|Home | About | Journals | Submit | Contact Us | Français|
The authors explored associations between blood gas abnormalities in more than 1,000 preterm infants during the first postnatal days and indicators of neonatal brain damage. During 2002–2004, women delivering infants before 28 weeks’ gestation at one of 14 participating institutions in 5 US states were asked to enroll in the study. The authors compared infants with blood gas values in the highest or lowest quintile for gestational age and postnatal day (extreme value) on at least 1 of the first 3 postnatal days with the remainder of the subjects, with separate analyses for blood gas abnormalities on multiple days and for partial pressure of oxygen in the alveolar gas of <35. Outcomes analyzed were ventriculomegaly and an echolucent lesion on an ultrasound scan in the neonatal intensive care unit, and cerebral palsy, microcephaly, and a low score on a Bayley Scale of Infant Development at 24 months. Every blood gas derangement (hypoxemia, hyperoxemia, hypocapnia, hypercapnia, and acidosis) was associated with multiple indicators of brain damage. However, for some, the associations were seen with only 1 day of exposure; others were evident with 2 or more days’ exposure. Findings suggest that individual blood gas derangements do not increase brain damage risk. Rather, the multiple derangements associated with indicators of brain damage might be indicators of immaturity/vulnerability and illness severity.
Brain damage and dysfunctions in preterm infants have been attributed to hypoxia (1), hyperoxia (and oxidative stress) (2–4), hypocapnia (5, 6), hypercapnia (7, 8), and acidemia (9, 10). Blood gas derangements might be indicators of causes of brain damage rather than causes of brain damage themselves (11). For example, in preterm infants, inflammation of the placenta has been associated with elevated concentrations of partial pressure of oxygen in the alveolar gas (PaO2) (12), and inflammation of the umbilical cord has been associated with reduced partial pressure of carbon dioxide (PCO2) (13). Intravenous endotoxin causes fetal hypoxemia without acidemia (14), whereas chronic antepartum vascular disease has been associated with acidemia (15). Thus, blood gas derangements might be indicators of earlier, potentially damaging exposures rather than contributors to brain damage.
In a cohort of extremely low gestational age newborns (ELGANs), defined as those born before the 28th week of gestation, we evaluated the association of blood gas abnormalities during the first 3 postnatal days with ultrasound-defined cerebral white matter damage and later motor and cognitive dysfunctions. By adjusting for antecedents (or their surrogates) of blood gas derangements in multivariable models of abnormalities of brain structure and function, we attempted to separate the contribution of what might have caused a blood gas derangement from the contribution of the blood gas itself.
The ELGAN study was designed to identify the risk factors for structural and functional neurologic disorders in extremely low gestational age newborns (16). Early postnatal physiologic derangements, such as blood gas extremes, are among these risk factors. The institutional review boards at all study centers approved the recruitment procedures.
During 2002–2004, women who gave (or were at risk of giving) birth before 28 weeks of gestation at one of 14 participating US medical centers in 11 US cities in 5 states were approached for consent either upon admission or shortly after delivery, depending on clinical circumstances and institutional preference. A total of 1,249 mothers of 1,506 infants consented (Table 1). A total of 1,172 newborns survived to postnatal day 7, had a protocol head ultrasound, and had blood gas measurements on 2 of the first 3 days of life. Of those, 998 survived to 24 months postterm equivalent, and 888 (89%) had at least 1 of 4 neurodevelopmental evaluations.
After delivery, a trained research nurse interviewed the mother in her native language by using a structured data collection form. The clinical circumstances that led to each maternal admission and ultimately to each preterm delivery were operationally defined by using data from both the maternal interview and the medical record abstraction (17).
Placentas were biopsied under sterile conditions. Eighty-two percent of the samples were obtained within 1 hour of delivery. The microbiologic (18) and histologic (19) procedures and definitions are presented elsewhere.
The gestational-age estimates were based on a qualitative hierarchy of available information. Included were dates of embryo retrieval or intrauterine insemination or fetal ultrasound before the 14th week (62%), fetal ultrasound at 14 or more weeks (29%), last menstrual period without fetal ultrasound (7%), and gestational age recorded in the log of the neonatal intensive care unit (1%). The birth weight z score reflects the number of standard deviations the infant's birth weight is above or below the median weight of infants at the same gestational age in a standard data set (20).
