Causal association between plasma vascular endothelial growth factor family proteins and placenta previa: A 2-sample Mendelian randomization study.

This study has several limitations. First, MR reflects lifelong, genetically proxied differences in plasma VEGFA and therefore informs etiology rather than the clinical predictive or screening value of measuring VEGFA during pregnancy; the modest effect size likewise should not be read as evidence of discriminative utility. Second, both GWAS datasets were restricted to individuals of European ancestry, limiting generalizability to other populations. Finally, although the VEGFA sensitivity analyses did not detect significant heterogeneity or directional horizontal pleiotropy, residual pleiotropy cannot be fully excluded, and prospective pregnancy-specific validation is needed before any clinical translation. The significant MR-Egger intercept for VEGFD warrants particular caution for that exposure. Furthermore, because screening against databases such as PhenoScanner V2 can interrogate only a limited set of previously studied phenotypes, and the exclusion-restriction assumption cannot be tested empirically, the nonsignificant VEGFA MR-Egger and MR-PRESSO results should be interpreted as being consistent with, rather than as proof of, the absence of horizontal pleiotropy.

Considering these findings, future research should explore the specific role of VEGFA’s downstream signaling pathways in the pathogenesis of placenta previa and clarify the interrelationships among the different VEGF family members in placental development. This research represents a pioneering application of MR methods to analyze the causal association between plasma VEGF family proteins and placenta previa, revealing a causal link between elevated plasma VEGFA levels and the increased risk of this condition. The results implicate VEGFA-mediated angiogenesis as a biologically plausible pathway in placenta previa and provide a foundation for further mechanistic investigation, rather than establishing the clinical predictive value of VEGFA measurement in pregnancy.

Placenta previa is defined as a pathological condition in which the placenta is abnormally positioned in the lower uterine segment, reaching or covering the cervical os during late pregnancy, with an incidence rate of 0.3% to 0.5%. [ 1 ] This complication frequently results in spontaneous, painless vaginal bleeding, particularly in late pregnancy or during labor. Placenta previa not only poses a risk of delivery complications but also threatens the safety of both the mother and fetus. Evidence suggests that the cesarean delivery rate in patients with placenta previa may be as high as 80% to 90%, [ 2 ] and this condition is associated with complications such as postpartum hemorrhage and fetal distress, with reported rates of postpartum hemorrhage ranging from 25% to 50%. [ 3 ] Furthermore, placenta previa is correlated with an increased fetal mortality rate, with certain studies indicating that fetal mortality among affected patients may reach 5% to 10%. [ 4 ] Numerous studies have identified various risk factors associated with placenta previa, including a history of induced or spontaneous abortion, male fetal sex, smoking, advanced maternal age, prior cesarean delivery, singleton pregnancies, and endometriosis. [ 5 ] While these findings provide critical insights into the etiology of placenta previa, a consensus on the exact mechanisms remains elusive. Abnormal angiogenesis has been identified as a significant factor in the development of placenta previa. [ 6 ] The vascular endothelial growth factor (VEGF) family plays a crucial role in placental development, comprising members such as VEGFA, VEGFB, VEGFC, VEGFD, and placental growth factor (PLGF), with VEGFA serving as the primary molecule involved in the regulation of angiogenesis. [ 7 ] Despite some studies reporting no significant differences in serum VEGF levels between patients with placenta previa and those with normal pregnancies, [ 8 , 9 ] other research has documented significantly reduced serum VEGF expression in individuals with placenta previa. [ 10 ] Additionally, Zhang et al [ 11 ] observed that plasma PLGF levels in patients with placenta previa were significantly elevated compared to those in normal pregnant women, while Biberoglu