Advancement in predictive biomarkers for gestational diabetes mellitus diagnosis and related outcomes: a scoping review Authors: Hasini Rathnayake, Luhao Han, Fabrício da Silva Costa, Cristiane Paganoti, Brett Dyer, Avinash Kundur, Indu Singh, Olivia J Holland Published: 2024 ABSTRACT Abstract AbstractObjectiveGestational diabetes mellitus (GDM) is a metabolic disorder associated with adverse maternal and neonatal outcomes. While GDM is diagnosed by oral glucose tolerance testing between 24–28 weeks, earlier prediction of risk of developing GDM via circulating biomarkers has the potential to risk-stratify women and implement targeted risk reduction before adverse obstetric outcomes. This scoping review aims to collate biomarkers associated with GDM development, associated perinatal outcome and medication requirement in GDM.DesignThe Preferred Reporting Items for Systematic Reviews and Meta-Analysis extension for scoping reviews was used to guide the study.Data sourcesThis review searched for articles on PubMed, Embase, Scopus, Cochrane Central Register of Controlled Trials, the Cumulative Index to Nursing and Allied Health Literature and the Web of Science from January 2013 to February 2023.Eligibility criteriaThe eligibility criteria included analytical observational studies published in English, focusing on pregnant women with maternal plasma or serum biomarkers collected between 6 and 24 weeks of gestation. Studies were excluded if they evaluated drug effects, non-GDM diabetes types or involved twin pregnancies, microbiota, genetic analyses or non-English publications.Data extraction and synthesisTwo independent reviewers extracted data. One reviewer extracted data from papers included in the scoping review using Covidence. From the 8837 retrieved records, 137 studies were included.ResultsA total of 278 biomarkers with significant changes in individuals with GDM compared with controls were identified. The univariate predictive biomarkers exhibited insufficient clinical sensitivity and specificity for predicting GDM, perinatal outcomes, and the necessity of medication. Multivariable models combining maternal risk factors with biomarkers provided more accurate detection but required validation for use in clinical settings.ConclusionThis review recommends further research integrating novel omics technology for building accurate models for predicting GDM, perinatal outcome, and the necessity of medication while considering the optimal testing time.Keywords: Diabetes in pregnancy, Prenatal diagnosis, OBSTETRICS, Pregnant Women, Pregnancy STRENGTHS AND LIMITATIONS OF THIS STUDY This review used a comprehensive search strategy across multiple databases to ensure a broad coverage of relevant literature. This review used a comprehensive search strategy across multiple databases to ensure a broad coverage of relevant literature. Adherence to established scoping review guidelines enhances the transparency and reliability of the review process. Adherence to established scoping review guidelines enhances the transparency and reliability of the review process. The inclusion of studies focusing on maternal plasma and serum biomarkers allows for targeted insights into early gestational diabetes mellitus identification. The inclusion of studies focusing on maternal plasma and serum biomarkers allows for targeted insights into early gestational diabetes mellitus identification. A limitation of the review is the exclusion of non-English publications, which may lead to potential language bias. A limitation of the review is the exclusion of non-English publications, which may lead to potential language bias. The heterogeneity in study designs and methodologies may limit the generalisability of the findings across different populations. The heterogeneity in study designs and methodologies may limit the generalisability of the findings across different populations. Introduction Gestational diabetes mellitus (GDM) is a pregnancy complication affecting a significant number of pregnant women. Epidemiological data suggests a wide prevalence range, with estimates fluctuating between 2%–38% depending on the population studied.1 GDM is characterised by impaired glucose tolerance arising during gestation. When poorly controlled, GDM is associated with increased incidence of a range of complications and poor perinatal outcomes including excessive weight gain, pre-eclampsia (PE), macrosomia, preterm delivery, caesarean section, neonatal jaundice, neonatal hypoglycaemia, stillbirth and neonatal death.2,4 The International Federation of Gynaecology and Obstetrics recommends an oral glucose tolerance test (OGTT) as the definitive diagnostic test of GDM at 24–28 weeks of gestation for all pregnant women. However, the timing of this screening offers a limited window for intervention. Currently, GDM diagnosis often relies on a combination of fasting plasma glucose and OGTT results, which may not be as sensitive for early detection in all populations,5 as these tests identify GDM after hyperglycaemia onset, missing subtle changes in glucose regulation before diabetes develops. It also overlooks important metabolic factors like insulin resistance (IR) and inflammation that play a role in diabetes risk.6 Early (eg, first trimester) prediction of the risk of GDM onset and associated outcomes could allow targeted interventions in at-risk individuals to improve outcomes.7 Recent advancements have opened doors to investigating various potential early GDM biomarkers. These include factors related to glycated haemoglobin (HbA1c), insulin and C-peptide levels, inflammatory markers and metabolic profiles. Studies suggest these markers may hold promise for early detection, potentially as early as the first trimester.8,10 Additionally, the development of predictive models incorporating these and other factors is being explored.11 Though HbA1c can be considered an acceptable test in GDM risk prediction, first-trimester HbA1c is not a useful marker to diagnose GDM due to insufficient sensitivity and specificity.12 Despite the established burden