Effect of gestational diabetes mellitus on maternal serum uric acid levels: a systematic review and meta-analysis

Abstract

Objective

This study aimed to investigate the effect of gestational diabetes mellitus (GDM) on maternal serum uric acid (SUA) levels.

Methods

We searched PubMed, Embase, Web of Science, Scopus and the Cochrane Library from inception to 31 January 2025. Cohort and case–control studies reporting SUA at or before 28 weeks’ gestation and subsequent GDM were eligible. Random effects meta analyses with Hartung–Knapp adjustment synthesised.

Results

Sixteen studies (19 reports) encompassing 25 212 GDM cases and 148 519 normoglycaemic controls met all criteria. In eight studies with continuous data (2 690 GDM vs. 6 250 controls), women who developed GDM had an average first trimester SUA 44.3 µmol L⁻¹ higher than controls (95% CI 10.3–78.3; I²=96.9%). Ten studies providing categorical estimates yielded a pooled adjusted OR of 2.35 (95% CI 1.54–3.58; I² = 98.7%), indicating a > two fold risk of GDM with elevated SUA. However, substantial heterogeneity was observed across studies (I² >95%). Sequential omission of individual studies, alternative τ² estimators and restriction to high quality cohorts all preserved statistical significance. Funnel plot asymmetry suggested mild small study effects for categorical data but trim and fill adjustment changed the OR by < 10%.

Conclusions

Elevated SUA measured as early as the periconceptional period is consistently associated with an approximately two fold increase in GDM risk. Given its low cost and routine availability, first trimester SUA may hold promise as an adjunctive early pregnancy risk stratification tool, though further validation studies are needed to confirm its clinical utility.

Introduction

Gestational diabetes mellitus (GDM) has become one of the most prevalent metabolic disturbances in pregnant women worldwide [1]. Its incidence continues to rise in parallel with the global epidemics of overweight and obesity, the growing proportion of advanced age pregnancies, and the progressive tightening of diagnostic criteria, so that current epidemiological models estimate that roughly one in seven pregnancies is now complicated by GDM [2]. Beyond doubling the maternal risks of hypertensive disorders, polyhydramnios and subsequent type 2 diabetes, GDM exerts intergenerational effects: in utero exposure programs offspring toward lifelong obesity, insulin resistance and cardiovascular disease [3–5]. Although international guidelines issued by the IADPSG, ADA and WHO 2013 recommend universal screening with an oral glucose tolerance test (OGTT) at 24–28 weeks’ gestation, the OGTT is resource intensive, poorly tolerated and temporally late; Screening revealed that a high-risk group of women had already experienced irreversible loss of β cells [6, 7]. Consequently, there is an urgent obstetric and public health imperative to identify a simple, inexpensive biochemical marker capable of flagging high risk much earlier in pregnancy.

Serum uric acid (SUA), the end product of purine catabolism, is regulated by glomerular filtration, tubular reabsorption/secretion and systemic oxidative stress [3, 5]. During normal early pregnancy, oestrogen driven renal clearance and haemodilution lower SUA by ≈ 25% relative to non pregnant values [8]. An “unexpected mild elevation” at this stage may therefore serve as an external sign of emerging insulin signalling defects, augmented hepatic gluconeogenesis or activation of the inflammation–oxidative stress axis [8, 9]. Experimental work demonstrates that SUA inhibits endothelial nitric oxide synthase, activates the NLRP3 inflammasome and increases reactive oxygen production, collectively reducing skeletal muscle glucose uptake, promoting β cell apoptosis and intensifying insulin resistance [8]. These mechanisms overlap substantially with the pathophysiology of GDM, giving SUA strong biological plausibility as a predictive factor. Epidemiological evidence, however, remains inconsistent: individual cohort and case control studies report associations ranging from null to odds ratios >5, often with limited sample sizes; two small 2023 meta analyses each synthesised fewer than ten studies and yielded pooled odds ratios of 1.67 and 2.58, respectively, with high heterogeneity and no resolution of gestational timing effects [10, 11]. Furthermore, these prior analyses lacked detailed investigation of gestational timing effects and extensive sensitivity analyses to test the robustness of their findings.

