Online Program

Sample Size Estimation for Multiply Matched, Noninferiority Case-Control Studies with Binary Exposures (303003)

Keywords: Case-control, non-inferiority, matched, sample size, power

Case-control studies often are used to reduce costs when studying the odds of being a case between two exposure groups. Matching can help control known sources of variation and reduce total sample size. Multiple controls per case also can help achieve a specified power level when cases are limiting. Noninferiority studies, where Ha is that the odds of exposure in cases is not significantly greater than in controls, are becoming increasingly popular. To our knowledge, no sample size estimates are available in the case of multiply matched case-control studies of noninferiority for binary exposures. We studied empirical power under a range of sample-size scenarios for such studies using Monte Carlo simulation and conditional logistic regression to estimate one-sided confidence intervals for odds ratios. Other input parameters included average risk (N increased with very high or very low risk), multiplicity of matching (M>5 not useful), variability of risk over matched sets (modest effect), and acceptable margin of noninferiority (N decreased with bigger MONI). Sample size also will be compared to unmatched designs under the same scenarios.