This article is more than 5 years old. Information may no longer be current.
DFS can serve as surrogate for OS in localized prostate cancer
Click Here to Manage Email Alerts
CHICAGO — DFS can act as a surrogate endpoint for OS in trials for localized prostate cancer, according to the results of a study presented at the ASCO Annual Meeting.
Further, time to disease recurrence can be used as a surrogate for disease-specific survival in this setting.
“Currently drugs are being developed to treat prostate cancer in later stages using OS as the endpoint, but the disease is most curable when it is still localized,” Christopher Sweeney, MBBS, associate professor at Harvard Medical School and medical oncologist at the Dana-Farber Cancer Center, told HemOnc Today. “There is a disconnect between the drugs and where they have the most effect. This disconnect prompted our investigation into possible surrogate endpoints.”
The Intermediate Clinical Endpoints in Cancer of the Prostate Working Group seeks to identify surrogate endpoints based on a meta-analysis of trial data.
Sweeney and colleagues hypothesized that DFS could stand as a surrogate endpoint for OS.
Researchers collected individual patient data from 102 randomized trials that compared treatments in localized prostate cancer. They evaluated DFS from the first randomization to the first evidence of clinical recurrence or death from any cause and OS from the first randomization to death from any cause.
The analysis included 21,140 men from 24 randomized trials. Thirty-two percent of the men were part of prostatectomy trials, approximately 30% had intermediate-risk disease, 50% had high-risk disease and 86% were aged younger than 74 years.
Median follow-up was 10.2 years.
Thirty-two percent (n = 5,370) of evaluated men died, 30% (n = 1,592) of whom died as a result of prostate cancer.
Researchers evaluated DFS-for-OS surrogacy using a meta-analytic two-stage validation model. Two conditions had to be met to claim DFS as a surrogate for OS: A correlation must exist between the intermediate clinical endpoint and the true endpoint, and a correlation must exist between study-specific treatment effect on the endpoints.
Researchers conducted two analyses to evaluate the first condition, including R2 from weighted linear regression of t-year Kaplan Meier estimates, and Plackett’s copula model estimate of Kendal Tau correlation.
When comparing 10-year OS vs. 5-year DFS, researchers calculated a 0.8 (95% CI, 0.67-0.86) Kaplan Meier estimate and 0.85 (95% CI, 0.85-0.86) Kendall Tau correlation. An analysis of 8-year OS vs. 5-year DFS yielded a 0.85 (95% CI, 0.76-0.9) Kaplan Meier estimate for the first condition.
For the second condition, researchers calculated a log HR for the treatment-effect correlation of 0.8 (95% CI, 0.6-0.87).
Researchers then conducted a secondary surrogacy evaluation to determine if time to disease recurrence serves as a surrogate for disease-specific survival. Researchers used the same two-condition evaluation method for the secondary analysis, but censored nonprostate cancer deaths.
For the first condition, 10-year disease-specific survival vs. 5-year time to disease recurrence had a 0.73 (95% CI, 0.56-0.81) Kaplan Meier estimate and 0.71 (95% CI, 0.7-0.72) Kendall Tau correlation. Eight-year disease-specific survival and 5-year time to disease recurrence had a 0.79 (95% CI, 0.67-0.85) Kaplan Meier estimate.
For the second condition, the log HR for the treatment-effect correlation was 0.61 (95% CI, 0.29-0.75).
“For patients with localized prostate cancer, trials could take as long as 10 to 15 years if the endpoint is OS,” Sweeney said. “Using DFS as the endpoint can expedite trials and provide further research opportunities.” – by Nick Andrews
Reference :
Sweeney C, et al. Abstract 5023. Presented at: ASCO Annual Meeting; June 3-7, 2016; Chicago.
Disclosure: Sweeney reports consultant/advisory roles with, institutional research funding from, and stock or other ownership in Astellas Pharma, AstraZeneca, Bayer, BIND Biosciences, Exelixis, Genentech/Roche, Janssen Biotech, Leuchemix and Sanofi. He also reports royalties from patents on parthenolide and dimethylaminoparthenolide (Leuchemix), and abiraterone plus cabozantinib (Exelixis). Please see the abstract for a list of all other researchers’ relevant financial disclosures.
Perspective
Back to Top
James Symanowski, PhD
Sweeney and colleagues reported results from research conducted by the Intermediate Clinical Endpoints in Cancer of the Prostate Working Group. The objective of the investigation was to identify proximal endpoints that may serve as surrogate endpoints for OS and for disease-specific survival (DSS) in the localized prostate cancer setting. This was accomplished with a meta-analysis of 24 randomized clinical trials and, importantly, based on individual patient data (n = 21,140). Although the identification and selection processes will need to be described in detail in a full publication, this represents a robust sample size for the evaluation of surrogacy.At the endpoint level, various correlation measures were presented and DFS correlated with OS (eg, Kendall’s Tau = 0.85; 95% CI, 0.85-0.86), although time to disease recurrence (TDR) showed a weaker correlation with DSS (Kendall’s Tau = 0.68; 95% CI, 0.67- 0.69).
However, the correlation between two endpoints is not a sufficient condition to establish the surrogacy of the proximal endpoint, as Fleming and others have clearly demonstrated. A surrogate endpoint must demonstrate a high correlation with the true endpoint in terms of the between-arm treatment effect sizes. The researchers have established this regarding DFS as a surrogate endpoint for OS, as the correlation between the DFS and OS log HRs was 0.73 (95% CI, 0.53-0.82).
Examination of the scatter plot of the log HRs reveals supportive findings. First, this correlation holds over a wide range of HRs for both endpoints (log HR for DFS ranging from approximately –1.2 to 0.6; log HR for OS ranging from approximately –0.6 to 0.6). Second, with the exception of one significantly outlying trial, the correlation appears to be very high for all other trials. The results from the outlying trial likely had a large detrimental effect on the overall correlation. These results support the conclusion that DFS appears to be a surrogate endpoint for OS. However, the correlation between the log HRs for TDR vs. DSS was markedly weaker (0.63; 95% CI, 0.36-0.75); therefore, the conclusion that TDR can be used as a surrogate for DSS is questionable.
Nonetheless, the results regarding the surrogacy of DFS may have profound impact on the design of clinical trials for men with early-stage prostate cancer in terms of the duration of the trials and, importantly, the ability to avoid the confounding impact of subsequent anticancer therapy.
Reference:
Fleming TR and Powers JH. Stat Med. 2012;doi:10.1002/sim.5403.Click Here to Manage Email Alerts