The immunophenotype of immune checkpoint-induced bullous pemphigoid: a cohort study

Abstract

Immune checkpoint inhibitors (ICIs) are increasingly used in cancer therapy and have been linked to adverse dermatologic effects, including bullous pemphigoid (BP), which may necessitate treatment interruption. This study aims to characterize the immunologic profile of ICI-induced BP (BPICI) and identify associated autoantibodies. We conducted a retrospective cohort study (2015–2022) at Rambam Health Care Campus, including patients aged ≥ 18 years who received anti-PD-1/PD-L1 therapy and developed BP. Controls included (1) patients with non-BP cutaneous adverse events (NBPICI), (2) ICI-treated patients without cutaneous immune-related adverse events (irAEs) (NirAEICI), and (3) BP patients not treated with ICIs (BPNICI). Sera were analyzed via ELISA for BP180NC16A, BP230, Dsg1, Dsg3, and collagen VII autoantibodies. Of 242 patients with cutaneous irAEs, 16 patients were diagnosed with BP (BPICI group), and 11 patients had sera available for analysis. ELISA detected BP180NC16A antibodies in 81.8% of BPICI cases, while other autoantibodies were negative in all but one case. BPICI patients were significantly older (p < 0.05), more often male (p = 0.0066), and predominantly had cutaneous squamous cell carcinoma (SCC, 43.8%). Patients with SCC had a significantly increased hazard ratio (HR 9.41, 95% confidence interval (CI) 3.43–25.77) for BP development. BP onset occurred later in BPICI (mean 61.3 weeks) than in other irAEs (mean 32.2 weeks), but latency did not differ according to cancer type. BPICI patients required higher average daily prednisone doses per kilogram body weight during the first six months compared to BPNICI (p < 0.05). Most patients (66.6%) permanently discontinued ICI therapy due to BP. Our findings suggest that ICI-induced BP shares a similar immunologic profile with classic BP, without evidence of antigen spreading, and that SCC patients are at greater risk.

Introduction

Cancer immunotherapy, particularly immune checkpoint inhibitors (ICI) targeting PD-1 and PD-L1, has revolutionized oncology by improving survival rates across various malignancies [1]. These therapies reactivate T cells to overcome immune evasion manipulated by cancer cells [2]. While highly effective, ICIs are associated with immune-related adverse events (irAEs), especially dermatologic toxicity, with up to 71.5% of patients experiencing skin-related side effects [3].

Bullous pemphigoid (BP), an autoimmune blistering disease, is an infrequent side effect of ICI therapy, reported in up to 1% of patients [4]. Checkpoint inhibitor-induced BP manifests with urticarial plaques, tense blisters, and pruritus, which generally require systemic corticosteroids and discontinuation of ICI treatment [5]. Despite its scarcity, the emergence of BP under ICI therapy (BPICI) warrants investigation. The immunologic basis for BP in ICI-treated patients remains unclear, though autoantibodies against basement membrane components, such as those against BP180NC16A, have been implicated in disease pathogenesis [6].

This study aims to identify the intermolecular cutaneous epitope spreading of patients with BP associated with ICI therapy, focusing on BP180NC16A, BP230, Desmoglein1 (Dsg1), Desmoglein3 (Dsg3), and Collagen VII, using enzyme-linked immunosorbent assay (ELISA). By characterizing the immunophenotype, we seek to understand the differences between ICI-induced BP and BP unrelated to immunotherapy and the potential for antigen spreading under immunotherapy.

Methods

Study design and setting

This retrospective cohort study was conducted at the Department of Dermatology, Rambam Health Care Campus (Haifa, Israel), and included patients treated with PD-1/PD-L1 immune checkpoint inhibitors (ICIs) between January 2015 and December 2022. The study aimed to evaluate the immunologic profile and clinical outcomes of bullous pemphigoid (BP) induced by ICIs (BPICI).

Participants and controls

Eligible participants were adults (≥ 18 years) who received PD-1/PD-L1 inhibitors and developed cutaneous immune-related adverse events (cirAEs). A total of 1391 patients received immunotherapy at Rambam during this period. 242 ICI-treated patients developed cirAEs; of 242 patients with cutaneous irAEs, 17 were suspected of developing an autoimmune bullous disease. One patient was diagnosed with EBA, and 16 patients were diagnosed with BP. Among these (BPICI group), sera were available for 11 patients.

