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

Competing event analysis with propensity score adjustment in registry data of patients with coronary heart disease

Meeting Abstract

  • Susanne Stampf - Institut für Medizinische Biometrie und Medizinische Informatik, Universitätsklinikum Freiburg, Freiburg
  • Nadine Grambauer - Institut für Medizinische Biometrie und Medizinische Informatik, Universitätsklinikum Freiburg, Freiburg
  • Miroslav Ferenc - Herz-Zentrum, Bad Krozingen
  • Franz-Josef Neumann - Herz-Zentrum, Bad Krozingen
  • Martin Schumacher - Institut für Medizinische Biometrie und Medizinische Informatik, Universitätsklinikum Freiburg, Freiburg

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. Doc11gmds082

doi: 10.3205/11gmds082, urn:nbn:de:0183-11gmds0822

Veröffentlicht: 20. September 2011

© 2011 Stampf et al.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen ( Er darf vervielfältigt, verbreitet und öffentlich zugänglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.



Background: The Bifurcation Bad Krozingen registry [1] contains information on 1,037 patients with bifurcation lesions in coronary vessels treated between 2002 and 2005 in “Herz-Zentrum” Bad Krozingen. The vessel lesion is provided with stents in main branch only (single stenting, n=662) or in main and side branches (double stenting, n=375). From those, 95 patients died, 29 suffered a myocardial infarction (MI) without later death and 189 needed a target lesion revascularization (TLR); after three years follow-up, 723 patients were event-free. This constitutes a competing events setting, with focus on investigating effects of treatment modalities on the time to death, MI and TLR, taking into account that comparison is based on an observational study.

Methods: For competing risks analysis, different approaches can be used [2]. Often, the standard Cox proportional hazards model for a composite endpoint is employed, where event types are simply combined. However, as there might be different effects on event types, event-specific hazards analyses offer more detailed insights, so separate Cox models for each endpoint should be preferred. If interest focusses on the probability to experience an event of interest over the course of time, cumulative incidence functions (CIFs) are considered. To quantify differences on this scale, a Cox model for the subdistribution hazard (Fine&Gray model) provides a summary analysis [3].

A further challenge is to cope with potential confounding. Adjustment by propensity score (PS) is done where PS is the probability to receive double stenting conditional on patients characteristics. Three different PS methods are applied to the competing event data: stratification, matching and direct adjustment by PS [4].

Results: Since the PS is estimated without using information on endpoints, it is identical for all competing event analyses. Compared to single stenting, double stenting has no effect on the risk of death. In contrast, the occurrence of MI without later death and time to TLR are markedly increased by double stenting. Adjusted effect estimates differ to unadjusted ones and underline the need to adjust for confounding which is visualized by contrasting plots of adjusted with unadjusted CIFs. Adjusted effects are compatible with effects observed in randomized trials that have recently been summarized in a meta-analysis [5].

Discussion: While the composite endpoint analysis already indicates an increased risk of double compared to single stenting, event-specific hazard analyses point to differential effects on endpoints. The subdistribution hazard analysis provides a summary analysis that also gives additionally insight in effects over time.


Ferenc M, Gick M, Kienzle RP, Bestehorn HP, Werner KD, Comberg T, Zhao M, Buettner HJ, Neumann FJ. Long-term outcome of percutaneous catheter intervention for de novo coronary bifurcation lesions with drug-eluting stents or bare-metal stents. Am Heart J. 2010;159(3):454-61.
Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data. Am J Epidemiol. 2009;170(2):244-56.
Grambauer N, Schumacher M, Beyersmann J. Proportional subdistribution hazards modeling offers a summary analysis, even if misspecified. Stat Med. 2010;29(7-8):875-84.
Schmoor C, Gall C, Stampf S, Graf E. Correction of confounding bias in non-randomized studies by appropriate weighting. Biom J. 2011;53(2):369-87.
Katritsis DG, Siontis GCM, Ioannidis JPA. Double versus single stenting for coronary bifurcation lesions: a meta-analysis. Circ Cardiovasc Interv. 2009;2:409-15.