Effectiveness of food supplements in increasing fat-free tissue accretion in children with moderate acute malnutrition: A randomised 2 × 2 × 3 factorial trial in Burkina Faso

RESEARCH ARTICLE E ec i eness o ood supplemen s in inc easing a - ee issue acc e ion in child en wi h mode a e acu e malnu i ion: A andomised 2 ×2×3 ac o ial ial in Bu kina Faso Ch is ian Fabiansen 1,2 , Cha les W. Yame ´ogo 1,3 , Ann-Sophie Iuel-B ockdo 1,2 , Be na de e Cichon 1,2 , Ma en J. H. Ry e 1 , Anu a Ku pad 4 , Jona han C. Wells 5 , Ch is ian Ri z 1 , Pe Asho n 6 , Suzanne Fil eau 7 , And e ´B iend 1,6 , Susan Shephe d 8 , Vibeke B. Ch is ensen 2,9 , Kim F. Michaelsen 1 , Hen ik F iis 1 * 1Depa men o Nu i ion, Exe cise and Spo s, Uni e si y o Copenhagen, Copenhagen, Denma k, 2Me ´decins Sans F on iè es–Denma k, Copenhagen, Denma k, 3De ´pa emen Biome ´dical e San e ´ Publique Ins i u de Reche che en Sciences de la San e ´, Ouagadougou, Bu kina Faso, 4Di ision o Nu i ion, S John’s Resea ch Ins i u e, Bangalo e, India, 5Childhood Nu i ion Resea ch Cen e, UCL G ea O mond S ee Ins i u e o Child Heal h, London, Uni ed Kingdom, 6Cen e o Child Heal h Resea ch, Uni e si y o Tampe e School o Medicine and Tampe e Uni e si y Hospi al, Tampe e, Finland, 7Facul y o Epidemiology and Popula ion Heal h, London School o Hygiene & T opical Medicine, London, Uni ed Kingdom, 8Alliance o In e na ional Medical Ac ion, Daka , Senegal, 9Depa men o Paedia ics, Righospi ale , Copenhagen, Denma k *[email p o ec ed] Abs ac Backg ound Child en wi h mode a e acu e malnu i ion (MAM) a e ea ed wi h lipid-based nu ien sup- plemen (LNS) o co n-soy blend (CSB). We assessed he e ec i eness o (a) ma ix, i.e., LNS o CSB, (b) soy quali y, i.e., soy isola e (SI) o dehulled soy (DS), and (c) pe cen age o o al p o ein om d y skimmed milk, i.e., 0%, 20%, o 50%, in inc easing a - ee issue acc e ion. Me hods and indings Be ween Sep embe 9, 2013, and Augus 29, 2014, a andomised 2 ×2×3 ac o ial ial ec ui ed 6- o 23-mon h-old child en wi h MAM in Bu kina Faso. The in e en ion comp ised 12 weeks o ood supplemen a ion p o iding 500 kcal/day as LNS o CSB, each con aining SI o DS, and 0%, 20%, o 50% o p o ein om milk. Fa - ee mass (FFM) was assessed by deu e ium dilu ion echnique. By di iding FFM by leng h squa ed, he p ima y ou come was exp essed independen o leng h as FFM index (FFMI) acc e ion o e 12 weeks. O he ou - comes comp ised eco e y a e and addi ional an h opome ic measu es. O 1,609 child en, 4 died, 61 we e los o ollow-up, and 119 we e ans e ed ou due o supplemen a ion being swi ched o non-expe imen al p oduc s. No child en de eloped alle gic eac ion. A inclu- sion, 95% we e b eas ed, mean (SD) weigh was 6.91 kg (0.93), wi h 83.5% (5.5) FFM. In PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 1 / 18 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Ci a ion: Fabiansen C, Yame ´ogo CW, Iuel- B ockdo A-S, Cichon B, Ry e MJH, Ku pad A, e al. (2017) E ec i eness o ood supplemen s in inc easing a - ee issue acc e ion in child en wi h mode a e acu e malnu i ion: A andomised 2 ×2× 3 ac o ial ial in Bu kina Faso. PLoS Med 14(9): e1002387. h ps://doi.o g/10.1371/jou nal. pmed.1002387 Academic Edi o : James K Tumwine, Make e e Uni e si y Medical School, UGANDA Recei ed: Ma ch 10, 2017 Accep ed: Augus 9, 2017 Published: Sep embe 11, 2017 Copy igh : ©2017 Fabiansen e al. This is an open access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal au ho and sou ce a e c edi ed. Da a A ailabili y S a emen : All ele an da a a e wi hin he pape and i s Suppo ing In o ma ion iles. Funding: The s udy was unded by Danish In e na ional De elopmen Assis ance (09-097 LIFE) (KFM); Me ´decins Sans F on iè es (Denma k, No way); A id Nilsson’s Founda ion; The Wo ld Food P og am, which was pa o a dona ion o he Wo ld Food P og am om he Ame ican people he whole coho , weigh inc eased 0.90 kg (95% CI 0.88, 0.93; p<0.01) comp ising 93.5% (95% CI 89.5, 97.3) FFM. As compa ed o child en who ecei ed CSB, FFMI acc e ion was inc eased by 0.083 kg/m 2 (95% CI 0.003, 0.163; p= 0.042) in hose who ecei ed LNS. In con as , SI did no inc ease FFMI compa ed o DS (mean di e ence 0.038 kg/m 2 ; 95% CI −0.041, 0.118; p= 0.35), i espec i e o ma ix. Ha ing 20% milk p o ein was associa ed wi h 0.097 kg/m 2 (95% CI −0.002, 0.196) g ea e FFMI acc e ion han ha ing 0% milk p o- ein, al hough his di e ence was no signi ican (p= 0.055), and he e was no e ec o 50% milk p o ein (0.049 kg/m 2 ; 95% CI −0.047, 0.146; p= 0.32). The e was no e ec modi ica ion by season, admission c i e ia, o baseline FFMI, s un ing, in lamma ion, o b eas eeding (p >0.05). LNS compa ed o CSB esul ed in 128 g (95% CI 67, 190; p<0.01) g ea e weigh gain i bo h con ained SI, bu he e was no di e ence be ween LNS and CSB i bo h con- ained DS (mean di e ence 22 g; 95% CI −40, 84; p= 0.49) (in e ac ion p= 0.017). Acco d- ingly, SI compa ed o DS inc eased weigh by 89 g (95% CI 27, 