Prevalence and predictors of overweight and obesity among Cameroonian women in a national survey and relationships with waist circumference and inflammation in Yaoundé and Douala

Abstract Information on the distribution and predictors of obesity in Africa is needed to identify populations at risk and explore intervention options. Our objectives were to (a) examine the prevalence and geographic distribution of overweight and obesity among Cameroonian women; (b) evaluate change in anthropometric indicators among urban women between 2009 and 2012; (c) examine associations between household and individual characteristics and overweight and obesity; and (d) examine relationships between body mass index (BMI), abdominal obesity, and inflammation. We analysed data from a nationally representative survey conducted in 3 geographic strata (North, South, and Yaoundé/Douala) in Cameroon in 2009 and a survey in Yaoundé/Douala in 2012. Participants selected for this analysis were nonpregnant women, ages 15–49 years (n = 704 in 2009; n = 243 in 2012). In 2009, ~8% of women were underweight (BMI < 18.5) and 32% overweight or obese (BMI ≥ 25.0). Underweight was most common in the North (19%) and overweight and obesity in the South (40%) and Yaoundé/Douala (49%). Prevalence of BMI ≥ 25.0 in Yaoundé/Douala did not differ in 2012 compared with 2009 (55.5% vs. 48.7%; P = 0.16). Residence in urban areas, greater maternal age, and TV ownership were independently related to overweight and obesity in national and stratified analyses. In Yaoundé/Douala in 2012, 48% (waist‐to‐hip ratio > 0.85) to 73% (waist circumference > 80 cm) had abdominal obesity. Body mass index was positively associated with abdominal obesity and inflammation. Though causal inferences cannot be drawn, these findings indicate population subgroups at greatest risk for overweight and associated health consequences in Cameroon.


| INTRODUCTION
The prevalence of overweight and obesity is increasing, even in lowincome countries (NCD Risk Factor Collaboration (NCD-RisC), 2016).
Projections are for an increased prevalence of overweight and obesity, from 1.3 billion people affected in 2005 to 3.28 billion in 2030; currently 57 of 129 countries have high prevalence of both underweight and overweight (Global Panel on Agriculture and Food Systems for Nutrition, 2016). The growing worldwide problem of overweight, obesity, and diet-related chronic diseases in the face of persistent poorquality diets is thought to be driven by urbanization and globalization.
Food system changes may still be able to prevent dire outcomes in countries in earlier stages of nutrition transition.
In West Africa, most countries are in the early stages of transitioning from undernutrition to overnutrition, though some (Ghana, Cape Verde, and Senegal) are in later stages (Bosu, 2015).
For example, Ghana has experienced dramatic increases in maternal overweight/obesity from 12.8% in 1993 to 29.2% in 2008. The nutrition transition in this region is associated with an increase in other risk factors, for example, hypertension in Cameroon (Bosu, 2015;Fezeu, Kengne, Balkau, Awah, & Mbanya, 2010). Diabetes prevalence is also increasing in some areas within countries, particularly urban Ghana and Senegal (Bosu, 2015), and incidence is projected to double between 2000 and 2030 (Haggblade et al., 2016). In a systematic review of African studies conducted between 1972 and 2011, maternal obesity (body mass index [kg/m 2 ] ≥ 30, based on pre-pregnancy or first trimester measurements) occurred in 9-17.9% of pregnancies and was associated with increased risk of C-section birth deliveries, macrosomia, gestational diabetes, pregnancy-induced hypertension, and preeclampsia but decreased risk of maternal anaemia and low birthweight (Onubi, Marais, Aucott, Okonofua, & Poobalan, 2016).
Although information is sparse, available data suggest that the burden of overweight is greatest in urban areas and among wealthier, older women (Jones, Acharya, & Galway, 2016;Kandala & Stranges, 2014).
In Cameroon, urban residents reported lower levels of physical activity than their rural counterparts, but the relationship between physical inactivity and obesity was more consistent for men than women (Sobngwi et al., 2002). A cross-sectional study among adults of African origin from Cameroon, Jamaica, and the United Kingdom also found gender to influence the relationship of dietary intakes to overweight or obesity (Jackson et al., 2007).
Since prevalence of maternal overweight and obesity varies geographically within and across countries in sub-Saharan Africa, more research is also needed in this region on the prevalence and distribution of these conditions. In addition, evaluating the relationship of BMI to other risk factors, including abdominal obesity, as measured by waist circumference, and inflammation, may provide insight regarding the risks of overweight in this context. In a cross-sectional study of community-based samples from urban and rural Cameroon and France, the relationship of waist circumference to blood pressure, blood lipids, fasting blood glucose, and other risk factors differed by ethnicity/race, gender, and urbanization (3). Elsewhere in Africa, Traissac et al. found abdominal obesity to be more prevalent and variable than overall obesity based on BMI and highlighted a need for more research among women living in a nutrition transition context (Traissac et al., 2015).
Timely, national studies of maternal weight status and child growth are urgently needed to guide policies and strategies to prevent nutrition-related chronic diseases within countries and regions. Many studies in Africa still focus on exclusively on undernutrition, despite the emerging problem of maternal obesity and overweight within this region. This paper uses data from a 2009 national survey and a 2012 regional survey in Cameroon, conducted for the purpose of designing and evaluating a food fortification program. In this paper, the objectives were to (a) examine the prevalence and geographic distribution of overweight and obesity among women of reproductive age; (b) evaluate the change in anthropometric indicators between 2009 and 2012 in two major cities (Yaoundé and Douala); (c) examine the associations between household and individual characteristics and overweight and obesity among women; and (d) assess the prevalence of abdominal obesity (elevated waist circumference) and relationships between BMI and abdominal obesity and inflammation among women in urban areas.

