Sat0588 Comparative Effectiveness Research in Observational Settings: Evaluating Two New Methods to Analyse Response Rates


Background: Researchers typically report the proportion of patients reaching a defined clinical threshold (e.g. EULAR response, low disease activity (LDA) rates) after a set time. Comparing response rates (%rr) in observational settings is hampered by two major threats. 1. Confounding: Patient, disease, and treatment characteristics often differ for each drug. 2. Attrition bias: Assessing %rr after a set time excludes patients who discontinued their treatment, which may overestimate drug effectiveness. Currently, no proposed method accounts for both confounding and attrition.Objectives pTo propose two new methods (3 and 4) to adequately compare %rr in patients with different baseline characteristics, while accounting for attrition and compare them to established methods (1 and 2). Methods Complete case (CC) %rr is computed as the percentage of responders on total patients still on treatment at the given time point LUNDEX %rr is corrected for attrition by multiplying it by the Kaplan Meier estimates of the survival (SKM), thus considering all patients not on treatment as non-responders, irrespective of the reason for drug discontinuation. PSM LUNDEX: Propensity Score (PS) Matching LUNDEXpStep A: Select patients in both exposure groups using propensity score matchingpStep B: Use the LUNDEX to compute the %rr. CARRAC: Confounder-Adjusted Response Rate with Attrition Correction by reason for drug discontinuationp Step A: Compute estimates of drug survival for the main reasons of drug discontinuation (e.g., ineffectiveness, adverse events, remission, other reasons). Step B: Estimate %rr using random effect IPD meta-analysis with estimates for each reason of drug discontinuation. Step C: Combine %rr estimates using weights of step A. The methods will be illustrated using CDAI LDA rates in data from a collaboration of registries. Results We used 3448 patients treated by a biologic, either intravenously (IV: n=2414) or subcutaneously (SC: n=1034). Before matching, the population differed in terms of body mass index, function (Health Assessment Questionnaire, HAQ), concomitant treatments and erythrocyte sedimentation rate (ESR). Adjustment was done using these potential confounders. For PSM LUNDEX, PS matching was done on a 1:1 basis, yielding 561 patients in each group with similar characteristics. Estimated %rr differed by more than 20%, depending on the method used (Figure). Compared to CC %rr, both PSM LUNDEX and LUNDEX methods yielded much lower %rr, while the CARRAC method estimated %rr was in between these estimates.pCompared to CC analysis, differences in %rr between the SC and IV groups were smaller for the LUNDEX methods, larger for PSM LUNDEX, and close to CC for the CARRAC method (Figure). Conclusion Both LUNDEX methods certainly underestimates the true response rates by considering all patients stopping treatment as non-responders, while complete case method certainly overestimates the %rr by considering patients stopping as having a similar %rr to patients continuing treatment. As expected, the CARRAC method, which accounts for attrition by reason for drug discontinuation, obtained %rr estimates in between complete-case and LUNDEX corrections. Simulation studies are needed to assess the most accurate estimation method.

Annals of the Rheumatic Diseases, 78 page 1386–1388