Predictive ability of lipid indices for large-for-gestational-age infants in pregnant females with gestational diabetes mellitus

Abstract

BACKGROUND

The primary complication associated with gestational diabetes mellitus (GDM) is delivery of an infant that is large for gestational age (LGA). Epidemiological findings have demonstrated that irregular lipid metabolism significantly contributes to insulin resistance, a key pathophysiological mechanism in GDM. However, the correlation between various lipid indices and the probability of delivering LGA infants remains inconsistent.

AIM

To explore the relationships between lipid indices and the possibility of having LGA infants among GDM-affected pregnant females.

METHODS

Binary logistic regression methods were employed to evaluate the odds ratios and corresponding 95% confidence intervals for LGA according to five lipid indices. Restricted cubic spline models were applied to investigate dose-response relationships. The association between lipid indices and the risk of delivering LGA infants was further investigated among different subgroups. Receiver operating characteristic curves were utilized to assess the diagnostic performance of lipid indices.

RESULTS

Across crude and adjusted models, females with lipid indices in the upper two tertiles presented a markedly elevated risk of delivering LGA infants compared with the lowest tertile category. Conversely, high-density lipoprotein cholesterol levels demonstrated the contrary trend. Restricted cubic spline analyses revealed linear associations between the five lipid indices, except triglyceride levels, and the prevalence of LGA. The subgroup analysis highlighted that the correlation between lipid indices and the probability of LGA was inconsistent. The five lipid indices presented significant diagnostic efficacy, as indicated by receiver operating characteristic curve areas.

CONCLUSION

Our research demonstrated that lipid indices were effective predictors of the incidence of LGA infants in GDM-affected pregnancies irrespective of potential confounding factors.

Key Words: Gestational diabetes mellitus; Large-for-gestational-age infant; Lipid indices; Restricted cubic spline; Subgroup analysis

Core Tip: Large-for-gestational-age (LGA) infants are a primary complication associated with gestational diabetes mellitus. While maternal hyperglycemia is a recognized risk factor for LGA, it may not be the sole contributor. Epidemiological studies have indicated that abnormal lipid metabolism significantly contributes to insulin resistance, a key pathophysiological mechanism in gestational diabetes mellitus. However, the link between lipid indices and LGA infant risk remains contentious, with inconsistent findings across studies. Consequently, the aim of the study was to assess the association between maternal lipid indices and LGA infant risk in our cohort.

INTRODUCTION

Large for gestational age (LGA) infants, whose birth weight exceeds the 90^th percentile for their gestational age[1], are frequently linked to gestational diabetes mellitus (GDM)[2]. LGA increases the likelihood of adverse perinatal outcomes[3], including preterm birth, cesarean delivery, shoulder dystocia, neonatal hypoglycemia, and hypoxic brain damage and predisposes infants to obesity, reduced glucose tolerance, and type 2 diabetes mellitus in the long term[4]. Global prevalence of GDM ranges from 1% to 30%[5], with rates in mainland China between 12.8% and 16.7%[6]. The rise in GDM incidence is largely driven by lifestyle changes, increasing obesity, and improved screening. While maternal hyperglycemia is a well-documented contributor to LGA, other factors like maternal overweight, above-recommended gestational weight gain (GWG), genetic predisposition, and ethnicity also contribute to fetal overgrowth[7]. Therefore, maternal hyperglycemia may not be the sole determinant of the increased chance of LGA infants within the GDM population.

Recent studies highlighted abnormal lipid metabolism as a key contributor to insulin resistance (IR), a central aspect of GDM[8]. Lipid indices include high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglyceride (TG), and total cholesterol (TC), apart from their calculated indices, such as TG/HDL-C, TC/HDL-C, TG-glucose (TyG) index, and LDL-C/HDL-C. Research has indicated that maternal hypertriglyceridemia is related to infant birth weight in pregnant females[9]. In a study by Marques Puga et al[10], third-trimester hypertriglyceridemia was an autonomous contributor to LGA in females with GDM, without regard to glycemic control. Nevertheless, the function of serum lipids in the occurrence of LGA has been overlooked in various subgroups, including body mass index (BMI) subgroups.

Olmos et al[11] discovered that maternal serum TG levels were unrelated to infant birth weight in normal-weight pregnant females with GDM but were connected to infant birth weight in GDM-affected pregnant females who were obese or overweight. The TG/HDL-C ratio has been documented as a valuable tool for monitoring IR during pregnancy[12-14] while forecasting the possibility of GDM in pregnant females and LGA in newborns[13,15,16].

