Maternal and fetal outcomes in women with uterine congenital abnormalities: a systematic review and meta-analysis

The database search yielded 10,561 records, of which 365 were duplicates. After screening titles/abstracts, 47 full-text articles were assessed for eligibility, and 30 studies met all inclusion criteria (Fig. 1 ). The most common reasons for exclusion at full text were: wrong population (e.g. anomalies not congenital or no appropriate comparator), ineligible study design (e.g. case series without controls), or publication outside the date range. The 30 included studies encompassed > 3.3 million total pregnancies (range per study: 21 to 1.4 million) and were published 2015–2025. Table 1 summarizes key characteristics. Nineteen studies were retrospective cohorts [ 6 , 15 – 18 , 21 , 28 , 31 – 38 , 40 , 42 , 46 , 50 ], five were prospective cohorts [ 30 , 44 , 47 , 48 , 53 ], one was an RCT of septum resection [ 5 ], one a non-randomized pre–post-surgical study [ 21 ], and four were cross-sectional or diagnostic accuracy studies [ 6 , 29 , 38 , 40 ]. Geographic representation was broad: 10 studies from Europe [ 5 , 6 , 16 , 21 , 36 – 38 ] , [ 32 ] , [ 39 ], 8 from Asia [ 15 , 18 , 31 , 33 , 34 , 44 , 47 , 48 ], 7 from North America [ 17 , 29 , 30 , 35 , 40 , 46 , 50 ], 2 from the Middle East [ 43 , 53 ], 2 from Australia [ 16 , 40 ], and 1 from Africa [ 42 ]. CUA subtypes evaluated included septate, bicornuate, unicornuate, didelphys, and (in some studies) arcuate uteri [ 5 , 6 , 16 – 18 , 39 ]. Diagnosis was confirmed by modern imaging (3D ultrasound or MRI) in most recent studies, sometimes combined with hysteroscopic or laparoscopic findings (Table 2 ). Several studies focused on specific subpopulations: two in ART populations (women undergoing IVF) [ 28 , 36 ], and six, in women investigated for RPL or infertility [ 6 , 21 , 34 , 39 , 46 , 48 ]—these often reported CUA prevalence or diagnostic test accuracy rather than pregnancy outcomes. Most studies compared outcomes in women with CUA to women with normal uteri (control groups drawn from the general obstetric population or matched fertility clinic patients) [ 15 – 18 , 30 , 31 , 36 , 42 , 47 , 50 ]. Three studies compared different anomaly subtypes without control groups [ 16 – 18 ]. Three surgical studies compared post-metroplasty outcomes to pre-surgery baseline or no surgery [ 5 , 21 , 39 ]. No study reported outcomes for untreated versus treated anomalies in a randomized fashion except the one RCT of septum resection [ 5 ]. Table 1 presents detailed study characteristics, including module/domain, design, setting, sample size, comparator, inclusion in quantitative synthesis, and notes. In brief, Module A (natural history of obstetric outcomes) included population-based and hospital-based cohort studies evaluating spontaneous pregnancy outcomes in women with CUA. Module B comprised studies in infertility or RPL settings (including prevalence surveys and diagnostic accuracy studies). Module C encompassed studies of surgical intervention (mostly septum resection) and subsequent reproductive outcomes. Table 2 provides an overview of each study’s population, exposure (anomaly or intervention), comparator category, exposure ascertainment, confounding control, and synthesis classification. Notably, 10 studies were considered suitable for meta-analysis of obstetric outcomes, while 20 were synthesized narratively due to incompatible outcomes or lack of comparators. Risk of bias: Risk-of-bias assessments are summarized in Table 3 and Fig. 2 . Overall, 13 studies were rated Moderate risk, 8 Serious, 6 Some concerns, 2 Low–Moderate, and 1 Critical/High. The sole RCT (septum resection) had “some concerns” [ 5 ] (open-label design with potential deviations). Common limitations in cohort studies were retrospective design, potential selection bias, and unmeasured confounding (only 12/28 observational studies adjusted for key confounders like age or IVF use). Many diagnostic studies had “some concerns” due to patient selection and lack of blind outcome assessment [ 29 , 38 ]. One technical case series was deemed at critical risk of bias (no comparator) [ 40 ]. These limitations temper the strength of causal inferences. Fig. 2 Risk of bias summary across studies Risk of bias summary across studies In the pooled analyses, women with congenital uterine anomalies had higher odds of adverse obstetric outcomes compared with women with normal uterine anatomy (Table 4 A). The pooled odds of preterm birth < 37 weeks were increased (OR 3.74, 95% CI 1.99–7.02; I 2  = 68.1%; k = 5; Fig. 3 A). Caesarean delivery was also more common among women with CUA (OR 3.52, 95% CI 1.11–11.09; I 2  = 94.9%; k = 5; Fig. 3 B). Fetal malpresentation was associated with the largest pooled effect estimate (OR 19.28, 95% CI 6.27–59.24; I 2  = 23.6%; k = 2; Fig. 4 A), although this estimate should be interpreted cautiously because one contributing study defined malpresentation as breech presentation only. Miscarriage was also more common among women with CUA (OR 1.45, 95% CI 1.13–1.86; I 2  = 0.0%; k = 3; Fig. 4 B). Fig. 3 A - B : forest plots showing study-level and pooled odds ratios (ORs) with 95% confidence intervals (CIs) for ( A ) preterm birth < 37 weeks ( B ) caesarean delivery Fig. 4 A - B : forest plots showing study-level and pooled odds ratios (ORs) with 95% confidence intervals (CIs) for ( A ) fetal malpresentation, and ( B ) miscarriage among women with congenital uterine anomalies compared with women with normal uterine anatomy A - B : forest plots showing study-level and pooled odds ratios (ORs) with 95% confidence intervals (CIs) for ( A ) preterm birth < 37 weeks ( B ) caesarean delivery A - B : forest plots showing study-level and pooled odds ratios (ORs) with 95% confidence intervals (CIs) for ( A ) fetal malpresentation, and ( B ) miscarriage among women with congenital uterine anomalies compared with women with normal uterine anatomy Squares represent individual study effect estimates, with size proportional to the inverse-variance weight of each study. Horizontal lines indicate 95% confidence intervals. Diamonds represent pooled estimates derived using random-effects models. The vertical dashed line represents the line of no effect (OR = 1). Values greater than 1 indicate higher odds of the outcome among women with congenital uterine anomalies. Between-study heterogeneity is quantified using the I 2 statistic and τ 2 . Variation in effect sizes across studies reflects differences in study design, population characteristics, and classification of uterine anomalies. Figures 3 , 4 illustrate the overall risk differences for key outcomes; notably, most pregnancies in women with CUA still resulted in live birth, but the absolute risk of preterm delivery was substantially higher (e.g. ~15% vs ~3% in one cohort【46†Outcome】【46†Relative effect】), and the absolute risk of malpresentation was markedly elevated (e.g. ~25% vs ~3–4%【46†Outcome】【46†Key evidence】). GRADE summary has been presented as Table 5 . This section summarises findings from studies not included in quantitative pooling and reports subtype-specific patterns described across the literature. A number of studies categorised outcomes by subtype of anomaly, but subtype meta-analysis was not conducted due to inconsistency in comparator groups. In most cases, unicornuate and didelphys uteri were associated with less favourable late-gestation outcomes. In the largest study [ 17 ], unicornuate uterus was associated with increased risks of preterm birth (aOR 1.48, 95% CI 1.35–1.62) and fetal malpresentation (aOR 1.39, 95% CI 1.28–1.50) compared with arcuate uterus. Septate uteri were more frequently associated with miscarriage and showed a smaller increase in the risk of preterm birth and malpresentation. Bicornuate uteri showed a moderate increase in risk. Two studies explicitly compared subtypes without using normal uteri as a reference group. Naeh et al. [ 18 ] reported higher risks of preterm birth and malpresentation in didelphys and unicornuate uteri compared with septate uteri. Subtype comparisons were limited by variability in study design, sample size, and adjustment for confounders. No head-to-head randomised trials comparing surgical correction across subtypes were identified. Two studies conducted in ART populations [ 34 , 35 ] reported associations between CUA and lower live birth rates and higher preterm birth rates in IVF pregnancies. One study reported that septate uterus was associated with early pregnancy loss in ART cohorts but not with late pregnancy complications, whereas studies in natural conception cohorts reported associations with both early and late adverse outcomes. Multiple sensitivity analyses were performed to investigate the strength of the findings (Table 4 B). To begin with, the omission of the hemi-uterus study [ 31 ] reduced the pooled OR of preterm birth to about 3.1 as compared to the initial pooled OR of about 3.9, but it was still statistically significant. Omission of the ART cohort [ 35 ] of the preterm analysis provided an OR of approximately 5.06 (95% CI 3.12–8.21) with no heterogeneity (I2 0%), which implies that in non-ART groups the relative