Transitions between body mass index categories, South Africa

Abstract Objective To profile the prevalence of the three body mass index (BMI) categories by sociodemographic characteristics, and to calculate the percentage transitioning (or not) from one BMI category to another, to inform South African health policy for the control of obesity and noncommunicable diseases. Methods We used data from the National Income Dynamics Study, including sociodemographic characteristics and BMI measurements collected in 2008, 2010, 2012, 2014 and 2017. For each data collection wave and each population group, we calculated mean BMI and prevalence by category. We also calculated the percentage making an upwards transition (e.g. from overweight to obese), a downwards transition or remaining within a particular category. We used a multinomial logistic regression model to estimate transition likelihood. Findings Between 2008 and 2017, mean BMI increased by 2.3 kg/m2. We calculated an increased prevalence of obesity from 19.7% (3686/18 679) to 23.6% (3412/14 463), with the largest increases in prevalence for those aged 19–24 years and those with at least high school education. The percentages of upwards transitions to overweight or obese categories increased sharply between the ages of 19 and 50 years. Once overweight or obese, the likelihood of transitioning to a normal BMI is low, particularly for women, those of higher age groups, and those with a higher income and a higher level of education. Conclusion In the development of national strategies to control obesity and noncommunicable diseases, our results will allow limited public health resources to be focused on the relevant population groups.


Introduction
The mean body mass index (BMI) of the African population is increasing, 1 resulting in a steady rise of the prevalence of people being overweight or obese across Africa, with the southern part of Africa being most affected. 1,2 In 2016, the prevalence of the population aged ≥ 15 years being overweight or obese in South Africa was 68% for women and 31% for men. 3 Global efforts to combat obesity include the World Health Organization (WHO) Global strategy on diet, physical activity and health, 4 the Global action plan for the prevention and control of noncommunicable diseases 2013-2020 5 and the United Nations (UN) High-level meetings of the General Assembly on prevention and control of non-communicable diseases. 6,7 The Global action plan proposes the promotion of healthy diets by Member States to halt the rise in the prevalence of school children, adolescents and adults being overweight or obese. Similarly, the 2011 Sixty-sixth session of the UN Political Declaration of the High-level Meeting of the General Assembly on the Prevention and Control of Non-communicable Diseases committed to strengthening national policies and health systems by promoting multisectoral and multistakeholder engagement to reverse, stop and decrease the rising trends of obesity in child, youth and adult populations. 6 In line with global strategies and policies, the South African Department of Health developed the Strategic plan for the prevention and control of non-communicable diseases 2013-2017 8 and the Strategy for the prevention and control of obesity in South Africa 2015-2020; 9 the targets of these strategic plans were to reduce obesity prevalence by 3% by 2017 and by 10% by 2020 in all age groups. These two strategic plans are aligned with the agenda of the country's 2030 National development plan for the promotion of healthy diets and physical activity at schools, workplaces and in the general community. 10 Promotion and support through research is an essential component of global and national strategies for prevention and control of obesity and noncommunicable diseases. 4,5,8,9 We anticipate that a better understanding of transitions between the BMI categories -normal, overweight and obese (Table 1) 11 -will allow the improvement of interventions to reduce the prevalence of obesity. Our objectives are: (i) to profile the prevalence of the three BMI categories within a study population according to various sociodemographic characteristics, and to estimate the percentage of these population groups that underwent transitions (or not) between BMI categories; (ii) to identify the factors associated with transitions between BMI categories; and (iii) to discuss the key public health implications of our findings for national obesity control strategies.

Study population
The National Income Dynamics Study, first conducted in 2008, is a nationally representative panel study that collects information on a wide variety of social, demographic, economic and health characteristics of the civilian non-institutionalized population. 12,13 We used data from the five completed waves of the panel survey (the subsequent four waves were conducted in 2010, 2012, 2014 and 2017) from participants for which anthropometric measurements had also been recorded. In the first wave in 2008, the survey recorded weight and height measurements for the calculation of BMI for 21 002 individuals; 2323 (11.1%) were immediately lost to follow-up.

