Transcutaneous Bilirubin Accuracy Before, During, and After Phototherapy: A Meta-Analysis

This systematic review and meta-analysis was registered in PROSPERO, an international prospective register of systematic reviews (CRD42022361932).¹³  This study was performed and reported using the reporting guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses and Meta-analysis of Observational Studies in Epidemiology.¹⁴ 

The literature search was performed in the electronic database PubMed (search string: “[bilirubinometry OR bilirubinometer OR transcutaneous] AND [hyperbilirubinaemie OR hyperbilirubinemia OR jaundice OR bilirubin] AND [neonatal OR newborn]”) and Cochrane (search string: hyperbilirubinemia) in December 2021 and the search was updated in July 2022. References of included studies were checked. We restricted our search to studies reported in English, German, and Dutch languages.

Prospective and retrospective cohort and cross-sectional observational studies evaluating the agreement between paired measurements of TcB and TSB in neonates across all gestational ages ≤28 days postpartum were eligible for inclusion. The following transcutaneous bilirubinometers were assessed as eligible: JM-103/105, BiliChek, Bilicare, MBJ-20, JH20-1C and KJ-8000. Solely studies using Bland-Altman analysis, including a Bland-Altman graph providing an MD of TcB and TSB measurements and SD or LoA, were included.

Two review authors (L.K., T.O.) independently screened all titles and abstracts identified by our search strategy for relevance to this systematic review. Disagreements were solved in consensus. Full-text articles were retrieved and independently assessed by 2 review authors on the above-mentioned predefined eligibility criteria (L.K., T.O.). Disagreements were solved through discussion with a third review author (J.B.).

Two review authors (L.K., T.O.) extracted data independently from the selected articles: Type of study (prospective, retrospective, cross-sectional); study size; number of paired measurements; gestational age; inclusion and exclusion criteria; the use of phototherapy on covered and/or uncovered skin; type of TcB device and location of measurement; method of TSB measurement; time between TcB and TSB measurement; MD (TcB − TSB) and SD, whether calculated from the LoA (LoA = MD ± 1,96 x SD). We conversed all data to μmol/L (1 mg/dL = 17.1 μmol/L). Data for meta-analysis were selected if they were clearly defined to the groups before, during, or after the application of phototherapy. After phototherapy was defined as at least 8 hours after discontinuing phototherapy.

Quality assessment was performed using the tool Quality Assessment of Diagnostic Accuracy Studies-2 by 2 review authors (L.K., J.B.).¹⁵  We evaluated the studies for patient selection, index test, reference standard, and flow and timing. For reference standard, the well-known TSB methods diazo, direct spectrophotometry, and high-performance liquid chromatography were described as low risk.¹⁶  In the section Flow and Timing, an interval <60 minutes between TcB and TSB measurements was described as appropriate. Uncertainties were solved in consensus.

To obtain a pooled MD and SD, we used the Mantel-Haenszel weighted approach described by Williamson et al: A statistical method for pooling Bland-Altman data that was used before in several systematic reviews assessing the agreement of 2 continuous measurements (Supplemental Fig 3).^17–19  The meta-analysis was executed in Excel. Forest plots were created using R software. We analyzed data before, during (covered and uncovered), and after (covered and uncovered) phototherapy separately. We hypothesized a priori the following sources for clinical heterogeneity and performed subgroup analyses accordingly: Covering of skin during phototherapy, location of TcB measurement, gestational age (<37 weeks and ≥37 weeks), and type of TcB device. We considered an MD between TcB and TSB ≥ ± 50 μmol/L (∼3 mg/dL) clinically relevant.⁷ 

The study selection is illustrated in Fig 1. The literature search identified 550 records, of which 201 studies potentially qualified for inclusion on the basis of screening of titles and abstracts. The 201 full-text records were screened and 67 studies met all inclusion criteria. Of these 67 studies, 54 studies reported results separately to a group before (48 studies), during (respectively, 14 and 13 studies in covered and uncovered skin), and after phototherapy (respectively, 5 and 2 studies in covered and uncovered skin), and were selected for final inclusion and analysis.

