We conducted a cross-sectional study at the neonatology Unit of the Essos Hospital Centre in Yaoundé from 28 January to 27 June 2019.

The following were involved in this study:

­ All full-term newborn infants (from 37 gestational weeks) aged between 0 - 7 days presenting clinical signs of jaundice, and whose parent or guardian agreed to participate in this study.

The following were excluded from this study:

­ Infants with prior phototherapy;

­ TcB measurements and /or sampling not done.

We conducted a consecutive sampling. The sampling size was calculated using

the COCHRAN formula N = Z 2 ∗ P ( 1 − P ) T 2 with “N” representing the sample

size, “z” the confidence interval at 95% , “p” the estimated prevalence of severe neonatal jaundice estimated at 6% and “t” the margin error risk at 5%. Based on this formula our sample size stood at a minimal of 87 full term newborn babies.

Recruitment and testing

Data was collected using a questionnaire designed by the main investigator and validated by the supervisor following testing amongst mothers for reliability and feasibility. After obtaining the parent or guardian’s informed consent, a short questionnaire was filled during face-to-face interview gathering information on each newborn prior clinical examination and screening.

➢ Clinical examination

This was a general physical examination to screen for physical signs of craniocaudal jaundice, in daylight or with a white radiation pocket torch.

Following a clinical evaluation of the newborn and confirmation of diagnosis by a paediatrician or health personnel neonatology unit, we carried out testing.

➢ Transcutaneous bilirubin measurement

All transcutaneous bilirubin measurements were done with a jaundice meter, the MBJ20^® transcutaneous bilirubin meter by Beijing M & B Electronic Instrument Co. Ltd. (Beijing, China). Measurements were taken over the forehead and the sternum of the newborn in the supine position. The bilirubin level in each measuring site was clearly displayed on the digital screen in mg/dl. Both values were reported on our data collection sheet and the highest value was chosen for the study in mg/l.

➢ Total serum bilirubin (TsB) dosage

Blood samples were collected from a peripheral vein; two millilitres (2 ml) of venous blood within 5 - 10 minutes following the transcutaneous bilirubin measurement and transferred in heparinized sample tubes or bilirubin tubes in aseptic conditions.

TsB dosages were carried out in the hospital’s clinical chemistry laboratory. The proportioning principle was calibrated daily based on the Diazo method (use of diazonium ion) with the Biotechnical 1500 biochemistry Automate.

Statistical method: the statistical analysis was carried out with the R 3.6.0 version software. Our results were presented in form of tables, plots and figures expressed per frequency, per enrolment for qualitative variables, per mean and standard deviation for quantitative variables. The R package EpiR was particularly useful. The association between the two measurements was evaluated by the Pearson correlation coefficient r, the Student test for paired data was used to compare the means of samples, the concordance correlation coefficient (CCC), and Bland and Altman plots were used to assess the agreement between TsB and TcB. In all statistical analysis, a significance threshold of 5% was retained.

Ethical Considerations

Ethical clearance was obtained from the Institutional Human Health Ethics and Research Committee of the University of Douala (CEI-UD) under the number 1878 CEI-UDo/05/2019/T, as well as a research authorisation from the administrative authorities of the EHC of Yaoundé. Participants were enrolled after written consent following the reading of the information sheet of the study. The identity of participants was kept confidential and the results obtained during the study were used for scientific purposes alone. The participants were involved after reading.

Table 1 summarizes the distribution of newborn infants per clinical and socio-demographic patterns. The sex ratio of the babies included was 1.25 favouring

male. Median Postnatal age was 3 days and 75% of the newborn had a birth weight ranged between 2500 and 4000 g, mostly full term (Mean gestational age was 38 weeks) and predominantly breastfed.

The mean TcB corresponding to the maximum average between frontal and sternal measurement was 153 mg/dl ± 48 and the average Tsb was 123.80 mg/dl ± 50.48 (Table 2).

The population under observation was concentrated in 100 - 150 mg/l and 150 - 200 mg/l ranges. For the Tsb variable, the population under observation was highly concentrated in the <100 mg/l and 100 - 150 mg/l range.

