Incidence and risk factors for pneumonitis in urothelial carcinoma treated with enfortumab vedotin plus pembrolizumab: retrospective study

Abstract

Background

Enfortumab vedotin plus pembrolizumab (EVP) is first-line standard therapy for locally advanced or metastatic urothelial carcinoma (LA/mUC), but real-world pneumonitis incidence and predictors remain poorly characterized.

Material and methods

We conducted a retrospective cohort study of patients with LA/mUC treated with EVP at MD Anderson Cancer Center. Pneumonitis was adjudicated by pulmonology and graded using CTCAE v5.0. Prior immune checkpoint inhibitor (ICI) exposure was defined as ICI treatment before EVP initiation. Risk factors were assessed using Cox proportional hazards and Fine-Gray competing-risk regression; associations with overall survival (OS) and progression-free survival (PFS) were evaluated using extended Cox models.

Results

Pneumonitis occurred in 22 patients (7.0%), with median onset at 77 days; 10 events (45%) were grade ≥3, including 1 fatal case. Prior ICI exposure was the strongest independent predictor of pneumonitis (any-grade adjusted HR 2.68; 95% CI, 1.03-6.98; P = .04; grade ≥2 adjusted HR 5.91; 95% CI, 1.94-18.02; P = .002). No other baseline variable was significantly associated with pneumonitis. Grade ≥2 pneumonitis was associated with increased mortality (HR 2.35; 95% CI, 1.02-5.45; P = .04), whereas any-grade pneumonitis was not. Twelve patients underwent treatment re-exposure; all had no clinically meaningful recurrent or worsening pneumonitis during subsequent evaluable follow-up, with 8 resuming EV plus pembrolizumab and 4 EV alone.

Conclusion

Pneumonitis occurred in 7.0% of EVP-treated patients and generally arose early. Prior ICI exposure predicted pneumonitis, particularly grade ≥2 events, which were associated with increased mortality; any-grade pneumonitis and PFS were not.

Implications for Practice.

Enfortumab vedotin plus pembrolizumab is increasingly used as first-line therapy for advanced urothelial carcinoma, yet real-world pneumonitis risk remains incompletely defined. In this cohort, pneumonitis occurred in 7% of patients and was frequently severe. Prior immune checkpoint inhibitor exposure identified patients at substantially higher risk, particularly for symptomatic grade ≥2 pneumonitis, which was associated with increased mortality. These findings support closer respiratory monitoring in patients with prior ICI exposure and careful, individualized treatment re-exposure after pneumonitis; among 12 selected patients re-exposed, no clinically meaningful recurrent or worsening pneumonitis was observed.

Introduction

Genitourinary (GU) malignancies account for a substantial global cancer burden, urothelial carcinoma is the most common malignancy of the urinary tract and a leading cause of cancer-related morbidity and mortality worldwide.¹ Despite initial sensitivity to platinum-based chemotherapy, the majority of patients with locally advanced or metastatic urothelial carcinoma (LA/mUC) experience disease progression, and historically, median overall survival in this population has remained limited.² The development of novel therapeutic agents, including immune checkpoint inhibitors and antibody-drug conjugates (ADCs), has substantially altered the treatment landscape for advanced urothelial carcinoma in recent years.²

Enfortumab vedotin (EV) is a first-in-class ADC that delivers the microtubule inhibitor monomethyl auristatin E (MMAE) directly to nectin-4 expressing tumor cells.³ Pembrolizumab is a highly active monoclonal antibody targeting the programmed cell death protein 1 (PD-1) immune checkpoint.⁴ Both agents have individually shown significant antineoplastic activity in advanced urothelial carcinoma (UC).^5–7 In the phase III EV-302/KEYNOTE-A39 trial, the combination of EV with pembrolizumab (EVP) demonstrated superior overall survival and progression-free survival (PFS) compared with platinum-based chemotherapy, leading to FDA approval as first-line therapy for LA/mUC.^8,9

Both immune checkpoint inhibitors (ICIs) and ADCs carry independent risks of pulmonary toxicity. ICI-related pneumonitis is a well-characterized immune-related adverse event that can be life-threatening, occurring in approximately 3%-5% of patients receiving anti-PD-1 monotherapy across tumor types.¹⁰ EV monotherapy has been associated with pneumonitis in fewer than 2% of patients in pivotal trial settings, although a real-world retrospective analysis from South Korea reported substantially higher rates, suggesting that the true incidence may be underestimated in controlled settings.^11,12 In the EV-302 trial, the combination of EV with pembrolizumab was associated with a pneumonitis rate of approximately 10%, with grade ≥3 events occurring in approximately 4% of patients.^8,9 The mechanisms underlying this increased risk with combination therapy have not been fully elucidated, and both overlapping and potentially additive pulmonary toxicity pathways have been proposed.¹¹

Despite the increasing use of EVP in advanced UC, real-world data characterizing pneumonitis incidence, clinical course, and risk factors in this population remain limited. In particular, it is unclear whether established risk factors for ICI-related pneumonitis—such as preexisting lung disease, smoking history, and prior thoracic radiation13—are applicable to patients receiving EVP. In this retrospective cohort study, we sought to determine the incidence, timing, severity, and mortality impact of pneumonitis, and to identify clinical predictors of pneumonitis risk, in patients with advanced UC treated with EVP at a single academic center.

