Accuracy of transcutaneous bilirubinometry in term infants after phototherapy: a prospective observational study

Abstract

Objective

To test the accuracy of transcutaneous bilirubinometry (TcB) in neonates 12 h after discontinuing phototherapy.

Study design

In a prospective study of 91 neonates at ≥35 weeks of gestation, paired measurements of total serum bilirubin (TSB) and TcB were obtained 12 h after discontinuation of phototherapy. TcB measurements were obtained on the uncovered skin of the sternum and the covered skin of the lower abdomen. Bland–Altman plots were used to evaluate agreement between TSB and TcB.

Results

TcB was found to systematically underestimate TSB on both covered and uncovered skin. The smallest but statistically significant difference between TSB and TcB was found on the covered lower abdomen (–1.03, p < .0001) compared with the uncovered skin of the sternum (–1.44, p < .0001). The correlation between TSB and TcB was excellent on both covered (r = 0.86, p < .001) and uncovered skin (r = 0.90, p < .001). Bland and Altman plots showed poor agreement between TcB and TSB.

Conclusions

This study demonstrated excellent correlation between TcB and TSB 12 h after phototherapy but poor TcB–TSB agreement. TcB cannot be reliably used in neonates exposed to phototherapy.

Keywords:

• Phototherapy
• transcutaneous bilirubin
• jaundice
• neonate
• bilirubin

Introduction

Neonatal jaundice is a common condition. It affects two-thirds of healthy infants. Extremely elevated levels of unconjugated bilirubin can result in neurological damage and kernicterus. Phototherapy is the treatment of choice for hyperbilirubinemia. One of the most common tests performed in neonatal units is the total serum bilirubin (TSB) performed by heel prick, which is considered painful.

Transcutaneous measurement of bilirubin concentration is a validated and recommended screening tool to assess bilirubin levels before phototherapy in infants more than 35 weeks of gestation [1]. Transcutaneous bilirubinometry (TcB) uses a handheld, noninvasive device to estimate serum bilirubin levels in infants. It is quick, painless, and does not require a blood test. TcB is considered unreliable for the assessment of recurrent hyperbilirubinemia in infants during and after phototherapy, because phototherapy blanches the skin [2]. If validated, TcB use after phototherapy could lead to less blood-draws, less pain for the neonate, reduced costs to the hospital and shorter hospital admissions since TcB is more readily available in the community. In addition, there is a growing trend for home phototherapy in low-risk infants [3–5] where performing a follow-up TSB is logistically difficult. Investigating the diagnostic accuracy of TcB in the post-phototherapy phase is therefore clinically useful.

An improvement in TcB–TSB correlation following phototherapy has been noted, although current data are inconclusive or inconsistent [2]. Some studies included a limited number of patients [6,7]. Others found a low correlation between TcB and TSB <12 h after phototherapy (r = 0.4–0.5) and suggested further study on temporal changes in TcB–TSB agreement during the post phototherapy phase [8]. Skin shielded from the effects of phototherapy may be able to predict TSB with sufficient reliability [7,9–13].

The accuracy of TcB measurements on covered and uncovered skin at 12 h after phototherapy is not known. The aim of the present prospective study was to determine the accuracy of TcB measurements on covered and uncovered skin 12 h after phototherapy in term/near infants.

Methods

This single-center prospective observational study was performed in the special-care nursery at Frankston Hospital, Victoria, Australia. Ethics approval was obtained from the Peninsula Health Human Research Ethics Committee, which is registered with the National Health and Medical Research Council (HREC/18/PH/33). This study was performed at the Frankston Hospital Special Care Nursery and Paediatric Ward. Informed written consent from the guardian or primary caregiver was obtained from all participants.

Study population

The inclusion criteria were gestational age of 35 weeks and above, and requirement of phototherapy for the treatment of hyperbilirubinemia. The study infants were born at Frankston Hospital between December 2018 and September 2020.

The exclusion criteria included infants requiring exchange transfusion, infants with edema and poor peripheral circulation, infants with perinatal infection and infants with major congenital malformations.

The requirement for phototherapy was based upon the National Institute for Health and Care Excellence (NICE) guidelines [14]. Infants were treated with continuous phototherapy. Both eyes were always protected, and nappies were always applied to well infants during phototherapy. The treatment was interrupted for feeding. Transcutaneous bilirubin was measured using the Konica-Minolta Air Shield JM-103 (Drager Medical Inc., Telford, PA). Twelve hours after the discontinuation of phototherapy, TcB and TSB were performed. Blood samples for TSB were obtained by heel prick and measured in a clinical laboratory using direct spectrophotometry (Beckman Coulter, O'Callaghansmills, Ireland). Transcutaneous measurements were performed within 10 min of blood sampling. TcB was measured at the uncovered mid-sternum and the covered lower abdomen, which was shielded by a nappy.

