2.1.

Study Protocol

The protocol was approved by the Comitato Etico Milano Area 2 and registered at ClinicalTrials.gov, identifier NCT03983694. Written informed parental consent was obtained before the infant’s enrolment.

The inclusion criteria were RBCT prescribed according to local guidelines and corrected age $<4$ weeks.

No exclusion criteria were defined.

Cerebral oxygenation was recorded with a commercially available SRS CW-NIRS device (INVOS 5100C, Covidien Inc., Minneapolis, United States). The device operates at 730 and 830 nm, with an integration time of 1 s, and it provides two source-detection separations: 3 and 4 cm. The sensor was placed on the prefrontal region of the neonate’s head and not moved during the study. The before and after measurements were averages of $\sim 30\min$ immediately before the start and after the end of the transfusion. Depending on the total volume of transfused blood and the weight of each neonate, the RBCT itself lasted at least 4 h. However, the CW-NIRS and BabyLux measurements were not conducted simultaneously. The CW-NIRS recording was paused for the entire duration of BabyLux measurements to prevent crosstalk between the two devices, which could lead to a non-negligible increase in the TD-NIRS background signal.

TD-NIRS and DCS measurements were performed with the BabyLux device to assess cerebral oxygenation and perfusion before and after RBCTs. The device operates at 690, 760, and 830 nm, with an integration time of 1 s. Measurements, consisting of five 60-s repetitions, were conducted both before and after the transfusion, with a maximum of 30 min between the measurements and the transfusion. A source-detection separation of 1.5 cm was used. The probe was placed in the same location on the fronto-parietal region of the head, but opposite side of the INVOS probe. During the measurements, the BabyLux probe was fixed with a black elastic bandage. To avoid direct light to affect measurements and to further reduce risks (of eye damage), a capacitive sensor was used to immediately stop the laser emission in case of no-contact with the skin is detected. Moreover, to assure the sterilization, a sterile ultrasound probe cover was used during measurement.

Heart rate (HR) and arterial saturation (${\mathrm{S}}_{\mathrm{p}}{\mathrm{O}}_2$) were continuously monitored according to clinical practice using Radical-7 Pulse CO-Oxymeter (Masimo, Irvine, CA, United States) or Philips Intellivue X2 (Philips, Amsterdam, The Netherlands) and continuously recorded. Moreover, invasive blood pressure was recorded if available; otherwise, non-invasive blood pressure measurements were performed three times before and after the transfusion.

Synchronization of peripheral saturation and BabyLux data was performed retrospectively by aligning the acquisition times from each device.

Capillary hemoglobin concentration was measured before and after the transfusion. The blood samples were within 6 h before the transfusion, and 1 h after the transfusion.

2.2.

Data Analysis

TD-NIRS data were analyzed using the semi-infinite homogeneous model for photon migration,^31,32 and the optical properties (absorption coefficient and reduced scattering coefficient) at the three wavelengths were retrieved. From the wavelength-dependent absorption coefficients, by means of Beer’s law, and assuming a water content of 90%, ${\mathrm{O}}_2\mathrm{Hb},\mathrm{HHb}$, $\mathrm{tHb}={\mathrm{O}}_2\mathrm{Hb}+\mathrm{HHb}$, and ${\mathrm{S}}_{\mathrm{t}}{\mathrm{O}}_2={\mathrm{O}}_2\mathrm{Hb}/\mathrm{tHb}$ were calculated.³³

Concerning BFI, the DCS data were analyzed using the semi-infinite homogeneous model for the electric field autocorrelation function (retrieved from the intensity autocorrelation function through the Siegert relation).^34,35 The optical properties used to retrieve the BFI were the ones measured with the TD-NIRS module at 760 nm and averaged among the measurements performed before and after the transfusion on the same neonate. However, BFI measured by DCS is expressed in a non-traditional unit ${\mathrm{cm}}^2/\mathrm{s}$; thus, to convert it into a more conventional unit for cerebral blood flow (CBF, ml/100g/min), the BFI measured by DCS was calibrated with the coefficient estimated by Giovannella et al. in newborn piglets.³⁶ The signal quality exclusion criteria are the ones reported in our previous work, where similar data analysis was performed.³⁷

These results were combined with the concentration of hemoglobin in the blood [Hb], and with ${\mathrm{S}}_{\mathrm{p}}{\mathrm{O}}_2$, to obtain $\mathrm{TOE}=({\mathrm{S}}_{\mathrm{p}}{\mathrm{O}}_2-{\mathrm{S}}_{\mathrm{t}}{\mathrm{O}}_2)$ and the index of the cerebral metabolic rate of oxygen, ${\mathrm{tCMRO}}_2=\mathrm{TOE}·\mathrm{CBF}·k·[\mathrm{Hb}]$, where $k$ is the oxygen-carrying capacity of hemoglobin ($k=1.39\mathrm{ml}$ of ${\mathrm{O}}_2$ per g of Hb).

