Objectives. Premature infants experience brain injury, ie, germinal matrix–intraventricular hemorrhage (GMH-IVH) and periventricular leukomalacia (PVL), in considerable part because of disturbances in cerebral blood flow (CBF). Because such infants are susceptible to major fluctuations in mean arterial blood pressure (MAP), impaired cerebrovascular autoregulation would increase the likelihood for the changes in CBF that could result in GMH-IVH and PVL. The objectives of this study were to determine whether a state of impaired cerebrovascular autoregulation could be identified reliably and conveniently at the bedside, the frequency of any such impairment, and the relation of the impairment to the subsequent occurrence of severe GMH-IVH and PVL.
Patients and Methods. To monitor the cerebral circulation continuously and noninvasively, we used near-infrared spectroscopy (NIRS) to determine quantitative changes in cerebral concentrations of oxygenated hemoglobin (HbO2) and deoxygenated hemoglobin (Hb) from the first hours of life. Our previous experimental study showed a strong correlation between a measure of cerebral intravascular oxygenation (HbD), ie, HbD = HbO2 − Hb, determined by NIRS, and volemic CBF, determined by radioactive microspheres. We studied 32 very low birth weight premature infants (gestational age: 23–31 weeks; birth weight: 605-1870 g) requiring mechanical ventilation, supplemental oxygen, and invasive blood pressure monitoring by NIRS from 1 to 3 days of age. MAP measured by arterial catheter pressure transducer and arterial oxygen saturation measured by pulse oximetry were recorded simultaneously. The relationship of MAP to HbD was quantitated by coherence analysis.
Results. Concordant changes (coherence scores >.5) in HbD and MAP, consistent with impaired cerebrovascular autoregulation, were observed in 17 of the 32 infants (53%). Eight of the 17 infants (47%) developed severe GMH-IVH or PVL or both. Of the 15 infants with apparently intact autoregulation, ie, coherence scores <.5, only 2 (13%) developed severe ultrasonographic lesions. Thus, for the entire study population of 32 infants, 8 of the 10 with severe lesions exhibited coherence scores >.5.
Conclusions. We conclude that NIRS can be used in a noninvasive manner at the bedside to identify premature infants with impaired cerebrovascular autoregulation, that this impairment is relatively common in such infants, and that the presence of this impairment is associated with a high likelihood of occurrence of severe GMH-IVH/PVL.
- near-infrared spectroscopy
- cerebrovascular autoregulation
- intraventricular hemorrhage
- periventricular leukomalacia
A large body of experimental and clinical observations indicates that disturbances in cerebral blood flow (CBF) are important in the pathogenesis of germinal matrix–intraventricular hemorrhage (GMH-IVH) and periventricular leukomalacia (PVL), the 2 most important forms of brain injury in premature infants.1–3The human infant seems to have a particular propensity for development of disturbances in CBF for 2 major reasons. First, because of their very high incidence of respiratory disease, need for mechanical ventilation, the complications of such ventilation, and a variety of other adverse medical conditions, alterations in mean arterial blood pressure (MAP) are very common in such infants. Second, considerable experimental data and some clinical results indicate that cerebrovascular autoregulation, the mechanism by which CBF is maintained constant despite alterations in MAP, is either defective or absent in at least some infants (see the “Discussion” section). Several excellent studies using the xenon-133 clearance technique to measure CBF have shown impaired autoregulation in sick premature infants and intact autoregulation in clinically stable infants. One such study has shown a clear relationship between impaired autoregulation and the subsequent occurrence of GMH-IVH.4,5 However, the xenon-133 clearance technique to determine CBF involves radiation exposure and cannot provide the minute-to-minute information required for optimal delineation of impaired cerebrovascular autoregulation.
