Questioning Our Approach to Hyperbilirubinemia

Kernicterus is a devastating disease demanding vigilance in our attempts to prevent its occurrence. The American Academy of Pediatrics (AAP) released clinical practice guidelines for the management of hyperbilirubinemia and prevention of kernicterus in 2004, which demonstrated clear and logical rationale based on limited data and information available at the time.¹  Since this guideline’s release >15 years ago, newer data suggest kernicterus occurs at bilirubin levels much higher than previously thought and challenges the current standards of practice.^2–5  AAP clinical practice guidelines expire after 5 years unless otherwise reaffirmed, updated, or removed. These guidelines were updated in 2009 with minimal changes and, despite expiring in 2014, the management approach laid out by these guidelines is still considered standard of care by most pediatricians.⁶  However, continuing to follow them ignores new information and results in substantial overtreatment and unnecessary harm to infants and families.^2–5,7  In this Perspectives piece, we will discuss the logistical framework of the 2004 AAP clinical guidelines on hyperbilirubinemia and its persistence in clinical practice despite evidence from large data sets that suggest that a change in practice is long overdue.

First, we must examine how the 2004 committee decided on the bilirubin levels that warranted treatment. The committee’s rationale for the thresholds for both phototherapy and exchange transfusion curves followed 2 guiding principles:

1. to begin exchange transfusions at or below levels at which cases of kernicterus had been reported; and

2. to begin phototherapy at a level where 5 to 10 infants would be treated to prevent 1 exchange transfusion.¹ 

Thus, the phototherapy treatment curve is inherently dependent on the exchange transfusion curve. Exchange transfusions are rare procedures but contain a significant mortality risk and should only be used as a last resort. Some studies have reported a mortality rate as high as 2% directly because of the procedure.⁸  At the time of their creation, the guidelines appropriately reflected a conservative approach to preventing both kernicterus and the harm of an exchange transfusion.

However, the guidelines were developed before electronic medical record systems were widely available, so the incidence of kernicterus within large health systems was difficult to determine. Without good population level data, the committee based their recommendations for the bilirubin level warranting treatment with an exchange transfusion on 123 cases of kernicterus reported from 14 countries from 1955 to 2002.⁹  These data were obtained solely from case reports and there are many concerns about this data set when viewed from today’s vantage point. Follow-up information is lacking on many of these cases, the definition of kernicterus was broad and included diagnoses based on only acute clinical findings, some of the peak bilirubin levels were obtained >24 hours after diagnosis of kernicterus, and there was no standardization of laboratory practices.⁹ 

With the addition of robust electronic medical records, we now have access to data from large populations of newborns with presumably improved laboratory quality. Reports from Kaiser Permanente, state databases from Utah, and national databases in Denmark and Sweden provide data on almost 3 million infants.^2,3,5,10  Within this collection of databases, a total of 34 infants were diagnosed with kernicterus and 32 of those infants had a peak bilirubin level >30 mg/dL. The study from the Kaiser Permanente system reported on 525 000 infants, of which 1833 infants had levels above the exchange transfusion level, including 47 with levels above 30 mg/dL. Only 4 of these infants were diagnosed with kernicterus and their peak bilirubin values ranged from 38.2 mg/dL to 48.5 mg/dL.² 

The broad swath of high-quality data produced since 2004 suggests that the foundation with which the exchange transfusion guideline was built, the bilirubin concentration at which infants develop kernicterus, is set at too low a level. Across the almost 3 million patients from the databases listed above, there was 1 infant >36 weeks’ gestation who developed bilateral hearing loss at a level of 26.8 mg/dL and 1 infant at 35 weeks’ gestation with a level of 28.5 mg/dL who lacked long-term follow-up data.^5,10  The rest were above 30 mg/dL, with most well above that level.^2,3,5,10  However, based on the exchange transfusion curves, a full-term infant with hemolysis would exceed the exchange transfusion level at 22 mg/dL at 3 days of life.¹  Performing a procedure with potentially catastrophic side effects at levels well below those associated with serious pathology is not good medicine.

