Articles The shift of obesity burden by socioeconomic status between 1998 and 2017 in Latin America and the Caribbean: a cross-sectional series study Safia S Jiwani, Rodrigo M Carrillo-Larco, Akram Hernández-Vásquez, Tonatiuh Barrientos-Gutiérrez, Ana Basto-Abreu, Laura Gutierrez, Vilma Irazola, Ramfis Nieto-Martínez, Bruno P Nunes, Diana C Parra, J Jaime Miranda Summary Background The burden of obesity differs by socioeconomic status. We aimed to characterise the prevalence of obesity Lancet Glob Health 2019; among adult men and women in Latin America and the Caribbean by socioeconomic measures and the shifting 7: e1644–54 obesity burden over time. This online publication has been corrected. The corrected version first appeared at Methods We did a cross-sectional series analysis of obesity prevalence by socioeconomic status by use of national thelancet.com/lancetgh on health surveys done between 1998 and 2017 in 13 countries in Latin America and the Caribbean. We generated Jan 24, 2020 equiplots to display inequalities in, the primary outcome, obesity by wealth, education, and residence area. We See Comment page e1589 measured obesity gaps as the difference in percentage points between the highest and lowest obesity prevalence within For the French translation of each socioeconomic measure, and described trends as well as changing patterns of the obesity burden over time. the abstract see Online for appendix 1 Findings 479 809 adult men and women were included in the analysis. Obesity prevalence across countries has For the Portuguese translation of increased over time, with distinct patterns emerging by wealth and education indices. In the most recent available the abstract see Online for appendix 2 surveys, obesity was most prevalent among women in Mexico in 2016, and the least prevalent among women in Haiti For the Spanish translation of in 2016. The largest gap between the highest and lowest obesity estimates by wealth was observed in Honduras the abstract see Online for among women (21·6 percentage point gap), and in Peru among men (22·4 percentage point gap), compared with a appendix 3 3·7 percentage point gap among women in Brazil and 3·3 percentage points among men in Argentina. Urban CRONICAS Center of Excellence residents consistently had a larger burden than their rural counterparts in most countries, with obesity gaps ranging in Chronic Diseases, Universidad from 0·1 percentage points among women in Paraguay to 15·8 percentage points among men in Peru. The trend Peruana Cayetano Heredia, Lima, Peru (S S Jiwani MSPH, analysis done in five countries suggests a shifting of the obesity burden across socioeconomic groups and different R M Carrillo-Larco MD, patterns by gender. Obesity gaps by education in Mexico have reduced over time among women, but increased among A Hernández-Vásquez MD, men, whereas the gap has increased among women but remains relatively constant among men in Argentina. Prof J J Miranda PhD); Department of International Health, Johns Hopkins Interpretation The increase in obesity prevalence in the Latin American and Caribbean region has been paralleled Bloomberg School of Public with an unequal distribution and a shifting burden across socioeconomic groups. Anticipation of the establishment Health, Baltimore, MD, USA of obesity among low socioeconomic groups could provide opportunities for societal gains in primordial prevention. (S S Jiwani); Department of Epidemiology and Biostatistics, School of Public Health, Funding None. Imperial College London, London, UK (R M Carrillo-Larco); Copyright © 2019 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Center for Population Health Research, National Institute of Introduction Public Health, Cuernavaca, and its transition between socioeconomic status groups Mexico The characterisation of the association between obesity has been proposed on the basis of national data from (T Barrientos-Gutiérrez PhD, and elevated risks of chronic conditions, such as 30 countries between 1975 and 2015.7 From a societal A Basto-Abreu DrPH); South diabetes, heart disease, and some cancers, and all-cause perspective, once the burden of obesity shifts to the most American Center of Excellence for Cardiovascular Health, mortality has been well researched.1 The prevalence of socioeconomically disadvantaged groups, it adds major Institute for Clinical obesity has dramatically increased globally in the past challenges to other coexisting health and societal Effectiveness and Health Policy, two decades,2 owing to the nutrition transition and conditions and to the possibility of reverting to a non- Buenos Aires, Argentina changes in dietary patterns, lifestyle, physical activity, obesity status. Such anticipation, or precision public (L Gutierrez BSc, V Irazola MD); South Florida Veterans Affairs and economic access.3,4 health, requires an understanding of context and trends. Foundation for Research and Although obesity has long been considered a condition Latin America and the Caribbean has the largest Education and Geriatric of the elite and a mark of wealth, published literature in income inequalities globally8 and has had an alarming Research, Education and Clinical the past decade suggests that it can no longer be increase in the prevalence of obesity since the 1990s, in Center, Miami VA Healthcare System, Miami, FL, USA attributed to higher socioeconomic status.4,5 The burden parallel with rapidly growing urbanisation and economic (R Nieto-Martínez MD); of obesity is not static over time and the magnitude of growth.4,9,10 By use of data from Mexico, Brazil, and Foundation for Clinical, Public such estimates are not necessarily the same across Colombia, one review7 suggests that Latin American Health, and Epidemiology socioeconomic groups or across countries.6 A four-stage countries are in stage 2 of the obesity transition, in which Research in Venezuela, Caracas, Venezuela (R Nieto-Martínez); framework to approximate the epidemiology of obesity obesity prevalence has increased among the lower www.thelancet.com/lancetgh Vol 7 December 2019 e1644 Articles Postgraduate Program of Nursing, Federal University of Research in context Pelotas, Rio Grande do Sul, Brazil (B P Nunes PhD); Program Evidence before this study (stage 4). Nevertheless, given that only three countries of the of Physical Therapy and Obesity has long been believed to affect the elite; however, Latin American and Caribbean region were represented in this Department of Surgery, research in the past decade suggests a rapid shift of the burden study, the evidence on the differential burden of obesity Institute for Public Health, of obesity towards lower socioeconomic groups. The prevalence between high and low socioeconomic groups in this region Washington University in St Louis School of Medicine, of obesity has been increasing in countries of the Latin America remains unclear. St Louis, MO, USA and Caribbean region since the 1990s. We did a PubMed (D C Parra PhD); and School of Added value of this study literature search for articles on adult obesity prevalence and Medicine, Universidad Peruana The published literature on this issue remains outdated with Cayetano Heredia, Lima, Peru trends in Latin America and the Caribbean published between the majority of studies covering a period in the early 2000s. (Prof J J Miranda) Jan 1, 2010, and April 1, 2019, with “adult obesity“ and “Latin In this analysis, we provide an update on the current Correspondence to: America“ or “South America“ or “Carribean“ in the title. On the distribution of the obesity burden across socioeconomic status Ms Safia S Jiwani, Department basis of title review we identified few articles that matched our of International Health, in Latin America and the Caribbean and the changing burden search criteria: the majority of articles were clinical or Johns Hopkins Bloomberg over time. In particular, our findings point to the shifting experimental in nature. Jaacks and colleagues explored the School of Public Health, patterns of the obesity burden across gender and Baltimore, MD 21205, USA epidemiology of obesity between 1975 and 2015 in socioeconomic status in the Latin American and Caribbean sjiwani1@jhu.edu 30 countries, representing more than 75% of the world’s region. Our findings serve as a call to action for tailored, population, including Mexico, Colombia, and Brazil. A four- equity-focused programmes and policies. stage obesity transition model was proposed, in which the obesity burden concentrated among women and higher Implications of all the available evidence socioeconomic status groups (particularly for women; stage 1) Latin America and the Caribbean is the region with the largest shifts towards the more disadvantaged, narrowing the gap income inequalities globally. These inequalities, coupled with between sexes and between socioeconomic status groups rapid urbanisation and economic growth, increase the risk of among women (stage 2), until a reversal of the burden occurs growing obesity rates. Therefore, up-to-date information on where obesity prevalence among lower socioeconomic status the magnitude of the problem by various socioeconomic surpasses that of the higher socioeconomic status groups measures can help guide and target prevention efforts. (stage 3), after which declines in obesity would be expected socioeconomic groups, and the gap by socioeconomic indicators. When Demographic and Health Surveys status groups has narrowed. The published literature on were not available, we used data from other nationally- this issue needs to be updated, with the majority of representative health surveys: Argentina’s National studies covering a period in the early 2000s, and lacks Survey for Risk Factors, Brazil’s National Health Survey, information about obesity inequalities, particularly in the Colombia’s National Survey on Nutritional Status, Latin American and Caribbean region. Mexico’s National Health and Nutrition Survey, We aimed to describe the obesity distribution and Paraguay’s Non-communicable Disease Risk Factor obesity gap by socioeconomic measures among adult survey, and the Venezuelan Cardiometabolic Health men and women in 13 countries in Latin America and Study. Each survey had a distinct sampling design as See Online for appendix 4 the Caribbean and evaluate the changing trend of obesity outlined in the appendix 4 (pp 2–3). distribution and gap by socioeconomic measures over The study population included individuals aged at least time in five countries with available data. 18 years with available data on obesity. We excluded pregnant women from the analysis in all countries, Methods except in Argentina where pregnancy status was not Study design and participants recorded in the dataset. Obesity data from Demographic We did a cross-sectional series analysis of obesity and Health Surveys covered women aged 18–49 years, prevalence by socioeconomic status by use of national whereas data from other surveys covered women and health surveys done in 13 countries in Latin America men (if available) aged 18 years or older. and the Caribbean. We used nationally representative All surveys used for analysis included de-identified health surveys done between 1998 and 2017 that data. Ethical approval was not sought for this analysis of included obesity and socioeconomic variables. Publicly secondary data. All surveys except Brazil’s National available Demographic and Health Survey datasets were Health Survey, Paraguay’s Non-communicable Disease retrieved for Bolivia, Dominican Republic, Guatemala, Risk Factor survey, and the Venezuelan Cardiometabolic Haiti, Honduras, Nicaragua, and Peru. Demographic Health Study were publicly accessible. and Health Surveys are nationally representative household surveys implemented in more than 90 low- Procedures income and middle-income countries that provide For the analysis on obesity patterns and gaps by information on standard global health and population socioeconomic status, we did a cross-sectional analysis e1645 www.thelancet.com/lancetgh Vol 7 December 2019 Articles using the latest national health surveys in 13 countries: Statistical analysis Argentina, Bolivia, Brazil, Colombia, Dominican Republic, All 13 countries were included in the analysis of obesity Guatemala, Haiti, Honduras, Mexico, Nicaragua, Paraguay, gaps, whereas only five countries that had three Peru, and Venezuela. For the trend analysis on obesity consecutive surveys, at least 4 years apart, were included gaps, we did a cross-sectional series analysis and included in the trend analysis of obesity gaps. We defined the five countries (Argentina, Bolivia, Haiti, Mexico, and gap in obesity prevalence as the absolute difference Peru) that had three consecutive surveys at least 4 years in percentage points between the highest and lowest apart, with the most recent published after 2000. most extreme obesity prevalence estimates within The primary outcome, obesity, was defined as having each socioeconomic status measure. For instance, if the a body-mass index (BMI) of 30 kg/m² or more. Measured highest obesity prevalence by wealth was observed or reported weight and height variables were used to among the third quintile, and the lowest among the fifth compute BMI. The three socioeconomic status quintile, the obesity gap by wealth was calculated as the measures of interest were wealth index (Q1–Q5, where arithmetic difference between the obesity estimate in Q1 is the poorest quintile and Q5 is the richest), the third quintile and that in the fifth quintile. For the education index (E1–E5, where E1 is the least educated five countries with available data, we assessed the trends quintile and E5 is the most educated), and area of in obesity gaps by socioeconomic status over three residence (rural or urban). For countries with available timepoints. For the most recent surveys, we also Demographic and Health Surveys, we used the reported the regional mean obesity prevalence within wealth quintiles existing in the datasets, generated by each socioe conomic status measure, computed as the principal component analysis, which includes house- arithmetic average of all countries’ estimate within each hold ownership of assets, materials used for household quintile. construction, and access to water and sanitation We ran separate stratified analyses by gender for each facilities. Wealth index in Mexico was previously country and survey. We calculated and reported the age- constructed using household characteristics (eg, number standardised obesity prevalence by each of the three of rooms, exclusive kitchen, bathroom, and type of fuel) socioeconomic status measures (wealth, education, and and household assets (eg, television, microwave, and area of residence) using the WHO standard population computer), through principal component analysis; age distribution.12 similarly, a wealth index based on a sum of asset All analyses and graphs were conducted on Stata ownership was developed in Brazil. For other national version 15. We used the svy command to