Association between weight gain during pregnancyand postpartum weight retention and obesity:a bias-adjusted meta-analysis

Munim Mannan, Suhail AR Doi, and Abdullah A Mamun

Gestational weight gain (GWG) is associated with postpartum weight retention(PPWR) in women. The strength of the association between GWG and long-termPPWR and body mass index (BMI), however, is still unclear. Publications fromdifferent databases were systematically extracted and the articles relevant to thisstudy were reviewed to quantify the effect estimate of GWG on PPWR and BMI usinga bias-adjusted method. The Institute of Medicine categories of “inadequate,”“adequate,” and “excess” were used to define GWG. The time span for PPWR wasdivided into three periods (<1 year, 1 year to 9 years, and �15 years) to determineoutcome at different times postpartum. Twelve studies met the eligibility criteria andwere included in the analyses. Women with an inadequate GWG had a significantlylower mean PPWR of -2.14 kg (95%CI, -2.61 to -1.66) than women with anadequate GWG, who had a mean PPWR of 3.15 kg (95%CI, 2.47 to 3.82) up to21 years postpartum. Over the postpartum time span, a U-shaped relationship wasobserved between the weighted mean difference calculated for women with excessGWG and the weighted mean difference calculated for women with adequate GWG,and this relationship was time independent between these two groups. PostpartumBMI showed a similar relationship and magnitude of change, but the exact lossor gain was difficult to assess due to fewer studies (n = 5) with considerableheterogeneity of BMI measurements. The findings of this study suggest that GWGoutside of the Institute of Medicine recommendations can lead to both short-termand long-term postpartum weight imbalance.© 2013 International Life Sciences Institute

INTRODUCTION

One of the natural and biological causes of weight recy-cling in the human population is pregnancy.1 Manywomen gain considerable weight during pregnancy and,for some of them, pregnancy substantially alters theirfuture weight-gain trajectory.1,2 Pregnancy has thus beenidentified as a critical period for the future developmentof weight gain and obesity in mothers3 and children.4,5

Gestational weight gain (GWG), defined by the Instituteof Medicine (IOM)6,7 and categorized as “inadequate,”

“adequate,” and “excess,” is a commonly used indicator topredict postpartum weight retention (PPWR) and obesityover short and long terms.1,3,8 Studies have reported thatGWG – compared with other factors such as smokingcessation, parity, lactation, maternal age, and race – has asignificant influence on the relationship between precon-ception weight and weight 6–18 months after delivery.3,9

However, it is unknown whether the strength of the asso-ciation between the IOM categories and PPWR and bodymass index (BMI) are different over the longer term com-pared with the shorter periods of time.

Affiliation: M Mannan, SAR Doi, and AA Mamun are with the School of Population Health, University of Queensland, Brisbane, Queensland,Australia.

Correspondence: AA Mamun, School of Population Health, University of Queensland, Brisbane, Queensland 4006, Australia. E-mail:mamun@sph.uq.edu.au, Phone: +61-7-33464689, Fax: +61-7-33655599.

Key words: body mass index, gestational weight gain, postpartum weight retention

bs_bs_banner

Special Article

doi:10.1111/nure.12034Nutrition Reviews® Vol. 71(6):343–352 343

A systematic review and meta-analyses of 21 articlespublished during the period 1995 to 2005 showed anoverall decline in mean body weight within the first yearpostpartum, but women were not categorized accordingto the IOM categories.10 A more recent meta-analysis ofnine observational studies that categorized women by theIOM GWG categories found that, compared with womenwith adequate GWG, those with excess GWG had gainedan average of 4.72 kg (95%CI, 2.94–6.50 kg) after 15 yearspostpartum, but women with inadequate GWG demon-strated lower average weight gain in the short andmedium terms but no weight loss beyond 15 years post-partum.11 Since many of the primary studies have meth-odological limitations, a bias-adjusted meta-analysis12,13

with an extension to more recent studies was designed inan attempt to verify whether the previously defined rela-tionships remain valid. This is important because a vali-dation of the association between GWG and PPWR hasthe potential to guide targeted intervention efforts for theprevention of obesity in women.

METHODS

Search strategy

A systematic computerized literature search of thePubMed (MeSH & Medline), CINAHL, Embase, BIOSISPreviews, and Current Contents Connect databases wasperformed, covering the period 1961 to 2011. Articlesexamining the impact of GWG on PPWR and BMI weresought.As part of the search strategy, key terms were usedin the following manner: (“Pregnant” OR “Pregnancy”OR “Prenatal” OR “Gestation” OR “Gestational”) AND(“Postpartum” OR “Post-partum” OR “Post partum” OR“Post pregnancy” OR “Post-natal” OR Puerperium OR“After birth” OR “After childbirth”) AND (“Weight gain”OR “Weight-gain” OR “Weight gains” OR “Weightgained” OR “Gained weight” OR “Weight increase” OR“Overweight” OR “Obese” OR “Obesity” OR “GWG”)AND (“Retention” OR “Retain” OR “Maintain” OR“Maintained” OR “Stabilization” OR “Stabilized” OR“Stabilisation” OR “Stabilised” OR “PPWR”). In addition,reference lists of eligible articles were hand searched toensure a complete collection of studies. To ensure allavailable studies were included, authors of articles werecontacted by e-mail when the required data was onlypartially reported in the published articles.

