Changes in body mass index, obesity, and overweight in Southern Africa development countries, 1990 to 2019: Findings from the Global Burden of Disease, Injuries, and Risk Factors Study

Abstract Background High body mass index (BMI) is associated with stroke, ischemic heart disease (IHD), and type 2 diabetes mellitus (T2DM). An epidemiological analysis of the prevalence of high BMI, stroke, IHD, and T2DM was conducted for 16 Southern Africa Development Community (SADC) using Global Burden of Diseases, Injuries, and Risk Factors (GBD) Study data. Methods GBD obtained data from vital registration, verbal autopsy, and ICD codes. Prevalence of high BMI (≥25 kg/m2), stroke, IHD, and T2DM attributed to high BMI were calculated. Cause of Death Ensemble Model and Spatiotemporal Gaussian regression was used to estimate mortality due to stroke, IHD, and T2DM attributable to high BMI. Results Obesity in adult females increased 1.54‐fold from 12.0% (uncertainty interval [UI]: 11.5–12.4) to 18.5% (17.9–19.0), whereas in adult males, obesity nearly doubled from 4.5 (4.3–4.8) to 8.8 (8.5–9.2). In children, obesity more than doubled in both sexes, and overweight increased by 27.4% in girls and by 37.4% in boys. Mean BMI increased by 0.7 from 22.4 (21.6–23.1) to 23.1 (22.3–24.0) in adult males, and by 1.0 from 23.8 (22.9–24.7) to 24.8 (23.8–25.8) in adult females. South Africa 44.7 (42.5–46.8), Swaziland 33.9 (31.7–36.0) and Lesotho 31.6 (29.8–33.5) had the highest prevalence of obesity in 2019. The corresponding prevalence in males for the three countries were 19.1 (17.5–20.7), 19.3 (17.7–20.8), and 9.2 (8.4–10.1), respectively. The DRC and Madagascar had the least prevalence of adult obesity, from 5.6 (4.8–6.4) and 7.0 (6.1–7.9), respectively in females in 2019, and in males from 4.9 (4.3–5.4) in the DRC to 3.9 (3.4–4.4) in Madagascar. Conclusions The prevalence of high BMI is high in SADC. Obesity more than doubled in adults and nearly doubled in children. The 2019 mean BMI for adult females in seven countries exceeded 25 kg/m2. SADC countries are unlikely to meet UN2030 SDG targets. Prevalence of high BMI should be studied locally to help reduce morbidity.

The SADC countries are facing challenges similar to other developing countries, that currently have rapid demography and lifestyle changes. [14][15][16][17][18] Five-fold regional and seven-fold in-betweencountry differences in the risk of stroke were observed among these countries, with the East and Central SSA, which include SADC countries, having a lifetime risk ranging from 28% to 37%. 14 While there was a significant reduction in the risk of stroke in all Global Burden of Disease (GBD) regions and high-income countries, and also in most countries between 1990 and 2015, there was a significant increase in the lifetime risk of stroke in seven SADC countries ranging from 4% to 32%, However, in three of the SADC countries, changes in the risk of stroke were not substantial. 14 The prevalence of T2DM, IHD, and cardiovascular disease (CVD) increased 10-fold in SSA between 1988 and 2008. 19 Between 2010 and 2030, the number of adults with T2DM in developing countries is projected to increase by 69%, compared to a 20% increase in developed countries. 20 The 16 SADC countries are a regional economic community whose aim is to increase regional socioeconomic integration to achieve greater economic growth and alleviate poverty (www.sadc.int/ about-sadc). Levels of development and poverty, social service delivery, and economic performance vary greatly among these member states. Key among the aims of SADC is the strengthening of economic cooperation and integration, providing for cross-border investment and trade, free movement of goods and services across national borders. Although this region is at increased risk, there is no readily available systematic understanding of the distribution of obesity and overweight and time trends across all 16 SADC countries. SDG Target 3.4 aims for a one-third reduction, relative to 2015 levels, in death rate due to CVD, cancers, chronic respiratory diseases, and diabetes by 2030 in populations aged 30-70 years. 21 Evidence-based information on the level and trends of BMI, obesity, and overweight is essential to support policy and goal settings, program evaluation, and decision-making. Despite the increased burden of high BMI in the region, no comprehensive estimates of the burden exist for the SADC countries. Therefore, the goal of this analysis was to highlight the burden of obesity and overweight in SADC region and its impact on health, as well as help inform strategies and interventions to reduce high BMI in turn reducing the burden of stroke, IHD, and T2DM attributable to high BMI, thus improving the chances for attaining SDG Target 3.4. This is a descriptive, epidemiological analysis of the prevalence of high BMI, stroke, IHD, and T2DM in 1990 and 2019 for the 16 SADC countries using secondary data from the Global Burden of Diseases, Injuries and Risk Factor (GBD) Study. 22

