CDKN2A/2B rs10811661 polymorphism and risk of gestational diabetes mellitus in Caucasian and Asian: a systematic review and meta-analysis

Abstract

Purpose

The correlation among the risk of gestational diabetes mellitus (GDM) and CDKN2A/2B rs10811661 polymorphism remains controversial, as previous studies have reported inconsistent findings. The potential relationship among the CDKN2A/2B rs10811661 polymorphism and GDM risk was examined by a meta-analysis.

Methods

In order to find relevant studies for our study, we performed an extensive search across several databases, including Embase, Pubmed, Web of science, Scopus, and China National Knowledge Infrastructure. Afterward, the link among the CDKN2A/2B rs10811661 polymorphism and the risk of GDM was then assessed using either random-effects models or fixed-effects to compute 95 percent confidence intervals (CIs) and pooled odds ratios (ORs).

Results

This meta-analysis comprised a total of 10 studies, and the results showed that the CDKN2A/2B rs10811661 polymorphism was linked to a decreased risk of GDM across all examined models. The pooled analysis demonstrated a substantial link, with the corresponding 95% CIs and the following ORs: Allele contrast: 0.68 (0.57–0.81), Homozygote 0.47,(0.32–0.69), Heterozygote 0.78, (0.68–0.88), Dominant model 0.69, (0.62–0.78), Recessive model 0.54(0.38–0.77). However, Trial Sequential Analysis (TSA) indicated that the Z-curve did not cross monitoring boundaries and the required information size (RIS) was not reached, suggesting that the evidence may be preliminary and should be interpreted with caution.

Conclusion

According to the current meta-analysis, the CDKN2A/2B rs10811661 variant may serve as a potential genetic biomarker for GDM, but additional large-scale studies are needed to confirm these findings.

Trial registration

The study was registered on PROSPERO (https://www.crd.york.ac.uk/prospero/), registration number: CRD42023484296.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-025-08439-6.

Introduction

During pregnancy, Gestational Diabetes Mellitus (GDM) is characterized by hyperglycemia and glucose intolerance [1]. According to the estimates, approximately 14 percent of pregnant women all around the world are glucose intolerant [2]. Fetal chronic metabolic diseases and adverse pregnancy outcomes are linked to GDM. GDM’s mechanism and etiology are yet to be known. Nevertheless, published evidences suggest that GDM is a clinical syndrome caused by both environmental as well as genetic factors [3, 4]. For GDM, genetic susceptibility is an important risk factor [5]. Recent meta-analyses have also highlighted the contribution of other loci such as KCNJ11 rs5219 in modulating GDM susceptibility [6], reinforcing the notion that multiple genetic pathways underlie inter-individual variation in GDM risk.

CDKN2B and CDKN2A genes, found on chromosome 9p21, encode the cyclin kinase inhibitors p15^INK4band p16^INK4a, respectively. These inhibitors prevent pancreatic islet beta cell proliferation by suppressing the activities of cyclin-dependent kinases 4 (CDK4) and CDK6 [7]. Animal studies have shown that mice lacking CDK4 suffer from diabetes because their pancreatic islet tissue and beta cell proliferation are diminished, whereas mice with CDK4 overexpression showed pathological proliferation of beta cells [8]. Studies have demonstrated a relationship between high expression of p15INK4b in mice and pancreatic islet development and insulin hyposecretion. One of the main contributing factors to type 2 diabetes is impaired insulin secretion [9]. Published data suggests that the rs10811661 variant in the CDKN2A/2B gene may enhance the susceptibility of developing type 2 diabetes by impacting the secretion function of pancreatic islet beta cells [10, 11]. The rs10811661 polymorphism is characterized by two alleles, T and C, which form three genotypes: TT, TC, and CC. In addition to genetic factors, accumulating evidence has demonstrated that inflammatory and oxidative stress pathways are closely involved in GDM and diabetes pathogenesis [12], suggesting a complex interplay between genetic predisposition and systemic metabolic stressors.Studies have indicated that the TT genotype may be associated with an increased risk of GDM, as it potentially influences insulin secretion and pancreatic beta cell function [13, 14].

