Bilirubin measurements in neonates: uniform neonatal treatment can only be achieved by improved standardization

Discussion

The nomenclature of indirect and direct bilirubin is based on the chemically reactive behavior in the diazo reaction, even if direct bilirubin is often incorrectly equated only with structurally defined conjugated bilirubin. Assays for Tbil measure fractions that react both directly and indirectly in the diazo reaction. Commercially available bilirubin assays for automated clinical chemistry analyzers are usually available for Tbil and direct bilirubin. The majority of automated clinical chemical analyzers use methods for the determination of bilirubin using diazo reagents. Bilirubin assays based on the azo-coupling principle are unable to distinguish conjugated or delta-bilirubin from other bilirubin species that react directly in the assay. Bilirubin assays using the vanadate oxidation method are also offered as total and direct bilirubin assays. The specificity for direct bilirubin or for TBil depends on the presence or omittance of detergent in the reagent formulation. The vanadate method does also potentially detect “all” bilirubin species [18, 19], however these assays are usually not designed to distinguish the individual bilirubin species.

A third option for routine bilirubin determination is multi-layer film technology where slides for TBil and specific slides for the differentiation of unconjugated and conjugated (not “direct”) bilirubin are available. While the TBil slide follows the diazo technique, the second slide (BuBc) uses spectrophotometry to differentiate unconjugated bilirubin from the sum of mono- and di-conjugated bilirubin by the application of different layers and an algorithm calculating the individual concentrations from the measured signals. Due to the nature of the layers, delta bilirubin cannot be detected with the BuBc slide [16, 20].

In general, the difference between the measurement of mono- and di-glucuronide from the TBil provides an indirect measurement of the delta bilirubin fraction. Additional structural and configurational bilirubin isomers may also be formed quickly under phototherapy but are not routinely measured [21].

Currently no analyzers directly measure the delta bilirubin form or the more polar bilirubin isomers formed during PT, however these have been assessed by HPLC methods [22, 23]. There are different clinical impacts from the differing forms of bilirubin, for example unconjugated hyperbilirubinemia and photo-isomers are relevant to bilirubin-induced neurological damage (kernicterus spectrum disorders) whereas conjugated forms are relevant to specific diseases including biliary atresia [24, 25]. Thus, different measurands would ideally be considered for different intended uses. Certainly, the specific measurands being determined by different methods may warrant further consideration. In the case of blood gas analyzers, it is likely that the analytes being actually measured may be different compared to traditional clinical analyzers and this will be of even greater complexity for hand-held near patient testing instruments. Thus, the effective linkage of measurements across these different platforms may be challenging.

Discussion

National Institute of Standards and Technology (NIST) Standard Reference Material^® (SRM) 916b bilirubin is a certified reference material consisting of neat unconjugated bilirubin with a certified mass fraction purity of (94.7±0.7) %. The mass purity determination was made by quantitative ¹H nuclear magnetic resonance spectroscopy using an internal standard (q¹H-NMR_IS) and is metrologically traceable to the International System of Units (SI). Comparison of qNMR results to diazo spectrophotometric reference measurement procedure (RMP) results demonstrated that the methods are consistent and molar absorptivity values calculated from the spectrophotometric RMP provide evidence for the continuity of the reference system initially established for bilirubin using NIST SRM 916 and NIST SRM 916a. The bilirubin IX-α isomer is predominant in NIST SRM 916b. Bilirubin III-α content is estimated to be 3 % (g/g, mol/kg) greater than that of the XIII-α isomer, which is consistent with the isomer composition observed for SRM 916a. Organic impurities structurally similar to bilirubin, including biliverdin or mesobilirubin, were not identified. Observations from ¹H(¹³C)-NMR spectra suggested that the NIST SRM 916b material contains a greater quantity of paramagnetic impurities vs. NIST SRM 916. The total content of paramagnetic and ferromagnetic elements is approximately 0.1 % (g/g). The total mass fraction of all elements readily attributable to impurity components, determined by semiquantitative ICP-MS and X-ray fluorescence spectroscopy methods, is approximately 0.6 %. Combustion analyses and elemental microanalysis suggest that the unidentified impurities are largely of organic nature and may include macromolecular species or larger organic particles not observable via ¹H-NMR [27].

While the previous batch of NIST SRM 916a bilirubin was listed in the JCTLM database, it was delisted in 2014 after it was out of stock. The new batch of NIST SRM 916b bilirubin has been available since 2021 and was submitted for the 2024 JCTLM review cycle. NIST also provides a frozen plasma matrix material with a non-certified reference value for bilirubin.