| Setting
Cameroon is classified by the World Bank as a lower middle income country, with slightly over half the population living in urban areas (54% in 2015) and a gross domestic product per capita (purchasing power parity) estimated at $2,829 in 2014 (Global Nutrition Report, 2015). There has been limited progress on indicators related to health and undernutrition over the past decade: The under-5 mortality rate was 95 per 1,000 in 2012, ranked 21st globally, and stunting prevalence was 33% in 2011 (Global Nutrition Report, 2015). Previous work revealed a high prevalence of anaemia and micronutrient deficiencies among women and children (Engle-Stone, Ndjebayi, Nankap, & Brown,

Key messages
• Overweight and obesity affect one third of women in Cameroon and are particularly common in the cities of Yaoundé and Douala and other urban areas.
• Overweight and obesity were more common among older and more educated women, women who consumed sweetened beverages, and women in households with TV ownership and greater socioeconomic status.
• BMI was positively associated with markers of abdominal adiposity and inflammation, suggesting increased cardiometabolic risk.
• Underweight and micronutrient deficiencies are still prevalent in Cameroon, particularly the North, so interventions to prevent malnutrition in all forms are needed. 2012), although large-scale fortification appears to have reduced this risk for selected micronutrients (Engle-Stone et al., 2017). At the same time, concerns about overweight and associated chronic disease risk in Cameroon have arisen. For example, a 2013 survey using the WHO STEPWISE approach to Surveillance (STEPS) methodology reported an age-standardized prevalence of 29.7% for hypertension among the 15,470 adult participants from urban areas (Kingue et al., 2015). This paper presents data from a nationally representative survey conducted in Oct-Dec, 2009, and a regional survey conducted in Oct-Nov, 2012 -Stone et al., 2012). Thus, the South (representing~45% of the national population) and North (~33% of total population) included both urban and rural areas. Within each stratum, 30 clusters were selected using probability proportionate to size sampling. Next, 10 households per cluster were selected by first identifying a random start point within the cluster and then conducting systematic sampling of nearby households.
Households were eligible for the survey if there was a child 12-59 months of age and a woman 15-49 years of age who was the child's primary female caregiver. If there were multiple children in the target age group in the household, a child was selected at random.
Households were excluded if the woman and child had not lived in the household for at least 1 month, or if either individual had experienced severe fever, diarrhoea with dehydration, or other severe illness during the 3 days prior to data collection. Women provided informed oral consent for themselves and the index child to participate. The study was approved by the Institutional Review Board of the University of California, Davis, and the Cameroon National Ethics Committee. For both surveys, only women who reported that they were not currently pregnant and for whom anthropometric data were available were included in the current analysis.

