The stakes in this comparison are high. Merkel cell carcinoma carries a five-year survival rate of roughly 40 to 45 percent in the United States and between 28 and 48 percent in Europe. The disease predominantly strikes men and patients aged 70 and older, and its incidence is climbing as populations age. In most cases, Merkel cell polyomavirus can be detected in the tumor, a viral signature that distinguishes this cancer from melanoma and other skin malignancies. While about 65 percent of patients present with disease confined to the original site, roughly 8 percent already harbor distant metastases at diagnosis, and among those with localized disease, distant spread occurs in nearly half of cases within a single year and in virtually all patients within five years. Once the cancer metastasizes, chemotherapy may initially shrink tumors, but responses typically last only three to six months, and the toxicities are often severe in an elderly patient population.

That therapeutic vacuum opened the door to immunotherapy. Avelumab, a monoclonal antibody that blocks the programmed death-ligand 1 protein, or PD-L1, became the first approved immunotherapy for metastatic Merkel cell carcinoma, gaining accelerated approval in the United States and authorization in the European Union in 2017 on the strength of the phase 2 JAVELIN Merkel 200 trial. Retifanlimab, a humanized immunoglobulin G4 antibody targeting the PD-1 receptor itself, followed more recently, receiving accelerated US approval in 2023 and European approval in 2024 based on the phase 2 POD1UM-201 trial, in which 101 patients with locally advanced or metastatic disease achieved an objective response rate of 54.5 percent. Both drugs work by releasing a molecular brake on T cells, allowing the immune system to recognize and attack tumor cells, but they have never been tested against each other in the same study.

The research team, led by Camilla Porta, Gianni Ghetti, and Massimiliano Povero of the health economics firm AdRes in Turin, Italy, confronted a fundamental problem in evidence-based medicine: how to compare treatments when only separate, single-arm trials exist. Traditional network meta-analyses require a connected web of trials sharing a common comparator, an assumption of consistency across studies, and broadly exchangeable patient populations. None of those conditions held here. Instead, the team used the MAIC approach, first formalized by James Signorovitch and colleagues, which applies statistical weights to individual patient data from one trial until its baseline characteristics match those reported for the aggregate population of another trial. The method directly balances observed covariates between populations and avoids the outcome-model misspecification biases that can afflict alternative techniques such as simulated treatment comparisons, which is why the authors selected it for these time-to-event endpoints.

The mechanics of the analysis were intricate. Individual patient data came from POD1UM-201, with a data cutoff of July 2024 and a median follow-up of 35.7 months, restricted to patients with metastatic disease to align with the eligibility criteria of the comparator. For avelumab, the researchers used aggregate data from part B of JAVELIN Merkel 200, a cohort of 116 previously untreated patients with metastatic disease, digitizing the published Kaplan-Meier survival curves and reconstructing pseudo-individual patient data using an algorithm developed by Pierre Guyot and colleagues. Matching covariates were chosen based on their established prognostic value in Merkel cell carcinoma: age below or above 65 years, Eastern Cooperative Oncology Group performance status, PD-L1 positivity, Merkel cell polyomavirus status, site of the primary tumor, and the presence or absence of visceral metastases at baseline.

Weighting came at a price. After the adjustment, the effective sample size of the retifanlimab population shrank to approximately 42 patients, just 47 percent of the original trial enrollment, a consequence of the right-skewed weight distribution in which a small number of patients carried disproportionate statistical influence. Some covariates simply could not be balanced. Ethnicity data showed an apparent imbalance, with Hispanic or Latino patients representing 1.1 percent of the retifanlimab trial but 25 percent of the avelumab trial, and adjusting for it would have collapsed the effective sample size to unusable levels. Geographic region posed a similar obstacle, since 10.3 percent of avelumab patients came from Australia or Asia, regions absent from the retifanlimab dataset. Race, sex, and time since initial diagnosis were explored in scenario analyses to test the robustness of the base case findings.

The headline result concerns progression-free survival, the length of time patients live without their disease worsening. After adjustment, the weighted hazard ratio of 0.590, with a 95 percent confidence interval of 0.396 to 0.881, statistically significantly favored retifanlimab, translating to roughly a 40 percent lower hazard of progression or death. Intriguingly, the proportional hazards assumption was violated for this endpoint, meaning the treatment effect changed over time. A stratified analysis revealed that the entire advantage was concentrated in the first two months of treatment, with no significant difference thereafter, so the overall hazard ratio should be read as a time-averaged summary rather than a constant benefit. For overall survival, the hazard ratio of 0.800 numerically favored retifanlimab but carried wide uncertainty and did not reach statistical significance, and exploratory piecewise modeling suggested the survival benefit, if real, emerged after the first five months. The odds ratio for overall response rate, 1.871, also pointed toward retifanlimab without achieving significance in the base case, though a scenario analysis found the odds of complete or partial response were approximately 2.4 times higher with retifanlimab.

