Electrical Cardiometry during transition and short-term outcome in very preterm infants: a prospective observational study

This prospective observational cohort study was performed in a tertiary level neonatal center (Cork University Maternity Hospital, Ireland). It was approved by the local Research Ethics Committee and registered at clinicaltrials.gov (NCT04538079). Infants were enrolled between November 2019 and May 2021.

The inclusion criteria were birth between 23 + 0 weeks and 31 + 6 weeks’ GA with the possibility to commence EC monitoring before 12 h of age. Infants were enrolled following informed parental consent. In cases of twins, only one infant was eligible due to the availability of a single device. Exclusion criteria were declined parental consent, major congenital cardiac defects (except intra-atrial shunts or patent ductus arteriosus (PDA)), severe congenital anomalies, and hydrops fetalis.

All participants underwent EC monitoring (ICON, Osypka Medical, Berlin, Germany), which was commenced as early as possible but within 12 h of delivery and continued up to the age of 48 h. The detailed technical background was published previously [5]. In short, four neonatal electrodes (iSense, Osypka Medical), placed on the forehead, right side of the neck, left hemithorax at xyphoid level, and left inner thigh, measure changes in electrical conductivity within the thorax which relate to the circulatory system. From this, an estimate of stroke volume and CO is calculated, and these are indexed for birthweight for analysis. The term cardiac output in neonates typically refers to cardiac index, as this is the term used for CO indexed to bodyweight. Hereafter, CO refers to cardiac index. Data processing for EC outputs is detailed in the supplementary material.

All participants underwent routine crUS performed by the attending radiologists as part of standard care (unit standard first crUS within 3 days of life, second between 7 and 10th day of life). The Philips HD-11XE with a C8-5 transducer (convex curved array 8–5 MHz probe, Philips, Netherlands) or GE Logiq P9 with an 8C transducer (micro convex curved array 4–8 MHz probe, GE Healthcare, USA) was used. Radiologists were unaware of the EC values. Papile et al. classification of brain injury was used [12]. The highest grade of ICH diagnosed and the diagnosis of PVL within the first 2 weeks was documented.

Clinical and neonatal data (including GA, birthweight, Apgar scores, CRIB-II score), data on respiratory support (including surfactant use and pneumothorax), and mortality were extracted from electronic charts. Functional echocardiography was performed at least once during the monitoring period (first 48 h) to ensure normal cardiac anatomy, and for PDA evaluation, PDA was classified according to its flow pattern: pulsatile non-restrictive flow considered a hemodynamically significant PDA (Vivid-I, 10S-RS sector array transducer, frequency range 4.5–11.5 MHz or Vivid-E9, 12S sector phased array transducer, 4.0–12.0 MHz) (both GE Healthcare, USA). Echocardiography was performed by neonatologists trained in neonatal echocardiography (EMD, CES, Neidin Bussmann). This institution’s general approach to PDA management is conservative, and none of the infants received treatment for PDA during the initial 48-h monitoring period.

Infants were divided into normal and adverse outcome groups. The adverse outcome was defined as a combined outcome including mortality and/or abnormal crUS (any grade ICH or periventricular leukomalacia) within the first 2 weeks of life. For a post hoc sensitivity analysis, infants with low-grade ICH (grade 1) were considered to have a normal outcome.

The planned target sample size was 100 participants to allow for multiple comparisons.

Categorical variables were described using frequencies and percentages and continuous variables using medians and interquartile ranges (IQRs). For comparisons between outcome groups (adverse, normal), the Mann–Whitney U test was used for continuous variables and the χ² test or Fisher’s exact test (in the case of small expected counts) was used for categorical variables. Logistic regression models were used to investigate the association between median CO across the first 48 h of life and outcome, both unadjusted and adjusted for GA group. GA was dichotomized into either extremely preterm (< 28 weeks) or very preterm (≥ 28 weeks and < 32 weeks) for the analysis. The Hosmer–Lemeshow goodness of fit test was used to determine the fit of the logistic regression models with p < 0.05 indicating lack of fit. Prior to performing the multivariable logistic regression analysis, multicollinearity among the independent variables was tested using the variance inflation factor (VIF) and VIF < 10 was considered acceptable. Linear mixed-effects models were fitted to the data to investigate if the CO trajectory over time differed by outcome group. The initial model included intercept, time (postnatal age), group, the interaction of time by group and GA group as fixed effects and intercept and time as random effects. The dependent variable was CO. Statistical significance for the fixed effects was determined using Satterthwaite’s method and 95% CI for fixed effects were computed using a bootstrap method with 1000 iterations. Model selection was based on a step-down model building approach, starting with a complex model (initial model), followed by backward elimination of random effects and then backward elimination of fixed effects (final model).