Material and methods

Study participants and clinical data collection

We conducted a retrospective cohort study of patients with GU malignancies treated with EVP between October 2023 and December 2025 at MD Anderson Cancer Center. Sequential initiation—defined as initiation of 1 agent followed by addition of the second at a later date—was permitted, provided both agents were concurrently administered as part of the EVP regimen. Patients with penile cancer were excluded due to distinct disease biology and treatment paradigms compared with other GU malignancies.

We collected clinical, imaging, and laboratory data from the electronic health record, including demographics, cancer subtype, disease stage, metastatic status, lung metastases, ECOG performance status, smoking history, pulmonary comorbidities, autoimmune disease, prior ICI therapy defined as exposure to ICIs before EVP initiation and excluding pembrolizumab administered as part of EVP, prior chest radiation, number of prior lines of therapy, and EVP treatment details. Baseline chest imaging was reviewed to identify interstitial lung abnormalities (ILAs), which were recorded as a distinct exposure variable. A dedicated thoracic radiologist reviewed pretreatment computed tomography (CT) of the chest or position emission per Fleischner guidelines.¹⁴ Abnormalities were only considered to be ILAs if they involved at least 5% of non-dependent lung regions. A secondary sensitivity analysis included possible ILA to capture subtle or equivocal interstitial abnormalities. This study was approved by the MD Anderson Institutional Review Board (2023-0447).

Study definitions

We reviewed clinical and imaging data for participants who developed new respiratory symptoms and/or new thoracic imaging abnormalities suspicious for pneumonitis following initiation of EVP. Pneumonitis was diagnosed by expert pulmonologists and a thoracic radiologist using clinical symptoms, chest imaging, relevant laboratory results (including microbiological data), and response to treatment (steroids and/or antibiotics). Alternative diagnostic considerations included infectious pneumonia, pulmonary edema, or malignant disease progression.

Pneumonitis was adjudicated based on new thoracic imaging abnormalities compatible with pneumonitis in the appropriate clinical context and after exclusion of a more likely alternative etiology, including respiratory infection, aspiration pneumonia, pulmonary edema, or progressive malignant disease. Respiratory manifestations specifically assessed included cough, shortness of breath/dyspnea, hypoxemia, and fever. Respiratory symptoms were not required for the diagnosis of grade 1 pneumonitis. Pneumonitis severity was graded according to CTCAE v5.0, with grade 1 representing asymptomatic disease and grade ≥2 representing clinically symptomatic pneumonitis.

The radiological patterns for pneumonitis were classified as nonspecific interstitial pneumonia (NSIP), cryptogenic organizing pneumonia (COP), diffuse alveolar damage (DAD), hypersensitivity pneumonitis (HP), or other patterns based on imaging characteristics.

Outcome definitions

For patients who developed pneumonitis, additional clinical characteristics were collected, including presenting symptoms (hypoxemia, cough, fever, shortness of breath), other immune-related adverse events, radiologic pattern, corticosteroid treatment, re-exposure to EVP and pneumonitis resolution status. Treatment re-exposure was categorized according to the component(s) resumed after pneumonitis as EV alone, pembrolizumab alone, or EV plus pembrolizumab. Tolerated treatment re-exposure was defined as the absence of clinically meaningful recurrent or worsening pneumonitis attributable to the resumed treatment during subsequent evaluable follow-up, based on recurrent respiratory symptoms, new or increased oxygen requirement, subsequent thoracic imaging when available, recurrent or escalated pneumonitis-directed corticosteroid treatment, hospitalization, and treatment discontinuation attributable to recurrent pneumonitis.

Secondary outcomes included time to pneumonitis onset, pneumonitis-related severity markers (intensive care unit admission, mechanical ventilation, corticosteroid use), and all-cause mortality.

Statistical analysis

We constructed univariable and multivariable models to assess associations of baseline variables with pneumonitis risk, using both standard Cox proportional hazards and Fine-Gray subdistribution hazard regression accounting for death without pneumonitis as a competing event. Univariable models were first fitted for each framework and visualized using forest plots. Multivariable models were constructed using an event-per-variable-aware strategy, with age and ECOG performance status forced into the model and prior ICI exposure selected based on clinical relevance and univariable signal strength. Risk factor analyses were performed for 2 pneumonitis definitions: any grade and clinically symptomatic grade ≥2. To assess the association of pneumonitis with overall survival, we constructed extended Cox proportional hazards models with pneumonitis modeled as a time-dependent covariate, adjusted for metastatic disease, age, and ECOG performance status. Survival analyses were likewise performed separately for each pneumonitis grade threshold. All statistical analyses were performed using R version 4.3.2.

