Risk of All-Cause Mortality, Cardiovascular Disease Mortality, and Cancer Mortality in Patients With Bullous Pemphigoid

Key Points

Question  Do patients with bullous pemphigoid (BP) have an increased risk of cardiovascular disease (CVD) mortality?

Findings  In this cohort study, 252 patients with BP confirmed by clinical, pathological, or immunofluorescent evidence were individually matched with 1008 patients without BP according to age, sex, and date of dermatology clinic visit. Bullous pemphigoid was associated with a 5-fold higher risk of 1-, 3-, and 5- year CVD mortality after adjustment for comorbidities; use of antidiabetic agents, insulin, and corticosteroids; glucose level; and white blood cell count.

Meaning  These findings suggest that more evidence is needed to determine guidance for optimizing CVD protection for patients with BP.

Importance  The role of bullous pemphigoid (BP) in cardiovascular disease (CVD) mortality remains controversial, and analyses of causes of death among patients with BP based on individual data remain lacking.

Objective  To evaluate the risk of all-cause mortality, CVD mortality, and cancer mortality in patients with BP.

Design, Setting, and Participants  This cohort study identified patients who received a diagnosis of and treatment for BP during their dermatology clinic visits at a tertiary medical center in central Taiwan between January 1, 2007, and December 31, 2017. Controls were patients without BP and were individually matched to cases (4:1) according to age, sex, and date of the dermatology clinic visit. Data were analyzed from March 6, 2019, to April 2, 2021.

Exposures  Bullous pemphigoid was confirmed pathologically with typical direct immunofluorescence findings or clinically with typical clinical presentation, positive findings of an anti–basement membrane zone antibody test, and corticosteroid use for at least 28 cumulative days.

Main Outcomes and Measures  Mortality outcomes confirmed by the National Death Registry.

Results  Of 252 patients with BP and 1008 matched control patients (N = 1260), 685 (54.4%) were men and the median age was 78.0 (IQR, 70.3-84.8) years. Patients with BP had higher CVD mortality at 1 year (20 [7.9%] vs 13 [1.3%]), 3 years (28 [11.1%] vs 24 [2.4%]), and 5 years (31 [12.3%] vs 39 [3.9%]) compared with matched control patients. After adjusting for potential confounding variables, patients with BP had a 5-fold higher risk of CVD mortality at 1 year (hazard ratio [HR], 5.29 [95% CI, 2.40-11.68]), 3 years (HR, 5.79 [95% CI, 3.11-10.78]), and 5 years (HR, 4.95 [95% CI, 2.88-8.51]). Subgroup analyses revealed that the CVD mortality risk associated with BP was higher in patients without a history of hypertension (HR, 7.28 [95% CI, 3.87-13.69]) or CVD (HR, 6.59 [95% CI, 3.40-12.79]) and in patients without prior diuretic use (HR, 5.75 [95% CI, 3.15-10.50]) compared with matched control patients. In addition, all-cause mortality associated with BP was higher in patients without prior corticosteroid use than in control patients (HR 5.65 [95% CI, 4.19-7.61]).

Conclusions and Relevance  The findings of this cohort study suggest that BP was associated with a 5-fold higher risk of CVD mortality, particularly in patients without underlying hypertension or CVD or those without prior corticosteroid or diuretic use. Future studies should investigate the benefits of routine monitoring and timely management of CVD symptoms and signs in patients with BP.

Bullous pemphigoid (BP) is the most common autoimmune blistering disorder affecting older adults, with the worldwide incidence ranging from 12.1 to 43.0 per 1 million persons per year.^1-7 Bullous pemphigoid is characterized by the presence of autoantibodies targeting BP antigens 180 and 230, leading to the degradation of the basement membrane zone.⁷ Bullous pemphigoid has a chronic relapsing course that requires long-term systemic corticosteroids or other immunosuppressants. Complications such as infections or adverse effects of medical therapies lead to 3.6-fold higher mortality rates in patients with BP compared with the age-matched general population.⁸

