Systematic Review of the Methodological Quality of Studies Aimed at Creating Gestational Weight Gain Charts^,

Abstract

A range of adverse outcomes is associated with insufficient and excessive maternal weight gain in pregnancy, but there is no consensus regarding what constitutes optimal gestational weight gain (GWG). Differences in the methodological quality of GWG studies may explain the varying chart recommendations. The goal of this systematic review was to evaluate the methodological quality of studies that aimed to create GWG charts by scoring them against a set of predefined, independently agreed-upon criteria. These criteria were divided into 3 domains: study design (12 criteria), statistical methods (7 criteria), and reporting methods (4 criteria). The criteria were broken down further into items, and studies were assigned a quality score (QS) based on these criteria. For each item, studies were scored as either high (score = 0) or low (score = 1) risk of bias; a high QS correlated with a low risk of bias. The maximum possible QS was 34. The systematic search identified 12 eligible studies involving 2,268,556 women from 9 countries; their QSs ranged from 9 (26%) to 29 (85%) (median, 18; 53%). The most common sources for bias were found in study designs (i.e., not prospective); assessments of prepregnancy weight and gestational age; descriptions of weighing protocols; sample size calculations; and the multiple measurements taken at each visit. There is wide variation in the methodological quality of GWG studies constructing charts. High-quality studies are needed to guide future clinical recommendations. We recommend the following main requirements for future studies: prospective design, reliable evaluation of prepregnancy weight and gestational age, detailed description of measurement procedures and protocols, description of sample-size calculation, and the creation of smooth centile charts or z scores.

Introduction

Monitoring maternal gestational weight gain (GWG) has been part of antenatal care since the early 20th century despite a lack of agreement on the optimal range of GWG (1–5).

Excessive GWG has been associated with a range of conditions, including gestational diabetes (1, 6–9), pre-eclampsia, pregnancy-induced hypertension (1–5, 10–12), macrosomia (1, 6–9), increased risk of Caesarean delivery (13–16), and postpartum weight retention (2, 17). Conversely, insufficient weight gain in pregnancy is associated with poor fetal growth, preterm birth, and difficulty establishing breastfeeding (1, 6–8, 18–20), which remain major problems in many world regions.

Recommendations are available in various formats: longitudinal charts (21, 22), single weight gain targets for the entire pregnancy (2, 17), targets based on BMI (kg/m²) (9), and proportional increases based on antenatal “weight for height” (23, 24). This makes comparisons and the process of evidence synthesis of methodological quality challenging. Despite these challenges, there is general agreement that prepregnancy BMI should be between 18.5 and 24.9 and that GWG should be lost in the postpartum period (3–5, 25, 26).

The literature on the subject is vast, with numerous reports published since the early 1970s, including the landmark reports Maternal Nutrition and the Course of Pregnancy (Food and Nutrition Board, 1970) and Nutrition During Pregnancy—Weight Gain (Institute of Medicine, 1990) (27, 28). In 2009, the Institute of Medicine published to our knowledge the most comprehensive synthesis of GWG recommendations to date (9). Total GWG targets are commonly used in clinical practice, but their use has been questioned because of their lack of precision and uncertain benefit in low-risk populations (29). The longitudinal presentation of GWG could offer advantages. First, it would enable women to track their own weight change visually and instigate health interventions as appropriate. Second, it would provide clinicians with a more precise tool for monitoring GWG, taking into account the length of gestation and rate of weight gain throughout pregnancy. However, there has been little attention to studies that created GWG charts or their methodology.

The number and quality of studies of GWG charts is unknown. Therefore, we aimed to identify and compare the methodological quality of published GWG charts and then make recommendations for the essential reporting criteria for future research in this field.

Methods

This review was conducted and reported using the checklist proposed by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (30). Electronic databases (MEDLINE, EMBASE, Web of Science, and Science Direct) were systematically searched to identify studies that aimed to create GWG charts between 1960 and 2014. Before 1960, patient characteristics, data analysis, and study design were deemed unlikely to be comparable to contemporary studies. The search was limited to full articles in English because it was not possible to identify 2 independent reviewers with a third mediator in other languages.

