Pemphigus variant associated with penicillin use: A case-cohort study of 363 patients from Israel

Objective: To determine whether medication use is associated with the development of a pemphigus variant. Design: Population-based case-cohort study. Setting: Health maintenance organization in Israel. Methods: All incident pemphigus variant cases diagnosed from January 1, 1997, through December 31, 2001, among 1.5 million members were identified. A cohort of 150 000 was randomly selected from the health maintenance organization population as the control group. Data on case patients and control subjects, including all medication purchased during the 6 months before the diagnosis, were obtained using the health maintenance organization's central database. Results: We identified a total of 363 case patients diagnosed as having pemphigus during the 5-year study (6 961 853 person-years of follow-up). The mean age at diagnosis was 49.8 (SD, 22.7) years, and 53% of the cases were women. Results of a multivariate analysis showed that increased risk for pemphigus was associated with purchasing penicillin during the 6 months before the diagnosis (odds ratio, 2.03; 95% confidence interval, 1.56-2.64). Compared with individuals with no penicillin purchases, we calculated increased risks of 1.84 (95% CI, 1.36-2.49) and 3.02 (95% CI, 1.41-6.49) in those with 1 and 3 or more purchases, respectively. None of the other examined medications, including cephalosporins, angiotensin-converting enzyme inhibitors, dipyrone, anticonvulsants, and nonsteroidal anti-inflammatory drugs, showed similar risks. Conclusions: To our knowledge, the present research is one of the largest published epidemiological studies on pemphigus variant. The use of computerized medical and administrative databases allowed the detection of case patients in the community, resulting in a higher calculated incidence rate than previously reported. The findings suggest a relationship between the use of penicillin and pemphigus variant. Further studies to assess the nature of this statistical association are warranted.