Over the last twenty years, a plethora of studies was performed to investigate the potential association among the risk of GDM and CDKN2A/2B rs10811661 polymorphism across various racial groups. However, the findings from these studies have been inconclusive and inconsistent [13, 15–17]. To address this issue, a meta-analysis was conducted to thoroughly assess the association between the CDKN2A/2B rs10811661 polymorphism and GDM.

Materials and methods

Publication search

A detailed search was done in PubMed, Embase, Web of science, Scopus, and CNKI (China National Knowledge Infrastructure) up until February 19th, 2025. The search employed the following keywords: ‘‘CDKN2A/2B” or “rs10811661″ and ‘‘variant’’ or ‘‘polymorphism’’ and ‘‘GDM’’. The detailed search strategy is presented in the Supplementary Table 1.The examination followed the PRISMA guidelines, ensuring methodological rigor and transparency in the research process.

Inclusion and exclusion criteria

Studies had to fulfill some requirements in order to be considered for inclusion in the investigation: (i) they needed to assess the link among the CDKN2A/2B rs10811661 polymorphism and GDM risk using case–control designs, regardless of sample size; (ii) they had to provide enough information to compute OR with a 95% CI. Investigations that were conference abstracts, conference reports, reviews, meta-analyses, or lacked adequate data were not included.

Data extraction

Two reviewers (DS and AW) extracted data independently and adhered to predetermined inclusion criteria. Any conflicts were settled by conferring with an arbitrator (KY). The extracted data included details such as the author's last name, ethnicity, the year of publication, participant country, minor allele frequency (MAF), sample size, and genotyping methods. The Newcastle–Ottawa Scale (NOS), which assesses exposure, comparability, and selection criteria, was employed to assess the research’s quality. The maximum quality assessment score ranges from 0 to 9 stars. Generally, studies with a score of at least 6 points are typically regarded as high-quality studies in meta-analyses.

Statistical analysis

For this part, ORs and their corresponding 95% CIs were employed for assessing the correlation among the risk of GDM and CDKN2A/2B rs10811661 polymorphism. Subgroup analyses stratified by ethnicity, diagnostic criteria for GDM, genotyping method, and HWE status were performed. Heterogeneity analysis was performed using the Cochran Q statistic and the I² statistic. A p-value > 0.10 for the Q statistic and I² < 31% suggests a lack of heterogeneity between investigations [18], and the ORs were pooled using the fixed-effects model (Mantel–Haenszel method) [19]. The ORs were computed using the random-effects model (DerSimonian and Laird method) [20].

Egger's weighted regression approach and Begg's rank correlation approach were employed to examine publication bias. A funnel plot was visually inspected, and a p-value of less than 0.05 showed great publication bias [21, 22]. The statistical examination was performed through STATA software, version 18.0.

Trial sequential analysis

To analyze the reliability of the findings and the RIS (required information size), trial sequential analysis (TSA) was done. The RIS was computed based on 80% power of the study (β = 20%), 5% risk of type I error (α = 5%), and a two-sided boundary type was utilized. TSA software from the Copenhagen Trial Unit was utilized for this analysis.

Result

Characteristics of studies

After performing a comprehensive literature search, 15 articles were initially identified for examination. Three papers were eliminated after their titles and abstracts were reviewed. Full texts of 13 articles were obtained and carefully reviewed, resulting in the exclusion of 2 article due to insufficient data [23, 24], and 1 article not related to the CDKN2A/2B rs10811661 polymorphism [25]. Ultimately, 10 case–control studies examining GDM risk and the CDKN2A/2B rs10811661 polymorphism were included, in accordance with MOOSE guidelines [13–17, 26–29]. Figure 1 illustrates the process of choosing the study and literature searches.

Fig. 1.

