Gemcitabine and Cisplatin for Advanced, Metastatic Bladder Cancer

Article Tools SPECIAL DEPARTMENTS Article Tools OPTIONS & TOOLS Export Citation Track Citation Add To Favorites Rights & Permissions COMPANION ARTICLES No companion articles ARTICLE CITATION DOI: 10.1200/JCO.2001.19.4.1229 Journal of Clinical Oncology - published online before print September 21, 2016 PMID: 11181690 Gemcitabine and Cisplatin for Advanced, Metastatic Bladder Cancer Martin H. CohenxMartin H. CohenSearch for articles by this author , Mark RothmannxMark RothmannSearch for articles by this author Hans von der MaasexHans von der MaaseSearch for articles by this author , Anna Marie HaydenxAnna Marie HaydenSearch for articles by this author Show More United States Food and Drug Administration, Rockville, MDAarhus University Hospital, Aarhus, DenmarkEli Lilly and Company, Indianapolis, IN https://doi.org/10.1200/JCO.2001.19.4.1229 First Page Full Text PDF Figures and Tables © 2001 by American Society of Clinical OncologyjcoJ Clin OncolJournal of Clinical OncologyJCO0732-183X1527-7755American Society of Clinical OncologyResponse15022001In Reply:Drs Cohen and Rothmann raise an important and relevant issue. However, analyzing noninferiority posthoc is difficult, and the methodologies are not clearly defined. At the very least, this requires a wealth of data on the control arm as compared with a placebo (or clearly defined standard). Our conclusion needs to be considered mainly on the results of our study.1 The hazard ratio for survival was 1.04 (95% confidence interval [CI], 0.82 to 1.32). The adjusted survival hazard ratio, taking significant prognostic factors into account, was 0.95 (95% CI, 0.74 to 1.22).The analyses performed by Cohen and Rothmann are based on two considerations that we do not agree with. The study comparing cisplatin, cyclophosphamide, and doxorubicin (CISCA) with cisplatin2 included only 109 patients and therefore had little power to confirm a statistically significant difference in survival in favor of the combination (29 weeks v 21 weeks). However, this does not mean that it can be concluded that CISCA does not provide a survival advantage over cisplatin. Therefore, we believe that any noninferiority analysis should not assume that CISCA and cisplatin are similar and can be analyzed together, as Cohen and Rothmann have done. However, the studies by Troner et al2 and Logothetis et al3 can be used together to better understand the effect of methotrexate, vinblastine, doxorubicin, and cisplatin (MVAC) over cisplatin, since Troner et al randomized patients to CISCA or cisplatin and Logothetis et al randomized patients to MVAC or CISCA.The second consideration of the analyses conducted by Cohen and Rothmann is the use of the unadjusted hazard ratio from our study in comparisons with the study by Loehrer et al.4 The Loehrer et al study reported an adjusted hazard ratio for overall survival; therefore, the adjusted hazard ratio from our study should be used to conduct such a noninferiority analysis with the Loehrer et al study. However, we have used both the unadjusted and the adjusted values to be consistent with their methodology.To further understand the benefit of the gemcitabine and cisplatin (GC) regimen, we conducted various meta-analyses of the randomized data on GC, MVAC, and cisplatin. While there does not exist a randomized trial comparing GC directly with cisplatin, an indirect comparison is possible by combining results from studies involving GC, cisplatin, and MVAC. We believe that there is strong evidence that gemcitabine provides an efficacy benefit when added to single-agent cisplatin and that there is a very high probability that GC preserves more the 50% of the survival effect of MVAC over cisplatin. Below, we present in detail the results of the meta-analyses we have conducted.Meta-Analysis of Survival Including the Loehrer Study and Our Study. The effect of the MVAC regimen on overall survival, relative to single-agent cisplatin, is best judged from the results of the large randomized trial that directly compared the two regimens. Loehrer et al4 reported the following results for overall survival: for the cisplatin regimen, 115 deaths and 8.2-month median survival; for the MVAC regimen, 106 deaths and 12.5-month median survival.An indirect comparison of GC to single-agent cisplatin is possible by pooling the results of the Loehrer et al4 study with those from our study.1 Specifically, if we estimate the hazard ratio of cisplatin to MVAC from the Loehrer et al study and divide that by the estimated hazard ratio of GC to MVAC, the result is an estimate of the