Table 5.

Summary of results to evaluate the causal association between smoking and tooth loss

Element	Description for consistency and study type	Evaluation of consistency	Evaluation of study type	Perceived shortcoming
Strength of association	All studies reported significant associations based on effect size: moderate association in 5 CSSs, 1 PCS and 1 CSS for males; weak association in 1 PCS and 1 CSS for females.	Evidence for weak or moderate association based on effect size is strong.	Evidence for strength of association is convincing.	Not applicable
Natural experiment	All studies reported smaller effect size in former smokers than in current smokers. The association between former smoking and tooth loss relative to non-smokers was not significant in 3 CSSs and 1 PCS, and was significant in 3 CSSs and 1 PCS. The description of association in former vs. current smokers decreased in 4 CSSs, 2 PCSs and 1 CSS for males, and remained at the same level in 1 CSS and 1 CSS for females. The hazard ratio decreased based on years of abstinence in 2 PCSs.	Evidence for natural experiment is strong. However, this interpretation does not mean that the risk in former smokers is lesser than that in current smokers.	Evidence for natural experiment is probable.	Control group did not comprise current smokers, and only a relative relationship was evaluated.
Dose-response relationship	Trend of the relationship between level of exposure and effect size, i.e. odds ratio or hazard ratio, was obvious in 3 CSSs and 1 PCS. This trend was highly significant in 2 CSSs.	Evidence for the dose-response relationship is strong. This interpretation is limited to populations assessed in 3 countries.	Evidence for dose-response relationship is probable.	Findings pertain to limited populations and 1 PCS.

Evaluation in each element was based on the consistency of findings and study types.

CSS: cross-sectional study, PCS: prospective cohort study