Our study’s main finding is that when the maintenance of anesthesia was performed by sevoflurane in living liver donors from the induction to the harvesting of the graft, significantly fewer children developed IPGF.

So far, only one study has evaluated the preconditioning effect of sevoflurane on EAD in liver transplantation (Minou et al. 2012). In a randomized controlled trial, sevoflurane preconditioning was administered to brain-dead donors and showed a significant decrease in EAD compared to the control group, an effect we also found in our study. However, the authors noticed that this positive effect was only apparent in the case of steatosis. In our cohort, sevoflurane appears superior in grafts originating from living donors, a population particularly selected for its good health and absence of liver steatosis.

Two studies compared the effect on early graft function of sevoflurane and propofol administered as postconditioning (Beck-Schimmer et al. 2015; Gajate Martin et al. 2016). In a retrospective study comparing sevoflurane to propofol anesthesia in recipients, Gajate Martín et al. found no difference regarding EAD (Gajate Martin et al. 2016). However, it is interesting that they used a more straightforward definition of EAD, which included only transaminase levels. In contrast, the composite definition typically used by the other studies investigating this topic includes INR and bilirubin levels (Olthoff et al. 2010; Beck-Schimmer et al. 2015; Minou et al. 2012; Nguyen et al. 2019). This lack of difference in transaminase level between both groups is in line with the results of a prospective randomized controlled trial (RCT) carried out a year earlier by Beck-Schimmer et al. (2015). However, they mentioned that the incidence of EAD, using the composite definition, was lower in the sevoflurane group, but without reaching statistical significance (Beck-Schimmer et al. 2015). Besides the small sample size, the authors highlighted two other possible explanations for their negative findings. Their sevoflurane regimen may have had no effect because it had only been applied as a postconditioning strategy. Indeed, due to logistic issues, the transplanted cadaveric grafts were not exposed to volatile anesthetics before cross-clamping. They also suggested that organ injuries in transplantation are caused by various factors, including donor characteristics that may constitute confounding factors. In our study, donors were a homogeneous group of young, healthy people instead of a brain-dead donor population. This may have led to better standardization, enabling the demonstration of a potential beneficial effect of sevoflurane on IRI. Of note, all the grafts benefited from pharmacologic postconditioning since all children received sevoflurane throughout the surgical procedure.

The effects of sevoflurane preconditioning have also been investigated in non-transplant liver surgery, with conflicting results. In an RCT, Beck-Schimmer et al. demonstrated that sevoflurane anesthesia significantly limited the postoperative increase of transaminase levels after hepatectomy with inflow occlusion (Beck-Schimmer et al. 2008). This benefit was more pronounced in patients with steatosis. On the contrary, in another RCT investigating pharmacological conditioning during hepatectomy with inflow occlusion, Song et al. did not show any difference in transaminase levels and clinical outcomes between the propofol and sevoflurane groups (Song et al. 2010). A possible explanation for these different findings may lie in the different durations of ischemia. Indeed, the average duration of Pringle’s maneuver was around 20 min in Song’s study versus 35 min in Beck-Schimmer’s trial. However, it must be emphasized that the ischemia-reperfusion insult observed during a short Pringle’s maneuver in liver resection surgery is less important than that observed in transplantation surgery. Hence, results from liver resection surgery must be extrapolated with caution. The longer duration of ischemic stress observed in transplantation surgery could have a greater potential to reveal the beneficial effects of an IRI prevention strategy.