Results

Characteristics of the study cohort

We identified 315 patients treated with EVP. Baseline characteristics are summarized in Table 1. The median age was 71 years. Metastatic disease was present in 220 patients (70%), and 81 patients (26%) had lung metastases prior to EVP initiation. Most patients had advanced urothelial cancers. One patient in whom EVP was initiated for a presumed urothelial primary while the site of origin remained clinically and pathologically indeterminate was retained in the analyses, but was subsequently found to have renal cell carcinoma (RCC) and transitioned RCC-directed therapies. A history of tobacco use was reported in 189 patients (60%), classified as ever smokers. Preexisting lung disease was present in 56 patients (18%), including chronic obstructive pulmonary disease (COPD) in 41 (13%), 2 (0.6%) had preexisting interstitial lung disease (ILD), and 16 (5.1%) had asthma. Autoimmune disease was documented in 11 patients (3.5%). Prior ICI therapy had been administered in 43 patients (14%). Among patients with prior ICI exposure, the most commonly received agents were nivolumab (39.5%), followed by pembrolizumab, avelumab, durvalumab, ipilimumab, atezolizumab, and tiragolumab. Sixteen patients (5.1%) received prior chest radiation. Interstitial lung abnormalities (ILAs) on baseline imaging were identified in 12 patients (3.8%).

Table 1.

Baseline characteristics of the study cohort.

Characteristic	Overall N = 315	No pneumonitis N = 293	Pneumonitis N = 22	P-value
Age at EVP initiation, years (Median (Q1, Q3))	71 (64, 78)	71 (65, 78)	69 (63, 78)	.745
Sex				.441
Female	77 (24%)	70 (24%)	7 (32%)
Male	238 (76%)	223 (76%)	15 (68%)
Cancer type				.428
RCC	1 (0.3%)	1 (0.3%)	0 (0%)
Urothelial	285 (90%)	263 (90%)	22 (100%)
Urothelial + prostate	26 (8.3%)	26 (8.9%)	0 (0%)
Urothelial + RCC	3 (1.0%)	3 (1.0%)	0 (0%)
Cancer stage				.527
Stage I	3 (1.0%)	2 (0.7%)	1 (4.5%)
Stage II	16 (5.1%)	15 (5.1%)	1 (4.5%)
III	23 (7.3%)	22 (7.5%)	1 (4.5%)
IVA (locally advanced)	52 (17%)	49 (17%)	3 (14%)
IVB (metastatic)	219 (70%)	203 (69%)	16 (73%)
Unknown	2 (0.6%)	2 (0.7%)	0 (0%)
Metastatic disease	220 (70%)	204 (70%)	16 (73%)	>.999
lung metastasis	81 (26%)	75 (26%)	6 (27%)	.805
ECOG performance status				.836
0-1	256 (81%)	237 (81%)	19 (86%)
2	42 (13%)	39 (13%)	3 (14%)
3	15 (4.8%)	15 (5.1%)	0 (0%)
4	2 (0.6%)	2 (0.7%)	0 (0%)
ECOG ≥2	59 (19%)	56 (19%)	3 (14%)	.777
Smoking status				.612
Never	126 (40%)	119 (41%)	7 (32%)
Current	24 (7.6%)	23 (7.8%)	1 (4.5%)
Ex smoker	165 (52%)	151 (52%)	14 (64%)
Ever smoker	189 (60%)	174 (59%)	15 (68%)	.502
Current smoker	24 (7.6%)	23 (7.8%)	1 (4.5%)	>.999
Any lung disease	56 (18%)	52 (18%)	4 (18%)	>.999
COPD	41 (13%)	37 (13%)	4 (18%)	.506
Preexisting ILD	2 (0.6%)	2 (0.7%)	0 (0%)	>.999
Asthma	16 (5.1%)	16 (5.5%)	0 (0%)	.614
Autoimmune disease	11 (3.5%)	10 (3.4%)	1 (4.5%)	.555
Prior ICI therapy	43 (14%)	37 (13%)	6 (27%)	.097
Atezolizumab	2 (4.7%)	2/37 (5.4%)	0 (0.0%)
Pembrolizumab	14 (32.6%)	14 (37.8%)	0 (0.0%)
Nivolumab	17 (39.5%)	13 (35.1%)	4 (66.7%)
Durvalumab	4 (9.3%)	3 (8.1%)	1 (16.7%)
Avelumab	6 (14.0%)	5 (13.5%)	1 (16.7%)
Ipilimumab	4 (9.3%)	4 (10.8%)	0 (0.0%)
Tiragolumab	2 (4.7%)	2 (5.4%)	0 (0.0%)
Prior chest radiation	16 (5.1%)	15 (5.1%)	1 (4.5%)	>.999
Interstitial lung abnormalities	12 (3.8%)	10 (3.4%)	2 (9.1%)	.201
Prior lines of therapy	1.00 (0.00, 2.00)	1.00 (0.00, 2.00)	2.00 (1.00, 3.00)	.090
Number of EV cycles	4.0 (2.0, 7.0)	4.0 (2.0, 7.0)	5.0 (3.0, 10.0)	.161
Number of Pembrolizumab cycles	4.0 (2.0, 8.0)	4.0 (2.0, 8.0)	4.0 (3.0, 8.0)	.898

Abbreviations: COPD, chronic obstructive pulmonary disease; EVP, enfortumab vedotin plus pembrolizumab; ILD, interstitial lung disease; ICI, immune checkpoint inhibitor; RCC, renal cell carcinoma.