| Questionnaires, dietary data
Interviewers administered questionnaires to the index woman to collect information on variables relating to household socio-economic status (SES), her exposure to media (originally for the purpose of assessing potential outlets for promotion of fortified foods), and frequency of intake of selected foods by the woman and child during the previous week. Pregnancy was determined by self-report. Interviewers were female, and most had some college education. All interviewers were multilingual (spoke French, English, and at least one other local language) and completed a 1-week study training on research ethics, interviewing techniques, and study procedures and instruments.
Within a questionnaire module on exposure to media, participants were asked how many days in the past 7 days they watched television and how many minutes they watched television on the last day they watched television (Helen Keller International Cameroun et al., 2011).
A food frequency questionnaire (FFQ), based on the Fortification Rapid Assessment Tool (Micronutrient Initiative & PATH Canada, 2003), was administered to collect information on selected foods or dishes prepared with any of four potentially fortifiable foods: refined cooking oil, wheat flour, sugar, or bouillon cube (Engle-Stone et al., 2012). For each item, participants were asked how many days in the previous week they consumed the item, and how many times per day the item was consumed on the last day on which it was consumed.
As indicators of processed snack food intake, the analysis described below included the mother's frequency of consuming (a) packaged biscuits (sweet and savoury); (b) sweets (hard candy, chocolate, and ice cream); (c) sweetened beverages (carbonated beverages, fruitflavoured juice, and popsicles); and (d) beignets (fried dough), which are prepared by local vendors and commonly consumed for breakfast or as a snack.
The original Fortification Rapid Assessment Tool questionnaire was field tested in multiple settings, but not formally validated. However, the adapted questionnaire used in this setting provided data consistent with the results of 24-hr dietary recalls (Engle-Stone et al., 2012). For example, both methods indicated almost ubiquitous consumption of bouillon cube (>90% of respondents consumed in the past day (24-hr recall) and past week [FFQ]) and more frequent consumption of refined vegetable oil and wheat flour in Yaoundé/Douala compared with the South and North.

| Anthropometry
In both surveys, women were asked to remove shoes, hats, and any heavy outer clothing (jackets, etc.) prior to anthropometric measurements. Measurements were conducted by anthropometrists who completed study-specific training and standardization exercises prior to the study. Height was measured in duplicate to the nearest 0.1 cm ENGLE-STONE ET AL. with a portable stadiometer (Seca Leicester Portable Height Measure, Seca Weighing and Measuring Systems), and weight was measured in duplicate to the nearest 0.1 kg with an electronic scale (Seca 899, Seca Weighing and Measuring Systems). For women with hairstyles that could not be compressed to measure the height of the crown of the head, the height of the hair was estimated and subtracted from the final measurement. If the difference between the two height or weight measurements was greater than 0.5 cm or 0.5 kg, respectively, a third measurement was taken and the closest two measures recorded. The mean difference in repeat measurements was 0.03 cm and 0.01 kg in 2009 and 0.04 cm and 0.01 kg in 2012.
In 2012 only, waist circumference (measured at the navel, over light or no clothing) and hip circumference (measured at the widest point over light clothing) were measured to the nearest 0.1 cm using nonflexible measuring tape. The tape was removed before taking a second measurement. If the two measurements were more than 0.5cm apart, a third measurement was taken and the closest two measurements were recorded. The mean difference in repeat measurements for hip and waist circumference in 2012 was 0.05 cm for both.
For women >18 years, underweight was defined as BMI < 18.5; overweight or obese as BMI ≥ 25.0; and obese as BMI ≥ 30.0 (World Health Organization, 2018). For females 15-18 years, modified BMI cut-offs to indicate obesity among adolescents were applied, ranging from 23.94 at 15 years of age to 25.0 at 18 years of age for overweight and 28.30 at 15 years to 30.0 at 18 years for obesity (Cole, Bellizzi, Flegal, & Dietz, 2000). However, these cut-offs did not affect the classification of any study participants as overweight or obese, compared with the adult cut-offs of 25.0 and 30.0 (i.e., the prevalence of overweight was identical, regardless of the cut-off applied for adolescents).

| Markers of inflammation
In both surveys, venous blood was collected and centrifuged withiñ 2 hr to separate plasma, which was frozen on the day of collection.