| METHODS
The GBD is a systematic, scientific effort by the Institute for Health Metrics and Evaluation (IHME) to quantify the comparative magnitude of health loss due to diseases, injuries, and risk factors by age, sex, and geographies for specific points in time. The GBD study estimates country-specific incidence, prevalence, mortality, years of life lost (YLLs), years lived with disability (YLDs), and disability-adjusted life-years (DALYs) due to diseases such as T2DM, IHD, and stroke and conditions such as high BMI. All GBD estimates adhere to the 14 Guidelines on Accurate and Transparent Health Estimate Reporting (GATHER). 23 GATHER recommends making available statistical codes, details on the use or non-use of specific sources, and how primary data are adjusted. In this study outcomes of interest were BMI, mortality, stroke, IHD, and T2DM morbidity attributable to high BMI.
Full details of data sources, availability of data, and methods of dietary risk factors and BMI are described in detail elsewhere, 24,25 However, to describe it briefly, GBD combines information from several sources including country-specific surveys and administrative databases, weighting it based on quality decided by a panel of experts. Where individual-level weight and height survey data were available the mean BMI was used to determine the prevalence of overweight and obesity. Medline for studies that provide nationally or sub-nationally representative estimates of BMI, overweight, or obesity among children or adults were systematically searched. The search terms, selection criteria, and flow diagrams of screening are provided elsewhere, 24 The search was conducted using the following  26 Also searched was the Global Health Data Exchange (http://ghdx.healthdata.org) for multi-country survey programs, national surveys, and longitudinal studies that provide selfreported or measured data on height and weight for children or adults. Full details of data sources, availability of data, and methods are described in detail elsewhere. 24 Mean prevalence of obesity and overweight were estimated from Spatiotemporal Gaussian process regression modeling. 25 The coefficients of this regression were applied to the prevalence of overweight and obesity to estimate the mean BMI for each country, according to age, sex, and year. Estimation was improved in datasparse countries using a wide range of covariates with plausible relationships to overweight and obesity. Three country-level covariates with best fit comprising of energy intake per capita, the absolute latitude of the country, and the proportion of persons living in urban areas were included as independent variables (Supplement 1). 24 CoD ensemble modelling (CODEm) framework was used to model cause-specific death rates in the GBD. 27 Mixed effects linear models and spatial-temporal Gaussian Process Regression models were used to estimate cause fractions and death rates. The current version of the data download tool is available in the GHDx and contains core summary results for the GBD 2019: http://ghdx. healthdata.org/gbd-results-tool. Input data for stroke, IHD, and T2DM was obtained from a combination of vital registration, verbal autopsy, and ICD codes where available. Verbal autopsy is a validated GONA ET AL.
-511 method for determining causes of death and cause-specific mortality fractions in populations without a complete vital registration system.
Verbal autopsies consist of a trained interviewer using a questionnaire to collect information about the signs, symptoms, and demographic characteristics of a recently deceased person from an individual familiar with the deceased. 28 The burden of disease related to high BMI for each of T2DM, IHD, and stroke attributable to high BMI was estimated by calculating the population attributable fraction (PAF) for each country, segregated by age, sex, and year. 29 Cause of deaths due to each of the three outcomes was modeled with the CODEm approach. Covariates included in the ensemble modelling process for IHD and stroke were total cholesterol, smoking prevalence, systolic blood pressure, trans fatty acid, mean BMI, and others.
Deaths attributed to Alzheimer's and other dementias, Parkinson's disease, or atrial fibrillation and flutter were not considered as stroke related. Country-level covariates with best fit and coefficients chosen for inclusion in the ensemble modelling process for T2DM were fasting plasma glucose, prevalence of diabetes mean BMI, mean cholesterol, mean systolic blood pressure, prevalence of obesity, and others. Complete lists and description of input data, covariates used in CODEm, model specifications and processing sequence are provided online in Supplementary Appendix 1. 25 CODEm relied on four key components: (a) all available data were identified and gathered to be used in the modelling process; (b) a diverse set of plausible models were developed to capture welldocumented associations in the estimates; (c) the out-of-sample predictive validity was assessed for all individual models, which were then ranked for use in the ensemble modelling stage; and (d) differently weighted combinations of individual models were evaluated to select the ensemble model with the highest out-of-sample predictive validity. Out-of-sample predictions are used to calculate mean squared error (MSE), root-mean-squared-error (RMSE), coefficient of variation (CV), and 95% coverage of predictive intervals representing the proportion of observed out-of-sample data that fall within the predicted 95% credible intervals. The variance of the random walk was determined by estimated residuals from the out-ofsample validation. The out-of-sample 95% coverage for stroke, IHD, and T2DM all exceeded 97.5%.
The total morbidity related to high BMI was calculated as the total of disease-specific burdens. PAF for BMI categories (20 to 24, 25 to 29, and ≥30) and each outcome, stroke, IHD, and T2DM were used to help understand locations with high concentrations of BMI burden. The Das Gupta method, 30 details of which are described elsewhere 30 was used to decompose the change in deaths and DALYs attributed to high BMI between population growth, population age structure, risk exposure to high BMI, and rates of risk-deleted mortality and DALYs. Rates per 100,000 population of YLL, YLD, and DALYs attributable to high BMI were computed.
The flowchart in Figure S5 describes the process of estimating the disease burden of high BMI (Supplementary Appendix). 26 The relative risk per 5-unit change in BMI for each disease endpoint was obtained from meta-analyses. Relative risks for all outcomes, by age and sex, are reported elsewhere (Table S2 of the Supplement 1). 24 The GBD study estimates country-specific incidence, prevalence, mortality, YLLs, YLDs, and DALYs due to stroke, IHD, and T2DM. Cause-specific crude and age-standardized death rates per 100,000 population were obtained from the Cause of Death Ensemble model (CODEm) and spatiotemporal Gaussian process regression (ST-GPR).
Deaths were multiplied by standard life expectancy at each 5-year age-group to calculate YLLs. Cause-specific mortality was estimated using a Bayesian meta-regression modelling tool, DisMod-MR.
Prevalence estimates were multiplied by disability weights for mutually exclusive sequelae of diseases to calculate YLDs. 24 YLLs were calculated using the product of age-specific life expectancy from the reference life table used in the GBD study. YLDs were calculated as a product of the prevalence of the disease-specific disability weights. 31,32 In this calculation, risk-deleted mortality is defined as the burden of disease in the absence of the risk factor, that is, rates of death from T2DM, IHD, and stroke assuming everyone is at the lowest-risk BMI category. Adding together YLLs and YLDs yields DALYs. Rates per 100,000 population of YLL, YLD, and DALYs attributable to stroke, IHD, and T2DM were computed.