150; p= 0.005) when combined wi h LNS, bu no when combined wi h CSB. A limi a ion o his and o he ood supplemen a ion ials is ha i is no possible o collec eliable da a on indi idual adhe ence. Conclusions Based on his s udy, child en wi h MAM mainly gain a - ee issue when ehabili a ed. Ne - e heless, LNS yields mo e a - ee issue and highe eco e y a es han CSB. Mo eo e , cu en LNSs wi h DS may be imp o ed by shi ing o SI. The ole o milk ela i e o soy me - i s u he esea ch. T ial egis a ion ISRCTN egis y ISRCTN42569496 Au ho summa y Why was his s udy done? • Mode a e acu e malnu i ion (MAM) is widesp ead among child en in low-income coun ies and is a isk ac o o mo bidi y, se e e acu e malnu i ion, and dea h. • WHO ecommends mo e esea ch on ood supplemen s o child en wi h MAM. Cu - en ly, he e is no consensus on ei he he e ec i eness o lipid-nu ien supplemen (LNS) compa ed o co n-soy blend (CSB), o on he ole o key ac o s like milk and soy. • The e is conce n ha child en ecei ing LNS migh accumula e oo much a issue. Wha did he esea che s do and ind? • In a la ge andomised ial o 1,609 child en wi h MAM, conduc ed in a humani a ian se ing in Bu kina Faso, we in es iga ed he e ec i eness o key ac o s in supplemen al Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 2 / 18 h ough he suppo o he US Agency o In e na ional De elopmen ’s O ice o Food o Peace; he Alliance o In e na ional Medical Ac ion; and he Eu opean Union’s humani a ian aid unds, in pa ne ship wi h Ac ion Con e la Faim. The unde s had no ole in s udy design, da a collec ion and analysis, decision o publish, o p epa a ion o he manusc ip . Compe ing in e es s: I ha e ead he jou nal’s policy and he au ho s o his manusc ip ha e he ollowing compe ing in e es s: KFM has ecei ed esea ch g an s om US Dai y Expo Council and he Danish Dai y Resea ch Founda ion, and also has esea ch collabo a ion wi h Nu ise , a p oduce o LNS p oduc s, and pa en owne ; HF has ecei ed esea ch g an s om ARLA Food o Heal h Cen e, and also has esea ch collabo a ion wi h Nu ise , a p oduce o LNS p oduc s, and pa en owne ; AB was he in en o o LNS, o which Nu ise has he pa en , bu abandoned claims o oyal ies in 2003. O he au ho s decla e no inancial ela ionships wi h any o ganisa ions ha migh ha e an in e es in he submi ed wo k in he p e ious h ee yea s, and decla e no o he ela ionships o ac i i ies ha could appea o ha e in luenced he submi ed wo k. Abb e ia ions: CHW, communi y heal h wo ke ; CSB, co n-soy blend; D 2 O, deu e ium oxide; DS, dehulled soy; DSM, d y skimmed milk; FFM, a - ee mass; FFMI, a - ee mass index; FM, a mass; FMI, a mass index; ITT, in en ion- o- ea ; LNS, lipid-based nu ien supplemen ; MAM, mode a e acu e malnu i ion; MUAC, mid-uppe a m ci cum e ence; PP, pe p o ocol; RUTF, eady- o- use he apeu ic ood; SAM, se e e acu e malnu i ion; SI, soy isola e; TBW, o al body wa e ; WHO, Wo ld Heal h O ganiza ion; WHZ, weigh - o - heigh z-sco e. oods, namely (a) he ma ix (LNS e sus CSB), (b) he soy quali y (soy isola e e sus dehulled soy), and (c) he le el o d y skimmed milk. • We e alua ed body composi ion o assess he quali y o weigh gain, and used a - ee mass issue acc e ion as ou p ima y ou come. • Fa - ee issue acc e ion as well as eco e y om MAM we e bo h highe when using LNS compa ed o CSB. • Child en gi en LNS did no pu on excessi e a . Wha do hese indings mean? • Ou indings suppo a wide use o LNS in he ea men o child en wi h MAM. A swi ch o LNS would lead o g ea e gain o a - ee issue and eco e y, and would ben- e i millions o child en. In oduc ion Mode a e acu e malnu i ion (MAM), de ined as mode a e was ing o low mid-uppe a m ci - cum e ence (MUAC), is widesp ead among child en in low-income coun ies. While bo h p e alence and incidence a e unknown, mode a e was ing alone a ec s 33 million child en a any ime [1]. Child en wi h MAM a e a isk o mo bidi y and mo ali y om in ec ious dis- eases o p og ession o se e e acu e malnu i ion (SAM), while hose who eco e may be a isk o elapse, impai ed physical and cogni i e de elopmen , and la e non-communicable diseases [2,3]. The e is limi ed e idence o in o m ecommenda ions on he composi ion o supplemen- a y oods o ea child en wi h MAM, and he Wo ld Heal h O ganiza ion (WHO) ecom- mends esea ch o de elop cos -e ec i e p oduc s [4]. Supplemen a y oods o malnou ished child en a e based on a ma ix o ei he co n-soy blend (CSB) o lipid-based nu ien supple- men (LNS) [5]. The e a e subs an ial di e ences be ween he 2 p oduc ypes, no only in nu i ional composi ion, bu also in cos , how hey a e consumed, and he logis ics needed o deli e y. A key sou ce o p o ein in CSB is soy, o en dehulled soy (DS), which con ains highe le els o an i-nu ien s (compounds impai ing abso p ion o mine als) compa ed o mo e expensi e soy isola e (SI) [6]. The i s LNS p oduc was de eloped o ea SAM and was based on he nu i ional composi ion o he he apeu ic milk F-100, con aining high amoun s o dai y