The study enrolled patients who underwent liver transplantation (LT) in two liver transplant centers, first in Shulan [Hangzhou] hospital (abbreviated as SL cohort) between July 2016 and October 2017 and second in First Affiliated Hospital, School of Medicine, Zhejiang University (abbreviated as ZY cohort), between May 2020 and April 2021. The enrollment was limited to subjects who received deceased donor liver transplantation (DDLT) and ranged between adult donors and recipients (aged >18 years). Accordingly, the following types of patients were excluded: (I) adolescent donor/recipients (aged <18 years); (II) multiorgan transplantation recipients (n ≥ 2); (III) living donor LT.

Consent was obtained from the enrolled participants. The study was performed under the Declaration of Helsinki and approved by the ethical institutional review board of Zhejiang Shuren University.

Indicators with the potential to affect post-transplant prognosis were collected for further analysis. Factors were generally categorized into the following: donor, recipient, graft, surgery, and interaction (Table 1). Routinely, recipient weights were measured before LT. Graft weights were also scaled at the end of the cold ischemia phase before implantation. Details of indicators for LT cases and prognostic information were obtained from medical record systems in each hospital.

Macrosteatosis (MaS) was assessed under microscopic observation in a double-blinded manner (6). Graft failure (GF) and patient death (PD) were the primary endpoints representing the prognosis. The model for end-stage liver disease (MELD) score was calculated by using the formula provided in a previous study (7). Child–Pugh scores were calculated based on the clinical factors of total bilirubin, albumin, prothrombin time, ascites, and encephalopathy (8). Diagnosis of early allograft dysfunction (EAD) was defined according to criteria updated in a prior study (9). Primary non-function (PNF) was defined as impaired liver function needing re-transplantation in an emergency ward within 72 h after LT (10).

Categorized by the GWRWR, quantitative data were presented as median [interquartile range (IQR)] and compared by using the Mann–Whitney U test. A comparison of the distribution of qualitative data in different groups was performed by using the χ² test. The hazard ratios (HRs) of sectionalized GWRWR and other potential indicators of GF/PD occurrence were assessed by using univariate and multicovariate COX regression models. Specifically, the HRs for prognosis were assessed between the selected and the remaining groups. Furthermore, graft/patient survival (GS/PS) rates were compared by using the log-rank test across groups by GWRWR classifications. Relative excess risk caused by interaction (RERI) and attributable proportion (AP) was applied to investigate the synergistic effects on LT prognosis for graft size with other relevant factors. To be specific, RERI, AP, and their confidence intervals (CIs) were calculated based on regression coefficients (B) and correlated covariance from different comparisons via scale provided in previous literature (11). RERI > 0 and AP > 0 meant positive interaction (11). The network was constructed by using Cytoscape (v3.9.0) to visualize the associations between the GWRWR and LT prognosis within risk profiles speculated from survival analysis.

A total of 362 LT cases included for 354 recipients were enrolled for analysis. Our study enrolled 243 cases from the SL cohort and 119 cases from the ZY cohort, accounting for 85.9% and 70.8% of all adult LT cases during the same period in each center. Eight patients underwent re-transplantation, and six remained alive at the end of the follow-up. The median of the GWRWR for the whole cohort was 2.15% with IQR (1.79%–2.49%) (Figure 1A). Most re-transplant patients used smaller (GWRWR < 1.5) or larger (GWRWR > 2.5) grafts during the first LT (Figure 1B).

LT cases were divided into three subgroups by the GWRWR at 2.5% and 3.0% (Table 1). As expected, body mass was higher in donors but lower in recipients, followed by GWRWR increment stepwise. The GWRWR was positively affected by graft weight. A sharp MaS increment was observed with GWRWR elevation (8.3% in the group with a lower GWRWR and 58.3% in the group with a higher GWRWR). Meanwhile, increased post-transplant liver enzymes [including alanine/aspartate aminotransferase (ALT/AST)], higher blood loss, and a higher volume of transfusions indicated more severe transplant complications in those who received LFS grafts (GWRWR > 3.0). Because of insignificant intergroup differences in recipient disease severity (similar MELD scores and primary disease), it was more challenging to treat LT patients using large-for-size grafts (LFSG, GWRWR ≥ 3.0%), and challenges were in the form of higher blood loss, higher transfusion volume, and longer surgical duration (P < 0.05, Table 1). Besides, it was comparable for other clinical indicators in groups categorized by GWRWR. More details about clinical indicators are given in Table 1.

Distribution of survival status for all enrolled subjects in the follow-up duration is presented in Figure 1B. No death occurred in patients whose GWRWR ranged between (2.8%–3.0%) and (1.0%–1.3%). COX regressions revealed that the LFSGs caused an inferior prognosis compared with the remaining groups.

Most patients (75%) used grafts with the GWRWR ranging between 1.0% and 2.5%. An insignificant HR was observed in the group with a lower intraheterogeneity (Figure 1C). A GWRWR between 2.5% and 3.0% seemed to be the most suitable with a lower risk of GF/PD. Moreover, these results remained consistent after adjusting for correlated donor/recipient BMI and graft MaS (Figure 1C and Table 2). Furthermore, survival curves also confirmed clear separations in groups with different GWRWRs (Figure 2). Considering the separated survival curves in groups categorized by the GWRWR, we defined the three groups [(1%–2.5%), (2.5%–3.0%), (3.0%–4.1%) on GWRWR] to have “normal,” “optimal,” and “inferior post-transplant prognoses, respectively. LT by SFSS organ (defined by GWRWR < 1%) seemed unsuitable, as highlighted by the absence of survivors during the follow-up period. However, the result was uncertain for a few cases (only two) reported in this cohort (Figure 1B).

Despite its close connection with the GWRWR, MaS also increased the risk of PD/GF in patients after LT. A further interactive analysis found that MaS exerted additive effects on LFSGs to cause poor prognosis with positive RERI and AP. But in the group with optimal prognosis, the GWRWR caused a lower GF/PD risk by initiating a negative interaction with MaS (Table 2). Meanwhile, surgical indicators [blood loss, pRBC/fresh frozen plasma (FFP) transfusion, and operational duration] associated with LFSGs also showed positive associations with inferior prognosis (Table 2). In addition, factors such as a higher Child–Pugh score, EAD status, and ABO mismatch were associated with poor prognosis (Table 2). However, these factors were less correlated with GWRWR variation (Table 1).

The profile of the risk factors for LT recipients is summarized in Figure 1E. The GWRWR was commonly determined by factors such as recipient/donor BMI, graft weight, and MaS status. LFSGs increased GF/PD risk by causing more surgical complications (prolonged surgical duration, higher blood loss/transfusion volume) and synergistic effects with concomitant graft MaS. The GWRWR was less related to the potential prognostic factors of the recipients (Child–Pugh score, EAD status) and to donor/recipient interaction (ABO mismatch).