Comparison of Cerebroplacental Ratio, Intergrowth-21st Standards, Customized Growth, and Local Population References for the Prediction of Fetal Compromise: Which Is the Best Approach?

An intense debate exists between those researchers who consider the existence of a similar fetal growth under similar and optimal environmental conditions(Intergrowth 21st standards; IG21) [1] and those who recognize the existence of inherent constitutional differences, proposing the adjustment to particular maternal characteristics (customized growth; CG) [2]. Despite its scientific nature, this controversy is partially philosophic as it raises the question of whether ethnic differences occur by nature or are influenced by the environment. This debate, far from being concluded, has intensified after the publication of two articles showing that IG21 standards might be less sensitive than customized centiles in identifying adverse perinatal outcome (APO) [3, 4]. In this scenario, finding the most useful estimated fetal weight (EFW) centiles for the detection of intrapartum fetal compromise (IFC) is of notable interest. However, it might be even more rewarding to evaluate the ability of EFW centiles in comparison with the cerebroplacental ratio (CPR), a hemodynamic parameter available to most ultrasonographers, which also predicts adverse outcome and has been proven to be a reliable and promising parameter in the evaluation of fetal wellbeing at the end of pregnancy [5, 6]. The purpose of this work was to compare the abilities of the CPR, IG21, CG, and local population references (LPR) for the detection of IFC.

This was a prospective study of 714 fetuses that underwent routine ultrasound at the public tertiary maternity of La Fe hospital. The ultrasound examination was performed between 34 and 40 weeks and included an EFW and Doppler evaluation of the umbilical (UA) and middle cerebral arteries (MCA) pulsatility indices (PI). The UA and MCA were recorded using color and pulse Doppler according to earlier descriptions [5, 6] and the CPR was calculated as the simple ratio between the MCA PI and the UA PI [5, 7]. All pregnancies were delivered in less than 2 weeks after the scan (14 days or less) and only one (the last) examination per fetus was included in the analysis.

If the examination showed that the EFW was below the 3rd centile, then an induction was planned at >37 weeks (37–39) depending on the severity of the case. In the minority of cases the severity of the FGR advised the performance of a cesarean section. In these cases, the patient was not included as we were only interested in those cases with spontaneous onset or an induction that entered into labor. In order to adjust for the effect of the GA, EFW, and BW values were converted into three different types of centiles (IG21, CG, and LPR centiles). This conversion was done according to references earlier published or using an online calculator or specific Excel calculators provided by the original authors if these were available. LPR were adjusted only for fetal gender [8-10]. Also, CPR values were converted into multiples of the median (MoM), dividing each value by the 50th centile at each gestational age (GA), as previously described [5]. CPR medians (50th centile) were those used in recent studies and were represented by the equation [5]:

CPR 50th centile = –3.814786276 + 0.36363249 × GA – 0.005646672 × GA [2],

where GA was in weeks with decimals.

All Doppler examinations were performed by the first author, a certified teaching expert in obstetric ultrasound by the Spanish Society of Obstetrics and Gynecology, using General Electric Voluson® (E8/E6/730) ultrasound machines with 2–8 MHz convex probes, during fetal quiescence, in the absence of fetal tachycardia, and keeping the insonation angle with the examined vessels as small as possible. GA was determined according to the crown-rump length in the first trimester. Multiple pregnancies and those complicated by congenital fetal abnormalities or aneuploidy were excluded from the study.

Gestational characteristics including parity, number of gestations, and maternal ethnicity, age, weight and height, were collected at examination. In addition, labor outcome data including BW, mode of delivery, Apgar score, cord arterial pH, and admission to the neonatal care unit were collected after birth to calculate a composite adverse outcome. IFC was considered when the composite adverse outcome was positive for any of these components: abnormal intrapartum fetal heart rate (according to the intrapartum fetal monitoring guidelines of the FIGO) [11], intrapartum fetal scalp pH <7.20 requiring cesarean section, neonatal umbilical cord pH < 7.10, 5-min Apgar score <7, and postpartum admission to the neonatal intensive care unit or special care baby unit. As per local protocol, all fetuses were initially treated as low-risk AGA fetuses, and were subsequently managed according to their progression in labor. Doppler examinations did not influence the management of labors, which were performed by different obstetricians. Cases with abnormal intrapartum fetal heart rate ending with instrumental delivery were not included if fetal scalp pH or neonatal pH were within the normal limits. Elective cesarean deliveries were discarded, as we were specifically interested in labor outcome, so only pregnancies initiating spontaneous or induced labor were included.