hazard ratio of cisplatin to GC. Equivalently, we can work with the natural logarithms of these hazard ratios so as to obtain a CI for the hazard ratio of cisplatin to GC. From this CI, we can judge the strength of the evidence from the two studies that GC has superior survival to cisplatin.An assumption of constant survival hazards for the Loehrer et al study is approximate at best, but it is necessary due to the limited information available from the article.4 The ratio of the medians under this assumption provides an estimate of the hazard ratio of cisplatin to MVAC (R1 = 1.524). The natural logarithm of this hazard ratio is L1 = 0.421, and the variance of the log hazard ratio can be approximated by the following formula:equation For our study, the corresponding statistics were directly estimated from the data using the Cox proportional hazards model (with therapy as the only covariate and without assuming constant hazards), as follows:equation Assumptions for pooling the results of these two studies include the following: (1) the two patient groups represent stochastically independent random samples from the same patient population, and (2) data from the two studies are equally weighted (ie, each observed death from either study contributes the same amount of information). A review of the inclusion/exclusion criteria for the two studies indicates that these assumptions are reasonable.Therefore, the estimated log hazard ratio of cisplatin to GC is:equation So the estimate of the hazard ratio of cisplatin to GC is 1.47, with a 95% CI of 1.03 to 2.09. Since the value "1" is excluded from this interval, we conclude with 95% confidence that gemcitabine provides an added survival benefit to single-agent cisplatin. Furthermore, we are estimating cisplatin to have a 47% greater risk of death than GC.One extension of this two-study analysis is to incorporate Bayesian distributions for the log hazard ratios of cisplatin to MVAC and cisplatin to GC. Using the methods described by Simon,5 we conducted this Bayesian analysis by assuming a noninformative prior distribution for the log hazard ratio of cisplatin to GC (ie, a normal distribution with infinite variance). The Loehrer data were used to construct the prior distribution for the log hazard ratio of cisplatin to MVAC. This Bayesian analysis resulted in the exact same point estimates as in our straightforward analysis above, but it allows for further interpretation. The point estimates can be interpreted as the mean and variance of a normal probability distribution. The results show that there is a 98% probability that GC has a superior survival compared with cisplatin alone. Furthermore, this Bayesian analysis shows that there is an 89% probability that GC preserves at least 50% of the survival effect of MVAC.Both the straightforward analysis and the Bayesian analysis above are conservative, in the sense that only Loehrer's and our results are incorporated, and no covariate adjustments are used. In fact, the results are stronger in favor of GC if covariate adjustment is used. The estimate of the hazard ratio of cisplatin to GC is 1.60, with a 95% CI of 1.07 to 2.30. Furthermore, the Bayesian analysis shows that there is a 97% probability that GC preserves at least 50% of the survival effect of MVAC.Meta-Analysis of Survival Including the Troner and Logothetis Studies. Additional randomized data on cisplatin and MVAC is available from two studies, the above-mentioned study from Troner et al2 (CISCA v cisplatin) and that from Logothetis et al3 (MVAC v CISCA). By combining the results from these two studies, one comparing MVAC to the CISCA regimen and the other comparing CISCA to cisplatin, it is possible to strengthen the estimate of the effect of MVAC over cisplatin. In fact, when the Troner and Logothetis results are pooled (analogously to the methods used above), we obtain estimates of the hazard ratios of cisplatin to MVAC for survival that are larger than that obtained from the Loehrer study alone (log hazard ratio = 0.422, variance = 0.018). So by pooling all three studies—the Loehrer, Troner, and Logothetis results—we obtain a larger estimate of the survival benefits for MVAC relative to cisplatin (hazard ratio, 1.58; 95% CI, 1.24 to 2.00). This in turn provides larger estimates of the survival benefits for GC over cisplatin (hazard ratio, 1.52; 95% CI, 1.08 to 2.12). In addition, the Bayesian analysis again shows that there is an 87% probability that GC preserves at least 50% of the survival effect of MVAC. So the analyses we have presented above in detail are the most conservative presentation of what is known about