Lichen sclerosus and data science: using an age, site and gender specific disease to define correlation and causality

Introduction: Lichen sclerosus (LS) is a site, gender and age specific disease that affects the vulvar region of menopausal women. The menopausal transition is characterized by physiologic changes that impact basic processes, such as weight, sleep and drug clearance. Our current pharmacopeia includes drugs that target basic metabolic processes such as cholesterol, sex steroidogenesis and the gut microbiome. Introduction of drugs engages complex pathways that govern bile salt metabolism and the provision of sex steroid substrate. We hypothesized that lichen sclerosus in the vulva may represent an off target effect of pharmaceutical alteration of sex steroid substrate in some women. We compared biometric and drug data in women with lichen sclerosus with an age matched control group without disease. Methods: 43 women with lichen sclerosus underwent chart review to determine BMI, Fitzpatrick level, and pharmaceutical burden. This data was compared to 106 randomly selected age matched controls. Logistic regression was adjusted for age, BMI and Fitzpatrick photo typing. The statistical software was R version 3.0.1. Cases were defined as those with diagnosis claims for lichen sclerosus and were obtained from electronic records in a single private practice. Results: Proton pump inhibitors were noted in 12/43 women (28%) cases and 15/106 (14%) controls: p=0.048, (confidence interval CI=-0.01 to 0.287), adjusted p value 0.33; Hormone therapy 5/43 cases (12%), 26/106 (25%) controls: p value 0.078, (CI=-0.25 to 0.002), adjusted p value 0.02. Antihypertensive medications 17/43 (40%) cases, 29/106 (27%) controls, p value 0.14, (CI=-0.047 to 0.29), adjusted p value 0.8. Statins 16/43 cases (37%), 28/106 controls (26%); p=0.19, (CI=-0.05 to 0.27), adjusted p value 0.89. Average age among cases was 63, average age in the control group was 61. Average BMI was 27.5 in the case group, 23.6 in the control group. Logistic regression analysis was adjusted for age, BMI and Fitzpatrick phototyping. Conclusions: Borderline statistical significance was seen with proton pump inhibitors and hormone replacement therapy (HRT). The significance seen with proton pump inhibitors went away with adjustment for Fitzpatrick, age and BMI. The significance seen with HRT withstood adjustment, with the control group featuring more women on hormone replacement therapy than the case group.