and prevalence of GDM, a critical research gap persists in the identification of reliable and readily available predictive biomarkers. While recent years have witnessed promising advancements in biomarker research, a single definitive biomarker for GDM risk prediction remains elusive.13 This lack of a robust and universally applicable biomarker hinders our ability to effectively identify at-risk women early in pregnancy, consequently limiting the potential benefits of timely interventions. Therefore, significant knowledge gaps remain regarding the efficacy and universality of these early detection strategies. Discrepancies exist across studies in identifying the most effective biomarkers and their performance in different populations.9 For example, some research suggests elevated C-reactive protein, an inflammatory marker, can predict GDM in the first trimester,8 while others report less conclusive results.8 Additionally, the optimal timing for biomarker assessment and the development of robust, universally applicable predictive models require further investigation.11 This scoping review aims to map the existing literature on candidate biomarkers for GDM. It will encompass a thorough analysis of the diverse range of biomarkers currently under investigation, the methodologies employed in various studies, and a critical evaluation of the strengths and limitations of the current body of research. By identifying research gaps and areas requiring further exploration, this scoping review will pave the way for future research efforts to identify robust and clinically applicable biomarkers for early GDM diagnosis. Ultimately, this has the potential to significantly improve pregnancy outcomes for countless women and their babies worldwide. Methods Study design The scoping review was carried out in accordance with the Joanna Briggs Institute methodology, the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) extension for Scoping Reviews.14 The review protocol was registered in Open Science Framework (registration DOI https://doi.org/10.17605/OSF.IO/KGFTD). Aims To examine the extent and nature of empirical evidence available on biomarkers for the prediction of GDM at <24 weeks of gestation, perinatal outcomes related to GDM, and the need for specific type of intervention. Review questions The following research question and selection criteria were defined by the PCC framework (‘Participants’, ‘Concept’, ‘Context’) Are there any effective biomarkers and prediction models, established to be used before 24 weeks of gestation, reported in the literature for the prediction of GDM, associated maternal/neonatal outcome, and the need for pharmacological treatment during pregnancy? Are there any effective biomarkers and prediction models, established to be used before 24 weeks of gestation, reported in the literature for the prediction of GDM, associated maternal/neonatal outcome, and the need for pharmacological treatment during pregnancy? What techniques have been used to study biomarkers? What techniques have been used to study biomarkers? Eligibility criteria Publications were included in this review if the study participants were pregnant women without glucose metabolic disorders at the time of recruitment, or specific conditions that can directly affect the levels of blood glucose across pregnancy including polycystic ovary syndrome, and the association of novel biomarkers (including metabolites, nucleic acid biomarkers and proteins) with later GDM development was addressed considering the obstetric outcome of GDM and the need for pharmacological treatment. Full-text articles published from January 2013 to February 2023 in English were included. Sample types were maternal blood, serum and plasma collected from the first trimester before the diagnosis of GDM. Systematic reviews, case reports and review articles were excluded. Search strategy A search strategy has found published studies through a comprehensive electronic literature review of the Cochrane Central Register of Controlled Trials, the Cumulative Index to Nursing and Allied Health Literature, PubMed, Embase, Scopus and Web of Science. This strategy was applied to terms within the titles, abstracts and keywords of articles; ‘gestational diabetes*’, ‘GDM’, ‘marker*’, ‘biomarker*’, ‘indicator*’ ‘test*’, ‘model’, ‘prognosis*’ and ‘predict*’, combined as appropriate using the Boolean operators ‘AND’ and ‘OR’. Additional details can be found in the protocol (https://doi.org/10.17605/OSF.IO/KGFTD) and in online supplemental appendix 6. Data extraction and synthesis Identified studies were collated in Covidence (Veritas Health Innovation) and duplicated articles were removed with EndNote (Clarivate). Titles and abstracts were screened by one reviewer (HR) for assessment against the inclusion and exclusion criteria. The full text of selected citations was assessed in detail against the eligibility criteria by two independent reviewers (HR and LH). Disagreements were resolved through discussion, or through a third reviewer and fourth reviewer (OJH and IS). Data were extracted by one author (HR), including publication information, methods and results, and confirmed by other reviewers (OJH, IS and AK). Patient and public involvement As this scoping review focuses on studies already published regarding circulating biomarkers for predicting the risk of GDM, no patients were directly involved in this research. The findings will be shared both with pregnant individuals and healthcare providers to ensure clarity and relevance, so as to guide future research that truly meets the needs of those people. Results A total of 8837 records were identified after a systematic search. The flow diagram (figure 1) shows the study selection and screening process following PRISMA guidelines.15 After the exclusion of 4823 duplicates, 4014 studies were screened by title and abstract, excluding 3340 studies. The reasons for excluding studies comprise conference abstract, studies with biomarker for postpartum diabetes or diagnosis of GDM, GDM prevalence etc. The studies eligible for full-paper screening were 659. A total of 522 articles were excluded at the full-text screening stage, leaving 137 articles with full text included for data extraction. PRISMA flow diagram. 