Descriptive statistics were performed evaluating maternal age, parity, GA at examination in weeks, GA at delivery in weeks, interval between ultrasound and delivery, EFW, EFW centile, Doppler parameters (UA PI, MCA PI, CPR, UA PI MoM, MCA PI MoM, CPR MoM), fetal gender, ethnicity, onset of labor (induction and spontaneous), mode of delivery (cesarean section, forceps, vacuum [ventouse], Thierry’s spatulas spontaneous vaginal delivery), Apgar scores at 5 min, neonatal cord arterial pH, and baby destiny (maternity ward, neonates ward, intensive care unit). Continuous variables are presented as the median and interquartile range, while categorical variables were presented as absolute and relative frequencies.

The accuracy of the CPR and the three different types of centiles (IG21, CG, and LPR) for the detection of IFC was evaluated with univariate analysis, describing the estimate, the exponent of the estimate, the p value, the ROC curve with the area under the curve (AUC) and the detection rate (DR; sensitivity) for a false positive rate (FPR; 1 – specificity) of 5 and 10%. However, in order to improve the predictive accuracy, we also applied multivariate logistic regression analysis combining CPR with centiles and clinical parameters to create models in which the same statistical parameters were calculated.

Comparisons were made with Mann-Whitney and χ² tests. The Akaike information criterion (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). 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. This is not usually a good approach; adding more parameters usually results in a good model fit for the data at hand, but that same model will likely be useless for predicting other data sets. The AIC allows a good balance between parsimony and goodness of fit. Statistical analysis and graphs were performed using R-software® (version 3.3.2). Significance was considered with a p value <0.05.

Table 1 shows the characteristics of the study population. In summary, the study included 714 fetuses, of which 53.1% were male and 46.9% were female, with most being (87.8%) of Spanish Caucasian origin. The mean maternal age was 32.5 years, and the mean GA at examination and delivery was 38.9 and 40.1 weeks. Most of the pregnancies had a spontaneous onset of labor (52.4%) and a spontaneous vaginal delivery (57.1%), with neonates born uneventfully and sent with the mother to the maternity ward (96.4%). The main indications for labor induction were GA >41 weeks and premature rupture of membranes. Most fetuses with induction of labor presented a favorable outcome (84.7%), which was slightly worse than those with spontaneous onset of labor (95.4%). Finally, most inductions (82%) were performed at >39 weeks. Concerning delivery, only 0.4% had an Apgar score <7 at 5 min, while 1.4% had a neonatal cord pH <7.10 and 3.6% needed admission to the neonatal ward. The proportion of fetuses below the 10th centile was 19.6% according to the tertiary referral nature of the center. However, only 9.7% presented IFC.

Table 2 compares the characteristics of the pregnancies according to the study outcome. In summary, mothers in the adverse outcome group were older (p = 0.01). In addition, when compared to the fetuses with a normal outcome, the UA PI MoM was significantly higher (p < 0.0001), while the MCA PI MoM (p = 0.014) and the CPR MoM (p < 0.0001) were significantly lower. There were also significant differences in the EFW (p = 0.001), BW (p < 0.0001), and in the frequency of cesarean sections, labor inductions, SGA, Apgar score, and neonatal cord pH (p < 0.0001).

Tables 3-5 and Figures 1-3 show several comparisons of prediction models. As indicated earlier, in order to evaluate the predominance of the models we took into account the AIC. A lower AIC meant a better prediction performance. Differences between models were significant when the difference in the AIC was 2 units. If the AIC was 2–4 units lower, the difference was highly significant.

Table 3 and Figure 1 show the univariate analysis for the four studied parameters (CPR, IG21, CG, LPR). The best prediction (lowest AIC) was achieved using the CPR MoM with an AIC of 431.99 (AUC 0.661, DR of 20.28/34.78% for an FPR of 5/10%). This difference was highly significant. The three EFW centile models obtained worse prediction abilities than the CPR. Among them, only LPR were significantly better. No significant differences were seen between IG21 and CG: LPR AIC = 445.09 (AUC 0.635, DR 14.49/31.88% for an FPR of 5/10%), CG AIC = 447.25 (AUC 0.625, DR 17.39/24.63% for an FPR of 5/10%), and IG21 AIC = 448.05 (AUC 0.615, DR 15.94/24.63% for an FPR of 5/10%).

Table 4 and Figure 2 show the multivariate analysis of the studied parameters (CPR, G21, CG, LPR) in combination with several gestational parameters that were known at the moment of examination (maternal age, parity, height, weight, GA at examination, and fetal gender). The best prediction was again achieved using the CPR MoM with an AIC of 414.287 (AUC 0.738, DR of 24.64/34.78% for an FPR of 5/10%). The difference was highly significant. The three EFW models had worse prediction abilities (higher AIC) than the CPR, but the differences in between were not significant (AIC differences <2 units): LPR AIC 421.919 (AUC 0.716, DR 26.09/31.88% for an FPR of 5/10%), CG AIC = 421.641 (AUC 0.717, DR 27.54/34.78% for an FPR of 5/10%), and IG21 AIC = 423.256 (AUC 0.713, DR 27.54/34.78% for an FPR of 5/10%).