The TyG index, which is calculated by using fasting plasma glucose (FPG) and serum TG levels, has lately been confirmed as a reliable substitute for identifying individuals with IR and metabolic disorders. This has attracted the attention of several investigators[17-21]. Although earlier studies have examined the link between the TyG index and GDM risk across various regions, including southeast[22] and northwest China[19], Mexico[23], Korea[24], Hungary[25] and Iran[15], there may be variations in this relationship between races and ethnicities. For example, Sánchez-García et al[26] found that in pregnant Latin American females the TyG index did not show significant predictive value for GDM. The findings from the meta-analysis by Song et al[27] showed that the higher TyG index had limited predictive power specifically for Asian females. While it could independently predict GDM in this group, it had no such effect on non-Asian women. The utility of the TyG index as a predictor of the risk of the likelihood of LGA infants in GDM pregnancies remains unknown. There is considerable inconsistency regarding the ties between different lipid indices and the chance of delivering LGA infants.

Therefore, considering the role of race and ethnicity in the lipid indices of pregnant females, improper management of confounding variables, and absence of dose-response analysis, we investigated the associations between lipid indices, including TG level, the ratio of TG/HDL-C, HDL-C level, the ratio of TC/HDL-C, and the TyG index, assessed in the second trimester and the likelihood of giving birth to LGA infants among GDM-affected pregnancies. We innovatively used restricted cubic spline (RCS) models to elucidate the non-linear interaction between lipid indices and the likelihood of giving birth to LGA infants in females with GDM in China and compared the predictive capacities of different lipid indices for LGA among different subgroups. Finally, we determined the diagnostic efficacy via receiver operating characteristic (ROC) curves and the corresponding area under the curve (AUC) to formulate a composite prediction model combining traditional risk factors [pre-pregnancy BMI (pre-BMI) and GWG] to increase the diagnostic efficacy of lipid indices.

MATERIALS AND METHODS

Study populations

A detailed description of the study design and methods was previously published[28]. This was a continuous retrospective study of pregnant females with GDM attending Nanjing Maternal and Child Health Hospital from January 2017 to September 2021. Initially, 10426 pregnant females who were diagnosed with GDM according to the criteria of the International Association of the Diabetes and Pregnancy Study Groups[29] were enrolled. The study was conducted with approval from the Nanjing Maternal and Child Health Hospital Medical Ethics Committee (No. 2022KY-068). The inclusion criteria were: (1) 18-49 years old; (2) Gestational age at birth > 35 weeks; and (3) Pregnancy with single fetus. The following criteria were used to exclude participants: (1) Diagnosis of diabetes mellitus before pregnancy; (2) Severe acute or chronic diseases, including liver disease, hypertension, kidney disease, cardiovascular disease, autoimmune disease, and serious psychiatric disorders; (3) Lack of complete lipid index records; and (4) Absence of data on the variables of interest. All related data were extracted from electronic medical records by trained researchers. We classified infants based on the Chinese neonatal birth weight standard curve: > 90^th percentile as LGA and between the 10^th and 90^th percentile as normal for gestational age (NGA), after adjusting for sex and gestational age[30]. Of 5499 pregnant females with GDM, 976 infants were in the LGA group and 4523 in the NGA group. Figure 1 details the flow chart for the selection of participants.

Figure 1 Study population selection and data analysis flowchart. GDM: Gestational diabetes mellitus; NGA: Normal for gestational age; LGA: Large for gestational age; RCS: Restricted cubic spline; NJMCHH: Nanjing Maternal and Child Health Hospital.

Data collection

Detailed medical histories and anthropometric measurements were obtained for all the enrolled pregnant females. The following baseline data were collected from participants during the first visit: Age; pre-pregnancy weight and height; gravidity; and parity. During antenatal visits, body weight and systolic and diastolic blood pressure were measured. GWG before the oral glucose tolerance test (OGTT) was defined as weight at the time of the OGTT visit minus pre-pregnancy weight. After delivery, information on maternal and perinatal outcomes, including gestational week at delivery, mode of delivery, infant sex, and infant weight at birth, was collected. Blood samples were collected from the participants’ peripheral veins after an overnight fast (≥ 8 h) at 24-28 weeks of gestation. For all participants, the parameters were obtained from routine laboratory biochemical tests, including FPG, glycated hemoglobin (HbA1c), TG, TC, HDL-C, LDL-C, aspartate aminotransferase (AST), lactate dehydrogenase (LDH), alkaline phosphatase, gamma-glutamyl transferase, alanine aminotransferase (ALT), thyroid-stimulating hormone, thyroid peroxidase antibody, free tetraiodothyronine (FT4), total bilirubin, urea, creatinine, total protein (TP), uric acid, albumin (ALB), total bile acid, and direct bilirubin. The Nanjing Maternal and Child Health Hospital clinical laboratory conducted the biochemical marker measurements. All available clinical and laboratory data for all participants were extracted from the electronic medical records and checked by two researchers to ensure data quality.