Control groups included:

1. 225 ICI-treated patients with cutaneous irAEs other than BP (NBPICI), 11 of whom had sera available;

2. 1149 ICI-treated patients without any cirAEs (NirAEICI); 5 of whom had sera available;

3. 276 patients diagnosed with BP who had no history of ICI exposure (BPNICI).

All BP diagnoses were confirmed by clinical criteria, direct immunofluorescence (DIF), indirect immunofluorescence (IIF), and ELISA testing. Sera were collected at presentation for the BPICI cohort, whereas for the control groups, they were retrospectively retrieved from the dermatology department’s biorepository, where they had been archived following routine clinical care. No patients were lost to follow-up during the study period.

Outcomes and variables

The primary outcome was the presence of BP-related autoantibodies, measured by enzyme-linked immunosorbent assay (ELISA). Secondary outcomes included latency to BP onset, ICI discontinuation due to BP, average daily prednisone dose per body weight (kilograms) during the first six months following diagnosis, and overall survival. Predictive variables included cancer type, age, sex, and the specific ICI agent.

Data collection and serological analysis

Clinical, demographic, and treatment data were extracted from electronic medical records. Sera were collected at the time of BP diagnosis, before the initiation of prednisone treatment. Sera were analyzed using the TECAN Infinite F50 ELISA reader. ELISA kits (EUROIMMUNE) included: BP180NC16A (Cat. No. 1502–4801-2G), BP230 (1502–4801-1G), desmoglein 1 (1495–4801 G), desmoglein 3 (1496–4801 G), and collagen VII (1947–4801 G). The cutoff for positivity was defined according to the manufacturer’s instructions, with titers > 20 IU considered positive. IIF was performed using monkey esophagus slides in earlier years, and more recently, using the BIOCHIP Mosaic method, known for its high sensitivity and standardized detection of multiple autoantibodies. The mean daily prednisone dose was calculated over the first six months of treatment and was normalized to body weight. Overall survival was defined as the time from BP diagnosis to death. The study was approved by the institutional review board (Approval No. 0483-22-RMB).

Bias and sample size

Selection bias was minimized by including all consecutive eligible patients within the study period. The retrospective design and variability in ICI treatment regimens may introduce a bias in BP latency and follow-up duration. The rate of BP180 positivity is 1% in the normal population, but can reach 37.5% in non-small cell lung cancer (NSCLC) [7]. As our cohort was heterogeneous, comprising patients with various cancers and with limited data, we assumed up to 20% positivity in the control groups. Based on expected ELISA positivity rates (80% in BPICI vs. 20% in controls), the required sample size per group was calculated as 8 (α = 0.05, power = 0.8). All BPICI and BPNICI patients were treated at Rambam Health Care Campus in the dermatology department by the same team of specialized physicians, following routine clinical practice.

Statistical analysis

Continuous variables (e.g., age, BP latency, prednisone dose) were summarized as means ± standard deviation. Categorical variables (e.g., sex and cancer type) were expressed as frequencies and percentages. Continuous variables with non-normal distribution, such as prednisone dose, were compared using the nonparametric Mann–Whitney U test. Age was compared across groups using one-way ANOVA, followed by Tukey’s post hoc test for pairwise comparisons. Categorical variables, such as sex and cancer type, were compared using the Chi-square test or Fisher’s exact test (contingency analysis). Differences in BP latency among patients receiving different immune checkpoint inhibitors were assessed using the Kruskal–Wallis test. Statistical analyses were performed using GraphPad Prism (version 8.4.3), with p-values < 0.05 considered statistically significant.

Results

We analyzed the records of 1,391 patients who received immunotherapy. Among those, 242 patients (17.3%) were referred to dermatologic consultation due to suspected cutaneous adverse events from PD-1 or PD-L1 immune checkpoint inhibitors (ICIs). Sixteen patients (1.15%) were diagnosed with ICI-associated bullous pemphigoid (BPICI), and 11 of them had sera available for immunologic analysis. 225 patients developed non-BP cutaneous immune-related adverse events (NBPICI), and 11 had available sera. Additionally, 1,149 ICI-treated patients without any dermatologic adverse events (NirAEICI) were included, of whom sera were available for five patients. A comparison cohort included 276 patients diagnosed with BP unrelated to immunotherapy (BPNICI), who were treated at our department during the same period (Fig. 1).

