BACKGROUND AND OBJECTIVE: Transcutaneous bilirubin (TcB) correlates positively with total serum bilirubin (TSB) across different racial populations. However, little is known about the pattern of divergence between TcB and TSB at individual patient-level and the relationship with TcB measuring techniques among African neonates. This study, therefore, investigates TcB–TSB discrepancies and the contribution of 2 models of transcutaneous bilirubinometers to the observed divergence in a black African population.
METHODS: Medical records were retrieved for late preterm and term infants with 1 to 3 pairs of TcB and TSB measurements between December 2011 to June 2015 in Nigeria. Divergence between TcB and TSB values for each infant was determined and the associated factors explored with generalized estimating equations for logistic regression. Contributions of BiliChek and JM-103 transcutaneous bilirubinometers to the divergence were further explored through linear regression and Bland-Altman analysis.
RESULTS: Overall, 2107 TcB/TSB measurements from 1553 infants were analyzed. TSB was overestimated by ≥2 mg/dL in 64.5%, ≥3 mg/dL in 42.7%, and ≥4 mg/dL in 25.7% of all measurements. In contrast, TSB was underestimated by ≥2 mg/dL in 1.1%, ≥3 mg/dL in 0.5%, and ≥4 mg/dL in 0.3% of all recordings. Postnatal age, feeding mode, and type of TcB instrument were predictive of TSB overestimation. The JM-103 was associated with greater imprecision than BiliChek at all TSB levels.
CONCLUSIONS: BiliChek and JM-103 bilirubinometers significantly overestimate TSB in black African neonates and may result in unnecessary or excessive treatments. Additional development of appropriate bilirubin determination devices for this racial group, especially in resource-limited settings, is warranted.
- AAP —
- American Academy of Pediatrics
- CI —
- confidence interval
- G6PD —
- glucose-6-phosphate dehydrogenase
- LOA —
- limits of agreement
- OR —
- odds ratio
- TcB —
- transcutaneous bilirubin
- TSB —
- total serum bilirubin
What’s Known on This Subject:
Transcutaneous bilirubin (TcB) is highly correlated with total serum bilirubin (TSB) across ethnic populations. TcB is a useful screening tool or indication for treatment in infants with severe hyperbilirubinemia where facilities for TSB are limited or lacking.
What This Study Adds:
Approximately 1 in 3 black African neonates with hyperbilirubinemia may be prone to TcB overestimation (≥3 mg/dL), resulting in unnecessary treatments where confirmatory TSB is not readily available. Improved bilirubin measuring devices are needed for this racial population, especially in resource-limited settings.
Studies among neonates of African-American ancestry have so far shown that the black race is unusually associated with a decreased risk of severe hyperbilirubinemia but an increased risk of kernicterus,1,2 underpinned by glucose-6-phosphate dehydrogenase (G6PD) deficiency.3 As a high-risk population, routine, predischarge bilirubin screening is warranted in this racial group to minimize the risk of kernicterus.4 The excellent correlation between transcutaneous bilirubin (TcB) and total serum bilirubin (TSB) tests5,6 has made TcB a valuable screening tool for TSB in the management of hyperbilirubinemia worldwide.4,7,8 However, TcB tends to underestimate TSB in whites or neonates with light or medium skin color and overestimate TSB in non-whites or neonates with dark skin color.9–11 Such discrepancies may compromise the effective management of infants at risk for bilirubin encephalopathy.12 Limited evidence exists on the pattern and predictors of individual TcB–TSB divergence, especially in settings with predominantly dark-skinned infants or limited access to timely and reliable TSB measurements.13–15 This study, therefore, set out to determine the prevalence and correlates of significant TcB overestimation in a black African population to facilitate optimal decisions for appropriate clinical intervention.
This prospective, observational study was conducted at Island Maternity Hospital, a leading specialist maternity hospital in Nigeria owned by the Lagos State Government. This 180-bed public hospital is a referral center for over 300 private and public hospitals in the Lagos metropolis and its environs.
All healthy late preterm and term neonates (gestational age ≥35 weeks or birth weight ≥2.2 kg) delivered between December 2011 and June 2015 with 1 to 3 pairs of TcB–TSB measurements in the first 120 hours of life were enrolled. Only measurements recorded before receiving phototherapy were eligible for analysis. All sick infants were excluded. The study was conducted under institutional ethical approval from the Lagos State Health Service Commission.
