Use of progression-free survival (PFS) as a clinical trial endpoint in first-line treatment of patients with metastatic urothelial carcinoma (UC) is attractive, but would be enhanced by establishing a correlation between PFS and overall survival (OS).
Data was pooled from 7 phase 2 and 3 trials evaluating cisplatin-based chemotherapy in metastatic UC. An independent cohort of patients enrolled on a phase 3 trial was used for external validation. Landmark analyses for progression at 6 and 9 months after treatment initiation were performed to minimize lead-time bias. A proportional hazards model was used to assess the utility of PFS for predicting OS.
A total of 364 patients were included in the initial cohort. The median PFS was 8.21 months (95% confidence interval = 7.43, 8.39) and the median OS was 13.50 months (95% confidence interval = 11.80, 15.67). In the landmark analysis, the median OS for patients who progressed at 6 months was 3.87 months compared with 15.06 months for those patients who did not progress (P < .0001) and the median OS for patients who progressed at 9 months was 5.65 months compared with 21.39 months for those patients who did not progress (P < .0001). A Fleischer model demonstrated a statistically significant dependent correlation between PFS and OS. The findings were externally validated in an independent cohort.
Metastatic urothelial cancer (UC) is a chemosensitive disease. With contemporary cisplatin-based combination chemotherapy, approximately 50% to 60% of patients will achieve an objective response to treatment and approximately 10% to 20% will achieve a complete response.[1, 2] Despite these relatively high response proportions in the context of other solid tumors treated with cytotoxic agents, response durations in metastatic UC are generally short-lived, and median overall survival is only approximately 14 months. Recent attempts to improve contemporary cisplatin-based regimens have proven unsuccessful, with no advances in the efficacy of therapy in the past 30 years.
The relatively high response rates achieved with first-line cisplatin-based chemotherapy in metastatic UC has created a challenge with regard to appropriate endpoints for phase 2 trials exploring novel therapeutic regimens. A high barrier must be overcome to detect a signal of enhanced antitumor activity (as measured by objective response rate) compared with historical controls. At the same time, tumor response rate has not convincingly proven useful in selecting phase 2 regimens worthy of moving forward to definitive randomized trials in this clinical disease state, likely contributing to the lack of progress in drug development.
Alternatively, time-to-event endpoints, particularly progression-free survival (PFS) endpoints, may be preferable to screen the activity of novel first-line treatment regimens in metastatic UC. PFS offers the potential to encompass an aspect of “disease control” not captured by focusing on tumor shrinkage alone. Compared with overall survival (OS) endpoints, PFS endpoints generally occur months to years earlier, particularly when evaluated at a fixed time point, shortening timelines for clinical trial completion, and also minimizing the risk of possible benefits being diluted as effective salvage regimens are introduced. There are multiple barriers to drug development in advanced UC, including inadequate funding and repeated difficulty with trial accrual and/or completion, and one can only speculate whether PFS-based endpoints could have identified active regimens that have since been abandoned. However, such endpoints may still facilitate more efficient drug development moving forward. In this context, understanding the relationship between PFS and OS, and establishing benchmarks of PFS at fixed time points, is critical to inform future drug evaluation and clinical trial design. In this study, we explored these issues in a pooled analysis of patients enrolled on phase 2 and 3 trials evaluating first-line, cisplatin-based, combination chemotherapy in metastatic UC.
MATERIALS AND METHODS
The initial cohort included 399 patients with unresectable and/or metastatic urothelial carcinoma enrolled on 7 phase 2 and 3 trials exploring first-line, cisplatin-based, combination chemotherapy from 1998 to 2011. A cohort of 186 patients enrolled on a phase 3 study comparing MVAC (methotrexate, vinblastine, doxorubicin, and cisplatin) with docetaxel plus cisplatin in patients with metastatic UC (patients enrolled from 1997 to 2002) was used for external validation. The details of each trial are provided in Table 1 and have been presented or published previously. Each study was approved by the institutional review board at the participating institutions and informed consent was obtained before treatment.
Table 1. Phase 2 and 3 Trials Included in the Current Analysis
Abbreviations: 5FU, 5-fluorouracil; cT4b, unresectable; dd, dose-dense; M+, metastatic; MVAC, methotrexate, vinblastine, doxorubicin, cisplatin; NCI CTC, National Cancer Institute Common Toxicity Criteria; RECIST, Response Evaluation in Solid Tumors; WHO, World Health Organization.
Each protocol required a histologic or cytologic diagnosis of UC. The pretreatment evaluations were similar among the protocols and included a complete history and physical examination and laboratory testing including a complete blood cell and platelet count, renal function, and hepatic function. Imaging studies were performed at baseline and every 6 to 12 weeks on treatment, depending on the study. Response assessments were performed by using either World Health Organization (WHO) criteria or Response Evaluation Criteria in Solid Tumors (RECIST), and toxicity assessments were performed by using either WHO criteria or the National Cancer Institute Common Toxicity Criteria.
