Real-world outcomes of second-line pembrolizumab in urothelial carcinoma: a multicenter analysis from the Campania Oncological Network (ROC)

Abstract

Background

Platinum-based chemotherapy followed by second-line immunotherapy has been traditionally the standard of care for locally advanced or metastatic urothelial cancer (La/mUC). However, checkpoint inhibitors are remarkably reshaping the therapeutic landscape for urothelial cancer. Despite these advances, La/mUC remains a highly lethal disease with poor prognosis and limited therapeutic response. Moreover, discrepancies between clinical trial populations and real-world patients often result in different outcomes and toxicity profiles. In this context, oncology networks have become critical in generating Real-World Data (RWD), which can be translated into Real-World Evidence (RWE) to support clinical decision-making in everyday practice. We retrospectively analysed La/mUC patients treated with second-line pembrolizumab within the Campania Oncological Network (Rete Oncologica Campana, ROC).

Patients and methods

This multicenter retrospective study included adult patients (≥ 18 years) with histologically or cytologically confirmed La/mUC, who had previously received chemotherapy and were treated with Pembrolizumab (200 mg every three weeks) across six ROC centres. Primary endpoints were progression-free survival (PFS) and overall survival (OS). Secondary endpoints included objective response rate (ORR), disease control rate (DCR) and safety.

Results

Between January 2021 and November 2023, 132 patients with La/mUC received at least one Pembrolizumab dose. The median age was 67 years (range 30–88), and most were male (73.5%). The majority had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 (20.5%) or 1 (62.1%). Histologically, 115 patients (87.1%) had pure urothelial carcinoma, while 17 (12.9%) had rare urothelial carcinoma subtypes, including squamous, sarcomatoid/carcinosarcoma, micropapillary or nested variants. At the time of initial diagnosis, 45 (34.1%) patients presented with metastatic disease (stage IV), and 72 (54.5%) had undergone radical cystectomy. Metastatic sites included lymph nodes 98 (74.2%), lung 51 (38.6%), bones 45 (34.1%), and liver 21 (15.9%). Liver metastases were more frequent in pure urothelial carcinoma (17.4% vs. 5.9%), while non-urothelial variants showed a higher incidence of peritoneal carcinomatosis (23.5% vs. 1.7%). After a median follow-up of 20 months (95% CI 12.7–32.6), 108 patients (81.8%) experienced disease progression or death. Median PFS was 3.75 months (95% CI: 3.4 to 4.7), while median OS was 7.3 months (95% CI: 6.05–9.33). The ORR was 13.5% (95% CI: 7.4% − 19.7%), DCR was 33.9% (95% CI: 25.6%-42.0%) (driven by a 20.4% stable disease rate). Among 17 patients with rare subtypes of urothelial carcinoma, the ORR was 23.5% (95% CI: 3.3% to 43.7%).

Multivariate analysis revealed that metastatic disease at diagnosis (HR = 2.01, p = 0.02) and liver metastases (HR = 2.11, p = 0.02) were independently associated with increased mortality risk (Table 3B). Lymph node-only metastases had a significantly lower risk of disease progression (HR = 0.46, p = 0.004) and death (HR = 0.52, p = 0.02). (). Univariate analysis indicated that prior cystectomy was associated with improved PFS (HR = 0.67, p = 0.04) and OS (HR = 0.54, p = 0.003), suggesting its role as a favourable prognostic factor in this cohort [1] (Table 2 A and 2B).

A total of 114 treatment-related AEs were reported, with 79.8% being grade 1–2 and 20.2% grade 3–4. The most common events were asthenia (20.0%), pruritus (10.0%), and myalgia (7.0%). No treatment-related deaths occurred. Pembrolizumab was discontinued in 7.6% of patients due to AEs. The safety profile was manageable; 7.6% of patients discontinued treatment due to AEs (predominantly drug-induced pneumonitis), and 1.5% required permanent corticosteroid replacement for immune-related adrenal insufficiency.

Conclusions

This real-world analysis supports the clinical applicability of second-line Pembrolizumab in La/mUC, demonstrating efficacy and acceptable safety. Prognostic indicators, including metastatic sites and prior cystectomy, significantly impacted outcomes. These findings highlight the value of oncology networks in generating real-world evidence to refine treatment approaches.

