Effect of diet and physical activity based interventions in pregnancy on gestational weight gain and pregnancy outcomes: meta-analysis of individual participant data from randomised trials

he bmj | BMJ 2017;358:j3119 | doi: 10.1136/bmj.j3119 1 RESEARCH E ec o die and physical ac i i y based in e en ions in p egnancy on ges a ional weigh gain and p egnancy ou comes: me a-analysis o indi idual pa icipan da a om andomised ials The In e na ional Weigh Managemen in P egnancy (i-WIP) Collabo a i e G oup ABSTRACT Objec i e To syn hesise he e idence on he o e all and di e en ial e ec s o in e en ions based on die and physical ac i i y du ing p egnancy, p ima ily on ges a ional weigh gain and ma e nal and o sp ing composi e ou comes, acco ding o women’s body mass index, age, pa i y, e hnici y, and p e-exis ing medical condi ion; and seconda ily on indi idual complica ions. DESIGN Sys ema ic e iew and me a-analysis o indi idual pa icipan da a (IPD). DATA SOURCES Majo elec onic da abases om incep ion o Feb ua y 2017 wi hou language es ic ions. ELIGIBILITY CRITERIA FOR SELECTING STUDIES Randomised ials on die and physical ac i i y based in e en ions in p egnancy. DATA SYNTHESIS S a is ical models accoun ed o clus e ing o pa icipan s wi hin ials and he e ogenei y ac oss ials leading o summa y mean di e ences o odds a ios wi h 95% con idence in e als o he e ec s o e all, and in subg oups (in e ac ions). RESULTS IPD we e ob ained om 36 andomised ials (12 526 women). Less weigh gain occu ed in he in e en ion g oup han con ol g oup (mean di e ence −0.70 kg, 95% con idence in e al −0.92 o −0.48 kg, I2=14.1%; 33 s udies, 9320 women). Al hough summa y e ec es ima es a ou ed he in e en ion, he educ ions in ma e nal (odds a io 0.90, 95% con idence in e al 0.79 o 1.03, I2=26.7%; 24 s udies, 8852 women) and o sp ing (0.94, 0.83 o 1.08, I2=0%; 18 s udies, 7981 women) composi e ou comes we e no s a is ically signi ican . No e idence was ound o di e en ial in e en ion e ec s ac oss subg oups, o ei he ges a ional weigh gain o composi e ou comes. The e was s ong e idence ha in e en ions educed he odds o caesa ean sec ion (0.91, 0.83 o 0.99, I2=0%; 32 s udies, 11 410 women), bu no o o he indi idual complica ions in IPD me a-analysis. When IPD we e supplemen ed wi h s udy le el da a om s udies ha did no p o ide IPD, he o e all e ec was simila , wi h s onge e idence o bene i o ges a ional diabe es (0.76, 0.65 o 0.89, I2=36.8%; 59 s udies, 16 885 women). CONCLUSION Die and physical ac i i y based in e en ions du ing p egnancy educe ges a ional weigh gain and lowe he odds o caesa ean sec ion. The e is no e idence ha e ec s di e ac oss subg oups o women. In oduc ion Hal o all women o childbea ing age wo ldwide a e o e weigh o obese.1-3 Obesi y and excessi e ges a ional weigh gain pu mo he and o sp ing a isk, bo h in p egnancy and in la e li e.4-6 The esul an cos s o he heal h se ice and socie y a e conside able.7 8 Inc easingly, heal hca e o ganisa ions and esea ch unding bodies p io i ise esea ch on in e en ions and s a egies o educe ma e nal weigh ela ed ad e se ou comes in p egnancy.9-12 Syn heses o s udy le el da a on e ec s o die and physical ac i i y based in e en ions in p egnancy13 ha e shown an o e all bene i on limi ing ges a ional weigh gain, bu he indings a ied o hei p o ec i e e ec on ma e nal and o sp ing ou comes.13 14 Impo an ly, he subg oups o women who may bene i he mos om such in e en ions a e no known.15 Fo his, p ima y s udies do no ha e su icien powe ,16 17 and me a-analyses o s udy le el da a a e limi ed by he absence o published de ails o subg oup e ec s,18 and by po en ial ecological bias.19 These p oblems can be add essed by e idence syn hesis using aw indi idual le el da a om ele an s udies.20 21 We unde ook an indi idual pa icipan da a (IPD) me a-analysis o assess he e ec s o die and physical ac i i y based in e en ions, p ima ily on ges a ional weigh gain and on ma e nal and o sp ing composi e ou comes, in subg oups de ined by body mass index (BMI), age, pa i y, e hnici y, and p e-exis ing medical Co espondence o: K S Khan k.s.kh[email p o ec ed] Addi ional ma e ial is published online only. To iew please isi he jou nal online. Ci e his as: BMJ 2017;358:j3119 h p://dx.doi.o g/10.1136/bmj.j3119 Accep ed: 15 June 2017 WHAT IS ALREADY KNOWN ON THIS TOPIC Inc eased weigh gain in p egnancy is associa ed wi h ma e nal and e al complica ions In e en ions based on die o physical ac i i y o bo h in p egnancy minimise ges a ional weigh gain In e en ions based on die and physical ac i i y may ha e a po en ial ole in p e en ing ad e se p egnancy ou comes WHAT THIS STUDY ADDS Die and physical ac i i y based in e en ions consis en ly educe ges a ional weigh gain ac oss a ious subg oups o women ca ego ised by age, pa i y, body mass index, e hnici y, and p e-exis ing medical condi ion The educ ion in odds o ad e se ma e nal and o sp ing composi e ou comes wi h die and physical ac i i y is no signi ican , and does no a y ac oss a ious subg oups o women In e en ions signi ican ly lowe he odds o caesa ean sec ion and ha e no e ec on o sp ing ou comes RESEARCH 2 doi: 10.1136/bmj.j3119 | BMJ 2017;358:j3119 | he bmj condi ion. Fu he mo e, we assessed he o e all e ec s, and hose o indi idual in e en ions (die , physical ac i i y, mixed), on c i ically impo an ma e nal and o sp ing complica ions. In addi