Objective. To compare estimates of serum bilirubin as determined by a transcutaneous device (BiliChek [BC]) with laboratory-measured total serum bilirubin (TSB) in a predominately Hispanic population in which a significant number of TSB values ≥15 mg/dL was anticipated.
Methods. A total of 248 Hispanic and 56 non-Hispanic neonates were studied. Transcutaneous measurements were performed by 1 investigator within 30 minutes of blood sampling for TSB; TSB was determined in a large clinical laboratory using the diazo Jendrassik-Grof with blank method. Agreement between BC and TSB determinations was assessed using Bland-Altman plots and the Bradley-Blackwood test. Interdevice comparisons were made among the BC devices. Predictive indices for TSB >10 mg/dL and >15 mg/dL were determined using various BC cutoff values.
Results. TSB was ≥15 mg/dL in 31% of the Hispanic neonates. BC generally tended to underestimate TSB determinations, and this trend was more pronounced when TSB was >10 mg/dL. Very high sensitivities were observed only when relatively low BC cutoff values were used to predict TSB >10 mg/dL or >15 mg/L. Relatively small numbers of infants had BC values in these low ranges.
Conclusions. The tendency of BC to underestimate TSB limits its usefulness in neonates with relatively high TSB. In this population, most infants would have required additional evaluation to ensure that TSB was not >10 mg/dL or >15 mg/dL. It seems that the discrepancy between this study and previous studies of BC is related to our relatively large number of TSB values ≥15 mg/dL.
The management of jaundice in term and near-term neonates is a challenging clinical problem, complicated by increased prevalence of breastfeeding, early hospital discharge, and difficulty in ensuring prompt postdischarge follow-up.1 Concerns have been raised regarding an apparent increase in kernicterus in this group of infants,2 and new approaches to prevention and diagnosis of hyperbilirubinemia have been sought.3 In particular, there has been considerable interest in the development of noninvasive techniques to estimate bilirubin production4,5 and the level of serum bilirubin.6–8 These techniques are pain-free and provide rapid availability of the desired result.
The Bili-Chek (BC; Respironics, Marietta, GA) is a hand-held device developed to provide a rapid, noninvasive transcutaneous estimation of total serum bilirubin (TSB). Light from the instrument is reflected from the infant’s skin, and, with the use of known characteristics of major skin components that affect spectral reflectance in the neonate (collagen, melanin, hemoglobin, and bilirubin), the principle of spectral subtraction can be used to derive bilirubin concentration.9,10 Previous studies of the performance of BC have shown a strong correlation between laboratory determination of bilirubin versus the transcutaneous value,11,12 and this device has been approved for clinical use in neonates, including those receiving phototherapy. However, these studies have included relatively few Hispanic infants and a small number of infants with serum bilirubin values ≥15 mg/dL.
The objectives of this study were 1) to compare estimates of serum bilirubin using the BC method with TSB values determined in a large clinical laboratory, 2) to determine BC cutoff values at which specificity and sensitivity are maximized, and 3) to perform this evaluation in a setting in which the majority of the study population would be Hispanic and a large number of TSB values ≥15 mg/dL was anticipated.
The study was conducted in the newborn nursery (NBN) at Parkland Memorial Hospital. The NBN, to which approximately 14 000 neonates are admitted each year, accepts infants with an estimated gestational age ≥35 weeks and a birth weight ≥2100 g. Infants in the NBN were identified as study candidates if the primary caregiver determined that clinically apparent jaundice necessitated serum bilirubin determination. Demographic data were collected primarily through review of the infants’ medical records. Infants who were inpatients were evaluated in the NBN while in a bassinet; outpatients, ie, infants who were recently discharged and returning for a follow-up TSB, were evaluated while on a radiant warmer located in a treatment room within the NBN.
BC determinations were made in accordance with the manufacturer’s recommendations, and the device was calibrated before each measurement. All determinations were obtained from the infant’s forehead, and readings were taken by 1 investigator only (D.M.) while the infant was in a quiet state. For each infant, readings were obtained with 2 of the 4 BC devices used in the study, and the first reading was used for data analysis. Only BC determinations performed within 30 minutes of blood sampling for TSB (blood sampling was performed first in 57% of the comparisons) were included in the analysis. No study patients were receiving phototherapy when the transcutaneous values or blood samples were obtained; 6 patients included in the analysis were studied 8 to 22 hours after phototherapy, and the skin sites analyzed were not patched during phototherapy. The 4 devices used during the study were designated 1 to 4. An abrupt mechanical failure occurred in device 2 during the study, and it was replaced; because of the comparability of the results before and after replacement, these values were grouped together under the designation “device 2.”