account for health surveys without existing wealth quintiles by complex survey sampling designs and the sampling principal component analysis, we computed wealth weights for all countries’ surveys. We generated equiplots quintiles using alternative measures of wealth—eg, we to display inequalities in obesity by socioeconomic status used average monthly household income in Argentina using the equiplot.ado file. For the equiplot.ado file see and Paraguay. In all surveys, the wealth index was https://www.equidade.org/ computed at the household level; therefore individuals Role of the funding source equiplot residing in the same household belonged to the same There was no funding source for this study. wealth index category. The Venezuelan Cardiometabolic Health Study 2014–17 did not include any wealth Results variables; therefore we did not estimate obesity Data from 23 health surveys were used for this analysis, prevalence by wealth for this survey. The education of which 13 were Demographic and Health Surveys. index was generated into quintiles using the total A total of 157 741 adult men and 322 068 adult women number of formal years of education, as reported in aged at least 18 years were included in the analysis: Demographic and Health Surveys. When a continuous 282 247 men and women were included in analysis of education variable was not available, we used an existing the most recent obesity prevalence in the region, and ascending categorical measure of education specified in 275 191 were included in the trend analysis of the change the survey, such as in Argentina, Mexico, and Venezuela in prevalence over time. (no education and primary, secondary, and higher The most recent data available for Latin America and education). Area of residence was defined as rural or the Caribbean corresponded to 2001–17, and the age- urban in all countries, except for Argentina’s National standardised obesity prevalence among adults varied Survey for Risk Factors, which only sampled urban greatly within the region (figure; tables 1, 2). Overall, the populations given that 91% of the Argentinian highest obesity prevalence was found among the fourth population reside in urban areas according to the 2010 richest quintile (26·1%), third education quintile (27·1%), census.11 In the case of Argentina, the obesity estimates and urban (26·0%) women (table 1), whereas among computed therefore reflect the prevalence among urban men, the highest burden was in the richest quintile populations. The data sources and socioeconomic status (24·5%), fourth education quintile (24·2%), and urban measure definitions used in each survey are summarised residents (22·0%; table 2). Mexico had the highest in the appendix 4 (pp 2–3). obesity prevalence by all three socioeconomic measures www.thelancet.com/lancetgh Vol 7 December 2019 e1646 Articles among men and women, whereas Haiti had the lowest the obesity prevalence was lowest among the highest obesity prevalence by wealth index among women and wealth and education quintiles (table 1). Colombia among men (table 1, 2). Among women in Bolivia, Peru, Mexico, and Colombia, The obesity prevalence varied by socioeconomic status the prevalence was highest in the third or fourth wealth measure and by country. Among women in Argentina, and third education quintiles. For example, among Brazil, Dominican Republic, Venezuela, and Paraguay, women in Mexico in 2016, the third wealth quintile had a the burden was concentrated among the poor and least 43·2% (95% CI 39·0–47·4) obesity prevalence compared educated, particularly in Argentina in 2013, where 23·2% with 37·2% (33·2–41·2) in the highest wealth quintile (95% CI 21·0–25·5) of the first wealth quintile and (table 1). Among men in Mexico, the prevalence was the 28·5% (24·3–32·6) of the first education quintile were highest in the fourth wealth and education quintiles obese compared with 13·1% (10·4–15·7) of the fifth (table 2). wealth quintile and 13·0% (11·6–14·3) of the fifth In all countries except Venezuela, and Argentina where education quintile (table 1). Among men in Brazil, the comparison was not possible, the most recent surveys Colombia, and Paraguay, the richest and most educated indicate that urban men had a higher obesity prevalence quintiles had a higher obesity prevalence compared with compared with their rural counterparts (table 2). This lower wealth and education quintiles (table 2). The finding was consistent among women in Bolivia, pattern was similar among women in Guatemala and Guatemala, Haiti, Honduras, Nicaragua, and Peru, and Haiti and reversed among women in Colombia, where in the remaining countries albeit with overlapping CIs A Wealth index Education index Area of residence Q1 Q2 Q3 Q4 Q5 E1 E2 E3 E4 E5 Rural Urban Argentina (2013) Bolivia (2008) Brazil (2013) Colombia (2010) Dominican Republic (2013) Guatemala (2015) Haiti (2016) Honduras (2012) Mexico (2016) Nicaragua (2001) Paraguay (2011) Peru (2017) Venezuela (2014–17) Regional mean B Argentina (2013) Brazil (2013) Colombia (2010) Mexico (2016) Paraguay (2011) Peru (2017) Venezuela (2014–17) Regional mean 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 Obesity prevalence (%) Obesity prevalence (%) Obesity prevalence (%) Figure: Most recent obesity prevalence by wealth, education, and residence (A) Among women. (B) Among men. Q=wealth quintile. E=education quintile. e1647 www.thelancet.com/lancetgh Vol 7 December 2019 Articles www.thelancet.com/lancetgh Vol 7 December 2019 e1648 Sample Wealth index Education index Area of residence size, N* Q1 Q2 Q3 Q4 Q5 Gap,† E1 E2 E3 E4 E5 Gap,† Rural Urban Gap,† (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) percent- (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) percent- (95% CI) (95% CI) percent- age age age points points points Argentina 2005 21 037 18·2% 16·8% 16·0% 12·2% 10·6% 7·6 21·6% 16·9% 15·6% 11·8% 9·6% 11·9 ·· 13·8% ·· (15·7–20·7) (14·7–18·9) (13·9–18·1) (10·2–14·1) (8·9–12·3) (18·2–24·9) (14·7–19·0) (13·3–17·9) (10·0–13·7) (8·2–11·1) (12·8–14·8) 2009 18 095 22·2% 19·4% 15·8% 16·2% 13·4% 8·8 29·1% 23·0% 19·0% 14·7% 10·2% 18·9 ·· 16·8% ·· (20·2–24·2) (14·9–23·9) (13·6–18·1) (13·4–19·0) (10·9–15·9) (24·1–34·0) (18·3–27·7) (16·1–21·9) (13·0–16·4) (7·5–12·8) (14·2–19·4) 2013 16 664 23·2% 21·3% 20·3% 15·3% 13·1% 10·2 28·5% 25·9% 22·4% 16·8% 13·0% 15·5 ·· 18·5% ·· (21·0–25·5) (18·1–24·5) (17·8–22·8) (14·1–16·5) (10·4–15·7) (24·3–32·6) (23·8–28·0) (19·4–25·3) (14·7–18·9) (11·6–14·3) (16·5–20·4) Bolivia 1998 4125 5·5% 10·1% 15·9% 19·3% 14·4% 13·9 9·4% 17·3% 15·7% 13·9% 12·5% 7·9 6·6% 16·1% 9·4 (3·6–7·3) (7·5–12·6) (12·9–19·0) (15·2–23·5) (10·9–18·0) (7·2–11·6) (14·0–20·7) (12·1–19·3) (10·6–17·1) (9·3–15·7) (5·1–8·2) (13·6–18·5) 2003 12 347 7·1% 14·9% 21·2% 22·5% 21·1% 15·5 15·2% 21·6% 24·2% 20·9% 13·5% 10·6 11·4% 21·5% 10·1 (5·6–8·5) (12·8–17·0) (19·0–23·4) (20·5–24·5) (19·3–22·9) (13·3–17·0) (19·6–23·6) (21·3–27·0) (18·8–23·0) (11·7–15·4) (9·7–13·0) (20·3–22·7) 2008 13 497 9·0% 18·3% 24·8% 27·9% 20·1% 19·0 20·4% 22·9% 24·9% 15·3% 12·6% 12·3 15·7% 23·2% 7·5 (7·5–10·4) (16·4–20·2) (22·8–26·8) (25·9–29·9) (18·3–21·9) (18·6–22·1) (20·8–25·1) (23·2–26·6) (10·1–20·6) (10·9–14·2) (14·3–17·1) (22·1–24·3) Brazil, 2013 33 482 24·4% 25·4% 25·5% 23·9% 21·8% 3·7 28·2% 27·0% 25·7% 23·4% 19·2% 9·0 23·8% 24·0% 0·2 (22·1–26·6) (23·8–27·0) (23·9–27·0) (22·4–25·5) (20·0–23·6) (25·6–30·9) (25·4–28·7) (23·8–27·6) (21·9–24·8) (17·2–21·1) (22·1–25·6) (23·1–24·8) Colombia, 47 272 18·3% 22·4 % 22·2% 20·9% 17·3% 5·1 20·8% 20·7% 21·8% 19·2% 