Page 106 - 16Neonatal Jaundice_compressed

Recognition

Recommendations
See the end of Section 5.2.
5.2.2 Transcutaneous bilirubinometers

Review of devices

Three devices are reviewed in this section. Since a large number of studies were identified that
evaluated transcutaneous bilirubinometers, it was decided to include studies with EL II or above
only. A large number of studies reported the sensitivity and specificity based on the ROC curves
without specifying exact values and therefore preference was given to those studies which
reported sufficient data for meta-analysis.
Description of included studies – Minolta JM-102

66
70
Seven studies 26;46;66-70 are included in this section – one each from Denmark, Turkey, the
46
68
26
69
UK, Spain, Saudi Arabia, the USA and Taiwan. The sample populations comprised term
67
babies in six studies, 26;66-70 while in the seventh study both term and near-term babies
46
> 34 weeks of gestational age were included. Five of the studies 26;46;66;68;69 are of EL II quality
with blinding not reported in most while two 67;70 are EL III. Exclusion criteria were not defined
in three studies. 26;68;70 Transcutaneous bilirubin levels were measured on the forehead in all
studies, while in two studies readings were also taken from the sternum and reported separately.
Although all studies reported diagnostic accuracy in terms of correlation coefficient and six
studies reported on sensitivity/specificity of the test for different thresholds, only four studies
gave sufficient data to be used for meta-analysis.
Review findings – Minolta JM-102
The sample size in the studies ranged from 76 to 2004. There was a statistically significant
positive linear correlation between the transcutaneous bilirubin reading at the forehead and
serum bilirubin levels in all the studies. The correlation coefficients ranged from 0.76 to 0.93.

In the two studies 26;46 for which detailed data were not available for meta-analysis, sensitivity
and specificity were reported separately. One study showed transcutaneous bilirubin (Minolta
46
JM-102 threshold value 19.9 reflectance units) to have a sensitivity of 86% and specificity of
78% for detecting serum bilirubin levels > 249 micromol/litre, while the other study reported
26
98% sensitivity and 72% specificity for detecting serum bilirubin levels > 222 micromol/litre.
Data from the other four studies 66-69 were pooled to examine the diagnostic accuracy of
transcutaneous bilirubin readings (with various thresholds) with the Minolta JM-102 in detecting
serum bilirubin levels > 220 micromol/litre in term babies. The pooled sensitivity was 85%
(95% CI 76% to 91%) and the pooled specificity was 83% (95% CI 79% to 86%) but there was
strong evidence of statistical heterogeneity for both results (I² = 78.5% and 92.8% for sensitivity
and specificity, respectively) (Figures 5.1 and 5.2). In the summary ROC curve, AROC was 0.93
but a threshold effect could not be seen, indicating further evidence of heterogeneity among the
included studies (Figure 5.3).

Evidence summary – Minolta JM-102
Results from EL II studies show great variation in the accuracy of transcutaneous bilirubin
measurement using the Minolta JM-102 at the forehead. The correlation coefficient ranged from
0.76 to 0.93 in the included studies. Meta-analysis showed transcutaneous bilirubin readings at
the forehead to have an AROC of 0.93 for the detection of serum bilirubin levels
> 220 micromol/litre in term babies. The pooled sensitivity was 85% and the pooled specificity
83%, but again these results were marred by strong evidence of statistical heterogeneity. The
studies on the Minolta JM-102 were confined to healthy term babies with light skin tones and
Chinese babies.

77