End-organ saturations correlate with aortic blood flow estimates by echocardiography in the extremely premature newborn – an observational cohort study

Extremely premature newborns undergo significant physiologic changes in their immediate post-natal life, with potential hemodynamic instability in the context of persistent fetal shunts. Neonatal care in this vulnerable population aims at maintaining adequate organ perfusion in the setting of changing physiology. Near infrared spectroscopy (NIRS) is a promising technology allowing for real-time monitoring of tissue oxygenation as a surrogate for end-organ perfusion. However, validation of NIRS measures in the premature population as a correlate of blood flow is needed, especially in the context of postnatal delays in transition, such as presence of a patent ductus arteriosus (PDA). Specifically, premature newborns are at risk of diastolic steal effect, which may adversely alter the post-ductal aortic blood flow [1].

NIRS monitoring estimates venous-weighted percentage of saturated hemoglobin in the monitored tissue [2‐4]. Regional saturations are affected by organ perfusion, hemoglobin concentration, arterial oxygen content, oxygen extraction and oxygen consumption. Perfusion is a reflection of local blood flow, which is dependent on cardiac output, as well as vascular distribution, filling and tone. As such, cerebral oxygen saturation (Csat) and renal oxygen saturation (Rsat) monitored by NIRS may indicate underlying perfusion when all other factors remain stable. Csat is primarily affected by pre-ductal systemic perfusion and cerebral oxygen extraction, while the Rsat is impacted by post-ductal diastolic steal, underlying renal oxygen extraction and altered systemic output. Csat correlates with cerebral perfusion [2, 5] and decreased values have been associated with neurological injury in premature newborns [6‐8]. We recently published that the velocity time integral (VTI) of the descending aortic flow by echocardiography (ECHO), a marker of stroke distance, correlated with Rsat at the time of the measurement in a population of single ventricle infants [9]. However, limited data exist on the correlation between ultrasound markers of perfusion and cerebral/renal NIRS values in the premature neonate. We sought to compare ECHO derived markers of aortic blood flow with end-organ saturation measurements obtained at the time of ECHO acquisition in extremely premature neonates. We hypothesized that in preterm infants, especially in first week of life when the duct is usually patent, resulting in diastolic runoff, Csat would reflect ECHO-derived markers of aortic pre-ductal blood flow and Rsat would reflect those of post-ductal flow.

From May 2013 to August 2017, 97 premature patients < 29 weeks gestation at birth were identified in the NIRS database with either Csat or Rsat monitoring results available. Of these, 50 infants had an available ECHO at the time of NIRS monitoring, of which 1 infant was excluded due to concomitant diagnosis of Tetralogy of Fallot. The final cohort included 49 neonates with 75 Csat-ECHO and 62 Rsat-ECHO observations available for analysis. During NIRS monitoring, a single ECHO was available in 30 patients, two ECHOs in 11 patients, three ECHOs in 7 patients and four ECHOs in 1 patient. Mean gestational age at birth was 25.8 ± 1.3 weeks, with a mean post-menstrual age of 28.3 ± 3.8 weeks at time of ECHO.

Demographic variables are shown in Table 1 and clinical variables pertaining to the ECHO-NIRS dyads in Table 2. There were no significant differences in clinical variables between those with Csat-ECHO measures and those with Rsat-ECHO measures. For those with available Rsat measurements, the average urine output was of 3.9. ± 1.7 mL/kg/min prior to ECHO. The average Rsat was of 58.6 ± 12.6% and the average rFTOE was 37.7 ± 13.3%. There was no difference between the Rsat in the hour following the onset of the ECHO compared to the Rsat in the hour preceding the ECHO (58.6 ± 12.6% vs 58.5 ± 12.3% vs, p = 0.90). For those with available Csat measurements, the average Csat during ECHO was 65.8 ± 8.6% and the average cFTOE was 30.1 ± 8.7%. Similarly, there was no difference between the Csat in the hour following the onset of the ECHO with the Csat in the hour preceding the ECHO (65.8 ± 8.6% vs 66.1 ± 9.1%, p = 0.43). The presence of an umbilical arterial line (UAL), all placed in a high-position (T6-T10), was not associated with a difference in Rsat at the start of echocardiography (54.1 ± 12.3% vs 58.6 ± 19.5%, p = 0.31).