et al [ 9 ] reported no significant differences in circulating PLGF levels between the 2 groups. The associations of VEGFB, VEGFC, and VEGFD with placenta previa remain inadequately characterized. These inconsistencies may be attributable to design limitations in existing studies, particularly the susceptibility of observational studies to biases and confounding variables. Randomized controlled trials (RCTs) are widely recognized as the gold standard for establishing causal relationships. However, the practical application of RCTs is frequently hindered by considerations such as study design, ethical constraints, and financial costs. As a result, researchers are compelled to explore alternative methodologies that are both ethically and economically viable. Mendelian randomization (MR) has emerged as a promising alternative, utilizing genetic variants as instrumental variables to effectively evaluate causal relationships between exposures and outcomes, thereby minimizing confounding factors and providing more robust evidence for causal inference. [ 12 , 13 ] In recent years, MR methodologies have gained prominence in clinical research across various diseases, highlighting their significant potential for advancing medical knowledge. This study aims to investigate the causal relationship between plasma VEGF family proteins and placenta previa through a 2-sample MR analysis using data from genome-wide association studies (GWAS). By employing rigorous study design and advanced statistical analysis, we seek to clarify the association between plasma VEGF family protein levels and the occurrence of placenta previa, thereby providing etiologic insight into the role of VEGF family proteins in this condition. We hypothesize that our findings will elucidate how plasma VEGF family protein levels influence the pathophysiology of placenta previa, ultimately offering novel insights and strategies for clinical practice. This research not only aims to mitigate the risks associated with placenta previa but also seeks to enhance maternal and fetal safety.

Conceptualization: Zhantong Hong. Data curation: Zhibin Xu, Zeqi Huang. Formal analysis: Zhantong Hong, Zhibin Xu, Weichao Li. Project administration: Qifeng Liu. Supervision: Qifeng Liu. Writing – original draft: Zhantong Hong. Writing – review & editing: Zhantong Hong, Zhibin Xu, Zeqi Huang, Weichao Li, Qifeng Liu.

This study utilizes a 2-sample MR approach to investigate the causal relationship between plasma VEGF family proteins and placenta previa. The validity of MR analysis is contingent upon the fulfillment of 3 critical assumptions (Fig. 1 ): Conceptual diagram of 2-sample Mendelian randomization analysis for plasma VEGF family proteins and placenta previa risk. VEGF = vascular endothelial growth factor. Association assumption : There must be a robust statistical association between the instrumental variable and the exposure factor, ensuring that the selected single nucleotide polymorphisms (SNPs) accurately reflect variations in plasma VEGF family protein levels. Independence assumption : The association between the instrumental variable and the outcome must be independent of any potential confounders. This assumption necessitates that the instrumental variable is not correlated with any variables that could influence the relationship between the exposure and the outcome, thereby safeguarding the integrity of causal inferences. Exclusivity assumption : The influence of the instrumental variable on the outcome should occur solely through the exposure factor, ensuring that any observed effects are indicative of a genuine causal relationship rather than being confounded by other mechanisms. To meet these assumptions, the study will employ publicly available summary statistics from GWAS to identify significant SNPs associated with plasma VEGF family protein levels as genetic instrumental variables. Additionally, data pertaining to placenta previa will be sourced from a separate GWAS to validate the association of the selected SNPs. Ultimately, a range of MR methodologies will be applied to assess the causal relationship between plasma VEGF family protein levels and the risk of developing placenta previa. This study was designed and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) guidelines. [ 13 ] A completed STROBE-MR