- All analysis will be performed using R software. Continuous variables will be described using medians with interquartile ranges and will be compared by the Wilcoxon rank sum test. Categorical variables will be described using frequencies and percentages, and differences will be assessed using Pearson’s chi-squared test or Fisher’s exact test. Categorization of continuous covariates, if non-linearity between the covariate and the log odds of the outcome is suspected, will be based on graphical assessment. Cutoff values will be based on either clinical parameters or sample percentiles.
- To ascertain the influence of intraoperative vasopressin administration on liver transplantation recipients postoperative AKI incidence, a propensity score-adjusted multivariate logistic regression will
be conducted to adjust for potential confounding factors. Multiplicative interaction terms will be tested in the regression model between hypotension time and vasopressors administered (intraoperative catecholamine dose and vasopressin use). This model will include perioperative variables we consider potential confounders that might influence both vasopressors use and renal outcomes, outside of the causal pathway, as described in the directed acyclic graph (Figure 1).
- The association between the vasopressin use and the eGFR at one week and at one year after liver transplant will be addressed with a linear multivariable regression model
including the same set of covariates used in the primary model.
- A sensitivity analysis will be performed running the same propensity score- adjusted model on the subpopulation with an intraoperative dose in the top quartile of this cohort.
All tests will be two-sided, alpha - value < 0.05 will be considered statistically significant. No statistical power calculation will be conducted before the study. The sample size will be solely based on the available data.
The propensity score adjustment method is a statistical technique for estimating treatment effects in non-randomized studies. In this technique, the propensity score of the treatment – vasopressin use – is included in the final regression model, in addition to treatment status and other covariates. Only covariates measured before exposure will be included for propensity score calculation, to avoid adjusting for mediators. Covariates included in the propensity score calculation model will not be used in the primary model to avoid collinearity. A logistic regression model will be employed to estimate the propensity score (i.e., the probability of receiving treatment):
- Indication for liver transplantation
- Model for End-stage Liver Disease Sodium (MELD-Na) score
- Preoperative intensive care unit admission
- Propensity score for vasopressin use
- Catecholamine vasopressors dose
- Total packed red blood cells transfused
- Warm Ischemic Time (per 10 min)
- Vena cava implantation surgical technique
- Intraoperative central venous pressure