Kaplan-Meier analysis of survival: overall death. Most patients died in the first 6 months after hospitalization (log-rank test: not feasible).
Kaplan-Meier analysis of survival with serum albumin levels of 3.6 g/dL or less (B) vs more than 3.6 g/dL (A) at baseline (log-rank test: P<.001).
Kaplan-Meier analysis of survival with a daily dosage of 37 mg or less (A) vs more than 37 mg (B) of prednisolone equivalent at discharge.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Rzany B, Partscht K, Jung M, et al. Risk Factors for Lethal Outcome in Patients With Bullous Pemphigoid: Low Serum Albumin Level, High Dosage of Glucocorticosteroids, and Old Age. Arch Dermatol. 2002;138(7):903–908. doi:10.1001/archderm.138.7.903
Although bullous pemphigoid (BP) is the most frequent autoimmune bullous disease and is associated with a considerable case-fatality rate, little is known about factors that influence its prognosis.
To identify prognostic factors for lethal outcome in the first year after the initial hospitalization in patients with BP.
A multicenter retrospective cohort study.
Seven dermatologic university hospitals in Germany.
A total of 369 patients diagnosed as having BP between January 1, 1987, and December 31, 1997.
Univariate (Kaplan-Meier) and multivariate (Cox regression) analysis.
Of the 369 patients with BP, 209 (57%) died, 106 (29%) within the first year after hospitalization. Fifty-four percent were women. The mean ± SD age at entry was 77.3 ± 11.1 years. The patients with BP were followed up to 10.5 years, with a median time of 1.8 years to death or interview (25th and 75th quartiles, 0.5 and 4.0 years). The major risk factors for lethal outcome in the first year after hospitalization were an increased age, with a multivariate risk estimate of 3.2 (95% confidence interval [CI], 1.9-5.2) for age greater than 80.4 years (median); a daily glucocorticosteroid dosage of more than 37 mg (75th quartile) at discharge, with a multivariate risk estimate of 2.5 (95% CI, 1.5-4.3); serum albumin levels of 3.6 g/dL or less (25th quartile), with a multivariate risk estimate of 2.6 (95% CI, 1.5-4.4); and an erythrocyte sedimentation rate greater than 30 mm/h (75th quartile), with a multivariate risk estimate of 1.7 (95% CI, 1.1-2.8).
There is a considerable case-fatality rate in patients with BP. Older patients who require a higher dosage of oral glucocorticosteroids at hospital discharge and who have low serum albumin levels are at greater risk of death within the first year after hospitalization. These prognostic factors should be considered in the care of patients with BP as well as in the design of future clinical trials.
BULLOUS PEMPHIGOID (BP) is by far the most frequent bullous autoimmune disease in central Europe.1,2 It is characterized by a high case-fatality rate in the first year after diagnosis.3-5 The incidence of BP increases rapidly for each decade beyond the age of 60 years. As the structure of the European population is shifting toward the aged, more people are expected to develop BP in the coming years.6 Therefore, it is important to investigate risk and prognostic factors of this disease.
However, despite recent progress in the analysis of cellular and molecular mechanisms,7-13 our knowledge of risk and prognostic factors in BP remains limited. The only data available so far on risk factors influencing the long-term prognosis of patients with BP are clinical trials covering only a limited period and comprising selected patients14-16 and 3 studies, mostly based on a univariate analysis, on prognostic factors from France3,5 and Great Britain.17
Therefore, in a multicenter effort of 7 German dermatology departments, a large retrospective cohort of patients with BP was recruited to study factors that may influence the lethal course of the disease.
The cohort comprised all patients with a discharge diagnosis of BP coded as 694 by the International Classification of Diseases, Ninth Revision (Technische Universität Dresden, Dresden, and Fakultät für Klinische Medizin Mannheim der Universität Heidelberg, Mannheim), or patients documented as having BP in the files from the immunohistochemical laboratories (Klinikum der Rheinisch-Westfälische Technischen Hochschule Aachen, Aachen; Universität Bonn, Bonn; Universität Kiel, Kiel; Ludwig-Maximilians Universität, München; and Universität Würzburg, Würzburg). As nearly all patients during this period were treated as inpatients, only hospitalized patients were included. Data on the first inpatient treatment were obtained from the charts and validated by interview with either the patient or a relative whenever possible. The data recruitment period varied between the different centers. For this study, only patients who were admitted between January 1, 1987, and December 31, 1997, were included (N = 432). The diagnosis was reevaluated according to clinical, histologic, and immunohistologic (direct and indirect immunofluorescence [IF] examinations) criteria. Only patients with autoantibodies against the basement membrane zone detected by direct or indirect IF were included (n = 402). Patients had to be hospitalized for at least 1 night (n = 399).
