OBJECTIVE. The goal was to study the incidence of kernicterus, developmental delay, autism, cerebral palsy, and hearing loss in infants with peak total serum bilirubin levels of ≥325 μmol/L (≥19 mg/dL), compared with infants with less-severe or no hyperbilirubinemia, in a population of healthy term and late preterm infants.
METHODS. Prospectively gathered, standardized, maternal and neonatal data for infants at ≥35 weeks of gestation who were born between January 1, 1994, and December 31, 2000, were extracted from the Nova Scotia Atlee Perinatal Database. Infants with Rh factor isoimmunization, significant congenital or chromosomal abnormalities, or severe peripartum asphyxia were excluded. Comparisons were made on the basis of peak total serum bilirubin levels. Diagnoses were obtained through data linkage with the Medical Services Insurance Database for office visits and the Canadian Institute for Health Information Database for hospital admissions. The registration file provided information allowing calculation of follow-up times, which were determined for each separate outcome. Follow-up periods ranged from 2 to 9 years, with the end point being the first time the diagnostic code was encountered in either database. Cox proportional-hazards regression analyses were used to examine the relationships between outcomes and total serum bilirubin levels.
RESULTS. Of 61238 infants included in the study cohort, 4010 (6.7%) did not have linkage data, which left 56019 infants for analysis. There were no cases of kernicterus and no significant differences in rates of cerebral palsy, deafness, developmental delay, or visual abnormalities between the groups. There were suggestions of associations with attention-deficit disorder in the severe hyperbilirubinemia group and with autism in the combined moderate and severe hyperbilirubinemia group.
CONCLUSIONS. There was no increase in adverse effects reported previously to be associated with bilirubin toxicity. Associations with developmental delay, attention-deficit disorder, and autism were observed.
Hyperbilirubinemia in term infants is common, with up to 65% of newborns developing clinical jaundice in the first week of life.1 This represents normal physiologic adaptation to extrauterine life for most infants, but extreme levels of hyperbilirubinemia have been causally linked with kernicterus causing permanent neurologic sequelae, including mental retardation, choreoathetoid cerebral palsy, gaze palsies, and sensorineural hearing loss.1–3 It is thought that more-subtle deficits in motor function, attention, and learning may be attributable to more-moderate levels of hyperbilirubinemia, although this has remained controversial.4–8 In 1994, the American Academy of Pediatrics recommendations for phototherapy for healthy term infants were published.9 Contemporaneous with the introduction of those guidelines, there were increased numbers of reports of kernicterus in the literature, although clear comparisons of incidence rates before and after guideline introduction are not available.10–16 A large cohort study suggested associations between high bilirubin levels and abnormal neurologic outcomes in children with bilirubin levels of >342 μmol/L (>20 mg/dL)17; however, more-recent reports have not shown significant increases in cerebral palsy rates among infants with bilirubin levels of >425 μmol/L (>25 mg/dL).18 The rates of cerebral palsy in the general population range from 1.2 to 2 cases per 1000.19 Population-based estimates of the prevalence of hearing loss in children <5 years of age range from 0.6% to 1.7%.17,20 In studies examining risk factors for hearing loss, the incidence rates of hearing loss varied with serum bilirubin levels, but rates were not statistically different.17,21,22 One report also suggested an association between hyperbilirubinemia and autism,23 although this was not found in a larger, population-based study.24
In Nova Scotia, the rate of documented hyperbilirubinemia with total serum bilirubin (TSB) levels of ≥230 μmol/L (≥13.5 mg/dL) increased by 23% after implementation of the American Academy of Pediatrics guidelines, but a 58% reduction in the use of phototherapy and a significant decrease in the length of hospital stay were seen in the same time period (K.A.J. and A.C.A., unpublished data, 2003). A study by Sgro et al25 documented that infants with severe hyperbilirubinemia often had blood group incompatibilities and predischarge assessment was important. For infants with less-severe TSB levels, however, the rates of readmission with hyperbilirubinemia, the number of infants developing severe hyperbilirubinemia with TSB levels of >325 μmol/L (>19 mg/dL) in the immediate postdischarge period, and the number of infants developing long-term adverse effects are not well documented. Because of the lack of population-based outcome information currently available, we investigated the incidence of adverse effects, including kernicterus, developmental delay, autism, choreoathetoid cerebral palsy, and hearing loss, in infants with TSB levels of ≥230 μmol/L (≥13.5 mg/dL) at any time in the neonatal period, compared with infants with no hyperbilirubinemia, in the population of healthy term and near-term infants born in Nova Scotia between 1994 and 2000.
