Select search scope: search across all journals or within the current journal
Restricted accessOtherFirst published August, 2018pp. 319-320
, Jonas H Ellenberg
Restricted accessResearch articleFirst published August, 2018pp. 321-328, David L DeMetsMaintaining confidentiality of emerging data and ensuring the independence of Data Monitoring Committees are best practices of considerable importance to the ability of these committees to achieve their mission of safeguarding the interests of study participants and enhancing the integrity and credibility of clinical trials. Even with the wide recognition of these principles, there are circumstances where confidentiality issues remain challenging, controversial or inconsistently addressed. First, consider settings where a clinical trial’s interim data could provide the evidence regulatory authorities require for decisions about marketing approval, yet where such a trial would be continued post-approval to provide more definitive evidence about principal safety and/or efficacy outcomes. In such settings, data informative about the longer term objectives of the trial should remain confidential until pre-specified criteria for trial completion have been met. Second, for those other than Data Monitoring Committee members, access to safety and efficacy outcomes during trial conduct, even when presented as data pooled across treatment arms, should be on a limited “need to know” basis relating to the ability to carry out ethical or scientific responsibilities in the conduct of the trial. Third, Data Monitoring Committee members should have access to unblinded efficacy and safety data throughout the trial to enable timely and informed judgments about risks and benefits. Fourth, it should be recognized that a mediator potentially could be useful in rare settings where the Data Monitoring Committee would have serious ethical or scientific concerns about the sponsor’s dissemination or lack of dissemination of information. Data Monitoring Committee Contract Agreements, Indemnification Agreements and Charters should be developed in a manner to protect Data Monitoring Committee members and their independence, in order to enhance the Data Monitoring Committee’s ability to effectively address their mission.
Restricted accessResearch articleFirst published August, 2018pp. 329-334With the initiation in the late 1960s of the data and safety monitoring board or equivalently the data monitoring committee in randomized clinical trials came the need for interim statistical reports for these committees to review for study conduct and early evidence of harm or overwhelming evidence of benefit, perhaps leading to early trial termination. Initially, the statistical team was part of the data coordinating center for the trial. Later, starting in the early 1990s in many industry-sponsored trials, this statistical unit was separated organizationally from the team that collected and managed the data. This unit, often referred to as the statistical data analysis center, prepares reports for the data monitoring committee, which cover study conduct, data quality and completeness, primary and secondary outcomes, and safety measures by study arm in an unblinded fashion. The role of the statistical data analysis center is critical to any well-functioning data monitoring committee. With the proliferation of data monitoring committees has grown the need for many more well-trained and experienced statistical data analysis centers. In my experience, some such units perform their tasks extremely well but many do not. There is a tremendous need and opportunity to provide training for statistical data analysis centers, and what sponsors and data monitoring committees should expect from statistical data analysis centers.
Restricted accessOtherFirst published August, 2018pp. 335-351
Restricted accessResearch articleFirst published August, 2018pp. 352-358Before a novel treatment can be deemed a clinical success, an assessment of its risk–benefit profile must be made. One of the inherent challenges for this assessment comes from the multiplicity that arises from comparing treatment groups across multiple outcomes. Composite outcomes that summarize a patient’s clinical status, or severity, across a prioritized list of safety and efficacy outcomes have become increasing popular. In this article, we review these approaches and illustrate through examples some of the challenges and complexities of a composite derived from prioritized outcomes, such as the win ratio. These challenges include the difficult tension between the analytical validity that comes from choosing a pre-specified outcome and an evaluation that is responsive to unexpected safety events that arise during the course of a trial. Other challenges include a sensitivity of the resulting test statistic to the underlying censoring distribution and other nuisance parameters. Approaches that resolve some of the difficulties of the analytical challenges associated with prioritized outcomes are then discussed. Ultimately, a composite outcome of net clinical benefit is another decision tool, but one to be used alongside more traditional analyses of efficacy and safety, and with the broader perspective that investigators, the data safety monitoring board, and regulators bring to an evaluation of risk–benefit.
Restricted accessResearch articleFirst published August, 2018pp. 359-365Data monitoring committees for randomized clinical trials have the responsibility of safeguarding interests of trial participants. To do so, the data monitoring committee must receive reports on safety and efficacy to assess risk/benefit and on trial conduct to ensure that the study can achieve its goals. This article outlines the key components of reports to the data monitoring committee and the important role of the unblinded statistician in preparing those reports.
Most data monitoring committee meetings include open and closed sessions. For each session, there is a report of interim results. The open session is attended by the sponsor and lead investigators, including the statistician(s) responsible for the trial design. These investigators are blinded to the interim treatment comparisons. The closed session is attended by the data monitoring committee members and by the statistician(s) who prepared the closed report. These individuals are unblinded to interim treatment comparisons and therefore are not involved in study design changes. The optimal content of data monitoring committee reports and qualifications of the unblinded statistician(s) are discussed.