42% of studies were case-control (figure 2A). In the diagnosis of GDM (figure 2B), 49% of studies used the International Association of Diabetes and Pregnancy Study Groups/WHO criteria. Logistic regression was the algorithm type used by the majority of studies (89%) in building multivariable models (figure 2C). Characteristics of included studies. (A) Proportion of study design types. (B) Proportion of diagnostic criterion types. (C) Proportion of algorithms used for multivariable model-building in multivariable model studies. IADPSG, International Association of Diabetes and Pregnancy Study Groups; ADA, American Diabetes Association; WHO, World Health Organization; AOOG, The American College of Obstetricians and Gynaecologists; NICE, National Institute for Health and Care A total of 313 biomarkers were addressed in different publications, with 278 biomarkers significantly altered in the individuals with GDM compared with the normal glucose tolerance (NGT) group. Biomarkers included 88 proteins, 83 nucleic acids and 107 metabolites (summarised in online supplemental appendix 1). GDM prediction in the included studies was achieved in two ways. More than half (83/137; 60.6%) used molecular biomarkers alone, reporting level change and association with disease. The remaining studies (54/137; 39.4%) used molecular biomarkers in combination with other predictors, such as maternal risk factors, to build multivariable models. The 60.6% of studies that investigated biomarkers alone (summarised in online supplemental appendix 2) identified 197 biomarkers. The type of molecular biomarkers varied, including protein, cell-free DNA (cfDNA), long non-coding RNAs (lncRNAs), microRNA and metabolites. The most frequently studied (58 studies) biomarkers were proteins, especially adipokines, angiogenic factors placenta-related proteins and glycoproteins (online supplemental appendix 2). Sensitivity and specificity for the outcome of NGT versus GDM in the 60.6% of studies that used individual molecular biomarkers were 14%–100% and 20.2%–100%, respectively, and AUC 0.337–0.980 (online supplemental appendix 2). Of these, 10 studies reported biomarkers with an AUC >0.9, while nine studies reported biomarkers with an AUC <0.6. The majority of studies reported biomarkers with an AUC between 0.6 and 0.7, and between 0.7 and 0.8, with 26 studies in each range. In 83 studies that used molecular biomarkers alone, ELISA (41 studies) and biochemical analyser (six studies) were the most popular methods for protein measurement. Atomic absorption spectroscopy (three studies) and liquid chromatography-mass spectrometry (LC-MS) (eight studies) were the main approaches for metabolite measurement. Reverse transcription PCR (RT-PCR) (three studies) and next-generation sequencing (two studies) were the two primary techniques for nucleic acid maker studies. 54 studies combine maternal factors with biochemical and/or biophysical markers to develop multivariable models to predict GDM. Only two studies identified biomarkers at more than one point in time (first and second trimester) before the diagnosis of GDM (online supplemental appendix 3), with both studies yielding higher prediction values of area under the ROC curve (AUC) at the second trimester compared with the first trimester. Online supplemental appendix 4 summarises studies focused on sample collection at a single time point. In studies with multivariable models, sensitivity and specificity for the outcome of NGT versus GDM were 21%–98.2% and 40%–98.1%, respectively, and AUC ranges of 0.602–0.985. The performance of the prediction models varied widely among the studies. The strong prediction accuracy was reflected in the AUCs of more than 0.9 reported by 10 investigations with 11 models. On the other hand, AUC <0.7 was reported by 15 studies. Most studies fell within the moderate to high predictive range; 29 reported models were within an AUC range of 0.7–0.8, and for 23 studies, models had AUCs ranging from 0.8 to 0.9 (online supplemental appendix 4). 11 studies identified biomarkers for 13 adverse outcomes, including maternal and neonatal outcomes and the necessity of medication. Two studies investigated RNA biomarkers and nine studies investigated protein biomarkers to predict adverse outcomes. Three studies developed a model to predict GDM treatment type, introducing novel biomarkers to a maternal risk factor-based model (table 1). Table 1Summary of test performance as a predictive indicator of GDM-related perinatal outcomesBiomarkerAvailability of prediction modelGA of screening (weeks)Study groupSensitivity (%)Specificity (%)AUC (95% CI)ReferenceGDM requiring pharmacological therapy Retinolna11–13GDMnana0.72 (0.55 to 0.89) 117 Maternal risk factors only*11–13GDMnana0.77 (0.60 to 0.94) 117 Maternal risk factors*+log(retinol) corrected for crown-rump length11–13GDMnana0.87 (0.76 to 0.98) 117 Afaminna<12APGnanaNa 45 SHBGMaternal risk factors only†14–17APG59.8‡na0.843 (0.80 to 0.89) 118 Maternal risk factors† +HbA1c + SHBG,14–17APG68.9‡na0.88 (0.84 to 0.92) 118 miR-20a-5p, miR-20b-5p and miR-195–5 pna10–13APG65.073.80.731 70 Maternal risk factors§+miR-20a-5p, miR-20b-5p and miR-195–5 p10–13APG89.586.30.949 70 Maternal risk factors ¶ + miR-20a-5p, miR-20b-5p and miR-195–5 p10–13APG90.089.50.957 70 Afaminna11–14APGnanana 47 GDM requiring diet only miR-1–3 pna13GDM13.8‡nana 70 miR-100–5 pna10–13GDM19.8‡nana 70 miR-125b-5pna10–13GDM14.8‡nana 70 miR-499a-5pna13GDM15.8‡nana 70 miR-574–3 pna10–13GDM21.7‡nana 70 pGCD60na<12APG47610.62 (0.55 to 0.68) 124 Oligohydramnios pGCD59na<12APG100.046.40.96 (0.93 to 0.98) 124 Polyhydramnios pGCD60na<12APG72.711.70.73 (0.59 to 0.86) 124 NGALna8–13APGnanana 57 Preterm delivery NGALna8–13APGnanana 57 Caesarean section NGALna8–13APGnanana 57 Pregnancy-induced hypertension pGCD60na<12APG85.030.40.85 124 Macrosomia pGCD60na<12APG64.546.00.66 (0.58 to 0.74) 124 pGCD60na<12GDMnanana 124 Birth weight PlGFna11–14APGnanana 40 Adiponectinna11–13APGnanana 24 Adiponectin/leptinna11–13APGnanana 24 LGA NGALna8–13APGnanana 57 NGALna<12GDMnanana 57 pGCD59na<12APG59.356.90.71 (0.62 to 0.81) 124 Retinolna11–13APGnanana 117 RBP/TTRna11–13APGnanana 117 TTRna11–13APGnanana 117 Shoulder dystocia pGCD60na<12APG100.083.30.99 (0.98 to 0.99) 124 First minute APGAR scores <7 Zonulinna11–14APGnanana 125 Foetal malformation circRNA LRP6naNot clearAPGnanana 66 Neonatal death pGCD59na<12APGnana0.91 (0.78 to 0.99) 124 Intrauterine death circRNA LRP6naNot clearAPGnanana 66 *Parity and familial history of diabetes mellitus.