Metabolomic and proteomic signatures of gestational diabetes predicting long-term risk of type 2 diabetes: a prospective cohort study in the UK Biobank

Metabolomic and proteomic signatures of gestational diabetes predicting long-term risk of type 2 diabetes: a prospective cohort study in the UK Biobank | Research Square Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Metabolomic and proteomic signatures of gestational diabetes predicting long-term risk of type 2 diabetes: a prospective cohort study in the UK Biobank Alvaro Torres-Martos, Moscho Michalopoulou, Lucas Jurado-Fasoli, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8700326/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Women who develop gestational diabetes mellitus (GDM) during pregnancy have up to a 20-fold increased risk of progressing to type 2 diabetes mellitus (T2DM) later in life. Despite this elevated risk, efforts to accurately predict which women with GDM will develop T2DM remain largely unsuccessful. Methods Using data from women in the UK Biobank, comprehensive metabolomic (n = 197,263; 685 with GDM) and proteomic (n = 17,895; 62 with GDM) profiling was performed. We derived GDM-associated metabolomic and proteomic signatures using metabolomic and proteomic features as predictors and GDM as a binary outcome in generalized linear models with elastic net regularization. In prospective analyses, we estimated hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between the GDM-associated metabolomic and proteomic signatures and incident T2DM. Results Distinct metabolomic signatures associated with prior GDM were identified, characterized by higher levels of alanine, glucose–lactate, lactate, isoleucine, and leucine, and lower levels of creatinine, glycine, glutamine, citrate, 3-hydroxybutyrate and docosahexaenoic acid. Proteomic profiling revealed lower concentrations of proteins involved in cholesterol metabolism, integrin and cytokine signaling, and higher concentrations of proteins involved in inositol phosphate metabolism and phosphatidylinositol signaling. Higher scores for the GDM-associated metabolomic and proteomic signatures were significantly associated with incident T2DM (HR per SD 1.82, 95% CI 1.73–1.91; and 1.27, 95% CI 1.10–1.47, respectively). Conclusions We identified distinct molecular signatures of GDM that were detectable years after pregnancy and predictive of T2DM risk. These findings may improve long-term risk stratification and inform early prevention strategies for T2DM. Gestational diabetes mellitus Metabolomics Proteomics Type 2 diabetes mellitus Risk prediction Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Gestational diabetes mellitus (GDM) is defined as hyperglycaemia which first develops or is first diagnosed during pregnancy 1 . It represents one of the most prevalent pregnancy complications, affecting > 17 million pregnancies per year globally 2 . GDM increases the risk of both maternal and neonatal medical complications 3 , although early identification and glycaemic management reduce these risks to those similar to the general maternity population 4 , 5 . However, the long-term risks associated with GDM still persist. Women with GDM have up to 20 times higher risk of developing future type 2 diabetes mellitus (T2DM) compared to women without GDM, with up to 50% progressing to T2DM within 5 years after GDM diagnosis 6 – 13 . Despite being an established risk factor for subsequent T2DM, not all women with GDM present the same risk. Systematic reviews have reported substantial heterogeneity in risk estimates for development of T2DM after a pregnancy affected by GDM, but evidence exploring potential individual factors that could explain it is limited 12 , 14 . Identifying women with GDM at the highest risk of future T2DM would help target and prioritise women for T2DM screening as well as interventions to prevent T2DM. Several T2DM risk prediction models have been developed which integrate conventional risk factors such as age, body mass index (BMI), waist circumference, hypertension, family history of diabetes, and fasting glucose 15 . However, most models are at high risk of bias, have limited accuracy and are rarely implemented in clinical practice 15 . Metabolomics and proteomics have emerged as useful approaches for phenotyping and risk stratification of hyperglycaemia, offering the opportunity to identify molecular signatures explaining postpartum progression to T2DM after GDM 16 . Previous cohort studies have identified a limited number of molecules associated with the transition from GDM to T2DM 17 – 28 , most of which are limited by relatively short periods of follow-up of 2 to about 8 years from a pregnancy affected by GDM until development of T2DM. Only few investigations have followed-up larger cohorts of women with a history of GDM for a decade or longer 24 , 26 . In addition, previous studies measured biomarkers during pregnancy or in the early postpartum period. It is plausible that levels of these biomarkers could have been affected by metabolic changes that happened during or shortly after pregnancy 29 – 33 . Profiling metabolites or proteins very close to pregnancy may limit the assessment of metabolomic and proteomic profiles as true mediators in the association of GDM with later T2DM. Using data from the UK Biobank cohort, we derived GDM-associated metabolomic and proteomic signatures and investigate if there was an association between the derived signatures and the risk of developing T2DM. Additionally, we investigated whether these omic signatures mediated the relationship between GDM and the development of future T2DM. Methods Ethics and governance The UK Biobank study was conducted according to the Declaration of Helsinki, and ethical approval was granted by the Northwest Multi-centre Research Ethics Committee (reference number 06/MRE08/65). All participants gave informed consent to participate and be followed up through data linkage to routine electronic healthcare records. Data sources We used data from the UK Biobank study, a large-scale population-based prospective cohort that recruited 502,128 participants aged 40-69 years across the UK (England, Scotland and Wales). During a baseline visit (between 2006 and 2010), participants were asked to report information on sociodemographic factors, medical history and health behaviours through self-completed touchscreen questionnaires. As part of this visit, participants underwent a physical examination and a verbal interview with a nurse who collected relevant medical history, blood pressure, height and weight measures as well as blood, saliva and urine samples. Data from the UK Biobank have been linked with the NHS Hospital Episode Statistics (HES) 34 and the death registries from the Office of National Statistics (ONS). Further details can be found in the latest available protocol (https://www.ukbiobank.ac.uk/wp-content/uploads/2025/01/Main-study-protocol.pdf). Population We included women who had a record of at least one birth, determined by a self-reported touchscreen questionnaire or a delivery episode in the HES maternity data. Women were excluded if they had pre-pregnancy type 2 diabetes mellitus (T2DM), prevalent T2DM, or cardiovascular disease (CVD) at the time of the baseline assessment; GDM diagnosed at the baseline assessment; a reported pregnancy at age 10; no recorded history of childbirth or missing information on having at least one delivery; or missing data on mother’s age at first birth. Exposure The exposures of interest comprised the metabolomic and proteomic signatures associated with GDM. GDM status was determined as follows: women with a self-reported diabetes occurring only during pregnancy via touchscreen questionnaire; GDM reported during the nurse interview; a diagnosis code for GDM, defined as the presence of ICD-10 codes O24.4 (diabetes mellitus arising in pregnancy) or O24.9 (unspecified diabetes in pregnancy) recorded in the linked HES data for the delivery episode. Metabolomic and proteomic data profiles We used data from metabolomic and proteomic determinations carried out on the blood samples collected from the UK Biobank participants at baseline (between 2006-2010). Plasma was extracted from non-fasting blood samples collected in EDTA blood collection tubes and frozen (-80ºC) in aliquots for later analysis. A sample of 275,000 plasma samples were selected for metabolomic analysis using high-throughput nuclear magnetic resonance (NMR) spectroscopy by Nightingale Health Plc (Finland); and a random sample of 53,058 samples were selected for proteomic analysis by Olink Analysis Service (Sweden). The process of sample selection ensured the samples were representative of the overall UK Biobank population. The technical details of metabolomic and proteomic profiling are described elsewhere 35,36 (https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/NMR_companion_phase2.pdf; https://biobank.ctsu.ox.ac.uk/ukb/ukb/docs/PPP_Phase_1_QC_dataset_companion_doc.pdf). The NMR platform quantifies 251 metabolites from each sample in a single experimental assay. The outcome is measures of 169 metabolites in absolute levels (expressed in molar concentration units) and 82 ratio measures. The metabolites in the profile include those routinely used in clinical practice (e.g. cholesterol) in addition to several emerging biomarkers increasingly measured in research studies, such as lipoproteins in 14 subclasses, fatty acids, amino acids, ketone bodies and glycolysis-related metabolites. We excluded 5,794 participants with a percentage of missing values >20%. The remaining missing values were imputed using a non-parametric method known as missRanger package 37 . Proteomic data was obtained from the UK Biobank Pharma Proteomics Project. The antibody-based Olink ExploreTM Proximity Extension Assay, in combination with Next-Generation Sequencing were used to determine the relative abundance of 2923 unique proteins, expressed as normalized protein expression values (log2-transformed) 36 . We excluded three proteins where the proportion of missing values was >50%, a total of 2920 proteins were included in our proteomic analyses. We excluded 4,259 participants with a percentage of missing values >30%. The remaining missing values were singularly imputed using missRanger package 37 . All the metabolites and proteins were scaled as z-scores. Outcome Eligible women were followed-up from enrolment into the UK Biobank until the outcome of interest or the end of follow-up.The primary outcome of interest was the earliest recorded date of a T2DM diagnosis, ascertained from the presence of ICD-10 code for non-insulin dependent diabetes (E11.0-E11.9), malnutrition-related diabetes (E12.0-E12.9), other specified diabetes (E13.0-E13.9) or unspecified diabetes (E14.0-E14.9) in the linked HES data (latest available follow-up data was 31 October 2022 for England, 31 May 2022 for Wales, and 31 August 2022 for Scotland). Covariate definition Covariates were defined consistently with our previous approach which used a directed acyclic graph 6 and included: age at first birth (years); ethnicity grouped as White, Black, Asian, Other; Townsend index of deprivation, a composite measure of deprivation based on unemployment, non-car ownership, non-home ownership and household overcrowding, split by quintiles 1-5 (higher quintiles indicating more deprivation); education grouped as higher degree, any school degree, vocational qualification, unknown; family history of diabetes grouped as yes, no, unknown; body mass index (BMI) categorised as <18.5, 18.5-24.9, 25-29.9, 30-34.9, ≥35 kg/m 2 ; use of cholesterol-lowering medication; smoking status grouped as never, current, previous, unknown; alcohol consumption grouped as none, occasional, moderate, heavy, unknown; physical activity classified as meeting recommendations (≥150 min of moderate-intensity activity and/or ≥75 min of vigorous activity per week), not meeting recommendations, unknown; fruit and vegetable consumption classified as meeting five portions a day, not meeting five portions a day, unknown; saturated fat intake classified in tertiles (low, moderate, high, unknown); menopause grouped as yes, no, unknown; and self-reported number of live births as a proxy for parity. Participating women were enrolled into the UK Biobank several years after their GDM diagnosis, hence several covariates (e.g. BMI and lifestyle factors) and blood samples were collected during the baseline assessment. Those covariates were included in the analysis because of their potential influence on the exposure and outcome of interest and to help remove some of the residual confounding 38–41 . Table S1 (Additional file 1) includesfurther details on the derivation and classification of covariates. Statistical and bioinformatic analysis Participant characteristics are reported descriptively. Categorical variables are reported as total numbers and percentages and continuous variables are reported as means and standard deviations. To compare categorical and continuous variables, Pearson's Chi-square and Wilcoxon rank sum tests were performed, respectively. For metabolomic and proteomic analyses, we first performed retrospective analyses separately on each metabolomic and proteomic dataset, applying two different multivariable approaches. Limma (linear models for microarray data) was used to assess the differential abundance of individual features between women with GDM and those without GDM. Nominal p-values were corrected for multiple testing with the false discovery rate (FDR) according to the Benjamin-Hochberg procedure 42 . To capture the combined effects and intrinsic relationships between statistically significant metabolites and proteins, we