risk of preterm birth could be even greater. The addition of the small Wang [ 54 ] twin study (as an exploratory sensitivity) to the preterm meta-analysis provided an OR of approximately 5.45 (95% CI 2.27–13.06) high heterogeneity indicating that the gestation of twins in a unicornuate uterus presents unprecedented risk. Table 4B Sensitivity analyses for preterm birth and caesarean delivery Outcome Sensitivity scenario k Pooled OR (95% CI) I 2 (%) P Preterm birth < 37 weeks Primary model 5 3.74 (1.99–7.02) 68.1 <0.001 Preterm birth < 37 weeks Exclude Qian 2023 4 3.12 (1.64–5.91) 62.5 <0.001 Preterm birth < 37 weeks Exclude Prior 2018 4 4.84 (3.15–7.43) 0.0 <0.001 Preterm birth < 37 weeks Exclude Wang 2025 4 3.92 (1.75–8.75) 76.5 0.001 Caesarean delivery Primary model 5 3.52 (1.12–11.07) 94.9 0.032 Caesarean delivery Exclude Qian 2023 4 2.00 (0.94–4.28) 85.8 0.073 Caesarean delivery Exclude Prior 2018 4 4.77 (1.32–17.22) 92.3 0.017 Caesarean delivery Exclude Wang 2025 4 3.71 (0.93–14.72) 96.2 0.062 Note: k refers to the number of studies included in each sensitivity model. The “Exclude Wang 2025” scenario corresponds to the previous singleton-focused model before inclusion of the twin gestation cohort. sensitivity analyses were performed by sequentially excluding individual studies from the primary meta-analysis model. The association between congenital uterine anomalies and preterm birth remained statistically significant across all sensitivity analyses. For caesarean delivery, pooled estimates varied according to study exclusion, consistent with the substantial between-study heterogeneity observed in the primary analysis Table 5 GRADE summary of findings Outcome Studies (k) Participants/pregnancies Pooled effect estimate Heterogeneity Certainty of evidence Main reasons for downgrading Miscarriage 3 As reported across contributing studies OR 1.45 (95% CI 1.13–1.86) I 2  = 0.0% Low Observational evidence; residual confounding and outcome-definition variation Preterm birth < 37 weeks 5 As reported across contributing studies OR 3.74 (95% CI 1.99–7.02) I 2  = 68.1% Very low Observational evidence; serious inconsistency; residual confounding; population heterogeneity Fetal malpresentation 2 As reported across contributing studies OR 19.28 (95% CI 6.27–59.24) I 2  = 23.6% Very low Few studies; imprecision; variable outcome definition; residual confounding Caesarean delivery 5 As reported across contributing studies OR 3.52 (95% CI 1.11–11.09) I 2  = 94.9% Very low Serious inconsistency; wide confidence interval; indication bias; institutional practice variation Note: Certainty of evidence was assessed using GRADE. Because the evidence was derived predominantly from observational studies, the initial certainty was rated as low and downgraded where applicable for risk of bias, inconsistency, imprecision, indirectness, and potential publication bias. OR = odds ratio; CI = confidence interval; CUA = congenital uterine anomalies; k = number of studies. Population: Women with congenital uterine anomalies. Comparison: Women with normal uterine anatomy or the reference comparator defined in the original study. Outcomes: Miscarriage, preterm birth (<37 weeks), fetal malpresentation, and caesarean delivery Sensitivity analyses for preterm birth and caesarean delivery Note: k refers to the number of studies included in each sensitivity model. The “Exclude Wang 2025” scenario corresponds to the previous singleton-focused model before inclusion of the twin gestation cohort. sensitivity analyses were performed by sequentially excluding individual studies from the primary meta-analysis model. The association between congenital uterine anomalies and preterm birth remained statistically significant across all sensitivity analyses. For caesarean delivery, pooled estimates varied according to study exclusion, consistent with the substantial between-study heterogeneity observed in the primary analysis GRADE summary of findings Note: Certainty of evidence was assessed using GRADE. Because the evidence was derived predominantly from observational studies, the initial certainty was rated as low and downgraded where applicable for risk of bias, inconsistency, imprecision, indirectness, and potential publication bias. OR = odds ratio; CI = confidence interval; CUA = congenital uterine anomalies; k = number of studies. Population: Women with congenital uterine anomalies. Comparison: Women with normal uterine anatomy or the reference comparator defined in the original study. Outcomes: Miscarriage, preterm birth (<37 weeks), fetal malpresentation, and caesarean delivery For caesarean delivery, removing the Qian [ 31 ] cohort dramatically lowered the pooled OR (from ~3.7 to ~1.8, 95% CI 0.70–4.66) and reduced heterogeneity (I 2 90% vs 96%), suggesting that the very high baseline caesarean rate in that study was an outlier influencing the meta-analysis. Removing the ART cohort increased the caesarean OR to ~5.62 (95% CI 1.02–30.86), but again with extreme heterogeneity. Notably, when including the twin data, the caesarean OR was ~3.26 (95% CI 0.96–11.06; k  = 5), which is similar to the primary estimate but still with wide CI. Overall, sensitivity analyses revealed that the relationship between CUA and preterm birth was strong in all assumptions (remaining 3- to 5-fold). The effect size of caesarean delivery was extremely case-mix dependent, as without the high-risk group cases the finding lost its statistical significance, which suggests that baseline risk factors (e.g. the use of prophylactic caesarean when malpresentation is observed in certain centers) have a potent impact on this outcome. Comprehensively, none of the sensitivity tests negated the direction of the original results. Publication bias was not formally assessed because there were few studies per outcome (k = 4 in both preterm birth and caesarean). Funnel plots did not show obvious asymmetry due to the limited number of points, and formal statistical assessment was not performed due to limited number of studies. We view these findings with caution; having less than 10 studies, the tests of publication bias are underpowered and can be unreliable. There might also exist smaller studies with negative findings that were unpublished. Nevertheless, this concern is partially addressed by the fact that a number of large population studies were included. We had no evidence of selective reporting in studies (data were typically reported once collected and the outcomes of interest were provided). Five studies investigated reproductive outcome following surgical repair of a uterine anomaly, mostly resection of the septum. The sole RCT [ 5 ] did not find any significant difference in live birth rate between women who received hysteroscopic septum resection and those who received expectant management during the 12-month period. We did not pool RCT and observational data of surgery in our meta-analysis, but rather narratively synthesized them. Most observational pre-post studies [ 36 ] have reported better pregnancy outcomes following resecting the septum, although they are highly susceptible to bias (no controls at the same time). An example of this is a pre- post study where live birth rate rose by approximately 30% prior to septoplasty to approximately 60% following the procedure although it lacked a control group, thus, could have been a regression to the mean or continued fertility treatment. It is therefore possible to draw only tentative conclusions about surgery: septum resection might possibly reduce the miscarriage rates in women who have recurrent loss with a septate uterus as several case series studies have shown but strong-evidence randomized controlled trials did not provide support to a definite benefit in the case of unexplained infertility. It was observed that all the surgical articles did not cover any interventions on other types of anomalies (e.g. Strassman metroplasty on bicornuate uteri) in the contemporary era. Not many studies were found to report on stillbirth, SGA/fetal growth restriction, or severe maternal morbidity, and the data were too sparse to be pooled. According to Tellum et al. [ 29 ], CUA increased the incidence of placenta previa and postpartum bleeding, which was not consistently recorded in other studies. Perinatal mortality was only reported in one study [ 28 ] with no significant difference between CUA and controls (probably because of a small sample)[(46+Outcome)]. Goyal et al. [ 55 ] reported those cases of surgical complications of metroplasty without outcome comparisons. Therefore, the main findings of our review are informative with regard to miscarriage, preterm birth, malpresentation, and caesarean delivery; we do not have enough evidence to make conclusions concerning other maternal or neonatal outcomes. Overall risk-of-bias assessment of the 30 studies included in the systematic review. Each horizontal bar represents the overall methodological quality judgment assigned to an individual study following critical appraisal using study-design–specific assessment tools. Green indicates low/moderate risk of bias, yellow indicates some concerns/moderate risk of bias, and red indicates high/serious risk of bias.