Fig. 1.

Sera available for autoantibody analysis. ICI (Immune Checkpoint Inhibitors), BP (Bullous Pemphigoid), EBA (Epidermolysis Bullosa Acquisita)

Clinical characteristics

All 16 patients with BPICI presented with a widespread rash of tense bullae and erosions over the trunk and limbs. The rash was more confluent on the limbs in six patients, and the palms and soles were involved in 3 patients. Most patients (14/16, 87.5%) also presented with erythematous and edematous urticarial plaques. The oral cavity was involved in 4 (25%) patients, and the epiglottis in 3 of those. Pruritus was a prodromal sign in 14/16 patients.

Table 1 provides a summary of the demographic and immunological data. The mean age of BPICI patients was 75.1 ± 8 years, similar to that of the BPNICI group (79 ± 12.5 years), but significantly older than the NBPICI+NirAEICI group (64.4 ± 12 years, p < 0.05; Fig. 2A). Male predominance was observed in the BPICI group compared to BPNICI (p = 0.0066; Fig. 2B). Notably, male predominance (69.6%) was also observed among the total cohort of patients receiving immunotherapy. There is no statistically significant difference in gender distribution between BPICI patients and the entire ICI-treated cohort (Fisher’s exact test, 0.42). Thus, the apparent male predominance in BPICI largely reflects the underlying gender distribution of ICI recipients.

Table 1.

Demographic, Clinical Profile, and Laboratory Results

Parameter	Subgroup	BPICI	NBPICI	NirAEICI	BPNICI	Overall ICI	p-value
Age (Years)		75.1 ± 8	64.4 ± 12		79 ± 12.5	64.8	BPICI VS NBPICI + NirAEICI = 0.0472 BPICI VS BPNICI = 0.4313
Gender	Male	13/16 (81.25%)	4/5 (80%)	7/11 (63.63%)	128/276 (46.37%)	969/1391 (69.66%)	*p < 0.0001
	Female	3/16 (18.75%)	1/5 (20%)	4/11 (36.36%)	148/276 (53.62%)	422/1391 (30.33%)
Cancer Type	SCC	7/16 (43.75%)	0/5	1/11 (9.1%)	N/A	112/1391 (8.05%)	**p = 0.0034
	Melanoma	4/16 (25%)	3/5 (60%)	4/11 (36.36%)	N/A	187/1391 (13.44%)
	RCC	3/16 (18.75%)	1/5 (20%)	0/11	N/A	228/1391 (16.4%)
	NSCLC	2/16 (12.5%)	1/5 (20%)	6/11 (54.54%)	N/A	362/1391 (26%)
Other side effects	Thyroid	4/16 (25%)	N/A	N/A	N/A	N/A
	GI Tract	2/16 (12.5%)	N/A	N/A	N/A	N/A
	Other	7/16 (43.8%)	N/A	N/A	N/A	N/A
	None	8/16 (50%)	N/A	N/A	N/A	N/A
DIF	IgG	12/15 (80%)	N/A	0/3	208/276 (75.36%)	N/A	***p = 0.0205
	C3	9/15 (60%)	N/A	1/3 (33.33%)	238/276 (86.23%)	N/A
	IgA	1/15 (6.66%)	N/A	0/3 (0%)	0	N/A
	IgM	0/15 (0%)	N/A	0/3 (0%)	0	N/A
IIF	Positive (basement membrane)	9/15 (60%)	N/A	N/A	110/276 (39.85%)	N/A	****NS
Permanent Discontinuation of immunotherapy#	Yes	10/15 (66.66%)	–	–	N/A	N/A
	No	2/15 (13.33%)	–	–	N/A	N/A

NS = non-significant; N/A = not available

*Fisher’s exact test (contingency analysis) of BPICI, NBPICI, NirAEICI, BPNICI, and overall ICI groups

**Fisher’s exact test (contingency analysis) of BPICI, NBPICI, NirAEICI, and overall ICI groups

***Fisher’s exact test (contingency analysis) of BPICI, NBPICI, and BPNICI groups

****Fisher’s exact test (contingency analysis) of BPICI and BPNICI groups

^#Three patients developed BP one to two months after cessation of immunotherapy

BPICI (Underwent immunotherapy and developed Bullous Pemphigoid), BPNICI (Did not undergo Immunotherapy but developed Bullous Pemphigoid), NBPICI (Underwent immunotherapy and developed different dermatological Immune-Related Adverse Event than Bullous Pemphigoid), NirAEICI (Underwent immunotherapy and did not develop Immune-Related Adverse Events), SCC (Squamous Cell Carcinoma), RCC (Renal Cell Carcinoma), NSCLC (Non–Small Cell Lung Carcinoma), GI tract (Gastrointestinal tract), DIF (Direct Immunofluorescence). IIF (Indirect Immunofluorescence)

Fig. 2.

