This was a five-year retrospective cohort study which involved primary school children initially aged 6 years attending government aided schools in the city of Shah Alam and mixed urban–rural districts of Seberang Prai, Malaysia. Both, Shah Alam and Seberang Prai districts have access to fluoridated public water supply at 0.5 to 0.8 ppm since 1976 as part of the national water fluoridation programme. Apart from water fluoridation, fluoridated tooth paste is widely available in both areas
. However, topical fluorides are not indicated and rarely used
A significance level of 95% and power of 80% was accepted for the purpose of sample size calculation. Using the proportion-based sample calculation formula
, a minimum sample size of 737 children was needed for the study. Another sample size calculation was done to test the association of the outcomes with exploratory variables. A sample size of 432 was calculated to detect an odds ratio of at least 1.1 with, significance level of 95% and 80% power, for a caries prevalence of 45% among the non-exposed. To compensate clustering, a design effect of 2 was adopted and the sample size was inflated to double. Moreover, a further 10% was added to compensate possibility of incomplete records and 20% for possible confounding factors in the analysis, totaling 1915 individuals. The initial sample included 1987 six year old children, but 157 records were excluded due to incomplete information throughout the 5 years study period. Therefore, the final sample size consisted of 1830 students.
A multi-stage cluster random sampling was used in the study. At the first stage, all the schools and number of standard six enrolment in each school from both the districts were listed. Calculated sample size (1915) was divided by total number of standard six enrolments in all the schools in both the districts (14035) to calculate probability of selection (0.136) of each student. To maintain equal weight for each student this probability of selection was implemented at each school level and number of sample from each school was calculated. E.g. a sample of 26 students was selected from the school with 192 enrolments (26=192*0.136), whereas, 46 students were selected from the school with 338 enrolments (46=338*0.136).
The research protocol for this study was approved by the Research Ethics Committee of Universiti Teknologi MARA. Parental consent was obtained at the beginning of this programme. Five years data was gathered retrospectively in 2010 from annual school dental records. These record cards included information about demographic data (age, gender and ethnicity), oral health condition (caries experience, gingivitis) and treatment need for each child. All these information in record cards were collected from 2004 to 2009 as part of the School Incremental Dental Care Programme run by Ministry of Health Malaysia. All clinical examination and records were done by trained dental therapists. The clinical data represents records of the dental status and treatment given to each child which was kept by the school administration. Caries experience of each child was recorded annually using World Health Organization (WHO, 1997) diagnostic criteria for caries namely, decayed (D), missing (M), filled (F) teeth (T) and surfaces (S)
The schools were contacted for permission to retrieve the dental records. To maintain confidentiality, the subject’s names were not disclosed. One of the authors (N.Y) examined all the dental records without knowing to which city or school the records belong to. All record cards with complete information from 2004 to 2009 were included for data entry and analysis until the minimum sample size was met. All data were entered by two dental nurses independently to ensure accuracy.
The data were analyzed using the "R-Project" statistical program. The rates of dental caries were presented in prevalence and incidence density of carious lesions from baseline to final examination. Risk assessment was done to assess relative risk for having caries after 5 years in children with baseline caries level. Additionally, bivariate and multivariate logistic regression analysis were performed to establish relationship between dental caries and selected explanatory variables (gender, ethnicity and rural/urban), and odds ratio (OR) and 95% confidence interval (CI) were calculated. Urban and rural classification followed the Malaysian statistics department guidelines. Schools in areas of less than 10,000 inhabitant were classified as rural school and schools in areas of more than 10,000 inhabitants were classified as urban school. In regression analysis, outcome variable was dichotomized according to individuals in the caries free group (DMFT=0, coded as 0) and individuals with caries (DMFT>0, coded as 1). Bivariate logistic regressions were performed individually in separate model with each explanatory variable; model 1 included gender (Female=0 and Male=1), model 2 included ethnicity (Indian=0, Chinese=1, Malay= 3) and model 3 included rural/urban (Rural=0 and Urban=1). Finally, all the explanatory variables were combined into a multivariate logistic regression model.
Incidence in the present study was calculated as incidence density for risk of caries of a person (IDp
) and of a tooth (IDt
) summarized by the formula below:
Person-time in IDp and tooth-time in IDt represent the sum of observation-time for individuals/tooth having no caries. Under School Incremental Dental Care Programme the schools were visited by the examiners at the same time each calendar year for screening and treatment. If a student in a particular school was examined in September 2004, he was followed up in September 2005, 2006, 2008 and 2009. An important assumption for incident density is the accurate recording of the disease-free state of each person at all stages of follow up throughout the duration of study. In case of caries for a person and tooth, it was not clinically feasible to record exact time of onset of caries in a tooth or person. Therefore, we adopted an approach used by Beck et.al. (1997) and Thitasomakul et.al (2006) to calculate caries incidence density in infants
[11, 12]. It was assumed that the teeth had a uniform risk rate within each follow-up interval. For teeth that were present as sound at baseline (t0) and final (t1), the time at risk was = t1 – t0. For teeth which were absent at t0, the time of eruption was assumed to be half way between two examinations (t1 – t0)/2. For teeth that were present as sound at t0 and became carious teeth at t1, the onset of caries was assumed to take place at halfway between baseline and final examination (t1-t0)/2. Similarly, if the teeth were absent at t0 and became carious teeth at t1, the eruption time was calculated to be at (t1 – t0)/2, whereas, the onset of caries was assumed to be at t0+3/4(t1-t0). In other words, the caries-free duration of these teeth was one-fourth of the interval. When a carious lesion was detected, that person/tooth was considered to be a new case for the period. It was then excluded from the at-risk status