Eligibility criteria

Studies were included if they 1) were published in theEnglish language; 2) reported GWG as classified by IOMrecommendations into “inadequate,” “adequate,” and

“excess” categories6,7; 3) documented PPWR as a continu-ous variable and reported it from 2 weeks postpartumonward; and 4) reported only singleton pregnancies ofwomen.

Data abstraction and quality appraisal

A standard data extraction form summarizing the studydesign, study population, and relevant raw data was com-pleted for each article by two independent reviewers.Requests for data not reported in a published article weree-mailed to the authors. In addition, weights were con-verted from pounds to kilograms in a few studies to keepthe reported data homogenous. A quality scoring instru-ment was devised on the basis of standard bias criteria inobservational studies,14 with consideration given tofactors that might result from design-specific bias, selec-tion bias, adjustment of potential confounders, informa-tion bias, and analytical methods. This quality scoringinstrument was then applied independently by tworesearchers (MM and AM), and any disagreement inscoring was resolved by consensus.

Statistical analysis

Data on PPWR and BMI were reported up to 21 yearsafter pregnancy. PPWR was determined by various crite-ria used in the individual studies. The weighted meandifference was used as a measure of the relationshipbetween GWG and PPWR. A value of 1 indicates equiva-lence, and a value greater than 1 indicates a higher PPWRthan that observed in the adequate GWG groups. Hetero-geneity was determined to be present when the value of t2

was greater than zero and/or the Q statistic was signifi-cant at a P < 0.10.15 Heterogeneous studies were com-bined in both quality-effects12,13 and random-effects(RE)16 models (for comparison, see Supporting Informa-tion), though only quality-effects results are reported. Allanalyses were carried out using MetaXL version 1.1(EpiGear International; Wilston, Queensland, Australia).

Sensitivity analysis and publication bias

To assess the robustness of the meta-analysis, sensitivityanalyses were performed by examining the effect of sub-grouping of the various studies on the pooled results.Since most of the studies reported short-term relation-ships of GWG and PPWR, the data was analyzed bothwithout any stratification and with stratification by post-partum time. Studies were stratified into three differentcategories (up to 1 year; 1 year to 9 years, and 15 years andabove) on the basis of the time postpartum when theweight retention was measured. For BMI, stratification ofthe postpartum time was not considered because only a

Nutrition Reviews® Vol. 71(6):343–352344

few studies reported BMI as an outcome. For PPWR andBMI, there was a discontinuity in time categories becauseof the unavailability of information.A one-way sensitivityanalysis was also carried out by excluding individualstudies from the meta-analysis. Finally, analysis of theeffect of potential unpublished studies or missed studieswith negative results was undertaken to evaluate therobustness of the meta-analysis, and funnel plot asymme-try was checked. The funnel plot was considered to beasymmetrical if the intercept of Egger’s regression devi-ated from zero with P < 0.10.

RESULTS

Study characteristics

After 700 articles were excluded on the basis of irrel-evance or an absence of information from the pregnancyor postpregnancy period (Figure 1), 98 abstracts werereviewed and 25 full-text articles screened.At this stage ofscreening, studies that did not follow the IOM guidelines(either the 1990 version or the updated recommendation)in the assessment of GWG were excluded. Twelve17–28 of25 studies met the inclusion criteria: nine were prospec-tive and three17,18,20 were retrospective. Eleven studieswere excluded for the following reasons: PPWR was

measured only for obese women (n = 2 studies); GWGwas an outcome (n = 1); the effect size of PPWR wasdefined as an odds ratio (OR) (n = 3); and data on GWGand PPWR were missing (n = 5). A further two studieswere excluded because either GWG was measured forconsecutive pregnancies or GWG and PPWR were cat-egorized arbitrarily. The final meta-analysis thus included>68,000 women from 12 studies, analyzed at differenttime points. Eight of the studies reported GWG as rec-ommended by the 1990 IOM guidelines and four studiesused the updated recommendations from 2009.17,20,27,28

One study applied the Canadian GWG recommendationsof 1999, which are similar to the 1990 IOM guidelines.18

Only five studies reported postpartum BMI and its cat-egories as an outcome.19,21,22,25,26

Six studies were conducted in NorthAmerica,17,18,20–22,24 three were from Europe,25,27,28 and theother three were from Brazil,23 Asia,26 and Australia.19

Socioeconomic status among the studies was thought tobe homogeneous across all studies except one.24 There wasa small variation in age among the study participants, withthe average age being approximately 27 years (mean agerange, 22.2–31.2 years). The sample size varied betweenthe studies, ranging from 63 to 32,920. Five studies hadfewer than 550 participants, and the rest had more than1,000 (Table 1). Most of the studies, with the exceptionof three,17,20,24 included a preponderance of Caucasian

Records identified through databasesearch

(n = 1149)

Scr

een

ing

Incl

ud

edE

ligib

ility

Iden

tifi

cati

on Additional records identified through

other sources (n = 10)

Records remaining after removal of duplicates(n = 798)

Abstracts screened (n = 98)

Records excluded for improper topic (n = 700)

Full-text articles assessed foreligibility(n = 25)

Full-text articles excluded

• PPWR measured only for obese (n=2)• GWG measured as outcome (n= 1) • PPWR measured as OR (n=3)• Missing data on GWG and PPWR (n=5)• GWG measured for consecutive

pregnancies (n=1)• GWG and PPWR categorized arbitrarily

(n=1)

Studies included in quantitative synthesis (meta-analysis)(n = 12)

Figure 1 Flow diagram illustrating the selection process for articles included in the meta-analysis.

