Inflammatory Biomarkers in Relation to Gestational Diabetes Mellitus: A Prospective Study and Meta-Analysis Authors: Ping Wu, Yi Ye, Yi Wang, Xue Yang, Jiaying Yuan, Jianguo Xu, Yi-Xin Wang, Jin Wu, Gang Liu, An Pan, Xiong-Fei Pan Published: 2026 ABSTRACT BackgroundThe prospective association between C-reactive protein (CRP) and gestational diabetes mellitus (GDM) is inconsistent, and the independent association of tumor necrosis factor-α (TNF-α) with GDM risk remains unclear. We aimed to prospectively evaluate associations of early-pregnancy CRP and TNF-α with incident GDM and updated the evidence through a meta-analysis.MethodsSerum CRP and TNF-α levels were measured at 6–15 weeks of gestation among 332 GDM and 664 matched controls. Conditional logistic regression was used to compute odds ratios (ORs) and 95% confidence intervals (CIs). PubMed and EMBASE were searched up to June 18, 2025, to identify prospective studies that investigated associations of CRP and TNF-α levels with GDM.ResultsAfter multivariable adjustment, OR for GDM was 1.75 (95% CI 1.07, 2.84) for the extreme-quartile comparison of CRP and a non-linear association was noted (P-nonlinear = 0.03). TNF-α was associated with a 1.82-fold (95% CI 1.07, 3.09; P-trend = 0.04) higher risk of GDM. CRP and TNF-α levels were positively associated with the risk of GDM, independent of each other and of pre-pregnancy body mass index (BMI). The meta-analysis confirmed the positive association between CRP and GDM risk, independent of BMI. CRP levels were associated with adverse metabolic traits (e.g. fasting insulin, homeostasis model assessment for insulin resistance [HOMA-IR], and HOMA-β), while TNF-α was not.ConclusionElevated CRP and TNF-α levels in early pregnancy were independently associated with increased risks of GDM. While our findings support potential roles of inflammation in GDM, specific pathways involved should be examined to identify novel targets for the prevention of GDM. Introduction Gestational diabetes mellitus (GDM), characterized by glucose intolerance first recognized during pregnancy, is one of the most common pregnancy complications, with a worldwide prevalence ranging from 1% to >30%.1 GDM is associated with adverse pregnancy outcomes, neonatal complications, and increased long-term risks of metabolic disorders and cardiovascular disease in mothers and their offspring.1 The development of GDM is associated with multiple risk factors, including advanced maternal age, ethnicity, obesity, family history of diabetes, and previous adverse pregnancy outcomes.1 Its pathophysiology is thought to involve increased insulin resistance and inadequate pancreatic β-cell compensation.1 However, the precise mechanisms underlying GDM remain incompletely understood. A variety of biomarkers have been investigated for their potential roles in early prediction and diagnosis of GDM, including biochemical indicators of glucose and lipid metabolism, hormonal markers related to insulin resistance, adipokines, and inflammatory markers.2 Among these, inflammatory markers have attracted increasing attention because chronic low-grade inflammation may contribute to insulin resistance and impaired glucose homeostasis.3,4 The role of inflammation in the pathogenesis of type 2 diabetes has been extensively studied,3–6 and inflammation may increase type 2 diabetes risk partly by aggravating insulin resistance and obesity-related metabolic dysfunction.3,4 Whether inflammation also predisposes pregnant women to the risk of GDM warrants further investigation. Adipose tissues may serve as an important biological link between inflammation and insulin resistance during pregnancy by secreting adipokines and inflammatory cytokines.7 Our previous study showed that early-pregnancy adipokines, including adiponectin and leptin, were significantly associated with subsequent GDM risk.8 Building on these findings, evaluating inflammatory markers in early pregnancy may provide useful insights into the pathophysiology of GDM and help identify women at higher risk before the clinical diagnosis of GDM. C-reactive protein (CRP) is an acute-phase protein released by the liver under the stimulation of cytokines (eg, tumor necrosis factor-α [TNF-α] and interleukin-6), and is widely recognized as a marker of low-grade systemic inflammation.9 TNF-α may be particularly relevant to GDM because it is a key pro-inflammatory cytokine implicated in pregnancy-related insulin resistance.10 Therefore, CRP and TNF-α, as inflammatory markers, are presumably linked to the risk of GDM. However, inconsistent associations between CRP in early pregnancy and risk of GDM have been reported.11–23 Especially, several studies observed a positive association between CRP levels in early pregnancy and GDM risk, although most of these positive associations were null after further adjustment for body mass index (BMI).11,12,14,18 Sample sizes in these studies ranged from 108 to 359, with GDM cases ranging from 35 to 86, which might generate insufficient statistical power. Several studies suggested that TNF-α levels in early pregnancy were higher among pregnant women with GDM than those without GDM,24–28 while others showed no differences10,29–31 and even contrary results.32–34 In addition, previous studies did not statistically assess the relation between TNF-α and GDM with adjustment for potential confounders, so the independent association was still unclear. Therefore, we examined the independent associations of two inflammatory biomarkers, CRP and TNF-α in early pregnancy with subsequent risk of GDM among Chinese pregnant women in a nested case-control study. In addition, we quantified these associations in a meta-analysis of available published data to corroborate our findings. Materials and