In 1982, the three maternity hospitals in Pelotas, a southern Brazilian city, were visited daily and 7392 births were identified. Those children whose parents lived in the urban area of the city (N
= 5914) were examined and the mothers interviewed. These individuals have been followed on several occasions (Barros et al. 2008
In 1997, a census was carried out in a systematic sample of 70 from the 259 census tracts located in the urban area of the city, and subjects born in 1982 were linked to the study records and those subjects belonging to the cohort were interviewed and examined (n
= 1,076). In 2004–2005, we tried to follow up the whole cohort and the subjects answered a questionnaire on sociodemographic, health and behavioural variables. At the end of the interview, the subjects were invited to visit the research laboratory to give a blood sample. Concerning the assessment of oral health, a sample of 900 individuals examined in 1997 was randomly selected to participate in the oral health study (OHS-97). Of the 900 selected subjects, 888 (98.7%) agreed to participate in the OHS-97 (Peres et al. 2011
). In 2006, those individuals who participated in the OHS-97 were again contacted, 720 of 888 individuals (81.1%) were located and participated in the OHS-06. In the OHS-06, all teeth were examined for the presence of periodontal diseases (gum bleeding, dental calculus and periodontal pocket), dental caries (DMFT), quality of restorative treatments, oral cavity lesions and use/needs of dental prosthesis. The interview covered issues related to hygiene, use of dental services and dental pain. In 2006, oral examination was performed at home by six dentists and four students at final year of graduation in dentistry, previously trained and calibrated. Examiners underwent theoretical and practical training covering diagnostic criteria and details on each index. Practical exercises were performed on 20 patients, and reproducibility assessment was performed on 25 volunteers. Diagnostic inter-examiner reproducibility was measured, and the lowest Kappa was 0.60 for gingival bleeding but most of the values were close to 1.00. More details were reported elsewhere (Peres et al. 2011
The phases of the study were approved by the Ethical Review Board of the Faculty of Medicine of the Federal University of Pelotas, and written informed consent was obtained from participating subjects.
Definition of outcomes
All teeth were examined at six different sites (mesiobuccal, mediobuccal, distobuccal, mesiolingual, mediolingual and distolingual). The following periodontal disease outcomes were assessed:
- Gingivitis: all sites were probed, waiting 10 s to verify the presence or absence of gingival bleeding. The variable was categorized as absent, a tooth and two or more teeth with gingival bleeding.
- Calculus: all sites were probed for detection of calculus. The variable was dichotomized in absence or presence of dental calculus. Presence was considered if calculus was present in at least one surface.
- Periodontal pocket: all sites were probed, pocket should have probing depth ≥4 mm in at least one site. The variable was dichotomized in absence or presence of periodontal pocket. Presence was considered if periodontal pockets were present in at least one surface.
For each outcome, the results were recorded in separate forms for each tooth in relation to the outcomes.
Definition of exposures
In each visit, the subjects were weighted and their height was assessed. BMI was estimated in kg/m2
. Waist circumference was measured in 2004–2005 and measured at the narrowest part of the trunk directly on the skin. For individuals with no visible waist circumference, this measure was made at the midpoint between the iliac crest and last rib. Measures of waist circumference were categorized according to sex in normal (men < 94 cm, women < 80 cm), level 1 (men ≥ 94 and <102 cm, women ≥80 and <88 cm) and level 2 (men ≥ 102 cm; women ≥ 88 cm; Lean et al. 1995
At the age of 15 years, the following cut-off was used to categorize the BMI: eutrophic (BMI in z
score for age and sex ≤1 SD), overweight (BMI > 1 and <2 SD) or obesity (BMI ≥ 2 SD; World Health Organization 2007
). At 18 and 23 years, the following cut-off were used to categorize BMI eutrophic (BMI < 25 kg/m2
), overweight (BMI ≥ 25 and ≤29.9 kg/m2
) and obesity (BMI ≥ 30 kg/m2
; World Health Organization 1998
In the follow-up visits at 15, 18 and 23 years, individuals were classified as obese or non-obese according to WHO criteria previously cited. We created a variable that summarized the number of times that each subject was considered obese. Therefore, the individuals could have been classified as obese in none, one follow-up and two or more follow-ups with obesity.
Confounders and mediators variables
The following variables were considered possible confounding factors:
Sex, skin colour [classified as white and black (black + brown)], smoking at 23 years (never smoked, former smoker and smoker), attained schooling at 23 years (classified according to the highest level of schooling in years at 0–4, 5–8, 9–11, 12 or more), family income at the age of 23 years (categorized into tertiles according to the total family income in the month preceding the interview), and the asset index at 23 years of age was estimated by factor analysis from scores of household goods such as vacuum cleaner, washing machine, DVD, fridge, freezer, microwave oven, computer, telephone, radio, television, automobile and air conditioning. This variable also included salaried housemaid, maternal education and family income. From the different components generated in the factor analysis, the first was used to create a continuous score that included all variables who contributed directly. This variable was subsequently divided into tertiles.
The following possible mediating factors were also included in the analysis: use of dental floss (yes/no); reported frequency of brushing (0–2, 3 and 4 or more times a day) at age 24 years; percentage of dietary energy intake from carbohydrates at age 23 years categorized in tertiles; and C-reactive protein level at age 23 years (low ≤ 1.0 mg/l, moderate from 1.01 to 3.0 mg/l and high 3.01–10.0 mg/l). C-reactive protein values greater than 10.0 mg/l were excluded from the analysis as they are related to acute inflammation, not chronic (Pearson et al. 2003
). In the 2004–2005 visit, the subjects answered a food frequency questionnaire, and the percentage of dietary energy from carbohydrates was estimated (Olinto et al. 2011
Descriptive analysis of the population was performed using absolute and relative frequencies and Fisher's exact test. Multinomial logistic regression was used to assess the relationship between number of teeth with gingivitis (none, 1 or ≥2). Poisson regression was used to estimate the prevalence ratio of calculus and periodontal pocket (Barros & Hirakata 2003
All potential confounding factors which showed p < 0.2 were selected to remain in the multivariable analysis.
Effect modification by gender, income and smoking was evaluated with interaction terms in the multivariate analysis. Analyses were performed with software STATA 11.0 for Windows, StataCorp., College Station, Texas.