The authors are candid about the limitations. An unanchored MAIC, used here because no common comparator links the two trials, rests on strong assumptions, including the conditional constancy of absolute treatment effects and adequate adjustment for every prognostic factor and effect modifier, assumptions that cannot be fully verified. Residual confounding by unmeasured characteristics remains possible, and the reduced effective sample size, together with sparse events in the tails of the survival curves, limits statistical precision. Differences in follow-up duration between the trials could also skew time-to-event estimates in either direction, although the authors argue these effects are likely confined to the curve tails. Post-progression therapies, which about 40 percent of patients in each trial received, could theoretically influence overall survival, but the broadly similar distribution of subsequent treatments, including somewhat more frequent subsequent immunotherapy in the avelumab trial, suggests any imbalance would if anything have attenuated rather than inflated the observed advantage for retifanlimab. The findings, the researchers conclude, should be treated as hypothesis-generating rather than confirmatory.

Nevertheless, the analysis carries real-world weight. A companion cost-utility analysis conducted in Italy, drawing on an earlier MAIC, found retifanlimab to be a cost-effective option relative to avelumab, with an incremental cost of 12,228 euros per patient and an incremental cost-utility ratio of 5,037 euros per quality-adjusted life year gained. Practical considerations reinforce the clinical picture: retifanlimab is administered once every four weeks, while avelumab requires infusions every two weeks, a meaningful difference for elderly patients who often face long travel distances to reach the specialist centers equipped to manage this rare cancer. For a disease in which diagnosis is frequently delayed and access to optimal care is constrained, a therapy that combines favorable efficacy signals with a convenient dosing schedule could reshape treatment decisions. As population-adjusted indirect comparisons become increasingly embedded in regulatory, reimbursement, and health technology assessment frameworks, this study offers both a template and a caution: statistical ingenuity can squeeze comparative insight from unconnected trials, but the resulting estimates are only as trustworthy as the assumptions beneath them. For now, the evidence tilts toward retifanlimab, pending the head-to-head data that rare-disease oncology so rarely delivers.

Subject of Research: Comparative effectiveness of the immune checkpoint inhibitors retifanlimab and avelumab in metastatic Merkel cell carcinoma using matching-adjusted indirect comparison methodology

Article Title: Retifanlimab Versus Avelumab for the Treatment of Metastatic Merkel Cell Carcinoma: A Matching-Adjusted Indirect Comparison Analysis

Article References: Porta, C., Ghetti, G., & Povero, M. (2026). Retifanlimab Versus Avelumab for the Treatment of Metastatic Merkel Cell Carcinoma: A Matching-Adjusted Indirect Comparison Analysis. Advances in Therapy. https://doi.org/10.1007/s12325-026-03802-1

Image Credits: AI Generated

DOI: 10.1007/s12325-026-03802-1

Keywords: Merkel cell carcinoma, retifanlimab, avelumab, immune checkpoint inhibitors, matching-adjusted indirect comparison, PD-1, PD-L1, immunotherapy, progression-free survival, POD1UM-201, JAVELIN Merkel 200, rare skin cancer

The research team, led by Yiyue Xu and corresponding author Linlin Wang, systematically searched three major biomedical databases, PubMed, Embase, and Web of Science, for phase III randomized controlled trials published between January 1, 2015, and August 5, 2026. Their inclusion criteria were deliberately strict: only trials evaluating first-line immune checkpoint inhibitor plus chemotherapy regimens in advanced solid tumors qualified. The effort yielded an extraordinary evidence base of 58 phase III randomized controlled trials encompassing 36,664 patients, with regimens combining chemotherapy with 20 distinct immune checkpoint inhibitors across multiple tumor types. That scale matters, because individual trials typically compare a single immunotherapy against chemotherapy alone within one cancer type, leaving clinicians to guess how, say, a PD-L1 inhibitor used in lung cancer might compare on safety with a CTLA-4 pathway agent deployed elsewhere.

The methodological engine of the study is the network meta-analysis, a statistical framework that allows simultaneous comparison of multiple treatments even when they have never been tested against each other directly. By building a network of evidence connected through common comparators, in this case chemotherapy plus placebo, the researchers could estimate odds ratios with 95 percent credibility intervals for every pairwise comparison in the network. They then applied Bayesian ranking probabilities and calculated the surface under the cumulative ranking curve, known as SUCRA, to produce a league table of regimens ordered by safety. The primary outcome was the occurrence of grade 3 or higher immune-related adverse events, the severe toxicities that arise when unleashed immune systems attack healthy tissue. Secondary safety endpoints included all-grade immune-related adverse events and both severe and all-grade treatment-related adverse events more broadly.

The headline finding concerns adebrelimab, an anti-PD-L1 antibody, which emerged as the most favorably ranked regimen for severe immune-related adverse events. Adebrelimab plus chemotherapy achieved the highest SUCRA value in this domain, at 76.27 percent, indicating the strongest cumulative ranking probability among the 17 immune checkpoint inhibitor and chemotherapy regimens analyzed for this outcome. Notably, its pooled odds ratio against placebo plus chemotherapy was 1.62 with a 95 percent credibility interval of 0.28 to 9.83, an interval that crosses one, meaning the analysis could not statistically distinguish its severe immune toxicity risk from chemotherapy alone. For patients and clinicians weighing the trade-off between efficacy and immune toxicity, that profile represents a meaningful signal, though the wide interval also underscores the statistical uncertainty inherent in even the largest networks.