Clinical characteristics, treatment, and outcomes of patients with immune checkpoint inhibitor-associated bullous pemphigoid (BPICI) and controls. A Age distribution among BPICI patients (patients who underwent immunotherapy and developed BP), BPNICI (patients who did not undergo immunotherapy but developed BP), and NBPICI + NirAEICI (patients who underwent immunotherapy and either developed a different dermatological irAE than BP or did not develop irAEs). B Gender distribution (F-female vs. M-male) in BPICI and BPNICI cohorts. C Latency (weeks) to BP development in BPICI and NBPICI groups. D Latency (months) to BP development by ICI agent: pembrolizumab, nivolumab, and cemiplimab. E Mean daily prednisone (mg) per kilogram (kg) body weight during the first six months in BPICI and BPNICI patients. F Kaplan–Meier survival curves comparing BPICI, BPNICI, and NBPICI + NirAEICI cohorts. Error bars represent standard deviations where applicable. Statistical significance is indicated as follows: *p < .0.05, **p < .0.01, ****p < .0.0001, ns = non-significant

DIF was performed in 15 of 16 BPICI patients and was positive in the majority (14/15 patients), with variable patterns of IgG and/or C3 deposition. One patient (1/15) had negative DIF yet tested IIF-positive for BP180, and another (1/16) did not undergo DIF but had a positive IIF for basement membrane zone antibodies, underscoring the heterogeneity of diagnostic profiles in BPICI.

Serologic analysis using ELISA (Table 2) showed BP180NC16A antibody positivity in 81.8% of BPICI patients. Only one patient exhibited multi-antigen positivity, possibly indicating intermolecular antigen spreading. Two patients were ELISA-negative (But positive for IgG deposition in DIF). All control sera from NBPICI and NirAEICI groups were negative for BP180NC16A, BP230, desmogleins, and collagen VII autoantibodies.

Table 2.

ELISA immunoassay before and after immunotherapy

	Groups	BP 180	BP 230	Dsg1	Dsg3	Collagen VII
BPICI	1	neg	neg	neg	neg	neg
	2	+  +  +	+  +	+  +	+  +	neg
	3	+  +	neg	neg	neg	neg
	4	+  +  +	neg	neg	neg	neg
	5	+  +	neg	neg	neg	neg
	6	+  +	neg	neg	neg	neg
	7	neg	neg	neg	neg	neg
	8	+  +  +	neg	neg	neg	neg
	9	+  +  +	neg	neg	neg	neg
	10	+  +  +	neg	neg	neg	neg
	11	+  +	neg	neg	neg	neg
NBPICI (n=11)	Serum before ICI (n=7)	neg	neg	neg	neg	neg
	Serum after ICI (n=4)	neg	neg	neg	neg	neg
NirAEICI (n = 5)	Serum before ICI (n = 3)	neg	neg	neg	neg	neg
	Serum after ICI (n = 4)	neg	neg	neg	neg	neg

Legend: – < 20 IU/ml; +  ≥ 20 to < 30 IU/ml; +  +  ≥ 30 to ≤ 200 IU/ml; +  +  +  > 200 IU/m

BPICI (Underwent immunotherapy and developed Bullous Pemphigoid), NBPICI (Underwent immunotherapy and developed different dermatological Immune-Related Adverse Event than Bullous Pemphigoid), NirAEICI (Underwent immunotherapy and did not develop Immune-Related Adverse Events), ICI (Immune Checkpoint Inhibitors), Dsg1 (Desmoglein 1), Dsg3 (Desmoglein 3), neg (negative)

Cancer type

The most common associated cancers in BPICI patients were cutaneous squamous cell carcinoma (SCC, seven patients, 43.8%) and melanoma (4 patients, 25%), followed by renal cell carcinoma (RCC,18.8%) and non-small cell lung cancer (NSCLC,12.5%).