Primary research articles were included if 1) participants were pregnant; 2) weight was reported in pounds or kilograms; and 3) the study had a longitudinal or cross-sectional design, the main objective of which was to create a GWG chart describing changes throughout gestation. Articles were excluded if 1) they were validation studies of previously described charts or review articles or recommendations synthesized from several sources [e.g., the Institute of Medicine 2009 guidelines (9)]; 2) the aim was to describe GWG limited to a very specific population (e.g., adolescents, women with medical conditions); 3) GWG recommendations were not provided in a format describing change throughout pregnancy; or 4) the primary outcome was not maternal.

A keyword search strategy was conducted in consultation with a specialist medical librarian using medical subject heading terms related to pregnancy (gestational, pregnancy, pregnancy outcomes, or pregnant women); weight (body weight, body weight changes, ideal body weight chart, or weight gain); and charts (charts, body weight charts, or reference values). Combinations of the following keyword search terms were also included: curves, centiles, maternal weight, pregn*, chart*, gestational weight gain, range, reference, trend, and gest*.

One reviewer screened the titles and abstracts of all identified citations and selected potentially eligible studies for which full-text articles were obtained and screened further for eligibility. Reference lists of retrieved articles were examined for additional and relevant citations. The list of articles was assessed for duplications, and any duplicates were removed. Two reviewers independently checked and scored the methodological quality of each eligible article against the criteria specified in Table 1. Discrepancies in scoring were discussed, and a third reviewer consulted if needed. Statistical criteria were independently checked and scored by a medical statistician.

TABLE 1.

Methodological quality criteria¹

Domain	Low risk of bias	High risk of bias	Rationale
Study design (maximum points)
Aim of publication (2)	Clearly stated to produce a chart or recommendations for maternal aims in abstract and paper are the same weight gain during pregnancy	Aim not clearly defined; different aims in abstract and paper	More reliable charts will be obtained if the primary purpose of the study was to create them
Definition of target population (1)	Target population clearly defined, i.e., geographical location, ethnic group	Not clearly defined	Determines population for which the chart can be appropriately applied
Definition of reference or standard (1)	Authors state if the trend in weight gain is a reference or standard	Not clearly defined	Reference and standards charts are different tools and, therefore, users must know which they are
Design (1)	Clearly described and longitudinal	Not reported or not longitudinal	Good scientific practice
Sample selection (6)	Population-based; clearly defined inclusion and exclusion criteria; women selected and enrolled consecutively and in accordance with target population; planned study before enrolment	Not population-based; inclusion/exclusion-based criteria not clearly defined; convenience sampling or not described; unplanned study: retrospective study or secondary analysis of data collected for another research study	Good scientific practice
Prepregnancy weight and measurements during pregnancy (2)	Prepregnancy weight recorded; each woman’s weight measured at least every mo	No prepregnancy weight taken or based on maternal recall alone; no clear documentation of intervals between weight measurements or less than monthly	Poor reliability of measurements increases dispersion
Inclusion/exclusion criteria (1)	Prescriptive approach to sample selection; i.e., only women at low risk of complications in a defined population included	Not clearly documented; general population sample	Enable creation of standard
Sample size (1)	Documented determination/calculation of sample size and justification	No documented determination/calculation of sample size and justification	Good scientific practice; more reliable charts created
Data collection (2)	Prospective study; data collected specifically for creating GWG charts	Retrospective study	Good scientific practice
Weight evaluation (3)	Weight measured using standardized instruments; instruments and protocols reported in detail, e.g., calibration schedule, training of staff, and gestation ages at which measurements taken documented	Standardized instruments not used or not described/reported	Poor reliability of measurements increases dispersion
Number of observations for each stage of pregnancy (1)	Reported	Not reported	Precision of estimates increases with the number of observations
Gestational age determination (1)	Clearly described and reliable; i.e., last menstrual period confirmed by ultrasound scan at 9–14 wk	Not described	Charts will be based on gestational age
Statistical methods
Number of measurements taken at each visit (1)	More than 1 measure of weight per woman at each visit	Single measure of weight per woman at each visit	Precision of estimates increases with the number of observation
Statistical methods (1)	Clearly described	Not clearly described/performed	Crucial to use an adequate statistical model for creating reliable charts
Assessment of goodness of fit (1)	Test of goodness of fit reported	Goodness of fit test not reported	Allows the reader to evaluate the adequacy of the model used to trace charts
Methods used to estimate the reference intervals/standards (1)	Mean and SD model and LMS method	Inadequate	Measure of reliability of results
Precision of estimates (1)	SEs or CIs reported	Not reported—only single mean or median value reported	Measure of reliability of results
Covariates (1)	Charts stratified by prepregnancy BMI	Not clearly described	Some covariates may exert effects on GWG
Smooth centiles (1)	Smooth centiles created	Raw centiles only presented	Smoothing reduces the fluctuations observed in raw centiles because of sampling variability
Reporting methods
Characteristics of study population (1)	Presented clearly or in a table	Not presented	Determines if the population studied belongs to the target population
Presentation of recommendations (1)	Presented as gestation-specific mean/SD in the form of centile charts or z score charts	Not clearly described or single recommendation for whole of pregnancy or trimester-specific recommendations	Aids interpretation of results
Report of regression equations for the mean and SD if relevant (1)	Reported	Not reported	Use of appropriate statistical model for creating reliable charts
Chart presentation (3)	Values presented as 10th, 50th, and 90th centiles or parameters that allow them to be computed are reported; z scores presented or computable; conditional and unconditional standards included	Not reported	Graphical method has a lower accuracy than use of numerical values; use of z scores increases the reliability of size evaluation and allows comparison among different subjects or population