Near-infrared spectroscopy (NIRS) is a relatively recently developed technique that can provide continuous bedside quantitation of changes in cerebral concentration of oxygenated hemoglobin (HBO2) and deoxygenated hemoglobin (Hb).6–9 We have shown recently in a piglet model that declines in CBF (measured by radioactive microspheres) caused by decreases in MAP can be detected by simultaneous measurements of cerebral intravascular oxygenation (HbD), measured by NIRS as the difference between HbO2 and Hb, ie, HbD = HbO2 − Hb.1 Thus, we undertook this study of premature infants from the first hours of life to determine whether we could identify infants with the hallmark of impaired cerebrovascular autoregulation, ie, concordant changes in cerebral perfusion, assessed by NIRS measurements of HbD, and in MAP, assessed by conventional techniques. Moreover, because there is increasing evidence that autoregulation is not a simple static system but rather a more complex rate-sensitive system, we used coherence analysis to evaluate autoregulation as a dynamic process. We then determined whether the likelihood of occurrence of cranial ultrasonographic evidence for severe hemorrhagic and/or ischemic disease (GMH-IVH/PVL) was related to the previous occurrence of impaired cerebrovascular autoregulation. The data indicate that NIRS can be used to identify premature infants with concordant changes in HbD and MAP, ie, impaired cerebrovascular autoregulation, that this abnormality is relatively common in such infants, and that the presence of the abnormality is associated with a high likelihood for development of GMH-IVH/PVL.
Thirty-two preterm infants with gestational age <32 weeks were enrolled in the study from the neonatal intensive care units of the Children's Hospital, Boston, and the Brigham and Women's Hospital, Boston, between November 1993 and August 1995. Total enrollment was determined by the entry criteria, consent rate, and enrollment period. Average gestational age was 27.1 ± 2.5 weeks, and average birth weight was 1023 ± 341 g. There were 10 female infants and 22 male infants. To study a cohort of premature infants at high risk for brain injury, infants were recruited on the day of birth from those admissions with gestational age <33 weeks and respiratory distress syndrome requiring intubation, mechanical ventilation, supplemental oxygen, and placement of umbilical arterial catheters for blood pressure monitoring and arterial blood gas sampling. All infants had lung disease by clinical examination and by chest radiographs consistent with hyaline membrane disease or pulmonary insufficiency of extreme prematurity. Infants with congenital anomalies or known intracranial hemorrhage or other lesions at the time of diagnosis were excluded. The study was approved by the human studies committees of each hospital. Informed consent was obtained from the parents of each infant.
NIRS measurements were collected with a 4-wavelength (776, 828, 848, 913 nm) spectrophotometer (NIRO-500, Hamamatsu Photonics, Hamamatsu, Japan). The instrument transmits laser diode light via a fiber-optic bundle to the head and thereby to the intracranial compartment. An identical bundle receives transmitted light from the intracranial compartment and conveys the light to a sensitive photomultiplier tube. The end of each fiber-optic bundle (optode) was affixed to a thin, flexible foam pad and the optodes were spaced 3 cm apart. The foam pads were applied to the skin of the scalp on a line 1 cm above the supraorbital ridge with the transmitting optode 1 cm lateral to the midsagittal plane on either the right or left forehead and the receiving optode 3 cm lateral to the transmitting optode. This placement ensured sampling of cerebral hemoglobin signals from fronto-parietal brain tissue and avoided the sagittal sinus. The optode pads were secured to the head with a flexible bandage wrapped around the head. An opaque hat was then placed on the infant's head to further shield the optodes from ambient light. Optode position was alternated between right and left scalp positions between recordings to minimize skin irritation. Because the skin of extremely premature infants is very thin and easily abraded, no adhesive tape was used.
Changes in the relative concentrations of HbO2and Hb were calculated from changes in light absorption at each of the transmitted laser frequencies according to the Beer-Lambert law.8,10 The calculations were corrected for the interoptode distance and the effect of light scattering, using a differential pathlength factor of 4.39.11 Concentration changes in HbO2 and Hb were expressed in Umol/L. HbD was calculated as HbO2 − Hb.
During monitoring periods, NIRS measurements were recorded simultaneously with arterial Sao2monitored by pulse oximetry (N-200, Nellcor, Hayward, CA) and with MAP (HP 78801A neonatal monitor, Hewlett Packard, Palo Alto, CA). Blood pressure was measured from a pressure transducer connected to the umbilical arterial catheter. Positioning of the tip of the umbilical catheter in the descending aorta was confirmed by chest radiograph. Analog outputs from the blood pressure monitor and pulse oximeter were sampled by an analog to digital converter built into the NIRO-500. The values were sampled by the NIRO-500 instrument every .5 seconds, and the average value for nonoverlapping 5- or 10-second intervals was written to a computer file with calculated values for HbO2 and Hb. The averaging intervals, which represented ∼10 to 30 heart beats, allowed accurate assessment of MAP with less high frequency noise. To eliminate episodes of artifact, the data were separated into 30-minute blocks of uninterrupted recordings for coherence analysis. Only periods in which Sao2 fluctuated <5% were considered to minimize the influence of Sao2 on HbD.