SUA was chosen over other potential biomarkers such as HbA1c and adiponectin due to several advantages: (1) it is routinely measured in clinical practice with standardized assays, (2) it requires no special storage conditions, (3) it is cost-effective and widely available, and (4) it has strong biological plausibility for predicting GDM based on its role in insulin resistance and oxidative stress pathways. Against this background, we prospectively registered our protocol in PROSPERO and, adhering to PRISMA 2020 and MOOSE guidelines, performed the largest meta analysis to date exploring the relationship between SUA measured from the pre conception period up to 28 weeks of gestation and the subsequent development of GDM. We aimed to determine whether elevated early pregnancy SUA constitutes a viable early predictor of GDM and to examine the influence of sampling time, assay methodology, population characteristics and underlying biological mechanisms, thereby providing an evidence base for refined first trimester risk stratification and targeted preventive strategies.

Methods

Protocol registration and reporting standard

The review protocol was prospectively registered in PROSPERO. All methods were developed a priori and are reported in accordance with the PRISMA 2020 statement and the MOOSE guideline for meta analyses of observational studies.

Eligibility criteria

Studies were eligible if they investigated the association between maternal serum (or plasma) uric acid concentrations obtained at, or before, the 28th gestational week and the subsequent development of gestational diabetes mellitus (GDM).

In detail, we included (i) cohort (prospective, retrospective, or nested) and case–control studies that recruited pregnant women without pre-existing diabetes; (ii) reported either mean ± SD UA values for GDM and normoglycaemic controls or an effect estimate [odds ratio (OR) or risk ratio (RR)] with 95% confidence interval; and (iii) diagnosed GDM with recognised criteria (IADPSG, ADA, Carpenter–Coustan, WHO 2013, NDDG). We excluded case series, editorials, conference abstracts, animal studies, and studies measuring UA postpartum, in cord blood, or in neonates.

The 28-week cutoff was chosen to balance the inclusion of early pregnancy measurements with the recognition that physiological SUA levels begin to rise after 24 weeks of gestation. This cutoff allowed us to capture the majority of first and early second trimester measurements while maintaining clinical relevance for early risk stratification. Sensitivity analyses were planned to examine the impact of restricting analyses to measurements obtained before 20 weeks of gestation.

Information sources and search strategy

A comprehensive search of PubMed, Embase, Web of Science, Cochrane Library and Scopus was conducted from inception to 31 January 2025 without language or date restrictions. The search strategy combined controlled vocabulary (MeSH, Emtree) and free text terms:

(“uric acid” OR “urate” OR hyperuricemia) AND (“gestational diabetes” OR GDM) AND (pregnancy OR pregnant OR “gestational week” OR prenatal).

Reference lists of all eligible articles and recent reviews were hand searched. ClinicalTrials.gov and the WHO ICTRP were screened for additional grey literature.

After automatic deduplication in EndNote X20, two reviewers (A.B., C.D.) independently screened titles and abstracts in Rayyan. Full texts of potentially eligible reports were retrieved and assessed against the inclusion criteria. Disagreements were resolved by discussion or, if required, adjudication by a third reviewer (E.F.). The final screening flow is summarised in a PRISMA 2020 diagram.

Data extraction and risk of bias assessment

A piloted extraction form was used. For each study we recorded: first author, year, country, study design, recruitment frame, sample size (GDM/controls), gestational age at UA sampling, assay method, unit of UA, diagnostic criteria for GDM, crude and/or adjusted effect estimates with 95% CI, covariates adjusted for, and Newcastle–Ottawa Scale (NOS) domain scores. When continuous UA values were reported in mg dL⁻¹ they were converted to µmol L⁻¹ (1 mg dL⁻¹ = 59.48 µmol L⁻¹). If multiple UA time points were provided, the earliest measurement ≤ 28 weeks was extracted.

Potential overlap in study populations was addressed by carefully reviewing study characteristics, recruitment periods, and geographical locations. When multiple publications from the same cohort were identified, we contacted the corresponding authors when possible to clarify potential overlap. In cases where overlap was confirmed or suspected, the most comprehensive or recent publication was included to avoid double-counting of participants.

Methodological quality was appraised with the Newcastle–Ottawa Scale—the cohort or case–control version as appropriate. Each study can receive up to nine stars across three domains: Selection (four stars), Comparability (two stars) and Outcome/Exposure assessment (three stars). Scores ≥ 7 stars were considered high quality, 5–6 moderate, and ≤ 4 low. Scores ≥ 7 stars were considered high quality based on commonly used conventions in systematic reviews, though we acknowledge this threshold is somewhat arbitrary.

