Early stopping for benefit in National Cancer Institute-sponsored randomized Phase III trials: the system is working.

  • Daniel J. Sargent
  • Published 2009 in
    Journal of clinical oncology : official journal…


The act of consent for a patient to enter a randomized, clinical trial requires a fundamental pact between physician and patient. For the patient, the choice to allow chance to determine his treatment rests on the assumption that his physician and the medical community at large believe that there is clinical equipoise between the treatments under investigation. This equipoise allows ethical randomization. In addition to the need for equipoise at the time of trial entry, a further expectation for patient participation in clinical trials requires that if any relevant information becomes available during the course of the trial, this information will be made available to trial participants. This need for constant reinforcement of the ethics of randomization raises the issue of when, and through what mechanisms, data from ongoing trials should be examined to determine if the trial’s hypothesis has been resolved and, therefore, that the result is known and should be communicated to the scientific and patient communities. Multiple valid reasons exist to consider premature termination of a clinical trial, several of which cannot be argued against (excessive toxicity, changes in standard of care, poor accrual and, somewhat more controversially, stopping for futility because the trial is highly unlikely to ever attain a statistically significant result). The reason for early stopping within an ongoing trial that generates the greatest controversy, however, is terminating a trial early due to the apparent efficacy of the experimental arm versus the control. Recently, considerable attention has been focused on issues surrounding the early release of efficacy data from ongoing clinical trials. Clearly, the premature termination of a clinical trial can have highly problematic consequences. Conclusions based on immature data may change with further follow-up or provide inadequate evidence to convince the broader scientific community. In addition, data from the initial patients enrolled on a trial may not be representative of data that would be obtained from the remainder of the trial, as the early patients (or early enrolling sites) may not be representative of the general population of patients or sites. Release of early data may also prohibit the collection of meaningful data on long-term benefits or adverse events from therapy if the data release results in a change in care for patients currently on the clinical trial (for example, with cross-over treatment). Finally, the estimation of the treatment effect has been shown to be biased away from the null hypothesis (and thus be overly optimistic) if the clinical trial is stopped early, even if stopped at a planned interim analysis, although recent work in oncology suggests that if the trial is stopped after at least 50% of the planned events have been observed, this bias is modest. Clearly, these considerations result in a complex set of factors that must be considered as clinical trials are ongoing, in particular if the interim data from the trial is in some sense promising. The need for regular, careful, and deliberate monitoring of data from ongoing clinical trials has led to the now widely established model of the independent data monitoring committee (IDMC; also often known as a Data Safety Monitoring Board); the need to confine data access in ongoing trials to such a committee has been clearly articulated. The IDMC is generally comprised of three to eight individuals with expertise in the field of the disease under study, who are not affiliated in any way with the ongoing trial that the IDMC is charged to monitor. These IDMCs typically consist of clinicians, statisticians, and often patient representatives or ethicists. The prominence (and sophistication) of IDMCs has become such that a book focused exclusively on this topic has recently been published. The US National Cancer Institute (NCI) has required IDMC monitoring of phase III clinical trials conducted through its cooperative group program since at least 1996. In this issue of Journal of Clinical Oncology, Korn et al present a thorough review of the outcomes associated with clinical trials stopped or reported early that have been sponsored by the NCI through the cooperative group program. Based on a comprehensive review, Korn et al identified 27 phase III clinical trials from 1990 to 2005 in which either the accrual was stopped early or outcome data were released before attaining the protocol-specified number of events. As expected, the trials had a wide variety of primary end points; in 70% of the trials, accrual was complete at the time of closing, which has been recommended by some as an important criterion before trial reporting should be considered. In only one case were data released before at least 25% of the protocol specified number required for the primary analysis had been observed, consistent with recommendations that trial termination based on very early and incomplete data should be avoided. The most important finding by Korn et al was that in 89% of trials (24 of 27) stopped or reported early, the conclusions based on the final analysis were consistent with those made at the time of the early trial reporting. It should be noted that in four of these 24 trials, the estimated treatment effect at the final analysis was more moderate than that observed at the trial stopping point, providing some data to support the concept that the observed treatment effect when a trial is stopped for success at an interim time point may be JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 27 NUMBER 10 APRIL 1 2009

DOI: 10.1200/JCO.2008.20.8611


Citations per Year

51 Citations

Semantic Scholar estimates that this publication has 51 citations based on the available data.

See our FAQ for additional information.

Cite this paper

@article{Sargent2009EarlySF, title={Early stopping for benefit in National Cancer Institute-sponsored randomized Phase III trials: the system is working.}, author={Daniel J. Sargent}, journal={Journal of clinical oncology : official journal of the American Society of Clinical Oncology}, year={2009}, volume={27 10}, pages={1543-4} }