Abstract
Background
Growing evidence suggests that metabolic disturbances underlie pregnancy losses, imposing a substantial public health burden among young women of reproductive age. Glycolipid metabolism 7 factors (GLM7) is a composite glycolipid metabolism index derived from routinely measured metabolic indicators; however, its relevance to pregnancy loss in young women remains unexamined.
Methods
The present study analyzed publicly available data from the 2009–2018 National Health and Nutrition Examination Survey (NHANES) cycles. Participants were categorized as having no pregnancy loss, single pregnancy loss, or recurrent pregnancy loss (≥ 2 losses). GLM7 was calculated as the logarithm of the product of age, body mass index, fasting blood glucose, fasting insulin, triglycerides, low-density lipoprotein cholesterol, and high-density lipoprotein cholesterol levels. The association between GLM7 and pregnancy loss categories was evaluated, potential dose–response patterns were explored, and effect modification by selected sociodemographic and behavioral factors was assessed.
Results
Of the 736 women included in the analysis, 300 (37%) reported a history of pregnancy loss. In binary analyses, higher GLM7 was initially associated with pregnancy loss (OR = 1.21, 95% CI: 1.03–1.44); however, the association was attenuated after adjustment for the number of pregnancies (OR = 1.19, 95% CI: 0.99–1.44). In multinomial analyses, higher GLM7 levels were not significantly associated with single pregnancy loss. In contrast, higher GLM7 remained independently associated with recurrent pregnancy loss after adjustment for all covariates (OR = 1.17, 95% CI: 1.02–1.42; P = 0.046). Women in the second to fourth quartiles had significantly higher odds of recurrent pregnancy loss than those in the lowest quartile. No significant nonlinearity was observed in the association between GLM7 and recurrent pregnancy loss (P for nonlinearity = 0.265). No statistically significant interaction was observed across the subgroups.
Conclusions
In this nationally representative sample of U.S. women aged 18–35 years, higher GLM7 levels were more robustly associated with recurrent rather than isolated pregnancy loss. The findings suggested that cumulative glycolipid metabolic burden may be particularly relevant to repeated reproductive failures and serve as indicators to improved strategies for early reproductive risk identification and preventive healthcare for women.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s12944-026-02991-6.
Keywords: Pregnancy loss, Glycolipid metabolism, Lipid metabolism, Nutrition surveys, Reproductive health
Background
Pregnancy loss, including miscarriage and stillbirth, is one of the most common adverse reproductive outcomes in women of reproductive age and represents a major public health concern worldwide [1]. Approximately 15–20% of clinically recognized pregnancies are estimated to result in loss [2], imposing substantial physical, psychological, and socioeconomic burdens on affected women and their families. Although a range of factors, such as genetic abnormalities, uterine structural conditions, infections, and endocrine disorders, have been implicated, a considerable proportion of pregnancy loss cases remains unexplained, thereby underscoring the importance of identifying potentially modifiable risk factors, particularly those associated with metabolic health.
Increasing evidence suggests a link between metabolic dysfunction and female reproductive health. Obesity, insulin resistance, impaired glucose metabolism, and dyslipidemia are associated with an increased risk of adverse pregnancy outcomes, including infertility, implantation failure, and pregnancy loss [3, 4]. However, most existing studies have focused on individual metabolic markers, such as body mass index or fasting glucose, which may not fully capture the complex and interrelated nature of metabolic disturbances. Metabolic abnormalities often coexist and interact synergistically, resulting in a cumulative metabolic burden that exerts greater biological effects on reproductive function than any single parameter alone. Consistent with this concept, metabolic syndrome has been linked to an increased risk of both early and late pregnancy loss [5].
Composite metabolic indices, based on routinely measured clinical parameters, have gained increasing attention as comprehensive tools for assessing metabolic health [6–8]. The glycolipid metabolism 7 factors (GLM7) is a recently developed composite glycolipid metabolism index that integrates age, body mass index (BMI), fasting blood glucose (FBG), fasting insulin, triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C). This index has a strong predictive value for cardiometabolic multimorbidity [9]. However, despite growing evidence linking metabolic dysregulation to pregnancy loss, little is known about the role of composite metabolic indices, such as GLM7, in pregnancy loss.
Evidence from life course epidemiology suggests that pregnancy loss may be associated with an increased risk of cardiometabolic diseases [10], indicating a potential link between metabolic health and reproductive outcomes. However, little is known about whether cumulative metabolic burden is associated with recurrent rather than isolated pregnancy loss. Based on previous evidence linking metabolic dysfunction with reproductive outcomes, we hypothesized that a comprehensive metabolic burden, reflected by the multicomponent GLM7 index, could be associated with increased odds of pregnancy loss, particularly recurrent pregnancy loss. To the best of our knowledge, this is the first study to evaluate the association between GLM7 and pregnancy loss. Data from the National Health and Nutrition Examination Survey (NHANES) 2009–2018 were utilized to investigate the relationship between the GLM7 index and pregnancy loss history in women aged 18–35 years, thereby offering a potential framework for preconception metabolic risk stratification.