When the researchers turned to the broader toxicity categories, different agents rose to the top. Avelumab plus chemotherapy ranked highest for both all-grade immune-related adverse events and all-grade treatment-related adverse events, suggesting that while this regimen may produce more frequent lower-grade side effects, its overall tolerability profile compared favorably across the network. Meanwhile, regimens built around durvalumab and penpulimab showed the most favorable safety profiles for grade 3 or higher treatment-related adverse events, the severe toxicities of any cause attributable to treatment. The fact that different drugs top different safety categories is itself an important result: it demonstrates that safety is not a single dimension but a multidimensional profile, and the optimal choice may depend on which toxicity domain matters most for a given patient.

Understanding why these differences exist requires a brief tour of immunopharmacology. Immune checkpoint inhibitors work by blocking inhibitory receptors, such as PD-1, PD-L1, and CTLA-4, that normally restrain T-cell activity. Removing these brakes allows immune cells to attack tumors, but the same mechanism can trigger inflammation in the skin, gut, liver, lungs, and endocrine organs, producing the immune-related adverse events that define this drug class. The degree and character of immune activation vary substantially between molecules, depending on target, binding affinity, antibody engineering, and pharmacokinetics. When combined with cytotoxic chemotherapy, which carries its own burden of myelosuppression, nausea, and neuropathy, the resulting toxicity profile reflects a complex interplay that randomized trials rarely dissect in comparative terms. A network meta-analysis of this scale is precisely the tool needed to bring that hidden structure into view.

The rigor of the underlying evidence was assessed using the Cochrane Risk of Bias 2 tool, the current standard for evaluating randomized trials, which judges bias across domains including the randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting. By restricting inclusion to phase III trials, the most definitive study design in oncology, the authors minimized the risk that their rankings reflect artifacts of small or preliminary studies. The Bayesian framework additionally allows uncertainty to propagate through the entire network, so that rankings are expressed as probabilities rather than false certainties. Still, the authors and outside observers alike should note the limitations inherent to any such synthesis: trials enrolled different tumor types, used varying chemotherapy backbones, and applied somewhat heterogeneous toxicity grading, all of which can introduce transitivity assumptions that may not hold perfectly across the network.

The clinical implications are considerable. Immune checkpoint inhibitor plus chemotherapy combinations are now standard first-line therapy for multiple advanced solid tumors, including non-small cell lung cancer, and their use continues to expand. With 20 distinct checkpoint inhibitors now validated in this setting, oncologists face a genuine selection problem, and until now the comparative safety evidence has been fragmented across hundreds of individual trials. Severe immune-related adverse events are not merely statistical inconveniences; they can require high-dose corticosteroids, permanent drug discontinuation, hospitalization, and in rare cases prove fatal. A regimen ranking that identifies adebrelimab, avelumab, durvalumab, and penpulimab as comparatively favorable across different safety domains gives clinicians an evidence-based starting point for individualizing treatment, particularly for patients with preexisting autoimmune disease, organ transplants, or frailty who may tolerate immune toxicity poorly.

The study also carries implications for drug development and health policy. Regulators and guideline committees increasingly demand comparative rather than merely placebo-controlled evidence, and SUCRA-based rankings of this kind can inform which agents deserve prioritization in head-to-head trials, which combinations warrant preferential pricing negotiations, and where post-marketing surveillance should concentrate. The authors report no funding or competing interests, and the analysis was conducted by a radiation oncology department team, reflecting the growing recognition that safety, not only survival, defines the value of modern cancer therapy.

Caveats remain, and the authors are careful not to overstate them away. The credibility interval around adebrelimab’s odds ratio is wide, rankings depend on the set of trials available through August 2026, and the analysis addresses safety rather than efficacy, meaning the safest regimen is not automatically the best one for a given cancer. The article was published as a peer-reviewed, citable version carrying a permanent DOI, subject to further editorial processing. Even so, the work stands as a landmark synthesis: by pooling 58 phase III trials and nearly 37,000 patients into a single coherent network, it converts a decade of scattered toxicity data into an actionable hierarchy, and it signals a maturing of immuno-oncology from the era of proving that these drugs work to the more nuanced task of choosing precisely the right one for each patient.

Subject of Research: Comparative safety of immune checkpoint inhibitor plus chemotherapy first-line regimens for solid tumors

Article Title: Comparative safety of immune checkpoint inhibitor combined with chemotherapy as first-line treatment for solid tumors: a systematic review and network meta–analysis

Article References: Xu, Y., Li, B., Gao, A., Zhao, Q., Yang, L., & Wang, L. (2026). Comparative safety of immune checkpoint inhibitor combined with chemotherapy as first-line treatment for solid tumors: a systematic review and network meta–analysis. BMC Cancer. https://doi.org/10.1186/s12885-026-17007-w

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17007-w

Keywords: immune checkpoint inhibitors, chemotherapy, network meta-analysis, immune-related adverse events, treatment-related adverse events, solid tumors, phase III trials, adebrelimab, avelumab, durvalumab, penpulimab, drug safety