Objective To examine whether neonatal non-hemolytic hyperbilirubinemia is associated with adult neuropsychiatric disability and cognitive function.
Methods The study included all men born as singletons ≥35 gestational weeks in two Danish counties from 1 January1977 to 31 December 1983 that registered at conscription in a Danish region. Their infant levels of hyperbilirubinemia was ascertained from hospital records. At conscription, the prevalence of neurologic conditions and performance on a standard group intelligence test (Boerge Prien test) was compared between men with and without neonatal non-hemolytic hyperbilirubinemia.
Results The study group consisted of 463 conscripts exposed to neonatal non-hemolytic hyperbilirubinemia and 12 718 unexposed conscripts. The median value of maximum serum bilirubin concentration was 256 µmol/l (range 105–482). Among the exposed, 5.6% were deemed unfit for military service due to a neurologic or a psychiatric condition, compared with 4.8% among the unexposed (prevalence ratio 1.18, 95% CI 0.81 to 1.73). Among men with Boerge Prien measurement, mean Boerge Prien test score among 391 exposed men was 42.4 points compared with 43.4 points among 11 248 unexposed men (mean difference 1.0 points, 95% CI 0.0 to 1.9). There was no association between level of hyperbilirubinemia and cognitive score. Adjusted prevalence ratio of obtaining a Boerge Prien test score in the lowest quartile was 1.04 (95% CI 0.87 to 1.23).
Conclusion The study found no evidence of an association between neonatal non-hemolytic hyperbilirubinemia and adult neurodevelopment and cognitive performance in male conscripts. Since cognitive performance was not associated with the severity of hyperbilirubinemia we ascribe the slightly lower cognitive scores among exposed to uncontrolled confounding.
Statistics from Altmetric.com
Most infants develop jaundice during their first days of life, with greater prevalence at shorter gestational ages. Occasionally infants develop severe hyperbilirubinemia, which in rare cases may progress to advanced bilirubin encephalopathy, a devastating condition that may cause serious neurodevelopmental impairment (kernicterus).
Cognitive outcome in full-term newborns with non-hemolytic hyperbilirubinemia without kernicterus has been investigated in several studies, but the conclusions have been conflicting.1,–,15 The generalisability of many studies has been limited by the inclusion of preterm infants and/or infants with hemolytic hyperbilirubinemia,1,–,7 9 or by a small number of subjects.8 9 Only two studies had a follow-up of more than 13 years.9 10
What is already known on this topic
▶. Neonatal hyperbilirubinemia may progress to chronic bilirubin encephalopathy (kernicterus).
▶. Published data on long-term cognitive consequences of neonatal hyperbilirubinemia are conflicting.
What this study adds
▶. The level of neonatal non-hemolytic hyperbilirubinemia was not associated with cognitive performance at conscription for military service.
▶. There was no association between recorded neuropsychiatric disability and neonatal non-hemolytic hyperbilirubinemia.
Seidman et al10 found an increased risk of obtaining an IQ <85 (one SD below the mean) at age 17 years among boys born at term and exposed to a maximum total serum bilirubin (TSB-max) level >340 µmol/l. This study has raised concern about a potential negative long-term effect of severe hyperbilirubinemia on cognitive performance.
To address this issue, we examined the prevalence of psychiatric and neurologic diagnoses identified at time of army conscription in a group of men exposed to neonatal non-hemolytic hyperbilirubinemia and compared it with corresponding prevalence among unexposed men. Cognitive performance was compared in the two groups and assessed according to level of hyperbilirubinemia.
We conducted this study in the Fifth Military Conscription District of Denmark, comprising Viborg and North Jutland counties, populated by approximately 700 000 inhabitants. As our study base we identified all men who were born as singletons between 1 January 1977 and 31 December 1983 and registered in the Conscript District at the age of 18 years.
