Diagnostic significance of fluorine-18-fluorodeoxyglucose-positron emission tomography/computed tomography in differentiating Edmondson grade II and III hepatocellular carcinoma

Abstract

BACKGROUND

Preoperative prognosis assessment of hepatocellular carcinoma (HCC) is crucial, and pathologic grading is a key prognostic determinant that affects patient prognosis. Therefore, accurate determination of pathological grading before surgical intervention is crucial for optimizing treatment strategies and improving prognostic outcomes.

AIM

To investigate the distinguishing capability of fluorine-18-fluorodeoxyglucose-positron emission tomography/computed tomography (¹⁸F-FDG PET/CT)-derived metabolic parameters between Edmondson grade II and III HCC and to assess their correlation with Ki67 expression levels.

METHODS

We retrospectively assessed the ¹⁸F-FDG PET/CT imaging datasets from 32 patients with solitary HCC, all of whom had pathological confirmation of their diagnosis. Patients were categorized into Edmondson grade II and III groups according to pathological grading criteria. Comparative analyses were conducted on metabolic parameters, including maximum standardized uptake value (SUV_max), mean standardized uptake value (SUV_mean), metabolic tumor volume (MTV), total lesion glycolysis (TLG), tumor-to-normal background ratio (TNR), and tumor-to-blood pool ratio (TBR), between the groups. Further, correlations between these parameters and Ki67 expression were investigated.

RESULTS

Significant differences were observed in SUV_max, SUV_mean, TLG, TNR, and TBR between Edmondson grade II and III HCC groups (P < 0.05), whereas MTV was not significantly different (P = 0.052). The maximum tumor diameter and Ki67 expression percentage significantly varied between the two groups (P < 0.05). SUV_max yielded the largest area under the receiver operating characteristic curve, measuring 0.853 (95% confidence interval: 0.709-0.997, P = 0.001). Using an optimal SUV_max cut-off of 10.95, the sensitivity and specificity for identifying Edmondson grade III HCC were 66.7% and 100%, respectively. Notably, significant positive correlations were identified in terms of SUV_max, SUV_mean, TNR, TBR, and the percentage of Ki67 expression (P < 0.01). Conversely, MTV, TLG, and maximum tumor diameter exhibited no significant association with Ki67 expression (P > 0.05).

CONCLUSION

¹⁸F-FDG PET/CT-derived metabolic parameters, particularly SUV_max, SUV_mean, TNR, TBR, and TLG, are valuable in differentiating Edmondson grade II and III HCC, with SUV_max showing the optimal differential diagnostic efficacy. TLG is a three-dimensional volumetric parameter that holds some differential diagnostic potential, but it fails to display a distinct advantage. Moreover, increased ¹⁸F-FDG uptake and Ki67 expression in tumor tissue correlate with poorer HCC prognoses, emphasizing their potential role in prognostic assessments.

Key Words: Hepatocellular carcinoma; Fluorine-18-fluorodeoxyglucose; Positron emission tomography-computed tomography; Metabolic parameters; Edmondson grading

Core Tip: Fluorine-18-fluorodeoxyglucose (¹⁸F-FDG)-positron emission tomography/computed tomography technology has established its indispensable position in numerous aspects of liver cancer diagnosis, staging, lymph node metastasis assessment, and prognosis prediction. The hypermetabolic nature of malignant cells, induced by accelerated glycolysis, increases ¹⁸F-FDG uptake. Results have revealed an intimate connection between ¹⁸F-FDG uptake and hepatocellular carcinoma differentiation status. This study assessed the diagnostic potential of conventional and three-dimensional metabolic ¹⁸F-FDG-positron emission tomography/computed tomography parameters in differentiating Edmondson grade II from grade III hepatocellular carcinoma.