derived GDM-associated metabolomic and proteomic signatures. Firstly, we fitted a logistic regularized regression model with metabolomic and proteomic features as predictors and GDM as a binary outcome. This generalized linear model with elasticnet regularization, which combines L1 and L2 penalties, requires a statistical framework to find the hyperparameters that optimize the feature selection and the predictive performance. Our study population presented a severe class imbalance, so the trained models can be prone to overfit the majority class. For this reason, a repeated k-fold cross validation (5 folds with 5 repeats) with down-sampling only in the training folds was used to estimate the hyperparameters with the highest values of G-mean 43 . Secondly, we fitted the final model with the hyperparameters using the whole dataset of women. The scaled probabilistic prediction of the final model was called GDM-associated signatures, expressed as z-scores ( Additional file 1: Table S2 ). In both multivariable analyses, we used a basic adjustment including maternal age at first birth and ethnicity. For these analyses, we used the limma 44 , caret and glmnet 45 packages. After, enrichment analyses of statistically significant metabolites and proteins were conducted using the fgsea 46 and clusterProfiler 47 packages in R software (version 4.2.2). In prospective analyses, we estimated hazard ratios (HR) with 95% CIs for the GDM-associated metabolomic and proteomic signatures and incident T2DM using multivariable Cox proportional hazard models with age (years) as the underlying time covariate. These analyses were performed both in women with GDM as well as in the total population of parous women (with and without a GDM diagnosis). Person-time of follow-up was calculated from the age at the baseline assessment visit, until the age at which participants were censored, using the date when the outcome occurred. Models were further adjusted for age at first birth, ethnicity, socio-economic status (Townsend index and education), medical conditions (family history of diabetes, BMI group and cholesterol lowering medication), menopausal status, number of live births and health behaviours (smoking, physical activity, alcohol, dietary scores). The proportional hazards assumption using Schoenfeld residuals' plots for the exposures was not violated. Events that happened within one year of follow-up from the UK Biobank baseline assessment centre were excluded from the analysis to reduce reverse causality. For this analysis, we used the survival package in R. Sensitivity analyses employing a waffle plot illustrated the proportion of women with and without GDM, across quintiles of the metabolomic signature, who later developed or did not develop T2DM. A causal mediation analysis based on a two-model approach was performed using a mediation and outcome model. For the mediation model, we used a linear model (mediation model) with GDM (past) as a predictor and GDM-associated omic signatures as an outcome. The mediation model was minimally adjusted for maternal age at first birth and ethnicity. For the outcome model, we used a logistic model with GDM (past) and GDM-associated omic signatures as predictors and T2DM as the outcome. In this case, demographics, socio-economic status, medical conditions, health behaviours and dietary intake variables were included as covariates. We estimated the average causal mediation effects (indirect effect), average direct effects, proportions mediated, and total effect using the mediation package in R. Results Participant characteristics After exclusions, the eligible study sample comprised parous women (n = 203,057), of whom n = 706 had a history of GDM. Of these, 197,263 women were included in the metabolomics analyses and 17,895 in the proteomics analyses ( Fig. 1 ). Overall, women with GDM were older and had a higher proportion of non-White ethnicity, a lower prevalence of healthy weight, and a greater prevalence of a family history of diabetes. Identification of individual metabolites and proteins associated with GDM Analysis of baseline metabolites yielded 182 metabolites that were associated with GDM after FDR correction ( Fig. 2A ). Overall, women with a history of GDM had higher concentrations of large VLDL particles and lower concentrations of small HDL and LDL than women with no GDM history. Additionally, GDM was associated with lower total concentrations of cholines, phosphocholines, phosphoglycerides, and sphingomyelins. There was also a lower degree of unsaturation of fatty acids. Several amino acids and metabolites related to glycolysis and inflammation showed differential associations, including lower levels of glycine and glutamine and higher levels of circulating glucose-lactate, alanine, total BCAAs and GlycA in women with GDM ( Fig. 2A ). The most significant differences in the enrichment set were observed within the lipidome, which revealed lower percentages of total lipids, cholesterol, cholesterol esters, and free cholesterol, but a higher triglyceride content across most lipoprotein subclasses in women with GDM ( Fig. 2B ). The assessment of baseline proteomics after FDR correction, identified 35 proteins associated with GDM ( Fig. 3A ). These proteins were involved in pathways such as cytokine-cytokine receptor interaction (EDA2R and GDF15), cAMP signaling pathway (CGA, FSHB and NPR1), hormone signaling (CGA, FSHB and NPR1), cytoskeleton in muscle cells (ELN and NEB), and regulation of Insulin-like Growth Factor (IGF) transport and uptake by Insulin-like Growth Factor Binding Proteins (IGFBPs) (IGFBP1 and IGFBP2). The gene-set enrichment analysis revealed differential protein levels activating inositol phosphate metabolism and phosphatidilinositol signaling and suppressing cholesterol metabolism, integrin signaling, cytokine-cytokine receptor interaction and cytoskeleton in muscle cells ( Fig. 3B ). Derivation of GDM-associated omic signatures To investigate the combined effects and intrinsic relationships between metabolomic and proteomic features associated with a history of GDM, we constructed metabolomic and proteomic signatures based on statistically significant metabolites and proteins, applying downweighting to reduce redundancy. The resulting signatures contained 32 metabolomic ( Fig. 2C ) and 18 proteomic features ( Fig. 3C ). The GDM-associated metabolomic signature was characterized by higher levels of alanine, glucose–lactate, isoleucine and leucine, and lower levels of creatinine, glycine, glutamine, citrate, and 3-hydroxybutyrate. In addition, this signature highlighted an imbalance between omega-6 and omega-3 fatty acids ( Fig. 2C ). In addition, the GDM-associated proteomic signature was defined by increased abundance of proteins related to cytokine-cytokine receptor interaction (GDF15 and EDA2R) and regulation of IGF and transport