Introduction

Bladder cancer (BC) is the ninth most commonly diagnosed cancer worldwide, with 614,298 new cases and an age-standardised incidence rate (ASR) of 5.6 in 2022 [2]. The incidence rises with age, and it is four times higher in men than in women [3]. In Europe, ASRs are 60.5 per 100,000 in men and 14.1 in women, with 165,684 new cases [4]. Italy reports higher rates: 84.8 in men and 18.5 in women (34,850 cases), with Campania showing the highest regional incidence (ASR 88.4 in men, 15.6 in women). Alongside lung and liver cancers, BC is one of the most common cancers in Campania [5].

Globally, BC ranks 13th in cancer mortality, with 220,596 deaths in 2022 and an ASR of 1.81 [2]. Europe reported 67,289 deaths (ASR 17.8 in men, 3.7 in women) [4], while Italy recorded 6,108 deaths (ASR 18.0 in men, 3.4 in women). Campania again exceeded national averages. The five-year survival rate is 79% in Italy and 76.5% in Campania [5].

Smoking remains the most significant modifiable risk factor, contributing to over 50% of cases [6]. Other established risks include poor diet, inactivity, exposure to aromatic amines, schistosomiasis, hair dyes, and water contaminants [7]. Additional factors are adverse drug reactions, geographic differences, and socioeconomic status [8, 9].

Up to 95% of BCs are urothelial carcinomas (UCs), with subtypes such as conventional UC, micropapillary, and squamous variants, the latter associated with worse prognosis [10, 11]. Urothelial carcinomas of the upper urinary tract (UTUC) represent 5–10% of cases, with a poorer prognosis than bladder UC. Standard of care for early UTUC includes radical nephroureterectomy (RNU), followed by intravesical or systemic chemotherapy depending on pathological staging [12].

Approximately 70% of BC cases are diagnosed as non-muscle-invasive bladder cancer (NMIBC), limited to the mucosa [8]. NMIBC is stratified into risk categories based on features such as size, grade, multifocality, and presence of carcinoma in situ [13–15]. Treatment consists of transurethral resection (TURB) and risk-adapted intravesical therapy.

Muscle-invasive bladder cancer (MIBC) accounts for about 25% of UC cases.

For cisplatin-eligible patients (based on Galsky’s criteria [16]), the standard treatment is neoadjuvant chemotherapy with Durvalumab and Cisplatin, followed by radical cystectomy and lymphadenectomy, and maintenance Durvalumab [17]. Adjuvant Nivolumab or cisplatin-based chemotherapy may be considered in non-perioperative treated, high-risk patients (pT3, pT4, or N+) [15]. Recently, the CheckMate-274 trial showed that 1 year of adjuvant Nivolumab improved disease-free and overall survival, especially in PD-L1–positive patients [18, 19].

Furthermore, the peri-operative landscape is rapidly shifting. Recent data from the EV-302 trial in the metastatic setting has now translated into the perioperative space. Specifically, recent data have highlighted the potential of EV plus pembrolizumab in both cisplatin-eligible and ineligible patients, potentially redefining the standard of care before and after radical surgery [20].

While < 5% of BCs are metastatic at diagnosis, up to 50% of patients treated for MIBC relapse within two years [21].

The treatment landscape for metastatic UC (mUC) has evolved significantly. First-line standards now include Enfortumab Vedotin (EV) plus Pembrolizumab, platinum-based chemotherapy followed by avelumab maintenance, or the combination of Nivolumab with Cisplatin and Gemcitabine for cisplatin-eligible patients [22].

For years, no global standard existed after platinum failure. Agents such as paclitaxel, docetaxel [23, 24], and vinflunine [25, 26] were commonly used. The KEYNOTE-045 trial established Pembrolizumab as superior to chemotherapy in overall survival (10.3 vs. 7.4 months; HR 0.73) and response rate (21% vs. 11%), with fewer grade 3–4 toxicities [27, 28]. A 5-year follow-up confirmed durable benefit [29]. Based on this, the FDA and EMA approved Pembrolizumab as a second-line treatment after platinum failure, independently of PD-L1 status.

Despite its efficacy, Pembrolizumab introduces a new spectrum of immune-related toxicities that require proactive management. Furthermore, clinical trials often exclude patients with poor performance status, comorbidities, or prior malignancies [30], limiting generalizability.

