Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128·9 million children, adolescents, and adults

A icles www. helance .com Vol 390 Decembe 16, 2017 2627 Wo ldwide ends in body-mass index, unde weigh , o e weigh , and obesi y om 1975 o 2016: a pooled analysis o 2416 popula ion-based measu emen s udies in 128·9 million child en, adolescen s, and adul s NCD Risk Fac o Collabo a ion (NCD-RisC)* Summa y Backg ound Unde weigh , o e weigh , and obesi y in childhood and adolescence a e associa ed wi h ad e se heal h consequences h oughou he li e-cou se. Ou aim was o es ima e wo ldwide ends in mean body-mass index (BMI) and a comp ehensi e se o BMI ca ego ies ha co e unde weigh o obesi y in child en and adolescen s, and o compa e ends wi h hose o adul s. Me hods We pooled 2416 popula ion-based s udies wi h measu emen s o heigh and weigh on 128·9 million pa icipan s aged 5 yea s and olde , including 31·5 million aged 5–19 yea s. We used a Bayesian hie a chical model o es ima e ends om 1975 o 2016 in 200 coun ies o mean BMI and o p e alence o BMI in he ollowing ca ego ies o child en and adolescen s aged 5–19 yea s: mo e han 2 SD below he median o he WHO g ow h e e ence o child en and adolescen s ( e e ed o as mode a e and se e e unde weigh he ea e ), 2 SD o mo e han 1 SD below he median (mild unde weigh ), 1 SD below he median o 1 SD abo e he median (heal hy weigh ), mo e han 1 SD o 2 SD abo e he median (o e weigh bu no obese), and mo e han 2 SD abo e he median (obesi y). Findings Regional change in age-s anda dised mean BMI in gi ls om 1975 o 2016 anged om i ually no change (–0·01 kg/m² pe decade; 95% c edible in e al –0·42 o 0·39, pos e io p obabili y [PP] o he obse ed dec ease being a ue dec ease=0·5098) in eas e n Eu ope o an inc ease o 1·00 kg/m² pe decade (0·69–1·35, PP>0·9999) in cen al La in Ame ica and an inc ease o 0·95 kg/m² pe decade (0·64–1·25, PP>0·9999) in Polynesia and Mic onesia. The ange o boys was om a non-signi ican inc ease o 0·09 kg/m² pe decade (–0·33 o 0·49, PP=0·6926) in eas e n Eu ope o an inc ease o 0·77 kg/m² pe decade (0·50–1·06, PP>0·9999) in Polynesia and Mic onesia. T ends in mean BMI ha e ecen ly la ened in no hwes e n Eu ope and he high-income English-speaking and Asia-Paci ic egions o bo h sexes, sou hwes e n Eu ope o boys, and cen al and Andean La in Ame ica o gi ls. By con as , he ise in BMI has accele a ed in eas and sou h Asia o bo h sexes, and sou heas Asia o boys. Global age-s anda dised p e alence o obesi y inc eased om 0·7% (0·4–1·2) in 1975 o 5·6% (4·8–6·5) in 2016 in gi ls, and om 0·9% (0·5–1·3) in 1975 o 7·8% (6·7–9·1) in 2016 in boys; he p e alence o mode a e and se e e unde weigh dec eased om 9·2% (6·0–12·9) in 1975 o 8·4% (6·8–10·1) in 2016 in gi ls and om 14·8% (10·4–19·5) in 1975 o 12·4% (10·3–14·5) in 2016 in boys. P e alence o mode a e and se e e unde weigh was highes in India, a 22·7% (16·7–29·6) among gi ls and 30·7% (23·5–38·0) among boys. P e alence o obesi y was mo e han 30% in gi ls in Nau u, he Cook Islands, and Palau; and boys in he Cook Islands, Nau u, Palau, Niue, and Ame ican Samoa in 2016. P e alence o obesi y was abou 20% o mo e in se e al coun ies in Polynesia and Mic onesia, he Middle Eas and no h A ica, he Ca ibbean, and he USA. In 2016, 75 (44–117) million gi ls and 117 (70–178) million boys wo ldwide we e mode a ely o se e ely unde weigh . In he same yea , 50 (24–89) million gi ls and 74 (39–125) million boys wo ldwide we e obese. In e p e a ion The ising ends in child en’s and adolescen s’ BMI ha e pla eaued in many high-income coun ies, albei a high le els, bu ha e accele a ed in pa s o Asia, wi h ends no longe co ela ed wi h hose o adul s. Funding Wellcome T us , As aZeneca Young Heal h P og amme. Copy igh © The Au ho (s). Published by Else ie L d. This is an Open Access a icle unde he CC BY 4.0 license. In oduc ion Being unde weigh , o e weigh , o obese du ing childhood and adolescence is associa ed wi h ad e se heal h consequences h oughou he li e-cou se. Unde - weigh among child en and adolescen s is associa ed wi h highe isk o in ec ious diseases, and o gi ls o childbea ing age, is associa ed wi h ad e se p egnancy ou comes including ma e nal mo ali y, deli e y complica ions, p e e m bi h, and in au e ine g ow h e a da ion.1,2 P e en ing and e e sing excess weigh in child en and adolescen s is also impo an o many easons;3,4 i s , weigh loss and main enance a e weigh loss a e ha d o achie e,5 he e o e gaining excess weigh in childhood and adolescence is likely o lead o li elong Lance 2017; 390: 2627–42 Published Online Oc obe 10, 2017 h p://dx.doi.o g/10.1016/ S0140-6736(17)32129-3 See Commen page 2607 *Membe s lis ed in he appendix Co espondence o: P o Majid Ezza i, School o Public Heal h, Impe ial College London, London W2 1PG, UK [email p o ec ed] See Online o appendix A icles 2628 www. helance .com Vol 390 Decembe 16, 2017 o e weigh and obesi y.6 Second, being o e weigh in childhood and adolescence is associa ed wi h g ea e isk and ea lie onse o ch onic diso de s such as ype 2 diabe es.3,4,7–9 Thi d, childhood