and uptake by IGFBPs (IGFBP1), and decreased abundance of proteins involved in cytoskeleton in muscle cells (NEB) and hormone/cAMP signaling pathway (NPR1) ( Fig. 3C ). Associations with incident T2DM and mediation analysis Among women with GDM and available metabolomic data (n = 685), 77 cases of T2DM were observed over an average follow-up of 13.5 years (median 13.7) from the baseline assessment. In this smaller sample, continuous scores of the GDM-associated metabolomic signature were significantly associated with incident T2DM (HR per SD 1.58, 1.13-2.19) ( Additional file 1: Table S3 ). In the total sample of parous women with available metabolomic data (n = 197,263 with and without GDM), 6,268 cases of T2DM were observed over an average follow-up of 13.5 years (median 13.7) of follow up. Continuous scores of the GDM-associated metabolomic signature in this sample showed a strong linear association with T2DM risk (HR per SD 1.82, 95% CI 1.73-1.91), with the women in the highest quintile showing a four-fold higher risk compared to those in the lowest quintile (HR in Q5 4.53, 3.92-5.24) ( Fig. 4A ). Stepwise models of the GDM-associated signatures are presented in Table S4 ( Additional file 1 ). In sensitivity analyses, nearly 80% of women with GDM who subsequently developed T2DM were classified within Q4–Q5 of the metabolomic signature score ( Additional file 1: Fig S1 ). Among women with GDM and available proteomic data (n = 62), six incident cases of T2DM occurred over a mean follow-up of 13.6 years (median 13.7) and no associations were estimated in this sample. In the total sample of parous women (n = 17,895 with and without GDM) with proteomic data, 607 cases of T2DM were observed over an average follow up of 13.5 years (median 13.7). Continuous scores of the proteomic signature were significantly associated with incident T2DM (HR per SD 1.27, 1.10-1.47) ( Fig. 4B ). The GDM-associated proteomic signature showed clear discrimination between women with and without a history of GDM, but limited ability to further stratify T2DM risk among women with GDM ( Additional file 1: Fig S2 ). We used a mediation model to quantify the associations between GDM (past) and omic-signatures (baseline), followed by an outcome model for GDM (past) and omic-signatures (baseline) as predictors and T2DM as outcome (follow-up). There was a strong direct association between prior GDM and incident T2DM, of which 17% (95% CI 13–22%) was mediated by the baseline metabolomic signature, with the remaining effect operating independently of the signature ( Fig. 4C ). In contrast, mediation through the GDM-associated proteomic signature did not reach statistical significance ( Fig. 4D ). Discussion This study describes novel metabolomic and proteomic signatures associated with a prior diagnosis of GDM in a large sample of parous women from the UK Biobank. Our findings provide large-scale molecular evidence that GDM may leave a durable imprint on the plasma metabolome and proteome, detectable years after pregnancy and before the onset of T2DM. We further show that these metabolomic and proteomic signatures predict future progression to T2DM and are shared by women with and without a history of GDM. Notably, the metabolomic – but not the proteomic – signature explained approximately one fifth of the association between GDM and T2DM, suggesting that persistent metabolic alterations may be key drivers of disease progression. Unlike previous studies 17 – 26 , our work shows that women with prior GDM exhibit a persistent metabolic signature years later, while still free of T2DM, that is predictive of future disease risk. In this study, women with history of GDM displayed a comprehensive metabolomic profile marked by high BCAAs, impaired lipid metabolism (triglycerides, phosphatidylcholines, cholesterol esters, and reduced degree of unsaturation), disrupted glucose metabolism (glucose and lactate) and inflammatory markers (high omega-6/omega-3 ratio, reduced degree of unsaturation and GlycA). This expands the evidence from earlier studies, which also observed elevated levels of BCAAs, which are strongly implicated in insulin resistance 48 , pancreatic β-cell dysfunction 49 and T2DM onset 50 . Clinical trials have further shown that BCAAs concentrations can increase up to seven years before T2DM onset, underscoring their predictive value and potential as preventive targets 51 , 52 , as they can be reduced through bariatric surgery 53 , Mediterranean diet 54 and lifestyle interventions 55 . We also replicate findings from studies exploring metabolomic profiles in women with a history of GDM in the early post-partum period that report higher levels of triglycerides and lower levels of sphingomyelins, phosphatidylcholines, and cholesterol esters, supporting the notion that lipid perturbations emerge soon after pregnancy and may persist in the long-term 18 , 23 . A prospective study of healthy women (N = 26,836) also reported a similar lipoprotein profile, with higher concentrations of large VLDL particles and lower concentrations of small HDL and LDL, associated with higher T2DM risk 56 . In addition, several studies have reported a positive association between BCAAs levels and triglycerides, suggesting a potential synergy in T2DM development 57 , 58 . Our analysis also highlights altered levels of non-essential amino acids, including lower glycine, previously associated with higher T2DM risk 59 , but potentially rectified by supplementation, which may reduce overall T2DM risk 60 , 61 , as well as higher alanine, a key gluconeogenic substrate associated with increased T2DM risk 16 , 62 . Only one previous study focused on mediation of the association between GDM and T2DM by metabolomic and proteomic signatures. This study found that early postpartum lipidomic profiles mediated 12% of the direct effect between GDM and T2DM 22 . In our analysis, spanning three temporally distant stages (GDM, mid-life, and long-term follow-up), we detected a slightly higher proportion of mediation. These findings highlight that women with a history of GDM may benefit from continued metabolic monitoring and preventive strategies both in the early postpartum period and in the long term. In this study of women without T2DM at baseline, those with prior GDM still displayed marked metabolic dysregulation years after pregnancy, underscoring the need for sustained follow-up beyond the immediate postpartum years. This reinforces the hypothesis that GDM leaves a lasting and progressive metabolic and molecular disturbance on maternal physiology. We also found that metabolomic disruptions partly mediate progression to T2DM later in life. Although this indirect pathway is weaker than the direct effect of GDM on T2DM, our findings suggest that circulating plasma metabolites measured post-partum may capture the metabolic disturbances preceding