5·6% 16·2 19·5% 20·5% 1·0 2010 (17·3–19·4) (21·2–23·6) (21·1–23·4) (19·6–22·1) (16·1–18·5) (19·9–21·8) (19·7–21·7) (20·1–23·6) (18·2–20·1) (5·4–5·8) (18·4–20·5) (19·8–21·1) Dominican 7655 21·6% 24·0% 24·7% 23·5% 20·9% 3·8 25·7% 23·9% 22·7% 26·3% 20·8% 5·5 21·9% 23·7% 1·9 Republic, (19·0–24·1) (21·1–27·0) (22·1–27·3) (20·8–26·2) (18·5–23·4) (23·5–27·9) (20·9–26·9) (20·3–25·1) (21·1–31·6) (18·2–23·4) (19·7–24·1) (22·3–25·2) 2013 Guatemala, 14 256 13·0% 17·4% 26·4% 29·3% 30·7% 17·7 19·9% 24·9% 27·1% 29·1% 26·4% 9·2 20·0% 29·0% 9·0 2015 (11·3–14·8) (15·6–19·2) (24·4–28·4) (27·2–31·4) (28·7–32·7) (18·1–21·6) (22·8–27·0) (24·7–29·5) (26·9–31·3) (23·8–28·9) (18·6–21·3) (27·5–30·5) Haiti 2006 2528 0·7% 2·4% 6·6% 11·3% 19·2% 18·5 5·0% 6·6% 10·1% 18·4% 16·6% 13·3 6·0% 14·7% 8·7 (-0·2–1·5) (0·6–4·1) (3·7–9·6) (7·9–14·7) (15·1–23·4) (2·8–7·2) (4·2–9·1) (6·4–13·7) (13·9–22·8) (13·3–19·9) (3·9–8·0) (11·6–17·7) 2012 7462 1·5% 4·9 % 6·7% 12·1% 19·3% 17·8 4·9% 9·2% 12·8% 16·5% 15·4% 11·7 7·0% 14·2% 7·2 (0·8–2·2) (3·4–6·3) (5·1–8·4) (10·0–14·2) (16·7–21·9) (3·7–6·1) (7·4–11·0) (9·7–15·9) (13·4–19·7) (12·5–18·4) (5·7–8·3) (12·4–16·0) 2016 7667 3·8% 5·6% 12·6% 17·5% 24·2% 20·4 8·0% 11·7% 15·5% 18·6% 19·3% 11·3 10·6% 18·5% 7·9 (2·6–5·0) (4·4–6·8) (10·6–14·7) (15·4–19·6) (21·6–26·9) (5·9–10·1) (9·8–13·6) (13·0–18·1) (15·5–21·7) (16·3–22·2) (9·1–12·0) (16·7–20·3) Honduras, 10 061 13·5% 20·4% 30·5% 35·1% 31·0% 21·6 22·1% 29·7% 37·0% 29·0% 27·4% 14·9 24·0% 31·4% 7·4 2012 (11·7–15·4) (18·4–22·5) (28·0–32·9) (32·0–38·3) (28·3–33·6) (19·9–24·2) (27·9–31·6) (29·7–44·2) (26·3–31·6) (23·8–31·0) (21·9–26·0) (29·6–33·2) Mexico 2006 21 334 32·2% 34·1% 36·2% 31·4% 31·1% 5·1 30·9% 36·3% 35·9% 29·7% 25·3% 11·0 29·8% 34·2% 4·4 (30·6–33·8) (32·2–36·1) (34·0–38·4) (28·8–34·0) (27·6–34·7) (27·6–34·2) (34·7–37·9) (33·6–38·2) (27·3–32·2) (22·5–28·1) (27·8–31·8) (32·9–35·5) 2012 23 707 30·7% 37·5% 39·0% 39·0% 35·3% 8·3 33·4% 39·7% 38·6% 32·3% 30·8% 8·9 32·7% 37·2% 4·4 (28·8–32·6) (35·6–39·5) (37·1–40·8) (36·8–41·2) (33·2–37·3) (29·9–36·9) (38·0–41·4) (36·6–40·5) (30·1–34·6) (28·2–33·4) (31·1–34·2) (36·0–38·4) 2016 5889 35·1% 39·4% 43·2% 36·8% 37·2% 8·1 40·4% 40·6% 40·8% 34·7% 32·1% 8·7 36·2% 38·5% 2·3 (31·1–39·1) (35·8–42·9) (39·0–47·4) (33·6–40·1) (33·2–41·2) (35·4–45·4) (37·4–43·8) (37·4–44·3) (30·8–38·6) (28·1–36·2) (33·4–39·0) (36·0–41·0) (Table 1 continues on next page) Articles (table 1). The largest obesity prevalence was observed in Mexico in 2016, with 38·5% (95% CI 36·0–41·0) among urban women, 36·2% (33·4–39·0) among rural women, 28·6% (25·1–32·1) among urban men, and 21·8% (18·8–24·8) among rural men (tables 1, 2). Multiple obesity patterns emerge by socioeconomic status (figure; table 1, 2): Bolivia, Guatemala, Haiti, Honduras, Peru, and Nicaragua had large inequalities in the distribution of obesity by wealth and education index; the widest obesity gaps among women were observed in Honduras, with a 21·6 percentage point difference in obesity prevalence by wealth, and Haiti, with a 20·4 percentage point difference, with the largest prevalence concentrated in the fourth wealth quintile in Honduras and the fifth in Haiti and the lowest prevalence among the poorest; and among men in Peru there was a 22·4 percentage point gap by wealth index between the highest prevalence among the fourth richest quintile and the lowest prevalence among the poorest quintile. Bolivia in 2008 had a bottom inequality pattern by wealth among women, in which large inequalities existed between the first and second poorest quintiles (9·0% [95% CI 7·5–10·4] vs 18·3% [16·4–20·2]), with smaller differences between subsequent quintiles (table 1). In Colombia, Brazil, Mexico, and the Dominican Republic, the prevalence of obesity among women was similar in all wealth and education quintiles (table 1); this was also true for men in Argentina, where the gap in obesity prevalence between the first and second wealth quintiles was 3·3 percentage points (table 2). By area of residence, the largest obesity gaps were observed in Peru, with an 11·7 percentage point gap between urban and rural women and a 15·8 percentage point gap between urban and rural men (table 1, 2). The smallest obesity gap by area of residence was in Paraguay, with a 0·1 percentage point gap between urban and rural women (table 1). The differential effect of socioeconomic status on obesity by gender was further confirmed in post-hoc analysis (appendix 4 p 4), in which gender modified the association between wealth index and obesity (in Argentina, Colombia, Paraguay, and Peru) and between education and obesity (in Colombia, Mexico, Paraguay, and Peru). The trend analysis indicated that the prevalence and gap in obesity among women increased between 2005 and 2013 in Argentina (table 1; appendix 4 p 5): the obesity gap increased from 7·6 percentage points by wealth and 11·9 percentage points by education in 2005 to 10·2 percentage points and 15·5 percentage points in 2013. Although the obesity prevalence increased across all socioeconomic status groups among men between 2005 and 2013, the gap by socioeconomic status has remained relatively constant over time (appendix 4 p 6). In 2016, Mexico had the highest obesity prevalence among men and women in the Latin American and Caribbean region, with women bearing a higher prevalence compared with men across socioeconomic status measures. Among women, the obesity prevalence e1649 www.thelancet.com/lancetgh Vol 7 December 2019 Sample Wealth index Education index Area of residence size, N* Q1 Q2 Q3 Q4 Q5 Gap,† E1 E2 E3 E4 E5 Gap,† Rural Urban Gap,† (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) percent- (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) percent- (95% CI) (95% CI) percent- age age age points points points (Continued from previous page) Nicaragua, 10 186 9·0% 16·6% 24·3% 28·2% 24·7% 19·2 17·0% 23·5% 25·1% 26·2% 19·6% 9·1 16·2% 25·0% 8·8 2001 (7·5–10·5) (14·6–18·6) (22·0–26·5) (26·0–30·4) (22·0–27·4) (15·0–19·0) (21·3–25·6) (22·8–27·4) (23·1–29·3) (17·2–22·0) (14·6–17·9) (23·4–26·5) Paraguay, 1509 24·0% 34·2% 25·1% 22·8% 31·7% 11·4 30·8% 30·3% 28·3% 33·8% 18·2% 15·6 28·3% 28·3% 0·1 2011 (19·7–28·4) (29·9–38·5) (20·1–30·2) (18·3–27·3) (26·1–37·2) (26·8–34·7) (26·1–34·6) (24·2–32·4) (28·7–38·8) (14·3–22·1) (24·9–31·7) (25·5–31·2) Peru 2005 4911 2·1% 10·7% 19·1% 17·6% 15·7% 17·0 11·3% 18·1% 18·4% 13·4% 12·3% 7·1 8·6% 17·0% 8·4 (1·0–3·2) (7·9–13·5) (15·7–22·5) (14·5–20·6) (13·1–18·4) (8·0–14·5) (14·4–21·7) (15·6–21·1) (8·9–17·9) (10·1–14·5) (5·7–11·5) (15·1–18·8) 2010 18 837 8·5% 15·2% 21·1% 21·8% 17·9% 13·4 16·5% 20·4% 20·6% 15·0% 14·6% 6·0 10·8% 19·9% 9·1 (7·4–9·5) (13·8–16·7) (19·5–22·7) (20·0–23·6) (16·1–19·8) (14·6–18·4) (18·6–22·1) (18·9–22·2) (11·8–18·2) (13·0–16·2) (9·9–11·8) (18·9–20·9) 2017 17 210 16·4% 28·8% 31·7 32·4% 27·3% 16·0 25·1% 29·8% 31·3% 26·2% 24·1% 7·2 18·4% 30·1% 11·7 (15·0–17·8) (27·0–30·7) (29·5–33·9) (30·1–34·8) (24·7–30·0) (23·2–27·1) (27·4–32·1) (29·0–33·6) (23·8–28·7) (21·5–26·7) (17·1–19·8) (28·8–31·5) Venezuela, 2337 ·· ·· ·· ·· ·· ·· 20·1% 38·2% 29·3% 23·2% NA 18·1 25·5% 27·3% 1·8 2014–17 (18·9–21·3) (32·8–43·6) (23·5–35·1) (17·8–28·5) (19·0–32·1) (23·2–31·4) Regional ·· 17·6% 22·8% 25·9% 26·1% 25·0% 8·5 23·6% 26·8% 27·1% 24·7% 19·8% 7·3 21·7% 26·0% 4·3 mean‡ NA=not applicable. *Unweighted. †Gap defined as the difference between the highest and lowest obesity prevalence within the socioeconomic status measure. ‡Regional mean is the arithmetic average of country estimates for the most recent surveys. Table 1: Age-standardised