Implementation of a transcutaneous bilirubinometer in a newborn nursery A randomized controlled trial

Transcription

1 Implementation of a transcutaneous bilirubinometer in a newborn nursery A randomized controlled trial Lieve den Boer Supervisor Dr. J. Bekhof, pediatrician Department Pediatrics Institution Isala, Zwolle Student number

2 ABSTRACT Objective: The aim of this study is to provide quantitative data on the reduction of blood sampling when using transcutaneous bilirubinometry to assess hyperbilirubinemia in healthy jaundiced newborns. Study-design: In a randomized controlled trial in jaundiced infants ( 36 weeks gestational age) in a Dutch newborn nursery ward, the need for total serum bilirubin (TSB) measurement to determine hyperbilirubinemia was assessed by either a transcutaneous bilirubin (TcB) measurement (JM-103) (intervention group) or visual assessment of jaundice (control group). When TcB was within 50µmol/L below the threshold for phototherapy on the bilirubin nomogram, a TSB was obtained. Results: A total of 254 neonates were randomized to either the intervention group (n=127) or the control group (n = 127). In the intervention group 43 neonates (33.9%) needed one or more TSB measurements taken versus 85 neonates (66.9%) in the control group (difference 33.0; 95% CI 21.4 to 44.6, P = <0.001). There was a significant reduction of 46.9% in blood sampling for TSB assays in the intervention group compared to the control group (0.50±0.80 vs 0.94±0.83, difference (-0.64 to -0.24), P = <0.001). The number needed to treat is 3.0. There were no differences between groups in treatment, treatment duration or hospitalization. Conclusion: Use of a transcutaneous device to assess hyperbilirubinemia significantly and safely reduced blood draws for TSB assays in 47% in healthy jaundiced near-term and term neonates in the newborn nursery. NEDERLANDSE SAMENVATTING Doel: Deze studie heeft als doel in kwantitatieve data te voorzien wat betreft de reductie van het aantal bloedafnames voor totaal serum bilirubine (TSB) bij het gebruik van een transcutane bilirubine meter bij gezonde neonaten met icterus. Methode: In een gecontroleerde gerandomiseerde studie bij icterische neonaten, geboren bij een zwangerschapsduur 36 weken, werd de indicatie om een TSB te laten bepalen gesteld met behulp van een transcutane bilirubine meter (JM-103) (interventie groep) of middels visuele inschatting van de icterus (controlegroep). Indien de transcutane bilirubine (TcB) minder dan 50 µmol/l onder de fototherapiegrens op de bilirubinecurve lag, werd er een TSB bepaald. Resultaten: Er werden 254 neonaten gerandomiseerd. In de interventiegroep werd er bij 43 pasgeborenen (43/127, 33.9%) 1 of meer TSB-bepalingen gedaan, in de controlegroep waren er 1 of meer bloedafnamen voor TSB nodig bij 85 neonaten (85/127, 66.9%; verschil 33.0; 95% CI 21.4 tot 44.6, P = <0.001). Er was een significante afname van 46.9% in bloedbepalingen voor TSB in de interventiegroep ten opzichte van de controlegroep (0.50±0.80 vs 0.94±0.83, verschil (-0.64 tot -0.24), P = <0.001). De number needed to treat is 3.0. Er waren geen verschillen in behandeling, behandelingsduur of opnameduur tussen beide groepen. Conclusie: Het gebruik van een transcutane bilirubinemeter om de mate van hyperbilirubinemie vast te stellen vermindert het aantal bloedafnames voor TSB-bepalingen significant met 47% bij gezonde a terme pasgeborenen en randprematuren met neonatale icterus, zonder negatieve effecten. L. den Boer 1

3 TABLE OF CONTENTS TABLE OF CONTENTS... 2 LIST OF ABBREVIATIONS INTRODUCTION Neonatal icterus Background Epidemiology Pathophysiology Bilirubin encephalopathy Treatment Transcutaneous bilirubinometry Objective MATERIALS AND METHODS Study design Patients Inclusion criteria Exclusion criteria Methods Randomization, allocation and blinding Data collection Clinical data Outcomes Data analysis Sample size calculation Statistical analysis Medical ethical considerations RESULTS Patients Patient characteristics Primary outcome Secondary outcomes Additional outcomes DISCUSSION Primary outcome L. den Boer 2

4 4.2 Comparison to existing literature Strengths and limitations Recommendations for future research CONCLUSION REFERENCES APPENDIX A BILIRUBIN NOMOGRAM APPENDIX B FLOWCHART HYPERBILIRUBINEMIA LIST OF ABBREVIATIONS AAP AGA CI G6PD GA JM-103 LGA SGA TcB TSB NICE NNT American Academy of Pediatrics appropriate for gestational age confidence interval glucose-6-phosphatedehydrogenase gestational age Dräger Jaundice Meter model 103 large for gestational age small for gestational age transcutaneous bilirubin total serum bilirubin The National Institute for Health and Care Excellence (UK) number needed to treat L. den Boer 3