Methods Study Population The design and procedures of the nested case-control study in the Tongji-Shuangliu Birth Cohort (TSBC) have been described previously.35,36 Briefly, the TSBC was launched in Chengdu, China, in March 2017, and enrolled 6143 pregnant women aged 18–40 years in early pregnancy (6–15 weeks of gestation) until June 2019. Pregnancy status was confirmed according to routine clinical practice through ultrasonographic evidence of intrauterine pregnancy and laboratory assessments, including serum human chorionic gonadotropin (hCG) and progesterone measurements. Participants self-reported no major chronic diseases or infectious diseases (eg, cancer, tuberculosis, or HIV infection). At enrollment, serum samples were extracted from peripheral blood samples collected after at least 8 hours of fasting and subsequently stored in −80°C freezers for future experiments. The study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of the Tongji Medical College, Huazhong University of Science and Technology (2017 [S225]-1). Written informed consent was obtained from all study participants before enrollment. This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.37 Ascertainment of GDM All participants underwent a 75 g oral glucose tolerance test (OGTT) at 24–28 weeks of gestation. GDM was ascertained according to the International Diabetes in Pregnancy Study Group (IADPSG) criteria: fasting plasma glucose ≥5.1 mmol/L, 1-hour ≥10.0 mmol/L, or 2-hour ≥8.5 mmol/L.38 As of June 2019, 347 pregnant women were diagnosed with GDM. Fifteen women with GDM who had insufficient baseline blood samples or no available information on covariates were excluded. Finally, 332 GDM cases and randomly selected 664 controls were individually matched at a 1:2 ratio on maternal age (±3 years) and gestational age at enrollment (±4 weeks). Participants’ characteristics between case and control groups have been described in a previous study.36 Laboratory Assessments All laboratory biomarkers were examined using fasting blood samples collected at enrollment. Serum CRP (mg/L) was measured by Human Quantikine ELISA kit (R&D Systems, Minneapolis, MN), with within-assay and between-assay coefficients of variations (CVs) of 3.7% and 17.8%, respectively. Serum TNF-α (pg/mL) and insulin were analyzed using U-PLEX Metabolic Assays (Meso Scale Diagnostics, Rockville, MD, US) according to the manufacturer’s instructions. The within-assay and between-assay CVs of TNF-α were 6.0% and 19.0%, respectively. Fasting plasma glucose was measured by Glucose Assay Kit (Sichuan Maccura Biotechnology, Chengdu, China). Homeostasis model assessment for insulin resistance (HOMA-IR) was calculated as fasting glucose (mmol/L) × fasting insulin (mIU/L)/22.5. HOMA-β was calculated as fasting insulin (mIU/L) × 20/(fasting glucose [mmol/L]-3.5). Assessment of Covariates Major covariates were collected through structured questionnaires and physical examination at enrollment, including maternal age, education, parity, history of GDM, family history of diabetes, smoking status, drinking status, physical activity, systolic blood pressure, and pre-pregnancy BMI. Systolic blood pressure was measured twice by Omron electronic sphygmomanometer and averaged for analyses. Height and weight were measured with light clothes and no shoes. Pre-pregnancy BMI was calculated as self-reported pre-pregnancy weight in kilograms divided by the square of height in meters. Early-pregnancy BMI was calculated as early-pregnancy weight in kilograms divided by the square of height in meters. Physical activity level was assessed by the Pregnancy Physical Activity Questionnaire.39 Statistical Analyses Study participants were divided into quartiles according to the distribution of CRP and TNF-α levels among the control group, respectively. Maternal characteristics were presented as median (interquartile range [IQR]) for continuous variables and frequency (percentage) for categorical variables. Comparisons between groups were made by Kruskal–Wallis test for continuous variables and Chi-square test for categorical variables. Conditional logistic regression was used to compute odds ratios (ORs) with 95% confidence intervals (CIs) for the associations of CRP and TNF-α with GDM risk. We tested the linear trend across quartiles by assigning the median value of each quartile as a continuous variable. We also calculated the ORs for each 1 natural log increment of CRP and TNF-α levels. Model 1 was adjusted for maternal age, gestational age at enrollment, education level, parity, history of GDM, and family history of diabetes. Model 2 was additionally adjusted for smoking status, drinking status, physical activity, and systolic blood pressure. Model 3 was further adjusted for pre-pregnancy BMI. We further adjusted for CRP (continuous) and TNF-α (continuous) mutually based on Model 3. Restricted cubic splines with 3 knots at the 10th (reference), 50th, and 90th percentiles of CRP levels or TNF-α levels were used to test the potential non-linear associations. We also stratified the analyses by maternal age (≥28 years based on median age in the control group) and pre-pregnancy BMI (≥23 kg/m2 based on the Asian cutoffs for overweight and obesity40). The multiplicative interactions of serum CRP and TNF-α with the stratifying factors on the risk of GDM were tested by including an interaction term. In addition, we performed several sensitivity analyses: 1) replacing pre-pregnancy BMI with early-pregnancy BMI in the multivariable model; 2) further adjusting for baseline fasting plasma glucose; 3) excluding those with a history of GDM; and 4) excluding outliers of CRP and TNF-α, defined as values greater than 4 standard deviations (SDs) above or below the mean. Meta-Analysis The meta-analysis