checklist is provided as a separate guidelines checklist file. The GWAS data for plasma VEGFA, VEGFB, VEGFC, and VEGFD used in this study (ukb-ppp-VEGFA- OID20650 , ukb-ppp-VEGFB- OID30554 , ukb-ppp-VEGFC- OID21524 , ukb-ppp-VEGFD- OID20662 ) were derived from the Pharma Proteomics Project (PPP; https://github.com/MRCIEU ). This dataset includes plasma proteomic features from 54,219 participants of European descent in the UK Biobank, [ 14 ] with an effective sample size of 33,918. The GWAS data for plasma PLGF (prot-b-66) were sourced from the IEU database ( https://github.com/MRCIEU ), also comprising individuals of European descent, with an effective sample size of 3394. [ 15 ] The GWAS data for placenta previa (finngen_R9_O15_PLAC_PRAVE) comprised 1232 placenta previa cases and 1,68,929 controls (1,70,161 individuals in total) of Finnish ancestry and was obtained from the FinnGen (Finnish Genomics database) project (release R9; https://www.finngen.fi ). This project aims to explore the relationship between genetic variations and various diseases by integrating genomic data with health records. The data resources from the FinnGen project have been widely utilized and recognized in the international academic community. Sample overlap . Two-sample MR assumes that the exposure and outcome associations are estimated in nonoverlapping samples, because participant overlap can bias effect estimates toward the confounded observational association, with greater distortion when the instruments are weak. In the present analysis, the exposure data were obtained from the UK Biobank (VEGFA–D) and the IEU/Folkersen cohort (PLGF), both comprising predominantly non-Finnish European individuals, whereas the outcome data were obtained from FinnGen, which is composed of Finnish biobank participants. UK Biobank participants are not included in FinnGen, and no overlapping participants between the exposure and outcome GWAS are known. Given the distinct national cohorts and independent recruitment frameworks, meaningful sample overlap is highly unlikely. Furthermore, because all instruments were strong (F ≫ 10), any bias arising from residual, unrecognized overlap would be expected to be negligible and, if present, would act conservatively – biasing estimates toward the observational association rather than generating a spurious causal signal. This study was based exclusively on publicly available, summary-level GWAS data from the UK Biobank, IEU, and FinnGen resources, in which all participating studies had obtained ethical approval and informed consent. As no individual-level or identifiable patient data were used, additional ethical approval and informed consent were not required for the present analysis. Furthermore, this study has been been registered with ClinicalTrials.gov ( https://register.clinicaltrials.gov/ ; NCT06620068 ). To adhere to the relevance assumption, this study conducted a genome-wide screening for SNPs that exhibit significant associations with the exposure factors, employing a significance threshold of 5 × 10 −8 . Independent instrumental variables were derived through linkage disequilibrium (LD) clustering, with a threshold of R 2  < 0.001 and a genetic distance of 10,000 kb. [ 14 ] To reduce the potential for violations of the independence assumption, SNPs associated with potential confounders, including prior miscarriage history, smoking, cesarean section history, and endometriosis, were systematically excluded using the PhenoScanner V2 database. [ 16 ] Instrument strength was quantified for each SNP individually using the exposure GWAS effect estimate (β), effect-allele frequency (EAF), and sample size (N). For each SNP, the proportion of variance in the plasma protein level explained was estimated as R 2  = 2 × EAF × (1 − EAF) × β 2 , where β is the per-allele effect estimate in standard-deviation units. The corresponding SNP-level F -statistic was calculated as F  =  R 2 (N − 2)/(1 −  R 2 ). Standard errors were reported as part of the GWAS summary statistics but were not inputs to this R 2 -based F -statistic formula. F  > 10 was used as the conventional threshold for adequate instrument strength. Complete instrument-level information, including rsID, chromosome and position, effect and other alleles, EAF, effect estimates, standard errors and P values in the