Association between the null polymorphisms of GSTT1 and GSTM1 and the risk of gestational diabetes mellitus: a systematic review and meta-analysis Authors: Zitong Liu, Chenxi Ji, Rui Wang, Qiyue Liu, Xiaoyin Guo, Li Fan, Youyi Kong, Lingxi Chen, Xiaoqin Yang, Shangshang Gao Published: 2025 ABSTRACT BackgroundThe relationship between the null polymorphisms of the GSTT1 and GSTM1 genes and the risk of gestational diabetes mellitus has been previously investigated. However, these scattered evidence remained controversial. Therefore, this systematic review and meta-analysis was performed to synthesize previous findings and statistically estimate whether these polymorphic variants are associated with the risk of this common pregnancy complication.MethodsA systematic literature search was conducted in PubMed, Clarivate Web of Science, Elsevier Scopus, Cochrane Library, and EBSCOhost (from inception to February 28, 2025), using a predefined search strategy. The methodological quality of each study was assessed by using the Newcastle-Ottawa Scale (NOS). Pooled odds ratio (OR) and 95% confidence interval (95% CI) were estimated to measure the effect size. In addition, heterogeneity, sensitivity, and publication bias were also examined. All statistical tests were performed using the R language (version 4.4.2) and STATA software (version 14.2). This study followed the PRISMA 2020 statement, and the protocol was prospectively registered in PROSPERO (CRD420250621255).ResultsFor the deletion polymorphism of GSTT1, seven studies with 1012 cases and 1081 controls were integrated into the pooled estimation. For the null variant of GSTM1, seven datasets with 1012 diabetic and 1081 control individuals were included in the meta-analysis. Pooled estimates showed statistical significance for the null polymorphisms of GSTT1 (OR: 1.43, 95% CI: 1.08–1.89, P = 0.01) and GSTM1 (OR: 2.01, 95% CI: 1.68–2.39, P < 0.01). The symmetric funnel plot shape and statistical tests of publication bias suggested no substantial file drawer problem. However, the pooled estimates of the GSTT1 null polymorphism lost statistical significance in the leave-one-out sensitivity analysis and Galbraith plot-guided heterogeneity adjustment.ConclusionThe results suggested that the null polymorphisms of GSTT1 and GSTM1 may be associated with an increased risk of gestational diabetes mellitus. Notably, the association for the GSTT1 null polymorphism was not robust and warrants further confirmation.Clinical trial numberNot applicable.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12902-025-02103-y. Background Gestational diabetes mellitus is defined as a form of hyperglycemia first recognized during pregnancy. This prevalent antepartum condition affects approximately 14% of pregnancies worldwide (accounting for over 18 million births annually) [1, 2]. It is associated with a significant chance of hypertension after the index pregnancy [3], an increased likelihood of developing type 2 diabetes mellitus in mothers later on [4], and a high risk of maternal obesity [5]. Moreover, this common medical complication in pregnancy poses an amplified possibility of adverse neonatal outcomes, including respiratory distress syndrome, large for gestational age, neonatal jaundice, and admission to the neonatal intensive care unit [6]. Several sources of risk that contribute to this endocrine complication of pregnancy have been identified, such as ethnicity (e.g., Asians) [7], obesity [8, 9], physical inactivity [10], and advanced maternal age [7, 11]. However, many factors (e.g., genetic determinants) that affect the susceptibility of gestational diabetes mellitus remain unclear or unknown. Previous studies have indicated that the evolution and outcomes of gestational diabetes mellitus arise from a disparity between cellular oxidants and antioxidants [12]. The noted manifestations can be ascribed to an excessive production of free radicals [e.g., reactive oxygen species (ROS)] inside the cells, together with impaired antioxidant defenses (e.g., functioning of antioxidant enzymes) [12, 13]. Glutathione S-transferases (GSTs) are recognized as a multigene enzyme family of phase II metabolic isoenzymes. These metabolic multigene catalysts serve a pivotal function in protecting cells by mediating detoxification and clearance of free radicals [14]. Therefore, the genetic defects in GST family members may compromise cellular antioxidant defenses, consequently facilitating the accumulation of free radicals and exacerbating oxidative stress in the context of gestational diabetes mellitus. Glutathione S-Transferase Mu 1 (gene symbol: GSTM1, location: 1p13.3, Exon