Association between maternal adiposity measures and adverse maternal outcomes of pregnancy: Systematic review and meta‐analysis

Summary Maternal obesity increases pregnancy‐related risks. Women with a body mass index (BMI) ≥ 30 kg/m2 are considered to be at risk and should receive additional care, although approximately half will have uncomplicated pregnancies. This systematic review aimed to identify early pregnancy measures of adiposity associated with adverse maternal health outcomes. Searches included six databases, reference lists, citations, and contacting authors. Screening and quality assessment were carried out by two authors independently. Random effects meta‐analysis and narrative synthesis were conducted. Seventy studies were included with a pooled sample of 89,588 women. Meta‐analysis showed significantly increased odds of gestational diabetes mellitus (GDM) with higher waist circumference (WC) categories (1.40, 95% confidence interval [CI] 1.04, 1.88) and per unit increase in WC (1.31, 95% CI 1.03, 1.67). Women with GDM had higher WC than controls (mean difference [MD] 6.18 cm, 95% CI 3.92, 8.44). WC was significantly associated with hypertensive disorders, delivery‐related outcomes, metabolic syndrome, and composite pregnancy outcomes. Waist to hip ratio was significantly associated with GDM, hypertensive disorders, and delivery‐related outcomes. Fat mass, neck circumference, skinfolds, and visceral fat were significantly associated with adverse outcomes, although limited data were available. Our findings identify the need to explore how useful adiposity measures are at predicting risk in pregnancy, compared with BMI, to direct care to women with the greatest need.


| INTRODUCTION
The prevalence of maternal obesity, usually defined as a prepregnancy body mass index (BMI) ≥ 30 kg/m 2 , has increased in recent decades. In the United Kingdom, recent data published in 2021 suggest that 22% of women start their pregnancy with a BMI in the obese range, 1 an increase from 7.6% in 1989 and 15.6% in 2007. 2 Obesity is associated with an increased risk of multiple adverse pregnancy outcomes that impact on maternal health. These include maternal mortality, gestational diabetes mellitus (GDM), and preeclampsia, as well as long-term health consequences including the development type 2 diabetes. 3,4 Guidelines recommend that women with an obese BMI receive additional antenatal care to reduce their risk of an adverse pregnancy outcome. [5][6][7][8][9] In the context of increasing maternal obesity prevalence, this presents a significant challenge for clinical practice, globally. For example, a national survey of maternity units in England, UK, found that 40% had not implemented guidance to screen all women with a BMI ≥ 30 kg/m 2 for GDM, primarily due to lack of capacity to do so given the high prevalence of maternal obesity. 10 Available evidence suggests that the risk of adverse pregnancy outcome associated with obesity has increased over recent years. A large US study using National Center for Health Statistics birth certificate data found the risk of adverse outcomes associated with obesity had increased between 2013 and 2018. In women from all ethnicities studied, odds ratios (ORs) ranged from 1.27 (95% confidence interval [CI] 1. 25, 1.29) in non-Hispanic Black to 1.94 (1.92, 1.96) in non-Hispanic white women. 11 These data suggest that current strategies for reducing the clinical risk for women with an obese BMI are not working. The reasons for the failure to reduce risk might be attributable, in part, to guidance using BMI to identify which women require additional routine clinical care during pregnancy, such as GDM screening, and to target behavior change interventions.
There has been an abundance of pregnancy weight management interventions that aim to reduce risk of adverse maternal health outcomes, such as GDM. While these interventions appear to be effective in changing maternal behaviors, particularly diet behaviors, 12 and limiting gestational weight gain and postnatal weight retention, 13 the evidence base for effectiveness of these interventions is conflicting relating to reducing the risk of maternal health outcomes such as GDM and preeclampsia. 13 Currently, all women with an obese BMI are considered as being at equal risk of having an adverse pregnancy outcome. However, many women with a BMI ≥ 30 kg/m 2 will not experience an adverse pregnancy outcome, while a substantial proportion of women with a BMI < 30 kg/m 2 will. 14 A multicenter study reported data for uncomplicated pregnancy (defined as normotensive, live birth at >37 weeks, not small for gestational age, and an absence of any other significant pregnancy complications) among 5628 women from the United Kingdom, Ireland, New Zealand, and Australia. 15 The authors found that 47% of women with an obese BMI had an uncomplicated pregnancy, whereas 42% of women with an overweight BMI (25-29.9 kg/m 2 ) did develop pregnancy complications. 15 The intervention and observational evidence base to date suggests that BMI is not a useful tool to use to predict which women are at high risk of an obesity-related adverse outcome of pregnancy and therefore require additional care. Body fat distribution was first identified as being important for health in the 1940s, 16 although there is still debate relating to which measures work best to predict risk. A meta-analysis identified that using BMI to diagnose obesity demonstrated low sensitivity to identify adiposity, failing to identify half of the people with excess body fat (pooled sensitivity 0.50, 95% CI 0.43, 0.57). 17 Waist circumference (WC) has been used as an alternative, or alongside, BMI for a number of years as it has been found to be highly correlated with visceral fat. 18 A large international cardio-metabolic study reported that the frequent discordance between BMI and WC was driven by the substantial variability in visceral fat for a given BMI. 19 Although body fat distribution is well established as being important in terms of degree of risk of experiencing a negative health outcome in the general population, it is less clear if body fat distribution is important in terms of predicting risk of an adverse pregnancy outcome. There is some evidence to suggest that central adiposity is important in terms of risk of GDM 20 and pregnancy hypertension, 21 but further work to confirm this and to establish which measures of body fat distribution are best at predicting the risk of an adverse pregnancy outcome is needed. This systematic review and meta-analysis aimed to identify measures of adiposity that are associated with adverse pregnancy outcomes relating to maternal health, in order to assess which may have potential to predict risk better than the current use of BMI.

