¹⁸F-FDG PET/CT Predicts the Prognosis of Patients with Hepatocellular Carcinoma Undergoing Liver Transplantation

Globally, liver cancer ranks as the third most common cause of cancer-related death [1, 2], with the majority of cases being hepatocellular carcinoma (HCC) [3]. Liver transplantation (LTx) has proved to be an effective treatment for HCC, since it allows radical removal of not only tumor lesions, but also the cirrhotic portions of the liver [4]. According to data from the China Liver Transplant Registry (CLTR), nearly 40,000 liver transplants have been performed in mainland China [5], with hepatocellular carcinoma (HCC) accounting for 36.8% of all LTx cases [6]. From 2015 to 2021, the cumulative survival rates at 1, 3, and 5 years post-LTx for recipients of deceased-donor liver transplants were 83.7%, 74.5%, and 68.9%, respectively, while for living-donor LTx recipients, the corresponding survival rates were 92.4%, 89.3%, and 88.2%, respectively [7]. However, tumor recurrence affects the clinical outcome of patients who undergo LTx, which also exacerbates, to some extent, the need for donor organs [8]. Therefore, it is crucial to develop effective strategies for selecting appropriate candidates for LTx and predicting postoperative tumor recurrence to enhance the efficacy of LTx in patients with HCC.

The uptake of ¹⁸F-fludeoxyglucose (¹⁸F-FDG) on positron emission tomography/computed tomography (PET/CT) can reflect the level of glucose metabolism of different areas of the organism, which has led to it being widely used in the diagnosis and staging of malignant tumors [9]. This is because the glucose metabolism level is generally related to cell proliferation [10], and the standardized uptake value (SUV) of lesions can be used to predict the prognosis of malignant tumors [11]. For HCC patients who are potential candidates for LTx, an ¹⁸F-FDG PET/CT scan is routinely conducted to evaluate whether there is extrahepatic metastasis before the LTx [12]. Studies conducted by various transplant groups have demonstrated that ¹⁸F-FDG PET/CT data, such as the maximum SUV of tumors, are closely associated with the staging of HCC [13]. The prognostic value of ¹⁸F-FDG PET/CT for HCC patients undergoing surgery has been investigated in several previous studies [14‒16]. However, the relatively small sample sizes and limited follow-up times have limited the utility of the results. Furthermore, the biological characteristics of the patients classified as having a high risk of tumor recurrence by ^18F-FDG PET/CT have not been well elucidated.

In this study, we aim to explore the prognostic value of preoperative ¹⁸F-FDG PET/CT in patients with HCC undergoing LTx and the molecular characteristics of the tumors by integrating DNA next-generation sequencing (NGS) and immunohistochemistry (IHC) analyses. By investigating the relationship between preoperative PET/CT parameters and tumor recurrence, we aimed to provide insights that could improve patient selection for LTx and aid in optimizing post-LTx management for HCC patients.

The HCC patients who underwent deceased-donor LTx (DDLT) at Zhongshan Hospital, Fudan University between January 2015 and December 2019 were retrospectively reviewed and included for the present study. The inclusion criteria were as follows: (1) pathological diagnosis of primary HCC; (2) no history of other malignances; (3) no evidence of extrahepatic metastasis; and (4) no perioperative mortality [17]. In this cohort, the liver transplant committee evaluated patients for LTx using the UCSF criteria [18] to assess tumor-related factors and conducted a multidisciplinary assessment to determine overall eligibility. Factors considered included hepatic functional reserve, absence of extrahepatic metastasis, waiting time, and organ allocation priorities, the patient’s overall condition and comorbidities, psychosocial factors such as support systems and adherence potential, and response to bridging therapies. LTx was also selected for cases where hepatic resection or ablation was contraindicated or considered less effective due to underlying liver dysfunction, nonanatomical resectability, inadequate ablation margins, multifocal or advanced disease, or other patient-specific considerations. The median waiting time for LTx in this cohort was 66.5 days (range: 13–124 days), which was managed through the China Organ Transplant Response System (COTRS, http://www.cot.org.cn). The tumor grade was assigned based on the Edmondson-Steiner grading system. The liver function was assessed according to the Child-Pugh classification. All of the recruited transplant recipients had received a ¹⁸F-FDG PET/CT examination within 90 days prior to transplantation to confirm the absence of extrahepatic metastasis. All PET/CT data were collected prior to any oncologic therapy to ensure accurate baseline assessment of tumor characteristics. Patients with a prior history of other malignancies or missing data were excluded from the study. The design of the study is shown in Figure 1. To identify a prognostic indicator and validate its performance over time, we divided all the included patients into two cohorts based on the surgery date of January 1, 2019. The rationale behind this temporal division was to ensure that the indicator could be both developed and temporally validated using separate datasets, minimizing potential biases from using the same data for model building and testing. This approach simulates real-world clinical scenarios, where historical patient data are used to predict outcomes in future patient populations. We defined the patients who underwent LTx between January 2015 and December 2018 as the derivation cohort, and patients who underwent LTx between January 2019 and December 2019 as the temporal validation cohort. This temporal division follows commonly accepted practices in the literature for selecting an appropriate ratio between derivation and validation cohorts in clinical predictive modeling [19]. A total of 170 patients received bridging therapies prior to LTx, including transarterial chemoembolization in 146 patients, radiofrequency ablation in 39 patients, microwave ablation in 5 patients, radiotherapy in 6 patients, immune checkpoint inhibitors in 8 patients, and tyrosine kinase inhibitors in 32 patients. Additionally, the formalin fixed and paraffin embedded (FFPE) samples from these patients were also collected for further analysis. All transplant recipients received grafts from deceased donors. Eighty-three FFPE tumor samples were collected for DNA NGS assays.