The leading cause of death in patients with BP remains undetermined. Most relevant studies have examined overall mortality rates but not causes of death. Three studies conducted in the UK,³ Singapore,⁹ and the US¹⁰ evaluated the cause-specific mortality rates of patients with BP and reported that bronchopneumonia³ and infectious diseases^9,10 were the most common causes of death after cardiovascular disease (CVD).^3,9

Whether cancer and BP affect the other’s clinical course is controversial. A meta-analysis¹¹ identified a possible association of BP with hematologic cancer but not with overall malignant disease, whereas another national database study in Taiwan¹² suggested that cancer was not a risk factor for BP. In a review article,¹³ cancer screening was suggested for patients with early-onset or refractory pemphigoid; however, cancer deaths did not rank among the top 3 causes of mortality in relevant studies evaluating cause-specific mortality.^3,9,10

The internal validity of existing evidence on mortality risk has been jeopardized because most studies have used standardized mortality ratios to compare the mortality of patients with BP with that of the general population instead of with that of patients from the same source population with comparable characteristics.¹⁴ In addition, control of confounders was mostly limited to an age-, sex-, or region-matching approach rather than a systematic regression adjustment for the control of strong confounders. To fill the research gaps, we conducted this matched cohort study to systematically assess the association of BP with all-cause, CVD, and cancer mortality.

The Clinical Research Data Repository of China Medical University Hospital, Taichung City, Taiwan, contains medical record data from 2 750 901 patients who sought care from January 1, 2003, to December 31, 2017. The repository includes administrative and demographic information and data on diagnoses, medical and surgical procedures, prescriptions, laboratory measurements, pathology reports, physiological monitoring, hospitalization, catastrophic illness status, and National Death Registry.^15-18 This study was approved by the research ethics committee and institutional review board of China Medical University Hospital, which determined that informed consent could be waived because this was a retrospective observational study using existing electronic medical records. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

The present study population included 252 patients with BP matched with 1008 patients without BP (1:4) who visited China Medical University Hospital’s dermatology clinic between January 1, 2007, and December 31, 2017 (Figure 1). Patients with BP had pathological confirmation with typical direct immunofluorescence findings or clinical confirmation with typical clinical presentation, positive finding of an anti–basement membrane zone antibody test, and corticosteroid use for at least 28 cumulative days. We matched each patient with BP with control patients without BP (1:4) for age (2 years before or after the index date), sex, and date of dermatology clinic visit (30 days before or after the index date). The detailed identification process of patients with BP and the selection process of control patients is described in eMethods in the Supplement. The quality of the matching procedure was verified by examining the distribution of the matching factors of the BP and non-BP cohorts (eTable 1 in the Supplement). The distribution of the matching factors between BP and non-BP cohorts did not have any meaningful differences. The dermatology visit date (index date) was a median of 5 (IQR, −12 to 20) days apart between the BP and non-BP cohorts. Detailed definitions of the covariates, including comorbidities, biochemical profiles, and medication history, are provided in eTable 2 in the Supplement.

Using the National Death Registry Database, we were able to capture the mortality outcomes to December 31, 2017. All-cause mortality was defined as deaths due to any cause. As causes of death, CVD and cancer were defined by the codes from the International Classification of Diseases, Ninth Revision, and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, recorded in the registry (eTable 2 in the Supplement).

Data were analyzed from March 6, 2019, to April 2, 2021. Continuous variables are expressed as medians and IQRs and were compared using the nonparametric Wilcoxon rank-sum test. Categorical variables are expressed as a frequency (percentage) and were compared using a χ² test.

Multivariable Cox proportional hazards regression models in which age was used as the time scale were used to estimate the associations between BP and risk of mortality due to all causes, CVD, or cancer. The models yielded hazard ratios (HRs) with corresponding 95% CIs. To account for the clustering within matched pairs, we used the Cox frailty model with gamma distribution in parallel with proportional hazards models and confirmed that the HRs were similar (eTable 3 in the Supplement). Model 1 was adjusted for baseline comorbidities. Model 2 was additionally adjusted for history of oral antidiabetic agent, insulin, and corticosteroid use. Model 3 was further adjusted for baseline white blood cell (WBC) count and fasting glucose level. Because the baseline WBC count and fasting glucose level were missing for a considerable proportion of the study population (WBC count missing for 66 [26.2%] with and 501 [49.7%] without BP; glucose level missing for 118 [46.8%] with and 462 [45.8%] without BP), we performed multiple imputation using an iterative Markov chain Monte Carlo procedure with 20 imputations and 100 iterations. The missing mechanism in this study was assumed to be missing at random, because variables with missing proportions of greater than 50% were highly dependent on comorbidity status (eFigure 1 in the Supplement).