This study aims to evaluate real-world outcomes of Pembrolizumab in platinum-refractory UC, with a focus on efficacy and safety.

Patients and methods

Study population

A retrospective study was conducted within the ROC, promoted by the Istituto Nazionale Tumori—IRCCS—Fondazione “G. Pascale” in Naples, Italy, and approved by the local Ethics Committee. The study was performed in accordance with the revised version of the Declaration of Helsinki (52nd WMA General Assembly, Edinburgh, Scotland, October 2000). Data were collected for patients aged 18 years or older, with a histological diagnosis of metastatic urothelial carcinoma who received Pembrolizumab 200 mg every 3 weeks, up to unacceptable toxicity or disease progression, after first-line platinum-based chemotherapy or recurring within < 12 months since adjuvant or neoadjuvant chemotherapy. All patients provided written informed consent.

Evaluation of outcomes

RECIST 1.1 criteria were used to assess tumour response as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD). The following parameters were collected: causes of treatment discontinuation, comorbidities, progression-free survival (PFS), overall survival (OS), objective response rate (ORR, defined as the sum of CR and PR) and treatment-related toxicities. Adverse events (AEs) were graded based on the Common Terminology Criteria for Adverse Events (CTCAE), version 4.0.

Statistical analysis

Demographic and clinical data were analysed using descriptive statistical analysis. Categorical variables were presented in terms of frequencies and proportions, either the means or standard deviation or median and interquartile ranges (IQRs) according to their distribution, as assessed by the Shapiro-Wilk normality test, were used for continuous variables.

To assess patient prognosis, we used Progression-Free Survival (PFS) and Overall Survival (OS) as primary outcomes. The inverse Kaplan-Meier method was used to estimate the median follow-up time. PFS was calculated from the start of Pembrolizumab treatment until evidence of PD or death, whichever occurs first. OS was calculated from the start of Pembrolizumab treatment to death or censoring at the last follow-up. Survival times were analysed using the Kaplan–Meier method. ORR and DCR were reported as percentages. A multivariate Cox regression model, based on baseline clinical characteristics of patients (Histology, Gender, Sites of Metastases, and Body Mass Index (BMI)), yielded adjusted Hazard Ratios (HRs).

To assess the prognosis according to Response to Pembrolizumab, patients who exhibited CR or PR as best response were considered as Responders, patients who exhibited SD or PD were considered as Non-Responders and PFS and OS were used as outcomes.

The differences in prognosis between responders and non-responders were tested using the log-rank test, and the results were displayed using Kaplan-Meier curves. The analyses were performed using the software R (version 4.3.3).

Ethical approval

This study was conducted in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Results

From January 2021 to November 2023, 132 patients treated with at least one dose of pembrolizumab were enrolled across six centres in the Campania Oncology Network.

Baseline characteristics are summarised in Table 1.

Table 1.

Baseline characteristics of the study population (N = 132) Summary of baseline demographic, clinical and pathological features of the enrolled patients

Characteristic	N (%) or Median (Range)
Age, years	67.0 (30.0–89.0)
≤ 65 years	58 (43.9%)
> 65 years	74 (56.1%)
Sex
Male	97 (73.5%)
Female	35 (26.5%)
Comorbidities
Hypertension	52 (39.4%)
Cardiovascular disease	31 (23.5%)
Diabetes	18 (13.6%)
Metabolic disease	13 (9.8%)
Kidney disease	6 (4.5%)
Lung disease	5 (3.8%)
Previous hepatitis	4 (3.0%)
Cystectomy performed	72 (54.5%)
Histology
Urothelial	115 (87.1%)
Minor/mixed variants	17 (12.9%)
Metastatic disease	45 (34.1%)
Site of metastasis
Lymph nodes	98 (74.2%)
Lung	51 (38.6%)
Liver	21 (15.9%)
Bones	45 (34.1%)
Peritoneal carcinomatosis	7 (5.3%)
Brain	6 (4.5%)
Body Mass Index (BMI)
≤ 25 kg/m²	44 (33.3%)
> 25 kg/m²	76 (57.6%)
Missing	12 (9.1%)
ECOG Performance Status
ECOG 0	25 (18.9%)
ECOG ≥ 1	51 (38.6%)
Unknown	56 (42.4%)

Median age was 67 years (IQR 30–89), 73.5% were male, and 38.6% had ECOG ≥ 1. Pure urothelial carcinoma was present in 115 patients (87.1%). Among the 17 patients (12.9%) with non-pure or variant histology, the most frequent subtypes were squamous differentiation (n = 7; 5.3%), followed by undifferentiated sarcomatoid carcinoma (n = 4; 3.0%). Other rare variants included the micropapillary (n = 3; 2.3%) and nested (n = 3; 2.3%) variants. (Table 2).