The JCTLM database lists two entries of RMPs published for bilirubin. However, both methods are based on spectrophotometry and provide the same reference so there is only one method on the JCTLM database [28]. In 2018, the original Doumas method was adapted to be applied without the use of the bilirubin calibrant due to the lack of available NIST SRM 916a bilirubin at that time [29]. This variation of the method relies on the use of a molar absorption coefficient for calibration, rather than a higher order reference material. It is not currently listed in the JCTLM database and thus it still needs to be formally assessed as a candidate RMP.

Historically, precision for the Doumas RMP has been reported to achieve a coefficient of variation of <1 % for values above 100 μmol/L (divide by 17.1 to convert to mg/dL) [28]. Inter-laboratory Doumas RMP coefficients of variation at concentrations >34.2 μmol/L were less than 2.4 % [30]. While many laboratories report use of the Doumas spectrophotometric method, only three of these laboratories are currently listed in the JCTLM database for providing Reference Measurement Services for bilirubin [26]. A list of all of the laboratories that participate in the “RELA External Quality Control for Reference Laboratories” scheme for total bilirubin and their performance is available on the RELA website [31]. Typically, around 23 laboratories participate each year (noting this is not a dedicated neonatal program), three of which are the JCTLM-listed reference service providers. The typical CVs for these results are ≈5–10 %.

Discussion

Although various EQA programs consistently report differences between methods and the assignment of RMP target values makes it possible to assess bias, there are some major prerequisites to this method of post market surveillance. The commutability of the samples used is often not, or not formally, demonstrated and typically, EQA providers are often limited to the use of adult human serum due to obvious practical restraints in obtaining neonatal material in sufficient amount at relevant concentrations. In EQA programs the serum is often spiked with unconjugated bilirubin, to obtain elevated concentrations encountered in the biological materials from neonates at risk for jaundice. As the conjugation function of neonates is not fully mature [32, 33], it remains uncertain to what extent serum from adult donors spiked with unconjugated bilirubin properly mimics the characteristics of serum from neonates. For TBil, the lack of commutable calibrators and EQA materials likely contributes to discrepancies reported in bias of bilirubin results between native neonatal material and in adult serum-based control samples. For instance, in a French study, it was shown that some EQA materials have insufficient commutability: in this case, results of an individual laboratory can only be compared with its peer laboratories using the same analyzer [34]. In this study, the authors also showed that addition of some cryoprotectants like polyethylene glycol (PEG) could alter commutability of calibration materials specifically for some assays. Additionally, effects of added conjugated bilirubin seems to also affect results provided by some specific platforms, which has not been well investigated [6, 7, 35].

Most EQA programs for bilirubin testing are established for all patient samples and not specific to neonates, which may be inappropriate given that differences may exist between samples from adults and from neonates (e.g., differences in the amount of conjugated vs. unconjugated bilirubin etc.). In addition, some EQA providers lack a target value as established by an RMP, allowing only comparability to other methods, rather than assessment of bias vs. the ‘true’ value. Ultimately, we advocate that where possible best practice would be the establishment of EQA programs specifically dedicated to neonatal TBil measurement with commutable control materials value assigned with an RMP.

Discussion

Over the last decade there has been development in near patient testing bilirubin devices, encompassing both blood based and non-invasive approaches, proposed for community screening of neonatal jaundice [36]. Appropriate screening tests will have a decision cascade of e.g., screen negative (i.e., no further action), uncertain (i.e., monitor with a repeat test) or screen positive (i.e., refer for tertiary clinical care). Considerations of traceability and how the decision limits are formed for these devices is beyond the scope of this position paper, but will continue to be an important consideration as more of these devices are adopted into patient care pathways. Furthermore, when wanting to integrate these devices into a patient care pathway, additional challenges on performing the traditional mandatory method comparisons prevail. Importantly, the accuracy of TBil results provided by the central laboratory is likely to be integral to the standardization of these devices.

Discussion

In 1999, the ‘Bhutani nomogram’, a tool for universal TBil screening in term and late preterm infants was published [37]. This nomogram consists of various TBil–based curves and four risk zones. The risk zones predict the likelihood of a subsequent TBil result falling into the high-risk zone of the nomogram, providing a clinical interpretation of the risk of developing severe hyperbilirubinemia. It has been instrumental in guiding the decision-making process on neonatal jaundice follow-up care in many newborns [37]. The Bhutani curve was originally constructed with TBil measurements from a diazo assay on the Hitachi 747 platform [37]. Whilst there were efforts to ensure accuracy of TBil measurement for the Bhutani curve, inter- and intra-laboratory variability in TBil measurements already existed at that time [41, 42]. Consequently, TBil test results obtained on different platforms may lead to different risk zone assignments and corresponding follow-up policy to prevent imminent severe hyperbilirubinemia. Although recently updated with data from 15 years of universal bilirubin screening, the ‘1999’ screening nomogram continues to be in use in countries outside the USA [40]. The newly revised hour-specific nomogram includes 140 times more infants compared to the previous nomogram, slightly increased the TBil-defined risk thresholds, and added the influence of neurotoxic risk factors [43] The update does not take into account the difference in TBil measurement methods; nearly all of the TBil measurements were done on Beckman-Coulter instruments which use a variation of the diazo method.