INTRODUCTION

Hepatocellular carcinoma (HCC), constituting approximately 90% of all primary liver malignancies, establishes itself as the predominant liver cancer type, with its continuously increasing global incidence[1]. The preoperative assessment of HCC prognosis holds crucial importance, with pathological grading serving as a pivotal prognostic determinant affecting patient outcomes[2,3]. Existing research has associated poorly differentiated tumors with an increased propensity for recurrence[4]. Consequently, accurately identifying the pathological grade before surgical intervention is crucial for optimizing treatment strategies and improving prognostic outcomes. Being a cutting-edge radiotracer imaging modality, positron emission tomography (PET)/computed tomography (CT) provides unique advantages by concurrently providing information about both the anatomical architecture of tissues and organs and the metabolic profiles of diverse substances. This technology has already established its indispensability in multiple aspects of liver cancer management, including diagnosis, staging, lymph node metastasis evaluation, and prognosis prediction[5,6]. Among various PET/CT radiotracers, fluorine-18-fluorodeoxyglucose (¹⁸F-FDG), which is a glucose analog, is most extensively used for tumor metabolic imaging. The hypermetabolic nature of malignant cells, driven by accelerated glycolysis, increases ¹⁸F-FDG uptake. Research findings indicate a strong correlation between ¹⁸F-FDG uptake degree in HCC and tumor differentiation status, with poorly differentiated tumors typically demonstrating increased maximum standardized uptake values (SUV_max)[5,7]. In recent years, three-dimensional (3D) metabolic parameters [e.g., total lesion glycolysis (TLG), metabolic tumor volume (MTV)] have gained attention for their tumor behavior prediction capacities. Multiple studies have corroborated their associations with microvascular invasion and prognosis in HCC, with increased MTV and TLG levels indicating increased risks of recurrence and extrahepatic metastasis[8,9]. However, the use of these parameters in distinguishing between different pathological grades of HCC remains underexplored. Originally, Ki67 was deemed an antigen discerned by a monoclonal antibody. However, in essence, Ki67 expression is present across every active cell cycle stage. It serves as a reliable biomarker reflecting cellular multiplication, with its expression elevation associated with greater tumor aggressiveness and malignancy[10]. Some studies have reported no significant differences in overall or tumor-free survival between Edmondson grade I vs grade II HCC or between grade III vs grade IV HCC, but patients with grade III HCC demonstrate markedly worse survival than grade I or II cases[11]. Despite these findings, limited research has focused specifically on differentiating Edmondson grades II and III HCC. This study assesses the diagnostic potential of conventional and 3D metabolic ¹⁸F-FDG PET/CT parameters in distinguishing between Edmondson grades II and III HCC. Further, it seeks to investigate the association between these imaging parameters and Ki67 expression, providing information about their combined use in prognostic assessment and therapeutic planning.

MATERIALS AND METHODS

Study cohort

This retrospective study enrolled 32 histologically verified solitary HCC cases from March 2021 to April 2023. The patients, who received ¹⁸F-FDG PET/CT scans at the First Affiliated Hospital of Zhengzhou University, comprised 26 males and 6 females aged between 28 and 72 (mean: 53.3 ± 10.3) years. Inclusion criteria were patients who underwent ¹⁸F-FDG PET/CT examination in our hospital, which were later confirmed as HCC by surgical or biopsy pathological results, and had not received any antineoplastic therapies before the imaging examination. Exclusion criteria: (1) Those with a pathological grade of Edmondson I; (2) Those with active tumors originating from other organs; and (3) Those with incomplete clinical, imaging, and pathological data. Of these, 25 patients underwent curative surgical resection, whereas the remaining 7 underwent percutaneous biopsy. Patients were grouped based on the Edmondson pathological grading system as grades II (n = 17) and III (n = 15)[12]. The predominant grade was documented in cases where intratumorally heterogeneity in grading was observed. Immunohistochemical staining was conducted on all pathological specimens to quantify Ki67 expression levels.