p oduc s and no soy [7]. This p oduc ype is also e e ed o as eady- o-use he apeu- ic ood (RUTF). Mo e ecen ly, a ange o LNS p oduc s, some con aining soy, has been de el- oped o MAM. The inclusion o d y skimmed milk (DSM) in supplemen s imp o es he amino acid p o ile, p o ides mine als wi h high bioa ailabili y, and educes he con en o an i-nu ien s when i eplaces ege able p o ein sou ces, bu also inc eases cos s [8,9]. T ials assessing he e ec s o supplemen s ypically use weigh gain o nu i ional eco e y as he p ima y ou come, and do no conside body composi ion. While nu i ional eco e y, i.e., inc ease in weigh - o -heigh z-sco e (WHZ) o MUAC abo e speci ic cu o s, is an es ab- lished p og amma ic ou come, he accompanying heal h bene i s a e no well known. Inade- qua e acc e ion o a - ee issue, i.e., muscle, o gans, and bone, impai s body unc ion and Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 3 / 18 heal h [10]. The weigh o a - ee issue, i.e., a - ee mass (FFM), can be measu ed unde ield condi ions using he deu e ium dilu ion echnique and can be exp essed independen o leng h as FFM index (FFMI) by di iding by leng h in me e s squa ed. We aimed o in es iga e he e ec i eness o ma ix, soy quali y, and pe cen age o o al p o- ein om d y skimmed milk in he ea men o MAM in inc easing FFMI acc e ion. Me hods E hics The ial was app o ed by he E hics Commi ee o Heal h Resea ch in Bu kina Faso (2012-8- 059), and consul a i e app o al was ob ained om he Danish Na ional Commi ee on Bio- medical Resea ch E hics (1208204). The ial was egis e ed in he ISRCTN egis y (ISRCTN42569496). In o ma ion and consen o ms we e ansla ed in o he local language (Mo e ´) and back- ansla ed o ensu e accu acy. Ca egi e s ga e e bal and w i en consen (signa u e o inge p in ) p io o en olmen . S udy design We conduc ed a andomised 2 ×2×3 ac o ial ial (T ea ood) assessing he e ec s o he ma ix (LNS e sus CSB), soy quali y (SI e sus DS), and pe cen age o p o ein om DSM (20% and 50% e sus 0%), esul ing in 12 di e en supplemen s, some o which co espond o exis ing p oduc s (Fig 1). Supplemen s we e gi en o 12 weeks. The ial was double-blinded wi h espec o soy quali y and milk con en , bu no ma ix. Pa icipan s This ial was conduc ed in he P o ince du Passo e ´in he no he n egion o Bu kina Faso, in an a ea quali ying o supplemen a y eeding p og ammes based on he p e alence o acu e malnu i ion [11]. The p e alence o mode a e was ing was 9% and o se e e was ing and/o oedema was 1.4% [12]. A 3- o 5-mon h-long ‘lean’ season usually s a s in June [13]. The p e - alence o HIV in ec ion among women 15–49 yea s o age was 0.8% [14]. The amily die was ce eal-based, and consump ion o cow’s milk and o he animal-sou ce oods was a e. S udy si es we e cons uc ed a 5 go e nmen al heal h cen es (Gomponsom, La oden, Baga e ´, Bokin, and Samba) and s a ed by he non-go e nmen al o ganisa ion Alliance o In e na ional Medical Ac ion (ALIMA, Senegal). The s udy ca chmen a ea was de ined by a maximum dis ance o 23 km om a s udy si e, and illages assumed o be inaccessible du ing he ainy season we e excluded. Popula ion size in he s udy ca chmen a ea was es ima ed a 258,000. Child en we e sc eened o MAM in illages by communi y heal h wo ke s (CHWs) o by designa ed mobile sc eening eams. Fu he mo e, child en could be e e ed o a s udy si e om a heal h cen e, o could p esen a a si e on ca egi e ’s ini ia i e. Child en we e ec ui ed i (a) hey we e esiden in he ca chmen a ea a he ime o inclusion, (b) a diagnosis o MAM was con i med, i.e., MUAC 115 mm and <125 mm and/o WHZ –3 and <–2, and (c) hey we e age 6–23 mon hs, based on a heal h ca d o a locally adap ed e en s calenda . Chil- d en we e no included i hey we e (a) ea ed o SAM (WHZ <−3 o MUAC <115 mm o oedema) o hospi alised wi hin he pas 2 mon hs, (b) al eady in a nu i ional p og amme, o (c) equi ing hospi alisa ion, e.g., haemoglobin <50 g/l. Child en wi h a se e e disabili y, lim- i ing he easibili y o in es iga ions, o wi h suspec ed alle gy o milk, peanu s, CSB, o LNS we e also excluded. Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 4 / 18 To minimise mixing o sha ing o expe imen al supplemen s, only he i s child wi h MAM iden i ied om a gi en household was included in he ial. Siblings wi h MAM o wins disco dan o MAM s a us ecei ed he same supplemen bu we e no included in he ial. Randomisa ion Child en we e indi idually assigned a andom o in e en ion g oups. Supplemen s we e des- igna ed by a 1-le e code by he manu ac u e , and a code-key was kep in a sealed en elope in a sa e un il comple ion o da a analysis. Random sequences, in blocks o 12 o 24 and s a i ied by si e, we e c ea ed by a pe son no in ol ed in he ial using Randomiza ion.com (h p:// www. andomiza ion.com). The supplemen s we e packed in indi idual boxes con aining a ull 12-week ea men o 1 pa icipan (ei he 6 bags o CSB o 84 sache s o LNS). Du ing p o- duc ion, each box and bag/sache o