The median number of EV cycles received was 4.0 (IQR, 2.0-7.0), and the median number of pembrolizumab cycles received was 4.0 (IQR, 2.0-8.0). By reverse Kaplan-Meier estimation, the median follow-up was 356 days. The observed median follow-up was 238 days (IQR 111.5-442.0; range 5-992 days). No statistically significant differences in baseline characteristics were observed between patients who developed pneumonitis and those who did not (Table 1).

Pneumonitis

Pneumonitis developed in 22 of 315 patients (7.0%). Nine patients (40.9%) had grade 1, 3 patients (13.6%) had grade 2, 6 patients (27.3%) had grade 3, 3 patients (13.6%) had grade 4, and 1 patient (4.5%) had grade 5 pneumonitis. The median time to pneumonitis onset was 77 days (IQR 62-96; range 35-504). The cumulative incidence of pneumonitis, accounting for competing risks of death, was 5.1% (95% CI, 2.6%-7.7%) at 90 days, 6.3% (95% CI, 3.5%-9.1%) at 180 days, and 6.7% (95% CI, 3.8%-9.6%) at 365 days. There were 93 competing deaths without pneumonitis.

Systemic corticosteroids were administered in 12 of 22 patients (54.5%). Steroid use varied by grade: 2 of 9 grade 1 patients (22.2%), all 3 grade 2 patients (100%), 5 of 6 grade 3 patients (83.3%), and 2 of 3 grade 4 patients (66.7%) received steroids. One patient (4.5%) required ICU admission; no patients required mechanical ventilation. Radiologically, 11 patients had an NSIP pattern, 8 had a cryptogenic organizing pneumonia (COP) pattern, 1 had a diffuse alveolar damage (DAD) pattern, and 1 had a hypersensitivity pneumonitis (HP) pattern. Representative CT examples of NSIP and OP/COP patterns are shown in Figure 1. Complete radiologic resolution of pneumonitis was observed in 16 of 22 patients (72.7%), partial resolution in 3 (13.6%), and no radiographic resolution in 3 (13.6%). None of the 3 patients without radiographic resolution underwent subsequent treatment re-exposure. Pneumonitis was recorded as the reason for EV discontinuation in 6 of 22 patients (27.3%) and for pembrolizumab discontinuation in 8 of 22 patients (36.4%). Concurrent immune-related adverse events at the time of pneumonitis included colitis (n = 5), skin toxicity (n = 2), and thyroiditis (n = 1).

Figure 1.

Representative examples of pneumonitis in 2 patients. (A) Nonspecific interstitial pneumonitis (NSIP) pattern. CT shows peripheral-predominant ground-glass opacities (arrows) that were basilar predominant. (B) Organizing pneumonia (OP) pattern. CT shows peripheral-predominant consolidative opacities (arrows) that were basilar predominant.

Twelve of 22 patients (54.5%) underwent treatment re-exposure following pneumonitis. Eight of 12 (66.7%) resumed EV plus pembrolizumab and 4 (33.3%) resumed EV alone. All 12 patients had no clinically meaningful recurrent or worsening pneumonitis during subsequent evaluable follow-up. At treatment re-exposure, 10 of 12 patients (83.3%) were asymptomatic or had clinically resolved respiratory symptoms, whereas 2 (16.7%) had residual pulmonary symptoms and/or oxygen requirements. Both patients with residual pulmonary impairment resumed EV alone rather than pembrolizumab, and neither subsequently experienced clinically meaningful recurrent or worsening pneumonitis. An explicit EV dose reduction during the post-pneumonitis treatment course was documented in 4 of the 12 re-exposed patients (33.3%). Among the 10 patients with an exact documented point interval from pneumonitis to treatment re-exposure, the median interval was 21 days (IQR, 11-52.5 days).

By initial pneumonitis grade, treatment re-exposure occurred in 8 of 9 patients (88.9%) with grade 1 pneumonitis, 1 of 3 (33.3%) with grade 2 pneumonitis, and 3 of 6 (50.0%) with grade 3 pneumonitis. No patients with grade 4 (0/3) or grade 5 (0/1) pneumonitis underwent treatment re-exposure. Among the 8 patients in whom pembrolizumab was resumed after pneumonitis, 3 (37.5%) had an additional non-pulmonary irAE documented before rechallenge: colitis (n = 1), skin toxicity (n = 1), and thyroiditis (n = 1).

