STUDY. Prognostic Factors of Paraneoplastic Pemphigus

Transcription

1 ONLINE FIRST STUDY Prognostic Factors of Paraneoplastic Pemphigus Sandy Leger, MD; Damien Picard, MD, PhD; Saskia Ingen-Housz-Oro, MD; Jean-Philippe Arnault, MD; François Aubin, MD, PhD; Francis Carsuzaa, MD; Gilles Chaumentin, MD; Jacqueline Chevrant-Breton, MD; Olivier Chosidow, MD, PhD; Béatrice Crickx, MD, PhD; Michel D incan, MD, PhD; Michel Dandurand, MD; Stéphane Debarbieux, MD; Emmanuel Delaporte, MD, PhD; Olivier Dereure, MD, PhD; Marie-Sylvie Doutre, MD, PhD; Gérard Guillet, MD; Denis Jullien, MD, PhD; Ingrid Kupfer, MD; Jean-Philippe Lacour, MD; Fabienne Leonard, MD; Catherine Lok, MD, PhD; Laurent Machet, MD, PhD; Ludovic Martin, MD, PhD; Carle Paul, MD, PhD; Jean-Michel Pignon, MD; Caroline Robert, MD, PhD; Luc Thomas, MD, PhD; Pierre-Jean Weiller, MD, PhD; Vincent Ferranti, MD; Danièle Gilbert, PhD; Philippe Courville, MD; Estelle Houivet, PhD; Jacques Benichou, MD, PhD; Pascal Joly, MD, PhD Objective: To identify the prognostic factors of overall survival in a series of patients with paraneoplastic pemphigus (PNP). Design: Multicenter retrospective cohort study. Setting: Twenty-seven dermatology departments in France. Patients: A total of 53 patients (31 men and 22 women; median age, 59 years; age range, years) were diagnosed as having PNP between 1992 and Main Outcome Measures: Overall Kaplan-Meier survival rates were estimated, and features associated with survival were assessed using univariate (log-rank test) and multivariate (Cox regression) analyses. Results: The study included 53 patients with PNP. Thirtysix patients (68%) died during the study. The 1-, 3-, and 5-year overall survival rates were 49%, 41%, and 38%, respectively. The main causes of death were infections (n=21) and evolution of neoplasia (n=6). In univariate analysis, the main detrimental prognostic factors identified were erythema multiforme like skin lesions (P=.05) and histologic keratinocyte necrosis (P=.03). None of the 5 patients with Castleman disease died during the study. After adjustment for age and sex in multivariate analysis, erythema multiforme like skin lesions remained predictive of fatal outcome, with a 2-fold increase in death rate (hazard ratio [HR], 2.3; 95% CI, ; P=.04). The prognosis of patients with PNP was even poorer when erythema multiforme like skin lesions were associated with severe skin or mucosal involvement at presentation (HR of death, 3.0; 95% CI, ; P=.049). Conclusion: Patients with PNP with erythema multiforme like skin lesions and histologic keratinocyte necrosis, especially when associated with extensive lesions at presentation, are likely to have a more severe and rapid fatal outcome and should be managed very carefully. Arch Dermatol. 2012;148(10): Published online July 16, doi: /archdermatol Author Affiliations are listed at the end of this article. PARANEOPLASTIC PEMPHIGUS (PNP) is a rare type of pemphigus that is characterized by the production of autoantibodies directed against a complex of desmosomal proteins, including desmoplakin I and II; bullous pemphigoid(bp) antigen 1 (BPAG1); envoplakin; and periplakin, a recently identified 170-kDa antigen, as well as desmoglein (Dsg)-1 and Dsg It is mainly associated with lymphoproliferative disorders such as non- Hodgkin lymphoma (NHL), chronic lymphocytic leukemia (CLL), and Castleman disease. Nonlymphoidmalignantneoplasms are more rarely associated. 1,4,8-19 The prognosis of PNP is severe, with a mortality rate ranging from 75% to 90% and a mean survival of less than 1 year. 4,19-21 The main causes of death are the evolution of an associated malignant neoplasm, sepsis, orrespiratoryfailure. 4,21,22 However,patients with PNP and a long-term survival have also beendescribed. 14,17,23,24 Moreover,thecourse of PNP does not necessarily parallel the evolution of the associated neoplasm. Because of a widely variable course and the absence of prognostic factors identified to date, the treatment of PNP remains controversial. Many treatments have been proposed, including oral corticosteroids, cyclosporine, azathioprine, cyclophosphamide, mycophe- Author Affil the end of th 1165

2 nolate mofetil, plasmapheresis, and rituximab. 4,13,25-29 The aim of this study was to identify prognostic factors for PNP to improve treatment of patients. METHODS This retrospective study was conducted in 27 dermatology departments in France between 1992 and Serum samples from all patients suspected of having PNP were sent to the Immunodermatology Laboratory of Rouen University Hospital, Rouen, which is the national reference center for autoimmune blistering skin disorders and for immunofluorescence, immunoblot, and enzyme-linked immunosorbent assay (ELISA) testing. According to the diagnostic criteria proposed by Anhalt et al 1 and their respective sensitivity and specificity, 30,31 patients were included in the study if at least 4 of the 7 following criteria were met: (1) skin and/or mucosal lesions suggestive of pemphigus vulgaris, erythema multiforme, BP, or lichen planus; (2) associated neoplasia or lymphoproliferative disorder, which is commonly described in patients with PNP 4 ; (3) histologic findings showing acantholysis, intraepithelial blister, keratinocyte necrosis, and/or an inflammatory infiltrate of small lymphocytes along the dermoepidermal junction (DEJ); (4) IgG and/or C3 deposits on the epithelial cell surface (ECS) and along the DEJ by direct immunofluorescence; (5) presence of circulating antibodies binding both the ECS and the DEJ by indirect immunofluorescence (IIF) on normal human skin; (6) IIF labeling of rat bladder epithelium by patient s serum sample; and (7) recognition of 1 or more of the following bands on immunoblot assays: the 250-kDa band of desmoplakin I, the 230- kda band of BPAG1, the 210-kDa band of envoplakin, the 190- kda band of periplakin, and the 170-kDa band corresponding to 2 -macroglobulin like 1 protein. 