The data extraction form was piloted on three studies and refined based on feedback from all reviewers. The form included standardized fields for study identification, population characteristics, methodological details, statistical results, and quality assessment criteria. Disagreements in data extraction were resolved through discussion between reviewers (A.B. and C.D.), with involvement of a third reviewer (E.F.) when consensus could not be reached. All extracted data were double-checked for accuracy before analysis.

Effect measures

For continuous outcomes, differences in UA were synthesised as Hedges’ g (standardised mean difference) using the inverse variance method. For binary outcomes, reported ORs or RRs were logarithmically transformed; when both were available, the multivariable adjusted OR was preferred. Pooled estimates were calculated with random effects models a priori, employing restricted maximum likelihood (REML) for τ², and Hartung–Knapp adjustment where applicable.

Assessment of heterogeneity

Statistical heterogeneity was quantified by the Cochran Q test (P < 0.10 considered significant), Higgins I² statistic, and between study variance (τ²). I² values of 25, 50, and 75% were interpreted as low, moderate, and high heterogeneity, respectively.

Sub group analyses and sensitivity analyses

Pre specified subgroup analyses examined differences by gestational age at UA sampling (trimester 1 vs. trimester 2), study design (cohort vs. case–control), and study quality (NOS ≥ 7 vs. < 7). Random effects mixed models were computed for each subgroup, and the χ² test compared subgroup estimates. Univariable random effects meta regression explored continuous modifiers (mean maternal age, body mass index, sample size) when ≥ 10 studies were available. Robustness was tested by (i) leave one out analyses (metainf), (ii) analysis restricted to high quality studies, (iii) comparison of random versus fixed effect pooling, and (iv) alternative τ² estimators (DerSimonian–Laird, Paule–Mandel).

Statistical software

All analyses were performed in R 4.4.2 using the meta 8.0.2 and metafor 4.8.0 packages. Unless otherwise specified, P < 0.05 were deemed statistically significant and all confidence intervals are two sided.

Results

Study selection

The electronic search returned 1982 records (PubMed 742, Embase 611, Web of Science 385, Cochrane 74, Scopus 170). After removal of 412 duplicates, 1 570 unique titles and abstracts were screened; 1 490 were excluded as clearly irrelevant. Eighty full text reports were retrieved, two could not be obtained, and 62 were excluded after detailed assessment—most frequently because the outcome was not serum uric acid (n = 25), the population involved pre existing diabetes (n = 14), the report was only an abstract (n = 12), or it represented a lower quality duplicate of an already included cohort (n = 11). Finally, 16 studies (19 reports) met all eligibility criteria and were included in the qualitative and quantitative syntheses (Fig. 1).

Fig. 1.

PRISMA 2020 flow diagram

Study characteristics

Key characteristics of the 16 included studies are summarised in Table 1. Eight were prospective cohorts (five conventional, one high risk, two nested), five were retrospective cohorts, and three employed a case control design. Sample sizes ranged from 105 participants in a high risk Egyptian cohort to 73 649 in a large Chinese database study; altogether the analysis comprised 25 212 women with GDM and 148 519 normoglycaemic controls.

Table 1.

Study characteristics

Reference (first author – year)	Country	Study design	Sample size (GDM/control)	Gestational age at UA sampling	GDM diagnostic criterion	Quality ★	SUA threshold (µmol/L)
Rehman 2021 [12]	Pakistan	Case–control	86/86	< 14 weeks	ADA 75 g OGTT	7	360
Mishu 2019 [13]	Bangladesh	Case–control	86/86	24–28 w & 34–36 w	WHO 2013	5	NR
Şahin Aker 2016 [14]	Türkiye	Retrospective case–control	66/202	6–12 weeks	Carpenter–coustan 100 g OGTT	8	309
Pang 2023 [15]	China	Retrospective cohort	1 396/16 854	11–20 weeks	IADPSG 75 g OGTT	8	360
Zhou 2012 [16]	China	Prospective cohort	100/900	20 weeks	Carpenter–coustan	7	NR
Amudha 2017 [17]	India	Prospective cohort	15/163	< 14 weeks	IADPSG 75 g OGTT	6	NR
Kanwar 2025 [6]	India	Nested case–control	55/110	< 14 weeks	IADPSG 75 g OGTT	7	357
Laughon 2009 [18]	USA	Prospective cohort	73/1 497	8–15 weeks	Carpenter–coustan	7	333
Yue 2023 [19]	China	Retrospective cohort	3 204/20 819	≤ 24 w (median 17 w)	IADPSG 75 g OGTT	8	360
Ganta 2019 [20]	India	Prospective cohort	88/224	< 12 weeks	IADPSG 75 g OGTT	5	NR
Zhang 2024 [8]	China	Prospective cohort	5 013/28 017	6–13 weeks	IADPSG 75 g OGTT	8	360
Li 2024 [21]	China	Retrospective cohort	1 744/4 256	8–12 weeks	IADPSG 75 g OGTT	8	360
Güngör 2006 [22]	Türkiye	Case–control	56/56	24–28 weeks	Carpenter–Coustan	5	NR
Zhao 2022 [23]	China	Retrospective cohort	11 960/73 649	≤ 24 w (median 17 w)	IADPSG 75 g OGTT	8	360
El-Gharib 2013 [24]	Egypt	Prospective high-risk cohort	105/145	≤ 13 weeks	NDDG 100 g OGTT	6	NR
Ghanei 2023 [25]	Iran	Case–control	478/454	Pre-pregnancy ≤ 6 months	ADA 75 g OGTT	7	NR