supplemen was labelled wi h a 12-le e sequence con- aining he ele an 1-le e code in a ixed posi ion and he 11 emaining le e s in andom o de . In Bu kina Faso, 1 indi idual no in ol ed in ec ui men and da a collec ion was awa e o he andom sequences and he 1-le e code. This indi idual elabelled boxes and supple- men s wi h indi idual s udy IDs, and all la e handling o supplemen s a he 5 s udy si es was based on s udy ID alone. A en olmen , child en we e gi en a s udy ID by s a wi hou access o he andom sequences o supplemen s. Nu i ional in e en ion All expe imen al supplemen s complied wi h WHO’s echnical no e on supplemen a y oods o he managemen o MAM [4]. All supplemen s con ained 500 kcal pe daily se ing (120 g o CSB o 92 g o LNS). The a con en was 11.4–11.7 g pe daily se ing in CSBs and 31.4– 32.1 g pe daily se ing in LNSs. The p o ein con en was 15.9–16.8 g in CSBs and 12.5–13.5 g in LNSs. All supplemen s had a p o ein diges ibili y—co ec ed amino acid sco e (PDCAAS) abo e 70%, as ecommended by WHO [4]. The shel li e o he CSBs was 9 mon hs, and o he LNSs was 2 yea s. The ing edien s in he 12 expe imen al supplemen s a e shown in Table 1. Soy and mal odex in eplaced milk in milk- ee p oduc s. Mo e mal odex in was added when SI eplaced DS. Fig 1. Expe imen al ood supplemen s. The 2 ×2×3 ac o ial design, showing he 12 expe imen al ood supplemen s based on co n-soy blends o lipid-based nu ien supplemen s, wi h ei he dehulled soy o soy isola e, and wi h 0%, 20%, o 50% o o al p o ein om milk. Two supplemen s co espond o cu en ly used p oduc s: “a” (CSB+) and “b” (CSB++). P oduc “i” is simila o Plumpy’Sup, con aining dehulled soy bu wi h d y skimmed milk ins ead o whey. h ps://doi.o g/10.1371/jou nal.pmed.1002387.g001 Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 5 / 18 The 6 LNSs had simila packaging and ex u e, and he 6 CSBs had simila packaging and ex u e. LNSs we e p o ided in sache s, each con aining a daily se ing eady o consump ion. CSBs we e p o ided in bags o 1.7 kg con aining a o nigh ly a ion (120 g/child/day ecom- mended o be di ided in 3 meals, 40 g/meal) o be cooked wi h wa e and consumed as a po - idge. Supplemen s we e manu ac u ed by GC Riebe Compac (No way), which was o he wise no in ol ed in he ial design o in e p e a ion o da a. The same supplie s o he main aw ma e ial, DS, SI, and milk we e used o bo h he LNSs and CSBs. The inal expe i- men al supplemen s we e mic obiologically es ed and app o ed o ood sa e y by ce i ied labs (Compac India, India; TU ¨V SU ¨D, India; Do dogne Labs, F ance). S udy isi s Child en we e scheduled o clinic isi s e e y 2 weeks du ing he 12-week supplemen a ion pe iod. FFM was assessed a en olmen and a e 12 weeks. Clinical s a us and an h opome y we e assessed a e e y isi , excep leng h, which was measu ed once a mon h. Child en who missed scheduled isi s we e isi ed by a CHW a hei homes and encou aged o e u n o ollow-up. To assess FFM, a dose o 5 g o deu e ium oxide (D 2 O) (99.8%; Camb idge Iso ope Labo a- o ies, US) dilu ed in 5 g o bo led wa e was weighed wi h 0.01-g p ecision and gi en o ally a e collec ion o p e-dose sali a samples. Pos -dose sali a samples we e collec ed a e a 3-hou equilib a ion pe iod, as es ablished du ing he pilo s udy [15]. Fo each assessmen , D 2 O abundance was measu ed in duplica e in p e- and pos -dose sali a samples and in a dilu ed sample o each child’s dose, using Fou ie - ans o m in a ed spec ome y (Agilen Technologies, US) [16] a S . John’s Resea ch Ins i u e, Bangalo e, India. A leas 60 μl o sali a was equi ed o analysis. Deu e ium dilu ion space was calcula ed as desc ibed p e iously [17], and con e ed o o al body wa e (TBW) using a ac o o 1.044 o adjus o p o on exchange. FFM was calcula ed as TBW/hyd a ion, using age- and sex-speci ic hyd a ion coe i- cien s [18]. Fa mass (FM) was calcula ed as weigh minus FFM. Da a we e cleaned o ypo- g aphical e o s and implausible TBW alues, based on he associa ion o TBW wi h leng h and cu o s o FM o <−0.1 ( o accoun o he no mal echnical a iabili y in deu e ium Table 1. Supplemen composi ion. Ing edien Supplemen a b c d e g h i j k l Co n 55.0 55.0 55.0 55.0 55.0 55.0 Peanu 27.0 27.0 27.0 27.0 27.0 27.0 Suga 9.0 9.0 9.0 9.0 9.0 9.0 27.0 27.0 27.0 27.0 27.0 27.0 Mal odex in 5.6 3.7 0.3 12.55 8.4 2.5 7.8 5.55 1.9 13.47 9.05 2.34 Skimmed milk powde 8.0 20.0 8.0 20.0 8.0 20.0 8.0 20.0 Vege able a 7.9 7.8 7.8 8.0 7.9 7.7 18.0 18.0 18.0 18.0 18.0 18.0 Soy p o ein isola e 11.4 8.1 3.2 8.6 5.5 0.9 Soy lou 19.7 14.0 5.6 15.0 9.5 1.5 Magnesium ci a e 0.5 0.5 0.6 0.85 0.8 0.6 0.2 0.2 0.25 0.33 0.3 0.26 Calcium phospha e 1.4 1.1 0.8 1.5 1.3 0.9 1.95 1.7 1.3 2.1 1.8 1.4 Po assium chlo ide 0.5 0.5 0.5 1.3 1.1 0.7 0.9 0.9 0.9 1.35 1.2 0.95 Vi amin p emix 0.4 0.4 0.4 0.4 0.4 0.4 0.5 0.5 0.5 0.5 0.5 0.5 Emulsi ie 1.65 1.65 1.65 1.65 1.65 1.65 Ing edien s gi en as pe cen age o o al weigh . h ps://doi.o g/10.1371/jou nal.pmed.1002387. 