Blood gases were measured when clinically indicated. The median number of blood gases obtained on each day declined rapidly from 8 on day 1 to 4 on day 3. Only arterial blood was used for oxygen measurements (indicated by the a in PaO2). Measurements of PCO2 and pH were from arterial blood except for 7 ELGANs on day 1, 15 on day 2, and 60 on day 3, for whom these measurements were made on venous blood.
Children who did not have a blood gas measurement on postnatal day 3 were more likely to be physiologically stable compared with children who had a set of blood gas measurements that day, and they were also much less likely than others to have postnatal day 2 measurements in any extreme quintile. Assigning these nonextreme measurements to postnatal day 3 enabled us to include these children in this sample and avoid inflating odds ratios inappropriately. Imputations were classified as arterial or venous based on the most recent measurement.
When a day 1 measure was not available (16 children lacked a PaO2, 1 lacked a PCO2, and 1 lacked a pH measurement), we substituted the day 2 measure for the missing day 1 measure. When no day 2 measure was available (23 lacked a PaO2 and 6 lacked a pH measurement), the average of the day 1 and day 3 measures was used. Finally, when a day 3 measure was missing (175 lacked a PaO2 and 56 lacked a pH measurement), we substituted the day 2 measure.
We classified ELGANs by their most extreme blood gas measurements on any of the first 3 postnatal days. For each of these days, we recorded the lowest and highest PaO2, PCO2, and pH values. In our sample, the blood gas measurement that defined the extreme quintile varied by gestational age and by postnatal day (Table 2) and minimally by whether the specimen was arterial or venous. Consequently, we classified infants by whether or not their extreme value each day was in the extreme quintile for their gestational age (23–24, 25–26, and 27 weeks) and whether the specimen was arterial or venous.
Because we collected the minimum and maximum blood gas values each day, we could not determine whether they were from the same specimen. Because the extreme pH and PCO2 measurements were not paired, we did not calculate base excess.
Procedures for obtaining and reading ultrasound scans are described elsewhere (21). Two independent readers had to agree on the presence of every lesion reported. A child was considered to have ventriculomegaly if, compared with templates (22), the ventricles were moderately or severely enlarged on the third protocol scan (obtained between the 15th postnatal day and the 40th postmenstrual week).
Fully 91% of the children returned for a developmental assessment close to the time that they would be a corrected age of 24 months. Of these children, 77% had their examination within the range of 23.5–27.9 months.
Those who performed the neurologic examinations demonstrated acceptably low variability (23). The topographic diagnosis of cerebral palsy (quadriparesis, diparesis, or hemiparesis) was based on an algorithm using these data (24).
The largest occipital-frontal circumference was measured to the nearest 0.1 cm. Because newborns were assessed at different approximations of 24 months corrected age, all head circumferences were converted to z scores based on standards in the Centers for Disease Control and Prevention data sets (25).
The generalized form of the null hypotheses we evaluated is that a blood gas extreme during the first 3 postnatal days is no more common among children who
Each of the outcomes did not occur more commonly in sibships than in singletons. In addition, the outcomes were no more concordant than could be explained by chance. Consequently, we viewed all births as independent of one another and included twins and higher-order births along with singletons in all analyses.
The first sets of analyses initially classified a blood gas abnormality as the highest/lowest quintile for gestational age on at least 1 of the 3 days assessed (Table 3) and on 2 or more days (Table 4).
Because the PaO2 that defined the lowest quintile (range, 39–46 mm Hg) might not be considered hazardous, we carried out separate, similar sets of analyses for hypoxemia defined as <35 mm Hg. All other derangements were defined as the lowest quintile for gestational age on at least 1 day (Table 5) and on 2 or more days (Table 6).
In this sample, high PCO2 and low pH levels occurred in the same children more frequently than would be expected if their occurrences were independent. To estimate the effect of each in light of the other gases (lowest quintile of PaO2, highest quintile of PaO2, and lowest quintile of PCO2), we fit them in 2 separate conditional logistic multivariable models, one that included low pH but not high PCO2 (model 1) and one that included high PCO2 but not low pH (model 2). Both models included terms for the other blood gases and potential confounders, as well as a hospital cluster term to account for the tendency for infants born and cared for at a particular hospital to be more like each other than similar infants born and cared for at other hospitals (28).
In this paper, we use the terms “to predict” and “to be associated with” interchangeably. With both, we refer to statistical associations in groups of individuals, not to indicate co-occurrence of 2 characteristics in one individual (29).
