Scolaris Content Display Scolaris Content Display

Base de datos Cochrane de Revisiones Sistemáticas Protocolo - Intervención

High‐ versus low‐dose conventional phototherapy for neonatal jaundice

Declaraciones de intereses de los autores

Versión publicada: 30 abril 2020 Historial de versiones

https://doi.org/10.1002/14651858.CD003308.pub2
Contraer todo Desplegar todo

Abstract

Objectives

This is a protocol for a Cochrane Review (intervention). The objectives are as follows:

  1. To assess the effects of high‐dose conventional phototherapy versus low‐dose conventional phototherapy on bilirubin level and associated clinical outcomes that constitute the major conditions in the spectrum of bilirubin‐induced neurological dysfunction (BIND), such as acute bilirubin encephalopathy and kernicterus, as well as cerebral palsy and neurodevelopmental disabilities in infants with hyperbilirubinaemia. For the purpose of this review, we will include only studies on treatment and will exclude studies on prophylactic use of phototherapy. Prophylactic phototherapy is covered in another Cochrane Review (Okwundu 2012)

  2. To assess high‐dose phototherapy defined as the use of high levels of measured spectral irradiance greater than 30 μW/cm²/nm over the same bandwidth (AAP 2004), as well as low‐dose phototherapy defined as measured levels of spectral irradiance below 30 μW/cm²/nm

  3. To restrict this review to studies that specifically state measured irradiance

Background

Description of the condition

Neonatal hyperbilirubinaemia is the term that is used to describe elevated levels of bilirubin in the blood of a neonate. It is clinically apparent as jaundice with yellowish discolouration of the skin and the sclera at serum bilirubin levels greater than 5 mg/dL (Porter 2002). Hyperbilirubinaemia occurs in up to 60% of term and 80% of preterm infants (Kumar 1999; Maisels 1988; Woodgate 2015).

Neonatal hyperbilirubinaemia results from increased production of bilirubin due to lysis of red blood cells, decreased ability of liver cells to clear bilirubin, and increased enterohepatic circulation (Maisels 1988). Unconjugated hyperbilirubinaemia is the most common form of neonatal hyperbilirubinaemia. Conditions that further increase bilirubin production, alter the transport or metabolism of bilirubin, and increase the severity of hyperbilirubinaemia include blood group and rhesus incompatibility, prematurity, instrumental delivery, non‐optimal breastfeeding, glucose‐6‐phosphate dehydrogenase (G6PD) enzyme deficiency, and other hereditary haemolytic anemias (Christensen 2015; Huang 2004; Maisels 1988; Woodgate 2015; Yusoff 2006). Blood group (ABO) and rhesus incompatibilities increase bilirubin production through haemolysis (Kaplan 2014). Preterm infants have more severe hyperbilirubinaemia than their term counterparts due to immaturity of red blood cells, liver, and gastrointestinal tract (Watchko 2003). Infants with mutations or polymorphisms of the genes encoding for enzymes involved in bilirubin metabolism and infants with G6PD deficiency and other forms of hereditary haemolytic anaemia have a higher incidence of severe neonatal jaundice (Christensen 2015; Huang 2004; Riskin 2012; Yusoff 2006). Non‐optimal breastfeeding often results in significant postnatal weight loss in the infant, which increases the risk of jaundice developed through increased enterohepatic circulation (Gartner 2001). Instrumental delivery, especially vacuum delivery, causes blood sequestration in the infant's scalp, can result in an increased bilirubin load, and can exacerbate the severity of neonatal hyperbilirubinaemia (Arad 1982).

Unconjugated bilirubin can cross the blood‐brain barrier and be deposited in the basal ganglia, resulting in kernicterus. The exact level of unconjugated serum bilirubin that is neurotoxic remains unclear. Kernicterus has been reported at autopsy in infants without markedly elevated levels (Turkel 1980); it has also been noted at relatively low levels of serum bilirubin (Maisels 1987). It is generally accepted that the preterm brain is more susceptible to potential bilirubin toxicity; therefore guidelines recommend more aggressive management of hyperbilirubinaemia in preterm infants.

In 1985, the National Institute of Child Health and Human Development (NICHHD) reported that phototherapy was as effective as exchange transfusion in preventing brain injury in hyperbilirubinaemia, and phototherapy has been widely adopted as the initial treatment of choice for hyperbilirubinaemia (NICHHD 1985).