5 1. INTRODUCTION 1.1 Neonatal icterus Up to 80% of all newborn infants develop some degree of visible jaundice due to hyperbilirubinemia in the first few days postpartum.(1,2) Hyperbilirubinemia is the most frequent cause of hospitalization or readmission in the first week of life.(3) In most newborns neonatal hyperbilirubinemia is transient, physiological and mild, and does not pose any health risks. However, approximately 2-5% of all newborns develop significant hyperbilirubinemia necessitating treatment to prevent irreversible neurological damage due to kernicterus.(4,5) Recognition of infants with severe hyperbilirubinemia that need phototherapy or exchange transfusion is crucial. However, the yellow color of the skin does not correlate with the serum bilirubin concentration, which makes visual assessment an unreliable method to assess severity of neonatal jaundice and identify which newborns are at risk for hyperbilirubinemia.(5-9) Determination of serum bilirubin levels is one of the most frequently performed laboratory tests in neonates. This procedure is painful, costly, time consuming and can even have consequences like osteomyelitis.(10) Ideally, the number of blood samples taken in neonates should be reduced as much as possible. For this purpose, a number of devices for noninvasive bilirubin determination have been developed.(11) Transcutaneous measurements of bilirubin are a promising alternative to visual assessment, avoiding unnecessary blood sampling in neonates, the associated stress and financial costs. In international guidelines, the use of transcutaneous bilirubin measuring is advocated.(3,8,12) However, despite these favorable outcomes, in the Netherlands bilirubinometers are not widely used. 1.2 Background Epidemiology In the Netherlands, approximately 1% of all term newborns develop clinically relevant hyperbilirubinemia requiring treatment, accounting for children of the roughly children born nationwide every year. An estimated newborns develop an extreme hyperbilirubinemia ( 420 µmol/l and/or signs of acute encephalopathy).(12) Roughly half of these children require an exchange transfusion ( % of all newborns). (5) Pathophysiology Hyperbilirubinemia appears clinically as jaundice. Jaundice refers to the yellow coloration of skin and sclera caused by deposition of bilirubin in the skin and mucous membranes.(3,13) The breakdown of erythrocytes produces unconjugated bilirubin, which circulates through the bloodstream bound to albumin. Unconjugated bilirubin is metabolized in the liver to form conjugated bilirubin which can be excreted in the bile.(14) In most newborns, hyperbilirubinemia results from increased turnover of fetal hemoglobin and transient immaturity of the liver in bilirubin conjugation and excretion. (11,15) Therefore, the majority of neonates show an increase in total serum bilirubin levels for the first two to six days followed by a decrease as bilirubin production decreases and conjugation improves. In most cases, this imbalance is physiological will resolves spontaneously after 1 to 2 weeks. (1,3) The major risk factors for severe neonatal hyperbilirubinemia include prematurity, breastfeeding, extravasated blood (i.e. bruising and cephalohematoma due to birth trauma), hemolysis, blood group incompatibility (Rhesus factor or ABO), infection, (obstructive) liver disease and metabolic disorders (i.e. glucose-6-phosphate-dehydrogenase (G6PD) deficiency).(14,16) Total serum bilirubin levels vary with the racial composition of a L. den Boer 4

6 population; bilirubin levels are higher in East-Asian populations than in Caucasian populations, and lower in black infants compared to white infants.(1) When jaundice appears during the first 24 hours postpartum hyperbilirubinemia is more likely to be pathological and underlying causes (i.e. isoimmune or other hemolytic diseases) should be investigated Bilirubin encephalopathy Severe neonatal hyperbilirubinemia is associated with chronic bilirubin encephalopathy, also known as kernicterus, a rare but invalidating neurological condition. Unconjugated bilirubin, which is lipid soluble, can cross the blood-brain barrier. Deposition of unconjugated bilirubin in the brain is toxic and can lead to acute and, subsequently, chronic bilirubin encephalopathy. The precise molecular mechanisms responsible for the cytotoxicity of bilirubin are not completely understood.(1,15) The early phase of acute bilirubin encephalopathy is characterized by the non-specific symptoms of lethargy, hypotonicity and poor feeding, later evolving to the mostly irreversible symptoms of chronic bilirubin encephalopathy. Chronic bilirubin encephalopathy is classified by severe choreoathetoid cerebral palsy, auditory neuropathy, upward gaze palsy and enamel dysplasia of the deciduous teeth.(1,8) The cognition is relatively spared.(5) The risk of kernicterus is increased in neonates with extreme hyperbilirubinemia (>425 µmol/l), and is known to occur at lower levels in pre-term infants and term infants with risk factors. However, some affected infants appeared to be otherwise healthy, breastfeeding newborns.(17,18) Although rare, bilirubin encephalopathy still causes mortality (10%) and severe morbidity (>70%) worldwide.(19) The exact incidence of kernicterus is unknown, it is estimated that kernicterus still occurs in approximately per infants born after 35 weeks of gestation.(1,20) Early discharge (often within 24 hours) and inadequate follow-up of healthy, term newborns is associated with an increase in the number of reported cases of severe hyperbilirubinemia and kernicterus.(18,21-23) In developing countries persistently high rates of significant neonatal hyperbilirubinemia are implicated to be a major contributor to the incidence of cerebral palsy and hearing impairment.(16,24) Treatment The aim of therapy for hyperbilirubinemia is to lower the concentration of circulating unconjugated bilirubin to prevent the risk for bilirubin encephalopathy. Hyperbilirubinemia can be treated easily and with minimal adverse effects with phototherapy.(1,5,14) Phototherapy uses blue light to convert bilirubin in the skin and subcutaneous tissue to watersoluble molecules that can be excreted in bile or urine without the need for conjugation.(14) The decision to start treatment is made by plotting the total serum bilirubin (TSB) value on a validated bilirubin nomogram, which takes gestational age, age of the infant in hours since birth and risk factors into account (shown in Appendix A).(5,8) Treatment is indicated at a lower threshold for premature infants or when risk factors are present. When intensive phototherapy is not sufficient in lowering the serum bilirubin level, exchange transfusion is indicated. It is faster in lowering the TSB but has more complications.(5) 1.3 Transcutaneous bilirubinometry The estimation of serum bilirubin by hand-held transcutaneous bilirubin (TcB) devices in jaundiced infants is becoming a frequently used method.(8) TcB devices analyze the light reflected from the skin using spectral reflectance techniques. The device exposes the skin to light of various wave lengths. The spectra of returned light will depend on the concentration of certain chromophores in the skin, mainly melanin, collagen, hemoglobin and bilirubin. The bilirubinometer calculates the concentration of unconjugated bilirubin (both free and bound to L. den Boer 5

