gms | German Medical Science

MAINZ//2011: 56. GMDS-Jahrestagung und 6. DGEpi-Jahrestagung

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V.
Deutsche Gesellschaft für Epidemiologie e. V.

26. - 29.09.2011 in Mainz

High dimensional versus conventional propensity score matching investigating gastrointestinal complications in coxib users vs. nonselective NSAIDS users

Meeting Abstract

  • E. Garbe - Bremer Institut für Präventionsforschung und Sozialmedizin, Bremen
  • S. Kloss - Bremer Institut für Präventionsforschung und Sozialmedizin, Bremen
  • W. Schill - Bremer Institut für Präventionsforschung und Sozialmedizin, Bremen
  • M. Suling - Bremer Institut für Präventionsforschung und Sozialmedizin, Bremen
  • S. Schneeweiss - Brigham and Women's Hospital / Harvard Medical School, Boston

Mainz//2011. 56. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 6. Jahrestagung der Deutschen Gesellschaft für Epidemiologie (DGEpi). Mainz, 26.-29.09.2011. Düsseldorf: German Medical Science GMS Publishing House; 2011. Doc11gmds243

DOI: 10.3205/11gmds243, URN: urn:nbn:de:0183-11gmds2431

Published: September 20, 2011

© 2011 Garbe et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.en). You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.


Outline

Text

Background: High dimensional propensity score (hd-PS) adjustment has been proposed to improve control for confounding in pharmacoepidemiologic studies using longitudinal claims databases, as large numbers of empirically identified covariates may collectively serve as proxies for unobserved confounding.

Objectives: To compare results based on hd-PS matching vs. conventional PS matching in a cohort study of coxibs vs. nonselective (ns)NSAIDS and their association with gastrointestinal (GI) complications.

Methods: A cohort of new users of coxibs or nsNSAIDs was identified in the German Pharmacoepidemiological Research Database (GePaRD) and followed until discontinuation or switch of the initial NSAID, disenrollment, hospitalization for GI complication, death or end of study period whichever was earliest. Conventional PS based on 79 investigator-identified covariates and hd-PS based on 200, 500, 700 or 900 empirically identified covariates were estimated by logistic regression analysis. Initiators of coxibs were 1:1 matched to initiators of nsNSAIDs using nearest neighbour matching. Rate ratios (RR) of GI complications (haemorrhage, perforation or obstruction) associated with coxibs or nsNSAIDs were calculated with Poisson regression analysis in the full cohort and in each of the resulting matched cohorts.

Results: 37,281 coxib users and 2,104,614 nsNSAID users were identified. The crude RR of upper GI complications for coxib users vs. nsNSAID users in this cohort was 1.21 [95% CI=0.91-1.61]. Conventional and hd-PS matched 37,281 nsNSAID to the 37,281 coxib users. The conventional PS resulted in a RR of 0.84 [0.56-1.26]. Using the hd-PS algorithm to identify 900 empirical covariates and additionally including age, sex, calendar year and the number of distinct drugs during 180 days before cohort entry as a measure of health service intensity resulted in a RR of 0.62 [0.43-0.91].

Conclusions: hd-PS matching compared to conventional PS matching resulted in improved point estimates for treatment effects of COX-2 selective vs nsNSAIDS when benchmarked against results expected from randomized clinical trials.