| METHODS
The systematic review was registered on PROSPERO (CRD42017064464) and the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines were followed. 22

| Searches and Screening
A rigorous search strategy was implemented to limit the effect of publication bias, as database searches alone for systematic reviews of observational studies are insufficiently rigorous. 23 An experienced information specialist (IS) developed the search strategy following an iterative process in consultation with the review team. The MEDLINE strategy was peer reviewed by another experienced IS using the PRESS checklist. 24 We searched MEDLINE, EMBASE, PsycINFO, CINAHL (EBSCO), JBI Database of Systematic Reviews and Implementation Reports, and Cochrane Library. Using a mixture of controlled search vocabulary (e.g., MeSH) and free text, search terms were derived using the following concepts: "Pregnancy," "Adiposity," "Prediction/Risk," and "Outcomes." "Outcomes" included generic vocabulary to capture all pregnancy outcomes, as well as specific outcomes of interest (Table S1). Following identification of studies that met the inclusion criteria, all reference lists were hand searched and citation searches were carried out using the Google Scholar cited by feature. Finally, authors of included studies were contacted when additional information was required to assess eligibility for inclusion, or for additional data when required for meta-analyses (Table S2). Database searches were completed between February 25 and April 2021. Citation and reference list searches and contacting authors were carried until December 2021.
Inclusion criteria were peer-reviewed studies reporting the association between maternal pre-or early-pregnancy measures of adiposity measured before 20 weeks' gestation and any pregnancy outcomes relating to maternal health, in singleton pregnancies. For the purpose of this review, we classed maternal health outcomes as those that were primarily diagnosed as being a risk to maternal health and well-being (e.g., GDM and preeclampsia), while recognizing that these outcomes also incur risks to the fetus. Mode of delivery outcomes were also classed as being maternal outcomes in this review.
Any outcomes that we classified as being primarily a risk to the fetus or new-born's health, such as gestational age at birth or birthweightrelated outcomes, will be reported elsewhere. Studies restricted to specific sub-populations (e.g., adolescents and those with pre-existing conditions such as polycystic ovarian syndrome or type 2 diabetes) were excluded, with the exception of those who had BMI inclusion criteria as we wanted to explore associations across a range of BMIs.
There were no restrictions applied to the country of study or date of publication. Results of screening are reported using the PRISMA statement. 25 Data extractions were carried out by one researcher using a standardized data extraction protocol (Supplement Information 1), and all data extraction tables were validated by a second researcher (NH, LN, AO, AF, LH, AS, LC, VS). Quality assessments were carried out independently by two researchers using the Newcastle-Ottawa Scales for cohort and case control studies to assess information bias, selection bias, and confounding. 26 Any conflicts in data extraction or quality assessment decisions were either resolved by discussion between the two researchers or by a third researcher. Where multiple publications reported data for the same study population, these were further assessed to ensure duplicate data were removed before anlaysis (Supplement Information 2).