7 albumin) in the skin and sub cutis by using the difference in absorption spectra and an algorithm.(25) Previous research shows TcB is a physiologically different parameter from TSB, since the contribution of intravascular bilirubin to the measured TcB is <1%.(6,25-27) TSB measures intravascular bilirubin (added conjugated and unconjugated bilirubin), whereas TcB measures both unconjugated bilirubin bound to albumin and free unconjugated bilirubin. Bilirubin encephalopathy is caused by deposition of free unconjugated bilirubin in the basal ganglia and brainstem nuclei. Some studies therefore suggest free bilirubin concentration measured by TcB is a better predictor for bilirubin encephalopathy than TSB.(6,26,28,29) Although transcutaneous bilirubin is not a measurement of the total serum bilirubin, TcB measurements appear reliable in identification of hyperbilirubinemia in both term and preterm infants, of different skin colours and ethnic backgrounds, prior to phototherapy. There is a linear relationship between TcB and TSB with published correlations of (13,30-33). More importantly, there is a good agreement between TcB and TSB. Transcutaneous devices are reported to underestimate TSB by µmol/l, particularly at higher bilirubin concentrations (> µmol/l). (15,30,32-35) For preterm neonates values for TcB were found to overestimate TSB levels, increasing the safety of TcB as a screening tool.(30,38) Skin tone has a significant effect on TSB and TcB agreement.(36,37) Because of the general underestimation of TSB levels by TcB, several decision rules were proposed by the American Academy of Pediatrics (AAP).(8,34) The AAP suggested to obtain a TSB level when 1) the TcB measurement is 75 th percentile on the Bhutani TSB nomogram; 2) the TcB measurement is 70% of the recommended phototherapy threshold for a particular infant; or 3) when the TcB measurement is 13 mg/dl (~222 µmol/l). Maisels et al.(47) proposed to measure a TSB when the TcB level is within 3 mg/dl (~50 µmol/l) of the phototherapy threshold. In a recent study, Taylor et al.(34) did not identify a superior decision rule. The NICE guidelines (3) recommend obtaining a TSB if the TcB is 250 µmol/l or at or above the phototherapy threshold. The two devices widely used at time of writing, Philips BiliChek and Dräger s Jaundice Meter model 103 (JM-103), have been studied extensively and are validated for clinical use.(11,31-33) The JM-103 was selected for this trial since it was noted to be easier in use and the cost reduction is expected to be higher than with BiliChek, since there is no need for disposable materials.(31) When obtaining a TcB measurement, birth marks, hairy areas and other skin defects should be avoided. The JM-103 performed best when measurements were made at the sternum.(13,39,40) In Northern America the bilirubinometer is widely used. However, the use of transcutaneous devices is not as common and widespread in Europe.(33) A recent telephonic review among the majority of Dutch hospitals showed that no more than 27% of all hospitals use a transcutaneous device in the screening for hyperbilirubinemia. There seems to be a lack of confidence among Dutch pediatricians that a transcutaneous bilirubinometer will safely reduce a clinically relevant amount of blood draws in real-life practice. Low income countries could also have many benefits from TcB screening; however the high cost of the transcutaneous devices is a barrier.(41,42) L. den Boer 6

8 1.4 Objective Prior studies demonstrated that transcutaneous bilirubinometers are a non-invasive, easy to use method to identify hyperbilirubinemia, with superiority over visual assessment of jaundice.(13,17,43) Observational and retrospective studies have shown a reduction of 20-50% in blood sampling for serum bilirubin (32,33,40), and a prior randomized controlled trial comparing TcB to visual assessment in an Indian population of healthy newborns found a 34% reduction of TSB measurements.(44) However, despite these promising outcomes, TcB devices are not widely used in the Netherlands. The objective of this study was to provide quantitative data on the effectiveness of a transcutaneous device as a screening method for severity of jaundice compared to visual appraisal of the skin by a healthcare professional, in healthy term and near-term jaundiced neonates. We hypothesized that the use of transcutaneous bilirubinometry decreases the need for blood sampling for TSB assays, number of complications and health care costs. This study aims to improve quality of care for healthy jaundiced newborns and to provide data on the advantages of a transcutaneous bilirubinometer in a Dutch setting, to lift the barriers concerning (Dutch) pediatricians and neonatologists. L. den Boer 7

9 2. MATERIALS AND METHODS 2.1 Study design We conducted a randomized controlled trial at the newborn nursery of Isala, between January 2013 and May Isala is a large general teaching hospital in Zwolle, the Netherlands, with a predominantly Caucasian patient population. There are approximately 3700 deliveries in the maternity ward every year. The objective was to compare the current standard care, visual assessment of jaundice, with the use of a transcutaneous bilirubinometer. 2.2 Patients Inclusion criteria All neonates born with a gestational age of >36 weeks, admitted at the newborn nursery, were eligible for inclusion if they developed clinical jaundice after 24 hours of age and before day 8 postpartum. Jaundice was assessed by nurses. Yellowness of the skin was evaluated in a well-lit room, blanching the skin of the undressed newborn to reveal the underlying colour Exclusion criteria Infants with active hemolytic disease, clinical signs of kernicterus, prior TcB and/or TSB measurements, prior phototherapy and/or a large congenital defect on the sternum were excluded. Jaundiced neonate Informed consent given No informed consent or exclusion Control Visual assessment Intervention Transcutaneous bilirubin 50 µmol below treshold >50 µmol below treshold TSB Observation TSB Observation Figure 1 Study design. TSB: total serum bilirubin. 2.3 Methods As shown in figure 1, after written parental informed consent was given, the infant was randomized to the control or intervention group. When allocated to the intervention group, a TcB was measured using the Dräger Jaundice Meter model 103 (JM-103) on the sternum. The machine averaged mean of 2 measurements was recorded. The device was maintained according to manufacturers instructions. Calibration on the docking station was done daily. The Dutch hour-specific, risk-factor dependent bilirubin nomogram (Appendix A, based on international guidelines,(5)) was used to determine the need for TSB measurement. A TSB L. den Boer 8

