Abstract

Aims:

This study investigated the associations of four insulin resistance (IR) indicators—the triglyceride-glucose (TyG) index, the TyG-body mass index (TyG-BMI), the metabolic score for IR (METS-IR), and the triglyceride to high-density lipoprotein cholesterol (TG/HDL-C) ratio—with subsequent pregnancy outcomes in women with recurrent pregnancy loss (RPL) and assessed their predictive value.

Methods:

This cohort study recruited RPL participants from the Chinese Pregnancy Loss Cohort. Enrollment occurred between September 2019 and December 2022. All participants were followed up every 6 months, with a minimum follow-up duration of 18 months, to document pregnancy outcomes (live birth or subsequent pregnancy loss). Univariate and multivariate logistic regression analyses were performed to assess the associations between four IR indicators (TG/HDL-C, TyG, TyG-BMI, METS-IR) and pregnancy outcomes. Receiver operating characteristic (ROC) analysis was conducted to determine the predictive efficacy of each indicator.

Results:

Among 2,454 screened participants, 897 RPL women were analyzed (638 live births, 71.1%; 259 pregnancy losses, 28.9%). In the fully adjusted model, the highest tertiles of TyG-BMI and METS-IR were associated with significantly elevated odds of pregnancy loss (OR = 1.52, 95% CI: 1.01–2.27, P = 0.044; OR = 1.49, 95% CI: 1.05–2.29, P = 0.045, respectively). METS-IR demonstrated the highest predictive efficacy for pregnancy outcomes (AUC = 0.710), followed by TyG-BMI, TG/HDL-C, and the TyG index.

Conclusions:

Among women with RPL, TyG-BMI and METS-IR are independently associated with increased pregnancy loss risk, with METS-IR demonstrating superior predictive performance.

1 Introduction

Recurrent pregnancy loss (RPL) is defined as two or more spontaneous pregnancy losses before 24 weeks of gestation (1). The etiology of RPL is complex and involves genetic predispositions, anatomical anomalies, endocrine disorders, immune dysfunction, and infectious agents (2). However, approximately 50% of RPL cases are classified as idiopathic, lacking a definitive etiological diagnosis or targeted intervention strategies, which poses a significant clinical challenge for reproductive-aged women worldwide (3).

Insulin resistance (IR), a prevalent metabolic disorder characterized by diminished sensitivity of target tissues to insulin, has gained recognition as a potential underlying pathogenic factor in unexplained recurrent pregnancy loss (URPL) (4). IR has been definitively linked to reproduction-related abnormalities, such as polycystic ovarian syndrome (PCOS), endometrial dysfunction (5), and compromised oocyte quality (6), each of which significantly contributes to adverse pregnancy outcomes.

The preconception period is a critical window for intervening in metabolic abnormalities and optimizing pregnancy outcomes (7, 8). However, current studies on whether IR levels during this crucial window influence subsequent pregnancy outcomes in patients with RPL remain scarce. Various methods are available for assessing insulin resistance, among which the Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) is widely used due to its simplicity requiring only fasting glucose and insulin levels. However, HOMA-IR cut-off values vary significantly across ethnicities, sexes, and ages. For instance, the 75th percentile values range from 1.6 in Iranians to 2.53 in Koreans, 2.0 in Swedish men, and 3.8 in French men (9). Among Asian populations, the 90th percentile was 1.7 in Japanese (10), while ROC-derived optimal cut-offs were 2.39-2.48 for Korean women (9), and 1.4-2.0 for Southern Chinese (11). Given these variations, the present study adopted a cut-off value of 2.5 based on previous studies in RPL patients as the diagnostic criterion for HOMA-IR (12).

As an indirect method, the measurement of IR indicators is derived from readily available and cost-effective clinical parameters such as blood glucose, triglycerides, and body mass index (BMI), without additional insulin testing. Common IR indicators include the triglyceride-glucose (TyG) index (13), TyG-body mass index (TyG-BMI) (14), metabolic score for insulin resistance (METS-IR) (15), and triglyceride-high-density lipoprotein cholesterol (TG/HDL) ratio (16). Although these IR indicators have been used to evaluate IR status in various populations (17–20). Whether they can effectively identify pre-pregnant RPL women at high risk of pregnancy loss has not been fully verified. Therefore, this study aims to enroll a prospective cohort of women with RPL to examine the association between the value of four pre-conception IR-related indicators (TyG index, TyG-BMI, METS-IR, and TG/HDL-C) and the risk of subsequent pregnancy outcomes.