We performed a systematic review and meta-analysis in accordance with the PRISMA 2020 statement [ 23 ]. The review protocol, detailing the study objectives, eligibility criteria, and planned analytical approach, was developed a priori before study selection and data extraction. Although the protocol was not prospectively registered in PROSPERO, the predefined methodology was followed throughout the conduct of the review. This report adheres to PRISMA guidelines, including a PRISMA flow diagram for study selection (Fig. 1 ). Fig. 1 PRISMA flow diagram PRISMA flow diagram Participants: Studies of women of reproductive age (≈15–49 years) who were pregnant or attempting conception (including those undergoing ART) with a confirmed diagnosis of a congenital uterine anomaly were eligible [ 4 , 22 ]. Exposure (CUA): Included studies had to classify CUAs using a recognized system (e.g. American Society for Reproductive Medicine [ASRM]; European Society of Human Reproduction and Embryology/European Society for Gynaecological Endoscopy [ESHRE/ESGE]) [ 20 , 22 ]. Congenital uterine anomalies were identified using recognized diagnostic modalities reported in the original studies, including hysterosalpingography (HSG), 3D transvaginal ultrasonography, hysteroscopy, magnetic resonance imaging (MRI), or combined imaging approaches. Studies using multimodal assessment were considered to provide greater diagnostic precision and more reliable subtype classification. The arcuate uterus was only considered when it was explicitly stated in the study as a congenital anomaly, and was treated independently in the sensitivity-analyses due to remaining controversy about its clinical significance). Comparators : Eligible comparator groups for quantitative synthesis (meta-analysis) included: (1) women with normal uterine anatomy (no anomaly), (2) comparisons between different CUA subtypes, and (3) surgical intervention versus no intervention or preoperative baseline (in studies of septum resection). Comparative studies contributed to pooled analyses where outcome definitions and study populations were sufficiently comparable; where pooling was not appropriate, these studies were synthesised narratively. Studies without comparator groups (e.g., single-arm cohorts) were included in the systematic review and synthesised narratively, where they provided relevant evidence on prevalence, classification, or intervention outcomes. Studies were grouped into predefined evidence domains according to their primary contribution to the review. Comparative obstetric outcome studies contributed to the quantitative synthesis and meta-analysis. Diagnostic, prevalence, infertility, and classification studies were retained to provide contextual evidence regarding anomaly identification, subtype characterization, and population characteristics and were synthesised narratively. Intervention studies evaluating surgical management of congenital uterine anomalies were also retained for structured narrative synthesis. Only comparative obstetric outcome studies meeting predefined eligibility criteria contributed to pooled effect estimates. Outcome : The studies had to have at least one prespecified outcome: miscarriage; live birth; preterm birth (<37 weeks); stillbirth; fetal growth restriction or small-for-gestational-age (SGA); fetal malpresentation (non-cephalic presentation at ≥37 weeks); caesarean delivery; or ART outcomes (clinical pregnancy, implantation rate, ongoing pregnancy, live birth in ART context). When reported, severe maternal morbidity was also taken into account. We accepted study-specific outcome definitions (e.g., each study SGA threshold) but attempted to use standard definitions where feasible. Designs of Study : We incorporated primary research studies; randomized controlled trials (RCTs), non-randomized intervention studies, prospective or retrospective cohort studies, and case-control studies. Case reports and small case series (less than 10 cases) were omitted as well as reviews, editorials, conference abstracts, and studies that studied acquired uterine pathology only (fibroids, adenomyosis, intrauterine adhesions). Studies were eligible for inclusion in the systematic review if they met the predefined population, exposure, and outcome criteria, regardless of study design. However, only studies with an appropriate comparator group were included in the quantitative synthesis (meta-analysis). Non-comparative studies (e.g., single-arm cohorts, diagnostic studies, and post-intervention series) were included in the review and synthesised narratively where they contributed relevant evidence on prevalence, classification, or intervention outcomes. Timeframe and Language : We restricted the search to articles published between January 2015 and December 2025 to identify the recent studies based on the use of modern classification and imaging. Only publications in English were included. The final database search was performed on 31 December 2025, and no additional eligible studies were identified thereafter. The search strategy was developed with the help of a medical librarian. We searched MEDLINE (Ovid), Web of Science, and Cochrane CENTRAL for studies published from January 2015 to December 2025. The search terms were used as a combination of controlled vocabulary and keywords of congenital uterine anomalies (e.g. Mullerian anomaly, septate uterus, unicornuate uterus) and pregnancy/reproductive outcomes (miscarriage, preterm birth, infertility, IVF etc.). The entire strategy of the search is presented in Supplementary Appendix. Reference lists of any other eligible studies were also hand-searched in the included studies and any other relevant reviews. All the records were identified and imported into reference management software and all duplicates were eliminated. Relevance Titles and abstracts were screened by two reviewers. Independent evaluation of full-text articles of potentially eligible studies was then conducted against the inclusion criteria. Any discrepancies on one of the stages were resolved either through discussion or through third party adjudication. The PRISMA flow diagram (Fig. 1 ) summarizes the process of selection of the studies. Data were extracted independently by two reviewers using a standardized form. Such extracted data were: study characteristics (First author, year, country, design, setting); participant characteristics (age, parity, infertility or RPL history); CUA subtype and classification; diagnostic modality; surgical intervention details (where appropriate); ART protocol details (where appropriate); outcome definitions; follow-up period; and outcome data (event rates or effect estimates with confidence intervals). When primary outcome was available or derivable, we obtained unadjusted 2 × 2 contingent data. In cases where several measures of the effects were reported (e.g., adjusted ORs), we collected those to be used in qualitative synthesis but used crude data in primary meta-analyses so as to maximize comparability. The second reviewer cross-checked all data extraction and consensus were used to resolve any discrepancies. Two reviewers independently assessed risk of bias for each study using design-appropriate tools. For RCTs, we used the Cochrane ROB 2 tool [ 25 ]; for non-randomized intervention studies, ROBINS-I [ 26 ]; for observational cohort and case–control studies, the Newcastle–Ottawa Scale [ 27 ]. We interpreted NOS scores ≥ 7/9 as indicative of lower risk of bias (higher quality). We also considered key biases such as confounding (e.g., failure to adjust for maternal age, ART use) and outcome misclassification. Disagreements in risk-of-bias ratings were resolved by discussion. Table 3 presents a summary of risk-of-bias assessments. We performed a qualitative narrative synthesis for all 30 included studies, structured by analytic domain (natural history vs infertility vs surgical) and by CUA subtype where applicable. Study characteristics and findings are summarized in Table 1 (study characteristics) and Table 2 (population, exposure, comparator, and outcome categories). Pre–post intervention studies evaluating surgical correction (e.g., septum resection) were included to capture interventional outcomes; however, these studies were not included in the quantitative synthesis and were analysed separately