|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2005.04.906 on July 18 2005 © 2005 American Society of Clinical Oncology.
Early Viewing of Noninferiority Trials in ProgressDepartment of Health Studies, University of Chicago, Chicago, IL Randomized clinical trials have a relatively brief history, and the implementation of structured interim-monitoring plans is even more recent, with the majority of developments related to group sequential methods taking place beginning in the 1970s and gaining wide adoption in the 1980s.1 These ideas were readily taken up in oncology2,3; indeed, early in its history the Journal of Clinical Oncology featured articles describing what were essentially policy statements regarding the conduct of interim monitoring in cancer clinical trials.4,5 In those articles, principles were laid out for incorporation of appropriate interim-monitoring plans into the trial design, establishment of a data-monitoring committee that is empowered to act with authority on information obtained from interim analyses, and—most relevant to this discussion—limits on access to the interim-analysis results to the data-monitoring committee until such time that results are definitive or other circumstances dictate that the trial should be closed. Such policies have been widely implemented throughout oncology clinical trials, and case studies indicate that these practices seem to have served well.6-8 Over time, however, a more open and perhaps less disciplined approach to interim analyses has crept into practice. The notion that interim analyses should be confidential is sometimes considered strictly in terms of how patients and investigators for the trial in question might be affected by the results; once patients have been enrolled and completed their assigned therapy, sharing results of interim analyses is permissible while the trial progresses toward the requisite accumulated information for definitive analysis. Hence, public conference presentations and even publications described as "interim analyses" occasionally appear. However, it is clear that, as originally framed, the confidentiality provision for interim data encompasses much broader concerns than the trial participants at hand, such as the avoidance of premature wide adoption of therapies that later may prove less than fully efficacious and safe. The position that access to interim-analysis results should be limited has been reinforced over time and in different disease-research settings,9,10 although it is not without its dissenting views.11 The basic stance is that findings should be disseminated for general scientific and clinical consideration only when the trial has satisfied criteria for disclosure, which in turn implies that the data can support clinical decisions. Note that we wish to draw a distinction here between dissemination of "interim-analysis findings," labeled as such, and early (with respect to the originally planned trial duration) disclosure of findings because either the trial has satisfied appropriate criteria to render the results sufficiently reliable to guide treatment decisions or there are other circumstances that justify early disclosure. In short, interim-analysis findings are generally deemed not fit for general consumption. In this issue of JCO, Korn et al12 introduce a new proposal suggesting that, in certain clinical trial settings, data from ongoing trials could be appropriately disclosed. Specifically, in the conduct of noninferiority trials, preliminary data could be released to aid in clinical decision making while not jeopardizing the successful completion of the trial or altering the operating characteristics of its formal monitoring and stopping rules. To define terms clearly, in a noninferiority trial, investigators aim to show that a treatment alternative is materially equivalent to an existing standard treatment. In statistical terms, this means that the alternative is not worse than the standard by more than a specified margin by which physicians participating in the trial (and, hopefully, later the larger community of physicians and patients) have determined to be the boundary of a clinically material difference. What renders this marginal compromise of efficacy worthwhile are improvements in other dimensions such as adverse-effect profile, cost, or convenience for the treatment alternative. For an excellent example, see the recent report in JCO by Jones et al13 of a trial comparing 30- v 20-Gy radiotherapy doses for stage I seminoma. The early-data-release proposal is intended specifically for situations in which the intervention under test against a current standard is already an option available outside clinical trials. This is an increasingly common scenario as technologically innovative treatment alternatives in areas such as surgery and radiotherapy become available and for which trials to demonstrate noninferiority to more traditional interventions may be underway but results are long in coming. Subject to a set of additional criteria that serve to protect the integrity of the trial, preliminary data regarding the relative efficacy of the interventions and additional information pertinent to decision making (such as adverse effects) might be provided from the ongoing trial in this case so that patients and caregivers could make maximally informed choices. The proposal (summarized in their Table 1) carefully outlines circumstances in which early data release might be justified and permitted,12 and some comments on specified points can be made. The first two conditions require that before any data are released, all patients have completed their assigned treatment and are unlikely to modify subsequent treatment, ensuring that the trial will not become compromised operationally. One might also consider the effect of the data release on similar trials that are still accruing or treating patients. In our experience with an efficacy trial of tamoxifen duration that shared some characteristics of trials considered by Korn et al (treatment in use outside of trials and embedded within trials of other agents, with a possibly lengthy interval to completion), early stopping for lack of benefit was additionally complicated by other ongoing high-profile trials, coupled with the findings that went against expectations, prompting questions about the reliability of the early results.14 Significant difficulties, including the loss of the opportunity to complete other trials in progress, can occur when a major trial is stopped with findings disclosed. Although this situation is unavoidable at times,15 in the circumstance in which the data released are