The LTx and postoperative immunosuppressive therapy were performed as reported previously [20]. After LTx, all of the included patients were followed up every month for the first 12 months and then every 3 months thereafter. For each follow-up, assessments of the liver function, tumor markers (alpha-fetoprotein [AFP] and protein induced by vitamin K absence or antagonist-II [PIVKA-II]), and abdominal ultrasonography were conducted. Computed tomography (CT) or magnetic resonance imaging (MRI) was performed to screen for recurrence every 3–6 months during the first 2 years after LTx, then yearly after that. When metastasis or recurrence was suspected, CT, MRI or a bone scan (or PET/CT, if necessary) was performed immediately. Recurrence and/or progression were diagnosed based on imaging evidence. Treatments for patients who experienced tumor recurrence or metastasis were chosen on a case-by-case basis. Generally, further resection, radiofrequency ablation, transcatheter arterial chemoembolization, targeted therapy or chemotherapy, and/or radiotherapy were offered. Recurrence was considered as the primary endpoint of the present study. The time to recurrence (TTR) was defined as the interval between LTx and any type of recurrence [21]. Follow-up was terminated in November 2022.

The IHC analyses were carried out using a two-step protocol (Novolink Polymer Detection System; Novocastra) as described previously [22]. The primary antibodies used for immunohistochemistry were anti-Ki-67 (monoclonal rabbit, clone SP6, diluted 1:100; Abcam), anti-CK19 (monoclonal mouse, clone A53-B/A2, diluted 1:600; Abcam), anti-4-hydroxy-2-nonenal (4-HNE; monoclonal rabbit, ab46545, diluted 1:400; Abcam), anti-8-hydroxy-2′-deoxyguanosine (8-OHdG; monoclonal mouse, ab48508, diluted 5 μg/mL; Abcam), and the anti-programmed cell death ligand 1 (PD-L1; monoclonal rabbit, ab205921, diluted 2 μg/mL; Abcam). Briefly, after microwave antigen retrieval, tissues were incubated with primary antibodies for 60 min at room temperature or overnight at 4°C. Following a 30-min incubation with secondary antibody (Novolink Polymer, RE7112), tissue sections were developed in 3,3′-diaminobenzidine solution under microscopic observation and counterstained with hematoxylin. Negative control slides with the omission of primary antibodies were included in all assays. The results were interpreted by two experienced pathologists. For 4-HNE, 8-OHdG, and PD-L1, IHC scores ≥4 were considered to indicate high expression, and IHC scores <4 were considered as low expression [23, 24]. Specifically, the IHC scores ranging from 0 to 9 were calculated by multiplying the staining intensity score by the stained area score. The staining intensity score was assessed on a scale from 0 to 3 as follows: 0 for no staining, 1 for mild staining (punctate labeling), 2 for moderate staining (dense labeling in a limited number of cells), and 3 for strong staining (dense and uniform labeling across numerous cells). The staining area score was evaluated based on the percentage of positively stained cells as follows: 0 for 0% staining, 1 for 1%–9% staining, 2 for 10%–49% staining, and 3 for 50%–100% staining.

Data were presented as means, medians, SDs, and ranges. Continuous data were compared using the Student’s t test, Wilcoxon signed rank test, or the Mann-Whitney U test, as appropriate. Categorical variables were compared using the Pearson χ² test. A Kaplan-Meier (K-M) analysis was used to calculate the cumulative recurrence rates and overall survival (OS) rates. The Cox proportional hazards regression model was applied to perform univariate and multivariate analyses. The predictive performance was calculated using a receiver operating characteristic curve (ROC) analysis and the cutoff values were determined by the Youden index. Correlation analysis of continuous variables was performed using Pearson’s correlation or Spearman’s correlation analysis, as appropriate. A two-tailed p value <0.05 was considered to be statistically significant. Statistical analyses were conducted with SPSS software version 26.0 (IBM, USA).

The baseline clinicopathological characteristics of the patients are summarized in Table 1. A total of 278 HCC patients with available preoperative ¹⁸F-FDG PET/CT data were enrolled in this retrospective study, including 246 males and 32 females (shown in Fig. 1; Table 1). The mean age of the patients was 54.0 (range: 28–73) years. Patients were divided into 2 cohorts: consecutive patients who received a transplant and had preoperative ¹⁸F-FDG PET/CT data available between January 2015 and December 2018 constituted the derivation cohort (n = 178), and those patients who received transplantation and had preoperative ¹⁸F-FDG PET/CT data available between January 2019 and December 2019 constituted the temporal validation cohort (n = 100). Fifty-seven patients (32%) in the derivation cohort and 35 (35%) in the temporal validation cohort fulfilled the Milan criteria for LTx [21]. The median follow-up period was 54.5 months (range 2.6–95.7 months) for the derivation cohort vs. 40.5 months (range 1.8–47.0 months) for the temporal validation cohort. The cumulative recurrence rate at 1-year post-LTx was 19.1% and was 25.3% at 2 years post-LTx in the derivation cohort; while in the temporal validation cohort these values were 13.0% and 23.0%, respectively (p = 0.193 and p = 0.671, Table 1). The baseline characteristics of the patients with HCC undergoing LTx between the derivation and temporal validation cohorts showed no statistically significant differences.

