Gestational weight gain (GWG) is an important indicator of health during pregnancy. Sub-optimal weight gain has been linked with adverse outcomes for both mother and child, including small- and large-for-gestational age birth, gestational diabetes and pre-eclampsia(Reference Goldstein, Abell and Ranasinha1–Reference Hutcheon, Stephansson and Cnattingius3). Interventions during pregnancy to optimise weight gain have been shown to improve several of these outcomes(Reference Cantor, Jungbauer and McDonagh4), highlighting the importance of monitoring weight gain in antenatal care.
Charts to monitor GWG can be standards (prescriptive charts of how weight gain ought to be) or references (descriptive charts of what weight gain was like in an unselected population)(Reference Ananth, Brandt and Vintzileos5,Reference Cole6) . From a theoretical perspective, standards are preferable for identifying individuals at risk of insufficient or excessive weight gain in clinical practice(Reference Ananth, Brandt and Vintzileos5). Yet, from a practical perspective, standards are more challenging to create because they require detailed information on all conditions that can adversely affect pregnancy weight gain, such as maternal diet, physical activity(7) and adverse postpartum outcomes such as excess postpartum weight retention and child overweight/obesity. This detailed information is often not available in cohorts used to create such charts: none of the nine pregnancy weight gain charts created in the past 10 years have incorporated diet, exercise or postpartum outcomes into their exclusion criteria(Reference Hutcheon, Platt and Abrams8–Reference Thiruvengadam, Desiraju and Natchu16). Most studies excluded individuals with hypertension, diabetes, hypertensive disorders of pregnancy and/or gestational diabetes(Reference Hutcheon, Platt and Abrams8–Reference Kac, Carrilho and Rasmussen15). Adverse neonatal outcomes, such as the birth of small-/large-for-gestational age infants, with low birth weight, macrosomia or with congenital abnormalities, were excluded in about half of the charts(Reference Xu, Luntamo and Kulmala9,Reference Santos, Eekhout and Voerman13–Reference Kac, Carrilho and Rasmussen15) .
Collection of this detailed information can be burdensome and is only warranted if the exclusion of individuals with the adverse conditions of interest will meaningfully impact the percentile values of the weight gain chart. Identifying which exclusion criteria meaningfully impact those percentiles will support efficient use of resources in the design of new pregnancy weight gain standards and help to identify the extent to which the utility of existing charts is decreased by the lack of information on many adverse lifestyle factors. Thus, we aimed to identify how normative values of pregnancy weight gain are affected by the exclusion of individuals with pre-pregnancy, pregnancy and postpartum factors adversely linked with pregnancy weight gain.
Methods
Study sample
We used data from the nuMoM2b-Heart Health Study (HHS), a 2–5-year follow-up study of the Nulliparous Pregnancy Outcomes Study: monitoring mothers-to-be (nuMoM2b) cohort. Briefly, the nuMoM2b was a cohort study following nulliparous pregnant women and other nulliparous pregnant individuals in eight US clinical centres, with the recruitment occurring from 2010 to 2013. Participants were recruited from 6 weeks + 0 d to 13 weeks + 6 d of pregnancy if they had a viable singleton pregnancy and no previous pregnancy lasting 20 weeks or more. Study visits occurred in each pregnancy trimester. At delivery, chart abstraction was performed by trained research team members(Reference Haas, Parker and Wing17). The HHS followed participating individuals after delivery, with continuous monitoring of eligible participants through telephone interviews and an in-person visit up to 7 years after delivery (median 3·2 years). In this in-person visit, information on psychosocial and medical characteristics was collected through questionnaires. Clinical measurements and biological specimens were also obtained. A detailed protocol of the HHS is available elsewhere(Reference Haas, Ehrenthal and Koch18).
For this study, we limited our analysis to HHS participants who did not have a stillbirth or neonatal death in their nuMoM2b pregnancy, had an in-person visit from 2 to 5 years postpartum and had complete data for our key variables of interest. The dataset of individuals meeting these minimum inclusion criteria is referred to as the ‘full cohort’ in this study.