Imaging apparatus and protocols

A Siemens Biograph Truepoint 64 PET/CT system and a HM-20 medical cyclotron accelerator of the Sumitomo Group of Japan were used to acquire all PET/CT scans. Fasting for ≥ 6 hours was enforced before imaging, and blood glucose levels were strictly regulated at < 11.1 mmol/L. After ensuring that the patient was in a resting state, ¹⁸F-FDG (synthesized in-house with a radiochemical purity exceeding 95%) was administered via the cubital vein at a standardized dose of 3.70-5.55 megabecquerel/kg. After a 60-minute uptake period, covering an anatomical range between the skull base and upper thighs, PET/CT imaging was performed. The acquisition used 120-kV tube voltage, auto-modulated tube current modulation (100-160 mA), 13 noise index, and a 3-mm slice thickness. Each bed position was imaged for 2.5 minutes. Post-processing included CT-based attenuation correction and PET image iterative reconstruction.

Image processing and quantitative analysis

PET and CT datasets were co-registered and fused utilizing the Siemens Syngo TrueD workstation. Two highly experienced nuclear medicine physicians, blinded to clinical and pathological details, independently assessed the original and fused images to determine and localize tumor lesions. Using PET/CT images, the tumor margins were manually outlined to define a 3D volume of interest. Conventional metabolic metrics, such as the SUV_max and mean standardized uptake value (SUV_mean), as well as volumetric parameters, including MTV and TLG, were then automatically calculated using the post-processing software. A region of interest measuring 1.5-2.0 cm² was delineated in the normal liver parenchyma of either the right or left lobe. Further, a region of interest of 1.0-1.5 cm² was placed on the abdominal aorta at the corresponding axial level of the lesion to serve as a blood pool reference. The tumor-to-normal background ratio (TNR) was computed as tumor SUV_max divided by normal liver SUV_max, whereas the tumor-to-blood pool ratio (TBR) was derived by dividing the tumor SUV_max by the SUV_max of the corresponding-level blood pool.

Statistical analysis

Statistical Package for the Social Sciences version 26.0 was used for data analysis. The Shapiro-Wilk test was applied for data distribution normality assessment. Continuous variables following and deviating from normal distribution were reported as mean ± SD and median (interquartile range), respectively. Group comparisons were performed using independent samples t-tests when variables followed a normal distribution and had equal variances; otherwise, the Mann-Whitney U test was employed. Receiver operating characteristic (ROC) curves were used to assess each parameter’s efficacy in distinguishing between Edmondson grades II and III HCC, with discriminative performance demonstrated by the area under the curve (AUC). The Youden’s index was used for optimal cut-off determination. We defined statistical significance as a P-value of < 0.05 (two-tailed). Associations between imaging parameters and Ki67 expression were assessed with Spearman’s rank correlation coefficients.

RESULTS

Comparative analysis of metabolic parameters between Edmondson grade II and III HCC groups

Statistically significant differences were determined between Edmondson grades II and III HCC groups for several metabolic parameters, including SUV_max, SUV_mean, TLG, TNR, and TBR (P < 0.05; Figures 1 and 2). However, MTV differed insignificantly (P = 0.052). Of the study cohort, the median maximum tumor diameter (MTD) was 5.65 cm (interquartile range: 4.40-9.20 cm), and the median Ki67 expression level was 30% (interquartile range: 21.25%-60%). Both the MTD and the Ki67 expression percentage demonstrated marked differences between the Edmondson grade II and III HCC cases (P < 0.05; Table 1).

Figure 1 Hepatocellular carcinoma in the right hepatic lobe, with an Edmondson grade of III (Case 1). The percentage of Ki67 expression was 30%, maximum standardized uptake value was 15.56, mean standardized uptake value was 6.24, metabolic tumor volume was 94.7, total lesion glycolysis was 590.5, tumor-to-normal background ratio was 5.02, and tumor-to-blood pool ratio was 8.64. A: Non-enhanced computerized tomography image depicted a slightly hypodense mass (blue arrow) beneath the capsule of the right hepatic lobe, with an approximate computerized tomography value of 32 Hounsfield unit, a three-dimensional volume of interest was manually outlined along the tumor periphery; B: Normal liver tissue region of interest (white arrow) and abdominal aorta region of interest (green arrow); C: Axial maximum intensity projection image; D: Coronal view; E: Sagittal view; F: Sagittal maximum intensity projection image. PT: Positron tomography; CT: Computed tomography; 3D: Three-dimensional.