Following informed consent, the NIRS instrument optodes were positioned on the study infant and recordings were initiated. During the study the infant and the NIRS instrument were continually observed by an investigator or study technician. Clinical care of the infant continued under the direction of the infant's attending physician without interruption or modification by the investigators. Clinical events that could influence the NIRS signal baseline, such as head positioning changes, suctioning, or spontaneous movement, were noted on a log sheet.
Infants were studied during the first 3 hospital days. Hemodynamic instability thought to be associated with GMH-IVH and PVL is often most acute during the first few days of age,2,4,12–14 and the majority of GMH-IVH events occur during the first 3 hospital days.2,14 Recording periods of up to 6 hours on each of the study days were conducted as nearly continuously as possible. However, NIRS recordings were interrupted if movement artifacts significant enough to cause a baseline shift occurred. In practice this constraint resulted in interruptions during physical examinations, chest radiographs, intravenous line procedures, and other clinical events requiring manipulation of the infant. Spontaneous head movements by the infants also interrupted the NIRS measurements and arterial blood gas sampling additionally interfered with recordings attributable to loss of the arterial pressure signal. To eliminate episodes of artifact, the data were separated into 30-minute blocks of uninterrupted recordings for coherence analysis.
Cranial Ultrasound Examinations
Cranial ultrasound examinations were performed routinely on day 3, at 10 to 14 days, and at 1 month of age according to usual clinical practice in the respective neonatal intensive care units. Weekly cranial ultrasound examinations were obtained if abnormalities were detected. Standard coronal and sagittal views were recorded. GMH-IVH was scored according to the system of Papile et al.15Ultrasound criteria for PVL consisted of bilateral echolucencies in white matter areas and/or the presence of ventriculomegaly without grade 3 or 4 GMH-IVH and without evidence of increased intracranial pressure (macrocephaly, abnormally rapid head growth, tense anterior fontanelle, rapid progressive ventricular dilation). Infants were classified by their ultrasonographic findings into 2 outcome groups: 1) those with severe abnormalities (grade 3 or 4 GMH-IVH or PVL); and 2) those with minimal (grade 1–2 GMH-IVH) or no abnormalities.
Examination of the raw recordings suggested that slow changes in HbD and MAP were correlated in some, but not all, records. Because the cerebrovascular autoregulatory response may be frequency specific,16,17 the correlation of HbD and MAP in each recording was quantitated in a frequency specific manner by calculating coherence scores. This approach uses transfer function analysis. Transfer function analysis uses Fourier transforms to compute the statistical correlation of the harmonic components of 2 waveforms. The results can be quantitated as a coherence score, which describes the degree of correlation between the waveforms at a given frequency or frequency range. A coherence of 1.0 indicates perfect frequency-specific correlation, and a coherence of 0 indicates a complete lack of frequency-specific correlation. Coherence scores between .5 and 1.0 indicate increasing concordance between the waveforms. According to the Nyquist theorem of digital sampling, a sampling interval of 5 seconds allows detection of frequencies up to .1 Hz and a sampling interval of 10 seconds allows detection of frequencies up to .05 Hz.18 Coherence scores were computed on each 30-minute block of continuous data for bandwidths covering the limits of the data-sampling interval (0–.01 Hz, .01–.05 Hz, .05–.1 Hz for ultralow, very low, and low frequency ranges, respectively). As described in the “Discussion” section, major physiologic importance can be attributed to the ultralow frequency bandwidth. The .05- to .1-Hz range was only computed when the sampling interval was at least 5 seconds. It should be noted that the nonoverlapping sampling intervals lead to a small potential for aliased information at lower frequencies. However, importantly, because the method was used uniformly for all subjects and epochs, this method should not introduce any intersubject/intrasubject or group bias. Computations were performed with a combination of custom software and a scientific computation software package (MatLab, Natick, MA).
The data were analyzed by grouping according to no/minimal versus severe ultrasound abnormalities (grades 0–2 GMH-IVH vs grades 3–4 GMH-IVH/PVL) and by high (>.5) versus low coherence scores (<.5). Differences between groups were compared with Student'st test or by χ2 testing using commercial statistical software (Systat, SPSS, Chicago, IL). AP value of <.05 was considered significant.