Weight and weight gain measurements
Weight gain during pregnancy (kg) was calculated as the difference between the weight measured in the study visits and self-reported pre-pregnancy weight. Four weight gain measurements were available for this study (at 1st, 2nd, 3rd trimesters and at delivery). Weight gain measurements > + 6 or < −6 sd for gestational age(Reference Hutcheon, Platt and Abrams8,Reference Hutcheon, Platt and Abrams10) were considered outliers and removed from the analysis. This conservative cut-off was chosen to minimise the loss of weight gain values that are extreme, but still plausible.
We selected specific windows for comparison, considering the available sample size in each gestational week: first visit weight gain (measurements obtained between 11 and 14 weeks); second visit weight gain (measurements from 27 to 30 weeks) and at delivery (measurements obtained between 26 and 42 weeks). The broader range of gestational ages for weight gain at delivery enabled us to examine the impact of excluding individuals with adverse perinatal outcomes such as preterm birth, which would not have been possible if we limited total weight gain to a narrower window (e.g. term only). However, because the window of gestational ages in which delivery happened varied from 26 to 42 weeks, we also compared z scores of weight gain at delivery according to gestational age and pre-pregnancy BMI(Reference Hutcheon, Platt and Abrams8,Reference Hutcheon, Platt and Abrams10) .
Maternal pre-pregnancy BMI (kg/m2) was calculated using self-reported pre-pregnancy weight and maternal height obtained in the first study visit. BMI was classified according to the WHO cut-offs: underweight, BMI < 18·5 kg/m2; normal weight, BMI ≥ 18·5 and < 25·0 kg/m2; overweight, BMI ≥ 25·0 and < 30·0 kg/m2; and obesity, BMI ≥ 30 kg/m2(19).
Pre-pregnancy, pregnancy and postpartum exclusion criteria
We identified pre-pregnancy, pregnancy and postpartum factors available in the HHS dataset that could be considered as exclusion criteria when creating a pregnancy weight gain standard based on the 2009 Institute of Medicine report’s chapter on weight gain determinants(7) and on adverse pregnancy outcomes related to weight gain(Reference Rogozinska, D’Amico and Khan20) (Table 1). We additionally included information related to sleep quality, as recent work suggests this may be a determinant of weight gain(Reference Hawkins, Pokutnaya and Bodnar36). The criteria were classified into five groups of associated characteristics for parsimony: Group 1 – Socio-economic and demographic characteristics; 2 – Pre-existing co-morbidities and maternal pregnancy complications; 3 – Maternal lifestyle and behaviour factors; 4 – Adverse neonatal outcomes; and 5 – Longer-term postpartum outcomes.
Pre-pregnancy, pregnancy and postpartum exclusion criteria
Statistical analyses
We used absolute (n) and relative (%) frequencies to describe the characteristics of study participants. We established the impact of different exclusion criteria for a weight gain standard by comparing the median, 25th and 75th percentiles of weight gain in the full cohort with the median, 25th and 75th percentiles obtained after applying each of the five groups of exclusion criteria. We removed all characteristics in each group together for ease of presentation but planned a priori to evaluate the impact of individual exclusion criteria if meaningful differences in weight percentiles were observed after excluding the group they were in. We did not explore more extreme percentiles (e.g. 5th or 10th percentiles) due to the small sample sizes remaining after applying many exclusion criteria.
A group of variables was considered to have a meaningful impact on weight gain if the difference between the median, 25th or 75th percentiles of the full cohort and the sub-cohort excluding that group was > 0·75 kg. This value was considered to be meaningful from a clinical/public health perspective, as systematic reviews and meta-analysis of pregnancy weight gain interventions have shown that the mean weight gain changes of this amount are not associated with a significant reduction in the risks of maternal and infant adverse outcomes(Reference Cantor, Jungbauer and McDonagh4,Reference Teede, Bailey and Moran37) . For the z scores of weight gain at delivery, differences of 0·15 z score were considered meaningful. This value corresponds to a 0·75 kg change in weight gain at 40 weeks based on charts created for normal-weight US women(Reference Hutcheon, Platt and Abrams8).