10 assay was done when the TcB value fell within 50 µmol/l of the phototherapy treatment line or above in the nomogram. This 50 µmol/l margin was set because of the known underestimation of serum bilirubin by the TcB device, mainly at higher values.(5,47) In the control group, standard care was given, which was visual assessment of jaundice by the consulting physician, when indicated by the nurse. In an adequately lit room, preferably in daylight, the infant was systematically assessed by a physician. The necessity for TSB measurement was decided by yellowness of the skin, clinical picture and risk factors, conform the Dutch guidelines.(5) The same bilirubin nomogram (Appendix A) was used to determine the need for phototherapy. When a measured TSB was below the threshold, the infant was observed. Blood samples for TSB were drawn by peripheral venipuncture in the heel. Serum bilirubin assays were performed by using a Radiometer ABL 800 series. To ensure not a single infant with significant hyperbilirubinemia would be missed, the clinical team had the authority at all times to order a TSB, even when TcB measurement deemed this not necessary. These escape decisions and their outcomes were recorded in the database. 2.4 Randomization, allocation and blinding All patients were randomized in either the intervention or control group and stratified in two groups according to gestational age (group 1: jaundiced neonates born at gestational age >36 weeks and <38 weeks, and group 2: jaundiced neonates born at gestational age >38 weeks). Randomization and allocation were done using web-based program Research Manager, ensuring concealment of allocation. Due to the nature of the intervention, blinding was not possible. 2.5 Data collection Clinical data All data were collected in a single database with a log of any events occurring during the study period. For each randomized newborn, the following data were collected: gestational age (estimated from ultrasound in first semester of pregnancy), date and time of birth, sex, race, birth weight, standard deviation category of birth weight for gestational age, type of feeding (exclusively breastfeeding, combination of breastfeeding and formula, or exclusively formula), risk factors for hyperbilirubinemia, lowest recorded weight and percentage of weight loss from birth weight. The most likely cause for hyperbilirubinemia was determined using a flowchart (Appendix B). The race of the newborn was noted if known; otherwise, the race of the mother was recorded Outcomes The primary study outcome was the number of neonates that had one or more blood samples taken for TSB measurements. Secondary outcome measures were: the number of blood samples for TSB measurement per patient (before phototherapy); duration of phototherapy in hours; highest recorded TSB; number of TSB above exchange transfusion threshold; total (estimated) costs; number of patients with kernicterus; and correlation of TcB and TSB in patients with escape decisions. 2.6 Data analysis Sample size calculation The primary outcome variable was the number of children that needed blood samples taken for TSB measurement in each group. For demonstrating 20% reduction in the need for blood L. den Boer 9

11 sampling (observed in Hartshorn (2)) with a power of 80% and two-sided significance of 5%, we needed to enroll 219 patients per group, a total of 438 patients Statistical analysis Data were entered and analyzed using IBM SPSS statistics for Windows Version 23. Dichotomous variables were analyzed by a Chi square test or in case of small numbers a Fisher s Exact test. Continuous variables with normal distributions were analyzed by two sampled Student s t test. When continuous data were not normally distributed, we used logarithmic transformation followed by a t-test. As the mean duration of phototherapy and the mean duration of admission in the pediatric ward were log transformed before analysis, the difference between two groups is expressed as a ratio of geometric means. In addition, the difference in median duration of phototherapy and median duration of admission with 95% confidence intervals were calculated.(45) For rare secondary outcome variables (i.e. kernicterus, exchange transfusion) descriptive statistics were used. A Bland-Altman plot was made for agreement between measurements of TcB and TSB values. All statistical tests were 2-sided and a P-value of <0.05 was considered statistically significant. Analysis was by intention to treat analysis. 2.7 Medical ethical considerations Written informed consent was obtained from the parents or caretakers of participating neonates. The trial protocol was approved by the institutional Medical Ethical Committee. This trial was registered with ClinicalTrials.gov, number NCT L. den Boer 10

12 3. RESULTS 3.1 Patients Between January 2013 and May 2016, approximately live births occurred in the maternity ward. In this study period, 293 neonates (2.4%) from the newborn nursery were approached for enrolment in the study and assessed for eligibility, as shown in figure 2. A total of 39 patients were excluded from the study for various reasons: declined to participate (n=11), not meeting inclusion criteria (n=8) and excluded for other reasons (n=20). Of these twenty, twelve neonates could not be included due to logistic reasons (absence of randomization personnel on weekends, nights and holidays), two were not randomized during a brief period where the JM-103 was out of order, in one case there was a language barrier, and one child was excluded because he required multiple lab tests for other symptoms. In four neonates the reason for missed inclusion is unknown. A total of 254 neonates were enrolled in the trial. Three children were allocated to the intervention group, but did not receive the intervention for unknown reasons. They were analyzed in the intervention group according to the intention to treat principle. Figure 2 Enrolment of patients. *Other reasons were logistic reasons (n=12), JM-103 out of order (n=2), language barrier (n=1), more laboratory assays required for other symptoms (n=1), unknown reasons (n=4). ** Allocated to the intervention group, but did not receive intervention (n=3). L. den Boer 11