Nutrition Reviews® Vol. 71(6):343–352 345

women, and most had a mix of participants that consistedof different ethnic groups.

The mean GWG ranged between 10.6 kg and15.8 kg across studies (Table 2). The mean PPWR rangedbetween 0.90 kg at 1 year to 16.14 kg at 21 years. Similarly,postpartum BMI ranged from 23.20 kg/m2 at 1 year to33.80 kg/m2 at 3 years.There was considerable variation inexcess weight gain, which ranged from 16% to 53%.

Study quality

Of the prospective studies identified, three reported thatthe percentage of women lost to follow-up exceeded20%,19,23,28 and another two21,22 did not report follow-upstatus. All but three of the studies had a well-documentedstudy design and well-defined inclusion criteria: theexceptions included two studies with partial reports17,19

and one study without an explicit report.25 Six of thestudies were designed for short-term follow-up (up to 1year), and the rest of the studies had long-term follow-upperiods, the longest being up to 21 years. Seven of thestudies used self-reported prepregnancy weight mea-sures, which may be a source of information bias. Post-partum body weight was measured in most17,19,21–26 butnot all studies and included two studies in which it wasself-reported.18,27 Five of the studies reported missingdata, and thus this was an additional source of bias. Allthese study qualities were combined into a score out of 20points, with scores ranging between 6 and 16 (Table S1).

Meta-analysis of PPWR

Compared with women who had an adequate GWG(within the limits specified in the IOM recommenda-tions), women with an inadequate GWG (below the IOMrecommendation) had a significantly lower PPWR of-2.14 kg (95%CI, -2.61 to -1.66 kg) (Figure 2). Thispooled estimate is based on all studies in which measure-ments were taken between 2 weeks postpartum and 21years postpartum. When postpartum time spans werestratified into <1 year, 1 year to <9 years, and �15 years,the respective estimates of PPWR were -2.25 kg (95%CI,-3.08 to -1.42 kg), -2.23 kg (95%CI, -2.56 to -1.90 kg),and -1.59 kg (95%CI, -2.59 to -0.60 kg).

Compared with women who had an adequate GWG,those with excess GWG (above that in the IOM recom-mendations) retained an additional 3.15 kg (95%CI,2.47 to 3.82 kg) by 21 years postpartum (Figure 3). Again,stratification of postpartum time spans into <1 year, 1 to<9 years, and �15 years revealed PPWR estimates of2.98 kg (95%CI, 1.77 to 4.18 kg), 2.89 kg (95%CI, 2.00 to3.77 kg), and 4.95 kg (95%CI, 3.90 to 6.00 kg), respec-tively. This difference in PPWR between women withexcess GWG and women with adequate GWG appears toTa

ble

1Ch

arac

teri

stic

sof

the

stud

ies

incl

uded

inth

em

eta-

anal

ysis

.Re

fere

nce

Stud

yde

sign

Anal

ytic

sam

ple

(no.

ofsu

bjec

ts)

Mea

nag

eat

entr

y(y

ears

)G

WG

PPW

RBM

IO

R(o

verw

eigh

tan

dob

esity

)Bu

tte

etal

.(20

03)22

Pros

pect

ive

coho

rtst

udy

6331

Da

NR

M–

Wal

kere

tal.

(200

4)24

Pros

pect

ive

coho

rtst

udy

419

22.2

Da

Db

––

Kac

etal

.(20

04)23

Pros

pect

ive

coho

rtst

udy

233

26.2

SRD

b–

–Kr

ause

etal

.(20

10)17

Retr

ospe

ctiv

eco

hort

stud

y14

,330

at3

mon

ths

&4,

922

at6

mon

ths

23.9

NR

D–

–

Low

ell&

Mill

er.(

2010

)18Re

tros

pect

ive

coho

rtst

udy

5,55

418

–50

SRN

R–

–Ro

deet

al.(

2012

)27Pr

ospe

ctiv

eco

hort

stud

y1,

840

29D

aD

b–

–O

stby

eet

al.(

2010

)20Re

tros

pect

ive

coho

rtst

udy

32,9

2022

.5N

RN

R–

–M

adda

h&

Nik

ooye

h(2

009)