The ¹⁸F-FDG PET/CT data from all enrolled patients were gathered, and a quantitative analysis of ¹⁸F-FDG SUVs was conducted to obtain ¹⁸F-FDG PET parameters, as reported previously (shown in online suppl. Fig. S1; for all online suppl. material, see https://doi.org/10.1159/000544966) [25]. Initially, we compared the maximum SUV of the tumor (T_SUVmax), the maximum SUV of the normal liver (L_SUVmax), the mean SUV of the normal liver (L_SUVmean), the maximum SUV of the inferior vena cava (IVC_SUVmax), and the mean SUV of the inferior vena cava (IVC_SUVmean) between the patients with or without tumor recurrence during follow-up after LTx. The results showed that the average T_SUVmax of the patients with tumor recurrence was significantly higher in the derivation cohort (6.73 ± 3.37 vs. 4.53 ± 2.80, p < 0.001, shown in Fig. 2a), and similar findings were observed in the temporal validation cohort (6.11 ± 3.75 vs. 4.28 ± 2.61, p = 0.007, shown in Fig. 2b). No statistically significant differences were observed for other parameters, including the L_SUVmax, L_SUVmean, IVC_SUVmax, and IVC_SUVmean. Moreover, we calculated the ratios of T_SUVmax to L_SUVmax (T_SUVmax/L_SUVmax), T_SUVmax to L_SUVmean (T_SUVmax/L_SUVmean), T_SUVmax to IVC_SUVmax (T_SUVmax/IVC_SUVmax), and T_SUVmax to IVC_SUVmean (T_SUVmax/IVC_SUVmean). These ratios were then further evaluated for their potential to predict post-LTx tumor recurrence in both the derivation and temporal validation cohorts.

ROC curve analyses revealed that the T_SUVmax/L_SUVmax had the highest area under the curve in both cohorts (0.785 and 0.757, respectively, shown in Fig. 2c, d). These results indicated that the T_SUVmax/L_SUVmax could be an optimal index for predicting post-LTx tumor recurrence. Based on the calculation of Youden’s index, a T_SUVmax/L_SUVmax >1.43 was found to be the optimal cutoff value for predicting post-LTx recurrence in the derivation cohort. Kaplan-Meier analysis demonstrated that a T_SUVmax/L_SUVmax >1.43 was associated with a higher recurrence rate (63.6% vs. 18.9%), shorter TTR, and shorter OS in the derivation cohort (both p < 0.001, shown in Fig. 3a, b). The median TTR and OS were not reached for patients with T_SUVmax/L_SUVmax ≤1.43, while for those with T_SUVmax/L_SUVmax >1.43, the median TTR was 25.6 months and the median OS was 56.8 months. We further investigated the prognostic value of T_SUVmax/L_SUVmax >1.43 in the temporal validation cohort. Kaplan-Meier analyses verified that a T_SUVmax/L_SUVmax >1.43 had similarly robust predictive value for a higher recurrence rate (47.5% vs. 15.0%) and shorter TTR and OS (p < 0.001 and p = 0.009, shown in Fig. 3c, d) compared to patients with a lower ratio. The median TTR was not reached for patients with T_SUVmax/L_SUVmax ≤1.43, while for those with T_SUVmax/L_SUVmax >1.43, the median TTR was 44.2 months. Although the OS for both the T_SUVmax/L_SUVmax ≤1.43 and >1.43 groups was not reached, patients in the T_SUVmax/L_SUVmax >1.43 group exhibited a significantly shorter OS after approximately 48 months of follow-up.

To further enhance existing LTx criteria, we proposed a novel selection approach that combines the “UCSF criteria” with “T_SUVmax/L_SUVmax ≤1.43,” which we have newly defined as the “UCSF-PET criteria.” Kaplan-Meier analysis revealed that patients meeting the UCSF-PET criteria had significantly lower recurrence rates in both the derivation cohort (8.6% [5/58]) and the temporal validation cohort (9.1% [3/33]), as well as markedly improved OS (Fig. 3e–h).

Univariate and multivariate analyses were performed to evaluate the prognostic significance of various clinicopathologic parameters and the T_SUVmax/L_SUVmax value. Results from the univariate analysis indicated that the ALT, GGT, AFP, PIVKA-II, tumor diameter, tumor number, tumor encapsulation, vascular invasion, Edmondson stage, and T_SUVmax/L_SUVmax were prognostic factors for the TTR and OS in the derivation cohort (Table 2). A T_SUVmax/L_SUVmax >1.43 was significantly associated with both a shorter TTR (hazard ratio [HR], 5.09; 95% confidence interval [CI], 2.95–8.80; p < 0.001) and a shorter OS (HR, 3.93; 95% CI, 2.18–7.07; p < 0.001). In the subsequent multivariate analysis, the T_SUVmax/L_SUVmax value was still a significant risk factor for the TTR (HR, 2.47; 95% CI, 1.18–5.18; p = 0.016).

