Abstract
Clinical trials in medicine often derive their financial support from the manufacturer of a device or pharmaceutical, but one must view with caution trial designs that allow an interested sponsor to terminate data collection.
I recently had the experience, while serving as a member of a human study section doing institutional review of a pre-existing multisite research protocol, of considering the implications of a decision by the pharmaceutical company sponsoring the study to discontinue not only the overall study, but also to discontinue follow-up of the survival results for those who had been previously randomized within the study.
The initial study design reflected a generic research question which arises in many clinical studies, specifically whether a drug, already legally (in this case, the US FDA) approved for one (oncology) indication A, offered benefit in another clinical setting B for which it was not currently FDA approved. The study sponsor subsequently decided to discontinue the study, reportedly because the initial data suggested that the study drug would not prove of benefit in indication B. From the perspective of the pharmaceutical company, presumably this would be characterized as a failed drug trial; from the perspective of an institutional review board, this was perhaps a study that no longer merited additional enrollment or ongoing study drug administration (except to responders, as was indeed planned).
The more troublesome aspect of the decision by the study sponsor was their action in discontinuing follow-up for survival data on those subjects who had already been enrolled but had not been among the responders. This decision by the study sponsor precluded any effort to compare overall survival on the two study arms based on intention to treat, as is commonly done in many oncology trials. The described initial study data suggest it is unlikely that the study drug arm would have demonstrated superior overall survival. On the other hand, the limited data presented to our institutional review board was consistent with the possibility that longer follow-up might have shown the study drug to have produced inferior overall survival for those subjects who had been in the arm randomized to receive the study drug.
Randomized clinical research trials can best be justified when there is uncertainty whether the standard therapy or the experimental therapy yields the better result, i.e. a state of equipoise. The premature termination of follow-up on overall survival may artificially maintain a state of clinical equipoise – a state of equipoise that could have been resolved if it were found that the study drug had an inferior outcome compared to the standard therapy. In this artificially maintained state of clinical equipoise, the study sponsor presumably would maintain the ability to propose a new trial of the same drug for the same indication B, perhaps with a slight modification of the study design.
A broader concern is that we now lack some of the potential data to assess the risks and benefits of use of the study drug in a variety of other indications (C, D, E, …). Many FDA approved drugs for oncologic and other indications are expected to carry significant side effects but are FDA approved for particular indications where it is demonstrated that, on balance, the benefits outweigh the risks associated with the respective drug’s side effects. Such demonstration of net benefit typically rests on improved overall survival among those receiving the drug (setting A). If a study happens to use the particular drug in another setting B where it is not very effective at treating the condition, so that there is little drug benefit, the overall survival of patients treated with the drug will reflect primarily the side effects of the drug. It is precisely data arising from a setting where the drug is least effective which is likely to give the best measure of the potential harm which the drug may inflict. This measure of potential drug harm, when the drug is given for a condition B for which it was not effective, is particularly relevant to evaluating the state of equipoise when it is proposed to administer the drug in studies for yet other indications (C, D, E, …) for which it may also prove to be as ineffective as it was for condition B.
Although the institutional review board at our site certainly would have had the negative ability and authority to prevent the study from the beginning, we lacked the effective means to positively require the study sponsor to continue to monitor for survival. Even if we had chosen to demand (which we did not) that such survival follow-up occur at our individual site, the data from a single site would be unlikely to have a sufficient sample size to generate statistically meaningful survival data. One presumes that institutional review boards at other sites would have been similarly limited in approaching this same issue.
In conclusion, there may often be sound ethical reasons for early discontinuation of a multisite clinical trial (Pocock, 2012). An individual site’s IRB will generally want to defer to an objective overall data safety monitoring board on the decision to stop a trial, because a data safety monitoring board can look at the evolving data across all trial sites, and recognize quickly situations where further study treatment of new or existing subjects is no longer warranted. There is generally no risk to subjects, however, from continuing to follow their survival outcomes after study discontinuation. A decision to terminate follow-up for survival is accordingly much more problematic, and especially so if the decision is made by an interested pharmaceutical sponsor (see Rodwin, 2012) rather than a neutral data safety monitoring board.