Methods
Study population
This cross-sectional analysis used publicly available data from NHANES, a nationwide survey designed to assess the health and nutritional status of the U.S. population. Five survey cycles spanning 2009–2018 were combined, during which complete reproductive health questionnaires and laboratory measurements required for the calculation of GLM7 were available. Among 49,693 participants initially screened, male participants were excluded (n = 24,533). Women were excluded if they were younger than 18 or older than 35 years (n = 20,457), had missing data on pregnancy history (n = 2,921), were pregnant at the time of the survey (n = 127), or had missing data on GLM7-related components (n = 919). The final analysis included 736 women (Fig. 1). Included and excluded participants showed generally similar demographic and clinical characteristics (Supplementary Table S1). A complete case-analysis approach was used.
Fig. 1.
Flow diagram of study participants from NHANES 2009–2018. Abbreviations: NHANES, National Health and Nutrition Examination Survey; BMI, Body mass index; FBG, Fasting blood glucose; TG, Triglyceride; HDL-C, High-density lipoprotein cholesterol; LDL-C, Low-density lipoprotein cholesterol; GLM7, Glycolipid metabolism 7 factors
Assessment of pregnancy loss
Pregnancy loss was defined as described in a previous study [11]. Reproductive history was obtained through computer-assisted personal interviews conducted at mobile examination centers (MECs). Information on the total number of pregnancies and live births was obtained from all participants. Pregnancy loss was identified when the total number of reported pregnancies exceeded the number of live births by at least one in non-pregnant participants. The outcome captured all non-live birth pregnancy terminations, including spontaneous miscarriages, induced abortions, stillbirths, and ectopic pregnancies, since these cannot be distinguished in the NHANES self-reported data. For primary analysis, participants were categorized into three mutually exclusive groups, namely no pregnancy loss, single pregnancy loss (exactly one loss), and recurrent pregnancy loss (≥ 2 losses). A binary pregnancy loss history (yes/no) was analyzed as a secondary outcome [12].
Exposure assessment: GLM7
The primary exposure variable was GLM7, a novel composite glycolipid metabolism index derived from seven routinely measured clinical indicators, as described previously by Wang et al. [9]. GLM7 integrates age, BMI, FBG, fasting insulin, TG, LDL-C, and HDL-C, and is calculated using the following formula: log10 [Age (years) * BMI (kg/m2) * FBG (mg/dL) * Insulin (pmol/L) * TG (mmol/L) * LDL‐C (mmol/L)/HDL‐C (mmol/L)]. All components were obtained from standardized NHANES physical examinations and laboratory assessments conducted at the MECs, following rigorous quality control procedures. To improve interpretability, GLM7 was standardized as a z-score, and all primary effect estimates were reported per standard deviation (SD) increase in GLM7.
Covariates
Covariates included: (1) demographic information, including age, race, education level, marital status, and family poverty income ratio (PIR); (2) reproductive characteristics, including age at menarche, age at first and last live birth, pregnancy number, and total number of live births; (3) lifestyle behaviors, including smoking status (≥ 100 cigarettes smoked during lifetime), drinking status (≥ 12 drinks within 1 year), vigorous physical activity (yes/no), and moderate physical activity (yes/no); and (4) clinical characteristics, including BMI, and histories of hypertension, hyperlipidemia, and diabetes (yes/no).
Statistical analysis
All analyses accounted for the complex NHANES sampling design using examination weights, strata, and primary sampling units implemented in the R survey package. A complete-case analytic strategy was adopted, with only participants containing complete information included in the analyses. Multiple imputation was not applied to the GLM7 components, since GLM7 is a multiplicative composite score (log of a product of seven variables); imputing individual components separately would not preserve the mathematical structure of the composite and could introduce bias into the derived score.
Continuous variables were summarized as mean ± SD and categorical variables as percentages. Differences across groups were assessed using chi-square and Kruskal–Wallis tests. Weighted logistic regression models were constructed to evaluate the association between GLM7 and pregnancy loss. The following five models were analyzed: Model 1 was the crude model without covariate adjustment; Model 2 incorporated demographic and socioeconomic factors, including race, marital status, educational attainment, and PIR; Smoking and alcohol consumption variables were further introduced in Model 3; Model 4 additionally accounted for vigorous and moderate physical activity; and Model 5 was further adjusted for the number of pregnancies as a sensitivity analysis to account for reproductive exposure (pregnancy-frequency confounding). To further evaluate the independent impact of adjusting for the number of pregnancies, additional sensitivity analyses were conducted by extending Model 3 with adjustment for the number of pregnancies, but without including physical activity variables. The results are presented in Supplementary Tables S2 and S3. The variables in Models 1–4 were selected as sociodemographic and lifestyle confounders that were plausible common causes of both GLM7 and pregnancy loss. Histories of diabetes, hypertension, and hyperlipidemia were not included in the primary models, since they could represent clinical manifestations downstream of the metabolic state captured by GLM7 (potential mediators).
Since GLM7 is a composite index that incorporates age and BMI, these were not additionally adjusted in the regression models to avoid potential overadjustment and multicollinearity. Restricting the study population to women aged 18–35 years further minimized age-related confounding factors. GLM7 was analyzed continuously, and quartile analyses were performed. Associations are presented as odds ratios (ORs) with 95% confidence intervals (CIs). Furthermore, potential nonlinear associations between GLM7 and recurrent pregnancy loss were evaluated using restricted cubic splines (RCS) with three knots located at the 10th, 50th, and 90th percentiles. Finally, subgroup analyses were conducted for recurrent pregnancy loss by comparing participants with recurrent pregnancy loss to those without; participants with single pregnancy loss were excluded from the subgroup analyses. Statistical analyses were performed in R version 4.5.1, with two-sided P < 0.05 considered statistically significant.