Description of the intervention

Phototherapy, or use of light, is the most common treatment for reducing bilirubin levels in neonates. A newborn infant receiving phototherapy typically lies enclosed within an incubator or placed on a nursing cot and is exposed to a light source. The infant is undressed to maximise exposure of the surface area to the light. It has been shown that the larger the surface area exposed to phototherapy lights, the faster the reduction in bilirubin levels (Hart 2005).

The dose of phototherapy is determined by the wavelength of the light, the intensity of the light (irradiance), the distance between the light and the infant, and the body surface area exposed to the light (Stokowski 2006). The effectiveness of phototherapy depends on energy output or irradiance of the light source. Irradiance (radiant power incident on a surface per unit area of the surface) is measured with a radiometer, and spectral irradiance (irradiance in a certain wavelength band) is measured with a spectro‐radiometer in watts per square centimetre per nanometre (W/cm²/nm), or in microwatts per square centimetre per nanometre (μW/cm²/nm) over a given wavelength band.

The level of irradiance in phototherapy has been reported to correlate with the rate of bilirubin decline (Vandborg 2012). Although the most effective and safest dosage of phototherapy is currently unknown (AAP 2004), use of blue light in the emission spectrum of 460 to 490 nm delivered at a light irradiance greater than 30 μW/cm²/nm to the largest possible body surface area has been recommended as the optimal way of administering phototherapy (Bhutani 2011).

High‐dose, or intensive, phototherapy has been defined as the use of high levels of spectral irradiance greater than 30 μW/cm²/nm over the same bandwidth delivered to as much of the infant’s body surface area as possible (AAP 2004). Low‐dose phototherapy has been defined as the use of levels below 30 μW/cm²/nm over the same bandwidth. The minimum irradiance for effective phototherapy has been shown to be 4.0 μW/cm²/nm (Bonta 1976).

Other considerations of the effectiveness of phototherapy include light source and distance to the infant. Conventional phototherapy lights consist of non‐light‐emitting diode (LED) units such as special blue fluorescent tubes, compact fluorescent tubes, and halogen spotlights. Use of a blue fluorescent tube as the phototherapy light source has been advocated (Tan 1992), as blue fluorescent lamps deliver 30 to 40 μW/cm²/nm, and standard daylight phototherapy units positioned 20 cm above the infant should deliver spectral irradiance (measured at the level of the infant) of 8 to 10 μW/cm²/nm in the 430‐ to 490‐nm band (Maisels 1996).

The distance of the light source from the infant affects the effectiveness of phototherapy. Bringing overhead units closer to the surface of the neonate increases measured irradiance, as irradiance from an extended source is approximately proportional to 1/distance from the source (Hart 2005). However, the ability of these light sources to provide intensive phototherapy and its effectiveness may be limited because of an inability to keep light sources close to the infant. This occurs because of heat emitted from the fluorescent tubes, which causes risk of overheating the infant at a reduced distance.

Factors intrinsic to the individual patient also contribute to the effectiveness of phototherapy. Rate of bilirubin production, physiologic milieu, coexisting pathologies, and functional status of the baby's endogenous clearance mechanisms all affect the ability of phototherapy to reduce serum bilirubin levels (Maisels 2008).

Complications from phototherapy are rare but are not entirely benign. Infants exposed to phototherapy have increased insensible water loss (Oh 1972). Exposure to light has been linked to persistent patent ductus arteriosus (PDA) in an animal model (Clyman 1978), as well as in human infants (Rosenfeld 1986). Phototherapy from overhead lights (e.g. "conventional" phototherapy) but not from a fibreoptic blanket was shown to blunt the postprandial increase in mesenteric blood flow (Pezzati 2000). Conventional phototherapy can also interrupt the maternal‐infant dyad at a time when breastfeeding and other mother‐infant interactions are occurring. The most important, albeit uncommon, complication is bronze baby syndrome, wherein infants with cholestatic jaundice, when exposed to phototherapy, develop a dark, greyish‐brown discolouration of skin, urine, and serum (Tan 1982a). Intensive phototherapy might also increase the number of atypical melanocytic nevi identified at school age (Csoma 2007). Reports have described a slight non‐significant increase in mortality amongst extremely low birth weight infants (501 to 750 g) (Morris 2008).

