Intraperitoneal therapy for ovarian cancer: a treatment ready for prime time.

In this issue of the Journal of Clinical Oncology, three experienced and respected European clinical researchers present their viewpoint that intraperitoneal (IP) chemotherapy remains experimental in the treatment of ovarian cancer. We appreciate that the authors, like many of us, work in an environment that sometimes imposes financial, administrative, or political restrictions on health care delivery. However, we do not think that such restrictions should color our evaluation of the data or the treatment recommendation for an individual patient. We respectfully disagree with their conclusions and present this rebuttal to the issues raised in their commentary. Gore et al criticize the Gynecologic Oncology Group (GOG) trial GOG-172, implying that the study was not reported based on an intent-to-treat analysis and they indicate their suspicion that the exclusion of ineligible patients may favor the IP regimen. Gore et al are in error in this regard. We recognize that the policy for central review of eligibility for clinical trials differs between European and North American cooperative groups. From previous experience, we know that prescreening by participating institutions is not perfect. It is thus the policy of the GOG, and of most North American cooperative groups, to require a central review of eligibility. The 14 patients excluded from GOG-172 were deemed ineligible during central review, 12 based on pathology review. This review, by necessity, occurs after patient registration but includes only material and information collected before registration and was blinded to the arm to which the patient was randomly assigned. The analysis as presented in the original article is in fact slightly more conservative. If the 14 ineligible patients had been included in the survival analysis, the hazard ratio, 95% CI, and P value would have been 0.73, 0.56 to 0.94, and .016, respectively, (compared with 0.75, 0.58 to 0.97, and .03, respectively). Because the intent of the study from its inception was to analyze only eligible patients, the manuscript presented results based on eligible patients only. Gore et al’s next concern focuses on there being only a 15 patient difference in the number of patients alive at the time of the analysis. They are further discouraged by the small difference in the number patients free from disease progression (nine patients). It is important to recognize that these comparisons are confounded with the cumulative patient-time at risk of death. Effective treatments can only delay death, not eliminate it. Patients, who receive treatments that delay death, accumulate more life-time (time at risk of death). Accounting for the difference between treatment groups in the cumulative times at risk, there were actually 33 fewer deaths in the IP treatment group than would have been expected if the IP regimen was only as effective as the intravenous (IV) regimen. Similarly, the adjusted progression-free survival (PFS) comparison indicates that there were 32 fewer patients experiencing either disease progression or death on the IP regimen. Since deaths are reported whether they are disease related or not, the absolute difference between treatment groups in number of patients alive will eventually become zero. Therefore, it is not an appropriate statistic for identifying treatments that delay death. Interestingly, the authors’ critique of GOG-172 resurrects a similar concern that was raised regarding the results from GOG111 10 years ago. They question whether it is possible that a 2.4to 2.9-month shift in median time to progression, could lead to a 12.5to 15.9-month increase in median survival. A review of all GOG randomized front-line ovarian cancer studies indicates that the treatment effects on overall survival are generally similar in size to the progression-free effect when they are assessed on relative hazards scale. Deviations have been observed in those trials in which a significant number of patients who were randomly assigned to the control regimen crossed over to the experimental arm. The same observation has been made using trials from other cooperative groups. These results indicate that the 20% reduction in PFS hazard seen in GOG-172 is consistent with the 25% reduction in the death rate. In Gore et al’s review of GOG 172, the authors state that “there are only two possible explanations for [the differences in overall survival], either (A) patients with relapse after IP therapy live longer because the nature of the treatment has altered the biology of their disease, or (B) patients who relapse after IP therapy are able to receive more effective second-line treatment.” We would argue that a treatment that provides either or both of these beneficial outcomes is desirable. However, we would add two additional potential explanations of the improved survival: (C) patients who relapse later are more likely to be sensitive to second-line treatment, and (D) some patients who would have relapsed after IV therapy are prevented from relapsing by the use of IP therapy. The JOURNAL OF CLINICAL ONCOLOGY COMMENTS AND CONTROVERSIES VOLUME 24 NUMBER 28 OCTOBER 1 2006