Gestational age at uric acid sampling varied from the pre conception period (Ghanei 2023) to 34–36 weeks (Mishu 2019); the majority (12/16) collected samples in the first or early second trimester (≤ 20 weeks). GDM was diagnosed most often with the IADPSG 75 g OGTT (nine studies), followed by Carpenter–Coustan (four), ADA criteria (two) and WHO 2013 or NDDG (one each).

Continuous outcome pooling was possible for the eight studies that reported mean ± SD uric acid concentrations (total = 2 665 GDM vs. 5 270 controls). Ten studies provided extractable adjusted odds ratios comparing elevated UA with the reference category; those data represented 25 048 GDM cases and 137 637 controls.

Overall, the analysis comprised 25,212 women with GDM and 148,519 normoglycaemic controls across all 16 included studies. However, not all studies contributed data to both analyses: eight studies provided continuous SUA data (2,665 GDM cases vs. 5,270 controls), while ten studies contributed categorical data for the odds ratio analysis (25,048 GDM cases vs. 137,637 controls). The difference in sample sizes reflects the different types of data available from each study.

Methodological quality

Detailed Newcastle Ottawa ratings are presented in Table 2. Twelve studies (75%) achieved high quality (≥ 7 stars); the median NOS score was 7 (range 5–8). Selection bias was generally low (median ★★★), whereas comparability showed greater variability—five studies failed to adjust for major confounders and received only one star. Follow up adequacy and exposure assessment were acceptable in most cohorts; case control studies occasionally lacked explicit non response information, yielding lower exposure scores.

Table 2.

Quality assessment of studies included

Study (first-author – year)	Design	Selection (max 4 ★)	Comparability (max 2 ★)	Outcome/Exposure (max 3 ★)	Total ★
Rehman 2021	Case–control	★★★	★	★★	7
Mishu 2019	Case–control	★★	★	★	5
Şahin Aker 2016	Case–control	★★★★	★★	★★	8
Pang 2023	Retro. cohort	★★★	★★	★★	8
Zhou 2012	Pros. cohort	★★★	★	★★	7
Amudha 2017	Pros. cohort	★★	★	★★	6
Kanwar 2025	Nested case–control	★★★	★★	★	7
Laughon 2009	Pros. cohort	★★★	★★	★	7
Yue 2023	Retro. cohort	★★★	★★	★★	8
Ganta 2019	Pros. cohort	★★	★	★★	5
Zhang 2024	Pros. cohort	★★★	★★	★★	8
Li 2024 (Sci Rep)	Retro. cohort	★★★	★★	★★	8
Güngör 2006	Case–control	★★	★	★★	5
Zhao 2022	Retro. cohort	★★★	★★	★★	8
El-Gharib 2013	Pros. high-risk cohort	★★	★	★★	6
Ghanei 2023	Case–control	★★★	★★	★	7

Meta analysis

Quantitative synthesis of continuous uric acid concentrations

Eight studies (GDM = 2,690; controls = 6,250) reported continuous data. Observed mean differences ranged from 2.4 µmol L⁻¹ (Mishu 2019 [13]) to 129.7 µmol L⁻¹ (Şahin Aker 2016 [14]). The random effects model indicated a pooled increase of 44.3 µmol L⁻¹ in women who developed GDM (95% CI 10.3 to 78.3). Heterogeneity was high (I² = 96.9%; τ² = 1 592.2; P < 0.0001) (Fig. 2).