Study variables
Our main study variable of interest is whether transition occurred from one BMI category to another during a particular period in time. This derived variable has seven possible outcomes: two downwards transitions (either from obese to overweight or from overweight to normal); two upwards transitions (either from normal to overweight or from overweight to obese); and three no-transition outcomes, when a person's BMI category does not change from either normal, overweight or obese between two waves of the survey. If a population is experiencing a higher number of upwards than downwards transitions, the prevalence of adverse conditions will increase.
Our independent time-invariant variables were sex and race, and baseline time-variant variables were age, whether urban or rural residence, education level, equivalized income level and frequency of physical exercise. To account for the economies of scale in household consumption, we used the equivalization method of the Organisation for Economic Co-operation and Develop-ment of dividing household income by the square root of the number of people within the household. 14

Statistical analysis
For each BMI category and for each of the five waves of data collection, we calculated both average BMI with 95% confidence intervals and prevalence. To visualize how prevalence varies with age, we calculated the prevalence of all three BMI categories for all ages and averaged over all five data collection waves (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017). We also calculated the percentage within each population subgroup either transitioning to an upwards or downwards BMI category or remaining within the same category during the four periods between subsequent waves (i.e. ending in 2010, 2012, 2014 and 2017). We used a multinomial logistic regression to model the probability of transitioning (or not) from one BMI category to another, relative to the probability of remaining within the normal category. We performed all statistical analyses using Stata SE Version 15.0 (StataCorp, College Station, United States of America).

Average BMI and prevalence
Between 2008 and 2017, mean BMI increased by 2.3 kg/m 2 , from 23.1 to 25.4 kg/m 2 . We observed that the age group 7-13 years experienced the highest increase (by 4.7 kg/m 2 ), followed by the age groups 14-18 years (3.3 kg/ m 2 ) and 19-24 years (3.2 kg/m 2 ; Table 2; available at: http:// www .who .int/ bulletin/ volumes/ 98/ 12/ 20 -255703). We note that the groups that demonstrated at least an average increase in mean BMI (i.e. ≥ 2.3 kg/m 2 ) include women, Africans and Caucasians, rural dwellers, those with some education and those whose level of physical exercise was unknown. When examining the data by income, those with the lowest income demonstrated the largest increase in mean BMI (2.5 kg/m 2 ). We also observed that women, those aged ≥ 25 years, Caucasians, urban dwellers, those with no formal schooling, those with a high school education or more, those within the highest-income tertile and those exercising < 3 times per week had an average BMI of > 25.0 kg/m 2 for most, if not all, of the study period, contributing to the high prevalence of people being overweight and obese ( Table 2).
Our data show an increase in the prevalence of obesity from 19.7% (3686/18 679) in 2008 to 23.6% (3412/14 463) in 2017. In terms of age group, we observed the highest increase in the prevalence of obesity over this period in those aged 19-24 years from 10.9% (227/2079) to 28.4% (449/1581). In terms of education level, those with the most education (high school and above) demonstrated the largest increase in the prevalence of obesity from 25.2% (708/2811) to 40.1% (840/2093). The prevalence of people being overweight and obese was lowest for those aged around 11-18 years (Fig. 1). The prevalence of obesity increases steeply after this age up to around 50 years, when it reaches a plateau (Fig. 1). Between the ages of 40 and 70 years, the prevalence of obesity is higher than that of normal BMI.

Transitions between categories
We provide percentages of the study population groups who transitioned (or not) from one BMI category to another during the four periods ending 2010, 2012, 2014 and 2017 in Table 3 (avail-

