Change of maternal weight gain trajectory following lifestyle interventions for gestational diabetes and its impact on abnormal infant birthweight: an observational study based on longitudinal weight measurements

Abstract

Background

Restricted gestational weight gain (GWG) following lifestyle interventions for gestational diabetes mellitus (GDM) has been observed in clinical trials but not been explicitly confirmed in real-world clinical settings. Furthermore, how such a restricted GWG affects adverse birthweight outcomes, i.e. infants born large/small for gestational age (LGA/SGA), remains unclear.

Method

In this retrospective study, based on longitudinal weight measurements from 33,515 Chinese women, including 5,932 with GDM diagnosed by 75-g oral glucose tolerance test (OGTT), we compared the adjusted GWG trajectories of women with and without GDM. We estimated weekly GWG before and after OGTT (WGWG_pre−OGTT and WGWG_post−OGTT) and calculated ΔWGWG as WGWG_pre−OGTT minus WGWG_post−OGTT. We then estimated the adjusted relative risks (aRR) of LGA and SGA for restricted vs. unrestricted WGWG_post−OGTT (ΔWGWG > 0 vs. ≤0 kg) among women with GDM. The analyses were stratified by prepregnancy body mass index (pBMI).

Results

There was a marked decrease between the mean estimated WGWG_pre−OGTT (0.50 kg) and WGWG_post−OGTT (0.36 kg) in women with GDM, in contrast to a negligible decrease from 0.52 to 0.51 kg in women without GDM. Using 14–19 gestational weeks as the baseline, the cumulative GWG since the baseline became statistically significantly less in women with than without GDM after the initiation of lifestyle interventions, regardless of their pBMI category. Women with an underweight, normal-weight, and overweight/obese pBMI gained by an adjusted median of 1.84 [95% confidence interval (CI): 1.50, 2.17], 2.53 (2.38, 2.68), and 2.99 kg (2.63, 3.35) less, respectively, than their non-GDM counterparts by the end of gestation (38–39 weeks). Among women with GDM, a restricted WGWG_post−OGTT showed no association with either LGA or SGA for underweight pBMI, a null association with SGA but a statistically significant association with a reduced risk of LGA for normal-weight pBMI [aRR (95% CI): 1.02 (0.82, 1.28) and 0.62 (0.49, 0.79), respectively], and opposite although non-significant associations for overweight/obese pBMI [ aRR (95% CI): 1.22 (0.66, 2.28) for SGA and 0.74 (0.52, 1.06) for LGA].

Conclusions

Lifestyle interventions for GDM are likely to cause a substantially restricted GWG, which however might only have a fully beneficial effect on infant birthweight for women with a normal-weight pBMI.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-026-08757-3.

Background

Gestational diabetes mellitus (GDM) is a prevalent pregnancy-related condition characterized by maternal hyperglycemia, which leads to an excessive transplacental glucose supply to the fetus and consequently causes fetal overgrowth and infants born large for gestational age (LGA), among other consequences [1]. The treatment of GDM involves lifestyle interventions (i.e., physical activity plus diet therapy) and if needed, insulin therapy, with a goal to obtain and maintain maternal euglycemia until childbirth [2]. Despite improved pregnancy outcomes including a reduced rate of LGA births [3–7], there are concerns that tight glycemic control may lead to unduly restricted fetal growth and infants born small for gestational age (SGA) [8, 9]. Therefore, attaining a normal birthweight for infants born to mothers with GDM is a tricky challenge for antenatal care professionals.

In observational studies [10, 11], excessive gestational weight gain (GWG) has been associated with increased risks of developing GDM and delivering an LGA infant. According to pooled data from clinical trials, lifestyle interventions initiated in early pregnancy limit GWG and, partly because of this, reduce the incidence of GDM [12–17]. Also on the basis of multiple meta-analyses of clinical trials, lifestyle interventions have been established as an effective treatment for GDM to prevent adverse pregnancy outcomes including LGA births [5–7]. Among the clinical trials included in these meta-analyses, two looked into GWG as an outcome [3, 4], both reporting a significantly less GWG in the lifestyle intervention groups than in the usual care groups. However, this restricting effect needs to be confirmed by observational studies conducted in real-world clinical settings. According to few small observational studies, lifestyle interventions following GDM diagnosis seem to decelerate the subsequent GWG [18, 19]. Several observational studies also suggest that a low GWG during GDM treatment, which does not necessarily represent a restricted GWG relative to the GWG before GDM diagnosis, might reduce women’s risk of delivering an LGA infant, but they did not produce clear findings on whether this benefit comes at the expense of increasing the risk of having an SGA infant [20–23]. In addition, these studies did not either consider GWG before GDM diagnosis or include women without GDM as the reference, and therefore it was impossible for them to accurately assess the restricting effect of the lifestyle interventions on the GWG after GDM diagnosis and how the potentially restricted GWG would affect the risks of adverse infant birthweight outcomes.