Individual patient data were collected from the principal investigators of each trial. The primary endpoint for the current analysis was OS, defined as the time from treatment initiation to the date of death from any cause, or censored at the last follow-up. PFS was defined as the time from treatment initiation until the time of disease progression or death or was censored at the last follow-up.
Landmark analysis of PFS at 6 months and 9 months after treatment initiation was performed to minimize lead-time bias. These time points were selected on the basis of the published historical median PFS of patients with metastatic UC treated with first-line cisplatin-based chemotherapy. Patients who died before 6 or 9 months were excluded in the 6-month landmark analysis or the 9-month landmark analysis, respectively, to provide the most conservative estimate of effect.
The method of Kaplan and Meier was used to estimate the OS of patients stratified by disease progression at 6 or 9 months. A proportional hazards model was used to assess the significance of progression at 6 or 9 months in predicting OS when adjusted for prognostic factors.[10, 11] The proportionality assumption was met by graphically assessing plots of log (−log[survival]) versus log of survival time. The case deletion method was used to handle missing values in all explanatory variables.
The correlation between PFS and OS was estimated using the statistical model for dependence between PFS and OS developed by Fleischer et al. The Fleischer approach models the maximal independence between OS and PFS by modeling the time to progression and OS as independent events, and PFS is the minimum of time to progression or OS. This maximal independence model, with the hazard rate for time to progression (λ1) independent of OS time, is further generalized so that the hazard rate for OS after progression (λ3) is a different hazard rate from the one experienced before progression (λ2). The Fleischer correlation is calculated as
The standard errors of the correlation statistics were estimated using the bootstrap method with 300 replications. The bootstrap confidence intervals were computed based on the percentiles of the bootstrap distribution of the statistic. For the landmark analysis, 300 bootstrap samples were generated similarly at each landmark time point.
Of the 399 patients in the initial cohort, 35 patients were excluded due to either missing date of last follow-up (n = 11), date of treatment initiation (n = 4), or progression status (n = 20). Therefore, the study population consisted of 364 patients. At the time of the analysis, 292 patients had progressed and 255 had died. Of the 186 patients in the validation cohort, 5 patients were excluded due to either missing progression status (n = 4) or date of progression (n = 1). Therefore, the validation study population consisted of 181 patients. At the time of the analysis, 161 patients had progressed and 170 had died. The baseline characteristics of patients in both cohorts are detailed in Table 2.
In the landmark analysis of the initial cohort, the median OS for patients who progressed at 6 months, compared with those patients who did not, was 3.87 months versus 15.06 months, respectively. OS was significantly better for patients who were progression-free at 6 months (log-rank test, P < .0001; Fig. 1A). For those who progressed at 9 months, compared with those patients who did not, the median OS was 5.65 months versus 21.39 months, respectively. OS was significantly better for patients who were progression-free at 9 months (log-rank test, P < .0001; Fig. 1B). For patients who progressed at 6 and 9 months, compared with those who did not progress, the hazard ratios (HRs) for death adjusted for known baseline adverse prognostic factors (presence of visceral metastases and performance status) were 2.49 (95% confidence interval [CI] = 1.55, 3.89) and 2.84 (95% CI = 1.81, 4.24), respectively (Table 3). Bootstrap analyses performed at 6 months and 9 months produced very similar HRs, confirming the validity of the results. In addition, the findings were externally validated in an independent cohort (Table 3).
Table 3. Results of Landmark and Bootstrap Analyses for OS by Progression Status at 6 and 9 Months After Initiation of Therapy
For patients in the initial cohort (n = 364), the median PFS was 8.21 months (95% CI = 7.43, 8.39) and the median OS was 13.50 months (95% CI = 11.80, 15.67). The median postprogression survival for patients who progressed prior to death was 5.09 months (95% CI = 4.27, 5.95). By using the Fleischer model, the estimated correlation between PFS and OS was 0.86 (bootstrap standard error of 0.001, 95% CI = 0.82, 0.90). This means that approximately 74% of the variability in OS can be explained by PFS (bootstrap standard error of 0.002, 95% CI = 0.67, 0.81).
There have been no improvements in the efficacy of first-line treatment for patients with metastatic UC since the introduction of cisplatin-based combination chemotherapy regimens in the 1980s.[2, 3, 13] The lack of progress in this disease is almost certainly multifactorial, but may be linked to disease biology, inadequate funding for research, lack of interest from investigators and industry, and poor accrual to clinical trials. Phase 2 efforts over the past few decades have focused on small trials, with response rate as a primary endpoint, which have been extremely difficult to interpret with regard to a “go/no-go” decision, and have not clearly moved the field forward.
