This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A comprehensive literature search was conducted across three major electronic databases: PubMed, Scopus, and Web of Science. The search strategy was meticulously developed based on preliminary scoping reviews and expert knowledge within the field, ensuring the inclusion of relevant terms and keywords. Boolean operators (AND, OR) were employed to combine the search terms: mirnas, microRNA, miRNA, Heart Defects, Congenital, congenital heart disease, ventricular septal defects, atrial septal defects, Patent ductus arteriosus, Tetralogy of Fallot, prenatal diagnosis, Prenatal Diagnosis, Noninvasive Prenatal Testing. Detailed search strings used in each database are shown in Table 1. Although this review followed the PRISMA 2020 methodology, a formal meta-analysis was not performed due to substantial heterogeneity in study design, miRNA detection methods, sample types, outcome definitions, and reporting formats. Instead, a structured narrative synthesis was conducted in accordance with current best practices for systematic reviews where meta-analysis is not feasible.

Only original research articles published in English and available in full-text format were included in the review. Publications that did not meet these criteria were excluded. Specifically, review articles, case reports, case series, opinion pieces, short communications, and conference abstracts were not considered for inclusion. Additionally, studies published in languages other than English or those for which the full text was inaccessible were excluded.

Study selection was performed in two phases. In the first phase, two authors independently screened the titles and abstracts of identified articles. Articles that did not apply to the research question were excluded. Abstracts that appeared eligible for full-text screening were retrieved. In the second phase, full-text articles were thoroughly assessed to determine whether they met the predefined inclusion and exclusion criteria. Two authors independently reviewed the full texts, and any discrepancies regarding eligibility were resolved through discussion or, when necessary, by consultation with a third author. This two-step process was conducted using Zotero (ver. 6.0.36) open-source software to facilitate the organization and management of references.

Data from studies meeting the inclusion criteria were extracted into a structured Excel spreadsheet by two independent reviewers. The following variables were collected: author, year, study design, study population, gestational age at sampling, CHD subtype, diagnostic methods, and miRNAs investigated. Performance indicators such as AUC, sensitivity, and specificity were also recorded when available.

In addition, we extracted methodological details relevant to data heterogeneity, including type of sample (e.g., maternal serum, amniotic fluid), type of experiment (discovery vs. targeted), discovery platform (e.g., NGS, microarray), validation method (e.g., qRT-PCR), dye system used, and normalization strategy (e.g., U6, cel-miR-39).

Due to heterogeneity in CHD classification, timing of sample collection, and technological platforms, pooling of data for meta-analysis was not appropriate. Therefore, results were summarized using a narrative synthesis approach, structured around clinical and experimental characteristics. Two detailed summary tables were created to support transparency and reproducibility: one presenting the clinical characteristics and study designs, and the second presenting technical and methodological variables.

This study used the Newcastle Ottawa Scale (NOS) to assess the potential for bias in observational studies. The overall risk of bias was categorized as low, moderate or high risk of bias. The quality of the individual studies was analysed and evidence certainty was measured indicating a lower bias (Lo et al., 2014).

A total of 36 articles were identified through database searches. After removing duplicates, 25 articles remained for title and abstract screening. Of these, 11 articles were deemed eligible for full-text review. After a comprehensive assessment of the full texts against the predefined inclusion and exclusion criteria, seven articles met the eligibility requirements and were included in the systematic review. (Figure 1).

The potential of miRNAs as biomarkers for detecting CHD in fetuses has been identified in seven studies included in this review. The studies included in this systematic review featured diverse designs and methodologies. The sample sizes ranged from 14 to 182 participants, with a balanced distribution of study and control groups across most studies. Diagnostic methods for CHD predominantly relied on fetal echocardiography, supplemented by postnatal surgery or autopsy in some studies. Among the CHD types, VSD was the most frequently studied, with some investigations also addressing TOF, persistent truncus arteriosus (PTA), single ventricle (SV), transposition of the great arteries (TGA), and other mixed or rare defects. Gestational age at sample collection varied, with most studies targeting the second trimester (16–27 weeks), although some included samples during labor or in narrower gestational ranges (e.g., 18–22 weeks). Notably, Gu et al. and Jin et al. enrolled large cohorts, while smaller case-control studies, such as Yang et al., focused on highly specific CHD subtypes (Gu et al., 2019; Jin et al., 2021; Yang et al., 2022). Detailed information has been provided in Table 2.