Literature search and study selection procedures used for a meta-analysis of CDKN2A/2B rs10811661 genetic polymorphism and GDM

Table 1 presents an outline of the characteristics of the included investigations. Four included studies had people who were Caucasian, and six contained subjects who were Asian. These investigations were carried out in a number of places, including China, Denmark, Egypt, Korea, Poland, and Romania.

Table 1.

Characteristics of studies included in this meta-analysis

Author	Year	Country	Ethnicity	Sample	Genotyping Methods	Diagnostic criteria for GDM	MAF in Controls	HWE
Cho	2009	Korea	Asian	869/632	Taqman	NDDG(1979)	0.488	0.822
Lauenborg	2009	Denmark	Caucasian	283/2446	Taqman	WHO(1999)	0.166	0.664
Lv	2012	China	Asian	120/100	PCR–RFLP	ADA(2010)	0.490	0.596
Deng	2016	China	Asian	500/100	PCR–RFLP	WHO(2011)	0.480	0.112
Hu	2016	China	Asian	65/65	PCR–RFLP	ADA(2010)	0.476	0.271
Tarmowski	2017	Poland	Caucasian	204/207	Taqman	IADPSG(2010)	0.193	0.145
Liu.J	2018	China	Asian	120/120	PCR–RFLP	IADPSG(2010)	0.471	< 0.001
Noury	2018	Egypt	Caucasian	47/51	PCR–RFLP	ADA(2017)	0.156	0.064
Liu.N	2019	China	Asian	459/571	MassArray	IADPSG(2010)	0.442	0.461
Muntean	2025	Romania	Caucasian	71/142	Taqman	IADPSG(2010)	0.190	0.118

Abbreviations: PCR–RFLP polymerase chain reaction-restriction fragment length polymorphism, MAF minor allele frequency, GDM gestational diabetes mellitus, NDDG national diabetes data group, WHO world health organization, ADA American diabetes association, IADPSG international association of the diabetes and pregnancy study groups

Quantitative synthesis

Quantitative synthesis was performed using the data from these ten studies, which included overall 2738 cases and 4434 controls. The meta-analysis revealed a huge decrease in GDM risk linked to the CDKN2A/2B rs10811661 polymorphism across all models including: homozygote comparison, allele contrast, recessive model, heterozygote comparison, and dominant model (Table 2). Figure 2 illustrates the forest plots illustrating the link among the GDM risk and CDKN2A/2B rs10811661 polymorphism. The NOS scores, which show the quality of the investigations added, are summarized in Table 3. In addition, risk of bias was further assessed using the Q-Genie tool (Supplementary table S3). Across the 10 studies, total Q-Genie scores ranged from 33 to 47, with most investigations falling in the moderate-to-high quality range. Domains such as “rationale of study” and “appropriateness of inferences drawn” generally scored higher, whereas “non-technical classification of the exposure” and “sample size and power” tended to show lower ratings, reflecting limitations in exposure definition and study precision.

Table 2.

Quantitative analyses of the CDKN2A2B rs10811661 polymorphism on the GDM risk

Genetic model			Allele contrast		Homozygote		Heterozygote		Dominant Model		Recessive Model
Variables	Sample size		C vs. T		CC vs. TT		TC vs. TT		CC + TC vs. TT		CC vs. TC + TT
	Na	Case/control	OR(95%CI)	Pvalueb	OR(95%CI)	Pvalueb	OR(95%CI)	Pvalueb	OR(95%CI)	Pvalueb	OR(95%CI)	Pvalueb
Total	10	2738/4434	0.68(0.57,0.81)	0.001	0.47(0.32,0.69)	0.001	0.78(0.68,0.88)	0.537	0.69(0.62,0.78)	0.102	0.54(0.38,0.77)	0.001
Caucasian	4	605/2846	0.88(0.60,1.28)	0.031	0.83(0.31,2.21)	0.084	0.79(0.63,1.00)	0.345	0.80(0.64,0.99)	0.160	0.95(0.42,2.13)	0.085
Asian	6	2133/1588	0.62(0.51,0.75)	0.011	0.39(0.27,0.56)	0.030	0.77(0.66,0.90)	0.465	0.65(0.57,0.75)	0.198	0.45(0.32,0.62)	0.019

^aNumber of comparisons

^bP value of Q-test for heterogeneity test. Random-effects model was used when P value for heterogeneity test < 0.05; otherwise, fixed-effects model was used

Fig. 2.