and adolescen obesi y has ad e se psychosocial consequences and lowe s educa ional a ainmen .3,4,10,11 Finally, child en and adolescen s a e mo e suscep ible o ood ma ke ing han adul s, which makes educing child en’s exposu e o obesogenic oods necessa y o p o ec hem om ha m.3,12 Al hough ends in child en’s and adolescen s’ weigh s a us ha e been documen ed in indi idual coun ies, li le compa able in o ma ion exis s on wo ldwide ends, and none o mean body-mass index (BMI) and unde weigh . We pooled popula ion-based da a o es ima e ends om 1975 o 2016 in mean BMI and in he p e alence o a comp ehensi e se o BMI ca ego ies ha co e he unde weigh o obese ange among child en and adolescen s o all coun ies in he wo ld. To compa e he ajec o y o obesi y and unde weigh in child en and adolescen s wi h ha o adul s, we also gene a ed upda ed es ima es o adul s.13 Me hods S udy design We pooled and analysed popula ion-based s udies ha had measu ed heigh and weigh in people aged 5 yea s and olde o es ima e ends om 1975 o 2016 in mean BMI and BMI ca ego ies in 200 coun ies and e i o ies (appendix). We s a ed ou analysis om 5 yea s o age because he de ini ions o unde weigh , o e weigh , and obesi y change a 5 yea s o age.14 Fu he , child en en e school a o a ound his age, which is associa ed wi h a change in hei nu i ion and physical ac i i y.15 We p esen da a on school-aged child en and adolescen s aged 5–19 yea s and on adul s aged 20 yea s and olde . We did sepa a e analyses o child en and adolescen s and o adul s o wo easons: i s , cu o s used o de ine unde weigh , o e weigh , and obesi y o child en and adolescen s a e di e en om hose o adul s and a y by age and sex because o he na u al g ow h in childhood and adolescence.16 Second, he ajec o y o he obesi y epidemic in child en and adolescen s migh be di e en om ha o adul s,17 mo i a ing sepa a e analyses o ends. Simila ly, unde weigh in child en and adolescen s is ypically a ge ed h ough school and communi y-based nu i ion p og ammes, decoupling i s ajec o y om ha o adul s. Fo child en and adolescen s, we analysed ends in mean BMI and p e alence o BMI in he ollowing ca ego ies: mo e han 2 SD below he median o he WHO g ow h e e ence o child en and adolescen s16 (he ea e e e ed o as mode a e and se e e unde weigh ), 2 SD o mo e han 1 SD below he median (mild unde weigh ), 1 SD below he median o 1 SD abo e he median (heal hy Resea ch in con ex E idence be o e his s udy We sea ched MEDLINE ( ia PubMed) o a icles published in any language be ween Jan 1, 1950, and July 12, 2017, using he sea ch e ms (“body size”[mh:noexp] OR “body heigh ”[mh:noexp] OR “body weigh ”[mh:noexp] OR “bi h weigh ”[mh:noexp] OR “o e weigh ”[mh:noexp] OR “obesi y”[mh] OR “ hinness”[mh:noexp] OR “Wais -Hip Ra io”[mh:noexp] OR “Wais Ci cum e ence”[mh:noexp] OR “body mass index”[mh:noexp]) AND (“Humans”[mh]) AND (“Heal h Su eys”[mh] OR “Epidemiological Moni o ing”[mh] OR “P e alence”[mh]) NOT Commen [p yp] NOT Case Repo s[p yp]. A icles we e sc eened acco ding o he inclusion and exclusion c i e ia desc ibed in he Appendix. We iden i ied h ee p io global analyses o mean body-mass index (BMI) o p e alence o o e weigh and obesi y among adul s. One o hese s udies also es ima ed he p e alence o unde weigh in adul s. Ano he s udy also included da a on o e weigh and obesi y in child en and adolescen s, using a combina ion o measu ed and sel - epo ed heigh and weigh , and analysed in he same model as adul s. A ew mul icoun y s udies and sys ema ic e iews ha e epo ed, quan i a i ely o quali a i ely, ends in o e weigh and obesi y in child en and adolescen s, some also epo ing unde weigh . To ou knowledge, he e is no global analysis o mean BMI, which is a summa y measu e o popula ion dis ibu ion, o p e alence o unde weigh among child en and adolescen s aged 5–19 yea s. Added alue o his s udy This s udy p o ides a comple e pic u e o ends in mean BMI and p e alence o BMI ca ego ies ha co e he unde weigh o obese ange among child en and adolescen s aged 5–19 yea s, o all coun ies in he wo ld wi h he longes obse a ion pe iod, and compa es ends wi h hose o adul s. I includes he i s global es ima es o mean BMI and unde weigh p e alence o child en and adolescen s. We also p esen ends in he numbe o child en, adolescen s, and adul s who a e mode a ely o se e ely unde weigh and obese, and hus a isk o ad e se heal h ou comes. Implica ions o all he a ailable e idence O e he pas ou decades, mean BMI and obesi y in child en and adolescen s aged 5–19 yea s ha e inc eased in mos egions and coun ies. Despi e his ise, mo e child en and adolescen s a e mode a ely o se e ely unde weigh han obese, wi h he bu den o unde weigh inc easingly concen a ed in sou h Asia and cen al, eas and wes A ica. The ise in child en’s and adolescen s’ BMI has pla eaued, albei a high le els, in many high-income coun ies bu has accele a ed in pa s o Asia. The e is a need o b idging he disconnec be ween policies ha add ess unde weigh and o e weigh in child en and adolescen s o cohe en ly add ess he la ge emaining unde weigh bu den while