In the present study, therefore, we analyzed longitudinally measured weight values before and after GDM screening in a large cohort of women who had received standard antenatal care or, in case of GDM, guideline-based lifestyle interventions, aiming to: (1) quantify the restricting effect of lifestyle interventions on the GWG in women with GDM; (2) investigate how this effect might affect women’s risk of delivering an SGA or LGA infant.

Methods

Study population

This retrospective study was conducted at Guangzhou Women and Children’s Medical Center (GWCMC), the largest maternal and child healthcare facility in South China. The institutional ethics committee approved the study protocol. Written informed consent from study subjects was waived because of the study’s retrospective design. The study was predetermined to include women who: (1) were 18–50 years old; (2) delivered a term, liveborn singleton; (3) underwent a routine 2-h 75-g oral glucose tolerance test (OGTT) at 24–28 weeks of gestation to screen for GDM; (4) did not have pre-existing hypertension or diabetes, either self-reported or diagnosed during pregnancy; (5) did not develop any type of gestational hypertensive disorders during pregnancy; (6) attended all the scheduled antenatal visits in the second and third trimesters (one visit during each of the 14–19, 20–24, 25–28, 29–32, and 33–36 gestational week intervals and thereafter weekly visits until childbirth), and had weight measured during each visit; and (7) had self-reported prepregnancy weight values.

As illustrated in Fig. 1, from 91,933 pregnancy women included in the GWCMC electronic obstetric database (2018–2023), we identified 37,143 pregnancies from 36,109 women who met the aforementioned inclusion criteria. Of the 54,790 pregnancies not meeting the inclusion criteria, 47,897 (87.4%) did not have a full set of gestational weight measurements as required by inclusion criterion 6, while 13,959 (25.5%) did not have an OGTT during 24–28 gestational weeks as required by inclusion criterion 3. Routinely measured gestational weight during antenatal care is subject to errors in measurement or data entry. For self-reported prepregnancy weight, recall errors are always inevitable. These errors, if large, might result in outlier values. In this study, we took two steps to identify in turn the outlier values on measured gestational weight and the outlier values on self-reported prepregnancy weight (see the Statistical analysis section for methodological details). After treating the identified outliers as missing values, we excluded 1,557 pregnancies because they no longer had a full set of weight values for analysis. We further excluded 7 pregnancies with maternal body height < 110 cm, 45 pregnancies that had only one single weight measurement before OGTT, and 1,120 pregnancies where the first and the last weight measurements before GDM screening were < 4 weeks apart, because calculation of the rate of GWG before OGTT among these women was either impossible or unreliable. In the remaining 34,414 pregnancies of 33,515 women, 898 women had two or more separate pregnancies. For each of these women only one pregnancy was randomly selected. Finally, a total of 33,515 independent pregnancies remained for analyses.

Fig. 1.

Flowchart of the study population

Information on covariates was extracted from the same obstetric database for multivariable analyses. The outcomes of this study were LGA and SGA births, defined as infant birthweight above the 90th and below the 10th gestational-age-specific population percentiles [24], respectively.