ion o using IPD, we also assessed he impac o inco po a ing s udy le el da a om o he s udies no p o iding IPD. Me hods The IPD me a-analysis was pe o med using a p especi ied p o ocol (PROSPERO CRD42013003804)22 and was epo ed in line wi h ecommenda ions o he P e e ed Repo ing I ems o Sys ema ic e iews and Me a-Analysis o Indi idual Pa icipan Da a (PRISMA- IPD).23 Li e a u e sea ch and s udy iden i ica ion We sea ched he majo elec onic da abases Medline, Embase, Coch ane Da abase o Sys ema ic Re iews, Da abase o Abs ac s o Re iews o E ec s, Coch ane Cen al Regis e o Con olled T ials, and Heal h Technology Assessmen Da abase om Oc obe 2013 o Ma ch 2015 o upda e ou p e ious sea ch in his opic o andomised ials on die and physical ac i i y based in e en ions in p egnancy.13 The sea ch was u he upda ed in Janua y 2016 and Feb ua y 2017 o iden i y new s udies. We sea ched he in e ne by using gene al sea ch engines, and con ac ed esea che s in he special y o iden i y ele an ials. The e we e no language es ic ions. Web appendix 1 p o ides de ails o he sea ch s a egy. Two independen esea che s (ER and NM, AAM, o EM) selec ed s udies in a wo s age p ocess. In he i s s age, po en ial ci a ions we e iden i ied. Nex , we did a de ailed e alua ion o he ull manusc ip s o po en ial pape s and selec ed a icles ha ul illed he eligibili y c i e ia. We included andomised ials ha assessed he e ec s o in e en ions based on die , physical ac i i y, and mixed in e en ions in p egnancy, on ma e nal and o sp ing ou comes. We classi ied complex in e en ions on die and physical ac i i y, including hose wi h beha iou al change componen s, as mixed in e en ions. We excluded s udies ha only included women wi h ges a ional diabe es a baseline, in ol ed animals, epo ed only non-clinical ou comes, and we e published be o e 1990. The p ima y ou comes we e ges a ional weigh gain, a composi e o ma e nal ou comes, and a composi e o o sp ing ou comes. The seconda y ou comes we e indi idual ma e nal and o sp ing complica ions. The componen s o he composi e ou comes we e de e mined by a wo ound Delphi su ey o esea che s in his special y, and we e conside ed o be c i ically impo an o clinical p ac ice.24 The ma e nal composi e ou come included ges a ional diabe es melli us, hype ensi e diso de s o p egnancy, p e e m deli e y, and caesa ean sec ion. The o sp ing composi e ou come included s illbi h, small o ges a ional age e us, la ge o ges a ional age e us, and admission o he newbo n o a neona al in ensi e ca e uni . We de ined ges a ional weigh gain as he di e ence be ween ma e nal weigh a an ena al booking and he las weigh measu ed be o e deli e y. We accep ed he p ima y au ho s’ de ini ion and epo ing o ges a ional diabe es melli us, p egnancy induced hype ension, p e-eclampsia, caesa ean sec ion, s illbi h, and admission o a neona al in ensi e ca e uni . We de ined p e e m deli e y as bi h be o e 37 weeks o ges a ion, and small o ges a ional age and la ge o ges a ional age as babies wi h a bi h weigh below he 10 h and a o o e he 90 h cen iles, espec i ely, adjus ed o mo he ’s BMI, pa i y, and ges a ional age a deli e y.25 Es ablishmen o IPD collabo a i e ne wo k and da abase—We es ablished he In e na ional Weigh Managemen in P egnancy IPD Collabo a i e G oup by con ac ing esea che s o eligible s udies.26 A bespoke da abase was de eloped, and we eques ed collabo a o s o ele an da a in any o ma . We sen h ee eminde s when he e was no esponse. Quali y assessmen o he included s udies Two independen e iewe s assessed he quali y o he andomised ials using a isk o bias ool o sequence gene a ion, alloca ion concealmen , blinding, incomple e ou come da a, selec i e ou come epo ing, and o he po en ial sou ces o bias.27 We conside ed a s udy o ha e a high isk o bias i i sco ed as such in a leas one o he ollowing domains: andomisa ion, alloca ion concealmen , blinding o ou come assessmen , o incomple e ou come da a; all i ems should be sco ed as low isk o a s udy o be classi ied as low isk o bias. Da a ex ac ion and assessmen o IPD in eg i y Two independen e iewe s (ER and NM) unde ook da a ex ac ion a s udy le el o inclusion and exclusion c i e ia, he cha ac e is ics o he in e en ion, and he epo ed ou comes. We sough o ob ain IPD om ele an s udies published un il July 2015, which was he endpoin o IPD acquisi ion, o allow su icien ime o da a cleaning, s anda disa ion, and amalgama ion o da ase s. We also ex ac ed he published s udy le el da a o all ele an s udies published un il Feb ua y 2017, including hose published beyond he indi idual da a acquisi ion imeline, and hose o which IPD we e no p o ided by s udy au ho s. We ob ained IPD o indi idual ma e nal cha ac e is ics ha we e de e mined a p io i, such as BMI, age, pa i y, e hnici y, socioeconomic s a us, and p e-exis ing medical condi ions. Con inuous a iables we e kep con inuous, bu some we e also ca ego ised when conside ed o be clinically use ul. These included ca ego isa ions based on BMI (no mal 18.5-24.9 kg/m2, o e weigh 25-29.9 kg/m2, obese ≥30 kg/m2) and age (cu -o 20 yea s). Mo he ’s e hnici y was classi ied as whi e o non-whi e. We used he mo he ’s educa ional s a us o indica e socioeconomic s a us: low s a us i he mo he did no comple e seconda y educa ion o A le el, medium i she comple ed seconda y educa ion (A le el equi alen ), and high i she comple ed any u he