cu bing and e e sing he ise in o e weigh and obesi y. A icles www. helance .com Vol 390 Decembe 16, 2017 2629 Age-s anda dised mean BMI (kg/m²) Age-s anda dised mean BMI (kg/m²)Age-s anda dised mean BMI (kg/m²)Age-s anda dised mean BMI (kg/m²) Age-s anda dised mean BMI (kg/m²) Cen al Asia 15 20 25 30 Sou h Asia Sou heas Asia Eas Asia High-income Asia-Paci ic High-income English-speaking coun ies 15 20 25 30 No hwes e n Eu ope Sou hwes e n Eu ope Cen al Eu ope Eas e n Eu ope Middle Eas and no h A ica 15 20 25 30 Wes A ica Cen al A ica Eas A ica Sou he n A ica Cen al La in Ame ica 15 20 25 30 Andean La in Ame ica Sou he n La in Ame ica Ca ibbean 15 20 25 30 Melanesia Polynesia and Mic onesia Wo ld 15 20 25 30 Adul s Child en and adolescen s 1980 1990 2000 Yea Yea Yea Yea 2010 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010 Figu e 1: T ends in age-s anda dised mean BMI by sex and egion in emales Child en and adolescen s we e aged 5–19 yea s and adul s we e aged 20 yea s and olde . The lines show he pos e io mean es ima es and he shaded a eas show he 95% c edible in e als. See appendix o ends by coun y. BMI=body-mass index. A icles 2630 www. helance .com Vol 390 Decembe 16, 2017 Age-s anda dised mean BMI (kg/m²)Age-s anda dised mean BMI (kg/m²)Age-s anda dised mean BMI (kg/m²)Age-s anda dised mean BMI (kg/m²)Age-s anda dised mean BMI (kg/m²) Cen al Asia 15 20 25 30 Sou h Asia Sou heas Asia Eas Asia High-income Asia-Paci ic High-income English-speaking coun ies 15 20 25 30 No hwes e n Eu ope Sou hwes e n Eu ope Cen al Eu ope Eas e n Eu ope Middle Eas and no h A ica 15 20 25 30 Wes A ica Cen al A ica Eas A ica Sou he n A ica Cen al La in Ame ica 15 20 25 30 Andean La in Ame ica Sou he n La in Ame ica Ca ibbean 15 20 25 30 Melanesia Polynesia and Mic onesia Wo ld 15 20 25 30 Adul s Child en and adolescen s 1980 1990 2000 Yea Yea Yea Yea 2010 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010 Figu e 2: T ends in age-s anda dised mean BMI by sex and egion in males Child en and adolescen s we e aged 5–19 yea s and adul s we e aged 20 yea s and olde . The lines show he pos e io mean es ima es and he shaded a eas show he 95% c edible in e als. See appendix o ends by coun y. BMI=body-mass index. A icles www. helance .com Vol 390 Decembe 16, 2017 2631 weigh ), mo e han 1 SD o 2 SD abo e he median (o e weigh bu no obese), and mo e han 2 SD abo e he median (obesi y). The cu o s o calcula ing p e alence in hese BMI ca ego ies we e all age-speci ic and sex-speci ic and we e applied o da a in 1-yea age bands. We used he WHO de ini ions because hey include a comp ehensi e se o BMI ca ego ies anging om mode a e and se e e unde weigh o obesi y, de ined based on symme ic h esholds o SDs om he e e ence popula ion median. Fo adul s, we analysed ends in mean BMI and p e alence o a comp ehensi e se o BMI ca ego ies as desc ibed in de ail elsewhe e and in he appendix.13 Resul s o child en and adolescen s a e p esen ed he e; upda ed esul s o adul s a e p esen ed in he appendix excep when compa ed wi h child en and adolescen s. Da a sou ces Ou me hods o iden i ying and accessing da a sou ces, and ou inclusion and exclusion c i e ia, a e desc ibed in he appendix. In summa y, we used a da abase o popula ion-based da a on ca diome abolic isk ac o s colla ed by he Non-communicable Disease Risk Fac o Collabo a ion (NCD-RisC), a wo ldwide ne wo k o heal h esea che s and p ac i ione s whose aim is o documen sys ema ically he wo ldwide ends and a ia ions in non-communicable disease isk ac o s. The da abase was colla ed h ough mul iple ou es o iden i ying and accessing da a. We accessed publicly a ailable popula ion-based mul i-coun y and na ional measu emen su eys, as well as he WHO STEPwise app oach o Su eillance (STEPS) su eys. We eques ed, ia WHO and i s egional and coun y o ices, help wi h iden i ica ion and access o popula ion-based su eys om minis ies o heal h and o he na ional heal h and s a is ical agencies. We also sen eques s ia he Wo ld Hea Fede a ion o i s na ional pa ne s. We made a simila eques o he au ho s o an ea lie pooled analysis o ca diome abolic isk ac o s,18–21 and in i ed hem o eanalyse da a om hei s udies and join NCD- RisC. To iden i y majo sou ces no accessed h ough he abo e ou es, we sea ched and e iewed published s udies as de ailed p e iously,13 and in i ed all eligible s udies o join NCD-RisC. Finally, NCD-RisC membe s we e pe iodically asked o e iew he lis o sou ces om hei coun y, o sugges addi ional sou ces no in he da abase, and o e i y ha he included da a om hei coun y me he inclusion c i e ia as lis ed in he appendix and ha he e we e no duplica es. The lis o da a sou ces and hei cha ac e is ics is p o ided in he appendix. In summa y, we included da a collec ed on samples o a na ional, subna ional (ie, co e ing one o mo e subna ional egions, o mo e han h ee communi ies), o communi y (one o a ew communi ies) popula ion, which had measu ed heigh and weigh . We did no use sel - epo ed heigh and weigh because hey a e subjec o biases ha a y by geog aphy, ime, age, sex, and