Statistical analyses

The required sample size was determined using acceptable limits of agreement on Bland–Altman analysis, via Lu et al.’s method [15]. Using pooled mean differences for the JM-103 device published in a prior systematic review (mean difference = 0.18 mg/dL, standard deviation (SD) of differences = 1.42) [2], and a maximum allowable difference of 3.7 mg/dL, a sample of 90 patients could estimate limits of agreement within these parameters, with 80% power.

Data are presented descriptively using means, medians (and interquartile range (IQR) and SDs) or percentages. Our design was informed by Newson’s suggested approach to method comparison, and by the approach of Grabenhenrich et al. [6,16]. The average difference between TSB and TcB was computed, along with 95% confidence intervals. The IQR of these differences is also reported to transparently reflect the range of differences observed in our study population.

Correlations are reported using Pearson’s correlation coefficient. As correlation reflects the relationship between one variable and another and not the differences between paired measurements, it is not recommended as a method for assessing the agreement between two measures [17]. We investigated the reliability of TcB as a representation of TSB via the estimation of mean differences, and also via the intra-class correlation (ICC). To compute the ICC, we used a linear mixed model with the subject as a random effect (i.e. we reported the ICC for agreement with systematic differences between methods incorporated into the error variability) [18].

Even reliable and highly correlated measures may be systematically biased. Bias and scale differentials were investigated via Bland–Altman plots [16]. It was suspected that the reliability of TcB may vary by ethnicity, or by the duration of phototherapy. To investigate this hypothesis, exploratory univariable analysis of the difference between measures was conducted using the t-test, and a multivariable test was performed via linear regression.

All data were analyzed in Stata, v16 and power calculation was conducted in R (Version 0.1.0.) using the blandPower package.

Results

Ninety-one term and near-term infants with jaundice who received phototherapy were enrolled in this study. Clinical characteristics of the enrolled infants are presented in Table 1.

Table 1. Baseline population characteristics.

Mean gestational age at birth, weeks (SD)	36.5 (1.2)
Mean postnatal age, hours (SD)	84 (34.7)
Mean birth weight, g (SD)	2821 (596)
Female sex (%)	39 (43%)
Ethnicity (%)
Caucasian	81 (89%)
Indian	3 (3.3%)
Aboriginal and Torres Strait Islander	2 (2.2%)
Asian	3 (3.3%)
Greek	1 (1.1%)
South American	1 (1.1%)
Length of phototherapy, hours (SD)	20 (9.5)
Mean TSB, mg/dL (95% CI)	8.82 (95% CI: 8.26–9.38)

Data are expressed as mean (SD), or number (%), as appropriate.

Birth weights ranged from 1676 g to 4810 g (mean ± SD, 2821 ± 596), gestational ages at birth ranged from 35 to 41 weeks (mean ± SD, 36.5 ± 1.2), and postnatal ages ranged from 24 to 192 h (mean ± SD, 84 ± 34.7). Most study participants were Caucasian (89%).

Table 2 shows the mean of the differences between TcB and TSB. On average, TcB underestimated TSB on covered skin by 1.03 mg/dL (95% CI: –1.36, −0.70), and on uncovered skin by 1.44 mg/dL (95% CI: −1.70, −1.19). Differences between TcB and TSB were statistically significant at both sites (p < .0001).

Table 2. Mean of the differences between TCB and TSB on covered skin and uncovered skin.

	Covered skin	95% CI	Uncovered skin	95% CI
Mean TCB	7.79	(7.15–8.42)	7.37	(6.80–7.95)
Mean difference (TCB–TSB)	−1.03	(–1.36, −0.70)	−1.44	(–1.70, −1.19)
IQR for difference (TCB–TSB)	−2.05, +0.12	–	−2.11, −0.64	–
Rate of overestimation^a		–		–
<2 mg/dL	23% (21/91)		12% (11/91)
≥2 mg/dL	2% (2/91)		0 (0/91)
≥3 mg/dL	0 (0/91)		0 (0/91)
≥4 mg/dL	0 (0/91)		0 (0/91)
Rate of underestimation^b
<2 mg/dL	47% (43/91)		59% (54/91)
≥2 mg/dL	15% (14/91)	–	23% [21/91]
≥3 mg/dL	8% (7/91)		3% [3/91]	–
≥4 mg/dL	3% (3/91)		2% [2/91]
Pearson’s r (correlation with TSB)	0.86		0.90
ICC	0.79	(0.71–0.86)	0.78	(0.69–0.85)

In this study population, mean TSB was 8.82 mg/dL (95% CI: 8.26–9.38).