For these meaningful variables, 95 % CI for the median, 25th and 75th percentiles of GWG were calculated using quantile regressions. We examined the distribution of weight gain at delivery for our primary analysis, and weight gain at the 1st and 2nd trimester visits in secondary analyses. We also constructed density plots for GWG at delivery in the full cohort and in the sub-cohorts created after the exclusion of individuals with an adverse characteristic in each group. These analyses were conducted in the complete dataset. We conducted a sensitivity analysis in which we stratified by pre-pregnancy BMI (participants with normal weight, overweight and obesity only). Our decision to combine all BMI groups in our primary analysis was based on the relatively small number of participants with overweight and obesity after exclusions.
Results
The initial sample of the HHS study included 4405 participants. Excluding nineteen stillbirth/neonatal deaths and 1208 individuals with missing data for our key variables left 3178 participants for analysis (online Supplementary Fig. 1). The proportion of missing data for each of the key characteristics is shown in online Supplementary Table S1. The variables with the highest degree of missing values were diet quality (15 %), sleep quality (11 %) and maternal anxiety (10 %). There were no meaningful differences in the distribution of weight gain at delivery between individuals excluded due to missing data and those retained in our analytic cohort (online Supplementary Fig. 2).
In the full cohort, 114 (3·6 %) participants were classified with underweight, 1753 (55 %) with normal weight, 670 (21 %) with overweight and 641 (20 %) with obesity. The majority of individuals with obesity (52·2 %) were classified in class 1 (BMI ≥ 30·0 and < 35·0 kg/m2). Only fifty (1·6 %) participants had experienced weight loss at delivery (data not shown in tables). As shown in Table 2, most participants had been to college (85 %) and had private health insurance (75 %). Nine (9 %) per cent of participants developed pre-eclampsia, 20 % had an unplanned caesarean delivery and 44 % had another disease during pregnancy, the most common of which was anaemia (approximately 11 %). Most participants (54 %) were sedentary (practiced < 450 MET h/week of physical activity) and had a diet quality score under 70 of 100 possible points (66 %). Twelve per cent experienced depression, while 19 % described their sleep in the previous four weeks as restless.
Description of the sample according to the selected groups of exclusion criteria (n 3178 individuals from the Heart Health Study) (Numbers and percentages; median values and interquartile ranges)
GWG, gestational weight gain.
* Refers to the category considered ‘sub-optimal’.
† Weight gain until diagnostic of pre-eclampsia: median: 9·3 (5·9, 13·3) kg; distribution of z scores of GWG at delivery: No: 0·05 (–0·54, 0·63)/Yes: 0·43 (–0·13, 0·99).
‡ Weight gain until diagnostic of gestational diabetes: median: 7·7 (3·6, 11·2) kg; distribution of z scores of GWG at delivery: No: 0·07 (–0·50, 0·66)/Yes: −0·15 (–0·75, 0·61).
§ Other diseases included anaemia, hypo/hyperthyroidism, heart diseases, kidney diseases, sickle cell disease, bleeding disorders, autoimmune diseases, mental health conditions, gynaecological conditions and sexually transmitted diseases.
The prevalence of adverse neonatal outcomes was below 10 % for each of the evaluated outcomes, while longer-term adverse outcomes were more common: 41 % of the mothers retained more than 5 kg at 2–5 years, with 21 % retaining more than 10 kg and 16 % had postpartum metabolic syndrome. Twenty-one per cent of the children were classified with overweight or obesity according to BMI.