001 Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 6 / 18 dilu ion s udies) and >2.4 kg. FFM and FM we e di ided by leng h in me e s squa ed o de i e indices independen o leng h, i.e., FFM index (FFMI) and FM index (FMI) [19]. Weigh was measu ed in duplica e o he nea es 100 g using elec onic scales (Seca model 881 1021659) wi h double weighing unc ion. Leng h was measu ed in duplica e wi h a wooden heigh boa d o he nea es 1 mm (sub ac ing a co ec ion ac o o 0.7 cm o chil- d en eaching an age o >24 mon hs du ing he in e en ion). WHZ was de e mined a he s udy si es using WHO ield ables, and his alue was used o ec ui men . In la e analysis, WHZ was ecalcula ed using he package “zsco e06” in STATA 12 (S a aCo p, US). MUAC was measu ed in duplica e o he nea es 1 mm, a he midpoin be ween he olec anon and he ac omion p ocess using a s anda d measu ing ape. T iceps skin old was measu ed in duplica e using a Ha penden callipe . The mean o he duplica e measu emen s was used o analysis. Knee-heel leng h was measu ed using a digi al callipe (Mi u oyo, Ge many), moun ed wi h knee and heel caps cas in ha d plas ic [20]. The ins umen has a esolu ion o 0.01 mm. The esul was exp essed as he mean o 5 consecu i e sepa a e measu emen s. An h opome ic measu emen s we e unde aken by ained s a , a e s anda disa ion sessions. A en olmen , a esea ch nu se collec ed da a on demog aphic cha ac e is ics, accina ion s a us, and 2-week e ospec i e mo bidi y and medical ea men s using a s uc u ed ques- ionnai e in he local language. All child en ecei ed i amin A supplemen a ion (100,000 IU i 4–8 kg; 200,000 IU i >8 kg) i hey had no ecei ed any supplemen s in he p e ious 6 mon hs, and albendazole (200 mg i 4–8 kg; 400 mg i >8 kg) and accina ions we e adminis- e ed acco ding o he na ional schedule a he heal h cen es. Medical ea men was p o ided based on In eg a ed Managemen o Childhood Illness guidelines. Venous blood was collec ed om he a m and es ed o Plasmodium alcipa um by apid es (SD Bioline Mala ia Ag P. ; S anda d Diagnos ics, Ko ea). Se um C- eac i e p o ein was de e mined using a combined sandwich enzyme-linked immunoso ben assay (Vi Min Lab, Ge many). Nu i ional eco e y was ca ego ised in o 3 g oups: eco e y a 12 weeks (WHZ −2 and MUAC 125 mm), MAM (non- esponse) a 12 weeks, o SAM a any poin du ing he in e - en ion. Child en de eloping SAM we e discon inued om expe imen al supplemen s, ea ed wi h RUTF, and ans e ed o inpa ien ea men i accompanied by medical complica ions. Child en who did no eco e a e 12 weeks o supplemen a ion ecei ed 2 weeks o RUTF. Child en who did no gain weigh du ing his pe iod we e examined by esea ch nu ses and i necessa y e e ed o hospi al. All child en who had no eco e ed om MAM a e 2 weeks o RUTF supplemen a ion ecei ed i o a u he 2 weeks. Child en who ailed o eco e a e he o al 4 weeks o RUTF supplemen a ion we e e e ed o hospi al o a mo e in-dep h clinical examina ion. Du ing he ial, due o suspec ed con amina ion o 1 ba ch o expe imen al supplemen , he comme cially a ailable LNS Plumpy’Sup (Nu ise , F ance) was gi en o some child en. Ou comes The p ima y ou come was FFMI inc emen measu ed using he deu e ium dilu ion echnique. D 2 O en ichmen was de e mined in a single labo a o y, and cleaning o da a was blinded. O he ou comes included inc emen s in FFM, FM, FMI, weigh , leng h, knee-heel leng h, MUAC, iceps skin old, and nu i ional eco e y. Se ious ad e se e en s we e no expec ed om in ake o he ood supplemen s, bu we e de ined as anaphylac ic eac ion and dea h. The e ec s we e exp essed as LNS compa ed o CSB, SI compa ed o DS, and 20% and 50% milk p o ein compa ed o 0% milk p o ein. Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 7 / 18 S a is ical analyses To de ec a 0.6-SD di e ence be ween any 2 combina ions o he 3 ac o s wi h 80% powe and 5% signi icance, while allowing o 20% loss o ollow-up, we aimed o ec ui in o al 1,608 (= 12 ×134) child en. Da a we e double-en e ed in EpiDa a 3.1 (Denma k). S a is ical analyses we e ca ied ou using STATA 12. In he absence o an unsupplemen ed g oup, baseline alues we e eg essed on age o es ima e a e age age-associa ed weigh and FFM acc e ion o e a pe iod equi alen o he leng h o he in e en ion. Desc ip i e s a is ics a e p esen ed as mean (SD). Mean changes in an h opome y and body composi ion a e gi en wi h 95% con idence in e als. The analysis was based on he in en ion- o- ea (ITT) p inciple using a ailable case da a. The 2 ×2×3 ac o ial design was analysed using a linea mixed model wi h si e-speci ic an- dom e ec s. Model educ ion was ca ied ou o educe he 3-way in e ac ion o 2-way in e ac- ions o main e ec s (using likelihood a io es s). Speci ically, he ull model including he 3-way in e ac ion co esponding o he 2 ×2×3 ac o ial design was compa ed o he model con aining only he main e ec s o he 3 ac o s; hus, we es ed o any kind o in e ac ion in he ac o ial layou in one go. Pai wise compa isons o means