2 Materials and methods2.1 Data source

Data for this study were extracted from the Pregnancy Loss Cohort of the Real-World Study on Recurrent Pregnancy Loss in China (Project number: YJS-BD-19). The cohort is registered in the Chinese Clinical Trial Registry (registration number: ChiCTR2000039414). This study was approved by the Ethics Committee of Lanzhou University Second Hospital (approval number: 2019A-231).

2.2 Participants and selection criteria

Between September 2019 and December 2022, we enrolled women with one or more prior pregnancy losses into the cohort. All participants provided written informed consent at enrollment, and information on potential etiologies, risk factors, and medical examinations was collected at recruitment. The inclusion criteria were a history of two or more pregnancy losses, age ranging from 18 to 42 years, and a clear intention to pursue future fertility, while the exclusion criteria included pregnancy loss caused by a single confirmed, corrected or irreversible etiology, a history of PCOS, a confirmed diagnosis of infertility, incomplete key medical record data precluding comprehensive analysis, and severe mental illness that prevents voluntary signing of informed consent for follow-up.

After the initial consultation, detailed disease histories were recorded for each participant, including age, ethnicity, education, height, weight, pregnancy losses, menstrual history, and other medical details. A comprehensive physical exam and uterine imaging were then conducted. Participants were monitored every six months for at least 18 months, ending by August 2024.

Blood samples were taken after an 8-hour fast and analyzed within 3 hours. Fasting blood glucose (FBG), fasting insulin (INS), fasting C-peptide (FCP), 2-hour postprandial blood glucose (2hPBG), 2-hour postprandial insulin (2hINS), 2-hour postprandial C-peptide (2hCP), and 25−hydroxyvitamin D [25(OH)D] were measured by chemiluminescent immunoassay (Roche Diagnostics GmbH, Mannheim, Germany). Furthermore, levels of total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), uric acid (UA), total bilirubin (TBIL), direct bilirubin (DBIL), and indirect bilirubin (IBIL), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and homocysteine (HCY) were quantified using the Beckman Coulter AU5800. Serum thyroid function indicators, including thyroid-stimulating hormone (TSH), free triiodothyronine (FT3), and free thyroxine (FT4), were quantified using an enzyme-linked immunosorbent assay (Roche Diagnostics GmbH, Mannheim, Germany).

2.3 Pregnancy outcomes

Women with RPL who became pregnant again after their initial visit experienced two outcomes: pregnancy loss and live birth. Pregnancy loss is defined as spontaneous pregnancy loss before 24 weeks, while live birth refers to the survival of the fetus at or after 24 weeks of gestation.

2.4 Definition of variables

In this study, IR was assessed based on the HOMA-IR, which is derived from fasting plasma glucose and fasting insulin concentrations. IR was diagnosed by the following indicators (1): impaired glucose tolerance (IGT) or diabetes mellitus (21) (2); fasting insulin levels (FINS) ≥15 mIU/L (22) (3); HOMA-IR = [FPG (mmol/L) × INS (mIU/L)]/22.5 > 2.5 (23).

Four indicators of IR were calculated using the following equations:

Laboratory measurements of FBG, TG, and HDL-C, initially recorded in mmol/L, were converted to mg/dL using the following conversion factors: HDL-C (mg/dL) = HDL-C (mmol/L) × 38.7; FBG (mg/dL) = FBG (mmol/L) × 18.0; TG (mg/dL) = TG (mmol/L) × 88.5.

2.5 Statistical analysis

Categorical variables were represented as percentages, while continuous variables with a normal distribution were expressed as mean ± standard deviation. Group comparisons were conducted using the independent samples t-test. For continuous variables with a skewed distribution, data were presented as median and interquartile ranges (IQR). Categorical data were also presented as counts with corresponding percentages. Missing data were addressed through multiple imputations. To investigate potential associations between four IR indicators and the HOMA-IR, a generalized additive model (GAM) with smooth curve fitting was employed. Optimal cutoff values were identified using the maximum Youden index. Univariate and multivariate logistic regression analyses were conducted to examine the associations between the four IR-related indicators and subsequent pregnancy outcomes, with results reported as odds ratios (ORs) and 95% confidence intervals (CIs). Additionally, receiver operating characteristic (ROC) curve analysis was performed to evaluate the discriminative ability of the TyG index, TyG-BMI, TG/HDL-C, and METS-IR in predicting pregnancy outcomes. Stratified sensitivity analyses were conducted according to age, number of pregnancy losses, and insulin resistance treatment status. All statistical analyses were performed using EmpowerStats software version 5.2 (https://www.empowerstats.com, X&Y Solutions, Inc., Boston, MA) and R language (http://www.R-project.org).

