Study design and settings
This historical cohort study was conducted at Paul-Brousse hospital (Villejuif, France), a high volume LT center. This study was approved by the Ethics Committee of the French Society of Anesthesia and Resuscitation (IRB# 00010254-2020-070) and is reported according to the STROBE guidelines.15 Written consent was waived by the Ethics Committee
We identified all patients who underwent a total or partial LT for ESLD between January 01, 2015 and December 31, 2020. We excluded patients who received a combined transplantation (liver-kidney, liver-heart or liver-lung transplants), a retransplantation or a transplantation for acute liver failure, neuropathic amyloidosis or primary liver cancer without known ESLD. We also excluded patients with known PoPH who were treated prior to transplantation and those for whom no baseline mPAP measurement was found on the intraoperative anesthesia health records. All perioperative data were searched using our electronic medical records and anesthesia sheet records.
Before surgery, all patients had a preoperative echocardiography and stress echography as well as a pulmonary examination including EFR
Patient care was standardized over the study period. Patients arrived in the operating room and were moved onto a heated mattress. The following noninvasive monitors were routinely used: 5-lead electrocardiogram, noninvasive blood pressure, rectal temperature probe, frontal electroencephalogram, Foley catheter, and neuromuscular blockage depth monitoring. Vascular accesses consisted of two large-diameter peripheral venous catheters, a radial arterial catheter, and a right jugular venous catheter placed under ultrasound guidance. The femoral vessels and the left internal jugular vein were left available for a possible veno-venous bypass. A pulmonary artery catheter (swan-ganz catheter) was also placed using ultrasound, and hemodynamic management was guided with a calibrated continuous cardiac index monitoring, calibrated venous central oxygen saturation, central venous pressure, and arterial pressure. Rapid infusers and infusion warmers were ready for use prior induction. General anesthesia was induced with sufentanil, propofol or etomidate, and neuromuscular blockade was achieved using succinylcholine (if potassium was in the normal range) and maintained intraoperatively with atracurium. Rapid sequence intubation was performed in all patients. Maintenance of anesthesia was achieved using a sufentanil infusion and inhaled sevoflurane. Fluid administration consisted of a balanced crystalloid infusion (Ringer's lactate) and correction of blood loss with either Ringer's lactate, 4% albumin or packed red blood cells if hemoglobin level was < 8 g/dl
In most cases, a total hepatectomy with vena cava preservation was performed associated with a temporary portacaval shunt. The standard technique used for vena cava reconstruction was the so-called “3-vein piggy-back” technique7. In rare cases of caval replacement, a veno-venous bypass was used in case of poor hemodynamic tolerance of caval clamping.
End-to-end porto-portal anastomosis was preferred in the vast majority of cases, whereas extra-anatomic anastomosis was reserved for more complex situations, as described previously. Hepatic artery anastomosis was usually designed using the junction of the common hepatic artery and gastroduodenal artery, or both recipient’s and graft’s sides. The splenic artery, the recipient's aorta or replaced arteries were considered only as second-intent alternative. Care was taken to avoid grafting an excessively long artery. End-to-end duct-to-duct anastomosis was the preferred technique for biliary reconstruction. A hepaticojejunal stoma with a Roux-en-Y anastomosis was reserved for patients with biliary pathologies or in cases of significant disparities in caliber between the native and grafted bile ducts¨.
Our exposure of interest was the presence of elevated mPAP (equal to or greater than 20 mmHg) at the beginning of surgery in patients with a pulmonary artery catheter inserted and PAP measurements recorded on anesthesia sheet. This threshold is the recognized threshold defining pulmonary hypertension in recent guidelines. 4
Our primary outcome was a composite outcome of postoperative pulmonary complications which included pneumonia, acute respiratory distress syndrome, acute pulmonary edema, and pleural effusion. This composite outcome is slighy different than other recognized postoperative pulmonary complications composite outcomes. 16 We excluded atelectasis because we hypothetized it might be less relevant for our exposure of interest and added pleural effusion as being more relevant for pulmonary congestive mechanisms.
Our secondary outcomes were intraoperative bleeding, need for postoperative renal replacement therapy (in the subgoup of patients who did not require renal replacement therapy in the preoperative period), graft dysfunction, infectious complications (urinary tract infection, sepsis, septic shock, superficial infection, and peritonitis) and 90-day mortality
Several preoperative variables were collected to describe the cohort and to adjust for potential confounders. The following variables were considered potential confounders as being potentially associated with a higher mPAP and a higher risk of postoperative complications: age, sex, MELD score, preexisting arterial hypertension (AH), chronic obstructive pulmonary disease (COPD), chronic renal failure (CRF), chronic atrial fibrillation (AF), left ventricular ejection fraction (LVEF) <50%, and baseline cardiac index.
Data source and measurement:
Some data was available in a database prospectively collected and maintained by our surgical team. We extracted data on missing variables, including our exposure and outcomes, directly from medical charts. MELD was calculated at the inscription on the waiting list and adjusted just before the liver transplant. Outcomes were classified based on reported complications in patients’ charts by treating physicians.
We described patients’ characteristics for the full cohort and for each exposure group. We presented categorical variables as frequencies with proportions and continuous variables as means with standard deviations (or medians with first and third quartiles for skewed distributions). We also graphically explored the potential association between baseline mPAP and both the MELD score and the initial cardiac index. We reported the intensive care unit (ICU) and hospital lengths of stay in a descriptive manner.
We estimated the effect of elevated mPAP on our primary outcome by fitting bivariable and multivariable logistic regression models. We fitted the multivariable model using all potential confounders.
For our secondary outcomes, we fitted similar logistic regression models except for blood loss, for which we estimated a log-transformed linear regression model since the distribution of blood loss was right skewed. We explored the statistical interaction between our exposure and the baseline cardiac index in all models. Homoscedasticity and linearity assumptions were explored by an analysis of the residuals for the linear model. The linearity assumption was also explored by fitting a quadratic term for every continuous variable in all models. When the linearity assumption was not met, we fitted polynomial models with quadratic terms to improve the fit of the models. We assessed for the presence of multicollinearity using the Variance Inflation Factor statistic for all models. We reported odds ratios for the logistic regression models and mean multiplicative factors for the log-transformed linear models as estimates. All estimates were reported with 95% confidence intervals. All statistical analyses were performed using the R software (R collaboration, version 4.0.3).
We conducted sensitivity analyses on the effects of the categorization threshold used for our exposure of interest. Since pulmonary hypertension was up to recently defined as a mPAP above 25 mmHg, we fitted all our multivariable models with such exposure dichotomization. We also conducted a second sensitivity analysis by fitting our multivariable models using a categorization threshold of 35 mmHg for our exposure, which defines moderate pulmonary hypertension and a fourth one with the mPAP as a continuous variable.4 For the latter, we used restricted cubic splines with 4 knots to explore potential non-linear associations and tested such non-linear associations by conducting either a general linear test or a likelihood ratio test.