The GUSTO cohort and study design
The GUSTO cohort is a mother-offspring cohort study that recruited over 1200 pregnant women aged 18–50 years delivering at the National University Hospital (NUH) and KK Women’s and Children’s Hospital (KKH) between June 2009 and September 2010. The participants were of self-identified homogenous ethnic background representing the 3 major ethnicities of Singapore: Chinese, Malay and Indian. Subjects were followed very closely from birth with detailed observations and samples collected to provide insights into the management and prevention of important diseases. The study received ethical approval from the institutional review boards of the National Healthcare Group and Singapore Health services. (DSRB reference D/09/021)(CIRB reference 2009/280/D) Informed consent was obtained from all participants and their legal guardians. Additionally, all experiments were performed in accordance with relevant guidelines and regulations.
This is a cohort study utilizing the GUSTO data where the various domains of the GUSTO study examined nested groups for different health parameters including nutrition and oral health. Demographics such as household income, maternal age, maternal education, and birthweight were recorded at the initial recruitment visit.
Infant diet data – independent variable
Infant diet intake data was collected from mothers or caregivers by mailing 3-day food diaries (Supplemental Methods) prior to their 6-, 9-, and 12-months postnatal visits. In cases where diaries were returned incomplete, a 24-hr recall interview by trained personnel was collected during the visit using a 5-stage, multiple-pass interviewing technique23. Data from either the 24-hr recalls or 1-day record from food diaries were used for dietary analyses to increase the sample size. A correlation (r ranging from 0.123 to 0.820) of the 1-day record with the 2 other days using a subset of infants with complete 3-day food diaries has been previously established21.
Each food item from the records was then assigned to one of the 72 sub-food groups within the 18 food groups based on type of food or similarities on nutrient content, conceptually similar to previous studies10, was used to categorize the foods consumed locally. Among the food groups, the confectionary and sugar sweetened beverages (SSB) groups were selected for analyses in this study. The confectionary food group consist of food such as chocolates, sweets, ice-cream, puddings and jellies while SSB consists of fruit drinks, carbonated soft drinks, sweetened soya milk, traditional drinks and other sweetened drinks like honey mixed with water.
Dietary pattern trajectories were calculated using multi-level mixed models generating intercepts and gradients of various patterns as described by Smithers et al.12. Further details can be found in Lim et al.21. Briefly, dietary trajectories were empirically constructed by mapping dietary patterns extracted by exploratory factor analysis at 6-, 9- and 12-months. The dietary patterns across the three time-points were examined for their similarity in the types of foods and their loadings to ascertain their suitability to be modelled as trajectories. The mapping of each pattern to a trajectory was based on similar key constituent foods (with high loadings) found in the dietary patterns across the three time-points and this correspondingly determined the name of each trajectory. In this study, 4 dietary pattern trajectories observed in the participants namely: 1) Predominantly breastmilk, 2) Guidelines, 3) Easy-to-prepare foods and 4) Noodles (in soup) and seafood21.
The multi-level mixed model generates estimates of intercepts and gradients. The intercept reflects the trajectory score at the start point of each pattern (6-months of age), while the gradient denotes the rate of change in trajectory scores over time from 6-months to 12 months of age. For example, a high Guidelines dietary pattern intercept score indicates close adherence to the Guidelines dietary pattern at 6 months of age, while a high Guidelines dietary pattern gradient score indicate continued high adherence to the Guidelines dietary pattern from 6 to 12 months of age.
Infants following the Predominantly breastmilk pattern, had more breastmilk than formula milk with feeding of fresh fruits at 9-months, and consumption of bean-curd, ethnic breads and starchy vegetables at 12-months. The Guidelines dietary pattern had high consumption of rice porridge, fish and meat, fresh fruit and vegetables, and followed the currently recommended WHO weaning guidelines15,16. The infants who followed the Easy-to-prepare pattern had foods such as infant cereals, juices, cakes and biscuits. The Noodles (in soup) and seafood pattern was a uniquely Asian pattern that was characterized by Asian adult foods such as noodles accompanied by eggs, seafood, bean curd and dried preserved fruits21.
Oral health data – primary outcome
Oral examinations were carried out by 3 trained calibrated dental professionals at ages 2 and 3 years (intraclass correlation coefficient, >0.80) using the modified International Caries Detection and Assessment System (ICDAS) criteria24. The participants were examined in the supine position using the knee-to-knee technique. Artificial illumination was used and tooth surfaces examined with sterile mouth mirror and blunt explorer. The tooth status was assessed by visual inspection, aided by tactile inspection if necessary after cleaning and drying of the surfaces with sterile gauze25. No radiographs were taken. Potential confounding factors in oral hygiene habits such as night-time bottle feeding, brushing frequency, and fluoride exposure were recorded as a self-administered questionnaire at the age 2 and age 3 oral examination.
As there is a lag time for the influence of dietary patterns to manifest into examinable changes in teeth, caries status at 2 to 3-years were examined for association with their dietary patterns profiled at 6 to 12-months.
Data analysis
The primary outcome, ECC, was coded in 3 ways and analysed independently as a binary variable (present/absent), number of decayed teeth (dt) and number of decayed surfaces (ds) with white spot/cavitation lesions (International Caries Detection and Assessment System (ICDAS) II codes 2–6). An initial univariate analysis of food groups most likely to be associated with ECC, namely sugar confectionary and sugar sweetened beverages (SSB) was conducted using a logistic regression. The amount of sugar consumption was estimated from the diet diaries. Furthermore, marginal associations of dietary pattern trajectories with early childhood caries at ages 2 and 3 years were estimated using generalized estimating equation (GEE) with exchangeable working correlation matrix. Negative binomial family distribution with log link function was used. Incidence rate ratio (IRR) for each variable in the model was estimated with the use of Huber-White sandwich robust estimates for standard errors. Interactions between age and dietary trajectories were not included in the models as they were not statistical significant (p-value > 0.10) as shown by Wald test, indicating similar effects of dietary trajectories on ECC at 2 and 3 years. Confounding factors found in the literature such as socio-demographic characteristics, oral hygiene habits, perinatal and postnatal characteristics were adjusted in the analyses.
Significance level was set at p-value < 0.05. Statistical software STATA/SE Version 14 (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP) was used to carried out all statistical analysis.