7 The following data were collected from medical charts: medical history, associated neoplasia, dates of onset and diagnosis, type and localization of skin and/or mucosal lesions, treatment used, adverse effects of treatment, follow-up, and outcome. The severity of lesions was assessed clinically at the time of diagnosis. Two recently published scoring sytems the autoimmune bullous skin disorder intensity score instrument and the pemphigus disease area index 32 could not be used in this retrospective study because most of our patients were assessed clinically before the systems were reported. 33 Paraneoplastic pemphigus was considered severe if 1 or both of the following 2 criteria were present: (1) extensive skin involvement (alone or associated with mucosal lesions) with at least 20 skin lesions, and/or (2) mucosal involvement affecting at least 2 mucosal areas or only oral mucosae responsible for dysphagia or weight loss, according to Harman et al. 33 In the absence of the above criteria, PNP was defined as moderate. The dead or alive status of patients at the time of the study was systematically verified from their birthplace city hall death register. A histologic analysis of a representative lesion was performed in all cases. Histologic slides were reviewed by 1 of us (P.C.), and the following lesions were analyzed: suprabasal acantholysis, keratinocyte necrosis, lichenlike inflammatory infiltrate in the superficial dermis, and dermoepidermal blister. The results of direct immunofluorescence were collected from the initial skin or mucosal biopsy specimen. The IIF studies were performed on monkey esophagus and rat bladder. Immunoblot analyses on human epidermal extracts were performed according to a previously described technique. 31 Anti Dsg-1 and -3 antibodies were detected using commercially available Dsg-1 and Dsg-3 ELISA tests (MESACUP Dsg; Medicare & Biological Laboratories) with 1:100 diluted serum. Two ELISA tests were used to detect antienvoplakin antibodies: a commercially available ELISA (IgG; EUROIMMUN AG) and a previously published specific bead-based assay. 34 The overall survival was estimated for all patients with PNP and for various subgroups using the Kaplan-Meier method. Baseline clinical, histologic, and immunologic features were considered alone or in combination for their prognostic significance. For univariate analysis, associations between these features and overall survival were assessed using the log-rank test. Features that appeared associated with overall survival at the 0.1 level were then assessed by multivariate analysis using the Cox proportional hazard regression. For each of these features, a separate Cox regression analysis that included age ( 59 years, 59 years) and sex was performed to determine which was associated with overall survival independently of age and sex. Each analysis produced a hazard ratio (HR) estimate as well as the corresponding 95% confidence interval (CI) and P value. All data management and data analysis were conducted with NCSS version statistical software. RESULTS BASELINE CHARACTERISTICS OF PATIENTS Fifty-three patients (31 men and 22 women) were included in the study. The median age was 59 years (age range, years). The associated neoplasms were as follows: CLL (n=16); NHL (n=14); carcinoma (n=10); Castleman disease (n=5); thymoma (n=4); Hodgkin lymphoma (n=1); sarcoma (n=1); essential thrombocytemia (myeloproliferative disorder) (n=1); Waldenström disease (n=1); dysglobulinemia (n=1); mediastinal follicular dendritic cell tumor (n=1); and cutaneous CD30 lymphoma (n=1). Two neoplasms were associated in 3 patients. The associated neoplasia was evident before the occurrence of PNP in 36 cases (68%), with a mean delay of 2.2 years. An associated neoplasia was found concomitantly or after the diagnosis of PNP in 8 (15%) and 9 (17%) cases, respectively. Mucosal lesions were present in 50 of the 53 patients (94%). Five patients had exclusive mucosal involvement without skin lesions. The mucosae involved were oral (n=47), conjunctival (n=28), genital (n=27), pharyngeal and/or laryngeal (n=12), esophageal (n=4), and bronchial (n=3). Forty-eight patients (90%) had pemphiguslike (n=28), erythema multiforme like (n=10), lichen planus like (n=8), BP-like (n=6), or linear IgA dermatosis like (n=1) lesions. According to the definition given herein, disease severity was considered moderate in 41 patients and severe in 12 patients. The most frequent histologic finding was acantholysis (n=35 [66%]). Other histologic features were keratinocyte necrosis (n=24 [45%]), lichenlike inflammatory infiltrate in the superficial dermis (n=24 [45%]), and dermoepidermal blister (n=12 [23%]). Direct immunofluorescence showed deposits of C3 or IgG on the ECS in 34 (64%) and 42 (79%) cases, respectively, and on the DEJ in 17 (32%) and 8 (15%) cases, respectively. Dual ECS and DEJ deposits of C3 and/or IgG were found in 17 cases (32%). Indirect immunofluorescence of patients serum samples on monkey esophagus was performed in 40 of the 53 cases, and the results were positive in 30 cases (75%). Intercellular staining and dual intercellular and 1166

