Age at surgery and native liver survival in biliary atresia: a systematic review and meta-analysis

Biliary atresia (BA) is a childhood rare disease of the liver and bile ducts that requires prompt surgical intervention. Age at surgery is an important prognostic factor; however, controversy exists with regard to the benefit of early Kasai procedure (KP). We aimed to conduct a systematic review and meta-analysis to examine the relationship between the age at KP and native liver survival (NLS) of BA patients. We performed the electronic database search using Pubmed, EMBASE, Cochrane, and Ichushi Web and included all relevant studies published from 1968 up to May 3, 2022. Studies that examined the timing of KP at ages 30, 45, 60, 75, 90, 120, and/or 150 days were included. The outcome measures of interest were NLS rates at 5, 10, 15, 20, and 30 years post-KP and the hazard ratio or risk ratio for NLS. The quality assessment was used using the ROBINS-I tool. Among 1653 potentially eligible studies, nine articles met the inclusion criteria for the meta-analysis. Meta-analysis for hazard ratios revealed that there was a significantly faster time to liver transplantation in the group of patients who had KP at later timing as compared with earlier KP (HR = 2.12, 95% CI 1.51–2.97). The risk ratio comparing KP ≤ 30 days and KP ≥ 31 days on native liver survival was 1.22 (95% CI 1.13–1.31). The sensitivity analysis showed that comparing KP ≤ 30 days and KP 31–60 days, the risk ratio was 1.13, 95% CI 1.04–1.22. Conclusion: Our meta-analysis showed the importance of early diagnosis and surgical interventions ideally before 30 days of life in infants with BA on native liver survival on 5, 10, and 20 years. Therefore, effective newborn screening of BA targeting KP ≤ 30 days is needed to ensure prompt diagnosis of affected infants. What is Known: • Age at surgery is an important prognostic factor. What is New: • Our study performed an updated systematic review and meta-analysis to examine the relationship between age at Kasai procedure and native liver survival in patients with BA. What is Known: • Age at surgery is an important prognostic factor. What is New: • Our study performed an updated systematic review and meta-analysis to examine the relationship between age at Kasai procedure and native liver survival in patients with BA.


Introduction
Biliary atresia (BA) is a childhood rare disease of liver and bile ducts presenting with biliary obstruction exclusively in the neonatal period [1]. BA is the most common cause of neonatal jaundice requiring prompt surgical intervention (the Kasai procedure [KP]) that aims to restore bile flow and is the most frequent indication for liver transplantation in children. Age at surgery is an important prognostic factor of BA [1,2]. The success of KP, such as the clearance of jaundice and native liver survival (NLS), is associated with infant age at the time of surgery [3]. Over the past decade, the average age at KP has remained between 60 to 70 days, and the 10-year NLS is 52.8% in Japan [4,5]. Due to medical advances enabling early diagnosis, the age at BA diagnosis and surgery have decreased in recent years. This has resulted in shifts in surgery being performed at earlier ages at less than 60 days with studies showing better prognosis of KP performed at less than 30 days [6,7].
The previous systematic review by Jimenez-Rivera et al. conducted in 2013 [8] investigated the relationship between the timing of KP and patient outcomes. The inclusion criteria for the review were that studies should be population-based and also include outcomes of interest, including overall survival, NLS, and the timing of KP. A total of 3128 reports were identified through searches; 149 articles were reviewed in full, and 36 articles qualified for inclusion. According to the results, there were 14 studies that examined the relationship between the timing of KP and patient outcomes. All studies, except for one [9], demonstrated improved NLS when surgery was performed at an earlier age. In addition to the fact that the review was done 10 years ago, a metaanalysis of study results was not conducted.
Our study aimed to conduct an updated systematic review and contribute a meta-analytic evaluation of the evidence of the relationship between the age at KP and NLS among BA patients.

Methods
This systematic review followed the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [10]. Our protocol was registered with the International Prospective Register of Systematic Reviews (CRD42022310131).

Search strategy
We performed the electronic database search on Pubmed, EMBASE, Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials, and Ichushi Web (Japanese). We included all relevant studies published from 1968 up to May 3, 2022. The lower range of the publication year was set as 1968 since KP became wellrecognized globally after a seminal publication was released during this year [11]. MeSH terms including "biliary atresia," "portoenterostomy," "hepatic," "Kasai procedure," "surgery," and other essential keywords were included in the search strategy. A medical information services/literature search expert developed the search strategy by combining conditions for disease name, intervention name, study design, and search period. The full search strategy used is shown in Supplementary Table S1.

