Bilirubin Screening for Normal Newborns: A Critique of the Hour-Specific Bilirubin Nomogram

Kernicterus is a devastating but rare disease with an incidence that ranges from 0.4 to 2.9 per 100000 live births.^1–6 Recently, a preventive strategy of measuring predischarge bilirubin levels in all infants discharged from the newborn nursery was proposed and included as an option in the American Academy of Pediatrics guidelines.⁷ The percentile location of these levels on an hour-specific “nomogram” of total serum bilirubin (TSB) levels first described by Bhutani et al⁸ is thought to predict the risk of severe hyperbilirubinemia. In this commentary we highlight the methodologic flaws in the development of the nomogram, the lack of evidence for efficacy of screening, and the potential undesirable consequences of such an approach. Multiple flaws in the methods used to create the nomogram⁸ and assess its accuracy generate serious questions about its validity. Some of these issues were emphasized previously by experts on hyperbilirubinemia who stated that “the study biases and interlaboratory variability in TSB measurements preclude general extrapolation of the quantitative results of this study… ”⁹

First, the applicability of these percentile curves to the overall US or international population is suspect, because they were developed in a retrospective study that used a small sample of infants from a single urban Philadelphia, Pennsylvania, hospital, the demographics of which (43% white, 41% black, 4% Hispanic, and 4% Asian) do not represent US demographics.^10–13

Second, the estimation of test characteristics (sensitivity, specificity, and positive and negative predictive values) in the original study⁸ is flawed. To accurately estimate such characteristics, all patients undergoing a diagnostic test should be followed adequately and all subjected to a gold-standard confirmatory test regardless of whether the diagnostic test is positive or negative. In this study, the presence or absence of hyperbilirubinemia after discharge was not ascertained in 10163 (78%) of the 13003 infants undergoing predischarge bilirubin testing. Only infants who returned to the hospital for follow-up had bilirubin levels checked (which occurred at the clinicians' discretion). Furthermore, there was not an adequate explanation of why the majority of infants were not analyzed. Thus, the curves were generated from only 22% of the infants enrolled in the study, which severely limits the generalizability of the results.

Third, the outcome targeted for prediction, a postdischarge bilirubin level at ≥95th percentile (termed “significant hyperbilirubinemia” in the article), is of uncertain clinical significance, and its relationship to kernicterus is unknown.

Fourth, the sample from which the percentile curves and the predictive criteria were derived was the same. Ideally, a predictive tool should be developed in 1 sample and validated in another.¹⁴

Even if the above-listed methodologic problems that undermine the study's validity are ignored, there is no evidence that the use of the nomogram in a screening program prevents kernicterus. The optimal study design for proving the efficacy of a screening test is a randomized trial.¹⁵ Studies that have reported the use of predischarge bilirubin screening were not randomized trials, used a variety of cointerventions in addition to screening, and in place of kernicterus used a surrogate outcome (severe hyperbilirubinemia, which does not always lead to kernicterus). Therefore, the use of the nomogram as part of a screening program does not satisfy World Health Organization criteria for screening tests.¹⁶

Finally, several undesirable consequences may occur from the use of nomogram-based risk screening. Infants designated at low risk on the nomogram may, in reality, be at high risk of kernicterus or severe hyperbilirubinemia. Such false-negative results have been described,^17–20 but their exact frequency is unknown. In a registry of 35 to 37 weeks' gestation infants with kernicterus, 1 (6.3%) of 16 infants whose bilirubin levels were measured during the first 4 days of life had a measurement below the 40th percentile (low-risk zone) on the Bhutani et al nomogram, and 2 (12.5%) had early bilirubin levels between the 40th and 75th percentiles (low-intermediate–risk zone).¹⁷ Two studies of infants readmitted for hyperbilirubinemia showed that 2.7% to 3.6% of the newborns were in the low-risk zone, and 13.5% to 43% were in the low-intermediate–risk zone, demonstrating a significant risk of false-negative results.^19,20 In addition, kernicterus has occurred in infants with TSB levels ranging from 4 to 54 mg/dL.²¹

Clinicians who are falsely reassured by such false-negative results may follow newborns inadequately after discharge, thereby delaying the detection and treatment of significant disease. We have also encountered many clinicians who believe that the Bhutani et al nomogram predicts the risk of kernicterus itself and many who use it to decide when to initiate phototherapy. Such misuse may arise because the Bhutani et al nomogram is unclear and prone to misinterpretation.

In this issue of Pediatrics, Anastasia et al²² describe a new transcutaneous bilirubin nomogram. Although their study had a lower attrition rate before hospital discharge at 72 to 96 hours of life, its flaws are similar to those of the Bhutani et al study: a highly selective patient sample (white infants from a single Greek hospital, of whom 45% were formula fed and 38% were born by cesarean delivery), selective follow-up after discharge, and lack of validation in an independent patient sample. The gold standard used to estimate test characteristics was the American Academy of Pediatrics phototherapy guideline, which is based solely on expert opinion. Therefore, this nomogram's clinical significance is unknown, as is its tendency toward misinterpretation.

The costs of a universal bilirubin screening program are high, ranging from $4 million to $78 million to prevent 1 case of kernicterus, if efficacy of screening is assumed.²³ Any prevention program this expensive requires high-level evidence to justify widespread implementation. Widespread screening without proof of efficacy can result not only in wasted health care resources but also in unnecessary testing of numerous newborns, “medicalization” of normal newborns, and adverse effects on the parents, including anxiety and a negative perception of the child's overall care.^24,25

Kernicterus is a serious complication of hyperbilirubinemia that should be prevented but, ideally, with interventions supported by high-level evidence from rigorous research. Until such evidence is available, methods that might be ineffective at best and harmful at worst should be avoided, and universal newborn follow-up and evaluation should be emphasized. The use of the hour-specific bilirubin nomogram, although well intentioned, not only is unjustified as a method of risk screening but also potentially encourages inadequate follow-up of newborns and exposes them to unnecessary risk.

• Copyright © 2009 by the American Academy of Pediatrics