26Pr

ospe

ctiv

eco

hort

stud

y1,

315

31.2

Da

Db

M–

Amor

imet

al.(

2007

)25Pr

ospe

ctiv

eco

hort

stud

y48

330

Da

Db

M–

Mam

unet

al.(

2010

)19Pr

ospe

ctiv

eco

hort

stud

y2,

026

25.7

Da

Db

MM

Fras

eret

al.(

2011

)28Pr

ospe

ctiv

eco

hort

stud

y2,

356

28.7

Da

––

MRo

oney

etal

.(20

02)21

Pros

pect

ive

coho

rtst

udy

540

28.6

DD

M–

aBo

dyw

eigh

tin

late

preg

nanc

ym

inus

prep

regn

ancy

body

wei

ght;

body

wei

ghta

tlas

tant

enat

alvi

sit–

body

wei

ghta

tfirs

tant

enat

alvi

sit;

max

imum

wei

ghti

npr

egna

ncy

min

uspr

epre

gnan

cyw

eigh

t;bo

dyw

eigh

tat3

7w

eeks

’ges

tatio

n–

prep

regn

ancy

wei

ght;

body

wei

ghta

tend

ofpr

egna

ncy

–pr

epre

gnan

cybo

dyw

eigh

t.b

Body

wei

ghta

fter

preg

nanc

y–

prep

regn

ancy

body

wei

ghte

qual

spo

stpa

rtum

wei

ghtg

ain.

Abbr

evia

tions

:BM

I,bo

dym

ass

inde

x;D,

defin

ed;G

WG

,ges

tatio

nalw

eigh

tgai

n;M

,mea

sure

d;N

R,no

trep

orte

d;SR

,sel

f-re

port

ed.P

PWR,

post

part

umw

eigh

tret

entio

n;O

R,od

dsra

tio.

Nutrition Reviews® Vol. 71(6):343–352346

Tabl

e2

Sum

mar

yst

atis

tics

for

GW

G,P

PWR,

and

BMIa

ccor

ding

toIO

Mca

tego

ries

.Re

fere

nce

Post

part

umtim

eM

ean

GW

G(k

g)M

ean

PPW

R(k

g)M

ean

BMI

Perc

enta

geof

part

icip

ants

with

inea

chIO

Mca

tego

ryIn

adeq

uate

Adeq

uate

Exce

ssBu

tte

etal

.(20

03)22

2w

eeks

14.2

5.68

24.7

527

5419

Butt

eet

al.(

2003

)226

wee

ks4.

8424

.44

Butt

eet

al.(

2003

)2227

wee

ks2.

6023

.47

Wal

kere

tal.

(200

4)24

6w

eeks

15.8

6.52

–21

2753

Kac

etal

.(20

04)23

2w

eeks

12.9

4.35

–36

3529

Kac

etal

.(20

04)23

2m

onth

s4.

49–

Kac

etal

.(20

04)23

6m

onth

s3.

46–

Kac

etal

.(20

04)23

9m

onth

s3.

11–

Krau

seet

al.(

2010

)173

mon

ths

13.6

5.28

–29

3338

Krau

seet

al.(

2010

)176

mon

ths

5.64

–Lo

wel

l&M

iller

(201

0)18

5–9

mon

ths

–2.

75–

2136

43Ro

deet

al.(

2012

)271

year

13.9

0.90

–23

3641

Ost

bye

etal

.(20

10)20

2.8

year

s13

.94.

90–

2828

44M

adda

h&

Nik

ooye

h(2

009)

261

year

10.6

1.75

23.2

043

4116

Mad

dah

&N

ikoo

yeh

(200

9)26

2ye

ars

3.59

27.4

0M

adda

h&

Nik

ooye

h(2

009)

263

year

s4.

4633

.80

Amor

imet

al.(

2007

)2515

year

s14

.27.

6624

.33

2842

30M

amun

etal

.(20

10)19

21ye

ars

14.7

16.1

427

.82

2641

33Fr

aser

etal

.(20

11)28

16ye

ars

13.0

6–

–35

4124

Roon

eyet

al.(

2002

)216

mon

ths

13.1

1.75

25.5

626

5024

Roon

eyet

al.(

2002

)218.

5ye

ars

6.34

27.3

1Ab

brev

iatio

ns:B

MI,

body

mas

sin

dex;

GW

G,g

esta

tiona

lwei

ghtg

ain;

IOM

,Ins

titut

eof

Med

icin

e;PP

WR,

post

part

umw

eigh

tret

entio

n.

Nutrition Reviews® Vol. 71(6):343–352 347

be dependent on postpartum time span. Therefore, theweighted mean difference in PPWR between women inthe excess GWG and adequate GWG categories wasfurther stratified by the length of the follow-up period: 6months, 1.5 years, 8.5 years, 15 years, and 21 years post-partum (Figure 4). Compared with PPWR in women withadequate GWG, PPWR in women with excess GWGshowed a U-shaped trend, declining in the early postpar-tum years and then increasing again later in the follow-upperiod.

Meta-analysis of BMI

Only five studies reported postpartum BMI. In a pooledanalysis, a trend similar to that observed with PPWR was

found, although the variations of the estimates weremuch wider. Those women with inadequate GWG, whencompared with women with adequate GWG, had adecline in BMI of -2.42 kg/m2 (95%CI, -3.03 to -1.80 kg/m2). In contrast, those with excess GWG gained an addi-tional 3.78 kg/m2 (95%CI, 3.14 to 4.41 kg/m2) over apostpartum period of 21 years (Figure S1).