3 Table 1. Criteria Used for Diagnosis of Paraneoplastic Pemphigus (PNP) According to Associated Neoplasm Criteria a Carcinoma NHL CLL (n = 10) b (n = 14) b (n = 16) Castleman Disease/Thymoma (n=9) b Other Hematologic Disorders (n=5) Other Neoplasia (n=2) b Mucosal and skin lesions suggestive of PNP Exclusive mucosal lesions Exclusive skin lesions PNP histologic findings Dual labeling (ECS BMZ) on DIF Positive IIF reaction on human skin Positive IIF reaction on rat bladder Recognition of 1 or more of PNP antigens on immunoblot assay and/or ELISA Patients with 4 or 5 PNP diagnostic criteria Patients with 6 or 7 PNP diagnostic criteria Abbreviations: BMZ, basement membrane zone; CLL, chronic lymphocytic leukemia; DIF, direct immunofluorescence; ECS, epidermal cell surface; ELISA, enzyme-linked immunosorbent assay; IIF, indirect immunofluorescence; NHL, non-hodgkin lymphoma. a From Anhault et al. 1 b Three patients had 2 neoplasms; therefore, the total number of neoplasms indicated in Table 1 is 56. DEJ staining were observed in 20 cases (50%) and 1 case, respectively. A positive IIF staining result on rat bladder was observed in 37 of 46 cases (80%). The results of immunoblot analyses of patients serum samples on human epidermal extracts are shown below. Antigen (kda) No. (%) Periplakin (190) 35 (66) Envoplakin/desmoplakin (210) 34 (64) DSG-3 (130) 12 (23) BPAG1 (230) 7 (13) Desmoplakin I (250) 7 (13) DSG-1 (160) 7 (13) Forty-two of 53 serum samples (79%) recognized at least 1 of the following 190-, 210-, 230-, or 250-kDa bands of periplakin, envoplakin, BPAG1, and desmoplakin I. Thirtynine serum samples (74%) recognized the 190- and/or 210-kDa bands of periplakin and envoplakin, including 30 serum samples that recognized the 190-kDa/210-kDa doublet and 9 serum samples that recognized only 1 of these 2 bands. Also, 3 serum samples recognized the 250-kDa (n=2) and/or 230-kDa bands exclusively. Thirty-one serum samples were available for retesting both by immunoblotting and by the commercially available antienvoplakin ELISA and our bead-based antienvoplakin assay. Of the 31 serum samples, 15 (48%) and 22 (71%) recognized envoplakin with the commercially available ELISA and our bead-based assay, respectively, and 24 (77%) recognized the 190- and/or 210-kDa bands of envoplakin and/or periplakin (including 20 serum samples [65%] that recognized the 190-kDa/210-kDa doublet). Also, 2 of the 31 serum samples recognized the 250-kDa band of desmoplakin alone. The last 5 serum samples (16%) did not react with any band of the PNP complex by immunoblotting. Interestingly, 2 of these 5 immunoblotting-negative serum samples were strongly positive (ELISA values, 5.65 and 3.26), and 3 serum samples had weak ELISA reactivity (ELISA values, 1.58, 1.32, and 1.16). Desmoglein- ELISA tests were performed in 42 patients. Anti Dsg-1 and -3 antibodies were detected in 17 (40%) and 30 (71%) cases, respectively. Overall, 2 patients with NHL whose serum sample was no longer available to be retested by ELISA had only 4 of the 7 diagnostic criteria (these 2 patients did not have circulating antibodies by IIF and immunoblotting and therefore lacked the 3 PNP diagnostic criteria based on serum analysis), 16 patients had 5 criteria, 20 had 6 criteria, and 15 had all 7 criteria. The PNP diagnostic criteria of subgroups of patients depending on the associated neoplasm are shown in Table 1. We carefully examined the PNP diagnostic criteria in the 3 patients with exclusive skin involvement. The associated neoplasms were 1 CLL, 1 NHL, and 1 cutaneous CD30 T-cell lymphoma. The 190- and 210-kDa bands of periplakin and envoplakin were recognized in 3 of 3 cases, and an IIF labeling of rat bladder was observed in 2 of the 3 serum samples. Two and 1 of these 3 patients had 5 and 6 PNP diagnostic criteria, respectively. We then examined the PNP diagnostic criteria in the 11 patients whose serum samples did not recognize the PNP complex by immunoblotting. Five of these 11 immunoblotting-negative serum samples were available for retesting by ELISA and had positive envoplakin ELISA values. The associated neoplasms were 5 NHL, 4 CLL, and 2 carcinomas (serum samples from the 2 latter patients showed a weak labeling on rat bladder by IIF). Two patients with NHL and no circulating antiepidermal antibodies had 4 of the 7 PNP diagnostic criteria (lacking the 3 criteria based on serum analysis), and 7 and 2 patients had 5 and 6 PNP diagnostic criteria, respectively. Finally, we examined the PNP diagnostic criteria in the 10 patients with an associated carcinoma. All 10 patients had mucosal lesions suggestive of PNP. Eight of the 10 serum samples (80%) recognized the 190- and/or 210-kDa bands of periplakin and envoplakin. Four patients had 5 PNP diagnostic criteria, and 5 and 1 had 6 and 7 diagnostic criteria, respectively. TREATMENT AND OUTCOME High doses of oral corticosteroids, eg, prednisone, 1 to 1.5 mg/kg/d, were used as the first line of treatment in 1167

