MFS Strong Surrogate Endpoint in Localized Prostate Cancer

Christopher J. Sweeney, MBBS

After a median of 10 years’ follow-up, investigators with the ICECaP Working Group have concluded that there is a strong correlation between metastasis-free survival (MFS) and overall survival (OS) in patients with localized prostate cancer.

At the patient level, Kendall’s correlation with OS was 0.91 for MFS. At the trial level, R² was 0.83 (95% CI, 0.71-0.88) from weighted linear regression of 8-year OS rates versus 5-year MFS rates. Treatment effects, as measured by log hazard ratios, were well correlated for OS and MFS (R², 0.92; 95% CI, 0.81-0.95).

“MFS is a strong surrogate for OS in clinically localized prostate cancer in a patient population with an approximate 15% chance of dying of prostate cancer over 10 years despite potentially curative local therapy,” Christopher J. Sweeney, MBBS, Dana-Farber Cancer Institute, and coinvestigators wrote. “Surrogacy is independent of primary local interventions and type of adjuvant therapy. Linear regression graphs used to generate the [surrogate threshold effect] can be used to define relative improvements in MFS that are associated with clinically meaningful improvements in OS.”

Researchers reviewed data collected from trials conducted from 1987 to 2011. They analyzed data from 24 trials (N = 21,140) to determine disease-free survival (DFS) and 19 trials (N = 12,712) to determine MFS. Disease-free survival has proven to be a surrogate for OS and is used as a primary endpoint in adjuvant trials of colon cancer, but there are no intermediate clinical endpoints accepted as robust surrogates for OS in prostate cancer trials.

Researchers determined the correlation of an intermediate clinical endpoint with OS and the correlation of treatment effects on both the intermediate clinical endpoint and OS using a 2-stage meta-analytic validation model.

At the individual patient level, the correlation between DFS and OS was 0.85 (95% CI, 0.85-0.86) and 0.91 (95% CI, 0.91-0.91) between OS and MFS. The correlation with disease-specific survival (DSS) was 0.68 (95% CI, 0.67-0.69) for time to disease recurrence and 0.91 (95% CI, 0.91-0.92) for time to metastases when nonprostate cancer deaths were censored.

Researchers said tight correlation between endpoints is reflected by the tight correlation between the trial and arm-specific Kaplan-Meier estimates of OS or DSS at 8 years versus the Kaplan-Meier estimates of the surrogates at 5 years. From the weighted linear regression (WLR), R² was 0.86 (95% CI, 0.78-0.90) between 8-year OS rates versus 5-year DFS and 0.83 (95% CI, 0.71-0.88) between 8-year OS rates versus 5-year MFS.

Treatment effects, as measured by log hazard ratios, for the surrogates and OS were well correlated for both DFS (R², 0.73; 95% CI, 0.53-0.82) and MFS (R², 0.92; 95% CI, 0.81-0.95).

The research team found that the results were largely consistent when analyzing the high-risk population alone, or in a subgroup analysis by type of primary therapy and by exposure to androgen deprivation therapy with radiation therapy-based trials at both the patient and trial level.

At the patient level, Kendall’s correlation between OS and DFS was 0.91 (95% CI, 0.90-0.92) in radical prostatectomy-based trials and 0.84 (95% CI, 0.83-0.84) in radiation therapy—based trials. At the trial level, R² from the WLR of log(HR)-OS versus log(HR)-DFS was 0.87 (95% CI, 0.31-0.93) for radical prostatectomy—based trials and 0.75 (95% CI, 0.48-0.84) for radiation therapy-based trials.

There was no separate analysis conducted for radical prostatectomy-based trials for the MFS endpoint because 90% of patients were treated in radiation trials.

The strong correlation between MFS and OS might mean that clinical trials can be designed with MFS as the final endpoint. Researchers said that trials have been historically designed with an HR for OS HR ranging from 0.71 to 0.75, with a study duration of 11.5 to 16.2 years, and 1000 patients enrolled over 5 years.

“Clearly, study durations would be shorter if the same treatment effects were assumed for MFS,” researchers wrote. “WLR analysis predicts that for OS HRs that range from 0.71 to 0.75, and corresponding MFS HRs would range from 0.65 to 0.7, so the benefit of using MFS instead of OS could be even greater.”

Xie W, Regan MM, Buyse M, et al. Metastasis-free survival is a strong surrogate of overall survival in localized prostate cancer [published online August 10, 2017]. J Clin Oncol. doi:10.1200/JCO.2017.73.9987.