was registered on PROSPERO (CRD42021265653). The meta-analysis was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.41 PubMed and EMBASE were searched up to June 18, 2025, with no language restriction. We included prospective studies that investigated relationships of maternal TNF-α levels or CRP levels in early pregnancy with GDM risks. All reference lists from relevant meta-analyses or reviews were screened to identify additional studies. The search strategy is shown in Table S1. Studies were excluded if they: 1) did not contain CRP/TNF-α levels in early pregnancy or GDM or make comparisons; 2) were not peer-reviewed publications (eg, conference abstract, editorial, comments) or not human studies; or 3) did not provide complete data (eg, mean [SD] or ORs [95% CIs]). Two authors (P.W. and Y.Y.) independently conducted a literature review with divergences resolved by discussions with a third reviewer (Y.W.). We assessed the quality of included studies using the Newcastle–Ottawa Scale,42 which covers the selection of study groups (two items), ascertainment of exposure, study design, the comparability of groups (two items), and the assessment of outcome (three items). If studies provided median and IQR or the maximum and minimum, means and SDs were estimated.43,44 The ORs (95% CIs) for each unit or each 1-SD increment of CRP associated with GDM were transformed to those for the extreme-tertile comparisons.45 We calculated standardized mean differences (SMDs) for CRP or TNF-α levels between women with and without GDM and pooled them with ours using random-effect models. We pooled ORs comparing the extreme tertiles in random-effect models. Several studies provided results from multiple models,11–15,18,20–23 thus making it possible to explore whether the association of CRP with GDM risk was independent of obesity. Therefore, we performed a subgroup analysis based on models adjusted for BMI/pre-pregnancy BMI or not. We also stratified our primary meta-analyses by region, assay methods of exposure, diagnostic criteria of GDM, and maternal characteristics (mean age and BMI/pre-pregnancy BMI). In addition, sensitivity analyses were performed by removing one study at a time to evaluate whether the results were robust. Publication bias was assessed by a funnel plot and Egger’s test. All analyses were conducted by Stata software version 18.0 (Stata Corp., College Station, TX, USA) and R software version 4.0.2 (R Project for Statistical Computing). Two-sided P values <0.05 was considered statistically significant. Results Characteristics of Participants The maternal characteristics across quartiles of CRP and TNF-α are shown in Table 1 and Table 2. The median (IQR) age of the pregnant women was 28.0 (25.0–30.0) years, and gestational age at enrollment was 11 (9.0–12.0) weeks. The median (IQR) levels of CRP and TNF-α were 2.8 (1.3–5.6) mg/L and 2.1 (1.7–2.8) pg/mL, respectively. Women with higher CRP levels had unfavorable metabolic profiles, including higher early-pregnancy BMI, pre-pregnancy BMI, systolic blood pressure, fasting insulin, HOMA-IR, and HOMA-β (all P <0.001); however, such associations were not observed among women with higher TNF-α levels.Table 1Characteristics of Participants Across Quartiles of CRP LevelsCharacteristicOverall (n = 996)aQuartiles of CRP, mg/LaP Value bQuartile 1 (<1.131)Quartile 2 (1.131–2.637)Quartile 3 (2.649–5.282)Quartile 4 (≥5.300)Age, years28.0 (25.0–30.0)26.0 (24.0–29.0)27.0 (25.0–30.0)28.0 (26.0–31.0)28.0 (26.0–31.0)<0.001Gestational age, weeks11.0 (9.0–12.0)10.0 (8.0–11.0)11.0 (9.0–12.0)11.0 (9.0–12.0)11.0 (9.0–12.0)<0.001Education level, n (%)0.12 Senior high school or below597 (59.9)115 (53.0)161 (62.7)155 (62.5)166 (60.6) Above senior high school399 (40.1)102 (47.0)96 (37.4)93 (37.5)108 (39.4)Parity, n (%)<0.001 0494 (49.6)153 (70.5)138 (53.7)106 (42.7)97 (35.4) ≥1502 (50.4)64 (29.5)119 (46.3)142 (57.3)177 (64.6)Family history of diabetes, n (%)78 (7.8)13 (6.0)19 (7.4)24 (9.7)22 (8.0)0.52History of GDM, n (%)27 (2.7)3 (1.4)7 (2.7)7 (2.8)10 (3.7)0.49Tobacco use, n (%)0.88 Current or Ever58 (5.8)14 (6.5)14 (5.5)16 (6.5)14 (5.1) Never938 (94.2)203 (93.6)243 (94.6)232 (93.6)260 (94.9)Alcohol use, n (%)0.92 Current or Ever191 (19.2)45 (20.7)49 (19.1)47 (19.0)50 (18.3) Never805 (80.8)172 (79.3)208 (80.9)201 (81.1)224(81.89)Total physical activity (MET-h/week)116.5 (70.5–171.0)111.3 (75.3–166.2)104.4 (59.7–158.9)124.1 (71.7–176.8)121.0 (74.5–178.0)0.06Pre-pregnancy BMI, kg/m221.0 (19.2–23.0)19.4 (18.3–21.1)20.7 (19.0–22.3)21.5 (19.8–23.1)22.4 (20.7–25.5)<0.001Early-pregnancy BMI, kg/m221.3 (19.5–23.5)19.7 (18.3–21.0)21.0 (19.3–22.7)22.0 (20.1–23.6)23.2 (20.8–25.6)<0.001Systolic blood pressure, mmHg109.0 (103.5–116.0)106.5 (101.5–114.0)109.0 (102.0–114.5)109 (103.0–115.5)112.0 (106.0–119.5)<0.001Fasting plasma glucose, mmol/L4.4 (4.2–4.7)4.4 (4.2–4.6)4.4 (4.2–4.7)4.4 (4.1–4.6)4.4 (4.2–4.7)0.34Fasting insulin, mIU/L8.0 (5.6–11.5)6.2 (4.7–8.6)7.6 (5.3–11.1)8.2 (5.8–11.5)9.8 (7.1–13.5)<0.001HOMA-IR1.6 (1.1–2.3)1.2 (0.9–1.7)1.5 (1.0–2.2)1.6 (1.1–2.3)2.0 (1.4–2.8)<0.001HOMA-β181.3 (124.2–261.8)145.1 (104.7–202.2)178.6 (118.0–240.3)197.5 (137.9–280.2)219.1 (148.5–333.8)<0.001TNF-α, pg/mL2.1 (1.7–2.8)2.2 (1.7–3.1)2.1 (1.7–2.7)2.1 (1.7–2.6)2.2 (1.8–2.9)0.38Notes: a Data was expressed as median (interquartile range) for continuous variables and frequency (%) for categorical variables. b P value was tested by Chi-square test for categorical variables and Kruskal–Wallis test for continuous variables.Abbreviations: BMI, body mass index; CRP, C-reactive protein; GDM, gestational diabetes mellitus; HOMA-IR, homeostasis model assessment for insulin resistance; TNF-α, tumor necrosis factor-α. Table 2Characteristics of Participants Across Quartiles of TNF-α LevelsCharacteristicOverall (n = 996)aQuartiles of TNF-α, pg/mLaP Value bQuartile 1 (<1.666)Quartile 2 (1.666–2.129)Quartile 3 (2.132–2.774)Quartile 4 (≥2.774)Age, years28.0 (25.0–30.0)28.0 (26.0–31.0)27.0 (25.0–30.0)28.0 (25.0–31.0)27.0 (24.0–30.0)<0.001Gestational age, weeks11.0 (9.0–12.0)11.0 (9.0–12.0)11.0 (9.0–12.0)10.5 (9.0–12.0)10.0 (9.0–12.0)0.22Education level, n (%)0.59 Senior high school or below597 (59.9)148 (62.5)148 (58.3)140 (57.4)161 (61.7) Above senior high school399 (40.1)89 (37.6)106 (41.7)104 (42.6)100 (38.3)Parity, n (%)0.5050.08 0494 (49.6)110 (46.4)129 (50.8)110 (45.1)145 (55.6) ≥1502 (50.4)127 (53.6)125 (49.2)134 (54.9)116 (44.4)Family history of diabetes, n (%)78 (7.8)27 (11.4)16 (6.3)20 (8.2)15 (5.8)0.09History of GDM, n (%)27 (2.7)11 (4.6)6 (2.4)7 (2.9)3 (1.2)0.12Tobacco use, n (%)0.44 Current or ever58 (5.8)14 (5.9)10 (3.9)15 (6.2)19 (7.3) Never938 (94.2)223 (94.1)244 (96.1)229 (93.9)242 (92.7)Alcohol use, n (%)0.74Current or ever191 (19.2)41 (17.3)51 (20.1)51 (20.9)48 (18.4)Never805 (80.8)196 (82.7)203 (79.9)193 (79.1)213 (81.6)Total physical activity (MET-h/week)116.5 (70.5–171.0)125.8 (76.6–183.2)111.8 (72.5–160.1)122.7 (69.6–179.0)110.4 (63.8–162.4)0.04Pre-pregnancy BMI, kg/m221.0 (19.2–23.0)20.9 (19.1–22.9)21.3 (19.5–23.1)21.1 (19.3–23.4)20.8 (19.0–22.7)0.22Early-pregnancy BMI, kg/m221.3 (19.5–23.5)21.3 (19.6–23.6)21.6 (19.70–23.5)21.2 (19.4–24.0)21.0 (19.4–23.2)0.25Systolic blood pressure, mmHg109.0 (103.5–116.0)109.0 (104.5–115.5)109.5 (103.0–116.5)109.5 (104.0–115.0)108.0 (101.5–115.5)0.24Fasting plasma glucose, mmol/L4.4 (4.2–4.7)4.4 (4.2–4.7)4.4 (4.1–4.6)4.4 (4.2–4.7)4.4 (4.2–4.7)0.10Fasting insulin, mIU/L8.0 (5.6–11.5)7.7 (5.2–10.3)7.8 (5.4–11.2)8.2 (5.9–12.6)8.3 (6.1–11.8)0.10HOMA-IR1.6 (1.1–2.3)1.5 (1.0–2.2)1.5 (1.0–2.3)1.6 (1.1–2.5)1.6 (1.2–2.3)0.09HOMA-β181.3 (124.2–261.8)179.1 (114.3–240.6)179.0 (127.9–263.4)189.7 (130.3–272.4)183.9 (126.9–261.3)0.26CRP, mg/L2.8 (1.3–5.6)2.5 (1.2–4.7)3.2 (1.4–5.7)3.0 (1.5–6.6)2.8 (1.0–6.0)0.10Notes: a Data was expressed as median (interquartile range) for continuous variables and frequency (%) for categorical variables. b P value was tested by Chi-square test for categorical variables and Kruskal–Wallis test for continuous variables.Abbreviations: BMI, body mass index; CRP, C-reactive protein; GDM, gestational diabetes mellitus; HOMA-IR, homeostasis model assessment for insulin resistance; TNF-α, tumor necrosis factor-α. Table 1Characteristics of Participants Across Quartiles of CRP LevelsCharacteristicOverall (n = 996)aQuartiles of CRP, mg/LaP Value bQuartile 1 (<1.131)Quartile 2 (1.131–2.637)Quartile 3 (2.649–5.282)Quartile 4 (≥5.300)Age, years28.0 (25.0–30.0)26.0 (24.0–29.0)27.0 (25.0–30.0)28.0 (26.0–31.0)28.0 (26.0–31.0)<0.001Gestational age, weeks11.0 (9.0–12.0)10.0 (8.0–11.0)11.0 (9.0–12.0)11.0 (9.0–12.0)11.0 (9.0–12.0)<0.001Education level, n (%)0.12 Senior high school or below597 (59.9)115 (53.0)161 (62.7)155 (62.5)166 (60.6) Above senior high school399 (40.1)102 (47.0)96 (37.4)93 (37.5)108 (39.4)Parity, n (%)<0.001 0494 (49.6)153 (70.5)138 (53.7)106 (42.7)97 (35.4) ≥1502 (50.4)64 (29.5)119 (46.3)142 (57.3)177 (64.6)Family history of diabetes, n (%)78 (7.8)13 (6.0)19 (7.4)24 (9.7)22 (8.0)0.52History of GDM, n (%)27 (2.7)3 (1.4)7 (2.7)7 (2.8)10 (3.7)0.49Tobacco use, n (%)0.88 Current or Ever58 (5.8)14 (6.5)14 (5.5)16 (6.5)14 (5.1) Never938 (94.2)203 (93.6)243 (94.6)232 (93.6)260 (94.9)Alcohol use, n (%)0.92 Current or Ever191 (19.2)45 (20.7)49 (19.1)47 (19.0)50 (18.3) Never805 (80.8)172 (79.3)208 (80.9)201 (81.1)224(81.89)Total physical activity (MET-h/week)116.5 (70.5–171.0)111.3 (75.3–166.2)104.4 (59.7–158.9)124.1 (71.7–176.8)121.0 (74.5–178.0)0.06Pre-pregnancy BMI, kg/m221.0 (19.2–23.0)19.4 (18.3–21.1)20.7 (19.0–22.3)21.5 (19.8–23.1)22.4 (20.7–25.5)<0.001Early-pregnancy BMI, kg/m221.3 (19.5–23.5)19.7 (18.3–21.0)21.0 (19.3–22.7)22.0 (20.1–23.6)23.2 (20.8–25.6)<0.001Systolic blood pressure, mmHg109.0 (103.5–116.0)106.5 (101.5–114.0)109.0 (102.0–114.5)109 (103.0–115.5)112.0 (106.0–119.5)<0.001Fasting plasma glucose, mmol/L4.4 (4.2–4.7)4.4 (4.2–4.6)4.4 (4.2–4.7)4.4 (4.1–4.6)4.4 (4.2–4.7)0.34Fasting insulin, mIU/L8.0 (5.6–11.5)6.2 (4.7–8.6)7.6 (5.3–11.1)8.2 (5.8–11.5)9.8 (7.1–13.5)<0.001HOMA-IR1.6 (1.1–2.3)1.2 (0.9–1.7)1.5 (1.0–2.2)1.6 (1.1–2.3)2.0 (1.4–2.8)<0.001HOMA-β181.3 (124.2–261.8)145.1 (104.7–202.2)178.6 (118.0–240.3)197.5 (137.9–280.2)219.1 (148.5–333.8)<0.001TNF-α, pg/mL2.1 (1.7–2.8)2.2 (1.7–3.1)2.1 (1.7–2.7)2.1 (1.7–2.6)2.2 (1.8–2.9)0.38Notes: a Data was expressed as median (interquartile range) for continuous variables and frequency (%) for categorical variables. b P value was tested by Chi-square test for categorical variables and Kruskal–Wallis test for continuous variables.Abbreviations: BMI, body mass index; CRP, C-reactive protein; GDM, gestational diabetes mellitus; HOMA-IR, homeostasis model assessment for insulin resistance; TNF-α, tumor necrosis factor-α. Characteristics of Participants Across Quartiles of CRP Levels Notes: a Data was expressed as median (interquartile range) for continuous variables and frequency (%) for categorical variables. b P value was tested by Chi-square test for categorical variables and Kruskal–Wallis test for continuous variables. Abbreviations: BMI, body mass index; CRP, C-reactive protein; GDM, gestational diabetes mellitus; HOMA-IR, homeostasis model assessment for insulin resistance; TNF-α, tumor necrosis factor-α. Table 2Characteristics of Participants Across Quartiles of TNF-α LevelsCharacteristicOverall (n = 996)aQuartiles of TNF-α, pg/mLaP Value bQuartile 1 (<1.666)Quartile 2 (1.666–2.129)Quartile 3 (2.132–2.774)Quartile 4 (≥2.774)Age, years28.0 (25.0–30.0)28.0 (26.0–31.0)27.0 (25.0–30.0)28.0 (25.0–31.0)27.0 (24.0–30.0)<0.001Gestational age, weeks11.0 (9.0–12.0)11.0 (9.0–12.0)11.0 (9.0–12.0)10.5 (9.0–12.0)10.0 (9.0–12.0)0.22Education level, n (%)0.59 Senior high school or below597 (59.9)148 (62.5)148 (58.3)140 (57.4)161 (61.7) Above senior high school399 (40.1)89 (37.6)106 (41.7)104 (42.6)100 (38.3)Parity, n (%)0.5050.08 0494 (49.6)110 (46.4)129 (50.8)110 (45.1)145 (55.6) ≥1502 (50.4)127 (53.6)125 (49.2)134 (54.9)116 (44.4)Family history of diabetes, n (%)78 (7.8)27 (11.4)16 (6.3)20 (8.2)15 (5.8)0.09History of GDM, n (%)27 (2.7)11 (4.6)6 (2.4)7 (2.9)3 (1.2)0.12Tobacco use, n (%)0.44 Current or ever58 (5.8)14 (5.9)10 (3.9)15 (6.2)19 (7.3) Never938 (94.2)223 (94.1)244 (96.1)229 (93.9)242 (92.7)Alcohol use, n (%)0.74Current or ever191 (19.2)41 (17.3)51 (20.1)51 (20.9)48 (18.4)Never805 (80.8)196 (82.7)203 (79.9)193 (79.1)213 (81.6)Total physical activity (MET-h/week)116.5 (70.5–171.0)125.8 (76.6–183.2)111.8 (72.5–160.1)122.7 (69.6–179.0)110.4 (63.8–162.4)0.04Pre-pregnancy BMI, kg/m221.0 (19.2–23.0)20.9 (19.1–22.9)21.3 (19.5–23.1)21.1 (19.3–23.4)20.8 (19.0–22.7)0.22Early-pregnancy BMI, kg/m221.3 (19.5–23.5)21.3 (19.6–23.6)21.6 (19.70–23.5)21.2 (19.4–24.0)21.0 (19.4–23.2)0.25Systolic blood pressure, mmHg109.0 (103.5–116.0)109.0 (104.5–115.5)109.5 (103.0–116.5)109.5 (104.0–115.0)108.0 (101.5–115.5)0.24Fasting plasma glucose, mmol/L4.4 (4.2–4.7)4.4 (4.2–4.7)4.4 (4.1–4.6)4.4 (4.2–4.7)4.4 (4.2–4.7)0.10Fasting insulin, mIU/L8.0 (5.6–11.5)7.7 (5.2–10.3)7.8 (5.4–11.2)8.2 (5.9–12.6)8.3 (6.1–11.8)0.10HOMA-IR1.6 (1.1–2.3)1.5 (1.0–2.2)1.5 (1.0–2.3)1.6 (1.1–2.5)1.6 (1.2–2.3)0.09HOMA-β181.3 (124.2–261.8)179.1 (114.3–240.6)179.0 (127.9–263.4)189.7 (130.3–272.4)183.9 (126.9–261.3)0.26CRP, mg/L2.8 (1.3–5.6)2.5 (1.2–4.7)3.2 (1.4–5.7)3.0 (1.5–6.6)2.8 (1.0–6.0)0.10Notes: a Data was expressed as median (interquartile range) for continuous variables and frequency (%) for categorical variables. b P value was tested by Chi-square test for categorical variables and Kruskal–Wallis test for continuous variables.Abbreviations: BMI, body mass index; CRP, C-reactive protein; GDM, gestational diabetes mellitus; HOMA-IR, homeostasis model assessment for insulin resistance; TNF-α, tumor necrosis factor-α. Characteristics of Participants Across Quartiles of TNF-α Levels Notes: a Data was expressed as median (interquartile range) for continuous variables and frequency (%) for categorical variables. b P value was tested by Chi-square test for categorical variables and Kruskal–Wallis test for continuous variables. Abbreviations: BMI, body mass index; CRP, C-reactive protein; GDM, gestational diabetes mellitus; HOMA-IR, homeostasis model assessment for insulin resistance; TNF-α, tumor necrosis factor-α. Associations of CRP and TNF-α with GDM Risk After adjustment for demographic and lifestyle factors, the ORs for the extreme-quartile comparisons were 2.11 (95% CI 1.33, 3.35; P-trend = 0.03) for CRP and 1.85 (95% CI 1.10, 3.13; P-trend = 0.04; Model 2 in Table 3) for TNF-α. Each 1 natural log increment of CRP and TNF-α was associated with a 1.20-fold (95% CI 1.05, 1.37) and 1.44-fold (95% CI 1.06, 1.95) higher risk of GDM, respectively. The positive associations of CRP and TNF-α with GDM remained significant after additional adjustment for pre-pregnancy BMI, with ORs of 1.75 (95% CI 1.07, 2.84) and 1.82 (95% CI 1.07, 3.09) for the extreme-quartile comparisons of CRP and TNF-α, respectively. After further mutual adjustment for CRP and TNF-α, the positive associations were not materially changed.Table 3Odds Ratios (95% Confidence Intervals) of GDM Associated with CRP and TNF-α Levels in Early PregnancyVariableOdds Ratios (95% Confidence Intervals) for GDMP for Trend aPer Log IncrementQuartile 1Quartile 2Quartile 3Quartile 4CRP, mg/L Median (range)0.595 (0.046, 1.130)1.768 (1.131–2.637)3.613 (2.649–5.282)8.887 (5.300, 261.610) Case/total (%)51/217 (23.5)91/257 (35.4)82/248 (33.1)108/274 (39.4)– Model 11.00 (reference)1.90 (1.24, 2.93)1.71 (1.08, 2.70)2.38 (1.53, 3.72)0.0031.25 (1.10, 1.42) Model 21.00 (reference)1.90 (1.22, 2.95)1.66 (1.04, 2.64)2.11 (1.33, 3.35)0.031.20 (1.05, 1.37) Model 31.00 (reference)1.79 (1.15, 2.79)1.49 (0.93, 2.40)1.75 (1.07, 2.84)0.221.13 (0.98, 1.30) Model 3+TNF-α1.00 (reference)1.77 (1.13, 2.76)1.49 (0.92, 2.40)1.69 (1.04, 2.75)0.301.11 (0.97, 1.29)TNF-α, pg/mL Median (range)1.403 (0.225, 1.665)1.903 (1.666, 2.129)2.386 (2.132, 2.774)3.863 (2.775, 65.014) Case/total (%)71/237 (30.0)88/254 (34.7)78/244 (32.0)95/261 (36.4)- Model 11.00 (reference)1.56 (1.00, 2.43)1.45 (0.90, 2.34)1.88 (1.13, 3.14)0.041.47 (1.10, 1.97) Model 21.00 (reference)1.52 (0.96, 2.39)1.38 (0.85, 2.25)1.85 (1.10, 3.13)0.041.44 (1.06, 1.95) Model 31.00 (reference)1.45 (0.92, 2.30)1.28 (0.78, 2.10)1.82 (1.07, 3.09)0.041.48 (1.09, 2.02) Model 3+CRP1.00 (reference)1.45 (0.92, 2.29)1.29 (0.79, 2.12)1.83 (1.08, 3.11)0.041.49 (1.09, 2.02)Notes: a P-trend values were obtained from the logistic regression models by using the median values of each quartile as continuous variables. Model 1: adjusted for maternal age (continuous), gestational age at enrollment (continuous), education (senior high school or below, above senior high school), parity (0, ≥1), family history of diabetes (yes, no), and history of GDM (yes, no). Model 2: adjusted for smoking (never, ever/current), drinking (never, ever/current), physical activity level (continuous), and systolic blood pressure (continuous) based on Model 1. Model 3: adjusted for pre-pregnancy BMI (continuous) based on Model 2. Table 3Odds Ratios (95% Confidence Intervals) of GDM Associated with CRP and TNF-α Levels in Early PregnancyVariableOdds Ratios (95% Confidence Intervals) for GDMP for Trend aPer Log IncrementQuartile 1Quartile 2Quartile 3Quartile 4CRP, mg/L Median (range)0.595 (0.046, 1.130)1.768 (1.131–2.637)3.613 (2.649–5.282)8.887 (5.300, 261.610) Case/total (%)51/217 (23.5)91/257 (35.4)82/248 (33.1)108/274 (39.4)– Model 11.00 (reference)1.90 (1.24, 2.93)1.71 (1.08, 2.70)2.38 (1.53, 3.72)0.0031.25 (1.10, 1.42) Model 21.00 (reference)1.90 (1.22, 2.95)1.66 (1.04, 2.64)2.11 (1.33, 3.35)0.031.20 (1.05, 1.37) Model 31.00 (reference)1.79 (1.15, 2.79)1.49 (0.93, 2.40)1.75 (1.07, 2.84)0.221.13 (0.98, 1.30) Model 3+TNF-α1.00 (reference)1.77 (1.13, 2.76)1.49 (0.92, 2.40)1.69 (1.04, 2.75)0.301.11 (0.97, 1.29)TNF-α, pg/mL Median (range)1.403 (0.225, 1.665)1.903 (1.666, 2.129)2.386 (2.132, 2.774)3.863 (2.775, 65.014) Case/total (%)71/237 (30.0)88/254 (34.7)78/244 (32.0)95/261 (36.4)- Model 11.00 (reference)1.56 (1.00, 2.43)1.45 (0.90, 2.34)1.88 (1.13, 3.14)0.041.47 (1.10, 1.97) Model 21.00 (reference)1.52 (0.96, 2.39)1.38 (0.85, 2.25)1.85 (1.10, 3.13)0.041.44 (1.06, 1.95) Model 31.00 (reference)1.45 (0.92, 2.30)1.28 (0.78, 2.10)1.82 (1.07, 3.09)0.041.48 (1.09, 2.02) Model 3+CRP1.00 (reference)1.45 (0.92, 2.29)1.29 (0.79, 2.12)1.83 (1.08, 3.11)0.041.49 (1.09, 2.02)Notes: a P-trend values were obtained from the logistic regression models by using the median values of each quartile as continuous variables. Model 1: adjusted for maternal age (continuous), gestational age at enrollment (continuous), education (senior high school or below, above senior high school), parity (0, ≥1), family history of diabetes (yes, no), and history of GDM (yes, no). Model 2: adjusted for smoking (never, ever/current), drinking (never, ever/current), physical activity level (continuous), and systolic blood pressure (continuous) based on Model 1. Model 3: adjusted for pre-pregnancy BMI (continuous) based on Model 2. Odds Ratios (95% Confidence Intervals) of GDM Associated with CRP and TNF-α Levels in Early Pregnancy Notes: a P-trend values were obtained from the logistic regression models by using the median values of each quartile as continuous variables. Model 1: adjusted for maternal age (continuous), gestational age at enrollment (continuous), education (senior high school or below, above senior high school), parity (0, ≥1), family history of diabetes (yes, no), and history of GDM (yes, no). Model 2: adjusted for smoking (never, ever/current), drinking (never, ever/current), physical activity level (continuous), and systolic blood pressure (continuous) based on Model 1. Model 3: adjusted for pre-pregnancy BMI (continuous) based on Model 2. Sensitivity Analysis and Subgroup Analysis Early-pregnancy BMI was highly correlated with pre-pregnancy BMI with a correlation coefficient of 0.922 in current study. After adjustment of early-pregnancy BMI instead of pre-pregnancy BMI, the results were not substantially changed (Table S2). In other sensitivity analyses, including additional adjustment for fasting plasma glucose, exclusion of participants with a history of GDM, or exclusion of participants with outliers of CRP and TNF-α (beyond the mean ± 4 SD), the positive associations were also not substantially changed (Table S2). Modeling based on restricted cubic splines showed a non-linear relation between CRP levels and GDM risk (P-nonlinear = 0.03; Figure S1). There were no significant interactions between CRP or TNF-α with the stratifying factors (maternal age or pre-pregnancy BMI) for risk of GDM (P-interaction ≥0.21; Table S3). Meta-Analysis of CRP and TNF-α in Early Pregnancy with GDM Figure S2 summarizes the process of literature search and study selection. A total of 1722 and 1577 records for CRP and TNF-α were identified, respectively. After we screened the titles, abstracts, and full texts, 36 studies for CRP and 12 studies for TNF-α were eligible. Characteristics of these studies are summarized in Table S4. The 36 studies for CRP included a total of 23,979 participants with 5,310 GDM cases. The 12 studies for TNF-α included 2,594 participants with 564 GDM cases. The Newcastle–Ottawa scores indicated a high quality of the included studies for CRP and TNF-α (6 or higher out of 9 points) (Table S5). Twenty eight of the 36 studies provided complete data for comparisons of CRP levels between groups after data conversion. Pregnant women with GDM had higher CRP levels than those without GDM, with an SMD of 0.46 (95% CI 0.12, 0.80; P-heterogeneity <0.001; Figure 1A). There were also elevated TNF-α levels in women with GDM versus those without GDM, with an overall SMD of 0.92 (95% CI 0.27, 1.57; P-heterogeneity <0.001; Figure 1B). Subgroup analyses found that the direction of SMDs for CRP and TNF-α levels between GDM and non-GDM groups was consistent across different subgroups, although most of the SMDs were not statistically significant (Figures S3 and S4). Figure 1Forest plots for comparisons of CRP or TNF-α levels in early pregnancy between participants with and without GDM. The SMD (95% CI) of CRP (A) and TNF-α (B) levels in early pregnancy between women with and without GDM.Image A shows a forest plot for C-reactive protein with columns for study, location, total, case, standardized mean difference (95% CI) and percentage of weight. The x-axis ranges from -2 to 4.68. Each study is represented by a square with a horizontal confidence interval line. The pooled diamond is at 0.46 (95% CI 0.12 to 0.80). Image B displays a forest plot for tumor necrosis factor-α, with similar columns. The x-axis ranges from -1 to 6. The pooled diamond is at 0.92 (95% CI 0.27 to 1.57).Two forest plots comparing standardized mean difference for C-reactive protein and tumor necrosis factor-α.Abbreviations: CI, confidence interval; CRP, C-reactive protein; GDM, gestational diabetes mellitus; SMD, standardized mean difference; TNF-α, tumor necrosis factor-α. Forest plots for comparisons of CRP or TNF-α levels in early pregnancy between participants with and without GDM. The SMD (95% CI) of CRP (A) and TNF-α (B) levels