How the intervention might work

In neonates with jaundice, transient deficiency of conjugation is combined with increased turnover of red cells. The goal of therapy is to lower the concentration of circulating bilirubin and/or prevent it from increasing by using phototherapy. Phototherapy is thought to work via several mechanisms, namely:

  1. absorption of light by the normal form of bilirubin (4Z,15Z form of bilirubin) causes bilirubin molecules to be in a transient excited state and to react with oxygen to produce a colourless product of lower molecular weight that is excreted in the urine; and

  2. some of these molecules also become structural isomers (also known as lumirubins) or, more commonly, form isomers with at least one of the two Z‐configurations double‐bonding to an E configuration. As all these isomers are more hydrophilic than the 4Z,15Z form of bilirubin, they can be excreted unchanged in the bile without the need for conjugation (Maisels 2008).

Through phototherapy, bilirubin is converted into water‐soluble photo‐products that can bypass the hepatic conjugating system and be excreted without further metabolism (Ennever 1990). This molecular conversion occurs when bilirubin accumulating in the skin of jaundiced infants is exposed to light of wavelengths 425 to 475 nm (blue‐green spectrum) (Tan 1982).

Factors affecting the photo‐induced isomeric conversion of bilirubin include wavelength of light, intensity or irradiance of light, surface area exposed, type of light source used, and duration of exposure. Other factors affecting phototherapy effectiveness include the distance of phototherapy from the neonate (AAP 2004).

Why it is important to do this review

It is crucial to understand the optimal dose and the various factors affecting the dose administered to the neonate through different phototherapy devices to enable a clear recommendation for clinical practice. Although several randomised controlled trials (RCTs) have been published to address this, no systematic review to date has pooled the findings of these trials to provide an overall picture of the benefits and harms of this intervention to inform practice and research.

Objectives

  1. To assess the effects of high‐dose conventional phototherapy versus low‐dose conventional phototherapy on bilirubin level and associated clinical outcomes that constitute the major conditions in the spectrum of bilirubin‐induced neurological dysfunction (BIND), such as acute bilirubin encephalopathy and kernicterus, as well as cerebral palsy and neurodevelopmental disabilities in infants with hyperbilirubinaemia. For the purpose of this review, we will include only studies on treatment and will exclude studies on prophylactic use of phototherapy. Prophylactic phototherapy is covered in another Cochrane Review (Okwundu 2012)

  2. To assess high‐dose phototherapy defined as the use of high levels of measured spectral irradiance greater than 30 μW/cm²/nm over the same bandwidth (AAP 2004), as well as low‐dose phototherapy defined as measured levels of spectral irradiance below 30 μW/cm²/nm

  3. To restrict this review to studies that specifically state measured irradiance

Methods

Criteria for considering studies for this review

Types of studies

Randomised, quasi‐randomised, and cluster‐randomised controlled trials will be included in this Review.

Types of participants

Infants (defined as ≥ 37 weeks' gestation) and preterm infants (defined as ˂ 37 weeks' gestation) with hyperbilirubinaemia (defined by Porter 2002 as serum bilirubin levels > 5 mg/dL) who require phototherapy.

Types of interventions

Levels of irradiance

We define high‐dose, or intensive, phototherapy as the use of measured spectral irradiance ˃ 30 μW/cm²/nm over the same bandwidth delivered to as much of the infant’s body surface area as possible; we define low‐dose phototherapy as the use of measured irradiance ≤ 30 μW/cm²/nm over the same bandwidth (AAP 2004).

However, for the purposes of this review, we will accept studies that compare different levels of irradiance that are both above or both below our pre‐defined cut‐off of 30 μW/cm²/nm.

For a study that sets out to compare different doses of phototherapy, for instance, by means to achieve different levels of irradiance, without specifying the level of irradiance achieved, we will contact the trial author to request further information with regards to whether measured irradiance was done while placing the study under the "awaiting classification" category. If the study author cannot or does not provide further information, we will exclude the study.

Means to achieve different levels of irradiance

Following are specific comparisons that will be included based on the means to achieve different dosages (levels of irradiance) of phototherapy. For a more detailed description of each comparison, please refer to Appendix 1.

