Liver transplantation continues to be associated with high risk for postoperative complications (1–3). Organ shortages have prompted greater utilization of higher risk grafts, in parallel normothermic and hypothermic oxygenation perfusion machines are today more common from cardiac-donor procurements (4, 5).

Scores as donor risk index (6, 7) have been published and models combining MELD score (Model for End-Stage Liver Disease) have been reported to predict complications, with lower mortality in recipients with high MELD scores (8–10).

We hypothesized that beyond the short- and long-term influence of the risk factors earlier studies have identified, certain other modifiable factors may be relevant although they have not been extensively analyzed in prospective series set in the context of today's advances in surgical and donor procurement processes. We aimed to explore all modifiable risk factors preoperative and intraoperative associated with major complications after LT in a series of consecutive recipients registered prospectively in three centers.

Data from all consecutive adults aged 18–80 years who were scheduled for LT in three centers were assessed from August 2, 2019, to November 2, 2021. The only criterion for excluding a patient was cancellation of the procedure. The extended protocol was approved by the institutional review board (IRB) of the lead hospital (University Hospital of Bellvitge, approval number AC 033/18). Human Ethics and Consent to Participate declarations are validated by the IRB. It was registered in the European Registry in 2018 (European Clinical Trials Database -EudraCT 2018-002510-13) and in ClinicalTrials.gov (NCT04405518, first registration date, 10-10.2019). Consent to participate: Every human participant has provided their consent. Methods were performed in accordance with the relevant guidelines and regulations. The work has been reported in line with the STROCSS guidelines (11).

A total of 318 patients on the LT waiting list were initially evaluated; seven patients were removed from the list, and six procedures were cancelled in the operating room. Finally, data for 305 patients were analyzed. Table 1 shows patient and surgical data for the cohort, stratified according to CDC status. The median age was 60 years (IQR, 55–64 years), 77.38% were male, and the most common diagnosis was alcohol cirrhosis 49.51%, followed by hepatocarcinoma (22.30%). The median MELD score was 17 (IQR, 11–21), 9.84% had partial portal thrombosis, and the median hemoglobin value was 108 g/L (IQR, 88–126). A prior history of abdominal surgery was present in 32.39% of patients.

Donor age ranged from 18 to 84 years, and 30.16% of the patient's received grafts from controlled cardiac-death donors. Vena cava preservation was performed in nearly all patients, and a temporary PCS was used in 41.91% of the patients. Intraoperative RBC transfusion was required by 51.48%, and 27.37% required additional transfusion in the first 24 h after LT. During LT, 54.43% required infusion of fibrinogen concentrate. At reperfusion stage, 27.87% received tranexamic acid because of bleeding. Reperfusion syndrome was present in 38.49%.

The primary composite outcome of major postoperative complications indicated by CDC 3–5 status occurred in 97 patients (31.80%, 95% CI, 26.58%–37.03%). Table 2 shows surgical events and outcome of patients. Miscellaneous complications (pneumothorax, cardiac arrythmia, neurotoxicity) were registered in 12 patients. One patient died during the surgical procedure, five patients died within 30 days, one patient died on day 60, and another patient died on day 75. Two patients were retransplanted because of primary graft failure; both were discharged alive.

The ICU length of stay was less than 1 week in 87.90%, between one and 2 weeks in 7.74%, and more than 2 weeks in 4.63%. Twenty patients (6.57%) were readmitted to the ICU. These patients were excluded from the binomial analysis, which was performed on data for 285 patients. The PT/INR ratio showed a variance inflation factor of 6.26 with the MELD score and was therefore excluded from the log-binomial regression analysis. Data used for the binomial regression model is presented in Table 3. The RR and aRR values derived from the binomial regression model for CDC 3–5 status are shown in Figure 1.

Sex, age, and preoperative co-morbidities were not associated with CDC 3–5 status. A MELD score was associated with risk for CDC 3–5, and a score cut point of >23 showed an even stronger association (aRR 1.92; 95% CI, 1.28–2.8). In contrast, a high plasma concentration of hemoglobin and fibrinogen levels (RR 0.77; 95% CI, 0.64–0.91) were not associated with risk for CDC 3–5, but after adjustments for MELD score, only a baseline hemoglobin concentration (aRR, 0.98; 95% CI, 0.97–0.99) continued to be associated with this outcome.

Zero RBC transfusion was not associated with risk for CDC 3–5 status (aRR 0.37; 95% CI, 0.28–0.72). The RBC transfusion cut point of >2.5 units during LT was found to be strongly associated with CDC 3–5 grade (aRR, 1.96; 95% CI, 1.29–2.96) (Table 3). We found no associations between CDC 3–5 status and use of a temporary PCS, CIT, WIT, surgical time, or intraoperative fluid administration. On the other hand, PRS (aRR, 2.11; 95% CI, 1.43–3.10) and fibrinogen administration (aRR, 1.07; 95% CI, 1.05–1.09) were associated with CDC 3–5 status.