Ethical approval was obtained from the Reproductive Care Program Joint Data Access Committee and the institutional research ethics boards of the Izaak Walton Killam Health Centre and of the regional district health authorities where health records were examined.
This population-based cohort study included infants with gestational ages of ≥35 weeks who were born in Nova Scotia, Canada, between January 1, 1994, and December 31, 2000. The Nova Scotia Atlee Perinatal Database was used to define the study cohort and to provide clinical information related to the prenatal period, labor and delivery, and the postpartum period. The infants identified in the Nova Scotia Atlee Perinatal Database as having a diagnosis of hyperbilirubinemia, defined as TSB levels of ≥230 μmol/L (≥13.5 mg/dL), were subsequently stratified into 2 groups, that is, those with moderate hyperbilirubinemia (with TSB levels of ≥230 μmol/L but <325 μmol/L, ie, 13.5–19 mg/dL) and those with severe hyperbilirubinemia (with TSB levels of ≥325 μmol/L, ie, >19 mg/dL), as determined through chart review. Infants were excluded if they had documented Rh factor isoimmunization, significant congenital anomalies, severe peripartum asphyxia with Apgar scores of <3 at 5 minutes of age, or documented hypoxic-ischemic encephalopathy. After exclusions, the study cohort contained 60029 infants.
The outcomes of interest consisted of the diagnoses of deafness, cerebral palsy, developmental delay, gaze palsy, attention-deficit disorder, autism spectrum disorders, and a composite of all outcomes. Information about these postneonatal diagnoses was obtained through data linkage of the study cohort with the Medical Services Insurance database for office visits and the Canadian Institute for Health Information Discharge Abstract Database for hospital admissions at the Population Health Research Unit of Dalhousie University. A registration file provided information regarding migration, death, or termination of eligibility for health care services in Nova Scotia for any reason, allowing calculation of follow-up times. There were 4010 infants (6.7%) who did not have linkage to the registration file, leaving 56019 infants available for analysis.
The χ2 test was used to compare characteristics between study groups. Follow-up time was determined separately for each outcome, with the end point corresponding to the first time the diagnostic code of interest was encountered in the Medical Services Insurance or Canadian Institute for Health Information databases or the end of the study follow-up period (December 31, 2002). Cox proportional-hazards regression analyses were used to examine the relationship between hyperbilirubinemia and each outcome. Initially, unadjusted relative risks (RRs) and 95% confidence intervals (CIs) were calculated for each outcome. The following potential confounders were tested for inclusion in a final adjusted model for each outcome: year of birth, region of residence, marital status, maternal age, parity (number of previous viable pregnancies), gravidity (number of previous pregnancies), multiple birth, any breastfeeding at discharge, maternal smoking at the time of birth admission, type of delivery (vaginal versus cesarean), preterm delivery (<37 weeks of gestation), infant gender, small-for-gestational age status (<10th percentile of weight for gestational age), infection, fetal malnutrition (clinical evidence of soft-tissue wasting), evidence of mild birth depression, respiratory distress syndrome, infant anemia, intrauterine hypoxia (defined as polycythemia or reticulocytosis), ABO isoimmunization, gestational diabetes mellitus, and preexisting maternal diabetes mellitus. Isoimmunization in this study referred to ABO isoimmunization and was defined as positive Coomb's test results in the setting of maternal-infant ABO incompatibility. We were unable to test for interactions between ABO isoimmunization and preterm delivery because none of the infants with isoimmunization was delivered before 37 weeks of gestation.
