Professional Documents
Culture Documents
E-Mail karger@karger.com
FR–38000 Grenoble (France)
www.karger.com/neo
E-Mail tdebillon @ chu-grenoble.fr
Introduction Grenoble Alpes and Saint-Denis de la Réunion, France, and the
Regional Hospital of Chambery, France.
Preterm babies <30 weeks’ GA, born between March 2013 and
Neonatal jaundice is an extremely frequent [1, 2] and August 2014 in these 3 hospitals, with one or more TSB determina-
potentially dangerous condition for preterm babies [3]. tions during the first 15 days of life, were included in this study.
Accurate determination of bilirubin levels is therefore es- Patients with conjugated hyperbilirubinemia were excluded. Ba-
sential for diagnosis and therapy. The reference method bies undergoing phototherapy or who had already received photo-
is the determination of total serum bilirubin (TSB). As therapy were not excluded.
Decisions regarding hyperbilirubinemia follow-up and treat-
jaundice is often prolonged and recurrent in preterm in- ment were taken by the NICU team, according to institution pro-
fants, this leads to repeated sampling, which can prove tocols based on the ‘National Institute for Health and Clinical Ex-
painful and time-consuming, and may lead to significant cellence’ (NICE) TSB normograms [15].
blood loss, especially in very-low-birth-weight infants.
Transcutaneous bilirubin (TcB) measurement has TcB and TSB Measurements
Our gold standard was the TSB measurement, determined ac-
been shown to be a convenient noninvasive, rapid, innoc- cording to the diazo reaction, which is currently the reference
uous, and reliable method for TSB estimation in term and method used in French laboratories. Significant hyperbilirubine-
near-term neonates [4, 5]. However, studies carried out in mia was defined by any TSB value that exceeded the hour-specific
preterm neonates have yielded varying results. TcB-TSB threshold for phototherapy. For each TSB quantification, a TcB
correlation seems to decrease when gestational age (GA) measurement was systematically immediately carried out by a
NICU nurse. The NICU clinicians, who are not used to monitoring
is low [6–9]. There are few studies of very preterm babies bilirubin levels through transcutaneous measurements, were
<30 weeks’ GA, and results have often proven contradic- blinded to each TcB result. Clinical decisions were therefore based
tory [10]. Only one study has exclusively explored this exclusively on bilirubinometry.
population, but the patient cohort was small (n = 24) [11]. The principle of the TcB measurement with BiliCheck® has
Preterm neonates requiring phototherapy also need been extensively described [11]. If the patient had received photo-
therapy within 12 h prior to the measurements, the scans were
frequent monitoring of bilirubin after treatment initia- performed between the eyebrows, on zones of skin which had been
tion to assess response. However, the performance of TcB shielded from phototherapy light by an opaque protection mask,
during phototherapy and in the hours after discontinua- as recommended for term babies. There was a safety margin of ≥1
tion of phototherapy is still debated, even in term and cm between the area of measurement and the border of the mask.
near-term neonates [12–14], and has never been explored The nurse was blinded to each TSB result. Each TcB value higher
or equal to the hour-specific threshold defined by the reference
in very preterm infants. curves was considered retrospectively as a positive result.
BiliCheck® (Philips, USA) is a widespread second-
generation bilirubinometer that has shown good correla- Data Collection
tions with conventionally determined TSB levels [4, 5]. Neonatal characteristics were collected (GA, birth weight, gen-
The intradevice imprecision has been deemed acceptable der, skin color, small for GA defined as a birth weight below the
10th percentile on the Audipog curves [16], mode of delivery, and
in the very preterm population (2.3 ± 13.5 μmol/l) [11]. antibiotic treatment). For each patient, intrinsic characteristics at
The purpose of this study was therefore to more pre- the time of measurement (postnatal age and weight, hemoglobin
cisely analyze the accuracy of BiliCheck® for the indica- value, skin condition in the TcB measurement zone), and admin-
tion of phototherapy in preterm patients <30 weeks’ GA, istered treatment (respiratory support, phototherapy within the
compared to TSB measurements. Diagnostic perfor- previous 12 h, and inotrope treatment) were collected.
mance was evaluated in two different contexts: for screen- Statistical Analysis
ing purposes (babies without phototherapy) and for bili- Quantitative data were expressed as means ± SD and range.
rubin-monitoring purposes (during phototherapy or up Qualitative variables included sample size and percentage of the
to 12 h after phototherapy). Secondary objectives includ- study population.
ed the evolution of this performance in the first 15 days First, to compare diagnosis accuracy of TcB among babies
without or under phototherapy, we selected the sample of new-
of life and the influence of the newborn’s characteristics borns for whom measurements in both conditions were available.
on TcB reliability. As the dates for TSB measurements were not standardized, their
number per patient varied. To avoid statistical dependence of re-
peated measurements and unequal weight of each patient in the
Patients and Methods database, one paired TSB-TcB measurement per child without and
under phototherapy was randomly selected. Subsequent analyses
Setting and Patients were thus done with only one dataset per patient and photothera-
Our prospective multicenter study was carried out in the neo- py condition (fig. 1). We estimated the sensitivity, specificity, pos-
natal intensive care units (NICU) of the University Hospitals of itive (PPV) and negative predictive values (NPV), positive (LR+)
130.237.122.245 - 8/7/2016 9:42:56 AM
Karolinska Institutet, University Library
Overall sample**
n = 167
M = 481
Fig. 1. Flow chart of the study population.
* Used to compare TcB diagnostic perfor-
mance with or without phototherapy.