†Body mass index (BMI), history of GDM, family history of diabetes and soft drink intake before pregnancy.‡At 10% FPR.§Maternal age, BMI at early stages of gestation, and an infertility treatment by assisted reproductive technology.¶Maternal age and BMI at early stages of gestation, an infertility treatment by assisted reproductive technology, history of miscarriage, the presence of thrombophilic gene mutations, positive first-trimester screening for pre-eclampsia and/or FGR by FMF algorithm, family history of diabetes mellitus in first-degree relatives.APG, all pregnancy group; APGARAppearance, Pulse, Grimace, Activity and RespirationAUCarea under the ROC curveFPRFalse positive rateGAGestational ageGDMgestational diabetes mellitusna, not available dataNGAL, neutrophil gelatinase-associated lipocalin; PlGF, placental growth factor; RBP, retinol binding protein; SHBG, sex-hormone binding globulin; TTR, transthyretin Summary of test performance as a predictive indicator of GDM-related perinatal outcomes Parity and familial history of diabetes mellitus. Body mass index (BMI), history of GDM, family history of diabetes and soft drink intake before pregnancy. At 10% FPR. Maternal age, BMI at early stages of gestation, and an infertility treatment by assisted reproductive technology. Maternal age and BMI at early stages of gestation, an infertility treatment by assisted reproductive technology, history of miscarriage, the presence of thrombophilic gene mutations, positive first-trimester screening for pre-eclampsia and/or FGR by FMF algorithm, family history of diabetes mellitus in first-degree relatives. APG, all pregnancy group; APGARAppearance, Pulse, Grimace, Activity and RespirationAUCarea under the ROC curveFPRFalse positive rateGAGestational ageGDMgestational diabetes mellitusna, not available dataNGAL, neutrophil gelatinase-associated lipocalin; PlGF, placental growth factor; RBP, retinol binding protein; SHBG, sex-hormone binding globulin; TTR, transthyretin The diagnostic performance and risk associations of different biomarkers from multiple studies to predict GDM were summarised in online supplemental appendix 5. The first table in online supplemental appendix 5 incorporates data from the studies that presented a point estimate of AUC and 95% CI. Biomarker types include adipokines and cytokines, placenta-related markers, other proteins (eg, glycoproteins, enzymes), metabolic markers and nucleic acid markers measured at different gestation stages. Of these biomarkers, adiponectin, leptin and placental growth factor (PlGF) achieved good predictive accuracy, with AUC values above 0.70. In particular, ApoE and the ficolin/adiponectin ratio obtained the highest diagnostic performance with AUC of 0.965 (95% CI 0.92, 1.01) and 0.97 (95% CI 0.94, 1.0), respectively. The second table in online supplemental appendix 5 presents OR and 95% CI for biomarkers for GDM. Studies had a significant inverse association between adiponectin, suggesting higher levels reduce GDM risk, and mixed results for leptin, with some studies indicating increased risk. The other biomarkers, including resistin, visfatin and placental proteins, have a positive correlation. Further, 25(OH)D and alanine as metabolic markers, and miR-148a-3p as nucleic acid markers, also show good correlation. This forest plot underscores the heterogeneity in biomarker performance and emphasises the need for further validation in diverse populations. Discussion In this scoping review, previous research on novel predictive biomarkers for the risk of GDM and related perinatal outcomes (eg, preterm delivery, necessity of medication, macrosomia) is summarised. 137 articles reported individual molecular biomarkers and biomarkers combined with maternal factors and/or biophysical markers. Biomarker studies for predicting the risk of GDM, related perinatal outcome, and medication necessity are still in the discovery and validation phase, with further development for clinical use. Of the 137 reviewed studies, only two validated their biomarkers externally, while 11 conducted internal validations. That is, basically, there is a wide gap for more rigid external validation that needs to be warranted in proving the clinical applicability of the biomarkers. Predictive molecular biomarkers Protein biomarkers Adipokines and cytokines The GDM groups showed adiponectin and nesfatin-1 levels decreased, and leptin, resistin, visfatin, vaspin and spexin increased, compared with the NGTs groups. Sensitivity of adipokines and cytokines to predict GDM were >28% and AUC ranges of 0.337–0.836, 10 (17%) of them have AUC more than 0.7.16,23 Among them, leptin and adiponectin are the most promising biomarkers in the prediction of GDM risk, achieving a sensitivity of 100% individually in studies with sample sizes of 21 GDM and 47 NGT24 and, 45 GDM and 45 NGT,18 respectively. Leptin is involved in regulating energy balance and insulin sensitivity and its elevated levels in GDM may reflect increased adiposity and IR. Adiponectin, on the other hand, enhances insulin sensitivity and its decreased levels in GDM suggest impaired glucose metabolism.22 Of note, one study identified that, within the same ethnic subcohort, there were differences in leptin levels associated with GDM diagnosis,25 suggesting that ethnicity should be considered in the development of predictive models since different ethnic groups have different degrees of