exposure and outcome datasets, and per-SNP R 2 and F -statistics, is provided in Table 1 . Instrument-level summary statistics for the plasma VEGF-family instruments (VEGFA, VEGFB, VEGFC, VEGFD, and PLGF) and placenta previa, showing for each SNP the rsID, chromosome and position, effect and other alleles, effect-allele frequency, and the effect estimate, standard error, and P value in both the exposure and outcome datasets, together with the per-SNP R 2 and F -statistic. Coordinates are on GRCh38. Subscripts denote the exposure (plasma protein) and outcome (placenta previa, FinnGen R9) datasets. rs7040440 (VEGFB) and rs10982164 (VEGFD) are absent from FinnGen R9 and their outcome fields are left blank. F = F -statistic, R 2 = variance explained, β, SE, P = per-allele effect estimate, standard error and P value. EA/OA = effect/other allele (aligned to the exposure), EAF = effect-allele frequency, IVW = inverse-variance weighted, PLGF = placental growth factor, SE = standard error, SNP = single nucleotide polymorphism, VEGF = vascular endothelial growth factor. This study employed R software (version 4.3.2; R Foundation for Statistical Computing) for statistical analysis. We implemented inverse-variance weighted (IVW) methods, MR-Egger regression, weighted median methods, and weighted mode methods for 2-sample MR analysis. The IVW method was designated as the primary analytical approach, utilizing odds ratios (OR) to assess the causal relationship between VEGF family proteins and placenta previa. In the MR analysis, we first evaluated the heterogeneity of instrumental variables using Cochran Q test. Heterogeneity may stem from variations in population characteristics, analytical processes, or experimental design, potentially impacting the validity of causal inference. A result of P  > .05 indicates that the heterogeneity among instrumental variables is not significant. Additionally, we employed the MR-PRESSO method to detect potential outliers; any identified outliers would be excluded from subsequent analyses. [ 17 ] Subsequently, we conducted MR-Egger regression to analyze the intercept term and to assess evidence of directional horizontal pleiotropy. A nonsignificant intercept test provides no statistical evidence of directional horizontal pleiotropy among the instrumental variables; however, because the exclusion-restriction assumption cannot be verified empirically, this should not be interpreted as proof that horizontal pleiotropy is absent. To further assess the sensitivity of the results, we employed a leave-one-out approach, systematically removing each SNP and recalculating the combined effect; if these results align closely with the combined values obtained from the IVW method, it suggests that no individual SNP significantly influences the overall analysis. Statistical significance was established at P  < .05. Because 5 VEGF-family protein exposures were analyzed, the primary IVW results were additionally interpreted against a Bonferroni-corrected significance threshold of P  < .01 (0.05/5) to account for multiple comparisons across exposures. In the interpretation of results, it is imperative to ensure that the assumptions underlying each analytical method are satisfied and to provide a comprehensive interpretation of the observed associations, particularly regarding potential confounding factors and their biological relevance. Effect estimates from all 4 MR methods (IVW, MR-Egger, weighted median, and weighted mode), together with the corresponding Cochran Q , MR-Egger intercept and MR-PRESSO statistics for each exposure, are summarized in Table 2 . Mendelian randomization estimates (inverse-variance weighted, MR-Egger, weighted median, and weighted mode) for plasma VEGF-family proteins on placenta previa, reporting the odds ratio, 95% confidence interval, β and P value for each method, together with Cochran Q , the MR-Egger intercept and MR-PRESSO global-test results for each exposure. IVW is the primary analysis. Values as reported in the main text. IVW = inverse-variance weighted, LCI/UCI = lower/upper 95% confidence limit, MR = Mendelian randomization, MR-PRESSO = MR Pleiotropy RESidual Sum and Outlier, OR = odds ratio, PLGF = placental growth factor, SE = standard error, SNP = single nucleotide polymorphism, VEGF = vascular endothelial growth factor.

