In defense of matching

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Epidemiologic studies often involve dichotomous disease, exposure, and confounding factors.
• Selecting an appropriate comparison group is crucial for valid study results.

Purpose of the Study:

• To compare random sampling versus matching on an extraneous factor for selecting comparison groups in epidemiologic studies.
• To evaluate the impact of matching on the precision of effect measure estimation in follow-up and case-control studies.

Main Methods:

• Analysis of a simplified epidemiologic study scenario with dichotomous variables.
• Application of a probability model to assess precision of effect measures under random sampling and matching.
• Illustrative example demonstrating differences in confounding control between study designs.

Main Results:

• Matching effectively controls confounding for risk ratio estimation in follow-up studies but not for odds ratio in case-control studies.
• Matching generally leads to more precise effect measure estimates than random sampling, despite potential sample size reduction.
• Loss of precision due to matching is unlikely to be significant in practically important situations.

Conclusions:

• Matching is a valuable strategy for selecting referent groups in both follow-up and case-control studies.
• Serious consideration should be given to matching to enhance the precision of epidemiologic effect estimates.
• Matching offers advantages in precision that often outweigh potential sample size considerations.