The outcome variable was death as reported by relatives, by the Städtisches Einwohnermeldeamt (municipal resident registration offices), or by the Bürgermeisteramt (office of the mayor). Only patients in whom the outcome variable (dead or alive) was known were included in the analysis (n = 369). As the aim of the study was to detect possible disease-relevant prognostic factors, the primary data analysis was restricted to prognostic factors for lethal outcome in the first year after the date of first hospitalization. In 2 further analyses, prognostic factors for lethal outcome after 1 to 3 years and more than 3 years after the first hospitalization were evaluated.
All data available at the time of the first hospitalization and believed to be important for the progress of the disease (based on expert opinion and literature) were collected. These data included demographic variables (sex and age), characteristics of the disease (erosions and/or blisters, mucosal involvement, and time between diagnosis of BP and admission), habits (smoking and alcohol), previous or present diseases (diabetes mellitus and malignancies), the treatment of BP (glucocorticosteroids with or without adjuvant therapy such as azathioprine, sulfones, and others, and the dosage of oral glucocorticosteroids at discharge). Furthermore, hematologic and immunologic measures were assessed: presence of autoantibodies, serum glucose levels, serum albumin levels, erythrocyte sedimentation rate, and other measures. For the Kaplan-Meier analysis, if no predefined cutoff points were available, continuous variables were categorized on the basis of the median or the 25th or 75th percentiles.
The primary analysis was a survival analysis using Kaplan-Meier plots, the log-rank test, and Cox proportional hazard regression analysis. The time scale was defined as time since first hospitalization.
Univariate analysis was performed with Kaplan-Meier survival curves. Curves were inspected graphically to visually check the proportionality assumption. Differences between the curves were assessed by means of the log-rank test. To document the similarities for univariate results, Cox proportional hazard regression analysis was calculated on single variables, with statistical significance of the model assessed with the –2 logarithm likelihood ratio. In the second step, all variables were adjusted for variables that were strongly associated with lethal outcome as well as region and sex.
In total, 3 multivariate analyses were performed on the basis of lethal outcome in the first year, between the first and third years, and after the third year. As the time of observation varied, the numbers of the patients included in the models also varied. Therefore, the analysis for the third year included only 130 patients and not 210.
All data management and data analysis were conducted with SAS statistical software (SAS Institute Inc, Cary, NC). Cox proportional hazard regression analysis was performed with the SAS PHREG procedure (SAS Institute Inc).
A total of 369 patients with BP were included in the study. Of these patients, 354 (96%) had cutaneous blisters and/or erosions at the time of first hospitalization. In addition, 74 patients (20%) had involvement of the oral mucosa. The median onset (25th and 75th quartiles) of the disease before hospitalization was 50 days (25 and 148 days); the mean ±SD onset was 140 ± 250 days.
All patients had either positive direct or indirect IF findings. Linear IgG and/or C3 deposits at the basement membrane zone were detected in 289 (91%) of 319 patients by direct IF microscopy of perilesional skin. Circulating autoantibodies against the basement membrane zone were detected in 266 (79%) of 338 patients by indirect IF on monkey or guinea pig esophagus and/or human sodium chloride–split skin. Indirect IF analysis on sodium chloride–split skin was performed with serum samples from 101 patients. Of these, serum samples from 89 (88%) contained autoantibodies that bound to the epidermal site of the blister. In 80 cases with missing (n = 50) or negative or noncharacteristic (n = 30) results of direct IF, circulating autoantibodies against the basement membrane zone were detectable by indirect IF.
The majority of patients with BP came from the departments of dermatology in Munich (33%) and Würzburg (27%). Of the 369 patients, 209 (57%) died; 106 (29%) died within the first year (Figure 1), 53 (14%) died between the first and the third years, and 50 (14%) died after the third year following the first hospitalization. The mean follow-up time to date of interview or of death was 2.7 ± 2.6 years; the median (25th and 75th quartiles) was 1.9 (0.5, 4.0) years.
Of the patients, 199 (54%) were female. The age distribution ranged from 33 to 99 years. The mean age at entry in the cohort was 77.3 ± 11.1 years; the median age (25th and 75th quartiles) was 80.4 (71.3, 84.8) years (Table 1).