^a % of infants for whom TcB > TSB.

^b % of infants for whom TcB < TSB.

None of the infants in our study met the treatment threshold for phototherapy at 12 h post phototherapy. The correlation between TSB and TcB was excellent on both covered skin (r = 0.86, p < .001) and uncovered skin (r = 0.90, p < .001).

The ICC analysis reflected good reliability at the covered abdomen (ICC 0.79 (95% CI 0.71–0.86)) and at the uncovered sternum (ICC 0.78 (95% CI 0.69–0.85)) [18].

Bland–Altman plots (Figures 1 and 2) suggest that the difference between TcB and TSB is unrelated to the magnitude of both measures. Bland–Altman analysis estimated that, on average, 95% of differences between TcB and TSB fell between +2.13 mg/dL and −4.18 mg/dL for covered skin, and +0.98 mg/dL and −3.87 mg/dL for uncovered skin. The limits of agreement were wider for the covered abdomen than the uncovered sternum. Table 2 reports the extent to which TcB overestimates and underestimates TSB.

Figure 1. Bland–Altman plots depicting agreement between TcB and TSB on uncovered skin.

Figure 2. Bland–Altman plots depicting agreement between TcB and TSB on covered skin.

Univariable analysis with t-tests did not demonstrate a statistically significant difference in TcB–TSB accuracy at either site for Caucasian babies compared with other babies; however, this study was insufficiently powered to analyze the relationship between skin tone and the accuracy of TcB after phototherapy. Multivariable linear regression did not demonstrate a relationship between phototherapy duration and TcB–TSB correlation.

Discussion

Infants who have received phototherapy require close monitoring to assess response to treatment. The American Academy of Pediatrics (AAP) guidelines recommend that a repeat bilirubin is needed after phototherapy if there are risk factors for rebound hyperbilirubinemia such as phototherapy treatment before 48 h of age, a positive DAT or suspected or known hemolytic disease [1]. The NICE guidelines suggest checking for rebound hyperbilirubinemia, with a repeat TSB 12–18 h after stopping phototherapy in all babies [14]. In practice, many health care facilities routinely test for rebound hyperbilirubinemia. TcB is widely used because of its ease of use, noninvasive nature, and immediate results. TcB is considered unreliable after phototherapy but if validated, would result in improved quality of care for infants. This study provides data on TcB measured using the Konica-Minolta Air Shield JM-103 TcB device on covered and uncovered skin in term and near-term infants in a predominantly Caucasian population 12 h after phototherapy.

The correlation coefficients between TSB and TcB on covered (r = 0.86) and uncovered skin (r = 0.90) found in this study are comparable to pre-phototherapy coefficients (i.e. pre-phototherapy coefficients typically ranged between 0.85 and 0.94) [19–23]. Tan and Dong [24] report TcB–TSB correlations of r = 0.80 on the uncovered chest and r = 0.70 on the covered forehead 18–24 h after cessation of phototherapy. Our study found a higher TcB–TSB correlation at a shorter post-therapy interval. Juster-Reicher et al. [25] reported that the TSB–TcB correlation is low on the exposed sternum in the first 8 h (r = 0.56) but increases between 9 and 16 h after phototherapy (r = 0.73), consistent with our findings.

Analyses of the Bland–Altman difference plots demonstrate that TcB underestimates TSB at both sites. For TcB to replace TSB in clinical practice after treatment with phototherapy, it must agree as closely as possible with TSB. It is difficult to identify a consensus regarding how close the limits of agreement ought to be to allow adoption into clinical practice. A value of ±3 mg/dL has been found to be acceptable in pre-phototherapy studies. In the present study, limits of agreement varied from +1 mg/dL to −3.9 mg/dL on uncovered skin and +2.1 mg/dL to −4.2 mg/dL on covered skin. TcB is generally within 3 mg/dL of TSB, where TSB concentrations are ≤15 mg/dL (94% in TcB on uncovered skin and 87% in TcB on covered skin). While improved precision was observed on the uncovered sternum, the limits of agreement are still relatively wide and fall outside the value of ±3 mg/dL, precluding its use in a clinical context. Other studies investigating accuracy of TcB in the post-phototherapy have reported correlation coefficients or mean differences only [7,24,25]. One study that presented Bland–Altman plots for TcB–TSB differences 24 h after phototherapy found limits of agreement that were similar to our study [26].