Inclusion and exclusion criteria and definition of bilirubin exposure
From this base we identified all live born singleton infants with a diagnosis of non-hemolytic hyperbilirubinemia (Eighth Revision of International Classification of Diseases (ICD-8), code 778.91) recorded in the National Registry of Patients, which stores data on all hospital discharges in Denmark.16 The diagnostic criteria for non-hemolytic hyperbilirubinemia were identical for all hospitals and defined by The Danish Paediatric Society. We retrieved the hospital records and extracted data on TSB-max (µmol/l); age at time of TSB-max (days); number of bilirubin measurements; phototherapy (yes/no); exchange transfusion (yes/no); and maternal diseases during pregnancy. TSB was measured either on heel capillary blood by the Jendrassik–Grof–Nosslin method17 or by the direct spectrometric method.18 TSB was determined instead of unconjugated bilirubin, as measurement of TSB is more precise; furthermore, the concentration of conjugated bilirubin is negligible in infants with uncomplicated hyperbilirubinemia.19
We excluded infants born with gestational age <35 weeks; age at time of TSB-max >14 days; hemolytic diseases due to Rhesus blood type isoimmunisation (ICD-8: 775.09), congenital sphaerocytosis, or ABO blood type incompatibility (ICD-8: 775.19) if the first haemoglobin value was <8.0 mmol/l. We also excluded infants with congenital hypothyreosis and those with maternal diabetes mellitus.20 From 1977 to 1983 there have been no recordings in Denmark of encephalopathy due to hyperbilirubinemia. Men presenting for conscription who had no record of neonatal hyperbilirubinemia in the National Registry of Patients constituted the reference group for all comparisons.
Outcomes measured at conscription
All Danish men are required to register with the military Draft Board between the ages of 18 and 20 years. At the time of conscription all young men complete a health questionnaire, in which they can report chronic health problems that could preclude military service. Men deemed ineligible for military service due to severe chronic diseases become exempt from the Draft Board's evaluation (rejected). Although these men (‘non-presenters’) do not present for examination or cognitive testing, diagnoses leading to rejection are recorded. The draft suitability of the remaining men (‘presenters’) is determined by a routine Draft Board evaluation consisting of a medical examination and an intelligence test. Diagnoses made at that examination are also recorded.
All presenters take the Boerge Prien test, a 45-min validated group intelligence test developed for the Danish Draft Board and used since 1957 in unmodified form. The final score is highly correlated with the verbal IQ (0.78), performance IQ (0.71) and full-scale IQ (0.82) on the Wechsler adult intelligence scale.21 In a validation study, the mean full-scale IQ was 105.8 and the mean Boerge Prien score was 44.2.22
Following the Draft Board examination, all draftees are categorised as fit, conditionally fit or unfit for military service.
Outcomes defined in the study
Among non-presenters and presenters we defined ‘neuropsychiatric diagnosis’ as a record with one of the following ICD-10 codes: F00–F99 (psychiatric diseases) or G10–G99 (neurologic diseases apart from meningitis) in the conscription file.
Among presenters we examined the distribution of the Boerge Prien test scores and defined two additional outcomes: (A) Boerge Prien scores in the bottom quartile of all observed scores (‘cognitive score 25%’) and (B) scores more than 1SD below the mean in the background population (‘cognitive score <1 SD’).
Data on maternal and perinatal covariates
Data on gestational age; birth weight; mode of delivery; 5-min Apgar score; maternal age, parity and civil status were collected from the Danish Medical Birth Registry,23 which tracks all births in Denmark since 1977.
Linkage of data sources
All Danes receive a 10-digit personal registration number at birth permitting unambiguous data linkage.24
We first examined the relation of neonatal hyperbilirubinemia (yes/no) with prevalence of army rejection (unfit vs fit or conditionally fit) for any cause and then of rejection combined with a neuropsychiatric diagnosis. Using the lowess procedure, we fitted a curve to display the relation between the level of bilirubin and Boerge Prien scores.25 We then examined the association between Boerge Prien score and exposure to hyperbilirubinemia, the latter used as a dichotomous (yes/no) and as categorical variable (≤239; 240–289; 290–339 and ≥340 µmol/l). We used 240 and 340 µmol/l as they are the recommended thresholds for phototherapy and for exchange transfusions. In the hyperbilirubinemia categories we compared the mean, median and IQRs of Boerge Prien scores as well as the prevalences of the outcomes ‘cognitive score 25%’ and ‘cognitive score <1 SD’.
