Surrogate Endpoints In Diffuse Large B-cell Lymphoma In The Rituximab Immunochemotherapy Era

Purpose:Overall survival(OS)is the gold standard to evaluate clinical benefit of novel drug or regimen in randomized controlled trials(RCTs),but it generally requires a large sample size and a long follow-up to demonstrate significant survival benefit.Event-free survival(EFS)or progression-free survival(PFS)as the surrogate endpoint of OS can help to identify treatment benefit earlier,reduce ineffective exposure and accelerate marketing.Diffuse large B-cell lymphoma(DLBCL)is the most common subtype of non-Hodgkin lymphoma(NHL),but the high-risk DLBCL patients can't have a satisfied response to standard immunochemotherapy currently.Relapsed and refractory patients lack effective second-line salvage therapy to prolong post-progression survival and experience early deaths.Surrogate endpoints help to accelerate the approval of more effective drugs or regimens,improve the long-term survival of high-risk patients as soon as possible.The purpose of this study was to investigate surrogate candidates EFS and PFS as efficacy endpoints in RCTs of DLBCL in the rituximab immunochemotherapy era,guiding RCT design and decision-making.Methods:A literature-based meta-analysis including phase ? RCTs and retrospective cohort studies in previously untreated DLBCL receiving rituximab-containing immunochemotherapy was conducted.Quality control was performed in potentially eligible studies,of which studies with high risk of bias were excluded.Surrogacy models were firstly established in RCTs,and then externally validated in retrospective populations.Trial-level surrogacy analysis was conducted by the correlation between logarithmic(lg)hazard ratio(HR)for EFS/PFS and lg HR for OS.Correlation analysis at treatment arm level was performed between survival rates of EFS/PFS at different year points(year 1,2,3 or 5)and 5-year OS.Coefficient of determination(R2)in weighted linear regression,with weight equal to patient size,was used to evaluate the correlation.We predefined the surrogacy criteria(strong correlation)as R2>0.8 with its lower boundary of 95%confidence interval(CI)>0.6.Results:Twenty-three RCTs were included in modeling,with 45 retrospective cohort studies enrolled in external validation.In trial-level surrogacy,EFS(R2=0.75;95%CI,0.48-0.91)and PFS(R2=0.65;95%CI,0.25-0.89)were associated with OS respectively,but neither met the predefined surrogacy criteria of strong correlation.The analysis of survival rate in treatment arms of RCTs demonstrated that only 3-year EFS(R2=0.82;95%CI,0.67-0.92)and 5-year EFS(R2=0.87;95%CI,0.75-0.94)were strongly correlated with 5-year OS.1-year EFS(R2=0.67;95%CI,0.44-0.84),2-year EFS(R2=0.78;95%Cl,0.62-0.88)and PFS at different year points(year 1,2,3,5)were correlated with 5-year OS,but without strong correlations.External validation suggested that PFS prediction models tended to overestimate 5-year OS in high-risk patients with actual 5-year OS lower than 80%.Conclusions:There was no strong association of EFS and PFS as earlier efficacy endpoints with OS in phase III RCTs in newly diagnosed DLBCL treated with rituximab-containing immunochemotherapy,indicating the clinical use in caution.More effective therapies to high-risk patients should be urgently explored in new trials to reduce refractory/relapse events and improve long-term EFS,which will finally improve survival.