not intended to be definitive, disruption of or other negative impact on ongoing trials would be particularly unfortunate. In light of this, consideration of the accrual or treatment status of other trials might also need to be incorporated into the timing of data release. Another condition (six) given by Korn et al deals with circumstances in which one intervention may seem inferior early in follow-up but later yields benefits, and thus the early data may seem particularly worrisome if released. This circumstance arises in surgical intervention trials for certain diseases16 and may be a less common scenario in oncology, but if so, might represent a case in which pressure may be high to release acute risk estimates even in the absence of the long-term benefit data. Thus, although the authors' recommendation that such a trial not be a candidate for early data release is a sound one, vigilance against requests for such information may be required. Finally, the authors mention the challenges that presentation of preliminary data may entail and specify that the information released must clearly communicate the uncertainty in the relative efficacy estimate. Innovative tools to effectively communicate treatment risk and benefit to patients would be particularly useful here. One critical issue regarding this proposal is the importance of seeing noninferiority trials through to planned completion. The key premise of the noninferiority trial is that there is a sufficiently small difference between the two treatment options in question to permit use of the one that provides other benefits such as a more favorable adverse-effect profile. The estimated efficacy difference between treatments becomes more precise as information accumulates, and thus early "stopping" is rarely warranted or desirable in such trials. Therefore, it would be unwise for those conducting a noninferiority trial to declare early victory and fail to collect additional long-term outcome information if they hope to motivate adoption of the treatment alternative in wider practice. In fact, it is ultimately the community of physicians and patients who will decide what constitutes noninferiority when they opt for the treatment based on trade-offs between risks and benefits, and some thoughtful investigators have attempted to directly measure this balance in the target population when designing a noninferiority trial so as to increase the chance that the trial results will be influential.13,17 It is unlikely that negligence in trial completion would be a problem in National Cancer Institute-sponsored or other high-visibility trials, but as the authors indicate, empirical evidence of successful completion and subsequent influence on practice of trials conducted in the manner proposed are needed before this concept is widely adopted. Finally, there is clearly a need for innovations in trial design and conduct of the type proposed here, because situations similar to those described by Korn et al are not uncommon. For example, trials to demonstrate noninferiority of partial breast irradiation to standard radiotherapy for breast cancer will necessarily be lengthy and must stretch the bounds of what might reasonably be called "equivalent" outcomes to be feasible. Inordinately long trials of procedures and devices run the additional risk of becoming scientifically and clinically defunct as the technology continues to evolve. Meanwhile, the techniques continue to move into the mainstream of treatment practice. In light of these challenges, the proposal offered deserves serious consideration. Author's Disclosures of Potential Conflicts of Interest The author indicated no potential conflicts of interest. REFERENCES 1. Jennison C, Turnbull BW: Group Sequential Methods With Applications to Clinical Trials. Boca Raton, FL, Chapman and Hall/CRC, 2000 2. Haybittle JL: Repeated assessment of results in clinical trials of cancer treatment. J Radiol 44:793-797, 1971 3. Peto R, Pike MC, Armitage P, et al: Design and analysis of randomized clinical trials requiring prolonged observation of each patient. I. Introduction and design. Br J Cancer 34:585-612, 1976[Medline]
4. Green SJ, Fleming TR, O'Fallon JR: Policies for study monitoring and interim reporting of results. J Clin Oncol 5:1477-1484, 1987
5. Geller NL: Planned interim analysis and its role in cancer clinical trials. J Clin Oncol 5:1485-1490, 1987 6. Parmar MK, Machin D: Monitoring clinical trials: Experience of, and proposals under consideration by, the Cancer Therapy Committee of the British Medical Research Council. Stat Med 12:497-504, 1993[Medline] 7. George SL: A survey of monitoring practices in cancer clinical trials. Stat Med 12:435-450, 1993[Medline] 8. Crowley J, Green S, Liu PY, et al: Data monitoring committees and early stopping guidelines: The Southwest Oncology Group experience. Stat Med 13:1391-1399, 1994[Medline] 9. Fleming TR, DeMets DL: Monitoring of clinical trials: Issues and recommendations. Control Clin Trials 14:183-197, 1993[CrossRef][Medline] 10. Fleming TR, Ellenberg S, DeMets DL: Monitoring clinical trials: Issues and controversies regarding confidentiality. Stat Med 21:2843-2851, 2002[CrossRef][Medline]
11. Lilford RJ, Braunholtz D, Edwards S, et al: Monitoring clinical trials: Interim data should be publicly available. BMJ 323:441-442, 2001
12. Korn EL, Hunsberger S, Friedlin B, et al: Preliminary data release for randomized clinical trials of noninferiority: A new proposal. J Clin Oncol 23:5831-5836, 2005
13. Jones WG, Fossa SD, Mead GM, et al: Randomized trial of 30 versus 20 Gy in the adjuvant treatment of stage I testicular seminoma: A report on Medical Research Council Trial TE18, European Organisation for the Research and Treatment of Cancer Trial 30942 (ISRCTN18525328). J Clin Oncol 23:1200-1208, 2005 14. Dignam JJ, Bryant J, Wieand HS, et al: Early stopping of a clinical trial when there is evidence of no treatment benefit: Protocol B-14 of the National Surgical Adjuvant Breast and Bowel Project. Control Clin Trials 19:575-588, 1998[CrossRef][Medline]
15. Bryant J, Wolmark N: Letrozole after tamoxifen for breast cancer: What is the price of success? N Engl J Med 349:1855-1857, 2003
16. Howard G, Chambless LE, Kronmal RA: Assessing differences in clinical trials comparing surgical vs nonsurgical therapy: using common (statistical) sense. JAMA 278:1432-1436, 1997
17. Stenning SP, Parmar MK: Designing randomised trials: Both large and small trials are needed. Ann Oncol 13:131-138, 2002 (suppl 4)
Related Article
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2005 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