count: 8) and Glutathione S-Transferase Theta 1 (gene symbol: GSTT1, location: 22q11.23, Exon count: 6) are the genes that encode the Mu and Theta classes of GST enzymes, respectively. These two enzymes eliminate intracellular ROS through the GSH conjugation and inhibit oxidative stress. Accumulated in vitro and in vivo evidence supported that GSTT1/GSTM1 deficiency is associated with increased intracellular ROS levels [15, 16]. In addition, both GSTM1 and GSTT1 were found to be significantly downregulated at the transcriptional level in venous blood leucocytes of gestational diabetic women compared to control individuals [17]. The deletion polymorphisms of the GSTT1 and GSTM1 genes, also referred to as null variants, can result in the absence of these two functional enzymes. Two meta-analyses suggested that the GSTM1 and GSTT1 null genotypes are associated with increase the risk for type 2 diabetes mellitus [18, 19]. Previous studies demonstrated that gestational diabetes mellitus and type 2 diabetes mellitus share a similar genetic background [20, 21]. The above evidence raised concerns about the effects of the two null polymorphisms on the predisposition to gestational diabetes mellitus. Several previous investigations have attempted to assess the association between the null polymorphisms of the GSTT1 and GSTM1 genes and the risk of gestational diabetes mellitus [17, 22–27]. However, these individual studies were plagued by the constraint of sample size, resulting in suboptimal statistical power and generalizability of results. By aggregating the scattered data, the meta-analysis method could substantially increase sample size and further strengthen statistical detection power. To obtain a more precise estimate of the association, we conducted this systematic review and meta-analysis to summarize the existing epidemiological evidence on the genetic impact of the polymorphisms in the GSTM1 and GSTT1 genes (null and present) on the risk of gestational diabetes mellitus. Materials and methods Protocol design and registration This study strictly complied with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) recommendations (Table S1) [28]. Two investigative reviewers independently performed the literature search and selection, data extraction, statistical analysis, results interpretation and visualization, and summarization of their findings. All discrepancies and differences in opinion were resolved through internal discussions. If necessary, a third investigative reviewer (the corresponding authors of this manuscript) would be consulted to resolve any discrepancies and finalize the decision. The protocol of this meta-analysis was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO, https://www.crd.york.ac.uk/PROSPERO/) under the registration number CRD420250621255. Data sourcing and search strategies To identify relevant literature, the following databases were searched: PubMed (https://pubmed.ncbi.nlm.nih.gov/), Clarivate Web of Science Platform (https://www.webofknowledge.com), Elsevier Scopus (https://www.scopus.com/), Cochrane Library (https://www.cochranelibrary.com/), and EBSCOhost (https://search.ebscohost.com/). The literature search was finally updated on 28 February 2025. Boolean queries employing standard logical operators were structured by linking official symbols, full names, and aliases for the GSTM1 and GSTT1 genes (“GST M1”, “GSTM1”, “glutathione S-transferase mu 1”, “GST1-1”, “GTH4”, “GTM1”, “H-B”, “MU”, “MU-1”, “GSTM1a-1a”, “GSTM1b-1b”, “GST T1”, “GSTT1”, and “glutathione S-transferase theta 1”), entire name and abbreviation for gestational diabetes mellitus (“gestational diabetes”, “GDM”, and “gestational diabetes mellitus”), and keywords for polymorphism (“polymorphism”, “variant”, and “polymorphic”). The detailed Boolean search string used was as follows: (“polymorphism” OR “variant” OR “polymorphic”) AND (“gestational diabetes” OR “GDM” OR “gestational diabetes mellitus”) AND (“GST M1” OR “GSTM1” OR “glutathione S-transferase mu 1” OR “GST1-1” OR “GTH4” OR “GTM1” OR “H-B” OR “MU” OR “MU-1” OR “GSTM1a-1a” OR “GSTM1b-1b” OR “GST T1” OR “GSTT1” OR “glutathione S-transferase theta 1”). Inclusion and exclusion criteria Studies were deemed eligible if they reported the association between the null polymorphism of GSTT1 or GSTM1 and the risk of gestational diabetes mellitus. To qualify, studies should adopt a retrospective or prospective cohort, cross-sectional, or case-control design and be published in peer-reviewed journals in either English or Chinese. In addition, to be included, studies had to provide data on GSTT1 or GSTM1 genotypes in both patients