13 3.2 Patient characteristics The characteristics of the 254 randomized neonates are summarized in Table 1. Baseline characteristics were comparable in both groups, confirming an adequate randomization. Patient Characteristics Intervention (n=127) Control (n=127) Gender (male) 61 (48.0) 69 (54.3) Age mother (years) (mean±sd) Gestational age (weeks +days ) (mean±sd) < ± ± (39.4) 77 (60.6) 30.7± ± (39.4) 77 (60.6) Birth weight (grams) (mean±sd) SGA (< -2 SD) AGA LGA (> +2 SD) Risk factors present Gestational age <38 +0 Asphyxia Active blood group incompatibility Temperature instability Hematoma Ethnicity Caucasian Non-Caucasian Unknown Feeding Exclusive breastfeeding Exclusive formula feeding Combination Weight loss (%) (mean±sd) <7% >7% Presumed diagnosis Physiological neonatal jaundice Insufficient feeding Hemolysis Other 3190±552 3 (2.4) 121 (95.3) 3 (2.4) 53 (41.7) 50 (39.4) 1 (0.8) 1 (0.8) 1 (0.8) 11 (8.7) 98 (77.2) 11 (8.7) 18 (14.2) 41 (32.3) 41 (32.3) 45 (35.4) 6.5± (59.1) 52 (40.9) 76 (59.8) 42 (33.1) 8 (6.3) 1 (0.8) 3278±580 0 (0) 121 (95.3) 6 (4.7) 50 (39.4) 49 (38.6) 0 (0) 0 (0) 1 (0.8) 10 (7.9) 103 (81.1) 14 (11.0) 10 (7.9) 32 (25.2) 28 (22.0) 67 (52.8) 6.8± (52.8) 60 (47.2) 65 (51.2) 52 (40.9) 7 (5.5) 3 (2.4) Table 1 Data expressed as n (%) unless stated otherwise. GA: gestational age; SGA: small for gestational age; AGA: appropriate for gestational age; LGA: larger for gestational age; SD: standard deviation. 3.3 Primary outcome In the intervention group, a total number of 154 TcB measurements were recorded. In four neonates (1.6%) no TcB measurements were done, due to various reasons; not receiving the allocated intervention (n=3), JM-103 out of order (n=1). The mean number of TcB measurements per neonate was 1.2 (SD±0.6), with a maximum of 4 measurements (n=1). The mean measured TcB level was (SD±43.7) µmol/l. The highest TcB level measured was 305 µmol/l. Following these 154 TcB measurements, 63 TSB measurements were obtained to confirm the TcB, a mean of 0.5 (SD±0.80) per neonate. Eighty-four neonates (66.1%) in the intervention group did not need any blood sampling for TSB measurement after transcutaneous measurement and one or more TSB measurements were performed in 43 neonates (33.9%). In the control group, after visual assessment one or more TSB measurements were obtained in 85 neonates (66.9%). A total of 119 measurements was done, with a mean of 0.94 (SD±0.83) per neonate. L. den Boer 12

14 TSB measurements Intervention Control Difference (95% P-value (before phototherapy) (n=127) (n=127) CI) Total number of TSB measurements Number of neonates needing 1 TSB 43 (33.9%) 85 (66.9%) 33% (21.4 to 44.6) <0.001* Number of TSB measurements per patient 0.50± ± (-0.64 to -0.24) <0.001* Table 2 Data expressed as n (%) or mean±sd. CI: confidence interval; SD: standard deviation.*statistically significant Between the number of TSB measurements in both groups a mean difference of was found (95% CI to -0.24). With a mean number of 0.94 measurements in the control group and a difference of , there was a significant reduction of 46.9% in blood sampling for TSB assays in the intervention group compared to the control group, as shown in table 2. The number needed to treat (NNT) is 3.0, which means TcB reduced the need for blood sampling in one in every 3 TcB assessments. 3.4 Secondary outcomes Total serum bilirubin A total of 182 TSB measurements were obtained. TSB values ranged from 85 to 382 µmol/l. The mean TSB value per neonate was significantly higher in the intervention group (213.2±49.9 µmol/l), compared to the control group (193.9±52.7 µmol/l, P = 0.049). This can be explained by the preceding TcB measurement, which only needs to be confirmed by TSB at higher values. The mean of the highest TSB value per neonate was (SD±54.0) µmol/l in the intervention group, not statistically significantly different from the control group (204.1±63.2 µmol/l, P =0.110). Although the mean TSB value is higher in the intervention group, the highest TSB values are found in the control group. As shown in table 3, the incidence of severe hyperbilirubinemia, defined as TSB value 342 µmol/l by Wainer et al.(46), was higher in the control group (5 vs 0 cases). None of the measured TSB values were extreme ( 425 µmol/l) or hazardous ( 512 µmol/l). Four of the five neonates with severe hyperbilirubinemia were (near) premature and exclusively breastfed which resulted in a weight loss of >7% in 3 of these 4 neonates. The fifth neonate with severe hyperbilirubinemia was term and had no risk factors. TSB levels before phototherapy Intervention Control Difference (95% CI) P-value (n=63) (n=120) Below 342 µmol/l 63 (100%) 115 (95.8%) 4.2% ( ) Severe ( 342 µmol/l) - 5 (4.2%) Extreme ( 425 µmol/l) - - Hazardous ( 512 µmol/l) - - Table 3 Data expressed as n (%). CI: confidence interval. *Statistically significant. The average age at the time of the first measurement (either TcB or TSB) was 50±15 hours (~2 days) in the intervention group, this was not statistically different from the control group (54±17 hours (~2.25 days), P = 0.121) Phototherapy Overall, 33 (13.0%) of the 254 enrolled neonates required phototherapy. The proportion of neonates requiring phototherapy in the intervention group (n=18 (14.2%)) is not significantly higher than in the control group (n=15 (11.8%), P =0.71). A t-test after logarithmic transformation revealed no significant difference in the phototherapy duration (P =.572), as shown in table 4. The age in hours at the start of phototherapy was significantly lower in the L. den Boer 13