Infants of consenting mothers were screened routinely for jaundice by using the BiliChek transcutaneous bilirubinometer (Philips Healthcare North America, Monroeville, PA) or the JM-103 transcutaneous bilirubinometer (Draeger Medical, Telford, PA). Measurements were derived from the infant’s sternum by specially trained nurse assistants. The BiliChek was used from December 2, 2011 to November 26, 2012 and the JM-103 was used from November 27, 2012 to June 23, 2015. Two units of each piece of equipment were available to provide back-up for maintenance. Both instruments were used in accordance with the manufacturers’ instructions for quality control and calibration. The algorithms and other technical features of the 2 devices are well described.5,6,9 Infants whose bilirubin levels at screening exceeded 3 mg/dL below the recommended postnatal age threshold for phototherapy based on the American Academy of Pediatrics (AAP) guidelines were assessed for TSB within ∼1 hour. This criterion was adopted because of the high prevalence of G6PD deficiency in this population.16 TSB measurements were performed on capillary blood samples drawn by heel puncture and analyzed by direct spectrophotometry by using the Advanced Bilirubin Stat-Analyzer (model BR2; Advanced Instruments, Inc, Norwood, MA).
The choice of putative predictor variables was guided by relevant literature,9–15 and included infant’s sex, birth weight, estimated gestational age (in completed weeks, based on maternal history of last menstrual period correlated with ultrasound scan by using Siemens Sonoline SI-450 as documented in the hospital records), postnatal age at screening, maternal self-report of ethnicity (based on the 3 most predominant tribes in Nigeria: Hausa, Ibo, and Yoruba, as well as others), skin color (light brown, medium brown, or dark brown), mode of feeding at the time of TcB screening (formula only, exclusive breastfeeding, or mixed feeding), G6PD status, and TcB model. Skin color was classified as previously reported in this population,14 and recorded by the trained nurse assistants as shown in a color guide (Supplemental Fig 5). G6PD evaluation was based on the fluorescent blood spot for all infants by using methods described by Beutler et al,17 as previously reported.18 Lack of fluorescence after 10 minutes was indicative of deficient levels.
Three levels of significant overestimation or underestimation were computed from the TcB–TSB pairs: TcB–TSB ≥2 mg/dL, ≥3 mg/dL, or ≥4 mg/dL. Because of the small amount of TSB underestimation by TcB, further analysis was not considered necessary. Univariate associations between the predictor variables and bilirubin overestimation (each defined as a binary outcome) were then explored with Pearson’s χ2. Strength of association was estimated by odds ratios (OR) and the corresponding 95% confidence intervals (CI) as an approximation of the relative risk. Statistical significance was set at a critical level of P < .05. Factors significantly associated with each outcome in the univariate analysis were entered into the respective multivariable logistic regression model. The generalized estimating equation method was used for the regression modeling to account for repeated measurements on the same infant. All models were adjusted for TSB level. Plausible interaction between TSB and each predictor variable was also investigated.
The contributions of BiliChek and JM-103 to TSB overestimation were examined further with Pearson’s correlation coefficients (r), scatter plots, and linear regression analyses. Because of the limitations of Pearson’s coefficients in determining the variability and linear correlation between 2 continuous variables, we explored the degree of agreement (bias) between TcB and TSB for each device using the Bland-Altman method.19 The mean bias was defined as the mean difference between each pair of TcB and TSB measurement obtained from each enrolled infant. TcB imprecision or variability compared with TSB was defined as 2 SD from the mean difference. The limits of agreement (LOA) of the mean differences and their 95% CIs were also determined. Additionally, the discriminatory power of the TcB devices in predicting infants who required phototherapy was determined by using receiver operating curve analysis.20 The criteria for phototherapy were based on the appropriate age-specific thresholds in the AAP guidelines.4 The interinstrument variability in a sample of infants tested simultaneously with BiliChek and JM-103 was also computed. IBM SPSS Statistics for Windows software, version 21.0 (IBM Corporation, Armonk, NY) was used for all statistical analyses.
A total of 1553 infants with 2107 paired TcB–TSB measurements were enrolled for the study. The infants had a mean TcB of 10.04 (SD, 3.40) mg/dL and a mean TSB of 7.28 (SD, 2.66) mg/dL. TcB ranged from 2.1 to 19.9 mg/dL, whereas TSB ranged from 0.3 to 19.5 mg/dL. The characteristics of the infants are presented in Table 1. Some 54.9% (853/1553) of the infants were male, 5.7% (89/1551) weighed <2.5 kg, and 9.8% (150/1533) had a gestational age of <37 weeks. The infants were predominantly of the Yoruba tribe (71.2%), had light or medium brown skin (91.6%), and were exclusively breastfed (69.1%). A total of 263 (16.9%) infants were G6PD deficient.
