2023 exceptional surveillance of Jaundice in newborns under 28 days (NICE guideline CG98)

Studies on risk factors for hyperbilirubinaemia

One study was included looking at risk factors for hyperbilirubinaemia. Hamadneh et al. (2016) compared 2 groups of women who were delivered by caesarean section and who have had at least 2 previous caesarean sections. Group 1 (n=505) was women who underwent caesarean section at 37 weeks and group 2 (n=381) was women who underwent caesarean section at 38 weeks or later. Results from multivariate analysis showed that neonatal jaundice was more prevalent in group 2 in which babies were delivered at 38 weeks or later (adjusted odds ratio 2.1, 95% CI 1.7–2.7; P=0.035).

Visual assessment of jaundice

Five studies used visual assessment for determining hyperbilirubinaemia, these included the use of visual icterometer (Bili-ruler, Bhutani 2019), Kramer's method (Dionis et al. 2021;), maternal visual assessment (Kittiarpornpon et al. 2020), colour card (Singh et al. 2022) or two colour icterometer (Bilistrip, Olusanya et al. 2017). These studies were carried out in Bangladesh, Tanzania, Nigeria, Thailand and India. We have assumed the majority of the babies in these studies have darker skin tones.

Bhutani (2019) reported the use of visual icterometer (Bili-ruler) at TSB ≥11 mg/dL, with 84.5% (95% CI, 79.1%–90.3) and 83.2% (95% CI, 76.1%–90.3%), sensitivity and specificity, respectively. For TSB >17 mg/dl, Bili-ruler performed moderately well, 87.8 (95% CI, 80.9–95) and 66.5 (95% CI, 59.6–73.3), for sensitivity and specificity, respectively.

Olusanya et al. (2017) utilised two colour icterometer; dark yellow and light yellow, (Bilistrip). The mean TcB in infants with dark yellow was significantly higher than those with light yellow. Bilistrip™ was associated with increasing sensitivity (47.0% - 92.6%) and NPV (91.4% - 99.9%) against increasing TcB thresholds (≥10mg/dL, ≥12mg/dL, ≥15mg/dL, and ≥17mg/dL) (n = 124).

Kittiarpornpon et al. (2020) compared maternal visual assessment with TSB for detecting hyperbilirubinaemia requiring phototherapy [sensitivity, 91.7%, (95% CI: 73.0–99.0)] and for identifying hyperbilirubinaemia [sensitivity, 92.9% (95% CI: 76.5–99.1)].

Singh et al. (2022) used "Colour Card" initially by yellow colour shades that fall into 4 bilirubin categories, i.e. TSB up to 7 mg/dl, 7.1 to 12 mg/dl, 12.1 to 18 mg/dl and >18 mg/dl. The overall accuracy of colour card in measuring various TSB ranges varied from 75% to 96.8%. The agreement between two observers was 85.6% (Cohen's kappa co-efficient: 0.61, p-value: .0001) overall and was 92.3%, 86%, 84%, 81.2% for each of the four bilirubin categories in ascending order.

Kramer's method was used by Dionis et al. (2021) in babies of black descent and showed sensitivity of 70.5% and specificity of 86.1%.

Limitations of the studies on visual assessment

One of the limitations of this evidence was that it is 'assumed' the study populations have darker skin tones as the studies were conducted in countries with majority of population as non-Caucasians. This assumption may not be true. Another limitation was that different devices or methods and different thresholds for TSB were used across the 5 studies to conduct visual assessment and therefore no direct conclusions can be drawn based on these different methods and different TSB thresholds. Also, a key limitation of these studies was that there was no comparator group of babies with lighter skin tones, hence it is impossible to determine the utility of visual assessment to identify jaundice in babies in darker skin tones and how it differs to babies with lighter skin tones. All the included study populations had high TSB, clinical opinion would indicate that jaundice would be visible at 5mg/dL and as such limit the external validity of findings from these studies to the UK context.

Impact statement for visual assessment studies

Due to variations in the studies, it is unclear whether the new evidence challenges the existing recommendations. All the studies were carried out in babies assumed to have darker skin tones; however, absence of comparator group with babies who have lighter skin tones makes it impossible to conclude that the accuracy of visual assessment will vary depending on skin tone based on this evidence. Therefore, the evidence does not support an update on existing recommendations.

Transcutaneous bilirubinometers to assess jaundice.