The experiments conducted in the included studies utilized diverse biological materials and methodologies to investigate miRNA expression patterns associated with CHD. The majority of studies analyzed maternal serum samples (Zhu et al., 2013; Kehler et al., 2015; Gu et al., 2019; Jin et al., 2021; Xi et al., 2024), while Kong et al. focused on umbilical cord blood and Yang et al. examined amniotic fluid (Kong et al., 2021; Yang et al., 2022). Discovery-phase experiments predominantly employed next-generation sequencing (NGS) technologies, including platforms such as Illumina HiSeq 2500 (Jin et al., 2021), Illumina NextSeq 500 (Xi et al., 2024), and Illumina NovaSeq 6000 (Yang et al., 2022). Zhu et al. utilized SOLiD sequencing, (Zhu et al., 2013), while Gu et al. performed microarray analysis using the miRCURY LNA™ microRNA Array system (Gu et al., 2019).

Targeted validation of miRNA expression was commonly carried out using quantitative reverse transcription PCR (qRT-PCR). SYBR Green systems were employed in most studies, including Gu et al. (2019), Jin et al. (2021), Yang et al. (2022), and Xi et al. (2024) whereas Zhu et al. utilized a TaqMan MicroRNA Assay (Zhu et al., 2013). Notably, Kehler et al. employed a miRCURY LNA™ Universal RT microRNA PCR Kit with LNA-based dye chemistry (Kehler et al., 2015), while Kong et al. used SYBR Green technology with U6 snRNA as an endogenous control (Kong et al., 2021). External controls, such as cel-miR-39, were used in studies by Jin et al., Xi et al., and Zhu et al. to normalize experimental variability (Zhu et al., 2013; Jin et al., 2021; Xi et al., 2024). The detailed experimental workflows and technical systems employed are summarized in Table 3.

Figure 2 illustrates upregulated (green) and downregulated (red) miRNAs identified in studies evaluating their potential as biomarkers for CHD. Each spoke represents a unique miRNA, with the corresponding study authors indicated in parentheses. Notably, no miRNAs overlapped between studies, which may be attributed to the variability in experimental methodologies, as discussed in the text. Upregulated miRNAs were predominantly observed in studies utilizing maternal serum and amniotic fluid, while downregulated miRNAs were also detected across diverse sample types, including umbilical cord blood. This visualization highlights the heterogeneity in miRNA biomarker discovery for CHD (Figure 2).

Several studies focused on the diagnostic potential of miRNAs by evaluating sensitivity, specificity, and area under the curve (AUC) metrics. Gu et al. reported that miR-142-5p (downregulated) achieved an AUC of 80.4% (95% CI: 72.1%–88.7%), with a combined panel of miR-142-5p, miR-1275, miR-4666a-3p, and miR-3664-3p increasing the AUC to 90.1% (95% CI: 84.0%–96.2%). Other miRNAs analyzed in this study, including miR-1275, miR-4426, and miR-4681, exhibited moderate diagnostic performance with individual AUCs ranging from 66.2% to 71.5% (Gu et al., 2019). Jin et al. demonstrated exceptional diagnostic accuracy for miR-146a-5p (downregulated), achieving sensitivity and specificity of 98.1%, with an AUC of 99.7% (95% CI: 99.2%–100%). Conversely, miR-199a-3p, also downregulated, showed lower performance with an AUC of 67.17% (95% CI: 56.0%–78.3%) (Jin et al., 2021; Table 4).