TcB recordings in 79.1% of the infants were with the JM-103. TSB was overestimated by ≥2 mg/dL in 64.5% (1358/2107), ≥3 mg/dL in 42.7% (899/2107), and ≥4 mg/dL in 25.7% (542/2107) of all measurements. TSB was overestimated by ∼8 mg/dL 32 times (19 neonates). In contrast, TSB was underestimated by ≥2 mg/dL in 1.1% (23/2107), ≥3 mg/dL in 0.5% (10/2107), and ≥4 mg/dL in 0.3% (7/2107) of all measurements. Of the 1774 (84.2%) overestimated TSB values ≥1 mg/dL, phototherapy was required in only 55 (3.1%) cases.
In the univariate logistic regression analysis, 4 factors (postnatal age, G6PD status, feeding mode, and TcB equipment) were significantly associated with the 3 levels of overestimation. Ethnicity or skin color was not significantly associated with overestimation. The independent predictors of significant bilirubin overestimation after multivariable logistic regression are presented in Table 2. Increasing postnatal age and use of the JM-103 were predictive of all levels of overestimation, whereas exclusive breastfeeding was only associated with ≥2 mg/dL overestimation. TSB level had a modifying effect on feeding mode, which prompted our adjusting all models with the interaction term, feeding mode*TSB.
The correlation between TcB and TSB was linear and significant for BiliChek (Pearson’s r = 0.83, R2 = 0.69, P < .001) or JM-103 (Pearson’s r = 0.85, R2 = 0.72, P < .001) as presented in Fig 1. The combined measurements also showed significant correlation (Pearson’s r = 0.81, R2 = 0.66, P < .001). As shown in Fig 2, the discriminatory power of JM-103 (area under the curve = 0.725; 95% CI, 0.665–0.786) was slightly higher than the value for BiliChek (area under the curve = 0.706; 95% CI, 0.625–0.787) in predicting the need for phototherapy. JM-103 had an optimal combination of sensitivity (64%) and specificity (72%) at a cut-off of 12.1 mg/dL in detecting infants who required phototherapy, compared with BiliChek, which had a sensitivity of 60% and a specificity of 74% at a cut-off of 10.3 mg/dL. JM-103 was associated with 100% sensitivity for TcB values up to 4.8 mg/dL, compared with 1.6 mg/dL for BiliChek.
The Bland-Altman plots are presented in Fig 3. The mean bias for JM-103 was 3.04 (SD, 1.89) mg/dL (95% LOA: –0.66 to 6.74 mg/dL) compared with 1.28 (SD, 1.77) mg/dL (95% LOA: –2.20 to 4.76 mg/dL) for BiliChek. The pattern of mean bias at various levels of TSB for all paired measurements is presented in Fig 4. The mean bias between JM-103 and BiliChek widens as TSB increases, with BiliChek underestimating TSB from roughly 13.5 mg/dL. The index of consistency based on the average measures of intraclass correlation between BiliChek and JM-103 in a sample of 276 neonates was 0.818 (95% CI, 0.769–0.856), which is indicative of a high degree of reliability.
TcB has strong positive linear correlation with TSB regardless of the choice of bilirubinometer. However, TcB is likely to overestimate TSB by at least 2 mg/dL in ∼6 out of every 10 black African neonates screened in the first 120 hours of life by either BiliChek or JM-103. Increasing postnatal age, exclusive breastfeeding, and type of bilirubinometer are predictive of significant overestimation. Ethnicity, G6PD status, and skin color are not significantly associated with overestimation in this population. TSB underestimation by both TcB devices is also minimal. However, JM-103 is associated with higher imprecision (mean bias) than BiliChek, which increases as TSB levels increase.