Forest plots of ORs with 95% CIs for CDKN2A/2B rs10811661 polymorphism and GDM risk. A, B, C, D, E show allelic, homozygous, heterozygous, dominant model and recessive models, respectively

Table 3.

Quality assessment of case–control studies included in this meta-analysis^a

Study	Adequate definition of cases	Representativeness of cases	Selection of control	Definition of control	Control for important factor or additional factorb	Exposure assessment	Same method of ascertainment for cases and controls	Nonresponse ratec	Total quality scoresd
Cho	★	★	★	★	★	-	★	★	7
Lauenborg	★	★	★	★	★	-	★	★	7
Lv	★	★	★	★	★★	-	★	★	8
Deng	★	★	★	★	★	-	★	★	7
Hu	★	★	★	★	★★	-	★	★	8
Tarmowski	★	★	★	★	★★	-	★	★	8
Liu.J	★	★	-	★	★	-	★	★	6
Noury	★	★	★	★	★★	-	★	★	8
Liu.N	★	★	★	★	★	-	★	★	7
Muntean	★	★	★	★	★★	-	★	★	8

^aA study can be awarded a maximum of one star for each numbered item except for the item Control for most important factor or second important factor

^bA maximum of two stars can be awarded for Control for most important factor or second important factor. Studies that controlled for maternal age received one star, whereas studies that controlled for high risk factor (diabetes or pre-pregnancy body mass index or family history of hypertension) received one additional star

^cOne star was awarded if there was no significant difference in the response rate between control subjects and cases in the chi-square test (P > 0.05)

^dThe studies are considered to be low-quality, when the sores were lower than six stars in quality assessment

Heterogeneity analysis

Heterogeneity analysis revealed substantial heterogeneity among studies in recessive model (P_{heterogeneity} = 0.001), allele contrast (P_{heterogeneity} = 0.001), and homozygote comparison (P_{heterogeneity} = 0.001). A random-effects model was used for OR and 95% CI estimation. Galbraith plot analyses identified two studies [15, 16], as sources of heterogeneity (Fig. 3). The removal of these two outlier studies led to a huge decline in heterogeneity. For the purpose of estimating OR and 95% CI, a fixed-effects model was employed following the removal of the outlier studies. The heterogeneity significantly decreased in recessive model (P_{heterogeneity} = 0.263), allele contrast (P_{heterogeneity} = 0.295), and homozygote comparison (P_{heterogeneity} = 0.357), while the overall conclusion remained unchanged.

Fig. 3.

Galbraith plots for heterogeneity test of CDKN2A/2B rs10811661 polymorphism

To further explore heterogeneity, we conducted subgroup analyses stratified by diagnostic criteria for GDM, genotyping methods, and HWE status. These analyses demonstrated that heterogeneity was particularly high under the IADPSG 2010 criteria (e.g., I² = 81.6% in allelic contrast; I² = 74.4% in homozygote comparison), whereas studies using ADA criteria showed negligible heterogeneity. The detailed thresholds and definitions applied across diagnostic guidelines are summarized in Supplementary table S2. By genotyping method, heterogeneity was pronounced in TaqMan-based studies (I² ≈ 70–77% in allelic and homozygote models), moderate in PCR–RFLP studies. With respect to HWE, studies in Hardy–Weinberg equilibrium contributed most of the heterogeneity (I² = 63.0% in allelic model; I² = 64.9% in recessive model). Between-subgroup heterogeneity was statistically significant for diagnostic criteria (p = 0.047 in allelic model), genotyping method (p = 0.036 in allelic model), and HWE status (p = 0.017 in allelic model). The corresponding stratified analyses are presented in the Supplementary Figures: S4–S8 (stratified by diagnostic criteria for GDM), S9–S13 (stratified by genotyping method), and S14–S18 (stratified by HWE status).