highe educa ion. We de ined p e-exis ing medical condi ions as diabe es melli us, ea ly onse o ges a ional diabe es, o hype ension. We conside ed pa icipan s o be adhe en o he in e en ion based on he ollowing c i e ia: RESEARCH he bmj | BMJ 2017;358:j3119 | doi: 10.1136/bmj.j3119 3 comple ion o a leas 70% o he in e en ion p o ocol, da ase p o ided in o ma ion on adhe ence in a yes o no o ma , o pa icipan was deemed o be adhe en as pe he s udy c i e ia. We pe o med ange and consis ency checks on all IPD and p oduced summa y ables. The andomisa ion a io, baseline cha ac e is ics, and me hod o analysis in he IPD da ase we e compa ed wi h he published in o ma ion. Any disc epancies, missing da a, ob ious e o s, and inconsis encies be ween a iables o ou lying alues we e que ied and ec i ied as necessa y wi h inpu om he o iginal au ho s. Da a syn hesis To ob ain summa y es ima es (mean di e ence o ges a ional weigh gain and odds a ios o bina y ou comes) and 95% con idence in e als o he in e en ion e ec s o each p ima y ou come we unde ook a wo s age IPD me a-analysis.21 We assessed he e ec s ac oss all in e en ions o e all and o indi idual in e en ions. A wo s age IPD me a- analysis was used o ob ain summa y es ima es o he subg oup e ec s (in e ac ions) o in e es , which compa ed di e en ial e ec s o in e en ions ac oss he p ima y ou comes. Addi ionally we e alua ed whe he he e a e any di e en ial e ec s o in e en ions o indi idual complica ions, acco ding o BMI (no mal, o e weigh , obese). All analyses we e designed o p ese e he in en ion o ea p inciple. The i s s age o he wo s age me a-analysis in ol ed analysing he IPD in each ial sepa a ely, o accoun o he clus e ing o pa icipan s wi hin ials, and o ob ain he es ima es o in e es and hei a iances. Fo he clus e andomised ials, we included a andom in e cep o a uni o andomisa ion o accoun o his u he clus e ing. Fo he ou come o ges a ional weigh gain, we used analysis o co a iance in each ial o eg ess he inal weigh alue agains he in e en ion while adjus ing o baseline weigh and cen es in clus e andomised ials. Fo ma e nal and o sp ing ou comes, we used a logis ic eg ession model o each ial sepa a ely, wi h he in e en ion as a co a ia e. We excluded women wi h con i med glucose in ole ance o a hype ensi e diso de a baseline, as de ined by he p ima y au ho s, in he analysis o composi e ad e se p egnancy ou comes. To assess po en ial in e en ion e ec modi ie s, we ex ended he a o emen ioned models o include in e ac ion e ms be ween pa icipan le el co a ia es and he in e en ion (ie, ea men -co a ia e in e ac ion e ms). In he second s age, we pooled he de i ed e ec es ima es (ie, ea men e ec s o ea men -co a ia e in e ac ions) ac oss ials using a andom e ec s model i ed using es ic ed maximum likelihood. The andom e ec s app oach allowed us o accoun o unexplained in e s udy he e ogenei y in e ec s ac oss s udies. This p oduced summa y es ima es and 95% con idence in e als o he in e en ion e ec s and he in e ac ions (subg oup e ec s). The Ha ung-Knapp co ec ion was applied when subsequen ly de i ing 95% con idence in e als o he ue mean e ec , o help accoun o he unce ain y o he es ima e o in e s udy he e ogenei y.28 29 We included s udies ha did no con ibu e IPD, by inco po a ing hei ex ac ed s udy le el da a wi hin he second s age o he IPD me a-analysis amewo k, o ob ain summa y es ima es o in e en ion e ec s ha combined IPD and non-IPD s udies. Sensi i i y analyses we e also pe o med by excluding s udies wi h high isk o bias, analysing he p ima y ou comes sepa a ely o each in e en ion ype (die , physical ac i i y, and mixed), excluding pa icipan s no adhe en o he in e en ion, by analysing change in BMI ins ead o weigh gain, and excluding ma e nal weigh gain es ima es om p egnancies ha ended be o e 37 comple ed weeks o ges a ion o a oid sys ema ic di e ences. He e ogenei y was summa ised using he I2 s a is ic, he es ima ed in e s udy a iance (τ2),30 and app oxima e 95% p edic ion in e als, which indica e he po en ial in e en ion (o in e ac ion) e ec in a new popula ion simila o hose included in he me a- analysis.31 Small s udy e ec s (po en ial publica ion bias) we e in es iga ed by using con ou enhanced unnel plo s alongside isual examina ion and s a is ical es s o asymme y (Egge ’s es o con inuous ou comes o Pe e ’s es o bina y ou comes).32 We assessed o IPD a ailabili y bias by compa ing he summa y esul s when including non-IPD s udies wi h hose om IPD s udies.33 Fu he mo e, we compa ed he symme y o unnel plo s be o e and a e inclusion o non-IPD s udies. All me a-analyses we e unde aken using S a a so wa e e sion 12.1 (S a aCo p, College S a ion, TX, USA), and s a is ical signi icance was conside ed a he 5% le el. Pa ien in ol emen No pa ien s we e in ol ed in se ing he esea ch ques ion o he ou come measu es, no we e hey in ol ed in de eloping plans o ec ui men , design, o implemen a ion o he s udy. A pa ien ep esen a i e p o ided an inpu o he in e p e a ion and w i ing up o esul s. The e a e no plans o dissemina e he esul s o he esea ch o s udy pa icipan s o he ele an pa ien communi y. I was no e alua ed whe he he s udies included in he e iew had any pa ien in ol emen . Resul s S udy selec ion We iden i ied 58 ials published up o June 2015, o which 36 s udies (62%) p o ided indi idual pa icipan da a (IPD),16 17 34-66 ha accoun ed o da a