Background

Obesity is an increasing global health concern, with more than 1.9 billion adults worldwide being overweight [1] and approximately one in two US women of childbearing age now being considered overweight or obese [2]. Considerable evidence exists showing serious perinatal complications associated with obesity in pregnancy including gestational diabetes (GDM), pre-eclampsia (PE) and neonatal death [3]. There is also an increased risk of complications such as fetal growth restriction and preterm birth amongst underweight women [4]. However, evidence is sparse about the effect of interpregnancy weight change on the risk of adverse outcomes in subsequent pregnancies. Current NICE guidelines in the UK recommend that overweight or obese women are referred for weight loss support at the 6–8 week postnatal check-up [5] despite limited evidence to support widespread implementation of such health promotion strategies and of benefit for future pregnancy outcomes [6].

The current study aimed to systematically synthesise the published evidence on the associations between interpregnancy weight change and common perinatal complications for both mother and child including GDM, PE, pregnancy induced hypertension (PIH), preterm birth (PTB), and delivery of a large- and small-for-gestational age neonate (LGA and SGA). Additionally, we compared the risk of these complications after interpregnancy weight change in women with a normal BMI and overweight or obese women, and where possible, we investigated the dose-response relationships.

Methods

Eligibility criteria, information sources, search strategy

The electronic databases PubMed, Ovid EMBASE, ClinicalTrials.gov and Cochrane Central were systematically searched until July 24th, 2019. The search strategy included terms relating to ‘interpregnancy’, ‘between pregnancy’, ‘weight change’ or ‘BMI’. These search terms were combined with the outcomes of interest (‘gestational diabetes’, ‘pre-eclampsia’, ‘pregnancy-induced hypertension’, ‘preterm birth’, ‘small-for-gestational age’ and ‘large-for-gestational age’) and synonyms of these outcomes (for full search string see Additional file 3: Table S1). Furthermore, we cross-referenced selected papers for additional articles to include. The studies identified were uploaded onto Covidence, an online tool for screening of papers for systematic reviews (www.covidence.org). The review protocol was designed a priori and registered with PROSPERO under registration number CRD42017067326.

Study selection

Studies were selected using the following predetermined inclusion criteria: [i] interpregnancy weight change reported in kilogram (kg), BMI units (kg/m2) or percentage body weight change in multiparous women with any time interval between the consecutive births, [ii] any of the perinatal outcomes of interest in the subsequent pregnancy, and [iii] observational, cohort or case-controlled human study design with a sample size ≥50, that were reported in English. When studies reported data from overlapping study populations, the study with the largest sample size was selected for inclusion. Information extracted from each study included country of research, study cohort name (if applicable), study period, sample size, study inclusion criteria, methods of weight reporting, definition of reference group, diagnostic criteria for perinatal outcomes and demographics that studies adjusted for. All study selection, full text screening, and data extraction was undertaken independently by two researchers (NEWDT and KLM), following PRISMA guidelines [7]. Disagreements were decided through a third opinion (AMW).

Data synthesis

Interpregnancy weight change was defined as the difference between pre-pregnancy weight in the index pregnancy, defined as the earliest recorded pregnancy, and pre-pregnancy weight in the subsequent pregnancy. Interpregnancy weight gain and loss were defined on two categorical scales; (i) for the meta-analysis we utilised categories of > 1 BMI unit interpregnancy weight loss, BMI gain between 1 and 2 units, BMI gain between 2 and 3 units or BMI gain of more than 3 units and (ii) for the dose-response analysis we utilised a BMI change of 0, 1, 2 or 3+ units. Crude odds ratios (calculated from studies providing relevant counts) and adjusted odds ratios for each outcome of interest were extracted from the selected publications.

Random effects meta-analysis was used to synthesize the odds ratios for weight change categories. To ensure a consistent reference group, only studies that employed a reference group of interpregnancy weight change between 1-unit weight loss and 1-unit weight gain were included. Heterogeneity was assessed using the I2 statistic.

We conducted a separate analysis comparing interpregnancy weight change and the risk of developing adverse perinatal outcomes in women with a normal BMI (< 25 kg/m2) versus women with an overweight BMI (≥25 kg/m2), at the start of their index pregnancy. To do so, adjusted odds ratios for both BMI categories were extracted from the publications and summarised by random effects meta-analysis.

Dose-response relationships were assessed by plotting association measurements from studies providing multiple weight gain categories. Where ranges of BMI changes were reported, the midpoint of the category was used (e.g. 1.5 BMI units change for the category weight change between 1 and 2 BMI units).

Statistical analysis and graphical presentation were performed using the metafor package in R for Windows, version 3.4.2.