Table 2.

Histological distribution of the study cohort

Hystology	N	%
Urothelial carcinoma	115	87,1%
Urothelial carcinoma with squamous differentiation	7	5,3%
Undifferentiated sarcomatoid	4	3,0%
Micropapillary variant	3	2,3%
Nested variant	3	2,3%

Comparison between histological groups revealed that patients with non-urothelial variants were generally younger (median age 60.0 vs. 68.0 years) and had a higher rate of cystectomy (70.6% vs. 52.2%). (Table 3).

Table 3.

Patient characteristics stratified by histological type (Urothelial vs. Non-Urothelial)

Characteristic	Urothelial (n = 115)	Non-urothelial (n = 17)
Age, years	68 (44–89)	60.0 (30–76)
≤ 65 years	46 (40.0%)	12 (70.6%)
> 65 years	69 (60.0%)	5 (29.4%)
Sex
Male	85 (73.9%)	12 (70.6%)
Female	30 (26.1%)	5 (29.4%)
Comorbidities
Hypertension	44 (38.3%)	8 (47.1%)
Cardiovascular disease	27 (23.5%)	4 (23.5%)
Diabetes	17 (14.8%)	1 (5.9%)
Metabolic disease	10 (8.7%)	3 (17.6%)
Kidney disease	6 (5.2%)	0 (0%)
Lung disease	5 (4.3%)	0 (0%)
Previous hepatitis	3 (2.6%)	1 (5.9%)
Cystectomy performed
	60 (52.2%)	12 (70.6%)
Site of metastasis
Lymph nodes	87 (75.7%)	11 (64.7%)
Lung	47 (40.9%)	4 (23.5%)
Liver	20 (17.4%)	1 (5.9%)
Bones	39 (33.9%)	6 (35.3%)
Peritoneal carcinomatosis	2 (1.7%)	4 (23.5%)
Brain	6 (5.2%)	0 (0%)
Body Mass Index (BMI)
≤ 25 kg/m²	37 (32.2%)	7 (41.2%)
> 25 kg/m²	69 (60.0%)	7 (41.2%)
Missing	9 (7.8%)	3 (17.6%)
ECOG Performance Status
ECOG 0	21 (18.3%)	4 (23.5%)
ECOG ≥ 1	42 (36.5%)	9 (52.9%)
Unknown	52 (45.2%)	4 (23.5%)

At diagnosis, 45 patients (34.1%) had stage IV disease. The most frequent metastatic sites were lymph nodes (74.2%), lung (38.6%), bone (34.1%), and liver (15.9%).

Liver metastases were significantly more frequent in the pure urothelial group compared to non-urothelial variants (17.4% vs. 5.9%). In contrast, the incidence of bone metastases remained remarkably consistent between the two cohorts (33.9% vs. 35.3%), suggesting that skeletal involvement is a common feature of metastatic progression in this disease, regardless of histological subtype. Remarkably, the non-urothelial group showed a higher incidence of peritoneal carcinomatosis compared to the pure urothelial group (23.5% vs. 1.7%, respectively).

Disease progression (PD) was the primary cause of treatment discontinuation (74.2%), followed by adverse events (AEs) in 7.6%.At the time of analysis, 24 patients (18.9%) were still undergoing treatment. Among those who discontinued Pembrolizumab, 23 patients received subsequent anticancer therapies as follows: 10 patients received Paclitaxel, 7 patients started Vinflunine, 2 patients initiated Enfortumab Vedotin, 2 patients began Erdafitinib, 1 patient received Docetaxel, and 1 patient received Carboplatin.Median follow-up, calculated via reverse Kaplan-Meier, was 20 months (95% CI 12.7–32.6). Overall, 108 patients (81.8%) experienced progression or death. Median PFS was 3.75 months (95% CI: 3.38–4.73), and median OS was 7.29 months (95% CI: 6.05–9.33) (Fig. 1A and B). ORR was 13.5% (95% CI: 7.4–19.7), including 3 complete responses and 13 partial responses. Disease control rate (DCR) was 33.9% (95% CI: 25.6–42.0) (Table 4). Subgroup analysis by histology showed an ORR of 10.8% and DCR of 31.5% in pure urothelial carcinoma (n = 115), compared to 23.5% and 29.4% in patients with mixed/variant histologies (n = 17), respectively.