Table 3 illustrates the results of the correlation between NIRS-derived measurements and ECHO-derived doppler flow measurements of the aorta. Although Rsat, Csat, rFTOE, and cFTOE did not correlate with markers of blood flow (VTI, cardiac output, and peak systolic velocity) obtained at the descending aorta and sampled with the subcostal view, the majority of these NIRS measurements did correlate with blood flow markers sampled with the suprasternal view. The unadjusted linear regressions correlating Rsat, Csat, rFTOE and cFTOE with the descending aortic VTI obtained in the suprasternal view are represented in Fig. 2. Rsat and rFTOE also were associated with the peak systolic velocity in that view. Furthermore, rFTOE, Csat and cFTOE correlated with the estimated descending aortic output measured with the suprasternal view. Csat was associated with the stroke distance estimated at the ascending aorta (suprasternal view) and estimated ascending aortic output correlated with cFTOE. Although the vast majority of the cohort had a patent ductus at evaluation, there was an association between PDA size at ECHO and Rsat/Csat values; and when stratifying for PDA presence at ECHO, we found similar results despite the decreasing number of observations (Supplemental Table A). Finally, none of the NIRS measurements were associated with the closest measurements of systolic/diastolic blood pressure, pH, hemoglobin, hematocrit, creatinine and urine output by mixed effect model.

In this retrospective study of NIRS measures in extremely premature neonates, we found an association between end organ saturations measured during ECHO with estimates of aortic flow by Doppler. As physiologically expected, markers of pre-ductal aortic flow correlated with Csat and those of post-ductal aortic flow with Rsat.

We previously described that descending abdominal aortic flow by subcostal view correlated with Rsat measures in a congenital heart disease term population [9]. In this current extreme premature population, markers of aortic blood flow in the suprasternal view were correlated with NIRS measures of end-organ saturations. Csat correlated with both ascending aortic measurements (pre-ductal) and descending aortic measurements while Rsat correlated with descending aortic measurements only. Other authors have described a correlation between cerebral saturation and estimated cardiac output in the extreme premature neonatal population [17]. In our study, measurements in the subcostal view were not associated with NIRS measurements. This may be explained by the increased angle of insonation in premature newborns, when evaluating the descending aorta in the subcostal window, compared to a term newborn where one may have more amplitude of manipulation of the probe. During Doppler acquisition, angle of insonation between the sampled vessel and the ultrasound beam must be minimized to avoid imprecision in estimating the Doppler shift [18, 19]. As such, the beam should interrogate the corresponding flow in a parallel fashion, as much as feasible [20]. Suprasternal view minimizes the angle of insonation for Doppler acquisition in the descending aorta, since it is aligned with the flow. Indeed, a previous report indicated that flow velocities for the descending aorta are best sampled from the suprasternal window [21]. Finally, estimation of stroke distance by VTI has been advocated as a simple and reproducible measurement of blood flow [22], which can be used to monitor the hemodynamic changes in critically ill patients [23] and, in select populations, has been associated with patient outcomes [24, 25]. It has previously been described as a marker of lower body perfusion in the pediatric and neonatal population [26]. As such, our data suggest that a suprasternal approach be considered when interrogating the post-ductal aorta to assess VTI in the premature neonate.