The second guiding principle of the committee was to develop phototherapy thresholds that would require the treatment of 5 to 10 infants with bilirubin levels between 15 and 20 mg/dL to prevent 1 infant from reaching 20 mg/dL (the exchange transfusion threshold).¹  The level of 20 is likely far too low to pursue exchange transfusion, but even if we accept this cutoff, the number needed to treat (NNT) was vastly underestimated. Just 5 years after the guidelines were published, a study from the Northern California Kaiser health system presented results from >22 000 infants with bilirubin levels within 3 mg/dL of phototherapy treatment level. The study looked at the actual NNT to prevent 1 infant from reaching exchange transfusion levels and found variability based on sex and gestational age and, overall, found NNTs that were vastly higher than those estimated by the AAP committee. A 3-day, 41-week–old female did not have an NNT of 5 to10 as predicted by the committee, but rather an NNT of 3041.⁷  Both treatment recommendations are presented in the form of a curve with increasing thresholds over time. There is no evidence or reasonable physiologic expectation for why a 5-day-old is better equipped to handle high bilirubin levels than a 3-day-old, yet the difference in recommended phototherapy levels among those 2 ages is 6 mg/dL.¹  The treatment curves mirror the Bhutani curves, which were developed through observational data of 2840 American infants with an aim to better define the risk that an infant would develop severe hyperbilirubinemia, defined as levels >95% for age.¹¹  However, the goal of treatment of jaundice is not to treat 5% of all infants, it is to prevent levels from reaching dangerous heights that may lead to kernicterus. Equating these 2 is not logically sound and results in treatment at levels that are far too low.

Given new evidence obtained as of 2009, continuing to use the phototherapy curves from the 2004 AAP practice guidelines in present times no longer upholds the principles from which these curves were initially created (Table 1).¹  Though the threshold for an appropriate NNT is obviously nuanced and subject to the perspective of individual patients and families, treating 3000 infants to prevent 1 exchange transfusion is inherently tied to the idea that phototherapy is “harmless” and therefore should be used in all cases where it has potential benefit. Unfortunately, that may not be the case. Studies from the Northern California Kaiser system have showed a small absolute increase in infantile cancers and in epilepsy for those infants who have received phototherapy.^12,13  In addition, treatment with phototherapy often requires a stressful readmission to the hospital, the separation of the baby and family, unnecessary and traumatic blood draws, potential disruption of breastfeeding, and a sense from the family that their newborn is sick in some way.^14–18 

With a long list of potential harms associated with treatment, the framework of how we consider infants with hyperbilirubinemia must be further investigated. Perhaps hyperbilirubinemia, in its less extreme forms, should not be considered a disease at all. Compared with adult levels, all babies have hyperbilirubinemia and levels are generally driven higher by exclusive breastfeeding, which, until the widespread use of formula 150 years ago, was essentially how all babies were fed.¹⁹  Something that every baby has (hyperbilirubinemia), exacerbated by something that most babies do (breastfeeding), is unlikely to be a disease but rather a potential evolutionary benefit. In fact, bilirubin is a potent antioxidant and is important in helping to mitigate the oxidative stresses that newborns experience.²⁰  It is a natural process that can, in rare circumstances, reach levels that can be harmful. Those levels constitute a major disease, but there is no logic to suggest that levels in the teens constitute a minor disease. The AAP practice guidelines state that implementing treatment should be done on the basis of estimates that the benefits to treatment exceed the risks.¹  The benefits to treating infants with bilirubin levels in the teens, as we have been doing, has not been demonstrated and may not exist, but the harms are clear.

As pediatricians, we have been massively overtreating hyperbilirubinemia in newborns. There are ∼4 million births each year in the United States. It is reasonable to estimate that 200 000 of those babies will be above the 95th percentile levels for bilirubin. Only a fraction of those are actually at risk for kernicterus, whereas the rest are only at risk for the harms that come with overtreatment. We have had the information to know that our approach has not been good medicine for over a decade, yet the standard has remained to use the treatment curves from these long-expired guidelines, even though a careful reading of new data would have told us to adjust. Guidelines from a panel of experts are a valuable tool but can have a chilling effect on the pace of change until the next set of guidelines is published. These guidelines no longer balance the large catastrophe of inaction (kernicterus) with the sum of thousands of smaller catastrophes of action (overtreatment), and haven’t for over a decade. As a profession, we must learn from this failure and strive to strike a more even balance. Of course, new guidelines are long overdue, but without internalizing the larger lesson, we risk falling into the same trap of overreliance on outdated guidelines. Guidelines can guide us astray if we do not, as a profession, remain vigilant and nimble in our approach.