Finally, Table 5 and Figure 3 show the same multivariate analysis of Table 4 and Figure 2, but adding the information of the CPR to the three EFW centile methods (IG21, CG, LPR). The AUC showed no improvement in comparison with the CPR alone. However, according to the AIC, the prediction ability improved for the CG and LPR and no significant differences existed between the three EFW models: LPR AIC = 411.787 (AUC 0.738 DR 33.33/40.58% for an FPR of 5/10), CG AIC 411.777 (AUC 0.737, DR 33.33/42.03% for an FPR of 5/10%), and IG21 AIC = 412.687 (AUC 0.738, DR 28.99/39.13% for an FPR of 5/10%).

The main goal of fetal surveillance is the prediction and avoidance of APO. Initial strategies designed to achieve this objective at the end of pregnancy are usually based on ultrasound biometry [12-15], probably because appropriate fetal weight has been commonly identified with adequate fetal health. Against this assumption, CPR has been recently proposed as a predictor of APO regardless of fetal weight [16, 17]. However, despite the growing evidence [18-24], clinical guidelines are reluctant to consider CPR evaluation as routine practice, and assessment of MCA Doppler is only proposed in cases of low birth weight, therefore making fetal biometry the first line of evaluation to the detriment of fetal hemodynamics [25]. In contrast with this approach, our results show that CPR yields more diagnostic accuracy than fetal biometry and that screening of APO might be approached with CPR instead of EFW.

Some earlier works have compared the performance of CPR and fetal biometry in the prediction of late APO and neonatal acid-base status. The majority concluded the existence of a CPR predominance [26-28], although one work suggested a similar prediction ability [29]. Interestingly, a common conclusion was that the best performance was achieved combining CPR and fetal biometry, therefore summing hemodynamic and ponderal information. Our results are in line with these studies, but highlight the importance of CPR over EFW. According to our data, the contribution of fetal biometry once CPR has been taken into account is small, making CPR by far the principal tool in the prediction of APO regardless of fetal weight.

Another conclusion of this work was that IG21 and CG yielded similar prediction abilities, a relevant conclusion considering the ongoing debate between these antagonistic approaches. To date, only two studies have compared IG21 with CG. In the first, CG identified more SGA infants at risk of APO than IG21 [3]. In the second, CG proved to be more accurate for the detection of SGA and stillbirth, identifying 60% more SGA cases at increased risk [4]. Our results are not completely in line with these works and suggest the existence of a similar performance using any of both fetal biometry approaches. Interestingly, to our surprise, the best EFW method according to the AIC was the LPR; however, considering LPR as a simplification of the CG model, adjusting only for fetal gender, this might be the result of a more parsimonious design, which included fewer explanatory variables.

The predominance of cerebroplacental hemodynamics over fetal biometry for the prediction of APO reflects the utility of the brain sparing mechanism for the interrogation of fetal wellbeing. Clinical studies have proven that fetuses tend to dilate the cerebral circulation reducing the MCA impedance when term or preterm labor approaches [30, 31] and that this mechanism is even more intense when the fetuses present complications at labor such as abnormal fetal monitoring, cesarean section for fetal compromise, and pathological acid base status [23, 32]. Our data agree with this evidence showing that regardless of fetal weight, CPR measurement excels fetal biometry in the prediction of APO, an advantage that could be applied to select those fetuses at the end of pregnancy that could be managed in a more conservative way [33]. This goal could be achieved by means of prediction models which include clinical parameters as recently proposed [34, 35].

The strengths of this study include the novelty of the hemodynamic versus ponderal comparison, the inclusion of the three weight prediction models currently described, and the high number of cases compared. Finally, shortcomings include the low prediction ability obtained with either of the methods and the absence of a comparison based on a postnatal neurocognitive follow-up.

In conclusion, we have shown that, although its accuracy is rather moderate, CPR provides the best diagnostic yield for the prediction of IFC alone or combined with clinical parameters, and that once CPR has been obtained, the information provided by EFW (IG21 standards, CG, or LPR) is small. In addition, when only fetal biometry is considered, IG21 and CG yielded similar prediction abilities while the best performance is obtained with the LPR centiles, a more parsimonious model with less explanatory variables. Although more studies will be needed to calibrate the real importance of CPR in the prediction of late APO, these results suggest CPR should be promoted as an initial tool for the prediction of APO.

IRB permission was obtained for this study (reference 2014/0063).

The authors report no conflicts of interest.

The authors received no funds to perform this research.