The distribution of underlying cancers differed significantly between BPICI patients, NBPICI, NirAEICI, and the overall ICI-treated cohorts (Fisher’s exact test, p = 0.0034). Among BPICI patients, cutaneous SCC (43.8%) and melanoma (25%) were more frequent compared with their prevalence in the overall ICI cohort (112 patients, 8.05%, and 187 patients, 13.44%, respectively). NSCLC was less common (12.5% in BPICI vs. 26% in overall ICI). RCC prevalence was similar between groups (18.8% in BPICI vs. 16.4% in overall ICI).

The hazard ratio (HR) for BP development in patients with SCC under immunotherapy was significantly higher than in other cancer types (HR 9.41, 95% confidence interval (CI) 3.43–25.77). Other cancers did not show increased risk for BP development (melanoma HR 2.19, CI 0.7–6.87; RCC HR 1.19, CI 0.34–4.2; NSCLC HR 0.41, CI 0.09–1.8).

Latency

In a subgroup of 59 NBPICI patients with available data, the mean latency to onset was 32.2 ± 43.2 weeks. In contrast, BPICI patients developed symptoms after a significantly longer latency of 61.3 ± 66.5 weeks (p < 0.05; Fig. 2C), consistent with prior findings that BP may be a delayed toxicity [8]. Latency did not differ according to cancer type.

ICI type

We further analyzed BP latency by ICI type (cemiplimab, nivolumab, pembrolizumab). One-way ANOVA revealed significant differences in latency among the three drugs (p = 0.0143; Fig. 2D). Post hoc Tukey’s multiple comparison test showed that latency was significantly shorter in cemiplimab-treated patients compared to those receiving nivolumab (p = 0.044), while differences between cemiplimab and pembrolizumab, as well as between pembrolizumab and nivolumab, were not statistically significant.

Prednisone dosage

Prednisone doses were analyzed and normalized by patient weight (mg/kg/day). BPICI patients required significantly higher mean daily prednisone doses per kilogram body weight during the first six months of treatment (p = 0.0195, Mann–Whitney U test; Fig. 2E) than BPNICI patients, suggesting greater disease severity. Earlier BP onset did not correlate with increased steroid requirement.

Discontinuation of immunotherapy

Ten patients (66.66%) permanently discontinued immunotherapy due to BP. The malignancy progressed in 2 of them, was stable in 4, and continued to improve/maintain complete remission in 4. One patient resumed immunotherapy 3 months after BP diagnosis but had to discontinue due to exacerbation of BP. Three patients developed BP one or two months after the end of the immunotherapy course. All three maintained complete remission. Only one patient continued immunotherapy after BP diagnosis at longer intervals and maintained stable disease.

Prognosis

Finally, we assessed whether BP development correlated with clinical outcomes. Analysis of disease-free and overall survival (Fig. 2F) revealed no significant differences between BPICI patients and those in each of the other control groups. However, these results are limited due to the small sample size, which may have limited the statistical power to detect survival differences. The follow-up data and mortality rates are summarized as follows:

• Group 1 (BPICI) Median survival 370.5 weeks, mean follow-up 6.48 years per patient, mortality rate 7.7 deaths per 100 person-years

• Group 2 (NBPICI + NirAEICI) Median survival 198 weeks, mean follow-up 5.66 years per patient, mortality rate 9.6 deaths per 100 person-years

• Group 3 (BPNICI) Median survival 88 weeks, mean follow-up 4.41 years per patient, mortality rate 17.9 deaths per 100 person-years

Discussion and summary

Our study provides a comprehensive analysis of the clinical and immunologic characteristics of bullous pemphigoid (BP) in patients receiving ICIs, to assess the potential for antigen spreading. Although BP is a relatively uncommon irAE, it poses significant clinical challenges due to its severity and the frequent need to discontinue anticancer therapy permanently.

Consistent with prior reports, our data show that BP predominantly affects older individuals, with most cases presenting in patients over 70 years of age [5, 9]. However, we observed a significant male predominance among BPICI patients compared to those with BPNICI, where the prevalence is generally equal between sexes [5, 9]. This finding may be explained by the male predominance observed in our cohort of patients receiving immunotherapy (Table 1). The male predominance in BPICI may also reflect the higher incidence of certain malignancies, such as squamous cell carcinoma (SCC) and non-small cell lung cancer (NSCLC), in men [10].