The study characteristics of the 54 included studies are reported in Table 1, including a total of 17 236 patients and 23 882 paired measurements. We cannot exclude that this number of paired measurements may not be completely correct because several studies compared >1 TcB measurement; for example, different locations or different kinds of TcB devices to 1 TSB measurement.^20–34  In most studies, TcB was measured on the forehead (37 of 54 studies; 68.5%) or sternum (30 of 54 studies; 55.5%), some in combination with other more caudal locations. In 1 study, only the hipbone was used,³⁵  in another the ear pinna,³⁶  and 1 study did not mention the location.³⁷  Six of 54 studies (11.1%) included only term infants (≥37 weeks of pregnancy) and 13 of 54 studies (24.1%) included only preterm infants. The other studies included both term and preterm neonates (33 of 54 studies; 61.1%) or didn’t report gestational age (n = 2). The most frequent reported TcB device was JM-103 (26 of 54 studies, 48.1%) and BiliChek (22 of 54 studies, 40.7%), followed by JM-105 (11 of 54, 20.4%). MBJ-20 and BiliCare were used in 2 studies, and JHC20-1C in 1. Excluded studies with reason for exclusion are presented in Supplemental Table 4.

Quality assessment according to Quality Assessment of Diagnostic Accuracy Studies-2 is summarized in Table 2. In general, quality of the included studies was good and risk of bias in most studies was estimated low. We considered consecutive or convenience sampling both as low risk of bias in the domain of patient selection. We assessed retrospective studies as unclear risk (8 of 54; 17.8%) because of the possibility of missing data. None of the studies had risk of bias in the domain of index test. In the domain reference test, 1 study (1.9%) was judged to have high risk of bias because the reference test was not based on 1 of the 3 well-known TSB methods.³⁸  Four of 54 studies (7.4%) had unclear risk of bias in this domain because the reference method was not mentioned.^22,24,25,39  Concerning flow and timing, 4 of 54 studies (7.4%) had high risk of bias because the time between TcB and TSB measurement was 60 minutes or more.^40–43  In another 4 studies, time between TcB and TSB measurement was not mentioned and risk of bias was therefore unclear.^28,38,44,45 

The pooled MD, pooled SD, and pooled 95% LoA before, during (covered and uncovered), and after (covered and uncovered) phototherapy including subgroup analyses for location of TcB measurement, gestational age, and device of the TcB measurement are listed in Table 3.¹⁷ Figures 2A–E depict the corresponding forest plots.

We identified 48 studies (18 630 paired measurements in 15 544 patients) applicable for meta-analysis before phototherapy, resulting in a pooled MD (TcB − TSB) of 2.5 μmol/L (SD 20.8; LoA −38.3 to 43.3 μmol/L). See Table 3 and Fig 2A.

Subgroup analysis revealed diminished agreement between TcB and TSB measurements, with LoA outside the safety margin of 50 μmol/L when TcB was measured on the sternum (MD 19.7 μmol/L; LoA −32.7 to 72.1 μmol/L) compared with TcB measurements on the forehead (MD −0.4 μmol/L; −45.5 to 44.8 μmol/L). In term infants, the LoA were just outside the 50 μmol/L range (−54.3 to 57.8 μmol/L), whereas in preterm infants, they were within this range (−43.1 to 38.5 μmol/L). Subgroup analyses comparing different TcB devices revealed adequate LoA, except for JM-105 (−39.2 to 65.7 μmol/L), MBJ-20 (−41.9 to 60.8 μmol/L), and JH20-1C (−63.3 to 60.3 μmol/L).

