AUTHORS
Qianying Pan, Jianwei Du2, quande Lin, PhD3, Zhihui Li, MD4, Hao Guo, MD, PhD5, Qing Zhang, MD6, Xueli Jiao, MD7, Jingyi Yang, MD8, Fenghua Gao9, Xingchen Liu, MD10, Ran Zhao, MD11, Keshu Zhou, PhD12
AFFILIATIONS
1 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
2 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
3 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
4 Xinxiang Central Hospital, Xinxiang, Henan, China
5 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
6 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
7 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
8 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
9 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
10 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
11 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
12 The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, Henan, China
ABSTRACT
Background
POLARGO (Matasar 2025 JCO; NCT04182204) established Pola-R-GemOx as an effective salvage regimen in relapsed/refractory (R/R) DLBCL. Real-world R/R practice employs heterogeneous pola-based backbones with no single dominant regimen, and whether these reproduce POLARGO-level benefit in Chinese patients is unknown. We performed a matching-adjusted indirect comparison (MAIC) against both POLARGO arms.
Methods
R/R patients receiving pola-based salvage therapy (n=128; 113 with complete matching covariates) were compared with POLARGO (24.6-month follow-up, ITT, IRC-PET-EOT response) using unanchored MAIC with entropy-balancing weights. Six covariates were matched: age>70, female sex, IPI 3–5 (a force-retained balancing covariate, not an eligibility criterion), stage III–IV, GCB, and ≥2 prior lines (POLARGO lacks a baseline LDH column, so LDH was not included; the age threshold >70 aligns with POLARGO stratification). The primary endpoints were the overall response rate (ORR) and complete response (CR) rate, compared between the weighted RW pola-based salvage group and each POLARGO arm (R-GemOx the primary comparison; Pola-R-GemOx a same-class reference comparison). Comparators were R-GemOx (standard salvage) and Pola-R-GemOx (same-class pola regimen). Because the RW cohort used heterogeneous backbones, it was evaluated as an aggregate "real-world pola-based salvage" population. Effective sample size (ESS) and post-weighting balance were assessed. A treatment-completion (≥6 cycles, completer) sensitivity analysis was prespecified to quantify potential immortal-time bias in the best-response estimand (see below).
Results
After weighting, all six covariates matched the POLARGO target; ESS=94.5 (84% retained)—the highest retention among our three MAICs, indicating limited weight concentration. Weighted RW pola ORR was 63.0% and CR 38.5%.
Weighted RW pola salvage approximately doubled response versus standard R-GemOx, and its CR was numerically close to the POLARGO pola arm (38.5% vs 40.3%).
Versus standard R-GemOx salvage, weighted RW pola raised ORR from 24.6% to 63.0% (RD+38.4pp, 95%CI 24.4–52.4; RR 2.56; P<0.0001) and CR from 19.0% to 38.5% (RD+19.4pp, 5.7–33.2; RR2.02; P=0.006)—both statistically significant, roughly doubling response. Versus the same-class POLARGO pola arm (ORR 52.7%, CR 40.3%), the weighted RW estimates were similar: ORR RD+10.3pp (−4.4–24.9; RR 1.20; P=0.17) and CR RD−1.8pp (−16.5–12.8; RR 0.95; P=0.80), with confidence intervals crossing the null—consistent with, but not proof of, comparable efficacy despite backbone heterogeneity.
Treatment-completion sensitivity analysis. The primary estimand uses full-population best response. R/R salvage cycle counts are inherently heterogeneous (distributed 1–8; median 3; only ~21% completed ≥6 cycles), with early discontinuation far more common than in first line. A completer-only estimand (restricted to patients completing ≥6 cycles) is therefore expected to overestimate response by systematically excluding early progressors/deaths—immortal-time/survivor bias. In the prespecified completer layer, only 24 of 128 patients qualified, with unweighted ORR 75.0% and CR 58.3%; the CR is ~20 percentage points higher than the weighted best-response CR of 38.5%, directly illustrating the completion-related selection effect. Because these 24 patients yield a post-weighting ESS<25, an indirect comparison in this layer is unstable and was reported descriptively rather than as a MAIC. Full-population best response thus remains the only robust estimand for this high-attrition R/R setting, and is directly comparable to the POLARGO IRC-PET-EOT CR comparator.
Conclusions
Real-world pola-based salvage therapy in R/R DLBCL roughly doubled ORR and CR versus standard R-GemOx, and its CR was numerically close to the POLARGO pola arm (38.5% vs 40.3%) despite differing backbones. This concordance was observed even though—unlike POLARGO, which excluded prior-pola patients—the RW cohort included patients who had received pola in the first line, a more treatment-exposed population. Because this is an unanchored MAIC in which residual confounding cannot be excluded, and the same-class comparison has wide confidence intervals crossing the null, these findings are hypothesis-generating: they suggest that adding pola to salvage regimens may confer benefit over standard chemoimmunotherapy consistent with same-class trial efficacy, warranting confirmation in adjusted or prospective comparisons.
DISCLOSURES
Qianying Pan: No relevant conflicts of interest to declare.