Inclusion and exclusion criteria
Studies that examined BA patients with regard to an intervention or exposure definition related to the timing of KP at ages 30, 45, 60, 75, 90, 120, and/or 150 days were included. The number of events at the above time of KP ± 5 days was combined into the closest cut-off points. The main outcome measures of interest were NLS rates at 5, 10, 15, 20, and 30 years post-KP and the hazard ratio (HR) or risk ratio (RR) for NLS in relation to the age of KP. The included studies had to be either a randomized controlled trial or an observational study. We included studies published in English and Japanese. Exclusion criteria included articles that were not published in a peerreviewed journal. Animal studies and case reports were excluded. Unpublished documents and grey literature including conference papers, dissertations, and patents were also excluded.

Study selection and data extraction
The relevant papers from the literature were managed in Rayyan [12] to facilitate the screening and selection process. Screening based on literature abstracts was performed blinded by two independent reviewers (E.H., Y.M.), and inconsistencies between the reviewers that occurred during the process were resolved through discussion or consulting the third reviewer (K.S.). The included studies were assessed for their quality of reporting evidence, and their data was extracted and summarized. Two reviewers (E.H., Y.M.) independently extracted data from the included studies, and disagreements were resolved by discussion or consulting the third reviewer (N.S.).

Quality assessment
The quality of the included studies was assessed based on the Cochrane tool known as the ROBINS-I tool [13]. Both E.H. and K.S. independently assessed the included papers against these checklists, and any discrepancies were resolved by discussion. The response options for an overall assessment were (1) low risk of bias: the study is comparable to a well-performed randomized trial; (2) moderate risk of bias: the study provides sound evidence for a non-randomized study but cannot be considered comparable to a wellperformed randomized trial; (3) serious risk of bias: the study has some important problems; (4) critical risk of bias: the study is too problematic to provide any useful evidence and should not be included in any synthesis; and (5) no information on which to base a judgment about risk of bias.

Data synthesis
If multiple articles were identified from the same country, the national data (e.g., nationwide registry) was given priority for data synthesis to avoid double counting of data. We included a general summary table that quantifies the main characteristics of the included studies, including study design, timing of KP, study year, country, number of patients, gender, mean age of patients, and outcome measures. Where appropriate, a meta-analysis was conducted using the DerSimonian and Laird random-effect model for HR analyses and the restricted maximum likelihood random-effect model for RR analyses using STATA (17, StataCorp LLC, College Station, TX, USA) [14]. Sensitivity analyses to examine the relationship between KP ≤ 30 with KP ≥ 31 days, ≥ 61 days, and ≥ 91 days were performed. To assess the degree of heterogeneity between studies, we visually inspected the forest plots and estimated the I 2 statistics; ≥ 75% of I 2 statistics was deemed considerable [15]. When the heterogeneity was considerable, we attempted to identify the source of heterogeneity.

Results
We identified 1653 potentially eligible studies through initial searches, screened 1261 titles and abstracts after removing duplicates, and reviewed the full texts of 33 reports. A PRISMA flow diagram of the study selection process is shown in Fig. 1. A list of excluded studies is shown in Supplementary Table S2. Finally, we included nine articles. Six articles reported the number of events, two articles reported HRs, and one article reported both the number of event and HR. The characteristics of the included studies are summarized in Table 1. There were no randomized controlled trials identified. Of the included studies, three were national registry studies, and six were retrospective cohort studies. The number of patients in each study varied; it ranged from a single institution study in Korea including 72 patients to a large national registry study in Japan including 3591 patients. The timing of KP was investigated as early as less than 30 days to as late as over 150 days after birth. Pakarinen et al. [16] had 65 days as their cut-off, and their data were included in the ≥ 61 ≤ 60 comparison. The NLSs were observed from

Risk of bias assessment
The quality assessment results are presented in Table 3. We rated the risk of bias of each study as either moderate or serious due to the insufficient information on the adjustment for potential factors assessed in the risk of bias domain. Five of the nine studies performed multivariate analyses; therefore, they were classified as having moderate bias, while other