From the exposure GWAS summary statistics, we identified 28, 13, 32, 13 and 6 independent SNPs ( P  < 5 × 10 −8 ; LD r 2  < 0.001 within 10,000 kb) associated with plasma VEGFA, VEGFB, VEGFC, VEGFD, and PLGF, respectively. After excluding SNPs associated with potential confounders (PhenoScanner V2) and harmonizing with the outcome dataset – removing variants absent from the FinnGen dataset and non-inferable palindromic SNPs – the final analyses included 27, 12, 24, 11 and 5 SNPs, respectively (Table 1 ). All instruments had F -statistics well above 10 (range 20.7–6.2 × 10 3 ), indicating that weak-instrument bias was unlikely. F -statistics were computed for each SNP individually using the exposure GWAS effect estimate (β), effect-allele frequency (EAF), and sample size (N), with R 2  = 2 × EAF × (1 − EAF) × β 2 and F  =  R 2 (N − 2)/(1 −  R 2 ) (Table 1 ). By exposure, the per-SNP F-statistic ranged from 20.7 to 6179.5 (median 41.3) for VEGFA, 29.2 to 197.5 (median 40.3) for VEGFB, 28.0 to 251.2 (median 43.3) for VEGFC, 28.4 to 344.2 (median 60.0) for VEGFD, and 47.3 to 93.9 (median 89.0) for PLGF. As every instrument exceeded the conventional threshold of 10, no SNP was excluded as a weak instrument. The IVW, MR-Egger, weighted median, and weighted mode point estimates were all positive, although not all estimators reached statistical significance (Fig. 2 and Table 2 ). The IVW analysis, used as the primary method, indicated that genetically predicted higher plasma VEGFA was associated with an increased risk of placenta previa (β = 0.231, OR = 1.26, 95% confidence interval [CI]: 1.06–1.51, P  = .008). The effect estimates derived from the different MR methods exhibited a uniform trend (Fig. 3 ). Leave-one-out sensitivity analyses indicated that no single instrumental variable significantly influenced the overall MR analysis results (Fig. 4 ). Cochran Q test results suggested no significant heterogeneity in the MR analysis outcomes (IVW: Q  = 21.257, P  = .729; MR-Egger: Q  = 19.368, P  = .779). Furthermore, MR-Egger regression did not detect directional horizontal pleiotropy among the instrumental variables (intercept = −0.026, P  = .181). The MR-PRESSO global test was not statistically significant ( P  = .692). Forest plots of Mendelian randomization analyses of plasma VEGF family proteins and placenta previa risk. CI = confidence interval, IVW = inverse-variance weighted, MR = Mendelian randomization, OR = odds ratio, PLGF = placental growth factor, VEGF = vascular endothelial growth factor. MR estimates for plasma VEGFA and placenta previa risk. The plot presents SNP–exposure and SNP–outcome associations together with estimates from IVW, MR-Egger, weighted median, and weighted mode analyses. The method-specific point estimates are positive, although not all estimators reach statistical significance. Causal interpretation and the assumptions required are discussed in the main text. IVW = inverse-variance weighted, MR = Mendelian randomization, SNP = single nucleotide polymorphism, VEGF = vascular endothelial growth factor. Leave-one-out analysis of plasma VEGFA-associated SNPs on placenta previa risk. Each VEGFA-associated SNP is sequentially omitted, and the IVW estimate is recalculated. The plot assesses whether the overall estimate is disproportionately influenced by any single SNP. Stability in this analysis does not independently establish causality or exclude horizontal pleiotropy. IVW = inverse-variance weighted, SNP = single nucleotide polymorphism, VEGF = vascular endothelial growth factor. The IVW point estimate suggested a positive direction of association between genetically predicted plasma VEGFB levels and placenta previa risk (β = 0.350, OR = 1.42, 95% CI: 0.95–2.12, P  = .088; Fig. 2 and Table 2 ). However, the confidence interval included the null, and the estimate was not statistically convincing. The weighted median and weighted mode estimates were also positive but nonsignificant (OR = 1.45, P  = .138 and OR = 1.38, P  = .320, respectively), whereas the MR-Egger estimate was in the opposite direction and nonsignificant (OR = 0.86, P  = .780). Thus, the IVW estimate was not consistently supported across sensitivity estimators, and these analyses did not provide robust evidence of a causal effect of VEGFB on placenta previa risk. The effect estimates for the instrumental variables utilized in the various MR analysis methods are presented (Fig. 5 A). Leave-one-out sensitivity analysis indicated that the MR findings were not attributable to any individual instrumental variable (Fig. 6 A). Cochran Q test demonstrated no significant heterogeneity within the MR analysis results (IVW: Q  = 10.129, P  = .519; MR-Egger: Q  = 6.745, P  = .749). Additionally, MR-Egger analysis did not reveal evidence of horizontal pleiotropy (intercept = 0.129, P  = .096). The MR-PRESSO global test was not statistically significant ( P  = .554). MR estimates for VEGFB, VEGFC, VEGFD, and PLGF on placenta previa risk. Method-specific estimates from IVW, MR-Egger, weighted median, and weighted mode analyses are shown for (A) VEGFB, (B) VEGFC, (C) VEGFD, and (D) PLGF. For VEGFB, the IVW point estimate is positive but nonsignificant, and the direction is not consistent across all sensitivity estimators. For VEGFC and PLGF, none of the 4 methods provides a statistically significant estimate. For VEGFD, the positive MR-Egger slope is nonsignificant and is not supported by the IVW analysis. The VEGFD MR-Egger intercept is significant (−0.155, P  = .029), indicating possible directional horizontal pleiotropy and warranting particular caution. The panels display the estimates and do not establish causal effects for these exposures. IVW = inverse-variance weighted, MR = Mendelian randomization, PLGF = placental growth factor, SNP = single nucleotide polymorphism, VEGF = vascular endothelial growth factor. Leave-one-out analyses of VEGF family proteins and placenta previa risk: VEGFB, VEGFC, VEGFD, and PLGF. Each SNP is sequentially omitted, and the IVW estimate is recalculated for (A) VEGFB, (B) VEGFC, (C) VEGFD, and (D) PLGF. These plots assess the influence of individual SNPs on each exposure-specific estimate. Stability of an estimate does not convert a nonsignificant association into evidence of a causal effect and does not exclude horizontal pleiotropy. IVW = inverse-variance weighted, PLGF = placental growth factor, SNP = single nucleotide polymorphism, VEGF = vascular endothelial growth factor. The analyses conducted using the IVW, MR-Egger regression, weighted median, and weighted mode methods consistently indicated a lack of significant association between plasma VEGFC levels and the risk of placenta previa (Fig. 2 ). The effects of the instrumental variables across the various MR methodologies are presented (Fig. 5 B). Sensitivity analysis using the leave-one-out method demonstrated that the overall MR results were not influenced by any single instrumental variable (Fig. 6 B). Cochran Q test revealed no significant heterogeneity among the MR findings (IVW: Q  = 20.060, P  = .638; MR-Egger: Q  = 16.127, P  = .810). Moreover, MR-Egger regression provided no evidence of directional horizontal pleiotropy, with an intercept of −0.090 ( P  = .060). The MR-PRESSO global test was not statistically significant ( P  = .648). Collectively, these findings indicate that plasma VEGFC levels do not exhibit a significant association with the risk of placenta previa within this study population. The MR-Egger point estimate for VEGFD was positive but nonsignificant (OR = 1.42, 95% CI: 0.42–4.79, P  = .610), whereas the IVW estimate was in the opposite direction and nonsignificant (OR = 0.86, 95% CI: 0.59–1.25, P  = .426). The weighted median and weighted mode estimates were also nonsignificant (Fig. 2 and Table 2 ). These analyses did not provide statistically convincing evidence of a causal effect of VEGFD on placenta previa risk, and the direction of the estimates was inconsistent across methods (Fig. 5 C). Leave-one-out analysis further confirms that no single instrumental variable can account for the overall MR analysis results (Fig. 6 C). Cochran Q test results indicate that the MR analysis does not demonstrate significant heterogeneity (IVW: Q  = 11.989, P  = .286; MR-Egger: Q  = 5.268, P  = .810). Additionally, the MR-Egger analysis reveals an intercept of −0.155 ( P  = .029), which is statistically compatible with directional horizontal pleiotropy among the VEGFD instruments; the exclusion-restriction assumption may therefore be violated for this exposure, and these VEGFD results should be interpreted with particular caution. The nonsignificant MR-PRESSO global test ( P  = .204) does not negate the significant MR-Egger intercept. IVW, MR-Egger, weighted median, and weighted mode analyses reveal no significant association between plasma PLGF levels and the risk of placenta previa (Figs. 2 and 5 D). The leave-one-out analysis further confirms that the overall MR analysis results are not influenced by any single instrumental variable (Fig. 6 D). Cochran Q test indicates no significant heterogeneity (IVW: Q  = 0.706, P  = .951; MR-Egger: Q  = 0.635, P  = .888), which supports the reliability of the analysis. Additionally, the MR-Egger analysis yields an intercept of 0.018 ( P  = .808), suggesting no evidence of horizontal pleiotropy. The MR-PRESSO global test was not statistically significant ( P  = .960). These findings underscore the limited association between plasma PLGF levels and the risk of placenta previa, highlighting the need for cautious interpretation of such exposure factors in related studies.