Likelihood of transitions
The probability of transitioning upwards or downwards to another BMI category decreased with time, while the probability of remaining within the overweight or obese categories increased with time.
The particular groups that demonstrated the largest probabilities of remaining within the overweight or obese categories, compared with other groups, also demonstrated the greatest likelihoods of transitioning either upwards or downwards. These groups included women, Caucasians, and those with at least a high school education and with a high income (Table 4). We note that the likelihood of either transitioning upwards or downwards, or of remaining within the overweight or obese categories, relative to that of retaining a normal BMI, increased with age. Our data show that women were 2.56 and 1.82 times more likely than men to transition upwards and downwards, respectively, and 4.81 times more likely than men to remain within the overweight or obese category. Men were therefore more likely to retain a normal BMI.
We note that Caucasians were 1.39 and 1.26 times more likely than Africans to transition upwards and downwards, respectively, and 1.21 times more likely than Africans to remain within the overweight or obesity category. Those of mixed ancestry and Asians demonstrated a higher probability of retaining a normal BMI than Africans and Caucasians. Urban dwellers were slightly more likely (1.08 times) than rural dwellers to remain within the overweight or obese categories, but less likely to transition upwards or downwards. We observed that individuals with a high school education and greater were more likely than the other education-level groups to either transition upwards or downwards or to remain within the overweight or obese categories. Compared with those with a low or middle income, high-income groups were more likely to transition upwards or downwards and remain within the overweight or obese categories; high-income groups were therefore less likely to retain a normal BMI than lower-income groups. Those who exercised ≥ 3 times per week and those whose frequency of physical exercise was unknown (data were unavailable for those of age < 15 years) were less likely to remain within the overweight or obese categories, and therefore more likely to retain a normal BMI, compared with those who exercised < 3 times per week.