Kong et al. highlighted the diagnostic capabilities of miR-1, miR-208, and miR-499 (all downregulated). Among these, miR-1 and miR-499 demonstrated the highest AUC values of 86% (95% CI: 76%–96%) and 84% (95% CI: 74%–95%), respectively, with perfect specificity (100%) for all three miRNAs (Kong et al., 2021). Xi et al. identified miR-122-5p and miR-3195 (both downregulated) as potential biomarkers, reporting AUC values of 70% (95% CI: 57%–83%) and 68% (95% CI: 54%–81%), respectively. The combination of these miRNAs improved the diagnostic performance, achieving a combined AUC of 73% (95% CI: 60%–86%) (Xi et al., 2024). In the study by Zhu et al., miR-19b (upregulated) showed the highest sensitivity (74.1%) and specificity (77.8%), with an AUC of 79.95% (95% CI: 66.6%–91.4%). The combined panel of miR-19b, miR-22, miR-29c, and miR-375 increased the AUC to 81.3% (95% CI: 69.5%–93.1%) (Zhu et al., 2013; Table 4).

Other studies primarily reported differences in miRNA expression levels without evaluating their diagnostic performance. For instance, Yang et al. highlighted upregulation of miR-1-3p, miR-184, and miR-206 (Yang et al., 2022), while Kehler et al. described increased expression of miR-99a (Kehler et al., 2015). These findings suggest the potential relevance of these miRNAs in CHD pathology but lack diagnostic validation (Table 4).

The risk of bias across the included studies was evaluated using the Newcastle-Ottawa Scale (NOS), which assesses methodological quality across three domains: selection (D1), comparability (D2), and exposure or outcome (D3). Table 5 provides a summary of the results. Among the seven studies assessed, four achieved a “good” rating across all domains, indicating a low risk of bias (Zhu et al., 2013; Gu et al., 2019; Jin et al., 2021; Xi et al., 2024). These studies demonstrated robust methodologies, including adequate selection of study participants, appropriate comparability between groups, and reliable measures of exposure or outcome. Two studies received an “overall middle” rating due to moderate concerns in the domain of comparability (D2) (Kong et al., 2021; Yang et al., 2022). These studies adequately selected participants and measured outcomes but demonstrated limitations in controlling for confounding variables, which may impact the reliability of their findings. Kehler et al. exhibited the highest risk of bias, with ratings of “middle” in selection (D1) and comparability (D2), and “poor” in exposure or outcome (D3) (Kehler et al., 2015). The “poor” rating in D3 suggests potential issues with the accuracy or consistency of exposure or outcome assessment methods, limiting the study’s internal validity. In summary, while most studies were of good methodological quality, potential risks of bias were identified in a minority of studies, particularly in the domains of comparability and exposure or outcome assessment.

The strengths of this review are underscored by its stringent methodology, comprehensive literature search, and systematic assessment of bias using the Newcastle-Ottawa Scale. By exclusively including studies with well-defined inclusion criteria, the reliability of the results was further bolstered, allowing for the evaluation of multiple miRNAs across various types of CHD.

However, several limitations must be acknowledged. Many of the included studies featured small sample sizes, which may compromise the generalizability of the findings. In addition, the studies employed diverse experimental methodologies for isolating and measuring miRNAs, contributing to variability in the reported results. The heterogeneity of biological materials analyzed, including maternal serum, amniotic fluid, and umbilical cord blood, further complicates the interpretation and synthesis of data. These differences preclude the possibility of conducting a comprehensive synthesis of results or a meta-analysis.

Moreover, CHDs represent a heterogeneous group of disorders with varying etiologies and pathophysiological mechanisms. Consequently, drawing generalized conclusions regarding the entire spectrum of CHDs based on the available data may introduce bias and lead to erroneous assumptions. These factors necessitate a cautious interpretation of the results and underscore the need for future studies with larger sample sizes, standardized methodologies, and a focus on specific CHD subtypes to enhance the reliability of findings related to miRNA biomarker utility in prenatal CHD diagnosis.