with gestational diabetes mellitus and non-diabetic subjects. Non-English/Chinese articles were not considered as they fell outside the predefined language scope of this analysis (which is beyond the investigators’ language capabilities). Additionally, non-human studies (e.g., preclinical, animal studies), non-research articles (e.g., case reports, letters, comments, review articles, abstracts), studies lacking publication in peer-reviewed outlets, and studies without full-text or valid data were excluded due to their inability to provide relevant data for this meta-analysis. Study selection procedure The initial search results from the five literature databases were exported to Clarivate EndNote citation and reference management software (version: 21.5) for screening, and duplicates were eliminated using the software’s capabilities and a further manual check. Then, two investigative reviewers independently reviewed the titles and abstracts of all retrieved articles to select relevant studies according to the inclusion and exclusion criteria listed above. Next, full texts of the remaining publications were assessed to confirm their relevance to the research topic and alignment with the predefined inclusion criteria. Data extraction Two investigative reviewers independently performed the data extraction and cross-checked the information to ensure accuracy. All data were extracted into a predefined spreadsheet in the MacOS iWork Numbers software. The following data were extracted: the family name of the first author, publication year of the data source, country, genotyping method, sample tissue source, and genotype data for both the gestational diabetes mellitus case and non-diabetic control groups. Study quality and risk of bias The Newcastle-Ottawa Scale (NOS) for non-randomised studies in meta-analysis (https://www.ohri.ca/programs/clinical_epidemiology/oxford.asp) was used to evaluate the risk of bias of the included studies, by assessing the quality of their design and methodology across three essential domains: the selection of case and control groups, the comparability between study groups, and the ascertainment of outcomes or exposure (e.g., polymorphism genotype). Scores between seven and nine stars indicated a high methodological quality (a low risk of bias), while scores of four to six stars suggested a moderate quality (a medium risk of bias). In comparison, scores of zero to three stars reflected a low quality (a high risk of bias). This process was independently completed by two reviewers. Statistical analysis and meta-analysis Above all, the degree of between-study heterogeneity was assessed by the P-value in the Cochrane Q statistical test and the inconsistency index (I2 statistics) in Higgins’ test [29]. If significant heterogeneity was detected (P < 0.10 or I2 >50%), odds ratio (OR) and corresponding 95% confidence interval (95% CI) were estimated using the random-effects model (DerSimonian-Laird method) [30]. Otherwise, pooled estimates were calculated using the fixed-effect model (Mantel-Haenszel method) [31]. The Galbraith plot was generated to detect the source of between-study heterogeneity [32–34]. Pooled estimates were recalculated after excluding identified outliers contributing to the heterogeneity. Subgroup analysis was performed to evaluate the potential influence of the study characteristics (ethnicity and sample size) on the outcome (risk of gestational diabetes mellitus). Forest plots were generated to summarize the results of individual studies and pooled estimates. “Leave-one-out” sensitivity analysis was performed to sequentially exclude individual datasets to identify those driving statistical significance and assess the robustness of the pooled results. The presence of publication bias was statistically inspected through the examination of funnel plot asymmetry by the Egger’s linear regression test and Begg’s rank correlation test (P < 0.05 indicating significant publication bias) [35, 36]. Standard methods for meta-analysis and supporting were implemented by “meta” (version: 8.0–2, https://cran.r-project.org/web/packages/meta/index.html) and “metafor” (version: 4.8-0, https://cran.r-project.org/web/packages/metafor/index.html) packages of R language (version: 4.4.2). In addition, the main results of this meta-analysis were also verified using STATA statistical software (version: 14.2, StataCorp LP, College Station, TX, USA). Results Study characteristics The flow diagram of the study selection for this systematic review and meta-analysis is illustrated in Fig. 1. The comprehensive literature search in five mainstream databases initially found a total