Blood was drawn by heel puncture, and serum bilirubin determinations were performed in the clinical chemistry laboratory at Parkland Health and Hospital System. This laboratory performs approximately 5000 total bilirubin assessments per month and uses the diazo Jendrassik-Grof with blank method (Olympus AU600; Olympus America Inc, Melville, NY). In quality control procedures performed during the study period, coefficient of variation was always <2%. Results of the College of American Pathologists Neonatal Bilirubin Proficiency Survey (NB-B 2000) demonstrated that determinations at our institution on samples with total bilirubin ranging from 4.3 to 21.0 mg/dL all were within 0.4 mg/dL of the mean values for 22 participating laboratories. In an analysis separate from the current study, samples were processed both in our laboratory (TSB range: 7.6–17.0 mg/dL) and by high-performance liquid chromatography (HPLC; TSB range: 7.8–18.9 mg/dL). The correlation coefficient was 0.96, and the difference was <1 mg/dL in 19 of 26 comparisons; although no values differed by >1.9 mg/dL, a tendency for values in our laboratory to underestimate HPLC values was noted (P < .001; G. R. Gourley and B. Kreamer, University of Wisconsin School of Medicine, personal communication, July 2001).
Demographic information as well as BC and TSB values were entered into a dedicated database (Microsoft Access 1997; Microsoft Corp, Redmond, WA) for later analysis. Characteristics of study neonates were compared for the 2 groups using the Wilcoxon rank sum test for quantitative measurements and the χ2 test of equality of binomial proportions for categorical observations. Comparisons between the BC and TSB determinations were made for all infants and within various ethnic groups using the Bradley-Blackwood test13,14; this procedure also was used to assess the relative agreement of the 4 devices. For TSB values >10 mg/dL and >15 mg/dL in Hispanic neonates, the sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios15 associated with various BC cutoff levels were determined, and receiver operator characteristic (ROC) curves were plotted.
This study was approved by the institutional review boards at the University of Texas Southwestern Medical Center at Dallas and Parkland Health and Hospital Systems.
A total of 404 comparisons between BC and TSB determinations were made in 304 infants, the majority of whom were inpatients (64%; Table 1). A total of 248 of the study patients (82%) were Hispanic. In this group, serum bilirubin was ≥15 mg/dL in 78 infants (31%). Fifty-six study infants were non-Hispanic (41 black, 10 white, 5 other). The postnatal age at which BC/TSB comparisons were made is shown in Table 2.
In Fig 1, 404 comparisons of TSB and BC in the 304 study patients are shown. It is apparent that the majority of points fall above the line of equality, particularly when TSB is >10 mg/dL. The equation for the least squares best fit line is TSB = 1.01 (BC) + 1.09; r = 0.84.
Assessment of equality of means and variances by the Bradley-Blackwood test for Hispanic neonates is shown in Fig 2. This plot of the average of BC and TSB versus the difference between the values demonstrates a highly significant (P < .0001) lack of agreement, with a tendency of the BC method to underestimate higher TSB values. A similar assessment for non-Hispanic neonates is shown in Fig 3. In the non-Hispanic infants, there were no differences in the slope or y intercept from 0, although relative to the Hispanic group, there were fewer infants and the percentage with TSB values ≥15 mg/dL was lower (9% vs 31%, respectively).
Although the tendency of BC measurements to underestimate TSB was noted with all devices, this tendency was more pronounced with 1 device used in this study. A relatively exaggerated tendency to compress values at higher TSB levels was observed with device 4 (1 vs 4, 3 vs 4: P < .001; 2 vs 4: P = .015). However, no significant changes in analytical results were noted when data from device 4 were excluded.
