Odds ratio (OR) or risk ratio (RR): Correct interpretation is key for clinical translation of treatment effectiveness, an example in psoriasis

Summary: Abstract Body: The OR is often the default summary measure of choice for dichotomous therapy response outcomes. However, the RR may be more appropriate in cohort studies of real-world effectiveness where risk has a more intuitive interpretation than odds, but requires careful consideration and accurate interpretation. To demonstrate how the OR may not reasonably approximate the RR in cohort studies of psoriasis patients, this study estimated the association between baseline Psoriasis Areas Severity Index (PASI) and PASI75 at 6 months among patients initiating biologic therapy. Adults who initiated a biologic and had a follow-up visit were selected from the CorEvitas Psoriasis Registry (N=5851). Modified-Poisson and logistic regression were used to estimate the RR and OR, respectively, for achieving PASI75 among baseline PASI groups: mild (ref; ≤5, n=2404), moderate (≤12, n=2210), severe (≤20, n=806), and very severe (≤72, n=431). Models were adjusted for age, race, sex, disease duration, psoriatic arthritis, and prior biologics. The percent of patients who achieved PASI75 increased with higher baseline PASI: 43.8 of mild, 59.8 of moderate, 71.0 of severe, and 74.0 of very severe. Adjusted RRs (95% CI) for achieving PASI75 for moderate, severe, and very severe baseline PASI were 1.30 (1.23-1.37), 1.50 (1.39-1.62), and 1.58 (1.48-1.70), respectively, while the corresponding ORs were of higher magnitude: 1.91 (1.63-2.23), 3.23 (2.55-4.10), and 4.16 (3.05-5.66), respectively. Thus, compared to patients with mild baseline PASI, those with very severe baseline PASI had 1.58 times higher risk and 4.16 higher odds of achieving PASI75. In real-world cohort studies evaluating response to therapy, the interpretation of OR estimates as RR is not recommended, as they produce different numerical estimates of the underlying associations. When selecting a measure of association, careful consideration of outcomes and accurate interpretation are necessary for appropriate inference to patient care. Alvin Li¹, Amy Scrader¹, Alicia Beeghly¹, Sarah Crisci¹, Oksana Pugach¹, Robert McLean¹ 1. CorEvitas LLC, Waltham, MA, United States. Clinical Research: Epidemiology and Observational Research