socioeconomic cha ac e is ics.22–24 Because o hese a ia ions, app oaches o co ec ing sel - epo ed da a lea e esidual bias. S a is ical analysis The s a is ical models used o es ima e mean and p e alence by coun y, yea , sex, and age a e desc ibed in de ail in a s a is ical pape and ela ed subs an i e pape s;13,25–28 he compu e code is a ailable om he NCD-RisC websi e. In summa y, we o ganised coun ies in o 21 egions, mos ly based on geog aphy and na ional income (appendix). The excep ion was high-income English-speaking coun ies (Aus alia, Canada, I eland, New Zealand, he UK, and he USA), g ouped oge he in one egion because BMI and o he ca diome abolic isk ac o s ha e simila ends in hese coun ies, which can be dis inc om o he coun ies in hei geog aphical egions.13,26–28 The model had a hie a chical s uc u e in which es ima es o each coun y and yea we e in o med by i s own da a, i a ailable, and by da a om o he yea s in he same coun y and om o he coun ies, especially hose in he same egion wi h da a o simila ime pe iods. The ex en o which es ima es o each coun y-yea we e in luenced by da a om o he yea s and o he coun ies depended on whe he he coun y had da a, he sample size o da a, whe he o no hey we e na ional, and he wi hin-coun y and wi hin- egion a iabili y o he a ailable da a. The model inco po a ed non-linea ime ends comp ising linea e ms and a second-o de andom walk, all modelled hie a chically. The age associa ion o BMI was modelled using a cubic spline o allow non-linea age pa e ns, which migh a y ac oss coun ies. The model accoun ed o he possibili y ha BMI in subna ional and communi y samples migh sys ema ically di e om na ionally ep esen a i e ones, and ha e la ge a ia ion han in na ional s udies. These ea u es we e implemen ed by including da a-d i en ixed-e ec and andom-e ec e ms o subna ional and communi y da a. The ixed e ec s adjus ed o sys ema ic di e ences be ween subna ional o communi y s udies and na ional s udies. The andom e ec s allowed na ional da a o ha e la ge in luence on he es ima es han subna ional o communi y da a wi h simila sample sizes. The model also accoun ed o u al–u ban di e ences in BMI, h ough he use o da a-d i en ixed e ec s o u al- only and u ban-only s udies. These u al and u ban e ec s we e weigh ed by he di e ence be ween s udy-le el and coun y-le el u banisa ion in he yea when he s udy was done. The s a is ical model included a co a ia e (p opo ion o na ional popula ion li ing in u ban a eas; da a om he Wo ld U baniza ion P ospec s, 2014 e ision) ha is associa ed wi h, and helps p edic , BMI.29 Resul s o model alida ion a e epo ed elsewhe e.13 We pe o med all analyses by sex, because he e a e di e ences in BMI le els and ends in ela ion o sex.13 We analysed he da a on mean BMI and on each o he abo e p e alence ca ego ies sepa a ely. We e-scaled he Fo he NCD-RisC websi e see www.ncd isc.o g A icles 2632 www. helance .com Vol 390 Decembe 16, 2017 Figu e 3: Age-s anda dised mean BMI, p e alence o obesi y, and p e alence o mode a e and se e e unde weigh by sex and coun y in 2016 in child en and adolescen s Child en and adolescen s we e aged 5–19 yea s. Obesi y was de ined as mo e han 2 SD abo e he median o he WHO g ow h e e ence. Mode a e and se e e unde weigh was de ined as mo e han 2 SD below he median. See appendix o esul s o adul s. BMI=body-mass index. AMean 8MI in gi ls Ca ibbean BMean 8MI in boys Ca ibbean Ame icanSamoa Ki iba i Samoa Age-s anda dised Ame ican Samoa Ki iba i Samoa Age-s anda dised Bah ain Maldi es Sao Tome and P incipe mean BMI (kg/m') Bah ain Maldi es Sao Tome and P ncipe mean 8MI (kg/m') Be muda Ma shall Islands Seychelles -24 Be muda Ma shall Islands Seychelles 24 B unei Da ussalam Mau i ius Solomon Islands B unei Da ussalam Mau i ius Solomon Islands CaboVe de Mic onesia FS Tokelau CaboVe de Mic onesia FS Tokelau Como os Mon eneg o Tonga -20 Como os Mon eneg o Tonga 20 CookIslands Nau u Tu alu Cook Islands Nau u Tu alu Fiji Niue Vanua u Fiji Niue Vanua u F enchPolynesia Palau -16 F enchPolynesia Palau 16 CObesi yp e alence in gi ls DObesi yp e alence in boys ICa ibbean Ca ibbean Ame ican Samoa Ki iba i Samoa P e alence o Ame ican Samoa Ki iba i Samoa P e alence o Bah ain Maldi es SaoTome and P incipe obesi y (%) Bah ain Maldi es SaoTome and P ncipe obesi y (%) Be muda Ma shall Islands Seychelles 30 Be muda Ma shall Islands Seychelles 30 B unei Da ussalam Mau i ius Solomon Islands B unei Da ussalam Mau i ius Solomon Islands CaboVe de Mic onesia FS Tokelau 20 CaboVe de Mic onesia FS Tokelau 20 Como os Mon eneg o Tonga Como os Mon eneg o Tonga Cook Islands Nau u Tu alu 10 Cook Islands Nau u Tu alu 10 Fiji Niue Vanua u Fiji Niue Vanua u F enchPolynesia Palau F enchPolynesia Palau E Mode a e and se e e unde weigh p e alence in gi ls F Mode a e and se e e unde weigh p e alence in boys ICa ibbean Ca ibbean Ame ican Samoa Ki iba i Samoa P e alence Ame ican Samoa Ki iba i Samoa P e alence Bah ain Maldi es SaoTome and P incipe o mode a e Bah ain Maldi es SaoTome and P ncipe o mode a e and se e e and se e e Be muda Ma shall Islands