4 Table 2. Main Treatments Used (First and Second Line) and Their Results, Depending on Paraneoplastic Pemphigus Severity Disease Severity Moderate (n = 41) Severe (n = 12) CS No. (%) First-Line Treatment Complete Epithelialization Second-Line Treatment CS IS Topical CS Other CS CS IS Patients Alive at the End of the Study Topical CS Other CS CS IVIG CS CIS CS Ritux CS CS IVIG CS CIS CS Ritux 28 (76) 3 (100) 8 (89) 2 (50) 10 (36) 3 (100) 4 (50) 1 (50) (75) 0 1 (8) 2 (17) 1 (8) (8) Abbreviations: CS, corticosteroids; CS CIS, CS combined with conventional immunosuppressants (azathioprine, cyclosporine, mycophenolate mofetil, methotrexate); CS IS, CS combined with immunosuppressive drugs; Other, thalidomide (n = 1) or rituximab alone (n = 1) or no treatment (n = 2); CS IVIG, corticosteroids combined with intravenous immunoglobulins; CS Ritux, corticosteroids combined with rituximab. 40 of the 53 patients (75%). Corticosteroids were used alone in 37 cases (70%). Four patients (8%) were treated with an immunosuppressant as first-line treatment, including 3 cases treated with corticosteroids and immunosuppressants (azathioprine [n=2] and rituximab [n=1]) and 1 case treated with rituximab alone without corticosteroids. Topical corticosteroids were used alone in 9 patients (17%) with a moderate type of PNP. Four of these 9 patients had limited skin lesions ( 5% of the body surface area) with limited or no mucosal involvement. One patient with isolated pemphigus vulgaris like oral lesions was treated with thalidomide alone. Two patients did not receive any specific treatment for PNP because of spontaneous healing of lesions in one case and early death after diagnosis in the other case. The main treatment regimens used as first line, and in some patients as second line, according to disease severity are shown in Table 2. Treatment of the associated neoplasm was performed in 49 patients (92%), including chemotherapy in 36 cases (68%), and/or surgery in 15 cases (28%), and/or radiotherapy in 9 cases (17%). Healing occurred in 32 of the 48 patients (67%) with skin lesions after a median delay of 45 days (range, 7 days to 18 months). Complete epithelialization of mucosal lesions was achieved in 27 of 50 patients (54%) after a median delay of 60 days (range, 7 days to 20 months). Of the 37 patients treated with corticosteroids alone as the first line of treatment, 26 (70%) did not achieve complete epithelialization. An immunosuppressant was added secondarily in 11 cases. The 15 other patients were treated secondarily with immunoglobulin infusions (n=2) or died rapidly before the initial treatment had been modified. The median follow-up of alive patients was 77 months (range, months). Overall, 36 patients (68%) died during the study period. The 1-, 3-, and 5-year overall survival rates were 49%, 41%, and 38% respectively (Figure). The main causes of death were infections in 21 cases (58%), evolution of neoplasia in 6 cases (17%), and respiratory failure in 3 cases (8%). The cause of death remained unknown in 6 cases. The diagnosis of bronchiolitis obliterans was suspected in 2 of the 3 patients who died of acute respiratory failure. This diagnosis was histologically confirmed in the third patient as well as in another patient with Castleman disease who was still alive at the end of the study after a 46-month follow-up period. The main causes of death according to disease severity and associated neoplasia are shown in Table 3. PROGNOSTIC FACTORS In univariate analysis, the main deleterious prognostic factors, which were associated with a shorter overall survival, were the presence of erythema multiforme like skin lesions (P=.05) and keratinocyte necrosis in skin biopsy specimens (P=.03) (Table 4). When the whole population of patients was considered, the type of associated neoplasia was not associated with overall survival (Figure). However, the patients with NHL had a significantly shorter survival when compared with patients with other types of neoplasia, excluding CLL and carcinoma (P=.03). Indeed, all 5 patients with Castleman disease were alive at the end of the study. There was no association between sex or age and patients prognosis whether age was coded as a continuous variable or dichotomized ( 59 years, 59 years). No particular PNP antigen patterns recognized by patients serum samples were found to be associated with prognosis on immunoblot analysis. Interestingly, the initial extent of skin and/or mucosal lesions at baseline was not associated with patients prognosis, although there was a nonsignificant trend for patients with initial lesions classified as severe to have shorter survival than patients with moderate lesions (P=.11). In age- and sex-adjusted multivariate analysis, only the presence of erythema multiforme like skin lesions (HR, 2.3; 95% CI, ; P=.037) remained associated with shorter survival. The overall survival of patients with PNP was even shorter when erythema multiforme like skin lesions were associated with severe skin and/or mucosal involvement at presentation (HR, 3.0; 95% CI, ; P=.049). The presence of histologic keratinocyte necrosis was no longer found to be associated with overall survival in multivariate analysis (HR, 1.8; 95% CI, ; P=.09). However, the presence of histologic keratinocyte necrosis when associated with severe disease at presentation (HR, 2.5; 95% CI, ; P=.05) or the presence of erythema multiforme like skin lesions (HR, 2.7; 95% CI, ; P=.02) was found to be associated with a fatal outcome. COMMENT The main findings of this study are (1) the demonstration that PNP has a highly variable course in both se- 1168