Open reports should include responses to data monitoring committee recommendations, a synopsis of the protocol, a review of the protocol history and amendments, and information on enrollment, baseline characteristics, completeness of follow-up, and data quality. The open report is also a vehicle through which the sponsor and investigators should inform the data monitoring committee of relevant external information. Data in the open report are pooled over the treatment groups. The open report should not include data summaries by treatment group. The closed report should include a written summary with references to key tables and figures and methods used to prepare them. Tables and figures should summarize baseline characteristics, follow-up completeness, treatment adherence, and major safety and efficacy outcomes by treatment group. Text summaries should accompany the tables and figures. The data monitoring committee monitoring history (e.g. treatment differences at previous meetings) should be summarized. The unblinded statistician preparing the closed report should be familiar with the protocol and data collection plan and be capable of customizing the report to the current stage of the trial. This includes anticipating questions that may arise during the data monitoring committee review and pro-actively including data summaries to address these questions.
There is considerable variation in the quality of open and closed data monitoring committee reports. Open and closed data monitoring committee reports should be concise, up to date, and informative. To achieve this, unblinded statisticians responsible for preparing closed data monitoring committee reports should be familiar with the statistical methods, the trial protocol, and the data collection plan. They should be capable of anticipating questions from the data monitoring committee and responding to requests for additional analyses.
Restricted accessOtherFirst published August, 2018pp. 366-385
Restricted accessResearch articleFirst published August, 2018pp. 386-397The goal of phase I clinical trials for cytotoxic agents is to find the maximum dose with an acceptable risk of severe toxicity. The most common designs for these dose-finding trials use a binary outcome indicating whether a patient had a dose-limiting toxicity. However, a patient may experience multiple toxicities, with each toxicity assigned an ordinal severity score. The binary response is then obtained by dichotomizing a patient’s richer set of data. We contribute to the growing literature on new models to exploit this richer toxicity data, with the goal of improving the efficiency in estimating the maximum tolerated dose.
We develop three new, related models that make use of the total number of dose-limiting and low-level toxicities a patient experiences. We use these models to estimate the probability of having at least one dose-limiting toxicity as a function of dose. In a simulation study, we evaluate how often our models select the true maximum tolerated dose, and we compare our models with the continual reassessment method, which uses binary data.
Across a variety of simulation settings, we find that our models compare well against the continual reassessment method in terms of selecting the true optimal dose. In particular, one of our models which uses dose-limiting and low-level toxicity counts beats or ties the other models, including the continual reassessment method, in all scenarios except the one in which the true optimal dose is the highest dose available. We also find that our models, when not selecting the true optimal dose, tend to err by picking lower, safer doses, while the continual reassessment method errs more toward toxic doses.
Using dose-limiting and low-level toxicity counts, which are easily obtained from data already routinely collected, is a promising way to improve the efficiency in finding the true maximum tolerated dose in phase I trials.
Restricted accessResearch articleFirst published August, 2018pp. 398-412Most deaths following severe traumatic brain injury follow decisions to withdraw life-sustaining therapies. However, the incidence of the withdrawal of life-sustaining therapies and its potential impact on research data interpretation have been poorly characterized. The aim of this systematic review was to assess the reporting and the impact of withdrawal of life-sustaining therapies in randomized clinical trials of patients with severe traumatic brain injury.
We searched Medline, Embase, Cochrane Central, BIOSIS, and CINAHL databases and references of included trials. All randomized controlled trials published between January 2002 and August 2015 in the six highest impact journals in general medicine, critical care medicine, and neurocritical care (total of 18 journals) were considered for eligibility. Randomized controlled trials were included if they enrolled adult patients with severe traumatic brain injury (Glasgow Coma Scale ≤ 8) and reported data on mortality. Our primary objective was to assess the proportion of trials reporting the withdrawal of life-sustaining therapies in a publication. Our secondary objectives were to describe the overall mortality rate, the proportion of deaths following the withdrawal of life-sustaining therapies, and to assess the impact of the withdrawal of life-sustaining therapies on trial results.
From 5987 citations retrieved, we included 41 randomized trials (n = 16,364, ranging from 11 to 10,008 patients). Overall mortality was 23% (range = 3%–57%). Withdrawal of life-sustaining therapies was reported in 20% of trials (8/41, 932 patients in trials) and the crude number of deaths due to the withdrawal of life-sustaining therapies was reported in 17% of trials (7/41, 884 patients in trials). In these trials, 63% of deaths were associated with the withdrawal of life-sustaining therapies (105/168). An analysis carried out by imputing a 4% differential rate in instances of withdrawal of life-sustaining therapies between study groups yielded different results and conclusions in one third of the trials.
Data on the withdrawal of life-sustaining therapies are incompletely reported in randomized controlled trials of patients with severe traumatic brain injury. Given the high proportion of deaths due to the withdrawal of life-sustaining therapies in severe traumatic brain injury patients, and the potential of this medical decision to influence the results of clinical trials, instances of withdrawal of life-sustaining therapies should be systematically reported in clinical trials in this group of patients.
Restricted accessResearch articleFirst published August, 2018pp. 413-423, Dikla Shmueli-BlumbergData sharing of de-identified individual participant data is being adopted by an increasing number of sponsors of human clinical trials. In addition to standardizing data syntax for shared trial data, semantic integration of various data elements is the focus of several initiatives that define research common data elements. This perspective article, in the first part, compares several data sharing platforms for de-identified clinical research data in terms of their size, policies and supported features. In the second part, we use a case study approach to describe in greater detail one data sharing platform (Data Share from National Institute of Drug Abuse). We present data on the past use of the platform, data formats offered, data de-identification approaches and its use of research common data elements. We conclude with a summary of current and expected future trends that facilitate secondary research use of data from completed human clinical trials.