om 80% o he pa icipan s (12 526/15 541); 22 s udies (3015 women) did no p o ide IPD ( ig 1).67-88 A u he 45 ials (9945 women)89-133 we e iden i ied a e he IPD acquisi ion imeline un il Feb ua y 2017. Cha ac e is ics o included s udies and pa icipan s IPD we e a ailable om 36 ials in 16 coun ies. Twen y wo s udies17 34 36-39 41 42 47 48 51-53 56-63 67 we e RESEARCH 4 doi: 10.1136/bmj.j3119 | BMJ 2017;358:j3119 | he bmj om Eu ope, ou each om No h Ame ica,4454 65 66 Aus alia,16 43 45 50 and B azil,35 49 55 64 and one s udy each om Egyp 40 and I an.46 Twen y h ee IPD s udies included women o any body mass index (BMI), 34-38 42 44-48 52 54-56 58-61 64-67 se en included only obese women,1739-41 50 62 63 and six included obese and o e weigh women.16 43 49 51 53 57 The in e en ions included hose mainly based on die ( ou IPD s udies)47 61 62 64 o physical ac i i y (16 IPD s udies),35-37 42 46 49-52 55 58 59 65 66 69 and hose based on a mixed app oach o die , physical ac i i y, o beha iou modi ying echniques, o all h ee oge he (15 IPD s udies).16 17 34 39-41 43-45 48 53 54 56 60 63 One s udy had a h ee a m design wi h in e en ion a ms being physical ac i i y only and a mixed app oach.57 The web appendix p o ides he cha ac e is ics o all IPD s udies, and also hose ha did no con ibu e IPD. Eligible s udies iden i ied beyond da a acquisi ion imeline (n=45; 9945 women) S udies o which IPD we e sough (n=58) S udies o which IPD da a we e no a ailable (n=67; 12 960 women) Ges a ional weigh gain (n=48; 8210 women) Ma e nal ou comes Ges a ional diabe es (n=32; 8033 women) Hype ensi e diso de o p egnancy (n=23; 5231 women) P e e m bi h (n=17; 2663 women) Caesa ean sec ion (n=34; 6631 women) O sp ing ou comes S illbi h (n=2; 815 women) Small o ges a ional age (n=11; 1271 women) La ge o ges a ional age (n=11; 1301 women) Admission o NICU (n=5; 1358 women) S udies ha p o ided IPD (n=36; 12 526 women) Ci a ions om p e ious sys ema ic e iew (incep ion o Jan 2012) and o he sou ces† (n=105) Ci a ions om elec onic da abases sea ch om Jan 2012 o Feb 2017* (n=11 815) Reco ds a ailable a e duplica es emo ed (n=7038) Full ex s udies assessed o eligibili y (n=218) Eligible s udies (n=103) S udies wi h IPD and wi hou lPD a ailabili y Ma e nal ou comes: ges a ional weigh gain (n=81; 17 530 women), ges a ional diabe es (n=59; 16 1885 women), hype ensi e disease (n=45; 14 849 women), p e e m bi h (n=49; 14 339 women), caesa ean sec ion (n=66; 18 041 women) O sp ing ou comes: sS illbi h (n=4; 4534 women), small o ges a ional age (n=44; 12 937 women), la ge o ges a ionaI age (n=45; 13 348 women), admission o NICU (n=21; 9498 women) Ci a ions excluded (n=6820) Excluded (n=115): Published be o e 1990 (n=4) Pilo ial inco po a ed in o main ial (n=2) P o ocol (ongoing ec ui men ) (n=37) I ele an s udy objec i e o in e en ion (n=15) Ac i e compa ison (n=28) No obs e ic ou comes (n=11) W ong s udy popula ion (n=14) Full ex o abs ac no a ailable (n=4) S udies ha did no p o ide IPD (n=22): Con lic o in e es (n=2; 610 women ) Lack o ime (n=2; 286 women) Da a sha ing issue (n=1; 132 women) Da a loss (n=2; 421 women) Con ac loss (n=4; 531 women) No esponse (n=11; 1035 women) Ges a ional weigh gain (n=33; 9320 women) Ma e nal composi e (n=24; 8852 women) Ges a ional diabe es (n=27; 9427 women) Hype ensi e diso de o p egnancy (n=22; 9618 women) P e e m bi h (n=32; 11 676 women) Caesa ean sec ion (n=32; 11 410 women) O sp ing composi e (n=18; 7981 women) S illbi h (n=2; 3719 women) Small o ges a ional age (n=33; 11 666 women) La ge o ges a ional age (n=34; 12 047 women) Admission o NICU (n=16; 8140 women) NICU=neona al in ensi e ca e uni *Da abase sea ch was upda ed in Oc obe 2013 (9359 eco ds), Ma ch 2015 (3551 eco ds), Janua y 2016 (1373 eco ds), and Feb ua y 2017 (1547 eco ds) †O he sou ces: e e ence sea ch, pe sonal communica ion, and Google sea ch Fig1 | Iden i ica ion and selec ion o s udies in indi idual pa icipan da a (IPD) me a-analysis o die and physical ac i i y based in e en ions on p egnancy ou comes a e ges a ional weigh gain RESEARCH he bmj | BMJ 2017;358:j3119 | doi: 10.1136/bmj.j3119 5 Mo e han 80% o women in he IPD me a-analyses we e o whi e o igin, and a leas hal we e classi ied as o high socioeconomic s a us. A ound 45% o women we e nullipa ous, 40% we e obese, and a simila p opo ion was classi ied as ha ing seden a y s a us wi h no exe cise a baseline ( able 1). IPD we e a ailable o assess he e ec s o in e en ions on ges a ional weigh gain (33 s udies, 9320 women), ma e nal composi e ou comes (24 s udies, 8852 women), and o sp ing composi e ou comes (18 s udies, 7981 women). The la ges IPD we e a ailable o he ou come o la ge o ges a ional age e us (34 s udies, 12 047 women), ollowed by p e e m deli e y (32 s udies, 11 676 women), small o ges a ional age e us (33 s udies, 11 666 women), any caesa ean sec ion (32 s udies, 11 410 women), hype ensi e diso de s o p egnancy (22 s udies, 9618 women), and ges a ional diabe es (27 s udies, 9427 women). We did no ha e access o IPD o 51% o all eligible women (12 960/25 486) om 67 s udies ( ig 1). Quali y o included s udies O e all, ials had a low isk o bias in andom sequence gene a ion (71%, 73/103). Mo e han 90% (34/36) o s udies ha con ibu ed o he IPD we e assessed as low isk o bias in his domain compa ed wi h 58% (28/67) o he non-IPD s udies. Two IPD s udies (2/36) and