Predictive indices of various BC cutoff values for TSB >10 mg/dL in Hispanic neonates are shown in Table 3, and the ROC curve is plotted in Fig 4A. As sensitivity decreases with increasing BC cutoff values, the ability to be certain that TSB is not >10 mg/dL also decreases. For example, a BC cutoff >7 mg/dL has a sensitivity of 100%, although specificity is relatively low. Although no neonates with TSB >10 mg/dL had a BC value ≤7 mg/dL, BC values in this range accounted for only 8.7% of the comparisons. Thus, if this cutoff were used, additional testing would be required in approximately 90% of the cases. When a BC cutoff >8 mg/dL was examined, TSB was >10 mg/dL in 4 of 41 comparisons in which BC was equal to or below the cutoff value. Furthermore, additional testing would be required in the other 294 cases (335 − 41; 88%) in which BC was >8 mg/dL.
Predictive indices of various BC cutoff values for TSB >15 mg/dL in Hispanic neonates are shown in Table 4, and the ROC curve is plotted in Fig 4B. As sensitivity decreases with increasing BC cutoff values, the ability to be certain that TSB is not >15 mg/dL also decreases. Although only 1 neonate with TSB >15 mg/dL had a BC value ≤8 mg/dL, BC values in this range accounted for only 11.9% of the comparisons. Thus, if this cutoff were used, additional testing would be required in approximately 88% of the cases. TSB was >15 mg/dL in 2 of 77 comparisons in which BC was ≤9 mg/dL. Additional testing would be required in the other 258 cases (335 − 77; 77%) in which BC was >9 mg/dL.
At least 2 factors have stimulated the recent renewed interest in hyperbilirubinemia in term and near-term neonates. First, reports of kernicterus in infants whose only risk factor was breastfeeding have challenged conventional wisdom regarding this potentially devastating problem.16 Second, early discharge from the hospital, at a time when serum bilirubin is still increasing in the vast majority of infants, essentially has transformed neonatal hyperbilirubinemia from an inpatient to an outpatient issue.1 Efforts such as that by Bhutani et al17 and others18 to develop risk profiles based on serum bilirubin values obtained at approximately 24 hours of age may represent important steps in improving the management of this difficult clinical problem. However, a recent multicenter study has raised questions about the usefulness of this approach.19,20
A major variable in the clinical management of neonatal hyperbilirubinemia is the well-documented unreliability of visual estimates of the degree of jaundice in an individual neonate.7 Thus, attempts have been made to provide other means to estimate noninvasively serum bilirubin, starting with the ictometer and progressing to the Minolta Air-Shields Jaundice Meter (Air-Shields, Hatboro, PA), a hand-held spectrophotometer.21–35 More sophisticated transcutaneous devices have been developed in recent years, including the TLC (Chromatics Color Sciences, Inc, New York, NY); the clinical usefulness of this method has been limited by the need to obtain a baseline measurement.7 The BC, which utilizes the principle of spectral subtraction,9,10 generally has been shown to agree with bilirubin measured by HPLC.11,12,36,37 A limitation of these studies has been the inclusion of relatively few Hispanic neonates and/or a small number of infants with TSB values ≥15 mg/dL.
The current study was conducted in a large NBN in which Hispanic infants are the predominant ethnic group, and a significant number of bilirubin values ≥15 mg/dL was anticipated. Transcutaneous measurements were performed by only 1 investigator, which helps to ensure continuity in technique; however, this could represent a source of error as well. The latter seems unlikely because the BC device is easy to operate. Similar to previous studies, a highly significant correlation between TSB values and estimates obtained with the BC was observed. However, as pointed out by Bland and Altman,38 correlation coefficients can be a poor indicator of the agreement between 2 clinical tests, and a much better measure of agreement is the plot of the differences of the values obtained by the 2 methods versus the means of the values. On analysis of our results using the Bradley-Blackwood test (which is based on the Bland-Altman concept), it was apparent that in the majority of comparisons, BC determinations underestimated TSB as determined in the laboratory. It also was clear that the difference between the 2 measuring techniques tended to increase in infants with relatively high TSB values.
Serum bilirubin determinations in the current study were performed in our clinical laboratory, whereas previous studies that evaluated BC used HPLC as the “gold standard.” However, Bhutani et al12 reported recently that agreement between BC values and TSB determined by a variety of laboratory methods was comparable to the agreement between BC and HPLC. Our laboratory performs approximately 5000 total bilirubin determinations per month and maintains rigorous quality control standards for bilirubin testing as required by the College of American Pathologists. In a limited comparison of our TSB method with the HPLC method, agreement was within 1.9 mg/dL in all cases, and the majority of the differences were <1 mg/dL. Our tendency to underestimate HPLC slightly suggests that the comments above regarding the underestimation of TSB by BC may be understated.