Data measures

The formulas were as follows: Divide weight (kg) by the square of height (m) to get BMI. TyG index was calculated as Ln [TG (mg/dL) × FPG (mg/dL)/2]. TG/HDL-C ratio was the ratio of TG to HDL-C. ALT/AST ratio was the ratio of ALT to AST, and TC/HDL-C ratio was the ratio of TC to HDL-C.

Covariates

Multivariate adjusted analyses included several covariates known to potentially influence the association between lipid indices and LGA risk, including maternal age, pre-BMI, gravidity (once or ≥ 2 times), parity (primiparous or multiparous), GWG, abdominal circumference (AC) at OGTT, and infant sex. We also adjusted for the following biochemical indicators as confounders: ALT/AST ratio and HbA1c; LDH; FT4; urea; TP; and ALB levels.

Statistical analysis

Continuous variables were tested for normality of distribution via the Kolmogorov-Smirnov test. Depending on their distribution, the variables were presented as the mean ± SD or median and interquartile range. Categorical variables were summarized as frequencies and percentages. For continuous variables the independent samples t-test or Mann-Whitney U test was used to compare groups and the χ²-test or Fisher’s exact test for categorical variables.

According to the criteria for Chinese adults, pre-BMI was classified into four categories: Underweight (< 18.5 kg/m²); normal weight (18.5-23.9 kg/m²); overweight (24.0-27.9 kg/m²); and obese (≥ 28.0 kg/m²)[31]. The following variables were stratified into tertiles using cutoff values determined from the population’s overall distribution. For trend analysis the first tertile served as the reference: TG level (tertile 1, ≤ 1.810; tertile 2, 1.811-2.430; tertile 3, ≥ 2.431); HDL-C level (tertile 1, ≤ 2.120; tertile 2, 2.121-2.490; tertile 3, ≥ 2.491); TG/HDL-C ratio (tertile 1, ≤ 0.750; tertile 2, 0.751-1.104; tertile 3, ≥ 1.105); TC/HDL-C ratio (tertile 1, ≤ 2.398; tertile 2, 2.399-2.781; tertile 3, ≥ 2.782); and TyG index (tertile 1, ≤ 8.804; tertile 2, 8.805-9.114; tertile 3, ≥ 9.115).

To investigate the associations between lipid indices and the risk of delivering LGA infants in pregnant females with GDM, binary logistic regression analyses were performed to determine the odds ratio (OR) and 95% confidence interval (CI) for LGA according to the respective lipid indices, with and without adjustment for potential confounders. The crude model lacked covariate adjustments. Model 1 was adjusted for maternal age, pre-BMI (underweight, normal weight, overweight, and obese), gravidity (1 or ≥ 2 pregnancies), parity (primiparous or multiparous), GWG before the OGTT (less than, within, and greater than recommended range), AC at the time of the OGTT, and infant sex. Model 2, the fully adjusted model, was further adjusted for biochemical indicators, including HbA1c, ALT/AST ratio, LDH, FT4, urea, TP, and ALB. The lowest tertile of each variable category was used as a reference. We plotted these associations via forest plots.

To explore the nonlinear relationship between lipid indices and the risk of LGA infant delivery among pregnant females with GDM, we performed RCS analysis, adjusting for the same covariates as in the binary logistic regression analyses described above. Three knots were selected at the 25^th, 50^th, and 75^th percentiles. Moreover, subgroup analyses were conducted on the basis of pre-BMI (< 24 or ≥ 24 kg/m²) and GWG (less than, within, and greater than recommended range) via stratified multivariate regression analysis. ROC curves and AUCs were used to determine the diagnostic efficacy of lipid indices in LGA infant delivery risk in GDM pregnancies. The optimal cutoff points of the lipid indices for predicting LGA were assessed via the maximum Youden index on the ROC curve. The Delong test evaluated differences in AUC values. Statistical analyses were performed using SPSS version 26.0 and R version 4.2.1. A two-sided P < 0.05 was the significance threshold.