15 intervention group (P=0.029), starting on an average age of approximately 2.25 days compared to roughly 2.75 days in the control group. In Isala, infants requiring phototherapy are admitted to the pediatric ward. The mean duration of admittance was 41 (SD±50.1) hours in the intervention group and 30.1 (SD±14.2) hours in the control group. Since length of stay was skewed to the right, we log transformed this measure. As shown in table 4, the duration of stay at the pediatric ward was not statistically different between both groups (P = 0.683). The large range of distribution in duration of admission in the pediatric ward is mainly due to four outliers. Three neonates (12.1%) were admitted for more than 48 hours, due to feeding difficulties. Another infant in the control group was already admitted in the pediatric ward for the evaluation of a systolic murmur, with a total stay at the pediatric ward of 219 hours (9 days 3 hours). The number of TSB measurements after phototherapy per neonate did not differ between groups (P=0.315). Phototherapy Intervention Control Difference (95% CI) P-value (n=127) (n=127) Neonates requiring phototherapy 18 (14.2%) 15 (11.8%) 2.4% (-5.86 to 10.67).709 Age at start phototherapy (hours) 55±15 67± (-22.7 to -1.3).029* Duration of phototherapy Median (hours) (-3.3 to 3.3) - Range of distribution (min-max) Log mean±sd 3.08± ± Geometric mean (hours) Ratio of geometric means (95% CI) 0.94 ( ) NA Stay in pediatric ward Median (hours) (-9.5 to 7.5) - Range of distribution (min-max) Log mean±sd 3.4± ± Geometric mean (hours) Ratio of geometric means (95% CI) 1.08 ( ) NA Number of TSB measurements per patient after phototherapy 1.67± ± (-1.20 to 0.40).315 Table 4 Data expressed as n (%) or mean±sd. The duration of phototherapy and the duration of stay in the pediatric ward were log transformed before analysis because these variables were positively skewed. The effect of the intervention refers to the ratio of geometric means for the duration of phototherapy/the duration of stay in the pediatric ward. CI: confidence interval; SD: standard deviation; NA: not applicable. *Statistically significant There were no differences between the intervention and control groups in number of neonates with rebound hyperbilirubinemia, readmission for phototherapy or TSB measurements after discharge, as shown in table 5. Intervention Control Difference (95% P-value (n=127) (n=127) CI) Rebound hyperbilirubinemia 1 (0.8%) 2 (1.6%) 0.8% (-1.87 to 3.47) Readmissions for phototherapy 1 (0.8%) 3 (2.4%) 1.6% (-0.91 to 4.11).622 TSB measurement(s) after discharge 7 (5.5%) 4 (3.2%) 2.3% (-2.72 to 7.32).538 Table 5 Data expressed as n (%). CI: confidence interval Exchange transfusion and kernicterus One neonate, in the control group, had a TSB level above the exchange transfusion threshold (363 µmol/l, threshold 350 µmol/l). Phototherapy was started while awaiting blood for exchange transfusion, resulting in a rapid decrease in TSB and therefore an exchange transfusion was no longer indicated at the time the exchange blood was available. None of the other neonates in either group required or underwent exchange transfusions. No neonates had clinical signs of kernicterus or additional complications. L. den Boer 14

16 3.4.4 Escape decisions The clinical team had the authority at all times to order a TSB, even when TcB measurements deemed this not necessary. In the intervention group, an escape decision was made by the consulting physician in 2 cases (1.6%). In one case the JM-103 was not available for use and a TSB was obtained without a preceding TcB. In the other case a TcB of 224 µmol/l was recorded, followed by a TSB of 243 µmol/l, a difference 19 µmol/l. The reason for this escape is not mentioned in the medical record. Since both the TcB and TSB were well below the phototherapy threshold, this appears to have been an unnecessary blood draw. There were no further TcB or TSB measurements in this case Agreement of TcB and TSB Of the 63 TSB measurements obtained in the intervention group, 48 TcB measurements from 39 neonates could be paired with a subsequent TSB measurement. The remaining 15 TSB measurements were not preceded by a TcB due to various reasons; 3 neonates did not receive the allocated intervention (figure 2) and 1 infant had an escape TSB without a TcB. In the remaining cases (n=10) the reason for obtaining a TSB without a TcB is not recorded; we suspect the consulting physician was not aware of the randomization or a TcB was not possible logistically. The Pearson s correlation coefficient between TSB and TcB was 0.83 (P=<0.001), however high correlation does not necessarily imply good agreement between TcB and TSB. A Bland- Altman plot was made to evaluate agreement between TcB and TSB, as shown in figure 3. The mean difference between TcB and TSB was µmol/l, in other words the mean underestimation of TSB by TcB is 3.43 µmol/l in our study population. TcB underestimated TSB with a maximum of 62 µmol/l or overestimated with a maximum of 83 µmol/l. The 95% limits of agreement between TcB and TSB ranged from to Overall, the JM-103 seems to underestimate the TSB. The lower limit of agreement of µmol/l lies within our clinical margin of 50 µmol/l below the phototherapy threshold, meaning with this margin there is a low potential risk of falsely low TcB values and therefore of missing significant hyperbilirubinemia. Of all combined measurements 91.7% lies within the 50 µmol/l clinical margin. In three patients (6.3%) TcB underestimated TSB by more than 50 µmol/l. However, the TcB values of these neonates were still within 50 µmol/l of their age- and risk factor dependent phototherapy threshold and they therefore had a TSB performed to confirm their TcB value. All three received phototherapy and there were no complications. In one paired measurement (2.1%) the TcB overestimated the TSB by more than 50 µmol/l (83 µmol/l). Retrospectively, a TSB was not necessary in this case. The amount of time between TcB and the subsequent TSB varied greatly. The mean duration between measurements was 79.08± minutes (1 hour 19 minutes ± 1 hour 44 minutes). The time interval between the four combined measurements that differed more than 50 µmol/l, was not statistically different from the paired values between the margins (P=0.162). L. den Boer 15

