Overall cohort characteristics
A total of 711 patients underwent major hepatectomy during the study period, including 476 right hemihepatectomies (RH) and 235 extended hemihepatectomies (EH; comprising 152 extended right, and 83 extended left). Of the entire cohort, 27 patients (3.8%) developed PHLF, with 5, 11 and 11 of grade A, B and C, respectively. The incidence of PHLF after RH and EH was 12/476 (2.5%) and 15/235 (6.4%), respectively. There were six PHLF-related deaths, including two deaths after RH.
Liver volumetry in patients developing PHLF
Preoperative CT scans were unavailable for 3/27 (11%) patients with PHLF; hence retrospective CT liver volumetric analysis could only be performed in the remaining 24 cases. The demographics of this group are reported in Table 1, with liver volumetry reported in Table 2. These patients had a median TLV of 1707 ml (IQR: 1361-2154), and median left and right lobe volumes of 656 ml (IQR: 499-848) and 932 ml (IQR: 839-1312), respectively. The left lobe volume was a median of 37.1% (IQR: 31.5-43.6) of TLV, and three patients (25%) who developed PHLF after RH had a left lobe volume of < 30% of TLV. The median FLRV% in patients undergoing RH (30.6%, IQR: 28.2-31.2) was significantly higher than in those undergoing EH (26.5%, IQR: 22.6-29.3; p=0.014), with FLRV% being <30% in 4/11 (36%) and 10/13 (77%) patients who developed PHLF after RH and EH, respectively. Three patients developed PHLF despite having FLRV% ≥ 40%, including two patients who developed grade B PHLF after right hemihepatectomy and had radiological drainage of intra-abdominal collections. The third patient developed severe venous congestion of the liver remnant following an extended left hemihepatectomy with significant intraoperative blood loss, and subsequently developed multi-organ failure and died on the third postoperative day.
Selection of matched controls
The 27 patients with PHLF were then matched in a 1:2 ratio to 54 control patients without PHLF. After excluding those for whom preoperative CT scans were unavailable, 24 patients with PHLF and 48 patients without PHLF were included in subsequent analysis. Comparisons between the matched controls and 630 unmatched controls found those included in the matched analysis to be significantly more likely to be male (75% vs. 54%, p=0.004), to have cholangiocarcinoma (35% vs. 8%, p<0.001) and to have undergone extended right hemihepatectomy (46% vs. 18%, p<0.001) than the remainder of the non-PHLF group (Table 1). As such, the matched controls were identified as being a biased sample of the cohort as a whole with very high risk of PHLF, as would be expected given the matching procedure; the median preoperative risk score for matched controls was 10.3 (IQR: 9.0-10.5), and 98% (47/48) had risk scores above the proposed high-risk threshold of >5.516.
Comparisons between patients developing PHLF and matched controls
Comparisons between the PHLF and matched control groups found these to have similar baseline characteristics, including similar PHLF risk scores (median: 10.5 vs. 10.3, p=0.247); 38% of patients in each cohort received chemotherapy prior to hepatectomy (Table 1). Analysis of liver volumetry (Table 2) found the TLV to be similar in the PHLF and matched control groups (median: 1707 vs. 1595 ml, p=0.685). However, the left liver volume was a significantly smaller percentage of TLV in those developing PHLF, compared to matched controls (median: 37.1 vs. 41.6%, p=0.025).
The average FLRV% was significantly lowerin the PHLF group compared to matched controls (median 28.7 vs. 35.2%; p=0.010), with 58% vs. 29% having FLRV% < 30%. The associated AUROC for the differentiation between the PHLF and matched control groups was 0.69 (95% CI: 0.56-0.82), with binary logistic regression returning an odds ratio of 0.92 (95% CI: 0.86-0.99) per unit increase in FLRV% (Figure 1). Based on the Youden’s J statistic, the optimum cut-off value of FLRV% for discriminating between the PHLF and matched control groups was 31.5%. Of those with FLRV% < 31.5%, 19/35 (54%) were in the PHLF group, compared to 5/37 (14%) of those with FLRV% ≥ 31.5%, yielding 79% sensitivity and 67% specificity.
Perioperative factors and postoperative outcomes
The proportions of patients who underwent a concomitant vascular resection or had inflow occlusion were similar in the PHLF and matched control groups (Table 1). The incidence of postoperative bile leak, intra-abdominal sepsis and portal vein thrombosis were also similar between groups. However, patients developing PHLF were significantly more likely to receive a perioperative blood transfusion (48% vs. 15%; p=0.005) and to require radiological drainage of intra-abdominal collections (46% vs. 10%; p=0.002) than matched controls.