Results
Participant characteristics
A total of 736 women aged between 18 and 35 years were included in the final analysis; 300 (37%) participants reported a history of pregnancy loss, and 436 (63%) did not. Baseline demographic and clinical features by pregnancy loss status are presented in Table 1. Compared to women without pregnancy loss, those with pregnancy loss were slightly older (29.49 ± 4.14 vs. 28.37 ± 4.33 years, P = 0.002) and had higher BMI (30.92 ± 8.11 vs. 29.00 ± 7.53 kg/m2, P = 0.009). Significant between-group differences were identified for race (P < 0.001), marital status (P = 0.028), drinking status (P = 0.041), age at last live birth (P = 0.017), number of pregnancies (P < 0.001), and number of live births (P = 0.009). Notably, women with pregnancy loss showed significantly higher GLM7 levels than those without (6.99 ± 0.66 vs. 6.86 ± 0.65, P = 0.038), whereas individual lipid components, such as TG, LDL-C, and HDL-C, showed no significant between-group differences. Supplementary Table S1 summarizes the baseline characteristics of participants included in and excluded from the analysis, showing that the two groups were broadly comparable across key demographic and clinical variables.
Table 1.
Baseline characteristics of participants
| Characteristic | Non-pregnancy loss, N = 436 (63%)1 | Pregnancy loss, N = 300 (37%)1 | P value2 |
|---|---|---|---|
| Demographic information | |||
| Age (years) | 28.373 (4.329) | 29.494 (4.138) | 0.002 |
| BMI (kg/m2) | 29.004 (7.534) | 30.923 (8.111) | 0.009 |
| Race | < 0.001 | ||
| Mexican American | 92 (14%) | 57 (15%) | |
| Non-Hispanic Black | 74 (11%) | 90 (22%) | |
| Non-Hispanic White | 176 (61%) | 89 (42%) | |
| Other | 94 (14%) | 64 (21%) | |
| Marital status | 0.028 | ||
| Married/Living with partners | 302 (73%) | 183 (64%) | |
| Widowed/Divorced/Separated/Never married | 134 (27%) | 117 (36%) | |
| Education level | 0.573 | ||
| Below high school | 91 (16%) | 67 (18%) | |
| High School or above | 345 (84%) | 233 (82%) | |
| PIR | 0.129 | ||
| Not poor | 297 (74%) | 189 (68%) | |
| Poor | 139 (26%) | 111 (32%) | |
| Lifestyle behaviors | |||
| Drinking status | 0.041 | ||
| Yes | 290 (73%) | 228 (80%) | |
| No | 146 (27%) | 72 (20%) | |
| Smoking status | 0.314 | ||
| Yes | 92 (22%) | 77 (25%) | |
| No | 344 (78%) | 223 (75%) | |
| Vigorous activity | 0.136 | ||
| Yes | 72 (15%) | 60 (20%) | |
| No | 364 (85%) | 240 (80%) | |
| Moderate activity | 0.682 | ||
| Yes | 184 (43%) | 135 (45%) | |
| No | 252 (57%) | 165 (55%) | |
| Pregnancy information | |||
| Age at first live birth (years) | 21.193 (4.061) | 21.245 (4.722) | 0.677 |
| Age at last live birth (years) | 24.895 (4.117) | 25.886 (4.361) | 0.017 |
| Age at menarche (years) | 12.565 (1.697) | 12.289 (1.629) | 0.051 |
| Number of pregnancies | 1.939 (1.048) | 3.900 (1.770) | < 0.001 |
| Number of live births | 1.939 (1.048) | 2.181 (1.185) | 0.009 |
| Clinical information | |||
| Diabetes history | 0.305 | ||
| Yes | 13 (2.5%) | 11 (4.1%) | |
| No | 423 (97%) | 289 (96%) | |
| Hypertension history | 0.074 | ||
| Yes | 47 (11%) | 46 (15%) | |
| No | 389 (89%) | 254 (85%) | |
| Hyperlipidemia history | 0.139 | ||
| Yes | 257 (56%) | 183 (62%) | |
| No | 179 (44%) | 117 (38%) | |
| Insulin (pmol/L) | 75.266 (60.837) | 82.855 (93.651) | 0.334 |
| FBG (mg/dL) | 94.495 (13.040) | 97.096 (15.643) | < 0.001 |
| TG (mmol/L) | 1.018 (0.612) | 1.082 (0.605) | 0.165 |
| HDL-C (mmol/L) | 1.387 (0.339) | 1.363 (0.325) | 0.572 |
| LDL-C (mmol/L) | 2.615 (0.765) | 2.683 (0.789) | 0.461 |
| GLM7 | 6.861 (0.647) | 6.988 (0.662) | 0.038 |
Abbreviations BMI Body mass index, PIR Poverty income ratio, FBG Fasting blood glucose, TG Triglycerides, HDL-C High-density lipoprotein cholesterol, LDL-C Low-density lipoprotein cholesterol, GLM7 Glycolipid metabolism 7 factors
1Mean (SD); n (unweighted) (%)
2Design-based KruskalWallis test; Pearson's X^2: Rao & Scott adjustment
Association between GLM7 and recurrent pregnancy loss (primary analysis)
No significant relationship between GLM7 and single pregnancy loss was observed across the survey-weighted multinomial logistic regression models. However, high GLM7 levels were associated with increased odds of recurrent pregnancy loss. In the crude model, higher GLM7 levels were associated with recurrent pregnancy loss (OR = 1.25, 95% CI: 1.03–1.50, P = 0.024). The association persisted after full adjustment, with each 1-SD increase in GLM7 corresponding to a 17% increase in the odds of recurrent pregnancy loss (OR = 1.17, 95% CI: 1.02–1.42, P = 0.046). When GLM7 was analyzed by quartiles, women in the second through fourth quartiles had significantly higher odds of recurrent pregnancy loss than those in the lowest quartiles. Detailed results are presented in Table 2. Although the continuous association reached statistical significance, the corresponding quartile-based trend test results in the fully adjusted model were not statistically significant. This discrepancy might reflect reduced statistical power due to the categorization of a continuous variable, the relatively limited sample size for recurrent pregnancy loss, and the absence of a strictly monotonic pattern across quartiles. Sensitivity analysis results, adjusting for the number of pregnancies without inclusion of physical activity variables, are presented in Supplementary Table S2 and were consistent with those of the main analyses.