Comparison With Other Studies
In general, the high and positive correlation between TcB and TSB is consistent with the findings in several studies from racially diverse populations.9,10,14,15,21–23 The TSB overestimation is also corroborated by several studies among multiracial populations10,13,24 and black neonates.14,15 For example, although Maisels et al10 found a close correlation between TcB and TSB values among 849 neonates across all ethnic groups, by using JM-103, TSB was overestimated by ≥2 mg/dL in 37.9% of blacks, compared with 1.8% of whites or 5.4% of other races. Among black infants, TSB was overestimated by ≥3 mg/dL in 17.4% and by ≥4 mg/dL in 6.7%. TcB was also consistently higher than TSB in the black population whenever the difference between TcB and TSB was ≥3 mg/dL. Similarly, in a multicenter study of 769 racially and ethnically diverse newborns with at least 1 TcB–TSB pair, Taylor et al found that African-American race was independently associated with a higher chance of TSB overestimation by at least 2 mg/dL (OR, 2.19; 95% CI, 1.09–4.39).13 In a predominantly black population, 1 study from Malawi reported TSB overestimation in term neonates not receiving phototherapy as indexed by a mean bias of 1.5 (SD, 4.2) mg/dL in the TcB measurements.15
JM-103 overestimated TSB in this population by a greater magnitude than BiliChek. Studies comparing BiliChek and JM-103 in predominantly black populations are rare. However, the significantly higher mean measurement bias associated with JM-103 compared with BiliChek appears consistent with findings in other studies among multiracial populations.10,25–28 For example, based on a limited sample of black infants, Maisels et al10 suggested that although JM-103 correlated with BiliChek, it had a higher tendency to overestimate TSB. In another multiracial population in the United Kingdom, JM-103 was also found to overestimate TSB in black infants (mean difference, 1.6 mg/dL; 95% CI, 1.0–2.3 mg/dL) and underestimate TSB in white infants (mean difference, –2.2 mg/dL; 95% CI, –2.7 to –1.7 mg/dL).25 However, despite the comparable correlations between BiliChek and JM-103 in an Asian population, the former had a higher tendency to overestimate TSB.27
To date, only 1 study is known to have explored the independent predictors of TcB–TSB divergence.13 In that study, a postnatal age of <48 hours was independently associated with TSB overestimation by ≥2 mg/dL (OR, 0.36; 95% CI, 0.22–0.59) and ≥3 mg/dL (OR, 0.38; 95% CI, 0.20–0.72). These findings are in agreement with our study, which showed that a postnatal age >48 hours was associated with higher odds of TSB overestimation at all levels. This may be due to the gradual reduction in the total blood flow through the skin and muscle during the first 7 days of life within the neonatal period.29 The study by Taylor et al13 also showed that TcB had a tendency to be less accurate at higher TSB levels, regardless of the type of bilirubinometer (Fig 4). Other studies among whites seem to support this finding,30 suggesting that it was not race dependent. This observation may possibly be attributable to the dynamic processes, such as changes in hemoglobin concentration, in the early hours of life.13
The association between exclusive breastfeeding and TSB overestimation appears to be supported by other reports on TcB patterns in relation to feeding mode.31–33 Itoh and colleagues,33 for example, proposed that increased unconjugated bilirubin in breastfed infants may be underpinned by the high enterohepatic circulation of unconjugated bilirubin from deconjugation by β-glucuronidase of the conjugated bilirubin in meconium. The resultant increase in circulating unconjugated bilirubin may therefore be related to the observed higher TcB/TSB discrepancy. The observed TSB overestimation by at least ≥4 mg/dL in this population warrants additional investigation.34
Our study also suggests that tribe or shades of skin color in the black race has little or no effect on TcB performance. This is in accordance with an earlier study that demonstrated strong correlations between TcB and TSB regardless of the degree of skin pigmentation.14
Clinical Implications of Key Findings
To optimize its utility, high TcB levels from any instrument should precipitate a valid concern about the risk of severe hyperbilirubinemia or kernicterus in the tested infant that requires a follow-up blood sampling for TSB or provide assurance where such risk does not exist. Within this context, the strong correlation between TcB and TSB confers some merit on the former as a screening tool as well as a potential alternative to TSB in resource-limited settings. TcB should, for example, be preferred to the highly subjective visual assessment that prevails in such settings. However, the evidence from the current study and similar reports10,13 would suggest some caution in placing sole reliance on TcB as a basis for clinical decisions to guide phototherapy or exchange transfusions in black African neonates.
Perhaps more notable is the ample evidence that the commonly reported satisfactory statistical correlation between TcB and TSB masks clinically significant discrepancies between the 2 measures at the individual patient level. TSB should therefore continue to be regarded as an absolute necessity, notwithstanding the common challenges in ensuring its availability routinely. In settings where a high proportion of infants are born outside hospitals and late presentation of high-risk infants is frequent, the utility of TcB is likely to be severely limited considering the significant imprecision of measurements at high TSB levels. This is without prejudice to studies that suggest improved predictive utility of TcB when combined with clinical risk factors.5 It may be argued that the divergence between TcB and TSB should not be unexpected after all, because the former measures the extravascular rather than the intravascular bilirubin concentration, and possibly better reflects the risk of kernicterus.35 However, additional studies are warranted to establish more precisely the clinical value of TcB as a predictor of kernicterus.