obesity prevalence among women by country and year Articles has been increasing over time within each socioeconomic status measure (appendix 4 pp 7–8). The gap between the quintiles with the highest and lowest prevalence has increased slightly by wealth index among women (5·1 percentage points in 2006 compared with 8·3 percentage points in 2012 and 8·1 percentage points in 2016), although the prevalence remains highest among the third wealth quintile and lowest among the first (table 1; appendix 4 p 7). The gap has decreased by education and area of residence (4·4 percentage points in 2006 to 2·3 percentage points in 2016 by area of residence), with the highest prevalence remaining among the second and third education quintiles, and among urban residents (table 1; appendix 4 p 7). Among men, however, the obesity prevalence is different (appendix 4 p 8): in the three poorest quintiles, obesity has decreased between 2006 and 2016, and it has increased across all education quintiles in the same period, with the largest increase occurring in the first education quintile (18·2% in 2006 vs 28·2% in 2016; table 2). Similarly, obesity gaps have widened over time by wealth (9·5 percentage points vs 14·5 percentage points) and by education (8·2 percentage points vs 10·8 percentage points) between 2006 and 2016 but narrowed by residence (8·4 percentage points vs 6·8 percentage points; table 2; appendix 4 p 8). In Peru, the prevalence of obesity among women increased in each group across all three socioeconomic status measures (appendix 4 p 9). The obesity gap by wealth reduced from 17·0 percentage points in 2005 to 13·4 percentage points in 2010, increasing to 16·0 percentage points in 2017, with substantial increases in the obesity prevalence among the poorest women (2·1% [95% CI 1·0–3·2] in 2005 to 16·4% [15·0–17·8] in 2017; table 1; appendix 4 p 9). However, the burden remains concentrated among the third and fourth wealth index quintiles. In terms of education, the gap in obesity prevalence between extreme quintiles has not varied substantially over time, but it has shifted; although the third education quintile retained the highest obesity prevalence between 2005 and 2017, the prevalence in the first quintile increased from 11·3% (95% CI 8·0–14·5) in 2005 to 25·1% (23·2–27·1) in 2017 (table 1; appendix 4 p 9). Both urban and rural women had an increasing obesity prevalence over time, with larger increases among urban residents (17·0% [95% CI 15·1–18·8] in 2005 to 30·1% [28·8–31·5] in 2017; table 1; appendix 4 p 9). Patterns in obesity in Haiti differ greatly from the rest of the region between 2006 and 2016 (appendix 4 p 10): the rich, more educated, and urban women had the highest obesity prevalence. The prevalence among each wealth quintile and education quintile increased most between 2012 and 2016 (appendix 4 p 10). Although the overall obesity gap by wealth increased in magnitude between 2006 and 2016, it narrowed by education index from 13·3 percentage points to 11·3 percentage points, with the highest prevalence remaining among the richest www.thelancet.com/lancetgh Vol 7 December 2019 e1650 Sample Wealth index Education index Area of residence size, N* Q1 Q2 Q3 Q4 Q5 Gap,† E1 E2 E3 E4 E5 Gap,† Rural Urban Gap,† (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) percent- (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) percent- (95% CI) (95% CI) percent- age age age points points points Argentina 2005 16 918 15·0% 16·5% 15·3% 17·7% 14·5% 3·2 19·8% 17·2% 17·8% 15·2% 12·3% 7·5 ·· 15·6% ·· (12·5–17·5) (13·9–19·0) (13·3–17·3) (15·2–20·2) (12·7–16·4) (16·8–22·7) (15·0–19·5) (15·4–20·3) (13·1–17·5) (10·6–13·9) (14·4–16·7) 2009 14 353 17·3% 21·1% 21·2% 20·4% 15·8% 5·4 19·1% 20·2% 22·3% 19·1% 15·6% 6·7 ·· 19·3% ·· (14·8–19·8) (19·4–22·9) (18·1–24·3) (18·6–22·2) (13·4–18·1) (16·5–21·6) (18·3–22·2) (19·2–25·4) (18·0–20·3) (14·3–17·0) (18·3–20·3) 2013 13 626 21·7% 25·0% 22·4% 23·9% 22·6% 3·3 24·4% 25·6% 25·1% 23·9% 18·4% 7·2 ·· 23·2% ·· (19·6–23·8) (21·9–28·0) (20·1–24·7) (20·7–27·1) (20·0–25·1) (20·3–28·6) (23·8–27·4) (22·0–28·2) (21·7–26·2) (15·9–20·9) (21·7–24·7) Brazil, 2013 25 920 10·2% 12·3% 16·9% 21·1% 21·7% 11·5 14·5% 14·0% 17·0% 18·9% 22·0% 8·0 11·9% 17·5% 5·6 (8·8–11·6) (11·0–13·6) (15·3–18·5) (19·0–23·1) (19·8–23·5) (12·5–16·4) (12·8–15·2) (15·1–18·8) (17·4–20·5) (20·1–24·0) (10·5–13·4) (16·5–18·4) Colombia, 36 544 5·5% 10·9% 12·5% 14·6% 15·9% 10·4 10·7% 11·1% 13·0% 14·3% 20·3% (NA) 9·6 7·0% 13·4% 6·4 2010 (4·8–6·2) (9·9–11·8) (11·5–13·5) (13·4–15·8) (14·6–17·2) (9·8–11·6) (10·2–12·0) (11·4–14·5) (13·3–15·2) (6·3–7·8) (12·8–14·0) Mexico 2006 14 394 19·7% 21·4% 29·2% 28·0% 25·6% 9·5 18·2% 23·8% 26·4% 25·7% 25·6% 8·2 17·1% 25·4% 8·4 (18·1–21·4) (18·9–23·2) (26·8–31·6) (25·2–30·9) (22·0–29·3) (15·1–21·3) (21·8–25·7) (24·1–28·7) (22·7–28·6) (22·7–28·4) (15·0–19·1) (24·1–26·8) 2012 17 514 17·4% 24·6% 26·5% 28·8% 30·4% 13·0 19·3% 23·5% 27·7% 29·1% 29·8% 10·6 19·7% 28·1% 8·4 (15·7–19·1) (22·6–26·6) (24·5–28·6) (26·4–31·2) (28·1–32·7) (15·9–22·7) (21·9–25·2) (25·6–29·7) (26·5–31·6) (27·2–32·5) (18·3–21·1) (26·8–29·4) 2016 3076 17·2% 19·2% 27·2% 31·7% 30·0% 14·5 28·2% 24·7% 26·5% 37·4% 26·9% 10·8 21·8% 28·6% 6·8 (13·8–20·6) (16·4–22·1) (23·2–31·2) (27·3–36·0) (24·9–35·0) (22·5–34·0) (20·8–28·5) (23·1–30·0) (33·1–41·6) (22·1–31·8) (18·8–24·8) (25·1–32·1) (Table 2 continues on next page) Articles groups, shifting from the fourth to the fifth education quintile, and the lowest prevalence remaining among the poorest and least educated (table 1; appendix 4 p 10). Additionally, the prevalence increased among both urban and rural women, shifting slightly towards rural residents, narrowing the obesity gap from 8·7 percentage points in 2006 to 7·9 percentage points in 2016 (table 1; appendix 4 p 10). In Bolivia in 2008, the obesity prevalence was highest among the fourth wealth quintile, the third education quintile, and urban women, and it was lowest among the poorest, the most educated, and the rural residents (appendix 4 p 11). Between 1998 and 2008, the obesity gap widened by wealth (13·9 percentage points to 19·0 percentage points) and by education (7·9 percentage points to 12·3 percentage points), but narrowed by area of residence (9·4 percentage points to 7·5 percentage points), with larger increases in obesity prevalence among rural women (table 1; appendix 4 p 11). Discussion Overall, our age-standardised obesity estimates suggest different obesity patterns across countries in the Latin American and Caribbean region, with the highest prevalence of obesity by socioeconomic status observed among women in Mexico in 2016 and the lowest among women in Haiti in 2016. We identified three distinct patterns for the distribution of obesity across socio- economic status: concentration in the low wealth and education groups (Argentina, women in Venezuela, and women in Mexico by education), concentration in middle wealth and education groups (women in Bolivia, Peru, Mexico by wealth, and Colombia), and concentration among the high-income and high- education groups (women in Guatemala and Haiti and men in Mexico, Brazil, Colombia, Paraguay, Peru, and Venezuela). Moreover, the prevalence of obesity remains consistently higher among urban compared with rural men and women in most countries included in this analysis. However, with the exception of Peru, we found that increases in obesity have been larger among rural populations, which is in line with a global analysis13 showing that obesity among rural populations is increasing