To account for the competing risk of non-CVD mortality in the analysis of CVD mortality and the competing risk of noncancer mortality in the analysis of cancer mortality, we used the Fine-Gray model to yield the subdistributions of HR. Subgroup analyses were performed using the full model (model 3) to estimate the mortality risk of BP. We conducted a test for interaction within the multivariable model to evaluate whether the effect of BP on mortality differed between subgroups. All statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc), and R, version 3.5.1 (R Program for Statistical Computing). A 2-sided statistical significance level of α was set at .05.

Of 252 patients with BP and 1008 matched control patients without BP (N = 1260), 685 (54.4%) were men and 575 (45.6%) were women; the median age was 78.0 (IQR, 70.3-84.8) years. Patients with BP were more likely to have comorbid diabetes (60 [23.8%] vs 160 [15.9%]), dementia (54 [21.4%] vs 93 [9.2%]), and organic psychosis (56 [22.2%] vs 94 [9.3%]) and less likely to have cancer (13 [5.2%] vs 116 [11.5%]) than patients without BP (Table 1). Patients with BP tended to have higher median levels of inflammatory markers such as WBC count (9.31 [IQR, 7.60-12.36] vs 6.62 [IQR, 5.30-8.70] × 10³/μL [to convert to × 10⁹/L, multiply by 0.001]), erythrocyte sedimentation rate (27 [IQR, 12-47] vs 15 [IQR, 7-33] mm/h), high-sensitivity C-reactive protein (2.02 [IQR, 0.74-4.59] vs 0.49 [IQR, 0.12-2.00] mg/dL [to convert to mg/L, multiply by 10]), immunity markers of eosinophil level (4.5% [IQR, 1.4%-11.1%] vs 2.3% [IQR, 1.0%-4.0%]), IgE (658.5 [IQR, 105.4-2027.3] vs 81.0 [IQR, 19.4-424.9] IU/mL), and fasting glucose (130 [IQR, 103-166] vs 109 [IQR, 98-131] mg/dL [to convert to mmol/L, multiply by 0.0555]). By contrast, patients with BP had lower median albumin levels (3.4 [IQR, 2.8-3.8] vs 4.0 [IQR, 3.5-4.3] g/dL [to convert to g/L, multiply by 10]) than patients without BP. Patients with BP were more likely to have taken antihypertensives, antidiabetic agents, and BP-inducing medications (dipeptidyl peptidase-4 inhibitor, dopaminergic, and furosemide) but had comparable corticosteroid use before the index date. In addition, the status of Charlson comorbidities was compared between cohorts with and without BP (eTable 4 in the Supplement). In patients with BP, the most commonly prescribed corticosteroid was prednisolone (233 [92.5%]), followed by methylprednisolone (23 [9.1%]) and dexamethasone (8 [3.2%]).