TcB at both covered and uncovered skin underestimated TSB and the difference was statistically significant. This could be because skin bilirubin concentrations decrease more rapidly than serum levels, resulting in a large bilirubin gradient between skin and blood [12]. The decrease in TcB at both covered and uncovered areas can be explained as follows. Bilirubin moves from plasma to the skin and subcutaneous tissues where phototherapy converts toxic unconjugated bilirubin into photoisomers by active photoisomerization [27]. Photoisomers then shift into the circulation, resulting in skin blanching, and are excreted in urine and the gut [27]. This movement occurs by diffusion, which is rapid and bidirectional. The departed photoisomers from the skin are replaced by more bilirubin from the circulation [12]. Covered skin is less affected by light exposure; therefore, dermal bilirubin concentrations fall at a slower rate over time. Bilirubin in the extravascular space of covered skin falls by diffusing into the intravascular space. In light-exposed skin, there is a more rapid decrease in dermal bilirubin levels [27].

Our study did not demonstrate improved reliability of TcB measurements on covered skin. Consistent with other studies, mean TcB in covered skin was higher (7.79, 95% CI 7.15–8.42) than covered skin (7.37, 95% CI 6.80–7.95) and demonstrated a smaller yet statistically significant mean TcB–TSB difference (–1.03) when compared to uncovered skin (–1.44) [7,8,27]. Skin sites not exposed to phototherapy have been previously found to demonstrate better agreement with TSB values during phototherapy. Two post-phototherapy studies found that TcB on forehead skin covered by a photo-opaque patch more closely approximated TSB as analyzed by correlation coefficients and analysis of mean differences [7,24]. There may be several reasons why these findings were not replicated in our study. First, nappies may be unable to completely shield underlying skin from the effect of phototherapy. Second, the lower abdomen is not the manufacturer’s recommended measurement site for TcB measurements by the JM-103 device and differences in dermal thickness and subcutaneous tissue are likely contributory. Lastly, neonatal jaundice is observed to progress in a cephalocaudal pattern with some cautioning against TcB measurements caudal from the sternum due to decreased reliability [28]. A falsely low TcB could result in failure to treat an infant who meets the criteria for phototherapy [29]

Our study, like many similar studies, has analyzed the relationship between TcB and a defined “standard” for bilirubin testing performed by one of many routine laboratory methods. However, applicability of these findings in other settings is contingent on two factors. First, significant interlaboratory variability in routine laboratory methods for TSB measurement is well documented [30]. Second, it is important to understand the nuances behind the term “accuracy” where TcB is being used as an estimate or proxy for TSB. While TSB measures intravascular bilirubin concentration, TcB measures extravascular bilirubin concentration. A serum bilirubin includes both toxic natural bilirubin and less toxic photoisomers [12]. TcB is therefore likely to better reflect the risk of kernicterus [31].

In summary, this study found that TcB performed 12 h after phototherapy on the uncovered sternum and covered abdomen demonstrates excellent correlation with TSB in contrast to previous studies but poor agreement. Our data suggest that TcB 12 h after phototherapy in uncomplicated jaundice in term and near-term infants is not a reliable substitute. Further large multicenter trials that investigate a different anatomical site for shielding skin, a different transcutaneous bilirubinometer and an alternative time interval after discontinuing phototherapy are needed.

Limitations

Our study had some limitations. A single brand of transcutaneous bilirubinometer was used, and the results may not be applicable to different brands of devices. The interaction between skin color and accuracy of TcB measurements could not be analyzed due to small subgroup sample sizes. Further studies in different settings are needed, as race-based variations in TcB accuracy prior to phototherapy have been reported [1,32,33]. While our cross-sectional study suggests no impact of phototherapy duration on TSB–TcB correlation, our study design could not rigorously assess this relationship, which may require serial measurements in the same infant. None of the participants in our study had TSB values ≥15 mg/dL, and the accuracy in this important group is unknown.

Acknowledgements

We thank Ms Alison Conroy-Joyce and her team of nurses on the wards for their enthusiastic participation in this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are openly available in figshare at http://10.6084/m9.figshare.23301335 [doi].

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.