Starting with a full model including hyperbilirubinemia and all potential confounders, we used a backward stepwise approach, removing the potential confounder with the largest P value first and evaluating the impact on the RR for hyperbilirubinemia. If the RR for either the moderate hyperbilirubinemia or severe hyperbilirubinemia group was not affected (ie, relative change of <5%) by removal of the potential confounder, then the potential confounder was excluded permanently from the model; the process was repeated until a parsimonious model was determined. All analyses were conducted with SAS 8.2 (SAS Institute, Cary, NC).
Overall, 3779 infants (6.7%) had TSB levels of ≥230 μmol/L (≥13.5 mg/dL), with 348 (0.6%) having TSB levels of ≥325 μmol/L (≥19 mg/dL). After implementation of the 1994 guidelines, the diagnosis of hyperbilirubinemia, defined in the database as a TSB level of ≥230 μmol/L (≥13.5 mg/dL), increased from 57.5 cases per 1000 in 1994 to a peak of 81.5 cases per 1000 in 1998 (Fig 1).
The demographic information for the mothers and their infants is presented in Table 1. Hyperbilirubinemia was significantly more common in breastfed infants and in those born to women who were ≥35 years of age, nonsmokers during pregnancy, and nulliparous (P < .0001). It was also significantly more common among infants delivered at <37 weeks of gestation, male infants, and those who experienced respiratory distress syndrome (P < .0001).
During the study period, there were no reported cases of kernicterus. There was no significant difference in the composite outcome in either the moderate hyperbilirubinemia group (adjusted RR: 1.1; 95% CI: 1.0–1.2) or the severe hyperbilirubinemia group (adjusted RR: 1.1; 95% CI: 0.8–1.4), compared with the no-hyperbilirubinemia group (Table 2). The risk of cerebral palsy was not increased significantly in the hyperbilirubinemia group, compared with the no-hyperbilirubinemia group (adjusted RR: 1.2; 95% CI: 0.5–2.9). Risks of abnormalities of hearing and vision were not different between the groups. The risk of developmental delay was significantly increased in the moderate hyperbilirubinemia group, compared with the no-hyperbilirubinemia group (adjusted RR: 1.6; 95% CI: 1.3–2.0), but a significant association was not seen for the severe hyperbilirubinemia group. There was a significant increase in the risk of attention-deficit disorder among infants exposed to TSB levels of ≥325 μmol/L (≥19 mg/dL; adjusted RR: 1.9; 95% CI: 1.1–3.3). A nonsignificant increase in the risk of autism was seen with exposure to moderate and severe bilirubin levels (adjusted RR: 1.6; 95% CI: 1.0–2.5).
This large, population-based, cohort study examined the incidence of adverse effects, including kernicterus, deafness, cerebral palsy, developmental delay, gaze palsy, attention-deficit disorder, and autism spectrum disorders, in infants with TSB levels of ≥230 μmol/L (≥13.5 mg/dL) at any time in the neonatal period, compared with infants with no hyperbilirubinemia, in the population of healthy term and near-term infants born in Nova Scotia between 1994 and 2000. Unlike other recent studies based on information from individual hospitals or health providers, this study represents 94% of the birth cohort for the entire province over several years. This allowed a robust examination of the effects of hyperbilirubinemia in a population with many caregivers and interpretations of the practice guidelines in use during that time period. Another advantage of this study method is the standardized format in which the information was collected and reported.
The implementation of the 1994 practice guidelines in Nova Scotia followed a period of intensive education; however, formal implementation of the guidelines on a provincial level was not undertaken. A review of practice patterns reported an increase in the diagnosis of hyperbilirubinemia but a concomitant decrease in the use of phototherapy, as would be expected on the basis of the guidelines.11 The current study confirms the increase in both moderate and more-extreme hyperbilirubinemia during the study period. Although a definition of severe hyperbilirubinemia based on a cutoff value of ≥325 μmol/L (≥19 mg/dL) is considered conservative by some, it was chosen to reflect the clinical practice approach to hyperbilirubinemia in the province during the study period.