¹Adapted with permission from reference 31. Low risk of bias (score = 1 for each point); high risk of bias (score = 0 for each point). GWG, gestational weight gain; LMS, lambda-mu-sigma.

Methodological quality criteria specific for studies that examined growth in the perinatal period have been developed and validated by our group (31–33). These criteria were modified and adopted in advance of this review. There were 23 quality criteria within 3 domains consisting of a total of 34 items: study design (12 criteria/21 items), statistical methods (7 criteria/7 items), and reporting methods (4 criteria/6 items). For each item, studies were scored as either a high (score = 0) or low (score = 1) risk of bias (Table 1). The data extracted from the studies are detailed in Table 1 along with the rationale for selecting these criteria. All scoring and study details were entered into Microsoft Excel (version 14.4.7).

The total quality score (QS) for each study was defined as the percentage of low risk of bias scores out of the total possible score. Mean scores for each domain were also calculated. If it was inappropriate make a judgment about an item because of the study design or if scoring an item would unfairly downgrade a study, then this item was excluded from the denominator when calculating the QS. The QS was also calculated with use of all items as a denominator, and these studies were not unfairly upgraded either. Finally, studies were ranked by QS, and the similarities and differences among the recommendations of the top-ranking studies were compared.

Results

Basic results and demographics

The search yielded 1599 citations, from which 116 full-text articles were obtained. Of these, 12 met the inclusion criteria. The reasons for exclusion are given in Figure 1. The detailed search strategy and results are shown in Table 2. The 12 studies published between March 1985 and February 2014 (21, 22, 24, 29, 34–41) provided data from 2,268,556 pregnancies. The median sample size was 1047 (minimum, 262; maximum, 2,241,487; IQR, 5730). The 12 studies originated from 9 countries (Argentina, Belgium, Germany, Hong Kong, Malawi, South Africa, Switzerland, the United Kingdom, and the United States).

FIGURE 1.

Flow chart of study selection process. GWG, gestational weight gain.

TABLE 2.

Searches and citations¹

Search terms	Citations obtained	Included after abstract and avoiding duplications
MEDLINE, EMBASE
Pregnancy kw + body weight af	588	15
Maternal weight kw	108	6
Maternal weight af + chart af	55	1
Pregnancy kw + gestational weight gain kw	108	13
Pregnancy kw + weight gain af +chart af	19	3
Pregnancy kw +weight gain af + curve af	20	3
Pregn* af + weight af + chart* af	150	0
Pregnancy kw + gestational weight gain af	309	8
Pregnancy af + gestational af + body weight change af	129	1
Pregnancy kw + body weight af + reference af	34	1
Pregn* af or gest* af + reference af + weight af + trend af	182	0
Pregnancy kw + range af + weight af + body af	206	1
Web of science
Weight gain + pregn* + chart	170	4
Gest* weight gain + women+ chart	88	2
Science Direct
Maternal weight kw	100	7
Pregnancy kw + gestational weight gain kw	400	7
Pregnancy kw + gestational weight gain af	200	1
Total	2765	73
Combined searches	1742
Duplicates removed	1556
Citations from other sources	43	43

¹af, all fields; kw, keywords.

