Methodological considerations in the evaluation of periodontal disease and their impact on the exploration of its link with other diseases.

摘要

Inaccuracies in data collection may lead to misclassification of disease. In periodontal disease research, methodological issues may affect our ability to characterize individuals concerning their periodontal status. The goals of this study were to characterize measurement variability in periodontal probing within and between examiners, and to demonstrate the impact of missing teeth on periodontal probing depth (PPD), clinical attachment loss (CAL), and alveolar crestal height (ACH), recorded using four periodontal indices. Aim 1 examined the intra- and inter-examiner variability. Three examiners collected PPD data from 20 individuals from one quadrant, on three different occasions, twice by each examiner. Standard of error of measurement (s.e.m.) was calculated to determine intra- and inter-examiner variability at the tooth and quadrant levels. All examiners provided reproducible measures under 0.5mm, and reproducibility did not vary when using quadrant or tooth mean. Intra-examiner variability was 0.40mm ± 0.02 at the quadrant level, and 0.38mm ± 0.07 at the tooth level. Inter-examiner variability at the quadrant level was 0.16mm ± 0.02, and 0.24mm ± 0.05 at the tooth level. Percentage of error at the quadrant level ranged from 6.5% to 19.5%, and at the tooth, from 6.6% to 23.9%. The estimated ratios between intra- and inter-individual variance at the quadrant level, ranged from 0.505 to 2.02. Aims 2 and 3 involved analyses of 759 individuals to determine the impact of missing teeth on misclassification of periodontal disease, measured by PPD, CAL and ACH, assessed by whole-mouth and three partial-mouth methods. Arbitrary values of PPD, CAL, and ACH were assigned to missing teeth, assuming the teeth were lost to periodontal disease. For whole-mouth method, we observed that an increase in numbers of missing teeth and disease severity led to an increase in misclassification (0%–38.9%), though lower for ACH than for PPD and CAL. Our results indicate that partial-mouth methods were highly correlated to whole-mouth method. However, partial-mouth indices may underestimate disease prevalence, regardless of the periodontal measure. In conclusion, misclassification of individuals was higher for all three partial-mouth methods compared to the whole-mouth method.