Our study also underscores the need to consider the performance and cost-effectiveness of available TcB equipment. Of the 3 predictors in this population, choice of TcB device is the most modifiable factor. Available cost estimates suggest that BiliChek is more expensive than JM-103 because of the added running cost for consumables. However, a more robust cost-effectiveness analysis should be considered for the latest models of all commercially available instruments to guide the expected trade-off among devices. To our knowledge, no instrument has been developed specifically to address TSB overestimation in the black population. Although there is a need to explore technological improvement for TcB devices for this racial group, efforts to develop low-cost, point-of-care TSB devices are currently in progress and, if successful, hold promise for low- and middle-income countries.8 It may also be helpful to consider the development of equipment-specific TcB nomograms in this population.
Strengths and Limitations of Study
A major strength of this study is the large sample of exclusively black African neonates. This is the first study, to our knowledge, to characterize the pattern and the risk factors of the divergence between TcB and TSB in this population. Unlike previous studies, we were able to compare the performance of 2 common TcB equipment models. However, only 2 TcB models were evaluated, which may restrict generalization to other TcB instruments. It was also unclear if the use of BiliChek for a significantly shorter duration impacted the findings. We could not determine the effect of deriving TcB measurements from the forehead compared with the sternum. It was unclear how the finding on skin color and ethnicity would have been moderated by comparison with white infants. Additionally, this study was conducted among inborn infants and may not provide an accurate picture of the pattern of divergence among outborn infants that are frequently exposed to hemolytic infant care products in this population.16 Notwithstanding, our study has provided valuable and rare insights to the potential limitations of TcB as a screening tool or proxy for TSB among black African neonates to inform clinical decisions and additional technological development of TcB devices for the optimal care of infants at risk for avoidable kernicterus.
TcB is noninvasive, simple to perform, less expensive, and has a satisfactory correlation with TSB. However, current TcB devices are highly sensitive only at lower TSB levels and are more likely to overestimate TSB with increasing age by at least 2 mg/dL in ∼6 out of every 10 black African neonates and up to 9 mg/dL in some neonates. The LOA depend on the type of bilirubinometer used, with JM-103 having a greater tendency to overestimate TSB than BiliChek. This divergence should be recognized in the management of infants at risk for kernicterus to avoid unnecessary blood sampling or excessive treatments where routine confirmatory TSB is impracticable. Technological advancement for improved bilirubin determination is warranted for this population.
We thank Tina Slusher for the provision of the bilirubinometers and G6PD test kits used in this study. We also thank the research team at the Centre for Healthy Start Initiative for assistance in data retrieval and management.
- Accepted June 30, 2016.
- Address correspondence to Bolajoko O. Olusanya, FRCPCH, PhD, Centre for Healthy Start Initiative, 286A Corporation Dr, PO Box 75130 VI, Dolphin Estate, Ikoyi, Lagos, Nigeria. E-mail:
FINANCIAL DISCLOSURE: The transcutaneous bilirubin devices used in this study were provided by Philips Healthcare North America (Monroeville, PA) and Draeger Medical (Telford, PA). The authors have indicated they have no other financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- American Academy of Pediatrics Subcommittee on Hyperbilirubinemia
- el-Beshbishi SN,
- Shattuck KE,
- Mohammad AA,
- Petersen JR
- National Institute for Health and Clinical Excellence
- Bhutani VK,
- Gourley GR,
- Adler S,
- Kreamer B,
- Dalin C,
- Johnson LH
- Maisels MJ,
- Ostrea EM Jr,
- Touch S, et al
- Samiee-Zafarghandy S,
- Feberova J,
- Williams K,
- Yasseen AS,
- Perkins SL,
- Lemyre B
- Olusanya BO,
- Ogunlesi TA,
- Slusher TM
- Taylor JA,
- Burgos AE,
- Flaherman V, et al; Better Outcomes through Research for Newborns Network
- Slusher TM,
- Angyo IA,
- Bode-Thomas F, et al
- Bland JM,
- Altman DG
- Thomson J,
- Culley V,
- Monfrinoli A,
- Sinha A
- Bosschaart N,
- Kok JH,
- Newsum AM, et al
- Copyright © 2016 by the American Academy of Pediatrics