one non-IPD s udy (3/67) we e conside ed high isk o alloca ion concealmen . Blinding o ou come assessmen was app op ia e in 44% (16/36) o IPD and 33% (22/67) o non-IPD s udies ( ig 2). Fewe IPD s udies (5/36) we e assessed as high isk o bias o incomple e ou come da a han non-IPD s udies (15/67). Figu e 2 shows he summa y o he isk o bias es ima es o all eligible s udies and hose ha did and did no con ibu e o IPD. We did no encoun e any issues ha we we e no able o cla i y wi h he IPD con ibu o du ing he IPD in eg i y check. Table1 | Baseline cha ac e is ics o women included in s udies ha con ibu ed o he me a-analysis o indi idual pa icipan da a on die and physical ac i i y based in e en ions in p egnancy. Values a e numbe s (pe cen ages*) unless s a ed o he wise Cha ac e is ics No o s udies (No o women) In e en ion Con ol Mean (SD) age (yea s) 35 (12 006) 30.0 (5.1) 30.1 (5.2) Weigh (body mass index): 34 (12 031) No mal (18.5-24.9 kg/m2)1974 (31.7) 1842 (31.8) O e weigh (25-29.9 kg/m2)1578 (25.3) 1523 (26.3) Obese (≥30 kg/m2)2680 (43.0) 2434 (42.0) Race/e hnici y: 27 (10 020) Whi e (including Russians and Aus alians) 4562 (88.0) 4217 (87.2) Asian 157 (3.0) 156 (3.2) Black 292 (5.6) 292 (6.0) Cen al and Sou h Ame ican 67 (1.3) 64 (1.3) Middle Eas e n (including I anian and Tu kish) 37 (0.7) 37 (0.8) O he 71 (1.4) 68 (1.4) Educa ional s a us o mo he †: 29 (8914) Low 722 (15.6) 724 (16.9) Medium 1372 (29.6) 1292 (30.2) High 2536 (54.8) 2268 (52.9) Smoke 29 (10 958) 875 (15.4) 865 (16.4) Pa i y: 33 (11 805) 0 3027 (49.5) 2692 (47.3) 1 2136 (34.9) 2083 (36.6) 2 647 (10.6) 634 (11.1) 3 179 (2.9) 165 (2.9) ≥4 129 (2.1) 113 (2) No exe cise o seden a y 27 (7583) 1761 (44.6) 1731 (47.6) P e-exis ing diabe es melli us 25 (9589) 6 (0.1) 9 (0.2) P e-exis ing hype ension 23 (5494) 73 (2.5) 54 (2.1) *P opo ion ou o obse a ions in con ol o in e en ion a ms, espec i ely. †Low=no comple ed seconda y educa ion o A le el; medium=comple ed seconda y educa ion (A le el equi alen ); high=any u he o highe educa ion. No o s udies assessed o isk o bias All eligible s udies IPD Non-IPD All eligible s udies IPD Non-IPD All eligible s udies IPD Non-IPD All eligible s udies IPD Non-IPD All eligible s udies IPD Non-IPD All eligible s udies IPD Non-IPD 0 20 40 60 80 100 Random sequence gene a ion High Unclea Low 4 26 73 2 34 4 24 39 49 23 26 60 17 43 38 16 22 71 28 42 5 49 3 11 2 38 43 19 24 5 49 14 6 2 43 20 12 5 3 15 10 73 23 50 20 11 7 6 5 12 Alloca ion concealmen Blinding o pa icipan s and s a Blinding o ou come assessmen Incomple e ou come da a Selec ing epo ing Fig2 | Assessmen o isk o bias in all eligible s udies (n=103), s udies wi h indi idual pa icipan da a (IPD) (n=36), and s udies wi hou access o IPD (n=67) RESEARCH 6 doi: 10.1136/bmj.j3119 | BMJ 2017;358:j3119 | he bmj E ec s o in e en ions on p egnancy ou comes Ges a ional weigh gain Based on IPD me a-analysis (33 s udies, 9320 women), die and physical ac i i y based in e en ions esul ed in signi ican ly less ges a ional weigh gain compa ed wi h con ol (summa y mean di e ence −0.70 kg, 95% con idence in e al −0.92 o −0.48 kg, I2=14.1%), a e adjus ing o baseline weigh and clus e ing. The app oxima e 95% p edic ion in e al o he in e en ion e ec in a new se ing was −1.24 o −0.16 kg ( able 2). Di e en ial e ec s in subg oups No s ong e idence was ound o a ea men -co a ia e in e ac ion o baseline BMI when ea ed as a con inuous co a ia e (−0.02 kg change in in e en ion e ec pe one uni inc ease in BMI, 95% con idence in e al −0.08 o 0.04 kg), o when compa ed as o e weigh e sus no mal (−0.11 kg, −0.77 o 0.55 kg), obese e sus no mal (0.06 kg, −0.90 o 1.01 kg), and obese e sus o e weigh (−0.09 kg, −1.05 o 0.86 kg). We also did no obse e e idence o a subg oup e ec o age (−0.03 kg pe one yea inc ease in age, 95% con idence in e al −0.08 o 0.02 kg), pa i y (0.10 kg change in e ec o mul ipa i y e sus nullipa i y, 95% con idence in e al −0.39 o 0.60 kg), e hnici y (0.05 kg change in e ec o non-whi e e sus whi e, 95% con idence in e al −1.27 o 1.37 kg), and unde lying medical condi ion (1.51 kg change in e ec o women wi h a leas one condi ion e sus none, 95% con idence in e al −2.01 o 5.02 kg). The indings we e consis en when con inuous co a ia es we e analysed as ca ego ical measu es based on clinically ele an cu poin s ( able 3). Sensi i i y analyses The educ ion in ges a ional weigh gain owing o he in e en ion was consis en ly obse ed when he analysis was es ic ed o s udies wi h low isk o bias (−0.67 kg, 95% con idence in e al −0.95 o −0.38 kg; 15 s udies, 5585 women), women adhe en o he in e en ion (−0.76 kg, −1.00 o −0.52 kg; 33 s udies, 8565 women), women ollowed up un il mo e han 37 weeks’ ges a ion (−0.91 kg, −1.17 o −0.66 kg; 28 s udies, 5324 women), and o BMI ins ead o ma e nal weigh as an ou come (−0.30 kg/m2, −0.39 o −0.21 kg/m2; 31 s udies, 9238 women). Table2 | E ec s o die and physical ac i i y based in e en ions on ges a ional weigh gain summa ised using indi idual pa icipan da a (IPD) alone, and by supplemen ing IPD wi h s udy le el da a om s udies ha did no con ibu e IPD. Values a e means (s anda d de ia ions) unless s a ed o he wise Ou comes No o s udies (No o women) In e en ion Con ol Mean di e ence (95% CI) I2 (%) IPD IPD and non-IPD IPD IPD and non-IPD IPD IPD and non-IPD IPD IPD and non-IPD IPD IPD and non-IPD O e all 33 (9320) 81 (17 530) 