11 studies investigated the accuracy of TcB in comparison to TSB using various devices (KJ-8000, Drager JM103, Drager JM 105, BiliChek, Bilistick)

Nine of these studies were conducted on babies with 'assumed' darker skin tones (conducted in countries with majority of population as non-Caucasians), with no comparator group of babies with lighter skin tones. Two studies compared babies with darker skin tones versus lighter skin tones.

Kejian 8000 device

One study used Kejian 8000 to determine TcB in Tehran (Afjeh et al. 2015).

The study was a prospective cross-sectional study that included 613 babies with gestational age of 35 weeks or above. It reported high TcB in 491 babies (those with TcB ≥ 5 mg/dL in first 24 hours and > 8 mg/dL in second 24 hours). Serum sample was taken from those 491 babies with high TcB, and 398 out of 491 babies also had high TSB, with a correlation of 72% (p<0.001).

The study selectively checked TSB in babies with high TcB in the first and second 24 hours and reported positive predictive values of 81% in diagnosis of hyperbilirubinaemia. The study does not suggest what TcB threshold is most accurate against TSB for babies with 'assumed' darker skin tones, or whether TcB differs compared to babies with lighter skin tones. These findings are not conclusive nor sufficient to determine the utility of KJ-8000 on babies with darker skin tones.

Drager JM 103 device

Four studies compared TcB measured by Drager JM 103 with TSB in babies with 'assumed' darker skin tones (Gunaseelan et al. 2018; Villanueva-Uy et al. 2022; Chimhini et al. 2018; Shihadeh et al. 2016). Two studies included babies of gestational age between 35 weeks and 37 weeks (Gunaseelan et al. 2018; Villanueva-Uy et al. 2022). One study included 283 babies with median gestational age of 38 weeks (25-42 weeks), of which 115/283 (41%) were preterm babies (Chimhini et al. 2018). The fourth study did not report gestational age (Shihadeh et al. 2016). These studies were carried out in India, Philippines, Bahrain and Zimbabwe, where the majority of populations are 'assumed' to have darker skin tones.

Three studies compared paired/simultaneous measurements between TcB and TSB (Gunaseelan et al. 2018; Villanueva-Uy et al. 2022; Shihadeh et al. 2016). The correlation measures between TcB (sternum) and TSB was 0.91, and between TcB (forehead) and TSB 0.88 (Villanueva-Uy et al. 2022). A study by Gunaseelan et al. (2018) did not report the correlation value for paired measurements of TcB and TSB but a statistical significance level of p<0.001 was reported for TcB at low-risk and medium-risk thresholds of phototherapy. Shihadeh et al. (2016), reported correlation between paired measurements of TcB and TSB as 0.75 (p<0.0005) with a mean difference of 1.09±2.16 mg/dL (range of differences: 6.18 mg/dL – 7.00 mg/dL) (Shihadeh et al. 2016).

Chimhini et al. (2018) also carried out subgroup analysis of correlation for those 115 preterm babies, with the correlation of 0.77 for sternum TcB and 0.70 for forehead TcB with TSB. For the other 168 term babies, correlation was 0.76 for sternum and 0.70 for forehead. Bland-Altman plot showed agreement between TcB and TSB in this study (no further details from abstract). Other two studies did not report Bland-Altman results.

Overall, the evidence identified using the Drager JM 103 device is unclear, inconsistent and does not suggest a clear threshold of TcB for babies with 'assumed' darker skin tones in comparison to babies with lighter skin tones. The populations are of heterogenous gestational age and paired correlations were sometimes only available for subsets of babies.

Drager JM 105 device

Two studies used Drager JM-105 to determine the accuracy of TcB against TSB in India and Malaysia (Sharma et al. 2022; Mohamed et al. 2022). The correlation between TcB and TSB was 0.892 (P<0.001) (Mohamed et al. 2022). Sharma et al. (2022) reported correlation of TSB separately for forehead TcB and sternum as 0.82 and 0.80 respectively.

The average error in evaluating hyperbilirubinemia with TcB compared to TSB was 0.101, with limits of agreement between −3.73 and +3.55 (Sharma et al. 2022). Mohamed (2022) also reported that TcB underestimates TSB with a mean difference of 10.10 µmol/L at the forehead and 9.27 µmol/L at the sternum.

The area under the curve at three TSB levels (>10 mg/dl, >12 mg/dl, and >15 mg/dl) was 0.860, 0.892, and 0.849 (Sharma et al. 2022). A good discriminations ability was observed for both the TcB forehead (ROC curve = 89.8%) and sternum (ROC curve = 89.7%) at a TSB level of 205 µmol/L (Mohamed et al. 2022).