Cumulative and sensitivity analyses

The results are stable, as shown by the cumulative meta-analysis (Supplementary figure S1) and sensitivity analyses (Supplementary figure S2).

Publication bias

To assess potential publication bias, Begg’s rank correlation test and Egger’s regression asymmetry test were performed. Begg’s funnel plots did not reveal obvious asymmetry (Fig. 4). The exact p-values for Begg’s test and Egger’s test, respectively, were as follows: allele contrast (Begg’s p = 0.47; Egger’s p = 0.74), homozygote comparison (Begg’s p = 0.59; Egger’s p = 0.61), heterozygote comparison (Begg’s p = 0.17; Egger’s p = 0.38), dominant model (Begg’s p = 0.72; Egger’s p = 0.93), and recessive model (Begg’s p = 0.47; Egger’s p = 0.50).

Fig. 4.

Begg’s funnel plot for publication bias test

Given the limited number of eligible studies (k = 10), the statistical power of Begg’s and Egger’s procedures to detect small-study effects is inherently constrained. Accordingly, these findings should not be interpreted as definitive evidence of the absence of publication bias; rather, they indicate that no statistically significant evidence of small-study effects was observed. To further corroborate these results, Peters’ test was performed as a sensitivity analysis tailored for binary outcomes, which similarly yielded non-significant results across all genetic models (allele contrast p = 0.35; homozygote comparison p = 0.13; heterozygote comparison p = 0.46; dominant model p = 0.64; recessive model p = 0.36).

Trial sequential analysis

TSA was employed to further investigate the relationship among CDKN2A/2B rs10811661 and GDM risk. The results indicated that the cumulative Z value (Z-curve) did not cross the TSA boundary, suggesting that the cumulative amount of information may not have reached the RIS (Supplementary figure S3). This indicates that the traditional meta-analysis may offer a false positive conclusion, and it would be necessary to do additional research to verify the association.

Discussion

This meta-analysis investigated the correlation among the GDM risk and CDKN2A/2B rs10811661 polymorphism based on ten case–control studies involving 2738 GDM subjects and 4434 control cases. It is the first meta-analysis that examined this correlation. The overall results demonstrated a significantly decreased GDM risk linked to the CDKN2A/2B rs10811661 polymorphism. Ethnicity-based subgroup analysis showed a substantially decreased risk among Asian people in all models. However, only the dominant model showed a significant reduction in risk among Caucasian groups. It is unclear why there is a discrepancy in this. One possible justification is that the genotype distribution of CDKN2A/2B rs10811661 TT differs between Caucasian and Asian populations. The study found that the distribution of CDKN2A/2B rs10811661 TT genotype frequency in the 3'untranslated region was 68.1% (364/534) among Caucasian patients, while among Asian patients, it was 38.2% (816/2133). This difference in genotype distribution may contribute to the varying associations among GDM risk and CDKN2A/2B rs10811661 polymorphism in different ethnic groups. The small number of Caucasian patients in the meta-analysis sample size may also be another possible explanation. The limited Caucasian patients may have resulted in insufficient statistical power to investigate the true associations.

Beyond differences in sample size and allele frequency, a constellation of population-specific determinants may also underlie the observed ethnic disparities. First, linkage disequilibrium (LD) structures surrounding the rs10811661 locus differ markedly between Asian and Caucasian populations, potentially influencing the extent to which this polymorphism serves as a proxy for causal variants. Second, environmental and lifestyle exposures—such as habitual dietary patterns, levels of physical activity, prevalence of obesity, and gestational weight trajectories—vary substantially across populations and may synergize with genetic susceptibility to modulate GDM risk. Third, heterogeneities in healthcare infrastructure, diagnostic criteria, and screening protocols across regions could further accentuate inconsistencies in reported associations. Taken together, these divergences in genetic architecture and environmental context underscore that the protective association observed in Asian populations may not extrapolate directly to Caucasian cohorts. Elucidating these ethnic-specific mechanisms will require rigorously designed, adequately powered, and multi-ethnic investigations that explicitly integrate gene–environment interactions and population genetic characteristics.