The mixed model was implemented in R (version 4.0, R Foundation for Statistical Computing, Vienna, Austria) using lme4 (version 1.1) and lmerTest (version 3.1) libraries. All other analyses were performed using IBM SPSS Statistics (version 28, IBM Corp, Armonk, NY, USA). All tests were two-sided and a p value < 0.05 was considered to be statistically significant.

Fifty-six infants were enrolled, of whom 53 infants were included in the analysis. Excluded infants had severe pulmonary stenosis (n = 1), a genetic syndrome (n = 1), and insufficient EC recording time < 12 h (n = 1) (Fig. 1S). Neonatal and demographic variables are summarized in Table 1 (see Table 1S for additional information). EC monitoring was commenced at a median (IQR) age of 3.2 (1.9 to 5.7) h for a duration of 44.3 (41.3 to 46.0) h.

The adverse outcome occurred in 21 (40%) of the infants (death (n = 4, of whom 3 had abnormal crUS) and/or abnormal crUS (any grade ICH (n = 19) or PVL (n = 1))). This group had lower GA and birthweight, a higher rate of hsPDA, and received more respiratory and circulatory support within the first 48 h.

Median (IQR) CO within the first 48 h of age was 241 (197 to 275) mL/kg/min for infants in the adverse outcome group and 198 (175 to 227) mL/kg/min for those in the normal group (odds ratio (OR) (95% CI), 1.01 (1.00 to 1.03); p = 0.028). After adjustment for GA, this difference was no longer significant (adjusted OR (95% CI), 1.01 (0.99 to 1.02; p = 0.373)) (Fig. 2S). The Hosmer–Lemeshow goodness of fit test indicated best fit with GA groups (compared to GA as continuous variable) and a VIF < 10 indicated no relevant multicollinearity.

Figure 1 illustrates the course of CO over the monitoring period. Linear mixed modeling was used to analyze differences between groups. Backwards and forwards selection resulted in the same model. The final model (Table 2) includes intercept and time (postnatal age) as random effects and intercept, time, and GA group as fixed effects. This model indicates that coming from a baseline (intercept) CO of 185 (95% CI 168 to 201) mL/kg/min, CO increased over time (+ 17 mL/kg/min per day; 95% CI 8 to 26). Furthermore, CO was higher in extremely compared with very preterm infants (+ 39 mL/kg/min; 95% CI 17 to 62) (Fig. 3S).

A post hoc analysis with re-classification of ICH grade 1 (n = 9) as a normal outcome was subsequently performed. This revealed a significantly greater increase in CI over time in infants with ICH ≥ grade 2 (+ 35 mL/kg/min/day compared to + 12 mL/kg/min per day, see Table 1S and Fig. 4S).

This is a single-center study and recruitment was stopped before reaching the original target sample size. Recruitment was below our estimation for several reasons, including inability to enroll multiple simultaneous participants and interruptions due to local COVID-19 restrictions (see Fig. 1S). Therefore, multifactorial and subgroup analysis is limited [13, 28].

Whereas normal crUS is predictive for a normal neuromotor and cognitive outcome in preterm infants [29], the ability of abnormal crUS to predict impaired outcome is limited. ICH severity is associated with mortality, disability, or cerebral palsy and potentially results in hydrocephalus [30, 31]. The predefined primary outcome included all grades of ICH. For ICH grade 1–2, the probability of abnormal neuromotor and cognitive outcome is only 9%, while for grade 3 this increases to 26% [29]. Hemorrhagic parenchymal infarction (grade 4) predicts abnormal neurodevelopmental outcome with a positive predictive value of 47% [32]. However, even low-grade ICH was found to affect long-term outcome [10, 33‐35]. Our primary endpoint was at 2 weeks, which includes the vast majority of ICH [8], but may not capture all PVL, some of which may only become apparent on ultrasound after this time [32, 36].