Premium
Collider bias in the association of periodontitis and carotid intima‐media thickness
Author(s) -
Leite Fábio R. M.,
Nascimento Gustavo G.,
Peres Karen G.,
Demarco Flávio F.,
Horta Bernardo L.,
Peres Marco A.
Publication year - 2020
Publication title -
community dentistry and oral epidemiology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.061
H-Index - 101
eISSN - 1600-0528
pISSN - 0301-5661
DOI - 10.1111/cdoe.12525
Subject(s) - medicine , confounding , periodontitis , collider , logistic regression , intima media thickness , carotid arteries , physics , nuclear physics
Objectives This cross‐sectional study tested the presence of collider bias in the relationship between periodontitis and the carotid intima‐media thickness (cIMT). Methods Data from 480 members of the 1982 Pelotas Birth Cohort, Brazil, were used. Periodontitis at the age of 24 years was determined as the main exposure. cIMT at the age of 30 years was set as the outcome. High‐sensitivity C‐reactive protein (hsCRP) was considered the mediator (collider). Confounding variables included sex, income, BMI and smoking. The association between cIMT and periodontitis was tested in conventional logistic regression stratified on hsCRP levels, marginal structural modelling and sensitivity analysis for collider stratification bias. Results Conventional adjusted logistic regression analysis showed a positive association between periodontitis and cIMT (OR 1.5; 95% CI 1.1; 2.3). Stratified analysis according to the hsCRP levels revealed that the magnitude of the association was even higher among participants with hsCRP ≥ 3 mg/L (OR 2.2, 95% CI 1.1; 4.2) with 36% collider bias probability. No association between periodontitis and cIMT was found among participants with hsCRP < 3 mg/L (OR 1.3; 95% CI 0.8; 2.1). The association was not detected using marginal structural modelling (OR 1.3, 95% CI 0.8; 2.0). Conclusions The association between periodontitis and surrogate markers of cardiovascular disease might be induced by collider bias stratification using conventional regression analysis.