We conducted a retrospective analysis of the clinical data of patients with UC aged ≥18 years with radiologically confirmed metastatic disease. Two cohorts were defined in this study. Cohort 1 included patients with either progressive or recurrent disease after platinum-based treatments who were subsequently treated with pembrolizumab, whereas in cohort 2, patients received avelumab as maintenance therapy after first-line platinum-based chemotherapy and achieved at least stable disease (Figure 1). Patients were treated with one of these two drugs between 1 January 2016 and 31 October 2024, and data were collected from 79 institutions worldwide.

All participants had available data on demographic factors (age and sex), tumor histology, Eastern Cooperative Oncology Group-Performance Status (ECOG-PS), metastatic sites, prior surgeries, chemotherapy treatments, treatment durations, and response to pembrolizumab or avelumab, as assessed by the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. Body Mass Index (BMI) was calculated as weight in kilograms divided by the square of height in meters.

Data on clinical and pathological characteristics were collected from the medical and pathology records of each center, as per routine clinical practice. Standard physical examinations, laboratory tests, and imaging studies (computed tomography and magnetic resonance imaging scans) were performed according to local guidelines.

Patients with incomplete clinical or outcome data were excluded.

The study protocol was approved by the Ethical Committee of the coordinating center (Marche Region, Italy; approval number 2022 39, Study Protocol “ARON 2 Project”) and by the Institutional Review Boards of the participating centers. This study adhered to the Good Clinical Practice (GCP), the Declaration of Helsinki, and international ethical standards for biomedical research. Informed consent was obtained from all patients, while consent was waived for patients who died or were lost to follow-up by the Institutional Review Board of the coordinating center.

The primary aim of this study was to evaluate the prognostic role of liver metastases in patients with metastatic UC treated with pembrolizumab (cohort 1) or avelumab (cohort 2). Tumor responses (progression disease [PD], stable disease [SD], partial response [PR], and complete response [CR]) were evaluated based on the RECIST version 1.1 criteria. We collected data on the overall response rate (ORR) and OS. OS was defined as the time from pembrolizumab or avelumab initiation until death from any cause.

To compare OS across groups, we used the Kaplan–Meier method and the log-rank test. The median follow-up time was determined using the Kaplan–Meier method. Cox proportional hazard models were employed to evaluate the multivariable impact on patient survival and estimate hazard ratios (HRs) with 95% confidence intervals (CIs). Fisher’s exact test was used for pairwise comparisons of categorical variables, and chi-square tests were used for multiple categorical comparisons. Statistical significance was set at p <0.05.

To mitigate confounding factors associated with the retrospective, non-randomized study design, a propensity score matching (PSM) approach was applied. Propensity scores were derived from a multivariable logistic regression model incorporating age, sex, ECOG performance status, tumor histology, primary tumor location, site of metastases, and type of immunotherapy. Patients receiving avelumab were matched in a 1:1 ratio with those treated with pembrolizumab using nearest-neighbor matching without replacement, applying a caliper of 0.2 standard deviations of the logit-transformed propensity score. Post-matching covariate balance was evaluated using standardized mean differences (SMDs), with thresholds below 0.1 indicating an adequate balance. The primary and secondary outcomes of the study were subsequently reassessed in the matched population using Kaplan–Meier survival analyses and Cox proportional hazards regression models, as appropriate.

We included 1,632 patients from the ARON-2 dataset; 1,064 (65%) were male and 568 (35%) were female. Two hundred and twenty (13%) patients had an ECOG-PS ≥2. Pure urothelial carcinoma histology was reported in 79% of patients, and 1,184 (73%) reported tumors of the lower urinary tract.

Cohort 1 included 1,341patients treated with pembrolizumab, and cohort 2 included 291 patients treated with avelumab. Patient’ characteristics are summarized in Table 1.

The median OS was 25.8 months (95%CI: 21.2–27.5) in cohort 2 and was significantly longer in patients without liver metastases (27.0 months, 95%CI: 23.2–29.1 vs. 16.4 months, 95%CI: 12.0–20.0; p < 0.001, Figure 2). In cohort 2, 122 patients (42%) were dead at the time of analysis, and the rates were 49% and 40% in patients with and without liver metastases, respectively (p = 0.255). A total of 204 patients (85%) out of 239 who progressed during avelumab treatment received further lines of treatment.

Pembrolizumab, an IgG4-κ humanized monoclonal antibody that targets PD-1, received Food and Drug Administration (FDA) approval in May 2017 for the treatment of pretreated mUC patients based on the positive results of the Keynote-045 (KN045) phase III trial. In this study, the presence of liver metastases was a stratification factor, and approximately 34% of patients with this characteristic were enrolled in each arm of the trial. Unfortunately, outcomes focusing on this specific population are not available for KN 045, analysis of OS in key subgroups shows an HR 0.85 (0.61–1.20) (23). In the present real-world dataset, only 17% of the 1,341 patients treated with pembrolizumab had liver metastases. The total population outcomes (OS and ORR) of ARON-2 were improved indirectly compared to those of KN045, confirming the efficacy of pembrolizumab in this setting.