10.1 (5.4) 10.6* 10.8 (5.4) 11.5* −0.70 (−0.92 o −0.48) −1.10 (−1.46 o −0.74) 14.1 73.8 Die 4 (1168) 12 (2017) 10.2 (4.4) 9.2* 11.0 (4.8) 11.7* −0.72 (−1.48 o 0.04) −2.84 (−4.77 o −0.91) 0.0 92.3 Physical ac i i y 15 (2915) 37 (7355) 9.8 (4.4) 11.3* 10.8 (4.8) 11.9* −0.73 (−1.11 o −0.34) −0.72 (−1.04 o −0.41) 0.0 45.4 Mixed app oach 15 (5369) 35 (8448) 10.2 (6.0) 10.3* 10.6 (5.9) 11.0* −0.71 (−1.10 o −0.31) −1.00 (−1.39 o −0.61) 34.9 54.6 *Recalcula ion using De Simonian-Lai d. Table3 | Di e en ial e ec s o die and physical ac i i y based in e en ions on ges a ional weigh gain in subg oups o p egnan women Ma e nal cha ac e is ics No o s udies (No o women) Mean di e ence* kg (95% CI) T ea men co a ia e in e ac ion Coe icien ; 95% CI (95% PI) I2 (%) Baseline body mass index: No mal 21 (3376) −0.77 (−1.15 o −0.39) −0.02; −0.08 o 0.04 (−0.21 o 0.17)† 39.8 O e weigh 28 (2574) −0.75 (−1.22 o −0.27) Obese 31 (3335) −0.85 (−1.41 o −0.29) Pa i y: Nullipa ous 27 (4513) −0.80 (−1.17 o −0.43) 0.10; −0.39 o 0.60 (−0.83 o 1.04)‡ 4.8 Mul ipa ous 27 (4548) −0.62 (−0.88 o −0.37) E hnici y: Whi e 21 (6814) −0.74 (−1.07 o −0.42) 0.05; −1.27 o 1.37 (−1.28 o 1.39)§ 26.1 Non-whi e 15 (621) −0.42 (−1.12 o 0.28) Age (yea s): ≥20 32 (9045) −0.72 (−0.95 o −0.50) −0.03; −0.08 o 0.02 (−0.14 o 0.09)¶ 25.9 <20 13 (232) 0.05 (−1.34 o 1.44) P e-exis ing medical condi ions**: None 18 (4335) −0.62 (−0.90 o −0.34) 1.51; −2.01 o 5.02 (−4.13 o 7.15)†† 28.4 ≥1 6 (128) 0.40 (−1.92 o 2.71) PI=p edic ion in e al. *Model accoun ed o baseline weigh and clus e ing e ec . †Pe uni o body mass index. ‡Mul ipa ous e sus nullipa ous. §Non-whi e e sus whi e. ¶Pe yea o age. **Diabe es melli us o hype ension. †† ≥1 medical condi ion e sus none. RESEARCH he bmj | BMJ 2017;358:j3119 | doi: 10.1136/bmj.j3119 7 Addi ion o s udies ha did no con ibu e IPD In me a-analysis unde aken by supplemen ing he IPD wi h s udy le el da a om s udies (48 s udies, 8210 women) ha did no con ibu e IPD, we obse ed a la ge bene icial in e en ion e ec o weigh gain (summa y mean di e ence −1.1 kg; 95% con idence in e al −1.46 o −0.74 kg; 81 s udies, 17 530 women). The bene i was also consis en ly obse ed o indi idual in e en ions based on die , physical ac i i y, o mixed app oach ( able 2). Ma e nal and o sp ing composi e ou comes In he IPD me a-analyses, he summa y es ima es a ou ed he in e en ion g oup o educ ion in odds o ma e nal (odds a io 0.90, 95% con idence in e al 0.79 o 1.03, I2=26.7%; 24 s udies, 8851 women) and o sp ing composi e ou comes (0.94, 0.83 o 1.08, I2=0%; 18 s udies, 7981 women), bu hese we e no s a is ically signi ican ( able 4). Di e en ial e ec s ac oss subg oups We obse ed no s ong e idence o di e en ial subg oup e ec s o ma e nal composi e ou come acco ding o ei he baseline BMI ( ea men -co a ia e in e ac ion 1.00, 95% con idence in e al 0.98 o 1.02), age (1.01, 0.99 o 1.03), pa i y (1.03, 0.75 o 1.39), e hnici y (0.93, 0.63 o 1.37), and unde lying medical condi ion (1.44, 0.15 o 13.74) ( able 5). A simila lack o di e en ial e ec was obse ed o o sp ing composi e ou come in mo he s g ouped acco ding o baseline BMI (in e ac ion 0.98, 95% con idence in e al 0.95 o 1.00), age (1.01, 0.98 o 1.04), pa i y (0.94, 0.64 o 1.37), e hnici y (1.12, 0.75 o 1.68), and unde lying medical condi ion (0.58, 0.03 o 9.81) ( able 4). The indings did no change o ma e nal and o sp ing composi e ou comes when BMI and age we e analysed as con inuous ins ead o ca ego ical a iables. Indi idual ma e nal ou comes O e all, in he IPD me a-analysis we obse ed a signi ican educ ion in caesa ean sec ion (odds a io 0.91, 95% con idence in e al 0.83 o 0.99, I2=0%; 32 s udies, 11 410 women) o in e en ions compa ed wi h ou ine ca e. The educ ion in o he indi idual ou comes such as ges a ional diabe es (0.89, 0.72 o 1.10, I2=23.8%; 27 s udies, 9427 women), hype ensi e diso de s o p egnancy (0.95, 0.78 o 1.16, I2=24.2%; 22 s udies, 9618 women), and p e e m deli e y (0.94, 0.78 o 1.13, I2=17.3%; 32 s udies, 11 676 women) we e no s a is ically signi ican in IPD me a-analyses ( able 5). We did no obse e any di e en ial e ec acco ding o baseline BMI ca ego y (no mal, o e weigh , obese) o any o he indi idual ma e nal ou comes (see web appendix 3). The indings we e consis en when s udy le el da a om non-IPD s udies we e me a-analysed wi h IPD, bu wi h a s onge e idence o bene i o ges a ional diabe es. The educ ion in ges a ional diabe es (0.76, 0.65 o 0.89, 36.8%; 59 s udies, 16 885 women) became signi ican ( able 5). Among indi idual in e en ions, hose based mainly on physical ac i i y showed a educ ion in ges a ional diabe es in bo h IPD (odds a io 0.67, 95% con idence in e al 0.46 o 0.99, I2=0%; 10 s udies, 2700 women) and in combined (IPD and non-IPD) me a-analyses (0.66, 0.53 o 0.83, I2=0%; 27 s udies, 6755 women). While he summa y es ima es o physical ac i i y based in e en ions a ou ed caesa ean sec ion (0.82, 0.67 o 1.01, I2=0%; 13 s udies, 3046 women) and hype ensi e diso de s o p egnancy (0.74, 0.42 o 1.33, I2=6.0%; 7 s udies, 2565 women) in IPD me a- analyses, he addi ion o non-IPD s udies esul ed in s onge e idence o bene i o hese complica ions, wi h educ ion in he espec i e odds by 17% (0.83, 0.73 o 0.95, I2=0%; 32 s udies, 6587 women) and 32% (0.68, 0.49 o 0.93, I2=0%; 20 s udies, 5125 women). A s ong e ec was obse ed o p e e m bi h wi h die based in e en