Both studies were carried out in babies with 'assumed' darker skin tones and neither included a comparator by skin tone. Overall, the sample size was relatively small in both studies (120 and 130 babies). Sharma (2022) included babies with visually identifiable jaundice and the other study had babies with jaundice that required TSB determination (Mohamed et al. 2022). Potential heterogeneity in outcomes arises from timing of assessment, gestational age and variation in TSB thresholds which are not aligned with UK clinical practice. Both studies reported an underestimation of 10µmol/L for accuracy of TcB against TSB.

BiliChek device

One study compared TcB measured by BiliChek with TSB in jaundiced term (37-42 weeks) babies (n=665) in Saudi Arabia (Alsaedi, 2016). Paired values of TcB and TSB indicated a correlation of r=0.84 (CI: 0.82-0.86; p<0.001). The results showed that TcB overestimated TSB with a mean difference of 17 µmol /L, 95% CI of 40 ± 77 µmol/L. The sensitivity was 83% and specificity 71%. Again the study did not report a specific cut-off of TSB but indicated that bilirubin level at low and above intermediate risk zone was considered significant (definition of low and intermediate risk zone was not reported in abstract). The findings are only reported in (assumed) darker skin tones babies and not compared with lighter skin tone babies, therefore it is uncertain whether the overestimation of TcB will apply specifically to babies with darker skin tones.

Bilistick system

One study used Bilistick system to measure TcB in 1458 babies in 17 hospitals from Nigeria, Egypt, Indonesia, and Vietnam (Greco et al. 2018). TSB level measured by Bilistick system correlated well with the lab result in all four countries, with positive predictive value (PPV) of 92·5% and a negative predictive value (NPV) of 92·8%. As with previous studies, it was 'assumed' that babies were with darker skin tones based on the country where data was collected and assessed. Results were not stratified based on skin tones or ethnicity. No other findings were reported. Due to the limitations of the study, especially reporting bias, the evidence on Bilistick system is not sufficient to conclude its accuracy.

Studies comparing babies with darker skin tone versus lighter skin tone

Two studies which compared TcB with TSB in different skin tone populations were identified, Maya-Enero et al. (2021) used the Drager JM105 device for TcB and Starwociz et al. (2019) used the KJ 8000 TcB device.

Maya-Enero et al (2021), included 1359 babies who were assigned to different groups at 24 hours of life according to Neomar's skin colour scale (Maya-Enero et al. 2020) into four categories: light (colour 1) n=337, medium-clear (2) n= 750, medium-dark (3) n=249, and dark (4) n=23. Correlation between TcB and TSB range was reported as r² = 0.908–0.956, with slight differences between darker and lighter skin tones. Correlation for colour 1 was 0.935 (95% CI 0.921; 0.947), for colour 2 was 0.924 (95% CI 0.913; 0.933), for colour 3 was 0.908 (95% CI 0.887; 0.926), and for colour 4 was 0.956 (95% CI 0.914; 0.978). Bland-Altman plots also showed differences in mean bilirubin bias depending on skin colour with the bias increasing from colour 1 to colour 4. The difference increased gradually from colours 1 to 4 and was statistically significantly different between colours 1 and 2, between colours 1 and 3, between colours 1 and 4, and between colours 2 and 3, and between colours 2 and 4 (p value < 0.001), but not between colours 3 and 4.

In a multiple linear regression, serum bilirubin was associated with TcB (β = 0.88, 95% CI 0.87; 0.90, p < 0.001). Considering colour 1 as reference, β coefficients were for colour 2 was β = − 0.22, [95% CI – 0.40; − 0.04], p = 0.019; for colour 3 β = − 0.81, [95% CI – 1.04; − 0.58], p < 0.001; and for colour 4 β = − 0.80, [95% CI – 1.32; − 0.28], p = 0.003. Bland Altman plot for all skin colours showed that generally TcB underestimated TSB for TcB > 15 mg/dl (for TcB 15: LoA, 95% CI – 1.73 (− 6.02; 2.55)) but overestimated it for TcB ≤ 15 mg/dL (for TcB 15: LoA, 95% CI 1.06 (− 3.87; 1.75)). The overestimation was greater in darker skin tones, with mean overestimation, 0.7 mg/dL for light; 1.08 mg/dL for medium light; 1.89 mg/dL for medium dark; and 1.86 mg/dL for dark; P < .001 for light vs medium dark or dark). This study concluded that TcB is more likely to overestimate TSB in darker skin tones babies compared with neonates with lighter skin tone.

