Have you noticed when you browse a book, journal, study report, or product label how your eye is drawn to figures more than to words and tables? Statistical graphs are powerful ways to transparently and succinctly communicate the key points of medical research. Furthermore, the graphic design itself adds to the clarity of the messages in the data. The goal of this paper is to provide a mechanism for selecting the appropriate graph to thoughtfully construct quality deliverables using good graphic design principles. Examples are motivated by the efforts of a Safety Graphics Working Group that consisted of scientists from the pharmaceutical industry, Food and Drug Administration, and academic institutions.

Acute pain pathways: protocol for a prospective cohort study.

INTRODUCTION: Opioid analgesics are often used to treat moderate-to-severe acute non-cancer pain; however, there is little high-quality evidence to guide clinician prescribing. An essential element to developing evidence-based guidelines is a better understanding of pain management and pain control among individuals experiencing acute pain for various common diagnoses. METHODS AND ANALYSIS: This multicentre prospective observational study will recruit 1550 opioid-naïve participants with acute pain seen in diverse clinical settings including primary/urgent care, emergency departments and dental clinics. Participants will be followed for 6 months with the aid of a patient-centred health data aggregating platform that consolidates data from study questionnaires, electronic health record data on healthcare services received, prescription fill data from pharmacies, and activity and sleep data from a Fitbit activity tracker. Participants will be enrolled to represent diverse races and ethnicities and pain conditions, as well as geographical diversity. Data analysis will focus on assessing patients' patterns of pain and opioid analgesic use, along with other pain treatments; associations between patient and condition characteristics and patient-centred outcomes including resolution of pain, satisfaction with care and long-term use of opioid analgesics; and descriptive analyses of patient management of leftover opioids. ETHICS AND DISSEMINATION: This study has received approval from IRBs at each site. Results will be made available to participants, funders, the research community and the public. TRIAL REGISTRATION NUMBER: NCT04509115.

Similar Breast Cancer Risk in Women Older Than 65 Years Initiating Glargine, Detemir, and NPH Insulins.

OBJECTIVE: To assess whether initiation of insulin glargine (glargine), compared with initiation of NPH or insulin detemir (detemir), was associated with an increased risk of breast cancer in women with diabetes. RESEARCH DESIGN AND METHODS: This was a retrospective new-user cohort study of female Medicare beneficiaries aged ≥65 years initiating glargine (203,159), detemir (67,012), or NPH (47,388) from September 2006 to September 2015, with follow-up through May 2017. Weighted Cox proportional hazards regression was used to estimate hazard ratios (HRs) and 95% CIs for incidence of breast cancer according to ever use, cumulative duration of use, cumulative dose of insulin, length of follow-up time, and a combination of dose and length of follow-up time. RESULTS: Ever use of glargine was not associated with an increased risk of breast cancer compared with NPH (HR 0.97; 95% CI 0.88-1.06) or detemir (HR 0.98; 95% CI 0.92-1.05). No increased risk was seen with glargine use compared with either NPH or detemir by duration of insulin use, length of follow-up, or cumulative dose of insulin. No increased risk of breast cancer was observed in medium- or high-dose glargine users compared with low-dose users. CONCLUSIONS: Overall, glargine use was not associated with an increased risk of breast cancer compared with NPH or detemir in female Medicare beneficiaries.

Clinical trials and statistical analyses: what should dermatologists look for in a report?

Clinicians need to evaluate the quality of individual clinical studies and synthesize the information from multiple clinical studies to provide insights in selecting appropriate therapies for patients. Understanding the key statistical principles that underlie a clinical trial and how they may be implemented can help clinicians properly interpret the efficacy and safety findings of clinical trials. Several factors should be considered when evaluating clinical studies reported in the literature, as important differences might exist among reported studies, thereby impacting the reliability of their findings. Studies vary in terms of study design, conduct, analysis, and presentation of findings. The key features to consider when evaluating clinical trials are inferential intent (exploratory versus confirmatory), choice of control group, randomization, extent of blinding, prespecification of analyses, appropriate handling of missing data, and multiple end points. Making comparisons across studies is extremely difficult and rarely statistically justified. However, this article will point out issues to keep in mind when evaluating multiple studies, such as variations in design and study populations.