Characteristics of studies

QSs ranged from 26% to 85%. Four studies had a QS >70% (21, 22, 34, 35). A summary of the assessed studies is shown in Table 3. The overall risk of bias scores across the studies for each of the criteria is shown in Figure 2.

TABLE 3.

Summary of assessed studies¹

Reference	Country	Sample size, n	Study design	First weight measurement, week of gestation	Weight measurements, n	Data	QS, %
34	Malawi	358	Longitudinal	14–25	1733 in total	Prospective	85
21	Argentina	1090	Longitudinal	12	8 per participant	Prospective	79
35	United States	648	Longitudinal	6	≥5 per participant	Retrospective	72
22	Belgium	605	Longitudinal	<12	5341 in total	Retrospective	71
36	Switzerland	4034	Cross-sectional	>5	Not applicable	Retrospective	59
37	Hong Kong	504	Longitudinal	ND	6 per participant	Retrospective	53
38	United States	10,418	Cross-sectional	14–26	Not applicable	Retrospective	53
39	South Africa	1003	Longitudinal	<24	Not documented	Prospective	50
40	United States	7002	Longitudinal	ND	2 per participant	Retrospective	50
24	United States	262	Longitudinal	12–14	Not documented	Prospective	48
29	United Kingdom	1145	Longitudinal	<20	6 per participant	Retrospective	46
41	Germany	2,241,487	Cross-sectional	ND	Not applicable	Retrospective	26

¹QS, quality score.

FIGURE 2.

Overall methodological quality of included studies assessed against quality criteria (corresponds to criteria in Table 1).

Main sources of bias

The most common potential source of bias was the lack of objective measurement of prepregnancy maternal weight used to calculate the baseline for GWG. In all studies, this value was obtained through maternal recall and never actually measured.

A further common potential source of bias was in the absence of anthropometric quality control, with a failure in regard to how measurements were validated. No studies performed more than 1 measurement at each visit to minimize intraobserver error. Only 3 studies named the instruments used to weigh women and described the weighing technique (e.g., with light clothing, shoes off). Choices of the weighing instrument and calibration frequency were not consistently described: 6 studies provided no information about which instrument and technique were used.

Study design domain

The median QS in this domain was 50% with a range of 19% to 76%. Of the 12 studies, 4 were prospective and planned. The remaining 8 presented data collected from other studies or were retrospective analyses of existing databases and birth records. Nine studies were longitudinal and 3 cross-sectional in design. Within the longitudinal studies, 6 documented the number of weight measurements taken throughout pregnancy, which ranged from 2 to 8. Five (56%) longitudinal studies specified an appropriate method of data analysis that accounted for repeated measurements within subjects.

There was a range in both the accuracy of gestational age assessment and the gestational age at the first weight that was recorded in each study. Gestational age was determined by a first-trimester ultrasound scan in 8 studies, by participant recall of the date of the last menstrual period in 2 studies, and undocumented in the remaining 2 studies. Four studies did not state the gestational age at which maternal weight was first measured. In the remaining studies, it ranged from 5 to 26 wk of gestation. Weight was measured monthly in 5 studies; however, in the remaining 7, the frequency of measurements was unclear or undocumented.

Inclusion and exclusion criteria were provided in 9 studies; however, there was little consensus on what these criteria should be (Figures 3 and 4). Women were most commonly excluded for multiple pregnancy or pre-existing hypertension or diabetes; however, none of the studies excluded all of these conditions. Seven studies were limited to women deemed to be “low risk,” defined either prospectively (i.e., women without any notable risk factors in their pregnancy or medical history) or retrospectively (i.e., women selected after delivery deemed to have experienced an uncomplicated pregnancy with a healthy-term baby). Baseline characteristics of women included in the study sample were presented in 8 of 12 studies.

FIGURE 3.

Pre-existing factors identified as exclusion criteria in 9 applicable studies.

FIGURE 4.

Pregnancy-related factors identified as exclusion criteria in 9 applicable studies.