adiposity.26 Furthermore, elevated proinflammatory cytokines such as TNFα, IL-6 and IL-4, along with imbalances in adipokines levels, contribute to IR and impaired glucose metabolism.27 The selected studies for this review identified increased first trimester circulating cytokines TNFα, IL-6 and IL-4 levels in mothers who developed GDM.28,30 The potential evaluation of inflammatory markers for early prediction of GDM still needs to be addressed more and validated in large cohort studies; some of these biomarkers have not yet been validated in different populations. Placental-derived biomarkers The placenta is essential to maternal physiological adaptations, including changing insulin sensitivity over pregnancy,31 and placental-derived biomarkers can provide information on pregnancy health.32 Placental dysfunction may lead to changes in the secretion of placental hormones, disrupting the balance between pregnancy-induced insulin secretion and IR.33 The present scoping review demonstrated that placenta-derived proteins, including pregnancy-associated plasma protein-A2 (PAPP-A2) (at 13.6±2.81 weeks of gestation) and placental protein 13 (at 16–20 weeks of gestation) could predict GDM diagnosis at a sensitivity of 71% and 92%, respectively.34 PAPP-A2 modulates insulin-like growth factor activity, which plays a crucial role in fetal growth and glucose metabolism. Elevated levels of PAPP-A2 may indicate placental stress or dysfunction, contributing to the development of GDM.35 Placental protein 13 is associated with placental development and vascularisation, and its altered levels may reflect impaired placental function, leading to GDM.36 Additionally, angiogenic factors that may be related to placental function, mainly prokineticin 137 and soluble fms-like tyrosine kinase-1 (sFlt-1)36 could also predict 88% and 95% of GDM onset in the early second trimester, respectively. Prokineticin 1 is involved in angiogenesis and placental development, and its dysregulation may impair placental blood flow and nutrient exchange, contributing to GDM.37 sFlt-1 is an antiangiogenic factor that can disrupt placental vascularisation, leading to placental insufficiency and GDM.36 PlGF, which is one of the key predictors of PE,38 is not a good early univariable predictor of GDM diagnosis, presenting with a sensitivity of only 51%.39 40 Other protein biomarkers Proteins that are not cytokines or placenta-derived are categorised in the present study as ‘other proteins’. Maternal circuiting glycoproteins in the first trimester play essential roles in cell signalling and immune response in IR and may also serve as predictive biomarkers for GDM. Sex-hormone binding globulin (SHBG), follistatin and myostatin showed a sensitivity superior to 80% for GDM diagnosis.36 41 42 SHBG regulates the bioavailability of sex hormones, with low levels of SHBG associated with IR and the development of GDM.43 Follistatin neutralises myostatin, which regulates muscle growth and glucose metabolism. Changes in their levels can affect insulin sensitivity and glucose metabolism, predicting GDM.36 Other glycoproteins, including fibronectin,44 afamin,45,49 glycosylated fibronectin,21 sCD16320 and sCD163,20 50 were also reported to be altered in the first trimester of GDM pregnancies compared with NGTs. Fibronectin aids in cell adhesion and repair, with glycosylated forms reflecting hyperglycaemia-related changes, indicating endothelial dysfunction and inflammation in GDM.44 Afamin, linked to oxidative stress, rises in GDM due to increased inflammation.45 sCD163, a marker of macrophage activation, signals chronic inflammation, contributing to IR and GDM.20 Apolipoprotein E (APOE),51 fatty acid-binding protein 4 (FABP4),52 Angiopoietin-like proteins (ANGPTL) 2,53 ANGPTL8,54 APOB51 and small- and low-density lipoprotein51 were involved in physiological processes associated with lipid metabolism, obesity, inflammation and IR. These biomarkers achieve an AUC >0.6.51,54 APOE demonstrated the highest AUC of 0.96555 among independent biomarkers reported in this study. FABP4 alone shows predictive AUC 0.733 at first trimester, and when combined with clinical risk factors, the AUC increases to 0.77.52 AUC increased to 0.82 when the same clinical risk factors were combined with early second trimester FABP4 dosage29 (online supplemental appendix 3). Moreover, iron dysregulation has been associated with IR and diabetes.56 Neutrophil gelatinase-associated lipocalin (NGAL)57 and ferritin58 are two molecules involved in iron transport and metabolism, and both biomarkers are promising in predicting GDM in the first trimester, showing a sensitivity >83%. Combined with clinical risk factors, NGAL showed predictive sensitivity, specificity and AUC of 91.2%, 90.8% and 0.923, respectively57 (online supplemental appendix 2). One study used the ficolin-3/adiponectin ratio for early GDM risk prediction and achieved a sensitivity of 91% (ficolin-3 sensitivity 51%; adiponectin sensitivity 81%).19 Ficolin-3 aids immune response, and adiponectin regulates glucose and fat metabolism. An altered ratio signals disruptions in immune and metabolic functions, a key in GDM development.19 Many of these proteins had good performance as individual biomarkers, which can be improved by combining clinical risk factors including age, body mass index (BMI), family history of diabetes and ethnicity; however, most biomarkers and prediction models have not been fully validated, limiting their ability to be translated into clinical practice. With regard to protein marker studies, only four studies have been internally validated. Nucleic acid biomarkers Cell-free DNA Measurement of the fetal (primarily placental-derived) cfDNA component in the maternal blood has been used for fetal aneuploidy screening.59 Yuan et al identified no significant changes in low fetal cfDNA levels in individuals who could develop GDM compared with NGTs.60 In contrast to measuring total levels, whole-genome promoter profiling of maternal plasma cfDNA found a set of 11 genes (CC2D2B, NAT10, SIPA1, ZNF565, ZNF552, WDR35, MICALL1, CTNNB1, CLOCK, BCKDHB