Maya-Enero et al (2021) included only 23 babies in the dark skin group (colour 4), and Bland-Altman results may not be comparable between colour 4 and colour 1, 2, and 3. The study included 1549 paired SB/TcB measurements (379 for colour 1, 828 for colour 2, 308 for colour 3, and 34 for colour 4) and mentioned that some babies had more than one determination of SB/TcB but all paired measurements were included. This could result in duplication of data points and might affect estimates of outcomes. Only healthy babies were included in the study with mean gestational age of 39 weeks. This study is based only in one hospital and hence results might only be relevant to representative of the population in the area.

Starwociz et al (2019) carried out a prospective study comparing TcB and serum bilirubin in Caucasian (n=76) and non-Caucasian (n=24) preterm (<32 weeks gestation) and term babies. Overall correlation was 0.8 (p<0.0001) between TcB and serum bilirubin levels. Correlation in term babies was r=0.82 and preterm babies r=0.49. The Caucasian group had stronger correlation r=0.84 compared to non-Caucasian group r=0.71. Overall, the bias was −5.9 μmol/L (95% CI: −101, 89) which was not evenly spread, and TcB tending to overestimate at lower serum bilirubin levels and underestimate at higher levels of serum bilirubin. TcB was overestimated and less precise in non-Caucasian babies.

The study sample size was small and thus outcome measures may not be comparable. There was also underrepresentation of term babies compared to preterm babies and multiple readings were collected more from preterm babies. Due to such imbalances, the bias and precision might have been affected.

Overall impact statement (TcB devices)

Currently the NICE guideline (CG98) recommends visual inspection as the initial assessment approach for suspected jaundice, followed by dependent on risk measurement of serum bilirubin or transcutaneous bilirubin. In light of the HSIB and NHS Race observatory reports, this exceptional review explored whether TcB could play a role as initial assessment for babies with darker skin tones in whom visual inspection may be difficult or unreliable.

No conclusive evidence was available on accuracy of TcB using various devices in babies with darker skin tones. The evidence had a lack of comparator groups of babies with lighter skin tones. Additionally, different thresholds for TSB across studies and a number of studies did not report cut-offs used limiting conclusions that can be drawn. It is worth noting that the study population of babies with darker skin tones was 'assumed' in these studies, given the countries where these studies were conducted.

Only two studies compared babies with darker versus lighter skin tones in the same setting. One used Drager JM 105 (Maya-Enero et al. 2021) and another used KJ 8000 (Starwociz et al. 2019) for TcB measurements. Maya-Enero et al. (2021) used Neomar's neonatal skin colour scale with four categories: light, medium-light, medium-dark and dark. These categories were defined solely based on skin colour regardless of ethnicity. Both studies reported a trend for overestimation of TSB in darker skin tones, however the evidence is of limited quality to allow definitive conclusions to be drawn on the role of TcB in identifying jaundice in babies with darker skin tones in whom visual inspection may be difficult or unreliable and thus at this point in time does not impact on the guideline.

Risk factors of risk of hyperbilirubinaemia

The section falls under the guideline review question 'What are the factors associated with an increased risk of hyperbilirubinaemia' and forms the recommendation 1.2.1, which was developed in 2010. During the development of this recommendation, 10 studies were included. Evidence from good-quality studies showed that four factors are independently associated with an increased risk of hyperbilirubinaemia; gestational age < 38 weeks, jaundice within 24 hours of birth, increase in severity of clinically apparent jaundice and intention to breastfeed exclusively. Other factors such as cephalohaematoma, vacuum delivery, male sex or race showed no statistically significant association with hyperbilirubinaemia. The committee noted that it is commonly believed that bruising, cephalohaematoma and vacuum delivery all contribute towards development of hyperbilirubinaemia, but the evidence was inconclusive to include those as risk factors.

This recommendation has not been updated since 2010.

Assessment of hyperbilirubinaemia

The guideline review question ''What is the accuracy of various tests (clinical history and examination, urine/stool examination, icterometer and transcutaneous bilirubin levels) in recognising neonatal jaundice or hyperbilirubinaemia?' was updated in 2016 as part of NICE guideline CG98. The evidence considered at that time comprised of 25 studies (with another 7 studies from 2010 original guideline, total 32 included studies), but only two studies specified babies of different skin tones (Wainer 2009; Karen 2009).

The committee noted that the evidence for the review question was limited: 1 study on clinical history and examination, no evidence on urine and stool examination: no evidence on icterometer and 31 studies on transcutaneous bilirubin. The committee noted that there were a number of limitations in the evidence base; 15/32 studies screened all newborns despite absence of clinical symptoms which resulted in lack of generalisability; inappropriate reference method; unclear information on whether phototherapy was given before the study; unclear reporting of infant's postnatal age; high uncertainty of the accuracy estimates due to wide confidence intervals. In addition to that, a range of diagnostic measures were reported using different devices measuring TcB. The diagnostic devices used in the studies varies and hence generates heterogenous readings of TcB with no obvious differences in accuracies when compared to TSB. Due to these reasons, the committee noted that the evidence was not sufficient to inform recommendations on diagnostic accuracies of tests to determine hyperbilirubinemia in newborns.

The committee also noted that the evidence on babies with different skin tones was still limited to inform recommendations based on skin tones.

Care of babies with prolonged jaundice

The guideline recommendations on care of babies with prolonged jaundice was largely based on clinical consensus at the time of development. Two case series (one UK and one Turkey) described some investigation and demographic correlations in prolonged jaundice cohorts. The Committee recognised the importance of clinical examination and key investigations including total and conjugated bilirubin, urine culture and where appropriate G-6-PD testing.

See ensuring that published guidelines are current and accurate in developing NICE guidelines: the manual for more details on our consultation processes.

Risk factors and diagnostics for neonatal jaundice

The HSIB report highlights concerns about risk factors for jaundice. Only one poor quality study was identified concerning risk factors for neonatal hyperbilirubinemia. As such, there is insufficient new evidence to update the recommendations.

Both the HSIB and NHS Race observatory reports raise questions about the diagnostic pathway for suspected jaundice. The guideline takes a stepped approach to diagnostic assessment from visual inspection to TcB or TSB dependent on the age and gestational age of the babies. The evidence identified in this review was inconclusive about the impact of skin tone on diagnostic accuracy of various approaches to detect jaundice/measure bilirubin, largely due to studies not having comparator groups. At present the recommendations in the guideline are uniform regardless of skin tone due to lack of evidence in 2016. Therefore, it remains uncertain whether the visual assessment pathway entry point results in different outcomes in babies with darker skin tones compared to those with lighter skin tones.

As babies with darker skin tones was an identified subgroup in the 2016 update, this exceptional surveillance review also explored whether there is new evidence to suggest the diagnostic accuracy of transcutaneous bilirubin measurement (TcB) may vary depending on different skin tones compared with standard reference to total serum bilirubin (TSB). Based on the evidence identified and assessed in this exceptional surveillance review, there is very limited evidence (2 studies) to suggest that TcB overestimates the TSB in babies with darker skin tones compared to those with lighter skin tones, although a conclusion cannot be drawn as to exactly how large this overestimation is. In conclusion there is currently insufficient new evidence to update the consensus-based bilirubin threshold table for a specific subgroup of babies with darker skin tones.

The HSIB report raised two areas of concern for NICE to explore, the reliability of visual signs to detect jaundice in newborns, particularly in babies with darker skin and risk factors for jaundice. The evidence identified in this surveillance review does not support changing the recommendations at this time. Given the overlap in concerns about case recognition potentially being variable dependent on skin tone raised by both NHS Race observatory and HSIB we propose to add a bullet to recommendation 1.2.5 to highlight that visual inspection may be harder in darker skin tones. The new recommendation proposed (underlined):

1.2.5 When looking for jaundice (visual inspection):

Prolonged jaundice

The BSPGHAN audit highlights that the guideline recommendations in section 1,7 are not being implemented uniformly in NHS trusts included in the audit for the management of prolonged jaundice. When the recommendations were developed it was an evidence sparse area searches were not performed as part of this surveillance review. The audit also highlights no impact on outcomes of differential implementation of investigations. The implementation challenges for urine culture raised by BSPGHAN are acknowledged, given the lack of evidence when the recommendations were developed it is proposed that the wording of recommendation 1.7.1 is amended to reflect the implementation challenges of carrying out urine culture, the lack of evidence and concerns of resource and impact on patient and carer experience . The proposed revised recommendation is (underlined);

In babies with a gestational age of 37 weeks or more with jaundice lasting more than 14 days, and in babies with a gestational age of less than 37 weeks and jaundice lasting more than 21 days:

Additionally, as part of their work, the BSPGHAN has developed a simplified pathway related to the audit which the RCPCH are considering endorsing into a clinical guideline. Given the specialist society interest in this area we propose to not update the recommendations on care of prolonged jaundice beyond the amendment regarding urine culture.

For further details and a summary of all evidence identified in surveillance, see appendix A.