** Used to evaluate TcB diagnostic perfor-
mance according to the time period. PT = Period 1: day 0–3 Period 3: day 4–7 Period 3: day 8–15
n = 113 n = 132 n = 90
Phototherapy; n = sample size; M = total M = 151, m = 113 M = 191, m = 132 M = 139, m = 90
number of measurements; m = number of
randomly selected measurements.
and negative likelihood ratios (LR–), and their 95% CI in these two Results
groups.
Secondly, we represented graphically the correlation between
TcB and TSB using Pearson linear regression analysis. TcB-TSB
Study Population
agreement was further assessed using the Bland-Altman technique A total of 167 patients were included. Characteristics
[17]. of the study population are presented in table 1. Mean GA
Considering the slope of the bilirubinemia curve, which rises was 27.6 ± 1.6 weeks (range: 24–29.9) and the mean birth
during the first 3 days and is stable thereafter, and infant matura- weight was 985 ± 248 g (range: 470–1,740). Thirteen pa-
tion between the first and second week of age, three distinct time
periods were established: from birth to day 3, from day 4 to day 7,
tients died in the NICU. All patients required photother-
and from day 8 to day 15. By the same random process described apy at least once; 52% of patients had hyperbilirubinemia
previously, one measurement per child and period was selected to which required phototherapy during the first time peri-
evaluate the consecutive performances of TcB in each period. od, 16% during the second time period, and 2% during
Thirdly, we also investigated whether neonatal characteristics the third time period. No patient presented with jaundice
might influence the discordance between TcB and TSB in very pre-
term newborns, using multivariate multilevel logistic regression
requiring exchange transfusion. At the time of bilirubin
analyses. A p value <0.05 was considered statistically significant. measurement, 97% of patients benefited from noninva-
The study sample size was determined according to the estimated sive (49%) or invasive (48%) respiratory support.
sensitivity of the BiliCheck®. Among the two possible types of errors, One hundred and ninety-seven measurements were
false-positive results (TcB value above the hour-specific threshold for taken while the patient received phototherapy in the
phototherapy when the TSB value was below) and false-negative re-
previous 12 h. Ninety newborns had both one measure-
sults (TcB value below the phototherapy threshold when the TSB
value was higher), false negatives are potentially more dangerous and ment without phototherapy and one measurement per-
expose preterm newborns to inadequate care. Consequently, we con- formed within 12 h of phototherapy, while 44 patients
sidered that BiliCheck® would be efficient if the sensitivity was high- had only measurements done without phototherapy and
er or equal to 95% (95% CI: 91–99). We estimated the incidence of 33 patients had only measurements during photothera-
jaundice in very preterm neonates as being 85% within the first week py. Figure 1 details the 3 groups of patients and how
of life [1]. Considering this prevalence and the expected precision,
the required number of 150 inclusions was retained. their measurements were used in the different analyses.
Data were analyzed using STATA 13.0 (StataCorp, College Sta- A total of 481 paired TcB-TSB measurements were ob-
tion, Tex., USA). tained, and the mean number of measurements per pa-
The study protocol was written and carried out in accordance tient was 3 ± 1 (range: 1–9) (fig. 1). According to the
with the ethical standards of the Helsinki Declaration of 1975, as timing of the measurement, 113 patients had at least one
revised in 1983. It was approved by the Rhone-Alpes Auvergne Re-
gional ethics committee on human experimentation (IRB No. 5891, paired measurement between birth and day 3, 132 be-
February 6th, 2013). All parents of the included neonates gave their tween days 4 and 7, and 90 during the second week of
informed consent for their child’s participation in the study. life (fig. 1).
130.237.122.245 - 8/7/2016 9:42:56 AM
Karolinska Institutet, University Library
400 200
300
100
200
0
100
–100
0
–200
Fig. 2. Relationship between TcB and TSB values. Fig. 3. Bland-Altman plots for TcB and TSB measurements.
ment with the TSB value for the absence of treatment re- ing phototherapy and in the hours following treatment
quirement in 211 of them (98%). These results were not discontinuation. These results had also been found by
significantly influenced by the patient’s intrinsic charac- others in term or near-term populations [13, 14], al-
teristics or treatment at the time of measurement. Relying though this is still debated [12].
solely on the TcB result for screening of very preterm ba- TcB sensitivity and LR– were excellent during the first
bies with hyperbilirubinemia requiring phototherapy days of life, but decreased over time, namely due to a re-
would have avoided 211 TSB measurements, with poten- duced requirement for phototherapy, while the NPV re-
tial avoidance of blood drawing. mained >95% throughout the 3 time periods. Our results
thereby indicate that the protocol for transcutaneous
screening of hyperbilirubinemia in term neonates could
Discussion also be applied, as the first line in the screening strategy,
to the extremely preterm population. Indeed, a negative
The purpose of our study was to analyze the accuracy TcB value would safely eliminate the need for TSB deter-
of TcB in very preterm patients <30 weeks’ GA, a popula- mination and phototherapy. This situation represented
tion that has scarcely been represented in previous studies 45% of the measurements performed in the first 15 days
[11, 18, 19] both for hyperbilirubinemia screening before of life in our study.
phototherapy and for bilirubin monitoring of babies un- This suggestion is in accordance with the recommen-
der phototherapy. In this population, we showed diag- dation made by Nagar et al. [10] in their review of the re-
nostic performance of TcB was comparable for hyperbil- liability of TcB devices in preterm infants. Our results fur-
irubinemia screening and for bilirubin monitoring dur- ther showed that this recommendation is also applicable
130.237.122.245 - 8/7/2016 9:42:56 AM
Karolinska Institutet, University Library