|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 23, No 31 (November 1), 2005: pp. 8124-8125 © 2005 American Society of Clinical Oncology. DOI: 10.1200/JCO.2005.03.2458
Randomized Phase II Trials and Prostate-Specific Antigen Endpoints in Prostate Cancer: Much Ado About Nothing?University of Chicago, Department of Medicine, Section Hematology Oncology, Chicago, IL To the Editor: We read with interest the randomized phase II trial of two schedules of docetaxel and estramustine versus mitoxantrone, as well as the accompanying editorial in the May 20, 2005, issue of the Journal of Clinical Oncology.1,2 These two articles illustrate some of the challenges and controversies in the development of novel therapies for patients with hormone refractory prostate cancer. We are concerned, however, that Roth's criticism of the Oudard trial threatens to indict both the randomized phase II trial design and standardized prostate-specific antigen (PSA) response criteria beyond the specific concerns raised in the article at hand. In regard to PSA response criteria, Dr Roth correctly points out that the arbitrary criteria of a 50% decline in the PSA is not an ideal marker, especially for phase III trials.3 Although this is due, in part, to the fact that many agents, especially hormonal agents, can affect PSA secretion, as noted by Dr Roth, a 50% decline does not correlate perfectly with clinical benefit even for cytotoxic drugs in hormone refractory disease. This same criticism can, however, be applied to the most common response marker used in solid tumor oncology, namely, a 30% reduction in unidimensional radiologic measurements. PSA kinetic measurements may be a superior marker, but are currently limited by lack of standardization for the frequency at which the assay should be performed and controversy as to how the kinetic parameter should be calculated. Rather than focusing on the inability of an arbitrary PSA decline to completely capture all of the associated clinical benefits from a treatment, it is perhaps more useful to consider it a response biomarker that has pragmatic clinical and clinical trial utility. Specifically, the PSA working group consensus criterion of a PSA response (ie, a 50% decline in PSA maintained for at least 4 weeks)4 does correlate with survival with multiple cytotoxic therapies. Furthermore, there are no examples of a cytotoxic therapy that fails to improve the PSA response rate, but leads to an improvement in survival. As was correctly pointed out by Dr Roth in his editorial, an effect on the PSA response rate will not necessarily translate into an improvement in survival. Therefore a 50% decrease in the PSA can be considered a necessary, but not sufficient, criterion for observing future clinical benefit with the caveat that this does not necessarily apply to agents interacting with the androgen receptor pathway or to cytostatic agents. The major question then becomes whether an observed PSA response rate is significant or an improvement over the expected. Oncology phase II trials have typically been conducted as uncontrolled single-arm trials designed to rule out an uninteresting response rate and recommend further testing if the response rate is interesting. It is typically forgotten that definitions of uninteresting and interesting are dependent on a historically implied control of expected outcome with either no therapy or with standard therapy. The known heterogeneity in patient responses to treatment makes the latter especially problematic. For a combination trial it thus makes more sense to randomize the investigational therapy against the standard. It has been argued that performing a randomized phase II trial comparing two treatments in a small number of patients is simply an inadequate phase III trial. However, it is illogical to then conclude that a single arm phase II trial with a historical control is thus a superior design. In keeping with phase II designs in other disease modalities, we would alternatively argue that the difference between a phase II and phase III trial is not the absence or presence of randomization, but is instead a matter of the trial's end points and objectives. Perhaps more accurate would be to define a phase II trial as a learning trial with flexibility in analyses to explore various hypotheses whereas a phase III trial is a confirming trial with rigorous end points and analyses designed to definitively answer a question.5
Specifically, a phase II trial should seek to determine whether a new intervention has sufficient activity to justify further study, whereas a phase III trial will determine whether there is a true clinical benefit. It is correct that a randomized phase II trial that can accurately rule out both false-positive and false-negative results using a PSA response end point is almost as large as a well-performed phase III trial with a survival end point. Smaller trials will not have sufficient power to adequately rule out a false-negative result. In an era when a limited number of agents were available for cancer treatment, this was a significant issue. However, in the current era, with the availability of hundreds of agents (many of which have similar mechanisms of action), and more concepts than can be possibly tested in the phase III setting, this becomes a less important issue. In fact, we would strongly argue that an appropriate clinical trial design for hormone refractory prostate cancer patients in the phase II setting is to randomly assign a relatively small group of patients to standard versus experimental therapy and use the 50% decline in PSA as the primary end point. For example, using a Fisher's exact test, to detect a difference of a 50% v 80% PSA response rate with a 2-sided In this context, the article by Oudard et al1 can be criticized not so much for the use of a randomized phase II design or for the PSA end point, but for the choice of performing another phase II trial rather than a phase III trial. One could argue that at the time the study was initiated there were sufficient uncontrolled phase II data already available with the experimental docetaxel based therapy such that another phase II trial, even if more rigorous than previously conducted trials, was highly unlikely to add additional information. The resources would have thus been better utilized to simply perform the phase III trial. In summary, with an increasing number of agents available and an ever greater segmentation of the cancer patient population into molecularly defined subgroups, it will become critical to conduct phase II clinical trials that will determine whether it is useful to perform a resource intensive large phase III study that is likely to have a major impact on clinical care, rather than simply identify agents with marginal benefit whose clinical utility, clinical cost benefit, or risk benefit will be questioned. It is our contention that randomized phase II designs comparing an experimental therapy to a control therapy with biomarker end points that are necessary, but not sufficient, for clinical benefit will go a long way towards meeting this goal. Authors' Disclosures of Potential Conflicts of Interest The authors indicated no potential conflicts of interest. REFERENCES
1. Oudard S, Banu E, Beuzeboc P, et al: Multicenter randomized phase II study of two schedules of docetaxel, estramustine, and prednisone versus mitoxantrone plus prednisone in patients with metastatic hormone-refractory prostate cancer. J Clin Oncol 23:3343-3351, 2005
2. Roth BJ: Prostate cancer chemotherapy: Emerging from the shadows. J Clin Oncol 23:3302-3303, 2005 3. Prentice RL: Surrogate endpoints in clinical trials: Definition and operational criteria. Stat Med 8:431-440, 1989[Medline]
4. Bubley GJ, Carducci M, Dahut W, et al: Eligibility and response guidelines for phase II clinical trials in androgen-independent prostate cancer: Recommendations from the Prostate-Specific Antigen Working Group. J Clin Oncol 17:3461-3467, 1999 5. Sheiner LB: Learning versus confirming in clinical drug development. Clin Pharmacol Ther 61:275-291, 1997[CrossRef][Medline]
Related Reply
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2005 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
of .05 and a power of 0.8 requires 45 patients in each group. If such a large improvement in the PSA response rate is not observed, further investigation of the experimental therapy should be dropped with the understanding that a small but potentially clinical interesting result would be missed. If such an effect is observed, then, as pointed out by Dr Roth in his editorial, the presumed superiority of the experimental therapy will need to be confirmed in a larger phase III setting with a clinically meaningful end point. 
