Effect of smoking cessation on tooth loss: a systematic review with meta-analysis

Background Smoking is a major risk factor for periodontitis and tooth loss. Smoking cessation has a positive impact in periodontal treatment. However, so far, no systematic review has evaluated the effect of smoking cessation on tooth loss. Therefore, this review aimed to evaluate if smoking cessation reduces the risk of tooth loss. Methods Observational (cross-sectional and longitudinal) studies that investigated the association between smoking cessation and tooth loss were included. MEDLINE, EMBASE and LILACS databases were searched for articles published up to November 2018. Pooled results for subgroups of current and former smokers were compared in meta-analysis. Meta-regression was used to test the influence of smoking status on estimates and explore the heterogeneity. Results Of 230 potentially relevant publications, 21 studies were included in the qualitative review and 12 in the quantitative analysis. Meta-analysis of cross-sectional studies did not show any differences between former and current smokers in the chance of losing 1 or more teeth (OR = 1.00; 95% CI = 0.80 to 1.24, I2 = 80%), losing more than 8 teeth (OR = 1.02; 95% CI = 0.78 to 1.32, I2 = 0%) or being edentulous (OR = 1.37; 95% CI = 0.94 to 1.99, I2 = 98%). Meta-analysis from longitudinal studies showed that, when compared to never smokers, former smokers presented no increased risk of tooth loss (RR = 1.15; 95% CI = 0.98 to 1.35, I2 = 76%), while current smokers presented an increased risk of tooth loss (RR = 2.60; 95% CI = 2.29 to 2.96, I2 = 61%). Meta-regression showed that, among former smokers, the time of cessation was the variable that better explained heterogeneity (approximately 60%). Conclusions Risk for tooth loss in former smokers is comparable to that of never smokers. Moreover, former smokers have a reduced risk of tooth loss, when compared to current smokers.

Tooth loss is the final outcome of periodontal disease. It is associated with loss of masticatory function [13], lack of self-esteem and impaired social interactions due to limited aesthetic appearance [14,15]. Consequently, tooth loss has a negative impact on oral health-related quality of life [16]. Smokers are more likely to lose their teeth than non-smokers [17,18], as a result of their increased severity of periodontitis. Cross-sectional [19,20] and prospective studies [21,22] have also suggested that former smokers have a significantly lower risk of tooth loss than current smokers.
Previous reviews have addressed the association between smoking and tooth loss [17,18]. However, so far, no review has focused on the effect of smoking cessation on tooth loss. Further, there is no meta-estimate to quantify the impact of smoking cessation on tooth loss. Therefore, the aim of this systematic review was to answer the following focused question: "Does smoking cessation reduce the risk of tooth loss in former smokers, when compared to current smokers?"

Methods
This review has been prepared according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [23] and Metaanalysis of Observational Studies in Epidemiology (MOOSE) guidelines [24]. The protocol was registered in the International Prospective Register of Systematic Reviews -PROSPERO (CRD42018085095).

Eligibility criteria
Only observational studies (cross-sectional and longitudinal studies) were included in this systematic review because, for ethical reasons, there are no randomized clinical trials with a control group that did not receive smoking cessation therapy. The inclusion criteria were as follows: a) original studies published in English; b) data comparing former smokers with current smokers and never smokers; c) studies that had tooth loss as an outcome.
We excluded narrative reviews, case series, case reports, in vitro and animal studies. Further, we excluded studies that did not include former smokers in the analysis, or that combined former smokers with never smokers or current smokers, or that did not associate smoking with tooth loss (e.g., smoking was used only for adjustment). . We also conducted a hand search of references lists from included publications.
In the first phase, two reviewers (MLSS and ESR) screened independently titles and abstracts identified by the search strategy. Disagreements were resolved by discussion or, if necessary, by the decision of a third reviewer (CMP). In the second phase, the same reviewers screened full texts of the studies that met inclusion criteria, or those with unclear information in the title and abstract. Reasons for rejection of studies were recorded for each report.

Data extraction
The following items were extracted from the publications that met inclusion criteria: author, year, country, study design, sample size, measures of exposure (smoking status), measures of outcome (tooth loss), results, conclusions, conflict of interest and source of funding. Authors of the included studies were contacted for missing, relevant data.