The relatively high response rates, yet inadequate response durations, achieved with cisplatin-based chemotherapy for metastatic UC suggest that time-to-event endpoints may be particularly well suited for studies in this clinical disease state. The increasing integration of “targeted” agents with potential cytostatic effects, into conventional cytotoxic regimens, further underscores this point. Use of PFS at a fixed time point (eg, 6-month PFS) as an intermediate endpoint is ideally informed by “benchmarking” of PFS from a large cohort of patients treated with cisplatin-based combination chemotherapy regimens and by understanding the relationship between PFS and OS. The current analysis reveals a median PFS of 7.75 months (95% CI = 7.06, 8.18) in a diverse population of patients with metastatic UC treated with cisplatin-based combination chemotherapy. In the landmark analyses at the 6- and 9-month time points, patients who progressed had a 2.5- to 2.8-fold increased risk of dying compared with patients at the same time point who remained progression-free, even when controlling for known prognostic factors. A significant positive relation between PFS and OS was observed as demonstrated by the Fleischer model (0.86).
These findings raise the important consideration of whether 6- or 9-month PFS should be considered appropriate endpoints for phase 2 trials in chemotherapy-naive patients with metastatic UC, and if so, in what context? The use of PFS as an endpoint for phase 2 trials has been proposed in multiple other malignancies.[14-16] However, there are clear limitations to the use of PFS in phase 2 trials, particularly with regards to the generally problematic comparison with “historical controls.”
Two general strategies have been proposed to minimize these limitations. First, large data sets demonstrating that PFS has been highly reproducible in multiple prior studies might provide sufficient justification for use of PFS as an endpoint, even in single-arm phase 2 studies. In the current study, analysis of the percentage of patients progression-free at 6 and 9 months across the 8 clinical trials (spanning approximately 13 years) reveals relatively similar findings, with overlapping confidence intervals, though ample variation is observed to suggest that single-arm phase 2 trials could still be prone to misinterpretation (Table 4). Importantly, concerns regarding comparisons to historical controls are not unique to PFS endpoints and similarly exist for single-arm trials in this clinical disease state that have employed response rate as a primary endpoint. Second, randomized phase 2 trials can be considered to mitigate the risk of inadvertent selection bias. Such trial designs, in an effort to build on cisplatin-based first-line treatment regimens are attractive, but clearly require greater patient and financial resources. Nonetheless, these resources are likely well invested if such designs minimize the risk of inappropriately discarding an effective regimen or prevent initiation of a large, costly phase 3 program for an ineffective regimen.
Table 4. Proportion of Patients Progression-Free at 6- and 9-Month Time Points Across Trials
There are potential limitations to the current analysis. Perhaps most importantly, approximately half of the studies used WHO criteria whereas the other half used RECIST, and the definitions for disease progression differ slightly between these response assessment criteria. However, prior studies have shown that these different response assessment systems are highly concordant, including in the assessment of progression.[20, 21] Despite these different assessment criteria, PFS was still highly correlated with OS and was externally validated in an independent data set. Although the data were initially collected prospectively, this is a retrospective study pooled from a series of diverse phase 2 and 3 trials exploring a variety of cisplatin-based treatment regimens. However, the multiple regimens employed in this analysis potentially increase the generalizability of the findings. The timing of restaging assessments was not uniform across all studies, but the landmark time points selected coincided with restaging evaluation time points in 7 of the 8 studies included, and the use of the landmark analyses mitigate the impact of this heterogeneity.
The current study was not designed to, nor establishes, 6- or 9-month PFS as a surrogate for OS. However, there are statistical and operational differences between intermediate and surrogate endpoints. Intermediate endpoints, events, or biomarkers that are early precursors to a given outcome need not necessarily fulfill the statistical criteria for “surrogacy” to serve as useful tools for screening the activity of regimens in the phase 2 setting and prioritizing such regimens for definitive evaluation. Surrogate endpoints, on the other hand, must fully capture the net effect of treatment on the clinical outcome of interest and as a result, may substitute for such outcomes in phase 3 trials as a basis for regulatory agency approval. Several steps, as outlined by Prentice, would ultimately be required to establish 6- and 9-month PFS as surrogate endpoints, and prospective integration of this analysis should be considered in future phase 3 trials. Notably, prior simulations suggest that the short median survival after progression observed in the current analysis makes dilution of OS improvements due to postprogression therapies unlikely in metastatic UC.
The current study reveals that 6- and 9-month PFS is highly correlated with OS in metastatic UC. Furthermore, this study provides benchmarks for 6- and 9-month PFS endpoints for future clinical trial design. Phase 2 (randomized) trials, using 6- or 9-month PFS as a primary endpoint, may overcome many of the limitations of current phase 2 clinical trial design in patients with chemotherapy-naive metastatic UC and better inform “go/no-go” decisions in this clinical disease state moving forward.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURE
Dr. Hahn has been on the speakers bureau and received honoraria from Janssen and Medivation, has been a consultant to Celgene and GlaxoSmithKline, and has received research support to the institution from Genentech, Novartis, Bristol-Myers Squibb, Pfizer, Celgene, Millennium, Merck, Sanofi, and Exelixis.