RESULTS

Study participant profiles

Among 5499 pregnant females with GDM, 17.75% (976/5499) delivered LGA infants. The demographic and anthropometric characteristics of the participants are shown in Table 1. Females who delivered LGA infants (LGA group) had significantly greater pre-BMI, gravidity, and parity than females who delivered NGA infants (NGA group). At the time of the OGTT, the weight, GWG, BMI, and AC were significantly greater in the LGA group than in the NGA group (all P < 0.05). The incidence of cesarean section and mother-infant separation was also greater in the LGA group. Among patients with GDM no significant differences were found in age, systolic and diastolic blood pressure at OGTT, gestational age at OGTT, and delivery type between mothers of NGA and LGA infants (all P > 0.05). The biochemical characteristics of the participants at the time of the OGTT are presented in Table 2. The LGA group exhibited significantly higher FPG, HbA1c, TG, TG/HDL-C ratio, TC/HDL-C ratio, TyG index, and LDH but lower HDL-C, ALT, AST, ALT/AST ratio, FT4, urea, TP, and ALB than the NGA group (all P < 0.05). There were no other statistically significant differences in the biochemical indicators between the groups (all P > 0.05).

Table 1 Maternal and neonatal characteristics.

Characteristics	NGA (n = 4523)	LGA (n = 976)	P value
Maternal characteristics before pregnancy
Age, years	29.90 ± 3.41	30.08 ± 3.53	0.132
Pre-pregnancy BMI, kg/m2	21.80 ± 2.89	22.78 ± 3.25	< 0.001
< 18.5	407 (9.00)	48 (4.92)
18.5-23.9	3224 (71.28)	628 (64.34)
24.0-27.9	729 (16.12)	231 (23.67)
≥ 28	163 (3.60)	69 (7.07)
Gravidity			< 0.001
First pregnancy	2377 (52.55)	424 (43.44)
Multiple pregnancies	2146 (47.45)	552 (56.56)
Parity			< 0.001
Primiparous	3376 (74.64)	629 (64.45)
Multiparous	1147 (25.36)	347 (35.55)
Maternal characteristics at the time of the OGTT
Gestational weeks at the time of the OGTT, weeks	25.72 ± 0.85	25.71 ± 0.84	0.522
Weight, kg	63.01 ± 8.44	67.70 ± 9.51	< 0.001
GWG, kg	6.00 (4.00-7.80)	7.00 (5.00-9.00)	< 0.001
BMI, kg/m2	24.04 ± 2.95	25.42 ± 3.29	< 0.001
SBP, mm/Hg	111.94 ± 10.73	111.75 ± 10.70	0.617
DBP, mm/Hg	69.76 ± 8.00	69.61 ± 8.06	0.596
AC, cm	89.08 ± 6.41	92.12 ± 6.94	< 0.001
Neonatal characteristics
Gestational weeks at delivery, weeks	39.52 ± 0.94	39.53 ± 0.91	0.752
Infant weight, kg	3.31 ± 0.28	3.96 ± 0.26	< 0.001
Infant sex, male	2154 (47.62)	490 (50.20)	0.143
Characteristics during delivery
Cesarean section	1579 (34.91)	540 (55.33)	< 0.001
Separation of the mother and infant	3466 (76.63)	800 (81.97)	< 0.001

NGA: Normal for gestational age; LGA: Large for gestational age; BMI: Body mass index; OGTT: Oral glucose tolerance test; GWG: Gestational weight gain; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; AC: Abdominal circumference.

Table 2 Comparison of biochemical characteristics at the time of the oral glucose tolerance test between the normal for gestational age and large for gestational age groups.