5 CSurvival Rate ASurvival Rate DSurvival Rate BSurvival Rate Other NHL Carcinoma CLL Years Years Presence of EM-like eruption Absence of EM-like eruption Presence of keratinocyte necrosis Absence of keratinocyte necrosis Years Years Figure. Overall survival distribution of the 53 patients with paraneoplastic pemphigus. A, Whole population. B, Patients with non-hodgkin lymphoma (NHL), chronic lymphocytic leukemia (CLL), carcinoma, or other types of neoplasm (Other). C, Patients with or without erythema multiforme (EM)-like lesions. D, Patients with or without keratinocyte necrosis. Table 3. Main Causes of Death According to Disease Severity and Associated Neoplasia Variable Death, No. (%) Infections Evolution of Neoplasia Respiratory Failure Other or Undetermined Disease severity Moderate (n = 41) 25 (61) Severe (n = 12) 11 (92) Associated neoplasia NHL (n = 14) 12 (86) CLL (n = 16) 11 (69) Carcinoma (n = 10) 6 (60) Other (n = 13) 7 (54) Abbreviations: CLL, chronic lymphocytic leukemia; NHL, non-hodgkin lymphoma. No. vere and indolent cases, and (2) the main deleterious prognostic factors are the presence of erythema multiforme like skin lesions and the presence of keratinocyte necrosis on histologic examination, especially when the latter is associated with the presence of extensive skin and/or mucosal lesions at presentation. The main clinical, histologic, and immunologic features of the 53 patients included in the present study were similar to those previously reported in different series of patients with PNP 31,35 : (1) frequent association with lymphoproliferative disorders (75%), eg, NHL and CLL; (2) high frequency of erosive mucosal lesions; (3) polymorphism of skin lesions; (4) histologic features associated with keratinocyte necrosis, acantholysis, and a lichen planus 1169

6 Table 4. Univariate Analysis of Prognostic Factors of Survival in Patients With Paraneoplastic Pemphigus Characteristic Dead Patients (n = 36) Alive Patients (n = 17) P Value a Sex.95 Male 22 9 Female 14 8 Age, y Associated neoplasia.15 NHL (vs Other) CLL (vs NHL) Carcinoma (vs Other) Other (vs CLL) Type of skin lesions Pemphiguslike Bullous pemphigoid like Erythema multiforme like Lichen planus like Disease severity.11 Moderate Severe 11 1 Histologic features Keratinocyte necrosis Lichenlike infiltrate Acantholysis Abbreviations: CLL, chronic lymphocytic leukemia; NHL, non-hodgkin lymphoma. a Calculated by the log-rank test. like inflammatory infiltrate; and (5) frequent recognition of envoplakin and periplakin by immunoblot analysis of patients serum samples. The prognosis of patients with PNP is usually considered poor, with a mortality rate ranging from 75% to 90%. 4,19-21 The 49%, 41%, and 38% 1-, 2- and 5-year overall survival rates based on the present study confirmed the severe prognosis of PNP, although this prognosis was rather better than previously reported. 4,19-21 Most deaths were observed during the first year after diagnosis, whereas survival curves only slightly decreased in patients who were still alive 1 year after the diagnosis of PNP. Interestingly, the survival of patients in the present series was longer than previously reported in the literature. 4,19-21 This discrepancy is likely attributable to the fact that the high mortality rates previously reported in patients with PNP were estimated from small series or even case studies, which probably reported the most severe cases, whereas the present study included all consecutive cases of PNP referred to the 27 dermatology departments of the French Study Group on Autoimmune Bullous Skin Disease during an 18-year period. Also, only a few patients in this series died of respiratory failure due to bronchiolitis obliterans, which is associated with a high death rate in some series of patients with PNP. 36 In fact, the frequency of bronchiolitis obliterans is highly variable in the literature, from 93% in a series of 28 patients with PNP, including 15 children and adolescents who all had Castleman disease, 36 to 18% and 17% in 2 series that included 10% and 6% of patients with Castleman disease 4,37 and 8% in the present series, which included 9% of patients with Castleman disease. Remarkably, the 9% rate of Castleman disease in the present study was very close to the 10% and 6% rates found in the 2 adult series of patients with PNP described by Anhalt 4 and Maldonado et al. 37 Although it has been suggested that bronchiolitis obliterans in patients with Castleman disease might reflect a more severe type of PNP, none of the 5 patients with Castleman disease in the present series died during the study. However, it should be emphasized that in all series of the literature, including the present one, the diagnosis of bronchiolitis obliterans is more frequently suspected than histologically proved. We observed a wide spectrum in the severity of skin and mucosal lesions in our series, as 41 (77%) and 