and TGIF2LY) that could be used to predict GDM before 20 weeks of gestation with good accuracy (72%), with many of these genes having functions potentially relevant to GDM development.61 For example, CTNNB1, encoding β-catenin, is crucial in the Wnt signalling pathway, which regulates cell proliferation and glucose metabolism.62 Disruptions in this pathway can lead to impaired glucose regulation. CLOCK, involved in circadian rhythms that influence metabolism, when altered, may contribute to metabolic imbalances, increasing the risk of GDM.61 Additionally, maternal blood CpG methylation was increased in one site (24837915) and decreased in two sites (157 130 156 and 89438648) at 10–15 weeks of gestation in individuals who would later develop GDM compared with NGTs.63 Changes in CpG methylation can affect gene expression, potentially leading to altered metabolic and inflammatory responses, which are key factors in GDM development.63 RNA Several RNA species, such as lncRNA, circular RNA (circRNA) and miRNA, are involved in gene regulation, with expression changing in pathologies, including diabetes, often before overt disease development.64 lncRNA MEG8 may influence insulin signalling pathways and β-cell function, contributing to GDM development. Overexpression of the lncRNA MEG8 1 month before GDM diagnosis achieved AUC 0.73 for GDM prediction,65 and overexpression of circRNA LRP666 achieved AUC 0.88. LRP6 is involved in the Wnt signalling pathway, which regulates glucose metabolism and insulin sensitivity.66 miRNAs can regulate gene expression post-transcriptionally, affecting pathways related to IR, inflammation, and lipid metabolism, which are crucial in GDM. Two studies developed prediction models using differentially expressed extracellular vesicle miRNAs from maternal blood at 10–16 weeks of gestation, achieving AUC 0.8267 and 0.96.68 Maternal circulating miRNAs obtained at 10–13 weeks of gestation have been used in models with an AUC of 0.91, incorporating two specific miRNAs.69 These miRNAs may target genes involved in glucose homeostasis and β-cell function.67,69 This model demonstrates the potential to detect 47.93% of pregnancies with GDM at an FPR of 10%. Furthermore, when 11 cardiovascular disease-associated miRNAs combined with maternal factors, the detection rate increases to 69.17%.70 Cardiovascular disease-associated miRNAs may reflect underlying metabolic and inflammatory changes that predispose to GDM. Metabolic biomarkers Considering GDM as a metabolic disorder, 18 studies explored metal ions, lipidomic, amino acids, metabolites and vitamins as predictive biomarkers. Maternal circulating metal ions evaluated before the second trimester, such as zinc,71,74 copper,72 74 selenium,74 75 magnesium,7274 76,78 iron,73 lead,72 calcium72 76 and manganese,72 were considered predictive biomarkers of GDM. Only iron73 and selenium75 had significantly different levels between GDM and NGT pregnancies, with >80% sensitivity. Elevated iron level can lead to an increase in the production of reactive oxygen species,79,81 contributing to oxidative stress and inflammation, which are linked to GDM, and selenium helps to reduce oxidative stress by acting as a cofactor for antioxidant enzymes, such as glutathione peroxidase, which is important to an adequate placental function;7982,84 thus, its deficiency or imbalance can be associated with GDM. Therefore, the association of these redox-sensitive elements with GDM development may be related to oxidative stress in GDM aetiology, potentially linked to placental and mitochondrial dysfunction.79 85 Adequate nutrition during pregnancy is of utmost importance to support maternal and fetal health.81 86 Decreased vitamin D/25-hydroxyvitamin D87,89 in first trimester showed a sensitivity of 81% and specificity of 44% for GDM diagnosis,87 and high levels of vitamin A90 at first trimester might be involved in the pathogenesis of GDM, with a prediction AUC of 0.649. Vitamins A and D may have roles in the immune response during pregnancy91 92 and its deficiency can impair insulin sensitivity. 28 metabolomic studies identified biomarkers for GDM prediction (online supplemental appendix 2). Several studies reported that circulating metabolic biomarkers are abnormal during GDM-affected pregnancies,93,98 and 17 multivariable model studies combined metabolic biomarkers with maternal factors and/or other biomarkers.7699,102 Valine, isoleucine and leucine achieve >76% sensitivity, and predictive values are ≥AUC 0.67. These branched-chain amino acids including leucine, isoleucine and valine have gained attention because they are linked with energy metabolism, and their elevated levels are associated with IR and metabolic dysfunction, which are key factors in GDM.103 Maternal plasma diacylglycerols (DGs) and triacylglycerols (TGs), which are involved in lipid metabolism, were explored at 11–13 weeks and 19–24 weeks of gestation, and seven TGs and five DGs were identified as potential biomarkers for GDM prediction with AUC 0.96 and AUC 0.97 at each gestational age, respectively.104 In validation data sets, the AUC reached 0.88 and 0.94, respectively. Their elevated levels can indicate dyslipidaemia and IR, contributing to GDM. Metabolic biomarkers have inadequate reproducibility as their levels can be influenced by dietary, lifestyle and environmental factors.81 Only four studies validated their potential metabolic biomarkers using internal104,106 or external107 cohorts. These validated studies represent a proof of concept that an algorithm including a metabolomic profile can effectively predict the risk of GDM in early pregnancy, although these findings need to be replicated and further validated in future research. Diagnostic performance and risk associations of biomarkers in GDM Forest plot analysis demonstrated key patterns in the replication efforts of biomarkers predicting GDM, underlining both strengths and gaps in the research evidence. Replication will be essential in confirming the first findings and determining the reliability of