Plausibility of periodontal disease estimates from NHANES III

OBJECTIVES: This study investigated possible reasons for observed discrepancies in prevalence estimates and measures of association for periodontal disease between phases (1988-91 and 1991-94) of the third National Health and Nutrition Examination Survey (NHANES III). METHODS: NHANES III data on CD-ROM were obtained from the National Center for Health Statistics. Accompanying documentation states that each phase and combined phases constitute national probability samples of the US population. Weighted estimates of prevalence (percent of persons affected) and extent (percent of sites affected) for previously reported thresholds of gingival bleeding (GB), attachment loss (AL), and probing pocket depth (PD) were generated using data from all 15,511 persons aged 13-90 years who received periodontal examinations. Odds ratios for associations between four selected risk indicators and both PD and AL were compared between phases. RESULTS: Phase 2 estimates of GB and PD were as much as 56 percent lower than phase 1 estimates and both were different from combined-phase estimates. However, AL prevalence was consistent between phases. Prevalence differences between phases could be explained in part by examiner variations. Odds ratios for PD differed between phases by as much as one-third, although the direction and precision of associations were not affected, and differences were reduced after controlling for examiner. CONCLUSIONS: Combined-phase estimates of GB and PD prevalence and extent differ from previously published estimates derived from Phase 1, apparently because estimates in at least one phase of the NHANES III study are biased. However, associations with selected risk indicators were fairly consistent between phases.