17 Figure 3 Bland-Altman plot of TcB versus TSB for combined measurements (n=48). Mean difference: µmol/l. Limits of agreement µmol/l to µmol/l. Dotted lines: mean difference and limits of agreement. Red lines: 50 µmol/l clinical margins Cost reduction The total cost of one bilirubin assay in the laboratory of Isala is approximately From figure 2, we estimated jaundiced neonates are encountered every year in the newborn nursery. With a mean of 0.94 TSB measurements per neonate in the control group (table 2), an estimated 89 TSB assays would be performed annually, with a total expense of approximately 800, -. As we found a reduction in blood sampling of 46.9% in the intervention group, with the implementation of the JM-103 the cost reduction would be more than 50%, an expected total of 425, -. The purchase costs for the JM-103 were 5.900, meaning it will take nearly 14 years to make the JM-103 have cost recovery in the newborn nursery. According to the manufacturer the JM-103 would have a service life of measurements, enough for performing 2 measurements on every newborn child in the Isala post-natal ward for more than 20 years. 3.5 Additional outcomes Decision rules In this trial we used a clinical margin of 50 µmol/l below phototherapy threshold on the ageand risk factor dependent bilirubin nomogram as the cut-off point for TcB values; within this 50 µmol/l margin a TSB was ordered. This margin corresponds with the pragmatic approach proposed by Maisels (47) of drawing blood for a TSB level in newborns whose TcB measurements are within 3 mg/dl (=51.3 µmol/l) of the phototherapy threshold. Would we have used a margin of 30 µmol/l instead of 50 µmol/l below threshold, there would have been 17 fewer blood draws for TSB assays, corresponding with a significant reduction (P =<0.001) of 61.2% in the need for blood sampling in the intervention group L. den Boer 16

18 compared to the control group. However, four (3.1%) infants in the intervention group had significant hyperbilirubinemia requiring phototherapy, that would have been missed had we used a margin of 30 µmol/l below threshold. The characteristics of these infants varied: two neonates were Caucasian and term with no risk factors; the other two infants were non- Caucasian and premature with no other risk factors. Of the 48 paired TcB and TSB measurements in the intervention group, 9 (18.9%) differ more than 30 µmol/l. Because of the known underestimation of serum bilirubin by transcutaneous devices, it is recommended by the British NICE guidelines and several studies that a TcB value of 250 µmol/l should be confirmed with a TSB.(3,11,13) Of the total number of 154 TcB measurements, 22 measurements (14.3%) were above 250 µmol/l. All 22 measurements were confirmed by a TSB in our study, since they were all within 50 µmol/l below the phototherapy threshold of the particular neonate. Other studies and guidelines suggest obtaining a TSB value when TcB measurement is 13 mg/dl (~ 222 µmol/l).(1,8,11,34) Retrospectively, there would have been 3 more TSB measurements in the intervention group, resulting in a significant reduction of 43.4% in number of blood draws in the intervention group compared to the control group (P=<0.001). Of these 3 neonates, none had a TSB assay after discharge, readmission for hyperbilirubinemia or any complications. L. den Boer 17

19 4. DISCUSSION 4.1 Primary outcome In neonatal hyperbilirubinemia the challenge is to avoid bilirubin encephalopathy while reducing the number of blood sampling as much as safely possible, without excessive testing or treatment. This randomized controlled study shows a marked reduction of almost 50% in blood sampling when using a transcutaneous bilirubin device compared to visual assessment of yellowness of the skin. We found a number needed to treat (NNT) of 3, meaning 3 transcutaneous measurements have to be performed to prevent one TSB measurement. Our findings suggest that TcB is a safe, easy and quick method to screen for significant hyperbilirubinemia in healthy newborn jaundiced infants, with a TSB measurement only necessary when the TcB is above a certain cut-off value considerably reducing the number of unnecessary blood sampling, without adverse effects. 4.2 Comparison to existing literature Reduction in blood sampling Our findings confirm earlier observational and retrospective reports on reduction of blood sampling of 20-50% when using a transcutaneous device.(32,33,40) The reduction we found was higher than in an earlier randomized controlled trial form Mishra et al.(44), which reported a decrease of 34%. There were certain differences between our trials. Mishra et al conducted his trial in an Indian population of newborns 35 weeks gestation, not comparable to our predominantly Caucasian population. Study design differed from ours in that study subjects were assessed by both visual appraisal and TcB (by BiliChek ) every 8-12 hours, instead of when indicated by nurses. Different cut-off values for TcB were used on what are likely different bilirubin nomograms. In Mishra et al, one pediatrician did all the visual appraisals, where in our trial this was done by different health care providers, ranging from pediatricians to interns. All in all, it seems like our trial was more like routine clinical practice in a European setting. The reduction in blood draws was also higher than reported by Hartshorn and Buckmaster (2). They conducted a before after implementation study in predominantly Caucasian neonates ( 36 weeks) in an Australian post-natal ward. Our TcB protocol closely resembles theirs, using the JM-103 on the sternum in jaundiced infants 24 hours of age with a margin on the routine hour-specific TSB nomogram. As in our study, nurses indicated which neonates were jaundiced. Hartshorn and Buckmaster showed a reduction of 20% in blood draws after implementation of a transcutaneous bilirubin meter. TcB and clinical outcomes Although Petersen et al.(48) found a reduction in hospital readmissions for hyperbilirubinemia and an increase in the incidence of phototherapy treatment before discharge after the introduction of a TcB protocol, we found no differences between groups regarding these outcomes. It has also been speculated that TcB measurements influence length of hospitalization and clinical outcome (40) but length of stay in the pediatric ward was not different between groups in this study, neither were the duration of phototherapy, the number of neonates with rebound hyperbilirubinemia after phototherapy or the number of TSB measurements after discharge. No infants in our study population underwent exchange transfusion or had (clinical signs of) kernicterus. In the intervention group, we found a significantly higher mean TSB per neonate. Since this difference was only 19.3 µmol/l, it does not seem to be of major clinical relevance. On the contrary, we found five neonates with severe hyperbilirubinemia ( 342 µmol/l) in the control L. den Boer 18