No cases of kernicterus were diagnosed during the study period; however, only 7% of the infants in the extreme hyperbilirubinemia group had TSB levels between 400 and 499 μmol/L, and <1% had levels of >500 μmol/L. The “softer” neurologic adverse effects previously associated with hyperbilirubinemia were not found in this study, with the exception of increases in the rates of diagnoses of developmental delay and attention-deficit disorder, both of which remained significant even when adjustment was made for confounders. Although it is not clear that the demonstrated association is causal, concerns remain that, even when a conservative definition is used, hyperbilirubinemia may have significant long-term effects. Although it is not life-threatening, the impact of attention-deficit disorder on the social and educational functioning of children can be considerable and the financial impact on society significant. Finally, although the CI for the association between hyperbilirubinemia and autism included the null value, the point estimate suggests that exposure to high levels of bilirubin in infancy may have a role in the development of this ever more commonly diagnosed condition, and this association requires closer examination.
There are several limitations to this study. First, the population of Nova Scotia is predominantly white and may not be representative of other populations that are more ethnically and racially diverse. Second, there might have been other factors (eg, socioeconomic status and race), for which information was not available in the perinatal database, that were confounders of the outcomes of interest. Third, some infants in the cohort might have had hyperbilirubinemia that was unrecognized and did not lead to investigation. Such infants would have been misclassified as part of the no-hyperbilirubinemia group. Infants also might have been misclassified if the testing for hyperbilirubinemia was performed late or in the community and the results were not available in the hospital chart. Both of these circumstances likely would have been rare but might have occurred; the result would have been to underestimate the impact of hyperbilirubinemia on the outcomes of interest. In addition, we cannot rule out a bias in diagnosis for some of the study outcomes, because we cannot be assured that the practitioners were blinded with respect to neonatal course and bilirubin levels. Furthermore, although the sample was population-based and of sufficient size to have the power to detect statistically significant differences for the moderate hyperbilirubinemia group, differences in the severe hyperbilirubinemia group might not have been found because the differences were not large enough and there were few patients in the group. Another limitation in this cohort analysis was the loss to follow-up monitoring of the 4010 infants for whom record linkage was not possible. Although the proportion of infants excluded from the analysis because of nonlinkage was relatively small (6.7%), these infants differed from the infants included in the analysis in that they were slightly less likely to have hyperbilirubinemia and were more likely to have been delivered to a younger smoking mother. Finally, although almost the entire cohort (97.4%) had ≥2 complete years of follow-up data, the proportions with ≥3 or ≥4 complete years of follow-up data (84.5% and 71.2%, respectively) were smaller; therefore, some of the study outcomes that are typically diagnosed after 2 years of age (such as attention-deficit disorder, autism spectrum disorder, and developmental delay) might be underrepresented among the more-recent members of the cohort.
Hyperbilirubinemia in term infants remains an area that requires attention and evaluation. It is reassuring that the “classic” adverse effects associated with high bilirubin levels were not seen in this large cohort; however, there is concern because associations with developmental delay, attention-deficit disorder, and autism were observed. These associations require additional prospective study to ascertain whether a causal relationship exists. In addition, more study of the characteristics, including genetic markers, of the infants who develop short- and long-term complications would be helpful for better identification of infants who would benefit from more-intensive treatment.
We thank the Nova Scotia Health Research Foundation for providing funding. Dr Dodds was supported by a Clinical Research Scholar Award from Dalhousie University and by a New Investigator Award from the Canadian Institutes of Health Research.
We thank the Reproductive Care Program of Nova Scotia and the Population Health Research Unit at Dalhousie University for facilitating access to the data.
- Accepted September 19, 2007.