NIRS is a promising tool to monitor changes in aortic blood flow, as may be seen with a significant PDA in the extremely premature neonate. Other clinical markers of perfusion commonly used in the clinical setting, including blood pressure, may not detect subtle changes [27, 28] and have not been correlated with echocardiographic markers of blood flow in preterm infants [28]. While indicators of aortic blood flow were associated with Rsat and Csat measurements in our study, they were not associated with other routinely used markers of end-organ hypoperfusion in clinical practice: blood pressure, urine output, pH and hemoglobin. While these traditional markers may be disturbed when tissue ischemia is ongoing [29, 30], they may be imperfect as early detectors of hypoperfusion. Low Csat values have been associated with increased likelihood for severe intraventricular hemorrhage or periventricular leukomalacia in premature newborns [6‐8]. Altered blood flow patterns may be associated with abnormal Csat values and contribute to the hemodynamic changes involved in the pathophysiology of neurological injury. Rsat values are usually higher than Csat values in term infants. However, in our population, Rsat values were lower, possibly reflecting the high incidence of a left-to-right PDA on the ECHO scans that were included. In the Rsat-ECHO dyads, 92% had a PDA with a significant proportion showing a diastolic steal (56% retrograde flow in the descending aorta), possibly compromising the renal perfusion and explaining the lower Rsat. Our group has previously described an association between a hemodynamically significant PDA and a tendency for a low Rsat [11]. In our cohort, we also found an association between PDA size and Rsat values (Supplemental Table A). Our data suggest that Csat and Rsat, as measured with NIRS, can be utilized as a correlate of pre- and post-ductal systemic blood flow. This adds to the small body of literature associating NIRS measures to markers of perfusion in the pediatric setting [9, 31‐33].

Several methodologic issues regarding the stability of NIRS measures during concurrent ECHO image acquisition and infusion through an indwelling UAL were investigated. The regional oxygen saturation/extraction values were similar in the hour prior and after the ECHO, suggesting that the image acquisition did not affect oxygen utilization. Furthermore, in our cohort, we did not detect any association between NIRS values and the presence of an UAL. However, the granularity of the NIRS measurements may not have detected subtle blood flow differences secondary to the presence of the line, and time periods of blood sampling from the UAL were not specifically captured. The stability of NIRS measures as described in this study may help to inform future design of simultaneous NIRS and ECHO studies in the preterm infant.

Our study was retrospective, single center and descriptive with a significant risk of being underpowered to detect meaningful associations. Additional limitations include lack of NIRS data for some infants with prematurity and the fact that clinicians were not blinded to the NIRS values. Also, not all infants had an ECHO during NIRS monitoring, and as such, were excluded from the analysis. Hence, the current cohort most likely represents the sickest premature newborns from a clinical perspective, potentially introducing selection bias that would limit generalizability. Arterial oxygen tension (PaO2) at the time of NIRS monitoring could have provided additional information and added accuracy compared to the use of the systemic oxygen saturation but were not consistently available in all patients. Creatinine measurements were only available for some infants and timing of collection was variable since retrospectively extracted from the chart. Precise position of renal parenchyma by ultrasound was not confirmed prior to sensor placement. However, previous studies have evaluated by ultrasound the position of the kidney in newborns and described its location as between the costal margin and the iliac crest [34‐36]. ECHO was limited by the use of multiple scanning platforms, multiple professionals acquiring the images, absence of some ECHO markers, or limited views in several patients. Indeed, as outlined in Table 3, some ECHO markers were missing in the included ECHOs (potentially introducing a measurement bias) and many newborns had only some of the ECHO markers available for analysis (number of observations available were presented). ECHO measures were not analyzed for intra-reader or inter-reader variability, although our methodology for VTI reproducibility has been validated in a previous publication by our group [24], and by other authors [22, 37, 38]. Furthermore, output estimation by ECHO has limitations regarding the assumption that the cross-sectional area has a circular shape and is calculated solely from the linear diameter measurement of the corresponding structure. However, a MRI study has indicated a strong correlation with ECHO-derived measurements of output, based on this methodology [39]. Finally, the ECHOs were not performed at standardized time-points.

Regional NIRS values are associated with aortic blood flow measurements by ECHO in the extremely premature population. Future studies should investigate if therapeutic approaches to maintain adequate end-organ saturations using NIRS monitoring may also favorably impact blood flow distribution and improve outcomes in the extremely premature population.