12 (23%) patients had the moderate or severe type of PNP, respectively. Accordingly, the treatment regimens used were highly variable, from topical corticosteroids to high doses of oral corticosteroids combined with immunosuppressants. Although, to our knowledge, no direct comparison of the combination of immunosuppressants and corticosteroids vs corticosteroids alone has been published in the literature to date, immunosuppressive drugs are the most often recommended treatment of PNP. 38 Only 4 patients in our study received an immunosuppressive drug as the first line of treatment. In view of the rather low mortality rate in this series compared with others in the literature, our findings suggest that treatment of PNP should preferentially be adapted to disease severity to limit severe adverse effects, such as infections, in these immunocompromised patients and not to favor the evolution of the associated neoplasia. Interestingly, 4 of 9 cases of PNP treated with topical corticosteroids alone could even be adequately controlled without systemic treatment. The prognostic factors identified in this study were in accordance with the main causes of death of patients in this series. Indeed, patients with Castleman disease or thymoma had a favorable prognosis compared with patients with other types of malignant neoplasms, particularly NHL, possibly because they were not treated with chemotherapy, which highly increased the risk of severe infections and death in patients with PNP who were concomitantly treated with high doses of corticosteroids. Moreover, Castleman disease occurred in younger patients (mean age, 44 years) compared with patients with NHL (mean age, 59 years), CLL (mean age, 65 years), or carcinoma (mean age, 63 years), which might also explain the better prognosis of these patients, although age was not associated with prognosis in univariate analysis. The presence of erythema multiforme like lesions and keratinocyte necrosis in association with extensive skin involvement was reminiscent of toxic epidermal necrolysis and might reflect a more severe type of PNP. Patients with erythema multiforme like lesions were more likely to die because of a higher risk of severe infections, especially patients who did not achieve disease control after initial treatment and had an increase in their immunosuppressive treatment. The poor prognosis of toxic epidermal necrolysis like forms of PNP has previously been reported in some cases. 39 It is possible that patients with PNP and erythema multiforme like lesions and keratinocyte necrosis might have high serum levels 1170

7 of tumor necrosis factor and Fas ligand, as previously reported in toxic epidermal necrolysis. 40,41 As in all retrospective studies, various biases are possible in this study. To avoid selection bias of cases, we aimed at exhaustively recruiting all consecutive cases of PNP from the 27 dermatology departments of the French Study Group on Autoimmune Bullous Diseases, which allowed us to include not only the most severe forms of PNP but also a significant proportion of patients with moderate disease, who have more rarely been described in the literature than patients with severe cases of PNP. This exhaustive recruitment, which included both severe and moderate cases of PNP, might be responsible for the better prognosis and the slightly lower rate of positive immunoblotting results as compared with series that included a limited number of patients or that used highly selected serum samples to characterize PNP antigens. 7,42,43 Information bias is a common problem in retrospective studies. However, the medical charts used in the centers belonging to our group, although not standardized, were quite precise. Moreover, the dead or alive status of patients at the time of the study was systematically verified from their birthplace city hall death register. Finally, despite the limited sample size of this study, which is attributable to the fact that PNP is an extremely rare disorder, we performed an age- and sexadjusted multivariate analysis, which confirmed the association between the presence of erythema multiforme like skin lesions and fatal outcome, especially in association with severe involvement of skin and mucosal lesions or with keratinocyte necrosis, suggesting that combinations of some markers of disease severity could define subgroups of patients with severe prognosis. In conclusion, this study showed that the prognosis of PNP is heterogeneous and rather better than previously reported. Treatment regimens should be adapted to disease severity, from mild treatments in patients with limited disease and in the most immunocompromised patients to aggressive regimens in patients with the most severe disease. The prognostic factors identified in the present study could be a useful tool for optimal treatment of patients with PNP. Accepted for Publication: April 26, Author Affiliations: Departments of Dermatology, Institut National de la Santé et de la Recherche Médicale (INSERM) U905, Institute for Research and Innovation in Biomedicine, Rouen University Hospital, University of Normandy, Rouen (Drs Leger, Picard, Ferranti, and Joly), Henri Mondor University Hospital, Assistance Publique Hôpitaux de Paris, Créteil (Drs Ingen-Housz-Oro and Chosidow), Institut Gustave Roussy, Villejuif (Drs