Non-participation and adjustment for bias in case-control studies of periodontitis.

Periodontal researchers frequently use case-control studies, but information on participation rates and the reasons for participation are often missing in the publications, thus hindering the assessment of the validity of those studies. A nested case-control study based on a well-defined population was used to (i) describe the patterns of participation; (ii) show how some associations can be biased; and (iii) illustrate how inverse probability weights can be applied to reduce bias. Differential subject participation was quantified using the ratio between participation for each level and the overall participation. Possible biased associations were illustrated using the odds ratios found for eligible and participant subjects. Finally, we used the estimated probability that an individual participates in the case-control study conditional on that individual's covariate pattern, as observed in the screening study to attempt bias reduction. Considerable differential participation was observed for selected factors, including age, annual tuitions and fees, parental income, and parental education. The strategy used for adjustment of bias resulted in some degree of bias reduction. These findings challenge the inferential validity of many studies on periodontitis. The design and conduct of these studies should aim to improve subject participation and must consider and minimize this potential source of bias.