AUTHORS
Andre Meireles, MD, Paloma Resende, MD2, Fabio Pires S Santos, MD, PhD3, Victor Cury, MD4, Carolina Feres, MD5, Larissa Lane Cardoso Teixeira, MD6, Leonardo Javier Arcuri Sr, LEONARDO JAVIER ARCURI7, Renata Stanzione, MD8, Gabriela Vieira, MD9, Kira Bucci, MD10, Nelson Hamerschlak Sr, MD PhD11, Guilherme Fleury Fleury Perini I, MD12
AFFILIATIONS
1 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
2 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
3 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
4 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
5 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
6 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
7 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
8 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
9 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
10 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
11 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
12 Einstein Hospital Israelita, Sao Paulo, SP, Brazil
ABSTRACT
Introduction: Patients with relapsed or refractory diffuse large B-cell lymphoma (R/R DLBCL) who are not candidates for autologous stem-cell transplantation or CAR T-cell therapy (transplant-ineligible, TI) have poor outcomes. Several combinations have now been tested against standard rituximab-based chemotherapy in randomized trials, but no head-to-head comparisons exist. We conducted a systematic review and network meta-analysis (NMA) of second-line or later therapy in this population.
Methods: Registered in PROSPERO (CRD420261418049). We searched PubMed, Embase, Cochrane and congress proceedings through July 11, 2026. TI status was operationalized using a prespecified tiered cascade requiring protocol-level ineligibility criteria verified against trial registries. Studies with CAR T-cell therapy as intervention or comparator were excluded. Of 12,426 records screened, 272 were assessed in full text and 6 randomized trials anchored on a rituximab-chemotherapy comparator (R-GemOx or BR) with published randomized time-to-event data were eligible: STARGLO, POLARGO, SUNMO and NIVEAU (R-GemOx control) and GO29365 and POLAROSE (BR control). The 6 trials informed 5 treatment contrasts, 4 of which were single-trial; only Pola-BR versus BR contributed 2 trials, leaving between-study variance non-estimable. A contrast-based fixed-effect Bayesian NMA with vague priors was fitted in multinma 0.8.1. The network is connected only if the R-GemOx and BR control arms are treated as exchangeable; these were therefore combined into a single rituximab-chemotherapy node for the primary analysis, and kept separate in a sensitivity analysis testing that assumption. Endpoints were progression-free survival (PFS) and overall survival (OS), reported as hazard ratios (HR) with 95% credible intervals (CrI).
Results: Six trials contributed 12 arms and 1,129 randomized patients. Four regimens improved PFS versus rituximab-chemotherapy: polatuzumab vedotin plus R-GemOx (Pola-R-GemOx) HR 0.37 (0.27-0.51), glofitamab plus GemOx (Glofit-GemOx) 0.41 (0.29-0.57), mosunetuzumab plus polatuzumab (Mosun-Pola) 0.41 (0.28-0.60) and polatuzumab plus BR (Pola-BR) 0.42 (0.28-0.66). Nivolumab plus R-GemOx showed no benefit, HR 1.06 (0.81-1.37). For OS, Pola-BR was 0.45 (0.28-0.72), Glofit-GemOx 0.60 (0.43-0.84), Pola-R-GemOx 0.60 (0.43-0.83), Mosun-Pola 0.80 (0.52-1.24) and nivolumab plus R-GemOx 0.98 (0.73-1.32). SUCRA ranked Pola-R-GemOx first for PFS (0.80) and Pola-BR first for OS (0.93). None of the 6 pairwise comparisons among the 4 effective regimens excluded unity for either endpoint; PFS point estimates differed by at most 13%, with CrIs spanning 0.51 to 1.49. Transitivity was questionable for OS. Median control OS was 4.7 (GO29365) and 6.5 months (POLAROSE) versus versus 12.5, 12.5 and 13.6 months in the 3 R-GemOx trials reporting it, a difference exceeding any treatment effect in the network and not attributable to age, since the oldest population in the network (median 76 years, R-GemOx) outlived both BR populations (medians 71 and 61 years). Control PFS was comparable once assessor was harmonized (2.0-3.8 months). Histologic composition, a prespecified effect modifier, varied across trials (high-grade B-cell lymphoma 0-20% of enrolment); within SUNMO, PFS point estimates differed by histology (0.38 for DLBCL versus 0.78 for high-grade B-cell lymphoma; interaction p=0.13). Keeping the 2 control regimens separate split the network into 2 disconnected subnetworks, rendering every cross-backbone contrast – including all comparisons of Pola-BR with the R-GemOx-based regimens — non-estimable.
Conclusions: In TI R/R DLBCL, 4 regimens incorporating an antibody-drug conjugate, a bispecific antibody, or both improved PFS over rituximab-based chemotherapy; 3 of these (Pola-R-GemOx, Glofit-GemOx and Pola-BR) also improved OS, whereas the OS estimate for Mosun-Pola was compatible with no effect and nivolumab improved neither endpoint. Ranking among these regimens is not supported by the randomized evidence: no pairwise contrast reached significance, and comparisons across chemotherapy backbones become non-estimable once the control arms are not assumed exchangeable – the same assumption on which network connectivity depends. Choice among these regimens should rest on toxicity, deliverability and subsequent-therapy considerations rather than on network rankings.
DISCLOSURES
Paloma Resende, MD: BeOne - Honoraria; CSL Behring - Honoraria