There was, however, considerable heterogeneity inBMI figures, and thus while the trend is clear, exact lossesor gains are difficult to assess. There were only two studiesthat examined the association of IOM GWG categorieswith maternal postpartum overweight and obesity.Mamun et al.19 reported that women who gained excessweight during pregnancy had an increased odds of beingoverweight (OR, 2.15; 95%CI, 1.64–2.82) or obese (OR,

Inadequate GWG by follow-up time

WMD20-2-4

Study or subgroup

Walker et al. (2004)24

Maddah & Nikooyeh (2009)26

Krause et al. (2010)17

Rooney et al. (2002)21

Rooney & Schauberger (2002)21

Ostbye et al. (2010)20

0-1 y subgroup

1+ to 9 y subgroup

0-1 y

Q=81.67, p=0.00, I2=91%

1+ to 9 y

Q=3.50, p=0.17, I2=43%

15+ y

Q=3.09, p=0.08, I2=68%

Overall

Q=99.06, p=0.00, I2=88%

Mamun et al. (2010)19

Rode et al. (2011)27

Kac et al. (2004)23

15+ y subgroup

Lowell & Miller (2010)18

Maddeh & Nikooyeh (2009)26

Amorim et al. (2007)25

Butte et al. (2003)22

WMD (95% CI) % Weight

-4.20 ( -5.17, -3.23) 4.49

-2.70 ( -3.25, -2.15) 6.55

-2.45 ( -2.65, -2.25) 14.70

-2.41 ( -2.88, -1.94) 8.27

-2.40 ( -2.85, -1.95) 8.48

-2.40 ( -2.62, -2.18) 14.52

-2.25 ( -3.08, -1.42) 59.60

-2.23 ( -2.56, -1.90) 28.29

-2.14 ( -2.61, -1.66) 100.00

-2.16 ( -3.27, -1.05) 6.95

-2.10 ( -2.62, -1.58) 4.72

-2.08 ( -3.32, -0.84) 5.28

-1.59 ( -2.59, -0.60) 12.11

-1.50 ( -1.67, -1.33) 12.71

-1.45 ( -2.43 -0.47) 5.29

-0.50 ( -1.98, 0.98) 5.17

-0.23 ( -5.79, 5.33) 2.87

Figure 2 Weighted mean difference between women with inadequate GWG and women with adequate GWG at threepostpartum time periods: <1 year, >1 year to 9 years, and �15 years.

Nutrition Reviews® Vol. 71(6):343–352348

4.49; 95%CI, 3.42–5.89) 21 years after the index preg-nancy, and Fraser et al.28 also found an increased odds ofbeing overweight (OR, 3.58; 95%CI, 2.61–4.93) at 16 yearspostpartum.

One-way sensitivity analysis and publication bias

One-way sensitivity analyses with each study removedindividually suggested that no study individually alteredthe pooled results of the meta-analyses of PPWR signifi-cantly. The 95%CI for the intercept (-0.819) on Egger’sregression for studies of inadequate GWG included zero(P = 0.541),suggesting symmetry and,thus,no publicationbias.For studies of excess GWG,a similar result was noted,with an intercept of 0.210 (P = 0.919) again suggesting

symmetry. Thus, the heterogeneity demonstrated for thisanalysis does not seem to be a consequence of publicationbias (Figure S2). The other meta-analyses did not havesufficient data to assess publication bias.

DISCUSSION

In this study, a series of meta-analyses were conducted toinvestigate the association of GWG with both PPWR andBMI over a 21-year period postpartum. The first meta-analysis showed that women with inadequate GWG hadless PPWR (-2.14 kg) over a 21-year period postpartum.All studies except three22,24,25 found that the estimatedPPWR during different postpartum time spans remainedconsistent over a 21-year period postpartum. Finer

Excess GWG by follow-up time

WMD50

Study or subgroup

Maddeh & Nikooyeh (2009)26

Rode et al. (2011)27

Maddah & Nikooyeh (2009)26

Rooney & Schauberger (2002)21

Butte et al. (2003)22

Rooney et al. (2002)21

Lowell & Miller (2010)18

1+ to 9 y subgroup

0-1 y subgroup

Kac et al. (2004)23

0-1 y

Q=158.65, p=0.00, I2=96%

1+ to 9 y

Q=56.55, p=0.00, I2=96%

15+ y

Q=3.09, p=0.08, I2=68%

Overall

Q=265.67, p=0.00, I2=95%

Krause et al. (2010)17

Amorim et al. (2007)25

Ostbye et al. (2010)20

15+ y subgroup

Mamun et al. (2010)19

Walker et al. (2004)24

WMD (95% CI) % Weight

1.55 ( 0.29, 2.81) 5.18

1.70 ( 1.22, 2.18) 4.931.90 ( 1.17, 2.63) 5.81

1.90 ( 1.40, 2.40) 7.84

2.13 ( -8.99, 13.25) 2.96

2.40 ( 1.87, 2.93) 7.58

2.50 ( 2.35, 2.65) 14.83

2.89 ( 2.00, 3.77) 27.26

2.98 ( 1.77, 4.18) 60.44

3.06 ( 1.74, 4.38) 5.37

3.15 ( 2.47, 3.82) 100.00

3.57 ( 3.38, 3.76) 14.44

3.70 ( 2.03, 5.38) 5.25

3.80 ( 3.59, 4.01) 14.25

4.96 ( 3.90, 6.00) 12.31

5.53 ( 4.37, 6.69) 7.05

6.69 ( 5.66, 7.72) 4.52

Figure 3 Weighted mean difference between women with excess GWG and women with adequate GWG at threepostpartum time periods: <1 year, >1 year to 9 years, and �15 years.