The CDKN2A/2B rs10811661 polymorphism is thought to modulate GDM susceptibility through several interconnected molecular pathways. Situated within the 9p21 locus adjacent to CDKN2A/2B and ANRIL, this variant has been reported to influence β-cell function via the cyclin-dependent kinase inhibitors p16^INK4a and p15^INK4b [30]. The T risk allele has been associated with reduced CDKN2A/2B expression, which may compromise β-cell proliferation and insulin secretory adaptation during pregnancy—a process of particular pathophysiological relevance given the heightened insulin resistance characteristic of GDM [31]. Beyond direct β-cell effects, the locus also regulates ANRIL, a long non-coding RNA that modulates CDKN2A/2B expression epigenetically through chromatin remodeling, an interaction potentially influenced by pregnancy-related hormones such as placental lactogen [32, 33]. In addition, reduced CDKN2A/2B activity may alter adipocyte differentiation and systemic insulin sensitivity, thereby amplifying metabolic stress during gestation [34]. Finally, ethnic heterogeneity in observed associations may reflect population-specific ANRIL isoforms or differences in 9p21 epigenetic regulation [35]; for example, stronger ANRIL suppression by the T allele has been reported in Asian populations, potentially contributing to the more pronounced protective association observed in this group [36]. Collectively, these multilayered mechanisms—encompassing β-cell biology, epigenetic regulation, and systemic metabolic effects—provide a coherent biological rationale linking rs10811661 to reduced GDM risk.

Beyond this locus, other candidate biomarkers and genetic variants have also been systematically investigated. For example, Galectin-3 has recently been associated with GDM risk through meta-analysis [37], while circulating nitric oxide levels were reported to be dysregulated in polycystic ovary syndrome, a recognized risk factor for GDM [38], underscoring the relevance of broader metabolic and inflammatory pathways in GDM pathophysiology. In addition, recent systematic reviews on diabetic complications, such as retinopathy, have identified biomarkers like pentraxin-3 as potential diagnostic tools [39]. Although not specific to GDM, these findings highlight the expanding biomarker landscape in diabetes, which may ultimately inform more precise prediction and management strategies for GDM.

Our results align with the meta-analysis by Guo et al. (2018) [40], as both studies identified a significant association between CDKN2A/B rs10811661 polymorphisms and GDM risk under the allelic model (C vs. T). However, our study advances beyond these findings by incorporating a multi-model analytical approach, including homozygous, heterozygous, dominant, and recessive genetic models. In contrast to the moderate association reported by the previous meta-analysis, our pooled analysis revealed a more robust and consistent association across all genetic models, underscoring a stronger genetic contribution to GDM susceptibility. A key methodological strength of our study is the inclusion of 10 studies, doubling the sample size of previous meta-analysis. This expansion significantly enhances the statistical power, reliability, and precision of our findings. Furthermore, by employing multiple genetic models, our study provides a more comprehensive evaluation of the relationship between CDKN2A/B polymorphisms and GDM.

