The impact of maternal cardiovascular status prior to labor on birth outcomes: an observational study

Statistical analysis

We divided our sample in two groups: patients with an APO and patients with uncomplicated ones.

Continuous variables were expressed as median with interquartile range, categorical variables were expressed as number and percentage. Group comparisons were performed using Student’s t-test or Mann-Whitney U test, as appropriate based on data distribution. We performed a Chi-Squared test for the comparison of two proportions (from independent samples). To account for multiple comparisons, p-values were adjusted using the Benjamini-Hochberg false discovery rate (FDR) procedure applied only to independent variables, with significance set at FDR <0.05.

A receiver operating characteristics (ROC) curve was constructed for each hemodynamic, cardiotocography and demographic parameter, to identify the best cut-off and convert the continuous variables into categorical ones. For each individual predictive variable, diagnostic performance metrics – including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) – were calculated.

We performed a univariate binary logistic regression analysis and odds ratios (ORs) were calculated. In conclusion, a multivariate binary logistic regression analysis was performed to identify independent predictors. To internally validate the predictive performance of the model, the area under the receiver operating characteristic curve (AUC) was calculated with 95 % confidence intervals estimated using bootstrap resampling (1,000 iterations). The Hosmer–Lemeshow goodness-of-fit test was also applied to assess model calibration.

To assess the clinical usefulness of the diagnostic tests and prediction models, the number needed to predict (NNP) was calculated. The NNP represents the average the number of patients who need to be examined in the patient population to correctly predict the diagnosis of one person [29].

Unlike the simple inverse of the Youden index, which does not consider disease prevalence, we calculated the NNP by incorporating the prevalence of the event in our study population to better reflect the real-world clinical context. The NNP was computed using the following formula [29]:

A post-hoc sample size assessment was performed using the formula proposed by Riley et al. [30] to verify whether the available sample size was sufficient to meet the criteria for model stability and minimal overfitting.

A secondary analysis was performed to evaluate the predictive performance of variables for “objective APO”, defined as the presence of at least one of the following: 5-min Apgar score <7, umbilical artery pH <7.1 and/or base excess >12 mmol/L at birth, or NICU admission. Operative delivery (caesarean section or instrumental vaginal delivery) for suspected fetal distress, included in the primary composite APO definition, was excluded from this endpoint.

The statistical analysis was performed using MedCalc Statistical Software (MedCalc Software bvba, Ostend, Belgium; https://www.medcalc.org; 2016). Statistical significance was set to p<0.05.

Primary analysis

Among the 307 patients with spontaneous onset of labor, APO were observed in 41 cases (13.36 %) (Table S1). Table 1 summarizes the maternal and fetal characteristics at the time of enrollment. There were no significant differences between patients with and without APO in terms of gestational age, maternal age, BMI, weight, height, smoking status, fetal Doppler parameters, or amniotic fluid index. However, the rate of multiparity was significantly higher in patients with uncomplicated outcomes (27.07 vs. 7.32 %, p=0.01) and remained significant after adjustment for multiple comparisons using the Benjamini-Hochberg correction. Among hemodynamic parameters, SVR was significantly higher in patients with APO [1,353 (1,209–1,498) dyn s cm⁻⁵ vs. 1,249 (1,071–1,438) dyn s cm⁻⁵, p=0.01] (Figure 1), and this difference also remained significant after Benjamini-Hochberg adjustment, while other maternal hemodynamic measures (CO, MAP, INO, PKR) did not differ significantly.

Figure 1:

Box-and-whisker plot of systemic vascular resistance in patients with uncomplicated delivery and with adverse perinatal outcome (APO).

Patients with APO also exhibited significantly lower STV [7.5 (6.2–10) ms vs. 9.05 (7.6–11) ms, p<0.01] compared to those with uncomplicated outcomes (Figure 2), and this difference also remained significant after Benjamini-Hochberg correction.

Figure 2:

Box-and-whisker plot of short term variability in patients with uncomplicated delivery and with adverse perinatal outcome (APO).

There were no significant differences between the two groups in terms of gestational age at birth, neonatal weight, or the use of epidural analgesia (78.05 vs. 80.83 %) (Table 1).