within the narrative synthesis. Table 1 Characteristics of included studies ( n  = 30) StudyID Citation Citation Module (Domain) Design Setting/Country Sample Size (N) Study Period Comparator Type Quantitative Synthesis? Primary Contribution to Review Buicu2025 Buicu et al., 2025 [ 28 ] Buicu et al., 2025 A (Obstetric outcomes) Retrospective cohort Romania 62 2010–2017 Normal uterus Yes Obstetric Outcome Study (Meta-analysis) Mandelbaum Mandelbaum et al. [ 17 ], Mandelbaum et al. A (Obstetric outcomes) Registry cross-sectional USA 50,180 2016–2019 CUA subtype comparisons Yes (Exploratory) Obstetric Outcome Study (Subtype Analysis) Tellum2022 Tellum et al., 2022 [ 29 ] Tellum et al., 2022 A (Obstetric outcomes) Population cohort Denmark >1.4 million births 1997–2018 Normal uterus Yes Obstetric Outcome Study (Meta-analysis) Khander2017 Khander et al. [ 30 ], Khander et al. A (Obstetric outcomes) Retrospective cohort USA 283 2005–2016 Normal uterus Yes Obstetric Outcome Study (Meta-analysis) Matthews2019 Matthews & Chasen [ 16 ], Matthews & Chasen A (Obstetric outcomes) Retrospective cohort USA 132 2005–2018 Normal uterus Yes Obstetric Outcome Study (Meta-analysis) Naeh2021 Naeh et al. [ 18 ], Naeh et al. A (Obstetric outcomes) Retrospective cohort Israel 167 pregnancies 2010–2017 Normal uterus No Obstetric Outcome Study (Narrative Subtype Comparison) Qian2023 Qian et al. [ 31 ], Qian et al. A (Obstetric outcomes) Retrospective cohort China 336 2015–2021 Normal uterus vs hemi-uterus Yes Obstetric Outcome Study (Meta-analysis) Żyła2015 Żyła et al. [ 32 ], Żyła et al. A (Obstetric outcomes) Retrospective cohort Poland 94 1994–2012 Normal uterus Yes Obstetric Outcome Study (Meta-analysis) Wang2025 Wang et al. [ 33 ], Wang et al. A (Obstetric outcomes) Retrospective cohort China 21 twin pregnancies 2013–2022 Normal uterus (twins) No Obstetric Outcome Study (Sensitivity Analysis) Cai2021 Cai et al. [ 34 ], Cai et al. B (Infertility/ART) Retrospective cohort China 348 NR IVF controls Yes ART-Related Reproductive Outcome Evidence Prior2018 Prior et al. [ 35 ], Prior et al. B (Infertility/ART) Retrospective cohort UK 2,375 NR IVF controls Yes ART-Related Reproductive Outcome Evidence Rikken2021_RCT Rikken et al. [ 5 ], Rikken et al. C (Surgical) Multicentre RCT Netherlands 80 NR Septum resection vs expectant No Surgical Intervention Evidence Rousseau2021 Rousseau et al. [ 36 ], Rousseau et al. C (Surgical) Pre-post study Belgium 31 NR Pre vs post surgery No Surgical Intervention Evidence Zizolfi2025 Zizolfi et al. [ 37 ], Zizolfi et al. C (Surgical) Retrospective cohort Italy 107 NR Pre-post No Surgical Intervention Evidence Chang2023 Chang et al. [ 15 ], Chang et al. C (Surgical) Retrospective cohort China 348 NR Single-arm No Surgical Intervention Evidence DiSpiezio2021 Di Spiezio Sardo et al. [ 21 ] Di Spiezio Sardo et al. D (Classification) Consensus/Narrative Europe NR NR None No Classification and Clinical Context Evidence Carriles2024 Carriles et al. [ 38 ], Carriles et al. E (Prevalence) Prospective cohort Spain 993 NR None No Prevalence Evidence Ludwin2019 Ludwin et al. [ 39 ], Ludwin et al. D (Diagnostic) Retrospective cohort Poland 261 NR None No Diagnostic/Classification Evidence Ludwin2020 Ludwin et al. [ 6 ], Ludwin et al. D (Diagnostic) Diagnostic accuracy Poland 100 2014–2016 Method comparison No Diagnostic/Classification Evidence McCormack2016 McCormack et al. [ 40 ], McCormack et al. D (Diagnostic) Cross-sectional cohort Australia 200 NR None No Diagnostic Evidence Priya2015 Priya & Vijayalakshmi [ 41 ], Priya & Vijayalakshmi E (Prevalence) Cross-sectional India 1,500 2009–2014 None No Prevalence Evidence Okonkwo2022 Okonkwo et al. [ 42 ], Okonkwo et al. D (Diagnostic accuracy) Diagnostic accuracy Nigeria 88 NR Hysteroscopy reference No Diagnostic Accuracy Evidence Shiva2018 Shiva et al. [ 43 ], Shiva et al. D (Diagnostic accuracy) Diagnostic accuracy Iran 789 NR Office hysteroscopy No Diagnostic Accuracy Evidence Tammam2015 Tammam et al., 2015 [ 44 ] Tammam et al. D (Diagnostic) Diagnostic implementation Egypt 226 NR Diagnostic reference No Diagnostic Evidence Russo2022 Russo et al. [ 45 ], Russo et al. E (Observational) Retrospective cohort Italy 664 NR None No Contextual Reproductive Outcome Evidence Busnelli2024 Busnelli et al. [ 46 ], Busnelli et al. E (RPL cohort) Retrospective cohort Italy 442 NR None No Recurrent Pregnancy Loss Context Evidence Anbumani2015 Anbumani et al. [ 47 ], Anbumani et al. E (Prevalence) Cross-sectional India 5,228 2009–2014 None No Prevalence Evidence Joshi2021 Joshi et al. [ 48 ], Joshi et al. E (Prevalence) Retrospective cohort India NR NR None No Prevalence and Infertility Context Evidence Liston2017 Liston et al., 2017 [ 49 ] Liston et al. E (Descriptive) Registry descriptive UK NR NR None No Descriptive Clinical Context Evidence Roepke2017 Rasmark-Roepke et al. [ 50 ], Rasmark-Roepke et al. E (Registry cohort) Registry-based cohort Sweden 7,842 2003–2012 None No Population Context Evidence Studies were classified according to their primary contribution to the review. Comparative obstetric outcome studies contributed to the quantitative synthesis and pooled meta-analysis. Diagnostic, prevalence, infertility, classification, and surgical intervention studies were retained to provide contextual, mechanistic, or methodological evidence and were synthesised narratively Table 2 Population, exposure, comparator, and outcome classification of included studies (structured summary of study context and analysis approach) StudyID Population Exposure/Intervention Comparator Exposure ascertainment Confounding control Synthesis contribution Buicu2025 Delivering women CUA (any) Normal uterus Imaging/surgical Adjusted Meta-analysis + Narrative Mandelbaum_NIS Registry deliveries CUA subtypes Subtype referent ICD coding Adjusted Meta-analysis + Narrative Tellum2022 National cohort CUA Normal uterus Registry coding Adjusted Meta-analysis + Narrative Khander2017 Pregnant women CUA subtype Normal uterus Imaging NR Meta-analysis + Narrative Matthews2019 Unicornuate pregnancies Unicornuate uterus Normal uterus Imaging Unadjusted Meta-analysis + Narrative Naeh2021 CUA pregnancies Subtypes Normal uterus Surgical/imaging NR Narrative (comparative; not pooled) Qian2023 Hemi-uterus pregnancies Hemi-uterus Normal uterus Imaging Adjusted Meta-analysis + Narrative Żyła2015 Uterine malformations CUA subtypes Normal uterus Imaging NR Meta-analysis + Narrative Wang2025 Twin pregnancies Unicornuate uterus Normal twins Imaging NR Narrative (sensitivity only) Cai2021 IVF pregnancies Didelphys uterus ART controls Imaging Matched Meta-analysis + Narrative Prior2018 ART cohort CUA ART controls Imaging Unadjusted Meta-analysis + Narrative Rikken2021_RCT Septate uterus Septum resection Expectant Diagnosis NR RCT Narrative (comparative; not pooled) Rousseau2021 Septate uterus Septoplasty Pre–post Imaging Unadjusted Narrative (comparative; not pooled) Zizolfi2025 Surgical cohort Metroplasty Pre–post/none NR NR Narrative (non-comparative) Chang2023 Post-surgery cohort Metroplasty None Hysteroscopy Adjusted Narrative (non-comparative) DiSpiezio2021 Complex anomaly Septum treatment None ESHRE/ESGE NA Narrative (non-comparative) Carriles2024 General cohort Septate prevalence None Multiple criteria NA Narrative (non-comparative) Ludwin2019 Infertility cohort Septate definitions None Classification systems Unadjusted Narrative (non-comparative) Ludwin2020 Imaging dataset T-shaped uterus None CUME criteria NA Narrative (non-comparative) McCormack2016 RPL clinic Uterine anomalies None Imaging NA Narrative (non-comparative) Okonkwo2022 Infertility cohort Diagnostic tests Reference standard Accuracy study NA Narrative (comparative; diagnostic) Shiva2018 RPL/IVF Diagnostic test Reference standard Imaging vs hysteroscopy NA Narrative (comparative; diagnostic) Tammam2015 Hospital cohort 3D-TVUS Reference/none Imaging NA Narrative (non-comparative) Russo2022 Infertility cohort Septum metrics None Imaging Unadjusted Narrative (non-comparative) Busnelli2024 RPL cohort Screening None Imaging NR Narrative (non-comparative) Anbumani2015 Infertility cohort Prevalence None Imaging NA Narrative (non-comparative) Joshi2021 Infertility cohort Müllerian anomalies None Imaging NR Narrative (non-comparative) Priya2015 Community sample