Hypoxemia, hypercapnia, and acidemia were strongly associated with ventriculomegaly, whereas only acidemia predicted an echolucent lesion, only hypoxemia predicted quadriparetic cerebral palsy, only hyperoxemia predicted diparetic cerebral palsy, and only hypercapnea predicted hemiparetic cerebral palsy and microcephaly. Hypoxemia accompanied by either hypercapnia or acidemia was associated with both low MDI and low PDI. Refer to Table 3.
Hyperoxemia, hypercapnia, and acidemia predicted ventriculomegaly, whereas no blood gas abnormality predicted an echolucent lesion. Only hypocapnia was associated with quadriparetic cerebral palsy, and only hypercapnea was associated with hemiparetic cerebral palsy, whereas hypocapnia accompanied by either hypercapnea or acidemia was associated with diparetic cerebral palsy. No blood gas predicted an MDI of <70, and only hypercapnea was associated with an increased risk of a PDI of <70, whereas hyperoxemia was associated with a reduced risk of a PDI of <70. Only hypocapnia predicted microcephaly. Refer to Table 4.
Replacing the lowest quintile of PaO2 with a PaO2 of <35, while retaining extreme quintiles for all other blood gases, eliminated hypoxemia as an antecedent of ventriculomegaly, quadriparetic cerebral palsy, and low MDI and PDI. Hypercapnia and acidemia, however, remained strong predictors of ventriculomegaly, hemiparetic cerebral palsy, and low MDI. Only hypercapnia predicted low PDI, whereas only hypocapnia predicted microcephaly. Refer to Table 5.
Having a PaO2 of <35 on 2 or more days was not associated with any of the brain structure or function outcomes. On the other hand, 2 or more days of hypocapnia accompanied by hypercapnia or acidemia predicted ventriculomegaly and diparetic cerebral palsy. In contrast, no blood gas abnormality predicted an echolucent lesion or low MDI. Isolated hypocapnia was associated with quadriparetic cerebral palsy and microcephaly. Refer to Table 6.
Our findings are best interpreted in light of 2 competing views of the physiologic instability of preterm infants. One view holds that blood gas abnormalities are characteristics of physiologic instability that contribute to brain damage. The other view holds that blood gas abnormalities are indicators of vulnerability and are unlikely to contribute to brain damage.
Some suggest that the physiologic instability limits cerebral blood flow, which, in turn, leads to what is described as hypoxia-ischemia or oxygen-nutrient deprivation (1). This has also been called the hemodynamic hypothesis of brain damage in the preterm newborn (30) and postulates that “insult occurs when oxygen delivery falls below oxygen demand” (31, p. 1069). Although we did not measure oxygen demand, we did not find that a PaO2 of <35 mm Hg on at least 1 day predicted any structural or functional brain disorder of statistical significance. A less extremely low PaO2, defined as a PaO2 in the lowest quintile for gestational age and postnatal day, however, predicted ventriculomegaly, quadriparetic cerebral palsy, and both low MDI and low PDI. Although some of this discrepancy reflects the size of the exposed groups (lowest-quintile PaO2 on only 1 day: n = 232; PaO2 of <35 on only 1 day: n = 134), some of the odds ratios really were a bit smaller for a PaO2 of <35.
Hypocarbia, especially if prolonged or recurrent, has been associated with periventricular leukomalacia in the past. Our observations that hypocarbia on a single day did not appear to have any adverse effect but that hypocarbia on multiple days predicted ventriculomegaly, quadriparetic cerebral palsy, and diparetic cerebral palsy extend these findings and raise the concept of sensitization. Findings from animal studies suggest that exposure to a low-intensity (i.e., nondamaging) insult can sometimes increase the probability that a subsequent low-intensity insult will result in brain damage, although, under different circumstances, it can reduce the likelihood of damage (32). In this multihit model of hypocarbia-induced brain damage, the first episode of hypocarbia does not do any damage but instead sensitizes the brain (i.e., making it more vulnerable) to the second episode of hypocarbia or one or more correlates.
The other interpretation of the associations between blood gas abnormalities and brain damage views the blood gas derangements as indicators of other risk factors for brain damage. We consider 2 variations on this theme.