Statistical methods domain

The median QS in this domain was 64% with a range of 0% to 85%. In 2 studies (29, 40), because GWG recommendations were based on crude statistical analytical summaries of observed associations between adverse neonatal or maternal outcomes and weight gain, scoring using the prespecified criteria was not possible. Of the remaining 10 studies, 8 clearly described the statistical methods used (longitudinal = 6 and cross-sectional = 2); 7 of these studies also reported the covariates in the analyses. Smooth centiles were created in 9 studies, with 5 reporting information on estimate precision (i.e., SEs or CIs).

Reporting methods domain

The median QS in this domain was 50% with a range of 30% to 83%. Regression equations were provided for 3 studies. One study reported both conditional and unconditional standards (34). The distribution of GWG at each gestation was presented as z scores in 2 studies (34, 35), with an additional 2 studies providing enough information to calculate z scores (21, 38).

Assessed studies recommendations

GWG recommendations from each study are provided in Table 4. Five studies (21, 24, 29, 37, 40) based GWG recommendations on adverse clinical outcomes; however, the outcomes selected varied. For example, Wong et al. (37) selected birth weight, term delivery, absence of gestational diabetes, and pregnancy-induced hypertension, whereas others selected only birth weight (21, 24).

TABLE 4.

Recommendations/results of assessed studies¹

		Recommendation, kg
Format of chart	Reference	Underweight	Normal	Overweight	Obese	All women	QS, %
Conditional and unconditional percentiles	34	—	—	—	—	GWG: 6 (unconditional chart)	85
Body weight and BMI centiles	212	12.2 ± 3.9	12.1 ± 4.3	12.1 ± 5.0	10.2 ± 4.8	Mean weight gain: 11.9 ± 4.4	79
Smoothed means, SD, centiles in all women	353	15.7 ± 4.6	15.5 ± 5.3	15.1 ± 6.2	9.9 ± 6.5	—	72
Centiles in all BMI categories	222	15.4 ± 4.1	15.1 ± 4.5	13.7 ± 5.3	12.0 ± 5.9	Mean weight gain: 14.8 ± 4.7	71
Centile charts for whites, Asians, and blacks	36	—	—	—	—	Mean weight gain: 15.5 ± 5.9	59
						5th Centile: 5.7
						95th Centile: 25.4
Centiles in all women and in underweight, normal, and overweight BMI groups	373	15.1 ± 3.8	13.8 ± 4.2	11.2 ± 5.2	—	—	53
Raw and fitted regression mean centiles in Asian, Hispanic, blacks, and whites	38						53
1st trimester/week						0.169 ± 0.268
2nd trimester/week		—	—	—	—	0.563 ± 0.236
3rd trimester/week		—	—	—	—	0.518 ± 0234
Centile chart for all women, for women >25 y + BMI >24 and women <25 y, or BMI <24	39	—	—	—	—	0.45	50
Centiles in all BMI groups	404	—	—	—	—		50
1st trimester total		1.92 ± 3.06	2.19 ± 3.47	2.16 ± 3.95	1.65 ± 3.94
2nd trimester/week		0.57 ± 0.20	0.58 ± 0.20	0.51 ± 0.24	0.41 ± 0.27
3rd trimester/week		0.48 ± 0.19	0.51 ± 0.21	0.49 ± 0.22	0.47 ± 0.24
Maternal body weight target as % of calculated standard weight	245	11.7 ± 8.3	10.4 ± 6.3	7.3 ± 6.6			48
Weight gain mean ± 1 SD in all women; weight gain in women with infants of birthweights >90th centile compared to 10–90th centile	29	—	—	—	—	Mean maternal weight gain: 10.71 ± 4.3; mean weekly gain: 0.38 ± 0.16	46
Centiles for individualized weight, height phenotype	41	—	—	—	—	Centile charts provided for 12 groups of women; largest group of women 161–171 cm in height and weighing ≤64 kg: 5th centile, 6.5; 50th centile, 14.0; and 95th centile, 21.0	26

¹All units of weight in kilograms or kilograms ± 1 SD unless otherwise indicated. GWG, gestational weight gain; QS, quality score.

²BMI categories per Institute of Medicine (9): underweight, <19.8; normal, 19.8–26; overweight, 26–29; and obese, >29.

³BMI categories per Asian BMI standard (37): underweight, <19; normal, 19–23.5; and overweight, >23.5.

