Evaluating CR as a surrogate endpoint for PFS in R/R chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL): A meta-analysis of randomized controlled trials (RCT).

Abstract

7046 Background: CLL/SLL is the most common leukemia in adults. Achieving a CR by International Workshop on CLL 2018 criteria indicates complete eradication of CLL/SLL in all disease compartments. Patients with R/R CLL/SLL who achieved CR tend to have delayed disease progression/death compared with patients who did not achieve CR. This is the first report to evaluate CR as a surrogate endpoint for PFS in R/R CLL/SLL using aggregate RCT data. Methods: A systematic literature review (SLR) identified published RCTs in R/R CLL/SLL from inception to 10/2023, reporting nonzero CR rates (CRR) and PFS. Association between treatment effects on CR and PFS across RCTs was estimated using a weighted linear model (WLM) in the primary analysis, and Daniels and Hughes (D&H) and Riley bivariate random-effects meta-analysis (BRMA) in sensitivity analyses. Strength of association was assessed based on surrogacy criteria described by D&H. Predictive performance of models was evaluated using leave-one-out cross-validation. Results: The SLR identified 13 RCTs with a total of 4388 patients with R/R CLL/SLL treated with a variety of therapies, including but not limited to BTKi, BCL2i, PI3Ki, CAR T cell therapy, anti-CD20 mAb, and chemotherapy. Across the evidence base, CRR ranged from 0.48% to 38.10% and median PFS was between 1 and 35.9 months (excluding unreached medians). Across RCTs, contrasting treatment and control arms, higher odds of CR resulted in lower hazards of disease progression/death (CRR improvement translated to PFS benefit). D&H surrogacy criteria were met for WLM: 1) a near-zero, statistically nonsignificant intercept, indicating that no treatment effect on CRR implies no effect on PFS; 2) a negative, statistically significant slope, suggesting that improved CRR correlates with longer PFS; and 3) a near-zero conditional variance, meaning that PFS HR variation was primarily explained by CR odds ratio. During cross-validation, treatment effects on CR were predictive of PFS benefits, as the 95% predictive interval of PFS HR contained the observed HR in all RCTs for WLM. Results were largely consistent across all 3 models(Table). Conclusions: Improved CRR corresponds to prolonged PFS across RCTs and treatment comparisons. The results support CRR as an important treatment goal for R/R CLL/SLL and a valid surrogate endpoint. Consistent results and predictive performance across different models indicate the robustness of the results. [Table: see text]