we e pe o med using model- based pos hoc es s in he educed models. The e ec s o supplemen a ion a e p esen ed in e ms o es ima ed means wi h 95% CIs. All analyses included adjus men o baseline measu e o he ou come [21], age, sex, MUAC, WHZ, and mon h o admission wi h he excep ion o unadjus ed ITT analyses, which included only baseline measu e o ou come and si e. A pe p o ocol (PP) analysis, including only child en who comple ed 12 weeks o supplemen a ion wi h he supplemen o which hey we e o iginally alloca ed, was also conduc ed. E ec modi ica ion was assessed o season, admission c i e ia, and baseline FFMI, s un - ing, in lamma ion, and b eas eeding. Model checking was based on esiduals and no mal p obabili y plo s. Reco e y was e alua ed by means o chi-squa ed es s. A signi icance le el o 0.05 was applied. Compa ed o he s a is ical plan ou lined in he p o ocol (S1 Tex ), we de ia ed on 2 issues. Fi s , we adjus ed FFM o leng h by di iding i by leng h in me e s squa ed, i.e., exp essing FFM acc e ion as an index, o be able o adjus o no only baseline leng h bu also changes in leng h o e ime. Second, we ini ially planned o adjus o mul iplici y, in case we had o com- pa e 12 g oups. Howe e , in he educed ac o ial model wi h pai wise compa isons wi hin ac o s, no mul iplici y adjus men s we e applied. Resul s F om Sep embe 9, 2013, o Augus 29, 2014, we assessed 3,398 child en, o whom 1,967 (58%) had MAM. O hese, 1,613 (82%) child en we e andomised, bu 4 excluded as ineligible. Hence, a o al o 1,609 child en we e andomised acco ding o he 2 ×2×3 ac o ial design (Fig 2). As shown in Table 2, he andomisa ion esul ed in baseline equi alence be ween ca ego ies o he expe imen al ac o s wi h espec o po en ial con ounde s. Du ing he in e en ion, 61 (3.8%) child en we e los o ollow-up, lea ing 1,548 (96.2%) o ITT analyses. Fou child en died and no child de eloped an alle gic eac ion o supple- men s. The in e en ion was discon inued be o e 12 weeks o 119 child en: 17 child en we e supplemen ed wi h Plumpy’Sup due o uncon i med suspicion o salmonella con amina ion o hei expe imen al supplemen , while 102 child en de e io a ing in o SAM we e swi ched o he apeu ic oods. Thus, he PP analyses included 1,429 (88.8%) child en. Baseline weigh mean (SD) was 6.91 kg (0.93). Weigh a baseline comp ised 83.5% (5.5) FFM. We es ima ed, based on eg ession o baseline da a on age, ha weigh would inc ease by Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 8 / 18 Fig 2. T ial p o ile. In all, 3,398 child en we e assessed o eligibili y a he 5 s udy si es, and 1,609 child en we e andomised and included in he s udy. CSB, co n-soy blend; D20, deu e ium oxide measu emen ; DS, dehulled soy; ITT, in en ion- o- ea ; LNS, lipid-based nu ien supplemen ; PP, pe p o ocol; PPS/Salmonella, ecei ed Plumpy’Sup due o suspicion o salmonella con amina ion; PPN/SAM, ecei ed Plumpy’Nu due o se e e acu e malnu i ion; SI, soy isola e. h ps://doi.o g/10.1371/jou nal.pmed.1002387.g002 Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 9 / 18 W i ing – e iew & edi ing: Ch is ian Fabiansen, Cha les W. Yame ´ogo, Ann-Sophie Iuel- B ockdo , Be na de e Cichon, Ma en J. H. Ry e , Anu a Ku pad, Jona han C. Wells, Ch is ian Ri z, Pe Asho n, Suzanne Fil eau, And e ´B iend, Susan Shephe d, Vibeke B. Ch is ensen, Kim F. Michaelsen, Hen ik F iis. Re e ences 1. Black RE, Vic o a CG, Walke SP, Bhu a ZA, Ch is ian P, de Onis M, e al. Ma e nal and child unde nu- i ion and o e weigh in low-income and middle-income coun ies. Lance . 2013; 382:427–51. h ps:// doi.o g/10.1016/S0140-6736(13)60937-X PMID: 23746772 2. Black RE, Allen LH, Bhu a ZA, Caul ield LE, de Onis M, Ezza i M, e al. Ma e nal and child unde nu i- ion: global and egional exposu es and heal h consequences. Lance . 2008; 371:243–60. h ps://doi. o g/10.1016/S0140-6736(07)61690-0 PMID: 18207566 3. Vic o a CG, Adai L, Fall C, Hallal PC, Ma o ell R, Rich e L, e al. Ma e nal and child unde nu i ion: consequences o adul heal h and human capi al. Lance . 2008; 371:340–57. h ps://doi.o g/10.1016/ S0140-6736(07)61692-4 PMID: 18206223 4. Wo ld Heal h O ganiza ion. Supplemen a y oods o he managemen o mode a e acu e malnu i ion in in an s and child en 6–59 mon hs o age. Gene a: Wo ld Heal h O ganiza ion; 2012 [ci ed 2017 May 22]. h p://apps.who.in /i is/bi s eam/10665/75836/1/9789241504423_eng.pd . 5. de Pee S, Bloem MW. Cu en and po en ial ole o specially o mula ed oods and ood supplemen s o p e en ing malnu i ion among 6- o 23-mon h-old child en and o ea ing mode a e malnu i ion among 6- o 59-mon h-old child en. Food Nu Bull. 2009; 30:S434–63. h ps://doi.o g/10.1177/ 15648265090303S305 PMID: 19998866 6. Michaelsen KF, Hoppe C, Roos N, Kaes el P, S ougaa d M, Lau i zen L, e al. Choice o oods and ing edien s o mode a ely malnou ished child en 6 mon hs o 5 yea s o age. Food Nu Bull. 2009; 30: S343–404. h ps://doi.o g/10.1177/15648265090303S303 PMID: 19998864 7. Wo ld Heal h O ganiza ion. Managemen o se e e malnu i ion: a manual o physicians and o he senio heal h wo ke s. Gene a: Wo ld Heal h O ganiza ion; 1999 [ci ed 2017 May 22]. h p://apps.who. in /i is/bi s eam/10665/41999/1/a57361.pd ?ua=1&ua=1. 