Seychelles unde weigh (%) Be muda Ma shall Islands Seychelles unde weigh (%) B unei Da ussalam Mau i ius Solomon Islands 30 B unei Da ussalam Mau i ius Solomon Islands 30 CaboVe de Mic onesia FS Tokelau CaboVe de Mic onesia FS Tokelau 20 20 Como os Mon eneg o Tonga Como os Mon eneg o Tonga Cook Islands Nau u Tu alu 10 Cook Islands Nau u Tu alu 10 Fiji Niue Vanua u Fiji Niue Vanua u F enchPolynesia Palau F enchPolynesia Palau A icles www. helance .com Vol 390 Decembe 16, 2017 2633 es ima ed p e alence ca ego ies so ha he sum o di e en ca ego ies was 1·0 in each age, sex, coun y, and yea . The a e age scaling ac o s ac oss d aws anged om 1·03 o 1·07, ie, he sum o he sepa a ely es ima ed p e alence ca ego ies was close o 1·0. We i ed he s a is ical model wi h he Ma ko chain Mon e Ca lo (MCMC) algo i hm, and ob ained 5000 pos -bu n-in samples om he pos e io dis ibu ion o model pa ame e s, which we e in u n used o ob ain he pos e io dis ibu ions o he abo e p ima y ou comes, ie, mean BMI and each o he p e alence ca ego ies. Fo model i ing, da a on pa icipan s aged 5–19 yea s we e included in he analysis o ends in child en and adolescen s, and on pa icipan s aged 18 yea s and olde in he analysis o ends in adul s. Da a on pa icipan s aged 18 and 19 yea s we e included in bo h se s o models because hese g oups o m a ansi ional age om adolescence o adul hood, and hence help in o m he es ima es in bo h g oups. Pos e io es ima es we e made in 1-yea age g oups o ages 5–19 yea s and in 5-yea age g oups o hose aged 20 yea s and olde . Fo p esen a ion, we used he pos e io es ima es o ages 5–19 yea s o child en and adolescen s, and o ages 20 yea s and olde o adul s. Age-s anda dised es ima es we e gene a ed by aking weigh ed a e ages o age- sex-speci ic es ima es, sepa a ely o child en and adolescen s and o adul s, wi h use o age weigh s om he WHO s anda d popula ion.30 Es ima es o egions and he wo ld we e calcula ed as popula ion- weigh ed a e ages o he cons i uen coun y es ima es by age g oup and sex. The numbe o child en, adolescen s, and adul s who we e unde weigh , o e weigh , o obese was calcula ed by mul iplying he co esponding age-speci ic p e alence by he popula ion by coun y, yea , and sex. The epo ed c edible in e als (C I) ep esen he 2·5 h o 97·5 h pe cen iles o he pos e io dis ibu ions. The unce ain ies o ou es ima es, ep esen ed by he wid hs o he c edible in e als, a ise om unce ain y due o sampling in each da a sou ce; unce ain y associa ed wi h he a iabili y o na ional da a beyond wha is accoun ed o by sampling; addi ional unce ain y associa ed wi h subna ional and communi y da a, and da a ha a e om u al-only o u ban-only samples; and unce ain y due o making es ima es by coun y, yea , and age when da a we e missing o sca ce, in he coun y-yea -age g oup uni o which es ima es a e made, in p oxima e ime pe iods and ages in ha coun y and in o he coun ies in he same egion. We also epo he pos e io p obabili y (PP) ha an es ima ed inc ease o dec ease ep esen s a uly inc easing o dec easing end. Role o he unding sou ce The unde o he s udy had no ole in s udy design, da a collec ion, analysis, in e p e a ion, o w i ing o he epo . Coun y and Regional Da a G oup membe s, JB, MDC, VB, HB, and BZ had ull access o he da a in he s udy. The co esponding au ho had inal esponsibili y o he decision o submi o publica ion. Resul s The esul s can be explo ed using dynamic isualisa ions and downloaded om he NCD-RisC websi e. We pooled 2416 popula ion-based da a sou ces wi h measu emen o heigh and weigh on 128·9 million people aged 5 yea s and olde om 1975 o 2016 (appendix). 1099 sou ces included da a on 24·1 million pa icipan s aged 5–17 yea s, 848 sou ces included da a on 7·4 million pa icipan s aged 18–19 yea s, and 1820 sou ces included da a on Figu e 4: Age-s anda dised mean BMI in child en and adolescen s in 1975 and 2016 Each line shows one coun y. BMI=body-mass index. Gi ls, 1975 14 16 18 20 22 24 Wo ld mean: 17·2 Gi ls, 2016 Wo ld mean: 18·6 Boys, 1975 Wo ld mean: 16·8 Boys, 2016 Wo ld mean: 18·5 Mean BMI (kg/m²) Sub-Saha an A ica Cen al Asia, Middle Eas , and no h A ica Sou h Asia Eas and sou heas Asia High-income Asia-Paci ic Oceania La in Ame ica and Ca ibbean High-income wes e n Cen al and eas e n Eu ope A icles 2634 www. helance .com Vol 390 Decembe 16, 2017 97·4 million pa icipan s aged 20 yea s and olde (appendix). Addi ional in o ma ion on he age dis- ibu ion o da a sou ces and pa icipan s is shown in he appendix. 1187 (49%) o 2416 da a sou ces we e om na ional samples, 390 (16%) co e ed one o mo e subna ional egions, and he emaining 839 (35%) we e om one o a ew communi ies (appendix). 