As expected, weight gain at delivery varied across categories of many of the pre-pregnancy, pregnancy and postpartum factors (Table 2). For example, median weight gain among participants reporting greater nausea/vomiting was 14·2 kg compared with those reporting the least symptoms (15·9 kg), while median weight gain was 16·4 kg among those with an unplanned caesarean delivery compared with 15·0 kg among those without. Nevertheless, the absolute differences in median weight gain between categories were small in magnitude, with differences < 2 kg for most variables. Of note, median weight gain among participants who self-reported a poor diet quality or a sedentary lifestyle was within 0·4 kg of the median weight gain among those who reported a high diet quality or the most active lifestyle. The largest differences in weight gain between categories were observed for gestational diabetes, preterm birth, low birth weight, macrosomia and excess postpartum weight retention.
Excluding participants with group 1, 2, 3 or 4 exclusion criteria had virtually no impact on the 25th, median or 75th percentiles of weight gain at delivery (Fig. 1). Weight gain percentiles were only meaningfully different after excluding participants in group 5 (longer-term adverse outcomes). Exclusion of participants in group 5 shifted the weight gain distribution to the left (Fig. 1), with differences most pronounced at the upper end of the distribution: the 75th percentile of weight gain decreased from 19·6 kg in the full cohort to 17·8 kg after participants with excess postpartum weight retention > 5 kg, metabolic syndrome or children with overweight/obesity were excluded. This difference was less pronounced for the weight gain at the second trimester and when excess postpartum weight retention was defined using 10 kg as a cut-off (Table 3). Results are similar when the 95 % CI are considered (online Supplementary Table 2).
Distribution of gestational weight gain at delivery (26–42 weeks) in the full cohort and after the exclusion of the participants with each of group of conditions. Notes: Group 1: Socio-economic and demographic variables, Group 2: Pre-existing co-morbidities and maternal pregnancy complications, Group 3: Maternal lifestyle and behaviour factors, Group 4: Adverse neonatal outcomes, Group 5: Longer-term postpartum outcomes.
Distribution of gestational weight gain at each study visit in the full and sub-cohorts – all individuals (Median values and interquartile ranges)
GWG, gestational weight gain (kg).
Group 1: Socio-economic and demographic variables.
Group 2: Pre-existing co-morbidities and maternal pregnancy complications.
Group 3: Maternal lifestyle and behaviour factors.
Group 4: Adverse neonatal outcomes.
Group 5: Longer-term postpartum outcomes (weight retention > 5 kg).
Group 5b: Longer-term postpartum outcomes (weight retention > 10 kg).
* Sample sizes for visits 1 and 2 vary due to the selection of weeks, missing data and outliers of GWG in each visit.
Examination of the individual exclusion criteria within group 5 (longer-term adverse outcomes) suggested that the exclusion of participants with excess postpartum weight retention was responsible for influencing weight gain percentiles. For example, the 75th percentile in the full cohort was 19·6, decreasing to 17·9 and 18·3 kg after excluding participants with excess postpartum weight retention > 5 kg > 10 kg, respectively. Exclusion of children with overweight/obesity or participants with metabolic syndrome did not meaningfully alter the weight gain distribution (Table 4).
Distribution of gestational weight gain at delivery according to the variables of the groups with meaningful differences (group 5 – longer-term postpartum outcomes) (Median values and interquartile ranges)
GWG, gestational weight gain (kg).
The exclusion of individuals with any of the pre-pregnancy, pregnancy or postpartum risk factors potentially related to sub-optimal weight gain (i.e. in any of groups 1–5) left only 137 participants (approximately 4 % of the original cohort). The median and the 75th percentile for weight gain at delivery between the full cohort and the sub-cohort with all exclusion criteria applied differed by > 0·75 kg (Table 3), and there was a shift to lower values in the weight gain distribution (Fig. 1(f)). These changes were due to the exclusion of participants in group 5 (longer-term adverse outcomes). Similar results were observed for the z scores of weight gain at delivery (online Supplementary Table 3).