Assessment of risk of bias

A sensitivity analysis was undertaken to assess potential impact of bias in individual studies by excluding studies that scored below 5 out of 9 points in the Newcastle-Ottawa Scale (NOS [8]) quality scoring assessment (Additional file 4: Table S2). Furthermore, leave-one-study-out analyses were conducted to identify whether one study leveraged the overall effect size estimate.

Results

Study selection

We identified and screened 4500 unique publications and included 194 articles for full text review (Fig. 1). A total of 27 studies were eligible for inclusion. Three studies were excluded due to overlapping study populations [9,10,11] and one was excluded because of a sample size < 50 women [12]. From the remaining 23 studies selected to take forward, a total of 671,906 women were identified for inclusion in the review (Table 1). Eighteen studies included only nulliparous women at the index pregnancy. The proportion of women older than 35 years varied between studies from 3 to 33%. All studies were conducted in Western populations, although this was not an inclusion criterion. Seven studies, comprising of 280,672 women, were included in the meta-analysis.

Fig. 1
figure 1

Flow diagram of study inclusion and exclusion

Full size image
Table 1 Characteristics of studies investigating the association between interpregnancy weight change and adverse perinatal outcomes
Full size table

Synthesis of results

Interpregnancy weight gain of between 1 and 2 BMI units was associated with a 51% higher risk of developing GDM (aOR 1.51 [1.22–1.80], I2 = 70.1%), whereas an increase of 2–3 or more than 3 BMI units was associated with an 81 and 137% higher risk (aOR 1.81 [1.20–2.41], I2 = 88.4% and aOR 2.37 [1.50–3.34], I2 = 91.0% respectively) (Fig. 2). Furthermore, interpregnancy weight gain of more than 3 BMI units was associated with a higher risk of PE or PIH (aOR 1.70 [1.50–1.91], I2 = 0.0% and aOR 1.71 [1.51–1.91] I2 = 0.0% respectively) (Figs. 3 and 4). The association between interpregnancy weight change and the risk of delivering an LGA neonate could only be estimated for a weight gain > 3 BMI units, and showed a 63% higher risk (aOR 1.63 [1.30–1.97], I2 = 85.6%) (Fig. 5). In contrast, interpregnancy weight loss of > 1 BMI unit was associated with a lower risk of delivering an LGA neonate, (aOR 0.79 [0.58–0.99], I2 = 86.1%) (Fig. 5), but there was no conclusive evidence of association of interpregnancy weight loss with the risk of developing GDM, PE or PIH (Figs. 2, 3 and 4). There was an insufficient number of studies to conduct a meta-analysis on adjusted odds ratios for the outcomes of SGA and PTB. A meta-analysis combining the crude odds ratios rather than adjusted ratios showed a significant higher risk of developing PE (cOR 1.31 [1.09–1.53], I2 = 75.1%), but showed similar results for the association between interpregnancy weight gain and the risk of developing GDM, PE or PIH (Additional file 1: Figure S1 for interpregnancy weight loss and Additional file 2: Figure S2 for weight gain). For the outcomes of SGA and PTB, meta-analyses of crude odds ratios showed interpregnancy weight loss of > 1 BMI unit was associated with a higher risk of delivering a SGA neonate or delivering preterm (cOR 1.53 [1.35–1.71], I2 = 0.0% and cOR 1.45 [1.21–1.69], I2 = 26.7%] respectively), but there was no evidence of association with interpregnancy weight gain (Additional file 1: Figure S1 and Additional file 2: Figure S2).

Fig. 2
figure 2

Forest plot from random effects meta-analysis showing the association between interpregnancy weight change and the risk of developing gestational diabetes in subsequent pregnancy. All adjusted odds ratios are relative to the reference category of interpregnancy weight change between − 1 and + 1 BMI units. BMI, body mass index (in kg/m2); aOR, adjusted odds ratio; CI, confidence interval

Full size image
Fig. 3
figure 3

Forest plot from random effects meta-analysis showing the association between interpregnancy weight change and the risk of developing pre-eclampsia in subsequent pregnancy. All adjusted odds ratios are relative to the reference category of interpregnancy weight change between − 1 and + 1 BMI units. BMI, body mass index (in kg/m2); aOR, adjusted odds ratio; CI, confidence interval

Full size image
Fig. 4
figure 4

Forest plot from random effects meta-analysis showing the association between interpregnancy weight change and the risk of developing pregnancy induced hypertension in subsequent pregnancy. All adjusted odds ratios are relative to the reference category of interpregnancy weight change between − 1 and + 1 BMI units. BMI, body mass index (in kg/m2); aOR, adjusted odds ratio; CI, confidence interval

Full size image
Fig. 5
figure 5

Forest plot from random effects meta-analysis showing the association between interpregnancy weight change and the risk of delivering a large-for-gestational age neonate in subsequent pregnancy. All adjusted odds ratios are relative to the reference category of interpregnancy weight change between − 1 and + 1 BMI units. BMI, body mass index (in kg/m2); aOR, adjusted odds ratio; CI, confidence interval

