Comparison of ductus venosus Doppler and cerebroplacental ratio for the prediction of adverse perinatal outcome in high‐risk pregnancies before and after 34 weeks

Abstract Introduction The objective of the study was to compare the accuracy of the ductus venosus pulsatility index (DV PI) with that of the cerebroplacental ratio (CPR) for the prediction of adverse perinatal outcome at two gestational ages: <34 and ≥34 weeks' gestation. Material and methods This was a retrospective study of 169 high‐risk pregnancies (72 < 34 and 97 ≥ 34 weeks) that underwent an ultrasound examination of CPR, DV Doppler and estimated fetal weight at 22–40 weeks. The CPR and DV PI were converted into multiples of the median, and the estimated fetal weight into centiles according to local references. Adverse perinatal outcome was defined as a composite of abnormal cardiotocogram, intrapartum pH requiring cesarean delivery, 5′ Apgar score <7, neonatal pH <7.10 and admission to neonatal intensive care unit. Values were plotted according to the interval to labor to evaluate progression of abnormal Doppler values, and their accuracy was evaluated at both gestational periods, alone and combined with clinical data, by means of univariable and multivariable models, using the Akaike information criteria (AIC) and the area under the curve (AUC). Results Prior to 34 weeks' gestation, DV PI was the latest parameter to become abnormal. However, it was a poor predictor of adverse perinatal outcome (AUC 0.56, 95% CI: 0.40–0.71, AIC 76.2, p > 0.05), and did not improve the predictive accuracy of CPR for adverse perinatal outcome (AUC 0.88, 95% CI: 0.79–0.97, AIC 52.9, p < 0.0001). After 34 weeks' gestation, the chronology of the DV PI and CPR anomalies overlapped, but again DV PI was a poor predictor for adverse perinatal outcome (AUC 0.62, 95% CI: 0.49–0.74, AIC 120.6, p > 0.05), that did not improve the CPR ability to predict adverse perinatal outcome (AUC 0.80, 95% CI: 0.67–0.92, AIC 106.8, p < 0.0001). The predictive accuracy of CPR prior to 34 weeks persisted when the gestational age at delivery was included in the model (AUC 0.91, 95% CI: 0.81–1.00, AIC 46.3, p < 0.0001, vs AUC 0.86, 95% CI: 0.72–1, AIC 56.1, p < 0.0001), and therefore was not determined by prematurity. Conclusions CPR predicts adverse perinatal outcome better than DV PI, regardless of gestational age. Larger prospective studies are needed to delineate the role of ultrasound tools of fetal wellbeing assessment in predicting and preventing adverse perinatal outcome.


| INTRODUC TI ON
Current consensus indicates that late-onset fetal growth restriction (LO-FGR) is caused by an imbalance between fetal demands and placental supply, [1][2][3] in which hemodynamic anomalies occur regardless of fetal weight. [4][5][6][7][8][9][10][11] In pregnancies complicated by LO-FGR, adverse perinatal outcome (APO) is optimally predicted with the cerebroplacental ratio (CPR), that reflects the degree of cerebral vasodilation in response to placental relative insufficiency. Conversely, early-onset fetal growth restriction (EO-FGR) is caused by a progressive placental failure where prediction of APO is achieved using the estimated fetal weight (EFW) 12 and umbilical artery (UA) Doppler, 13,14 although with progressive hemodynamic deterioration, cerebral vasodilation also occurs, making CPR a potentially useful tool for the prediction of APO, along with EFW. 1 In pregnancies complicated by EO-FGR, the ductus venosus (DV) Doppler has been considered as a marker of progression, but not in fetuses with LO-FGR, in which hemodynamic progression does not reach such degrees of severity. A drawback of this approach has been the paucity of information concerning the role of the DV Doppler at the end of pregnancy. Moreover, no study has clearly evaluated and compared the ability of the DV Doppler along with CPR in the prediction of APO prior to and beyond 34 weeks' gestation. Finally, even though the DV Doppler might be the last hemodynamic parameter to be altered, this does not necessarily make it the most accurate parameter for the prediction of APO, such that the veracity of such assertion remains to be proved.
The aim of this study was to compare the predictive accuracy of ductus venosus pulsatility index (DV PI) and CPR in the prediction of APO alone and combined with other sonographic or clinical parameters.

| MATERIALANDME THODS
This was a retrospective study of 169 high risk pregnancies who attended La Fe public tertiary maternity hospital, between 2012 and 2022. These included cases with chronic hypertension, preeclampsia, previous stillbirth and reduced fetal growth. Other pregnancies were considered at risk for different reasons, such as reduced amniotic fluid, advanced maternal age, assisted reproduction, and reduced fetal movements.
The ultrasound examinations were performed between 22 and 41 weeks' gestation (72 prior to 34 and 97 at or beyond 34 weeks) and included an EFW, a Doppler evaluation of the UA and middle cerebral artery (MCA) pulsatility indices (PI), and an evaluation of the DV PI. The UA, MCA and DV were recorded for fetal surveillance using color and pulse Doppler according to earlier descriptions [15][16][17] and the cerebroplacental ratio (CPR) was calculated as the ratio between the MCA PI and the UA PI. 18 Only one (the last) examination per fetus was included. Pregnancies were followed until the onset of spontaneous labor, induction of labor or elective cesarean section for obstetric indications. Management of fetuses, hospital protocols, and rate of inductions and cesarean sections did not change during the study period.
EFW and birthweight values were converted into centiles, 19 while CPR and DV PI values were converted into multiples of median (MoM) dividing each value by the 50th centile at each gestational age as earlier described. 15