Fig. 1.

Progression free survival and overall survival

Table 4.

Response outcomes in the overall population and by histological subtype. Distribution of best tumour response and disease control following pembrolizumab treatment in the overall cohort and by histological subtype

Outcome	Overall (n = 132)	Urothelial (n = 115)	Minor or mixed variants (n = 17)
ORR, % (95% CI)	13.5% (7.4%–19.7%)	10.8% (5.0%–16.6%)	23.5% (3.3%–43.7%)
DCR, % (95% CI)	33.9% (25.6%–42.0%)	31.5% (23.0%–40.0%)	29.4% (7.8%–51.0%)
CR, n (%)	3 (2.5%)	2 (1.8%)	1 (5.9%)
PR, n (%)	13 (11.0%)	10 (9.0%)	3 (17.6%)
SD, n (%)	24 (20.3%)	23 (20.7%)	1 (5.9%)
PD, n (%)	78 (66.1%)	66 (59.4%)	12 (70.6%)
Not evaluable	14	14	0

Univariate analysis identified cystectomy as protective for both PFS (HR 0.67; p = 0.04) and OS (HR 0.54; p = 0.003), while metastatic disease at diagnosis was associated with worse outcomes (PFS: HR 1.79; p = 0.03; OS: HR 2.23; p < 0.005) (Tables 5 and 6, Figs. 2 and 3).

Table 5.

Univariate and multivariate Cox regression analysis PFS. Association between baseline clinical variables and progression-free survival as assessed by univariate and multivariate Cox regression analysis

	HR Univ. (95% CI)	p Univ.	HR Multiv. (95% CI)	p Multiv.
Age >65 yrs.	1.01 (0.69-1.48)	0.95	-	-
Gender: Male	1.15 (0.66–2.03)	0.08	-	-
N. of Comorbidities ≥1	1.01 (0.68-1.51)	0.94	-	-
Cystectomy performed	0.67 (0.46-0.98)	0.04	0.92 (0.58-1.48)	0.79
Histology: Pure urothelial carcinoma	1.21 (0.67-2.15)	0.52	-	-
BMI ≤ 25 kg/m²	1.37 (0.91-2.05)	0.13	-	-
Brain metastases	0.55 (0.22-1.38)	0.20	-	-
Bone metastases	1.40 (0.94-2.07)	0.09	-	-
Lung metastases	1.27 (0.86-1.89)	0.22	-	-
Lymph nodal metastases	0.77 (0.50-1.19)	0.24	0.46 (0.27-0.78)	0.004
Liver metastases	1.14 (0.69-1.90)	0.59	1.63 (0.92-2.90)	0.09
Carcinosis	1.44 (0.58-3.57)	0.79	-	-
Metastatic disease at diagnosis	1.79 (1.21-2.64)	0.03	1.69 (0.99-2.89)	0.053

Table 6.

Univariate and multivariate Cox regression analysis of OS. Association between baseline clinical variables and overall survival as assessed by univariate and multivariate Cox regression analysis

	HR Univ. (95% CI)	p Univ.	HR Multiv. (95% CI)	p Multiv.
Age > 65 yrs.	1.13 (0.75–1.71)	0.54	-	-
Gender: Male	1.23 (0.79–1.96)	0.39	-	-
N. of Comorbidities ≥ 1	0.79 (0.52–1.20)	0.28	-	-
Cystectomy performed	0.54 (0.36–0.81)		0.72 (0.43–1.20)	0.21
Histology: Pure urothelial carcinoma	1.39 (0.72–2.69)	0.32	-	-
BMI ≤ 25 kg/m²	0.88 (0.56–1.39)	0.60	-	-
Brain metastases	0.47 (0.17–1.29)	0.14	-	-
Bone metastases	1.33 (0.84–1.93)	0.18	-	-
Lung metastases	1.28 (0.86–1.89)	0.24	-	-
Lymph nodal metastases	0.74 (0.48–1.67)	0.20	0.52 (0.29–0.92)	0.02
Liver metastases	1.48 (0.86–2.54)	0.15	2.11 (1.13–3.92)
Carcinosis	0.93 (0.29–2.97)	0.91	-	-
Metastatic disease at diagnosis	2.23 (1.47–3.37)		2.01 (1.11–3.66)

Fig. 2.