Full size image

Figures 6, 7 and 8 show the odds ratios for the risk of developing an adverse perinatal outcome after interpregnancy weight gain, stratified by BMI category in the index pregnancy (normal weight; BMI < 25 kg/m2 versus overweight; BMI ≥25 kg/m2). Women with a normal weight at the start of the index pregnancy had a higher risk of developing GDM after interpregnancy weight gain > 3 BMI units (aOR 4.36 [2.29–6.44], I2 = 81.6%) compared to women with an overweight BMI (aOR 2.26 [1.40–3.12], I2 = 74.4%) (Fig. 6a and b). Similarly, women with a BMI < 25 kg/m2 were at higher risk of delivering a LGA neonate after interpregnancy weight gain > 3 BMI units than women with BMI ≥25 kg/m2 (aOR 1.80 [1.24–2.35], I2 = 87.2% versus aOR 1.50 [1.35–1.66], I2 = 0.0% respectively) (Fig. 7a and b). Women with a normal BMI at the start of their index pregnancy were at higher risk of developing PIH in a subsequent pregnancy after interpregnancy weight gain of 2–3 BMI (aOR 1.60 [1.04–2.16, I2 = 54.6%) and > 3 BMI units (aOR 2.21 [1.81–2.60], I2 = 0.0%), compared to women with an overweight BMI (2–3 units gain; aOR 0.95 [0.73–1.17], I2 = 0.0%, > 3 units gain; aOR 1.37 [1.16–1.59], I2 = 0.0%) (Fig. 8a and b). We did not find differential effects of interpregnancy weight loss between women with a normal BMI and women with an overweight BMI on the risk of developing GDM, PIH or delivering an LGA neonate.

Fig. 6
figure 6

Forest plot from random effects meta-analysis showing association between interpregnancy weight change and the risk of developing gestational diabetes, stratified by BMI category at the start of index pregnancy. a Normal weight classified as BMI < 25 kg/m2; b Overweight classified as BMI ≥25 kg/m2. All adjusted odds ratios are relative to the reference category of interpregnancy weight change between − 1 and + 1 BMI units. BMI, body mass index (in kg/m2); aOR, adjusted odds ratio; CI, confidence interval

Full size image
Fig. 7
figure 7

Forest plot from random effects meta-analysis showing association between interpregnancy weight change and the risk of delivering a large-for-gestational age neonate, stratified by BMI category at the start of index pregnancy. a Normal weight classified as BMI < 25 kg/m2; b Overweight classified as BMI ≥25 kg/m2. All adjusted odds ratios are relative to the reference category of interpregnancy weight change between − 1 and + 1 BMI units. BMI, body mass index (in kg/m2); aOR, adjusted odds ratio; CI, confidence interval

Full size image
Fig. 8
figure 8

Forest plot from random effects meta-analysis showing association between interpregnancy weight change and the risk of developing pregnancy induced hypertension, stratified by BMI category at the start of index pregnancy. a Normal weight classified as BMI < 25 kg/m2; b Overweight classified as BMI ≥25 kg/m2. All adjusted odds ratios are relative to the reference category of interpregnancy weight change between − 1 and + 1 BMI units. BMI, body mass index (in kg/m2); aOR, adjusted odds ratio; CI, confidence interval

Full size image

There was an approximate log-linear association between interpregnancy weight gain and the risk of developing GDM (Fig. 9a), PE (Fig. 9b) or PIH (Fig. 9c) and delivering a LGA neonate (Fig. 9d).

Fig. 9
figure 9

Dose-response curve with line of best fit for the increase in odds ratio of developing perinatal complications after interpregnancy weight gain. a Gestational Diabetes. b Pre-eclampsia. c Pregnancy Induced Hypertension d Large-for-gestational age. Where ranges of BMI changes were reported, the midpoint category was utilised (e.g. 1.5 BMI units change for the category weight change between 1 and 2 BMI units). aOR, adjusted odds ratio. BMI, body mass index (in kg/m2)

Full size image

Risk of bias of included studies

After assessing the study selection criteria, comparability of cases and controls and outcome assessments through the NOS, we identified four studies of poor quality (NOS score < 5, Additional file 4: Table S2). However, as these studies did not employ a reference group of ±1 BMI unit, they were already excluded from the meta-analyses. Leave-one-out-analyses showed that removing the study by Villamor et al. made the association between GDM and interpregnancy weight change of 2–3 or > 3 BMI units not significant. We did not find evidence that the outcomes for PE or PIH were driven by one study. For the outcome of delivering an LGA neonate, leave-one-out analyses could not be conducted due to only two studies being included in the meta-analysis.