The prognostic value of the T_SUVmax/L_SUVmax was further confirmed in the temporal validation cohort (Table 3). Univariate analyses showed that a T_SUVmax/L_SUVmax >1.43 remained significantly associated with both a shorter TTR (HR, 4.82; 95% CI, 2.12–10.97; p < 0.001) and shorter OS (HR, 5.91; 95% CI, 1.65–21.19; p = 0.006). The subsequent multivariate analysis validated that the T_SUVmax/L_SUVmax value remained the most significant risk factor for the TTR (HR, 2.86; 95% CI, 1.06–7.69; p = 0.038). Notably, in both the derivation and temporal validation cohort, univariate analysis showed that T_SUVmax/L_SUVmax >1.43, as a predictor of TTR, had a higher HR value than all the other factors (Tables 2, 3).

To further evaluate the predictive value of ¹⁸F-FDG PET for tumor recurrence after LTx in patients with different clinical features, we divided the patients into subgroups according to different potential risk factors for post-LTx recurrence, including age, AFP level, HBsAg level, liver function, tumor pathological features, and transplant criteria. Consistent with our established knowledge, patients with a T_SUVmax/L_SUVmax >1.43 had a HR >1 across various high-risk subgroups for tumor recurrence. Strikingly, we found that a T_SUVmax/L_SUVmax >1.43 retained its prognostic significance in subgroups generally considered to have a low risk of recurrence, including those with an AFP ≤20 ng/mL (HR = 4.16; 95% CI, 1.36–12.78), no liver cirrhosis (HR = 7.09; 95% CI, 1.59–31.56), a single tumor (HR = 11.89; 95% CI, 2.63–53.7), a maximum tumor diameter ≤5 cm (HR = 3.15; 95% CI, 1.55–7.88), with a tumor capsule (HR = 3.82; 95% CI, 2.09–6.99), and no vascular invasion (HR = 5.87; 95% CI, 1.57–2.97); all p < 0.05, shown in Fig. 4a, b).

We further explored the relationship between the T_SUVmax/L_SUVmax and the clinical and histopathological features of patients with HCC. As shown in online supplementary Tables S1 and S2, the group with a T_SUVmax/L_SUVmax >1.43 exhibited significantly higher recurrence rates in both the derivation cohort (63.6% vs. 18.9%, p < 0.001) and the temporal validation cohort (48.8% vs. 13.6%, p = 0.001). Additionally, markers associated with tumor malignancy and aggressiveness were compared across both cohorts [26]. The patients with a T_SUVmax/L_SUVmax >1.43 displayed notably higher preoperative AFP levels (>400 ng/mL), larger maximum tumor sizes (>5 cm), a greater incidence of vascular invasion, and more advanced tumor cell differentiation (all p < 0.01). In terms of tumor staging, patients with a T_SUVmax/L_SUVmax >1.43 were significantly more likely to have advanced CNLC stages and exceed the Milan criteria in both the derivation cohort (all p < 0.001) and the temporal validation cohort (p = 0.002). Additionally, patients with T_SUVmax/L_SUVmax >1.43 had significantly higher PIVKA-II levels compared to those with T_SUVmax/L_SUVmax ≤1.43 in both cohorts (both p < 0.001, online suppl. Fig. S2a). Furthermore, we observed a positive correlation between PIVKA-II and T_SUVmax/L_SUVmax values (both p < 0.001, online suppl. Fig. S2b).

To gain deeper insights into the characteristics of HCC with a T_SUVmax/L_SUVmax >1.43, we examined the protein expression profiles using IHC. The expression of Ki-67 and CK19, two well-established markers of tumor proliferation, differentiation, and aggressiveness in HCC, was evaluated in all included patients [27, 28]. For Ki-67, a positive rate of staining in more than 30% of tumor cells was defined as high expression [29]. HCC tumors with a T_SUVmax/L_SUVmax >1.43 exhibited significantly higher proliferation activity, as indicated by high Ki-67 expression, in both the derivation cohort (68.2% vs. 31.8%, p < 0.001) and the temporal validation cohort (68.3% vs. 42.4%, p = 0.01, shown in online suppl. Fig. S2c). Additionally, the CK19 expression was markedly increased in the T_SUVmax/L_SUVmax >1.43 group, reinforcing the association of a higher T_SUVmax/L_SUVmax with greater tumor aggressiveness and a poorer prognosis in patients with HCC (shown in online suppl. Fig. S2d).

To explore the relationship between T_SUVmax/L_SUVmax and oxidative stress or immune evasion, we screened a new cohort of untreated HCC patients who underwent PET/CT scans within 90 days before surgical resection, with postoperative pathology confirming the tumor diagnoses. A total of 76 FFPE tumor samples from 76 patients were included in this study and divided into two groups based on the T_SUVmax/L_SUVmax = 1.43 threshold: 24 patients in the >1.43 group and 52 patients in the ≤1.43 group. We performed IHC analyses of 4-HNE and 8-OHdG as markers of oxidative stress [30, 31] and PD-L1 as a marker of immune evasion [32]. The results revealed that HCC tumors with a T_SUVmax/L_SUVmax >1.43 exhibited lower 4-HNE, 8-OHdG and higher PD-L1 expression (p = 0.032, p = 0.530, and p = 0.100, online suppl. Fig. S2e–g).