Table 2.
Multinomial logistic regression of GLM7 in relation to single and recurrent pregnancy loss
| Exposure | Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| OR (95%CI) | P value | OR (95%CI) | P value | OR (95%CI) | P value | OR (95%CI) | P value | OR (95%CI) | P value | |
| Single pregnancy loss | ||||||||||
| GLM7 | 1.20 (0.96, 1.50) | 0.115 | 1.18 (0.93, 1.50) | 0.166 | 1.21 (0.95, 1.54) | 0.114 | 1.21 (0.95, 1.53) | 0.120 | 1.20 (0.94, 1.53) | 0.144 |
| Q1 | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | |||||
| Q2 | 1.08 (0.62, 1.89) | 0.786 | 0.99 (0.57, 1.74) | 0.984 | 1.06 (0.61, 1.84) | 0.844 | 1.03 (0.60, 1.78) | 0.902 | 1.03 (0.60, 1.77) | 0.919 |
| Q3 | 0.95 (0.53, 1.71) | 0.868 | 0.87 (0.49, 1.55) | 0.628 | 0.91 (0.51, 1.61) | 0.744 | 0.88 (0.50, 1.55) | 0.665 | 0.88 (0.50, 1.53) | 0.646 |
| Q4 | 1.71 (0.94, 3.12) | 0.081 | 1.63 (0.87, 3.05) | 0.129 | 1.73 (0.92, 3.24) | 0.087 | 1.69 (0.91, 3.12) | 0.096 | 1.67 (0.89, 3.14) | 0.113 |
| P for trend | 0.108 | 0.161 | 0.116 | 0.126 | 0.145 | |||||
| Recurrent pregnancy loss | ||||||||||
| GLM7 | 1.25 (1.03, 1.50) | 0.024 | 1.24 (1.02, 1.49) | 0.029 | 1.22 (1.01, 1.47) | 0.044 | 1.21 (1.01, 1.46) | 0.040 | 1.17 (1.02, 1.42) | 0.046 |
| Q1 | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | |||||
| Q2 | 2.57 (1.41, 4.68) | 0.002 | 2.22 (1.25, 3.94) | 0.006 | 2.35 (1.33, 4.12) | 0.003 | 2.39 (1.34, 4.26) | 0.003 | 2.32 (1.30, 4.15) | 0.005 |
| Q3 | 2.27 (1.24, 4.15) | 0.008 | 2.19 (1.20, 4.02) | 0.011 | 2.22 (1.21, 4.06) | 0.010 | 2.22 (1.21, 4.07) | 0.010 | 2.09 (1.14, 3.85) | 0.018 |
| Q4 | 2.14 (1.10, 4.14) | 0.024 | 2.02 (1.05, 3.90) | 0.037 | 1.96 (1.01, 3.79) | 0.046 | 1.97 (1.02, 3.78) | 0.043 | 1.81 (1.00, 3.65) | 0.048 |
| P for trend | 0.037 | 0.049 | 0.074 | 0.068 | 0.165 | |||||
Single pregnancy loss was defined as exactly one loss; recurrent pregnancy loss as ≥ 2 losses
Model 1, crude model
Model 2, adjusted for race, marital status, education level, and PIR
Model 3, adjusted for race, marital status, education level, PIR, smoking status, and drinking status
Model 4, adjusted for race, marital status, education level, PIR, smoking status, drinking status, vigorous activity, and moderate activity
Model 5, adjusted for race, marital status, education level, PIR, smoking status, drinking status, vigorous activity, moderate activity, and number of pregnancies
Abbreviations: CI Confidence Interval, OR Odds Ratio
Association between GLM7 and binary pregnancy loss (secondary analysis)
In secondary binary analyses, GLM7 was found to be positively associated with pregnancy loss in the crude model (OR = 1.21, 95% CI: 1.03–1.44, P = 0.026). The positive association persisted after adjustment for demographic, socioeconomic, lifestyle, and physical activity variables in Model 4 (OR = 1.21, 95% CI: 1.02–1.45, P = 0.036). Following additional adjustment for the number of pregnancies in Model 5, the association became weaker and did not retain statistical significance (OR = 1.19, 95% CI: 0.99–1.44, P = 0.074). indicating that pregnancy-frequency confounding partly accounted for the binary association. The results are summarized in Table 3. Results from sensitivity analyses are also provided in Supplementary Table S3.
