Page 108 - 16Neonatal Jaundice_compressed

Recognition

GDG translation from evidence – Minolta JM-102
Forehead measurement of transcutaneous bilirubin using the Minolta JM-102 is more accurate
than visual assessment for the recognition of jaundice in babies with light skin tones and in
those with yellow skin tones.
The Minolta JM-102 is no longer available for purchase from the manufacturers.

Recommendations
See the end of Section 5.2.

Description of included studies – Minolta JM-103
Of the six included studies in this section, three were conducted in the USA, 71-73 two in
75
73
Thailand 74;75 and one in Taiwan. The study population in one study from Thailand and one
76
from the USA comprised healthy preterm babies with gestational age < 36 weeks, while all the
other studies 71;72;74;76 included either term babies or both term and near-term babies. In the
71
studies from the USA, the population was multi-ethnic in one study, while in two 72;73 it was
predominantly Hispanic. No exclusion criteria were specified in two studies. 72;76
Transcutaneous bilirubin was measured at the forehead in three studies 72;74;76 in term babies and
75
in one study in preterm babies , while the sternum was used as the only site in two studies. 71;73
Detailed data for meta-analysis were available from three studies, 72;74;76 but they all reported
different thresholds and thus a summary ROC was developed. All the studies are of EL II.
Review findings – Minolta JM-103

The sample size in the studies in term babies ranged from 90 to 849 babies, while there were
196 babies with mean birthweight of 1887 ± 344 g in one study on preterm babies and in the
75
other study of preterm babies the birthweight ranged from 370 g to 2989 g. All the studies
73
showed a statistically significant linear correlation between the transcutaneous bilirubin
observations and serum bilirubin levels. In the term babies, correlation coefficients ranged from
0.77 to 0.93 and one study from USAreported variable coefficients for different ethnicities:
71
71
0.95 for white babies, 0.82 for black babies and 0.92 for all other babies. This study also
reported the difference between the laboratory serum bilirubin levels and transcutaneous
bilirubin readings in different ethnicities. The results showed that transcutaneous bilirubin
values overestimated serum bilirubin levels by ≥ 51 micromol/litre in 17.4% of the black
babies compared with 2.0% of white babies and 3.3% of other babies. However, in the other
three studies in term babies, transcutaneous bilirubin readings were found to underestimate
serum bilirubin levels by a mean of 12, 17 and 27 micromol/litre. This discrepancy did not
74
72
76
75
increase with a rise in the serum bilirubin levels in two of the studies. 72;76 One study on
preterm babies reported a correlation coefficient of 0.79 and reported that the JM-103
overestimated serum bilirubin levels in the first 3–4 days of life but underestimated the serum
73
bilirubin after this age. The second study reported r = 0.92 for gestational age between 24 and
28 weeks, r = 0.91 for gestational age between 29 and 31 weeks and r = 0.82 between
73
gestational age 32 and 34 weeks. This study also noted that the JM-103 underestimated the
serum bilirubin by 19 ± 32 micromol/litre in babies with gestational age between 24 and
28 weeks, by 14 ± 22 micromol/litre in babies with gestational age between 28 and 31 weeks
and by 17 ± 27 micromol/litre in babies with gestational age between 32 and 34 weeks. Data
from three studies 72;74;76 in term babies were pooled to calculate the predictive accuracy of the
device in detecting serum bilirubin levels > 255 micromol/litre when transcutaneous bilirubin
was measured from the forehead with threshold level > 200–204 micromol/litre. The pooled
sensitivity and specificity were 85% (95% CI 78% to 91%) and 80% (95% CI 77% to 82%),
respectively (Figures 5.4 and 5.5). There was strong evidence of statistical heterogeneity for both
results (I² = 55% and 93% for sensitivity and specificity, respectively). The summary ROC curve
showed an AROC of 0.87 but there was variation in the individual study results and it showed
no indication of a threshold effect (Figure 5.6).

79