GDM diagnosis and lifestyle interventions

Diagnosis of GDM was made on a routine OGTT screening at 24–28 weeks of gestation, following the criteria from the International Association of Diabetes and Pregnancy Study Group [25]. Upon diagnosis, women were referred to our diabetes clinic, where they were provided with guideline-based counseling on how to control hyperglycemia by means of diet modification and physical activities. In short, women were recommended to increase their consumption of food items having a glycemic index (GI) < 55 (low-GI food items, such as whole grain porridge, buckwheat noodles, and green leafy vegetables), yet with a total daily energy intake of no less than 1,600 kcal. Provided that there was no contraindication to physical activities, women were recommended to have 30 min of moderate, aerobic exercise, such as brisk walking and jogging, at least five days a week. Insulin therapy was administered only if lifestyle interventions, after a test period of 1–2 weeks, failed to achieve the glycemic targets [< 5.3 mmol/L (95 mg/dL) preprandial and < 6.7 mmol/L (120 mg/dL) 2-h postprandial] without causing ketoacidosis. As we have reported elsewhere [26], less than 2% of women with GDM were treated with insulin therapy; therefore, these women were simply included in our study population and treated the same as those who only received lifestyle interventions.

Statistical analysis

We took two steps to identify in turn the outliers on measured gestational weight and the outliers on self-reported prepregnancy weight. We first identified outliers on measured gestational weight according to the size of the standardized residuals, i.e., the standardized differences between the recorded and predicted gestational weight values. The predicted gestational weight values were obtained from a mixed-effects model that fits gestational weight as a function of gestational age, with randomly varying intercept and slope across individuals, as well as a restricted cubic spline term to fit non-linear trajectories. Such a model was built for women with and without GDM separately because of the two groups’ potentially different GWG patterns. Among outliers, those extreme ones can pull the entire projected gestational weight trajectories toward themselves and thus cause other plausible values to be wrongly classified as outliers. Therefore, we first removed these extreme outliers, defined as if the standardized residual was larger than 5 in absolute value, and then refitted the models. Based on the refitted models, we further removed the outliers defined as if the standardized residual was larger than 3 in absolute value. In the second step, we built a linear model to predict self-reported prepregnancy weight as a function of maternal age, height, the earliest measured gestational weight during 14–19 gestational weeks, and the gestational age at that weight measurement. Then, similarly to the first step, a refitted model after removal of the extreme outliers, defined as self-reported prepregnancy weight values with a standardized residual > 5 in absolute value, was used to identify self-reported prepregnancy weight values with a standardized residual > 3 in absolute value. The identification of outliner values based on this two-step process is illustrated in Supplementary Material Figure S1.

We described maternal and infant characteristics using mean, median, or frequency, whichever was appropriate. We created plots to compare the medians of weight measurements between women with and without GDM across two-week intervals throughout the second and third trimesters, ranging from 14–15 to 38–39 gestational weeks. This comparison was further refined by adjusting for maternal age, height, prepregnancy weight, and parity, using median regression tailored to longitudinal data [27]. We stratified these comparisons by women’s prepregnancy body mass index (pBMI), which was categorized using the thresholds for Chinese adults [< 18.5 kg/m² (underweight), 18.5–23.9 kg/m² (normal-weight), and ≥ 24 kg/m² (overweight/obese)].

We built a linear mixed-effects model with randomly varying intercept and slope to estimate women’s weekly GWG before the OGTT screening for GDM (termed WGWG_pre−OGTT), which were equivalent to the model’s slope estimates. In the same way but with a separate model we obtained the weekly GWG after the OGTT screening (termed WGWG_post−OGTT). These were done separately for women with and without GDM. We calculated ΔWGWG as the estimated WGWG_pre−OGTT minus WGWG_post−OGTT and defined unrestricted and restricted WGWG_post−OGTT in women with GDM as ΔWGWG ≤ 0 and > 0, respectively. The continuous and dichotomized ΔWGWG were both examined in multivariable Poisson regression models for their risk associations with SGA and LGA in women with GDM. Maternal age, pBMI, height, parity, gestational age at delivery, family history of diabetes, and the estimated WGWG_pre−OGTT were also included in the models as covariates. These analyses were conducted separately for each pBMI category.

All the statistical tests were two-sided, with P < 0.05 considered statistically significant. All the statistical analyses were conducted using the R software (version 4.0.2, R foundation, Vienna, Austria).

Results

Maternal and infant characteristics, overall and by GDM status, are summarized in Table 1. The prevalence of prepregnancy overweight or obesity was relatively low in the entire cohort (10.4%), but was notably higher in women with than without GDM (16.6% vs. 9.1%). The mean estimated WGWG_pre−OGTT and WGWG_post−OGTT for the entire cohort were 0.52 and 0.48 kg, respectively. There was a marked decrease between the mean estimated WGWG_pre−OGTT (0.50 kg) and WGWG_post−OGTT (0.36 kg) in women with GDM, in contrast to a negligible decrease from 0.52 to 0.51 kg in women without GDM. The SGA rate was similar between the GDM and non-GDM groups GDM (9.7% vs. 10.2%. P = 0.22), while the LGA rate was statistically significantly higher in women with than without GDM (8.6% vs. 7.1%. P < 0.01).

