[Skip to Content]
[Skip to Content Landing]
Views 246
Citations 0
Research Letter
June 5, 2019

Distribution of Self-reported Hidradenitis Suppurativa Age at Onset

Author Affiliations
  • 1Department of Dermatology, University of California, San Francisco, San Francisco
  • 2Division of Dermatology, Department of Medicine, Albert Einstein College of Medicine, Bronx, New York
  • 3Division of Dermatology, Department of Medicine, Women’s College Hospital, University of Toronto, Toronto, Ontario, Canada
  • 4York Dermatology Center, Richmond Hill, Ontario, Canada
  • 5Department of Population Health and Science Policy, Icahn School of Medicine at Mount Sinai, New York, New York
  • 6Department of Genetics and Genomics, Icahn School of Medicine at Mount Sinai, New York, New York
  • 7The Rockefeller University, New York, New York
JAMA Dermatol. Published online June 5, 2019. doi:10.1001/jamadermatol.2019.0478

Although hidradenitis suppurativa (HS) has a reported mean age at onset of 21.8 years,1 we have observed initial onset of HS lesions in individuals older than 40 years more commonly than expected. In this retrospective study, we investigated age at HS onset and its distribution.

With ethics board approval from all participating institutions, deidentified data from the medical records of 1203 patients with dermatologist-confirmed cases of HS from 3 North American HS practices were aggregated. Collected data included age at visit, sex, race, self-identified ethnicity, self-reported age at HS onset, and disease severity by Hurley stage. All analyses were conducted using R software, version 3.5.0. Wilcoxon tests were used to compare age at onset across sex, race, and disease severity. Cullen-Frey plots were used to identify age at onset as a log-normal distribution, and χ2 tests were used to test for scale and location mixtures if the onset distribution came from a unique population or was a mixture of distributions representing distinct subpopulations. Because sex was significantly associated with age at onset, mixture of regression with sex as a covariate was used, allowing identification of subpopulations considering covariates. Parametric bootstrapping was used to sequentially test the number of components in various models, and 2 subpopulations were identified. Estimation of distribution of subpopulations was carried out using the expectation maximization algorithm for the mixture of regressions with sex as a covariate.2

×