20 group compared to none in the intervention group. Comparable results were previously reported by Hartshorn and Buckmaster (2), who found a significant reduction of neonates with a TSB of µmol/l after implementation of the transcutaneous bilirubin device. Secondly, in our trial the age in hours at the start of phototherapy was significantly lower in the intervention group compared to the control group. As mentioned before by Petersen et al.(48), a possible explanation for both the reduction in severe hyperbilirubinemia and the earlier start of phototherapy, is that due to the convenience and non-invasive nature of the JM- 103, the screening was more effective in the intervention group and therefore hyperbilirubinemia was noticed earlier. Subsequently, treatment was initiated earlier. Agreement between TcB and TSB Although TcB and TSB are two different physiological parameters, TcB appears to have a good agreement to TSB, with our clinical margin of 50 µmol/l (~3 mg/dl) within the 95% limits of agreement. As in multiple other reports, we found an underestimation of TSB by the transcutaneous device.(32,49) Our mean underestimation was 3.4 µmol/l, with 95% limits of agreement of approximately 50 µmol/l. This is similar to other studies in predominantly white infants.(11) In our study population, 6.3% of combined measurements have an underestimation of >50 µmol/l. These infants did have a TSB measured since their TcB was still within the 50 µmol/l clinical margin on their age- and risk factor dependent nomogram. However, this can be a clinically relevant difference, requiring caution. Our finding is somewhat higher than previously reported; Taylor et al.(50) reported 2.2% of TcB measurements underestimated TSB 3mg/dL and Maisels (37) found TcB underestimated TSB more than 3mg/dL in 0.6% of patients. Nevertheless, there appears to be ample margin for error with the current recommended levels for phototherapy and the even lower cut-off point for TcB values, which are both well below the bilirubin levels bilirubin encephalopathy generally occurs at.(15) Missing a child with significant hyperbilirubinemia is however a serious consequence of falsely low TcB values and should be avoided. We suggest an approach wherein not only TcB but also risk-factors and clinical picture are assessed. When in doubt, a TSB should be obtained. Since all TcB values above 250 µmol/l (~15 mg/dl) and nearly all (98.1%) TcB values above µmol/l (~13 mg/dl) were confirmed by a TSB with our current protocol, we would not recommend these margins to complement our current 50 µmol/l clinical margin. Only one combined measurement (2.1%) had an overestimation of TSB >50 µmol/l, making it likely that there were only few unnecessary blood draws due to overestimation in our patient population. Cost-Effectiveness Would we continue our current TcB protocol in our hospital, it is estimated that the potential cost savings compared to serum bilirubin measurements are 425, - annually for the newborn nursery. However, the cost-effectiveness of transcutaneous bilirubinometry is difficult to generalize, since the costs of TSB measurements vary greatly between countries and even between hospitals within a country. Other expenses (i.e. cost of readmission, duration of hospital stay) are not considered in this equation. A number of studies suggest the cost of a TcB device is offset by the decrease in TSB measurements.(28,40,51) 4.3 Strengths and limitations The main strength of our study is the design as a randomized controlled trial. While there are many studies addressing various aspects of TcB devices, to our knowledge, this is the first randomized controlled trial in a European population studying the implementation of a transcutaneous bilirubin device in the newborn nursery setting. L. den Boer 19

21 This report contains the results of a scheduled interim-analysis, to monitor the recruitment and outcomes of this trial. Since the results of this analysis demonstrated a both statistically and clinically relevant primary outcome, it was decided to discontinue the trial. We believe no additional clinical implications are to be expected when continuing this trial to the calculated sample size. Our outcomes suggest we would submit the patients in the control group to an inferior diagnostic method with unnecessary blood draws, which is unethical in our opinion. Although we have a reasonably large study population of 254 neonates, this study is technically underpowered. Outcomes that show no statistically significant difference between groups in our data should be interpreted with caution, since our study population could be too small to detect these differences.(52) Ideally, the clinically relevant parameter for hyperbilirubinemia would be the risk of developing bilirubin encephalopathy. Since bilirubin encephalopathy fortunately is rare, we have to assume a high TSB level is the predictor of adverse neurological outcome. We also make the assumption that any intervention that lowers high bilirubin levels will also prevent bilirubin encephalopathy.(53) However, it is hypothesized TcB is a better predictor for bilirubin encephalopathy since it measures bilirubin in tissue, a hypothesis that remains untested.(15) A limitation of our study-design was the subjectivity of the visual assessment of jaundice in the control group. Firstly, since the study could not be blinded and personnel were informed of the aim of the study, there is a risk for bias. Secondly, neonates were assessed for eligibility when they were jaundiced as indicated by nurses. The infants were assessed by nurses with different degrees of training and experience, in varying lightning conditions leading to interobserver variation. Consequently, it is very likely not all infants that were actually jaundiced were also regarded as jaundiced by the nurses. Furthermore, the post-natal ward is at times a very busy department, making it likely study protocol was not always adhered to completely, as can be expected of a trial in a routine clinical setting. There was also an inter-observer variation and bias in attending physicians when assessing the need for a TSB in the control group, since this was also based on assessment of the yellow colour of the skin, along with clinical picture and risk factors. On the other hand, a positive consequence of this real-life setting is that our results can be translated to a routine clinical setting, where many different health care providers asses jaundiced neonates. A further limitation is the time delay between TcB and TSB in paired measurements. Ideally, the TSB would be taken at the same time, or at least within minutes of the TcB, but since agreement between TcB and TSB was not the primary aim of our study this criterion was not met. A final drawback of this study is the limited generalizability of our results. Since our patient population is predominantly Caucasian (~80%), our findings should not be generalized to populations with more diverse ethnicities. Also the studied neonates were al healthy infants in the post-natal newborn nursery, for findings on (very) premature and sick neonates our study is not informative. Finally, we exclusively use the JM-103 device in our study, this prevents us from generalizing these results to other transcutaneous bilirubinometers. 4.4 Recommendations for future research This study investigated the implementation of the use of a transcutaneous device in a newborn nursery in jaundiced infants before phototherapy. Since a systematic review and meta-analysis by Nagar et al.(4) found an absence of evidence regarding the use of TcB devices during and after phototherapy, further research in needed in this area. L. den Boer 20