The main cancer types associated with BPICI include melanoma, NSCLC, SCC, and renal cell carcinoma (RCC) [8, 11]. The same cancer types were observed in our cohort. Interestingly, SCC was more common than melanoma in our population, whereas melanoma predominated in most previous reports [5, 8, 11]. This discrepancy could be influenced by differential reporting of BP cases, as immunotherapy received earlier approval in melanoma (2012) compared to SCC (2018) [12]. Moreover, when considering all patients who received immunotherapy at our institution, we demonstrated that patients with SCC had a significantly higher hazard ratio for ICI-induced BP compared to patients with other cancer types (HR = 9.41, 95% CI 3.43–25.77). This may be explained by the overexpression of BP180 in SCC tumor cells [13, 14]. Possibly, the immune response targeting SCC antigens can evolve into an autoreactive response against BP antigens that are overexpressed on SCC cells. An additional explanation for the increased risk is the older age of SCC patients (mean 77.57 years). Older individuals are more likely to harbor circulating BP autoantibodies without clinical disease, potentially predisposing them to ICI-induced BP [15]. An association between BP (not related to immunotherapy) and SCC has also been previously shown by Albadri et al. [16], who compared bullous pemphigoid patients to age-matched controls and found that SCC was the most commonly associated cancer [16].

We also observed a notable prevalence of RCC, which has been reported in seven previous cases of BPICI [9, 11]. Our RCC patients were the oldest in the cohort (mean 79.3 years), raising the possibility that age-related subclinical autoimmunity could contribute to disease onset. Moreover, BP180 has been detected in the kidney glomerular basement membrane and in podocyte foot processes [17], suggesting that the kidney glomerulus may represent a novel site of relevance in BP pathogenesis. It is conceivable that ICI-induced immune activation in RCC patients could unmask or amplify an autoimmune response against BP180 expressed in renal tissue, which might cross-react with cutaneous BP180 and precipitate BP [17].

The key objective of our study was to characterize the clinical and immune phenotype of BPICI. We evaluated direct immunofluorescence (DIF), indirect immunofluorescence (IIF), and ELISA profiles for BP180NC16A, BP230, desmoglein 1 (Dsg1), desmoglein 3 (Dsg3), and collagen VII. Among 95 BPICI cases reported in the literature, 74 had available DIF results, and only 31 had IIF results [6, 18]. In our cohort, DIF was considered positive if we detected IgG and/or C3 deposition. Of the 16 patients, 15 underwent DIF testing; 14 of these were positive, yielding a positivity rate of 93.3%, comparable to 78.3% (58/74) in the literature [19, 20]. The DIF-negative patient had positive IIF for BP180, so all had diagnostic evidence of BP [20, 21]. IIF was positive in only 60% of our patients, compared to 17–87% previously reported [18, 19, 22]. This might suggest technical differences due to the use of monkey esophagus slides earlier, while adopting the BIOCHIP Mosaic method recently, which provides higher sensitivity and standardization [23, 24]. The mixed testing approaches in the BPICI group (and even more so in the BPNICI group) likely contributed to the lower IIF positivity rate observed.

Epitope spreading in BP patients was reported in up to 49% of the patients, but mainly occurred within the BP180 and BP230 antigens [25]. Desmoglein antibodies in BP sera are considered rare and were only reported in case reports [26, 27]. However, demonstrating significant differences in epitope spreading between BPICI and BP unrelated to immunotherapy would require substantially larger cohorts.

We hypothesized that BPICI might express broader antigenic reactivity due to enhanced immune activation and potential antigen spreading. However, most patients demonstrated isolated BP180NC16A positivity, with only one patient exhibiting reactivity to multiple autoantibodies. Among 95 published BPICI cases, 52 had ELISA data: 40 were positive for BP180, and 16 were positive for BP230 [28, 29]. Dsg3 was assessed in two patients and was positive in only one case [30]. These results suggest that BPICI is immunologically similar to BPNICI, with limited evidence of antigen spreading. This interpretation is further supported by the negative serologic findings in a control group of 16 patients who received ICI therapy, did not develop BP, and had available sera for analysis. However, given the small size of our study population, these findings should be interpreted cautiously and cannot definitively exclude broader antigenic reactivity.