Risk of bias
Risk of bias of cohort studies was assessed using a modified version of the Newcastle-Ottawa scale (NOS) [25]. For cross-sectional studies, we adapted the Modesti et al. (2016) version of the NOS scale [26].
The NOS for cohort studies comprised 10 questions about selection of the study groups (i.e. representativeness of current and former smokers), comparability of the groups, outcome (criteria used to assess tooth loss and adequacy of follow-up) and statistical analysis. The scores ranged from 0 to 11. Studies with 9-11 stars were arbitrarily rated as low risk of bias, 6-8 stars moderate risk of bias and < 6 high risk of bias.
The NOS for cross-sectional studies comprised 07 questions about selection of the study groups (i.e. representativeness of the sample), comparability of the groups, outcome (criteria used to assess tooth loss) and statistical analysis. The scores ranged from 0 to 10. Studies with 7-10 stars were arbitrarily rated as low risk of bias, 5-6 stars moderate risk of bias and < 5 high risk of bias.

Summary measures and synthesis of results
Analyses of data extracted from cross-sectional studies were carried out using software Review Manager (Rev-Man) (Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Randomeffects meta-analyses were conducted for the following outcomes: loss of one or more teeth, loss of more than eight teeth and being edentulous. The estimates were presented as pooled odds radios (ORs) and their respective 95% confidence intervals (CIs). Heterogeneity was tested using the Cochran's Q test and quantified using the I-square test (level of inconsistency) and Tau 2 (estimate of between-study variance).
Meta-analyses were performed to assess the risk of tooth loss among subgroups of former and current smokers, compared to the control group (never smokers). OR values for studied groups (former and current smokers) were converted into LogOR and results from individual studies were pooled using a randomeffects model. Former and current smokers were considered as different subgroups and contrasted with never smokers. Differences between subgroups (subsets) were also tested based on random-effects models. Metaanalysis used the inverse variance method and the DerSimonian-Laird estimator for Tau 2 . The pooled results were estimated using the Risk Ratio (RR), Relative Risk and 95% CIs. Heterogeneity was tested similarly to the cross-sectional studies.
Meta-regressions were used to test the influence of different moderators (age, time of cessation and dropout rates) on pooled estimates. After testing each variable in the model, residual heterogeneity (I 2 ) and amount of heterogeneity accounted for each variable (R 2 ) could be calculated. Funnel plot visual analysis and linear regression test of funnel plot asymmetry were used to assess publication bias of the longitudinal studies. Both metaregressions and publication bias investigation were performed considering subgroups separately.
Data analyses of longitudinal studies were performed using the "meta" and "metafor" packages, R software (R Studio, Version 1.0.143).

Longitudinal studies
The risk of tooth loss among those who quit smoking was not significantly different from never smokers. Contrarily, current smokers presented a risk of tooth loss twice higher than never smokers. Moderate to high level of heterogeneity was found even within the subgroups (Fig. 5).
Among former smokers, the time of cessation was the variable that explained a considerable part of the heterogeneity (around 60%), resulting in a low residual heterogeneity (around 20%) when included in the regression model (Table 5). However, the level of significance was not inferior to 5%. On the other hand, for current smokers, no moderator was significantly associated with the estimates ( Table 5). The funnel plots showed no visual or statistically tested asymmetries (Additional file 1 a and b).