Patients with BP had higher all-cause 1-year (73 [29.0%] vs 62 [6.2%]), 3-year (116 [46.0%] vs 124 [12.3%]), 5-year (132 [52.4%] vs 170 [16.9%]), and overall (136 [54.0%] vs 219 [21.7%]) mortality than patients without BP (Table 2). In addition, patients with BP had higher CVD mortality at 1 year (20 [7.9%] vs 13 [1.3%]), 3 years (28 [11.1%] vs 24 [2.4%]), and 5 years (31 [12.3%] vs 39 [3.9%]) and higher cancer mortality at 1 year (7 [2.8%] vs 14 [1.4%]), 3 years (10 [4.0%] vs 30 [3.0%]), and 5 years (13 [5.2%] vs 44 [4.4%]) compared with patients without BP (Table 2). Based on a multivariable analysis of all-cause mortality associated with BP, the fully adjusted HR (model 3) for 1-year mortality was 4.59 (95% CI, 3.09-6.82); 3-year mortality, 4.56 (95% CI, 3.42-6.08); and 5-year mortality, 4.33 (95% CI, 3.34-5.61) (Table 2). Cardiovascular mortality accounted for 32 of 136 deaths (23.5%) among patients with BP and 52 of 219 deaths (23.7) among patients without BP, and cancer mortality accounted for 13 of 136 deaths (9.6%) among patients with BP and 50 of 219 deaths (22.8%) among patients without BP. According to model 3 (Table 2), patients with BP had a 5-fold higher risk of CVD mortality overall (HR, 3.76 [95% CI, 2.27-6.22]) and at 1 year (HR, 5.29 [95% CI, 2.40-11.68]), 3 years (HR, 5.79 [95% CI, 3.11-10.78]), and 5 years (HR, 4.95 [95% CI, 2.88-8.51]) compared with patients without BP. By contrast, an association between BP and cancer mortality was not observed in the fully adjusted model. After competing risks were accounted for, the effect size in the HRs decreased, but the risk of CVD mortality (HR for 1-year mortality, 4.45 [95% CI, 1.46-3.81]; HR for 3-year mortality, 4.27 [95% CI, 3.22-5.67]; HR for 5-year mortality, 4.10 [95% CI, 3.18-5.30]) remained significant (eTable 5 in the Supplement). Except for BP, the covariates that remained associated with overall mortality were cancer history (HR, 1.58 [95% CI, 1.10-2.27]), prior use of insulin (HR, 2.00 [95% CI, 1.34-2.98]) or corticosteroids (HR, 1.42 [95% CI, 1.10-1.82]), and increased WBC count (HR, 1.03 [95% CI, 1.00-1.07]). The covariate associated with CVD mortality was CVD history (HR, 1.67 [95% CI, 1.02-2.75]) (eTable 6 in the Supplement).

We conducted a few sensitivity analyses to support the robustness of study findings (eResults in the Supplement). These included (1) adjustment for the baseline BP-inducing medications (eg, dipeptidyl peptidase-4 inhibitor, dopaminergic, and furosemide) in the multivariable analysis (eTable 7 in the Supplement),¹⁹ (2) a complete-case analysis (eTable 8 in the Supplement), (3) verification of the stability of parameter estimation by performing pattern mixture kernel submodels assuming the underlying missing mechanism as missing not at random,²⁰ (4) use of alternative case definition when not considering corticosteroid criteria (eTable 9 in the Supplement), and (5) performance of data-driven variable selection procedures. Results based on these sensitivity analyses were consistent with the primary analyses.

In the subgroup analysis, the effect of BP on overall mortality was modified by hypertension (HR, 5.87 [95% CI, 4.35-7.93]; P = .002 for interaction) and the use of corticosteroids (HR, 5.65 [95% CI, 4.19-7.51]; P < .001 for interaction) or diuretics (HR, 5.53 [95% CI, 4.06-7.51]; P = .04 for interaction), indicating that the fully adjusted HRs were significantly increased in patients without baseline hypertension and in patients without prior corticosteroid or diuretic exposure (Figure 2). Similarly, CVD mortality was modified by hypertension (HR, 7.28 [95% CI, 3.87-13.69]; P = .001 for interaction), CVD (HR, 6.59 [95% CI, 3.40-12.79]; P = .08 for interaction), and diuretic use (HR, 5.75 [95% CI, 3.15-10.50]; P = .07 for interaction). The association between BP status and cancer mortality was modified by sex (HR for women, 4.46 [95% CI, 1.49-13.33]; P = .03 for interaction) and the use of corticosteroids (HR, 2.94 [95% CI, 1.23-7.02]; P = .03 for interaction).

This retrospective matched cohort study revealed patients with BP had a 4- to 5-fold higher risk of all-cause and CVD mortality compared with matched control patients. Subgroup analyses revealed that the effect of BP on mortality was particularly high among patients without hypertension (overall and CVD mortality), CVD (CVD mortality), or a history of corticosteroid use (overall mortality) or diuretic use (overall and CVD mortality). Despite BP not being associated with cancer mortality, patients without baseline use of corticosteroids had a higher cancer mortality rate.