Fig. 2.

Quantitative synthesis of continuous uric acid forest plot

Quantitative synthesis of categorical high uric acid exposure

Twelve comparisons from 10 studies provided adjusted odds ratios. Individual ORs ranged from 1.01 (Li 2024 [20], per µmol L⁻¹ increase) to 81.8 (Amudha 2017 [17]). The random effects model yielded a summary OR = 2.35 (95% CI 1.54 to 3.58), signifying more than twice the risk of GDM among women with elevated uric acid. Heterogeneity remained considerable (I² = 98.7%; τ² = 0.542; P < 0.0001) (Fig. 3).

Fig. 3.

Quantitative synthesis of categorical high uric acid forest plot

Sensitivity analysis

Leave one out testing for the continuous outcome (Fig. 4) showed that removal of any single study changed the pooled standardised mean difference (SMD) only modestly, from 0.64 after omitting Sahin Aker 2016 to 0.92 after omitting Mishu 2019. The corresponding 95% confidence intervals all excluded zero and the P values remained below 0.03. The re calculated random effects summary was 0.82 (95% CI 0.26 to 1.37; P = 0.010) with tau squared 0.413 and I squared 94.8%, indicating that the overall elevation in uric acid was robust.

Fig. 4.

Leave one out analysis for the continuous outcome

For the binary outcome (Fig. 5) the leave one out odds ratios ranged between 2.37 (after omitting Amudha 2017 [17]) and 3.30 (after omitting Pang 2023 [15]). The pooled estimate after sequential omission was 2.96 (95% CI 1.44 to 6.08; P = 0.008) with tau squared 0.658 and I squared 97.1%. Thus, no individual comparison explained the association between high uric acid and gestational diabetes.

Fig. 5.

Leave one out analysis for binary outcome

Subgroup analysis

Figure 6 compares studies that sampled uric acid during the first or early second trimester (n = 7) with the single study that measured levels before conception. The first trimester group showed an SMD of 0.85 (95% CI 0.19 to 1.51; I squared 87.1%). The pre-pregnancy study reported an SMD of 0.60 (95% CI 0.46 to 0.73). The chi square test indicated a statistically significant difference between subgroups (chi squared = 5.35, df = 1, P = 0.021), but heterogeneity within the first trimester strata remained high.

Fig. 6.

Subgroup analysis for the continuous outcome

For the binary outcome (Fig. 7), studies with mixed or undefined sampling points (n = 10) yielded a pooled odds ratio of 3.08 (95% CI 1.15 to 8.23; I squared 95.7%). The two comparisons limited to second trimester samples gave an odds ratio of 3.02, but the confidence interval was extremely wide (95% CI 0.00 to 1876.19) because of large between study variability (I squared 98.4%). The test for subgroup difference was significant on the common effect scale (chi squared = 85.55, df = 1, P < 0.001) yet the random effects model showed that heterogeneity persisted, suggesting that timing of sampling alone does not account for the observed dispersion.

Fig. 7.

Subgroup analysis for the binary outcome

Assay methodology significantly moderated pooled estimates (Fig. 8). Specifically, enzymatic assays yielded smaller, less heterogeneous effect sizes compared to colorimetric assays (subgroup difference: χ²=21.31, df = 1, p < 0.0001 for continuous outcomes; χ²=3.85, p = 0.0498 for categorical outcomes). The diagnostic thresholds defining elevated SUA varied from 300 to 360 µmol/L, substantially influencing observed heterogeneity. Ethnicity did not notably affect continuous outcome heterogeneity (χ²=0.24, p = 0.63) but significantly moderated categorical outcomes (χ²=23.76, p < 0.0001).

Fig. 8.

Methodological moderators: assay type and SUA threshold

Meta-regression

Meta-regression analyses showed no significant relationship between effect sizes and publication year or maternal age (Fig. 9). Maternal BMI exhibited a non-significant positive correlation trend with effect size (95% CI crossed zero). Sampling trimester (timing of SUA measurement) varied markedly across studies, potentially contributing further to heterogeneity. Additionally, confounder adjustment was inconsistent; notably, only 62% of studies controlled for both BMI and parity, with rare adjustment for family history, dietary factors, or physical activity.

Fig. 9.