As a reminder, the threshold value was 100 mg/l for diagnosis of hyperbilirubinemia as suggested by the National Institute for Health and Care Excellence (NICE). The highest average bilirubin level was obtained with transcutaneous bilirubin from sternum) (TcBs) 149.25 ± 47.64 mg/l, followed by transcutaneous bilirubin from forehead (TcBf) 143.32 ± 45.31 mg/l. The lowest average bilirubin level was found in total serum bilirubin, Tsb 123.80 ± 50.48 mg/l. The bilirubin intervals measured with these methods were 45 - 327 mg/dl for Tsb, 61 - 328 mg/dl and 59 - 329 mg/dl for TcBs and TcBf, respectively.

Table 3 presents the correlation coefficient between Tsb, TcBf and TcBs for all paired measurements. We have noted a strong correlation between the two TcB measurement sites (forehead and sternum) r = 0.91 (95% CI = [0.86 - 0.94]). A positive correlation was observed between Tsb and TcBf (r = 0.78, 95% CI = [0.68 - 0.85]) and Tsb and TcBs (r = 0.86, IC at 95% = [0.79 - 0.90]). Figure 1 shows a scatter plot between TcB values and Tsb. The right linear regression line is also represented. The Pearson’s linear correlation between TcB and Tsb was r = 0.86 [0.80; 0.91] showing a good linear correlation. The concordance correlation coefficient estimates in the general sample was 0.2 [0.05, 0.33]. Even though its 95% confidence interval does not contain 0, one would conclude on a poor agreement between TsB and TcB measurements.

Figure 2 represents the Bland and Alman Plot of TsB measurements versus TcB. The dotted lines (red and green) represent the upper and lower limits of the deviations 21.14 mg/l and −80.50 mg/l respectively. The blue line represents the bias = −29.68 mg/l. It is observed that the limits are wide and the bias −29.68 mg/l is statically significant (p < 0.01). This result indicates systematic bias, as TcB measurement overestimates the TsB mesurements.

In Table 4, it is presented CCC and bias estimates according to the levels of TcB measurements. Basically, for TcB measurements <100 mg/l, there is a fair agreement between TcB and TsB measurements as the CCC estimate was 0.59 [0.11, 0.85] and the bias (12.33, p = 0.08) was not significant. For TcB measurements > 100 mg/l, the bias was significant and a poor agreement was observed. At postnatal age more than 2 days, it is observed a fair agreement between TcB and TsB as the CCC estimate was 0.61 [0.47, 0.73], although the bias was significant. However, for postnatal age of 1 day and 2 days the CCC confidence interval estimates contains 0, indicating poor agreement for newborns with these postnatal ages. Basically, Table 4 indicates poor agreement between TcB and TsB, and that the TcB would overestimate the true TsB measurements.

The sample of this study was mostly constituted of male infants at 55.6%. This profile is similar to those of Oyapero et al. [9] and Madubuike [10] who had a high incidence of male infants in Nigeria.

We found a strong correlation between the two TcB testing sites (forehead and sternum) r = 0.91; our data are consistent with similar studies and protocols [8] [11] [12]. Conversely, for others, transcutaneous bilirubin measurements on the forehead had a better correlation with serum bilirubin [9] [13] [14]. The correlation

between TcB/TsB in this study is consistent with other sites, which, although using other devices, have indicated a strong association between TcB and TsB measurements, with correlation coefficients ranging from 0.75 to 0.95 [9] [14] [15].The correlation coefficient as used in many literature works to investigate agreement between TsB and TcB may be unappropriated to decide which tools to use. In fact, correlation does not imply agreement, and that constitute a pitfall for many research works [16] [17]. This study is among the firsts investigating the agreement between TsB and TcB in dark skinned population.

The average difference between mean TsB and TcB according to the Bland and Altman method was −29.68 mg/l [14]. This inaccuracy was greater than the observation of Oyapero, Raimondi, Romagnoli [6] [9] [18]. It is worth noting that we worked with a black population and previous studies have shown that the bias between TcB and TsB values was higher in black subjects [15].

Based on concordance correlation coefficient estimate for different levels of TcB thresholds, for TcB values below 100 mg/l, a fair agreement was observed. This finding is similar to those found in similar studies [14] [15] [19] [20] [21].

In this setting, the majority of the babies aged more than 3 days at inclusion, the potential applications of our findings may be scarce for early screening. However, as the agreement was higher after 2 days of life, a recommendation of systematic screening of jaundice using TcB in case of early discharge can be raised [22] [23]. In addition, this strategy could also target and prioritize all the babies at risk of hyperbilirubinemia born from mothers of O+ blood group and those mothers with a past story of babies having suffered of neonatal jaundice.