We summarized the 1-year mortality rate of patients with BP in the existing literature (eTable 10 in the Supplement). The mortality rate of patients with BP found in our study was similar to those ranges reported in the studies from Singapore⁹ (1-year mortality rate, 26.7%; 2-year mortality rate, 38.54%; and 3-year mortality rate, 45.7%) and Israel⁸ (1-year mortality rate, 26.9%; 5-year mortality rate, 56.9%; and 10-year mortality rate, 69.5%). In all related studies, the estimated 1-year mortality rate of patients with BP varied from 12.9% in Spain²¹ and China²² to 38% in France.⁴ The standardized mortality ratios of patients with BP ranged from 1.83²³ to 9.56²⁴ compared with those of age- and sex-matched individuals from the general population.^{4,6,8,9,21-23,25-29} In the past 2 decades, several studies^3,9 found an association between BP and CVD mortality. A study from Singapore⁹ reported 3-year standardized mortality ratios that were significantly elevated for heart failure (68.17 [95%, CI, 27.41-140.46]) but not for infectious diseases and cancer. Another study from North East Scotland³ found that the main causes of death in BP were respiratory disease (44%), CVD (27%), neurological disease (10%), and cancer (8%).

Roujeau et al³⁰ provided one of the first specific descriptions of the high number of deaths due to CVD among older adult patients with extensive BP. The investigators proposed that these harmful effects may result from systemic glucocorticoids, because most of the patients who died had been taking active oral glucocorticoids.³⁰ Souverein et al³¹ identified an association between oral glucocorticoid use and adverse CVD or cerebrovascular outcome in patients 50 years or older with at least 1 prescription for oral or nonsystemic glucocorticoids. A review by Milani-Nejad et al³² revealed an increased risk of stroke among patients with BP. In addition, glucocorticoids have been associated with increased CVD mortality in patients with other corticosteroid-treated diseases³³ such as rheumatoid arthritis³⁴ and in kidney transplant recipients with a functioning graft.³⁵

Ample evidence has shown the association between autoimmune-related inflammatory diseases and CVD risk. The underlying autoimmune mechanisms of BP could partially explain the increased risk of CVD mortality.³⁶ Several tissue-specific isoforms of BP antigen 1 (BPAG1), which are encoded by the gene DST (for dystonin), have been identified. One of the isoforms, BPAG1-b, was detectable in vitro and in vivo as a high-molecular-mass protein in murine striated and heart muscle cells.³⁷ All signs of cardiac muscle stress, including increased transcript levels of atrial natriuretic factor and β-myosin heavy chain and decreased levels of sarcoplasmic reticulum calcium pump isoform 2A, were noted in the BPAG1-deficient mouse.³⁸ More mechanistic research is required to better understand how inflammation in BP may also impact cytokine regulation and thrombosis risk in affected individuals. Because the link between BP and adverse CVD outcomes has not been investigated thoroughly, more mechanistic research is required to strengthen the potential causal connections of BP with CVD risk.

According to the subgroup analysis, the effect of BP on CVD mortality tended to be more pronounced among patients who did not receive pharmacological CVD protection at baseline, including those without baseline hypertension or diuretic use, and among patients without a history of anti-inflammatory or corticosteroid use. These findings confirm the importance of minimizing CVD risk in patients with BP. Future prospective studies are needed to evaluate the benefits of routine monitoring and timely management of CVD risk in patients with BP, particularly for patients with BP who do not have prior CVD comorbidities.