Cerebral Hemodynamic Features
We studied 32 premature infants, as characterized in the “Methods” section. After exclusion of records containing movement artifacts and other baseline shifts, an average of 6.9 ± 4.5 30-minute records were examined by coherence analysis for each infant over the first 3 hospital days. Considered as a total group, average coherence values were low for all infants (.32 ± .10, .32 ± .12, and .33 ± .13 for ultralow, very low, and low frequency ranges). However, examination of the data indicated the presence of concordant HbD and MAP changes in the ultralow frequency band for some infants. Ultralow frequency coherence values >.5, indicative of concordant MAP and HbD changes, were observed in 17 of the 32 infants (53%). Notably, in these 17 infants, not all records exhibited high coherence values, suggesting that the degree of CBF regulation in response to blood pressure changes varied at different times in the same infant.
An example of NIRS and MAP data recorded from a premature infant and indicative of ultralow coherence is shown in Fig 1. The blood pressure recording shows spontaneous fluctuations in MAP over the 30 minutes of recording. Although modest variability in the NIRS HbD recording is also apparent during this period, changes in MAP are not associated with changes in HbD. The ultralow frequency coherence score for this recording is .22. In contrast, the recordings in Fig 2 show spontaneous fluctuations in MAP that are associated with parallel changes in HbD. Thus, pronounced increases and decreases in HbD occurred in parallel with identical changes in the systemic circulation. The ultralow frequency coherence score for this recording is .75, reflecting the concordant nature of the changes in HbD and MAP. This infant subsequently developed ultrasonographic evidence of PVL. Similar concordance of changes in HbD and MAP in 2 other infants are shown in Figs 3 and 4. For the infant whose data are shown in Fig 4, selected time segments showed a high correlation (r values: >.90) by linear regression, although the slope and intercept varied over time. Coherence analysis yielded a markedly elevated ultralow frequency coherence score of .87. During the sampling intervals in all infants, simultaneous measurements of Sao2 by transcutaneous monitoring showed normal values and no appreciable changes (see Figs 1and 3 for examples).
Ultrasonographic Abnormalities as a Function of Cerebral Hemodynamic Characteristics
Cranial ultrasonographic abnormalities developed in 15 of 32 infants. Four infants had grade 1 GMH-IVH, 1 had grade 2 GMH-IVH, 6 had grade 3 GMH-IVH, and 1 had grade 4 GMH-IVH. Two infants developed ultrasonographic features of PVL. One infant had both grade 4 GMH-IVH and PVL. Thus, of the 32 infants, 22 had minimal or no abnormalities (grades 0–2 GMH-IVH), whereas 10 infants developed severe abnormalities (grades 3–4 GMH-IVH or PVL) or both. Average gestational age and birth weight for the 10 infants with severe cranial ultrasound abnormalities were 25.7 ± 1.5 weeks and 878 ± 217 g, and for the 22 infants with minimal or no abnormalities, 27.7 ± 2.6 weeks and 1089 ± 370 g (P values were not significant for both gestational age and birth weight).
To evaluate a possible relationship between impaired cerebrovascular regulation and the occurrence of severe lesions, we compared the maximum ultralow frequency coherence scores in the infants with and without severe ultrasonographic abnormalities. The maximum ultralow frequency coherence score for infants with severe ultrasound abnormalities was .64 ± .19, and the maximum score for infants with normal or minimal ultrasound findings was .45 ± .19 (P = .023). The infants who had coherence values >.5 had a higher rate of severe ultrasonographic abnormalities (8/17 or 47%), compared with infants with coherence scores <.5 (2/15 or 13%;P = .04; Fig 5). For the entire population, 8 of the 10 infants (80%) with severe GMH-IVH or PVL or both exhibited coherence scores >.5.
This study of 32 premature infants, assessed from the first hours of life, demonstrates concordant changes in HbD, assessed by NIRS, and MAP, assessed by intravascular catheter, in approximately one half of the population. Moreover, a strong relationship between this concordance and the development of ultrasonographic evidence for severe GMH-IVH/PVL was documented. The findings have implications for the occurrence of impaired cerebrovascular autoregulation in the premature newborn and the relation of such impairment to the development of severe GMH-IVH/PVL.