biomarkers; however, the extent of replication varies highly across biomarkers. For example, adiponectin has been very well replicated. Following the findings of Rasanen et al 2014, (OR 0.62, 95% CI 0.46 to 0.84),21 four studies, from 2019 to 2022,19 23 34 108 confirmed their results. This increases the confidence of adiponectin as a predictor of GDM. However, variability in results suggests some confounders that could affect its reliability. Leptin and vitamin D showed moderate replication; however, their results are inconsistent. In the case of leptin, after the first report by Bawah et al in 2019,17 the overall OR was 1.17, 95% CI 1.10 to 1.23, and three reports presented ORs from 1.06 to 2.9;23 24 108 thus, more studies have to be conducted to identify those factors affecting the predictive capacity of leptin. A similar case is vitamin D, which also had at least three replications since the first report, with inconsistent outcomes indicating further investigation about its role in GDM risk. In contrast, many other biomarkers with strong individual associations remain unreplicated and hence unvalidated. For instance, FABP4 had a high OR of 6.91 (95% CI 3.56 to 13.44),52 yet no subsequent studies have been conducted. Other biomarkers, including microRNAs such as miR-148a-3p and miR-192–5 p, are promising but not replicated. Inconsistencies like these are indicative of the systematic validation of GDM biomarkers that is required, particularly those initially associated strongly with disease development. Standardisation of methods and multicentre trials would enable better comparisons to be made, leading to stronger evidence and more predictive models for GDM. Combined risk prediction models The screening for GDM uses a risk-factor-based approach to selectively offer screening tests, but this targeted screening has limited predictive power.109 Maternal age, BMI, ethnicity, family history of diabetes and history of GDM are commonly used risk factors to predict GDM development. Based on BMI, there was a sensitivity, specificity and AUC of 33%,110 111 88%110 and 0.6,48 110 respectively. When age and BMI were evaluated together, there was a sensitivity, specificity and AUC of 22%,106 92%106 and 0.67,106 107 respectively. The addition of biomarkers to clinical risk factors may improve the prediction of GDM. In particular, first trimester adiponectin, leptin, lipocalin-2 and plasminogen activator inhibitor-2 appeared as promising predictors, as combining these four biomarkers with clinical risk factors leads to a sensitivity of 76.8% at 10% FPR and AUC of 0.91.25 In another study, the addition of maternal first trimester metabolites and miRNAs to maternal risk factors reached an AUC of >0.9.51 68 102 112 113 Given the vast physiological changes that occur over pregnancy, the timing of applying specific prediction models will be important. When identifying the best sample collection time point in GDM biomarker studies, only two studies identified more than one time point (first and second trimesters) (online supplemental appendix 3). Both studies yield high prediction values in the second trimester compared with the first trimester. Measurement of biomarkers in the second trimester likely exhibits greater predictive potential since the phenotypic beginning of GDM occurs in the second trimester or later. However, earlier diagnosis is preferable because it allows prompt intervention that may help avoid adverse pregnancy outcomes. 83 studies that measured individual molecular biomarkers reported a sensitivity of >14% and an AUC range of 0.337–0.980, and studies with multivariable models reported a sensitivity of >21% and an AUC range of 0.602–0.985. Although models combining biomarkers and risk factors show the most consistent ability to predict GDM, the clinical application of these multiparametric prediction models is influenced by a plethora of aspects, such as practical acceptance, ability to lead to considerable reduction in unfavourable pregnancy outcomes and cost-effectiveness.114 Biomarker studies for GDM-related perinatal outcomes and necessity of medication Biomarkers alone or in combination with clinical variables may provide early risk prediction, not only for GDM but also for related adverse outcomes (table 1). Birth weight negatively correlates with maternal adiponectin levels and adiponectin/leptin ratio at 11–13 weeks of gestation.24 In GDM, first trimester plasma glycated CD59 (pGCD59) was higher in individuals whose infants were macrosomic.115 The same study determined that pGCD59 generated a very good AUC for oligohydramnios (AUC: 0.96), neonatal death (AUC: 0.91) and shoulder dystocia (AUC: 0.99). Similarly, many adverse pregnancy outcomes like LGA, caesarean section, preterm delivery and polyhydramnios were associated with middle and highest tertials of serum NGAL levels.57 Furthermore, circRNA LRP6 was associated with fetal malformation and intrauterine death.66 These findings indicate that specific biomarkers associated with GDM can be related to pathology and severity, allowing them to be potentially useful in the prediction of both GDM and associated adverse outcomes. GDM can be classified as responsive to nutritional therapy (less severe) compared with requiring medication to achieve adequate glycaemic control (more severe).116 Many studies identified candidate biomarkers for the early prediction of the necessity of medication. One study found lower values of retinol (derivative of vitamin A) at 11–13 weeks of gestation in the insulin-dependent GDM group (iGDM) compared with the diet-controlled group, and the predictive value of log(retinol) to forecast the requirement of insulin use was AUC 0.77, which was increased to AUC 0.87 after including maternal risk factors.117 Three microRNAs (miR-20a-5p, miR-20b-5p and miR-195–5 p) at 10–13 weeks of gestation combined with maternal risk factors and clinical variables could detect 90% of GDM requiring medication.70 One study reported a multivariable model (maternal factors+HbA1 c + SHBG) at 14–17 weeks of gestation to forecast the iGDM