Qian Shi, PhD
Progression-free survival (PFS) appears to be a surrogate endpoint for overall survival (OS) in patients undergoing first-line treatment for diffuse large B-cell lymphoma (DLBCL), according to results from a large pooled analysis of 13 multicenter, randomized, controlled trials.1
Investigators with the Surrogate Endpoints for Aggressive Lymphoma (SEAL) group used linear regression (R2WLS
) and Copula bivariable (R2Copula
) models to assess the correlation of trial-level surrogacy with treatment effect estimates for PFS, PFS at 24 months (PFS24), and OS. To demonstrate surrogacy, either R2WLS
had to be ≥0.80 and and neither estimate could be <0.7. The prespecified lower-bound of the 95% CI was set at >0.6.
Trial-level surrogacy for PFS met the predefined criteria (R2WLS
= 0.83; 95% CI, 0.57-0.94 and R2Copula
= 0.85; 95% CI, 0.73-0.98), and investigators found that PFS strongly correlated with OS at the patient level. The rank correlation coefficient ρ between PFS and OS was 0.85 (95% CI, 0.84-0.86) for patient-level correlation.
The HR for the surrogate threshold effect was 0.89, indicating that, in a future trial, an observed HR for PFS of ≤0.89 “would predict a significant treatment effect,” first author Qian Shi, PhD, department of Health Science Research, Mayo Clinic, and colleagues wrote.
Findings were consistent between the observed and predicted OS treatment effects for each comparison unit on the basis of PFS on “leave-one-out cross-validation” except when excluding the PIX203 comparison (R2WLS
= 0.87; 95% CI, 0.70-0.95 and R2Copula
= 0.88; 95% CI, 0.76-0.99).
When investigators evaluated the re-estimated R2
, they found that PIX023 was also a “high influence outlier” when excluding a single comparison at a time.
Trial-level surrogacy for PFS24 did not meet prespecified criteria (R2WLS
= 0.77; 95% CI, 0.51-0.92 and R2Copula
= 0.78; 95% CI, 0.59-0.96). However, investigators found a significant correlation between PFS24 and OS at the patient level (odds ratio [OR], 1.51). The global OR was 61.1 (95% CI, 52.6-69.6) at the patient level, meaning that, at 24 months after initiation of induction treatment, patients who were alive and disease free have “substantially higher odds of remaining alive beyond a particular time point,” Shi et al wrote.
OR for the surrogate threshold effect was 1.51, indicating that an observed OR ≥1.51 for PFS24 would foresee a significant treatment effect on OS in a future trial. Investigators found that restricting the analysis to induction comparisons improved surrogacy performance.
“Our analysis demonstrates that treatment effect on PFS is a strong predictor of treatment effect on OS. These results were consistent across different surrogacy estimation methods and sensitivity analyses,” Shi et al wrote.
“The strong association between PFS and OS was maintained irrespective of the inclusion of rituximab in the induction and/or maintenance regimen. Together, these results indicate that PFS serves as a strong surrogate for OS in trials evaluating first-line therapy for DLBCL,” the researchers added.
While OS remains the gold standard for efficacy in DLBCL, the investigators wrote that “[a]n earlier endpoint assessed posttreatment would expedite clinical trial conduct and accelerate patient access to effective new therapies.”
SEAL investigators reviewed individual patient data from 7507 adults enrolled in large (N = ≥100) clinical trials published after 2002 to October 2015. The trials included 17 two-arm comparisons, 15 of which included rituximab. Fourteen trials compared induction regimens and 3 evaluated maintenance regimens.
The median patient age was 62.8 years (range, 18.0-92.2) and 58% were aged 60 years or older. Fourteen percent of patients had ECOG performance status ≥2, and men made up 54% of the cohort. Forty percent of patients had Ann Arbor stage IV disease, 36% had stage I/II, and 24% had stage III. The median follow-up for OS was 52 months.
These findings update results first presented at the 2016 ASH Annual Meeting. Shi et al reviewed data from large, multicenter, randomized controlled clinical trials of this same patient population published after January 1, 1995. The prespecified criteria for surrogacy were the same.2
At the patient level, investigators observed a strong correlation between longer PFS and longer OS (ρ = 0.85; 95% CI, 0.84-0.86). The same was true for PFS24 (global OR, 61.1; 95% CI 52.6-69.6).
The correlation between PFS and OS remained strong at the trial level, meeting the prespecified surrogacy criteria (R2WLS
= 0.83; 95% CI, 0.58-0.94 and R2Copula
= 0.85; 95% CI, 0.72-0.99). The correlation between PFS24 and OS was slightly lower at the trial level (R2WLS
= 0.77; 95% CI 0.49-0.92 and R2Copula
= 0.78; 95% CI 0.59-0.96).
As with the more recent results, investigators observed consistency across comparisons, including rituximab or only induction regimens, and investigators found that restricting analysis to induction studies improved PFS24 performance.
- Shi Q, Schmitz N, Ou FS, et al. Progression-free survival as a surrogate end point for overall survival in first-line diffuse large B-cell lymphoma: an individual patient–level analysis of multiple randomized trials (SEAL) [published online July 5, 2018]. J Clin Oncol. doi: 10.1200/JCO.2018.77.9124.
- Shi Q, Schmitz N, Flowers C, et al. Evaluation of progression-free survival (PFS) as a surrogate endpoint for overall survival (OS) in first-line therapy for diffuse large B-cell lymphoma (DLBCL): findings from the surrogate endpoint in aggressive lymphoma (SEAL) analysis of individual patient data from 7507 patients. Presented at: 58th American Society of Hematology Annual Meeting; San Diego, CA; December 2-6, 2016. Abstract 4196.