When analyses were stratified according to pre-pregnancy BMI, similar results were also observed for participants with normal weight (online Supplementary Fig. 3 and online Supplementary Table 4). For participants with overweight and obesity, the exclusion of participants with group 3 characteristics (maternal lifestyle and behaviour) caused a shift in the distribution of weight gain at delivery (online Supplementary Figures 4 and 5). However, these differences may be chance findings given the lower sample size in these two BMI categories (online Supplementary Table 4). Only group 5 exclusion criteria (longer-term adverse outcomes), and more specifically, excess postpartum weight retention (> 5 and > 10 kg), changed the weight gain distribution in all BMI categories (online Supplementary Tables 5–7).
Discussion
In this large US cohort of nulliparous pregnancies, we found that excess postpartum weight retention (> 5 or > 10 kg) was the only exclusion criterion to meaningfully impact weight gain percentiles. Exclusion of participants with excess postpartum weight retention lowered 75th percentile values by approximately 2 kg. Otherwise, pregnancy weight gain percentiles were not meaningfully impacted by the exclusion of participants with adverse pre-pregnancy, pregnancy or postpartum factors linked with sub-optimal weight gain. This suggests that most potential exclusion criteria for a pregnancy weight gain standard may have little practical impact on the chart percentiles, although our analysis cannot rule out an impact on the extremes of the distribution.
Our finding that most exclusion criteria have little practical impact on the percentile values of a pregnancy weight gain chart was unexpected, given that the selected factors are recognised in the literature as determinants or outcomes of weight gain. Yet, similar findings have been observed in the context of fetal weight charts(Reference Koren, Boskovic and Hard23). Hutcheon and Liauw observed that the distribution of estimated foetal weight at 32–33 weeks’ gestation, including the value of the 10th percentile used to define small-for-gestational-age birth, was virtually identical in a reference population and a standard population to which an extensive list of exclusion criteria had been applied(Reference Hutcheon and Liauw38).
Our team has recently evaluated the impact of excluding pregnancies with adverse neonatal outcomes on the resulting weight gain percentiles using studies conducted in low- and middle-income countries(Reference Carrilho, Wang and Hutcheon39). We found that chart percentiles for normal and overweight pregnant women and individuals was unchanged when those who gave birth to neonates with adverse outcomes were removed from the sample(Reference Carrilho, Wang and Hutcheon39). Our team also evaluated the impact of including women and individuals with high interpregnancy weight change (as a proxy for excess postpartum weight retention) on the percentile values of GWG charts in a large, population-based cohort from Sweden(Reference Socha, Johansson and Bodnar40). In contrast with our findings in this cohort, exclusion of pregnant women and other individuals with high interpregnancy weight change had no impact on chart percentiles. The prevalence of interpregnancy weight change ≥ 5 kg in the Swedish data was lower than we observed in this study (34 v. 40 %), potentially explaining the discrepancy in findings.
We speculate that the extent to which a potential exclusion criterion influences the values of percentiles depends on both its prevalence and strength of association with GWG. Although several of the groups of variables in our study had high combined prevalences (e.g. groups 2 and 3 excluded > 70 % of the cohort), they did not meaningfully impact the distribution of weight gain throughout pregnancy due to their low strength of the association with weight gain (as seen in Table 2). The sole exclusion criterion to meaningfully impact weight gain percentiles, excess postpartum weight retention, was both strongly associated with weight gain and had a high prevalence among the participants of the current study.
Another important finding of this study was that only 4 % of the original cohort remained after applying all possible exclusion criteria. This decrease in sample size highlights the practical burden that extensive exclusion criteria create at the data collection stage when using a prospective cohort for chart creation. It also raises concerns about the generalisability of the cohort that remains after applying extensive exclusion criteria. Although pregnancy weight gain standards are not intended to be constructed with representative samples, a chart created using only 4 % of the original population may raise concerns with face validity.
Existing pregnancy weight gain charts(Reference Hutcheon, Platt and Abrams8–Reference Thiruvengadam, Desiraju and Natchu16) have varied in their inclusion/exclusion criteria. However, none of the existing pregnancy weight gain charts, including the standards proposed by the INTERGROWTH-21st project(Reference Cheikh Ismail, Bishop and Pang11), excluded participants with excess postpartum weight retention. Our findings suggest that published charts may therefore normalise higher weight gains that may in practice increase an individual’s risk of excess postpartum weight retention.