Figure 2 Left hepatic lobe hepatocellular carcinoma, with an Edmondson grade of II (Case 2). The percentage of Ki67 expression was 10%, maximum standardized uptake value was 5.37, mean standardized uptake value was 2.9, metabolic tumor volume was 28.55, total lesion glycolysis was 82.7, tumor-to-normal background ratio was 1.79, and tumor-to-blood pool ratio was 2.24. A: Non-enhanced computerized tomography scan identified a slightly hypodense mass (blue arrow) under the capsule of the left hepatic lobe, with a computerized tomography value of approximately 34 Hounsfield unit; a volume of interest was manually drawn along the tumor margins; B: Region of interests were delineated in normal liver tissue (white arrow) and the abdominal aorta (green arrow); C: Axial maximum intensity projection image; D: Sagittal view; E: Coronal view; F: Coronal maximum intensity projection image. PT: Positron tomography; CT: Computed tomography; 3D: Three-dimensional.

Table 1 Comparative analysis of metabolic parameters between Edmondson grade II and III hepatocellular carcinoma groups, maximum tumor diameter (interquartile range).

	Edmondson grade II (n = 17)	Edmondson grade III (n = 15)	Z	P value
SUVmax	6.43 (4.36-7.73)	13.79 (9.04-17.99)	3.399	0.001
SUVmean	3.24 (2.83-4.145)	6.24 (3.99-6.83)	3.040	0.002
MTV	31.23 (18.91-120.52)	94.7 (41.9-280.0)	1.945	0.052
TLG	107.3 (54.94-384.825)	590.5 (208.9-1828.0)	2.776	0.006
TNR	2.18 (1.405-2.605)	4.76 (2.58-5.62)	2.964	0.003
TBR	3.18 (2.015-4.18)	6.27 (2.93-8.64)	2.663	0.008
Maximum tumor diameter (cm)	5.4 (4.15-7.05)	6.9 (4.8-11.3)	1.984	0.047
Ki67 expression percentage (mean ± SD)	26.76% ± 13.22%	52.3% ± 20.5%	3.356	0.001

SUV_max: Maximum standardized uptake value; SUV_mean: Mean standardized uptake value; MTV: Metabolic tumor volume; TLG: Total lesion glycolysis; TNR: Tumor-to-normal background ratio; TBR: Tumor-to-blood pool ratio.

Assessment of the discriminatory efficacy of each parameter in differentiating between Edmondson grade II and grade III HCC via ROC curve analysis

Among the assessed parameters, SUV_max demonstrated the highest discriminative ability, evidenced by an AUC of 0.853 [95% confidence interval (CI): 0.709-0.997, P = 0.001]. The Youden’s index-determined optimal SUV_max cut-off of 10.95 provided 66.7% sensitivity and 100% specificity for identifying Edmondson grade III HCC (Table 2, Figure 3).

Figure 3 Receiver operating characteristic curves in differentiating Edmondson grade II and III hepatocellular carcinoma. A: Receiver operating characteristic curve of maximum standardized uptake value, mean standardized uptake value; B: Receiver operating characteristic curve of total lesion glycolysis, tumor-to-normal background ratio, and tumor-to-blood pool ratio. SUV_max: Maximum standardized uptake value; SUV_mean: Mean standardized uptake value; TLG: Total lesion glycolysis; TNR: Tumor-to-normal background ratio; TBR: Tumor-to-blood pool ratio.

Table 2 Diagnostic performance of various parameters based on receiver operating characteristic curve analysis.