| RESULTS
Age standardized prevalence of overweight and obesity for adoles-   There was a positive linear association between the change in BMI from 1990 to 2019 in males and females, both with Pearson correlation coefficient r = 0.74, p < 0.001 suggesting that countries with the greatest increase in mean BMI in females tended to be the same countries with the greatest increases in mean BMI in males as well.

| Mean BMI
Notably, the DRC and Madagascar did not experience an increase in BMI in both adult females and adult males ( Table 2).

| Rates of deaths, YLLs, YLDs, and DALYs due to high BMI
Age-standardized mortality attributed to high BMI increased by   (Table 4). Table 5 shows the results of mortality rates due to IHD, stroke, and T2DM attributed to high BMI among persons 15 years or older.    A previous GBD study reported an increase in prevalence of obesity globally, more in adults than in children and greater in women than in men. 24 As found in the present study, the NCD Collaboration, 34 BMI increased in all regions and age groups, and was higher, for women than men for the period 1975-2016. 35

| Study limitations
Our study is not without limitations. General limitations of the GBD methodology are described elsewhere. 13,14,24 In this particular case, limitations include incomplete medically certified cause of death systems in SADC countries and the non-availability of traditional individual, social, and environmental risk factors for T2DM, IHD, and stroke such as health care access, urbanization, employment, physical inactivity, dietary risks, alcohol and cigarette use, and lipids, among others. These factors would make estimates presented in this paper more robust.

| Recommendations
Individual and population-wide interventions need to be established around strategies with a strength in disease management. Individual interventions must consist of prevention and treatment. 9,37,38 For individuals with overweight or obesity, lifestyle modification targeting changes in diet and physical activity through a multidisciplinary approach could be the cornerstone of interventions for weight management. 38 Health researchers and policymakers should develop targeted interventions to reduce consumption of certain foods classified as "harmful" and encourage people to eat healthpromoting foods. Where applicable, referral of patients for further interventions such as behavioral therapy and treatment-including pharmacotherapy and bariatric surgery-is recommended. 9,37,38 Prevention is the ideal option in the SADC at present. Population approaches for overweight/obesity recommended for children should include restricting advertisements that promote consumption of unhealthy foods, providing healthy school meals, and encourage school-based good nutrition education. 9,37 Other strategies should include use of the WHO STEPwise approach to NCD risk factor surveillance (STEPS) tool. 34 The tool is used to collect data and measure NCD risk factors within the WHO STEPwise approach to surveillance. The STEPS Instrument covers three different levels or "steps" of risk factor assessment: Step 1 (questionnaire), Step 2 (physical measurements), and Step 3 (biochemical measurements).

| CONCLUSION
The burdens of high BMI (overweight and obesity), stroke, IHD, and T2DM increased dramatically by 2019 compared to 1990. The prevalence of obesity more than doubled in adults and increased 1.7-fold in children. The 2019 mean BMI for adult females in 7 of the 16 countries was greater than the overweight cut-point of >25 kg/m 2 . The countries with greatest increase in mean BMI in females tended to be the same countries with the greatest increases in mean BMI in males. SADC countries do have a "window "of opportunity to halt and reverse these negative health outcomes before they become endemic and out of control.