Multivariate Cox regression analysis PFS

Fig. 3.

Multivariate Cox regression analysis of OS

Although not statistically significant, male sex (HR 1.15; p = 0.08) and bone metastases (HR 1.40; p = 0.09) showed trends toward worse PFS. In multivariate analysis (Tables 5 and 6), metastatic disease at diagnosis was associated with borderline worse PFS (HR 1.69; p = 0.053) and significantly reduced OS (HR 2.01; p = 0.02). Liver metastases were independently associated with poor OS (HR 2.11; p = 0.02), while lymph node metastases were linked to improved outcomes (PFS: HR 0.46; p = 0.004; OS: HR 0.52; p = 0.02).

Overall, 114 any grade adverse events occurred, where 91 (79.8%) were G1-G2, and 23 (20.2%) were G3-G4 AEs. No treatment-related AEs resulted in death. Ten (7.6%) patients discontinued pembrolizumab because of treatment-related AEs. Among these, there were 6 cases of drug-induced pneumonitis (4 cases of Grade 2 and 2 cases of Grade 3). Additionally, 3 patients developed symptomatic hypothyroidism and 2 patients presented with adrenal insufficiency.

The most common treatment-related AEs were asthenia (20.0%), itching (10.0%), and myalgias (7.0%). As shown in Fig. 4, the overall safety profile was characterised by a predominance of grade 1–2 adverse events, with no treatment-related deaths.

Fig. 4.

Adverse events

Regarding toxicity management, systemic steroid administration was required in 7.6% (n = 10) of patients to manage Grade 2 and 3 immune-related AEs (irAEs). Of these, 2 patients (1.5%) required permanent corticosteroid replacement therapy due to treatment-induced adrenal insufficiency.

Discussion

Our retrospective, multicenter real-world study in the Campania Oncology Network (ROC) represents a real-world evaluation that adds to the growing body of evidence supporting pembrolizumab as an effective and well-tolerated second-line therapy in patients with advanced or metastatic urothelial carcinoma (UC).

This work is essential because real-world clinical cohorts tend to include patients who are more heterogeneous and fragile than those typically enrolled in rigorous registration trials.

In this unselected population, our study reported a median overall survival (OS) of 7.29 months and a median progression-free survival (PFS) of 3.75 months.

The numerically lower efficacy compared to the data from the Phase III KEYNOTE-045 trial [28, 29], which led to the drug’s approval, reflects the nuanced differences in our regional population characteristics and broader patient eligibility. In comparison, the pivotal phase III KEYNOTE-045 trial [28, 29] reported a median OS of 10.3 months and PFS of 2.1 months and an Objective Response Rate (ORR) of 21% compared to 13.5% in our study. This difference reflects real-world conditions.

Our cohort included a significant proportion of patients with a poorer ECOG performance status (≥ 1 in 38.6% of our patients), an advanced median age of 67 years, a higher proportion of patients with visceral metastases (including liver: 15.9% and bone: 34.1%), and fewer exclusions based on comorbidities or prior therapies. These factors are often excluded or underrepresented in clinical trials but are known predictors of a less favourable prognosis. Our study, therefore, does not negate the drug’s efficacy but realistically defines its value in a more fragile and representative population.

Yet the study’s disease control rate (DCR) of 33.9% supports pembrolizumab’s role in stabilising disease, particularly in a less-selectively recruited population.The observed DCR of 33.9% is primarily driven by a significant proportion of patients (n = 27, 20.4%) achieving stable disease (SD) as their best response according to RECIST 1.1. The inclusion of SD in the Disease Control Rate (DCR) reflects a subset of patients who achieve a plateau in disease burden. In immunotherapy trials, including the pivotal KEYNOTE-045, it has been observed that patients achieving SD can derive a survival benefit compared to those with progressive disease, suggesting that disease stabilization may represent a clinically relevant state of immune-mediated equilibrium [27].