In analyzing the GWAS data of plasma VEGF family proteins and placenta previa from European populations using 2-sample MR methods, a causal association was identified between elevated plasma VEGFA levels and an increased risk of placenta previa (IVW P  = .008, which remains below the Bonferroni-corrected threshold of .01 for the 5 exposures tested), whereas the analyses did not provide statistically convincing evidence of a causal effect of plasma VEGFB, VEGFC, VEGFD, or PLGF. Given the limited power of these individual analyses, these null findings should not be equated with evidence of no effect. The placenta is a crucial organ for sustaining fetal growth and development within the uterus, characterized by a rich vascular network. [ 18 ] Recent studies have highlighted the link between placental dysfunction and pregnancy complications, further underscoring the significance of the VEGF family in gestation. [ 19 ] VEGF family proteins play a pivotal role in angiogenesis. During the early stages of vascular development, VEGFA binds to VEGFR-2 on endothelial cells, promoting their proliferation and migration, and later interacts with VEGFR-1 to facilitate capillary lumen formation. [ 20 ] The VEGF family, including VEGFA, VEGFB, VEGFC, and VEGFD, plays a crucial role in angiogenesis, particularly during placental development. [ 21 ] In the absence of VEGFA, VEGFB and VEGFD serve as essential molecules for maintaining angiogenesis, providing vital blood supply to the placenta and fetus during early embryonic development, indicating their functional differences at various stages. [ 22 ] The roles of VEGFA, VEGFB, and VEGFC in placental development are well-documented, with VEGFA primarily recognized for its pro-angiogenic properties, promoting the formation of new blood vessels essential for supporting the growing placenta. [ 23 ] Conversely, PLGF, another member of the VEGF family, has also been shown to facilitate angiogenesis and normal placental development through mechanisms that do not rely on VEGFA. [ 24 ] This VEGFA-independent pathway highlights the versatility of the VEGF family in regulating placental health and function. However, the activation of VEGF signaling must be tightly regulated. Excessive activation can lead to an increased density of newly formed blood vessels, compromising the integrity of the endothelial barrier, which may result in vascular leakage and reduced oxygen delivery to vital organs, potentially leading to placental developmental abnormalities. [ 25 ] Such dysregulation is associated with various pregnancy complications, including placental abruption and gestational hypertension. [ 26 ] Elevated expression levels of VEGF family proteins have been closely linked to conditions like placenta previa, where abnormal placentation can exacerbate these risks. Supporting this inference, Xie et al [ 27 ] found significantly increased VEGF expression levels in placental tissues from patients with placenta previa, indicating a potential link between elevated VEGF levels and the pathophysiology of this condition. This finding aligns with the understanding that heightened VEGF signaling may contribute to the complications associated with placenta previa, including increased vascularization and the potential for hemorrhage. The conclusions regarding the causal relationship between circulating VEGF family proteins and the risk of placenta previa remain inconsistent. [ 8 , 9 , 11 , 22 ] The findings of this study indicate a causal association between plasma VEGFA levels and an increased risk of placenta previa (IVW P  = .008, OR = 1.26), while plasma VEGFB, VEGFC, and VEGFD did not show statistically convincing evidence of a causal association. Conversely, some studies suggest that VEGFB may indirectly influence placental function by modulating maternal vascular remodeling. In exploring these findings, it is essential to consider the potential mechanisms between VEGFA and placenta previa. As a primary regulator of angiogenesis, VEGFA plays a crucial role in normal placental development. However, excessive activation of VEGFA may lead to over-angiogenesis, resulting in abnormalities in placental tissue