- Address correspondence to Krista A. Jangaard, MD, Department of Neonatal-Perinatal Medicine, 5850/5980 University Ave, PO Box 9700, Halifax, Nova Scotia, Canada B3K 6R8. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
Although this research is based in part on data obtained from the Population Health Research Unit, the observations and opinions expressed are those of the authors.
What's Known on This Subject
Studies have shown conflicting results regarding the severity and kinds of adverse outcomes that can be causally linked to hyperbilirubinemia. Treatment guidelines have been formulated to standardize approaches to care, but the impact of implementation has not been well studied.
What This Study Adds
This population-based study examines the associations between serum bilirubin levels and neurologic and developmental outcomes for >50 000 infants in the era after implementation of treatment guidelines in a Canadian province. Associations with milder forms of neurologic impairment are suggested.
- ↵Holtrop PC, Maisels MJ. Hyperbilirubinemia in Intensive Care of the Fetus and Neonate. Philadelphia, PA: Mosby-Year Book; 1996:888–898
- Scheidt PC, Graubard BI, Nelson KB, et al. Intelligence at six years in relation to neonatal bilirubin levels: follow-up of the National Institute of Child Health and Human Development Clinical Trial of Phototherapy. Pediatrics.1991;87 (6):797– 805
- ↵Scheidt PC, Bryla DA, Nelson KB, Hirtz DG, Hoffman HJ. Phototherapy for neonatal hyperbilirubinemia: six-year follow-up of the National Institute of Child Health and Human Development clinical trial. Pediatrics.1990;85 (4):455– 463
- ↵American Academy of Pediatrics, Provisional Committee for Quality Improvement and Subcommittee on Hyperbilirubinemia. Practice parameter: management of hyperbilirubinemia in the healthy term newborn. Pediatrics.1994;94 (4):558– 562
- ↵Joint Commission on Accreditation of Healthcare Organizations. Sentinel Event Alert. Oak Brook Terrace, IL: Joint Commission on Accreditation of Healthcare Organizations, Issue 18, April 1, 2001
- ↵Johnson L, Brown AK. A pilot registry for acute and chronic kernicterus in term and near-term infants. Pediatrics.1999;104 (suppl 3):736
- Maisels MJ, Newman TB. Kernicterus in otherwise healthy breast-fed term newborns. Pediatrics.1995;96 (4):730– 733
- ↵Harris MC, Bernbaum JC, Polin JR, Zimmerman R, Polin RA. Developmental follow-up of breastfed term and near-term infants with marked hyperbilirubinemia. Pediatrics.2001;107 (5):1075– 1080
- ↵Newman TB, Klebanoff MA. Neonatal hyperbilirubinemia and long-term outcome: another look at the Collaborative Perinatal Project. Pediatrics.1993;92 (5):651– 657
- ↵Boo NY, Oakes M, Lye MS, Said H. Risk factors associated with hearing loss in term neonates with hyperbilirubinemia. J Trop Pediatr.1994;40 (4):194– 197
- ↵Wong V, Chen WX, Wong KY. Short- and long-term outcome of severe neonatal nonhemolytic hyperbilirubinemia. J Child Neurol.2006;21 (4):309– 315
- ↵Juul-Dam N, Townsend J, Courchensne E. Prenatal, perinatal and neonatal factors in autism, pervasive developmental disorder-not otherwise specified, and the general population. Pediatrics.2001;107 (4). Available at: www.pediatrics.org/cgi/content/full/107/4/e63
- ↵Croen LA, Yoshida CK, Odouli R, Newman TB. Neonatal hyperbilirubinemia and risk of autism spectrum disorders. Pediatrics.2005;115 (2). Available at: www.pediatrics.org/cgi/content/full/115/2/e135
- ↵Sgro M, Campbell D, Shah V. Incidence and causes of severe neonatal hyperbilirubinemia in Canada. CMAJ.2006;175 (6):587– 590
- Copyright © 2008 by the American Academy of Pediatrics