Risk factors for pneumonitis

Table 2 presents univariable predictors of any-grade pneumonitis using standard Cox proportional hazards regression. Prior ICI exposure increased the risk for pneumonitis (HR 2.46; 95% CI, 0.96-6.28; P = .061), but did not reach statistical significance. Results were consistent in Fine-Gray competing-risk regression accounting for death without pneumonitis as the competing event (sHR 2.35; 95% CI, 0.94-5.89; P = .069; Table S1). Other evaluated variables—including age, sex, ECOG performance status, smoking history, COPD, any lung disease, autoimmune disease, metastatic status, lung metastasis, prior chest radiation, prior lines of therapy, and number of EV cycles—were not associated with any-grade pneumonitis in either modeling framework (Table 2 and Table S1).

Table 2.

Univariable predictors of pneumonitis by grade threshold: standard Cox proportional hazards regression.

Variable	Any grade (n = 22)		Grade ≥2 (n = 13)
	HR (95% CI)	P-value	HR (95% CI)	P-value
Age (per year)	0.98 (0.96-1.00)	.055	0.99 (0.96-1.02)	.383
Male sex	0.69 (0.28-1.69)	.412	0.51 (0.17-1.56)	.239
Metastatic disease	1.21 (0.47-3.09)	.69	1.01 (0.31-3.27)	.993
Lung metastasis	1.21 (0.47-3.08)	.696	0.26 (0.03-2.02)	.199
ECOG PS ≥2	0.93 (0.28-3.16)	.912	0.49 (0.06-3.81)	.499
Ever smoker	1.35 (0.55-3.30)	.517	0.73 (0.25-2.18)	.577
COPD	1.68 (0.57-4.97)	.348	2.24 (0.62-8.13)	.221
Any lung disease	1.08 (0.37-3.19)	.891	1.47 (0.40-5.33)	.561
Autoimmune disease	1.53 (0.20-11.48)	.678	2.82 (0.36-22.07)	.322
Prior ICI exposure	2.46 (0.96-6.28)	.061	5.70 (1.91-17.00)	.002*
Prior chest radiation	0.91 (0.12-6.79)	.929	1.65 (0.21-12.72)	.63
Baseline ILAs	3.21 (0.75-13.76)	.116	2.33 (0.30-17.93)	.417
Prior lines of therapy	1.15 (0.98-1.34)	.086	1.01 (0.78-1.31)	.963
Number of EV cycles	1.01 (0.92-1.09)	.895	0.92 (0.79-1.07)	.255

Abbreviations: CI, confidence interval; COPD, chronic obstructive pulmonary disease; ECOG PS, Eastern Cooperative Oncology Group performance status; EV, enfortumab vedotin; HR, hazard ratio; ICI, immune checkpoint inhibitor; ILAs, interstitial lung abnormalities; NE, not estimable.

^* Bold values indicate statistical significance (P < .05).

In multivariable Cox regression adjusted for age and ECOG ≥2 (Table 3), prior ICI exposure was independently associated with any-grade pneumonitis (HR 2.68; 95% CI, 1.03-6.98; P = .04), and older age was associated with a modestly lower risk (HR 0.98 per year; 95% CI, 0.96-1.00; P = .03). In multivariable Fine-Gray regression adjusted for the same covariates, prior ICI retained an elevated point estimate of similar magnitude (sHR 2.50; 95% CI, 0.96-6.53; P = .06), though this did not reach statistical significance (Table 3).

Table 3.

Multivariable predictors of pneumonitis by grade threshold.

Panel A: standard Cox proportional hazards (adjusted)
	Any grade (n = 22)		Grade ≥2 (n = 13)
Variable	HR (95% CI)	P-value	HR (95% CI)	P-value
Age (per year)	0.98 (0.96-1.00)	.037*	0.98 (0.95-1.01)	.272
ECOG PS ≥2	1.10 (0.32-3.78)	.876	0.65 (0.08-5.07)	.679
Prior ICI exposure	2.68 (1.03-6.98)	.043*	5.91 (1.94-18.02)	.002*
Panel B: Fine-Gray competing-risk regression (adjusted)
	Any grade (n = 22)		Grade ≥2 (n = 13)
Variable	sHR (95% CI)	P-value	sHR (95% CI)	P-value
Age (per year)	0.98 (0.96-1.00)	.069	0.98 (0.94-1.02)	.351
ECOG PS ≥2	0.83 (0.25-2.82)	.77	0.53 (0.07-4.03)	.537
Prior ICI exposure	2.50 (0.96-6.53)	.061	5.47 (1.74-17.25)	.004*

Abbreviations: HR, hazard ratio; sHR, subdistribution hazard ratio; CI, confidence interval; ECOG PS, Eastern Cooperative Oncology Group performance status; ICI, immune checkpoint inhibitor.

Models adjusted for age (continuous), ECOG PS ≥2, and prior ICI exposure. Panel B accounts for death without pneumonitis as a competing event.

^* Bold values indicate statistical significance (P < .05).

The association between prior ICI exposure and pneumonitis was more pronounced when restricted to symptomatic pneumonitis. For grade ≥2 pneumonitis (13 events), prior ICI was strongly and significantly associated with risk across all modeling approaches: Cox univariable (HR 5.70; 95% CI, 1.91-17.00; P = .002), Cox adjusted (HR 5.91; 95% CI, 1.94-18.02; P = .002), Fine-Gray univariable (sHR 5.43; 95% CI, 1.87-15.73; P = .002), and Fine-Gray adjusted (sHR 5.47; 95% CI, 1.74-17.25; P = .004). No other baseline variable was significantly associated with grade ≥2 pneumonitis in any model (Table 3).