and yielded an AUC 0.88 with 68.9% sensitivity at 10% FPR.118 Biomarkers and models that can predict the necessity of medication could improve the clinical management of GDM by allowing timely interventions, like lifestyle modification and medication (eg, insulin). Cost-effectiveness and standardisation of GDM risk prediction The identification of an optimal prediction model for the GDM diagnosis is challenging and should take into account the screening strategy (type and time of diagnostic test), sociodemographic characteristics of the population being evaluated, availability of resources to implement such strategy and the drawbacks of a risk-based population approach. Moreover, it should be highlighted that when effective strategies to identify women at high risk of developing GDM and its related complications are properly applied, the financial impact can be lower than that spent to treat the adverse outcomes when GDM is not adequately recognised and treated.119 Early identification of GDM risk factors that could accurately predict high risk of GDM in the second trimester in women who have not yet been diagnosed with the disease could serve as the foundation for a practical and affordable biomarker-led GDM screening approach.119 In addition to allowing for increased surveillance and timely management of GDM, early detection of women at risk for GDM would aid in the implementation of a focused antenatal care model that permits targeted dietary and lifestyle interventions in the first trimester to prevent GDM.120,122 Thus, the capacity to determine a woman’s specific absolute risk of unfavourable pregnancy outcomes due to GDM would facilitate collaborative decision-making and an individualised care plan. In such a scenario, those women identified at highest risk of adverse obstetric outcomes would benefit from high-intensity care at tertiary hospitals, while those labelled as low risk can be managed with routine antenatal care in the community, avoiding unnecessary intervention and allowing optimised resource allocation.120 121 123 Strengths and limitations To the best of our knowledge, this scoping review is a comprehensive presentation of first and early second trimester predictive biomarkers for GDM, its associated perinatal outcomes, and the necessity of medication. A strength is that our study described early predictive biomarkers without limiting them to a certain type of biomarker. The findings of this study should be perceived in light of its limitations. Due to the extensive nature of the topic, we only considered research that used maternal circulating novel biomarkers, excluding routine blood tests which may limit the scope of risk prediction insights. It is important to note the exclusion of various cellular markers, such as DNA methylation tests, which may limit the scope of the analysis. Future research could benefit from exploring additional cellular markers to enhance the understanding of GDM risk. In this review, quality assessment was not performed to determine the clinical performance of biomarkers and prediction models. Conclusion Accurate early detection of pregnant individuals at risk of GDM would allow prompt intervention to potentially improve outcomes. Although this review found evidence of substantial research effort in identifying biomarkers and prediction models, validated biomarkers/models for predicting the risk of GDM are scarce. This review has identified many promising biomarkers that could potentially predict GDM and related adverse outcomes as well as treatment type. Predictive values are essential in understanding the diagnostic performance of a biomarker and its utility in clinical practice. However, many studies only presented the level of biomarkers in a GDM group compared with NGTs, not the predictive values. The findings of this review emphasise the necessity for further validation and optimisation of prediction strategies through integrating new biomarkers and through combining biomarkers with other known risk factors to create robust predictive models. Additionally, future research should validate biomarkers/models in large multiethnic prospective cohorts with patient characteristics well recorded through collaborative work with clinical teams. To pave the way for future studies on predictive biomarkers of GDM, it is essential to focus on certain research areas. First, it is important to standardise biomarker measurements across different populations and clinical settings to ensure consistent and useful results. Additionally, conducting cost-effectiveness studies will help determine the economic feasibility of including biomarker testing in routine early pregnancy screening programmes. Future studies could also focus on developing and validating multibiomarker panels, including protein, metabolic and nucleic acid markers, to enhance predictive accuracy. Lastly, advancements in omics technologies, particularly proteomics and metabolomics, have led to the identification of new biomarkers that shed light on the molecular pathways involved in the development of GDM. Exploring these research areas will contribute to the improvement of biomarker utility and personalised clinical prediction and management of GDM. supplementary material Footnotes Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2024-089937). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: Not applicable. Patient and public involvement: Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research. Contributor Information Hasini Rathnayake, Email: hasini.rathnayake@griffithuni.edu.au. Luhao Han, Email: luhao.han@griffithuni.edu.au. Fabrício da Silva Costa, Email: fabricio.dasilvacosta@health.qld.gov.au. Cristiane Paganoti, Email: cristiane.paganoti@health.qld.gov.au. Brett Dyer, Email: b.dyer@griffith.edu.au. Avinash Kundur, Email: a.kundur@griffith.edu.au. Indu Singh, Email: i.singh@griffith.edu.au. Olivia J Holland, Email: o.holland@griffith.edu.au. Data availability statement No data are available. References Supplementary Materials Data Availability Statement No data are available.