3 Results3.1 Baseline characteristics

A total of 2,454 women were recruited between September 10, 2019, and December 31, 2022. After excluding 812 individuals (due to one previous pregnancy loss, polycystic ovary syndrome, uterine malformation, or other comorbidities), 1,642 women with recurrent pregnancy loss were identified. During follow-up (until August 2024), an additional 745 participants were excluded (including loss to follow-up, infertility, missing key laboratory data, etc.), resulting in 897 eligible participants (Figure 1). Within this cohort, 638(71.1%) participants achieved a live birth outcome, while 259(28.9%) experienced subsequent pregnancy loss.

Flowchart of participant enrollment and follow-up in women with recurrent pregnancy loss.

A total of 897 patients with RPL were classified into two groups: the non-IR group (n= 472, 52.6%) and the IR group (n=425, 47.4%) (Table 1). Age showed a significant between-group difference (P = 0.005), with no significant differences observed in other demographic and reproductive baseline characteristics, including educational level, ethnicity, menstrual status, pregnancy loss type, and number of previous pregnancy losses (all P > 0.05). The clinical intervention rate differed significantly between the two groups (P< 0.001): no participants in the non-IR group received intervention, while 60.47% (257/425) of the IR group underwent clinical intervention.

ParametersNon-IR groupIR groupP-valueN472425Age(years)30.07 ± 4.0630.86 ± 4.290.005BMI (kg/m2)21.61 ± 2.5523.33 ± 3.21<0.001Education (n, %)0.440 Primary School29 (6.14)18 (4.24) High School128 (27.12)117 (27.53) College315 (66.74)290 (68.24)Race (n, %)0.962 Han nationality422 (89.41)382 (89.88) Hui nationality31 (6.57)26 (6.12) Other minority nationality19 (4.03)17 (4.00)Menstrual cycles (n, %)0.529 Regular410 (86.86)363 (85.41) Irregular62 (13.14)62 (14.59)Pregnancy loss types (n, %)0.963 Primary367 (77.75)331 (77.88) Secondary105 (22.25)94 (22.12)Previous pregnancy losses (n, %)0.713 2 times304 (64.41)272 (64.00) 3 times118 (25.00)101 (23.76) ≥ 4 times50 (10.59)52 (12.24)Clinical intervention (n, %)<0.001 No472 (100.00)168 (39.53) Yes0 (0.00)257 (60.47)25(OH)D (nmol/L)12.82 ± 5.2612.95 ± 4.850.713FT3 (pmol/L)5.21 ± 0.585.17 ± 1.130.501FT4 (pmol/L)16.22 ± 2.5316.08 ± 3.180.490TSH (mIU/L)2.43 (1.70-3.10)2.46 (1.78-3.17)0.140HCY (μmol/L)11.50 ± 4.1611.18 ± 4.180.248FPG (mmol/L)4.84 ± 0.465.31 ± 1.04<0.001INS (mIU/L)7.38 (5.71-9.11)14.64 (12.34-18.99)<0.001FCP (ng/ml)1.05 (0.87-1.28)1.74 (1.46-2.16)<0.0012hPG (mmol/L)5.54 (4.93-6.20)6.40 (5.63-7.61)<0.0012hINS (mIU/L)30.49 (20.93-45.38)58.45 (36.51-86.84)<0.0012hCP (ng/ml)3.82 (2.98-5.10)5.83 (4.16-7.70)<0.001TC (mmol/L)3.90 ± 0.734.06 ± 0.760.002TG (mmol/L)1.02 ± 0.531.34 ± 0.79<0.001HDL-C (mmol/L)1.40 ± 0.341.31 ± 0.36<0.001LDL-C (mmol/L)2.41 ± 0.582.59 ± 0.68<0.001Cr (μmol/L)53.12 ± 7.7853.25 ± 8.220.808UA (μmol/L)254.93 ± 57.08275.55 ± 67.34<0.001TBIL (μmol/L)12.83 ± 5.3612.47 ± 4.800.290DBIL (μmol/L)3.11 ± 2.242.75 ± 1.690.008IBIL (μmol/L)9.69 ± 4.499.69 ± 4.380.986ALT(U/L)14.00 (10.00-19.00)18.00 (12.00-26.00)<0.001AST(U/L)20.00 (17.00-23.09)21.46 (18.00-26.00)<0.001TyG index8.17 ± 0.448.51 ± 0.51<0.001TG/HDL-C1.80 ± 1.252.61 ± 1.98<0.001TyG-BMI176.70 ± 24.41198.82 ± 32.03<0.001METS-IR30.46 ± 4.6834.47 ± 5.99<0.001

Baseline clinical and biochemical characteristics of participants.