cisplatin, MVAC, and GC.It should be noted that in this study, the median survival on the MVAC arm of 14.8 months is the highest reported in studies randomizing patients to MVAC. Other studies have reported median survival times of 62.6 weeks, 12.5 months, and 14.5 months.3,4,6Although the primary aim of cancer treatment is prolongation of life, the progression-free interval is also an important end point and should lend to the understanding of the impact of therapy on the patient's life. Comparisons of time to progressive disease in a meta-analysis may be used for the understanding of the impact of the chemotherapy regimen in bladder cancer. Therefore, the meta-analysis and the Bayesian analysis were also applied to the results on time to progressive disease. Loehrer et al4 reported the following results for time to progressive disease: for cisplatin, 113 progressions and 2.4-month time to progressive disease; for MVAC, 108 progressions and 6.6-month time to progressive disease.By constructing point estimates for the log hazard ratio and its variance (analogously to our approach above for survival), we obtained the following results: the hazard ratio of cisplatin to GC for time to progressive disease is estimated to be 2.62, with a 95% CI of 1.86 to 3.68. In other words, single-agent cisplatin has roughly between double and triple the risk of progression compared with GC. Furthermore, under the Bayesian analysis of these point estimates, there is an almost 100% probability that GC is superior to cisplatin in time to progressive disease. Also, there is a 91% probability that GC preserves at least 80% of the time-to-progressive-disease effect of MVAC. We have not performed an analysis by pooling all three studies, since the data on time to progressive disease were not available for the other studies.The results of these meta-analyses demonstrate that GC is superior to cisplatin in both survival and time to progression and that GC preserves most of the survival and time-to-progressive-disease effects of MVAC over cisplatin. Given the fact that a true noninferiority study would have needed impractical numbers of patients (eg, 1,900 patients for an upper limit of the hazard ratio of 1.15; this is assuming a two-sided test with alpha = 0.2, beta = 0.05, and null hypothesis of hazard ratio = 1.15 v alternative hazard ratio = 1.0),7 and that the risk-to-benefit ratio favors GC over MVAC,1 we still agree with our conclusion that GC should replace MVAC as the standard of care for patients with locally advanced and metastatic bladder cancer. This conclusion has also been accepted by many major cooperative groups, who are now conducting the intergroup (Europe and North America) study comparing GC with the gemcitabine, cisplatin, and paclitaxel combination, as well as by several regulatory agencies worldwide. The views expressed are the result of independent work and do not necessarily represent the views and findings of the United States Food and Drug Administration. 1. Von der Maase H, Hansen SW, Roberts JT, et al: Gemcitabine and cisplatin versus methotrexate, vinblastine, doxorubicinadriamycin, and cisplatin in advanced or metastatic bladder cancer: Results of a large randomized multinational, multicenter, phase III study. J Clin Oncol 18:: 3068,2000-3077, Link, Google Scholar2. Troner M, Birch R, Omura GA, et al: Phase III comparison of cisplatin alone versus cisplatin, doxorubicin and cyclophosphamide in the treatment of bladder (urotherlial) cancer: A Southeastern Cancer Study Group trial. J Urol 137:: 660,1987-662, Crossref, Medline, Google Scholar3. Logothetis CJ, Dexeus FH, Finn L, et al: A prospective randomized trial comparing MVAC and CISCA chemotherapy for patients with metastatic urothelial tumors. J Clin Oncol 8:: 1050,1990-1055, Link, Google Scholar4. Loehrer PJ, Einhorn LH, Elson PJ, et al: A randomized comparison of cisplatin alone or in combination with methotrexate, vinblastine, and doxorubicin in patients with metastatic urothelial carcinoma: A cooperative group study. J Clin Oncol 10:: 1066,1992-1073, Link, Google Scholar5. Simon R: Bayesian design and analysis of active-control clinical trials. Biometrics 55:: 484,1999-487, Crossref, Medline, Google Scholar6. Sternberg CN, de Mulder PH, Schornagel J, et al: Randomized phase III trial in advanced urothelial tract tumors of high dose intensity M-VAC chemotherapy and G-CSF versus classic M-VAC. Proc Am Soc Clin Oncol 19:: 329a,,2000 (abstr 1292) Google Scholar7. Schoenfield D: The asymptotic properties of comparative tests for comparing survival distributions. Biometrika 68:: 316,1981-319, Crossref, Google Scholar