Interestingly, in our study, the association of propofol and sevoflurane in the donor did not demonstrate any advantage in preserving the graft’s function compared to sevoflurane alone, suggesting, among other hypotheses, that the effect of sevoflurane might be dose-dependent. Several preclinical studies have investigated this question and the optimal timing of sevoflurane administration to protect graft function. In an animal model, Zhou et al. demonstrated a protective effect of sevoflurane preconditioning against hepatic IRI, but no significant dose-response relationship was found (Zhou et al. 2013). The benefit appeared to be the same for 1, 1.5, and 2 MAC. The authors concluded that a dose-response relationship might exist at lower concentrations, evoking a threshold effect that had previously been demonstrated by Obal et al. in a rat heart model, highlighting that preconditioning with sevoflurane at 1.0 MAC offered better protection than 0.75 MAC but that there was no additional benefit to increase the dose beyond 1.0 MAC (Obal et al. 2001). In addition to the dose, the duration of exposure to sevoflurane could also play an important role, and there may be an additive effect of pre- and postconditioning (De Hert et al. 2004; Zitta et al. 2010). Nevertheless, no clear consensus can currently be reached on the optimal timing of sevoflurane administration, and the dose-dependency of the pharmacological preconditioning effect still needs to be determined.

Another interesting finding of our study is that preoperative NLR in the recipient is also an independent predictor of IPGF. For several years, NLR has been known as an indicator of systemic inflammatory status, reflecting the balance between innate and adaptive immune function. The association between preoperative NLR and EAD was already known and described in adult liver transplantation after both cadaveric and living donation (Kwon et al. 2019; Nylec et al. 2020). It is explained by the crucial role of the inflammatory response and the preoperative immune status of the recipient in the development of EAD (Oweira et al. 2016). This is the first time these results have been confirmed in a pediatric cohort.

Another factor associated with IPGF in our study is donor age, a result already reported in several studies. In a retrospective study of 300 deceased donors, Olthoff et al. found an adjusted OR of 3.12 for the development of EAD in donors aged > 45 years (Olthoff et al. 2010). These results are consistent with earlier findings identifying donor age > 49 as an independent risk factor for EAD and primary non-function (Ploeg et al. 1993). It must be noted that deceased donors are often older than patients selected for a living donation. Interestingly, our study shows that donor age is significantly associated with IPGF, even in a young population with a median age of 34. These findings align with the results of another retrospective study that found an association between donor age and EAD in the adult-to-adult LDLT setting (Pomposelli et al. 2016).

Finally, we identified the duration of transplant surgery as a predictive factor for IPGF. At the same time, prolonged warm and cold ischemia times are known to induce more severe IRI (Ito et al. 2021). Surgery time as such is rarely mentioned in the literature. Still, in a retrospective study published in 2007, the authors found an association between operating room time and primary graft nonfunction, often leading to retransplantation (Uemura et al. 2007). It can be assumed that the harmful effects of prolonged surgery time are at least partially linked to prolonged ischemia time but also to more complex surgical conditions, as found in patients with PHT or severe coagulopathy.

This study suffers from several limitations inherent to its retrospective nature. First, we had a relatively high percentage (17.8%) of patient pairs with missing data. However, we analyzed the missing values in the dataset according to recently published recommendations. Since missingness was not associated with the outcome or the predictors, a complete case analysis has been shown to provide unbiased estimates (Bartlett et al. 2015; Lee et al. 2021). A second limitation—common to most studies investigating this topic—is the choice of surrogate biological markers instead of a clinical outcome. Indeed, if the release of liver transaminase is a clinical marker for acute graft injury, its clinical significance remains unclear. Third, model overfitting is a common potential problem in risk factor analyses, especially when the sample size is relatively limited. Hence, we acknowledge that our findings should be validated on an external database in order to support their robustness and generalizability. Further clinical studies, especially prospective randomized controlled trials, are needed to confirm the positive effect of sevoflurane preconditioning in liver transplantation. Fourth, the collected data covers a decade during which practices have constantly evolved, both from a surgical and an anesthetic point of view. Finally, in our cohort, there were differences between the sevoflurane, propofol, and propofol-sevoflurane groups both in donor and receiver intraoperative variables. However, none of these variables were associated with the occurrence of IPGF in multivariate logistic regression.

Nevertheless, the strength of this study lies in the fact that it is the first to investigate pharmacological preconditioning with sevoflurane in LDLT and also the first study investigating the impact of sevoflurane preconditioning on graft function in a pediatric population.