The first variation postulates that the blood gas derangement is a consequence of an earlier brain-damaging exposure (11). Such an exposure can be inflammation (12–14) or placenta vascular disease (15). In our multivariable models of the contribution of blood gas derangements to indicators of brain damage, we adjusted for just these types of potential confounders (maternal fever during pregnancy, use of nonsteroidal antiinflammatory drugs during pregnancy, birth weight z score <−1, delivery for preeclampsia or fetal indications, recovery of Mycoplasma from the placenta, and thrombosis of fetal stem vessels in the placenta). Consequently, we are reluctant to attribute the associations we found between each blood gas derangement and each indicator of brain damage to confounding by antecedents of both.
The second variation postulates that blood gas derangements are indicators of immaturity that supplement maturation information provided by gestational age. Among the characteristics of this immaturity is a paucity of substances that protect the brain from potential damaging exposures (33, 34). If concentrations of proteins with neurotrophic properties increase with increasing gestational age, then the least mature will have the lowest concentrations of the protective proteins and perhaps thereby be at heightened risk of brain cell damage (35). To the extent that the least mature newborns are also the most likely to be physiologically unstable, blood gas derangements, even when they do not contribute to brain damage, will be associated with heightened risks of brain damage.
The most physiologically unstable babies were the ones most likely to have a day 3 blood gas assessment. To minimize this informative missingness, we imputed missing blood gas information. Although doing so is likely to distort results toward the null, it is unlikely to have eliminated “exposure suspicion bias” (36).
Our blood gas measures were defined for each gestational age category (and postnatal day). Consequently, what we found is highly unlikely to reflect gestational age. More likely, however, the blood-gas abnormalities are functional indicators of immaturity not entirely captured by gestational age.
Hypercarbia or acidemia predicted every one of our indicators of brain damage except quadriparetic cerebral palsy and microcephaly. These 2 closely associated derangements, acidemia and hypercarbia, might have contributed to brain damage and functional impairments. On the other hand, the preterm newborn with both acidemia and hypercapnia probably is more likely than others to have respiratory dysfunction that is poorly responsive to efforts to eliminate hypercapnia and acidosis. It is possible that respiratory dysfunction in the preterm newborn serves as an indicator of brain vulnerability associated with immaturity beyond the information conveyed by gestational age. If this hypothesis is correct, the neurodevelopmental dysfunctions seen in children who had acidemia during the first postnatal days would be a consequence of immaturity-related phenomena, such as inadequate concentrations of neurotrophic proteins (33, 34).
Some doubt that the risk/benefit ratio justifies permissive hypercapnia (37). We do not think that our data should be used to support use or avoidance of permissive hypercapnia.
When we began our analyses, we did not know what value of any blood gas was likely to be most discriminating. Quintiles are unbiased, arbitrary dichotomies that provide an adequate number of children considered “exposed” to the extreme, which is especially important when dealing with relatively infrequent outcomes. Because the lowest quintile of PaO2 ranged from 36 mm Hg to 43 mm Hg, a level some might not consider sufficiently low to result in brain damage, we carried out additional analyses with a more extreme PaO2 value, <35 mm Hg, regardless of gestational age.
To avoid a fleeting event being considered capable of damaging the brain, we required that a blood gas measurement in the extreme quintile recur on another day. We did so recognizing that a brief interval of instability at the junction of 2 adjacent days will be identified as occurring on 2 separate days.
Looking at an individual blood gas derangement ignores the contribution of co-occurring blood gas derangements. Consequently, we created multivariable models that included all blood gas variables, as well as models that dropped 1 or 2 of the 5 blood gas variables. These efforts enabled us to determine which blood gas variables shared discriminating information and which ones competed to have the odds ratio with the highest point estimate and narrowest confidence interval.
Our study has several strengths, including a large number of infants (with resulting statistical power), selection of infants based on gestational age and not birth weight (38), prospective collection of data, efforts to minimize “informative missingness” (39) and observer variability (21, 24), and, finally, relatively high rates of follow-up. The weaknesses of our study are those attributed to all observational studies. In addition, the sickest infants were probably more likely than others to have frequent blood gas measurements and to be treated with potentially hazardous interventions, such as high ventilator pressures (40).
The associations we found between blood gas derangements and indicators of brain damage provide some support for the hemodynamic (hypoxic-ischemic) hypothesis of white matter injury in the preterm newborn. The multiplicity of derangements in our study associated with both lung immaturity and indicators of brain damage, however, prompts us to favor the hypothesis that blood gas abnormalities are indicators of (brain and lung) immaturity/vulnerability not captured by gestational age. This is the most parsimonious explanation for what we found.