	Cutoff	AUC (95%CI)	Sensitivity (%)	Specificity (%)	P value
SUVmax	10.95	0.853 (0.709-0.997)	66.7	100	0.001
SUVmean	4.32	0.816 (0.655-0.976)	73.3	88.2	0.002
TLG	248.1	0.788 (0.630-0.947)	71.5	76.5	0.006
TNR	2.53	0.808 (0.643-0.973)	80	76.5	0.003
TBR	4.46	0.776 (0.597-0.956)	70.2	82.4	0.008

SUV_max: Maximum standardized uptake value; SUV_mean: Mean standardized uptake value; TLG: Total lesion glycolysis; TNR: Tumor-to-normal background ratio; TBR: Tumor-to-blood pool ratio; AUC: Area under the curve; CI: Confidence interval.

Association between metabolic parameters and Ki67 expression levels

A significant positive correlation was identified between Ki67 expression percentage and several metabolic parameters, including SUV_max, SUV_mean, TNR, and TBR, with Spearman’s rank correlation coefficient (r_s) of 0.549, 0.658, 0.505, and 0.499, respectively (P = 0.001, < 0.001, 0.003, and 0.004, respectively). Conversely, Ki67 did not correlate strongly with MTV (r_s = 0.042, P = 0.820), TLG (r_s = 0.245, P = 0.176), or MTD (r_s = 0.156, P = 0.394).

DISCUSSION

HCC represents a prevalent gastrointestinal malignancy in clinical settings, which is associated with high recurrence, metastasis, and mortality[13]. The metabolic behavior of HCC is closely associated with its malignant potential, as poorly differentiated tumors demonstrate reduced glucose-6-phosphatase activity, which decreases dephosphorylation and consequently increases ¹⁸F-FDG uptake[14]. This metabolic alteration enables ¹⁸F-FDG PET/CT to effectively assess tumor aggressiveness by quantifying the ¹⁸F-FDG uptake. Previous studies have consistently revealed that Edmondson grade I + II HCC demonstrates lower ¹⁸F-FDG PET/CT positivity rates, along with reduced SUV_max and TNR values, compared with Edmondson grade III + IV tumors[15]. However, limited research has specifically addressed the differentiation between Edmondson grades II and III HCC using ¹⁸F-FDG PET/CT. Herein, we analyzed data from patients with solitary HCC and revealed that Edmondson grade III tumors exhibited significantly higher SUV_max, SUV_mean, TLG, TNR, and TBR levels compared with grade II tumors (P < 0.05). Notably, this study is among the first to identify TLG, which is a 3D volumetric parameter, as a potential discriminator between grade II and III HCC. However, MTV did not reach significant inter-group differences (P = 0.052). This finding may be affected by limited samples and the inclusion of a patient with grade II HCC presenting a massive tumor (MTD > 10 cm; MTV of 581). This outlier could potentially account for nonsignificant differences in MTV. Noteworthily, the pathological grading in this case was derived from biopsy, which may be subject to sampling bias due to tumor heterogeneity, potentially affecting the grading accuracy. These results align with the findings of Zhang et al[16], who conducted an in-depth ¹⁸F-FDG PET/CT-based investigation on moderately and poorly differentiated HCCs, revealing that poorly differentiated HCCs displayed higher SUV_max and TNR levels compared with moderately differentiated tumors. Further, they highlighted a positive correlation, wherein an increase in MTD was accompanied by corresponding SUV_max and TNR elevations. Our study did not perform subgroup analyses based on tumor size due to the limited cohort; however, we revealed a significantly larger MTD in patients with Edmondson grade III HCC compared with grade II cases (P < 0.05). To further assess the diagnostic efficacy of various parameters in distinguishing grade II from grade III HCC, we conducted a ROC curve analysis. Among all parameters assessed, SUV_max provided the strongest discriminative power (AUC = 0.853, 95%CI: 0.709-0.997). When SUV_max was set at an optimal cut-off value of 10.95, the sensitivity for identifying grade III HCC was 66.7%, whereas the specificity reached 100%. Interestingly, the diagnostic performance of 3D volumetric parameters, such as TLG, did not surpass that of conventional metabolic indices such as SUV_max and SUV_mean. This indicates that traditional metabolic markers remain highly effective in differentiating HCC grades, whereas volumetric parameters provide additional insights.