Heterogeneity among the included studies is a critical issue in meta-analyses, as it can impact the dependability of the findings. In our primary analyses, substantial heterogeneity was observed in several genetic models. While Galbraith plots identified two potential outlier studies whose exclusion reduced heterogeneity, we have now reframed these results as sensitivity analyses rather than default analyses. To further address this issue, we performed additional stratified analyses based on prespecified study-level characteristics, including ethnicity, diagnostic criteria for GDM, genotyping methods, and HWE status. These subgroup analyses revealed that heterogeneity was particularly pronounced under the IADPSG 2010 diagnostic criteria (I² > 70% in both allelic and homozygote models), whereas heterogeneity was minimal under ADA or NDDG criteria. In addition, genotyping method contributed to variability, with heterogeneity highest in TaqMan-based studies (I² ≈ 70–77%) and moderate in PCR–RFLP studies. Moreover, violation of Hardy–Weinberg equilibrium in one study was associated with increased between-group heterogeneity. Between-subgroup heterogeneity was statistically significant for diagnostic criteria (p = 0.047 in allelic model), genotyping method (p = 0.036 in allelic model), and HWE status (p = 0.017 in allelic model), suggesting that these methodological differences are plausible contributors to the observed inconsistency. Taken together, these findings indicate that heterogeneity is partly attributable to study design and methodological quality, beyond ethnicity alone.

Publication bias, which arises from the selective publication of studies, is another concern in meta-analyses. To address this issue, Begg's and Egger's tests were done, and funnel plots were examined. Based on the outcomes of these examinations and the graphical representations of the data, it can be inferred that there is no evidence of any publication bias in the present study.

Despite the significant findings, this investigation has certain limitations. First, the inclusion of a small number of studies and the limited number of samples may have limited the statistical power to examine the associations comprehensively. Second, the analysis utilized unadjusted odds ratios (ORs) due to the lack of availability of adjusted ORs in some studies. Furthermore, in cases where adjusted ORs were provided, they were adjusted for a variety of variables, including race, age, or body mass index (BMI), which could introduce confounding effects. Lastly, there was a great deal of variation among the included investigations, and one study was not in Hardy–Weinberg equilibrium (HWE). Although we explored heterogeneity through stratified analyses by ethnicity, diagnostic criteria, genotyping methods, and HWE status, residual unexplained heterogeneity cannot be ruled out. The outlier exclusions identified by Galbraith plots should therefore be regarded only as exploratory sensitivity analyses. Moreover, due to the limited number of studies available per stratum, we were unable to conduct formal meta-regression to quantify the contribution of each moderator. Methodological factors such as inconsistent diagnostic criteria, differences in genotyping techniques, and population stratification across studies may still contribute to heterogeneity and variability in effect estimates. Importantly, although Trial Sequential Analysis (TSA) was undertaken, the required information size (RIS) threshold was not attained. This indicates that the cumulative evidence may still be underpowered, raising the possibility of spurious associations. Accordingly, the present findings should be interpreted with heightened caution until validated by future high-quality research.

Implications for future research. To reduce confounding and enhance comparability, future primary studies should report covariate-adjusted ORs using a standardized core set of confounders (at minimum: maternal age, pre-pregnancy BMI, parity, family history of diabetes, prior GDM/PCOS, smoking, and socioeconomic indicators), apply uniform GDM diagnostic criteria and timing of OGTT, and include ancestry principal components to address population stratification. Pre-registered analysis plans with transparent covariate selection, and the routine reporting of both crude and adjusted estimates, would facilitate evidence synthesis. Where feasible, individual participant data (IPD) meta-analyses will allow harmonized adjustment, exploration of non-linearities, and formal assessment of effect modification by BMI, age, and co-inherited loci.

In conclusion, this meta-analysis provides evidence that the CDKN2A/2B rs10811661 variant appears to be associated with a reduced risk of GDM, particularly in Asian populations. Nevertheless, as the TSA demonstrated that the RIS has not yet been achieved, the statistical robustness of the current evidence remains insufficient. To strengthen the reliability and generalizability of these observations, further large-scale, rigorously designed, and ethnically diverse studies are imperative.

Authors’ contributions

DS and AW interpreted and analyzed the data, and drafted the manuscript. DS and KY concepted and designed the study and revised the manuscript. All authors contributed to and approved the final manuscript.

Funding

This research did not receive any specifc grant from funding agencies in the public, commercial, or not-for-proft sectors.

Data availability

Original data generated and analyzed during this study are included in this published article or supplementary material.

Declarations

Ethics approval and consent to participate

No ethical approval was required for this review, as all data were already published in peer-reviewed journals.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.