at a faster pace than that among urban populations. These patterns also suggest that countries in the Latin American and Caribbean region are in different stages in the transition of obesity as described by Jaacks and colleagues,7 according to socioeconomic groups and gender, thus tailoring of policies is required to adequately tackle the obesity epidemic in Latin America and the Caribbean. In the early 2000s, obesity was believed to be a problem of the elite.4,5 However, evidence suggests a rapidly shifting prevalence towards lower socioeconomic groups, fueling inequalities in developing countries. This shift is believed to be associated with countries’ economic development,14,15 although the evidence remains unclear. e1651 www.thelancet.com/lancetgh Vol 7 December 2019 Sample Wealth index Education index Area of residence size, N* Q1 Q2 Q3 Q4 Q5 Gap,† E1 E2 E3 E4 E5 Gap,† Rural Urban Gap,† (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) percent- (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) percent- (95% CI) (95% CI) percent- age age age points points points (Continued from previous page) Paraguay, 876 12·8% 14·7% 22·9% 26·1% 29·7% 16·9 15·8% 21·9% 20·7% 24·7% 26·8% 11·0 14·4% 25·3% 10·9 2011 (9·4–16·2) (10·7–18·8) (18·4–27·3) (19·9–32·3) (24·6–34·8) (11·6–19·9) (16·6–27·3) (16·0–25·4) (20·1–29·3) (21·9–31·7) (10·6–18·2) (21·8–28·8) Peru 2017 13 466 5·3% 14·6% 23·3% 27·7% 27·4 22·4 10·5% 16·2% 22·6% 23·9% 24·0 13·5 7·2 23·1% 15·8 (4·4–6·1) (13·1–16·2) (20·9–25·7) (24·7–30·6) (24·6–30·1) (8·2–12·7) (13·9–18·4) (20·5–24·6) (21·6–26·1) (21·0–26·9) (6·2–8·2) (21·7–24·4) Venezuela, 1054 ·· ·· ·· ·· ·· ·· ·· 16·2% 22·7% 26·3% NA 10·2 13·8% 23·2% 9·4 2014–17 (13·9–18·4) (16·1–29·4) (16·6–36·1) (9·2–18·4) (15·8–30·6) Regional ·· 12·1% 16·1% 20·9% 24·2% 24·5% 12·4 17·4% 18·5% 21·1% 24·2% 23·1% 6·8 12·7% 22·0% 9·3 mean‡ NA=not applicable. *Unweighted. †Gap defined as the difference between the highest and lowest obesity prevalence within the socioeconomic status measure. ‡Regional mean is the arithmetic average of country estimates for the most recent surveys. Table 2: Age-standardised obesity prevalence among men by country and year Articles Some studies in middle-income and high-income poorest and most educated women being shielded while settings have suggested a reverse gradient, where the those in the middle wealth and education groups have wealthier are more likely to be obese3,5,16 and where the highest prevalence. education protects against the obesogenic wealth effect,17 The third pattern, characterised by a high prevalence of whereas other studies predict that the poor will eventually obesity among the high socioeconomic status groups, fits have a higher burden of chronic conditions, particularly with the first stage of the obesity transition among in lower-income countries, where the prevalence of women, whereby the burden is still concentrated among obesity seems to be shifting to the most disadvantaged the higher socioeconomic status groups and has not groups as the country develops14,15 and the nutrition yet shifted towards the lower socioeconomic status transition unfolds.16,18 The CARMELA study,18 a cross- groups.7 This pattern was clearly observed among women sectional population-based observational study done in in Guatemala, classified as medium in the human seven Latin American cities between April, 2004, and development index,19 and Haiti, classified as low. It was August, 2005, found an inverse relationship between also found among men in Brazil, Colombia, Mexico, and socioeconomic status and obesity in adult women, Peru, which is in line with the proposed second stage of particularly in the higher-income countries. Our results the obesity transition for men.7 among women have now expanded this observation Beyond differences by wealth and education, urban by indicating that, in lower-income settings, such as populations uniformly have a higher obesity prevalence Haiti, Honduras, Nicaragua, and Guatemala, obesity is compared with their rural counterparts, regardless of concentrated among the richer groups for women. gender. However, the prevalence among rural residents However, in middle-income countries, such as Mexico, has increased more rapidly than among urban residents, Colombia, Peru, and Brazil, the prevalence is highest in leading to narrower gaps in obesity prevalence between middle wealth groups among women and in wealthier, the two groups.13,21–23 A cross-sectional analysis of obesity more educated groups among men. prevalence among 147 938 non-pregnant women of The first pattern we observed, in which obesity is reproductive age, using nationally representative data concentrated in the low education and wealth quintiles, is from between 1987 and 2000 in 38 countries, including in line with a review of articles published between nine in Latin America, indicated a scenario where 1989 and 2003 by Monteiro and colleagues,14,15 which obesity was equally distributed among the population in suggests that the prevalence of obesity was shifting more the Latin American countries.9 In contrast, an earlier rapidly towards the lower socioeconomic status groups. analysis4 using survey data from between 1982 and 1996 Argentina, a country with very high human development showed that a third of obese women in the region came index,19 fits this pattern. This result also fits Jaacks and from poor rural areas, indicating a changing obesity colleagues7 obesity transition, with a reversal of the burden burden, which is more in line with our results. Moreover, towards lower socioeconomic status groups (stage 3). changes in policies in the past decade might also have However, the hypothetical stage 4 proposed by Jaacks and affected the shifting burden of obesity in this region. colleagues,7 in which obesity declines among all groups Since 2006, 14 Latin American countries have adopted and the gap in obesity burden across socioeconomic policies to reduce the consumption of sugar-sweetened groups narrows, was not observed in our study. beverages,24 including taxation in Mexico and Brazil.25 The second pattern was characterised by a high obesity However, although the obesity epidemic is multifactorial, prevalence in the third or fourth quintile for wealth, and the effectiveness of such policies in reducing the obesity in the third quintile for education, particularly in burden has not been well established,25,26 nor is a potential women. This pattern was observed in countries with heterogenic effect across socioeconomic status well high or medium human development index,19 such as understood. Evidence suggests that such policies might Peru, Colombia, Mexico, and Bolivia. We hypothesise be most effective in settings with high obesity prevalence that these countries have entered the third stage of and consumption of sugar-sweetened beverages.25 the obesity transition, whereby the prevalence of obesity We also found that obesity prevalence in Latin America is in the process of shifting towards the lower socio- and the Caribbean appears to have distinct patterns by economic status groups, possibly going through the gender. With the exception of Argentina, the prevalence middle socioe conomic status groups first. The two most among men appears to be predominantly concentrated recent surveys in Peru, Bolivia, and Mexico depict a among the wealthier and the more educated groups, similar situation of lowest obesity prevalence among whereas this is not the case for women in the same the least socially advantaged