⁴BMI categories per WHO (46): underweight, <18.5; normal, 18.5–24.9; overweight, 25–29.9; and obese, >30.

⁵Weight categories per reference 24: underweight, <89% of standard weight; normal, 90–110% standard weight; and overweight, >111% standard weight.

Discussion

We have demonstrated considerable heterogeneity in the methodology of studies used to create GWG charts. The median QS was 53%, with 5 studies scoring below 51% (24, 29, 39–41) and only 4 scoring above 70% (21, 22, 34, 35). Given the wide range of scores (26–85%), we believe it is likely that some of the observed variations in recommendations may have been caused by methodological differences rather than underlying discrepancies in GWG patterns between populations.

This review has numerous strengths. First, we conducted the systematic literature search using explicit and reproducible methodology and used well-established tools (including a Preferred Reporting Items for Systematic Reviews and metaanalysis checklist and flowchart) to guide the review process to ensure reliability. By use of a predefined set of methodological quality criteria, studies could be compared objectively. These criteria were formulated a priori and adapted from previously validated methods (31–33). Second, historical quasiscientific studies were excluded. In addition, we aimed to reduce bias when scoring studies. For example, unfair downgrading of studies was avoided by omitting inapplicable criteria from the score calculation. Finally, the scoring criteria were designed to identify important factors for judging the validity and reliability of research guiding GWG recommendations, and to minimize potential unconscious bias the scoring system was deliberately kept simple, with each item scored only as low or high risk of bias.

Limitations of the study include the decision to exclude non–English-language publications, which resulted in the exclusion of some important studies in this field (23, 42, 43). Because our review aimed to assess the methodological quality of GWG charts, we feel a sufficient number of studies were included to provide a range of authors, locations, times, and populations and highlight the main methodological strengths and weaknesses of research in this field. A further limitation of methodological quality-scoring systems is the assumption that each item has equal importance and with the same potential to produce biased results. This is clearly an oversimplification because, for example, poor study design could substantially reduce the overall quality of the study from the outset, despite all other elements being performed appropriately.

This review has demonstrated that GWG studies are prone to errors in measuring baseline (usually prepregnancy) weight and gestational age estimation, both of which are essential for accurately determining GWG. Maternal recall of prepregnancy weight reporting can potentially cause bias. Russell et al. (44) reported that women systematically underreport their prepregnancy weight; however, the issue is how big the magnitude of bias is. An alternative would be to use weight measured in the first trimester as the baseline, thereby providing an objective and replicable measurement (45). Accurate knowledge of gestational age is also essential for determining the trend in GWG, which may not be linear. Without an accurate gestational age, there may be a failure to recognize women delivering preterm, thus underestimating total GWG (35). We found that 4 studies used methods of gestational age assessment prone to error and bias (36, 37, 40, 41), with 2 providing no information about how gestational age assessment was performed. This finding is consistent with those of a systematic review published in 2012 on fetal growth assessment in pregnancy (33) and highlights an important potential source of bias in pregnancy research.

The 4 studies with QSs >70% were from diverse geographic locations: Argentina (21), Belgium (22), Malawi (34), and the United States (35), which arguably limits their potential for international use. International studies will be required for creating international GWG standards to determine whether geographic or ethnic differences exist in patterns of GWG (46).

z Scores were presented only in studies conducted after 2013 (34, 35), indicating that research in this field is continuing to develop. The use of z scores is well established in fetal, infant, and child studies for classifying growth during and after pregnancy (46, 47). They provide a method of describing weight gain independently of gestation (34, 35) and make it easier to describe the severity of abnormal weight gain patterns; they are also useful for epidemiological analyses because their statistical characteristics make them less prone to bias and difficulties in interpretation because of nonlinearity (21, 48). Furthermore, using a z score allows adherence to the reference distribution, provides a linear scale permitting summary statistics, has uniform criteria across indexes, and is useful for detecting changes at extremes of distributions.

The application of GWG charts in clinical practice implies that there are known thresholds above or below which the risk of adverse outcomes increases. Ideally, these thresholds should only be determined by assessing the association with maternal and neonatal outcomes that are specifically related to excessive or insufficient GWG. We found that several studies included outcomes unlikely to be causally linked to GWG (e.g., stillbirth), which makes it difficult to compare recommendations across studies. Interestingly, none of the studies considered postpartum weight retention as an outcome, although it is an important adverse effect of excessive GWG (2, 9, 17, 35).