| Analysis
Each combination of early pregnancy adiposity measure (e.g., WC) and pregnancy outcomes (e.g., GDM) was assessed for ability to pool data in a meta-analysis. Meta-analysis was carried out when there were at least three studies reporting data suitable for pooling.
Studies that reported binary or continuous exposure variables were synthesized into separated pooled effect meta-analyses. Similarly, studies that reported mean differences of the adiposity exposure variable within the pregnancy outcome levels were synthesized into a single meta-analysis. When a categorical adiposity variable had more than two levels (e.g., WC < 80 cm compared with 80-88 and >88 cm), the method proposed by Greenland and Longnecker 27 was applied to pool estimates for responses at different levels of the adiposity variable. For each category, the respective OR was assigned to each midpoint (the average of the lower and upper bound). The summary ORs were calculated using the random effects model by restricted maximum likelihood. 28,29 The I 2 statistic was used to assess the heterogeneity among studies, 30 with a threshold of >75% representing significant heterogeneity. 31 Egger's test was used to test publication bias 32 when the meta-analysis included at least 10 studies. 33 Sensitivity analyses were performed by excluding one study at a time from meta-analysis with at least 10 studies. The statistical analyses were conducted using dosresmeta 34 (Table S4). The majority of outcome data related to GDM (n = 45 studies), followed by hypertensive disorders (n = 20; including preeclampsia and pregnancy-induced hypertension), measures of insulin and glucose (in the absence of reporting GDM diagnosis, n = 7), maternal lipids (n = 6), caesarean delivery (n = 5), composite outcomes (n = 4), induction or assisted deliveries (n = 3), metabolic syndrome (n = 2), non-spontaneous labor (n = 1), and gestational weight gain (n = 1) (Table S4).
The quality of studies ranged from a score of five to eight for both cohort and case control study designs (Table S5). No studies were rated as low quality, and the majority of studies were rated as high quality (76.3% for cohort and 72.7% for case control). Cohort studies consistently scored highly (all >70%) on the representativeness of the exposed cohort (Q1), selection of the non-exposed cohort (Q2), ascertainment of exposure (Q3), assessment of outcome (Q5), adequate length of follow up (Q6), and adequacy of follow up (Q7) (Table S5A). However, less than half of the cohort studies controlled for gestational weight gain or any other factors in their analysis (Q4, 42.4%). For case control studies, 100% scored highly for questions relating to case definition (Q1), selection and definition of controls (Q3 and Q4), ascertainment of exposure (Q6), and using the same method of ascertainment for cases and controls (Q7) ( Table S5B). The lowest scoring question related to representativeness of the cases (Q2, 27.3%), followed by controlling for weight gain or additional factors (Q5, 45.5%) and non-response rate (Q8, 63.6%). eight could be pooled in the meta-analysis ( Figure 1). There was a significantly increased odds of developing GDM in categories of higher WC (defined as >80, >78.5, and >84.5 cm) compared with lower categories (OR 1.40, 95% CI 1.04, 1.88) with significant heterogeneity (I 2 99.8%) ( Figure 1). The study 86 that was not pooled in the metaanalysis reported the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for WC to predict GDM (Table S6A).