Table 3.
The association between GLM7 and overall pregnancy loss: binary logistic regression models
| Exposure | Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| OR (95%CI) | P value | OR (95%CI) | P value | OR (95%CI) | P value | OR (95%CI) | P value | OR (95%CI) | P value | |
| GLM7 | 1.21 (1.03, 1.44) | 0.026 | 1.21 (1.01, 1.44) | 0.038 | 1.22 (1.02, 1.45) | 0.034 | 1.21 (1.02, 1.45) | 0.036 | 1.19 (0.99, 1.44) | 0.074 |
| Q1 | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | |||||
| Q2 | 1.52 (0.96, 2.39) | 0.073 | 1.35 (0.86, 2.13) | 0.187 | 1.43 (0.91, 2.24) | 0.117 | 1.43 (0.91, 2.12) | 0.117 | 1.40 (0.89, 2.18) | 0.141 |
| Q3 | 1.34 (0.86, 2.09) | 0.193 | 1.25 (0.80, 1.95) | 0.329 | 1.28 (0.82, 2.01) | 0.271 | 1.27 (0.82, 1.98) | 0.282 | 1.23 (0.80, 1.90) | 0.339 |
| Q4 | 1.83 (1.10, 3.06) | 0.021 | 1.76 (1.04, 2.97) | 0.036 | 1.80 (1.06, 3.05) | 0.031 | 1.79 (1.05, 3.04) | 0.033 | 1.71 (0.98, 2.99) | 0.059 |
| P for trend | 0.028 | 0.044 | 0.042 | 0.045 | 0.080 | |||||
Model 1, crude model
Model 2, adjusted for race, marital status, education level, and PIR
Model 3, adjusted for race, marital status, education level, PIR, smoking status, and drinking status
Model 4, adjusted for race, marital status, education level, PIR, smoking status, drinking status, vigorous activity, and moderate activity
Model 5, adjusted for race, marital status, education level, PIR, smoking status, drinking status, vigorous activity, moderate activity, and number of pregnancies
Abbreviations CI Confidence Interval, OR Odds Ratio
RCS analysis
Potential nonlinear association between GLM7 and recurrent pregnancy loss was evaluated using RCS (Fig. 2). The overall association reached statistical significance (P-overall = 0.046), whereas no significant nonlinearity was detected (P-nonlinear = 0.265). The results did not suggest a strong deviation from linearity across the observed range of GLM7.
Fig. 2.
RCS analysis of the association between GLM7 and recurrent pregnancy loss in reproductive-aged women. The solid line represents the adjusted odds ratio (OR) for recurrent pregnancy loss according to GLM7 levels and the shaded area indicates the 95% confidence interval (CI). The reference value was set at the median GLM7 level. A significant overall association was observed (P for overall association = 0.046), whereas no evidence of nonlinearity was detected (P = 0.265), suggesting an approximately linear relationship between GLM7 and recurrent pregnancy loss
Subgroup analyses
No significant effect modification was observed across subgroups stratified by race, marital status, smoking, or alcohol consumption (all P for interaction > 0.05; Fig. 3). Although a positive association was observed among married or cohabiting women (OR = 1.70, 95% CI: 1.13–2.56, P = 0.014), the interaction by marital status did not reach statistical significance (P for interaction = 0.055), and hence, the subgroup finding should be interpreted with caution.
Fig. 3.
Subgroup analyses of the association between GLM7 and recurrent pregnancy loss. Forest plots showing adjusted odds ratios (ORs) and 95% confidence intervals (CIs) across predefined subgroups including race, marital status, smoking status, and drinking status. No statistically significant interaction was observed between GLM7 and the subgroup variables (all P > 0.05), indicating that the association between GLM7 and recurrent pregnancy loss was generally consistent across subgroups. Subgroup analyses were restricted to participants without pregnancy loss and those with recurrent pregnancy loss (n = 567; 436 without pregnancy loss and 131 with recurrent pregnancy loss), excluding participants with single pregnancy loss (n = 169)
Discussion
In this cross-sectional study, higher GLM7 levels were more consistently associated with higher odds of recurrent pregnancy loss than with a binary history of pregnancy loss among women aged 18–35 years. After accounting for the number of pregnancies, the binary association with overall pregnancy loss attenuated, although the association with recurrent pregnancy loss remained significant. RCS analysis suggested a linear association between GLM7 and recurrent pregnancy loss. Subgroup analysis indicated no significant interaction across race, marital status, smoking status, or drinking status regarding the observed relationship between GLM7 and recurrent pregnancy loss. These findings suggested that an adverse composite glycolipid metabolic profile may be an important factor for recurrent pregnancy loss during early reproductive life.