of 58 pieces of literature. After a thorough and careful literature review, 52 studies were excluded as 30 were duplicate records; 19 studies did not address the association between the null polymorphisms of the GSTT1 and GSTM1 genes and gestational diabetes mellitus risk; two were not human research; one was a review article. The remaining six studies were subjected to further review. Additionally, an extra study was identified by exploring Google Scholar citations of these six articles. Finally, seven case-control studies with 2093 participants (1012 diabetic cases and 1081 controls) were included for the systematic review and meta-analysis. Fig. 1Workflow of the literature selection procedure Workflow of the literature selection procedure The publication years of the selected literature ranged from 2014 to 2024. The genotyping techniques varied across the studies, with the most common being the polymerase chain reaction (PCR) method (five studies); [17, 22, 23, 25, 26] the remaining two studies utilized the polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) technique [24, 27]. A total of two articles reported participants from the northeast African country of Egypt [22, 23], and five studies originated from Asian nations [17, 24–27]. In addition, five studies reported a sample size of 150 or larger [17, 23–25, 27], and two had fewer than 150 participants [22, 26]. A detailed summary of the characteristics of the eligible literature is presented in Table 1. Table 1Characteristics of the included articlesStudyGenotypingCountryAge (Case/Control)BMI (kg/m2) (Case/Control)Glucose level (Case/Control)Orhan 2014PCRTurkey(32.2 ± 5.1)/(28.6 ± 5.7)(30.2 ± 4.6)/(28.7 ± 5.0)First hour blood glucose value in 50 g glucose challenge test (mg/dl): (187 ± 27)/(104 ± 16)Li 2015PCR-RFLPChina(32.2 ± 10.4)/(27.4 ± 11.5)(30.6 ± 6.1)/(27.7 ± 5.3)FPG (mg/dl): (92.3 ± 24.3)/(75.7 ± 15.3)Qiu 2016PCR-RFLPChina(31.5 ± 9.3)/(28.2 ± 9.8)(30.7 ± 5.6)/(27.2 ± 4.8)NABarseem 2021PCREgypt(30.2 ± 5.4)/(26.4 ± 5.6)NAGlycosylated hemoglobin (%): (6.8 ± 1.3)/(5.2 ± 0.4)Barseem 2022PCREgypt(32.2 ± 5.4)/(30.4 ± 4.6)NA2 h. PP (mg/dl): (201.8 ± 80.4)/ (70 ± 4.3)Montazeri-Najafabady 2022PCRIran(32.5 ± 5.7)/(30.5 ± 5.4)(27.9 ± 4.3)/(26.1 ± 4.1)FBS (mg/dl): (101.4 ± 33.6)/(80.7 ± 15.9)Shamsad 2024PCRIndia(30.7 ± 5.2)/(28.3 ± 5.2)(25.3 ± 2.9)/(23.2 ± 2.7)FPG (mg/dl): (87.2 ± 10.2)/(75.7 ± 12.9)FPG: Fasting Plasma Glucose. FBS: Fasting Blood Sugar. 2 h.PP: 2 h post prandial blood glucose Table 1Characteristics of the included articlesStudyGenotypingCountryAge (Case/Control)BMI (kg/m2) (Case/Control)Glucose level (Case/Control)Orhan 2014PCRTurkey(32.2 ± 5.1)/(28.6 ± 5.7)(30.2 ± 4.6)/(28.7 ± 5.0)First hour blood glucose value in 50 g glucose challenge test (mg/dl): (187 ± 27)/(104 ± 16)Li 2015PCR-RFLPChina(32.2 ± 10.4)/(27.4 ± 11.5)(30.6 ± 6.1)/(27.7 ± 5.3)FPG (mg/dl): (92.3 ± 24.3)/(75.7 ± 15.3)Qiu 2016PCR-RFLPChina(31.5 ± 9.3)/(28.2 ± 9.8)(30.7 ± 5.6)/(27.2 ± 4.8)NABarseem 2021PCREgypt(30.2 ± 5.4)/(26.4 ± 5.6)NAGlycosylated hemoglobin (%): (6.8 ± 1.3)/(5.2 ± 0.4)Barseem 2022PCREgypt(32.2 ± 5.4)/(30.4 ± 4.6)NA2 h. PP (mg/dl): (201.8 ± 80.4)/ (70 ± 4.3)Montazeri-Najafabady 2022PCRIran(32.5 ± 5.7)/(30.5 ± 5.4)(27.9 ± 4.3)/(26.1 ± 4.1)FBS (mg/dl): (101.4 ± 33.6)/(80.7 ± 15.9)Shamsad 2024PCRIndia(30.7 ± 5.2)/(28.3 ± 5.2)(25.3 ± 2.9)/(23.2 ± 2.7)FPG (mg/dl): (87.2 ± 10.2)/(75.7 ± 12.9)FPG: Fasting Plasma Glucose. FBS: Fasting Blood Sugar. 2 h.PP: 2 h post prandial blood glucose Characteristics of the included articles FPG: Fasting Plasma Glucose. FBS: Fasting Blood Sugar. 2 h.PP: 2 h post prandial blood glucose Quality scores Based on the assessment items rated by the NOS scoring system, five studies were assessed as high quality (7–9 stars), and two were evaluated as medium quality (4–6 stars), indicating an overall good methodological quality of the studies incorporated in the interpretation of the meta-analytic results. Table 2 presents a summary of quality control procedures. Table 2Quality assessment using the Newcastle-Ottawa scaleIncluded studyIIIIIIIVVVIVIIVIIINOS ScoresOrhan 2014100111116Li 2015111111118Qiu 2016111111118Barseem 2021110111117Barseem 2022110111117Montazeri-Najafabady 2022110111117Shamsad 2024000111115The Newcastle-Ottawa scale (NOS) was used to assess the quality of the non-randomized studies included in this meta-analysis in the following three aspects: selection, exposure, and comparability. I: Is the case definition adequate; II: Representativeness of the cases; III: Selection of Controls; IV: Definition of Controls; V: Comparability of cases and controls on the basis of the design or analysis; VI: Ascertainment of exposure; VII: Same method of ascertainment for cases and controls; VIII: Non-Response rate Table 2Quality assessment using the Newcastle-Ottawa scaleIncluded studyIIIIIIIVVVIVIIVIIINOS ScoresOrhan 2014100111116Li 2015111111118Qiu 2016111111118Barseem 2021110111117Barseem 2022110111117Montazeri-Najafabady 2022110111117Shamsad 2024000111115The Newcastle-Ottawa scale (NOS) was used to assess the quality of the non-randomized studies included in this meta-analysis in the following three aspects: selection, exposure, and comparability. I: Is the case definition adequate; II: Representativeness of the cases; III: Selection of Controls; IV: Definition of Controls; V: Comparability of cases and controls on the basis of the design or analysis; VI: Ascertainment of exposure; VII: Same method of ascertainment for cases and controls; VIII: Non-Response rate Quality assessment using the Newcastle-Ottawa scale The Newcastle-Ottawa scale (NOS) was used to assess the quality of the non-randomized studies included in this meta-analysis in the following three aspects: selection, exposure, and comparability. I: Is the case definition adequate; II: Representativeness of the cases; III: Selection of Controls; IV: Definition of Controls; V: Comparability of cases and controls on the basis of the design or analysis; VI: Ascertainment of exposure; VII: Same method of ascertainment for cases and controls; VIII: Non-Response rate Meta-analysis of the association between the null polymorphism of GSTM1 and gestational diabetes mellitus susceptibility Between-study heterogeneity was not statistically evidenced (P = 0.30, I2 = 16.9%) in the meta-analysis for the null polymorphism of the GSTM1 gene (Table 3). Therefore, the fixed effect model was used to summarize the meta-analytic results. A statistically significant association was detected between the null polymorphism and susceptibility to this metabolic disorder (OR: 2.01, 95% CI: 1.68–2.39, P < 0.01), suggesting that individuals with this genotype are more likely to experience gestational diabetes mellitus. Further stratified analysis by ethnicity revealed significant relevance in all subgroups (Fig. 2). The estimates from studies with small sample sizes are more likely to be distorted and cannot be accurately and confidently extrapolated to the whole population. The synthesized results retained statistical significance even after excluding datasets with fewer than 150 participants, indicating the robustness of our findings. Table 3Meta-analytic results of overall and subgroup analysis for the null polymorphisms of GSTT1 and GSTM1 and gestational diabetes mellitusStudy numberSample size (Case/Control)OR (95% CI) P P h I2 GSTM1 Total71012/10812.01 (1.68, 2.39)< 0.010.3016.9%EthnicityWest-Asian2199/1881.96 (1.30, 2.96)0.030.2135.2%East-Asian2543/6231.77 (1.39, 2.24)< 0.010.740South-Asian1150/1503.36 (2.08, 5.44)< 0.01NANANorth-African2120/1202.08 (1.23, 3.51)< 0.011.000Sample size>=1505922/9912.05 (1.71, 2.47)< 0.010.2132.4%< 150290/901.58 (0.87, 2.86)0.130.430 GSTT1 Total71012/10811.43 (1.08, 1.89)0.010.0748.3%EthnicityWest-Asian2199/1881.05 (0.68, 1.62)0.820.870East-Asian2543/6231.58 (0.97, 2.57)0.070.0574.1%South-Asian1150/1501.06 (0.67, 1.68)0.81NANANorth-African2120/1202.44 (1.29, 4.61)< 0.010.530Sample size>=1505922/9911.39 (1.03, 1.89)0.030.0558%< 150290/901.75 (0.83, 3.67)0.140.1747.4%OR: odds ratio. 95% CI: 95% confidence interval. Ph: P-value of the heterogeneity test Table 3Meta-analytic results of overall and subgroup analysis for the null polymorphisms of GSTT1 and GSTM1 and gestational diabetes mellitusStudy numberSample size (Case/Control)OR (95% CI) P P h I2 GSTM1 Total71012/10812.01 (1.68, 2.39)< 0.010.3016.9%EthnicityWest-Asian2199/1881.96 (1.30, 2.96)0.030.2135.2%East-Asian2543/6231.77 (1.39, 2.24)< 0.010.740South-Asian1150/1503.36 (2.08, 5.44)< 0.01NANANorth-African2120/1202.08 (1.23, 3.51)< 0.011.000Sample size>=1505922/9912.05 (1.71, 2.47)< 0.010.2132.4%< 150290/901.58 (0.87, 2.86)0.130.430 GSTT1 Total71012/10811.43 (1.08, 1.89)0.010.0748.3%EthnicityWest-Asian2199/1881.05 (0.68, 1.62)0.820.870East-Asian2543/6231.58 (0.97, 2.57)0.070.0574.1%South-Asian1150/1501.06 (0.67, 1.68)0.81NANANorth-African2120/1202.44 (1.29, 4.61)< 0.010.530Sample size>=1505922/9911.39 (1.03, 1.89)0.030.0558%< 150290/901.75 (0.83, 3.67)0.140.1747.4%OR: odds ratio. 95% CI: 95% confidence interval. Ph: P-value of the heterogeneity test Meta-analytic results of overall and subgroup analysis for the null polymorphisms of GSTT1 and GSTM1 and gestational diabetes mellitus OR: odds ratio. 