Discussion
The aim of this systematic review was to assess if smoking cessation reduces the risk of tooth loss in former smokers, when compared to current smokers. Our results showed that smoking cessation may reduce the risk of tooth loss. Meta-analysis of data from longitudinal studies showed that the rate of tooth loss in former      smokers is similar to that of never smokers. Moreover, current smokers had a risk of tooth loss twice higher than never smokers. These results are consistent with a previous systematic review that found a causal relationship between smoking and tooth loss and a decreased risk of tooth loss in former smokers [17]. However, the effect of smoking cessation on tooth loss had not been explored in this previous review. To the best of the authors' knowledge, this is the first systematic review with meta-analysis that included never, former and current smokers, as different levels of exposures to smoking and investigated their association to tooth loss. The most plausible biological explanation for the increased risk of tooth loss in smokers is the destruction of the periodontal supporting tissues [17]. A recent systematic review showed that the risk for periodontitis incidence and progression could be reversed after smoking cessation to the same level as that of never smokers [12]. These results are in agreement with our findings that the risk of tooth loss between former smokers and never smokers were not significantly different.
In contrast to the results from longitudinal studies, the meta-analysis of data from cross-sectional studies did not show significant differences between former smokers and current smokers in relation to the risk of being edentulous, losing one or more teeth and losing eight or more teeth. The possible reason for this lack of effect is the inherent limitations of cross-sectional studies, especially the absence of information about the temporal relationship between cause (smoking cessation) and effect (tooth loss). For example, none of the included crosssectional studies assessed the time of tooth loss. Thus, it is possible that former smokers lost their teeth before stopped smoking. Moreover, most of these studies failed to report the non-smoking duration for subjects that ceased the habit. It is possible that many quitters in the included studies have stopped smoking for less than 5 years. Considering that it may take at least 10 to 20 years of abstinence for the risk to return to the level of never smokers [21,22,40], the inclusion of recent quitters in the analysis could have reduced the effect size of smoking cessation on tooth loss in the cross-sectional studies.
Even though our meta-estimates were derived from observational studies, which usually present high heterogeneity, the results provided from the analysis of this kind of studies can be considered similar to those of randomized trials [50]. Besides, heterogeneity was considered in our meta-analyses (random-effects models) and explored (subgroup and meta-regression analyses), contributing to an appropriate judgment about the findings and helping in identifying potential sources of heterogeneity. In addition, another point that should be emphasized is that the sample size in observational studies is frequently larger than that of clinical trials. Altogether, the studies included in this review enrolled 638,389 individuals (567,491 subjects from cross-sectional studies and 70,898 participants from longitudinal studies). Moreover, there are methodological difficulties in using tooth loss as an outcome in interventional studies. It would be necessary a very long follow-up time and a very large sample size to analyze this outcome. Some methodological differences between the studies should be pointed out. For example, while some studies [22,40] asked the time since smoking cessation at the baseline questionnaire, others [3,21,[41][42][43] considered as former smokers the subjects that stopped smoking during the follow-up time. These methodological differences could have influenced the results since in the first case [22,40] participants could have stopped smoking many years before tooth loss. Time since cessation was the variable that better explained the heterogeneity in meta-regression. However, the effect was not significant, which could be related to the small number of studies included in this analysis. Another important consideration that has to be pointed out is that all included studies were carried out in high-income countries. It is necessary to be carefully to extrapolate our results to low-middle-income countries because it is known that socio-economic differences have an important role in oral health status, tooth loss and smoking status. Wellconducted studies with these populations are necessary.
None of the studies have used an objective measure of smoking status (e.g. salivary levels of cotinine or levels of carbon monoxide exhaled). Self-reported smoking status has been associated with underestimated smoking prevalence [51]. Along with the same lines, self reported tooth loss may not be accurate. Although clinical examination is the best method to determinate tooth loss, some studies [27,28,53,36,22,[40][41][42] used self-report to determinate this outcome. This method could have been chosen because of the high number of participants or the long follow-up time (longitudinal studies). Another shortcoming in the included studies was that the reason of tooth loss was not considered. This information could help to better explain the relation between smoking and tooth loss.
Despite the methodological limitations of the included studies, the findings of this systematic review support a beneficial effect of smoking cessation on the risk of tooth loss. Considering the benefits of quitting tobacco for the general health, and that smoking cessation Fig. 2 Forest plot for meta-analysis of being edentulous in former smokers compared to never-smokers in cross-sectional studies (n = 04 studies, association measure: Odds ratio) Fig. 3 Forest plot for meta-analysis of losing 1 or more teeth in former smokers compared to never-smokers in cross-sectional studies (n = 03 studies, association measure: Odds ratio) Fig. 4 Forest plot for meta-analysis of losing more than 8 teeth in former smokers compared to never-smokers in cross-sectional studies (n = 02 studies, association measure: Odds ratio) interventions conducted by oral health professionals are effective [52], the dental setting seems to be appropriate to implement smoking cessation therapy.

Conclusions
This systematic review indicates that risk for tooth loss in former smokers is comparable to that of never smokers. Moreover, current smokers present a higher risk of tooth loss than former smokers.
Additional file 1. Funnel plots of longitudinal studies investigating the effect of smoking and smoking cessation on tooth loss. (a) former smokers vs. never smokers (Egger's test for asymmetry, p = 0.06), (b) current smokers vs. never smokers (Egger's test for asymmetry, p = 0.79).
Abbreviations CI: confidence interval; MOOSE: Meta-analysis of Observational Studies in Epidemiology; NOS: Newcastle-Ottawa scale; OR: odds ratio; PRISMA: Preferred reporting items for systematic reviews and meta-analyses; PROSPERO: International Prospective Register of Systematic Reviews; RR: risk ratio