To investigate the molecular alterations in the tumors associated with the ¹⁸F-FDG PET SUV values, we obtained 83 FFPE HCC tumor tissue samples from 83 patients in the temporal validation cohort and performed genomic DNA NGS analysis. We explored the mutation rates and types in 508 cancer-related genes [33]. The sequencing depth of the targeted regions was 145.6-fold ± 28.2-fold, with a mapping rate of 99.6% ± 0.1%. Collectively, the samples contained 1,043 mutations, including 671 nonsynonymous SNV mutations, 164 CNV amplifications, 85 CNV deletions, 56 stopgains, 22 frameshifts deletions, 17 splicing mutations, 11 frameshifts insertions, 8 nonframeshift deletions, 6 fusions, and 3 stoploss mutations.

To explore the relationship between the tumor mutational profiles and T_SUVmax/L_SUVmax, the 83 samples were divided into two groups based on the T_SUVmax/L_SUVmax values: a >1.43 group (n = 38) and ≤1.43 group (n = 45). Using an overall mutation frequency of >5% as the threshold [34], we identified 52 genes that were frequently mutated in this 83 patients cohort (shown in online suppl. Fig. S3a). The top 30 mutated genes enriched in the T_SUVmax/L_SUVmax ≤1.43 and T_SUVmax/L_SUVmax >1.43 groups are shown in the oncoplots (shown in Fig. 5a). The top 5 mutated genes in the T_SUVmax/L_SUVmax ≤1.43 group were TP53, BRCA1, CTNNB1, FAT3, and PMS2, while those in the T_SUVmax/L_SUVmax >1.43 group were TP53, PMS2, CTNNB1, FAT3, RB1, and ARID1A, respectively.

The characteristic mutated genes significantly enriched in the T_SUVmax/L_SUVmax >1.43 group included TP53, EPPK1, FCGR2A, FCGR2B, FCGR3A, FCGR3B, SDHC, B4GALT3, RXRG, MDM4, and SLAMF7 (all p < 0.05, shown in online suppl. Fig. S3b, highlighted in red in shown in Fig. 5a). These genes are associated with classic tumor suppression (TP53 [35]), metabolic reprogramming (B4GALT3 [36], RXRG [37], SDHC [38]), immune evasion and suppression (EPPK1[39], FCGR family [40], and SLAMF7 [41]), and anti-apoptosis (TP53 [35], MDM4 [42]). In contrast, only PCM1, a key gene involved in regulating cell division [43], was identified in the T_SUVmax/L_SUVmax ≤1.43 group (shown in highlighted in red in Fig. 5a). Of interest, 9 of the top 30 mutated genes were shared between the two groups. Although only TP53 mutation was significantly enriched in the T_SUVmax/L_SUVmax >1.43 group, the same tendency was seen for most of remaining mutated genes (75%, 6/8, shown in online suppl. Fig. S3c).

The tumor mutational burden (TMB) was also evaluated, which was defined by dividing the total number of nonsynonymous and indel somatic mutations (excluding known driver genes) by the amount of the genome covered by whole exome sequencing [44]. Notably, a positive correlation was observed between the T_SUVmax/L_SUVmax ratio and TMB in the overall cohort (R = 0.298, p = 0.006, shown in Fig. 5b). We compared different mutation types between patients with different T_SUVmax/L_SUVmax values and found that patients with a T_SUVmax/L_SUVmax >1.43 tended to have more amplification, deletion, and CNV mutations than patients with a T_SUVmax/L_SUVmax ≤1.43 (all p < 0.05, shown in Fig. 5c).

Multiple genes reported to be associated with oncogenic processes according to the gene set described by Chahoud et al. [45] were highly mutant in samples with a T_SUVmax/L_SUVmax >1.43. The TP53 gene set was the most frequently and significantly affected in the T_SUVmax/L_SUVmax >1.43 group (76.3% vs. 42.2%, p = 0.008, shown in Fig. 5d), in which 76.3% of the samples (n = 29) had mutations in at least one of the three key TP53 related genes; TP53 (n = 25, 65.8%), MDM4 (n = 4, 10.5%) or ATM (n = 4, 10.5%). Similarly, the PI3K gene set showed a significantly higher mutation frequency in the T_SUVmax/L_SUVmax >1.43 group (44.7% vs. 22.2%, p = 0.036, shown in Fig. 5d), in which 44.7% of the samples (n = 17) had mutations in at least one of the 12 key PI3K signaling-related genes, including TSC2 (n = 5, 13.2%), TSC1 (n = 5, 13.2%), PTEN (n = 5, 13.2%), PIK3R1 (n = 4, 10.5%), RICTOR (n = 3, 7.9%), MTOR (n = 2, 5.3%), RPTOR (n = 2, 5.3%), PIK3CA (n = 2, 5.3%), PPP2R1A (n = 2, 5.3%), AKT3 (n = 1, 2.6%), AKT2 (n = 1, 2.6%), or STK11 (n = 1, 2.6%). Furthermore, the T_SUVmax/L_SUVmax >1.43 group showed a higher proportion of samples with mutations in genes associated with WNT signaling (39.5% vs. 28.9%), cell cycle regulation (36.8% vs. 26.7%), Hippo signaling (21.1% vs. 15.6%), and TGF-β signaling (10.5% vs. 4.4%, online suppl. Fig. S3d). To investigate the genetic features associated with recurrence, we compared the mutational profiles of patients with and without recurrence in the T_SUVmax/L_SUVmax ≤1.43 and >1.43 subgroups (online suppl. Fig. S3e, f). In the T_SUVmax/L_SUVmax ≤1.43 subgroup, patients with recurrence exhibited more frequent mutations in CTNNB1, ARID2, ATR, FGFR3, ARID1A, LRRK2, and ROS1. Similarly, in the T_SUVmax/L_SUVmax >1.43 subgroup, patients with recurrence showed higher mutation frequencies in TP53, AXIN1, B4GALT3, DDR2, FCGR family genes, MCL1, SDHC, and SLAMF7 compared to those without recurrence.