The Ki67 protein expression level is intricately related to tumor cell proliferative activity, differentiation, aggressiveness, and patient outcomes[17,18]. Supporting this, Zhang et al[16] reported that HCC cases with high Ki67 expression demonstrated substantially increased SUV_max and TNR values compared with those having low expression, alongside a robust correlation between SUV_max and Ki67 expression levels. In line with these results, our study identified significant positive associations between the Ki67 expression percentage and metabolic parameters such as SUV_max, SUV_mean, TNR, and TBR (P < 0.01). However, volumetric parameters, such as MTV or TLG, exhibited no meaningful correlations with Ki67 expression (P > 0.05). These observations reinforce the concept that increased ¹⁸F-FDG uptake in HCC indicates poorer tumor differentiation and heightened proliferative activity. The absence of a correlation between MTV/TLG and Ki67 expression may imply that HCCs with low Ki67 expression remain present as large-sized lesions, leading to substantial MTV and TLG. To gain a more comprehensive understanding of this complex association, future investigations that involve multi-center collaborations and large-scale sample sizes are warranted for further exploration.

Our studies highlighted the clinical use of ¹⁸F-FDG PET/CT in guiding treatment strategy formulation for HCC, as well as its valuable role in staging and treatment response assessments in certain cancers. A preliminary retrospective study in France demonstrated that ¹⁸F-FDG PET/CT detected new HCC lesions in 21% (26/122) of cases, which prompted tumor stage reclassification in 11% (14/122) and treatment plan alteration in 14% (17/122)[18]. Another study revealed that combined ¹⁸F-FDG and ¹⁸F-fluorocholine (dual tracer) PET-CT revealed new lesions in 14% (39/273) of patients, 12% (32/273) of patients with Barcelona Clinic Liver Cancer advanced stage, and 8% (21/273) of patients had treatment changes[19].

Several limitations should be acknowledged in this study. The first limitation is the single-center design and modest cohort size. Second, pathological data for seven cases were obtained via biopsy, which may introduce sampling bias due to tumor heterogeneity. Further, previous evidence has indicated a potential connection between ¹⁸F-FDG PET/CT metabolic measurements and HCC differentiation levels[20]. In hyperdifferentiated HCC, the high gluconeogenesis rate may lead to ¹⁸F-FDG PET/CT uptake levels comparable to those of unaffected liver tissue. Therefore, certain hyperdifferentiated HCC lesions can present with low metabolic burden despite high tumor burden on PET, possibly impacting clinical assessments. The accuracy of pathological grading and Ki67 expression quantification could be affected by the specific region of the tumor sampled, thereby potentially affecting the reliability of the results.

CONCLUSION

The differentiation between Edmondson grade II and grade III HCC involved ¹⁸F-FDG PET/CT-derived metabolic parameters, including SUV_max, SUV_mean, TNR, TBR, and TLG. Among these, SUV_max appeared as the most effective diagnostic parameter. Meanwhile, TLG, which is a 3D volumetric metric, demonstrated some discriminative capacity. However, it did not surpass the performance of conventional metabolic indices. Importantly, the strong positive correlations between SUV_max, SUV_mean, TNR, TBR, and Ki67 expression levels indicate that higher ¹⁸F-FDG uptake reflects increased Ki67 expression and greater proliferative activity, which are associated with poorer clinical outcomes in patients with HCC. These results highlight how ¹⁸F-FDG PET/CT can help in tailoring treatment strategies and prognostic assessment in HCC management.