Concerning the occurrence of impaired cerebrovascular autoregulation in the premature newborn, the results of this study suggest that such impairment can be identified by simultaneous measurements of HbD and MAP. Thus, in 53% of the infants an ultralow frequency coherence score of >.5 could be identified between HbD and MAP in the ultralow frequency bandwidth. When the coherence function between 2 parameters, here HbD and MAP, is >.5, the changes in 1 signal are linearly related to changes in the other signal at that frequency. We chose to focus on the ultralow frequency (0–.01 Hz) response range because these frequencies correspond to changes occurring over several minutes. Although higher frequency coherence of CBF velocity and MAP have been noted in previous studies of premature infants with relatively brief cerebral Doppler ultrasound recordings,19 NIRS allows analysis of longer recording intervals and much lower frequencies. Changes in MAP increase in amplitude as frequency decreases.20 Because prolonged cerebral ischemia (eg, several minutes to hours) is likely to be more injurious than transient (eg, several seconds) reductions in CBF, prolonged correlations of CBF to these larger changes in MAP are more likely to be pathologically significant. These prolonged correlations are captured by ultralow frequency coherence. A relatively high incidence of ultralow frequency CBF and MAP correlation in premature infants is suggested also by the study of von Siebenthal et al17 of cerebral blood volume (as measured by total Hb [HbT] = HbO2 + Hb) and MAP. However, the coherence scores were not as high when HbT was used to monitor cerebral perfusion in that study as when HbD was used in the present study. This discrepancy is consistent with the findings of our previous experimental study in piglets which showed that HbD is a better measure of CBF than is HbT during hypotensive episodes.21
We propose that the concordance between HbD and MAP reflects impaired cerebrovascular autoregulation in considerable part because of our previous study of these parameters in the neonatal piglet.21 Thus, as noted earlier, under conditions of hypotension severe enough to cause a decrease in CBF, simultaneous measurements of HbD showed a parallel decrease. Importantly, the decrease in HbD was shown to correlate with a decrease in volemic CBF measured simultaneously by radioactive microspheres. The fact that HbD, a measure of cerebral intravascular oxygenation and thereby a reflection of cerebral oxygen delivery, correlates with CBF per se is supported further by the demonstration in our infants by transcutaneous monitoring that arterial Sao2(changes in which could alter HbD) was in the normal range and did not change appreciably during the study. Thus, under such conditions, changes in cerebral oxygen delivery should reflect changes in CBF. This conclusion is stated with the assumption that major changes in cerebral oxygen utilization did not occur concordantly with the changes in MAP. We have no reason to believe that this was likely in our infants.
Our demonstration of impaired cerebrovascular autoregulation or a pressure-passive cerebral circulation in the human premature newborn is not novel but is important because of the relative ease of application of the methodology. The first indication of impaired cerebrovascular autoregulation in at least a subset of premature infants was established in 1979 by Lou and coworkers22 using the xenon-133 clearance technique for measurement of CBF. Subsequent studies with this technique have shown, in general, intact cerebrovascular autoregulation in clinically stable premature infants and a pressure-passive cerebral circulation in sick preterm infants.5,23–27 However, the xenon-133 clearance technique involves radiation exposure and is difficult to use routinely for detection of impaired cerebrovascular autoregulation because only a very few time points can be used. Indeed, any technique for measuring cerebral perfusion with such limited temporal features is of somewhat limited value in detection of a dynamic process such as cerebrovascular autoregulation. This limitation may underlie the lack of detection of a concordance between MAP and single measurements of CBF by NIRS in a recent study.28 Utilization of the continuous measurement of HbD by NIRS, simultaneous with measurement of MAP, circumvents the limitations of previous approaches and suggests that impaired cerebrovascular autoregulation is relatively common in the premature newborn.