Developing optimal prediction models for cancer classification using gene expression data.

Microarrays can provide genome-wide expression patterns for various cancers, especially for tumor sub-types that may exhibit substantially different patient prognosis. Using such gene expression data, several approaches have been proposed to classify tumor sub-types accurately. These classification methods are not robust, and often dependent on a particular training sample for modelling, which raises issues in utilizing these methods to administer proper treatment for a future patient. We propose to construct an optimal, robust prediction model for classifying cancer sub-types using gene expression data. Our model is constructed in a step-wise fashion implementing cross-validated quadratic discriminant analysis. At each step, all identified models are validated by an independent sample of patients to develop a robust model for future data. We apply the proposed methods to two microarray data sets of cancer: the acute leukemia data by Golub et al. and the colon cancer data by Alon et al. We have found that the dimensionality of our optimal prediction models is relatively small for these cases and that our prediction models with one or two gene factors outperforms or has competing performance, especially for independent samples, to other methods based on 50 or more predictive gene factors. The methodology is implemented and developed by the procedures in R and Splus. The source code can be obtained at http://hesweb1.med.virginia.edu/bioinformatics.

No association of abacavir use with myocardial infarction: findings of an FDA meta-analysis.

BACKGROUND: Several studies have reported an association between abacavir (ABC) exposure and increased risk of myocardial infarction (MI) among HIV-infected individuals. Randomized controlled trials (RCTs) and a pooled analysis by GlaxoSmithKline, however, do not support this association. To better estimate the effect of ABC use on risk of MI, the US Food and Drug Administration (FDA) conducted a trial-level meta-analysis of RCTs in which ABC use was randomized as part of a combined antiretroviral regimen. METHODS: From a literature search conducted among 4 databases, 26 RCTs were selected that met the following criteria: conducted in adults, sample size more than 50 subjects, status completed, not a pharmacokinetic trial, and not conducted in Africa. The Mantel-Haenszel method, with risk difference and 95% confidence interval, was used for the primary analysis, along with additional alternative analyses, based on FDA-requested adverse event reports of MI provided by each investigator. RESULTS: The 26 RCTs were conducted from 1996 to 2010, and included 9868 subjects (5028 ABC and 4840 non-ABC). Mean follow-up was 1.43 person-years in the ABC group and 1.49 person-years in the non-ABC group. Forty-six (0.47%) MI events were reported [24 (0.48%) ABC and 22 (0.46%) non-ABC], with no significant difference noted between the 2 groups (risk difference of 0.008% with 95% confidence interval: -0.26% to 0.27%). CONCLUSIONS: To the best of our knowledge, our study represents the largest trial-level meta-analysis to date of clinical trials in which ABC use was randomized. Our analysis found no association between ABC use and MI risk.

Robust classification modeling on microarray data using misclassification penalized posterior.

MOTIVATION: Genome-wide microarray data are often used in challenging classification problems of clinically relevant subtypes of human diseases. However, the identification of a parsimonious robust prediction model that performs consistently well on future independent data has not been successful due to the biased model selection from an extremely large number of candidate models during the classification model search and construction. Furthermore, common criteria of prediction model performance, such as classification error rates, do not provide a sensitive measure for evaluating performance of such astronomic competing models. Also, even though several different classification approaches have been utilized to tackle such classification problems, no direct comparison on these methods have been made. RESULTS: We introduce a novel measure for assessing the performance of a prediction model, the misclassification-penalized posterior (MiPP), the sum of the posterior classification probabilities penalized by the number of incorrectly classified samples. Using MiPP, we implement a forward step-wise cross-validated procedure to find our optimal prediction models with different numbers of features on a training set. Our final robust classification model and its dimension are determined based on a completely independent test dataset. This MiPP-based classification modeling approach enables us to identify the most parsimonious robust prediction models only with two or three features on well-known microarray datasets. These models show superior performance to other models in the literature that often have more than 40-100 features in their model construction. AVAILABILITY: Our MiPP software program is available at the Bioconductor website (http://www.bioconductor.org).