structure, which could cause abnormal placental attachment and increase the risk of bleeding and other complications. Recent studies indicate that VEGFA promotes the proliferation and migration of vascular endothelial cells through its binding to VEGFR-2, a process hypothesized to contribute to the pathogenesis of placenta previa. Abnormal vascular distribution not only hinders placental function but may also exacerbate pregnancy complications associated with the placenta. It should be emphasized that our MR analysis supports an association compatible with a causal effect of genetically proxied circulating VEGFA on placenta previa, but does not itself demonstrate that elevated VEGFA causes abnormal placental attachment through these angiogenic mechanisms; the mechanistic accounts discussed here represent biological plausibility rather than experimentally established causation and require dedicated functional validation. Although the role of VEGFA is relatively straightforward, other members of the VEGF family (such as VEGFB, VEGFC, and VEGFD) did not show statistically convincing evidence of causal effects in this study. These nonsignificant estimates do not establish the absence of an effect. This phenomenon may be due to the fact that these proteins play more auxiliary or indirect roles during placental development. For instance, VEGFB may be more involved in maternal vascular remodeling than directly regulating placental angiogenesis. This indirect effect may become significant under specific pathological conditions, aligning with some studies that propose a secondary role for VEGFB in placental function regulation. In this study, the analysis did not provide statistically convincing evidence of a causal effect of PLGF on placenta previa; this does not establish the absence of an effect. PLGF is another important member of the VEGF family, primarily functioning to support normal placental angiogenesis through its interaction with VEGFR-1. However, the role of PLGF may be more focused on maintaining placental health during normal pregnancies, whereas its significance may not be as pronounced in pathological conditions such as placenta previa compared to VEGFA. Future research could explore the relationship between PLGF and VEGFA in regulating placental angiogenesis to uncover their differential roles in pregnancy complications. For VEGFA, the sensitivity analyses included Cochran Q test, MR-Egger regression, and MR-PRESSO analysis, which showed no evidence of statistical heterogeneity or directional horizontal pleiotropy, while leave-one-out analysis indicated that no single SNP substantially influenced the estimate. These findings should not be generalized to VEGFD, for which the MR-Egger intercept indicated possible directional horizontal pleiotropy. By employing Mendelian randomization, this study is less susceptible to the confounding biases that commonly affect traditional observational research, which are particularly important in the complex relationships between angiogenesis and placental development during pregnancy. Therefore, future research should focus on the specific roles of VEGFA downstream signaling pathways in the pathogenesis of placenta previa, particularly the interplay between VEGFA and other angiogenic factors. Because MR estimates the effect of lifelong, genetically proxied differences in plasma VEGFA, our results speak to the etiology of placenta previa rather than to the predictive performance of measuring VEGFA at a given point in pregnancy. The genetically estimated effect is modest and should not be interpreted as evidence of discriminative or screening utility, which would require formal evaluation of predictive metrics (e.g., sensitivity, specificity, and discrimination) in prospective pregnancy cohorts with serial measurements. Accordingly, any clinical translation, including the potential value of VEGFA-based testing, will require dedicated pregnancy-specific validation and triangulation with other study designs.

We would like to thank Dr Peihang Xu, PhD, from the First Affiliated Hospital of Guangzhou Medical University for assisting with the preparation and English revision of this manuscript.