To avoid overlap with another Cochrane Review that evaluates the level of phototherapy received by infants (Thukral 2015), this review focuses on the light source, instead of the recipient, in determining dosage and intensity variation of phototherapy.

  1. Comparison 1: use of one phototherapy device versus multiple identical phototherapy devices.

  2. Comparison 2: use of one type of light source versus a different type of light source.

  3. Comparison 3: use of one measured distance of light source from the infant versus a different distance from the light source.

For all three comparisons, study authors must show a difference in measured irradiance between groups for the studies to be eligible.

All included clinical trials will follow pre‐specified protocols for both comparison groups in all aspects of care, including standard treatment for neonatal hyperbilirubinaemia such as the use of feeding regimens (breastfeeding or fluids), and will follow the same or a similar protocol for both intervention and control groups.

Types of outcome measures

Primary outcomes

  1. Incidence of acute bilirubin encephalopathy (defined as acute clinical manifestations of bilirubin toxicity seen in the first weeks after birth, which is characterised by signs of irritability and hypertonia along with early retrocollis and opisthotonus, together with any one of the following: drowsiness, poor feeding, alternating tone, high‐pitched cry, or a failed auditory brainstem response hearing screen), measured at pre‐defined intervals (e.g. at 6, 12, 18, and 24 months)

  2. Incidence of kernicterus (defined as the long‐term sequelae of bilirubin toxicity, which is characterised by extrapyramidal movement disorders, gaze abnormalities, auditory disturbances, intellectual deficits, and posticteric sequelae of enamel dysplasia of the deciduous teeth) (Johnson 2002; AAP 2004), measured at pre‐defined intervals (e.g. at 6, 12, 18, and 24 months)

  3. Proportion of infants with moderate or severe cerebral palsy, defined as a non‐progressive disorder with abnormal muscle tone in at least one arm or leg associated with abnormal control of movement or posture and a modified Gross Motor Function Classification System (GMFCS) score ≥ 2 (Palisano 2008; Rosenbaum 2007), measured at pre‐defined intervals (e.g. at 6, 12, 18, and 24 months). We included as secondary outcomes results from individual assessments that measure other specific aspects of neurodevelopmental outcome, if available (see Secondary outcomes numbers 8 to 11)

  4. Mortality rate measured at pre‐defined intervals (e.g. at 6, 12, 18, and 24 months)

Secondary outcomes

  1. Bilirubin level (serum bilirubin) reported as follows: (a) absolute bilirubin level measured at specific time points and expressed in mmol/L or mg/dL; (b) change in bilirubin level measured as the difference between bilirubin readings at two time points and expressed in mmol/L or mg/dL; or (c) rate of change in bilirubin expressed as mmol/L/h or mg/dL/h. We grouped absolute bilirubin levels measured at the same time points (e.g. 6, 12, 24, 48, or 72 hours) after initiation of phototherapy and the difference in bilirubin levels between the same time points among included trials under the same outcomes

  2. Duration of phototherapy (in hours)

  3. Proportion of infants who require exchange transfusion

  4. Proportion of infants with clinical or echocardiographic evidence of significant PDA

  5. Proportion of infants with necrotising enterocolitis, classified via modified Bell's clinical staging (Bell 1978; Walsh 1986)

  6. Proportion of infants with bronchopulmonary dysplasia (BPD). We accept both "classical" and "physiologic" definitions of BPD. In the "classical" definition, BPD is defined by a sustained need for any supplemental oxygen at 28 days' postnatal age (Northway 1967), or at 36 weeks' postmenstrual age (Shennan 1988). In the "physiologic" definition, infants who require mechanical ventilation, continuous positive airway pressure (CPAP), or supplemental oxygen exceeding 0.30 fraction are diagnosed with BPD without further testing. Infants who require lower supplemental oxygen who fail to sustain desirable oxygen saturation during a timed stepwise reduction in oxygen concentration down to room air are also diagnosed with BPD (Walsh 2003)

  7. Weight, reported as changes in weight or number of infants with significant weight loss (e.g. > 10% of birth weight), measured at a defined interval such as weekly or twice weekly

  8. Proportion of infants with neurodevelopmental impairment, as indicated by a score ˂ 70 on the Bayley Scales of Infant and Toddler Development, Third Edition (BSID‐III) (Albers 2007), measured at pre‐defined intervals (e.g. at 6, 12, 18, and 24 months)