Data are shown as mean ± standard deviation, median(min-max) or frequency with percentages.

IR, insulin resistance; BMI, body mass index; 25(OH)D, 25-hydroxyvitamin D; FT3, free triiodothyronine; FT4, free thyroxine; TSH, thyroid-stimulating hormone; HCY, homocysteine; INS, insulin; FPG, fasting plasma glucose; FCP, fasting C-peptide; 2hINS, 2-hour postprandial insulin; 2hPG, 2-hour postprandial plasma glucose; 2hCP, 2-hour postprandial C-peptide; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; UA, uric acid; TBIL, total bilirubin; DBIL, direct bilirubin; IBIL, indirect bilirubin; ALT, alanine transaminase; AST, aspartate transaminase; TyG, triglyceride-glucose index; TG/HDL-C, triglyceride/high-density lipoprotein cholesterol; TyG-BMI, triglyceride-glucose body mass index; RPL, recurrent pregnancy loss; IR, insulin resistance; METS-IR, metabolic score for insulin resistance.

In contrast, compared with the non-IR group, the IR group exhibited significantly elevated levels of BMI, FBG,2hPBG, INS, FCP, TC, TG, LDL-C, UA, ALT, AST, and all IR-related indicators (TyG index, TyG-BMI, METS-IR, and TG/HDL-C) (all P< 0.05). Meanwhile, the IR group had significantly lower levels of HDL-C (P< 0.001) and DBIL (P = 0.008). No significant differences in 25(OH)D, thyroid function indicators (TSH, FT3, FT4), HCY, Cr, TBIL, and IBIL levels were observed between the two groups (all P > 0.05).

3.2 The association between four IR Indicators and IR

Restricted cubic spline (RCS) analysis with smooth curve fitting revealed nonlinear increasing associations of TG/HDL-C ratio, TyG-BMI, and METS-IR with HOMA-IR, whereas the TyG index showed a linear increasing trend. As shown in Figure 2, threshold effect analysis identified cutoff values of 1.91 for TG/HDL-C ratio (Figure 2A), 7.88 for TyG index (Figure 2B), 144.68 for TyG-BMI (Figure 2C), and 26.75 for METS-IR (Figure 2D).

Smooth curve fitting and threshold effect analysis of four IR indicators and IR (A) TG/HDL-C index (cut-off: 1.91); (B) TyG index (cut-off: 7.88); (C) TyG-BMI index (cut-off: 144.68); (D) METS-IR index (cut-off: 26.75). The solid red line represents the ROC curve, the dashed blue lines represent the 95% confidence interval, and the red dashed vertical line indicates the optimal cut-off value for each index.

3.3 Univariate analysis of subsequent pregnancy outcomes

A positive correlation was identified between age, BMI, a history of ≥ 4 pregnancy losses, TG, TG/HDL-C, TyG-BMI, and METS-IR and the risk of pregnancy loss, with all associations reaching statistical significance (all P< 0.05). The TyG index was marginally associated with pregnancy loss risk (P = 0.057). Higher HDL-C levels were linked to a lower risk of pregnancy loss (OR = 0.62, 95% CI: 0.40–0.97, P = 0.035). No significant correlations were found with educational level, ethnicity, menstrual cycle, type of pregnancy loss, hyperuricemia, and hyperhomocysteinemia (all P>0.05). All these results are summarized in Table 2.