Among the 43 patients with ICI exposure before EVP initiation, 10 (23.3%) had a documented prior non-pulmonary irAE. Subsequent EVP-associated pneumonitis occurred in 1 of 10 patients (10.0%) with a prior non-pulmonary irAE compared with 5 of 33 patients (15.2%) without a documented prior non-pulmonary irAE (Fisher exact P = 1.00).

Association of pneumonitis with survival

One hundred one of 315 patients (32.1%) had died. The Kaplan-Meier estimated median overall survival for the entire cohort was 712 days (95% CI, 553-not reached). Patients who developed pneumonitis had a numerically shorter median OS (608 days; 95% CI, 528-not reached) compared with those who did not (712 days; 95% CI, 539-not reached); however, this difference was not statistically significant (log-rank P = .79).

In extended Cox regression with pneumonitis modeled as a time-dependent covariate and adjusted for metastatic disease, age, and ECOG performance status, any-grade pneumonitis was not significantly associated with mortality (HR 1.39; 95% CI, 0.66-2.91; P = .38). However, when the analysis was restricted to grade ≥2 pneumonitis, a significant association with mortality emerged (HR 2.35; 95% CI, 1.02-5.45; P = .04). In all models, metastatic disease (HR 3.38-3.42; P < .001) and ECOG ≥2 (HR 2.67-2.72; P < .001) were independently and consistently associated with increased mortality, whereas age was not (Table 4).

Table 4.

Association of pneumonitis with overall survival: adjusted extended Cox regression with pneumonitis as a time-dependent covariate.

	Any grade		Grade ≥2
Variable	HR (95% CI)	P-value	HR (95% CI)	P-value
Pneumonitis (time-dependent)	1.39 (0.66-2.91)	.385	2.35 (1.02-5.45)	.046*
Metastatic disease	3.42 (1.82-6.40)	<.001*	3.38 (1.81-6.34)	<.001*
Age (per year)	1.00 (0.98-1.01)	.762	1.00 (0.98-1.01)	.698
ECOG PS ≥2	2.67 (1.73-4.11)	<.001*	2.72 (1.76-4.20)	<.001*

Abbreviations: CI, confidence interval; ECOG PS, Eastern Cooperative Oncology Group performance status; HR, hazard ratio.

N = 315; 101 deaths. Model adjusted for metastatic disease, age, and ECOG. Pneumonitis was modeled as a time-dependent covariate.

^* Bold values indicate statistical significance (P < .05).

Progression-free survival events occurred in 156 of 315 patients (49.5%). The median PFS for the entire cohort was 443 days (95% CI, 330-539). Median PFS was similar in the pneumonitis group (608 days; 95% CI, 528-not reached) than in the non-pneumonitis group (431 days; 95% CI, 304-539) (log-rank P = .4). In time-dependent Cox regression, pneumonitis was not associated with PFS either unadjusted (HR 1.06; 95% CI, 0.56-2.03; P = .85) or after adjustment for metastatic disease and age (HR 1.00; 95% CI, 0.52-1.95; P = .99).

Discussion

In this retrospective cohort study of 315 patients with LA/mUC treated with enfortumab vedotin plus pembrolizumab (EVP), we observed a 7.0% incidence of pneumonitis, with nearly half of events reaching grade ≥3 severity. Prior exposure to ICIs emerged as the strongest independent predictor of pneumonitis, with a particularly robust association for clinically symptomatic (grade ≥2) events. Grade ≥2 pneumonitis was associated with increased mortality, whereas any-grade pneumonitis was not significantly associated with either overall survival or PFS. These findings provide real-world evidence that complements existing clinical trial safety data and identify a risk factor that is potentially actionable at the point of treatment selection and sequencing in patients initiating EVP.

The observed pneumonitis rate in our cohort is broadly consistent with, though slightly lower than, rates reported in the pivotal EV-302/KEYNOTE-A39 trial, in which pneumonitis of any grade occurred in approximately 10% of patients receiving EVP, with grade ≥3 events in approximately 4% and fatal events in a small minority.^8,9 This modest difference may reflect several factors inherent to real-world practice, including variation in surveillance intensity, differences in adjudication methodology, and heterogeneity in patient populations. In our study, pneumonitis was adjudicated by a pulmonologist and thoracic radiologist using rigorous clinical, radiologic, and microbiological criteria, which may have yielded a more conservative diagnosis compared with investigator-reported events in clinical trials.