  9. Proportion of infants with motor impairment, as indicated by a score ˃ 2 on the modified GMFCS evaluation (Palisano 2008), measured at pre‐defined intervals (e.g. at 6, 12, 18, and 24 months)

  10. Proportion of infants with bilateral visual impairment, defined as vision worse than 20/200 (Vohr 2012), measured at pre‐defined intervals (e.g. at 6, 12, 18, and 24 months)

  11. Proportion of infants with hearing impairment, defined as the inability to understand oral directions of the examiner and to communicate, with or without hearing amplification (Vohr 2012), measured at pre‐defined intervals (e.g. at 6, 12, 18, and 24 months)

  12. Caregiver satisfaction, measured on a validated scale, such as the Health‐Related Quality of Life Tool (HRQoL) or health outcome rating scales, or self‐reported parental dissatisfaction using scales such as any adapted version of the Parent Satisfaction Scale (Gerkensmeyer 2005; Stallard 1996)

  13. Staff satisfaction, measured by a validated scale, such as HRQoL or health outcome rating scales, or self‐reported staff dissatisfaction

  14. Length of hospital stay (in days)

  15. Effect on breastfeeding, determined by the number of breastfeeds per day while receiving phototherapy

  16. Electrolyte imbalance, as reported during or after phototherapy

  17. Re‐admission rates after discharge from hospital after phototherapy

Outcomes will be measured at the following points in time.

  1. During hospital stay (acute bilirubin encephalopathy, bilirubin levels as specified above, duration of phototherapy, requirement for exchange transfusion, and other clinically important outcomes such as PDA, necrotising enterocolitis, BPD, electrolyte imbalance, weight, caregiver and staff satisfaction).

  2. At discharge (weight, caregiver and staff satisfaction).

  3. For neurodevelopmental outcomes, at defined points beyond discharge (e.g. at 6, 12, 18, and 24 months)

Search methods for identification of studies

We will use the criteria and standard methods of Cochrane and Cochrane Neonatal (see the Cochrane Neonatal search strategy for specialized register).

Electronic searches

We will follow the search strategy as used by Cochrane Neonatal and will search the following databases.

  1. MEDLINE (PubMed (National Library of Medicine)) (1950 to present).

  2. Embase (1980 to present).

  3. Cochrane Central Register of Controlled Trials (CENTRAL), in the Cochrane Library (current issue).

  4. Cumulative Index to Nursing and Allied Health Literature (CINAHL) (1982 to present).

We have outlined detailed search strategies for each of the above databases in Appendix 2.

We will also search ongoing clinical trials, unpublished studies, and trial protocols via the following sites.

  1. www.clinicaltrials.gov.

  2. www.controlled‐trials.com.

  3. www.clinicalstudyresults.org.

We will apply no language restrictions.

Searching other resources

We will search the references cited in relevant studies including Cochrane Reviews, guidelines, review articles, and conference proceedings, including abstracts from the Annual Meetings of the Pediatric Academic Societies (American Pediatric Society/Society for Pediatric Research (https://www.aps‐spr.org/) and European Society for Paediatric Research (www.espr.info/)) and the Perinatal Society of Australia and New Zealand (http://www.psanz.com.au/). We will search available archives of these conferences through the conference websites or via journal supplements for the last five years, or up to inception of the conferences. We will also contact expert informants if necessary to identify further relevant studies.

Data collection and analysis

We will follow standard Cochrane methods as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019).

Two review authors (KY and EW) will screen the titles and abstracts of articles independently and will exclude articles that clearly are not relevant. Two other review authors (YC and SL) will assess the short‐listed studies to determine final eligibility and to document reasons for exclusion. Any differences in decisions between two review authors at each of the two stages mentioned above will be discussed, leading to consensus, with involvement of an arbiter (NL) if necessary.

Selection of studies

We will accept published and unpublished studies, in both full‐text article and abstract forms. We will contact the authors of unpublished studies and studies available only as abstracts to request further information about the studies, including specific details, such as methods of sequence generation, allocation and blinding, participant withdrawal, pre‐specified outcomes, and full outcome data, to enable them to be included in our meta‐analysis. We will include only the final data from each included study ‐ not data from interim analyses.