| Data analysis
SAS software (Version 9.4, SAS Institute, Cary, NC) was used for data analysis. To characterize SES, factor analysis (PROC FACTOR) was used to combine variables related to household possessions; occupation, employment, and education of the index woman and household head; house construction materials; and sanitation facilities. The continuous score generated from the primary factor was then categorized into quintiles. SES scores were calculated separately for the 2009 and 2012 surveys.
The average age of the breastfeeding child among the 29% of women who were lactating in 2009 was 10 months (median: 9.9 months; range: 1.0 to 24.0 months). In 2012, in Yaoundé/Douala, 32% were lactating, and average age of their children was 6.7 months (median 5.6 months; range: 1.3 to 22.5 months). In 2012, mean parity was 2.6 births (data not collected in 2009).
In 2009, about 8% of women were underweight (BMI < 18.5) and 32% were overweight or obese (BMI ≥ 25.0) nationally, but the patterns differed widely by region. Underweight was more common in the North region (19%), and overweight and obesity were more common in the South (40% BMI ≥ 25.0) and Yaoundé/Douala (49% BMI ≥ 25.0). Regional differences in BMI were explained by differences in weight, which varied from a mean of 55.1 kg in the North to 67.2 kg in Yaoundé/Douala (P < 0.0001 overall), rather than height, which did not differ on average by region (P = 0.12 overall).  (Table 1), but the differences were not statistically significant (P = 0.16).

| Factors associated with overweight/obesity among women
In bivariate comparisons, maternal age was the most consistent factor associated with increased odds of overweight/obesity (Figure 1; Table 2; Table S1). The mean BMI and prevalence of overweight did not differ among women who were or were not currently lactating (nationally, within geographic stratum, and within socio-economic quintiles). However, among lactating women, BMI was negatively correlated with the infant's age (Spearman's r = −0.13, P = 0.02).
Nationally, the prevalence of overweight and obesity varied by SES, from 8.6% in the lowest quintile to 51.5% in the highest quintile. Nationally, the prevalence of overweight was greater among women who consumed sweetened beverages in the past week compared with those who did not consume sweetened beverages in the past week (43% vs. 26%, P < 0.0001), but this difference did not persist within geographic strata. Consumption of sweets or biscuits in the past week was not associated with prevalence of overweight, but prevalence of overweight was lower among women who consumed beignets in the past week (29% vs. 40%, P = 0.028). Sweetened beverage consumption was strongly related to SES, with the proportion of women who consumed sweetened beverages in the previous week increasing monotonically from 12% in the lowest SES quintile to 67% in the highest SES quintile (overall Rao Scot Chi-square P < 0.0001). The proportion of women who consumed beignet in the previous week was 72-79% in the three lowest SES quintiles and 63-66% in the highest two SES quintiles (Rao Scott Chi-square P = 0.03 overall). Consumption of sweets or biscuits in the past week was not associated with SES quintile (P > 0.4).  (17), but these did not change the classification of individuals as overweight or obese compared with the adult cut-offs.
In multivariable models, factors that were independently associated with greater prevalence of overweight were residence in Yaoundé/Douala or in an urban area, older age, and household TV ownership (Table 3)

| Associations between BMI and markers of abdominal obesity and inflammation
In Yaoundé/Douala in 2012, BMI was positively associated with waist circumference, waist-to-hip ratio, and waist-to-height ratio (

| DISCUSSION
This study presents estimates on the prevalence and geographic distribution of overweight and obesity among women in Cameroon, a transitioning country, and examines demographic and behavioural risk factors. Though causal inferences cannot be made between individual behaviours/characteristics and overweight/obesity, the findings of this cross-sectional national study are useful to identify subgroups at greatest risk for overweight and associated health consequences in Cameroon. Wide regional variation exists in the prevalence of overweight and obesity in Cameroon, with a large burden among women in the South region and in Yaoundé and Douala, whereas underweight among women remains prevalent in the North region. In addition to urban residence, overweight/obesity was independently associated with older maternal age and ownership of a TV.
TV ownership could be an indicator of physical inactivity but also may serve as a proxy for SES or preferred leisure activities. An    proportion of households with a TV ranged from 22% in the North to around 90% in Yaoundé/Douala. The corresponding population attributable fraction for overweight and obesity among women was 42-44%, with or without adjusting for region. Although causality cannot be attributed, presence of a TV in a household was a strong predictor of overweight in this study and may be an important avenue for targeting households or delivering messages related to prevention and management of overweight.
Sweetened beverage consumption was associated with overweight nationally, but not within each of the three survey strata. The overall association may reflect the contribution of sweetened beverages to total energy intake but also an association between sweetened beverage consumption and SES. The lack of association within strata may reflect reduced statistical power due to the lower sample size and/or lower variation in sweetened beverage consumption and overweight within strata (compared with between strata). Consumption of beignets was associated with lower prevalence of overweight, but this relationship was not significant in multivariable models and may reflect confounding by SES since there was a trend toward greater consumption of beignets among women in the three lowest SES quin-