LTx is the one of the most effective treatment modalities for select HCC patients, but 8–20% of patients suffer from tumor recurrence and metastasis after LTx [21]. Although an increasing number of patients with HCC now have access to LTx, post-LTx recurrence reduces the long-term efficacy of LTx and exacerbates the shortage of available liver donors [46]. Therefore, it is crucial to accurately identify HCC patients who are most suitable for LTx, as well as those at high risk of post-LTx tumor recurrence. Since the uptake of ¹⁸F-FDG in PET can reflect the glucose metabolism level of the tissue, which is closely correlated with the tumor’s proliferative activity [9], it may have profound implications for the clinical outcome evaluation of patients with malignant tumors, especially HCC.

In this study, we demonstrated that ¹⁸F-FDG PET/CT had remarkable predictive value for post-LTx tumor recurrence in patients with HCC. According to the analysis of the survival data, HCC patients with a T_SUVmax/L_SUVmax >1.43 were have a significantly poorer prognosis. Furthermore, we identified that a T_SUVmax/L_SUVmax >1.43 in HCC patients was significantly associated with more aggressive clinical and histopathologic features. Further DNA NGS analyses of 83 HCC tumor samples revealed that tumors with a T_SUVmax/L_SUVmax >1.43 demonstrated a positive correlation between the T_SUVmax/L_SUVmax and TMB, and harbored significantly more amplification, deletion and CNV mutations, particularly in the TP53 and PI3K signaling-related gene sets.

Previous findings obtained from living-donor LTx [47] cohorts showed that the T_SUVmax could be a predictor for post-LTx recurrence. However, inter-machine and inter-patient variability, at least partially due to the use of different PET/CT machines, as well as inter-individual differences in the expression of glucose-6-phosphatase, glucose metabolism, and blood glucose level [48], might lead to differences in T_SUVmax measurements. In this study, we comprehensively measured the ¹⁸F-FDG PET/CT SUV values of tumor, normal liver, and IVC tissue to evaluate its predictive value for determining the prognosis of patients with HCC after DDLT. We found that T_SUVmax/L_SUVmax, T_SUVmax, T_SUVmax/IVC_SUVmean, and T_SUVmax/IVC_SUVmax were potential indices that might be used to predict post-LTx tumor recurrence, while the T_SUVmax/L_SUVmax showed the best predictive efficacy, with an area under the curve value of 0.785 in the derivation cohort and 0.757 in temporal validation cohort. Utilizing the largest DDLT cohort known to us, we proposed that the ratio of the tumor SUV_max to normal liver SUV_max could help alleviate the potential bias due to inter-machine and inter-patient variability. More importantly, a T_SUVmax/L_SUVmax >1.43 showed excellent predictive performance for both the TTR and OS and was superior to conventional clinical indicators in both the univariate and multivariate cox regression analyses. We observed that T_SUVmax/L_SUVmax was an independent prognostic factor for TTR but not for OS (Tables 2, 3), likely due to differences in the determinants of these outcomes. TTR is primarily driven by tumor-specific factors, such as biological aggressiveness, intrahepatic metastasis, and the tumor microenvironment, all of which are closely linked to T_SUVmax/L_SUVmax. In contrast, OS is influenced by a broader range of factors, including comorbidities, liver function, postoperative complications, and treatments administered after recurrence, which dilute the impact of tumor-specific markers. Moreover, recurrent disease management with interventions such as ablation, TACE, or systemic therapies can further modify survival outcomes [49], reducing the direct relevance of pre-LTx T_SUVmax/L_SUVmax for OS. These findings emphasize that while T_SUVmax/L_SUVmax is highly valuable for predicting tumor recurrence, its prognostic utility for OS may be limited by the multifactorial nature of OS in post-LTx patients.