583 (24%) o 2416 da a sou ces we e om be o e 1995 and ano he 1833 (76%) we e om 1995 and la e (appendix). The numbe o da a sou ces pe coun y in di e en egions anged om 3·4 in he Ca ibbean o 54·0 in he high- income Asia-Paci ic egion (appendix). In 1975, he global age-s anda dised mean BMI o child en and adolescen s aged 5–19 yea s was 17·2 kg/m² (95% C I 16·8–17·6) o gi ls ( igu e 1) and 16·8 kg/m² (16·3–17·2) o boys ( igu e 2). Mean BMI was lowes in sou h Asia, wi h an age-s anda dised mean o 15·8 kg/m² (15·2–16·3) o gi ls and 15·0 kg/m² (14·5–15·6) o boys, ollowed by eas A ica (16·5 kg/m² [14·8–18·2] o gi ls and 15·5 kg/m² [13·6–17·4] o boys). Gi ls in Melanesia, Polynesia and Mic onesia, and he high- income English-speaking egion had he highes age- s anda dised mean BMI in 1975, all abo e 19·0 kg/m². The highes mean BMIs o boys we e hose in Polynesia and Mic onesia (19·1 kg/m², 18·0–20·2), ollowed by he high-income English-speaking egion. F om 1975 o 2016, child en’s and adolescen s’ age- s anda dised mean BMI inc eased globally and in mos egions ( igu es 1 and 2). The global inc ease was 0·32 kg/m² pe decade (95% C I 0·23–0·41, PP o he obse ed inc ease being a ue inc ease>0·9999) o gi ls and 0·40 kg/m² pe decade (0·30–0·50, PP>0·9999) o boys, leading o i ually iden ical age-s anda dised mean BMIs o 18·6 kg/m² (18·4–18·7) o gi ls and 18·5 kg/m² (18·3–18·7) o boys in 2016. The co esponding igu es o adul s we e 24·8 kg/m² (24·6–25·0) in women and 24·5 kg/m² (24·3–24·6) in men. Regional change in gi ls anged om i ually no change (–0·01 kg/m² pe decade [95% C I –0·42 o 0·39; PP o he obse ed dec ease being a ue dec ease=0·5098]) in eas e n Eu ope o 1·00 kg/m² inc ease pe decade (0·69–1·35, PP>0·9999) in cen al La in Ame ica and 0·95 kg/m² pe decade (0·64–1·25, PP>0·9999) in Polynesia and Mic onesia. The ange o boys was om 0·09 kg/m² pe decade (–0·33 o 0·49; PP=0·6926) in eas e n Eu ope o 0·77 kg/m² pe decade (0·50–1·06, PP>0·9999) in Polynesia and Mic onesia. In some egions, child en’s and adolescen s’ BMI inc eased g adually o e he ou decades o analysis ( igu es 1 and 2). Howe e , he e has been a ecen la ening o ends in no hwes e n Eu ope and he high-income English-speaking and Asia-Paci ic egions o bo h sexes, sou hwes e n Eu ope o boys, and cen al and Andean La in Ame ica o gi ls. Wi h he excep ion o women in he high-income Asia-Paci ic egion, adul mean BMI con inues o inc ease in all o hese egions and sexes ( igu es 1 and 2). By con as wi h his pla eauing, he ise in mean BMI has accele a ed since a ound 2000 in eas and sou h Asia o bo h sexes, and in sou heas Asia o boys. The lowes mean child and adolescen BMIs in 2016 we e s ill hose in sou h Asia and eas A ica, wi h age- s anda dised mean BMIs be ween 16·9 and 17·9 kg/m² o gi ls and boys; he highes we e hose in Polynesia and Mic onesia o bo h sexes, ollowed by Melanesia and he high-income English-speaking egion. Age- s anda dised mean BMIs o gi ls and boys in Polynesia and Mic onesia, which we e 23·1 kg/m² (95% C I 22·4–23·8) and 22·4 kg/m² (21·6–23·1), espec i ely, we e highe han hose o adul s in some egions. Child en’s and adolescen s’ age-s anda dised mean BMI was also mo e han 20 kg/m² in Melanesia and many pa s o La in Ame ica and he Ca ibbean. The egional ankings in 2016 di e ed sligh ly be ween child en aged 5–9 yea s and adolescen s aged 10–19 yea s (appendix). Fo example, he lowes mean BMI in child en aged 5–9 yea s was seen in eas A ica in bo h sexes, whe eas in hose aged 10–19 yea s, sou h Asian gi ls and boys had lowe mean BMI han hei A ican pee s. Polynesians and Mic onesians had he highes mean BMI in hose aged 5–9 and 10–19 yea s, wi h he 14 16 18 20 22 24 14 16 18 20 22 24 Age-s anda dised mean BMI in child en and adolescen s (kg/m²) Females, 1975 Females, 2016 20 25 30 35 Age-s anda dised mean BMI in child en and adolescen s (kg/m²) Males, 1975 20 25 30 35 Males, 2016 Age-s anda dised mean BMI in adul s (kg/m²) Age-s anda dised mean BMI in adul s (kg/m²) Sub-Saha an A ica Cen al Asia, Middle Eas , and no h A ica Sou h Asia Eas and sou heas Asia High-income Asia-Paci ic Oceania La in Ame ica and Ca ibbean High-income wes e n Cen al and eas e n Eu ope Figu e 5: Compa ison o age-s anda dised mean BMI in child en and adolescen s and in adul s Child en and adolescen s we e aged 5–19 yea s and adul s we e aged 20 yea s and olde . Each poin shows one coun y. The do ed line shows he linea associa ion be ween he wo ou comes. BMI=body-mass index. A icles www. helance .com Vol 390 Decembe 16, 2017 2635 subsequen spo s held by he high-income English- speaking egion, egions in La in Ame ica and he Ca ibbean, and Melanesia. Among hese egions, cen al La in Ame ica had a poo e anking (ie, highe BMI ela i e o o he egions) a age 10–19 yea s han a age 5–9 yea s, as did boys in he high-income English- speaking egion. By con as , eas Asia pe o med wo se in anking in 5–9 yea s o age han i did in 10–19 yea s. The lowes age-s anda dised mean BMI o e he 42 yea s o analysis among gi ls was in Bangladesh in 1975 (15·6 kg/m², 95% C I 13·2–17·9), and among boys was in E hiopia in 1975 (14·4 kg/m², 11·9–17·0; igu es 3 and 4). Age-s anda dised mean BMI in 1975 was less han 21 kg/m² in e e y coun y, excep o gi ls in Ame ican Samoa, who had an age-s anda dised mean BMI o 21·2 kg/m² (20·6–21·9). F om 1975 o 2016, age-s anda dised mean BMI inc eased by mo e han 0·25 kg/m² pe decade in 155 coun ies in gi ls, wi h he ise mo e han 1·0 kg/m² pe decade in some coun