Hidradenitis Suppurativa Phenotypic Heterogeneity: A Single Center Latent Class Analysis of 532 patients
Hidradenitis suppurativa (HS) is a heterogenous chronic inflammatory skin disease that lacks consensus regarding phenotypic classification. A stratification system of HS phenotypes may help individualize treatment and ensure accurate and timely diagnosis. This study aims to identify HS phenotypes using latent class analysis (LCA) to contribute to understanding of the disease. A total of 532 patients with HS were included from the outpatient dermatology clinics of Michigan Medicine in Ann Arbor, Michigan, USA. Two distinct HS phenotypes were identified. Latent class 1 (Axillary-Genital-Mammary) comprised 29.1% of patients and showed higher probabilities for severe axillary, genital, and mammary involvement. Latent class 2 (Axillary-Genital-Gluteal) included 70.9% of patients and had higher probabilities for mild axillary, genital, and gluteal involvement. Axillary-Genital-Mammary had an earlier age of onset, shorter disease duration, and a higher proportion of former smokers. This study identified two distinct HS phenotypes: a severe Axillary-Genital-Mammary and milder Axillary-Genital-Gluteal phenotype. Tailoring treatments based on phenotype-specific characteristics may improve outcomes. The study contributes to the growing understanding of HS phenotypes in the US population, offering insights for more individualized disease management.
Steps to reproduce
This was a retrospective review of consecutive patients diagnosed with and evaluated for HS at the outpatient dermatology clinics of Michigan Medicine in Ann Arbor, Michigan, USA. This study was deemed IRB exempt. We excluded identifiable variables. For questions please contact email@example.com and firstname.lastname@example.org To provide descriptive information, mean values with standard deviations were used to represent continuous data, while numbers accompanied by percentages were used for categorical data. Additionally, continuous variables such as duration of HS were converted into clinically relevant categorical variables. We used the following preselected variables to predict the Latent Classes (LCs): sex, BMI, age of HS onset, and duration of HS. Age of HS onset and duration of HS were determined by patient reporting that was documented in the chart. We selected the LC with the lowest Bayesian Information Criterion, as an indicator of best fit model. We calculated the conditional class probabilities of HS localizations for each LC. To evaluate the differences among the LCA classes, both categorical and continuous variables were subjected to the Pearson's chi-square test and one-way ANOVA test, respectively. We conducted multinomial logistic regressions to determine how certain variables of interest were associated with the odds LC class membership. The influence of variables on each class membership was quantified in terms of odds ratios (ORs) along with their corresponding 95% confidence intervals and P-values. Statistical significance was determined at a threshold of P-value < 0.05. Statistical analysis was performed with RStudio Version 1.4.1717.