Pregnancy loss is widely recognized as a complex reproductive outcome influenced by metabolic, hormonal, inflammatory, and social factors [12, 13]. Previous studies had primarily focused on individual metabolic indicators or environmental exposures, such as heavy metals [11], cardiovascular health [14], and inflammatory markers [15], in relation to pregnancy loss. Although these studies have provided important insights, they have often evaluated isolated biomarkers that may not fully capture the complex metabolic context underlying reproductive outcomes.
GLM7, a composite index integrating age, adiposity, glucose metabolism, insulin resistance, and lipid profiles, offers a more holistic assessment of glycolipid metabolic status [9]. In this study, although individual lipid components, such as TG, LDL-C, and HDL-C, showed no differences between women with and without pregnancy loss, GLM7 remained independently associated with recurrent pregnancy loss; its association with binary pregnancy loss was attenuated after adjustment for the number of pregnancies. The findings suggested that the cumulative metabolic burden captured by a composite index may be more informative than a single lipid or glucose parameter alone when evaluating reproductive risks. Composite metabolic indicators such as the triglyceride-glucose index and atherogenic index of plasma have shown predictive value in earlier studies involving cardiometabolic and reproductive outcomes [16–19].
Rather than representing an isolated reproductive event, pregnancy loss may reflect broader underlying metabolic susceptibility. Previous research had suggested that women who experience pregnancy loss may have an increased risk of future cardiometabolic diseases, including cardiovascular disease and type 2 diabetes [10, 20, 21]. The present findings aligned with those from previous studies, which suggested that adverse metabolic profiles may already be present during the early reproductive years, potentially contributing to pregnancy loss and later cardiometabolic risk. From a life-course perspective, impaired metabolic health before conception might influence implantation success, placental development, and early embryonic viability, thereby increasing susceptibility to pregnancy loss [3, 22]. Pregnancy loss may serve as an early clinical indicator of long-term cardiometabolic risk. Therefore, integrating reproductive history with metabolic profiling might provide valuable insights into early risk identification and prevention strategies.
Several biological mechanisms may explain the observed association between elevated GLM7 levels and pregnancy loss; however, these could not be directly evaluated in the present study. Insulin resistance and impaired glucose metabolism, the core components of GLM7, have been implicated in adverse pregnancy outcomes through altered endometrial receptivity, placental dysfunction, and impaired trophoblast invasion [23]. Hyperinsulinemia and chronic low-grade inflammation may also disrupt the delicate hormonal and immunological balance required for successful implantation and early placental development [24, 25]. In addition, dyslipidemia and excess adiposity may contribute to oxidative stress and endothelial dysfunction [26, 27], which can compromise uteroplacental blood flow and increase the susceptibility to early pregnancy losses. These pathways are still speculative in the context of the data in this study, but provide biologically plausible explanations for the observed association.
Notably, the association between GLM7 and pregnancy loss appears to be driven primarily by recurrent rather than single pregnancy loss. This pattern is biologically plausible, since an isolated pregnancy loss is frequently attributed to sporadic embryonic chromosomal abnormalities, whereas recurrent loss is more likely to reflect persistent maternal factors [12]. Although these pathways could not be directly evaluated in the present cross-sectional study, it is biologically plausible that metabolic dysregulation, including insulin resistance, chronic inflammation, endothelial dysfunction, and impaired endometrial receptivity, exerts sustained adverse effects on implantation and early placental development, thereby increasing susceptibility to repeated reproductive failure [4, 28]. The findings of this study suggested that cumulative metabolic burden may be more relevant to repeated reproductive failure than to isolated pregnancy loss events. Composite indices, such as GLM7, may better capture persistent systemic influences than biomarkers alone. The attenuation of the binary association after adjusting for the number of pregnancies highlights an important methodological issue. Women with more pregnancies had a greater chance of experiencing at least one pregnancy loss. Therefore, recurrent pregnancy loss may represent a more informative epidemiological phenotype than a binary history. Nevertheless, recurrent pregnancy loss in the NHANES remains based on the same broad non-live birth definition and should not be interpreted as a clinically confirmed recurrent miscarriage.