Meta-regression diagnostics

Publication bias assessment

The funnel plot for categorical data (Fig. 10) showed asymmetry, with a notable absence of small studies showing null or negative associations. Egger’s regression test confirmed significant small-study effects (p = 0.03).While this suggests some influence of small-study effects, the association remained statistically significant and clinically meaningful.

Fig. 10.

Funnel plot

Discussion

Our pooled analysis spanning more than 173 000 pregnancies shows that women who ultimately develop gestational diabetes mellitus (GDM) already exhibit markedly higher uric acid concentrations well before conventional screening: on average an excess of about 44 µmol L⁻¹ in the first half of gestation, and a more than two fold increase in the adjusted odds of GDM when uric acid exceeds study specific thresholds. Sequential exclusion of each comparison barely moved the point estimates, indicating that no single dataset—however extreme—drove the finding. These effect sizes mirror those of earlier but much smaller meta analyses, one of which combined 23 cohorts (≈ 105 000 participants) and produced a pooled OR of 2.58 (95% CI 1.89–3.52) [10]; adding six large 2023–2025 datasets therefore strengthens rather than inflates the signal. The association is detectable as early as the periconceptional period and within the first trimester, corroborating recent Chinese and Iranian cohort work that identified elevated uric acid before 14 weeks—or even within six months pre pregnancy—as an independent predictor of GDM [25]. Yet a Mendelian randomisation study published in 2024, which paired extensive GWAS data with observational cohorts, found no genetic evidence for a direct causal pathway, implying that early hyperuricaemia is more likely an epiphenomenon or downstream mediator than a primary driver of dysglycaemia [5]. Nevertheless, even if SUA is not directly causal in GDM pathogenesis, it remains a potentially valuable clinical risk marker. Effective biomarkers need not be causal; they only need to reliably identify at-risk individuals early enough to enable preventive interventions. Our analysis demonstrates that SUA can accomplish this goal, providing a simple, cost-effective tool for risk stratification that could complement existing screening protocols.

Physiologically, serum uric acid should decline during early gestation owing to oestrogen induced uricosuria and haemodilution; paradoxical elevation therefore flags underlying metabolic stress [25]. Experimental models offer several intersecting mechanisms: uric acid inhibits endothelial nitric oxide synthesis and impairs muscle glucose uptake, activates the NLRP3 inflammasome to produce IL 1β–mediated β cell damage, heightens hepatic gluconeogenesis via xanthine oxidase derived reactive oxygen species, and polarises adipose macrophages toward an inflammatory M1 phenotype, amplifying lipolysis and free fatty acid flux [15, 26]. Together these pathways sketch a biologically plausible chain in which modest hyperuricaemia precedes and potentiates insulin resistance, setting the stage for overt GDM once placental hormones surge in mid pregnancy.

Marked statistical heterogeneity persisted across both continuous and categorical analyses (overall I²≈95%, τ² >0.4, p < 0.0001), likely due to several contributing factors. Studies using enzymatic assays yielded smaller and less variable effect sizes compared to colorimetric assays, with subgroup differences being significant for both continuous (χ² = 21.31, df = 1, p < 0.0001) and categorical outcomes (χ² = 3.85, p = 0.0498). Variability in defining elevated SUA thresholds (300–360 µmol/L) further complicated between-study comparisons. While ethnicity minimally impacted heterogeneity in continuous analyses (χ² = 0.24, p = 0.63), it significantly moderated categorical outcomes (χ² = 23.76, p < 0.0001), potentially reflecting underlying genetic differences in renal urate transporters like URAT1 and GLUT9. Additionally, inconsistent confounder adjustments—particularly BMI, parity, dietary purine load, physical activity, and family history of diabetes—likely contributed to residual variance. Maternal age and publication year were not significant moderators, although maternal BMI exhibited a non-significant trend toward positive association. Differences in sampling trimester further introduced variability. Thus, key contributors to observed heterogeneity appear to be assay methodologies and inconsistent SUA thresholds, underscoring the need for harmonized protocols and comprehensive confounder adjustment in future studies. The use of different GDM diagnostic criteria across studies (IADPSG, Carpenter–Coustan, ADA, WHO 2013) may represent a significant and potentially unresolvable source of heterogeneity. These criteria differ in their glucose thresholds and testing protocols, potentially identifying different populations of women with varying degrees of glucose intolerance, which could contribute to the observed variation in effect sizes. Among these multiple sources of heterogeneity, we speculate that the variation in SUA threshold definitions and underlying population differences (particularly ethnicity and genetic variations in urate transport) may be the most significant contributors. The wide range of thresholds used to define ‘elevated SUA’ (from approximately 200 to > 300 µmol/L across studies) likely explains a substantial portion of the observed heterogeneity, combined with population-specific factors that influence baseline SUA levels and their relationship with GDM risk.