Of the studies that produced charts of GWG as a continuous variable through pregnancy, most are references as opposed to standards. A reference is defined as a tool for grouping and analyzing locally acquired data; it is descriptive. A standard embraces the concept of a healthy or aspirational target and involves a value judgement; it is prescriptive. Although reference data have been widely used to make inferences about the health or nutrition of populations and individuals, the WHO has endorsed the use of standards for human anthropometry (46, 47). This would mean weight gain is compared to a physiological ideal, enabling a valid comparison between populations and a direct comparison of the efficacy of nutritional interventions in pregnancy.

In other areas of medicine, normative standards have been adopted to quantify changes over time in an anthropometric index (such as weight) in relation to a physiological, or optimal, standard (46). Where standards are not available or accepted, population-based references, derived from a defined population at a given time and place, have been used.

Future studies need to move toward creating standards instead of references to enhance quality and applicability and to recognize the use of reference values in making value judgements and recommending interventions.

We observed no common exclusion criteria across studies (Figures 3 and 4). Moreover, few studies excluded women with well-recognized factors to adversely affect GWG, such as preexisting diabetes, smoking, alcoholism, illicit drug use, and inflammatory bowel disease (1). Agreement on standard criteria to define healthy, low-risk pregnancies would enable valid comparisons between studies. The INTERGROWTH-21st project consortium has provided a detailed list of such criteria, which could be applied for future research in this field. Rather than exclude all possible risk factors, it was decided to limit them to factors that would prevent women from being classified as low risk for antenatal care (47).

From this review, we recommend that GWG should adhere to a minimum set of criteria in future studies (Table 5). It is recommended that anthropometric studies should apply standardized internationally agreed-upon protocols for measuring techniques to ensure that results are accurate, reliable, and comparable (46, 49, 50). We propose that the minimum requirements for creating future GWG charts should be that 1) data collection is population-based and prospective; 2) there is a detailed description of and attention to a reliable evaluation of gestational age by ultrasound examination before 14 wk of gestation; 3) the selection and measurement of the baseline weight is fully described; 4) there are comprehensive descriptions of measurement procedures, instruments, and protocols, including efforts to minimize intra- and interobserver bias; 5) methods to calculate adequate sample size are provided; and 6) appropriate longitudinal statistical modeling techniques are used with the creation of smoothed centiles and z scores. Adherence to these criteria would represent considerable progress in the rigor of reporting in this field and help to address the question of how much of the differences in GWG charts are caused by local factors.

TABLE 5.

Recommendations to create accurate GWG charts in the future¹

Domain	Requirements
Study design	Prospective study; planned study before enrolment; women enrolled consecutively and in accordance with target population; population-based study with efforts to exclude those with conditions known to affect GWG; longitudinal design; multiple measurements from each individual on multiple occasions throughout the pregnancy; reliable evaluation of prepregnancy weight; weight objectively measured with standardized scales by trained data collectors; weight taken with light indoor clothing— footwear, coats, sweaters, and heavy clothing removed; reliable evaluation of gestational age; early pregnancy (9–13-wk) scan to determine gestational age; detailed description of measurement procedures and protocols; standardized training of data collectors; data collected specifically for creation of GWG charts; standardized instruments with specific calibration schedule; gestation at which measurements taken accurately recorded
Statistical methods	Creation of smooth centiles; use of appropriate statistical methods to convert raw centiles to smooth centiles; adequate sample size for each range of measurements to allow presentation by z scores and centiles; regression-based methods used
Reporting methods	Centile charts or z scores; charts presented with at least 10th, 50th, and 90th centiles or parameters that allow them to be computed; z scores presented or computable

¹GWG, gestational weight gain.

In conclusion, there is a lack of international agreement on what constitutes adequate GWG. We have demonstrated substantial heterogeneity in the methodological quality of studies used to make recommendations for GWG. It is possible that this heterogeneity has contributed to some of the variation in current GWG recommendations around the world. High-quality international studies at a low risk of methodological bias are needed to guide future clinical recommendations and counseling of women.