Table 1.

Maternal and infant characteristics, overall and stratified by GDM status

	Overall (n = 33,515)	GDM status
		Non-GDM (n = 27,583)	GDM (n = 5,932)	P *
Maternal age (y), mean (SD)	30.8 (3.9)	30.5 (3.8)	32.1 (4.0)	< 0.01
Multipara, n (%)	13,708 (40.9)	10,963 (39.7)	2,745 (46.3)	< 0.01
pBMI (kg/m2), mean (SD)	20.6 (2.6)	20.4 (2.5)	21.4 (2.8)	< 0.01
pBMI category, n (%)
<18.5 kg/m2 (underweight)	6,731 (20.1)	5,911 (21.4)	820 (13.8)
18.5–23.9 kg/m2 (normal-weight)	23,301 (69.5)	19,174 (69.5)	4,127 (69.6)
24–27.9 kg/m2 (overweight)	3,047 (9.1)	2,229 (8.1)	818 (13.8)
≥28 kg/m2 (obese)	436 (1.3)	269 (1.0)	167 (2.8)	< 0.01
WGWGpre−OGTT (kg/w), mean (SD)	0.52 (0.13)	0.52 (0.13)	0.50 (0.14)	< 0.01
WGWGpost−OGTT (kg/w), mean (SD)	0.48 (0.18)	0.51 (0.16)	0.36 (0.18)	< 0.01
Family history of diabetes, n (%)	1,176 (3.5)	672 (2.4)	504 (8.5)	< 0.01
Male fetal sex, n (%)	17,793 (53.1)	14,672 (53.2)	3,121 (52.6)	0.42
Gestational age (w) at birth, median (IQR)	39.3 (38.7, 40.0)	39.3 (38.7, 40.0)	39.1 (38.6, 39.9)	< 0.01
Neonate birthweight (g), mean (SD)†	3,203 (368)	3,205 (368)	3,193 (373)	0.03
SGA, n (%)†	3,391 (10.1)	2,817 (10.2)	574 (9.7)	0.22
LGA, n (%)†	2,457 (7.3)	1,949 (7.1)	508 (8.6)	< 0.01

*P values were based on t-test, Wilcoxon rank sum test, or χ² test. †Six women, all in the non-GDM group, were excluded due to missing information on neonate birthweight

GDM gestational diabetes mellitus, IQR interquartile range, LGA large for gestational age, pBMI prepregnancy body mass index, SD standard deviation, SGA small for gestational age, WGWG weekly gestational weight gain

As shown in Fig. 2, women with GDM in each pBMI category had approximately the same weight medians as did women without GDM from 14–15 until 24–25 gestational weeks, when GDM screening and lifestyle interventions, in case of GDM, were first initiated. Thereafter, the weight medians in women with GDM were notably lower than in those without GDM. The results from the multivariable median regression models (Table 2) confirmed this change after controlling for potential confounding factors: the adjusted medians of the cumulative GWG since 14–15 gestational weeks (the baseline) between women with and without GDM showed no statistically significant difference until 26–27 gestational weeks in the underweight pBMI subgroup and until 24–25 gestational weeks in the normal-weight and overweight/obese pBMI subgroups. Since then, the cumulative GWG in women with GDM became statistically significantly less than in women without GDM. In the underweight pBMI subgroup at 26–27 gestational weeks and normal-weight and overweight/obese pBMI subgroups at 24–25 gestational weeks, for example, the cumulative GWG since 14–15 gestational weeks in women with GDM was 0.37 [95% confidence interval (CI): 0.14, 0.61], 0.15 (95% CI: 0.02, 0.28), and 0.35 kg (95% CI: 0.04, 0.66) less, respectively, than in women without GDM. By the end of gestation (38–39 gestational weeks), women with GDM in the underweight, normal-weight, and overweight/obese pBMI subgroups gained 1.84 (95% CI: 1.50, 2.17), 2.53 (95% CI: 2.38, 2.68), and 2.99 kg (95% CI: 2.63, 3.35) less, respectively, than their non-GDM counterparts. When the international cut-offs were used to define normal pBMI (18.5–24.9 kg/m²) and overweight/obese pBMI (≥ 25 kg/m²), the corresponding figures were 2.62 (95% CI: 2.46, 2.78) and 3.04 kg (95% CI: 2.51, 3.57), respectively.