95% CI: 95% confidence interval. Ph: P-value of the heterogeneity test Fig. 2Forest plot of the association between the null polymorphism of GSTM1 and gestational diabetes mellitus. Odds ratios (OR) and 95% confidence intervals (CI) were estimated to assess the magnitude of the effect size. The area of each box represented the relative weight assigned to each study under the fixed effect model framework Forest plot of the association between the null polymorphism of GSTM1 and gestational diabetes mellitus. Odds ratios (OR) and 95% confidence intervals (CI) were estimated to assess the magnitude of the effect size. The area of each box represented the relative weight assigned to each study under the fixed effect model framework Meta-analysis of the association between the null polymorphism of GSTT1 and gestational diabetes mellitus susceptibility Significant between-study heterogeneity was found in the pooled estimation (P = 0.07, I2 = 48.3%) for the null polymorphism of the GSTT1 gene (Table 3). The random effects model was used to summarize the meta-analytic results. Summarized estimates revealed a statistically significant association between this polymorphism and susceptibility to gestational diabetes mellitus (OR: 1.43, 95% CI: 1.08–1.89, P = 0.01), suggesting a diabetogenic effect of the GSTT1 null genotype. When stratified by ethnicity, significance was observed only in the North African subgroup (Fig. 3). Additionally, the pooled estimates remained statistically significant after removing studies with fewer than 150 samples, thereby underscoring the robustness of our findings. Fig. 3Forest plot of the association between the null polymorphism of GSTT1 and gestational diabetes mellitus. Odds ratios (OR) and 95% confidence intervals (CI) were estimated to assess the magnitude of the effect size. The area of each box represented the relative weight assigned to each study under the fixed effects model framework Forest plot of the association between the null polymorphism of GSTT1 and gestational diabetes mellitus. Odds ratios (OR) and 95% confidence intervals (CI) were estimated to assess the magnitude of the effect size. The area of each box represented the relative weight assigned to each study under the fixed effects model framework Publication bias To visually diagnose the magnitude of publication bias, funnel plots were generated to depict the effect size of each study against its corresponding standard error. For the pooled estimations of the null polymorphisms of the GSTM1 (Fig. 4A) and GSTT1 (Fig. 4B) genes, symmetry around the triangular funnel indicated the absence of publication bias. Further statistics yielded from Begg’s test (the null polymorphism of GSTT1: P = 1.00; the null polymorphism of GSTM1: P = 0.76) and Egger’s test (the null polymorphism of GSTT1: P = 0.90; the null polymorphism of GSTM1: P = 0.78) suggested that the tendency of publication bias is extremely low. Fig. 4Funnel plots for the association between the null polymorphisms of the GSTM1 and GSTT1 genes and gestational diabetes mellitus risk. A The null polymorphism of GSTM1; B The null polymorphism of GSTT1 Funnel plots for the association between the null polymorphisms of the GSTM1 and GSTT1 genes and gestational diabetes mellitus risk. A The null polymorphism of GSTM1; B The null polymorphism of GSTT1 Evaluation of heterogeneity In the pooled estimation for the null polymorphisms of the GSTT1 gene, statistically significant between-study heterogeneity was identified. To trace the source of heterogeneity, a Galbraith plot (Fig. 5) was generated to identify studies that fall outside the expected confidence interval. The results showed that one Chinese study with 678 participants was marked as the outlier [24]. After removing this dataset, the between-study heterogeneity became statistically non-significant (P = 0.34, I2 = 11.8%). However, the pooled estimates also turned formally non-significant (OR: 1.24, 95% CI: 0.99–1.55, P = 0.06). This implied the vulnerability of the meta-analytic results. Fig. 5Galbraith plot for the association between the null polymorphism of GSTT1 and gestational diabetes mellitus Galbraith plot for the association between the null polymorphism of GSTT1 and gestational diabetes mellitus Sensitivity analysis The pooled estimates for both polymorphisms were subjected to sensitivity analysis to rule out the possibility of uncertainty or controversy. The effect of each dataset on the summarized statistics was assessed by executing the meta-analysis repeatedly, excluding one study at a time in each iteration. For the null polymorphism of GSTM1 (Fig. 6A), no substantial change in the statistical significance of the pooled results was observed, indicating