Data extraction and management

Two review authors (YC and SL) will code all data from each included study independently using a pro forma designed specifically for this review and generated by the Covidence platform (www.covidence.org/). The intervention defined in the study (definition of high‐ and low‐dose phototherapy) will be compared against our definition of high‐ versus low‐dose phototherapy. In addition, we will record data regarding trial participants, including gestational age, age and serum bilirubin level at study entry, and other inclusion or exclusion criteria. We will screen for duplicate entry of participants by matching initial numbers of participants recruited against total numbers at each step in the study. If we discover a discrepancy, we will look for an explanation in the article (e.g. multiple enrolment of the same infants in different hospital admissions). Any disagreement among review authors will be resolved by discussion leading to consensus.

Assessment of risk of bias in included studies

Two review authors (SL and KY) will independently assess each included trial for risk of bias according to the following six major criteria, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019).

  1. Sequence generation.

  2. Allocation concealment.

  3. Blinding of patients and personnel.

  4. Blinding of outcome assessors.

  5. Incomplete outcome data.

  6. Selective outcome reporting.

  7. Other issues (e.g. extreme baseline imbalance).

A detailed description for each of the 'Risk of bias' criteria is provided in Appendix 3.

We will assign a judgement of 'low', 'high', or 'unclear' risk with justification for each criterion by completing a 'Risk of bias' table for each included trial. We will discuss any disagreement among review authors and will involve a third review author (NL) if necessary.

Measures of treatment effect

We will report outcome estimates for categorical data using risk ratios (RRs), risk differences (RDs), and number needed to treat for an additional beneficial outcome (NNTB) or number needed to treat for an additional harmful outcome (NNTH). For continuous data, we will use mean differences (MDs) when we include two or more trials in an analysis. We will report all effect estimates with their respective 95% confidence intervals (CIs). If pooled analyses are not possible due to reasons such as major discrepancies in study characteristics or in outcome reporting, as detailed under Assessment of heterogeneity, we will report the results of studies individually.

Unit of analysis issues

If we include cluster‐RCTs (e.g. trials in which assignment to intervention and control groups is made at the Neonatal Intensive Care Unit (NICU) level), we will adopt the following approach.

First, we will assess whether trial authors adjusted for the effects of clustering by using appropriate analysis methods, such as generalised estimating equation (GEE) modelling. If no adjustment has been made, we will perform adjustment by calculating the design effect based on a fairly large assumed intracluster correlation (ICC) of 0.10, which has been shown to be a generally realistic estimate from studies on implementation research (Campbell 2001). If the unit of analysis is not stated in the trial, we will inspect the width of the standard error (SE) or 95% CIs of estimated treatment effects. If we find an inappropriately small SE or a narrow 95% CI, we will ask the trial authors to provide information on the unit of analysis. We will follow methods for calculations presented in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019).

To address the issue of repeated measures in included trials that arises from having multiple recordings of serum bilirubin from the same infant, we will follow the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). We will create separate subgroups of outcomes for serum bilirubin, with recordings obtained at different time points (e.g. at 6, 12, 24, 48, and 72 hours) after initiation of phototherapy. We will not total the participants among different subgroups so as to avoid multiple counting.

Dealing with missing data

We will determine the dropout rates from each study and will assess whether intention‐to‐treat analysis was performed. We will do this by comparing the number of infants initially randomised versus the total number analysed. We will consider an absolute dropout rate of 20% or higher as significant. Additionally, we plan to adopt a 'worst‐case approach' in judging the dropout rate: if we find that the direction of the outcome estimate is reversed after a worse‐case scenario, we will consider the dropout rate as significant; if we find a significant dropout rate with no reasonable explanation, we will assign the study as having high risk of bias with regard to the criterion 'incomplete outcome data'; if we consider the missing data to be critical to the final estimates in our meta‐analysis, we will contact the authors of individual studies to request further data.

If multiple studies are included in any comparison, we will perform sensitivity analyses to assess how the overall results are affected by including studies with high risk of attrition bias from incomplete outcome data.