Fig. 2.

The medians and IQRs of gestational weight measurements throughout the second and third trimesters in women with GDM (in black) and those without (in grey), divided into subgroups with an underweight pBMI (< 18.5 kg/m², Plot A), a normal-weight pBMI (18.5–23.9 kg/m², Plot B), and an overweight/obese pBMI (≥ 24.0 kg/m², Plot C). GDM gestational diabetes mellitus, IQR interquartile range, pBMI prepregnancy body mass index

Table 2.

The adjusted medians of cumulative GWG (kg) since 14–15 weeks of gestation (baseline) for women without GDM, and the adjusted median differences between women with (n = 5,932) and without GDM (n = 27,583), stratified by pBMI category

Gestational week interval	Underweight pBMI (< 18.5 kg/m2)		Normal-weight pBMI (18.5–23.9 kg/m2)		Overweight/obese pBMI (≥ 24.0 kg/m2)
	Adjusted median (95% CI) of the cumulative GWG since baseline in women without GDM*	Adjusted median difference (95% CI) between women with and without GDM*	Adjusted median (95% CI) of the cumulative GWG since baseline in women without GDM*	Adjusted median difference (95% CI) between women with and without GDM*	Adjusted median (95% CI) of the cumulative GWG since baseline in women without GDM*	Adjusted median difference (95% CI) between women with and without GDM*
14–15	Baseline	Baseline	Baseline	Baseline	Baseline	Baseline
16–17	0.81 (0.74, 0.89)	0.18 (-0.04, 0.41)	0.74 (0.70, 0.78)	0.10 (-0.02 0.22)	0.71 (0.58, 0.85)	-0.07 (-0.40, 0.26)
18–19	1.78 (1.68, 1.87)	0.14 (-0.06, 0.34)	1.77 (1.72, 1.82)	0.09 (-0.02, 0.22)	1.57 (1.44, 1.70)	-0.05 (-0.30, 0.20)
20–21	2.92 (2.84, 3.00)	0.18 (-0.03, 0.39)	2.93 (2.88, 2.98)	0.00 (-0.12, 0.11)	2.60 (2.45, 2.75)	-0.11 (-0.43, 0.21)
22–23	4.17 (4.09, 4.25)	0.16 (-0.11, 0.44)	4.16 (4.12, 4.20)	-0.03 (-0.15, 0.08)	3.63 (3.51, 3.75)	-0.18 (-0.45, 0.09)
24–25	5.11 (5.03, 5.18)	-0.09 (-0.29, 0.10)	5.08 (5.04, 5.12)	-0.15 (-0.28, -0.02)†	4.47 (4.33, 4.61)	-0.35 (-0.66, -0.04)†
26–27	6.18 (6.09, 6.27)	-0.37 (-0.61, -0.14)†	6.30 (6.25, 6.35)	-0.78 (-0.92, -0.64)†	5.62 (5.46, 5.77)	-1.07 (-1.36, -0.78)†
28–29	7.23 (7.15, 7.32)	-0.94 (-1.17, -0.70)†	7.31 (7.26, 7.36)	-1.26 (-1.39, -1.12)†	6.51 (6.34, 6.68)	-1.49 (-1.80, -1.18)†
30–31	8.19 (8.11, 8.27)	-1.18 (-1.42, -0.93)†	8.30 (8.24, 8.36)	-1.67 (-1.81, -1.53)†	7.46 (7.32, 7.60)	-2.06 (-2.38, -1.75)†
32–33	9.19 (9.11, 9.28)	-1.37 (-1.64, -1.10)†	9.29 (9.23, 9.34)	-2.00 (-2.14, -1.86)†	8.35 (8.19, 8.51)	-2.32 (-2.61, -2.02)†
34–35	10.30 (10.21, 10.38)	-1.57 (-1.86, -1.28)†	10.40 (10.34, 10.46)	-2.26 (-2.39, -2.13)†	9.37 (9.21, 9.53)	-2.65 (-3.00, -2.32)†
36–37	11.34 (11.25, 11.42)	-1.73 (-2.03, -1.42)†	11.41 (11.35, 11.47)	-2.44 (-2.58, -2.30)†	10.35 (10.17, 10.53)	-2.85 (-3.18, -2.53)†
38–39	12.16 (12.06, 12.26)	-1.84 (-2.17, -1.50)†	12.21 (12.15, 12.28)	-2.53 (-2.68, -2.38)†	11.13 (10.93, 11.32)	-2.99 (-3.35, -2.63)†