Nutrition Reviews® Vol. 71(6):343–352 349

stratification of postpartum follow-up time also revealedconsistent estimates, suggesting that the difference inPPWR in those women with inadequate GWG persistslong term. There were only two studies conducted after15 years postpartum, with the reported effect sizes being-0.50 (95%CI, -1.98 to 0.98 kg) at 15 years and -2.16(95%CI, -3.27 to -1.05 kg) at 21 years. The meta-analysisof these two studies suggested a consistent, albeit slightlydecreased, effect size (-1.53; 95%CI, -2.53 to -0.54 kg),but this remained statistically significant. This resultdoes not support the findings of another recent meta-analysis,11 which reported that the effect size of thisassociation appeared to decrease over postpartum time,becoming statistically nonsignificant after 15 yearspostpartum. In contrast, the present meta-analysis sug-gests that women with inadequate GWG had a lowerPPWR, independent of the postpartum time span up to21 years.

The second meta-analysis showed that women withexcess GWG retained an additional 3.18 kg up to 21 yearspostpartum. Stratification of postpartum time spansshowed this association to be dependent on the length ofthe postpartum follow-up period. A further finer stratifi-cation of postpartum follow-up times showed a U-shapedtrend for this additional weight retention, being higher atthe early years postpartum, then declining by 8.5 yearspostpartum, and then increasing again by 21 years post-partum. However, this result should be viewed withcaution, as it is based on the results of only one study at8.5 years postpartum and two studies at �15 years post-partum. In a broader perspective, this finding is similar tothat of the previous study.11 The present meta-analysis,however, adds to the finding that finer stratification of

postpartum follow-up periods suggests a time-dependentrelationship between excess GWG and PPWR that maydiffer between short-term and long-term follow-upperiods.

Overall, the third and fourth meta-analyses of post-partum BMI showed a trend similar to that of PPWR.However, only five studies were found in which the BMIvalues reported over periods of up to 21 years postpartumshowed considerable heterogeneity.

To verify the effect of other known determinants ofPPWR, a few of the studies attempted to adjust for poten-tial confounders. One study found no confounding effecton the associations, even two decades after pregnancy.19

In another study, high-intensity and frequency of breast-feeding was associated with less PPWR at all follow-uptimes.25 Rooney and Schauberger21 revealed that,although breastfeeding may not have an impact onshort-term weight loss, women who breastfed theirinfants for at least 3 months had a significantly lowerweight gain at 8 years postpartum, and this associationlasted until 15 years postpartum.29 Two other studies alsofound a significant relationship between breastfeedingand PPWR.17,20 Ethnicity24 and exercise were reported tohave minor, nonsignificant effects on short-term weightloss but significant impacts on long-term BMI.21 Lessmaternal education has also been shown to be signifi-cantly associated with weight retention,25 but the oppositeof this association was found in another study.26 A moth-er’s occupation at the time of pregnancy was not relatedto long-term weight change or BMI.21 Primiparouswomen retained more weight at 3 years postpartum thandid multiparous women,26 and adjustment for smokingstatus and parity did not alter the results considerably.

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0 5 10 15 20 25

Mea

n w

eigh

t gai

n (k

g)

Time postpartum (years)

InadequateExcessMean

Figure 4 Long-term trend of weighted mean difference (WMD) between women with excess GWG and women withadequate GWG at 6 months, 1.5 years, 8.5 years, 15 years, and 21 years postpartum.

Nutrition Reviews® Vol. 71(6):343–352350

Most of the studies included in the meta-analysis did notconsider reliable sources of information on diet andphysical activity, even though these two factors are themain ones that influence the association between GWGand PPWR or BMI.

Although underlying mechanisms of the associationbetween GWG and PPWR and BMI are unknown, thepresent meta-analyses suggest that inadequate GWGhas an impact, both short term and long term, on bothPPWR and BMI. These associations were independent ofthe postpartum follow-up time. However, the potentialbeneficial effect of inadequate GWG on PPWR and BMIneeds to be balanced against the potential risks of inad-equate GWG.30 Conversely, women who gained excessweight during pregnancy had a U-shaped trend of PPWRand tended to have a greater PPWR long term. Over thelong term, these women were also at greater risk of havingan increased BMI and of becoming overweight or obesepostpartum.

For statistical analysis, the quality-effects model wasused to adjust for bias because it had an added advantageover the RE model in observational studies. The method-ology in the RE model is grossly flawed, however, to theextent that, even in standard meta-analyses, there is a lackof interpretation of an RE summary. Peto31 called use ofthe RE model in meta-analysis “wrong” because itanswers a question that is “abstruse and uninteresting.”Moreover, use of the RE model requires strong assump-tions that are unlikely to be valid in practice. Mostnotably, the RE analysis is based on “the peculiar premisethat the trials done are representative of some hypotheti-cal population of trials, and that the heterogeneity can berepresented by a single variance.”32 In the present analy-sis, the approach used was that the detection of heteroge-neity presented a strong case for investigating its possibleorigin and for redistributing the weights based on theoutcome of such a determination. Weight redistributionbased on the common between-study variance will not behelpful in such studies because it is doubtful that anyuseful question can be formulated that an RE modelanalysis could answer. The RE results are, however, notedfor comparative purposes (Figure S3).

