Menu
Menu
Menu
Menu
Background: Childhood obesity is one of the most serious public health challenges of the 21st century. The prevalence of overweight and obesity among children ( 2) and obesity (BMI for age > 3). Regression analyses were performed to investigate risk factors of overweight/obesity. Results: The prevalence of overweight and obesity was 8% (1.7% for obesity alone). Boys were more affected by overweight than girls with a prevalence of 9.7% and 6.4% respectively. The highest prevalence of overweight was observed in the Grassfield area (including people living in West and North-West regions) (15.3%). Factors that were independently associated with overweight and obesity included: having overweight mother (adjusted odds ratio (aOR) = 1.51; 95% CI 1.15 to 1.97) and obese mother (aOR = 2.19; 95% CI = 155 to 3.07), compared to having normal weight mother; high birth weight (aOR = 1.69; 95% CI 1.24 to 2.28) compared to normal birth weight; male gender (aOR = 1.56; 95% CI 1.24 to 1.95); low birth rank (aOR = 1.35; 95% CI 1.06 to 1.72); being aged between 13-24 months (aOR = 1.81; 95% CI = 1.21 to 2.66) and 25-36 months (aOR = 2.79; 95% CI 1.93 to 4.13) compared to being aged 45 to 49 months; living in the grassfield area (aOR = 2.65; 95% CI = 1.87 to 3.79) compared to living in Forest area. Muslim appeared as a protective factor (aOR = 0.67; 95% CI 0.46 to 0.95).compared to Christian religion. Conclusion: This study underlines a high prevalence of early childhood overweight with significant disparities between ecological areas of Cameroon. Risk factors of overweight included high maternal BMI, high birth weight, male gender, low birth rank, aged between 13-36 months, and living in the Grassfield area while being Muslim appeared as a protective factor. Preventive strategies should be strengthened especially in Grassfield areas and should focus on sensitization campaigns to reduce overweight and obesity in mothers and on reinforcement of measures such as surveillance of weight gain during antenatal consultation and clinical follow-up of children with high birth weight. Meanwhile, further studies including nutritional characteristics are of great interest to understand the association with religion, child age and ecological area in this age group, and will help in refining preventive strategies against childhood overweight and obesity in Cameroon.
The current study is based on publicly available data from the fourth Demographic Health Survey Cameroon Database. This Demographic Health Survey fulfilled all ethical requirements and was approved by the Cameroon Ministry of Public Health. Participants’ records in the database are anonymized, thus we did not have access to the identity of participants. The retrospective use of these data was approved by the Institutional Review Board of the School of Health Sciences, Catholic University of Central Africa, Yaoundé, Cameroon. DHS provides data from representative cross-sectional studies that are carried out in the general population, using validated questionnaires. DHS surveys are regularly implemented in over 50 countries worldwide and permit for the main basic demographic indicators and the health situation in those countries to be estimated and updated. In Cameroon, the fourth DHS was conducted from January to August 2011 together with the “Multiple Indicator Cluster Survey” (MICS). Three standard questionnaires (women, men and household) were used [19]. A stratified national sample of 15,050 households was selected randomly in two stages: 580 clusters (or enumeration areas) were drawn for the first stage which included 291 and 289 clusters from urban and rural areas respectively. A count within each cluster led to a list of households from which was conducted a systematic sampling with equal probability. The final survey was carried out in 578 clusters and had included a national representative sample of 11,732 children less than five years from 14,214 households. Eligible households for the mothers’ and children’s anthropometric measurements (weight and height) were chosen by a systematic random draw conducted amongst selected household for the final survey (one household out of two) [19], which resulted to merely 50% of children in the database who had their weight and height recorded and their BMI z-score calculated. We used the kids file dataset of the 2011 DHS database. we excluded children aged less than 6 months since some interventional studies that used an intensive, multidisciplinary approach or parental coaching demonstrated a significant decrease of adiposity over 6 months or at 6 months of age [20]. Were also excluded those whose BMI z-score were missing or was recorded in the database as “Height out of plausible limits” or “Age in days out of plausible limits” or “Flagged cases”, as their values were unusable since they were recorded in the database under special codes which corresponded either to responses that were considered inconsistent with other response in the questionnaire and thought to be probably an error, or to responses which value was “Don’t know”[21]. Finally, a total of 4518 children aged 6–59 months including 2313 girls and 2205 boys living with their mothers were included in our analysis. The authorization for the access to the whole DHS database were obtained after a request explaining what we were intending to do with and which we sent through DHS program website at the address https://dhsprogram.com/ The BMI z-score based on WHO 2006 reference population [22] was used to assess overweight and obesity among children. The thresholds were defined according to WHO recommendations [23]: children who had a BMI z-score less than -2 were considered as thin; those who had a BMI z-score ranging from -2 to 2 were classified as normal weight children; those who had a BMI z-score over 2 were classified as overweight and children who had a BMI z-score greater than 3 were considered as obese. We excluded any BMI z-score recorded under the categories “Height out of plausible limits”, “Age in days out of plausible limits”, “Flagged cases”. Our independent variables were identified using the standard questionnaires and the DHS recoding manual (version 1.0, August 2012) [21]. Children, mothers’ and household characteristics that we considered to be relevant based on literature were chosen amongst the existing variables in the database. A minimal dataset was obtained after having applied exclusion criteria on BMI. We considered as “missing”, any missing case or any case that was recorded in the dataset as “Flagged”. Maternal BMI, initially a quantitative variable was recoded into four classes, based on the international classification of adult overweight and obesity [24]: thin (<18.5kg/m2), normal weight (18.5–24.9kg/m2), overweight (25–29.9kg/m2) and obese (≥30kg/m2). Cases that were recorded as ‘Flagged’ were classified as missing. Implausibility of maternal BMI values was checked using BMI70kg/m2 as threshold of exclusion according a Canadian study [25], since any standard cut off point were found in Sub-Saharan studies. Highest level of education comprised four categories: had never been to school, primary, secondary and higher education. Mothers’ religion was regrouped into Christian (catholic and protestant), animist, Muslim, or other religions categories. For mother’s marital status, the initial classes “married” and “cohabiting with a partner” were regrouped into “couple” category; the other initial classes (“single”, “widowed”, “divorced” and “separated”) were put together in “single parent” category. Mother’s occupation derived from Respondent currently working variable and comprised two classes: currently employed and not employed. Type of place of residence was urban or rural. Wealth index of household, a variable indicating an economic well being score based on housing characteristics and ownership of sustainable goods [26], initially comprised fives levels: poorest, poorer, middle, richer, richest. Given some similarities between “poorest” and “poorer”, and between “richer” and richest”, the first two classes were aggregated and renamed “low economic status” and the last two classes were also merged and renamed “high economic status”. The class “middle” was kept as initially. Number of children under 5 years in the household was broken into two classes based on its gross distribution of frequency: 1–2 children and >2 children Age was recoded in 12 months groups: 6–12, 13–24, 25–36, 37–48, and 49–59 months. Birth weight was categorized into three groups, based on clinical cut off points and WHO definition of low birth weight: low (4000g). As we did not have any data on gestational age in the dataset to deal with implausible birth weight values according to one of the current rules of exclusion [27], all birth weight > 5500g or 3). The variable “region” comprised 12 categories, corresponding to the 10 administrative regions of Cameroon except “Douala” and “Yaoundé” which are the two greatest urban cities, respectively the economic and the political capital, and which belong to Littoral and Center region respectively. Each of these regions is considered as a study area according to DHS survey plan. As regions may be grouped according to their lifestyle and other geographical considerations, this variable was recoded in four ecological areas: Savannah area (Far North, North and Adamawa Regions), Forest area (Centre East and South Regions), Coastal area (Littoral, South-West Regions) and Grassfield area (North-West, West Regions).We kept “Douala” and “Yaoundé” in the new variable since they are the most urbanized and cosmopolite settings in Cameroon. Variables of interest were selected using the Software Package for Statistics and Simulation SPSS version 16.0 for Windows (SPSS, Chicago, Illinois, USA), since DHS databases exist in SPSS format. Data were imported into R software (version 3.1.3) for analysis. A univariate descriptive analysis was performed to assess the structure of all selected variables and to estimate the overall proportions of obese, overweight, normal weight and thin children. A descriptive bivariate analysis was used to examine disparities in overweight and obesity prevalence pertaining to ecological areas, sex and type of place of residence. A Chi square test or eventually a Fisher test and Student t test were used to compare mothers, children and households characteristics amongst thin, normal weight and overweight/obese children. Birth weight, mother religion, Maternal BMI and occupation had respectively almost 40%, 0.5%, 0.4% and 0.2% of missing values, thus we looked at bias issue focusing our interest on birth weight as this variable had the highest and a huge percent of missing values. We found that almost 88% of children with missing value in birth weight had normal weight. Since normal weight children were more likely to have normal birth weight, we assumed that birth weight might been missing in children who were more likely to have normal birth weight and we made the missing at random assumption as this explanation of missing data emerged from the dataset. Thus, imputation of missing data was needed to prevent bias associated with birth weight missing data. Also, we observed that children with birth weight missing values were more likely to live in Savannah area (67%), in rural setting (81%) and in households with low economic status (almost 71%) and to have mothers with lowest level of education (91% of none and primary as the highest level). In addition, children whom mother had lower level of education were more likely to live in savannah area, in rural setting and in households with low economic status. Finally, children whom mother had lover level of education or who lived in rural settings or in household of low economic status were less likely to be obese and overweight, Consequently, to be consistent with the plausibility of the missing at random assumption [28], all these covariates and our outcome variable as well as all other variables were fitted into our imputation model. Data on maternal BMI and religion were also imputed using the same model although these variables relatively had small amount of missing data. Data were imputed using the imputation method of incomplete categorical datasets with a Multiple Correspondence Analysis model [29]. Also, we checked out multicolinearity issue between economic status, parent highest level of education and number of children less than five years in a household as they were a correlation between these variables and found no issue of multicolinearity between them (all their Variance Inflation Factors were less than 3 in multicolinearity test) Odds ratios (OR) with their 95% confidence intervals (CI) served to investigate the factors associated with overweight and obesity. They were calculated by both univariate and multivariate logistic regression analyses while adjusting for potential confounders. We included in the multivariate model all variables with a p value ≤ 0.20 in univariate analyses with child’s age and sex and maternal BMI to be forced in the model even if above the selection limit. Thin children (n = 204) were excluded from the univariate and multivariate regression analysis since we were comparing normal weight to overweight and obese children. Children with birth weight > 5500g (n = 19) were also excluded. Estimates that were obtained from imputed data, both in univariate and multivariate regression analysis were compared to the results of complete case analysis to ascertain their consistency with what were assumed or expected. A p value < 0.05 was set as statistically significant.
N/A