tiles. Other dietary variables (consumption of sweets and biscuits)
were not associated with overweight, possibly because the FFQ was designed for another purpose (estimating intake of selected fortifiable foods) rather than focusing on processed, sweetened snack foods specifically. However, another cross-sectional study that administered a FFQ among adults of African origin from Cameroon, Jamaica, and the United Kingdom did not find any dietary variables, except for higher protein intake, to be related to overweight or obesity in Cameroon women (Jackson et al., 2007).
The observation that mean BMI and prevalence of overweight did not differ among lactating women compared with nonlactating women may be explained by the wide age range of infants of lactating women in this sample (in 2009: 1 to 24 months of age; mean of 10 months). In the early postpartum period, lactating women are likely to have higher BMI than nonlactating women, reflecting weight gain during pregnancy. However, prolonged lactation may contribute to reduced risk of overweight; our observation of a negative correlation between BMI and infant age among lactating women supports this. The contribution of postpartum weight retention to overweight among women and the potential role of breastfeeding support to weight management in this setting should be explored further. Among women in Yaoundé/ Douala, higher BMI was associated with greater waist circumference and greater waist-to-hip ratio, similar to a previous study among adults in Yaoundé (Pasquet, Temgoua, Melaman-Sego, Froment, & Rikong-Adié, 2003). These findings suggest that overweight in this context may be associated with increased cardiometabolic risk. We measured waist circumference at the navel, rather than midway between the lowest rib and iliac crest, which may introduce error into the measures  (17), but these did not change the classification of individuals as overweight or obese compared with the adult cut-offs. Subgroups with n < 10 individuals were excluded from models. Nonsignificant variables (P > 0.05) were removed using a stepwise backward regression method. In the South (2009) In this study, the prevalence of overweight was greatest in the stratum composed of the two largest cities, and urban residence remained significantly related to overweight in multivariable models.
Other African studies have noted greater risk of overweight in urban areas (Jones et al., 2016;Kandala & Stranges, 2014 (Cohen, Boetsch, Palstra, & Pasquet, 2013). More than 90% of urban adults in Cameroon reported sedentary activity during their leisure time, and many younger adults are also sedentary at work (Kengne, Awah, Fezeu, & Mbanya, 2007). To launch social marketing campaigns, more qualitative and quantitative research needs to explore social norms and perceptions about healthy diets, health risks of physical inactivity (including too much TV/screen time), culturally acceptable leisure time physical activities, and the relationship of body weight to health.

| CONCLUSION
Overweight and obesity are common among women in Cameroon, particularly in the largest cities of Yaoundé and Douala and in urban areas generally. Overweight was more common among households with greater socio-economic status and among older women and women with more education. Household and individual characteristics such as TV ownership and sweetened beverage consumption were also associated with prevalence of overweight. Because of the study design, we cannot determine whether the observed associations between overweight and behaviours such as TV ownership are causal in nature. However, these observations are valuable in identifying subgroups of the population who are at highest risk for overweight and should be targeted for related interventions. Geographic region, in particular, is likely to be an important consideration for program planning. BMI was associated with markers of abdominal obesity and inflammation, suggesting an increased risk of adverse health consequences if trends toward increased BMI continue.
Finally, although our analysis focused on predictors of overweight, it is important to note that underweight was common among women in the northern regions (19%), in addition to anaemia, stunting, and micronutrient deficiencies. Efforts to address undernutrition and micronutrient malnutrition should continue, but evolve toward prevention of malnutrition in all its forms (overnutrition and undernutrition) throughout the life cycle to address the double burden of malnutrition.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.