Variables	NGA (n = 4523)	LGA (n = 976)	P value
Glucose indices
FPG, mmol/L	4.64 ± 0.29	4.68 ± 0.26	< 0.001
HbA1c, %	5.05 ± 0.24	5.11 ± 0.25	< 0.001
Lipid indices
TG, mmol/L	2.06 (1.65-2.61)	2.26 (1.82-2.89)	< 0.001
TC, mmol/L	6.01 ± 0.97	5.95 ± 1.00	0.112
HDL-C, mmol/L	2.34 ± 0.44	2.26 ± 0.43	< 0.001
LDL-C, mmol/L	3.01 ± 0.71	3.01 ± 0.74	0.916
TG/HDL-C ratio	0.90 (0.67-1.21)	1.00 (0.75-1.41)	< 0.001
TC/HDL-C ratio	2.62 ± 0.48	2.69 ± 0.48	< 0.001
TyG index	8.95 ± 0.36	9.05 ± 0.35	< 0.001
Liver enzymes
ALT, U/L	16.55 (13.95-16.76)	14.30 (10.60-21.10)	< 0.001
AST, U/L	17.80 (15.10-21.80)	16.70 (14.50-20.25)	< 0.001
ALT/AST ratio	0.94 (0.74-1.18)	0.86 (0.69-1.10)	< 0.001
ALP, U/L	58.30 ± 13.52	58.39 ± 13.81	0.865
LDH, U/L	165.06 ± 25.70	167.44 ± 30.45	0.011
GGT, U/L	13.10 (10.10-17.60)	12.50 (9.7017.10)	0.560
Thyroid function index
TSH, mIU/L	2.00 (1.46-2.72)	1.99 (1.37-2.77)	0.910
FT4, pmol/L	12.67 ± 1.76	12.20 ± 1.70	< 0.001
TPOAb, IU/mL	13.44 (9.73-18.34)	13.69 (9.78-18.66)	0.192
Other indices
Urea, mmol/L	3.01 ± 0.73	2.92 ± 0.72	<0.001
Cr, μmol/L	39.42 ± 6.25	39.05 ± 6.35	0.093
UA, μmol/L	229.98 ± 46.99	231.79 ± 50.77	0.282
TP, g/L	65.76 ± 3.31	65.44 ± 3.21	0.007
ALB, g/L	39.41 ± 2.29	39.09 ± 2.32	< 0.001
TBIL, μmol/L	5.67 ± 2.13	5.59 ± 2.08	0.287
DBIL, μmol/L	1.78 ± 0.65	1.75 ± 0.60	0.200
TBA, mmol/L	1.50 (1.00-2.20)	1.50 (1.10-2.20)	0.814

NGA: Normal for gestational age; LGA: Large for gestational age; FPG: Fasting plasma glucose; HbA1c: Glycosylated hemoglobin; TG: Triglycerides; TC: Total cholesterol; HDL-C: High-density lipoprotein cholesterol; LDL-C: Low-density lipoprotein cholesterol; TyG: Triglyceride-glucose index; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; ALP: Alkaline phosphatase; LDH: Lactate dehydrogenase; GGT: Gamma-glutamyl transferase; TSH: Thyroid-stimulating hormone; FT4: Free tetraiodothyronine; TPOAb: Thyroid peroxidase antibody; Cr: Creatinine; UA: Uric acid; TP: Total protein; ALB: Albumin; TBIL: Total bilirubin; DBIL: Direct bilirubin; TBA: Total bile acid.

Lipid indices-LGA delivery risk associations in females with GDM

The values of the lipid indices were divided into tertiles, and the lowest tertiles were used as references. The risk of LGA infants in pregnant females with GDM was analyzed in association with lipid indices in crude and adjusted models via binary logistic regression, as shown in Supplementary Table 1 and Figure 2. In the crude model for the TG level, TG/HDL-C ratio, TC/HDL-C ratio, and TyG index, females with indices in the upper two tertiles had a significantly greater risk of delivering LGA infants than those with indices in the lowest tertile (all P < 0.05). In contrast the HDL-C level showed a negative correlation with LGA (P < 0.05). All lipid indices showed a significant trend in P values (all P < 0.05).

Figure 2 Odds ratios and 95% confidence intervals for association of lipid indices with the risk of delivering large-for-gestational-age infants in pregnant females with gestational diabetes mellitus. OR: Odds ratios; CI: Confidence interval; TG: Triglyceride; HDL-C: High-density lipoprotein cholesterol; TC: Total cholesterol; TyG: Triglyceride-glucose index.

After adjustment in Models 1 and 2, the associations were broadly similar to those in the crude model. In Model 1 compared with females with values in the lowest tertile, females with values in the highest tertile had a greater risk of delivering LGA infants for each lipid index after adjusting for maternal covariates, including pre-BMI, gravidity, parity, GWG, AC at the time of the OGTT, and infant sex (all P < 0.05). Model 2 also revealed a significantly greater risk of delivering LGA infants (adjusted OR = 1.462, 95%CI: 1.166-1.833 for TG; OR = 0.800, 95%CI: 0.645-0.992 for HDL-C; OR = 1.359, 95%CI: 1.086-1.700 for TG/HDL-C ratio; OR = 1.370, 95%CI: 1.107-1.697 for TC/HDL-C ratio; OR = 1.445, 95%CI: 1.151-1.815 for TyG index, all P < 0.05) after further adjustment for HbA1c, ALT/AST ratio, LDH, FT4, urea, TP, and ALB following the above maternal covariates (Figure 2).

