From 2009 to 2019, 103 patients underwent 18−F-FDG-PET/CT in our hospital before liver transplantation for HCC. Patients’ age at transplantation was median 62 years (23–71 years). Morphological tumor load was inside Milan in 54 (52%) patients and outside Milan in 49 (48%) patients. 79 (77%) patients received a liver from deceased donors. 24 patients got a split from a living donor (all of them were right lobes). The waiting time was median 9 months (0–45 months) for LT from deceased donor and 6 months (0–34 months) for the living donations. The median interval between PET/CT scan and liver transplantation was 6 months (0–41 months). Further details on patients, tumor load and treatment are shown in Table 1.

Of all 103 patients, 25 had a bridging procedure before PET/CT, another 42 patients had at least one bridging procedure after 18F-FDG PET/CT and before liver transplantation. Three of the 25 patients who were bridged before 18F-FDG PET/CT had a complete pathological response (no vital tumor in the explanted liver), but none of these three patients had a complete radiological response before 18F-FDG PET/CT.

Median follow-up time after LT was 79 months (0–139 months). During the interval 48 patients died, 9 of them in the postoperative interval, 18 due to HCC recurrence. Three patients died from malignant second tumor (lung cancer in 1, ENT area in 2), and 18 died from tumor unrelated causes.

SUVmax in tumor tissue ranged from 1.1 to 23.6 with a median of 2.6. SUVmean in non-tumor liver tissue ranged from 1.1 to 3.6 with a median of 2.3. Median tumor to liver SUV ratio was 1.0 (1.0–8.74).

In Table 1, Quartiles of Tumor to liver SUV ratio are given for all listed subgroups. The Tumor to liver SUV ratio values were compared by Mann–Whitney U test. Tumor to liver SUV ratio (TLR) of the primary tumor was statistically significant higher in Milan out tumors (p = 0.018), “up-to-seven” out tumors (p = 0.015), grade 3 (p = 0.023), patients with AFP level > 400 ng/ml (p < 0.001) and lesions of a diameter of 5 cm and more (p = 0.007).

All other factors (age, sex, bridging therapy before transplantation, type of transplantation, microvascular invasion, number of tumors, Child–Pugh-stage, pT-category, necrosis in the tumor) did not show a statistically significant dependence on the Tumor to liver SUV ratio.

A ROC analysis was performed to define the optimal cut-off for the Tumor to liver SUV ratio to predict tumor recurrence. In the present study, we chose a cut-off value of > 1.38, giving a sensitivity of 70.0% and a specificity of 67.6%. A cut-off value of 1.80 or 2.00 gives sensitivity of 57% and 48% and specificity of 79% and 83%, respectively.

Patients who died in the first 3 months were excluded from survival and recurrence analysis resulting in 94 patients for long-term analysis. All 94 patients were followed up until death or until 31st December 2020. To date, 5 patients lived for more than 10 years after transplantation, 34 for more than 5 years. All living patients have been followed up for at least 1 year. Five patients died from HCC recurrence during the first year after LTX.

Median survival time after transplantation was 106 months, the overall 5- and 10 year-survival rates were 66% and 34%, respectively.

Univariate analysis found only four factors with statistically significant influence on 10 year overall survival: Milan (p = 0.018), “up-to-seven” (p = 0.044), number of lesions (p = 0.011) and pT-category (p = 0.047). Milan, number of lesions and pT-category were included in a multivariate COX regression analysis, which did not show independent statistically significant factors for 10 year overall survival.

A second multivariate COX analysis including the preoperative accessible Milan, AFP level and TLR showed only Milan to be an independent statistically significant factors for 10 year overall survival (p = 0.044, Exp(B) 2.127 (1.020–4.436)).

The majority of the 23 recurrences (70%) occurred in the first two years after transplantation, but there was also a substantial number of later recurrences. The median interval to tumor relapse was 15 months (2–84 months).

Recurrence was intrahepatic in 6 patients and extrahepatic in 17 patients. Sites of extrahepatic recurrence were lung (6 patients), bones (5 cases), adrenal gland (2 patients), peritoneum (2 patients), abdominal wall (1 patient) and lymph nodes (1 patient). Tumor recurrence was treated with curative intent in 8 patients. Pulmonary metastases were resected in 3 patients, adrenal metastases in 2 patients, and metastases in lymph nodes and metastases in the abdominal wall and local recurrence in the liver in one patient each.

5- and 10 year cumulative recurrence rates were 28% and 34%.

Age, sex, bridging before PET/CT and Child stage did not influence cumulative 10-year recurrence rates statistically significant but Milan, “up-to-seven” grade, microvascular invasion, AFP-level, number of lesions, size of lesion, pT-category, Tumor to liver SUV ratio did (Fig. 1). Details for all recurrences are shown in Table 2a.

Cumulative recurrence rate according to tumor to liver SUV ratio

The results for univariate and multivariate COX analyses depended on the three different cut-off values are presented in Table 2a, b.

16 of the 23 recurrences were early recurrences, that means, they occurred in the first 2 years after transplantation. We repeated the analyses given in Table 2a for early recurrences. In univariate as well as in multivariate analyses we saw only marginal differences in the results (Table 2b).

To achieve reliable results in multivariate COX analyses for the 23 patients with recurrence, a maximum of three factors should be used (Peduzzi et al. 1995). pT-category (because this factor takes number and size of lesions and vascular invasion into account), tumor to liver SUV ratio and grade were chosen. All three factors proved to be independent statistically significant factors for 10 year cumulative recurrence rates (Table 2a). For a second multivariate COX analysis, we chose pT-category, tumor to liver SUV ratio and pre-transplant AFP-level. In this analysis, pT-category and pre-transplant AFP-level were independent statistically significant factors for 10 year cumulative recurrence rates but tumor to liver SUV ratio was not (Table 2a).

A multivariate COX analysis only including the preoperative accessible Milan, AFP level and TLR showed only Milan to be an independent statistically significant factors for 10 year cumulative recurrence rates (p = 0.043, Exp(B) 2.838 (1.035–7.781)).

After stratification for Milan criteria, we repeated the univariate Cox Analysis for cumulative recurrence rates depending on the TLR. For a limited numbers of patients, we saw no statistically significant differences (p = 0.069, 4.047 (0.898–18.236) and p = 0.234, 1.999 (0.640–6.247), respectively) between patients Milan in and Milan out.

A score including the preoperative accessible values Diameter, number of lesions and AFP level was calculated. 40 cases with TLR < 1.38, Diameter < 5 cm, 1 to 6 lesion and AFP < 400 ng/ml were defined as to be “low risk”, 54 others “high risk”. All early recurrences and only three patients with recurrences at 25 months, 38 months and 40 months were classified into the low risk group. Therefore, the classification for risk of recurrence was slightly better than the grouping given by Milan or UTS (Fig. 2).

Cumulative recurrence rates according to Milan, UTS and risk score