Various malignancies, mostly of epithelial origin, at or before the hospitalization were reported in 58 patients with BP (16%). In an additional 14 patients (4%), malignancies occurred after first hospitalization. Diabetes mellitus was documented in 86 patients (23%). The mean serum glucose level (average of the first 3 measurements after first hospitalization) was 130 mg (7.2 mmol/L) in 117 patients (32%).
The mean ±SD serum albumin level at baseline was 4.3 ± 0.9 g/dL; the median (25th and 75th quartiles) was 4.1 (3.6, 4.7) g/dL. The mean ±SD serum sodium level was 139 ± 8 mEq/L; the median (25th and 75th quartiles) serum sodium level was 140 (138, 143) mEq/L.
Of the 369 patients, 159 (43%) were treated solely with systemic glucocorticosteroids, 89 (24%) with azathioprine and glucocorticosteroids, 42 (12%) with sulfones with or without glucocorticosteroids, 45 (12%) with tetracycline hydrochloride, 6 (2%) with cyclophosphamide with or without glucocorticosteroids, and 7 (2%) with methotrexate with or without glucocorticosteroids or other immunosuppressive drugs. Twenty-one (6%) of the patients with BP received other drugs or other combinations of drugs as mentioned above.
In the univariate analysis, several variables were found to increase the risk of death in the BP cohort within the first year after first hospitalization. Age greater than 80.4 years vs 80.4 years or less (P<.001), serum albumin levels of 3.6 g/dL or less vs greater than 3.6 g/dL (P<.001) (Figure 2), prednisolone equivalent greater than 37 mg/d vs 37 mg/d or less at discharge (P<.001) (Figure 3), and an erythrocyte sedimentation rate at admission of greater than 30 mm/h vs 30 mm/h or less (P = .002) were found to be highly significant. In addition, a duration of the disease before hospitalization of 50 days or less vs more than 50 days was significant (P = .02). In contrast, the presence of oral lesions did not significantly increase the risk of death (P = .12). Neither sex of male vs female (P = .67) nor status of partnership (P = .17) was a significant risk factor. Other nonsignificant variables were treatment with corticosteroids alone vs corticosteroids and at least 1 adjuvant immunosuppressive drug (P = .35) and the presence of malignancies at or before the hospitalization (P = .17) (Table 1).
In the multivariate model, 4 significant prognostic factors could be determined: age, with a relative risk (RR) (95% confidence interval [CI]) of 3.2 (1.9-5.2) for greater than 80.4 years vs 80.4 years or less; the dosage of prednisolone equivalent (>37 mg/d vs ≤37 mg/d) at discharge, with 2.5 (95% CI, 1.5-4.3); a reduced serum albumin level (≤3.6 g/dL vs >36 g/dL), with 2.6 (95% CI, 1.5-4.4); and an increased erythrocyte sedimentation rate (>30 mm/h vs ≤30 mm/h), with 1.7 (95% CI, 1.1-2.8). All variables were additionally controlled for sex, duration of disease before hospitalization, and region (Table 2).
As fewer patients were available for analysis of risk factors for lethal outcome between 1 and 3 years or after 3 years of admission, multivariate models based on different combinations of variables were performed. For lethal outcome between 1 and 3 years after the first hospitalization, a past or present history of malignancies became significant (adjusted RR, 1.1 [95% CI, 1.1-1.2]). However, for both additional periods investigated, old age remained the most prominent prognostic factor (Table 2).
This study confirms a high case-fatality rate for BP and presents multivariate results for prognostic factors of lethal outcome in a retrospective cohort of patients with BP in Germany. With lethal outcome defined as death within the first year after hospitalization, several important risk factors were found. In addition to greater age, a higher dosage of oral glucocorticosteroids at discharge, an increased erythrocyte sedimentation rate, and a low serum albumin level at admission could be identified.
Despite a different time frame, these results are mostly in accordance with results of a French cohort study that showed that a greater age, with an RR up to 7.1 (95% CI, 2.0-25.4) for patients aged 86 years and older; a poor general condition, with an RR of 5.8 (95% CI, 1.7-20.0); and a generalized distribution of lesions, with an RR of 4.4 (95% CI, 1.4-13.7) were associated with an increased risk of lethal outcome within the first 6 months after the beginning of treatment.5 Unlike the French investigators, we did not look at distribution of lesions, as this is a variable known to show a high variability between centers and might introduce information bias, especially in a retrospective setting. In addition, in the French study, an increased risk for female sex was found, with an RR of 2.4 (95% CI, 1.1-5.4). However, we were not able to confirm this borderline significant observation in our cohort. When we limited our analysis to lethal outcome in the first 6 months after hospitalization, our results for age, dosage of glucocorticosteroids at discharge, serum albumin levels, and erythrocyte sedimentation rate did not differ much compared with the 12-month analysis (data not shown).