ions in bo h IPD (odds a io 0.28, 95% con idence in e al 0.08 o 0.96, I2=0%; 4 s udies, 1344 women) and combined analyses (0.32, 0.14 o 0.70, I2=0%; 7 s udies, 1696 women), bu he o e all sample sizes we e ela i ely small ( able 5). The e was no e idence o bene i wi h mixed in e en ions o any ma e nal ou comes. Indi idual o sp ing ou comes No s ong e idence was ound ha in e en ions had an e ec on indi idual o sp ing ou comes such as s illbi h (odds a io 0.81, 95% con idence in e al <0.001 o 256.69, I2=0%; 2 s udies, 3719 women), small o ges a ional age e us (1.06, 0.94 o 1.20, I2=0%; 33 s udies, 11 666 women), la ge o ges a ional age e us (0.90, 0.76 o 1.07, I2=38.0%; 34 s udies, 12 047 women), and admission o a neona al in ensi e ca e uni (1.01, 0.84 o 1.23, I2=0%; 16 s udies, 8140 women) based on he IPD me a-analyses. The signi icance o he indings did no change when non-IPD s udies we e added o he IPD me a-analyses ( able 5). The numbe s o eligible pa icipan s o whom da a we e ob ained, e ec es ima es, and con idence in e als o all abo e analyses a e a ailable om he s udy au ho s on eques . The e was no di e en ial e ec o any indi idual o sp ing ou come acco ding o he BMI ca ego y (see web appendix 3). Small s udy e ec s We ound isual and s a is ical e idence (Egge ’s es P=0.04) o small s udy e ec s in he con ou enhanced unnel plo s o he IPD me a-analysis o he o e all e ec on ges a ional weigh gain. The asymme y o he plo was no imp o ed by he addi ion o s udy le el da a om non-IPD s udies o he me a-analysis. When s udies wi h high isk o bias we e excluded om he analysis, he symme y o he unnel plo imp o ed (Egge ’s es P=0.61). We ound signi ican e idence o small s udy e ec s o he ma e nal composi e ou come (Pe e ’s es P=0.04), bu no o he o sp ing composi e ou come (P=0.85) (see web appendix 4). RESEARCH 8 doi: 10.1136/bmj.j3119 | BMJ 2017;358:j3119 | he bmj Discussion Ou la ge, collabo a i e indi idual pa icipan da a (IPD) me a-analysis con i ms ha die and physical ac i i y based in e en ions in p egnancy educe ges a ional weigh gain. This bene icial e ec was consis en ly obse ed i espec i e o ma e nal body mass index (BMI), age, pa i y, e hnici y, o p e-exis ing medical condi ion; and emained when s udies a high isk o bias we e excluded. The indings a e gene alisable, wi h he 95% Table4 | E ec s o die and physical ac i i y based in e en ions on p egnancy ou comes summa ised using indi idual pa icipan da a (IPD) alone, and by supplemen ing IPD wi h s udy le el da a om s udies ha did no con ibu e IPD Ou comes No o s udies (No o women) In e en ion: e en /No e en Con ol: e en /No e en Odds a io (95% CI) I2 (%) IPD IPD and non-IPD IPD IPD and non-IPD IPD IPD and non-IPD IPD IPD and non-IPD IPD IPD and non-IPD Ma e nal Composi e ou come: O e all 24 (8851) NA 1896/2728 NA 1837/2390 NA 0.90 (0.79 o 1.03) NA 26.7 NA Die 3 (397) NA 42/137 NA 84/134 NA 0.60 (0.20 o 1.75) NA 0.0 NA Physical ac i i y 9 (2311) NA 346/850 NA 367/748 NA 0.81 (0.61 o 1.09) NA 10.8 NA Mixed app oach 13 (6259) NA 1508/1742 NA 1438/3009 NA 0.97 (0.84 o 1.12) NA 34.9 NA Ges a ional diabe es: O e all 27 (9427) 59 (16 885) 584/4333 974/7764 571/3939 1046/7101 0.89 (0.72 o 1.10) 0.76 (0.65 o 0.89) 23.8 36.8 Die 4 (490) 8 (1106) 13/208 57/476 19/250 75/498 1.03 (0.30 o 3.61) 0.79 (0.37 o 1.69) 0.0 0.0 Physical ac i i y 10 (2700) 27 (6755) 90/1300 240/3153 121/1189 347/3015 0.67 (0.46 o 0.99) 0.66 (0.53 o 0.83) 0.0 0.0 Mixed app oach 14 (6355) 27 (9342) 481/2825 677/4135 441/2608 672/3858 1.02 (0.79 o 1.32) 0.88 (0.72 o 1.07) 35.2 10.8 Hype ensi e diso de s o p egnancy: O e all 22 (9618) 45 (14 849) 432/4586 559/7130 423/4177 592/6568 0.95 (0.78 o 1.16) 0.85 (0.71 o 1.00) 24.2 21.5 Die 3 (397) 5 (729) 18/161 23/322 39/179 49/335 0.59 (0.07 o 4.65) 0.57 (0.18 o 1.79) 35.8 38.0 Physical ac i i y 7 (2565) 20 (5125) 55/1242 106/2513 73/1195 147/2359 0.74 (0.42 o 1.33) 0.68 (0.49 o 0.93) 6.0 0.0 Mixed app oach 13 (6797) 21 (9136) 359/3183 430/4295 322/2933 407/4004 1.05 (0.86 o 1.28) 1.01 (0.87 o 1.17)* 19.4 16.3 P e e m bi h: O e all 32 (11 676) 49 (14 339) 332/5713 414/6971 345/5286 443/6511 0.94 (0.78 o 1.13) 0.92 (0.79 o 1.08) 17.3 8.7 Die 4 (1344) 7 (1696) 9/647 13/819 35/653 45/819 0.28 (0.08 o 0.96) 0.32 (0.14 o 0.70) 0.0 0.0 Physical ac i i y 13 (3249) 23 (5149) 95/1541 160/2431 73/1540 148/2410 1.29 (0.90 o 1.85) 1.09 (0.84 o 1.41) 0.0 0.0 Mixed app oach 16 (7219) 20 (7630) 228/3525 241/3721 243/3223 256/3412 0.91 (0.73 o 1.12) 0.92 (0.75 o 1.12) 0.0 32.3 Caesa ean sec ion: O e all 32 (11 410) 66 (18 041) 1525/4385 2373/6860 1506/3994 2440/6368 0.91 (0.83 o 0.99) 0.89 (0.83 o 0.96) 0.0 16.2 Die 4 (1340) 7 (1732) 117/535 238/610 149/539 264/620 0.78 (0.50 o 1.22) 0.88 (0.65 o 1.17) 0.0 0.0 Physical ac i i y 13 (3046) 32 (6587) 306/1230 648/2646 349/1161 746/2547 0.82 (0.67 o 1.01) 0.83 (0.73 o 0.95) 0.0 0.0 Mixed app oach 16 (7160) 28 (9858) 1102/2620 1487/3604 1059/2379 1481/3286 0.95 (0.84 o 1.08) 0.92 (0.80 o 1.06) 17.6 21.9 O sp ing Composi e ou come: O e all 18 (7981) NA 1007/3172 NA 951/2851 NA 0.94 (0.83 o 1.08) NA 0.0 NA Die 2 (346) NA 34/132 NA 48/132 NA 0.71 (0.03 o 18.23) NA 0.0 NA Physical ac i i y 5 (1274) NA 138/495 NA 143/498 NA 0.99 (0.67 o 1.46) NA 0.0 NA Mixed app oach 12 (6494) NA 835/2545 NA 797/2317 NA 0.95 (0.81 o 1.11) NA 4.7 NA S illbi h†: O e all 2 (3719) 4 (4534) 9/1858 12/2261 11/1841 14/2247 