To our knowledge, this is the largest number of Hispanic neonates in which the BC technique has been evaluated, and racial differences could account for the less favorable results in our study compared with those in previous studies. Bhutani et al12 reported that results in Hispanic neonates were comparable to those obtained in other groups. However, in their study, only 17 (3.5%) of 490 study patients were Hispanic; in the recent study by Rubaltelli et al,37 only 6.7% of the patients were Hispanic. Furthermore, because of the scientific principles on which the BC device was designed, which include correction for melanin pigmentation, we believe that it is unlikely that ethnic considerations account for the differences in our results and those previously reported. A more likely explanation is our inclusion of a significant number of study patients with bilirubin values ≥15 mg/dL; 31% of the Hispanic neonates had a TSB ≥15 mg/dL in the current study versus 1.1% of the patients in the study reported by Bhutani et al.12
In determining the predictive indices for various BC cutoff values, we chose TSB levels of >10 mg/dL and >15 mg/dL because of their apparent clinical relevance. For example, a TSB value >10 mg/dL certainly would be cause for concern in a 24- to 36-hour-old infant being discharged from the hospital, whereas the higher value (>15 mg/dL) might be of more relevance in an older infant being seen in follow-up. As pointed out by Sackett et al,39 a very sensitive test will allow the clinician to rule out safely the condition of interest; thus, as shown in Tables 3 and 4, BC ≤7 mg/dL or ≤8 mg/dL effectively negated the possibility of TSB >10 mg/dL or >15 mg/dL, respectively, in the 2 analyses. The problem is that using BC cutoff levels in this relatively low range captured a relatively small percentage of patients (approximately 10%–25%); thus, additional testing would be necessary in the majority of infants. Furthermore, high values for sensitivity and negative predictive value (and low negative likelihood ratios) can be misleading because missing even a single infant with a high TSB is clinically unacceptable.
Transcutaneous bilirubin sampling is easy to perform and pain-free for the infant, and the test gives immediate results. Also, as Bhutani et al12 suggested, values obtained noninvasively at early discharge may assign infants with varying degrees of visible jaundice to risk categories, thus assisting the clinician in determining the need for additional follow-up. However, whatever advantages a new test may have, the clinician must be confident that the results are accurate, ie, good agreement with the established test. The findings of our study indicate that, for Hispanic neonates, and particularly those with relatively high bilirubin values, transcutaneous bilirubin determination by the BC method is not an appropriate substitute for serum bilirubin testing in the majority of instances. Additional study and perhaps modification in the design of this device may be necessary to establish its role in clinical practice.
This study was funded in part by Respironics, Inc.
We appreciate the excellent secretarial work of Marilyn Dixon and the assistance of Lani Martin, MT (ASCP), in obtaining quality control information regarding the Parkland Health and Hospital System Clinical Chemistry Laboratory.
- Received August 28, 2001.
- Accepted January 17, 2002.
- Reprint requests to (W.D.E.) University of Texas Southwestern Medical Center at Dallas, Department of Pediatrics, 5323 Harry Hines Blvd, Dallas, TX 75390-9063. Email:
Presented, in part, at the 2001 Pediatric Academic Societies Meeting, Baltimore, MD, April 30, 2001.
