The present study is a retrospective, registry-based cohort study using information collected from data sources at the Public Health Care Administration in Stockholm as well as from national registries at the National Board of Health and Welfare and at Statistics Sweden (SCB). The Regional Ethics Board in Stockholm and the Swedish Data Inspection Board, a national agency that serves as an institutional review board for studies using database linkage, approved the protocol for this study.
The study included all children born in 2000–2003 who resided in Stockholm County, Sweden at age 3 years (n = 83,147). We followed this cohort until the individuals reached age 7 years. During this period the subjects received regular dental check-ups at the Public Dental Service, with private practitioners, or at the Division of Pediatric Dentistry, Department of Dental Medicine, Karolinska Institutet. The final study cohort comprised 65,259 children (33,423 boys and 31,836 girls) who had been examined at both 3- and 7 years of age. The drop-out rate was 22%, and the most common reason for dropping out was that the child had moved out of the area.
Collection of Dental Caries Data
Data on manifest caries lesions were collected from clinical and radiographic examinations as decayed teeth (dt), extracted teeth (et), and filled teeth (ft). Manifest caries was defined as caries on smooth surfaces at the lowest level that can be verified as a cavity and is detectable by probing or, in fissures, by a catch of the probe under slight pressure. Proximal caries on radiographs was defined as manifest caries in which the lesion clearly extends into the dentin . Only children with clinical indications received a radiographic examination. The decayed, extracted, and filled primary teeth (deft) index was then calculated to determine the severity of the caries experience at ages 3- and 7 years. Caries increment from age 3- to 7 years comprised all new caries lesions in primary teeth (i.e., the difference between the deft values at age 3- and 7 years). No permanent teeth were included in the 7-year outcomes.
In Sweden, the personal identity number (PIN), a 10-digit number unique for each resident and used for indexing in all health and census registries, makes the national registries extremely useful in epidemiological research . Information on individuals is easily extracted from the various registries. The present study used information from the Medical Birth Registry (MBR), the Total Population Registry (TPR), the Income and Taxation Registry (IoT), and the Registry of Education.
Medical Birth Registry
The Centre for Epidemiology at the Swedish National Board of Health and Welfare maintains the Swedish Medical Birth Registry (MBR). We collected the following information from the MBR: gender, birth order, maternal age, family situation, maternal smoking habits during early pregnancy, and the mother’s height and weight at the first visit to the public maternity healthcare clinic. We chose “birth order” as our key exposure. Due to their low number, we merged families with five or more children into one group; thus, “birth order” had five subgroups: first-born, second-born, third-born, fourth-born, and fifth- or later-born children. Maternal age was categorized into three subgroups: <25 years, 2534 years, and ≥35 years or older. Family situation was dichotomized into cohabiting or single parents. Smoking habits during early pregnancy were dichotomized into no or daily smoking. Body mass index (BMI) of the mother was calculated and classified as BMI < 25.00, BMI = 25.0029.99 and BMI ≥ 30.00. Table 2 presents all collected child and parental characteristics.
Total Population Registry
The SCB maintains the TPR. We collected data on maternal and paternal country of birth from the TPR and dichotomized birth country as “Sweden” or “abroad” for the mother and the father.
Income and Taxation Registry
The Swedish National Tax Board collects data on individuals’ annual income tax based on income tax returns and tax-authority decisions. The Board then sends summary statistics to the SCB. From this registry, we collected information regarding the family’s disposable income from the 2004 survey. We categorized and analyzed the variable “family income” into five quintiles from highest to lowest income in the statistical analysis.
Registry of Education
We obtained data on maternal education level from the Registry of Education, which is updated each year in April. In the statistical analysis, we classified “education level” according to the number of years of schooling in 2004 as low (≤ 9 years), intermediate (10–12 years), and high (≥ 13 years), (Table 2).
The Statistical Package for the Social Sciences (IBMSPSS® Statistics version 22.0, [SPSS, Inc., IBM Corp., Armonk, NY, USA]) and STATA 13 for Windows (StataCorp. 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP) was used for data analysis. Descriptive analyses included relative and absolute frequencies, means, and standard deviations. Differences between categorical variables were assessed using the chi-square test.
Forward stepwise binary logistic regression was done to analyze the associated factors of the outcome “caries increment from age 3- to 7 years” (Δ deft > 0). The five subgroups of “birth order”, the key exposure, were analyzed in an univariate analysis (Table 2) and a multivariate analysis (Table 3). The analyses dichotomized all outcomes to distinguish between subjects with and without caries increment (Δ deft > 0 or = 0, respectively).
The final regression models were adjusted for potential confounders. A combination of methods was used to select the confounders included in the models; confounders were selected based on their association with the outcome, as well as based on their association with the key exposure and their subsequent influence on the outcome . Model I was adjusted for gender and maternal age; Model II, for maternal smoking and maternal BMI; and Model III, for socio-demographic factors (parent’s country of birth, maternal educational level, family situation, and family income). Model IV included all groups of variables as adjusted. The odds ratios (OR) with a 95% CI and statistical significance set at α = 0.05 was used to estimate associations.