The incidences of CDC 3–5 status and AKI were similar to findings reported for other series (17, 18).

A higher MELD score was associated with CDC 3–5 status, whereas baseline creatinine, glomerular filtration rate, and plasma sodium level were not associated with risk for the outcome. Higher baseline hemoglobin concentration protected against both CDC 3–5 status.

After MELD adjustment, our findings confirmed the relevance of preoperative anemia, consistent with two retrospective series in which anemia was linked to major complications and 90-day and 1-year mortality (19–21). Although mortality was associated with hemoglobin concentration in a series of cirrhotic patients, MELD score was a stronger predictor of death or need for LT in that study (22). In our cohort, the median for the highest quartile of patients with CDC 3–5 status was 111 g/L; in contrast the median value in CDC 0–2 patients was 115 g/L. This value for the patients with fewer complications suggests a target for a testable preoperative optimization strategy, even though optimization is clinically challenging in patients with liver disease (23).

CDC 3–5 patients had altered hemostatic and coagulation and thromboelastometry profiles. High values of PT/INR were associated with CDC 3–5 grading. However, because PT/INR was collinear with the MELD score, it was excluded from analysis, even though both can be considered to confer patient risk for CDC 3–5 status. Baseline plasma concentration of fibrinogen was also associated with CDC 3–5 status, but none of the variables derived from ExTem or FibTem were confirmed.

Duration of surgery, surgical technique, ischemia times, and donor characteristics were not associated with CDC 3–5 status. Because a temporary PCS was performed in nearly 42% of the cohort overall (without differences between the two CDC strata), it is uncertain whether this technique could have influenced the outcome. A meta-analysis of studies of LT using grafts from cadaveric donors reported that using a temporary PCS reduced blood component usage but had no effect on postoperative outcomes (24). However, a study reported that this procedure led to better intraoperative hemodynamic parameters and a decrease in the incidence of reperfusion syndrome (25). The duration of cold ischemia and the absence of a PCS were independent predictors of PRS in that cohort; nevertheless, the percentage of PRS in patients who underwent PCS was higher than expected, at 65% (26). A very recent randomized trial in living donor LT recipients claimed clear hemodynamic advantages for patients with a PCS (27); however, in the control group, inferior vena cava injury was present in a higher-than-expected percentage (26.67% vs. 3.3% in the intervention group), and consequently, intraoperative blood loss, hepatectomy time, total duration of surgery, and blood component requirements were significantly different in the PCS and control groups, calling into question the generalizability of the conclusions to other populations.

In patients with CDC 3–5 status, more patients in this stratum of the cohort also required administration of tranexamic acid at reperfusion of the graft. The occurrence of PRS, which developed in 54.26% of patients in the CDC 3–5 stratum. PRS is caused by liver graft and recipient risk factors, which can coexist. Avoiding hypervolemia and maintaining an optimal plasma calcium level before reperfusion of the graft are keys to mitigating these effects. In addition, the degree of steatosis of the graft is a determining factor in the incidence of PRS and subsequent complications (28). One group's retrospective study found that preconditioning grafts with hypothermic oxygen perfusion machines was associated with a lower incidence of PRS (29).

The relatively low number of LTs analyzed represents a partial limitation of this study. Nevertheless, the size of our series of patients waiting for liver grafts reflects numbers that are typical for most European registries (30, 31). We did not calculate the Comprehensive Complication Index (CCI) which allows a continuous stratification of the outcome and compared with the CDC; the CCI allows a grading system for surgical complications, its predictive value is superior for hospital stay and surgical strategies such as the use of temporary PCS (32). However, the objective of our study was to assess complications and their relationship with baseline patient's characteristic and intraoperative incidents. For this reason, we evaluated the simplicity of data collection of the CDC concomitantly with the course of patients in the critical care unit, seeking factors of association.

Certain variables (portal thrombosis, previous sodium values, PT/INR, and ascites) can be related to the MELD score; however, only PT/INR needed to be excluded because of collinearity. We found no other collinearities and are therefore confident that baseline preoperative hemoglobin and perioperative hemodynamic instability and blood component requirements are the most important factors affecting outcome in this prospective cohort. A third limitation is related to the timing of events. The 20 patients in whom complications occurred after discharge from the ICU and who were readmitted to the ICU were not included in the logistic regression analysis, thus simplifying the model and avoiding the need for right-censoring or time-to-event analysis. However, we note that we found similar results when we did a post-hoc analysis of the full group of 97 patients with CDC 3–5 status.

Strengths of the study are the participation of three high-volume LT hospitals, prospective data collection, high adherence to protocols, and the monitoring of data quality by an independent committee.

We hypothesize that despite the challenges of correcting preoperative anemia in liver disease, intravenous iron infusion in wait-listed patients may be feasible. Whether this strategy could possibly reduce blood transfusion remains to be tested in a randomized controlled trial.

We conclude that PRS at reperfusion of the liver graft and the volume of RBCs transfused are the main modifiable factors that influence major complications reflected by CDC 3–5 status after LT.