We used linear regression to estimate mean difference in Boerge Prien score and log-binomial regression to estimate prevalence ratios (PRs) for the dichotomous outcomes.26 We report crude estimates and estimates adjusted for gestational age, gestational-age-adjusted birth weight, maternal age, parity and civil status. The latter three variables were included as indicators for socioeconomic status of the mother. In order to enable comparison with other studies, we transformed the Boerge Prien scores to conventional IQ scale, with mean 100 and SD 15, and estimated mean difference in IQ between the exposed and the unexposed by linear regression. We analysed data using SAS software V.9.1.
The study was approved by the Danish Registry Board. An informed consent was not required for this study of routine records.
We identified 13 370 singleton men born in the two counties between 1977 and 1983 with a subsequent registration in the Fifth Conscription District. Among those, 652 had a diagnosis of neonatal non-hemolytic hyperbilirubinemia in the National Registry of Patients. Hospital records were traced and reviewed in 506 cases, of which 43 either did not satisfy our eligibility criteria or did not have valid data on bilirubin exposure or on fitness at conscription. Overall, 13 181 men were included in the study; the exposed group included in the analyses thus consisted of 463 men with documented exposure to various levels of bilirubin within the first two weeks of life and 12 718 unexposed controls. Median age at conscription was 18.8 years (quartiles: 18.6 and 19.6 years).
Median TSB-max value was 256 µmol/l (quartiles: 221 and 293, range 105–482) occurring on the fifth (median) day of life (quartiles: fourth to sixth day, range 1–14 days). Median number of measurements was 7 (range 1–16). Phototherapy was instituted in 66% of the infants, one infant was treated with exchange transfusion.
Army fitness and neuropsychiatric disability
Of the 13 181 men, 13 (0.1%) had a missing entry on army fitness; 4742 (36.0%) men were deemed unfit either before or after the Draft Board evaluation. Overall, 1167 (8.8%) men were recorded with a psychiatric (n=901) or neurologic (n=276) diagnosis in the draft file; and 630 (4.8%) men were deemed unfit with one of these diagnoses. Among exposed the PR of being declared unfit for military service was 1.13 (95% CI 1.01 to 1.27); the PR of obtaining a neuropsychiatric diagnosis was 1.08 (95% CI 0.81 to 1.43); and the PR of the combination of these two outcomes was 1.18 (95% CI 0.81 to 1.73) compared with unexposed. The associations did not change materially when psychiatric and neurologic diagnoses were used as separate outcomes. There was no significant difference in categories of diagnoses leading to rejection among presenters and non-presenters were similar (data not shown).
Among 11 680 participating in the Draft Board evaluation including the Boerge Prien test, we excluded 41 records (0.35%) with Boerge Prien scores below 15 (deemed to be invalid), leaving 11 639 men for the analysis of cognitive performance (see table 1). Perinatal characteristics of the conscripts with available intelligence scores according to exposure to neonatal hyperbilirubinemia are shown in table 2. The mean difference in Boerge Prien score was 1.0 point (95% CI 0.0 to 1.9 points), but as seen in table 3 there was no evidence of differences in cognitive score between the four categories of bilirubin exposure, and likewise the lowess plot did not show any association between level of bilirubin and cognitive score (figure 1). We found a crude PR of 1.09 (95% CI 0.92 to 1.28) of obtaining a ‘cognitive score 25%’ among those exposed to hyperbilirubinemia compared with the unexposed. After adjustment for potential confounders this PR decreased to 1.04 (95% CI 0.87 to 1.23). There was no association between the PRs of obtaining an IQ in the lower 25th centile and the serum bilirubin level (table 4). On the conventional IQ scale, any exposure to bilirubin was associated with an adjusted mean difference of 1.4 points (95% CI −0.1 to 2.9) compared with absence of exposure. The pattern of association did not change when ‘cognitive score <1 SD’ was used as the outcome, nor were the associations affected by combining ‘unfit’ and ‘conditionally fit’ men in a single category (data not shown).