TA B L E 2 (Continued)
studied parameters for the prediction of CS-IFC, in appropriate and small for gestational age (SGA) fetuses.
Comparisons were made using Mann-Whitney and Fisher's exact tests. It is important to underline that AIC was used to select the best prediction model by means of a lower AIC, which indicated the presence of a higher accuracy (a difference in the AIC of 2 units indicated significant differences and a difference of 2-4 units indicated highly significant differences). 21 There is generally a tradeoff between goodness of fit and parsimony: low parsimony models (i.e., models with many parameters) tend to have a better fit than high parsimony models.
However, adding more parameters usually results in a good model fit for the data at hand, but that same model will probably be less useful in  In addition, they were more frequently delivered via elective cesarean section, and admitted to neonates ward (p < 0.0001) and

| RE SULTS
Neonatal Intensive Care Unit (p < 0.01). Only 10 (13.8%) were delivered beyond 34 weeks. Figure 1 shows the three sonographic parameters examined at or beyond 34 weeks' gestation plotted against the interval to delivery. Interpolation curves and 95% confidence intervals (CI) showed a poor R 2 and did not suggest a clear chronology of progression to abnormality. Table 2 and Figure 3 show a comparison of different models for the prediction of APO at or beyond 34 weeks' gestation. Prediction of APO   addition of the DV PI MoM information (models 1-5) did not improve prediction of the above-mentioned models that included CPR MoM.
To address the potential confounding effect of prematurity on APO prior to 34 weeks' gestation, we performed a secondary analysis in which we added the information of GA at delivery to the multivariable analysis (

| DISCUSS ION
According to our results, the DV PI was a poor predictor of APO before and after 34 weeks' gestation, while the best individual prediction was achieved measuring the CPR MoM. These results persisted even when considering the potential confounding effects of prematurity.
Our results were consistent with previous findings showing that early in pregnancy the DV PI tended to be the last hemodynamic parameter to become abnormal. [23][24][25] Conversely, in the last weeks of pregnancy, no clear chronology of progression to abnormality could be detected. Unfortunately, the progression of DV doppler anomalies according to the interval to labor has not been evaluated in the third trimester, so we did not find references to support these data.
Our results were also consistent with previous studies demonstrating the poor predictive accuracy of DV PI (AUC 0.66-0.67). 26,27 Moreover, another study 28  is tempting to think that the last anomaly to appear should be the one with the highest correlation with adverse outcome. However, that is TA B L E 4 Logistic regression models for the prediction of adverse perinatal outcome prior to 34 weeks, adjusting for gestational age (GA) at delivery to address the potential confounding effect of prematurity (N = 72). were not blinded to the examination, we tried to diminish this influence with additional analysis that discarded elective cesarean deliveries. Second, not only the DV Doppler, but also the CPR performs better with short intervals to labor. 43 Third, to adjust for the effect of prematurity, additional multivariable analyses were performed, including GA to rule out the effect of prematurity, or evaluating only CS-IFC to discard unspecific neonatal admissions. In all cases the results persisted, and CPR remained as the only significant parameter for the prediction of APO or CS-IFC.

| CON CLUS ION
The best prediction of APO both prior to and beyond 34 weeks' gestation was achieved by means of CPR MoM evaluation. Addition of DV Doppler did not improve the predictive accuracy. However, prior to 34 weeks, it might provide useful information about fetal hemodynamic progression. Future studies are needed to determine if CPR evaluation might be useful in determining the ideal timing of delivery in pregnancies complicated by FGR.

AUTH O RCO NTR I B UTI O N S
José Morales-Roselló designed the study, performed the ultrasound examinations, did the statistical analysis, and wrote the manuscript.
Asma Khalil supervised the final manuscript and made notable contributions to the final text. Rohan Bhate and Nashwa Eltaweel suggested valuable inputs to the text.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors report no conflict of interests.
TA B L E 5 Logistic regression models for the prediction of CS-IFC, adjusting for gestational age (GA) at delivery to address the potential confounding effect of prematurity. To increase the consistency of the results, DV PI MoM and CPR MoM were compared at all gestational ages discarding elective cesarean sections. Model 16 (N = 103) included AGA and SGA, while model 17(N = 74) included SGA. Multivariable model for the prediction of cesarean section for intrapartum fetal compromise in SGA fetuses at 22-41 weeks. Gestational age at delivery was included to discard the effect of prematurity. Elective cesarean sections were discarded.
Abbreviations: AGA, appropriate for gestational age; AIC, Akaike information criteria; AUC, area under the curve; CPR MoM, cerebroplacental ratio multiples of the median; DV MoM, Ductus venosus pulsatility index multiples of the median; EFW, estimated fetal weight; GA at delivery, gestational age at delivery; NPP, negative predictive power; OR, odds ratio; PPP, positive predictive power; SGA, small for gestational age.