We identified 14 studies (2355 paired measurements in 1613 patients) applicable for analysis during phototherapy on covered skin, with a pooled MD of −0.3 μmol/L (SD 17.6 μmol/L; LoA −34.8 to 34.2 μmol/L). The forest plot is presented in Fig 2B.

Subgroup analysis of studies reporting comparisons of TcB and TSB measured during phototherapy on the covered sternum showed LoA outside the 50 μmol/L range (−67.7 to 69.3 μmol/L) compared with LoA well within this range when TcB was measured on the forehead (−21.0 to 32.9 μmol/L). In term infants, the LoA of TcB measurements during phototherapy on covered skin were marginally outside the 50 μmol/L range (−51.4 to 45.9 μmol/L) and within this range in preterm infants (−27.2 to 25.8 μmol/L). Subgroup analysis within the category during phototherapy on covered skin per TcB devices all showed LoA agreement outside the 50 μmol/L (Table 3).

The 13 studies (2796 paired measurements in 1806 patients) applicable for analysis during phototherapy on uncovered skin showed large heterogeneity (Fig 2C). Although strictly pooling should not be done in case of substantial heterogeneity, the calculated pooled MD was −28.6 μmol/L (SD 39.3 μmol/L; LoA −105.7 to 48.5 μmol/L).

All subgroup analyses including TcB measurements during phototherapy on uncovered skin gave LoA outside the 50 μmol/L safety margin (Table 3).

We pooled 5 studies (648 paired measurements in 555 patients) applicable for analysis after phototherapy on covered skin, resulting in a pooled MD of −34.3 μmol/L (SD 26.7 μmol/L; LoA −86.7 to 18.1 μmol/L). In this category, we observed an outlier affecting the results clearly in case of exclusion (MD −1.4 μmol/L; SD 28.7 μmol/L; LoA −54.9 to 57.7 μmol/L).³⁵  This specific study measured the TcB on the hipbone covered by a diaper.

The 2 studies (142 measurements in 142 patients) applicable for analysis after phototherapy on uncovered skin revealed a pooled MD of −21.1 μmol/L (SD 34.4 μmol/L; LoA −88.6 to 46.4 μmol/L). All LoA of the different subgroup analyses in this category showed LoA outside the 50 μmol/L margin (Table 3).

All measurements after phototherapy for both covered and uncovered were at least 8 hours after discontinuing phototherapy. Not enough data were available to access the agreement separately for timing after discontinuing phototherapy.

This systematic review included 54 studies with >17 000 term and preterm neonates reporting >22 000 paired measurements of TcB and TSB depicted in Bland-Altman plots. The results show good agreement before and during phototherapy, provided that TcB is measured on covered skin. The pooled LoA of ± 40.6 μmol/L, respectively, before and ± 34.5 μmol/L during phototherapy on covered skin are within the safety margin of 50 μmol/L. This safety margin of 50 μmol/L is in accordance with the current guidelines of the American Academy of Pediatrics, where TSB measurements are advised when TcB is within 3 mg/dL (∼50 μmol/L) from the phototherapy threshold.⁷ 

In contrast, the results after phototherapy on covered skin show that TcB underestimates TSB, with LoA >50 μmol/L. However, the number of studies after phototherapy on covered skin are relatively small, with only 5 studies reporting 648 paired measurements in 555 patients. Sensitivity analysis showed that 1 study acted as outlier with deviating results from the other 4 studies.³⁵  In this study, TcB was measured on the hipbone, which may explain the lower TcB values compared with measurements on the forehead or sternum because of cephalocaudal progression.⁴⁶  The role of the method used for skin covering may also have played a role in the underestimation of TcB in this study in which a diaper was used. We are not aware of studies comparing the effect of different methods of skin covering during phototherapy on TcB measurements; for example, between a specially designed photo-opaque patch (eg, BiliEclipse) and a diaper. However, Abdulkadir et al researched the irradiance under different kinds of eye protecting during phototherapy and found a difference in shielding from phototherapy with different kinds of material.⁴⁷  More studies investigating the agreement between cranial and sternal TcB and TSB measurements using different methods of covering are needed for this purpose. The results during, as well after, phototherapy on uncovered skin show that TcB underestimates TSB. A point to mention is that we did not account for the length of time after discontinuing phototherapy. Time of discontinuing phototherapy may have an influence on the difference between TcB on uncovered skin and TSB measurements. It is assumed that this difference will be negligible after 8 to 18 hours of discontinuing phototherapy.^48–50  Therefore, in several guidelines, the use of TcB measurements after phototherapy is restricted to at least 24 hours after discontinuation phototherapy.⁷  Because of lack of studies and original data, we were not able to provide subgroup analysis for this aspect.