A receiver operating characteristic (ROC) curve analysis was performed to determine the optimal cut-off values for variables showing significant differences between the two groups (Figure 3). The best thresholds identified were SVR >1,135 dyn s cm⁻⁵ [AUC 0.62 (95 % CI, 0.56–0.67), sensitivity 92.68 % (95 % CI, 80.08–98.46 %), specificity 35.34 % (95 % CI, 29.60–41.41 %), PPV 18.10 % (95 % CI, 13.13–23.98 %), NPV 96.91 % (95 % CI, 91.23–99.36 %), p<0.01] and STV ≤7 ms [AUC 0.644 (95 % CI, 0.588–0.698), sensitivity 43.90 % (95 % CI, 28.47–60.25 %), specificity 83.46 % (95 % CI, 78.44–87.72 %), PPV 29.03 % (95 % CI, 18.20–41.95 %), NPV 90.61 % (95 % CI, 86.25–93.96 %), p<0.01]. Parity showed a sensitivity of 92.68 % (95 % CI, 80.08–98.46 %), a specificity of 27.07 % (95 % CI, 21.82–32.83 %), a PPV of 16.38 % (95 % CI, 11.86–21.78 %), and a NPV of 96.00 % (95 % CI, 88.75–99.17 %) in identifying adverse outcome.

Figure 3:

ROC curve illustrating the predictive performance of systemic vascular resistance (A) and short-term variability (B) for adverse perinatal outcomes (APO).

Univariate binary logistic regression analysis was conducted, and ORs were calculated (Table 2). A multivariate logistic regression analysis was then performed, incorporating multiparity, STV ≤7 ms, and SVR ≥1,135 dyn s cm⁻⁵. The final multivariate model demonstrated a moderate diagnostic performance [AUC 0.761 (95 % CI, 0.709–0.807)] (Table 2). Internal validation using bootstrap resampling (1,000 iterations) yielded a 95 % confidence interval for the AUC of 0.682–0.818, confirming the model’s stability. The Hosmer–Lemeshow test indicated excellent calibration of the model (χ²=0.29, p=0.990), suggesting an adequate fit between observed and predicted outcomes (Supplementary Figure 1).

To test the robustness of our multivariate model, we performed an additional sensitivity analysis excluding borderline cases, defined as values within ±5 % of the cut-off thresholds for SVR and STV. This analysis confirmed the stability of the model, with an AUC of 0.738 (95 % CI 0.676–0.795). The odds ratios for the three variables included in the model are provided in Supplementary Table 2.

Based on a Cox & Snell R² of 0.1177 estimated from our data, an outcome prevalence of 13.36 %, and a model including three predictor parameters, the minimum required sample size according to the Riley et al. formula was 215 patients [30]. Our study population included 307 patients, confirming that the sample was adequate for reliable multivariable model development.

The NNP was calculated for individual and combined predictors. For SVR >1,135 dyn s cm⁻⁵, the NNP was 6.66; for STV ≤7 ms, 5.09; and for parity alone, 8.08. In multivariable models, the NNP was 6.90 when at least one of the three parameters (SVR, STV, or parity) was present, 5.61 when at least two were present, and 2.89 when all three predictors were simultaneously present.

Secondary analysis

A secondary analysis was also performed focusing exclusively on “objective” APO, the results of which are provided in Table 3. In this analysis, both SVR and STV proved to be strong predictors of objective APO, whereas parity did not emerge as a significant predictor of objective endpoints. The multivariable model including parity, SVR, and STV demonstrated excellent predictive performance [AUC 0.941 (95 % CI, 0.908–0.964), sensitivity 100 % (95 % CI, 47.80–100 %), specificity 72.85 % (95 % CI, 67.5–77.8 %)], comparable to the model including only SVR and STV [AUC 0.931 (95 % CI, 0.896–0.957), sensitivity 100 % (95 % CI, 47.80–100 %), specificity 72.85 % (95 % CI, 67.5–77.8 %)].

Main findings

This study highlights the importance of antepartum computerized cardiotocography and maternal hemodynamic assessment in identifying patients at risk of adverse perinatal outcomes (APO) within a low-risk population for intrapartum hypoxia. Specifically, our findings indicate that SVR >1,135 dyn s cm⁻⁵, STV ≤7 ms, and multiparity are independent predictors of adverse outcomes.

Interpretation and comparison with existing literature

Our results are consistent with existing scientific literature on this topic. Multiparity has been consistently identified as a protective factor against adverse maternal and fetal outcomes, with our findings yielding an OR comparable to that reported by Kalafat et al. (OR 0.20, 95 % CI 0.06–0.56, p=0.003) [31].