| GDM
Six studies reported WC as a continuous measure and the associ- Sixteen studies reported case control data and all could be included in a meta-analysis. 37 Figure 3). There was significant heterogeneity (I 2 96.9%) and evidence of publication bias (p = 0.02; Table S6B and Figure S2).

| WHR and GDM
There were six studies 43,65,72,95,102,103 reporting categories of WHR and associations with GDM and all were included in the meta-analysis.
There was a significant increase in odds of GDM for women in the category of high WHR compared with low (OR 2.73, 95% CI 1.67, 4.45) with no significant heterogeneity (I 2 43.5%) ( Figure S3). One F I G U R E 1 Meta-analysis of the association between waist circumference categories and gestational diabetes mellitus. Categories of high waist circumference reported by the included studies were >80 cm (Popova et al., 83 Gao et al., 55 Ebrahimi-Mameghani et al., 51 Zhu et al., 103 87.5%) ( Figure S4).
Six studies reported associations between early pregnancy FM and diagnosis of GDM, 40 79.1%) ( Figure S7). One study reported significantly higher lean leg and arm mass in women with GDM compared with controls and a significantly increased odds of GDM with increasing FFM (kg), and lean arm, leg, and trunk mass 101 (Table S6B) 84.4%) ( Figure S8). Five studies 58,60,66,69,76 reported associations between early pregnancy neck circumference and GDM but could not be pooled in meta-analysis (Table S6A). Three studies 60 (Table S6B). Two studies reported total adipose tissue and GDM.
One reported significantly increased odds for total adipose tissue >7 versus <4.5 cm, but not for measures between 4.6 and 7.0 cm. 20 One reported an AUROC of 0.70 (95% CI 0.62, 0.77) 44 (Table S6A) (Table S6A). There was also a significantly increased mean triceps, bicep, subscapular, suprailiac, abdominal, and sum of SFT in cases of GDM compared with controls (Table S6B).

| Composite adiposity measures and GDM
One study 49

| Insulin-and glucose-related outcomes in the absence of a GDM diagnosis
Seven studies 42,45,50,74,79,82,93 reported data relating to glucose or insulin measures that did not also report a diagnosis of GDM and it was not possible to pool these data in a meta-analysis ( whereas one reported significant associations with HOMA-IR and insulin sensitivity index. 50 There were conflicting data across the four studies 42,50,74,82 reporting visceral fat, and there was no association reported for total adipose tissue. However, visceral fat to subcutaneous fat ratio was significantly correlated with insulinemia and HOMA-IR in one study. 42 One study reported significant associations between bicep and triceps SFT and blood glucose following adjustments for confounding factors. 93

| Hypertensive disorders of pregnancy
There were 20 studies reporting data relating to hypertensive disorders of pregnancy including preeclampsia, pregnancy-induced hypertension, and systolic and diastolic blood pressure, 21 38.1%) (Figure 4). One study 87 reported significantly increased odds of hypertensive disorders per SD increase in WC (AOR 1.78, 95% CI 1.10, 2.89) (Table S8A). There was also a significant positive correlation reported between WC and hypertension 72 and diastolic blood pressure, but not systolic blood pressure 56 (Table S8A). One study 86 compared using Asian specific and general population criteria to predict gestational hypertension and complications (Table S8A).
Four studies 63,87,90,91 were pooled in meta-analysis that showed a significantly higher mean WC among cases of hypertensive disorders compared with controls (mean difference 7.83 cm, 95% CI 3.95, 9.23) with significant heterogeneity (I 2 79.5%) ( Figure 5). There were also three studies 21 51 and Sattar et al. 21 ) and ≥65 cm (Wen et al. 104 ). Data marked as (2) were for preeclampsia; other data were pregnancy-induced hypertension. CI, confidence interval; OR, odds ratio; RE, random effect

| Circumference measures and hypertensive disorders
One study 53 (Table S8A). There was conflicting evidence for FFM. One study reported significantly lower mean muscle and water mass percentage among women with preeclampsia compared with controls, but no difference in bone density 100 (Table S8B), and another 96 found no significant association between high FFM index categories and preeclampsia (Table S8A). Whereas, one study 81 reported significantly increased FFM, and total body water, among women who developed hypertensive disorders of pregnancy (Table S8B).