the robustness of the meta-analytic evidence. For the null polymorphism of GSTT1 (Fig. 6B), the originally significant association lost statistical significance when the study by Li et al. [24]. was excluded from the pooled estimation, which coincided with the findings from the heterogeneity analysis. Fig. 6Sensitivity analysis for the association between the null polymorphisms of the GSTM1 and GSTT1 genes and the risk of gestational diabetes mellitus. A The null polymorphism of GSTM1; B The null polymorphism of GSTT1 Sensitivity analysis for the association between the null polymorphisms of the GSTM1 and GSTT1 genes and the risk of gestational diabetes mellitus. A The null polymorphism of GSTM1; B The null polymorphism of GSTT1 Discussion This systematic review and meta-analysis was performed to achieve superior objectivity, generalizability, and precision by integrating all available epidemiological evidence from genetic association studies relevant to the issue. The pooled estimates demonstrated that carriers of the null polymorphism of GSTT1 or GSTM1 are prone to a higher risk of developing gestational diabetes mellitus. It should be noted objectively that confounding factors across studies (such as diet, socioeconomic status, maternal age, diagnostic criteria variations for gestational diabetes mellitus, or access to healthcare) could introduce heterogeneity and obscure the true association between exposure and outcome. Therefore, future high-quality prospective research is required to better account for potential confounding factors and substantiate these findings in this meta-analysis. Stratified analysis by ethnicity revealed a significant association between the GSTM1 null polymorphism in all subgroups. However, the association between the GSTT1 null polymorphism and the risk of gestational diabetes mellitus was only statistically significant in the North African subgroup, but not in other ethnic demographics. While ethnic differences (e.g., genetic background), participant heterogeneity, and interactions with other factors (e.g., environmental exposure) may indeed contribute to the observed differences, it should be acknowledged that the limited sample size in the aggregated cohort and specific subgroups within our meta-analysis could result in limited representativeness and reduced statistical power, potentially affecting the detection of significant associations. Therefore, further large-scale and well-designed validation studies across different ethnicities are required to evaluate the potential ethnic specificity. Publication bias is a distortion of the meta-analytic results because not all studies are integrated (especially those with negative or non-significant findings), which is subject to the tendency among investigators, reviewers, and journals to report or accept studies for publication based on the significance and direction of observed effects [37, 38]. Therefore, tackling publication bias is crucial to maintaining the integrity and credibility of scientific results. In this meta-analysis, no apparent signs of publication bias were observed for either null polymorphism. However, it should be noted that fewer than 10 studies were integrated in the meta-analysis for both polymorphisms, which may result in compromised test power, making it difficult to statistically distinguish between random variation and true asymmetry [39]. The pooled statistics may be biased in case individual observations are given excessive weight in the meta-analytic calculations, or if extreme values are introduced into the sample. To address these issues, sensitivity analysis was conducted in this meta-analysis. The pooled estimates of the null polymorphism of GSTM1 remained consistent, which substantially enhanced the reliability and credibility of the findings. However, the statistical significance of the association between the null polymorphism of GSTT1 and the risk of gestational diabetes mellitus diminished after excluding a particular study [24] during the sensitivity analysis. This finding actually weakened the inferences drawn from the pooled estimation. Further large-scale validation is warranted to generalize the findings. Although the primary conclusion of a systematic review and meta-analysis is usually a statistically quantitative summary of the results, a thorough investigation of the sources of between-study heterogeneity is likely to enhance both the scientific significance and clinical applicability of the conclusions reached [40]. Statistically significant between-study heterogeneity was observed in the pooled estimation of the GSTT1 null polymorphism. The meta-analytic