Subsequent pregnancy outcomesParametersNOR (95%CI)P-valueAge(years)30 ± 41.06 (1.02, 1.10)0.001BMI (kg/m2)22.43 ± 3.001.06 (1.01, 1.11)0.022Education (n, %) Primary School47 (5.24)1.0 High School245 (27.31)0.71 (0.37, 1.36)0.302 College605 (67.45)0.61 (0.33, 1.13)0.114Race (n, %) Han nationality804 (89.63)1.0 Hui nationality57 (6.35)0.94 (0.52, 1.72)0.852 Other nationality36 (4.01)0.69 (0.31, 1.54)0.367Menstrual cycles (n, %) Regular773 (86.18)1.0 Irregular124 (13.82)1.20 (0.80, 1.81)0.371Pregnancy loss types (n, %) Primary698 (77.81)1.0 Secondary199 (22.19)1.12 (0.79, 1.57)0.530Previous pregnancy losses (n, %) 2 times576 (64.21)1.0 3 times219 (24.41)1.21 (0.86, 1.71)0.267 ≥ 4 times102 (11.37)1.97 (1.27, 3.05)0.002Clinical intervention No640 (71.35%)1.0 Yes257 (28.65%)0.89 (0.65, 1.23)0.49325(OH)D (nmol/L)12.88 ± 5.061.00 (0.97, 1.03)0.967FT3 (pmol/L)5.19 ± 0.880.72 (0.56, 0.94)0.013FT4 (pmol/L)16.15 ± 2.851.00 (0.95, 1.05)0.982TSH (mIU/L)2.88 ± 2.731.01 (0.96, 1.06)0.709HCY (μmol/L)11.35 ± 4.170.99 (0.96, 1.03)0.657FPG (mmol/L)5.05 ± 0.720.93 (0.74, 1.15)0.485INS (IU/L)11.83 ± 6.890.99 (0.97, 1.01)0.349FCP (ng/ml)1.50 ± 0.851.02 (0.86, 1.21)0.8232hPG (mmol/L)6.23 ± 1.841.05 (0.97, 1.14)0.1952hINS (IU/L)51.28 ± 40.631.00 (1.00, 1.01)0.3442hCP (ng/ml)5.13 ± 2.521.04 (0.98, 1.10)0.225TC (mmol/L)3.98 ± 0.761.12 (0.92, 1.35)0.253TG (mmol/L)1.19 ± 0.851.28 (1.06, 1.54)0.011HDL-C (mmol/L)1.36 ± 0.350.62 (0.40, 0.97)0.035LDL-C (mmol/L)2.49 ± 0.641.25 (1.00, 1.57)0.050Cr (μmol/L)53.19 ± 7.990.99 (0.98, 1.01)0.502UA(μmol/L)265.03 ± 64.091.00 (1.00, 1.00)0.252TBIL (μmol/L)12.74 ± 5.361.00 (0.98, 1.03)0.752DBIL (μmol/L)2.96 ± 2.080.98 (0.91, 1.05)0.566IBIL (μmol/L)9.80 ± 4.681.01 (0.98, 1.04)0.584ALT(U/L)19.51 ± 16.501.00 (0.99, 1.01)0.786AST(U/L)22.95 ± 11.420.99 (0.98, 1.01)0.248TG/HDL-C2.18 ± 1.681.13 (1.04, 1.23)0.003TyG index8.33 ± 0.501.32 (0.99, 1.75)0.057TyG-BMI187.18 ± 30.341.01 (1.00, 1.01)0.009METS-IR32.36 ± 5.701.04 (1.01, 1.06)0.006

Univariate associations between variables and subsequent pregnancy loss.

Note: Data are shown as mean ± standard deviation, median(min-max) or frequency with percentages.

OR, dds ratio; 95% CI, 95% confidence interval. BMI, body mass index; 25(OH)D, 25-hydroxyvitamin D; FT3, free triiodothyronine; FT4, free thyroxine; TSH, thyroid-stimulating hormone; HCY, homocysteine; INS, insulin; FPG, fasting plasma glucose; FCP, fasting C-peptide; 2hINS, 2-hour postprandial insulin; 2hPG, 2-hour postprandial plasma glucose; 2hCP, 2-hour postprandial C-peptide; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; UA, uric acid; TBIL, total bilirubin; DBIL, direct bilirubin; IBIL, indirect bilirubin; ALT, alanine transaminase; AST, aspartate transaminase; TyG, triglyceride-glucose index; TG/HDL-C, triglyceride/high-density lipoprotein cholesterol; TyG-BMI, triglyceride-glucose body mass index; RPL, recurrent pregnancy loss; METS-IR, metabolic score for insulin resistance.

3.4 Association between four IR indicators and subsequent pregnancy outcomes

For evaluating the relationship between four IR indicators and pregnancy outcomes, we conducted a multivariate logistic regression analysis; participants were divided into three groups with 299 individuals in each group, with the low group serving as the reference for subsequent statistical comparisons. For TG/HDL-C: low group (≤ 1.33), middle group (1.33< value ≤ 2.28), and high group (>2.28); for TyG index: low group (≤ 6.76), middle group (6.76< value ≤ 7.16), and high group (>7.16); for TyG-BMI: low group (≤ 144.35), middle group (144.35< value ≤ 164.61), and high group (>164.61); for METS-IR: low group (≤ 25.79), middle group (25.79< value ≤ 30.17), and high group (>30.17).

Three models were used: Model I was an unadjusted model, Model II adjusted for age, and Model III adjusted for indicator-specific sets of covariates (detailed in the Table 3 footnote). In the TG/HDL−C subgroup, the high group showed a significantly higher risk of pregnancy loss in Model I (OR = 1.44, 95% CI: 1.00–2.06,