| Ratios and hypertensive disorders
One study 87 reported no significant association between waist to height ratio and any gestational hypertensive disorders (AOR 1.44, 95% CI 0.83, 2.51) (Table S8A), and two studies 63,87 reported mean waist to height ratio for cases of preeclampsia and gestational hypertension with conflicting results (Table S8B).

| SFTs and hypertensive disorders
One study 81 reported significantly higher median sum of SFTs for women who developed hypertensive disorders of pregnancy with appropriate gestational age (Table S8B).

| Heterogeneity, publication bias, and sensitivity analysis
There was heterogeneity in 10 out of the 14 meta-analyses (I 2 79.1% to 99.8%). However, given that in most of the analyses there were very few studies, no further analyses were performed to identify factors explaining observed heterogeneity. Sensitivity analyses were performed for meta-analyses comprising at least 10 studies.
The analyses showed that none of the studies did substantially influence the overall direction of association, effect size, statistical significance, or heterogeneity. There was evidence of publication bias in the analyses of WC (mean differences) and GDM (p = 0.024).
F I G U R E 5 Meta-analysis of the association between waist circumference (mean differences) and hypertensive disorders. Kausar et al. 63 and Sina et al. 87 reported combined category of preeclampsia or gestational hypertension; Sween et al. 90 and Taebi et al. 91 reported preeclampsia. CI, confidence interval; MD, mean difference (cm); RE, random effect

| Narrative synthesis
It was not possible to conduct any meta-analysis for delivery-related outcomes, maternal lipids, metabolic syndrome, composite pregnancy outcomes, or gestational weight gain. Data for these outcomes have been synthesized narratively.

| Delivery-related outcomes
Seven studies 55,64,72,75,77,85,89 reported outcomes relating to the mode delivery including caesarean delivery, 55,64,75,77,89 instrumental or caesarean delivery (defined as abnormal delivery), 72 and induction or non-spontaneous birth 64,72,85 (Table S9). High category of WC (≥80 cm) was significantly associated with caesarean delivery (AOR 1.71, 95% CI 1.11, 2.63). 55 Increasing WC was also significantly correlated with abnormal delivery and induction 72  total cholesterol, and free fatty acids) and waist and neck circumference, subcutaneous and visceral fat, and WHR. 42,56,65,66,79,82 The data reported were primarily correlations with mixed results. Women with a higher early pregnancy WC had significantly positive correlation and increased TGs (g/L) before and after an OGTT, but no significant correlation with HDL-C, LDL-C, or total cholesterol (Table S10).
There was no significant correlation between neck circumference and  (Table S10).

| Metabolic syndrome
Two studies reported case control data for maternal metabolic syndrome during pregnancy and in the immediate postpartum period 56,70 and waist, arm, and leg circumference, subcutaneous and visceral fat, and triceps and suprailiac SFTs (Table S11). Women who developed metabolic syndrome in pregnancy and postpartum had significantly increased early pregnancy measures of WC and SFT, but mixed results for all other measures. One study 56 found that both visceral and subcutaneous fat thickness were significantly higher among cases than controls, whereas the other 70 only found a significant association with subcutaneous fat thickness and postpartum metabolic syndrome. There was a significantly increased arm circumference among women with metabolic syndrome diagnosed in pregnancy but not postpartum, and no significant association with leg circumference. 70

| Composite adverse pregnancy outcomes
There were four studies that reported composite outcomes 54,55,65,89 (see Table S12 54 Case control analysis showed no significant difference between early pregnancy FM of women who developed adverse pregnancy outcomes compared with those who did not (Table S12).