Previously, our group reported a higher incidence of complications in women with SVR >1,069 dyn s cm⁻⁵ during early labor [21]. Notably, this threshold is close to the value identified in this study (1,135 dyn s cm⁻⁵), where SVR emerged as the strongest independent predictor of APO. The slightly higher cut-off observed here is likely attributable to methodological differences: in this study, measurements were obtained at term in a resting, pre-labor condition, whereas the previous study assessed women at the onset of active labor, when CO is physiologically increased, resulting in lower SVR values. Importantly, these differences are minimal and do not alter the clinical interpretation, as both studies consistently demonstrate that elevated SVR is strongly associated with adverse perinatal outcomes.

Moreover, Kalafat et al. demonstrated that elevated SVR, assessed using USCOM, correlates with an increased risk of operative delivery due to presumed fetal compromise in women undergoing labor induction [31].

Women who developed complications during labor exhibited a distinct hemodynamic profile characterized by high vascular resistance and low CO, indicative of cardiovascular maladaptation. This hypodynamic circulation was significantly associated with the development of fetal distress during labor. Cardiovascular adaptation is essential for maintaining adequate uteroplacental perfusion and ensuring sufficient fetal oxygen supply during pregnancy and labor [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32].

Labor represents a major physiological challenge for the feto-placental unit, as uterine contractions transiently reduce blood flow to the fetus [33]. Consequently, the onset of labor may further compromise placental function, increasing the risk of hypoxic perinatal events. Our previous studies demonstrated a relationship between hypodynamic circulation and subclinical placental dysfunction, suggesting that maternal hemodynamic abnormalities may be clinically silent until labor begins.

In uncomplicated pregnancies, vascular remodeling results in SVR reduction and CO augmentation, which are critical for maintaining adequate uterine perfusion and preventing fetal hypoxia. During labor, CO increases by 30 % in the first stage and by over 50 % in the second stage. Women with elevated SVR and poor cardiac performance at term face an increased risk of developing intrapartum complications. Subclinical alterations in cardiac function may underlie reduced uteroplacental perfusion, leading to fetal distress.

These findings suggest that maternal hemodynamic alterations, characterized by elevated SVR and decreased CO, reduce uteroplacental perfusion and compromise fetal oxygenation, particularly during labor when metabolic demands increase and placental circulation is further stressed.

In our study, fetal Doppler assessment before delivery was not significantly associated with subsequent adverse outcomes. Previous studies have reported that antenatal and early labor Doppler evaluation of cerebral redistribution is linked to an increased likelihood of vacuum extraction or cesarean section due to suspected fetal distress, though with poor predictive performance [34], [35], [36].

No significant differences have been reported between the use of antenatal CTG vs. no CTG in terms of perinatal mortality, cesarean section rate, Apgar scores <7 at 5 min, NICU admission, or gestational age at birth. However, computerized CTG, as opposed to traditional CTG, has been associated with a significant reduction in perinatal mortality (RR 0.20, 95 % CI 0.04–0.88), while no significant differences were found for cesarean section rate, Apgar scores <7 at 5 min, NICU stay, or gestational age at birth [37]. Notably, most studies evaluating CTG involve high-risk pregnancies [37], whereas few observational studies have focused on antenatal CTG in low-risk pregnancies, often with limited sample sizes. To date, the use of traditional CTG in low-risk pregnancies has demonstrated poor predictive capability for APO [38].

A prior study indicated that computerized CTG during labor in low-risk pregnancies effectively identifies fetal distress, with STV emerging as a strong independent predictor of fetal acidemia (OR 20.96, p<0.001) [39].

In our study, antenatal computerized CTG was predictive of perinatal complications, including cesarean section or operative vaginal delivery for pathological CTG, Apgar scores <7 at 5 min, neonatal metabolic acidosis, and NICU admission. Our findings suggest that computerized CTG may serve as a valuable tool for identifying pregnancies with subclinical maternal hemodynamic compromise, which are at higher risk of intrapartum hypoxic events.

Clinical implications

Intrapartum fetal distress in apparently low-risk pregnancies is a critical issue, as many vulnerable fetuses remain undetected until labor, when reduced placental reserve can lead to hypoxia, urgent delivery, and serious neonatal complications. In most cases, fetal distress during labor is the result of preexisting placental dysfunction, which limits the fetus’s ability to tolerate the stress of uterine contractions [40]. Identifying these pregnancies before the onset of labor could allow targeted surveillance and even the consideration of preventive interventions aimed at improving placental perfusion. However, current clinical trials exploring potential cardiovascular therapies, such as phosphodiesterase-5 inhibitors, do not incorporate maternal hemodynamic assessment into their inclusion criteria – potentially overlooking a subgroup of patients who might derive the greatest benefit.