Concerning the relationship of impaired cerebrovascular autoregulation to the development of severe GMH-IVH/PVL, the results of this study suggest a direct relationship. Thus, approximately one half of the infants with the hemodynamic impairment developed severe ultrasonographic abnormalities, and among all infants with the severe lesions, 80% had maximum coherence scores consistent with impaired cerebrovascular autoregulation. Because infants did not have systematic ultrasound scans before the third day of life, our findings do not prove that the impaired cerebrovascular autoregulation was present before the occurrence of severe GMH-IVH/PVL. However, a decisive demonstration of a relationship between a pressure-passive cerebral circulation and the occurrence of IVH was made initially in a study of 57 preterm infants by the xenon-133 clearance technique.4–5 This cerebrovascular regulatory abnormality may account in considerable part for the previously reported relationships between the subsequent occurrence of IVH and increases in MAP or decreases in MAP followed by reperfusion.13,29–33A similar explanation may account for the relations shown between the subsequent occurrence of PVL and decreases in MAP or the presence of related markers of circulatory insufficiency (eg, hypoplastic left heart syndrome, extracorporeal membrane oxygenation, fetal/neonatal acidemia, hypovolemia, elevations of plasma uric acid on the first postnatal day).32,34–42
The findings of this preliminary study suggest that continuous measurements of HbD by NIRS, simultaneous with measurements of MAP, provide the capability to define reliably and conveniently the presence of impaired cerebrovascular autoregulation in the human premature newborn. Moreover, the data indicate that this abnormality is relatively common in premature infants, and more importantly has a strong relationship with the subsequent occurrence of severe cranial ultrasonographic abnormalities. Taken together, the results raise the possibilities that the infant at highest risk for the development of GMH-IVH/PVL can be identified by NIRS before the occurrence of the lesions and that correction of the cerebral circulatory disturbance might lead to prevention of the lesions. Clearly what is needed now is confirmation of these findings in a larger study; determination of whether the temporal characteristics and the severity of the cerebral circulatory abnormality provide more refined predictive value for the development of the lesions; definition of the causes of the cerebral circulatory abnormality and the means of correction thereof; and determination of whether correction of the abnormality leads to prevention of the brain injury.
This work was supported by research Grants NS32570, HD18655, and NS-38475 (to J.J.V.), Grant HD01010 (to M.T.); Grant NS01721 (to A.D.P.); and Grants HL-59212 and HD-96007 (to J.P.S.) from the National Institutes of Health.
The near infrared spectroscopy instrument was provided by Hamamatsu Photonics, KK, Hamamatsu, Japan.
Dr Timothy Watkins contributed to the coherence analyses.
- Received March 8, 2000.
- Accepted June 22, 2000.
Reprint requests to (J.J.V.) Neurology Department, Fegan 1103, Children's Hospital, 300 Longwood Ave, Boston, MA 02115. E-mail:
Dr Tsuji is currently with St Joseph's Hospital, Milwaukee, Wisconsin.
Dr Saul is currently with Children's Heart Center of South Carolina, Charleston, South Carolina.
Jamil Sobh is currently with Microwave Medical Systems, Acton, Massachusetts.
- CBF =
- cerebral blood flow •
- GMH-IVH =
- germinal matrix–intraventricular hemorrhage •
- PVL =
- periventricular leukomalacia •
- MAP =
- mean arterial blood pressure •
- NIRS =
- near-infrared spectroscopy •
- HbO2 =
- oxygenated hemoglobin •
- Hb =
- hemoglobin •
- HbD =
- cerebral intravascular oxygenation •
- HbT =
- total hemoglobin •
- Sao2 =
- oxygen saturation
- Volpe JJ. Neurology of the Newborn. Philadelphia, PA: WB Saunders; 1995
- Tsuji M
- Lou HC. Pathogenesis of hemodynamic brain lesions in the newborn. In: Wasterlain CG, Vert P, eds. Neonatal Seizures. New York, NY: Raven Press; 1990:123–128
- Bracewell RN. The Fourier Transform and Its Applications. New York, NY: McGraw-Hill; 1978
- Parati G,
- Saul JP,
- Di Rienzo M,
- Mancia G
- Muller AM,
- Morales C,
- Briner J,
- Baenziger O,
- Duc G,
- Bucher HU
- Tyszczuk L,
- Meek J,
- Elwell C,
- Wyatt J
- Perry EH,
- Bada HS,
- Ray JD,
- Korones SB,
- Arheart K,
- Magill HL
- Kuban K,
- Volpe JJ
- Miall-Allen VM,
- de Vries LS,
- Whitelaw AG
- Glauser TA,
- Rorke LB,
- Weinberg PM,
- Clancy RR
- Copyright © 2000 American Academy of Pediatrics