Uterine anomalies None Imaging NA Narrative (non-comparative) Liston2017 Caesarean cohort Incidental CUA None Intraoperative NA Narrative (non-comparative) Roepke2017 Registry cohort RPL patterns None Registry coding Adjusted Narrative (non-comparative) Table 3 Risk-of-bias assessment of included studies (overall judgments and primary concerns) StudyID Risk of Bias Tool Overall RoB Key concerns (briefly) Buicu2025 NOS (cohort) Serious Small single-center retrospective; likely residual confounding; limited adjustment Mandelbaum ROBINS-I (registry) Moderate ICD coding misclassification; delivery-only data; residual confounding Tellum2022 ROBINS-I (registry cohort) Moderate Registry exposure/outcome misclassification possible; most key confounders adjusted Khander2017 NOS (cohort) Moderate Retrospective design; heterogeneity in anomaly types; minimal adjustment for confounding Matthews2019 NOS (cohort) Serious Very small sample; incidental anomaly detection bias; limited control of confounding Naeh2021 NOS (cohort) Moderate Retrospective; comparator is general population (possible residual confounding) Qian2023 NOS (cohort) Moderate Retrospective; selected population (hemi-uterus); some matching but residual confounding likely Wang2025 NOS (cohort) Serious Extremely small N (case series of twins); high imprecision; likely selection bias Żyła2015 NOS (cohort) Serious Older data; limited reporting detail; potential selection bias (single hospital) Cai2021 NOS (cohort, ART matched)** Moderate Well-matched controls; however, only ART context – residual ART-related confounding Prior2018 NOS (cohort, ART)** Moderate Unadjusted comparisons; potential ART confounding; anomaly ascertainment NR Rikken2021_RCT Cochrane ROB 2 (RCT) Some concerns Open-label design; possible deviations from intended intervention; potential selective reporting Rousseau2021 NOS (pre–post study) Serious No concurrent control; confounding by indication (only poor prognosis patients); regression to mean Zizolfi2025 NOS (cohort) Serious Limited outcome reporting; likely selection bias (who underwent surgery); unclear comparator group Chang2023 NOS (cohort) Serious Post-surgery cohort without internal control; possible prognostic-factor bias; no randomization DiSpiezio2021 JBI case series/tech note Critical/High Technical step-by-step report; no comparator or outcomes; cannot infer effect on fertility Carriles2024 JBI prevalence (cross-sec) Moderate Large prospective sample; some selection bias possible (volunteers); verification bias in diagnostics Ludwin2019 JBI/QUADAS-2 (classification) Low–Moderate Within-cohort comparisons; potential measurement error in classification; moderate representativeness Ludwin2020 JBI/QUADAS-2 (criteria) Low–Moderate Standardized imaging study; applicability issues to general population; measurement cutoff subjectivity McCormack2016 JBI cross-sectional Moderate Clinic-based RPL cohort (not generalizable); no external control; possible selection bias Okonkwo2022 QUADAS-2 (diag. accuracy) Some concerns Limited sample; patient spectrum bias (infertile only); flow/timing of reference standard not fully clear Shiva2018 QUADAS-2 (diag. accuracy) Some concerns Selected high-risk population; unclear blinding of index vs reference test; acceptable overall Tammam2015 QUADAS-2 (implementation) Some concerns Single-center implementation study; incomplete reference standard for all; otherwise acceptable Russo2022 JBI/QUADAS-2 (classif/outcome) Some concerns Retrospective; multiple anomaly definitions (complex analysis); confounding by reproductive history Busnelli2024 NOS (cohort; RPL clinic) Some concerns Retrospective single-cohort study; selection bias (RPL patients); heterogeneity in outcomes definitions Anbumani2015 JBI cross-sectional Moderate Large descriptive study; potential selection bias (hospital-based screening); no reference standard confirmation Joshi2021 NOS/JBI (retro cohort) Moderate Infertility clinic cohort; anomaly ascertainment heterogeneity; confounding (e.g. age) not clearly addressed Priya2015 JBI cross-sectional Moderate Community ultrasound survey; no comparator; limited verification of anomaly findings; large sample helps reliability Liston2017 JBI case series (descriptive) Serious Incidental findings at surgery; likely incomplete case capture; no comparison; high risk of selection bias Roepke2017 ROBINS-I (registry) Moderate National register data; some confounders adjusted; coding validity concerns; outcomes broad (RPL incidence) Note: Overall risk-of-bias judgments: Low–Moderate (between low and moderate), Some concerns (applies primarily to QUADAS-2 diagnostic studies), Moderate, Serious, or Critical/High. Key concerns highlight dominant potential biases per study Table 4A Pooled odds ratios for primary obstetric outcomes (CUA vs normal uterus) Outcome k Pooled OR (95% CI) Heterogeneity (I 2 , τ 2 ) P Miscarriage 3 1.45 (1.13–1.86) I 2  = 0.0%; τ 2  = 0.000 <0.05 Preterm birth < 37w 5 3.74 (1.99–7.02) I 2  = 68.1%; τ 2  = 0.313 <0.001 Malpresentation 2 19.28 (6.27–59.24) I 2  = 23.6%; τ 2  = 0.232 <0.0001 Caesarean delivery 5 3.52 (1.11–11.09) I 2  = 94.9%; τ 2  = 1.583 0.034 Note: Random-effects model (DerSimonian–Laird). k  = number of studies. Malpresentation defined variably (Buicu: breech only; Qian: any non-cephalic) – interpret pooled estimate with caution Characteristics of included studies ( n  = 30) Studies were classified according to their primary contribution to the review. Comparative obstetric outcome studies contributed to the quantitative synthesis and pooled meta-analysis. Diagnostic, prevalence, infertility, classification, and surgical intervention studies were retained to provide contextual, mechanistic, or methodological evidence and were synthesised narratively Population, exposure, comparator, and outcome classification of included studies (structured summary of study context and analysis approach) Risk-of-bias assessment of included studies (overall judgments and primary concerns) Note: Overall risk-of-bias judgments: Low–Moderate (between low and moderate), Some concerns (applies primarily to QUADAS-2 diagnostic studies), Moderate, Serious, or Critical/High. Key concerns highlight dominant potential biases per study Pooled odds ratios for primary obstetric outcomes (CUA vs normal uterus) Note: Random-effects model (DerSimonian–Laird). k  = number of studies. Malpresentation defined variably (Buicu: breech only; Qian: any non-cephalic) – interpret pooled estimate with caution For quantitative synthesis, we conducted meta-analyses when at least two studies provided comparable outcomes and study populations. Primary meta-analyses focused on the four principal obstetric outcomes: miscarriage, preterm birth < 37 weeks, fetal malpresentation, and caesarean delivery. We used DerSimonian–Laird random-effects models to account for between-study heterogeneity [ 51 ]. Pooled effect sizes were expressed as odds ratios (OR) with 95% confidence intervals. We calculated crude ORs from raw event counts for each study (2 × 2 tables) to ensure consistency. Between-study heterogeneity was assessed using Cochran’s Q and the I 2 statistic (with I 2  > 50% indicating substantial heterogeneity) [ 52 ]. We also reported τ 2 as the estimate of between-study variance. We performed prespecified subgroup and sensitivity analyses to explore heterogeneity. Subgroup analyses (where data allowed) included stratification by CUA subtype, diagnostic method (imaging-only vs surgical confirmation), and study risk of bias rating. Sensitivity analyses were conducted to investigate the effects of the exclusion of outlier studies (e.g. cohorts with very high baseline risk) and studies conducted on ART populations. We also performed sensitivity analyses excluding all studies that included arcuate uteri to assess the impact of classification variability on pooled estimates. Although fetal growth restriction (FGR) and small-for-gestational-age (SGA) are clinically important outcomes, they were not included in the primary meta-analyses due to substantial heterogeneity in their definitions across studies. SGA was variably defined using different percentile thresholds (e.g., <10th, <5th), while FGR was inconsistently reported and often lacked standardized diagnostic criteria incorporating fetal Doppler parameters or consensus definitions. This variability limited comparability and increased the risk of outcome misclassification in pooled analyses. Therefore, these outcomes were synthesised narratively to preserve interpretability. We evaluated the certainty of evidence for each primary outcome using GRADE. Evidence was downgraded for risk of bias, inconsistency, indirectness, imprecision, or publication bias as applicable (Guyatt et al., 2008). All statistical analyses were conducted using Review Manager (RevMan 5.4) and R (v4.0) software (R Core Team, 2020) Publication bias was assessed for outcomes with a sufficient number of studies (k ≥ 10) using visual inspection of funnel plots and Egger’s regression test. For outcomes with fewer studies, formal assessment was not performed due to limited statistical power, and any interpretation of potential small-study effects was undertaken with caution.