Our findings also compare consistently with the large-scale global ARON-2 study by Massari et al. [31], which included 836 patients across 23 countries and reported a median OS of 10.5 months and PFS of 6.2 months. Their ORR was notably higher at 31%, possibly reflecting a more heterogeneous global cohort and inclusion of patients relapsing post-adjuvant/neoadjuvant therapy (cohort B), who showed better outcomes than those progressing during first-line treatment (14.6 vs. 9.1 months OS, respectively). In our cohort, all patients were treated in a strictly platinum-refractory setting, likely aligning more closely with ARON-2’s cohort A, where the ORR was 29% and OS was 9.1 months.

The prognostic factor analysis yielded critical insights for risk stratification and treatment optimisation. Metastatic disease at diagnosis, liver, and bone metastases were associated with worse outcomes, consistent with both Santoni et al. and ARON-2 findings [32].

Liver metastatic involvement emerged as the most powerful independent predictor of negative outcome in the multivariate analysis (OS: HR 2.11; p = 0.02), consistent with both Santoni et al. [32]and the ARON-2 findings [31]. This finding supports the biological notion that liver metastases often indicate a disease with higher aggressiveness and a particularly immunosuppressive tumour microenvironment, which hinders the efficacy of checkpoint inhibitors.

In contrast, primary lymph node involvement resulted in a significant protective factor for outcomes (PFS: HR 0.46; OS: HR 0.52). This result suggests that disease predominantly limited to the lymph node compartment, compared to visceral disease, has a less aggressive biology, and perhaps it is a more fertile ground for triggering an effective immune response. Lymph nodes, being key sites for immune activation, may facilitate antigen presentation and the success of immunotherapy. This subgroup, therefore, represents the population that derives the greatest and most durable benefit from pembrolizumab therapy.

Our data highlights a divergent metastatic behavior: while bone involvement represents a common clinical burden across all histologies, liver metastases appear more characteristic of pure urothelial carcinoma. Conversely, non-urothelial variants show a tropism for the peritoneum (23.5%)., suggesting a more ‘sarcomatous’ or ‘epithelial-to-mesenchymal’ spread pattern rather than the standard hematogenous route. This highlights the need for dedicated clinical trials for non-pure histologies.

The numerically higher ORR in patients with mixed/variant histologies (23.5%) compared to pure urothelial carcinoma (10.8%), although in a limited cohort, is consistent with the hypothesis that these variants may harbour a higher tumour mutational burden (TMB), making them particularly targetable by immunotherapy.

However, given the small sample size and the absence of a significant survival advantage, this observation should be considered hypothesis-generating and requires validation in larger studies.”

Prior radical cystectomy seems to be a significant protective factor (OS: HR 0.54). This likely reflects a patient selection with originally less advanced disease and better general health, allowing for a longer follow-up.

Importantly, the ARON-2 study [31]developed a prognostic scoring system based on ECOG ≥ 2, liver/bone metastases, and line of pembrolizumab therapy. Patients with all adverse factors had a median OS of just 4.1 months, compared to 29.4 months in those without any. While our study did not formally stratify patients using this model, our data showed similar trends, highlighting the urgent need for predictive tools to guide immunotherapy use more precisely in clinical practice.

In our study, the observed safety profile was consistent with known data, with the majority of adverse events (AEs) classified as Grade 1–2. The low rate of treatment discontinuation due to AEs (7.6%) confirms that, in an experienced oncology centre, immune-mediated toxicities (irAEs) are manageable.

The documentation of subsequent therapies after pembrolizumab failure is also important. The predominance of standard chemotherapies (paclitaxel and vinflunine) still reflects the standard of practice during the enrollment period, particularly in regions such as ours, where third-line agents like EV were limited during data collection.

The emergence of innovative-targeted treatments like enfortumab vedotin and erdafitinib indicates that the therapeutic landscape is rapidly evolving. Moreover, the recent establishment of EV plus pembrolizumab as a first-line standard of care (EV-302/KEYNOTE-A39) significantly impacts second-line strategies. Interestingly, our observation of a numerically higher ORR in variant histologies (23.5%) might suggest that these aggressive subtypes may derive substantial benefit from pembrolizumab-based combinations even in earlier lines of therapy. As EV plus pembrolizumab moves to the first-line setting, understanding the specific sensitivity of mixed and variant histologies to this combination remains a priority for future research.