Assessment of heterogeneity

We will use the I² statistic as a guide to determine our action with regards to heterogeneity of each outcome (Higgins 2019). In accordance with the recommendations of Cochrane Neonatal, we will use the following cut‐offs for reporting of heterogeneity: ˂ 25%, negligible heterogeneity; 25% to 49%, low heterogeneity; 50% to 74%, moderate heterogeneity; and 75% or higher, high heterogeneity. If a moderate or high degree of heterogeneity is found, we will evaluate studies in terms of their clinical and methodological characteristics using the criteria listed as follows to determine whether the degree of heterogeneity may be explained by differences in those characteristics, and whether a meta‐analysis is appropriate.

Criteria that we will assess include the following.

  1. Baseline characteristics of participants (postmenstrual age, birth weight, haemolytic versus non‐haemolytic hyperbilirubinaemia).

  2. Clinical settings of studies (e.g. tertiary or secondary neonatal ICU).

  3. Presence of co‐interventions that are not pre‐specified under subgroup analysis and investigation of heterogeneity. Co‐inteventions include feeding (breast or formula) regimen for infants, fluid regimen therapy, positioning of infants, or any co‐interventions employed in addition to the primary intervention.

  4. Risk of bias (as detailed in the assessment of 'Risk of bias' section).

Assessment of reporting biases

We will evaluate possible publication bias by inspecting the shape of the funnel plot if 10 or more studies are included in the analysis of relevant outcomes. If publication bias is suggested by significant asymmetry of the funnel plot, we will include a statement in our results with a corresponding note of caution in our discussion.

Data synthesis

We will perform meta‐analyses using Review Manager 5.3 with a fixed‐effect model (RevMan 2014), according to the recommendations of Cochrane Neonatal. When possible, our primary data analyses will follow the intention‐to‐treat principle, namely, the original number of infants allocated to each study arm will be used as the denominator in subsequent analyses. We will express our results as RRs, RDs, NNTBs or NNTHs, and weighted mean differences (WMDs) with their respective 95% CIs, as detailed under Measures of treatment effect.

Certainty of evidence

We will use the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, as outlined in the GRADE Handbook (Schünemann 2013), to assess the certainty of evidence for the following major outcomes in the 'Summary of findings' table: incidence of acute bilirubin encephalopathy, incidence of kernicterus, proportion of infants with moderate or severe cerebral palsy, as detailed under the ’Primary outcomes’ section, proportion of infants with neurodevelopmental impairment, proportion of infants with hearing impairment, and bilirubin levels at time points measured.

Two review authors (NL and YC) will independently assess the certainty of evidence for each of the outcomes above. We will consider evidence from RCTs as high certainty but will downgrade the evidence by one level for serious (or by two levels for very serious) limitations based upon the following: design (risk of bias), consistency across studies, directness of evidence, precision of estimates, and presence of publication bias. We will use the GRADEpro GDT Guideline Development Tool to create a ‘Summary of findings’ table to report the certainty of evidence.

The GRADE approach results in an assessment of the certainty of a body of evidence as one of four grades.

  1. High: we are very confident that the true effect lies close to that of the estimate of the effect.

  2. Moderate: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

  3. Low: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.

  4. Very low: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

Subgroup analysis and investigation of heterogeneity

We plan to undertake the following subgroup analyses if relevant data are available. We will analyse three primary comparison groups: one versus multiple of the same light sources, one type of light source versus a different type of light source, and one measured distance of light source from the infant versus a different distance from the light source. For each of the primary comparisons, we plan to assess the following via subgroup analysis if sufficient data are available.

  1. Gestational age, including ˂ 32 weeks' gestational age, 32 weeks to ˂ 37 weeks, and ≥ 37 weeks (term infants).

  2. Birth weight ≤ 1500 g, ˃ 1500 g to ˂ 2500 g, and ≥ 2500 g.

  3. Causes of hyperbilirubinaemia: pathological versus physiological (as defined by the principal investigator).

  4. Severity of hyperbilirubinaemia: studies that define severe hyperbilirubinaemia as ˃ 340 µmol/L at any time during the first 28 days of life (Canadian Paediatric Society 2007).

Sensitivity analysis

If a sufficient number of studies are available, we will perform sensitivity analyses for primary outcomes and for any secondary outcomes to assess the impact of excluding studies with high risk of the following.

  1. Selection bias (for either criterion or for both criteria of random sequence generation and allocation concealment).

  2. Attrition bias (incomplete outcome data).