women by wealth, as well countries. Among Argentinian men, the prevalence is as among the most socially advantaged women by concentrated among the third or fourth wealth quintiles, education. This scenario confirms that the pathways by and shifts between the first and second education quintiles which socioe conomic indicators are associated with with the obesity gap remaining relatively constant between health outcomes differ depending on the indicator being 2005 and 2013. Mexico is another example where women used;20 therefore, wealth and education might be bear a larger prevalence of obesity compared with men: operating differently in the obesity epidemic, with the among women, we observed increasing trends and small, www.thelancet.com/lancetgh Vol 7 December 2019 e1652 Articles albeit increasing, obesity gaps by wealth, the prevalence education socioeconomic measures in Latin America being concentrated among middle-income groups, and the Caribbean, whereas urban populations still whereas men had a lower prevalence, concent rated among maintain a larger prevalence than rural populations the richer and more educated, with larger obesity gaps. overall. Our findings also indicate that the prevalence of Our post-hoc statistical analyses confirmed that the obesity is increasing in the region, with larger increases association between socioeconomic status and obesity among rural residents and the most disadvantaged varies by gender. Beyond socioeconomic status, the groups. However, the prevalence of obesity has been differential effect of gender on obesity can be further increasing not only among the poor, least educated, rural explained by physiological and biological factors. Studies populations but also among the rich, highly educated, done in the USA, India, and China27–30 have reported a and urban populations. Among women, the obesity gap larger biological predisposition towards abdominal obesity by wealth, education, and area of residence has stayed and a higher prevalence of metabolic syndrome among constant or widened in Argentina, Bolivia, Peru, and women compared with men. In Peru and Brazil, studies Mexico but has narrowed in Haiti by education and area have found a positive association between parity and of residence. BMI,31,32 and additional factors, such as environmental, Ideally, a situation of low obesity prevalence within each genetic,33 hormonal, and non-hormonal, have also socioeconomic status group and minimal obesity gaps been suggested to differentially affect cardiov ascular would indicate that prevention and action should target ageing mechanisms34 and metabolism33 between men and the entire population. However, our analyses indicate that women. we are far from reaching this goal and that the obesity Our study has several strengths, including the use of epidemic in Latin America and the Caribbean is complex, nationally representative surveys spanning a 20-year with distributions and trends varying across measures of period, and could aid in informing more precise policy socioeconomic status. In other words, wealth, education, responses. It also has some limitations that stem from its or place of residence alone do not capture the full picture cross-sectional design—ie, the trends we observed are of the burden of obesity. To contain this epidemic and its based on estimates computed at specific timepoints and heterogeneous spread, population-wide strategies are are not obtained from individual-level longitudinal data. needed alongside programmes and policies that focus Moreover, we compared obesity prevalence using the preventive interventions by socioeconomic status and by latest available health surveys, and the last survey for gender, advocating a more effective precision public each country might cover a different period and sample health, rather than using a single approach. Adequate and size; this comparison is not ideal, and we ought to frequent monitoring of the obesity epidemic is also keep in mind contextual country-specific factors, such needed in the region. Without updated data sources, as differing periods of economic growth and develop- countries will not be able to prioritise programmes and ment. Changes in obesity might not change drastically policies in the fight against obesity. Anticipation of the in the study periods, allowing a meaningful comparison establishment of obesity among the low socioeconomic across countries. Rather than making inferences com- status groups offers opportunities for societal gains in paring estimates between socioeconomic groups across primordial prevention. These findings can support efforts countries, our analyses aim to descriptively show the towards adequate monitoring of obesity by socioeconomic changing distribution of obesity across socioeconomic status groups that would allow anticipation of the status within countries. Similarly, we used different transitions in obesity across societies and, thus, the measures of wealth and education across countries, formulation of tailored, equity-focused policy responses based on the variables collected, and we are not by any to the burden of obesity in the region. means comparing estimates in specific socioeconomic Contributors status groups between countries.20,35 In Argentina’s JJM conceived the paper. SSJ, RMC-L, and JJM developed the analysis National Survey for Risk Factors, height and weight were plan. SSJ did the analysis and wrote the initial draft. RMC-L, AH-V, self-reported by the respondent, whereas in all other TB-G, AB-A, and BPN contributed to the analysis. All authors reviewed earlier versions of the draft and approved the final manuscript. surveys they were measured; hence obesity prevalence estimates for Argentina may bear recall-bias effects and Declaration of interests We declare no competing interests. lower accuracy. We did not compute absolute inequalities, instead we used equiplots to display the inequalities Acknowledgments RMC-L is supported by a Wellcome Trust International Training observed in the distribution of obesity across socio- Fellowship (214185/Z/18/Z). TB-G acknowledges receiving support from economic status and their directionality. Our sample the Lown Scholars Program at Harvard University. TB-G and AB-A have included a much larger proportion of women than men, received additional support from Bloomberg Philanthropies (principal investigator Juan Rivera-Dommarco). JJM acknowledges having received because the Demographic and Health Surveys mostly support from the Alliance for Health Policy and Systems Research collect height and weight variables for women of (HQHSR1206660), the Bernard Lown Scholars in Cardiovascular Health reproductive age and children. Program at Harvard T.H. Chan School of Public Health (BLSCHP-1902), In conclusion, our analyses suggest great variability in Bloomberg Philanthropies, the National Fund for Scientific and Technological Development through CIENCIACTIVA and the National the age-standardised obesity prevalence by wealth and e1653 www.thelancet.com/lancetgh Vol 7 December 2019 Articles Commission for Scientific and Technological Research, British Council, 15 Monteiro CA, Moura EC, Conde WL, Popkin BM. Socioeconomic British Embassy and the Newton-Paulet Fund (223-2018, 224-2018), status and obesity in adult populations of developing countries: DFID/MRC/Wellcome Global Health Trials (MR/M007405/1), Fogarty a review. Bull World Health Organ 2004; 82: 940–46. International Center (R21TW009982, D71TW010877), Grand Challenges 16 Jones-Smith