ies in Polynesia and in Mexico (PP o being a ue ise>0·9999); in boys, he ise was mo e han 0·25 kg/m² pe decade in 189 coun ies and mo e han 1·0 kg/m² pe decade in he Cook Islands. When subse s o he analysis pe iod a e conside ed, be o e he yea 2000, age- s anda dised mean BMI inc eased in almos e e y coun y. A e 2000, he e we e non-signi ican declines in mean BMI in 29 coun ies o gi ls and 12 (mos ly high-income) coun ies o boys. In 2016, E hiopia had he lowes age-s anda dised mean BMI o bo h sexes, 16·8 kg/m² (95% C I 15·6–17·9) o gi ls and 15·5 kg/m² (14·4–16·6) o boys ( igu es 3 and 4). O he coun ies wi h low BMI in bo h sexes in 2016 we e Nige , Senegal, India, Bangladesh, Myanma , and Cambodia. A he o he ex eme, age- s anda dised mean BMI was mo e han 24 kg/m² in gi ls and boys in he Cook Islands and Niue and gi ls in Samoa, which was g ea e han ha o adul s o he same sex in 36 coun ies o gi ls and 59 coun ies o boys. Age-s anda dised mean BMI was be ween 22 and 24 kg/m² in ano he 11 coun ies o gi ls and 10 coun ies o boys including in Polynesian and Mic onesian islands, gi ls in he Bahamas and Chile, and boys in Qa a and Kuwai . The age-s anda dised mean BMI o child en and adolescen s and o adul s we e co ela ed in 1975 and 2016 (co ela ion coe icien s 0·80 and 0·85 o emales and 0·92 and 0·89 o males; igu e 5). Changes in age- s anda dised mean BMI we e mode a ely co ela ed be ween he wo age g oups be o e 2000 (co ela ion coe icien 0·52 o emales and 0·51 o males) bu only weakly a e 2000 (co ela ion coe icien 0·14 o emales and 0·21 o males; igu e 6). The decoupling o BMI ends in child en and adolescen s and hose o adul s is due o a se o dis inc egional phenomena: adul s con inued o gain weigh in mos wes e n coun ies, whe e child en’s and adolescen s’ mean BMI s opped ising. By con as , he ise in adul BMI seems o ha e pla eaued in Oceania, albei a high le els, whe eas child en’s and adolescen s’ BMI con inues o ise. In La in Ame ica and he Ca ibbean, he e is mo e a ia ion in he a e o BMI inc ease in child en and adolescen s han in adul s. In 1975, gi ls had highe age-s anda dised mean BMI han boys in mos coun ies in sub-Saha an A ica, sou h Asia, and he Middle Eas and no h A ica, and lowe age-s anda dised mean BMI han boys in many coun ies in Eu ope and La in Ame ica and he Ca ibbean ( igu e 7). Highe BMI in gi ls han boys was s ill seen in 2016 in many sub-Saha an A ican and sou h Asian coun ies. By con as , he gap be ween sexes in BMI in he Middle Eas and no h A ica sh ank o e e sed as boys gained mo e weigh han gi ls. In Eu ope and La in Ame ica, gi ls gained mo e weigh han boys, closing he gap be ween sexes in BMI. O e he 42 yea s o analysis, he global age- s anda dised p e alence o obesi y in child en and adolescen s inc eased om 0·7% (95% C I 0·4–1·2) in 1975 o 5·6% (4·8–6·5) in 2016 in gi ls ( igu e 8), and om 0·9% (0·5–1·3) in 1975 o 7·8% (6·7–9·1) in 2016 in boys ( igu e 9). Obesi y inc eased in e e y egion, wi h p opo ional ise being smalles in high-income egions Sub-Saha an A ica Cen al Asia, Middle Eas , and no h A ica Sou h Asia Eas and sou heas Asia High-income Asia-Paci ic Oceania La in Ame ica and Ca ibbean High-income wes e n Cen al and eas e n Eu ope 1·5 2·0 −0·5 Females, 1975−2000 Females, 2000−2016 −0·5 0 0·5 1·0 1·5 2·0 Change in age-s anda dised mean BMI in adul s (kg/m²/decade) −0·500·5 1·0 1·5 2·0 Change in age-s anda dised mean BMI in adul s (kg/m²/decade) −0·5 0 0·5 1·0 0 0·5 1·0 1·5 2·0 Change in age-s anda dised mean BMI in child en and adolescen s (kg/m²/decade) Change in age-s anda dised mean BMI in child en and adolescen s (kg/m²/decade) −0·5 0 0·5 1·0 1·5 2·0 −0·5 0·0 0·5 1·0 1·5 2·0 Males, 2000−2016Males, 1975−2000 Figu e 6: Compa ison o change in age-s anda dised mean BMI in child en and adolescen s and in adul s Child en and adolescen s we e aged 5–19 yea s and adul s we e aged 20 yea s and olde . Each poin shows one coun y. The do ed line shows he linea associa ion be ween he wo ou comes. BMI=body-mass index. A icles 2642 www. helance .com Vol 390 Decembe 16, 2017 20 Danaei G, Finucane MM, Lu Y, e al. Na ional, egional, and global ends in as ing plasma glucose and diabe es p e alence since 1980: sys ema ic analysis o heal h examina ion su eys and epidemiological s udies wi h 370 coun y-yea s and 2·7 million pa icipan s. Lance 2011; 378: 31–40. 21 Fa zad a F, Finucane MM, Danaei G, e al. Na ional, egional, and global ends in se um o al choles e ol since 1980: sys ema ic analysis o heal h examina ion su eys and epidemiological s udies wi h 321 coun y-yea s and 3·0 million pa icipan s. Lance 2011; 377: 578–86. 22 Go be SC, T emblay M, Mohe D, Go be B. A compa ison o di ec s. sel – epo measu es o assessing heigh , weigh and body mass index: a sys ema ic e iew. Obes Re 2007; 8: 307–26. 23 Ezza i M, Ma in H, Skjold S, Vande Hoo n S, Mu ay CJ. T ends in na ional and s a e-le el obesi y in he USA a e co ec ion o sel - epo bias: analysis o heal h su eys. J R Soc Med 2006; 99: 250–57. 