Neonatal exposure to non-hemolytic hyperbilirubinemia was not associated with an increased risk of obtaining a neuropsychiatric diagnosis at time of conscription. Cognitive scores were slightly lower among exposed, but there was no evidence of a dose–response association regardless of which exposure or outcome measure was used. Therefore the slight association is likely to be produced by uncontrolled confounding.
The main strengths of our study are the population-based design, prospective recording of outcome relative to exposure, the documented magnitude and timing of bilirubin exposure, the relatively large sample with an almost complete follow-up into young adulthood regarding neuropsychiatric conditions and army fitness, and comparison with a concurrent, population-based control group.
However, several factors could affect the validity of our estimates. Differences in reasons for rejection between presenters and non-presenters could introduce selection bias, but the observation of identical proportions of rejection due to neuropsychiatric conditions in exposed and unexposed reduces this concern. Some draftees may intentionally have performed worse on the test than would be expected according to their IQ, but such behaviour is unlikely to be related to neonatal hyperbilirubinemia. The only unintended exclusions of boys in the original birth cohort were those who either died or emigrated out of Denmark before becoming draft-liable. This situation could have caused selection bias in our study if occurring differentially among those exposed and unexposed to non-hemolytic hyperbilirubinemia. Our results apply only to men surviving to conscription.
Registry-based data lack detail, such as information about several factors that could confound the observed associations, for example, parental education and socioeconomic status, maternal smoking, and duration of breastfeeding, which have been shown to be associated with cognitive performance.27 28 The absolute difference of 4.8% in overall rejection rates may indicate higher general morbidity in infants with neonatal hyperbilirubinemia. It would therefore be of interest to have more detailed information about morbidity in this group. Owing to the Danish policy of instituting phototherapy and exchange transfusions if bilirubin values exceed, respectively, 240 and 340 µmol/l, only 25 newborns in our data were in the highest exposure category. However, as we found little evidence of an association between hyperbilirubinemia and adult cognitive function, the low number does not invalidate our conclusion.
Seidman et al10 examined IQ scores at 17 years of age in a cohort of conscripts of which 452 had been exposed to at least moderate hyperbilirubinemia. Contrary to our results, they found a statistically significant negative association between TSB >340 µmol/l and an IQ score <85. One possible reason for the different findings could be the exclusion criteria of our study, yielding a sample with fewer newborns with maternal comorbid diseases that may adversely affect cognitive performance. Similarly to our study, Nilsen et al9 found no association of bilirubin concentration with either army rejection rate or IQ in a cohort of 55 males.
Several studies have used shorter observational periods and assessed various outcomes when looking for a relation between hyperbilirubinemia in neonates and IQ.1,–,8 11,–,13 Some of the studies were conducted within fractions of the Perinatal Collaborative Project comprising 56 990 pregnant women.1,–,7 The findings have not been uniform, but most often no association was found. A possible reason for divergent observations may be inclusion by many studies of preterm infants and/or infants with Rhesus isoimmunisation.1,–,7 Four studies have restricted the analyses to full-term newborns with non-hemolytic hyperbilirubinemia.8 11,–,13 The largest of these studies comprised 21 324 newborns,12 the others varied from 60 to 226.8 11 13 Results from these studies did not show an association between hyperbilirubinemia and IQ recorded at the age of 2–13 years. Recently Jangaard et al15 examined neurologic outcome at the age of 2–9 years in 56 019 infants with gestational age ≥35 weeks according to exposure to non-hemolytic hyperbilirubinemia. The only positive association was a higher risk of developmental delay in infants with TSB-max 230–324 μmol/l compared with unexposed.
We found no evidence of increased risk of neuropsychiatric morbidity or worse cognitive performance among male conscripts exposed to hyperbilirubinemia as neonates.
The authors thank Dr Peer Fisher and Dr Jørgen Lindhardt at the Fifth Military Conscription District for essential collaboration in collecting data from the Draft Board evaluation and Professor Henrik Toft Sørensen for valuable comments on the manuscript.
Funding The study received grants from the Western Danish Research Forum for Health Sciences, Department of Clinical Epidemiology's Research Foundation and Hertha Christensens Fond. The funding source did not affect any aspect of the study design, analysis or interpretation of findings.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.