We used both pack-years and years of smoking to explore the relation between smoking history and tooth loss, and found a clear exposure-dependent association with tooth loss among middle-aged Finnish adults. Among those middle-aged Finnish adults with fairly good oral health, the risk for tooth loss increased significantly from 11 pack-years and with a history of 21 or more years of smoking, especially among males irrespective of their socio-economic background. Those who had stopped smoking or smoked only occasionally were at no higher risk for tooth loss than never smokers, thus substantiating the benefit of smoking cessation.
Comparison with other studies
Despite differences in studies of smoking and tooth loss in various populations, previous studies have pointed to an exposure-dependent relation between tobacco smoking history and tooth loss. Our findings are in line with those of previous research on smoking and tooth loss [4, 6, 9, 10]. Ojima et al.  found an exposure-related association between smoking and tooth loss among young Japanese adults. Previously, with the investigation of the association between smoking and tooth loss in the 31-year-old NFBC1966 cohort, smoking showed an exposure-dependent relation with tooth loss . However, that study involved no oral health examinations, so self-reported postal questionnaires provided data on the number of teeth.
Most previous studies on smoking and tooth loss have not assessed the intensity and duration of smoking in the same study and have used a binary variable rather than the original scale of the number of teeth for tooth loss as an outcome [9, 10]. Our study takes advantage of both of these measures of smoking, which strengthens our findings over those of previous studies. Moreover, we used a count outcome for tooth loss and relative risk (RR), which are considered more accurate measures than the more commonly used odds ratio (with binary outcome) to assess association in cross-sectional studies. Mai et al.  used multiple measures of smoking history to investigate its association with tooth loss, but the study was limited to postmenopausal women, and ‘any tooth loss’ served as a binary outcome. Although the study on middle-aged Danes shares similarities with our study, Morse et al.  did not use pack-years (or any other corresponding measure) to assess smoking intensity and they used only a binary outcome for tooth loss (6+ versus <6 teeth lost).
Education and tooth brushing frequency also associated significantly with tooth loss in our study. In particular, the impact of socio-economic status (SES), measured as education level, on tooth loss seemed closely resembled that of smoking. Previous studies have also revealed a strong association between low education and tooth loss [4, 8, 17]. In addition, a low level of oral self-care, commonly assessed as a tooth brushing frequency of once daily or less, has shown a stronger connection with a lower number of present teeth than has a high level of oral self-care [9, 18]. However, this connection has proved to be relatively weak, and some studies have found no significant association between tooth brushing frequency and tooth loss among either gender [10, 19].
Alcohol use is known to accompany smoking habit often and it may act as a confounder for the association between smoking and oral diseases . However, in our study alcohol use was only weakly associated with the outcome. Moreover, we performed interaction analyses with alcohol use and smoking (pack-years and current smoking habit) and did not observe statistically significant interaction. Previous studies have shown inconsistent results for how alcohol use associates with tooth loss [11, 21].
The rate of tooth loss in our study was similar among both men and women, a finding in line with those of previous studies [9–11, 22]. However, this association depends strongly on other factors, such as age and population, which have led to a wide range of observations across studies [8, 18, 23]. Since all the NFBC1966 members were the same age, we were unable to investigate the influence of age beyond reporting the prevalence of tooth loss at this age. Nevertheless, aging is known associate strongly with tooth loss.
In our study, the cohort members were among the first people to receive comprehensive Finnish public health care, including free oral care from childhood, as a result of the National Health Act of 1972 . Those born in the late 1960s were the first to benefit from this new, free-of-charge public oral health care from the beginning of school (i.e., the age when the first permanent teeth erupt) through adolescence. Since then, the cohort participants have enjoyed heavily subsidized dental care throughout their lifetime. Generally, the oral health of these 46-year-old adults was better than that of the roughly same-age participants from the previous Health 2000 Survey in Finland . In our study, 53 % of the participants had experienced tooth loss (not counting third molars).
Strengths and limitations
The comprehensive and representative data on the 46-year-old Finnish adult population is one strength of this study. The NFBC1966 cohort study has been a unique research project, collecting detailed information on cohort members across several life stages. We had information (about tooth loss) from 1946 participants, with fairly even representation of men (47 %) and women (53 %). Smoking variables were sufficiently thorough to assess the participants’ smoking history in detail (smoking duration, amount, and intensity were calculable). Moreover, for the first time, the follow-up of 46-year-olds included complete clinical oral examinations, which enabled the clinical measurement of the number of missing teeth rather than self-reported measurements. Self-reported tooth loss has been considered an acceptable substitute for clinically measured tooth loss, but some validity issues remain for participants with certain characteristics .
Due to the cohort study design, with its several follow-ups and health monitoring, the prevalence of many oral health-related diseases and symptoms, as well as the presence of associated lifestyle-related risk factors, may be lower among cohort members than in the general Finnish population . Another weakness relates to the follow-up data on the 46-year-olds: we were able to examine only the association between smoking and tooth loss at this age and thus cannot predict how strong the association might have been in a younger or older study population.
In this study, the cause of tooth loss was unavailable. In addition, we do not know the exact moment of tooth loss, which raises the possibility that some of the individuals’ outcomes may have preceded their exposure. However, this scenario among these middle-aged adults with fairly good oral health is unlikely. Although we adjusted for common confounding factors, some problems related to the accuracy of the results may persist, possibly due to unknown risk factors or errors in the self-reported or other measurements. For instance, because we used only education as a measure for socio-economic status and tooth brushing for oral health behavior [9, 18], one should exercise caution when interpreting the findings.