Strengths and limitations
Several notable strengths should be acknowledged in the present study. The analysis was based on nationally representative NHANES data and incorporated standardized measurements of metabolic biomarkers and reproductive history. In addition, survey-weighted analytical methods were applied to account for the complex sampling structure. Another important feature was the restriction of the study population to women aged 18–35 years, which is a critical reproductive window in which pregnancy loss is less likely to be influenced by age-related reproductive decline. This age-focused approach may reduce recall bias, minimize the influence of comorbid conditions that are more prevalent at older ages, and enhance the relevance of the findings of this study to preconception health. Additionally, GLM7 is derived entirely from routine clinical measurements commonly obtained during health examinations, making it a low-cost and easily scalable marker. From a public health perspective, these findings suggested that composite metabolic indices, such as GLM7, may be useful for generating hypotheses in studies of reproductive and metabolic health, although their clinical utility requires validation in prospective cohorts. Although the findings did not establish causality, they highlighted the importance of comprehensive metabolic health assessments beyond single risk factors in pregnant women. However, several limitations should also be considered when interpreting the findings. Most importantly, the cross-sectional nature of the study prevents conclusions regarding causality. GLM7 was measured during the survey examination, whereas pregnancy losses may have occurred years earlier, and reverse causality could not be excluded, since pregnancy itself may alter metabolic parameters such that GLM7 partly reflects post-pregnancy metabolic changes. Second, pregnancy loss was defined as pregnancies not resulting in a live birth in which spontaneous miscarriage could not be distinguished from induced abortion, ectopic pregnancy, stillbirth, or other heterogeneous outcomes. In the United States, induced abortion rates vary systematically according to race, income, and marital status, factors correlated with the GLM7, which may introduce misclassification. If induced abortion was unrelated to GLM7, non-differential misclassification would attenuate associations toward the null, suggesting that the true association with spontaneous pregnancy loss may be at least as strong. If abortion rates co-vary with GLM7-correlated sociodemographic factors, differential misclassification could bias the estimates in either direction. NHANES does not record the gestational timing of loss, thus preventing early versus late loss stratification. This limitation may have introduced an outcome misclassification, although the approach has been widely adopted and validated in previous NHANES studies [11]. Third, the complete case exclusion of participants with missing GLM7-related data may have introduced selection bias. Fourth, unmeasured factors, including genetic susceptibility, dietary habits, psychosocial stress, and inflammatory markers, may still have influenced the results.
Conclusions
In conclusion, higher GLM7 levels were associated with higher odds of recurrent pregnancy loss among American women aged 18–35 years in this cross-sectional analysis; however, the association with history of binary pregnancy loss was attenuated after accounting for reproductive history. The findings suggested that cumulative metabolic burden may be more relevant to recurrent pregnancy loss than to isolated pregnancy loss events. From a clinical perspective, comprehensive metabolic profiling before conception may help identify women at elevated risk of repeated pregnancy loss and support early lifestyle or metabolic interventions in reproductive healthcare settings. Given that GLM7 is derived from routinely available clinical indicators, the index may have potential utility as an accessible metabolic risk stratification tool for future reproductive medicine research and preconception care strategies. Further prospective studies with clinically validated outcome definitions would be required to confirm causality and determine whether improving glycolipid metabolic status before conception could reduce the odds of recurrent pregnancy loss.
Supplementary Information
Acknowledgements
The authors sincerely acknowledge the NHANES program for establishing and maintaining the publicly accessible database, which provided invaluable data for the present study.
Abbreviations
- GLM7
Glycolipid metabolism 7 factors
- NHANES
National Health and Nutrition Examination Survey
- BMI
Body mass index
- TG
Triglyceride
- LDL-C
Low-density lipoprotein cholesterol
- HDL-C
High-density lipoprotein cholesterol
- RCS
Restricted cubic spline
- FBG
Fasting blood glucose
- OR
Odds ratio
- CI
Confidence interval
Authors’ contributions
Z.J.Z. contributed to the conceptualization of the study, performed the data analysis and validation, and drafted the original manuscript. R.A. and L.X. contributed to data curation, methodology support, and manuscript revision. Z.H.N. contributed to the conceptualization of the study, supervised the research, and critically reviewed and edited the manuscript. All authors have read and approved the final manuscript.
Funding
None.
Data availability
The datasets used and evaluated in this study are available from the corresponding author upon reasonable request.
Declarations
Ethics approval and consent to participate
The NHANES study protocols were approved by the National Center for Health Statistics Research Ethics Review Board, and written informed consent was obtained from all participants. All procedures were conducted in accordance with the principles of the Declaration of Helsinki.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Duffy JMN, Bhattacharya S, Bhattacharya S, Bofill M, Collura B, Curtis C, et al. Standardizing definitions and reporting guidelines for the infertility core outcome set: an international consensus development study. Fertil Steril. 2021;115(1):201–12. [DOI] [PubMed] [Google Scholar]