To evaluate the clinical applicability of our findings, we calculated the NNP, a metric that quantifies the efficiency of a diagnostic tool in identifying true positive cases. The relatively low NNP values observed for both individual parameters and combined models suggest that these tests may hold practical relevance even in a low-risk population. In particular, the progressive reduction of NNP when multiple predictors are simultaneously present underscores the additive value of combining maternal hemodynamic profiling with computerized CTG. Beyond statistical significance, the effect sizes observed showed clear clinical relevance: multiparity reduced the risk of APO by ∼80 %, elevated SVR conferred a two- to three-fold increase, and reduced STV tripled to quadrupled the likelihood of APO. Consistently, SVR >1,135 dyn s cm⁻⁵ demonstrated >90 % sensitivity, while STV ≤7 ms showed >80 % specificity, further reinforcing the potential utility of these markers for clinical risk stratification. This finding supports the potential implementation of a multimodal approach in routine antenatal assessment to better stratify the risk of intrapartum complications.

An important consideration when evaluating the clinical implications of this study relates to the potential costs associated with implementing routine screening based on maternal hemodynamics and computerized cardiotocography in all low-risk pregnancies. Although the acquisition and maintenance of non-invasive hemodynamic devices such as USCOM may initially appear costly, their integration into routine antenatal care at term could prove cost-effective when balanced against the potential reduction in emergency obstetric interventions, neonatal morbidity, and NICU admissions. Early identification of at-risk pregnancies through maternal cardiovascular assessment may improve perinatal outcomes and optimize the use of healthcare resources. Future prospective studies are warranted to formally assess the cost-effectiveness of such an approach in low-risk populations.

Limitations

The main limitation of this study is that one of the primary outcomes – obstetric intervention due to intrapartum fetal distress – was subjectively defined based on cardiotocography (CTG) interpretation, which has a well-documented high interobserver and intraobserver variability [41]. According to Rhöse et al., interobserver agreement for CTG classification and clinical management decisions is poor (κw range 0.31–0.50 and 0.20–0.45, respectively) [41]. However, the highest agreement was found for abnormal CTG patterns (Pa range 0.28–0.36), which are the primary indications for cesarean section or operative delivery [41]. In our study, no formal inter-observer agreement analysis was performed, and this may represent a potential source of bias. However, to reduce misclassification, only cases with clearly documented indications for operative delivery due to pathological CTG were included as adverse perinatal outcomes. Cases with multifactorial or unclear indications were excluded. While this approach reflects real-world clinical decision-making, it underlines the need for further prospective studies with centralized CTG review or predefined adjudication protocols to improve reproducibility.

As with all non-invasive hemodynamic methods, USCOM measurements may be subject to operator dependency. However, in this study, all assessments were performed by a single, trained operator under standardized conditions, minimizing variability and ensuring measurement consistency. Furthermore, USCOM has been validated against both echocardiography – the current gold standard for maternal cardiovascular assessment – and invasive techniques, and several studies have demonstrated its good reproducibility even in obstetric populations [8].

Additionally, this was a single-center study conducted in a tertiary care setting, which may limit the generalizability of the findings to other clinical contexts or patient populations. Although the exclusion of induced labors may reduce external applicability, this was a deliberate methodological choice to eliminate a major source of confounding related to pharmacologic and clinical variability.

Taken together, these limitations underline the need for a prospective, multicenter validation study to assess the reproducibility and predictive performance of the proposed model. This investigation should be considered a pilot study, aimed at exploring associations between maternal hemodynamic parameters, computerized CTG analysis, and adverse perinatal outcomes, providing the rationale and framework for future confirmatory research.

Strengths

A major strength of this study is that the hemodynamic assessment was conducted under standardized conditions, minimizing potential confounders such as labor pain and variations in room conditions that could affect hemodynamic parameters. Unlike our previous study, which assessed maternal hemodynamics during labor, this research examined maternal cardiovascular status before the onset of labor, allowing for a standardized evaluation in a non-stress condition. The 8-day window between assessment and delivery ensured clinical feasibility, while still capturing a physiologically stable and relevant maternal profile close to birth. This approach enabled us to determine whether the resting hemodynamic profile differs in women at risk of APO.

Additionally, we excluded patients undergoing labor induction to minimize confounding factors. Labor induction is known to significantly increase the likelihood of emergency cesarean section due to intrapartum fetal compromise [3].