Characteristics of the derivation cohort
A total of 749 HCC patients who underwent LR met the inclusion criteria and were enrolled in the derivation cohort. Overall, 138 patients (18.4%) had a BCLC score of 0, 542 (72.4%) had a BCLC score of A, and 69 (9.2%) had a BCLC score of B (Table 1). Of the 749 patients, 394 (52.6%) were HBsAg positive, 252 (33.6%) were anti-HCV antibody positive, 352 (47.0%) had a major resection, 254 (33.9%) were > 65 years of age, 576 (76.9%) were male, 120 (16.0%) had an AFP level of > 400 ng/ml, and 737 (98.4%) were Child–Pugh class A. Pathological examination showed 290 patients (38.7%) with cirrhosis, 29 (3.9%) with poorly differentiated tumors, and 63 (8.4%) with pathology T stage 3 or 4. The highest proportion of BCLC stage A patients were in the > 65-year age group. The highest proportion of BCLC stage 0 patients had cirrhosis, while the BCLC stage B group included the highest number of patients with AFP ≥ 400 ng/ml, major resection, and pathology T stage 3 or 4.
Cutoff values of tumor burden score
Patients were divided into three groups according to the TBS: high TBS (> 7.9; n = 81 [10.8%]); medium TBS (2.6–7.9; n = 474 [63.3%)); and low TBS (< 2.6; n = 194 [25⋅9%)) (Fig. 1). All 138 patients with BCLC stage 0 HCC had a low TBS. Patients with BCLC stage A tumors had a low (n = 82 [15.1%]), medium (406 [74.9%]), or high (54 [10.0%]) TBS, whereas those with BCLC stage B HCC had a medium (42 [60.9%]) or high (27 [39.1%]) TBS.
TBS could be used to stratify the OS of the derivation cohort
After a median follow-up of 19.9 (inter quantile range (IQR) = 10.9–85.9) months, the 5-year OS following LR of the derivation cohort was 70.6% (95% CI = 66.3–75.3%). The five-year OS varied based on BCLC stage (85, 73, and 48% for BCLC stages 0, A, and B, respectively; p < 0.001). The TBS could be used to stratify OS (5-year OS: 87.5, 72, and 48% for low, medium, and high TBS, respectively; p < 0⋅001) (Fig. 2).
TBS could be used to stratify OS at the BCLC stages A and B
To evaluate whether TBS could be used to stratify the OS at BCLC stages A and B, subgroup analyses were performed for BCLC stage A and B patients. In the BCLC stage A group, OS worsened progressively with increasing TBS (5-year OS: 87.5, 70, and 63% for low, medium, and high TBS, respectively; p = 0.04) (Fig. 3a). Of the patients with BCLC stage B HCC, those with a high TBS had a worse 5-year OS than those with a medium TBS (23% vs. 65%; p = 0.002) (Fig. 3b).
Multivariate analysis of prognostic factors associated with overall survival in BCLC stages A and B
Multivariate analysis showed that an age of > 65 years (HR = 2.195, 95% CI = 1.461–3.298, p < 0.001), the presence of cirrhosis (HR = 1.978, 95% CI = 1.309–2.988, p = 0.001), medium TBS vs. low TBS (HR = 2.390, 95% CI = 1.024–5.581, p = 0.04), and high vs. low TBS (HR = 3.885, 95% CI = 1.443–10.456, p = 0.007) were associated with OS in BCLC stage A (Table 2). Multivariate analysis showed that an age of > 65 years (HR = 3.269, 95% CI = 1.297–8.237, p = 0.012), AFP > 400 ng/ml (HR = 2.938, 95% CI = 1.138–7.589, p = 0.026), and high vs. medium TBS (HR = 2.542, 95% CI = 1.077–6.002, p = 0.033) were associated with OS in BCLC stage B (Table 2).
Validation of TBS
Fifty-six patients (20.4%) had BCLC stage 0, 196 (71.3%) had BCLC stage A, and 23 (8.4%) had BCLC stage B HCC in the validation cohort. In this cohort, 72 patients (26.2%) had a high TBS, 163 (59.3%) had a medium TBS, and 40 (14.5%) had a high TBS (Table 3). In addition, TBS could be used to stratify the OS of the validation cohort. Patients with a low, medium, and high TBS had a 5-year OS of 88, 78, and 50% respectively (p < 0.001) (Fig. 4a). The TBS could also be used to stratify OS in BCLC stage A (5-year OS = 83, 82, and 60% for low, medium, and high TBS, respectively; p = 0.03) (Fig. 4b).
Correlation between radiographic and pathological tumor size and number
The characteristics of this subgroup of 180 patients are shown in Table 4. Thirty-nine patients (21.7%) were BCLC stage 0, 129 (71.7%) BCLC stage A, and 12 (6.7%) BCLC stage B. In addition, 161 (89.4%) patients had a radiographic solitary tumor and 19 (10.6%) had radiographic multiple tumors; 160 (88.9%) patients had a pathological solitary tumor and 20 (11.1%) had pathological multiple tumors. The median (IQR) radiographic tumor size was 29 (22–50) mm. The median (IQR) pathological tumor size was 28.5 (21–50) mm. Kappa was 0.665, indicating good agreement between radiographic and pathological tumor number. The discrepancy between radiographic and pathological tumor numbers was mainly in nodules < 1 cm, found either by radiographic or pathological examination. Of the 180 patients, 12 (6.7%) showed a discrepancy between radiographic and pathological tumor numbers. Of these patients, seven had satellite nodules < 1 cm that were identified on pathological examination but were not detected in imaging studies; five patients had radiographic tumors < 1cm which were not found on pathological examination. The correlation between radiographic and pathological tumor size was almost perfect (r = 0.973; p < 0.001).