- 2.Mehra VM, Farooqi S, Sriram P, Tunde-Byass M. Diagnosis and management of early pregnancy loss. CMAJ. 2024;196(34):E1162–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Boxem AJ, Blaauwendraad SM, Mulders A, Bekkers EL, Kruithof CJ, Steegers EAP, et al. Preconception and early-pregnancy body mass index in women and men, time to pregnancy, and risk of miscarriage. JAMA Netw Open. 2024;7(9):e2436157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Cai WY, Luo X, Lv HY, Fu KY, Xu J. Insulin resistance in women with recurrent miscarriage: a systematic review and meta-analysis. BMC Pregnancy Childbirth. 2022;22(1):916. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Li S, Ma S, Yao X, Liu P. Effects of Metabolic Syndrome on Pregnancy Outcomes in Women Without Polycystic Ovary Syndrome. J Endocr Soc. 2024;8(10):bvae143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Song Z, Fang F, Liu Z, Liang R, Pan Y, Meng W, et al. Elevated triglyceride glucose-body mass index increases risk of miscarriage in women undergoing in vitro fertilization and embryo transfer. Fertil Steril. 2026;125(3):466–76. [DOI] [PubMed]
- 7.Meng Z, Lin M, Song L, Chen Y, Deng S, Xia S, et al. The first-trimester triglyceride glucose-body mass index is a valuable predictor for adverse pregnancy outcomes. BMC Pregnancy Childbirth. 2025;25(1):142. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Guo Y, Lu J, Bahani M, Ding G, Wang L, Zhang Y, et al. Triglyceride-glucose index in early pregnancy predicts the risk of gestational diabetes: a prospective cohort study. Lipids Health Dis. 2024;23(1):87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Wang Z, Chen S, Feng X, Chen X, Evans PC, Strijdom H, et al. GLM7 - A Novel Composite Glycolipid Index Derived from Routine Health Indicators for Enhanced Diagnosis and Prediction of Multimorbidity. Adv Sci (Weinh). 2025;12(42):e10552. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Muehlschlegel C, Kyriacou H, Al-Mohammad A, Foster-Davies LA, Simmons-Jones F, Oliver-Williams C. The risk of cardiovascular disease in women after miscarriage, stillbirth, and therapeutic abortion: a protocol for a systematic review and meta-analysis. Syst Rev. 2020;9(1):234. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Yu C, You Q, Bai X, Mu F. Association between heavy metal exposure and pregnancy loss: evidence from NHANES 2011-2016. Reprod Biol Endocrinol. 2025;23(1):87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Dimitriadis E, Menkhorst E, Saito S, Kutteh WH, Brosens JJ. Recurrent pregnancy loss. Nat Rev Dis Primers. 2020;6(1):98. [DOI] [PubMed] [Google Scholar]
- 13.Quenby S, Gallos ID, Dhillon-Smith RK, Podesek M, Stephenson MD, Fisher J, et al. Miscarriage matters: the epidemiological, physical, psychological, and economic costs of early pregnancy loss. Lancet. 2021;397(10285):1658–67. [DOI] [PubMed] [Google Scholar]
- 14.Yan Y, Chen J, Qin J, Yu M, Du M. Association of cardiovascular health with reproductive lifespan and pregnancy loss: insights from NHANES 2005-2018. Front Endocrinol (Lausanne). 2025;16:1597097. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yang S, Shi H, Ren S, Chen Y, Jin X, Han L, et al. Association of inflammatory markers with pregnancy loss: global burden (1990-2021) and NHANES survey (2005-2016). BMC Pregnancy Childbirth. 2025;25(1):996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.He X, He X, Zhou Y, Liu J, Wang G. Associations between triglyceride-glucose index in the early trimester of pregnancy and adverse pregnancy outcomes. BMC Pregnancy Childbirth. 2025;25(1):986. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mei Y, Li W, Wang M. Association of triglyceride-glucose index and platelet-to-lymphocyte ratio with adverse pregnancy outcomes in women with polycystic ovary syndrome: a retrospective cohort study. Ann N Y Acad Sci. 2025;1551(1):189–200. [DOI] [PubMed] [Google Scholar]
- 18.Zhu H, Gong L, Yan H. Threshold effects of atherogenic index of plasma on female infertility risk: evidence from NHANES 2013-2018. BMC Womens Health. 2025;25(1):564. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Bao Z, Zhang Y, Zhou J, Dai Z. Association between atherogenic index of plasma and infertility: a cross-sectional study based on U.S. women. Lipids Health Dis. 2025;24(1):51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Wang YX, Mínguez-Alarcón L, Gaskins AJ, Wang L, Ding M, Missmer SA, et al. Pregnancy loss and risk of cardiovascular disease: the Nurses’ Health Study II. Eur Heart J. 2022;43(3):190–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Khan SS, Petito LC, Huang X, Harrington K, McNeil RB, Bello NA, et al. Body Mass Index, Adverse Pregnancy Outcomes, and Cardiovascular Disease Risk. Circ Res. 2023;133(9):725–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Loy SL, Chan DWK, Ku CW, Cheung YB, Godfrey KM, Tan KML, et al. Metabolic health status and fecundability in a Singapore preconception cohort study. Am J Obstet Gynecol. 2022;226(5):714.e1-.e16. [DOI] [PMC free article] [PubMed]
- 23.Lei R, Chen S, Li W. Advances in the study of the correlation between insulin resistance and infertility. Front Endocrinol (Lausanne). 2024;15:1288326. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Ibrahim A, Engku Ismail EH, IrwanKhoo M, Yusuf L, Nik Hussain NH, Mat Zin AA, et al. Impaired implantation as a major upstream pathway of preeclampsia: a narrative synthesis of mechanistic, epidemiological and biomarker evidence. Front Reprod Health. 2025;7:1743504. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Menzies FM. The placenta as an immunological environment. Br J Biomed Sci. 2025;82:14910. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Echeverria C, Eltit F, Santibanez JF, Gatica S, Cabello-Verrugio C, Simon F. Endothelial dysfunction in pregnancy metabolic disorders. Biochim Biophys Acta Mol Basis Dis. 2020;1866(2):165414. [DOI] [PubMed] [Google Scholar]
- 27.Reijnders D, Olson KN, Liu CC, Beckers KF, Ghosh S, Redman LM, et al. Dyslipidemia and the role of adipose tissue in early pregnancy in the BPH/5 mouse model for preeclampsia. Am J Physiol Regul Integr Comp Physiol. 2019;317(1):R49-r58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Amrane S, McConnell R. Endocrine causes of recurrent pregnancy loss. Semin Perinatol. 2019;43(2):80–3. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used and evaluated in this study are available from the corresponding author upon reasonable request.