T2DM, providing potential biomarkers for T2DM monitoring, risk stratification and targets for preventive interventions. Proteomic studies with designs comparable to ours are scarce and typically limited to postpartum cohorts with small samples and short follow-up periods 27 , 28 . We identified several potential biomarkers of prior GDM within a signature that may reflect early subclinical insulin resistance 63 – 65 . Our findings align with the PONCH cohort, which followed women with GDM for up to 10 years 28 . More than half of participants developed prediabetes or T2DM, accompanied by higher insulin resistance (HOMA-IR), circulating lipid and chronic inflammation (CRP), despite half of the sample not having obesity. Importantly, HOMA-IR rose more sharply at 10 than at 6 years, underscoring the progression toward T2DM and its reflection in protein abundance. Similarly, reduced levels of IGFBP2 in our study mirrored PONCH findings 28 . The convergence of these results underscores the central role of IGFBP2 in the metabolic trajectory from GDM to T2DM. Importantly, IGFBP2 also appears relevant during pregnancy itself, as pregnancy and postpartum levels have been associated with a higher risk of both GDM and T2DM 66 – 68 . It remains unclear whether T2DM arising in women with GDM shares the same pathophysiology as T2DM in the general population 28 . Our study shows that GDM-associated omic signatures predict incident T2DM in women with and without a history of GDM, potentially indicating a common metabolic pathway to T2DM (Figs. 4 A and B ). A proteomic study in the UK Biobank identified adiposity- and T2DM-associated signatures that overlapped with ours, including IGFBP2, IGFBP1, and GDF15 69 . Among these, IGFBP2 emerged as a particularly strong candidate 70 , and its circulating levels were inversely associated with polygenic risk scores for T2DM 71 . The associations of IGFBP2 with T2DM risk have been validated in large-scale observational cohorts such as the China Kadoorie Biobank and the UK Biobank 72 . Low circulating IGFBP2 has been consistently reported as a biomarker of T2DM, establishing it as an promising therapeutic target 73 – 76 . This translational potential is supported by clinical evidence showing that semaglutide-induced weight loss was accompanied by reductions in IGFBP2 levels in individuals with obesity 77 . Taken together, these findings position IGFBP2 as a robust biomarker that bridges molecular evidence about the transition from GDM to T2DM. The strengths of our study include the derivation of metabolomic and proteomic composite profiles in women with a history of GDM. Examining whole metabolomic and proteomic profiles, rather than individual biomarkers, provides a more comprehensive view of disease associations and risk prediction, as these profiles integrate signals across pathways and capture heterogeneity that may be missed when focusing on single biomarkers. Omic signatures can reflect broader system-level metabolic disruption, reduce the noise from individual variation, capture different patterns within the same condition, and improve discrimination and predictive accuracy 78 , 79 . Our agnostic approach allowed data-driven identification of disease-associated biomarkers and molecular patterns across thousands of biomarkers. The longitudinal study design allowed for the long-term follow-up of a large cohort of parous women to explore whether these omic signatures predicted T2DM in the entire population. Unlike most studies, we were able to investigate if those signatures mediated the progression from GDM to T2DM given that those metabolites were measured in mid-life, many years after GDM and prior to T2DM diagnosis. The UK Biobank dataset also enabled us to adjust analyses for a variety of socio-demographic, anthropometric, clinical, and behavioural factors, and to ascertain the exposure and outcome from a variety of sources, including self-reported and hospital data. On the other hand, the study has several limitations. We identified less than 1% of the cohort as having had a history of GDM, which is lower than the overall GDM prevalence in the UK. First, the analysis excluded women with T2DM at the baseline assessment, and a proportion of them could have been previously diagnosed with GDM. Also, most women in the UK Biobank gave birth before the 1990s when women in the UK were not tested for GDM unless there was a clear clinical indication and different diagnosis thresholds were applied. This may have led to exposure misclassification, potentially underestimating the association of GDM and omic signatures with T2DM, and subsequently the mediation analyses. We used HES data to ascertain incident T2DM cases, which could have missed cases diagnosed in primary care. Covariates were measured many years after pregnancy, and we were unable to adjust for important confounders such as pre-pregnancy BMI. However, a long time elapsed between first birth and T2DM diagnosis, and we excluded participants being diagnosed with T2DM before the first birth to mitigate this bias. Finally, although we analysed data from a large sample of women in the UK, the generalisability of our findings to the wider UK population is limited by the fact that participants in the UK Biobank tended to have lower rates of deprivation and obesity, and were less likely to smoke and consume alcohol 80 . This study has direct implications for clinical practice. Our findings indicate that metabolomic and proteomic signatures can capture a persistent, subclinical state of dysmetabolism years after pregnancy and well before T2DM is diagnosed. Such signatures – together with a small set of individual biomarkers – could enhance risk stratification, guide the offer and uptake of postpartum prevention programs, and provide personalized feedback to women and clinicians on risk trajectories. Evidence from dietary, physical-activity, and weight-loss interventions shows improvements in high-risk metabolites and concomitant reductions in T2DM risk 53 , 54 , 81 – 84 . Building on this, future trials should test whether targeting whole omic signatures (rather than single biomarkers) amplifies risk reduction, and whether adding accessible biomarker panels to existing risk models improves discrimination and calibration while remaining feasible in clinical settings. As assays become standardized and clinical-grade panels mature, a subset of non-conventional metabolites/proteins could reasonably transition into routine risk assessment for women with prior GDM. Screening pathways can also be streamlined. Uptake of guideline-recommended oral glucose tolerance test (OGTT) every 1–3 years is low, largely due to time and burden 18 , 85 . Because the omic signatures identified here are tightly linked to insulin resistance, integrating simple biochemical indices – fasting glucose and triglycerides combined with