Arnault and Robert), Besançon University Hospital, Besançon (Dr Aubin), Sainte Anne Hospital, Hôpital d Instruction des Armées Sainte Anne,Toulon (Dr Carsuzaa), Rennes University Hospital, Rennes (Dr Chevrant-Breton), Bichat University Hospital, Assistance Publique Hôpitaux de Paris, Paris (Dr Crickx), Clermont-Ferrand University Hospital, Clermont-Ferrand (Dr D incan), Nîmes University Hospital, Nîmes (Dr Dandurand), Lyon Sud University Hospital, Pierre Bénite (Drs Debarbieux and Thomas), Lille University Hospital, Lille (Dr Delaporte), Montpellier University Hospital, Montpellier (Dr Dereure), Bordeaux Haut Levesque University Hospital, Bordeaux (Dr Doutre), Poitiers University Hospital, Poitiers (Dr Guillet), Lyon Edouard Flerriot University Hospital, Lyon (Dr Jullien), Cornouaille Hospital, Quimper (Dr Kupfer), Nice University Hospital, Nice (Dr Lacour), Clinique Courlancy, Reims (Dr Leonard), Amiens University Hospital, Amiens (Dr Lok), Tours University Hospital, Tours (Dr Machet), Angers University Hospital, Angers (Dr Martin), and Toulouse Purpan University Hospital, Toulouse (Dr Paul), France; Department of Hematology, Dunkerque Hospital, Dunkerque, France (Dr Pignon); Departments of Internal Medicine, Roanne Hospital, Roanne (Dr Chaumentin), and Marseille Timone University Hospital, Assistance Publique Hôpitaux de Marseille, Marseille (Dr Weiller), France; and INSERM U905 (Dr Gilbert), Department of Pathology, INSERM U905 (Dr Courville), and Department of Biostatistics, INSERM U657 (Drs Houivet, and Benichou), Rouen University Hospital, University of Rouen, Rouen, France. Published Online: July 16, doi: /archdermatol Correspondence: Pascal Joly, MD, PhD, Clinique Dermatologique, Hôpital Charles Nicolle, 1 Rue de Germont, F Rouen CEDEX, France Author Contributions: Drs Leger, Picard, and Joly 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. Study concept and design: Leger, D incan, and Joly. Acquisition of data: Leger, Ingen-Housz-Oro, Arnault, Aubin, Carsuzaa, Chaumentin, Chevrant-Breton, Chosidow, Crickx, D incan, Dandurand, Debarbieux, Delaporte, Dereure, Doutre, Guillet, Jullien, Kupfer, Lacour, Leonard, Lok, Machet, Martin, Paul, Pignon, Robert, Thomas, Weiller, Ferranti, Gilbert, Courville, and Joly. Analysis and interpretation of data: Leger, Picard, Lok, Houivet, Benichou, and Joly. Drafting of the manuscript: Leger, Picard, Dandurand, Kupfer, Ferranti, and Joly. Critical revision of the manuscript for important intellectual content: Ingen-Housz-Oro, Arnault, Aubin, Carsuzaa, Chaumentin, Chevrant-Breton, Chosidow, Crickx, D incan, Debarbieux, Delaporte, Dereure, Doutre, Guillet, Jullien, Lacour, Leonard, Lok, Machet, Martin, Paul, Pignon, Robert, Thomas, Weiller, Gilbert, Courville, Houivet, Benichou, and Joly. Statistical analysis: Leger, Houivet, Benichou, and Joly. Obtained funding: Courville and Joly. Administrative, technical, and material support: Aubin, Dereure, Jullien, Paul, Robert, Thomas, Ferranti, and Joly. Study supervision: Leger, Picard, Crickx, Lacour, Lok, and Joly. Financial Disclosure: None reported Additional Contributions: Richard Medeiros, Rouen University Hospital, assisted in editing the manuscript. REFERENCES 1. Anhalt GJ, Kim SC, Stanley JR, et al. Paraneoplastic pemphigus: an autoimmune mucocutaneous disease associated with neoplasia. N Engl J Med. 1990; 323(25): Joly P, Thomine E, Gilbert D, et al. Overlapping distribution of autoantibody speci- 1171

8 ficities in paraneoplastic pemphigus and pemphigus vulgaris. J Invest Dermatol. 1994;103(1): Horn TD, Anhalt GJ. Histologic features of paraneoplastic pemphigus. Arch Dermatol. 1992;128(8): Anhalt GJ. Paraneoplastic pemphigus. Adv Dermatol. 1997;12: Oursler JR, Labib RS, Ariss-Abdo L, Burke T, O Keefe EJ, Anhalt GJ. Human autoantibodies against desmoplakins in paraneoplastic pemphigus. J Clin Invest. 1992;89(6): Borradori L, Trüeb RM, Jaunin F, Limat A, Favre B, Saurat JH. Autoantibodies from a patient with paraneoplastic pemphigus bind periplakin, a novel member of the plakin family. J Invest Dermatol. 1998;111(2): Schepens I, Jaunin F, Begre N, et al. The protease inhibitor alpha-2-macroglobulin like-1 is the p170 antigen recognized by paraneoplastic pemphigus autoantibodies in human. PLoS One. 2010;5(8):e Lemon MA, Weston WL, Huff JC. Childhood paraneoplastic pemphigus associated with Castleman s tumour. Br J Dermatol. 1997;136(1): Jansen T, Plewig G, Anhalt GJ. Paraneoplastic pemphigus with clinical features of erosive lichen planus associated with Castleman s tumor. Dermatology. 1995; 190(3): Kaplan I, Hodak E, Ackerman L, Mimouni D, Anhalt GJ, Calderon S. Neoplasms associated with paraneoplastic pemphigus: a review with emphasis on nonhematologic malignancy and oral mucosal manifestations. Oral Oncol. 2004; 40(6): Camisa C, Helm TN. Paraneoplastic pemphigus is a distinct neoplasia-induced autoimmune disease. Arch Dermatol. 1993;129(7): Dega H, Laporte JL, Joly P, et al. Paraneoplastic pemphigus associated with Hodgkin s disease. Br J Dermatol. 1998;138(1): Wang J, Zhu X, Li R, et al. Paraneoplastic pemphigus associated with Castleman tumor: a commonly reported subtype of paraneoplastic pemphigus in China. Arch Dermatol. 2005;141(10): Barbetakis N, Samanidis G, Paliouras D, et al. Paraneoplastic pemphigus regression after thymoma resection. World J Surg Oncol. 