Statistical analysis revealed a significant nonlinear relationship between TG level and LGA risk (P = 0.038). In contrast HDL-C, TG/HDL-C, TC/HDL-C, and TyG demonstrated linear relationships with LGA risk, as demonstrated by nonsignificant nonlinearity tests (P > 0.05). The RCS analysis identified a distinct J-shaped association between TG and LGA risk, characterized by a positive correlation at TG levels < 2.205 mmol/L, followed by a plateau effect at TG concentrations exceeding this threshold (Figure 3). The comprehensive dose-response relationships of all lipid indices, including 95%CIs (shaded areas) are shown in Figure 3. All models were adjusted for maternal age, pre-BMI, gravidity, parity, GWG, AC at OGTT, infant sex, HbA1c, ALT/AST ratio, LDH, FT4, urea, TP, and ALB.

Figure 3 Dose-response relationships between lipid indices and the risk of delivering large-for-gestational-age infants in pregnant women with gestational diabetes mellitus. TG: Triglyceride; HDL-C: High-density lipoprotein cholesterol; TC: Total cholesterol; TyG: Triglyceride-glucose index.

Subgroup analysis

Subgroup analysis stratified by pre-BMI and GWG was performed to explore the associations between lipid indices and LGA infant delivery risk in GDM pregnancies (Figure 4 and Supplementary Figure 1). The results of our subgroup analysis indicated that the associations between the lipid indices and LGA infant delivery risk were inconsistent. Nevertheless, tests for interaction showed that subgroup variables and the associations between lipid indices and LGA risk had no significant effects (all interaction P values were greater than 0.05).

Figure 4 Subgroup analysis of lipid indices with respect to large-for-gestational-age infants risk stratified by pre-pregnancy body mass index. OR: Odds ratios; CI: Confidence interval; TG: Triglyceride; HDL-C: High-density lipoprotein cholesterol; TC: Total cholesterol; TyG: Triglyceride-glucose index.

Model prediction performance

ROC curve analyses for predicting the potential for pregnant females with GDM to deliver LGA infants using different indicators are shown in Supplementary Table 2 and Figure 5. The ROC curve areas for the five lipid indices were significant (all P < 0.05). The AUC and 95%CI for the TG/HDL-C ratio (0.584, 0.564-0.603) was almost the same as for the TyG index (0.580, 0.561-0.60) but were greater than for TG level (0.575, 0.556-0.595), HDL-C (0.557, 0.537-0.577), and TC/HDL-C ratio (0.547, 0.527-0.567). These indices were also compared in terms of their AUC value differences. In accordance with the Delong procedure (Supplementary Table 3), there were no significant differences in TG, HDL-C, TG/HDL-C ratio, TC/HDL-C ratio, TyG index, or composite indicator (TG vs HDL-C, P = 0.115; TG vs TG/HDL-C ratio, P = 0.056). HDL-C vs TC/HDL-C ratio, P = 0.278; TG/HDL-C ratio vs TyG index, P = 0.464; TG/HDL-C ratio vs composite indicator, P = 0.056). When conventional risk factors (pre-BMI and GWG) were combined, the ROC‐AUC of this model increased by 0.655 (95%CI: 0.636-0.673). The combined indicator outperformed other single lipid indices in predictive ability (all P < 0.05).

Figure 5 Receiver operating characteristic curves of lipid indices were used to assess the risk of delivering large-for-gestational-age infants in females with gestational diabetes mellitus. TG: Triglyceride; HDL-C: High-density lipoprotein cholesterol; TC: Total cholesterol; TyG: Triglyceride-glucose index; BMI: Body mass index; GWG: Gestational weight gain; AUC: Area under the curve.