This review synthesizes current evidence indicating that congenital uterine anomalies are associated with adverse pregnancy outcomes and highlights that persistent gap in the existing literature remain. Major congenital uterine anomalies, including septate, bicornuate, unicornuate, and didelphys uteri, are associated with increased risks of miscarriage, preterm birth, malpresentation, and caesarean delivery. Nevertheless, the severity of these risks, as well as the effectiveness of interventions (septum resection), is rather ambiguous because of the limitations in the number of available robust studies. The additional high-quality studies, such as larger prospective cohorts and randomized trials are urgently required to sharpen our perception of the outcome variability between the subtypes of anomalies and what kind of therapeutic intervention could possibly raise reproductive success in this group. In the future, enhanced standardization of the diagnosis and classification of uterine anomalies [ 39 ] will be essential; this will allow researchers and clinicians to align terminology when evaluating the effects of these rare yet significant disorders. By addressing these evidence gaps, we can better counsel patients with congenital uterine anomalies and develop management strategies to optimize their chances for a healthy pregnancy outcome.

The systematic review confirms that congenital uterine anomalies (CUA) are associated with increased risks of adverse pregnancy outcomes. Women with uterine malformations have an increased risk of miscarriage, preterm birth, and other adverse pregnancy outcomes [ 28 , 32 ]. In a cohort of 62 women (26 with CUA and 36 controls) it was found that approximately 54% of pregnancies in the anomaly group proceeded to a live birth and 46% to miscarriage or loss compared to a much greater live birth in the control group. Similarly, a comparative study of unicornuate uteri revealed significantly low live-birth rates and high risks of ectopic pregnancy, miscarriage, and preterm birth in women with unicornuate uterus compared to controls [ 29 ]. The results highlight that congenital uterine defect - in particular, more severe anomalies - have a significant adverse impact on reproductive outcomes, despite the fact that the prognosis of an individual patient may differ based on the type of anomaly and patient history. In our findings, preterm birth is one of the key complications related to anomalies of the uterus. In the literature, about 25–30% of women with CUA give birth to preterm babies, which is which is substantially higher than the overall obstetric population [ 17 , 18 ]. An example is a national inpatient sample study of more than 50,000 births of women with CUA: the rate was 26.8% preterm births, with a high rate of very preterm (<33 week) births in some subgroups of anomalies. Smaller clinical series also note that preterm labor is one of the most frequent complications in anomaly pregnancies: In one of the single-center studies, 55% of pregnancies with different uterine anomalies were associated with preterm delivery. Among women with uterine anomalies with a history of preterm birth, the risk of recurrence is particularly great − 60% in one group of studies [ 30 ]. The high risk of preterm birth is biologically plausible due to the small size of the uterine cavity and cervical insufficiency that is common in the malformations [ 56 – 58 ]. There is also some emerging evidence that anomalies can also interfere with normal uterine contractility and blood supply, also leading to early labor [ 37 ]. Fetal malpresentation was another consistently elevated risk in women with congenital uterine anomalies. Abnormal uterine anatomy often constrains fetal movement and growth, leading to a high incidence of non-cephalic presentations (such as breech) at term [ 18 , 31 ]. In a tertiary care cohort, 42–43% of pregnancies in women with CUA had a malpresentation at delivery, a striking contrast to the ~4% breech rate in unselected populations. Malpresentation in turn contributes to the markedly high rates of cesarean delivery observed in this patient group. Virtually all large studies of obstetric outcomes in women with uterine malformations report a 65–80% overall cesarean section rate [ 18 , 28 , 32 ]. Although fetal malpresentation is a recognised indication for caesarean delivery, the magnitude of association between these outcomes is not necessarily proportional. Caesarean delivery is influenced by multiple obstetric factors beyond presentation, including prior caesarean section, fetal condition, and clinical practice patterns, which may contribute to the observed differences in effect size. Indicatively, cesarean section has been reported to be 63.5% in the case of anomaly-related pregnancy in the case of an Israeli single-center study [ 18 ], and 78% among women with Mullerian duct anomalies in a Polish series. Even within the framework of the contemporary practice in obstetrics, the majority of clinicians choose surgical delivery due to the high rate of breech presentation and other complications in such pregnancies. Interestingly, a recent study that involved anatomically “hemi-uterus” anomalies (unicornuate, didelphic, complete bicornuate) reported a cesarean section rate of more than 90% in the anomaly cohort with malpresentation being the most frequent manifestation [ 31 ]. These findings suggest the clinical reality congenital uterine anomalies, often require operative delivery as a measure of their optimization of maternal-fetal outcomes. The stronger association observed for fetal malpresentation compared with caesarean delivery likely reflects differences in the clinical nature of these outcomes. Malpresentation represents a more direct anatomical consequence of uterine cavity distortion, whereas caesarean delivery is influenced by multiple additional obstetric and institutional factors, including prior caesarean delivery, gestational age, fetal condition, and local delivery practices. Consistent with this pattern, Qian et al. [ 31 ] reported markedly increased risks of malpresentation in women with hemi-uterus, while caesarean delivery reflected broader obstetric decision-making beyond fetal presentation alone. Similarly, Tellum et al. [ 29 ] observed increased caesarean delivery rates in unicornuate uterus, but with more moderate adjusted effect estimates than those observed for malpresentation. Although external cephalic version may be technically challenging in structurally distorted uterine cavities, this outcome was inconsistently reported across studies. Collectively, these findings suggest that congenital uterine anomalies may have a more direct anatomical relationship with fetal presentation than with delivery mode, which remains influenced by broader obstetric and clinical decision-making factors. There were interesting intergroup differences - and inconsistencies - of obstetric risk profile between different subtypes of anomalies. In the national sample, every subtype posed unique risks: e.g. uterine didelphys had the highest preterm birth rate (almost 35% of births were preterm), but septate uteri presented a higher risk of maternal complications (such as hemorrhage). The review revealed that anomalies with the unilateral or hemi-uterus configuration (unicornuate and didelphic uteri, and complete bicornuate uteri) are likely to have a similar pattern of unfavorable obstetric outcomes, with