Patients with liver metastases had a median OS of 9.4 months compared with 20.1 months in patients without liver metastases (p <0.001). Available historical data on vinflunine in the same setting showed the poorest prognosis in these patients. In the CURVE trial, mUC patients with liver metastases had a mOS of 5.6 months vs. 9.4 months in patients without, and in the Spanish dataset, mOS was 6.1 months vs. 11.7 months, respectively (24, 25). These data confirm that the presence of liver metastases is still a negative prognostic factor in the immunotherapy era, even if outcomes in this setting of patients improved compared to historical controls with chemotherapy.

Further subgroup analysis of the present dataset confirmed the negative impact of other known prognostic factors, with significantly lower OS in patients with ECOG-PS ≥2 and those with concomitant bone metastases. In the present analysis, a BMI ≥25 kg/m² emerged as an independent favorable prognostic factor in patients with liver metastases who were treated with ICIs. While this observation should be interpreted cautiously, given the retrospective nature of the study, it is consistent with the growing body of evidence describing an “obesity paradox” in immuno-oncology (26, 27). Multiple retrospective and pooled analyses across tumor types, including melanoma, non-small-cell lung cancer, renal cell carcinoma, and UC, have reported improved survival outcomes in overweight and mildly obese patients receiving ICIs compared with normal-weight individuals (28). Several non-mutually exclusive mechanisms have been proposed from a biological perspective. Obesity is associated with chronic low-grade inflammation and altered adipokine secretion, including increased leptin and decreased adiponectin levels, which may promote T-cell activation and survival (29, 30). Preclinical models have demonstrated that leptin-driven signaling enhances T-cell exhaustion while simultaneously increasing PD-1 expression, paradoxically rendering tumors more susceptible to PD-1/PD-L1 blockade (31). In addition, obesity-related metabolic reprogramming may modulate cytokine profiles and immune cell trafficking within the tumor microenvironment, potentially amplifying the therapeutic effect of immune checkpoint blockade (32). Importantly, recent translational studies suggest that patients with higher adiposity exhibit increased engagement of the PD-1/PD-L1 axis and enhanced reinvigoration of exhausted CD8+ T cells after ICI therapy (33). These findings provide biological plausibility for our clinical observations and support the hypothesis that host metabolic status may act as a modifier of immunotherapy efficacy, even in traditionally immunosuppressive contexts, such as liver metastases (34). However, BMI is an imperfect surrogate for body composition and metabolic health. Future prospective studies incorporating direct measurements of adipokines (e.g., leptin and adiponectin), insulin resistance markers, and sarcopenia indices are warranted to validate the obesity paradox hypothesis and disentangle the relative contributions of adiposity, nutrition, and muscle mass to immunotherapy outcomes.

In our pembrolizumab-treated cohort with liver metastases, male patients showed a numerically longer OS than females, with a difference approaching statistical significance. Although this finding did not reach conventional levels of statistical significance and should be interpreted cautiously, it is consistent with emerging evidence suggesting that sex may influence the response to ICIs.

Biological differences between males and females, particularly in hormonal regulation and immune function, may contribute to the divergent immunotherapy outcomes. Estrogens and androgens differentially modulate T-cell activation, regulatory T-cell expansion, cytokine production, and immune checkpoint expression, thereby shaping the tumor immune microenvironment (35–37). In addition, sex-related differences in hepatic immune regulation may be especially relevant in the context of liver metastases, given the intrinsically tolerogenic role of the liver and its impact on systemic antitumor immunity.

Taken together, these observations suggest that sex may act as a potential modifier of immunotherapy efficacy in patients with liver metastases. However, given the limited statistical power of the subgroup analyses, our results should be regarded as hypothesis-generating. Future studies should systematically incorporate sex as a biological variable and explore the interaction effects between sex and metastatic sites in multivariable models.

Avelumab, a human IgG1 monoclonal antibody that binds PD-L1, received FDA approval in June 2020 for maintenance treatment of patients with mUC who have not progressed during first-line platinum-based chemotherapy (38). Approval was based on the positive results of the phase III JAVELIN Bladder 100 trial, which showed a significantly prolonged OS and progression-free survival (PFS) of avelumab versus Best Supportive Care alone in this patient population. There are no available data on efficacy outcomes in patients with liver metastases; in the trial, this characteristic was included in “visceral metastases” which were present in 55% of patients enrolled (39).

In ARON-2, 18% of patients had liver metastases, a slightly higher percentage compared to other real-world experiences, such as READY and AVENANCE, which included 12.8% and 15% of liver as metastatic sites, respectively. The READY trial reported only an HR of 2.85 (95%CI: 1.75–4.62) comparing OS in patients with liver metastasis versus those without, and no specific data have been reported in the AVENANCE trial (40, 41). In the avelumab-treated group in ARON-2, patients with liver metastases exhibited significantly lower OS than those without hepatic involvement (16.4 months vs. 27.0 months, p <0.001). However, in the subgroup of patients with liver metastases, a higher BMI and better ECOG-PS remained favorable prognostic factors.