The paradigm shift introduced by the EV-302/KEYNOTE-A39 trial, which established Enfortumab Vedotin plus Pembrolizumab as the new first-line standard of care, inevitably reduces the future frequency of single-agent pembrolizumab use in the second-line setting.

Our study provides a pivotal ‘historical’ benchmark of pembrolizumab performance in a real-world, platinum-refractory population, These data will become increasingly difficult to prospectively collect as immunotherapy moves to earlier lines. Understanding these outcomes is essential for interpreting future sequencing strategies, especially for patients who still receive frontline chemotherapy due to regional reimbursement policies, contraindications to EV, or resource-limited settings.

Despite the importance of the real-world data generated, this study has limitations inherent to its retrospective design, potential selection bias, and lack of centralised radiological review.

The relatively small sample size, particularly for patients with variant histologies or subsequent therapies, also limits the generalizability of subgroup analyses. However, the multicenter design and representation of a real-world population (often underrepresented in randomised controlled trials) strengthen its relevance.

Although the median follow-up duration was calculated using the reverse Kaplan-Meier method (20.0 months), the variability across the six centres and the relatively short enrollment period may lead to heterogeneity in the follow-up data.

Analyses conducted on specific subgroups, particularly those referred to the variant histology (n = 17) or various subsequent treatments, are limited by the small number of patients. Conclusions drawn from these subgroups, such as the higher ORR in variant histologies, must be considered as hypothesis generating and require validation in larger cohort studies.

Despite these limitations, the strength of our study lies in its ability to capture the daily clinical reality of a large Italian region, providing essential survival data and identifying valuable prognostic factors for the management of platinum-refractory metastatic urothelial carcinoma.

Conclusions

This regional real-world study confirms the clinical utility of pembrolizumab as a second-line treatment for platinum-refractory UC, including in patients with poor performance status and variant histologies. Prognostic factors such as liver and bone metastases, ECOG status, and surgical history should guide treatment decisions. In light of evolving standards such as EV + pembrolizumab in first-line therapy, continued generation of real-world data is crucial to contextualise these developments and inform practice, especially in resource-limited settings.

Ultimately, this real-world study provides the basis for designing prospective studies focused on tumour biology, predictive biomarkers, and optimal therapeutic sequences to maximise survival across all subgroups of mUC patients.

Abbreviations

La/mUC

Locally advanced or metastatic urothelial carcinoma

ROC

Campania Oncological Network (Rete Oncologica Campana)

PFS

Progression-free survival

OS

Overall survival

ORR

Objective response rate

DCR

Disease control rate

ECOG

Eastern Cooperative Oncology Group

AE

Adverse event

CTCAE

Common Terminology Criteria for Adverse Events

Authors’ contributions

Conceptualization: Rosa Tambaro; Methodology: Rosa Tambaro; Data curation: Elisabetta Coppola, Davide Limongello; Investigation: Rosa Tambaro; Formal analysis: Elisabetta Coppola; Writing – Original Draft: Rosa Tambaro; Writing – Review & Editing: Rosa Tambaro, Gabriele Calvanese, Andrea Muto, Luigi Formisano, Giuseppe Lorenzo, Carlo Buonerba, Davide Bosso, Francesco Sabbatino, Elisabetta Coppola, Marilena Napoli, Sabrina Rossetti, Carmela Pisano, Sabrina Chiara Cecere, Anna Passarelli, Jole Ventriglia, Lorenzo Lobianco, Davide Limongello, Erica Perri, Mariarosaria Lamia, Chiara D’Alessio, Roberto Contieri, Florinda Feroce, Sandro Pignata; Project administration: Rosa Tambaro; Resources: Davide Limongello; Software: Elisabetta Coppola; Visualization: Davide Limongello; Validation: Rosa Tambaro; Supervision: Sandro Pignata.All authors have read and approved the final manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Istituto Nazionale Tumori IRCCS “Fondazione G. Pascale” of Naples (registration number 32/22 oss, 19 Oct 2022).

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.