This study was a secondary analysis of data generated through a household survey that determined the association between digit sucking, caries and periodontitis in 6-months-olds to 12-year-olds resident in Ile-Ife . The primary data were collected using a three-level cluster sampling procedure. This involved an initial random selection of enumeration areas within the Ife Central local government area, then the selection of every third household on each street in the randomly selected enumeration areas in the local government area, and finally the selection of an eligible individual from each household. Children who fell within the target age group, were living with their biological parents or legal guardians, and who consented to participate in the study were recruited for the study. Details on the sampling process had been discussed in the primary study .
Ethics consideration: Ethical approval for the study was obtained from the Health Research Ethics Committee of the Obafemi Awolowo University Teaching Hospitals’ Complex Ile-Ife (ERC/2013/07/14). The primary study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects/patients were approved by the Institute of Public Health-Health Research Ethics Committee. Written informed consent was obtained from the parents or legal guardian of the study participants.
Sample size and data collection procedures: The estimated sample size for this secondary data analysis was based on the ECC prevalence of 6.6% generated from the primary study . With a 3% margin of error and a confidence level of 95% precision, a sample size of 263 study participants will be adequate to determine the association between ECC and malnutrition for this secondary data analysis. There were 370 data points extracted from the primary study database for this analysis. This exceeded the needed minimum sample required for this study. We therefore had adequate sample size to determine an association between ECC and malnutrition.
Socioeconomic Status: Socioeconomic status was assessed because of its role as a determinant of malnutrition. It was measured according to a classification validated for use in Nigeria. The index combines the mother’s education and the father’s occupation to obtain five socio-economic classes for children . Class 1 indicates upper class, class II upper middle class, class III middle class, class IV lower middle class, and class V lower class. For the present analysis, these classes were regrouped into three: high (upper and upper middle classes), middle (middle class) and low (lower middle and lower classes).
Mothers' knowledge of oral health: Folayan et al  described how data on mothers’ knowledge of oral health were collected and handled. The same data were used for this study. Briefly, data were collected using a tool with a possible score ranging from 8 to 40. Scores of 21 and above were categorized as good oral health knowledge, and scores of 20 and below were categorized as poor oral health knowledge.
Cariogenic diet: The frequency of consumption of sugary snacks between main meals was collected with a tool described in detail by Folayan et al . Consumption of sugary snacks between main meals three times a day or more is a significant risk factor for ECC .
Nutritional status: Nutritional status was assessed based on weight and height of each child according to the International Society for the Advancement of Kinanthropometry standard protocol . Weight was measured in kilograms with an electronic weighing scale and recorded to the nearest tenth of a kilogram. The weighing scale was zero-balanced before the child stepped on it. Each child removed heavy items of clothing and mounted the scale bare footed, standing still and looking straight. Measurements were recorded after fluctuations on the digital screen had stopped.
Children’s height was measured in meters with an anthropometer-calibrated customized stadiometer. Each child was asked to step on the standiometer bare footed. Height was the maximum vertical distance from the feet to the vertex of the head, with the head held parallel to the Frankfort plane . In this position, the child looked straight, with arms hanging naturally by the sides and both heels together touching the base of the stadiometer. The heels, buttocks and upper part of the back and back of the head were in contact with the stadiometer. The headpiece was brought down to contact the vertex of the head. The measurement was read to the nearest tenth of a centimeter and converted to a fraction of meter.
Nutritional status was determined according to criteria used by the WHO: height (H), age (A) and weight (W) . Children whose H/A Z-score was below minus two standard deviations from the median of the WHO reference population were considered stunted. Children whose W/H Z-score was below minus two standard deviations from the reference population median were considered wasted. Children whose W/A was below minus two standard deviations from the reference population median were classified as underweight. Children whose W/Z-score was plus one standard deviations from the reference median were considered overweight, and those who were two standard deviations from the reference median were considered obese.
Oral examination for each child was conducted in the homes of the study participants. The participants were examined sitting, under natural light, using sterile dental mirrors and probes by trained dentists who were calibrated on the use of the WHO criteria for caries diagnosis, and the OHI-S index for assessment of the oral hygiene status. Radiographs were not used in this study. The intra-examiner correlation scores ranged between 0.89–0.94, while inter-examiner kappa Cohen scores ranged between 0.82–0.90 for caries detection and OHI-S [37, 38].
ECC status: The dmft was identified according to the World Health Organization (WHO) criteria . The dmft score was obtained by adding the d, m and f scores for each child less than 6 years of age. The dmft score was dichotomized into 0= ECC absent and >0= ECC present.
Oral hygiene status: Oral hygiene status was assessed with the index of Greene and Vermillion . Good oral hygiene was graded from 0.0 to 1.2; fair oral hygiene from 1.3 to 3.0; and poor oral hygiene from 3.1 and above.
Data analysis: The mean(SD) age, dmft, oral hygiene score, and proportion of those with ECC, and malnutrition were calculated. Socio-demographic profile of participants in the present study and the original study were compared using chi-square test to establish comparability and ensuring that the present sample is representative of the target population like the original sample.
The association between nutritional status and ECC presence (yes/ no) and confounding variables (age, sex, socio-economic status, frequency of sugar consumption between meals, oral hygiene status and maternal knowledge of oral health) was determined with the chi-square test or Fisher’s exact test where appropriate.
Multivariable Poisson regression models were used to assess the relationship between exposure (nutritional status) and confounders on the outcome variable (presence of ECC measured by prevalence ratio: ECC PR). A modified theoretical model outlined by Folayan et al. , was used to assess how various indicators explained ECC PR when they were grouped into blocks. Factors were grouped under three blocks and were successively introduced, one block at a time. Each model included the block of nutritional status (WAZ, HAZ and WHZ) as this was the variable assessed for being a risk indicator for ECC. Other blocks moved into the next model if the p value was < 0.2. Model 1 included only the block of nutritional status, while Model 2 included the block of nutritional status and the block of socio-economic and demographic factors (age was excluded because of possible collinearity since it was used to compute nutritional status) since these factors could influence caries and nutritional status. Model 3 included variables that met the inclusion criteria for the next model and they were known moderator variables of the association between ECC and malnutrition namely mother’s knowledge of oral health, frequency of daily consumption of sugar and oral hygiene index. At the introduction of a new block with its factor(s), the variables of all blocks were adjusted simultaneously for each other to produce adjusted prevalence ratios (APRs). To avoid sparse data bias  introduced by the low prevalence of ECC, underweight, overweight or their combination, we used robust variance estimation .
The cross tabulations were generated with IBM SPSS version 23. The test of association was and the Poisson regression analysis was conducted using Stata 15. The test of significance was set at P<0.05.