24 Hayes AJ, Cla ke PM, Lung TW. Change in bias in sel - epo ed body mass index in Aus alia be ween 1995 and 2008 and he e alua ion o co ec ion equa ions. Popul Heal h Me 2011; 9: 53. 25 Finucane MM, Pacio ek CJ, Danaei G, Ezza i M. Bayesian es ima ion o popula ion-le el ends in measu es o heal h s a us. S a Sci 2014; 29: 18–25. 26 NCD Risk Fac o Collabo a ion (NCD-RisC). Wo ldwide ends in diabe es since 1980: a pooled analysis o 751 popula ion-based s udies wi h 4·4 million pa icipan s. Lance 2016; 387: 1513–30. 27 NCD Risk Fac o Collabo a ion (NCD-RisC). A cen u y o ends in adul human heigh . eLi e 2016; 5: e13410. 28 NCD Risk Fac o Collabo a ion (NCD-RisC). Wo ldwide ends in blood p essu e om 1975 o 2015: a pooled analysis o 1479 popula ion-based measu emen s udies wi h 19·1 million pa icipan s. Lance 2017; 389: 37–55. 29 Danaei G, Singh GM, Pacio ek CJ, e al. The global ca dio ascula isk ansi ion: associa ions o ou me abolic isk ac o s wi h mac oeconomic a iables in 1980 and 2008. Ci cula ion 2013; 127: 1493–502. 30 Ahmad OB, Boschi-Pin o C, Lopez AD, Mu ay CJ, Lozano R, Inoue M. Age s anda diza ion o a es: a new WHO s anda d. Gene a: Wo ld Heal h O ganiza ion; 2001: 1–14. 31 Ng M, Fleming T, Robinson M, e al. Global, egional, and na ional p e alence o o e weigh and obesi y in child en and adul s du ing 1980–2013: a sys ema ic analysis o he Global Bu den o Disease S udy 2013. Lance 2014; 384: 766–81. 32 Wabi sch M, Moss A, K omeye -Hauschild K. Unexpec ed pla eauing o childhood obesi y a es in de eloped coun ies. BMC Medicine 2014; 12: 17. 33 Olds T, Mahe C, Zumin S, e al. E idence ha he p e alence o childhood o e weigh is pla eauing: da a om nine coun ies. In J Pedia Obes 2011; 6: 342–60. 34 Rokholm B, Bake JL, Sø ensen TIA. The le elling o o he obesi y epidemic since he yea 1999–a e iew o e idence and pe spec i es. Obes Re 2010; 11: 835–46. 35 Mu hu i SK, F ancis CE, Wachi a L-JM, e al. E idence o an o e weigh /obesi y ansi ion among school-aged child en and you h in Sub-Saha an A ica: a sys ema ic e iew. PLoS One 2014; 9: e92846. 36 Ri e a JÁ, de Cossío TG, Ped aza LS, Abu o TC, Sánchez TG, Ma o ell R. Childhood and adolescen o e weigh and obesi y in La in Ame ica: a sys ema ic e iew. Lance Diabe es Endoc inol 2014; 2: 321–32. 37 Cole TJ, Bellizzi MC, Flegal KM, Die z WH. Es ablishing a s anda d de ini ion o child o e weigh and obesi y wo ldwide: in e na ional su ey. BMJ 2000; 320: 1240–43. 38 Cole TJ, Lobs ein T. Ex ended in e na ional (IOTF) body mass index cu -o s o hinness, o e weigh and obesi y. Pedia Obes 2012; 7: 284–94. 39 Ri o A, Wijnho en TM, Ru e H, e al. P e alence o obesi y among Po uguese child en (6-8 yea s old) using h ee de ini ion c i e ia: COSI Po ugal, 2008. Pedia Obes 2012; 7: 413–22. 40 Song Y, Wang H-J, Dong B, Ma J, Wang Z, Aga dh A. 25-yea ends in gende dispa i y o obesi y and o e weigh by using WHO and IOTF de ini ions among Chinese school-aged child en: a mul iple c oss-sec ional s udy. BMJ Open 2016; 6: e011904. 41 Bahk J, Khang Y-H. T ends in measu es o childhood obesi y in Ko ea om 1998 o 2012. J Epidemiol 2016; 26: 199–207. 42 Ah ens W, Pigeo I, Pohlabeln H, e al. P e alence o o e weigh and obesi y in Eu opean child en below he age o 10. In J Obes 2014; 38: S99–107. 43 Wa e s E, de Sil a-Sanigo ski A, Hall BJ, e al. In e en ions o p e en ing obesi y in child en. Coch ane Da abase Sys Re 2011; 12: CD001871. 44 Wang Y, Cai L, Wu Y, e al. Wha childhood obesi y p e en ion p og ammes wo k? A sys ema ic e iew and me a-analysis. Obes Re 2015; 16: 547–65. 45 Sobol-Goldbe g S, Rabinowi z J, G oss R. School-based obesi y p e en ion p og ams: a me a-analysis o andomized con olled ials. Obesi y 2013; 21: 2422–28. 46 Lobs ein T, Jackson-Leach R, Moodie ML, e al. Child and adolescen obesi y: pa o a bigge pic u e. Lance 2015; 385: 2510–20. 47 Go make SL, Wang YC, Long MW, e al. Th ee in e en ions ha educe childhood obesi y a e p ojec ed o sa e mo e han hey cos o implemen . Heal h A (Millwood) 2015; 34: 1932–39. 48 Popkin BM. Rela ionship be ween shi s in ood sys em dynamics and accele a ion o he global nu i ion ansi ion. Nu Re 2017; 75: 73–82. 49 Hawkes C, Ha is JL. An analysis o he con en o ood indus y pledges on ma ke ing o child en. Public Heal h Nu 2011; 14: 1403–14. 50 Bleich SN, Rimm EB, B ownell KD. US nu i ion assis ance, 2018—modi ying SNAP o p omo e popula ion heal h. N Engl J Med 2017; 376: 1205–07. 51 Da mon N, D ewnowski A. Con ibu ion o ood p ices and die cos o socioeconomic dispa i ies in die quali y and heal h: a sys ema ic e iew and analysis. Nu Re 2015; 73: 643–60. 52 Powell LM, Ch iqui JF, Khan T, Wada R, Chaloupka FJ. Assessing he po en ial e ec i eness o ood and be e age axes and subsidies o imp o ing public heal h: a sys ema ic e iew o p ices, demand and body weigh ou comes. Obes Re 2013; 14: 110–28. 53 Die z WH, Bau LA, Hall K, e al. Managemen o obesi y: imp o emen o heal h-ca e aining and sys ems o p e en ion and ca e. Lance 2015; 385: 2521–33. 54 US P e en i e Se ices Task Fo ce. Sc eening o obesi y in child en and adolescen s: US P e en i e Se ices Task Fo ce ecommenda ion s a emen . JAMA 2017; 317: 2417–26.