in early pregnancy between women with and without GDM. Eighteen of the 36 studies provided ORs (95% CIs) for CRP in relation to GDM. However, one study reported diagnostic OR based on the receiver operating characteristic curve,46 one reported OR for CRP dichotomous, and three studies reported OR without information on models and units,47–49 so, we pooled our results with 13 remaining studies of CRP with ORs (95% CIs) in random-effect models. Since only one study reported the association between TNF-α in early pregnancy and risk of GDM,29 we did not conduct the meta-analysis. The pooled OR for extreme-tertile comparison of maternal CRP levels was 2.21 (95% CI 1.71, 2.84; P-heterogeneity <0.001) in relation to GDM risk (Figure 2). The magnitude of the positive association was slightly attenuated for models adjusted for BMI, with a pooled OR of 2.19 (95% CI 1.64, 2.91; Figure 2). Sensitivity analyses after omitting one study at a time indicated consistent findings (Figure S5). Consistent positive associations between CRP and risk of GDM were observed among different subgroups (Figure S6). There was no evidence of publication biases according to Egger’s tests (all P ≥ 0.08; Figure S7). Figure 2Association of CRP levels in early pregnancy with risk of GDM in a meta-analysis using random-effect models. The forest plot shows the ORs and 95% CIs for risk of GDM comparing the highest and lowest tertiles of CRP levels in final models or models stratified according to adjustment for BMI or pre-pregnancy BMI.A forest plot summarizing study-specific odds ratios for gestational diabetes mellitus by C-reactive protein tertiles. Three grouped sections are listed: final model; adjustment for body mass index or pre-pregnancy body mass index; no adjustment for body mass index or pre-pregnancy body mass index. A table at left lists study, location, total and case. A right column lists odds ratio with 95% confidence interval and percentage of weight. The x-axis label is not shown; tick labels include 0.1, 1, and 40. The y-axis is a study list. Final model pooled diamond: 2.21 (1.71, 2.84). Adjustment pooled diamond: 2.19 (1.64, 2.91). No adjustment pooled diamond: 2.71 (2.07, 3.54).A forest plot of odds ratio estimates for gestational diabetes mellitus across studies.Abbreviations: BMI, body mass index; CI, confidence interval; CRP, C-reactive protein; GDM, gestational diabetes mellitus; OR, odds ratio. Association of CRP levels in early pregnancy with risk of GDM in a meta-analysis using random-effect models. The forest plot shows the ORs and 95% CIs for risk of GDM comparing the highest and lowest tertiles of CRP levels in final models or models stratified according to adjustment for BMI or pre-pregnancy BMI. Discussion In this prospective nested case-control study, higher CRP and TNF-α levels in early pregnancy were associated with increased risks of GDM, although the positive associations were attenuated after additional adjustment for pre-pregnancy BMI. In addition, meta-analyses suggested that women with GDM had higher levels of both CRP and TNF-α in early pregnancy compared to women without GDM. The meta-analysis combining data from our study with 13 others also supported the positive association between CRP and risk of GDM, independent of maternal BMI and other confounding factors. Two recent meta-analyses and reviews have summarized evidence for early-pregnancy CRP in predicting risk of GDM.50,51 One of them reported that pregnant women with GDM had higher levels of CRP than those without GDM,50 which was consistent with the finding from our meta-analysis. While prospective studies observed positive associations between CRP levels in early pregnancy and GDM risk only in models without adjustment for BMI,11–14,16,18–23 several of these associations became null after further adjustment for BMI11,12,14,18 and the other studies retained consistent positive associations.13,17,20–23 This implies that obesity could be a major confounder for their association and should be controlled in similar analyses. Of note, similar conflicting results were reported in the associations between CRP and type 2 diabetes as well.52,53 The inconsistent associations in existing studies might be partially explained by differences in study populations, confounder adjustments, small sample sizes, and variations in assay methodology. However, we did not find heterogeneity in subgroup analyses. In our analysis among Chinese pregnant women, the positive CRP-GDM association was substantially attenuated but remained significant after additional adjustment for pre-pregnancy BMI. In addition, the association of CRP levels with GDM was still observed after additional adjustment for baseline fasting plasma glucose, which further consolidates the role of CRP. In our review to assess current evidence, the overall and subgroup analyses supported the positive association between CRP and GDM, although the association was weaker in the subgroup with adjustment for BMI, which was similar to what was observed in the meta-analysis of CRP-type 2 diabetes association.54 Due to limited sample sizes and different adjustments of confounders, the findings from our meta-analysis should be interpreted cautiously, although subgroup analyses and sensitivity analyses collectively corroborated the robust positive relationship. Studies that examined the role of TNF-α in GDM are very sparse. To date, no meta-analysis has been performed to evaluate the available evidence on the differences of early-pregnancy TNF-α in women with and without GDM. Our meta-analysis suggested that women with GDM had higher circulating TNF-α levels compared with those without GDM. We found a significant heterogeneity across studies for SMD, which may result from differences in sample