Preface

Last February I issued a call for short case reports on the deliberations of data and safety monitoring boards (DSMBs). We received nine papers in response to that call and six of them are published in this issue of Controlled Clinical Trials. Three other articles that describe data monitoring issues for specific studies, two articles on general statistical methodology for data monitoring, and a review of a recent book on the functions and operations of data monitoring committees round out this issue. If, like me, you have enjoyed the sessions at the Society meetings that give us a glimpse of what happens during the sometimes complex deliberations of DSMBs in their review of interim results, I think you will enjoy the case reports in this issue.

Often you have but one chance to do a major trial and Wheatley and Clayton emphasize the importance of not stopping early unless you have “proof beyond reasonable doubt.” I think reports like this are a must-read for new DSMB members. Chen-Mok et al. describe how they agonized over a subgroup finding—I suspect many of us can relate to their story. Hillman and Louis describe the decision making and steps that were taken in an AIDS trial when the interim results were different than a similar study stopped early for efficacy. The DSMB's decision to partially unblind patients in the ongoing trial to interim results of that trial raises the general question of when, how much, to whom, and under what conditions, interim data should be released in an ongoing trial. Of note, this specific example is also discussed in the book by Ellenberg, Fleming, and DeMets that is reviewed in this issue.

Gilpin et al. describe a number of interesting issues around the communication of results to investigators and trial sponsors when a DSMB (PDMB in their case) thinks a trial should be terminated. Their experience emphasizes the importance of having a charter for the DSMB that clearly outlines reporting responsibilities. The case report by Collins et al. illustrates how a DSMB can be effective, at least in part, by providing administrative oversight for a trial. In this case, the DSMB appeared to play a significant role in monitoring recruitment. Interestingly, the DSMB for this study, like the one described by Wheatley, also had to cope with some early worrisome trends that ultimately disappeared. Finally, the report by Scheifele et al. describes the role of a DSMB in reviewing the results of two pre-influenza season vaccine trials aimed at assessing vaccine safety. This may represent a record for the time a DSMB functioned—the study was completed in 1 month and the results were analyzed and reported in 10 days. Perhaps the time period the DSMB advised should have been longer as the authors indicate that continuation of the DSMB might have been helpful in ensuring more complete access to the study data. This report raises the general question of what responsibility a DSMB should take in ensuring that trials they monitor are fully reported.

Five other articles complement these six case reports. Under what circumstances should the investigators be allowed a “peek” at the interim data to determine whether their trial is properly powered? Proschan et al. describe a statistical approach for midcourse sample size modification (increase) based on the treatment effect for which easy computation of p-values and confidence intervals upon completion is possible. They conclude by stating: “there is nothing to lose and power to gain by this procedure.” I am not so sure. I like the methods these authors have developed but am concerned about their indiscriminate use. I hope their well-written paper spurs much-needed debate on this topic.

Chen et al. propose a plan to monitor mortality at interim analyses when the primary endpoint may or not be mortality and cite recent heart failure trial examples. The paper raises the general question of monitoring multiple endpoints in clinical trials, some of which are related to efficacy and some to safety. This is an area that requires more research. Chen et al. have given us a useful start for two outcomes.

The other three articles provide additional insights into DSMB's decision making. Bowen et al., like Hillman and Louis, describe a trial where data from another trial had to be carefully weighed by the DSMB. The procedures implemented for ultimately stopping the CARET study and informing participants about the results will be a useful guide to many investigators. Oden et al. describe the difficult decision making involved in a trial with a borderline result (one that nearly crosses the prespecified boundary), but with slowing recruitment and perhaps a reduced treatment effect with increasing follow-up. The authors conclude that stopping earlier than they did would have been a mistake, but did they stop too soon? Delgado-Herrera and Anbar describe how, as a sponsor, they organized a DSMB. They then go on to describe how the DSMB decided to stop the trial.

Our final piece on DSMBs is a review of a book by Ellenberg, Fleming, and DeMets titled “Data Monitoring Committees in Clinical Trials: A Practical Perspective.” It is a fitting close for this issue as the intent of my call for papers and the other papers in this issue is to expand the literature on practical aspects of data monitoring.

I would like to acknowledge the contributions of Marcus Kjelsberg who oversaw the editorial review of the nine papers submitted in response to the call last year. We look forward to receiving case reports in response to our 2003 call for case reports on follow-up in clinical trials.