*The values were adjusted for maternal age, height, prepregnancy weight, and parity. †P < 0.01 CI confidence interval, GDM gestational diabetes mellitus, GWG gestational weight gain, pBMI prepregnancy body mass index

A random sample of the estimated WGWG_pre−OGTT and WGWG_post−OGTT based on linear mixed-effects models is demonstrated in Supplementary Material Figure S2. The risk associations of ΔWGWG, both continuous and dichotomized, with SGA and LGA among GDM-complicated pregnancies are reported in Table 3. In women with an underweight pBMI, neither continuous nor dichotomized ΔWGWG showed statistically significant risk associations with SGA or LGA, while in women with a normal-weight pBMI, they both showed a statistically significant risk association with a reduced risk of LGA: the adjusted relative risks (aRRs) of LGA was 0.88 (95% CI: 0.84, 0.93) for 0.1-kg increment in ΔWGWG and 0.62 (95% CI: 0.49, 0.79) for restricted vs. unrestricted WGWG_post−OGTT (i.e. ΔWGWG ≤ 0 vs. >0 kg). In women with an overweight/obese pBMI, continuous ΔWGWG showed no association with SGA [aRR (95% CI): 0.99 (0.86, 1.13)] but a statistically significant inverse association with LGA [aRR (95% CI): 0.92 (0.86, 0.99)], while restricted vs. unrestricted WGWG_post−OGTT demonstrated contrary albeit statistically non-significant risk associations, with an aRR of 1.22 (95% CI: 0.66, 2.28) for SGA and 0.74 (95% CI: 0.52, 1.06) for LGA. Re-classification of pBMI based on the international cut-offs only led to an appreciably changed aRR of SGA among women with an overweight/obese pBMI (≥ 25 kg/m²), which was 0.94 (95% CI: 0.80, 1.12) for continuous ΔWGWG and 0.98 (95% CI: 0.49, 1.95) for restricted vs. unrestricted WGWG_post−OGTT.

Table 3.

Associations of continuous ΔWGWG and dichotomized ΔWGWG with the risks of SGA and LGA in women with GDM, stratified pBMI category

	N	SGA			LGA
		n	aRR (95% CI)*	P	n	aRR (95% CI)*	P
Underweight pBMI (<18.5 kg/m 2 )
Continuous ΔWGWG, per 0.1-kg increment			0.95 (0.85, 1.06)	0.36		0.94 (0.68, 1.28)	0.68
Dichotomized ΔWGWG
≤0 kg (unrestricted WGWGpost-OGTT)	230	51	1.00		4	1.00
>0 kg (restricted WGWGpost-OGTT)	619	80	0.73 (0.52, 1.03)	0.07	7	0.56 (0.15, 2.08)	0.39
Normal-weight pBMI (18.5-23.9 kg/m 2 )
Continuous ΔWGWG, per 0.1-kg increment			0.94 (0.89, 0.99)	0.03		0.88 (0.84, 0.93)	<0.01
Dichotomized ΔWGWG
≤0 kg (unrestricted WGWGpost-OGTT)	960	109	1.00		92	1.00
>0 kg (restricted WGWGpost-OGTT)	3,143	278	1.02 (0.82, 1.28)	0.84	257	0.62 (0.49, 0.79)	<0.01
Overweight/obese pBMI (≥24.0 kg/m 2 )
Continuous ΔWGWG, per 0.1-kg increment			0.99 (0.86, 1.13)	0.85		0.92 (0.86, 0.99)	0.02
Dichotomized ΔWGWG
≤0 kg (unrestricted WGWGpost-OGTT)	236	13	1.00		39	1.00
>0 kg (restricted WGWGpost-OGTT)	744	43	1.22 (0.66, 2.28)	0.52	109	0.74 (0.52, 1.06)	0.10

