12-14 million annual covered lives in the United States. The patient population included in the dataset are broadly representative of the demographics of the United States population with respect to gender, age, and geographic distribution. These data include both medical and pharmacy claims, as well as patient demographic information such as age and gender.
Study Design and Study Population
Inclusion criteria were (1) women who were enrolled in the OptumInsight Clinformatics DataMart with at least two claims associated with an International Classification of Diseases (ICD) Ninth or Tenth Revision code for acne separated by 12 months, (2) at least 6 months of continuous enrollment in the dataset before the start date of the index course of therapy without any prescriptions for spironolactone, an oral antibiotic, or isotretinoin during this period, (3) at least 12 months of continuous enrollment in the dataset after the start of the index course of therapy, and (4) who were between 16 and 35 years of age at the start date of the index course of therapy were included in the study. Previous studies have validated the accuracy of ICD codes to identify patients with acne.29 A 6-month period of continuous enrollment before the index course of therapy was chosen to increase the probability that the index course of therapy was the initial treatment for these patients. An age cutoff of 35 years was chosen since acne is less common after this age and to reduce the likelihood of including patients with rosacea rather than acne.2 Exclusion criteria included an ICD code for rosacea, since systemic antibiotics are also commonly used for this condition. Prescriptions for spironolactone, oral antibiotics, and isotretinoin were identified by their National Drug Codes. Prescriptions were consolidated into courses of therapy with the start date defined as the date of the first prescription of the series and the end date defined as the date of the last prescription in the series plus the number of days of medication supplied. To account for non-adherence that could result in delays between prescriptions, prescriptions separated by fewer than 30 days were considered to be part of the same course of therapy.30–36 Courses shorter than 30 days were excluded to avoid including prescriptions for acute illnesses (eg, Lyme disease, Rocky Mountain Spotted Fever). The index therapy was defined as the first course of therapy with either spironolactone or an oral tetracycline-class antibiotic prescribed to patients who had not received any previous courses of oral antibiotics, spironolactone, or isotretinoin.The primary outcome was the frequency with which patients were switched to a course of therapy or to combination therapy with a different systemic agent used in the treatment of acne (ie, oral antibiotic, spironolactone, or isotretinoin) within the first year after starting treatment with either spironolactone or an oral tetracycline-class antibiotic.28,37 Since switching may reflect treatment failure, either due to lack of efficacy, side-effects, cost, or other reasons, this outcome was selected as a surrogate marker for clinical effectiveness. Given that patients with acne are often young and without significant comorbidities, and since systemic acne treatments typically have narrow indications, any observation of switching is likely to have clinical relevance. In addition, prior work to evaluate clinical effectiveness of antibiotics using large administrative datasets has taken a similar approach.28 Secondary outcomes included the timing of the change in therapy and the percentage of patients on each systemic therapy at time points 90, 180, 270, and 360 days after starting the index medication. In addition, subgroup analyses were conducted among adolescents, who were defined as being age 16 to 20 at the start of treatment, and adults, who were defined as being age 21 or older as the start of treatment.35 Subgroup analyses were also performed comparing the three most commonly prescribed oral tetracycline-class antibiotics to spironolactone.
Statistical Analysis
Descriptive statistics are presented using means, medians, and percentages as appropriate for the outcomes of interest. In addition, the odds ratio and absolute risk difference for switching to a different systemic agent within the first year of therapy were calculated using logistic regression. Since topical retinoids and oral contraceptive pills are effective agents in acne and are also often prescribed to women on spironolactone and oral antibiotics, we adjusted for whether the patient had received any prescriptions for topical retinoids or oral contraceptives prior to starting the index therapy or concurrently with the index therapy, in addition to adjusting for age at the start of treatment, using a multivariable logistic model.38–41 In any observational study, there is a concern that unmeasured confounding could lead to biased estimates. Since claims data do not contain information about acne severity, which could be a potential confounding variable, a formal sensitivity analysis was performed to assess the impact of differing acne severity between the spironolactone and oral tetracycline-class antibiotic groups on the primary outcome of interest.42 Statistical analyses were performed in Stata 14 (StataCorp, College Station, Texas) and R version 3.1 (The R Foundation for Statistical Computing, Vienna, Austria). The Institutional Review Board of the University of Pennsylvania has granted a blanket exemption for all research completed at the University of Pennsylvania using OptumInsight data.
RESULTS
Cohort
The characteristics of the study population are summarized in Table 1. There were 6,684 and 31,614 unique women who were started on spironolactone and oral tetracycline-class antibiotics as the index systemic therapy, respectively. The most frequent tetracycline-class antibiotics prescribed were
