Dear all, I would like to draw your attention to two statistical seminars which will take place next week in the Salle Delachaux (Biopôle 2, first floor, Route de la Corniche 10, 1010 Lausanne, M2: Vennes). ------------------------------------------------------------------------------------------------------------------------------------ Monday 27 June 11:00: The "Missing Cause" Approach to detect and reduce bias due to unmeasured confounding in pharmacoepidemiology. By Prof. Michal Abrahamowicz, Department of Epidemiology and Biostatistics, McGill University, Montréal, Canada (organized by Arnaud Chiolero, IUMSP). Objectives: To propose a new method for reducing the impact of unobserved confounding in large observational pharmacoepidemiological studies [1], validate it in simulations and apply it in a real-life drug safety study. Methods: Under assumptions similar to the prescribing preferences-based instrumental variable (IV) approach, our new "missing cause" method relies on discrepancies between a) treatment actually received by individual patients and b) treatment they would be expected to receive based on their observed characteristics and their physicians' prescribing preferences. Specifically, we use the treatment-by-discrepancy interaction to test for presence of unmeasured confounding and correct the treatment effect estimate for the resulting bias. We implement the method within risk difference (RD) multivariable linear regression modeling of the effect of a binary exposure on a binary outcome. We use simulations to compare our missing-cause estimates with (i) conventional estimates, adjusted only for measured covariates, and (ii) IV estimates. We apply the method to compare gastrointestinal safety of COX-2 inhibitors vs. traditional NSAIDs in a large cohort of elderly NSAID users in Quebec, Canada, and to assess the potential benefits of influenza vaccine in preventing influenza. Results: In simulations, our estimates had four times smaller bias than conventional estimates, much smaller variance than unbiased but unstable IV estimates, and uniformly best overall accuracy (lowest mean square error). In the first application, our method suggested a slight reduction of gastrointestinal risks for COX-2 inhibitors (RD=-2.6% (-8.4, 3.2%)), in contrast to risk increase suggested by IV estimates, with much wider CIs. In the second application, the method detected very serious bias, due to unmeasured confounding by indication, which resulted in the conventional estimates, adjusted only for the confounders measured in the study database, suggesting that influenza vaccine is associated with an increase risk of influenza. In contrast, once we used the missing cause method to correct for this bias, the corrected estimate indicated a significant risk reduction. Conclusion: Our "missing cause" method provides a potentially useful alternative to IV approaches for reducing unmeasured confounding bias in pharmacoepidemiology. ------------------------------------------------------------------------------------------------------------------------------------ Thursday 30 June, 11:00: Appropriate methods for rare outcomes and/or rare exposures. By Prof. Anirban Basu, Departments of Pharmacy, Health Services and Economics, University of Washington, Seattle (organized by Prof. Mark Dusheiko, IUMSP, in collaboration with the platform "Economie de la Santé" and the seminar « Health and Labour Economics » from the Faculty of Business and Economics, University of Lausanne). Abstract: Using Monte-Carlo simulations, we compare the two-stage least-squares (2SLS) estimator with two-stage residual inclusion (2SRI) estimators, with varying forms of residuals, to estimate the local average treatment effect parameter for a binary outcome and endogenous binary treatment model in the presence of binary covariates and a binary instrumental variable. We vary the rarity of both the outcome and the treatment and find different estimators to produce the least bias in different settings. We develop guidance for applied researchers and illustrate the utility of this guidance with estimating the effects of long-term care insurance on a variety of binary health care use outcomes among the near-elderly using the Health and Retirement Study. We would be pleased to see you there! Valentin CHUV centre hospitalier universitaire vaudois Valentin ROUSSON - Professeur mathematicien/statisticien Département universitaire de medecine et santé communautaires (DUMSC) Médecine sociale et préventive (IUMSP) Biostatistique et méthodes quantitatives +41 (0)21 314 73 28 TEL Valentin.Rousson@chuv.chmailto:Valentin.Rousson@chuv.ch www.chuv.chhttp://www.chuv.ch/