sizes, assay methodology, diagnostic criteria, and participants’ characteristics as indicated in subgroup analysis. To our knowledge, only one study investigated the prospective association between TNF-α in early pregnancy and risk of GDM among 232 Saudi pregnant women in a cohort study.29 The positive association in the age and BMI adjusted model was not significant after further adjustment for maternal risk factors, including lifestyle factors.29 However, the small sample size in this study may limit the statistical power. In our original study, elevated TNF-α levels were associated with an increased risk of GDM, independent of lifestyle factors, pre-pregnancy BMI, and fasting blood glucose. Although the prevailing evidence suggests that low-grade inflammation is involved in the development of GDM, the biological mechanisms underlying the links between CRP and risk of GDM are not well understood. Recent evidence indicates that chronic low-grade inflammation contributes to diabetes-related metabolic dysfunction by promoting insulin resistance, β-cell dysfunction, and tissue damage.55 These molecular insights provide biological plausibility for the observed association between early-pregnancy inflammatory markers and subsequent GDM risk. Specifically, evidence suggests that CRP may influence insulin resistance through altered innate immune response56 and endothelial dysfunction.57 In our study, HOMA-IR and HOMA-β were significantly higher across quartiles of CRP levels, suggesting that CRP may have a role in the development of GDM through insulin resistance and insulin secretion. While the mechanisms underlying the association between TNF-α and GDM are not well recognized, TNF-α has been reported as a predictor of insulin resistance in human pregnancy,10,31 and may mediate insulin resistance by increasing fatty acid oxidation through suppression of AMP-kinase signaling,58 stimulating autoimmunity,59 and inhibiting the tyrosine kinase activity of the insulin receptor.56 We did not observe significant differences in HOMA-IR between higher and lower TNF-α levels. This may partly reflect the timing of measurement and the limitation of HOMA-IR as a fasting insulin resistance index, which may not fully capture dynamic changes in insulin sensitivity or inflammatory activity in adipose and placental tissues during pregnancy. GDM diagnosis is usually conducted at 24–28 weeks of gestation, which leaves a short time window for interventions. Thus, early identification of potential risk factors for GDM is crucial for risk stratification. Results in the current study advance our understanding of the complex relationships of inflammation markers with GDM risk and provide evidence for CRP and TNF-α in early pregnancy as robust predictors for subsequent risk of GDM. In particular, since CRP is a sensitive marker of low-grade systemic inflammation and TNF-α is a cytokine released by adipocytes and inflammatory cells in response to chronic inflammation,9 TNF-α and CRP seem closely related. However, we observed that elevated TNF-α and CRP were significantly associated with increased risks of GDM after mutual adjustments. In addition, CRP levels were largely associated with adverse metabolic traits (eg, fasting insulin and HOMA-IR), while TNF-α was not. These marker-specific metabolic profiles may help guide future mechanistic investigations. Further studies are warranted to investigate the specific pathways of different inflammatory markers in the pathogenesis of GDM, especially for TNF-α. Strengths and Limitations The strengths of our original study include a moderate sample size, a prospective design, and extensive clinical and phenotypic data that allow for assessing independent temporal relationships of CRP and TNF-α with GDM. In addition, we updated the evidence through a comprehensive meta-analysis on the topic. However, several limitations need to be recognized. First, baseline CRP and TNF-α levels may not capture the dynamic inflammatory changes during pregnancy. Although CRP and TNF-α are widely studied inflammatory markers, they cannot fully represent the complex inflammatory network involved in GDM. Future prospective studies with repeated measurements and broader inflammatory marker panels are needed to better assess the role of inflammatory markers in GDM development. Second, although we adjusted for a comprehensive list of potential confounders, we cannot rule out the possibility of residual confounding due to the observational study design. Third, the studies included in the meta-analysis varied in CRP measurement methods, diagnostic criteria for GDM, and covariate adjustment strategies. However, we did not observe significant differences in subgroup analyses and sensitivity analyses for the effect estimates, and our results from the meta-analysis should be reliable. Fourth, while our original study among Chinese, along with the meta-analysis, showed consistent positive associations of two inflammatory markers with GDM, our findings should be confirmed by mechanistic evidence. Conclusions In conclusion, we found that elevated serum levels of inflammatory markers in early pregnancy, especially TNF-α, were independently associated with an increased risk of incident GDM, even after adjustment for major confounders, including pre-pregnancy BMI. Our findings further support the hypothesis that inflammation may be involved in the pathogenesis of GDM and add novel prospective evidence for the potential role of TNF-α in early GDM risk stratification. Further studies with repeated measurements during pregnancy, broader inflammatory marker panels, and mechanistic investigations are warranted to examine the specific inflammatory pathways and to identify potential targets for the prevention and control of GDM.