Association Between Smoking Intensity and Duration and Tooth Loss Among Finnish Middle-aged Adults

The Northern Finland Birth Cohort 1966 Project

Toni Similä; Jorma I. Virtanen


BMC Public Health. 2015;15(1141) 

In This Article


Study Design

This cross-sectional study uses data from the longitudinal Northern Finland Birth Cohort 1966 Project (NFBC1966), which comprises a comprehensive sample of babies from the provinces of Lapland and Oulu whose expected birth year was 1966 (12 068 mothers, 12 231 children, 96.3 % of all births in this region).[12] The Ministry of Social and Health Affairs in Finland approved the data collection, and the Ethics Committee of the Northern Ostrobothnia Hospital District in Oulu, Finland approved the study protocol. We used information from the follow-up study of 46-year-olds (carried out in 2012–2014), which included a postal survey and a comprehensive clinical health examination. In addition, in the Oulu region, the examination also included the full inspection of the mouth and teeth, and during the examination day, participants completed two additional questionnaires. Participation in the follow-up study of 46-year-olds was voluntary, and the participants provided their informed written consent. Altogether 1946 participants (participation rate 62 %) provided information on the number of missing teeth.

Smoking Variables

In 2012, participants received postal questionnaires, which included several questions on previous and current smoking habits, to be returned prior to the clinical health examinations. We calculated pack-years (based on 20 cigarettes per pack) to measure smoking history among those who reported having smoked at least five days a week. In addition, years of smoking served as an alternative measure of smoking history. Both measures of smoking history use equal numbers of participants with available information. We weighted the use of any tobacco product (filtered cigarettes, n = 240; other cigarettes, n = 22; cigars, n = 10; pipe smoking, n = 1) equally when calculating pack-years.

Apart from daily smokers, we used separate categories for occasional, former and never smokers. Here, 'never smokers' includes all participants who had smoked daily for less than one year in their lifetime and did not smoke at the time of the follow-up. 'Former smokers' includes those who had smoked daily for at least one year, but had quit smoking and did not smoke at the time of the study. Those who smoked, but no more than four days a week at the time of the study were considered as 'occasional smokers'.


Seven dentists who underwent specific training and calibration for this purpose performed the clinical oral examinations. For each participant, the dentists recorded all missing teeth. Excluding third molars, we formed the number of missing teeth variable to measure tooth loss in this study. For the assessment of missing teeth, the kappa values for inter- and intra-examiner agreements were 1.00 and 0.97, respectively. In addition, we defined a dichotomous variable: 'missing one or more teeth' or 'none' (53 % were missing at least one tooth).

Explanatory Variables

The demographic variables included gender and education. Postal questionnaires inquired about education with two questions: one for comprehensive school and the matriculation exam, and the other for vocational training. Based on these questions, we defined a three-class ordinal variable. 'Basic education' included those who had not graduated from high school and had no formal vocational qualifications. 'Secondary education' included those who had graduated from high school or vocational school. 'Higher education' comprised participants with a university degree or those who had graduated from a polytechnic or equivalent school.

From the postal questionnaires we also determined tooth brushing frequency and use of alcohol. For tooth brushing frequency, we dichotomized the information on the original variable (with five categories) according to the general recommendation of brushing twice daily: 'once daily or less' or 'at least twice daily'.[13] To avoid missing data, we used similar questions in an additional questionnaire (completed on the day of the health examination), in case corresponding information was missing in the postal questionnaire, which served as the primary source for all the information. Alcohol use was inquired with several questions on the number of consumed standard doses and events of different beverages separately (mild alcoholic beverages: beer, cider and long drink; wine and spirits). We used the classification for alcohol contents per standard doses of different beverages by Sundell et al.[14] and calculated a continuous grams per week variable. Alcohol drinkers were defined as those consuming >230 g/wk for men and >150 g/wk for women, and the rest were defined as light drinkers.

We determined diabetes status using numerous sources: self-reported physician-diagnosed diabetes and medications from the postal questionnaires, hospital outpatient and inpatient registers, and medication registers from the Social Insurance Institution of Finland. The definition of the dichotomous variable (yes/no) did not distinguish between types 1 and 2 diabetes.

Dental plaque served as an indicator of oral health. During the oral health examination, the dentists recorded plaque status ('none', 'visible plaque or plaque detected while probing') for all visible teeth (excluding third molars) and then, for simplicity, we dichotomized this information (yes/no).

Statistical Analysis

We used a negative binomial regression model for the number of missing teeth as a count variable and then calculated unadjusted and adjusted relative risks (RR) with 95 % confidence intervals (CI) for each explanatory variable in the model in question.[15] In addition, we checked over all two-way interaction terms for explanatory variables.

For the count outcome, we performed stratified analyses by gender, education and tooth brushing frequency. Here, we narrowed education as a stratification variable to low and high so that the low stratum included basic and secondary education, and the high stratum included those with a higher education level in our original education variable.

We performed additional analyses to illustrate the correlation between continuous pack-years and the number of missing teeth as the count variable among those who had smoked. For visual preference, this illustration was based on an unadjusted negative binomial model.

We used the statistical package R environment version 3.1.2 for all statistical analyses.[16] For negative binomial modeling, we used the glm.nb function (with no offset option) in the MASS package.