Previous research has concentrated on the predictive value of ¹⁸F-FDG PET/CT for HCC recurrence after hepatectomy [50], the prediction of the therapeutic response to targeted therapy [51], and the correlation with tumor characteristics like microvascular invasion [14]. Here, in both derivation and temporal validation cohorts, we confirmed the predictive value of the T_SUVmax/L_SUVmax for tumor recurrence following LTx. In the subgroup analysis, the T_SUVmax/L_SUVmax consistently demonstrated strong predictive value for post-LTx recurrence in patients who met or exceeded the Milan criteria (shown in Fig. 4). This suggests that combining the T_SUVmax/L_SUVmax with the Milan criteria can more effectively identify HCC patients who are optimal candidates for LTx. Notably, even in subgroups traditionally considered at low risk for post-LTx recurrence based on conventional factors, the T_SUVmax/L_SUVmax effectively predicted patients who experienced recurrence. This suggests that HCC patients traditionally assessed as being at low risk, but with a high T_SUVmax/L_SUVmax, should be considered for more intensive and frequent postoperative surveillance or adjuvant therapy to reduce the likelihood of tumor recurrence. Building on these findings, we proposed a novel selection framework for LTx, termed the “UCSF-PET criteria,” which combines the UCSF morphological criteria with the T_SUVmax/L_SUVmax ≤1.43 threshold. This integrated approach demonstrated promising results in both the derivation and temporal validation cohorts, with recurrence rates below 10%, among patients meeting the UCSF-PET criteria (Fig. 3e–h. The UCSF-PET criteria provide a straightforward and effective strategy for identifying patients at lower risk of post-LT recurrence, offering potential improvements in the allocation of scarce liver grafts. Future studies should validate the UCSF-PET criteria across larger, multicenter cohorts to further refine its clinical utility.

Immunohistochemistry and DNA NGS assays were undertaken to reveal the molecular features and intrinsic mechanism underlying the differences between the two groups. The IHC studies on FFPE tumor tissue showed that the T_SUVmax/L_SUVmax was closely related to high expression of Ki-67 and positive expression of CK19, two canonical markers related to tumor proliferation, aggressiveness, and cancer cell stemness [10, 52]. IHC analysis of 4-HNE, 8-OHdG, and PD-L1 revealed that a T_SUVmax/L_SUVmax ≤1.43 was associated with higher expression of 4-HNE and 8-OHdG, the classical markers of oxidative stress [30, 31], whereas a ratio >1.43 was linked to increased PD-L1 expression, indicative of immune evasion [32]. These findings suggest that PET/CT-derived metabolic parameters may serve as useful indicators of the tumor microenvironment, reflecting distinct biological processes like oxidative stress and immune escape in HCC. However, due to the exhaustion of the FFPE samples from the original cohort, we were unable to conduct further analyses to validate these findings in the original cohort. This is a limitation of the current study, and we suggest that future research with larger cohorts be conducted to further validate these findings and explore the relationship between PET/CT parameters and the tumor microenvironment in HCC. Based on a panel that included 508 cancer-related genes, the mutation pattern of tumors samples from patients with a T_SUVmax/L_SUVmax >1.43 was characterized by TP53, EPPK1, FCGR2A, FCGR2B, FCGR3A, FCGR3B, SDHC, B4GALT3, RXRG, MDM4, and SLAMF7 mutations, and alterations in the TP53, PI3K, WNT, cell cycle, Hippo and TGF-β signaling-related gene sets, which have been reported to be involved in the proliferative, aggressive, and invasive tumor properties, as well as the poorer prognosis, of HCC.

The TP53 signaling plays a pivotal role in maintaining the genomic integrity, and its disruption can lead to unchecked cellular proliferation, reduced apoptosis, and increased tumor aggressiveness [53]. Similarly, mutations in the PI3K signaling-related genes are associated with enhanced cellular growth, survival, and metabolic reprogramming, driving tumor progression and metastasis [54]. Notably, tumors with a T_SUVmax/L_SUVmax >1.43 exhibited a higher TMB, along with more amplification and deletion mutations, suggesting that these tissues have a more unstable genomic profile marked by increased chromosomal instability (CIN). As a hallmark of human cancer, CIN is often associated with increased tumor aggressiveness, metastasis, poorer prognosis, and resistance to treatment [55]. Together, these lines of evidence from IHC and NGS analyses were consistent with the observation that the tumors from patients with a T_SUVmax/L_SUVmax >1.43 had more malignant clinicopathologic tumor features.

In addition to the mutational landscape differences between the T_SUVmax/L_SUVmax >1.43 and ≤1.43 groups, we further explored the genetic features associated with recurrence within these subgroups. Tumors in the T_SUVmax/L_SUVmax >1.43 subgroup that recurred showed higher frequencies of mutations in TP53, AXIN1, B4GALT3, DDR2, FCGR family genes, MCL1, SDHC, and SLAMF7, all of which are implicated in immune evasion [56], metabolic reprogramming [57], and tumor suppression [41]. Conversely, tumors in the T_SUVmax/L_SUVmax ≤1.43 subgroup that recurred were more likely to harbor mutations in CTNNB1, ARID2, ATR, FGFR3, ARID1A, LRRK2, and ROS1, which are involved in WNT signaling [58], DNA damage response [59], receptor tyrosine kinase signaling [60], and tumor suppression [61]. These findings suggest that the molecular drivers of recurrence differ between the two T_SUVmax/L_SUVmax subgroups and highlight the potential for combining PET/CT imaging with genomic profiling to improve risk stratification and guide personalized follow-up strategies.