DISCUSSION

This marks an initial large-scale cohort research (n = 5499) from Nanjing, China using rigorous stratification and adjustment methods to systematically investigate the dose-response linkage between lipid indices and the likelihood of giving birth to LGA infants among women with GDM. The chance of having LGA infants increased with an increase in four lipid indices (TG level, the ratio of TG/HDL-C, the ratio of TC/HDL-C, and TyG index) and a decrease in HDL-C levels throughout the second trimester of pregnancy (24-28 gestational weeks) in 5499 pregnant females with GDM in Eastern China. Adjustment for all possible confounding factors, such as maternal age, pre-BMI, gravidity, parity, GWG, AC, infant sex, HbA1c, ALT/AST ratio, LDH, FT4, urea, TP, and ALB did not alter our main results. Notably, this study was the first to report an innovative application of the RCS model to investigate how lipid indices relate to the potential of having LGA infants in Chinese pregnant females after adjusting for confounding variables. Subgroup analyses demonstrated the robustness of this relationship in the presence of various confounders, while the AUCs of these five lipid indices when used individually to predict LGA risk ranged from 0.576 to 0.581. Importantly, the combination of the lipid indices with pre-BMI and GWG demonstrated greater predictive ability than the lipid indices alone in predicting the likelihood of LGA infants among GDM-affected pregnancies.

Thus far, several studies have reported associations between abnormal lipid indices and the likelihood of delivering LGA infants in various research subjects. Studies have shown that females with GDM coupled with elevated TG levels have a higher likelihood of giving birth to LGA infants than other females[26,32]. A recent study reported that elevated TG levels were an independent contributor for LGA in 35914 pregnant females in Northern China[33]. Evidence suggests that maternal TG levels have a robust and clear relationship with birth weight regardless of GDM status[10,32,34-36], which is similar to our results.

Our findings support that an increased TG/HDL-C ratio[13,15,16,37], reduced HDL-C level[38], and increased TyG index[13,15,26,37] are connected to the chance of delivering LGA infants. However, Ji et al[39] noted that TG levels during the first trimester but not throughout the second or third trimesters showed a positive correlation with neonatal birth weight. Several studies did not report independent associations of TG[40,41], HDL-C[32,40], the ratio of TG/HDL-C[41], the ratio of TC/HDL-C[13], and the TyG index[41] with the probability of delivering LGA infants, which is inconsistent with our findings. The observed discrepancy may be attributed to several elements, such as variations in the study population’s characteristics (e.g., maternal age, ethnicity, gestational age, and lifestyle), study design, diagnostic criteria employed, sample size, infant sex, and test results obtained from females at different gestational ages. With respect to the aforementioned studies, the results also revealed the complexity and diversity of the potential for LGA births among females with GDM as well as interactions between GDM in pregnant females and risk factors.

We used a quantitative method, the RCS method, to appraise the correlations among various lipid factors and the predicted likelihood of delivering LGA infants more accurately in pregnant females with GDM. Based on available information our study is the first to assess the dose-response relationships between these variables. Our study demonstrated that among the five lipid indices, only the TG level showed a nonlinear association with LGA outcomes, and a TG threshold of 2.205 mmol/L may be the appropriate cutoff for predicting the link to LGA births in pregnancies with GDM in China. In addition, stratified analysis and interaction assessments consistently showed consistent associations between groups. These findings have the potential to inform public health and clinical practice.

It is well known that IR is the primary pathophysiological feature of GDM, and females with GDM and prominent IR face an increased risk of adverse outcomes, with a particular association with the risk of delivering LGA infants[42,43]. Some researchers have proposed that TG level[44,45] and TG/HDL-C ratio[12,13] may function as basic clinical indicators of IR. In addition, following the first observation of the TyG index as an indicator of IR in healthy individuals in 2008[12,13], many studies have demonstrated that the TyG index is a reliable sign of IR and highlighted its role in forecasting the progression of diabetes[46] and cardiovascular diseases[47,48]. Lipid dysregulation may contribute to IR, β cell dysfunction, inflammatory pathway activation, oxidative stress, and enhanced placental lipid transport, promoting fetal overgrowth[49,50]. In our study the occurrence of LGA showed a progressive increase with rising TG level, TG/HDL-C ratio, TC/HDL-C ratio, and TyG index. The possibility of delivering LGA infants exhibited a significantly greater likelihood in females with lipid indices in the highest tertile than in those with indices in the lowest tertiles (P < 0.05) after accounting for confounding factors. These results support previous findings[13,15]. Overall, these findings demonstrate that these lipid indices are effective indicators of IR, rendering them highly suitable for predicting the hazard of bearing LGA infants among pregnant females with GDM.