TyG index – offers a faster, scalable, and cost-effective option for periodic surveillance, reserving OGTT for equivocal cases or treatment decisions 86 , 87 . In parallel, IGFBP2 emerged as an early biomarker that could be incorporated alongside these indices to refine risk classification and trigger timely preventive interventions. Taken together, these steps would bridge omics-derived insights with pragmatic tools, enabling earlier and more precise prevention of T2DM in women with a history of GDM. Conclusions In conclusion, these results highlight a persistent subclinical state of metabolic dysregulation in women with prior GDM diagnosis, particularly linked to insulin resistance, offering novel opportunities for early targeted intervention. Metabolomic and proteomic signatures capture these molecular imprints and hold promise for integration into precision medicine approaches, supporting more accurate monitoring, risk stratification, and prevention of T2DM not only in women with a history of GDM but also in the wider population. Abbreviations GDM - Gestational Diabetes Mellitus T2DM - Type 2 Diabetes Mellitus BMI - Body Mass Index CVD - Cardiovascular Disease NHS - National Health Service HES - Hospital Episode Statistics ONS - Office of National Statistics ICD - International Classification of Diseases NMR - Nuclear Magnetic Resonance FDR - False Discovery Rate HR - Hazard Ratios VLDL - Very Low-Density Lipoprotein HDL - High-Density Lipoprotein LDL - Low-Density Lipoprotein BCAAs - Branched-Chain Amino Acids GlycA - Glycated Albumin cAMP - Cyclic Adenosine Monophosphate EDA2R - Ectodysplasin A2 Receptor GDF15 - Growth Differentiation Factor 15 CGA - Glycoprotein Hormones, Alpha Polypeptide FSHB - Follicle-Stimulating Hormone Beta Subunit NPR1 - Natriuretic Peptide Receptor 1 ELN - Elastin NEB - Nebulin IGF - Insulin-like Growth Factor IGFBPs - Insulin-like Growth Factor Binding Proteins Declarations Ethics approval and consent to participate The UK Biobank study was conducted according to the Declaration of Helsinki, and ethical approval was granted by the Northwest Multi-centre Research Ethics Committee (reference number 06/MRE08/65). All participants gave informed consent to participate and be followed up through data linkage to routine electronic healthcare records Consent for publication Not applicable. Availability of data and materials The data supporting the findings from this study are available within the manuscript and its Supplementary Information. Researchers can apply to use the raw UK Biobank resource for health-related research and public interest via the UK Biobank Access Management System (https://ams.ukbiobank.ac.uk/ams/). The code used in this study was deposited to a GitHub repository. Competing interests The authors declare no competing interests Funding The UK Biobank was supported by the Wellcome Trust, Medical Research Council, Department of Health, Scottish Government, the Northwest Regional Development Agency, the Welsh Assembly Government and the British Heart Foundation. This research has been conducted using the UK Biobank resource under application number 124330. ATM is funded by Instituto de Salud Carlos III i-PFIS contract: IIS doctorates - company in health sciences and technologies of the Strategic Health Action (IFI22/00013). NMA is supported by SPCR Cross-School Project Exploring the Long-term Outcomes following PrEGnangy affected by gestational diabetes (ELOPE-GDM) Ref: 562, and the NIHR Applied Research Collaboration Oxford and Thames Valley at Oxford Health HS Foundation Trust. CP is funded by RYC2020-028818-I, MCIN/AEI/10.13039/501100011033 and “ESF Investing in your future” (Ministry of Science and Innovation, Spain), which also paid for the access to the data. The views expressed are those of the authors, and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. Author contributions A.T.M., M.M, N.M.A. and C.P conceived and designed the study. A.T.M., M.M., L.J.F., J.C., C.C, N.M.A. and C.P. participated in data cleaning and preprocessing. A.T.M and L.J.F. performed the bioinformatic and statistical analysis. A.T.M., M.M., N.M.A. and C.P. wrote the original draft. S.A.J., C.M.A., N.M.A. and C.P. participated in supervision, project administration and funding acquisition. All authors critically revised drafts and approved the manuscript. Acknowledgements Not applicable References ElSayed, N. A. et al. 2. Classification and Diagnosis of Diabetes: Standards of Care in Diabetes-2023. Diabetes Care 46 , S19–S40 (2023). Wang, H. et al. IDF Diabetes Atlas: Estimation of Global and Regional Gestational Diabetes Mellitus Prevalence for 2021 by International Association of Diabetes in Pregnancy Study Group’s Criteria. Diabetes Res. Clin. Pract. 183 , 109050 (2022). HAPO Study Cooperative Research Group et al. Hyperglycemia and adverse pregnancy outcomes. N. Engl. J. Med. 358 , 1991–2002 (2008). Landon, M. B. et al. A multicenter, randomized trial of treatment for mild gestational diabetes. N. Engl. J. Med. 361 , 1339–1348 (2009). Crowther, C. A. et al. Effect of Treatment of Gestational Diabetes Mellitus on Pregnancy Outcomes. N. Engl. J. Med. 352 , 2477–2486 (2005). Michalopoulou, M., Piernas, C., Jebb, S. A., Gao, M. & Astbury, N. M. Association of gestational diabetes with long‐term risk of premature mortality, and cardiovascular outcomes and risk factors: A retrospective cohort analysis in the UK Biobank. Diabetes Obes. Metab. 26 , 2915–2924 (2024). Song, C. et al. Long‐term risk of diabetes in women at varying durations after gestational diabetes: a systematic review and meta‐analysis with more than 2 million women. Obes. Rev. 19 , 421–429 (2018). Vounzoulaki, E. et al. Progression to type 2 diabetes in women with a known history of gestational diabetes: systematic review and meta-analysis. BMJ m1361 (2020) doi:10.1136/bmj.m1361. Li, Z. et al. Incidence Rate of Type 2 Diabetes Mellitus after Gestational Diabetes Mellitus: A Systematic Review and Meta-Analysis of 170,139 Women. J. Diabetes Res. 2020 , 1–12 (2020). Nielsen, K. K., Kapur, A., Damm, P., De Courten, M. & Bygbjerg, I. C. From screening to postpartum follow-up – the determinants and barriers for gestational diabetes mellitus (GDM) services, a systematic review. BMC Pregnancy Childbirth 14 , (2014). Daly, B. et al. Increased risk of ischemic heart disease, hypertension, and type 2 diabetes in women with previous gestational diabetes mellitus, a target group in general practice for preventive interventions: A population-based cohort study. PLoS Med. 15 , e1002488 (2018). Dennison, R. A. et al. The absolute and relative risk of type 2 diabetes after gestational diabetes: A systematic review and meta-analysis of 129 studies. Diabetes Res. Clin. Pract. 171 , 108625 (2021). Bellamy, L., Casas, J.