The analyses here were limited to those studies thatmeasured GWG in relation to IOM categories thatincluded prepregnancy BMI. This approach allowed theGWG to be assessed independently of prepregnancyBMI. It did, however, restrict the analysis to a limitednumber of studies and forced the exclusion of studiesthat reported GWG unrelated to the IOM recommenda-tions. The present investigation was also limited by thelow number of long-term studies, which weakens thereliability of the long-term outcomes found. Further-more, long-term BMI values were available from onlyfive studies.

CONCLUSION

Inadequate GWG was associated with a decreased PPWRor BMI, and this association was independent of the post-partum time span. Excess GWG, while associated withincreased PPWR or higher BMI, was dependent on thepostpartum follow-up time, with a U-shaped trendobserved. The findings of this study indicate that weightgain in pregnancy can lead to short- and long-termweight imbalance postpartum. The association betweenexcess GWG and either short-term or long-term PPWRand BMI was stronger than the association between inad-equate GWG and PPWR. Therefore, it is important tounderstand what differentiates women with excess GWGfrom those who are able to keep the amount of weightgained within the guidelines. Further studies shouldexamine the weight-change trajectories of women fromthe preconception period to conception to the long-termpostpartum period and should determine whether thesetrajectories can be differentiated by IOM categorizationof GWG. In addition, such studies should not be limitedto IOM-recommended categories of GWG. Rather, GWGon a continuous scale and independent of prepregnancyweight or BMI may be considered. Studies should alsoexamine whether the U-shaped association betweenGWG and PPWR remains or if there is actually anoptimal recommended duration of postpartum follow-upin order to establish this link. More information regard-ing the influence of socioeconomic factors, cultural prac-tices, lifestyle changes, breastfeeding, and other relatedbehaviors on the association between weight gain duringpregnancy and PPWR and obesity in later years is essen-tial. It is strongly recommend that future prospectivestudies look at serial measures of anthropometry alongwith sociodemographic and lifestyle changes over thelong term to assess this association further. Given thesubstantial association between excess GWG and long-term PPWR and obesity, there is a clear need for high-quality confirmatory studies followed by studies ofinterventions based on diet and physical activity thatcould address excess weight gain during pregnancy.

Acknowledgments

AM and MM were responsible for literature review, dataextraction, analyses, and drafting of the manuscript. SDcontributed to the methodological development of themeta-analysis and critically reviewed the final draft of themanuscript. Justin Clark is thanked for his assistance insystematically searching the computerized literature.

Funding. For the work presented here, AM receivedgrants from the National Health and Medical ResearchCouncil (NHMRC) of Australia and the University of

Nutrition Reviews® Vol. 71(6):343–352 351

Queensland Foundation Research Excellence Awards.AM is supported by a Career Development Awards fromthe NHMRC (ID 519756).

The views expressed in the paper are those of theauthors and not necessarily those of any funding body,and no funding body influenced the way in which thedata were analyzed and presented.

Declaration of interest. The authors have no relevantinterest to declare.

REFERENCES

1. Gunderson EP, Murtaugh MA, Lewis CE, et al. Excess gains in weight and waistcircumference associated with childbearing: the Coronary Artery Risk Develop-ment in Young Adults Study (CARDIA). Int J Obes Relat Metab Disord.2004;28:525–535.

2. Harris HE, Ellison GTH, Clement S. Do the psychosocial and behavioral changesthat accompany motherhood influence the impact of pregnancy on long-termweight gain? J Psychosom Obstet Gynaecol. 1999;20:65–79.

3. Gunderson EP, Abrams B. Epidemiology of gestational weight gain and bodyweight changes after pregnancy. Epidemiol Rev. 2000;22:261–274.

4. Mamun AA, O’Callaghan M, Callaway L, et al. Associations of gestational weightgain with offspring body mass index and blood pressure at 21 years of age:evidence from a birth cohort study. Circulation. 2009;119:1720–1727.

5. Hochner H, Friedlander Y, Calderon-Margalit R, et al. Associations of maternalpre-pregnancy body mass index and gestational weight gain with adult off-spring cardio-metabolic risk factors: the Jerusalem Perinatal Family Follow-upstudy. Circulation. 2012;125:1381–1389.

6. Institute of Medicine (US), National Research Council (US), Committee toReexamine IOM Pregnancy Weight Gain Guidelines, Rasmussen KM, Yaktine AL,eds. Weight Gain During Pregnancy: Reexamining the Guidelines. Washington, DC:National Academies Press; 2009.

7. Subcommittee on Nutritional Status and Weight Gain During Pregnancy,Subcommittee on Dietary Intake and Nutrient Supplements During Pregnancy,Committee on Nutritional Status During Pregnancy and Lactation, Food andNutrition Board, Institute of Medicine. Nutrition During Pregnancy. Part I: WeightGain. Washington, DC: National Academy Press; 1990.

