Multivariable analyses (MVAs) of overall survival (OS) in the phase 3 SUNMO, STARGLO, and POLARGO trials in relapsed/refractory large B-cell lymphoma (LBCL).

7093 Background: Phase 3 trials, notably SUNMO (NCT05171647), STARGLO (NCT04408638) and POLARGO (NCT04182204), have shown superior efficacy of combination treatments over rituximab (R) + gemcitabine and oxaliplatin (GemOx) in second line or later LBCL. However, baseline imbalances may impact the robustness of efficacy results, even with stratified randomization. We conducted MVAs of SUNMO, STARGLO and POLARGO to assess OS. As the target number of OS events was not reached at the SUNMO interim OS analysis, to increase the sample size we also conducted a pooled analysis of mosunetuzumab (Mosun) + polatuzumab vedotin (Pola; data from SUNMO and GO40516 [Phase 2; NCT03671018]) vs R-GemOx (comparators from SUNMO, STARGLO and POLARGO). Methods: A multivariable Cox regression model was used to estimate treatment effect while adjusting for potential confounding factors selected by a systematic process involving: 1) univariable analysis of prespecified baseline factors; 2) multicollinearity assessment of pairwise correlation among selected factors; 3) MVA of the experimental arm (SUNMO: Mosun-Pola; STARGLO: glofitamab [Glofit]-GemOx; POLARGO: Pola-R-GemOx) vs the control arm (R-GemOx) and selected non-collinear prognostic baseline factors from steps 1 and 2. Results: In SUNMO, patients treated with Mosun-Pola had more high-risk features, such as a higher ECOG performance status and greater bulky disease, vs those treated with R-GemOx. The MVA-adjusted hazard ratio (HR, 0.65 [95% confidence interval [CI]: 0.41–1.02]) was more favorable than the unadjusted interim OS HR (0.80 [95% CI: 0.54–1.20]) for Mosun-Pola vs R-GemOx. The MVA-adjusted HR for POLARGO was also more favorable than the unadjusted HR; the HR for STARGLO remained largely unchanged (Table). A pooled MVA was used to increase the sample size, noting baseline differences between trials (e.g. ≥2 prior lines of therapy: SUNMO, 56%; STARGLO, 37%; POLARGO, 35%). In this pooled analysis, OS HRs favored Mosun-Pola vs R-GemOx in the unadjusted (HR, 0.62 [95% CI: 0.48–0.81]) and MVA-adjusted (HR, 0.59 [95% CI: 0.44–0.81]) models, with lower HRs and narrower 95% CIs vs the MVA of SUNMO alone. Conclusions: The MVAs support favorable OS benefits in SUNMO, STARGLO and POLARGO, all with an MVA-adjusted OS HR ≤0.65 vs R-GemOx. The pooled MVA further supports the OS benefit of Mosun-Pola vs R-GemOx. Overall, the MVAs show that baseline imbalances can influence OS estimates, thus multivariable adjustment is important when interpreting pooled or cross-trial comparisons. Clinical trial information: NCT05171647 , NCT04408638 , NCT04182204 , NCT03671018 . SUNMO n=208 STARGLO n=274 POLARGO n=255 Pooled Mosun-Pola vs R-GemOx n=465 Median follow-up, months (range) 23.2 (0–32) 20.7 (0–36) 24.6 (0–34) – MVA-adjusted OS HR (95% CI) 0.65 (0.41–1.02) 0.63 (0.44–0.90) 0.52 (0.36–0.74) 0.59 (0.44–0.81) Unadjusted OS HR (95% CI) 0.80 (0.54–1.20) 0.62 (0.43–0.88) 0.60 (0.43–0.83) 0.62 (0.48–0.81)