The most notable finding of this study is the identification of prior ICI exposure as a strong and grade-dependent predictor of pneumonitis risk. This association was most pronounced for grade ≥2 pneumonitis and remained significant across both standard Cox and competing-risk modeling frameworks, supporting its robustness. Although direct evidence linking prior ICI exposure to increased pneumonitis risk during subsequent EVP therapy is limited, our findings are biologically plausible and align with emerging data from related clinical contexts. In lung cancer, therapies administered after ICI exposure have been associated with increased drug-induced pneumonitis, supporting the possibility that persistent immune modulation after checkpoint blockade may increase susceptibility to subsequent pulmonary injury.^15,16 Mechanistically, persistent activation of memory and tissue-resident CD8+ T-cell populations and other inflammatory pathways has been implicated in ICI pneumonitis,^17,18 Subsequent exposure to the overlapping pulmonary toxicities of EV and pembrolizumab could therefore increase vulnerability to lung injury, although this hypothesis requires validation in UC.^9,11

Among patients previously exposed to ICI, pneumonitis occurred in 0 of 14 patients previously treated with pembrolizumab and 4 of 17 treated with nivolumab. However, these small, partly overlapping subgroups preclude meaningful comparative inference. A 2024 meta-analysis in metastatic NSCLC reported any-grade pneumonitis rates of 4.5% with nivolumab and 5.1% with pembrolizumab, with grade ≥3 rates of 1.3% and 2.4%, respectively.¹⁹ Thus, the numerical difference observed in our cohort should not be considered definitive evidence for a difference in pneumonitis rates with nivolumab compared to pembrolizumab.

Our observation that clinically symptomatic (grade ≥2) pneumonitis, rather than any-grade pneumonitis, was associated with increased mortality adds important nuance to the understanding of EVP-related pulmonary toxicity. This finding is consistent with the broader ICI pneumonitis literature, in which checkpoint inhibitor-related pneumonitis has been recognized as one of the most common causes of immune-related fatalities, accounting for an estimated 35% of deaths associated with anti-PD-1/PD-L1 therapy.²⁰ Severe pneumonitis in particular carries high mortality and poor treatment efficacy when it occurs in the context of non-small cell lung cancer.²¹ Our finding that grade ≥2, but not any-grade, pneumonitis was associated with mortality suggests clinically important differences according to severity. Grade 1 events constituted a substantial fraction of cases and may be more clinically manageable, whereas symptomatic disease may warrant more intensive monitoring and intervention. Prospective validation of these grade-specific outcomes is needed.

We assessed baseline ILAs as a potential predictor of pneumonitis, given the biological rationale and prior literature supporting their relevance. Preexisting ILAs have been associated with increased ICI-related pneumonitis risk in NSCLC, particularly with subpleural fibrotic patterns.^22,23 In our cohort, however, ILAs were not significantly associated with pneumonitis. This null finding should be interpreted cautiously given the low ILA prevalence (3.8%), limited representation of higher-risk ILA subtypes, and inconsistent associations reported across solid tumor populations.²⁴ Larger cohorts with systematic ILA characterization are needed to clarify their relevance to EVP-associated pneumonitis.

Other commonly proposed risk factors for ICI-related pneumonitis—including smoking history, COPD, prior chest radiation, autoimmune disease, and lung metastases—did not emerge as significant predictors in our models. This lack of association should not be interpreted as definitive evidence against their relevance. A recent meta-analysis of risk factors for checkpoint inhibitor pneumonitis in lung cancer identified advanced age, male sex, smoking status, preexisting interstitial lung disease, COPD, and prior thoracic radiation as significant predictors across pooled analyses, though considerable heterogeneity was noted across individual studies.¹³ The inconsistency of these risk factors across populations and tumor types is well recognized, and our null findings may reflect limited power in a cohort of 315 patients with relatively few pneumonitis events, cohort-specific demographic characteristics, or genuine differences between EVP-associated pneumonitis and classic ICI pneumonitis.^10,13 EV itself has been associated with clinically significant pneumonitis as monotherapy,²⁵ while proposed mechanisms including nectin-4-mediated alveolar injury or bystander MMAE toxicity remain unconfirmed.²⁶ These mechanistic differences may alter the relevance of traditional ICI pneumonitis risk factors and argue for dedicated risk-profiling studies specific to ADC-ICI combinations.

Treatment re-exposure was feasible in a carefully selected subset of patients. Twelve patients underwent treatment re-exposure after pneumonitis; 8 resumed EV plus pembrolizumab and 4 resumed EV alone, with no clinically meaningful recurrent or worsening pneumonitis during subsequent evaluable follow-up. Two patients had residual pulmonary symptoms or oxygen requirements at re-exposure; both resumed EV alone without clinically meaningful recurrence. Prior ICI literature supports selective rechallenge after immune-related adverse events.^27,28 A meta-analysis of ICI rechallenge after immune-related adverse events reported an overall recurrence rate of approximately 34% across irAE types, with pneumonitis as the index event associated with a significantly higher risk of all-grade recurrence relative to other irAE types (pooled OR 4.09); high-grade recurrence was substantially less common.²⁹ A recent retrospective lung-cancer cohort study reported more favorable survival outcomes in patients who resumed ICI treatment after pneumonitis compared with those who permanently discontinued, with the caveat that such comparisons are susceptible to immortal-time bias and confounding by indication.³⁰ However, these data derive predominantly from ICI monotherapy settings, and the applicability to ADC-ICI combinations remains uncertain. Our observations should be considered descriptive and hypothesis-generating. Prospective studies with standardized rechallenge criteria and toxicity monitoring are needed.

