OBJECTIVES:
The main objective of this study is to describe and measure the effects of intraoperative hemodynamic management on postoperative outcomes in liver transplantation in Canada and France.
The specific objectives are:
1. To describe the overall incidence of postoperative complications and graft outcomes in adult liver transplant recipients in Canada and France and across different recipients' characteristics.
2. To measure the effects of fluid balance, within an overall hemodynamic instability and management strategy, on the postoperative outcomes of liver transplant recipients in Canada and France.
The central hypotheses of this study are that an intraoperative hemodynamic management based on a restrictive fluid administration may improve postoperative outcomes in this population.
EXPOSURE VARIABLES:
The exposures of interest will be the intraoperative fluid balance and the intraoperative doses and types of vasopressors used. Fluid balance will be defined as the sum of the volume of administered crystalloids, colloids and blood product transfusions minus drained ascites, intraoperative diuresis and bleeding. Doses of administered vasopressors will be converted into a time-weighted total norepinephrine equivalent dose calculated as the total equipotent units of norepinephrine administered as an infusion divided by the number of hours the infusion was given during the period between entrance in the operating room and end of surgery. Types of vasopressors used will also be collected. Since a restrictive fluid management strategy is associated with a higher use of vasopressors and vice versa, both variables may covary together. However, more difficult surgeries will be associated with higher blood loss, higher volume of administered fluid and higher doses of vasopressors. The combination of these two variables will help estimate the intraoperative hemodynamic management strategy used and delineate the effects of each component.
OTHER DESCRIPTIVE VARIABLES:
Other patients' characteristics and perioperative practices variables will be captured for descriptive purposes. Recipients' demographic characteristics (age, sex), anthropometric variables (body mass index (BMI)), the presence of any previous abdominal surgery, presence of hepatic encephalopathy, need for preoperative organ support or any preoperative hospital admission will be collected as well. Data on the perioperative use of invasive cardiac output monitoring (thermodilution catheter, transoesophageal echocardiography, etc.), as well as coagulation management (use of thromboelastometry, tranexamic acid, transfusion thresholds, etc.) will be collected. Other donor and graft variables, such as donor sex, donor's BMI, the type of vascular and biliary anastomoses, the use of extended-criteria donor graft, the use of ex vivo perfusion, ischemia time, the practice of donation after circulatory death (donor characteristics, heparin usage, etc.) and the perioperative use of liver biopsies will also be collected.
DESCRIPTIVE AND STATISTICAL ANALYSIS:
Missing data The investigators will train all sites regarding quality of data collection and will emphasize the importance of complete data collection. In case a complete case analysis would exclude more than 5% of the observations, the investigators will use multiple imputations by chained equations using 5 to 20 imputated datasets assuming MAR (missingness at random) on the covariables and the outcomes.
Descriptive analyses and complications incidences across centers The investigators will first describe the cumulative incidences of different complications or categories of complications as previously defined. Their main descriptive analysis will be to report these incidences overall as well as across denominated centers. Then, they will estimate multivariable determination models for each complication category by fitting generalized mixed effect models using a logit link adjusted for age, sex, disease severity (MELD), transplantation indication, requirement for preoperative organ support and donor characteristics with an estimated random-effect for clustering within centres. The investigators will then compute intra-class correlation (ICC) coefficients to better define the variability of complications across centres (heterogeneity). However, they will fit a Cox regression with a shared frailty factor for biliary complications. Finally, they will fit a generalized mixed effect model using a log link and quasi-Poisson distribution for the total number of complications up to 30 days per patient with an estimated random effect for clustering within centres. Their secondary exploratory analyses will be to report the cumulative incidence of each complication according to age categories (\< 50, 50-60, \>60), sex, nature and severity of liver disease (chronic liver disease (CLF), acute liver failure (ALF), retransplantation and MELD categories among CLD patients (\< 20, 20-30, \>30). They will report such incidences per subgroups with 95% confidence intervals. They will also report survival up to 6 months over and across centres by fitting a Cox regression with a frailty factor.
Association between hemodynamic management and postoperative complications The investigators' primary analysis will be the association between fluid volume and primary graft dysfunction by 7 days after transplantation using a multivariable generalized mixed-effect model using a logistic link and random intercepts, adjusted for the time-weighted dose of vasopressors and the aforementioned confounders, and estimated using residualized penalized quasi-likelihood. They will consider any patient needing a retransplantation within 7 days after the index transplantation as having a primary graft dysfunction. Statistical interaction will be explored between fluid balance and the vasopressors dose variable.
The investigators' other secondary analyses will be the association between fluid volume and other outcomes, including AKI and the total number of complications up to 30 days, using similar adjusted survival models (with frailty factors) or generalized mixed effect models (with random intercepts). Among them, the investigators will analyze the association between fluid volume and time to biliary non-anastomotic strictures using a multivariable Fine and Gray model, adjusted for the same confounders, with retransplantation and death (not caused by the biliary non-anastomotic strictures) considered as competing risks. They will address intra-cluster correlation using a frailty factor. Risk proportionality over time will be explored using the Harrel and Lee test and a visual inspection of the Schoenfeld residuals.
Subgroup analysis The investigators will conduct a subgroup analysis comparing the effect of their exposures on the primary outcome in the subgroup of patients transplanted for end-stage liver disease and those transplanted for causes other than end-stage liver disease. The investigators will thus estimate the same model but with an interaction term between the fluid volume and a dummy variable for end-stage liver disease. They will report stratified results with 95% CI and the interaction test p value.
Sensitivity analyses For the primary outcome, the main secondary outcome and the total number of complications at 30 days, the investigators will first conduct a sensitivity analysis using a generalized propensity score for fluid volume (based on probability density) that will include all confounders and the vasopressor exposure and estimate a marginal effect of fluid balance using an inverse probability of treatment weighting with stabilized weights. The investigators will evaluate the confounding effect of hypotension by removing it from the models as well as by a sensitivity analysis restricted in the subgroup of patients without significant residual hypotension. Finally, they will conduct all analyses by adding data from the French centre to explore the variation in the estimates created by adding a non-Canadian centre.
Economic analysis The investigators will conduct an economic analysis to evaluate hospital costs associated with the observed incidences of postoperative complications. They will also analyze costs associated with their fluid balance exposure effect. Costs will be those associated with the actual procedures, as determined by every hospital using its account systems, and include both fixed and variable components previously identified as major cost drivers (mechanical ventilation days, LOS, reoperation and ICU readmission).