Discussion
Our data show a sharply rising prevalence of obesity coinciding with entry into adulthood in the community at large. We note that the prevalence of obesity increased by the greatest amount for the group aged 19-24 years between 2008 and 2017. This age group also reported the highest percentage of upwards transitions, as well as the largest difference between the percentage transitioning upwards and the percentage transitioning downwards. Although this age group had a normal-category mean BMI during the first two study periods, the subsequent sharp increase in BMI resulted in an increased prevalence of being overweight and obese with time.
The South African Department of Health targeted a 3% reduction in the prevalence of obesity by 2017 for all age groups through diet and physical activity. 8,9 To determine whether this target was achieved, we must compare data from different individuals of the same age at different times. However, by examining how BMI values change as a particular cohort ages, we were able to identify periods of increased obesity prevalence within a person's lifetime during which interventions may be critical. Our data show that there exists a decreasing trend in prevalence of being overweight or obese from birth to the age of 18 years, when most children usually leave high school. As our results are not capable of attributing this downwards trajectory from birth to age 18 years to the effectiveness of national strategies for obesity prevention and control (our data were obtained from an observational study), further research is required.
Guidance for healthy eating for the population aged ≤ 18 years is provided in South Africa by the Guidelines for early childhood development services, 15,16 and the National school nutrition programme annual report 17 and Guidelines for tuck shop operators. 18 The teaching of life skills in primary and high schools, which includes nutrition education, is aimed at educating students to make nutritious food choices and develop healthy eating habits. 9 In averaging our data for all ages over the period 2008-2017, we found that the lowest combined prevalence of being overweight and obese was for those aged 7-18 years; the fact that prevalence rises steadily after this age highlights that, despite these strategies on controlling obesity, there remain substantial problems in communicating healthy food choices to children and adolescents. Other studies have also shown that these strategies are not functioning optimally. [19][20][21] A synthesis of studies conducted between 2006 and 2014 on the South African school food environment revealed that over half (from 51.1%; 1233/2412 to 69.3%; 330/476) of the students bought available and unhealthy foods from either tuck shops or vendors in their neighbourhood. 22 The population of age 15-24 years has also been identified as the largest consumers of sugarsweetened beverages. 23 Low levels of physical activity, especially in female adolescents (aged 10-19 years) 24 and in older teenagers (aged 16-18 years) of both sexes, could also be contributing to the existing prevalence of being overweight and obese. 25 Physical activity policies include the Strategic plan 2009-2013: An active and winning nation, 26 of which one of the aims was to facilitate the implementation of sports in schools. In the National sport and recreation plan, one of the objectives is to maximize access to sport, recreation and physical education in every school. 27 Consolidated findings on the level of physical activity in children aged 3-19 years (e.g. early childhood physical activity, organized sport participation, active play and active transportation) showed an average participation of around 50%, although the percentage of those in early childhood (age 3-6 years) participating in physical activity was generally found to be high (~80%). 25 Since pre-school children are naturally active, a physical activity intervention for this age group was only deemed necessary for promoting the engagement of teachers and parents or caregivers, and for outcomes such as cognitive development. 28 The South African noncommunicable disease strategy 8 targeted an increase in the prevalence of physical activity (150 minutes of moderate-intensity physical activity per week, or equivalent) by 10% by 2020, and the WHO global action plan for noncommunicable diseases 5 targeted a 10% relative reduction in the prevalence of insufficient physical activity by 2020. Although our data do not allow a detailed analysis of types of physical activity, we demonstrated the importance of physical exercise. About half of those who exercised ≥ 3 times per week maintained a normal BMI even as the cohort aged. However, the fact that the proportion of this group who maintained a normal BMI decreased slightly over time may reflect decreasing levels of energy expenditure with age. Interventions that increase physical activity may therefore be necessary to reduce the large percentages of groups who remain either overweight or obese. Studies on obesity, especially in women, have identified obese participants who meet the WHO guidelines on physical activity but have very poor cardiorespiratory fitness, possibly attributable to a lack of high-intensity physical activity. 29,30 Suggested interventions include encouraging participation in high-intensity activities, such as sports and aerobics, and discouraging sedentary behaviour. 31 The recent introduction of higher tax on products such as sugar-sweetened beverages 32 will probably reduce consumption and consequently reduce the percentages transitioning upwards. However, the life course approach, the basis of many WHO strategies and recommendations for disease prevention and control, 5,33 may be a better strategy for maintaining normal BMI from this age. The life course approach stresses the importance of early intervention by considering which stages, transitions and settings of a person's life are critical for promoting or restoring health. 34 For example, the workplace has been found to be an important setting for health promotion during adulthood. 35 Effective interventions include the promotion of healthy food options in canteens, and the provision of nutrition education and counselling. 36 Similar interventions in settings such as community centres, churches and recreational facilities should also be promoted. 37 Barriers to leisure activity participation, including lack of time and facilities, safety issues and negative community perceptions regarding weight loss, may also need to be addressed, 38 along with the accessibility and affordability of healthy food options through subsidies. 37 Population groups that demonstrated a higher probability of remaining within the obese category included women, higher age groups (> 35 years), and those with a higher income and higher level of education. These groups are usually the target of public interventions and of the weight management programmes of private health-care providers. 39 However, our data show a high resistance to a downwards transition from either overweight or obese categories, and this resistance increases with time. Other studies have shown that although weight loss can be achieved through lifestyle changes, 40,41 maintaining this weight loss over the longer term can be difficult; only approximately one-fifth (20.6%; 47/228) 42 of those who achieve weight loss maintain it for ≥ 1 year, 42,43 and the majority of those who embark on a weight-loss programme give up before achieving success. [44][45][46][47] Our results indicate that, despite considerable percentages of the cohort transitioning downwards, any reductions in the prevalence of obesity were cancelled out by the larger percentages of these population groups transitioning upwards; we therefore believe that the South African target 9 of achieving a 10% reduction in the prevalence of obesity by 2020 will not be met.
The main strength of our study was our analysis of BMI transitions within a nationally representative sample of participants from a panel survey spanning birth to adulthood. However, our study had limitations. Some of the risk factors associated with higher BMI categories in this study, such as household income, education level and amount of physical exercise, were self-reported and therefore prone to minor errors that may have affected our results. A second limitation was loss to followup from one wave of data collection to another. Since we assume those lost to follow-up were randomly distributed throughout the study population, and the remaining numbers at each subsequent data collection wave were sufficiently high, we consider our results to be free of bias.
To conclude, we have demonstrated that the proportion of upwards transitions to overweight and obese categories in the South African population increases sharply between the ages of 19 and 50 years. Once overweight or obese, the likelihood of transitioning to a normal BMI is low, particularly for women, those of higher age groups, and those with a higher income and a higher level of education. With this evidence, we provide essential guidance for the South African Department of Health in the development of strategies to control obesity and noncommunicable diseases.
Our study results will allow limited public health resources to be focused on the necessary population segments in which the largest reductions in the prevalence of obesity can be made. ■ Funding: MEW received funding from SAMRC and University of the Witwatersrand.