8. Linne Y, Barkeling B, Rossner S. Long-term weight development after pregnancy.Obes Rev. 2002;3:75–83.

9. Nohr EA, Vaeth M, Baker JL, et al. Pregnancy outcomes related to gestationalweight gain in women defined by their body mass index, parity, height, andsmoking status. Am J Clin Nutr. 2009;90:1288–1294.

10. Schmitt NM, Nicholson WK, Schmitt J. The association of pregnancy and thedevelopment of obesity – results of a systematic review and meta-analysis onthe natural history of postpartum weight retention. Int J Obes. 2007;31:1642–1651.

11. Nehring I, Schmoll S, Beyerlein A, et al. Gestational weight gain and long-termpostpartum weight retention: a meta-analysis. Am J Clin Nutr. 2011;94:1225–1231.

12. Doi SA, Barendregt JJ, Mozurkewich EL. Meta-analysis of heterogeneous clinicaltrials: an empirical example. Contemp Clin Trials. 2011;32:288–298.

13. Doi SA, Thalib L. A quality-effects model for meta-analysis. Epidemiology.2008;19:94–100.

14. Shamliyan T, Kane RL, Dickinson S. A systematic review of tools used to assessthe quality of observational studies that examine incidence or prevalence andrisk factors for diseases. J Clin Epidemiol. 2010;63:1061–1070.

15. Takkouche B, Cadarso-Suarez C, Spiegelman D. Evaluation of old and new testsof heterogeneity in epidemiologic meta-analysis. Am J Epidemiol.1999;150:206–215.

16. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials.1986;7:177–188.

17. Krause KM, Lovelady CA, Peterson BL, et al. Effect of breast-feeding on weightretention at 3 and 6 months postpartum: data from the North Carolina WICProgramme. Public Health Nutr. 2010;13:2019–2026.

18. Lowell H, Miller DC. Weight gain during pregnancy: adherence to HealthCanada’s guidelines. Health Rep. 2010;21:31–36.

19. Mamun AA, Kinarivala M, O’Callaghan MJ, et al. Associations of excess weightgain during pregnancy with long-term maternal overweight and obesity: evi-dence from 21 y postpartum follow-up. Am J Clin Nutr. 2010;91:1336–1341.

20. Ostbye T, Krause KM, Swamy GK, et al. Effect of breastfeeding on weight reten-tion from one pregnancy to the next: results from the North Carolina WICprogram. Prev Med. 2010;51:368–372.

21. Rooney BL, Schauberger CW. Excess pregnancy weight gain and long-termobesity: one decade later. Obstet Gynecol. 2002;100:245–252.

22. Butte NF, Ellis KJ, Wong WW, et al. Composition of gestational weight gainimpacts maternal fat retention and infant birth weight. Am J Obstet Gynecol.2003;189:1423–1432.

23. Kac G, Benicio MH, Velasquez-Melendez G, et al. Gestational weight gain andprepregnancy weight influence postpartum weight retention in a cohort ofBrazilian women. J Nutr. 2004;134:661–666.

24. Walker LO, Timmerman GM, Sterling BS, et al. Do low-income women attain theirpre-pregnant weight by the 6th week of postpartum? Ethn Dis. 2004;14:119–126.

25. Amorim AR, Rossner S, Neovius M, et al. Does excess pregnancy weight gainconstitute a major risk for increasing long-term BMI? Obesity (Silver Spring).2007;15:1278–1286.

26. Maddah M, Nikooyeh B. Weight retention from early pregnancy to three yearspostpartum: a study in Iranian women. Midwifery. 2009;25:731–737.

27. Rode L, Kjaergaard H, Ottesen B, et al. Association between gestational weightgain according to body mass index and postpartum weight in a large cohort ofDanish women. Matern Child Health J. 2012;16:406–413.

28. Fraser A, Tilling K, Macdonald-Wallis C, et al. Associations of gestational weightgain with maternal body mass index, waist circumference, and blood pressuremeasured 16 y after pregnancy: the Avon Longitudinal Study of Parents andChildren (ALSPAC). Am J Clin Nutr. 2011;93:1285–1292.

29. Linne Y, Rossner S. Interrelationships between weight development and weightretention in subsequent pregnancies: the SPAWN study. Acta Obstet GynecolScand. 2003;82:318–325.

30. Beyerlein A, Schiessl B, Lack N, et al. Optimal gestational weight gain ranges forthe avoidance of adverse birth weight outcomes: a novel approach. Am J ClinNutr. 2009;90:1552–1558.

31. Peto R. Why do we need systematic overviews of randomized trials? Stat Med.1987;6:233–244.

32. Senn S. Trying to be precise about vagueness. Stat Med. 2007;26:1417–1430.

SUPPORTING INFORMATION

Additional Supporting Information may be found in theonline version of this article at the publisher’s web-site:

Figure S1 Weighted mean difference of BMI: categoriesof inadequate and excess weight gain from 02 weeks to21 years.Figure S2 Funnel plot for postpartum weight retention.Figure S3 Weighted mean difference between womenwith inadequate gestational weight gain and womenwith excess gestational weight gain at three post-partum time periods (<1 year, >1 year to 9 years, and�15 years) using the random-effects model.Table S1 Quality of the studies included in themeta-analysis.

Nutrition Reviews® Vol. 71(6):343–352352