Switch maintenance in mUC is a “new” validated treatment strategy, and there is a known positive selection of patients who benefit from it. In addition, to better understand whether immunotherapy can improve outcomes in patients with liver metastases in this setting, we may consider a few historical data. In 1999, Bajorin validated “the Bajorin prognostic factors” by analyzing five trials of the MVAC regimen, where 11.3% of patients with liver metastases were included, showing a mOS of 9.87 months vs. 15.45 months without (16). A Japanese group showed an mOS of 9 months vs. 17 months in the same population (42). ARON-2 showed that in patients with liver metastases treated with avelumab, an improvement in OS was achieved.

When interpreting the outcomes of both cohorts, it is essential to contextualize our findings with the historical chemotherapy results in patients with liver metastases. Prior studies reported markedly inferior outcomes in this subgroup, with median OS ranging between 5.6 months and 9.8 months in patients receiving vinflunine or MVAC-based regimens. In contrast, ARON-2 demonstrated a median OS of 9.4 months in post-platinum patients treated with pembrolizumab and 16.4 months in those receiving avelumab maintenance, suggesting that immune checkpoint inhibitors provide a clinically meaningful survival improvement even in this poor-prognosis population. Although liver involvement remains a strong adverse prognostic factor, these comparisons underscore that immunotherapy has shifted the therapeutic benchmark for patients with mUC with hepatic metastases. Our results complement those of Zacchi et al., who reported comparable OS between pembrolizumab and avelumab in a real-world population (20). Although our study did not aim to directly compare the two agents, the differential prognostic impact of liver metastases across cohorts highlights the importance of disease biology and treatment timing. Consistently inferior outcomes in patients with hepatic metastases, regardless of the immunotherapy used, underscore that liver involvement reflects a biologically aggressive disease subset rather than a lack of activity of ICIs.

Numerous studies have investigated the mechanisms by which liver metastases influence the effectiveness of immune checkpoint inhibitors (ICIs). Yoshida et al., suggest that hepatic microenvironment may promote immunosuppression through mechanisms involving myeloid cells and immunosuppressive factors, thereby limiting the efficacy of immunotherapy. Furthermore, the lower response observed could reflect reduced lymphocytic infiltration in liver metastases compared with other metastatic sites (43). Lee et al. demonstrated that liver metastases significantly impair tumor-specific immunity through an antigen-specific PD-1-dependent pathway. This process is associated with the coordinated activation of regulatory T-cells and alterations in intratumoral CD11b+ monocytes (44). Similarly, Tumeh et al. found that liver metastases were correlated with a reduced presence of CD8+ T-cells at the invasive margin of distant tumors (45). Furthermore, Yu et al. revealed that liver metastases recruit immunosuppressive macrophages that facilitate antigen-specific T cell apoptosis, leading to systemic depletion of T cells and diminished immunotherapy efficacy (19).

Despite the robustness of the collected data, this study has some limitations that should be considered. First, the retrospective nature of the analysis introduces potential biases, limiting the ability to establish definitive causal relationships between the variables. Additionally, variability in clinical practices across participating centers may have influenced the results, although the adoption of standardized data-collection criteria has mitigated this risk. The limited sample size of the sex-stratified subgroups, particularly among patients with liver metastases, may have reduced the statistical power to fully assess sex-specific differences in immunotherapy outcomes. Another limitation is the lack of detailed data on predictive biomarkers, such as PD-L1 expression levels and molecular characteristics, or concomitant treatments that may influence the response to immunotherapy. Finally, a limitation of the present analysis is the absence of systematic data on patients’ pre-existing hepatic conditions, such as fatty liver disease, cirrhosis, or chronic inflammatory liver disease. These conditions may influence both tumor biology and the tolerance or response to systemic therapy.

Although liver metastases remain a strong adverse prognostic factor, our findings suggest that ICIs provide clinically meaningful benefits compared to historical chemotherapy data. These results support the need for more individualized therapeutic strategies for this high-risk population.

Based on the present data, several hypotheses can be proposed. First, the integration of ICIs with loco-regional treatments for liver metastases, such as stereotactic radiotherapy or ablative techniques, may enhance systemic antitumor immunity by promoting antigen release and overcoming liver-mediated immune tolerance. Second, the combination of immunotherapeutic approaches, including dual immune checkpoint blockade or the addition of agents targeting immunosuppressive pathways (e.g., CTLA-4 or TGF-β inhibition), may represent rational strategies to improve outcomes in patients with hepatic involvement.

Prospective biomarker-driven studies are warranted to validate these approaches and refine treatment selection for patients with metastatic urothelial carcinoma and liver metastases.