Comparison of the missing-indicator method and conditional logistic regression in 1:m matched case-control studies...

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Matched case-control studies are crucial for epidemiological research.
• Missing exposure values pose analytical challenges in these studies.
• The missing-indicator method and conditional logistic regression are proposed solutions.

Purpose of the Study:

• To evaluate the performance of the missing-indicator method and conditional logistic regression.
• To compare these methods under various missing data scenarios in matched case-control studies.
• To identify optimal analytical strategies for handling missing exposure data.

Main Methods:

• Monte Carlo simulation was employed to generate data.
• A 1:m matched design based on McNemar's 2x2 tables was used.
• Four missing value scenarios were simulated: completely-at-random, case-dependent, exposure-dependent, and case/exposure-dependent.

Main Results:

• Conditional logistic regression provided less bias and better confidence interval coverage than the missing-indicator method under the first three missingness scenarios.
• Increased matched controls slightly increased bias and reduced coverage.
• Neither method performed well under the case/exposure-dependent missing data scenario.

Conclusions:

• Conditional logistic regression offers a slight advantage over the missing-indicator method for matched case-control studies with missing exposure data.
• More advanced statistical methods are needed for complex missing data patterns (case/exposure-dependent).
• The choice of method impacts bias, coverage, power, and efficiency.