*All the models were adjusted for maternal age, pBMI, height, parity, fetal sex, gestational age at delivery, family history of diabetes, and WGWG_pre-OGTT

aRR adjusted relative risk, CI confidence interval, LGA small for gestational age, pBMI prepregnancy body mass index, SGA small for gestational age, WGWG weekly gestational weight gain

Discussion

Based on longitudinal weight measurements throughout the second and third trimesters, this study showed that women with GDM underwent a significantly restricted GWG compared with those without GDM, regardless of their pBMI category. Nonetheless, depending on women’s pBMI, a restricted WGWG_post−OGTT might influence infant birthweight differently, with no effect on either the SGA or LGA risk in women with an underweight pBMI, a fully beneficial effect in women with a normal-weight pBMI that would reduce the LGA risk without increasing the SGA risk, and a seemingly mixed effect in women with an overweight/obese pBMI that would reduce the LGA risk but increase the SGA risk.

This study suggests a substantially reduced GWG in women with GDM after the initiation of lifestyle interventions by comparing their weight gain trajectories with those of their non-GDM counterparts. Two noteworthy advantages of this study are the adjustment for several factors that also affect GWG (such as prepregnancy weight) and stratification by women’s pBMI category. A pooling analysis of randomized clinical trials indicates that exercise or diet, or both, can significantly reduce the risk of excessive GWG [15]. However, that analysis did not estimate the magnitude of the reduction. In addition, the lifestyle interventions in those clinical trials were usually initiated much before 24 − 28 weeks of gestation when the routine OGTT screening started. According two clinical trials among women with GDM [3, 4], the GWG in the lifestyle intervention groups was 2.2 and 1.4 kg, respectively, less than the GWG in the usual care groups. However, because these clinical trials were conducted in settings that were rather different from the present clinical practices (such as with regard to the diagnostic criteria for GDM and the intensity of the interventions and patient management), evidence from real-world, observational studies is also of crucial importance and perhaps higher relevance. In a small Australian cohort that overall was overweight, women with GDM gained approximately 2.5 kg less weight than those without GDM during the period from GDM test up to childbirth [18], comparable to our estimates. However, that study was different from ours in several aspects, including a much smaller sample size, no adjustment for covariates, no stratification by pBMI, and different diagnostic criteria for GDM.

Despite the evidence that restricting GWG during GDM treatment may reduce the risk of LGA, it remains inconclusive whether this benefit is obtained at the cost of having more infants born SGA [20–23]. There have been two studies reporting an increased SGA rate in women with a low GWG during GDM treatment [20, 21]. However, as neither considered the GWG before GDM diagnosis, it was difficult to conclude that the low GWG after GDM diagnosis was responsible for the increased SGA rate, because women with a low GWG after GDM diagnosis are also likely to have had a low GWG during the early pregnancy, which is a more important risk factor for infant low birthweight [28–30]. After controlling for the estimated WGWG_pre−OGTT, the present study suggests that a restricted WGWG_post−OGTT in women with GDM might “normalize” infant birthweight by decreasing the LGA risk without increasing the SGA risk, but this fully beneficial effect was only confined to women with a normal-weight pBMI. A restricted GWG after GDM diagnosis is considered largely an indicator of good adherence to the lifestyle interventions [21]. In addition, evidence from clinical trials suggests that medical nutrition therapy can limit maternal GWG and fetal fatness without compromising fetal growth [31, 32], providing an explanation for our finding. In women with underweight pBMI, the restricted WGWG_post−OGTT showed no statistically significant association either with an increased SGA risk or with a decreased LGA risk, but the present study itself apparently had a very limited statistical power to detect a significant association for LGA due to the fewness of the LGA births in this subgroup. Similarly, the fewness of SGA births in the overweight/obese subgroup compromised the reliability of the seemingly increased SGA risk associated with the restricted WGWG_post−OGTT.

