Multivariate multilevel models for repeated measures in the study of smoking effects on the association between plaque and gingival bleeding

Müller HP, Stadermann S
Clinical Oral Investigations, 10, 311-316

Multivariate multilevel modeling was applied to analyze repeated measures data on the influence of heavy smoking on the association between the amount of supragingival plaque and gingival bleeding on probing (BOP) in a steady-state plaque environment. Data acquired in 65 systemically healthy young adults with mild plaque-induced gingivitis were analyzed. 33 heavy smokers consumed at least 20 cigarettes per day while 32 were non-smokers. Periodontal examinations at the outset consisted of periodontal probing depth, clinical attachment level, BOP, plaque index, and presence of calculus at 6 sites of every tooth present. They were repeated 3 times every 8 weeks. A multivariate 4-level variance component model revealed that the odds of BOP was twice as high in smokers. In addition, females had a lower likelihood for BOP but, with increasing bleeding scores during the course of the study, this effect attenuated. Low biserial correlations for BOP at the site level of between 0.11 and 0.2 were found. At the tooth level, correlations were moderate (0.2-0.5), and highest at the subject level (0.8-0.9). Variations at subject and tooth levels were very large at the outset but notably attenuated in the course of the study. Plaque consistently influenced the tendency for BOP with an odds ratio of about 1.7-1.8 for each increase in score in both smokers and non-smokers. The present study did not reveal evidence for attenuation of the plaque/gingival bleeding relationship in heavy smokers.

Number of levels
Model data structure
Response types
Multivariate response model?
Longitudinal data?
Substantive discipline
Paper submitted by
Hans-Peter Müller, Institute of Clinical Dentistry, Tromsø University,
Edit this page