Subgroup analysis for location of TcB measurement before phototherapy showed better agreement between TcB and TSB on the forehead compared with the sternum. However, because the sternal TcB measurements tend to overestimate the TSB before phototherapy, the use of sternal TcB measurements is still acceptable and safe at least. We found that TcB is reliable in term, as well as in preterm, infants, with LoA just outside the 50 μmol/L range in term infants and thus better agreement in preterm infants. Concerning the type of TcB device, BiliChek and JM-103 have smaller pooled MDs and SDs compared with JM-105. However, more inclusions are available with BiliChek and JM-103 in comparison with JM-105. For BiliCare, MBJ-20 and JH20-1C, only few studies were available.

Our study represents an up-to-date review of TcB agreements, with TSB using preferred methodology. One of the strengths of this systematic review is the extensive search of studies and the pooling of Bland-Altman statistics. Results of Bland-Altman analysis can easily be translated to the clinical practice and are therefore useful for formulating new policy.

Our study results are difficult to compare with the results of the other already existing systematic reviews of TcB measurements during phototherapy because of different statistical methods used for pooling. As described above, correlation coefficients indicate the strength of a linear relationship and not the agreement of 2 variables. A high correlation does not imply good agreement between 2 variables.

Our systematic review has some limitations. First, our literature search was limited to English, German, and Dutch languages, leading to the exclusions of 9 possible relevant studies. Secondly, several of the included studies have >1 TcB measurement at 1 moment of paired measurement, mostly TcB measurements at different locations or with different kinds of TcB devices compared with 1 serum bilirubin measurement.^20–34  Also, several of the included studies have repeated paired measurements per neonate.^{23,28–30,33,35,38,39,44,45,50–63}  In earlier literature, it is postulated that TcB values in the higher ranges (above 256 μmol/L [3 mg/dL]) are not reliable.⁷  We were not able because of lack of original data to separately assess the agreement in the higher ranges. Furthermore, despite save recommendations in the American Academy of Pediatrics, the influence of the amount of skin melanin and therefore different skin color stays controversial. In our review, we could not address the influence of skin color. Most inclusions were neonates from Caucasic and Asian origin, and most studies with mixed inclusions did not deliver data separately for different skin colors.

In conclusion, this systematic review shows that TcB measurements before and during phototherapy on covered skin show acceptable accuracy compared with TSB measurements in both term and preterm newborns. This suggests that, in addition to the use of TcB for screening of hyperbilirubinemia before the application of phototherapy, the use of TcB may be extended to the use during phototherapy, provided it is measured on covered skin. Evidence of the best method of skin covering does not exist yet, but the use of photo-opaque patches can be recommended. TSB measurements are advised when TcB measurements are above or below but within 50 μmol/L of the phototherapy threshold.⁷  The forehead is the preferred location for measurement of TcB; however, the sternum is also safe to use. The use of TcB during phototherapy on uncovered skin is not advised. More studies are needed to evaluate the accuracy of TcB compared with TSB after the application of phototherapy on covered skin.

We thank Joyce Peper, MSc, for her contribution.

LoA

limits of agreement

MD

mean difference

TcB

transcutaneous bilirubin

TSB

total serum bilirubin