-P., Hingorani, A. D. & Williams, D. Type 2 diabetes mellitus after gestational diabetes: a systematic review and meta-analysis. The Lancet 373 , 1773–1779 (2009). Tuomilehto, J. et al. Prevention of Type 2 Diabetes Mellitus by Changes in Lifestyle among Subjects with Impaired Glucose Tolerance. N. Engl. J. Med. 344 , 1343–1350 (2001). Belsti, Y. et al. Models Predicting Postpartum Glucose Intolerance Among Women with a History of Gestational Diabetes Mellitus: a Systematic Review. Curr. Diab. Rep. 23 , 231–243 (2023). Guasch-Ferré, M. et al. Metabolomics in Prediabetes and Diabetes: A Systematic Review and Meta-analysis. Diabetes Care 39 , 833–846 (2016). Lai, M. et al. Amino acid and lipid metabolism in post-gestational diabetes and progression to type 2 diabetes: A metabolic profiling study. PLOS Med. 17 , e1003112 (2020). Allalou, A. et al. A Predictive Metabolic Signature for the Transition From Gestational Diabetes Mellitus to Type 2 Diabetes. Diabetes 65 , 2529–2539 (2016). Andersson-Hall, U. et al. Higher Concentrations of BCAAs and 3-HIB Are Associated with Insulin Resistance in the Transition from Gestational Diabetes to Type 2 Diabetes. J. Diabetes Res. 2018 , 1–12 (2018). Huhtala, M. S. et al. Prediction of pre-diabetes and type 2 diabetes nine years postpartum using serum metabolome in pregnant women with gestational diabetes requiring pharmacological treatment. J. Diabetes Complications 37 , 108513 (2023). Wang, X.-M. et al. Metabolic signatures in the conversion from gestational diabetes mellitus to postpartum abnormal glucose metabolism: a pilot study in Asian women. Sci. Rep. 11 , 16435 (2021). Wang, G. et al. Gestational Diabetes Mellitus, Postpartum Lipidomic Signatures, and Subsequent Risk of Type 2 Diabetes: A Lipidome-Wide Association Study. Diabetes Care 46 , 1223–1230 (2023). Khan, S. R. et al. The discovery of novel predictive biomarkers and early-stage pathophysiology for the transition from gestational diabetes to type 2 diabetes. Diabetologia 62 , 687–703 (2019). Tobias, D. K. et al. Type 2 diabetes metabolomics score and risk of progression to type 2 diabetes among women with a history of gestational diabetes mellitus. Diabetes Metab. Res. Rev. 40 , (2024). Lappas, M. et al. The prediction of type 2 diabetes in women with previous gestational diabetes mellitus using lipidomics. Diabetologia 58 , 1436–1442 (2015). Liu, Y. et al. Maternal Metabolites Associated With Gestational Diabetes Mellitus and a Postpartum Disorder of Glucose Metabolism. J. Clin. Endocrinol. Metab. 106 , 3283–3294 (2021). Van, J. A. D. et al. Postpartum defects in inflammatory response after gestational diabetes precede progression to type 2 diabetes: a nested case-control study within the SWIFT study. Metabolism 149 , 155695 (2023). Chung, H. S. et al. Longitudinal clinical and proteomic diabetes signatures in women with a history of postgestational diabetes. JCI Insight 10 , e183213 (2025). Bränn, E., Edvinsson, Å., Rostedt Punga, A., Sundström-Poromaa, I. & Skalkidou, A. Inflammatory and anti-inflammatory markers in plasma: from late pregnancy to early postpartum. Sci. Rep. 9 , 1863 (2019). Waters, T. P. et al. Longitudinal changes in glucose metabolism in women with gestational diabetes, from late pregnancy to the postpartum period. Diabetologia 63 , 385–394 (2020). Ross, K. M. et al. Inflammatory and immune marker trajectories from pregnancy to one-year post-birth. Cytokine 149 , 155758 (2022). Chen, L. et al. Plasma lipidomic profiling reveals metabolic adaptations to pregnancy and signatures of cardiometabolic risk: a preconception and longitudinal cohort study. BMC Med. 21 , 53 (2023). Padilha, M. et al. Prepregnancy Body Mass Index and Lipoprotein Fractions are Associated with Changes in Women’s Serum Metabolome from Late Pregnancy to the First Months of Postpartum. J. Nutr. 153 , 56–65 (2023). Sudlow, C. et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLOS Med. 12 , e1001779 (2015). Würtz, P. et al. Quantitative Serum Nuclear Magnetic Resonance Metabolomics in Large-Scale Epidemiology: A Primer on -Omic Technologies. Am. J. Epidemiol. 186 , 1084–1096 (2017). Li, R. et al. Modifiable risk factors and plasma proteomics in relation to complications of type 2 diabetes. Nat. Commun. 16 , 2896 (2025). Stekhoven, D. J. & Bühlmann, P. MissForest—non-parametric missing value imputation for mixed-type data. Bioinformatics 28 , 112–118 (2012). Visseren, F. L. J. et al. 2021 ESC Guidelines on cardiovascular disease prevention in clinical practice: Developed by the Task Force for cardiovascular disease prevention in clinical practice with representatives of the European Society of Cardiology and 12 medical societies With the special contribution of the European Association of Preventive Cardiology (EAPC). Rev. Espanola Cardiol. Engl. Ed 75 , 429 (2022). Rezende, L. F. M. et al. Lifestyle risk factors and all-cause and cause-specific mortality: assessing the influence of reverse causation in a prospective cohort of 457,021 US adults. Eur. J. Epidemiol. 37 , 11–23 (2022). Bellou, V., Belbasis, L., Tzoulaki, I. & Evangelou, E. Risk factors for type 2 diabetes mellitus: An exposure-wide umbrella review of meta-analyses. PloS One 13 , e0194127 (2018). Mills, K. T., Stefanescu, A. & He, J. The global epidemiology of hypertension. Nat. Rev. Nephrol. 16 , 223–237 (2020). Benjamini, Y. & Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J. R. Stat. Soc. Ser. B Methodol. 57 , 289–300 (1995). Torres-Martos, Á. et al. Multiomics and eXplainable artificial intelligence for decision support in insulin resistance early diagnosis: A pediatric population-based longitudinal study. Artif. Intell. Med. 156 , (2024). Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43 , e47–e47 (2015). Tay, J. K., Narasimhan, B. & Hastie, T. Elastic Net Regularization Paths for All Generalized Linear Models. J. Stat. Softw. 106 , (2023). Korotkevich, G. et al. Fast gene set enrichment analysis. Preprint at https://doi.org/10.1101/060012 (2016). Wu, T. et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. The Innovation 2 , 100141 (2021). Newgard, C. B. et al. A Branched-Chain Amino Acid-Related Metabolic Signature that Differentiates Obese and Lean Humans and Contributes to Insulin Resistance. Cell Metab. 9 , 311–326 (2009). Lynch, C. J. & Adams, S. H. Branched-chain amino acids in metabolic signalling and insulin resistance. Nat. Rev. Endocrinol. 10 , 723–736 (2014). Vanweert, F., Schrauwen, P. & Phielix, E. Role of branched-chain amino acid metabolism in the pathogenesis of obesity and type 2 diabetes-related metabolic disturbances BCAA metabolism in type 2 diabetes. Nutr. Diabetes 12 , 35 (2022). Wang, T. J. et al. Metabolite profiles and the risk of developing diabetes. Nat. Med. 17 , 448–453 (2011). Bloomgarden, Z. Diabetes and branched‐chain amino acids: What is the link? J. Diabetes 10 , 350–352 (2018). Barati-Boldaji, R. et al. Bariatric surgery reduces branched-chain amino acids’ levels: a systematic review and meta-analysis. Nutr. Res. 87 , 80–90 (2021). Ruiz-Canela, M. et al. Plasma branched chain/aromatic amino acids, enriched Mediterranean diet and risk of type 2 diabetes: case-cohort study within the PREDIMED Trial. Diabetologia 61 , 1560–1571 (2018). Kivelä, J., Meinilä, J., Uusitupa, M., Tuomilehto, J. & Lindström, J. Longitudinal Branched-Chain Amino Acids, Lifestyle Intervention, and Type 2 Diabetes in the Finnish Diabetes Prevention Study. J. Clin. Endocrinol. Metab. 