The hour-specific treatment nomograms of CGPs are not based on a single specific measurement method of bilirubin [39]. In fact, treatment thresholds vary between guidelines, and countries, reflecting the lack of evidence. The threshold TBil concentration at which treatment is initiated is largely based on observational studies that assess TBil concentrations associated with an increased risk of neurodevelopmental impairment. These studies are derived from a variety of populations, and hospitals using different measurement platforms and methods, whereas fixed medical decision limits require harmonized results between methods. With the advent of new data since the 2004 AAP guidelines, the graphs for initiation of PT and ET have been updated and are described in the 2022 AAP guidelines [2]. In all CGPs, PT is the treatment of choice if the TBil exceeds the corresponding PT threshold based on the newborn’s gestational age and specific risk factors. Additionally, PT may be considered even when the TBil is below the threshold if the rate of TBil increment is higher than expected. When single-sided PT with normal intensity fails to reduce TBil, escalation of care, such as double-sided PT, an increase in PT intensity and even ET, may be considered. An ET is an invasive procedure initiated when TBil is higher than predefined TBil values, often ≥ 85 μmol/L above the PT threshold [2], or in cases of acute bilirubin encephalopathy regardless of TBil test results. PT is usually discontinued when the TBil is well below the hour-specific treatment threshold, usually ≥35–50 μmol/L lower, to prevent the risk of rebound hyperbilirubinemia [44], again underlining the importance of any measurement of TBil being close to the true value.

It is known that differences in treatment thresholds result in variations in PT rates: up to 11 % of newborns born at ≥35 weeks’ gestation and up to 97 % of extremely low birth weight infants (birthweight <1,000 g) require PT for hyperbilirubinemia [11, 12, 38, 41, 45]. But even when similar screening or treatment nomograms are followed, disparity in neonatal jaundice management may occur due to lack of comparability and inaccuracy of bilirubin assays from different manufacturers. Imprecision within instrument peer groups is typically less than 4 %. And while many peer group means are within 10 % of the target value, others exceed this, resulting in variation of care [46].

Ultimately, biased TBil test results will influence the overall burden of TBil determinations, PT utilization, hospital visits and admissions, and healthcare costs, potentially affecting patient outcomes and family well-being. Whilst the International Consortium for Harmonization of Clinical Laboratory Results considers that an adequate reference measurement system is in place for TBil, they do not specifically consider the intended use for neonatal care [47]. Therefore, we consider improvement of TBil standardization an urgent and important matter.

Discussion

So far, only a few studies have described differences between bilirubin measurement methods leading to substantially different clinical decisions in neonatal care (Table 1). A national study in 36 French laboratories with 11 measurement platforms reported on accuracy in a reference standard, control and patient samples, the maximal TBil concentration being 315 μmol/L [34]. Whereas variation in TBil concentration <150 μmol/L were considered acceptable, variation in TBil values above 150 μmol/L were 10 % or even higher. The authors proposed using a correction factor for TBil test results >150 μmol/L for particular instruments to prevent undertreatment. Australian colleagues studied interlaboratory variability of 11 neonatal samples with TBil concentrations between 75 and 360 μmol/L across seven platforms in four laboratories [6]. Mean differences ranged from −12 and +20 % relative to a particular platform, resulting in altered screening and treatment decisions in three and four of 11 individual neonatal samples, respectively. Canadian researchers studied analytical differences across four different instruments for 40 adult samples as well as the impact of switching TBil assays on neonatal hyperbilirubinemia risk classification and PT rates for >65,000 Canadian neonates [50]. Mean differences of up to 17 % were observed between two platforms. These differences in TBil concentrations resulted in PT rate increases of up to 22 % in the neonate cohort. Subsequent switching of platforms led to reductions in PT rates.

Finally, a recent Dutch study revealed a 3-fold increase in PT when switched from a diazo to a vanadate oxidase method within the same laboratory [7]. The mean differences of the vanadate vs. the previously used diazo method was much higher for neonatal samples when compared with adult ones: ≈+17 % vs. ≈+4 %, respectively. External quality assessment (EQA) data analysis of six samples supplemented with unconjugated bilirubin (range 233–435 μmol/L) from 66 laboratories (131 analyzers from 9 manufacturers) showed that the bias of commercial TBil methods varied from ≈−4 to ≈+20 %. In addition to these studies, an Italian case report of a preterm infant with (un)conjugated hyperbilirubinemia highlighted discrepant TBil measurements up to 244 μmol/L between two platforms [51]. Differences in results from different platforms are clearly altering some clinical outcome decisions and authors of these publications have highlighted that working to resolve this needs to be a priority.