Most of our patients (66.6%) permanently discontinued immunotherapy due to BP. Similarly, in previously published BPICI cases, 73.7% (70/95) discontinued ICI for the same reason, and in seven cases, treatment continuation status was missing [28–30]. We did not find evidence that BP development was associated with improved oncologic outcomes. Thus, decisions to stop immunotherapy were based on BP severity, oncologic disease status, and the availability of alternative treatments [31]. For patients already in remission, the decision to discontinue ICI was easier. The decision to continue or discontinue ICI therapy should be individualized, considering the overall clinical context [31, 32].

In our cohort, the latency from the start of immunotherapy to rash onset in NBPICI patients averaged 32.2 weeks (≈7 months), which is longer than previously reported for cutaneous irAEs. Prior studies report a median onset of 4 weeks, with a range of 2–150 weeks [3], reflecting the variable timing of these reactions. Additionally, while our study did not demonstrate differences in latency to BPICI development across cancer types, patients treated with cemiplimab exhibited a shorter latency compared with those receiving nivolumab (p = 0.044). Differences between cemiplimab and pembrolizumab, as well as between pembrolizumab and nivolumab, were not statistically significant. However, the retrospective design may have introduced potential misclassification or delays in diagnosis, and variability in ICI initiation dates could have influenced latency analysis. Thus, our findings underscore the importance of ongoing dermatologic monitoring throughout immune checkpoint inhibitor therapy.

This study has several limitations. The cohort size was limited by the infrequency of BPICI. Selection bias may have occurred, as only patients referred for dermatologic evaluation were included. Additionally, some patients lacked serum samples for a complete immunologic analysis. Specifically, we lacked sera of BPICI patients before BP development, which could have clarified whether BP represents a flare (induced by ICIs) of an underlying subclinical condition. This is indeed a possibility that is supported by the specific immunophenotype of most BPICI patients and the lack of antigen spreading. The retrospective design may have introduced potential misclassification or delays in diagnosis, and variability in ICI initiation dates could have influenced latency analysis. As a single-center study, our findings may not be generalizable; however, the consistency of our results with previously published data supports their broader relevance.

In summary, BPICI develops in approximately 1% of patients who receive immunotherapy. This study provides an immunological and clinical characterization of BPICI. Analyzing sera from these patients, as well as from other patients who received immunotherapy and did not develop BP, demonstrated the lack of anti-skin autoimmunity and intermolecular antigen spreading under immunotherapy. Our study also highlights that patients with SCC are at an increased risk for BP development compared to patients with other cancers. Additionally, a comparison to 276 BP patients who did not receive immunotherapy showed that BPICI patients required significantly higher mean daily prednisone doses during the first six months of treatment. Further studies are warranted to better define other risk factors and predictive markers for BP in patients undergoing ICI therapy.

Abbreviations

BP

Bullous pemphigoid

BPNICI

Did not undergo Immunotherapy but developed bullous pemphigoid

ICI

Immune checkpoint inhibitors

irAEs

Immune-related adverse events

NSCLC

Non–small cell lung carcinoma

RCC

Renal cell carcinoma

SCC

Squamous cell carcinoma

BPICI

Underwent immunotherapy and developed bullous pemphigoid

NBPICI

Underwent immunotherapy and developed different dermatological immune-related adverse event than bullous pemphigoid

NirAEICI

Underwent immunotherapy and did not develop immune-related adverse events

Author contributions

N.S and M.D collected data and analysed it. I.B Oversight of laboratory procedures. Z.A and I.B performed ELISA experiments. N.S, I.C and D.Y did the literature review N.S, E.AH and Z.KH did the drafting and revising of the manuscript I.C did the manuscript editing for grammar and language. N.S, IC and D.Y did the management of references and citations. N.S and E.AH prepared all the figures, tables and diagram E.AH Conceptualization of the study, Supervision, analysis of results, manuscript editing, revision, and final approval.

Funding

No funding was received for this study.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Conflict of interest

The authors declare no competing interests.

Ethics approval

This study was reviewed and approved by the institutional Helsinki Committee (Approval No. 0483-22-RMB) in accordance with the Declaration of Helsinki. Written informed consent was waived as the study was conducted using anonymized sera, as approved by the ethics committee.