To compute TOE and ${\mathrm{tCMRO}}_2$, all signals were synchronized (after recording), and the ${\mathrm{tCMRO}}_2$ was estimated for each second of measurement, overall results (optical and hemodynamic parameters) for “before” and “after” transfusion were obtained as the grand mean of measurement during the five repetitions of 1-min recordings by the BabyLux device. The results obtained before and after the transfusions were compared with the paired Wilkinson signed-rank test.

4.1.

Strengths and Limitations

The sample size is limited. Although the confirmation of the increase in cerebral oxygenation can be considered firm, the results as regards the effects on the blood flow and metabolic rate had very wide confidence limits. The sample size was too small to consider subgroup analysis or any other exploring of the inter-individual differences in the response to transfusion, so the aim of the study to help define a better indication for RBCT in well, growing preterm infants was not reached.

Due to the current construction of the TD-NIRS and DCS optodes (with light fibers), it is not possible to do effective long-term monitoring and studying effects of interventions over hours or days, and therefore, the estimate of the effect of transfusion includes the error introduced by removal/replacement of the probe. By contrast, the CW-NIRS sensor could stay on for the entire study period; however, the variability in responses to transfusion appeared higher, rather than lower, than by TD-NIRS, suggesting that some time-dependent, unexplained variation was present in the CW-NIRS signal.

The re-siting imprecision of DCS, unfortunately, was much higher than that of TD-NIRS in the term newborn study, as already established by Giovannella et al.,²⁸ and the data in this study confirm that this is a problem. A previous study performed on anesthetized and paralyzed piglets³⁵ showed a lower error in probe repositioning; thus, one likely reason is the movement of healthy infants in clinical studies.

For the purpose of the present study, the problem of the probe replacement variability could have been reduced by not removing the BabyLux sensor during the study, measuring cerebral hemodynamic variations also during the transfusion. However, for the purpose of clinical use, using cerebral oximetry and CBF to judge the need for a red blood cell transfusion, this source of imprecision might remain a limitation.

The CBF calculation is strongly influenced by the reduced scattering coefficient, which is affected by the structure of the neonatal head. In particular, in a previous study performed by the same authors,³⁸ it has been demonstrated that when homogeneous models are used to analyze TD-NIRS data, the reduced scattering coefficient is underestimated, and the error is larger in the case of preterm neonates with wider subarachnoidal space. However, by performing a Wilcoxon unsigned test (applying the Bonferroni correction), the CBF measured in previous studies performed on term neonates with the same device^39,40 is not statistically significantly different from that reported in this study.

4.2.

Physiological Interpretation

The increase in cerebral oxygenation would suggest a benefit of the transfusion in terms of increased oxygen availability and/or reduced risk of cerebral hypoxia if stressed by apnea or bradycardia with systemic hypoxemia, which is common in this population.

In the clinical situation of stable, growing preterm infants, however, it is surprising that correction of anemia results in increased cerebral oxygenation. The coupling between cerebral metabolism and cerebral blood flow, as well as the decreased blood viscosity in anemia, would both cause cerebral blood flow to increase with anemia to compensate for the reduced oxygen-carrying capacity and so, at least from a gross perspective, to maintain capillary oxygen tension—which is the driving force for the diffusion of oxygen toward the mitochondria of brain cells. Cerebral tissue oxygenation as measured by NIRS, however, is not capillary, and it is dominated by the hemoglobin in the blood in larger vessels, of which most is venous. Is it possible that transfusion induces a major shift in the proportion of arterial, capillary, and venous blood? If so, it would rather be in the direction of less arterial blood because it is the arteries that dilate in response to the decreased oxygen extraction due to the reduced oxygen-carrying capacity.

Another factor is the oxygen-hemoglobin dissociation curve, which may be left-shifted in bank blood due to GDP-6-phosphate depletion. This will indeed “require” a higher saturation to maintain the capillary oxygen tension.

The blood flow retrieved by DCS could be affected by Hct level; indeed, BFI is dependent on both the average flow and density of red blood cells in the interrogated tissue volume.⁴¹ In the case of RBCT, Hct level strongly increases after transfusion, biasing a before–after comparison. A correction factor between Hct level and BFI has been proposed.⁴² In a model, the Hct correction factor was used, $c=1/((1-\mathrm{Hct}·\phi ))$, where $\phi$ was empirically determined to be 1.8. However, this factor was tested only on phantoms, and no in vivo validation has been reported. For the sake of completeness, we corrected the measured BFI and the corresponding ${\mathrm{tCMRO}}_2$ for the change in Hct as a result of the transfusion, obtaining BFI* and ${\mathrm{tCMRO}}_2$*. Both now increased, $+31\%$ for BFI* and $+36\%$ for ${\mathrm{tCMRO}}_2$*. These changes, however, are in the range of BFI repositioning imprecision (28%)²⁹ and, more importantly, are very unexpected from physiological reasoning.