2008;6: Krunic AL, Kokai D, Bacetic B, et al. Retroperitoneal round-cell liposarcoma associated with paraneoplastic pemphigus presenting as lichen planus pemphigoideslike eruption. Int J Dermatol. 1997;36(7): Schlesinger T, McCarron K, Camisa C, Anhalt G. Paraneoplastic pemphigus occurring in a patient with B-cell non-hodgkin s lymphoma. Cutis. 1998;61(2): Perniciaro C, Kuechle MK, Colón-Otero G, Raymond MG, Spear KL, Pittelkow MR. Paraneoplastic pemphigus: a case of prolonged survival. Mayo Clin Proc. 1994;69(9): Herrada J, Phillips DK, Hebert AA, Cabanillas F, Jordon RE. A progressive blistering eruption in a patient with lymphoma: paraneoplastic pemphigus. Arch Dermatol. 1997;133(1):97-, Rodot S, Botcazou V, Lacour JP, et al. Paraneoplastic pemphigus: review of the literature, apropos of a case associated with chronic lymphoid leukemia. Rev Med Interne. 1995;16(12): Mutasim DF, Pelc NJ, Anhalt GJ. Paraneoplastic pemphigus. Dermatol Clin. 1993; 11(3): Nousari HC, Deterding R, Wojtczack H, et al. The mechanism of respiratory failure in paraneoplastic pemphigus. N Engl J Med. 1999;340(18): Fullerton SH, Woodley DT, Smoller BR, Anhalt GJ. Paraneoplastic pemphigus with autoantibody deposition in bronchial epithelium after autologous bone marrow transplantation. JAMA. 1992;267(11): Camisa C, Helm TN, Liu YC, et al. Paraneoplastic pemphigus: a report of three cases including one long-term survivor. J Am Acad Dermatol. 1992;27(4): Mascaró JM Jr, Ferrando J, Solé MT, et al. Paraneoplastic pemphigus: a case of long-term survival associated with systemic lupus erythematosus and polymyositis. Dermatology. 1999;199(1): Nousari HC, Brodsky RA, Jones RJ, Grever MR, Anhalt GJ. Immunoablative highdose cyclophosphamide without stem cell rescue in paraneoplastic pemphigus: report of a case and review of this new therapy for severe autoimmune disease. J Am Acad Dermatol. 1999;40(5, pt 1): Borradori L, Lombardi T, Samson J, Girardet C, Saurat JH, Hügli A. Anti-CD20 monoclonal antibody (rituximab) for refractory erosive stomatitis secondary to CD20 follicular lymphoma associated paraneoplastic pemphigus. Arch Dermatol. 2001;137(3): Becker LR, Bastian BC, Wesselmann U, Karl S, Hamm H, Bröcker EB. Paraneoplastic pemphigus treated with dexamethasone/cyclophosphamide pulse therapy. Eur J Dermatol. 1998;8(8): Heizmann M, Itin P, Wernli M, Borradori L, Bargetzi MJ. Successful treatment of paraneoplastic pemphigus in follicular NHL with rituximab: report of a case and review of treatment for paraneoplastic pemphigus in NHL and CLL. Am J Hematol. 2001;66(2): Izaki S, Yoshizawa Y, Kitamura K, et al. Paraneoplastic pemphigus: potential therapeutic effect of plasmapheresis. Br J Dermatol. 1996;134(5): Helou J, Allbritton J, Anhalt GJ. Accuracy of indirect immunofluorescence testing in the diagnosis of paraneoplastic pemphigus. J Am Acad Dermatol. 1995; 32(3): Joly P, Richard C, Gilbert D, et al. Sensitivity and specificity of clinical, histologic, and immunologic features in the diagnosis of paraneoplastic pemphigus. J Am Acad Dermatol. 2000;43(4): Rosenbach M, Murrell DF, Bystryn JC, et al. Reliability and convergent validity of two outcome instruments for pemphigus. J Invest Dermatol. 2009;129(10): Harman KE, Seed PT, Gratian MJ, Bhogal BS, Challacombe SJ, Black MM. The severity of cutaneous and oral pemphigus is related to desmoglein 1 and 3 antibody levels. Br J Dermatol. 2001;144(4): Mouquet H, Drenovska K, Lartigue A, et al. Detection and characterization of antienvoplakin linker autoantibodies in paraneoplastic pemphigus using specific beadbased assay. Clin Immunol. 2008;129(2): Ohyama M, Amagai M, Hashimoto T, Nousari HC, Anhalt GJ, Nishikawa T. Clinical phenotype and anti-desmoglein autoantibody profile in paraneoplastic pemphigus. J Am Acad Dermatol. 2001;44(4): Nikolskaia OV, Nousari CH, Anhalt GJ. Paraneoplastic pemphigus in association with Castleman s disease. Br J Dermatol. 2003;149(6): Maldonado F, Pittelkow MR, Ryu JH. Constrictive bronchiolitis associated with paraneoplastic autoimmune multi-organ syndrome. Respirology. 2009;14(1): Mignogna MD, Fortuna G, Leuci S, Adamo D, Ruoppo E. Metastatic prostate cancer presenting as paraneoplastic pemphigus: a favourable clinical response to combined androgen blockade and conventional immunosuppressive therapy. Br J Dermatol. 2009;160(2): Rybojad M, Leblanc T, Flageul B, et al. Paraneoplastic pemphigus in a child with a T-cell lymphoblastic lymphoma. Br J Dermatol. 1993;128(4): Nassif A, Moslehi H, Le Gouvello S, et al. Evaluation of the potential role of cytokines in toxic epidermal necrolysis. J Invest Dermatol. 2004;123(5): Murata J, Abe R, Shimizu H. Increased soluble Fas ligand levels in patients with Stevens-Johnson syndrome and toxic epidermal necrolysis preceding skin detachment. J Allergy Clin Immunol. 2008;122(5): Probst C, Schlumberger W, Stöcker W, et al. Development of ELISA for the specific determination of autoantibodies against envoplakin and periplakin in paraneoplastic pemphigus. Clin Chim Acta. 2009;410(1-2): Nagata Y, Karashima T, Watt FM, Salmhofer W, Kanzaki T, Hashimoto T. Paraneoplastic pemphigus sera react strongly with multiple epitopes on the various regions of envoplakin and periplakin, except for the c-terminal homologous domain of periplakin. J Invest Dermatol. 2001;116(4):