Related Experiment Videos
An empirical investigation on matching in published case-control studies
O Gefeller1, A Pfahlberg, H Brenner
1Department of Medical Statistics, University of Göttingen, Germany. Olaf.Gefeller@Unix.AMS.Med.Uni-Goettingen.DE
European Journal of Epidemiology
|August 5, 1998
Summary
Case-control study methods have evolved, with a decrease in individual matching and an increase in frequency matching. Statistical analysis has shifted towards logistic regression, with less use of traditional methods.
Area of Science:
- Epidemiology
- Biostatistics
Background:
- Methodological debates surrounding control selection in case-control studies have persisted.
- The choice of matching strategies significantly impacts study validity and analysis.
Purpose of the Study:
- To assess the evolution of control selection and statistical analysis methods in case-control studies.
- To determine the practical impact of methodological discussions on study design and analysis.
Main Methods:
- A literature review of 266 case-control studies published between 1955 and 1994 in three major epidemiological journals.
- Analysis of trends in control recruitment (individual vs. frequency matching) and statistical analysis techniques (logistic regression, Mantel-Haenszel, conditional likelihood).
Main Results:
- Individual matching decreased from 71.7% to 46.4%, while frequency matching increased from 5.0% to 26.2%.
- Logistic regression use surged from 18.4% to 87.2%, with Mantel-Haenszel analysis becoming rare.
- Conditional likelihood methods for matched data in logistic regression remain underutilized.
Conclusions:
- Case-control study practices regarding control selection and statistical analysis have undergone significant changes.
- The adoption of advanced statistical techniques has not always been accompanied by appropriate application to matched data.