Discussion

Main findings

This study systematically summarises and examines the published literature on the associations between interpregnancy weight change and several common perinatal outcomes. Our main findings are that interpregnancy weight gain is associated with a higher risk of developing GDM, PE, PIH and delivering an LGA neonate, while interpregnancy weight loss is associated with a lower risk of delivering an LGA neonate. BMI at the start of the index pregnancy possibly modifies the risk of developing GDM, PIH or delivering an LGA neonate after interpregnancy weight gain. Furthermore, we identify a positive approximately log-linear relationship between interpregnancy weight gain and the risk of developing GDM, PE, PIH or delivering a LGA neonate.

Comparison with existing literature

Our study confirms the associations between interpregnancy weight gain and the risk of developing GDM and LGA, as also shown in a recent meta-analysis [13]. Our research additionally summarises the effect of interpregnancy weight change on the risk of developing hypertensive disorders in pregnancy. However, our meta-analysis is to the authors knowledge the first study to show that gaining weight between pregnancies increases the risk of developing hypertensive disorders in the subsequent pregnancy. The observation that starting BMI possibly modifies this association is important for women with a healthy BMI at the start of their index pregnancy, as research often emphasises the risk associated with being overweight or obese, and women with a healthy BMI might not be aware of the risk that comes with (small) interpregnancy weight gain. Although the risks of (excessive) gestational weight gain [14] and high pre-pregnancy BMI [15, 16] on perinatal outcomes are well understood, the effects of interpregnancy weight gain are relatively unknown and are essential to understand in order to guide women in periconception and perinatal weight management.

Our study shows an approximate log-linear association between BMI gain and the risk of developing GDM or hypertensive disorders in pregnancy. This result contributes towards understanding the association between maternal weight and pregnancy complications. Linear dose-response associations are established between obesity and the incidence of GDM, PE and PIH [17], between adiposity and pre-eclampsia [18], as well as maternal weight and pre-eclampsia [15] and GDM [16]. Our identified associations emphasise the detrimental effects of (small amounts of) weight gain, additional to the influence of absolute BMI. This can contribute towards understanding the importance of postpartum weight management and highlights the need for the development of clinical guidelines.

Strengths and limitations

A strength of our study is we ensured a homogenous reference group (i.e. a BMI change ≤1 kg/m2) for our meta-analysis rather than including studies with very different reference groups [13]. Furthermore, we only harmonised studies reporting adjusted odds ratios, which all considered maternal age, country of origin, social economic status and smoking status as potential confounders. Nevertheless, our study has several limitations. First, between-study heterogeneity remained, arising from differences in outcome definitions and demographics, such as parity and age, and potentially differences in length of interpregnancy intervals and prevalence of perinatal complications. Of the studies selected for meta-analysis, only Lynes et al. did not restrict to nulliparous women, although removing this study had little impact on the results. Second, GDM, PE and PIH were either not defined in publications or the definitions of these adverse outcomes differed between studies, hence caution is needed when comparing effect estimates between studies. Third, it was not possible to consistently assess the impact of previous pregnancy complications, which may lead to excessive interpregnancy weight changes and a higher risk of subsequent pregnancy complications. Fourth, studies varied in the way they measured pre-pregnancy weight, with the majority of studies using self-reported weight (and height) to calculate BMI and interpregnancy weight change. Although evidence suggests that maternal reports of pre-pregnancy weight are in general consistent with clinical records [42], bias due to systematic over- or underreporting cannot be excluded. We can also not exclude the possibility of publication bias, as this could not be assessed due to the small number of studies available per adverse outcome and funnel plot assessment is generally not recommended with less than 10 studies [43]. Lastly, we were unable to make the distinction between spontaneous preterm birth and medically induced preterm birth. We hypothesise that an increased risk of preterm birth is at least partly related to the increased risk of carrying an SGA neonate, as (suspected) growth restriction is one of the main causes of medically induced premature birth [44]. However, inadequate nutrition in the context of severe maternal weight loss could also contribute to a higher risk of both SGA and preterm birth [45].

Conclusions

Our study highlights the importance of postpartum weight management, but also identifies opportunities for future research. There is a need to capture the typical weight change profiles of women in various BMI classes, to further elucidate risk groups. This will support further research into weight management strategies, eventually aiming to implement evidence-based weight control interventions to benefit maternal and offspring health. It is particularly important to elucidate strategies for postpartum weight loss in normal weight women, as this group might not be the focus of current research and interventions, yet may be at highest risk of adverse outcomes from interpregnancy weight gain.

In conclusion, we show that interpregnancy weight gain impacts on the risk of developing perinatal complications in a subsequent pregnancy and it is possible that BMI at the index pregnancy modifies these associations. These findings highlight the need to encourage women to return to their pre-pregnancy weight before conceiving again in an effort to reduce the risk of perinatal complications. Future work should focus on defining the most effective strategies to achieve this outcome.