We conducted a prospective comparative study in the University Hospital of Ioannina, Greece, during an eight-month period, 10/2022–5/2023. The Ethics Committee of the Institution approved the study (No: 950) and written consent from all parents was obtained at enrollment.

Neonates were eligible for enrollment if they were over 34 weeks of gestational age, over 72 h of age, and required evaluation of bilirubin levels as per clinical practice either in the Neonatal Unit or the Maternal Ward.

In each neonate, a venous blood sample of 250–500 μl was obtained for the measurement of bilirubin levels in the biochemical laboratory of our institution, as per routine clinical practice (reference measurement). Any decisions regarding the clinical management of each neonate were based on the laboratory bilirubin measurement only. The venous blood sample was obtained from each neonate by clean venipuncture with a 21-French gauge needle, allowing free dripping of blood into a BD Vacutainer® SST II Advance tube (BD-Beliver Industrial, Plymouth, UK). A minimum amount of 70 μl from the same venous blood sample was used for the evaluation of bilirubin with the POC analyzer, which was performed in the settings of the Neonatal Unit. At the same time, 70 μl of a capillary blood sample was obtained by heel prick into a similar collecting tube. The heel of the neonate was adequately warmed before pricking, avoiding the squeezing of blood and allowing the free drainage of blood. The capillary bilirubin measurement was performed with the POC analyzer in the setting of the Neonatal Unit, as well. We constantly analyzed the venous samples first, followed by the capillary samples, to keep a standardized analysis, and have potential tracking of the results available, if needed.

The Beckman Coulter AU5820 analyzer is used in our laboratory for bilirubin (total and direct) measurement (Beckman Coulter, Inc., Brea, CA, USA). A stabilized diazonium salt, 3,5-dichlorophenyldiazonium tetrafluoroborate, reacts with bilirubin to form azobilirubin which absorbs at 570/660 nm. Caffeine and a surfactant are used as reaction accelerators. The absorbance at 570/660 nm is proportional to the bilirubin concentration in the sample. A separate serum blank is performed to eliminate endogenous serum interferences. Direct bilirubin couples directly with a diazonium salt of 3,5-dichloroaniline in an acid medium to form azobilirubin. The direct bilirubin in serum is directly proportional to the color development of azobilirubin which is measured bichromatically at 570/660 nm. According to the manufacturer, the criteria for non-significant interference is recovery within 10% of the initial value, and are estimated for hemolysis to a non-significant interference up to 500 mg/dL hemolysate for the total and up to 10 mg/dL for the direct bilirubin, and for lipemia to a non-significant interference up to 500 mg/dL intralipid for the total and up to 300 mg/dL for the direct bilirubin.

The POC Calmark Neo-Bilirubin analyzer is tested and controlled according to the International Electrotechnical Commission 61010–2-101 and fulfills the requirements for in vitro diagnostics Medical Device Directive 98/79/EC. The test cassette is constituted by four filters with different functions. The first two filters which are in contact with the blood are responsible for the blood filtration. Further, the plasma is separated and migrated through the third filter into the fourth (detection filter) by lateral flow. A color appears in the detection filter. A camera, inside the instrument, takes photos during the test. The color in the detection filter is further analyzed and converted through software into the corresponding numeric value of the bilirubin concentration in the blood sample and presented on the screen. The measuring concentration range of bilirubin is 9–29 mg/dL. The POC analyzer can analyze samples with hemolysis up to 2 g/L and level of hematocrit up to 60%.

The within-run precision using 3 different blood samples and 20 replicates according to the Clinical and Laboratory Standards Institute document EP05-A: Evaluation of Precision of Quantitative Measurement Procedures (CLSI EP5-A), has a coefficient of variation (CV) of 6.9% for bilirubin < 9 mg/dL, 5.2% for 9–15 mg/dL, and 4.4% for > 15 mg/dL. Imprecision was estimated over 20 consecutive working days for commercially available bilirubin controls according to the CLSI EP5-A to CV of 14.0% for low control, and 11.8% for medium control. Bias has been calculated and grouped into three concentration groups, at 12.9% for bilirubin < 9 mg/dL, 8.8% for 9–15 mg/dL, and 4.0% for > 15 mg/dL.

The perinatal data including gestational age, sex, birth weight, delivery method, and the clinical outcomes of the neonates were collected.

Perinatal data and clinical outcomes were expressed with descriptive statistics. Continuous variables were expressed as mean (standard deviation, SD) or median (interquartile range), whereas categorical variables as n (percentage, %). The normality of the distribution was examined with the Kolmogorov-Smirnoff test or the Kruskal–Wallis test, as appropriate.

The agreement between the POC capillary and venous bilirubin measurements, and between both POC measurements and the reference laboratory venous bilirubin measurements was examined with the Bland–Altman plot. The Bland–Altman method of agreement between two quantitative measurements determines the bias (mean difference between the reference and alternative methods) and limits of agreement (LOA) resulting from the bias ± 1.96 times the SD of the bias. The systematic and the proportional bias were examined with the Passing-Bablok regression analyses, evaluating the intercept (a) and the slope (b) with their 95% confidence intervals (CI), respectively, of the regression formula: y = a + b*x (where y is the tested, and y is the reference method). If 95% CI for intercept includes value zero, there is no constant difference between the two methods. Accordingly, if 95% CI for slope includes value one, there is no proportional difference between the two methods. Statistical significance was defined for p < 0.05. The total error between the POC capillary and venous bilirubin measurements and the reference laboratory venous bilirubin measurements was examined and evaluated according to the total allowable error suggested by Clinical Laboratory Improvement Amendments of 1988 (CLIA) Proficiency Testing regulations, that was for bilirubin ± 20%, or 0.4 mg/dL [18]. Finally, we examined whether there was any difference between the POC and the reference methods in detecting neonates with bilirubin levels necessitating phototherapy. We calculated the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of both POC capillary and venous methods for the correct identification of neonates necessitating phototherapy. The receiver operating characteristic (ROC) curve analyses and the Area Under the Curve (AUC) were constructed.

A power analysis revealed that to achieve a power of 80% and a type I error α of 0.05, considering the study design, a pair of more than 45 samples was adequate to detect the desired absolute difference of bilirubin levels between any two methods of 10%. Data were analyzed with Microsoft Excel 365 (Microsoft, Redmond, WA, USA), and SPSS statistical software version 25.0 (SPSS Inc., Chicago, IL, USA). Plots were done using Prism 9.0c (GraphPad Software, La Jolla, CA, USA).