We did not observe an association between BP and cancer mortality, a finding that aligns well with the literature.^3,9,10,27 Moreover, increasing evidence supports the neutral association between BP and incident or prevalent cancer despite the dispute of cancer screening for patients with BP.^11,39,40 However, the findings of the present study suggest that patients with BP and without baseline corticosteroid use may be particularly vulnerable to cancer mortality. More research is required to verify these findings.¹³

The strengths of the present study are the rigorous case definition of BP based on pathological findings, results of immunofluorescent examination and anti– basement membrane zone testing, and the systemic corticosteroid treatment; the comparable matched control patients selected from the same source population; the robust analytical methods; and the complete follow-up of mortality using the National Death Registry Database. This study also has several limitations. First, selection bias could not be entirely excluded because the source population consisted of individuals who attended a single tertiary hospital. However, the matched cohort design minimized this issue; we matched control patients according to age, sex, and the date of the dermatology clinic visit. The selection bias resulting from matching should be minimal because we excluded only 8 of 260 patients with BP (3.1%) (Figure 1) owing to our inability to find a matched patient without BP. Our approach also ensured the random sample of control patients were from the same source population as the cases. Second, the causes of death obtained from the National Death Registry may not be accurate because of miscoding and because the resolution of mortality was limited to only 40 causes. A relevant study⁴¹ reported favorable agreement between the original coder and a coding reviewer, resulting in a κ value as high as 0.85 for heart disease–related mortality. Third, the wide 95% CI for the outcome of CVD mortality, despite statistical significance, leaves uncertainty about the actual magnitude of the effect. Future study with a much larger sample size is warranted to verify the actual magnitude. Fourth, investigating the impact of disease severity is important among patients with BP but is beyond the design of the present study. However, our study did not have standardized and prospective data to evaluate BP severity. Further study is warranted to address this issue. Fifth, our study results could be generalized to patients who seek care in the dermatology clinic in a tertiary medical center in the Chinese Han population. Whether our findings can be generalized to other populations with different ethnic and environmental backgrounds warrants further research. However, in our minireview of 15 prior studies related to BP-associated outcomes (eTable 10 in the Supplement), we found that the age and sex distribution of these studies was similar to the distribution of our study population. The pooled mean (SD) age was 77.2 (5.4) years, which was similar to that of our population (median age, 78.0 years); the mean proportion of men was 44.1%, which was slightly less than that of our population (54.4%). The comparability of the age of BP onset and sex proportion in the present population with prior studies mitigates the concern of external validity bias.

The findings of this matched cohort study suggest that individuals with BP have a 5-fold higher risk of CVD mortality compared with individuals without BP. Individuals with BP who are naive to CVD risk control may be particularly vulnerable to CVD mortality. Further research is required to confirm our findings; large, multicenter prospective studies should be conducted to evaluate the feasibility of conducting universal CVD risk assessment and management for patients with BP.

Accepted for Publication: October 18, 2021.

Published Online: December 29, 2021. doi:10.1001/jamadermatol.2021.5125

Corresponding Authors: Chin-Chi Kuo, MD, PhD, Big Data Center, China Medical University Hospital, No. 2, Yude Road, North District, Taichung 404, Taiwan (chinchik@gmail.com); Po-Yuan Wu, MD, PhD, Department of Dermatology, China Medical University Hospital, No. 2, Yude Road, North District, Taichung 404, Taiwan (wu.poyuan@gmail.com).

Author Contributions: Drs Kuo and Wu had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Shen and Chiang contributed equally as co–first authors.

Concept and design: Shen, Kuo, Wu.

Acquisition, analysis, or interpretation of data: Chiang, Chen, Lin, Kuo, Wu.

Drafting of the manuscript: Shen, Kuo, Wu.

Critical revision of the manuscript for important intellectual content: Chiang, Chen, Lin, Kuo, Wu.

Statistical analysis: Chiang, Chen, Kuo.

Obtained funding: Kuo, Wu.

Administrative, technical, or material support: Shen, Kuo, Wu.

Supervision: Chiang, Kuo, Wu.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was partially supported by grants 108-2314-B-039-038-MY3 and 110-2321-B-468-001 from the Ministry of Science and Technology.

Role of the Funder/Sponsor: The sponsor had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: The iHi Clinical Research Platform from the Big Data Center of China Medical University Hospital provided data exploration, statistical analysis, and manuscript preparation support. The Health and Welfare Data Science Center, Ministry of Health Welfare, and Health Data Science Center, China Medical University Hospital, provided administrative and technical support.