- ↵Brown AK, Johnson L. The Year Book of Neonatal and Perinatal Medicine. St Louis, MO: Mosby-Year Book, Inc; 1996
- ↵Making advances against jaundice in infant care: quality improvement in newborn care. MAJIC Newsletter 1998;1
- ↵Stevenson DK, Vreman HJ. Carbon monoxide and bilirubin production in neonates. Pediatrics.1997;100 :252– 254
- ↵Stevenson DK, Fanaroff AA, Maisels MJ, et al. The incidence of increased bilirubin production in hyperbilirubinemic neonates as indexed by breath carbon monoxide (CO). Presented at the First International Symposium on Heme Oxygenase and Carbon Monoxide; July 14–17, 2000; New York, NY
- ↵Maisels MJ, Engle WD, Jackson GL, et al. Noninvasive measurement of serum bilirubin [abstract]. Pediatr Res.1999;45 :209A
- ↵Tayaba RD, Gribetz D, Gribetz I, et al. Noninvasive estimation of serum bilirubin. Pediatrics.1998;102 :1– 5
- ↵Jacques SL, Saidi I, Ladner A, et al. Developing an optical fiber reflectance spectrometer to monitor bilirubinemia in neonates. Society of Photo-Optical Instrumentation Engineers Proceedings.1997;2975 :115– 124
- ↵Jacques SL. Reflectance spectroscopy with optical fiber devices and transcutaneous bilirubinometers. In: Verga Scheggi AM, Martelluci S, Chester AN, Pratesi R, eds. Biomedical Optical Instrumentation and Laser-Assisted Biotechnology. Netherlands: Kluwer Academic Publishers; 1996:83–94
- ↵Bhutani V, Johnson L, Gourley G, et al. Non-invasive measurement of total serum bilirubin by multi-wavelength spectral reflectance by BiliCheck in newborn patients [abstract]. Pediatr Res.1998;43 :167A
- ↵Bhutani VK, Gourley GR, Adler S, et al. Noninvasive measurement of total serum bilirubin in a multiracial predischarge newborn population to assess the risk of severe hyperbilirubinemia. Pediatrics.2000;106(2) . Available at: http://www.pediatrics.org/cgi/content/full/106/2/e17
- ↵Maisels MJ, Newman TB. Kernicterus in otherwise healthy, breast-fed term newborns. Pediatrics.1995;96 :730– 733
- ↵Bhutani VK, Johnson L, Sivieri M. Predictive ability of a predischarge hour-specific serum bilirubin for subsequent significant hyperbilirubinemia in healthy term and near-term newborns. Pediatrics.1999;103 :6– 14
- ↵Alpay F, Sarici SÜ, Tosuncuk HD, et al. The value of first-day bilirubin measurement in predicting the development of significant hyperbilirubinemia in healthy term newborns. Pediatrics.2000;106 :1– 5
- ↵Stevenson DK, Fanaroff AA, Maisels MJ, et al. Prediction of hyperbilirubinemia in near-term and term infants. Pediatrics.2001;108 :31– 39
- ↵Valaes T. Problems with prediction of neonatal hyperbilirubinemia. Pediatrics.2001;108 :175– 177
- ↵Maisels MJ, Conrad S. Transcutaneous bilirubin measurements in full-term infants. Pediatrics.1982;70 :464– 467
- Engel RR. Nuances or nuisances for cutaneous bilirubinometry? Pediatrics.1982;69 :126– 127
- Hegyi, T. Transcutaneous bilirubinometry: a new light on an old subject. Pediatrics.1982;69 :124– 125
- Hannemann RE, Schreiner RL, DeWitt DP, et al. Evaluation of the Minolta bilirubin meter as a screening device in white and black infants. Pediatrics.1982;69 :107– 109
- Schumacher RE, Thornbery JM, Gutcher GR. Transcutaneous bilirubinometry: a comparison of old and new methods. Pediatrics.1985;76 :10– 14
- Yamanouchi I, Yamauchi Y, Igarashi I. Transcutaneous bilirubinometry: preliminary studies of noninvasive transcutaneous bilirubin meter in the Okayama National Hospital. Pediatrics.1980;65 :195– 202
- Goldman SL, Penalver A, Penaranda R. Unreliability of the transcutaneous bilirubinometer in the newborn [abstract]. Clin Res.1981;29 :896A
- ↵Smith DW, Inguillo D, Martin D, et al. Use of noninvasive tests to predict significant jaundice in full-term infants: preliminary studies. Pediatrics.1985;75 :278– 280
- ↵Gourley G, Bhutani V, Johnson L, et al. Measurement of serum bilirubin in newborn infants: common clinical laboratory methods versus high performance liquid chromatography (HPLC) [abstract]. Pediatr Res.1999;45 :283
- ↵Rubaltelli FF, Gourley GR, Loskamp N, et al. Transcutaneous bilirubin measurement: a multicenter evaluation of a new device. Pediatrics.2001;107 :1264– 1271
- ↵Sackett DL, Richardson WS, Rosenberg W, Haynes RB. Evidence-Based Medicine: How to Practice and Teach EBM. London, England: Churchill Livingstone; 1998:121
- Copyright © 2002 by the American Academy of Pediatrics