0.81 (<0.01 o 256.69) 0.85 (0.24 o 3.02) 0.0 0.0 Small o ges a ional age: O e all 33 (11 666) 44 (12 937) 709/5324 773/6018 632/5001 685/5461 1.06 (0.94 o 1.20) 1.05 (0.94 o 1.18) 0.0 0.0 Die 4 (1337) 6 (1628) 41/610 56/746 47/639 55/771 0.92 (0.45 o 1.88) 1.05 (0.62 o 1.77) 0.0 0.0 Physical ac i i y 14 (3272) 21 (3955) 243/1402 274/1740 232/1395 271/1670 1.05 (0.84 o 1.34) 1.01 (0.83 o 1.24) 12.3 51.7 Mixed app oach 16 (7193) 20 (7670) 425/3312 443/3532 370/3086 386/3309 1.08 (0.92 o 1.28) 1.08 (0.93 o 1.27) 0.0 0.0 La ge o ges a ional age: O e all 34 (12 047) 45 (13 348) 744/5492 820/6185 759/5052 833/5510 0.90 (0.76 o 1.07) 0.86 (0.71 o 1.04) 38.0 41.0 Die 4 (1408) 6 (1699) 155/529 172/663 176/548 203/661 0.91 (0.60 o 1.37) 0.82 (0.54 o 1.22) 0.0 0.0 Physical ac i i y 15 (3330) 21 (3930) 121/1557 159/1842 124/1528 161/1768 0.96 (0.59 o 1.54) 0.96 (0.67 o 1.37) 34.3 6.9 Mixed app oach 16 (7450) 21 (8040) 468/3406 489/3680 481/3095 523/3348 0.89 (0.67 o 1.17) 0.83 (0.62 o 1.10) 51.0 4.3 Admission o neona al in ensi e ca e uni : O e all 16 (8140) 21 (9498) 302/3973 406/4543 279/3586 400/4149 1.01 (0.84 o 1.23) 0.97 (0.82 o 1.14) 0.0 0.0 Die 1 (289) 2 (389) 3/137 11/179 13/136 29/170 NA‡ 0.33 (<0.01 o 47.97) NA 0.0 Physical ac i i y 3 (1166) 4 (1240) 31/552 34/586 40/543 43/577 0.77 (0.21 o 2.81) 0.79 (0.35 o 1.78) 20.8 0.0 Mixed app oach 13 (6818) 15 (7771) 268/3284 360/3626 230/3036 332/3453 1.10 (0.89 o 1.35) 1.05 (0.88 o 1.25) 0.0 0.0 *Recalcula ion using De Simonian-Lai d. †All da a come om s udies wi h mixed app oach in e en ions. ‡No possible o es ima e s anda d de ia ions. RESEARCH he bmj | BMJ 2017;358:j3119 | doi: 10.1136/bmj.j3119 9 p edic ion in e al sugges ing a bene icial e ec on ges a ional weigh gain when he in e en ion is applied in a new popula ion o se ing. The e is no s ong e idence ha in e en ions educe he isk o ma e nal and o sp ing composi e ou comes, wi h no a ia ion in e ec obse ed ac oss he subg oups. Fo indi idual ou comes, in e en ions educe caesa ean sec ion wi hou a s a is ically signi ican educ ion in o he ma e nal and o sp ing complica ions. The e ec s o in e en ions o indi idual ma e nal and o sp ing complica ions a e consis en i espec i e o he BMI o he mo he . Addi ion o s udy le el da a om non-IPD s udies o he IPD me a-analysis inc eased he p ecision o es ima es, wi hou a change in he di ec ion o e ec , and showed addi ional bene i o ges a ional diabe es. Among indi idual in e en ions, hose mainly based on physical ac i i y lowe ed he odds o ges a ional diabe es. S eng hs and weaknesses o his s udy To ou knowledge his is he i s IPD me a-analysis o assess he di e en ial e ec s o die and physical ac i i y based in e en ions o impo an , clinically ele an ou comes, in subg oups o women who we e iden i ied a p io i. Es ablishmen o he In e na ional Weigh Managemen in P egnancy IPD Collabo a i e G oup acili a ed he collabo a ion o key esea che s in his a ea and p o ided access o he la ges IPD in his special y. This allowed us o ex ac da a ha we e no published, wi h la ge sample sizes o ou comes such as p e e m bi h, small and la ge o ges a ional age e uses, and admission o he neona al in ensi e ca e uni o IPD han o s udy le el me a- analysis. Fu he mo e, we we e able o minimise he he e ogenei y in he popula ion by excluding indi idual women who did no ul il he inclusion c i e ia. We compa ed he quali y o s udies ha con ibu ed o he IPD, which we e gene ally o highe quali y han hose ha did no con ibu e IPD. Table5 | Di e en ial e ec s o die and physical ac i i y based in e en ions on ma e nal and o sp ing composi e ou comes in subg oups o p egnan women Composi e ou comes No o s udies (No o women) Odds a io* (95% CI) T ea men co a ia e in e ac ion Coe icien ; 95% CI (95% PI) I2 (%) Ma e nal Baseline body mass index: No mal 12 (2445) 0.91 (0.65 o 1.28) 1.00; 0.98 o 1.02 (0.98 o 1.02)† 0 O e weigh 19 (2222) 1.04 (0.86 o 1.26) Obese 20 (4181) 0.92 (0.80 o 1.05) Pa i y: Nullipa ous 21 (4613) 0.87 (0.71 o 1.07) 1.03; 0.75 o 1.39 (0.53 o 2.00)‡ 34.0 Mul ipa ous 22 (4186) 0.92 (0.78 o 1.07) E hnici y: Whi e 15 (6510) 0.92 (0.79 o 1.07) 0.93; 0.63 o 1.37 (0.62 o 1.38)§ 0 Non-whi e 11 (917) 0.86 (0.63 o 1.17) Age (yea s): ≥20 24 (8656) 0.91 (0.81 o 1.02) 1.01; 0.99 o 1.03 (0.99 o 1.03)¶ 0 <20 9 (172) 1.57 (0.66 o 3.71) P e-exis ing medical condi ion**: None 15 (3135) 0.85 (0.66 o 1.09) 1.44; 0.15 o 13.74 (0.03 o 76.75)†† 24.9 ≥15 (89) 1.65 (0.36 o 7.51) O sp ing Baseline body mass index: No mal 7 (1843) 0.93 (0.60 o 1.43) 0.98; 0.95 o 1.00 (0.94 o 1.02)† 18.5 O e weigh 12 (2065) 0.83 (0.61 o 1.13) Obese 13 (4327) 0.92 (0.72 o 1.19) Pa i y: Nullipa ous 16 (4152) 0.97 (0.80 o 1.17) 0.94; 0.64 o 1.37 (0.39 o 2.28)‡ 35.5 Mul ipa ous 15 (4048) 0.91 (0.72 o 1.15) E hnici y: Whi e 11 (6018) 0.93 (0.79 o 1.08) 1.12; 0.75 o 1.68 (0.74 o 1.69)§ 0 Non-whi e 9 (939) 1.10 (0.78 o 1.54) Age (yea s): ≥20 16 (8061) 0.95 (0.82 o 1.09) 1.01; 0.98, 1.04 (0.97 o 1.05)¶ 4.1 <20 7 (162) 1.01 (0.34 o 2.98) P e-exis ing medical condi ion**: None 12 (3407) 0.89 (0.74 o 1.08) 0.58; 0.03, 9.81 (<0.001 o 2440.15)†† 0 ≥13 (63) 0.54 (0.04 o 7.52) PI=p edic ion in e al. *Model accoun ed o baseline weigh and clus e ing e ec . †Pe uni o body mass index. ‡Mul ipa ous e sus nullipa ous. §Non-whi e e sus whi e. ¶Pe yea o age. **Diabe es melli us o hype ension. ††≥1 medical condi ion e sus none.