Previous studies have investigated the efficacy of lipid indices alone for predicting GDM or LGA; however, few studies have used combined lipid indices, especially the TyG index, to predict the possibility of delivering LGA infants among the patients with GDM. In a cohort of 352 individuals, Liu et al[13] highlighted that the AUCs of the ratio of TG/HDL-C and TyG index for detecting LGA were 0.646 (95%CI: 0.559-0.734) and 0.643 (95%CI: 0.552-0.735), respectively. Similarly, Wang et al[16] demonstrated the capability of the ratio of TG/HDL-C in predicting the likelihood of LGA, with an AUC of 0.668 (95%CI: 0.514-0.823). We found that lipid indices alone have limited utility in predicting the likelihood of having LGA infants in GDM pregnancies, with AUCs ranging from 0.547 to 0.584, which combined with pre-BMI and GWG demonstrated the highest AUC and exceeded those of the other single biomarkers, showing a predictive capacity represented by the AUC value of 0.655 (95%CI: 0.636-0.673). The variations in ROC curves for predicting the odds of LGA births in females with GDM via lipid indices might be attributed to diverse contributors, including disparities in the study population, individual variations, sample size, covariates, and diagnostic criteria used to identify LGA and GDM. In the current study utilizing the Chinese neonatal birth weight curve[30] as the diagnostic reference for LGA might have potentially played a role in the observed difference. The predictive ability of existing lipid indices may not be sufficiently robust. Compared with other lipid indices, the TG/HDL-C ratio and TyG index demonstrated better predictive abilities for LGA. However, the ratio of TG/HDL-C is a more accessible and cost-efficient detection method than the TyG index. Consequently, in future research it is imperative to fully explore the capacity of integrating the ratio of TG/HDL-C with other clinical and biochemical indicators to appraise the risk of LGA.

Strengths and limitations

This study possessed some notable strengths. Our retrospective study had a substantial participant count that was representative of pregnant females living in urban Nanjing. Second, all participants’ morning fasting blood samples were collected simultaneously and analyzed within 2 h of sampling. This method improved the accuracy of glucose and lipid measurements by minimizing potential confounding factors and the use of freeze-thaw processes. We explored not only individual lipid indices but also several combined lipid indices. Third, traditional confounders during pregnancy were carefully and accurately recorded and adjusted for in the analysis. Finally, we used RCS analysis to ensure and improve the fit of the relationships between the lipid indices and the predicted chance of having LGA infants in GDM pregnancies. As far as we know, this study breaks new ground by using an RCS model to independently rank related risk factors for predicting GDM.

Despite this some limitations still existed in this study. First, the data source was conducted as a single-center retrospective analysis with a homogeneous ethnic population, potentially causing selection bias. Second, although rigorous inclusion/exclusion criteria and extensive adjustment for confounders, such as maternal age, pre-BMI, gravidity, parity, GWG, AC, infant sex, and some biochemical characteristics, were implemented to minimize bias owing to the retrospective design of the study, other potential confounders, like economic status, dietary habits, and physical activity level, were not recorded. Third, the cohort in this study was from Nanjing, which is a more developed economic region; therefore, our results cannot be generalized to other regions and populations. It is crucial to substantiate the robustness of our findings and to perform multicenter studies in different geographical locations and in individuals with different ethnic backgrounds. Future prospective multicenter studies are planned to validate the results and causal inferences. Finally, our study lacked biomarkers (e.g., β cell function, adipokine levels, oxidative stress, and inflammatory factor levels) to explore the biological mechanisms involved and define the straightforward causal link between the lipid indices and the chance of giving birth to LGA infants in GDM pregnancies.

CONCLUSION

The current study demonstrated that increased TG, ratio of TG/HDL-C, ratio of TC/HDL-C, and TyG index along with decreased HDL-C during the second trimester were good predictors of the probability of having LGA infants in Chinese GDM-affected pregnancies, independent of potential confounding factors. Lipid indices are economical and routinely measured and can aid primary healthcare in screening high-risk pregnancies for LGA. In addition, we identified a TG threshold of 2.205 mmol/L as potentially suitable for risk prediction in the Chinese population. This study confirmed that there were no variations among the values of TG, HDL-C, and the ratio of TC/HDL-C in the prediction of the risk of delivering LGA infants. Compared with these lipid indices, the TG/HDL-C ratio and TyG index exhibited better predictive abilities for LGA. All five lipid indices, combined with pre-BMI and GWG, presented the highest AUC for predicting the odds of having LGA infants in females with GDM. Nevertheless, the effectiveness of these lipid indices in diagnosing LGA was restricted. Therefore, lipid indices are valuable screening tools that can be integrated into routine obstetric clinical evaluation processes and early identification can facilitate community-based interventions, reducing adverse maternal and neonatal outcomes. Further investigation is essential to understand the causes and impacts of these relationships, shed light on the biological mechanisms at play, and boost maternal and infant outcomes via interventions.