The primary strength of this study is the availability of detailed information on maternal psychosocial factors, diet quality, physical activity and longer-term maternal and child health outcomes. These variables allowed us to explore the impact of a detailed list of possible exclusion criteria on weight gain percentiles. Nevertheless, our sample size was not large enough to evaluate the impact of those variables in underweight individuals, and the final sample size for women and individuals with overweight and obesity was also low (< 20). Further, the NuMoM2b/HHS only included nulliparous women, potentially limiting generalisability if the association between possible exclusion criteria and weight gain differs by parity, or the prevalence of the exclusion criteria is meaningfully different in parous women. Finally, regarding diet and physical activity, although they were obtained based on self-reported information and are prone to error, those types of instruments are likely what would be used in a real-world prospective study seeking to create a standard, due to practical challenges in obtaining more objective measures in large population cohorts.
Most potential inclusion/exclusion criteria for a pregnancy weight gain standard will have little practical impact on the resulting chart percentiles. However, pregnant women and other individuals with excess postpartum weight retention should be excluded from cohorts used to create pregnancy weight gain standards, as retaining these individuals will result in chart percentiles that are up to 2 kg higher. Future simulation studies may be useful in identifying the combination of a characteristic’s prevalence and its strength of association with pregnancy weight gain necessary for an exclusion criterion to influence the percentile values of pregnancy weight gain standards. Repeating this analysis in studies from different global settings and integrating findings is a fundamental step towards informing the inclusion/exclusion criteria for global weight gain standards.
Acknowledgements
The authors acknowledge NICHD DASH for providing the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-be data that was used in early stages of this research.
This work was support by an R01 HD094777 to L.M.B. and J.A.H. T.R.B.C. is a Michael Smith Health Research BC Research Trainee (RT 2022-2582). J.A.H. holds a Canada Research Chair in Perinatal Population Health. K.J. holds an Assistant Professor position from Karolinska Institutet. The NuMoM2b and the HHS were supported by grant funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD): U10 HD063036; U10 HD063072; U10 HD063047; U10 HD063037; U10 HD063041; U10 HD063020; U10 HD063046; U10 HD063048; and U10 HD063053. The studies were also supported by cooperative agreement funding from the National Heart, Lung, and Blood Institute and the Eunice Kennedy Shriver National Institute of Child Health and Human Development: U10 HL119991; U10 HL119989; U10 HL120034; U10 HL119990; U10 HL120006; U10 HL119992; U10 HL120019; U10 HL119993; and U10 HL120018. Support to the NuMoM2b and the HHS was also provided by the National Institutes of Health: Office of Research on Women’s Health through U10 HL119991; Office of Behavioral and Social Sciences Research through U10 HL119991 and U10-HL119992; and the National Center for Advancing Translational Sciences through UL 1TR000124, UL 1TR000153, UL 1TR000439 and UL 1TR001108; and the Barbra Streisand Women’s Cardiovascular Research and Education Program, and the Erika J. Glazer Women’s Heart Research Initiative, Cedars-Sinai Medical Center, Los Angeles. Support was also provided by respective Clinical and Translational Science Institutes to Indiana University (UL1TR001108) and University of California Irvine (UL1TR000153). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.
T. R. B. C. and J. A. H. planned the data analyses. T. R. B. C. analysed the data and wrote the manuscript with input from L. M. B., K. J., G. K. and J. A. H. T. R. B. C. and J. A. H. had primary responsibility for the final content. All authors read and approved the final manuscript.
The authors declare none.
The nuMoM2b and the HHS were conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving research study participants were approved by the Institutional Review Boards from the participating institutions. Written informed consent was obtained from all subjects/patients. This secondary analysis of de-identified data was waived from ethics review by the Research Ethics Board of the BC Children’s & Women’s Hospital.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114524001855
Open access
