An evaluation of potential sources of systematic error in the estimation of periodontal disease associations

Abstract

AIM: To investigate the effect of different definitions of periodontitis, using partial-mouth measurements instead of full-mouth ones, and unmeasured confounders on periodontitis’ associations with diabetes and cardiovascular disease (CVD). METHODS: Adults 30-79 years with ≥1 or ≥2 teeth in the 2009-2014 National Health and Nutrition Examination Survey for study 1 or 2, respectively. For study 1: periodontitis was defined using different thresholds while for study 2: it was defined using the CDC/AAP 2007 categorical definition and continuous measures were estimated using mean clinical attachment loss (CAL), Estimates of periodontitis were derived based on the full-mouth protocol and three partial-mouth protocols (PMPs), including the Ramfjörd teeth, the Community Periodontal Index for Treatment Needs teeth, and the random half-mouth. Effects of exposure and outcome misclassification of periodontitis were evaluated in relation to diabetes. Diabetes and CVD were ascertained using self-report. Percent relative bias (%RB) was calculated by comparing the odds ratios/beta estimates obtained from the full-mouth and PMPs. Study 3 used the dental longitudinal study to look at the effects of simulated unmeasured confounders on survival analysis in the periodontitis-diabetes/CVD and diabetes-periodontitis associations. RESULTS: For study 1: the effects of clinical severity on the odds ratios were association dependent. Clinical measures and extent did not depend on the association. For study 2: percent relative bias was generally less than 10% for the severe categories while it tended to exceed 10% for moderate categories. Mean clinical attachment loss resulted in minimal bias. For study 3: presence of one source of unmeasured confounding (one confounder) showed that the diabetes-periodontitis association was robust to it unlike the periodontitis-diabetes/CVD associations. CONCLUSION: These sources played a role in the periodontitis-diabetes, periodontitis-CVD, and diabetes-periodontitis associations. The associations were affected differently by each source. Some of these sources of systematic errors may change the conclusions of the associations.