Most other previous studies aimed at determining prognostic factors in patients with BP4,17 relied only on univariate analysis. Thus, no comparison with our multivariate results should be drawn, as a univariate analysis does not control for confounding. Nevertheless, our univariate results are in accord with those by Venning and Wojnarowska,17 who reported an increased risk of death in a British cohort within the first year for increased age. In contrast to their findings, we could not detect an increased risk for patients receiving adjuvant immunosuppressive therapy. Even when looking for a possible interaction between medication with azathioprine in patients with low serum albumin levels—assuming an increased risk of adverse events due to a relative overdose—we were not able to discover a significant risk (data not shown).
As recent univariate studies point to an association of disease activity with the presence and, presumably, the serum levels of BP180 autoantibodies,4,18 it would be tempting to investigate levels of BP180 antibodies as a prognostic factor for lethal outcome. However, as the present study was based on a retrospective design, serum samples from only a few patients were available at baseline. Therefore, we were not able to investigate these and other disease-specific biomarkers of BP.8,13,19
Because of the retrospective design, we believed it was not feasible to include severity and extent of BP lesions (erosions and blisters) as a risk factor. However, it is possible that a higher dosage of glucocorticosteroids at discharge may be a surrogate marker for more severe disease. The same might be postulated for the increased erythrocyte sedimentation rate. However, the latter factor might be influenced by various other factors, including other acute or chronic diseases.
Serum albumin levels can be influenced by various diseases and conditions. Low serum albumin levels were also found to be a significant negative prognostic marker in women with coronary heart disease,20 in patients with hepatocellular cancer,21 in older patients undergoing dialysis,22 and generally in hospitalized patients.23 From the data presented herein, it is not clear whether decreased albumin levels are a marker of the severity of disease, ie, cutaneous and extensive mucosal erosions, a marker of other underlying diseases, or even a marker of malnutrition. In a recent study from the United States on predictors of survival for older hospitalized patients with various diagnoses, serum albumin levels did predict survival.24 However, as the albumin values were missing in 56% of the 1266 patients, serum albumin level was not included in the final model. In the final model, survival was predicted by the Acute Physiology Score of the Acute Physiology and Chronic Health Evaluation III, modified Glasgow Coma Scale score, major diagnosis, age, activities of daily living, exercise capacity, weight loss, and global quality of life.
What do these results imply? First, patients with severe BP (higher daily doses of glucocorticosteroids needed to suppress the disease and lower serum albumin levels), generally decreased health (elevated erythrocyte sedimentation rate after 1 hour and lower serum albumin levels), and greater age are at an increased risk of death within the first year after the first hospitalization. Consequently, these patients should be monitored carefully. Furthermore, the need for a high dosage of glucocorticosteroid therapy in patients with BP should be critically evaluated. This is supported by the fact that some patients with BP—presumably with milder disease—might be treated successfully with topical glucocorticosteroids only.25
The results of the present study have to be confirmed by a prospective trial that includes new biomarkers for the disease or its severity as well as the markers of survival defined by Teno et al.24 Furthermore, in future clinical trials, to decrease the risk of unweighted treatment groups, patients should be stratified for age and possibly for low serum albumin levels, as these factors were strong predictors of a higher mortality.
Accepted for publication July 25, 2001.
This study was supported in part (costs of meetings and travel) by the Verein zur Förderung Dermatologischer Forschung, Heidelberg, Germany, and Hoffman-LaRoche GmbH, Grenzach-Wyhlen, Germany.
This study was presented, in part, at the joint meeting of the Society of Investigative Dermatology and the International DermatoEpidemiology Association, Chicago, Ill, June 5, 1999.
Corresponding author and reprints: Berthold Rzany, MD, ScM, Center for Evidence Based Medicine in Dermatology, Department of Dermatology, Charité University Hospital, Schumannstr 20/21, D-10117 Berlin, Germany (e-mail: firstname.lastname@example.org).
Create a personal account or sign in to: