Ursodeoxycholic acid for preventing parenteral nutrition‐associated cholestasis in neonates: A systematic review and meta‐analysis

While ursodeoxycholic acid is used in treating parenteral nutrition‐associated cholestasis (PNAC) in neonates, its role in prevention is unclear.


| INTRODUCTION
The incidence of preterm births is on the rise across the globe.Nearly 10% of all births are preterm, and about 1% are extremely low birth weight (<1000 g) neonates [1].These babies require special care and tend to stay in the neonatal intensive care unit (NICU) for many weeks with life support and nutritional therapy.Along with other organs, their digestive system is immature, hampering immediate oral and enteral nutrition delivery.Preterm neonates often require total parenteral nutrition (TPN) to complement enteral nutrition and, at times, as the only source of nutrition.Similarly, various digestive system disorders, like necrotising enterocolitis (NEC) in preterm neonates and surgical disorders (intestinal atresia, tracheoesophageal fistula, gastroschisis, etc.), necessitate longer durations of parenteral nutrition.
Parenteral nutrition has significantly impacted the nutrition of preterm babies and those with surgical disorders of the gastrointestinal tract [2].In preterm neonates, the use of early parenteral nutrition in the first seven postnatal days has been shown to reduce neonatal mortality [3], time to regain birth weight, and weight at discharge or 36 weeks postmenstrual age [4].In preterm neonates with NEC, parenteral nutrition provides optimal nutrition and results in better neonatal outcomes [5].Similarly, in critically ill term neonates, early initiation of parenteral nutrition has been shown to reduce mortality [6].However, parenteral nutrition was associated with complications like metabolic derangements, parenteral nutrition-associated liver disease (PNALD), and central line-associated bloodstream infections [7].Over a period, several advances were made in the components of TPN and its mode of delivery to prevent complications.These include using SMOF lipids instead of intralipid [8], early initiation of enteral feeding, and good supportive care to reduce central line-associated infections [9].Despite all the development, PNALD remains an important complication of prolonged parenteral nutrition (defined as >14 days), sometimes permanently damaging the liver [10,11].Parenteral nutrition-associated cholestasis (PNAC), defined as conjugated or direct bilirubin (CB or DB) levels >2 mg/dL and elevated liver enzymes, is considered a surrogate marker of PNALD, whose confirmation requires a biopsy [7].It can result in liver damage due to alteration in the gut-liver axis, ending in significant liver inflammation, cholestasis, cirrhosis, liver failure, and the need for liver transplantation [12].Recent data highlighted high mortality and morbidity on follow-up in neonates with PNAC and emphasised the need for better preventive and treatment strategies in such infants [13].
Various studies have shown that erythromycin and amino acid-free parenteral nutrition with enteral whey protein supplementation effectively prevented PNAC [10].Other interventions like cholecystokinin and phenobarbital were not found to be effective [14,15].Ursodeoxycholic acid (UDCA) improves hepatic ductular bile flow, reduces bile acid cytotoxicity, and improves renal excretion of bilirubin [16].In addition, it stabilises plasma membranes and reduces apoptosis.Few randomised trials and meta-analyses showed that UDCA is effective in treating established PNAC [17].However, studies on the role of UDCA in preventing PNAC yielded mixed results [18][19][20][21].In this systematic review and metaanalysis, we attempted to study the role of UDCA in preventing PNAC in preterm neonates on prolonged parenteral nutrition.

| METHODS
The meta-analysis was conducted following the Cochrane Handbook recommendations [22], and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were used for reporting [23].The protocol was prospectively registered in the PROSPERO database with an identification number CRD42022343104 and can be accessed online (https://www.crd.york.ac.uk/PROSPERO/).Our research question was whether ursodeoxycholic acid (UDCA) (I) compared with placebo (C) prevents/delays the development of parenteral nutrition-associated cholestasis or PNALD (O) (defined as conjugated bilirubin or direct bilirubin >2 mg/dL [34.1 μmol/L]) in neonates on parenteral nutrition.We also planned to assess the effect of UDCA prophylaxis on the critical outcomes of mortality, neurodevelopmental impairment in early childhood, liver failure, and the need for liver transplantation and the important outcomes of neonatal sepsis and NEC.We searched the databases PubMed, Embase, CINAHL, Cochrane CENTRAL, and Scopus on September 15, 2022.The search was updated on September 16, 2023.The search strategy is shown in Table S1.We searched for citations and reviewed the eligible articles' references to identify further studies.To identify ongoing trials, we searched ClinicalTrials.gov and the ISRCTN registry.For grey literature and unpublished data, we searched on Google and dissertation websites.We also contacted researchers working on these topics.

| Inclusion and exclusion criteria
We included all interventional studies comparing ursodeoxycholic acid with placebo or no treatment in preventing cholestasis in neonates requiring prolonged parenteral nutrition.We decided to include both randomised controlled trials (RCTs) and nonrandomised studies of intervention (NRSIs) because of the small number of RCTs available.As the randomised trials are selective in recruiting participants and assessing predefined outcomes, the inclusion of well-designed NRSIs can provide real-life data and, hence, a better understanding of the benefits and harms [24,25].We considered the study eligible if the design was an RCT or NRSI (prospective or retrospective).We excluded review articles, case reports, and unpublished data.Studies should have described at least one prespecified outcome for inclusion in this review.The critical outcomes of interest were cirrhosis/liver failure, the need for liver transplantation, and death.The other important outcomes were the incidence of cholestasis, direct bilirubin levels, and time taken to develop cholestasis.

| Study selection
Two reviewers (S.K. P. and A. S. A.) independently performed the title and abstract screening and full-text screening.Any disagreements were resolved by mutual discussion or involvement of the third reviewer (R. P. A.).

| Data extraction
After retrieving the full texts, data extraction was done by two reviewers (R. P. A. and A. S. A.) in a blinded manner, with discrepancies being resolved by involving the third reviewer (T.P. O.).The following data were extracted: author, year of study conduct and publication, place of study, study design, inclusion and exclusion criteria, details of the intervention, co-interventions (strategy of parenteral nutrition, feeding, type of milk used, and fortification), sample size, mean gestational age, mean birth weight, gender distribution, the proportion of preterm neonates as per the gestation weeks (<35 and 35-37 weeks), duration of hospital stay, parenteral nutrition days, dose and duration of UDCA, placebo used, direct/conjugated bilirubin values, the proportion of neonates not developing cholestasis (DB/CB < 2.0 mg/dL) outcome, and follow-up details.

| Assessment of risk of bias
We assessed the risk of bias (ROB) of included studies using the Cochrane Risk of Bias tool (RoB-2) for RCTs [26] and the Risk Of Bias In Nonrandomised Studies of Interventions (ROBINS-I) tool for NRSIs [27].We assessed the risk of bias in the five domains of random sequence generation, allocation concealment, selective reporting, blinding, and completeness of outcome data for RCTs.For NRSIs, we assessed the risk of bias due to confounding, in the selection of participants for the study, in the classification of interventions, due to departures from intended interventions, due to missing data, in the measurement of outcomes, and in the selection of reported results.

| Data synthesis
We decided to conduct a meta-analysis if there were at least two studies of similar study design comparing ursodeoxycholic acid with placebo and measuring the outcome(s) in a similar way.Heterogeneity was explored through consideration of the study populations (e.g., differences in gestational age), interventions (e.g., dose of UDCA), outcome definitions (e.g., definition of cholestasis), and in statistical terms, by the I 2 statistic.The I 2 statistic, with a level of >50%, was considered to indicate moderate levels of heterogeneity and I 2 > 80% as significant heterogeneity.The random and fixed effects models were used for metaanalysis.The random effects model meta-analysis results were presented when the heterogeneity was substantial.The meta-analysis was performed using the Cochrane statistical package, RevMan 5.4 software.A meta-analysis was performed separately for RCTs and observational studies due to the distinct risk of bias among two different study designs and to avoid significant heterogeneity.The effects of the intervention were expressed as risk ratio (RR), odds ratio (OR), and number needed to treat (NNT) and risk difference (RD) for dichotomous data and as mean difference (MD) and standardised mean difference (SMD) for continuous data.
If permitted, we planned to compare overall estimates with estimates from studies at a low risk of bias.We explored the reasons for significant heterogeneity when I 2 was >50%.We planned subgroup analyses for different birth weight groups (<1000 g vs. >1000 g), gestational age groups (<28 weeks vs. >28 weeks), dose and duration of UDCA, co-interventions (e.g., type of lipid used), and settings (high vs. low-and middleincome countries).The certainty of evidence was downgraded when the heterogeneity was unexplained.We planned to assess the publication bias by using funnel plots if the number of studies were more than 10.
The Cochrane Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach was used to assess the level of evidence (LOE) [28].

| Study selection
The database search yielded 6180 articles.After removing duplicates (n = 1684), we performed title and abstract screening of 4455 articles.Full-text screening of 22 articles was performed, and five studies were found eligible [18][19][20][21]29].We searched the references of the shortlisted full-text articles and found 33 articles.None of these were eligible for inclusion in the systematic review.The details of excluded studies are shown in Table S2.The PRISMA flow diagram is shown in Figure 1.Of the included studies, three were RCTs [18,19,29], and two were NRSIs, of which one was a prospective [21] and one a retrospective observational study [20].The characteristics of the participants in the included studies are shown in Table S3, and the details of the interventions and nutritional strategies are shown in Table S4.

| Quality assessment
On quality assessment, all the RCTs were judged to have some concerns in the risk of bias assessment.In one study, we noted problems with the randomisation process, as alternating allocation was used [29].In the other two studies, as the trial protocols were not published, we judged them to be at high risk for selecting the reported result [18,19].Among the NRSIs, both had concerns about blinding outcome assessors.One study was judged to be at moderate risk of bias [21] and the other to be at serious risk of bias, as the bias due to confounding was not addressed using appropriate statistical methods, that is, matching or regression [20].The risk of bias assessment is shown in Tables 1 and 2.
Meta-analysis of data from NRSIs did not show any decrease in the duration of parenteral nutrition, peak conjugated bilirubin levels, and time to full enteral feeds.The results for cholestasis, mortality, and liver failure were not estimable.While there was no heterogeneity for the outcome of time to full enteral feeds, significant heterogeneity was noted for the outcomes of parenteral nutrition duration and peak conjugated bilirubin levels.The details of the meta-analysis are shown in Table 3.The forest plots are shown in Figure 2.
We could not perform subgroup and sensitivity analyses because of a limited number of studies.Publication bias was not assessed because of the limited data.

| Certainty of evidence
On assessing the certainty of the evidence using the GRADE methodology, we noted a low certainty evidence for the outcomes of feed intolerance, peak conjugated bilirubin levels, and time to full enteral feeds.For the outcomes of mortality, NEC, cholestasis, neonatal sepsis, hospital stay duration, and parenteral nutrition duration, the certainty of the evidence was very low.Observational studies provided very low certainty evidence for the outcomes studied.The details of the certainty assessment are provided in Table S5.

| Summary of key findings
This systematic review and meta-analysis evaluated the benefits and harms of ursodeoxycholic acid used as prophylaxis for preventing parenteral nutritionassociated cholestasis in neonates requiring parenteral nutrition for >14 days.Five studies (three randomised Randomisation was inappropriate; an alternating method where odd numbers were allotted to intervention and even numbers to placebo.However, there were no differences in the baseline characteristics. b The protocol was not published; hence, it is difficult to predict if there was a selection of the reported result. T A B L E 2 Risk of bias: Nonrandomised studies of intervention (ROBINS-I).

Author, year
Bias The confounding factor of surgical patients was not appropriately controlled for (the control group had more surgical patients who have an inherently higher risk of cholestasis).
and two nonrandomised) with 286 infants met the inclusion criteria.Prophylactic UDCA did not decrease parenteral nutrition-associated cholestasis, mortality, neonatal sepsis, NEC, hospital stay duration, and parenteral nutrition duration.There was a decrease in feed intolerance, peak conjugated bilirubin levels, and time to reach full enteral feeds.No harm was noted.However, the confidence intervals of estimates were wide for all the outcomes studied.

| Overall completeness and applicability of evidence
None of the studies have reported the data on critical outcomes of liver cirrhosis, the need for liver transplantation, and neurodevelopmental outcomes.A single RCT reported the other critical outcome of mortality.In addition, the dose of UDCA was variable and ranged from 5 to 30 mg/kg/day in the included studies.The available evidence is insufficient to understand the benefits and harms of prophylactic ursodeoxycholic acid in neonates requiring prolonged parenteral nutrition.

| Quality of evidence
The overall certainty of the evidence was low.The main factors limiting the quality of evidence were the risk of bias and imprecision in estimates.The risk of bias in RCTs was predominantly due to problems with the randomisation process and selective outcomes reporting.
In observational studies, the risk of bias was due to problems controlling the confounding factors and a lack of blinding of outcome assessment.

| Comparison with published literature
None of the systematic reviews published to date evaluated the role of ursodeoxycholic acid in the prevention of parenteral nutrition-associated cholestasis in neonates [30][31][32].On the other hand, the role of UDCA in the treatment of PNAC is well-studied.A recent systematic review and meta-analysis has reported that UDCA was safe and effective as a treatment for cholestasis of varied aetiologies in children [17].There was a reduction in symptoms, liver enzyme levels, and serum bilirubin levels.On the subgroup analysis of infants and children with PNAC, UDCA was found to be beneficial.
In another systematic review published by Lauriti et al., various interventions for treating and preventing parenteral nutrition-associated cholestasis were studied [10].Erythromycin and amino acid-free parenteral nutrition with enteral whey protein were found to be effective in preventing PNAC.Although ursodeoxycholic acid was also evaluated, only one study evaluating UDCA for preventing cholestasis in neonates on parenteral nutrition was included [21].In comparison, we included five studies in this systematic review and meta-analysis.

| Strengths and weaknesses
We performed an extensive literature search and will unlikely miss any published literature.To reduce bias, two reviewers independently performed the steps of data screening, data extraction, and risk of bias assessment and the discrepancies were resolved in group meetings involving the third reviewer.
There are several limitations in this review.First, the number of studies included, and the sample size was small.As a result, optimal information size is unavailable for the reported outcomes.If an RCT was to be planned to assess the role of UDCA in preventing PNALD in neonates, assuming the prevalence of PNAC to be about 20% in neonates [33,34] and UDCA reduces the prevalence by 25%, the sample size required is about 943 in each group (1886 in total) to achieve a power of 80% and a two-sided Pvalue of 0.05 [35].Second, the studies were heterogeneous with respect to the characteristics of the neonates (gestational age and birth weight varied across studies), the dose and duration of ursodeoxycholic acid (ranged from 5 to 30 mg/kg/day), the parenteral nutrition, and feeding strategies used.Although there was no statistical heterogeneity (I 2 statistic) on meta-analysis of RCTs, these baseline differences are likely to impact the results.Third, the quality of the included studies was poor, both for RCTs and NRSIs.

| Implications for practice
There is low-certainty evidence that prophylactic ursodeoxycholic acid reduces feed intolerance, peak conjugated bilirubin levels, and time to reach full enteral feeds.There is very low-certainty evidence that prophylactic ursodeoxycholic acid does not decrease cholestasis, mortality, necrotising enterocolitis, neonatal sepsis, hospital stay duration, and parenteral nutrition duration.As we do not have enough evidence on critical outcomes of mortality, liver failure, liver cirrhosis, and the need for liver transplantation, it is impossible to recommend the use of prophylactic ursodeoxycholic acid for preventing parenteral nutrition-associated neonatal cholestasis.

| Implications for research
Future well-designed randomised controlled trials could be undertaken in neonates on prolonged parenteral nutrition to assess prophylactic ursodeoxycholic acid's effect on critical mortality outcomes and liver failure/ need for transplantation.

F I G U R E 2
Forest plots.Forest plots summarising the results of the meta-analysis.(a) Neonatal sepsis, (b) hospital stay duration, (c) parenteral nutrition duration, (d) peak conjugated bilirubin levels, and (e) time to full enteral feeds.CI, confidence interval; IV, inverse variance method; M-H, Mantel-Haenszel method; NRSIs, nonrandomised studies of intervention; RCTs, randomised controlled trials; UDCA, ursodeoxycholic acid.
Risk of bias: Randomised controlled trials.
T A B L E 1 Meta-analysis.
T A B L E 3