Although a restricted GWG during GDM treatment seems advisable at least for women with a normal-weight pBMI, it might be wiser in clinical practice to first assess the mothers’ GWG in early pregnancy upon GDM diagnosis and then decide on an appropriate GWG target that should be achieved during the subsequent GDM treatment. There is no doubt that women with an excessive GWG before GDM diagnosis should be recommended to restrict their GWG after GDM diagnosis given their high risk of delivering an LGA infant, but women with a low or moderate GWG before GDM diagnosis might benefit less from a restricted GWG during GDM treatment, considering their relatively low risk of having an LGA infant. Therefore, to further normalize infant birthweight, there is a necessity to investigate whether the GWG before GDM diagnosis should also be considered in addition to pBMI.

As one of the few studies thus far to seek evidence for proper maternal weight management after GDM diagnosis, the present study has some strengths as compared with the previous ones. The first one is its considerably larger study population. Second, the present study carefully considered GWG before GDM diagnosis in its analyses. Third, the present study was based on intensively measured serial weight values, which allowed us to eliminate bias caused by erroneous weight values and thus conferred an advantage over the other studies that relied on only two weight measurements to calculate total GWG or WGWG [20–23]. Of note, among a total of 412,400 longitudinal weight measurements, we only identified 2,857 as outliers, suggesting that maternal weight in this study was monitored with good consistency throughout pregnancy.

Several limitations of the present study should be acknowledged. First, exclusion of women who had fewer antenatal visits than desired might limit the representativeness of our study population and thus the generalizability of our findings. As we mentioned before, 47,897 (87.4%) of the 54,790 pregnancies excluded for not meeting all the inclusion criteria did not meet inclusion criterion 6, which required a full set of gestational weight measurements. Of these 47,897 pregnancies, 28,496 pregnancies (59.5%) met all the other inclusion criteria. Based on the 37,143 pregnancies with a full set of weight measurements combined with the 28,496 pregnancies with an incomplete set of weight measurements, we conducted a logistic regression to explore whether the two groups were different in clinical characteristics. The results show that older women, nulliparous women, and women with GDM were statistically more likely to have a complete set of weight measurements than younger women, parous women, and women without GDM, while the two groups were similar in pBMI and the rates of SGA and LGA births. Therefore, exclusion of women with an incomplete set of weight measurements surely introduced a selection bias. Although the impact of this selection bias might be mitigated by our multivariable statistical analyses, our findings may not be generalizable to women who attended antenatal care infrequently and, given our study’s single-center design, to women in other populations. Second, the present study was a real-world observational study. Unlike clinical trials, where the interventions were thoroughly defined and closely monitored, the present study was conducted in a clinical setting and the lifestyle interventions were given in the form of general clinical guidelines. More often than not, physicians may also follow their personal experience or preference when executing these guidelines. In addition, the degree of glycemic control may affect both GWG during GDM treatment and infant birthweight, but we were unable to adjust for it due to lack of data. These flaws together undermined our ability to prove a causal effect of lifestyle interventions on GWG. Third, the use of Chinese national rather than the international BMI cutoffs to categorize prepregnancy weight may be appropriate for the local population but compromised cross-study comparability. Last, it is known that maternal GWG in the early pregnancy is a more impotant factor associated with infant birthweight [28–30], but we were unable to further stratify the analyses by the estimated WGWG_pre−OGTT as it would further reduce the numbers of LGA and SGA births in each stratum.

Conclusions

The present study suggests a substantially restricted GWG in women with GDM following guideline-based lifestyle interventions, regardless of their pBMI category. However, depending on pBMI, the restricted GWG during GDM treatment is likely to affect infant birthweight differently, with a fully beneficial effect only for women with a normal-weight pBMI.

Abbreviations

aRR

Adjusted relative risk

CI

Confidence interval

GDM

Gestational diabetes mellitus

GWG

Gestational weight gain

LGA

Large for gestational age

OGTT

Oral glucose tolerance test

pBMI

Prepregnancy body mass index

SGA

Small for gestational age

WGWG

Weekly gestational weight gain

Authors’ contributions

CY, JL, and KL conceptualized the study. CY and JL conducted the statistical analysis and drafted the manuscript. LK reviewed and revised the manuscript. All authors read and approved the final version of the manuscript.

Funding

Not applicable.

Data availability

The datasets analyzed during the current study are not publicly available due to data protection regulations but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted according to the guidelines laid down in the Declaration of Helsinki and was approved by the ethics committee of Guangzhou Women and Children’s Medical Center. The need for informed consents was waived because of the retrospective nature of the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.