Our study has several strengths. First, we analyzed and enrolled all eligible patients in our institution with detailed clinical phenotyping and longitudinal follow-up. Second, pneumonitis cases were diagnosed by experienced pulmonologists, after the exclusion of alternative etiologies. Third, baseline imaging was reviewed by an expert thoracic radiologist, ensuring consistent identification and classification of ILAs. Fourth, because our study focused on readily available clinical data as well as use of electronic medical record, there was minimal missing data. However, this study has several important limitations. First, its retrospective, single-center design at a high-volume academic referral center limits generalizability to community practice settings, where patient demographics, comorbidity profiles, and monitoring practices may differ substantially. The reverse Kaplan-Meier estimated median follow-up was 356 days, which limits ascertainment of later-onset pneumonitis among patients remaining on therapy. Residual confounding by unmeasured variables, including concomitant medications, granular details of prior ICI exposure (agent, duration, interval to EVP initiation), and subclinical autoimmune predisposition, cannot be excluded. Second, in a real-world EVP cohort receiving both an ADC and an ICI concurrently, it is inherently difficult to disentangle the contribution of each agent to the observed pneumonitis events. Both enfortumab vedotin and pembrolizumab independently carry pneumonitis risk, and in combination their toxicities may overlap or act additively through distinct pathways, making definitive attribution to either agent challenging without controlled comparisons or translational biomarker studies. Third, while ILAs were systematically assessed by a thoracic radiologist using Fleischner Society criteria, the low prevalence of ILAs in this cohort may have precluded detection of a meaningful association. Finally, prospective, multicenter studies with larger sample sizes and predefined toxicity management algorithms are needed to validate these findings and refine risk-stratification approaches for EVP-treated patients with UC.

In conclusion, this study provides real-world evidence that pneumonitis occurs in approximately 7% of patients with UC receiving EVP, typically presents early in the treatment course, and can be clinically consequential. Prior ICI exposure is a strong and grade-dependent predictor of pneumonitis risk, representing a novel and potentially actionable finding in this treatment context. Clinically significant pneumonitis is associated with increased mortality, reinforcing the need for vigilant respiratory monitoring in patients receiving EVP. Cautious rechallenge after pneumonitis appears feasible in selected patients, though this requires prospective validation. These data support the integration of prior ICI exposure into clinical risk assessment and argue for the development of evidence-based pneumonitis surveillance and management protocols tailored to the EVP-treated population.

Author contributions

Luis Angel Diaz Tejada (Conceptualization, Data curation, Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Fabiana Alexandra Príncipe Osorio (Conceptualization, Data curation, Formal analysis, Visualization, Writing—original draft, Writing—review & editing), Abdelrahman M. Attia (Conceptualization, Data curation, Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Ivan Alexander Flores Hernandez (Conceptualization, Data curation, Writing—original draft), Emilio Jacques Villarreal (Data curation, Writing—original draft), Omar Alhalabi (Conceptualization, Supervision, Writing—original draft, Writing—review & editing), Matthew T. Campbell (Conceptualization, Supervision, Writing—original draft, Writing—review & editing), Cindy Y. Jiang (Conceptualization, Supervision, Writing—original draft, Writing—review & editing), Zachariah Thomas (Conceptualization, Formal analysis, Writing—original draft, Writing—review & editing), Girish S. Shroff (Conceptualization, Methodology, Supervision, Validation, Writing—original draft, Writing—review & editing), Mehmet Altan (Conceptualization, Formal analysis, Supervision, Validation, Writing—original draft, Writing—review & editing), Jianjun Gao (Conceptualization, Data curation, Investigation, Supervision, Validation, Writing—original draft, Writing—review & editing), and Ajay Sheshadri (Conceptualization, Data curation, Formal analysis, Investigation, Project administration, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing)

Concept and design: All authors. Data acquisition, analysis, and/or interpretation: Luis Angel Diaz Tejada, Fabiana Alexandra Príncipe Osorio, Abdelrahman M. Attia, Ivan Alexander Flores Hernandez, and Emilio Jacques Villarreal. Drafting the manuscript: Luis Angel Diaz Tejada, Fabiana Alexandra Príncipe Osorio, Abdelrahman M. Attia, Ivan Alexander Flores Hernandez, and Emilio Jacques Villarreal. Critical revision: Ajay Sheshadri, Jianjun Gao, and Mehmet Altan. Final approval and agreement: All authors.

Supplementary material

Supplementary material is available at The Oncologist online.

Funding

None declared.

Conflicts of interest

The authors have NO affiliations with or involvement in any organization or entity with any financial interest (such as honoraria; educational grants; participation in speakers’ bureaus; membership, employment, consultancies, stock ownership, or other equity interest; and expert testimony or patent-licensing arrangements), or non-financial interest (such as personal or professional relationships, affiliations, knowledge or beliefs) in the subject matter or materials discussed in this manuscript.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.