Clinically, first trimester uric acid testing is inexpensive, requires neither fasting nor glucose loading, and could be integrated into existing early pregnancy risk scores. Observational dose–response curves, particularly from the large-scale analysis by Zhao et al. [22], point to a pragmatic threshold near 220–230 µmol L⁻¹, above which GDM risk rises steeply. This threshold would need to be validated in diverse populations and integrated with existing risk stratification tools. The performance characteristics (sensitivity, specificity, positive and negative predictive values) of this threshold should be evaluated in prospective cohorts, and cost-effectiveness analyses are needed to determine its clinical utility compared to current screening approaches. Identifying such women could trigger intensified lifestyle counselling or continuous glucose monitoring months before the conventional 24 to 28 week oral glucose tolerance test. Pharmacological modulation is another frontier: xanthine oxidase inhibitors are already under investigation for pre eclampsia, and small proof of concept trials could clarify whether lowering uric acid before mid pregnancy can delay or blunt the emergence of GDM [27]. However, the safety profile of such interventions during pregnancy requires careful evaluation, as allopurinol and other urate-lowering therapies have limited safety data in pregnancy. Randomized controlled trials with appropriate safety monitoring are essential before any clinical implementation can be considered. Even if SUA is not directly causal in GDM pathogenesis, it remains a potentially valuable clinical risk marker. Effective biomarkers need not be causal; they only need to reliably identify at-risk individuals early enough to allow for preventive interventions. Our analysis suggests SUA can accomplish this goal, providing a simple, cost-effective tool for risk stratification that could complement existing screening protocols.

This review was prospectively registered, adhered to PRISMA 2020/MOOSE standards, used dual independent extraction, and applied Hartung–Knapp random effects models alongside multiple τ² estimators to gauge robustness. Several important limitations must be acknowledged. First and most fundamentally, all included evidence is observational in nature, which limits our ability to infer causality despite the consistent associations observed. The potential for residual confounding remains substantial, as unmeasured factors such as dietary patterns, physical activity levels, genetic predisposition, and socioeconomic status may influence both SUA levels and GDM risk. Second, GDM diagnostic thresholds varied across regions and time periods, potentially introducing heterogeneity in outcome definitions and limiting the generalizability of findings. Third, funnel plot asymmetry suggests mild small-study effects for categorical data, though trim-and-fill adjustment changed the summary OR by less than 10%. Fourth, the definition of ‘elevated SUA’ varied considerably across studies, making it difficult to establish universal clinical thresholds. Fifth, the high heterogeneity observed (I² >95%) could not be fully explained by our subgroup analyses, suggesting additional unmeasured sources of variation that may limit the reliability of pooled estimates. Sixth, most studies were conducted in specific ethnic populations (primarily Asian and Caucasian), potentially limiting the generalizability of findings to other ethnic groups. Finally, temporal changes in GDM diagnostic criteria and clinical practice over the study period may have introduced additional heterogeneity. Furthermore, to harmonize units across studies, we converted serum uric acid concentrations from mg/dL to µmol/L using a standard factor of 59.48, acknowledging that methodological differences between colorimetric and enzymatic assays may introduce additional analytical variation. Future work should follow multi ethnic, serial sampling cohorts with harmonised confounder adjustment and integrate metabolomic or genomic data to untangle whether uric acid is mediator, moderator or mere marker within the insulin resistance cascade; ultimately, randomised trials of urate lowering therapy initiated before 20 weeks are needed to test whether this readily measurable metabolite is simply predictive or genuinely modifiable in the pathogenesis of gestational diabetes.

Author contributions

Wang Y conceived of the study, and Wang Y and Miao X participated in its design and data analysis and statistics and Miao X and Hu AP helped to draft the manuscript. All authors read and approved the final manuscript.

Funding

This study did not receive any funding in any form.

Data availability

All data generated or analysed during this study are included in this article. Further enquiries can be directed to the corresponding author.

Declarations

Ethics approval and consent to participate

An ethics statement is not applicable because this study is based exclusively on published literature.

Consent for publication

Not applicable.

Clinical trial number

Not applicable.

Competing interests

The authors declare no competing interests.