| Gestational weight gain
Only one study reported gestational weight gain as an outcome, 68 which was significantly negatively correlated with FM (Pearson's r À0.24, p < 0.0001) (Table S13).  107 The heterogeneity in methods of analysis and reporting presents challenges when trying to pool data to directly compare different adiposity measures. Using an individual participant data (IPD) meta-analysis approach could help to overcome some of these challenges by obtaining the raw data to standardize analysis approaches across studies. 108,109 IPD meta-analysis would also facilitate the incorporation of data from additional studies that have not published associations between maternal adiposity and pregnancy outcomes, addressing potential implications of publication bias.

| DISCUSSION
For example, there were many studies excluded from this systematic review as they did not report associations between adiposity measures and outcome variables despite collecting these data 110,111 ; an IPD meta-analysis could incorporate the inclusion of these datasets.
This alternative approach to meta-analysis would enable a direct comparison of adiposity measures to determine which might be best at predicting risk of a range of adverse pregnancy outcomes. 108 We also transformed data where possible to increase the number of studies possible to be pooled in meta-analysis. However, a key limitation relates to the significant heterogeneity that was present in all but four meta-analyses. We had a limited number of studies in each metaanalysis, which meant we were not able to explore sources of heterogeneity using meta-regression as was planned. The low number of studies that could be pooled in each individual meta-analysis also meant that the usefulness of exploring publication bias and performing sensitivity analysis was limited. Finally, although we did not limit our search by type of pregnancy outcome, we identified only a few studies reporting associations between maternal adiposity and delivery outcomes or gestational weight gain, and no studies reporting maternal mental health, hemorrhage, infection, or breastfeeding outcomes, which are all significantly associated with maternal BMI.
Future adiposity studies should explore a wider range of outcomes relating to maternal health and well-being.
The evidence base to date shows that large-scale behavioral interventions that aim to reduce the risks associated with maternal obesity have been successful at improving maternal behavior and weight-related outcomes, 12 which may be viewed as being a public health success, but have yet to consistently significantly reduce the impact of obesity on clinical outcomes such as GDM. 13 However, there is a consistent direction of effect across multiple meta-analyses of interventions, which suggests potential for a reduction in risk, although there is a lack of statistical significance. 13 Therefore, interventions may be more successful in consistently preventing adverse outcomes associated with obesity with better targeting.
A primary aim of prenatal care is to improve health outcomes for both mother and baby. Clinicians have a role to assess the degree of risk for each pregnant woman they see and plan patient centered and individualized care with them. Current clinical guidelines use BMI to determine individual risk in pregnancy, which does not provide an accurate measure of adiposity or individual health risks, and this practice is unlikely to be cost-effective at preventing adverse outcomes. A large proportion of women will not experience the adverse pregnancy outcomes that population studies show they are significantly at risk of developing with a BMI ≥ 30 kg/m 2 . Yet BMI is used in the clinical context as a screening tool to determine individual risk and which women need additional antenatal care. This could result in unnecessary clinical intervention and reduced birth and care choices for these women. Importantly, this also potentially overlooks women with a BMI < 30 kg/m 2 who have high adiposity but are not currently deemed to need additional care. This systematic review and meta-analysis has identified a number of potential early pregnancy adiposity measures that could be used in routine clinical care to identify women at increased risk of adiposity-related adverse outcomes. Our meta-analysis has identified some promising evidence to help inform clinical practice, for example, relating to WC and WHR and the risk of GDM. However, further research is needed to explore whether these measures work better than BMI at predicting risk of adverse pregnancy outcomes, or if they could be used in combination with BMI or other predictor variables in a risk prediction model. It is essential that future studies prioritize adiposity measures that can be easily implemented into routine maternity care. Further research should compare these measures to determine which could be used most effectively to direct early intervention to women who need it most, to support the best chance of good pregnancy outcomes.