the risks of preterm birth, prelabor rupture of membranes, malpresentation, and following cesarean birth being notably high [ 31 ]. This is in line with the fact that, a small and single-sided uterine cavity has restricted space and irregular vascular supply to a developing fetus. A septate uterus - an abnormality in the middle of an otherwise normally sized uterus - could, however, have different risk profile, being more closely linked to first-trimester miscarriage and potentially less late-pregnancy complications such as malpresentation. In fact, septate uteri are traditionally associated with frequent pregnancy loss (they are usually mentioned as the most frequent uterine anomaly with the highest miscarriage rate) and the relatively low frequency of preterm birth or fetal growth retardation [ 36 ]. It should be stated though that the evidence regarding the difference of subtypes is not consistent. In a single retrospective study that directly compared the outcomes of women with unification defects of the bicornuate type ( n  = 4061), didelphys type ( n  = 1758), and unicornuate type ( n  = 2365), and women with canalization defects of the septate uterus type ( n  = 1758), the majority of the obstetric outcomes were not statistically different, despite controlling confounders [ 18 ]. In that study, the initial difference in very preterm birth and placental abruption was higher in septate uteri compared with fusion defects but was not observed when adjusted with multivariate analysis. Therefore, whereas large population data indicate that each subtype of anomaly is associated with distinct risks, smaller controlled studies suggest that patient factors and underlying risk (not necessarily the subtype itself) can contribute to a significant number of the outcome differences. And this is a position that our review endorses: high-risk obstetric care should be given to women with any major congenital uterine anomaly, but the type of complication to expect may differ to some degree depending on the type of anomaly. The managerial implications of these results are multifaceted. In the case of septate uterus - the most prevalent malformation in women with recurrent spontaneous abortion - the research question has always been whether surgery can enhance its prognosis. The evidence we included contradicted on this fact. On the one hand, several observational studies and case series indicate that hysteroscopic septum resection has the potential to significantly lower the rate of miscarriage and the number of live births in women with a septate uterus [ 36 ]. A recent pre-post study of 101 patients, e.g., showed a decreased miscarriage rate from 35.6% pre-surgery to 14.8% after septoplasty and the live-birth rate doubled (18% to 37%) in the first year after surgery. These results have encouraged a number of clinicians to recommend septum resection in patients with recurrent pregnancy loss or infertility due to a uterine septum. The first randomized controlled trial (the TRUST trial) reported no superiority of septum resection over expectant management in support of better reproductive outcomes [ 5 ]. In that multicenter RCT, live birth was seen in 31% of women who were randomized to the septum resection and 35% of those who were randomized to the expectant management and there was no significant difference in miscarriage or preterm birth in both groups. The null finding of the trial has been disputed, partly because of the characteristics of its inclusion criteria (many participants had small, partial septa) and sample size, but offers the strongest evidence to date challenging the utility of elective septoplasty in the behavior of pregnancy [ 5 , 36 ]. So, the importance of septum resection is controversial: although the results provided by observational studies are rather encouraging, high-quality evidence has not proved the existence of an overt reproductive advantage. In the case of other types of anomalies, evidence-based interventions are even more limited. Such operations as surgical “unification” of bicornuate or didelphys uteri (metroplasty) are now rarely performed, and none of the included studies assessed such operations. Some patients with unicornuate uteri or other anomalies at risk of second-trimester loss may benefit from cervical cerclage or close cervical surveillance, as suggested by case series [ 28 ]. Nevertheless, there are no strong data regarding preventive measures (prophylactic cerclage) in conditions where congenital anomalies of the uterus are present, and this monitoring should be personalized in the absence of guidelines. One complication in the synthesis of the literature regarding the congenital anomalies of the uterus is the non-homogeneity of the methods and classification systems of diagnosis across the investigations. Most of the earlier researches employed hysterosalpingography or two-dimensional ultrasound to detect abnormalities, which are not effective in describing the anatomy of the uterus and unable to differentiate between some of the subtypes [ 44 ]. Conversely, three-dimensional ultrasound and MRI have been applied in much more recent studies, and they significantly advance the accuracy of the diagnoses of Mullerian abnormalities [ 42 , 43 ]. The extensive use of 3D transvaginal ultrasounds has probably elevated the number of previously unknown anomalies and enabled more optimal subclassification of the anomaly types [ 42 ]. Inconsistent reporting may occur even when there is improved imaging as a result of variation in the criteria used to classify the image. The ESHRE/ESGE classification system employs stricter anatomical criteria, which may result in a greater proportion of uteri being classified as septate compared with the older ASRM system, where similar morphologies may have been categorised as normal (arcuate). [ 6 , 22 , 39 ]. Ludwin et al. [ 39 ] compared the two classification systems on the same population and concluded that the ESHRE/ESGE criteria recognized septate uteri in much more women than the traditional criteria, which may be over diagnosing minor variants of septum. This lack of consistency has practical, clinical consequences: it can result in over-treatment (surgery on milder abnormalities, which might not have posed any problems), or, conversely, in being unable to compare studies that defined what a significant uterine abnormality is differently. According to some authors, the absence of a universal definition of anomalies, such as a septate uterus, is one of the problems preventing the formulation of effective conclusions regarding the effectiveness of management [ 36 ]. Thus, future research must endeavor to employ standardized, evidence-based diagnostic criteria of congenital uterine anomalies and present findings in defined subtype, to increase comparability and application of findings. The findings of this review should be interpreted in light of both study-level and review-level limitations. Limitations of included studies : Most included studies were observational in design, with inherent risks of residual confounding and selection bias. There was considerable heterogeneity in study populations, classification of congenital uterine anomalies, and outcome definitions. In addition, variability in diagnostic modalities and inconsistent reporting of adjusted effect estimates may have influenced the pooled estimates. Some of them were small-scale studies or single-center series that reduced the statistical power and generalizability (e.g., some of the cohorts contained less than 30 cases of anomalies [ 16 , 28 ]. Some studies did not have a proper internal control group of women without anomalies, which makes it hard to measure the excess risk due to the anomaly itself [ 15 , 37 ]. Inconsistency in outcome definitions and patient population was also high. For example, some studies were based exclusively on women with recurrent miscarriage or infertility only [ 46 , 48 ], and others on all pregnant women with an anomaly irrespective of obstetric history [ 17 ]. Such variability may introduce selection bias and limit comparability across studies. Also, a study relying on administrative data or registries can be misclassified; one validation study reported concerns over the quality of diagnostic coding of uterine anomalies, and pregnancy loss in regular