results derived from a random effects model indicated that the null genotype is prone to an increased risk of gestational diabetes mellitus. However, the corrected pooled statistics based on a fixed effect framework were no longer significant. The discrepancy in statistical significance between the meta-analytic results before and after outlier correction suggests a lack of robustness in the summary statistics. Therefore, several future actions are essential to advance this line of research. Initially, additional extensive validation of this observed trend in disease vulnerability in larger, more diverse populations is necessary to verify its universality across different ethnicities or demographics. Furthermore, assessing the interplay between genetic predispositions and environmental exposures across different ethnic populations can provide deeper insights into the delicate mechanisms behind gestational diabetes mellitus susceptibility. Additionally, robust longitudinal studies in broader, more varied cohorts are also necessary to track diabetic vulnerability over time to measure the stability of the observed trend. In this study, the following merits should be considered. Firstly, this PRISMA-based systematic review and meta-analysis was prospectively registered in PROSPERO, and the analytic protocol has been approved. Secondly, a comprehensive and rigorous search was conducted across five common academic databases to gather extensive evidence from the published literature. Thirdly, the overall degree of methodological rigor in the included studies was satisfactory, ensuring the validity and reliability of the selection of diabetic case and control groups, the comparability between groups, and the ascertainment of the exposure of interest (the null polymorphisms of the GSTT1 and GSTM1 genes). Finally, neither the funnel plot nor the publication bias detection methods provided statistical evidence of a systematic bias in the pooled estimations, indicating the meta-analytic results were not likely to be distorted by the partial disclosure of findings. It is essential to acknowledge that this systematic review and meta-analysis may have several potential limitations. First of all, all the included studies derived from Asian and North-African populations, which significantly affected the generalizability of the meta-analysis results. Further large-scale validation tests are crucial for assessing the validity of the findings across different ethnic groups (e.g., Caucasian populations). Furthermore, statistically significant between-study heterogeneity was detected for the pooled estimate of the GSTT1 null polymorphism. This posed a pressing challenge to the appropriateness of aggregating the integrated studies and the trustworthiness of the meta-analytic findings [41]. Moreover, although we conducted a comprehensive literature search, only a relatively small number of datasets for null polymorphisms met our inclusion criteria in this meta-analysis, which limited the power to achieve genuine statistical significance. In addition, subgroup analysis according to some characteristics could not be accomplished due to the inclusion of only a single dataset in several subgroups, which substantially restricted the reliability of the summary statistics. Besides, the studies included in this meta-analysis were all of case-control design by nature, which was particularly prone to selection bias than other epidemiologic approaches because it requires that both cases and controls adequately represent the same source population [42]. Of particular concern is the absence of rigorous genotyping quality assessments (e.g., Hardy-Weinberg equilibrium test) due to the methodologies employed in the included studies, which significantly undermines the validity of the research findings. Last but not least, the pooled estimation for the impact of the genetic variants of other Glutathione S-Transferase genes (e.g., the GSTP1 A313G and null polymorphisms) was not performed due to a limited number of studies. Further meta-analysis should be performed when enough datasets have been accumulated. Conclusion This systematic review and meta-analysis indicated that the null polymorphisms of the GSTT1 and GSTM1 genes may be potential biomarkers for gestational diabetes mellitus risk. However, the clinical value of these polymorphisms should be validated by further large-scale, multicentre, and prospective studies, particularly in light of the non-robustness of the findings for the GSTT1 null polymorphism indicated by sensitivity analysis and heterogeneity adjustment. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1