Discussion
We performed a systematic review and meta-analysis of the available evidence examining the relationship between the age at KP and NLS among BA patients. A total of nine studies were included, all of which were retrospective studies and included three national registry studies. The quality assessment of included studies was either moderate or serious due to the insufficient information on the adjustment for potential confounders. Study heterogeneity was not evident in our meta-analyses, and there showed a consistent pattern of results supporting the conclusion that earlier KP leads to better NLS.
In the Japanese clinical guideline, there is a clinical question stating "Should the Kasai procedure be performed within the first 30 days after birth?" [17]. The result is given as "It is recommended that the Kasai procedure should be performed within the first 30 days after birth, considering the survival rate of the native liver in biliary atresia. Strength of recommendation: Weakly recommended (agreement rate = 94%)." However, this was based on meta-analysis findings that did not include the results from the Japan Biliary Atresia Registry. Moreover, comparisons between KP ≤ 30 days and ≥ 31 days were not performed. In their meta-analysis, only three studies were included [6,18,19], and a search on Ichushi-shi (Japanese database) was not performed. To our knowledge, there are no other clinical guidelines with a graded recommendation scale such as the Japanese guidelines to evaluate this clinical question. The practice guideline of the American Association for the Study of Liver Diseases states that diagnosis of BA and performance of KP by 8 to 10 weeks of age is optimal for transplant-free survival [20]. The New Zealand guideline suggests that KP is most successful when performed early, preferably before 6 weeks of age [21]. The guideline from France shows that if KP is performed before the age of 30 days, 40% of the operated children will be alive with their native liver at the age of 20 years [22]. The results of this study support the Japanese guideline recommendation by showing that performing KP earlier, even before 30 days of age, contributes to increased NLS. In particular, this study is the first meta-analysis to incorporate comparisons between ≤ 30 and 31-60 days.
In our meta-analyses, only observational studies were identified. Our risk of bias assessment showed that the overall quality of evidence was at the level of moderate to serious risk of bias. This was influenced by the fact that among the nine studies included in our analyses, only five studies [16,[23][24][25][26] performed multivariate analyses which additionally controlled for post-operative variables. The remaining four other studies did not control for any potential confounding variables in their analyses [6,9,27,28]. Of those five studies, four estimated the HRs. The most frequently controlled variable was the clearance of jaundice [16,23], although the study by Lee et al. [26] controlled for bilirubin normalization at 2 months after surgery. The anatomic pattern and biliary atresia splenic malformation syndrome were controlled in one study [16]. The use of prophylactic antibiotics [23], center caseload [16], and postoperative cholangitis [25] were controlled in one study each. Controlling for clinical variables that may be mediating the effect between the timing of KP and NLS may introduce bias; however, the baseline patient characteristics such as anatomic pattern and biliary atresia splenic malformation syndrome influencing the clinical outcome [29] should be controlled. Other variables that could potentially affect NLS, such as surgical techniques, were not included in any of the multivariate analyses. It is possible that different sets of variables adjusted in multivariate analyses and other unadjusted variables may have influenced the results on NLS; however, given the difficulty of conducting randomized controlled trials, the results suggested by this study must be interpreted with the limitations of existing observational studies. Our results suggest that NLS was improved at 5, 10, and 20 years with early KP; however, the effect on NLS at 30 years and some of the results of the sensitivity analyses such as < 30 and 30-60 groups comparisons at 10 and 20 years were not statistically significant in our analyses. This might be due to the limited sample size, but also may be influenced by other long-term prognostic factors. In addition to issues of NLS among BA patients, qualityof-life is an important consideration which requires additional studies, particularly among long-term postoperative patients 20 to 30 years since KP; there are also risks for complications such as esophageal varices, cirrhosis, and hepatocellular carcinoma [30]. Further research is needed to determine the effect of the early timing of KP on the potential longer-term benefits.
Even though KP performed within 30 days is desirable, it may be difficult to diagnose the condition within this time window. In Japan, the nationwide stool color card screening was introduced in 2012; yet, according to the 2019 Japan Biliary Atresia Registry results, the average age at KP was 59.2 days [28]. Efforts are needed to ensure the prompt recognition, referral, and diagnosis of BA patients, and future research is needed to determine the optimal screening strategies with the most effective impact on populations. The BA diagnostic algorithms must ensure that screen-positive infants are efficiently evaluated and receive KP in a timely manner.

Conclusion
This report represents the first meta-analytic evaluation of the evidence addressing the effect of the age at KP on NLS among BA patients. Our meta-analysis showed the importance of early diagnosis and surgical interventions ideally before 30 days of life in infants with BA for improving NLS at 5, 10, and 20 years. Our results highlight the need for effective newborn screening of BA to allow for prompt identification and diagnosis and the possibility for KP to occur before 30 days.
Authors' contributions Eri Hoshino designed the study, performed literature screening and analysis, drafted the manuscript, and revised the manuscript. Yamato Muto, Kotomi Sakai, and Nobuyuki Shimohata performed literature screening and data extraction and revised the manuscript. Kevin Urayama and Mitsuyoshi Suzuki performed critical revision of the manuscript. All authors have approved the final manuscript.
Funding This work was supported by JSPS KAKENHI Grant Number 20K18896.

Data availability
The data that support the findings of this study are available from the corresponding author, [EH], upon reasonable request.