Study population and inclusion criteria
727 patients were diagnosed as liver cirrhosis (LC) in our institution during years 2006 to 2017. Our institution is a public hospital located in suburban area of Tokyo metropolitan area, with 376 beds and an outpatient ward. Etiology of LC was investigated by trained hepatologists by laboratory findings such as low Platelet and Albumin, and imaging test including ultrasound, computed tomography, or magnetic resonance image. 258 participants that had had history of excessive alcohol consumption and were diagnosed as alcoholic LC by the hepatologists were left. 12 patients that had only visited our outpatient ward once, and 2 that further medical record review revealed no presence of alcoholic LC were excluded from the analysis. Finally, a total of 244 patients were left and included for the present study. Inclusion and exclusion flowchart of participants is described in figure 1.
The research protocol was approved by Institutional Review Board of St Marianna University Hospital (approval number 4995).
Basic demographics at point of diagnosis were obtained through medical record review. Age, sex, body mass index (BMI) calculated by height and weight, chief complaint at first hospital visit, and emergency hospitalization were recorded. Coexistence of viral hepatitis (HBV and HCV), and laboratory findings including Asparate aminotransferase (AST), Alanine aminotransferase (ALT), Creatinine, total Bilirubin, Albumin, Prothrombin time- international normalized ratio (PT-INR), and Platelet count, were obtained from medical record review.
Liver cirrhosis severity was measured by MELD score  and albumin-bilirubin (ALBI) score . MELD score is an objective scoring calculated by creatinine, total Bilirubin, Albumin and PT-INR . Whether the discrimination power of MELD score in predicting mortality is superior to that of Child Pugh Score, a classical scoring for liver cirrhosis, is controversial; however, a systematic review in 2016 reports that MELD score is as reliable a prognostic score as the Child Pugh score . A study conducted in 2019 reported ALBI to be the optimal predictor of mortality compared to MELD, MELD-Na and Child Pugh . Furthermore, a study among Asians have reported that ALBI was provided particularly reliable prediction for short-term outcomes, whereas the MELD score was better in terms of assessing long-term outcomes . Considering that Child Pugh scoring is rather subjective in terms of ascites and encephalopathy scoring, we employed the MELD and ALBI scores in the current study.
Regarding socioeconomic status, whether the participant was a PA beneficiary was obtained from insurance information in medical record. Residing alone and having no key person (a family member or someone in close relationship to contact in case of emergency) were other SES measurements obtained. Information on SES variables mentioned above were obtained from medical record of the participants.
The primary outcome for our study was survival time after diagnosis of alcoholic liver cirrhosis. The patients were followed until death, dropout or continuous outpatient visit at March 2020 or later, whichever came first. Data for those that death was not recorded (either dropped out or were continuing hospital visit) was handled as censored data.
As secondary outcome for our study, the incidence of ascites, esophageal varices, hepatocellular carcinoma (HCC), encephalopathy and spontaneous bacterial peritonitis (SBP) were obtained. Presence of abovementioned complications observed at our hospital at any point during follow up period was counted as positive.
Basic demographics were depicted using median and interquartile range, or by percentage, depending on whether the variable was linear or bivariate. In order to test differences between the two subgroups, Chi squared analyses were applied for binary variables. Mann Whitney U tests were applied for continuous variables since all continuous variables had non-normal distribution. We computed survival probabilities using the Kapan Meier method and used Cox proportional hazards regression to estimate hazard ratios. Using the Schoenfeld residuals, we determined that hazard ratios were constant over the follow-up time. Mean survival time, survival rate at median follow-up, and five-year survival rate was obtained from survival curve for the two subgroups. Statistical difference of crude hazard ratio for being a PA recipient was obtained by Log rank test. Then, we included receiving PA, together with age, HCV infection, and ALBI score in one cox hazard model to obtain the adjusted hazard ratio. Significance level was set at the alpha value of 0.05. All statistical analysis was computed using STATA version 16 software (Stata Corporation, College Station, TX, USA).