8. Hoppe C, Ande sen GS, Jacobsen S, Mølgaa d C, F iis H, Sangild PT, e al. he use o whey o skimmed milk powde in o i ied blended oods o ulne able g oups. J Nu . 2008; 138:145S–61S. PMID: 18156417 9. Hoppe C, Mølgaa d C, Michaelsen KF. Cow’s milk and linea g ow h in indus ialized and de eloping coun ies. Annu Re Nu . 2006; 26:131–73. h ps://doi.o g/10.1146/annu e .nu .26.010506.103757 PMID: 16848703 10. F iis H, Michaelsen KF, Wells JC. Choice o design and ou comes in ials among child en wi h mode - a e acu e malnu i ion. Food Nu Bull. 2015; 36:S35–40. h ps://doi.o g/10.1177/ 15648265150361S106 PMID: 25902612 11. Wo ld Heal h O ganiza ion. The managemen o nu i ion in majo eme gencies. Gene a: Wo ld Heal h O ganiza ion; 2000 [ci ed 2017 May 22]. h p://www.who.in /nu i ion/publica ions/eme gencies/ 9241545208/en/. 12. Bu kina Faso Minis è e de la San e ´Di ec ion de la Nu i ion. Rappo Enquê e Nu i ionnelle Na ionale (Na ional Nu i ion Su ey Repo ). Ouagadougou: Bu kina Faso Minis è e de la San e ´; 2013. 13. Famine Ea ly Wa ning Sys ems Ne wo k. Wes A ica: Bu kina Faso—seasonal calenda : ypical yea . 2013 [ci ed 2017 May 22]. h p://www. ews.ne /wes -a ica/bu kina- aso/seasonal-calenda /decembe - 2013. 14. Ins i u Na ional de la S a is ique e de la De ´mog aphie. Enquê e De ´mog aphique e de San e ´e àIndi- ca eu s Mul iples (EDSBF-MICS IV) 2010. Ouagadougou: Ins i u Na ional de la S a is ique e de la De ´mog aphie; 2012 [ci ed 2017 May 22]. h ps://www.unice .o g/b a/ ench/b _eds_2010.pd . 15. Fabiansen C, Yame ´ogo CW, De i S, F iis H, Ku pad A, Wells JC. Deu e ium dilu ion echnique o body composi ion assessmen : esol ing me hodological issues in child en wi h mode a e acu e malnu i ion. Iso opes En i on Heal h S ud. 2017; 53:344–55. h ps://doi.o g/10.1080/10256016.2017.1295043 PMID: 28276731 16. In e na ional A omic Ene gy Agency. In oduc ion o body composi ion assessmen using he deu e ium dilu ion echnique wi h analysis o sali a samples by Fou ie ans o m in a ed spec ome y. Vienna: In e na ional A omic Ene gy Agency; 2010 [ci ed 2017 May 22]. h p://www-pub.iaea.o g/MTCD/ Publica ions/PDF/Pub1450_web.pd . 17. Da ies PS, Wells JC. Calcula ion o o al body wa e in in ancy. Eu J Clin Nu . 1994; 48:490–5. PMID: 7956990 Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 16 / 18 18. Lohman TG. Ad ances in body composi ion assessmen . Cu en Issues in Exe cise Science Se ies, Monog aph 3. Champaign (Illinois): Human Kine ics Publishe s; 1992. 19. VanI allie TB, Yang MU, Heyms ield SB, Funk RC, Boileau RA. Heigh -no malized indices o he body’s a - ee mass and a mass: po en ially use ul indica o s o nu i ional s a us. Am J Clin Nu . 1990; 52:953–9. PMID: 2239792 20. Skau JKH, Touch B, Chhoun C, Chea M, Unni US, Maku a J, e al. E ec s o animal sou ce ood and mic onu ien o i ica ion in complemen a y ood p oduc s on body composi ion, i on s a us, and linea g ow h: a andomized ial in Cambodia. Am J Clin Nu . 2015; 101:742–51. h ps://doi.o g/10.3945/ ajcn.114.084889 PMID: 25833972 21. Ba ne AG. Reg ession o he mean: wha i is and how o deal wi h i . In J Epidemiol. 2004; 34:215– 20. h ps://doi.o g/10.1093/ije/dyh299 PMID: 15333621 22. Wo ld Heal h O ganiza ion. WHO child g ow h s anda ds: leng h/heigh - o -age, weigh - o -age, weigh - o -leng h, weigh - o -heigh and body mass index- o -age: me hods and de elopmen . Gene a: Wo ld Heal h O ganiza ion; 2006. 23. WHO, UNICEF, WFP and UNHCR Consul a ion on he p og amma ic aspec s o he managemen o mode a e acu e malnu i ion in child en unde i e yea s o age. Gene a: Wo ld Heal h O ganiza ion; 2010 [ci ed 2017 May 22]. h p://www.who.in /nu i ion/ opics/mode a emalnu i ion_consul a ion_ p og amma icaspec s_MM_ epo .pd . 24. Ma ilsky DK, Male a K, Cas leman T, Mana y MJ. Supplemen a y eeding wi h o i ied sp eads esul s in highe eco e y a es han wi h a co n/soy blend in mode a ely was ed child en. J Nu . 2009; 139:773–8. h ps://doi.o g/10.3945/jn.108.104018 PMID: 19225128 25. Nacke s F, B oille F, Ouma ou D, Djibo A, Gaboulaud V, Gue in PJ, e al. E ec i eness o eady- o- use he apeu ic ood compa ed o a co n/soy-blend-based p e-mix o he ea men o childhood mod- e a e acu e malnu i ion in Nige . J T op Pedia . 2010; 56:407–13. h ps://doi.o g/10.1093/ opej/ mq019 PMID: 20332221 26. Lag one LN, T ehan I, Meuli GJ, Wang RJ, Thakwalakwa C, Male a K, e al. A no el o i ied blended lou , co n-soy blend “plus-plus”, is no in e io o lipid-based eady- o-use supplemen a y oods o he ea men o mode a e acu e malnu i ion in Malawian child en. Am J Clin Nu . 2012; 95:212–9. h ps:// doi.o g/10.3945/ajcn.111.022525 PMID: 22170366 27. Nikièma L, Huyb eg s L, Kols e en P, Lanou H, Tiend ebeogo S, Bouckae K, e al. T ea ing mode a e acu e malnu i ion in i s -line heal h se ices: an e ec i eness clus e - andomized ial in Bu kina Faso. Am J Clin Nu . 