Author affiliations: Neuroepidemiology Unit, Neurology Department, Children's Hospital Boston, and Harvard Medical School, Boston, Massachusetts (Alan Leviton, Elizabeth Allred); Division of Pediatric Neurology, Department of Pediatrics, Boston Medical Center, Boston University, Boston, Massachusetts (Karl C. K. Kuban); Floating Hospital for Children at Tufts Medical Center, Sackler School of Graduate Biomedical Sciences, Tufts University, Boston, Massachusetts (Olaf Dammann); Department of Pediatrics, Wake Forest University, Winston-Salem, North Carolina (T. Michael O'Shea); Division of Extramural Research, National Institute of Neurological Disorders and Stroke, Bethesda, Maryland (Deborah Hirtz); Department of Pediatrics, University of Chicago Comer Children's Hospital, Chicago, Illinois (Michael D. Schreiber); and Department of Epidemiology, Michigan State University, East Lansing, Michigan (Nigel Paneth).
This study was supported by a cooperative agreement with the National Institute of Neurological Disorders and Stroke (5U01NS040069-05) and a program project grant from the National Institute of Child Health and Human Development (NIH-P30-HD-18655), Bethesda, Maryland.
The authors gratefully acknowledge the contributions of their colleagues.
ELGAN study collaborators who made this report possible: Children's Hospital, Boston, Massachusetts (Kristen Ecklund, Haim Bassan, Samantha Butler, Adré Duplessis, Cecil Hahn, Catherine Limperopoulos, Omar Khwaja, Janet S. Soul); Baystate Medical Center, Springfield, Massachusetts (Bhavesh Shah, Frederick Hampf, Herbert Gilmore, Susan McQuiston); Beth Israel Deaconess Medical Center, Boston, Massachusetts (Camilia R. Martin, Jane Share); Brigham and Women's Hospital, Boston, Massachusetts (Linda J. Van Marter, Sara Durfee); Massachusetts General Hospital, Boston, Massachusetts (Robert M. Insoft, Sjirk J. Westra, Kalpathy Krishnamoorthy); Floating Hospital for Children at Tufts Medical Center, Boston, Massachusetts (Cynthia Cole, John M. Fiascone, Roy McCauley, Paige T. Church, Cecelia Keller, Karen J. Miller); U Mass Memorial Health Care, Worcester, Massachusetts (Francis Bednarek, Jacqueline Wellman, Robin Adair, Richard Bream, Alice Miller, Albert Scheiner, Christy Stine); Yale University School of Medicine, New Haven, Connecticut (Richard Ehrenkranz, Cindy Miller, Nancy Close, Elaine Romano, Joanne Williams); Wake Forest University Baptist Medical Center and Forsyth Medical Center, Winston-Salem, North Carolina (T. Michael O'Shea, Barbara Specter, Deborah Allred, Robert Dillard, Don Goldstein, Deborah Hiatt, Gail Hounshell, Ellen Waldrep, Lisa Washburn, Cherrie D. Welch); University Health Systems of Eastern Carolina, Greenville, North Carolina (Stephen C. Engelke, Ira Adler, Sharon Buckwald, Rebecca Helms, Kathyrn Kerkering, Scott S. MacGilvray, Peter Resnik); North Carolina Children's Hospital, Chapel Hill, North Carolina (Carl Bose, Lynn A. Fordham, Lisa Bostic, Diane Marshall, Kristi Milowic, Janice Wereszczak); Helen DeVos Children's Hospital, Grand Rapids, Michigan (Mariel Poortenga, Bradford W. Betz, Steven L. Bezinque, Joseph Junewick, Wendy Burdo-Hartman, Lynn Fagerman, Kim Lohr, Steve Pastyrnak, Dinah Sutton); Sparrow Hospital, Lansing, Michigan (Ellen Cavenagh, Victoria J. Caine, Nicholas Olomu, Joan Price); Michigan State University, East Lansing, Michigan (Nigel Paneth, Padmani Karna); University of Chicago Medical Center, Chicago, Illinois (Michael D. Schreiber, Kate Feinstein, Leslie Caldarelli, Sunila E. O'Connor, Michael Msall, Susan Plesha-Troyke); William Beaumont Hospital, Royal Oak, Michigan (Daniel Batton, Karen Brooklier, Beth Kring, Melisa J. Oca, Katherine M. Solomon); Arkansas Children's Hospital, Little Rock, Arkansas (Joanna J Seibert); and Children's Hospital of Atlanta, Atlanta, Georgia (Robert Lorenzo).
Conflict of interest: none declared.