107 , 2844–2853 (2022). Mora, S. et al. Lipoprotein Particle Size and Concentration by Nuclear Magnetic Resonance and Incident Type 2 Diabetes in Women. Diabetes 59 , 1153–1160 (2010). Hamaya, R. et al. Association of Plasma Branched-Chain Amino Acid With Biomarkers of Inflammation and Lipid Metabolism in Women. Circ. Genomic Precis. Med. 14 , e003330 (2021). Sankanagoudar, S., Shukla, R., Shukla, K. K. & Sharma, P. Positive association of branched-chain amino acids with triglyceride and glycated haemoglobin in Indian patients with type 2 diabetes mellitus. Diabetes Metab. Syndr. Clin. Res. Rev. 16 , 102481 (2022). Alves, A., Bassot, A., Bulteau, A.-L., Pirola, L. & Morio, B. Glycine Metabolism and Its Alterations in Obesity and Metabolic Diseases. Nutrients 11 , 1356 (2019). Yan-Do, R. & MacDonald, P. E. Impaired “Glycine”-mia in Type 2 Diabetes and Potential Mechanisms Contributing to Glucose Homeostasis. Endocrinology 158 , 1064–1073 (2017). Rasmussen, B. F., Ennis, M. A., Dyer, R. A., Lim, K. & Elango, R. Glycine, a Dispensable Amino Acid, Is Conditionally Indispensable in Late Stages of Human Pregnancy. J. Nutr. 151 , 361–369 (2021). Guasch-Ferré, M. et al. Glycolysis/gluconeogenesis- and tricarboxylic acid cycle–related metabolites, Mediterranean diet, and type 2 diabetes. Am. J. Clin. Nutr. 111 , 835–844 (2020). Cai, Z., Deng, X., Jia, J., Wang, D. & Yuan, G. Ectodysplasin A/Ectodysplasin A Receptor System and Their Roles in Multiple Diseases. Front. Physiol. 12 , 788411 (2021). Sjøberg, K. A. et al. GDF15 increases insulin action in the liver and adipose tissue via a β-adrenergic receptor-mediated mechanism. Cell Metab. 35 , 1327-1340.e5 (2023). Gladych-Macioszek, A. et al. Insulin‐Like Growth Factors, Binding Proteins and Their Role in Pregnancy in Patients With Diabetes. J. Diabetes Res. 2025 , 3330482 (2025). Zhu, Y. et al. Insulin-Like Growth Factor Axis and Gestational Diabetes Mellitus: A Longitudinal Study in a Multiracial Cohort. Diabetes 65 , 3495–3504 (2016). Lappas, M. et al. Postpartum IGF-I and IGFBP-2 levels are prospectively associated with the development of type 2 diabetes in women with previous gestational diabetes mellitus. Diabetes Metab. 42 , 442–447 (2016). Yang, M.-N. et al. Prediction of gestational diabetes mellitus by multiple biomarkers at early gestation. BMC Pregnancy Childbirth 24 , 601 (2024). Gupte, T. P. et al. Plasma proteomic signatures for type 2 diabetes and related traits in the UK Biobank cohort. Diabetes Res. Clin. Pract. 224 , 112194 (2025). Molvin, J. et al. Using a Targeted Proteomics Chip to Explore Pathophysiological Pathways for Incident Diabetes– The Malmö Preventive Project. Sci. Rep. 9 , 272 (2019). Loesch, D. P. et al. Identification of plasma proteomic markers underlying polygenic risk of type 2 diabetes and related comorbidities. Nat. Commun. 16 , 2124 (2025). Yao, P. et al. Proteomic Analyses in Diverse Populations Improved Risk Prediction and Identified New Drug Targets for Type 2 Diabetes. Diabetes Care 47 , 1012–1019 (2024). Rajpathak, S. N. et al. Insulin-Like Growth Factor Axis and Risk of Type 2 Diabetes in Women. Diabetes 61 , 2248–2254 (2012). Haywood, N. J., Slater, T. A., Matthews, C. J. & Wheatcroft, S. B. The insulin like growth factor and binding protein family: Novel therapeutic targets in obesity & diabetes. Mol. Metab. 19 , 86–96 (2019). Elhadad, M. A. et al. Deciphering the Plasma Proteome of Type 2 Diabetes. Diabetes 69 , 2766–2778 (2020). Rooney, M. R. et al. Proteomic Predictors of Incident Diabetes: Results From the Atherosclerosis Risk in Communities (ARIC) Study. Diabetes Care 46 , 733–741 (2023). Maretty, L. et al. Proteomic changes upon treatment with semaglutide in individuals with obesity. Nat. Med. 31 , 267–277 (2025). Buergel, T. et al. Metabolomic profiles predict individual multidisease outcomes. Nat. Med. 28 , 2309–2320 (2022). Carrasco-Zanini, J. et al. Proteomic signatures improve risk prediction for common and rare diseases. Nat. Med. 30 , 2489–2498 (2024). Fry, A. et al. Comparison of Sociodemographic and Health-Related Characteristics of UK Biobank Participants With Those of the General Population. Am. J. Epidemiol. 186 , 1026–1034 (2017). Tulipani, S. et al. Metabolomics‐guided insights on bariatric surgery versus behavioral interventions for weight loss. Obesity 24 , 2451–2466 (2016). Duft, R. G. et al. Altered metabolomic profiling of overweight and obese adolescents after combined training is associated with reduced insulin resistance. Sci. Rep. 10 , 16880 (2020). Papandreou, C. et al. Plasma metabolites predict both insulin resistance and incident type 2 diabetes: a metabolomics approach within the Prevención con Dieta Mediterránea (PREDIMED) study. Am. J. Clin. Nutr. 109 , 626–634 (2019). Walford, G. A. et al. Metabolite Profiles of Diabetes Incidence and Intervention Response in the Diabetes Prevention Program. Diabetes 65 , 1424–1433 (2016). Russell, M. A., Phipps, M. G., Olson, C. L., Welch, H. G. & Carpenter, M. W. Rates of Postpartum Glucose Testing After Gestational Diabetes Mellitus: Obstet. Gynecol. 108 , 1456–1462 (2006). Kim, J. A. et al. Triglyceride and glucose index and the risk of gestational diabetes mellitus: A nationwide population-based cohort study. Diabetes Res. Clin. Pract. 171 , 108533 (2021). Sergi, D. et al. HOMA-IR and TyG index differ for their relationship with dietary, anthropometric, inflammatory factors and capacity to predict cardiovascular risk. Diabetes Res. Clin. Pract. 222 , 112103 (2025). Additional Declarations No competing interests reported. Supplementary Files Additionalfile1.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information

Related documents

• Systematic variant-to-gene mapping highlights TGFB2 and VEGFA as adipokine-coding genes with non-obese, insulin-resistance-like characteristics and distinct disease risks#249466
• Hormonal and Inflammatory Changes During Pregnancy in Women With Glucose Metabolic Disorders.#1057554
• Progression of Metabolic Syndrome Components During Pregnancy#660973
• Gut microbiome in type 2 diabetes: insights from metagenomics, multi-omics, and diet-microbe interactions#4547
• Bump2Baby & Me+ (B2B&Me+): Protocol for a multi-country, European hybrid type 3 implementation-effectiveness study to reduce the incidence of gestational diabetes mellitus and improve maternal and child health.#631091