2014; 100:241–9. h ps://doi.o g/10.3945/ajcn.113.072538 PMID: 24808482 28. Ka akochuk C, an den B iel T, S ephens D, Zlo kin S. T ea men o mode a e acu e malnu i ion wi h eady- o-use supplemen a y ood esul s in highe o e all eco e y a es compa ed wi h a co n-soya blend in child en in sou he n E hiopia: an ope a ions esea ch ial. Am J Clin Nu . 2012; 96:911–6. h ps://doi.o g/10.3945/ajcn.111.029744 PMID: 22952175 29. Acka ia-A mah RS, McDonald CM, Doumbia S, E ha d JG, Hame DH, B own KH. Malian child en wi h mode a e acu e malnu i ion who a e ea ed wi h lipid-based die a y supplemen s ha e g ea e weigh gains and eco e y a es han hose ea ed wi h locally p oduced ce eal-legume p oduc s: a commu- ni y-based, clus e - andomized ial. Am J Clin Nu . 2015; 101:632–45. h ps://doi.o g/10.3945/ajcn. 113.069807 PMID: 25733649 30. Medoua GN, N sama PM, Ndzana ACA, Essa’a VJ, Tsa ack JJT, Dimodi HT. Reco e y a e o child en wi h mode a e acu e malnu i ion ea ed wi h eady- o-use supplemen a y ood (RUSF) o imp o ed co n—soya blend (CSB+): a andomized con olled ial. Public Heal h Nu . 2016; 19:363–70. h ps:// doi.o g/10.1017/S1368980015001238 PMID: 25939394 31. S obaugh HC, Ryan KN, Kennedy JA, G ise JB, C ocke AH, Thakwalakwa C, e al. Including whey p o- ein and whey pe mea e in eady- o-use supplemen a y ood imp o es eco e y a es in child en wi h mode a e acu e malnu i ion: a andomized, double-blind clinical ial. Am J Clin Nu . 2016; 103:926– 33. h ps://doi.o g/10.3945/ajcn.115.124636 PMID: 26864368 32. Fomon SJ, Haschke F, Ziegle EE, Nelson SE. Body composi ion o e e ence child en om bi h o age 10 yea s. Am J Clin Nu . 1982; 35:1169–75. PMID: 7081099 33. Ba z S, Mody A, Ho nik C, Bain J, Muehlbaue M, Kiyimba T, e al. Se e e acu e malnu i ion in child- hood: ho monal and me abolic s a us a p esen a ion, esponse o ea men , and p edic o s o mo ali y. J Clin Endoc inol Me ab. 2014; 99:2128–37. h ps://doi.o g/10.1210/jc.2013-4018 PMID: 24606092 34. B ooke OG, Wheele EF. High ene gy eeding in p o ein-ene gy malnu i ion. A ch Dis Child. 1976; 51:968–71. PMID: 827979 35. Fjeld CR, Schoelle DA, B own KH. Body composi ion o child en eco e ing om se e e p o ein-ene gy malnu i ion a wo a es o ca ch-up g ow h. Am J Clin Nu . 1989; 50:1266–75. PMID: 2512801 Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 17 / 18 36. Radhak ishna KV, Kulka ni B, Balak ishna N, Rajkuma H, Omka C, Sha ugna V. Composi ion o weigh gain du ing nu i ion ehabili a ion o se e ely unde nou ished child en in a hospi al based s udy om India. Asia Pac J Clin Nu . 2010; 19:8–13. PMID: 20199982 37. Golden BE, Golden MH. Plasma zinc, a e o weigh gain, and he ene gy cos o issue deposi ion in child en eco e ing om se e e malnu i ion on a cow’s milk o soya p o ein based die . Am J Clin Nu . 1981; 34:892–9. PMID: 7195146 38. Golden MH. P oposed ecommended nu ien densi ies o mode a ely malnou ished child en. Food Nu Bull. 2009; 30:S267–342. h ps://doi.o g/10.1177/15648265090303S302 PMID: 19998863 39. Bahwe e P, Balaluka B, Wells JC, Mbi ibindi CN, Sadle K, Akomo P, e al. Ce eals and pulse-based eady- o-use he apeu ic ood as an al e na i e o he s anda d milk- and peanu pas e-based o mula- ion o ea ing se e e acu e malnu i ion: a nonin e io i y, indi idually andomized con olled e icacy clinical ial. Am J Clin Nu . 2016; 103:1145–61. h ps://doi.o g/10.3945/ajcn.115.119537 PMID: 26984485 40. Iuel-B ockdo A-S, D æbel TA, Fabiansen C, Cichon B, Ch is ensen VB, Yameogo C, e al. Accep abil- i y o new o mula ions o co n-soy blends and lipid-based nu ien supplemen s in P o ince du Passo e ´, Bu kina Faso. Appe i e. 2015; 91:278–86. h ps://doi.o g/10.1016/j.appe .2015.04.058 PMID: 25913687 41. Iuel-B ockdo A-S, D aebel TA, Ri z C, Fabiansen C, Cichon B, B ix Ch is ensen V, e al. E alua ion o he accep abili y o imp o ed supplemen a y oods o he ea men o mode a e acu e malnu i ion in Bu kina Faso using a mixed me hod app oach. Appe i e. 2016; 99:34–45. h ps://doi.o g/10.1016/j. appe .2015.12.030 PMID: 26752599 42. Iuel-B ockdo A-S, Oued aogo A, Ri z C, D aebel TA, Asho n P, Fil eau S, e al. Feeding beha io s du - ing home-based ea men o mode a e acu e malnu i ion using co n-soy blends o lipid-based nu ien supplemen s. Ma e n Child Nu . 2016 Dec 2. h ps://doi.o g/10.1111/mcn.12399 PMID: 27910283 43. UNICEF Supply Ca alogue. P oduc s: nu i ion—supplemen a y ood. 2016 [ci ed 2016 Oc obe 26]. h ps://supply.unice .o g/. 44. Maus A, Ko oma AS, Abla C, Molokwu N, Ryan KN, Singh L, e al. Se e e and mode a e acu e malnu- i ion can be success ully managed wi h an in eg a ed p o ocol in Sie a Leone. J Nu . 2015; 145:2604–9. h ps://doi.o g/10.3945/jn.115.214957 PMID: 26423737 Food supplemen s in mode a e acu e malnu i ion PLOS Medicine | h ps://doi.o g/10.1371/jou nal.pmed.1002387 Sep embe 11, 2017 18 / 18