JC, Gordon-Larsen P, Siddiqi A, Popkin BM. Is the Canada (0335-04), International Development Research Center Canada burden of overweight shifting to the poor across the globe? Time (IDRC 106887, 108167), Inter-American Institute for Global Change trends among women in 39 low- and middle-income countries Research (IAI CRN3036), Medical Research Council (MR/P008984/1, (1991–2008). Int J Obes 2012; 36: 1114–20. MR/P024408/1, MR/P02386X/1), National Cancer Institute 17 Aitsi-Selmi A, Bell R, Shipley MJ, Marmot MG. Education modifies (1P20CA217231), National Heart, Lung and Blood Institute the association of wealth with obesity in women in middle-income (HHSN268200900033C, 5U01HL114180, 1UM1HL134590), National but not low-income countries: an interaction study using seven national datasets, 2005–2010. PLoS One 2014; 9: e90403. Institute of Mental Health (1U19MH098780), Swiss National Science Foundation (40P740-160366), Wellcome Trust (074833/Z/04/Z, 18 Boissonnet C, Schargrodsky H, Pellegrini F, et al. Educational inequalities in obesity, abdominal obesity, and metabolic syndrome 093541/Z/10/Z, 107435/Z/15/Z, 103994/Z/14/Z, 205177/Z/16/Z, in seven Latin American cities: the CARMELA study. 214185/Z/18/Z), and the World Diabetes Foundation (WDF15-1224). Eur J Cardiovasc Prev Rehabil 2011; 18: 550–56. SSJ acknowledges Daniel Antiporta for statistical support. 19 United Nations Development Programme. Human development VI acknowledges the local teams in Marcos Paz, Bariloche, Temuco, and indices and indicators: 2018 statistical update. United Nations Uruguay for making the study possible. RN-M acknowledges the Internal Development Programme, 2018. http://hdr.undp.org/sites/default/ Medicine Venezuelan Society and Venezuelan Cardiometabolic Health files/2018_human_development_statistical_update.pdf (accessed Study Research Group, especially Eunice Ugel for data management and Feb 4, 2019). statistical support. DCP acknowledges the Instituto Colombiano de 20 Howe LD, Galobardes B, Matijasevich A, et al. Measuring Bienestar Familiar and PROFAMILIA for providing support and allowing socio-economic position for epidemiological studies in low- the use of Encuesta Nacional de Demografía y Salud and Colombia’s and middle-income countries: a methods of measurement in National Survey on Nutritional Status datasets and Daniel Ayala for epidemiology paper. Int J Epidemiol 2012; 41: 871–86. invaluable support and input in data management and statistical analysis. 21 Parra DC, Gomez LF, Iannotti L, Haire-Joshu D, Sebert Kuhlmann AK, Brownson RC. Maternal and familial References correlates of anthropometric typologies in the nutrition transition 1 Global BMI Mortality Collaboration, Di Angelantonio E, of Colombia, 2000–2010. Public Health Nutr 2018; 21: 2584–94. Bhupathiraju S, et al. Body-mass index and all-cause mortality: 22 Carrillo-Larco RM, Miranda JJ, Gilman RH, et al. Trajectories of individual-participant-data meta-analysis of 239 prospective studies body mass index and waist circumference in four Peruvian settings in four continents. Lancet 2016; 388: 776–86. at different level of urbanisation: the CRONICAS cohort study. 2 NCD Risk Factor Collaboration. Trends in adult body-mass index in J Epidemiol Community Health 2018; 72: 397–403. 200 countries from 1975 to 2014: a pooled analysis of 1698 population- 23 Parra DC, Iannotti L, Gomez LF, et al. The nutrition transition in based measurement studies with 19.2 million participants. Lancet Colombia over a decade: a novel household classification system of 2016; 387: 1377–96. anthropometric measures. Arch Public Health 2015; 73: 12. 3 Jeon H, Salinas D, Baker DP. Non-linear education gradient across 24 Bergallo P, Castagnari V, Fernández A, Mejía R. Regulatory the nutrition transition: mothers’ overweight and the population initiatives to reduce sugar-sweetened beverages (SSBs) in Latin education transition. Public Health Nutr 2015; 18: 3172–82. America. PLoS One 2018; 13: e0205694. 4 Martorell R, Khan LK, Hughes ML, Grummer-Strawn LM. Obesity 25 Jou J, Techakehakij W. International application of sugar-sweetened in Latin American women and children. J Nutr 1998; 128: 1464–73. beverage (SSB) taxation in obesity reduction: factors that may 5 Poterico JA, Stanojevic S, Ruiz-Grosso P, Bernabe-Ortiz A, influence policy effectiveness in country-specific contexts. Miranda JJ. The association between socioeconomic status and Health Policy 2012; 107: 83–90. obesity in Peruvian women. Obesity 2012; 20: 2283–89. 26 Fernandez MA, Raine KD. Insights on the influence of sugar taxes 6 Carrillo-Larco RM, Miranda JJ, Bernabé-Ortiz A. Wealth index and on obesity prevention efforts. Curr Nutr Rep 2019; published online risk of childhood overweight and obesity: evidence from four June 8. DOI:10.1007/s13668-019-00282-4. prospective cohorts in Peru and Vietnam. Int J Public Health 2016; 27 Mauvais-Jarvis F. Epidemiology of gender differences in diabetes 61: 475–85. and obesity. Adv Exp Med Biol 2017; 1043: 3–8. 7 Jaacks LM, Vandevijvere S, Pan A, et al. The obesity transition: stages 28 Aguilar M, Bhuket T, Torres S, Liu B, Wong RJ. Prevalence of the of the global epidemic. Lancet Diabetes Endocrinol 2019; published metabolic syndrome in the United States, 2003–2012. JAMA 2015; online Jan 28. https://doi.org/10.1016/S2213-8587(19)30026-9. 313: 1973–74. 8 Dansereau E, McNellan CR, Gagnier MC, et al. Coverage and 29 Gu D, Reynolds K, Wu X, et al. Prevalence of the metabolic syndrome timing of antenatal care among poor women in 6 Mesoamerican and overweight among adults in China. Lancet 2005; 365: 1398–405. countries. BMC Pregnancy Childbirth 2016; 16: 234. 30 Gupta R, Deedwania PC, Gupta A, Rastogi S, Panwar RB, Kothari K. 9 Martorell R, Khan LK, Hughes ML, Grummer-Strawn LM. Obesity in Prevalence of metabolic syndrome in an Indian urban population. women from developing countries. Eur J Clin Nutr 2000; 54: 247–52. Int J Cardiol 2004; 97: 257–61. 10 Uauy R, Albala C, Kain J. Obesity trends in Latin America: 31 Huayanay-Espinoza CA, Quispe R, Poterico JA, Carrillo-Larco RM, transiting from under- to overweight. J Nutr 2001; 131: 893S–99S. Bazo-Alvarez JC, Miranda JJ. Parity and overweight/obesity in 11 Ministerio de Defensa. Población de la República Argentina. Peruvian women. Prev Chronic Dis 2017; 14: E102. Instituto Geográfico Nacional. http://www.ign.gob.ar/ 32 Reis-Santos B, Barros FC, Horta BL. Is there a causal effect of parity NuestrasActividades/Geografia/DatosArgentina/Poblacion2 on body composition: a birth cohort study. BMC Public Health 2018; (accessed Oct 30, 2018). 18: 215. 12 Ahmad OB, Boschi-Pinto C, Lopez AD, Murray CJL, Lozano R, 33 Link JC, Reue K. Genetic basis for sex differences in obesity and Inoue M. Age standardization of rates: a new WHO standard. lipid metabolism. Annu Rev Nutr 2017; 37: 225–45. World Health Organization, 2001 https://www.who.int/healthinfo/ paper31.pdf (accessed Oct 8, 2019). 34 Merz AA, Cheng S. Sex differences in cardiovascular ageing. Heart 2016; 102: 825–31. 13 NCD Risk Factor Collaboration. Rising rural body-mass index is the main driver of the global obesity epidemic in adults. Nature 2019; 35 McCartney G, Bartley M, Dundas R, et al. Theorising social class 569: 260–64. and its application to the study of health inequalities. SSM Popul Health 2018; 7: 100315. 14 Monteiro CA, Conde WL, Lu B, Popkin BM. Obesity and inequities in health in the developing world. Int J Obes Relat Metab Disord 2004; 28: 1181–86. www.thelancet.com/lancetgh Vol 7 December 2019 e1654