Globally, metabolic dysfunction-associated steatotic liver disease (MASLD)-related HCC is becoming an increasingly important cause of liver cancer, driven by rising rates of obesity and metabolic syndrome [62]. While the tumor microenvironment and molecular mechanisms in MASLD-related HCC differ from those in HBV-related HCC, both etiologies share a similar pattern of poor-prognosis tumors exhibiting higher ¹⁸F-FDG uptake on PET/CT [63]. This pattern is supported by prior in vivo studies on ¹⁸F-FDG PET/CT imaging in MASLD-related HCC [64], indicating that ¹⁸F-FDG PET/CT holds prognostic value across diverse HCC etiologies. Despite this similarity in imaging features, MASLD- and virus-related HCC harbor distinct mutational profiles [65]. MASLD-related HCC is more frequently associated with TERT promoter, ACVR2A, and WNT signaling pathway mutations, whereas HBV-related HCC typically involves TP53 mutations and viral genome integration [65]. Notably, 90.4% (75/83, Fig. 5a) of patients in our sequencing cohort were HBV-related HCC, reflecting the etiology distribution of our population. As such, the PET/CT-correlated DNA sequencing findings in this study may not completely apply to MASLD-related HCC, highlighting the need for future studies focusing on MASLD-specific cohorts to validate and expand upon these observations. Given the rising incidence of MASLD-related HCC, further research into the interplay between molecular mechanisms and ¹⁸F-FDG PET/CT imaging findings in this patient population is critically important for advancing personalized prognostic tools and therapeutic strategies.

Since TKIs are more effective in treating highly metabolic tumors [66], a T_SUVmax/L_SUVmax >1.43 might indicate that these tumors would be better suited for TKI therapy. Additionally, the positive correlation between the T_SUVmax/L_SUVmax and TMB is noteworthy because a higher TMB has been linked to improved responses to immune checkpoint inhibitors (ICIs) in several cancer types [67]. Tumors with a greater mutation load are more likely to generate neoantigens, enhancing immune recognition [68]. Thus, the observed relationship suggests that patients with an elevated T_SUVmax/L_SUVmax may have a more immunogenic tumor environment and might be better candidates for ICIs therapy, although further validation is required to confirm this possibility.

The main limitations of this work include the retrospective nature of the study, the cohort size, and the somewhat limited follow-up time. To overcome the limitations associated with the retrospective nature of the study, we divided the recruited patients into derivation and temporal validation cohorts simply according to the time of LTx, which made the study results analogous to a prospective study. Another important limitation was that the DNA NGS was carried out with a predesigned 508-gene panel, which may have led to some genetic mutations being missed that contributed to the higher T_SUVmax/L_SUVmax. Additionally, the lack of multicenter validation and the underlying mechanisms still require further elucidation.

To our knowledge, this is the first study illustrating the clinical significance of PET/CT findings in HCC patients who underwent LTx in well-designed derivation and temporal validation cohorts. It is also the largest LTx cohort studied to date for the combination of PET/CT with immunohistochemistry and DNA NGS assays intended to reveal the mechanism at the molecular level. Our study suggests that T_SUVmax/L_SUVmax values obtained from preoperative ¹⁸F-FDG PET/CT scans hold great promise in predicting the posttransplantation prognosis and can be used to improve the selection of LTx recipients. This approach could be particularly valuable when combined with existing criteria for LTx in HCC patients, enabling the identification of recipients at high risk of recurrence and potentially reclassifying patients who would conventionally undergo LTx to receive treatment with ICIs or targeted therapies. Therefore, besides the diagnosis and staging of HCC, pre-LTx PET/CT would act as a potential method to provide an early warning of a high risk for tumor recurrence after surgery, which can help clinicians to design appropriate post-LTx management strategies.

We sincerely thank Dr. Li-Hong Huang from the Department of Biostatistics, Zhongshan Hospital, Fudan University, for her valuable guidance and expertise in addressing statistical issues. Her insights were instrumental in refining the statistical analyses and enhancing the overall quality of this study.

This study was approved by the Zhongshan Hospital Research Ethics Committee (Approval No. B2024-343R) and was conducted in accordance with the ethical principles of the World Medical Association Declaration of Helsinki. Written informed consent was obtained from all patients included in the study.

The authors have no conflicts of interest to declare.

This study was supported by grants the National Natural Science Foundation of China (82341027 and 82072715), the Project of Shanghai Municipal Health Commission (2022LJ005), the Eastern Talent Program (Leading project), the Projects from the Shanghai Science and Technology Commission (21140900300, 22S31901800), Shanghai Municipal Science and Technology Major Project, the project from Shanghai Hospital Development Center (SHDC2023CRD025), the Projects from Science Foundation of Zhongshan Hospital, Fudan university (2021ZSCX28, 2020ZSLC31, ZP2023-017), National Ten-thousand Talent Program, the Project of the Chinese National Key Clinical Specialty Construction (YWP2022-007) and Shanghai Rising-star Program (22QA1408600).

Wen-Jing Zheng, Yang Xu, and Hui Tan contributed equally to data acquisition, analysis and interpretation of data, drafting of the manuscript, and statistical analysis. Shu-Guang Chen, Peng-Xiang Wang, Hai-Xiang Sun, Rui-Zhe Li, and Jian-Wen Cheng contributed to statistical analysis. Hai-Ying Zeng and Yu-Chen Zhong contributed to IHC experiment performing. Jia Fan, Jian Zhou, Hong-Cheng Shi, and Xin-Rong Yang contributed to study concept and design.

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants but are available from the corresponding author Xin-Rong Yang upon reasonable request.