medical records [ 50 ]. In a large U.S. hospital data set analysis, bicornuate uteri represented a surprisingly large percentage of cases (more than 70), probably due to over-coding of bicornuate cases against other anomalies in practice. These peculiarities of coding and identifying the cases may bias risk estimates by the subtype. Considering such constraints, quantitative estimates of risk used in this review (e.g., a three times increased risk of preterm delivery or miscarriage due to a uterine defect) be should be interpreted with caution. The actual extent of risk can vary in more strictly defined groups. Limitations of the present review : At the review level, the number of studies contributing to some of the assessed outcomes was limited, restricting the precision of pooled estimates and precluding robust assessment of publication bias. Differences in study design and reporting also limited the ability to perform subgroup and sensitivity analyses for certain outcomes. Furthermore, inclusion of studies across diverse clinical settings may have contributed to inter-study heterogeneity. Data on the risk of uterine rupture in subsequent pregnancies following caesarean delivery in women with congenital uterine anomalies were limited and inconsistently reported across studies. However, despite the uncertainty that must be admitted, the pattern of directional results in different studies increases the conclusion that the congenital uterine anomalies produce a clinically significant effect on the results of the pregnancies. An additional limitation is that the review protocol was not prospectively registered in PROSPERO. Although the methodology, eligibility criteria, and planned analyses were defined a priori and adhered to throughout the review process, the absence of formal protocol registration may limit external verification of methodological decisions and introduces the potential for unrecognized reporting bias.

Congenital uterine anomalies (CUAs) arise from abnormal Müllerian duct development and encompass a heterogeneous group of structural variants – most commonly septate, bicornuate, unicornuate, and didelphys uteri [ 1 , 2 ]. The estimated prevalence of CUA in the general female population is approximately 3–6% [ 3 , 4 ], but CUAs are substantially overrepresented among women with infertility, recurrent pregnancy loss (RPL), and adverse obstetric outcomes [ 5 ]. Importantly, the pattern and magnitude of risk vary by anomaly subtype. Septate and unicornuate uteri are consistently associated with poorer reproductive and perinatal outcomes than fusion anomalies such as bicornuate or didelphys uteri [ 4 , 6 ]. Multiple mechanisms have been proposed to link CUAs with adverse outcomes. Structural distortion of the uterine cavity, reduced functional volume, impaired endometrial receptivity, abnormal placentation, and altered uterine contractility may all contribute to higher risks of miscarriage, fetal growth restriction, preterm birth, malpresentation, and caesarean delivery [ 7 ]. These pathophysiological differences across subtypes likely underlie the observed heterogeneity in outcomes, underscoring the clinical importance of subtype-specific risk stratification [ 8 ]. Previously conducted observational studies and meta-analyses have shown a relationship between CUA and adverse outcomes. Among these adverse outcomes, fetal growth abnormalities represent a particularly complex and less well-characterized domain. While associations between congenital uterine anomalies and reduced fetal size are increasingly reported, the underlying biological mechanisms remain insufficiently defined and are often inconsistently interpreted across studies. A key challenge in this area is the distinction between fetal growth restriction (FGR) and small-for-gestational-age (SGA), which are frequently used interchangeably in the literature but represent fundamentally different entities. FGR reflects a pathological process, most commonly driven by placental insufficiency, whereas SGA is a statistical definition based on fetal size and may include constitutionally small but otherwise healthy fetuses [ 9 ]. Emerging evidence suggests that congenital uterine anomalies are associated with increased risks of both SGA and clinically defined FGR, with adjusted risk estimates for growth restriction approaching two-fold in some large cohort studies [ 10 , 11 ]. However, the underlying mechanisms remain uncertain. While abnormal uterine morphology may impose mechanical constraints on fetal growth, leading to SGA phenotypes, it is unclear whether these anomalies also contribute to placental dysfunction sufficient to result in true FGR. Notably, the majority of studies in this field report SGA outcomes without standardized diagnostic criteria for FGR, limiting mechanistic interpretation [ 11 , 12 ]. In non-anomalous pregnancies, FGR and severe SGA are strongly associated with placental lesions such as maternal vascular malperfusion and fetal vascular malperfusion, which reflect impaired uteroplacental and fetoplacental circulation [ 13 ]. However, studies of congenital uterine anomalies rarely include systematic placental histopathological evaluation. As a result, direct evidence linking specific uterine anomaly subtypes to defined placental lesions is lacking [ 14 ]. This represents a critical gap in the literature, as it remains unclear whether growth abnormalities in these pregnancies are primarily driven by placental pathology or by structural and biomechanical constraints of the uterine environment Chang et al. [ 15 ] found high rates of first trimester miscarriage and preterm birth, and reduced live birth rates in women with CUA. Recent studies with national data have identified specific risks of fetal malpresentation, operative birth and preterm birth that are notably high in CUA pregnancies [ 16 – 18 ]. However, it is still unclear which are clinically significant. A large part of the evidence is retrospective and inconsistent in its anomaly classification and variable outcomes. Past syntheses often precede the contemporary classification systems (ESHRE/ESGE 2013 consensus and the revised 2021 ASRM schema) and lack the ability to incorporate the data of the assisted reproductive technology (ART) cohorts or large-scale population databases [ 1 , 19 , 20 ]. Previous reviews have frequently focused on isolated reproductive endpoints (such as miscarriage or infertility outcomes) rather than providing an integrated assessment of both maternal and fetal outcomes across the continuum of pregnancy [ 4 ]. In the meantime, the growing awareness of CUA in the normal antenatal medical check-up has increased the demands of evidence-based information on risk stratification, sensitization, and delivery planning. Recent European guidelines suggest that pregnancies with uterine anomalies are to be considered at risk but no specific management-informing data are available [ 21 , 22 ]. The presence of more modern, larger cohorts and better methodological provides an opportunity to have a more rigorous, up-to-date synthesis of outcomes in women with CUA. This study aimed to systematically evaluate the association between congenital uterine anomalies and adverse obstetric outcomes, including miscarriage, preterm birth, fetal malpresentation, and caesarean delivery, compared with women with normal uterine anatomy, and to explore differences across anomaly subtypes where data were available. We thus performed a systematic review and meta-analysis of maternal and fetal outcome in pregnancy complicated by congenital uterine anomalies. The review was conducted using PRISMA 2020 guidelines. 17 and observational study recommendations (MOOSE) [ 23 , 24 ]. In addition to observational studies, we included randomized and non-randomized intervention studies to capture evidence on surgical management, particularly septum resection; however, these were synthesised narratively and not pooled with observational data to avoid methodological heterogeneity. This approach enabled a comprehensive evaluation of both the natural history and interventional outcomes associated with congenital uterine anomalies while maintaining analytical rigor and producing clinically relevant estimates to inform patient counselling, antenatal management, and future guideline development.