Metastasis in Patients With Hepatocellular Carcinoma

Prevalence, Determinants, Prognostic Impact and Ability to Improve the Barcelona Clinic Liver Cancer System

Chia-Yang Hsu; Po-Hong Liu; Shu-Yein Ho; Yi-Hsiang Huang; Yun-Hsuan Lee; Rheun-Chuan Lee; Teddy S. Nagaria; Ming-Chih Hou; Teh-Ia Huo


Liver International. 2018;38(10):1803-1811. 

In This Article


Patient Characteristics and Prevalence of Metastasis

The baseline demographics of study patients with and without metastasis are shown in Table 1. There were 357 (10%) patients who had documented metastasis at the time of enrolment. Patients with metastasis were significantly younger, more often had hepatitis B and alcoholism, and less frequently had hepatitis C. In addition, they had significantly more advanced cirrhosis, larger tumour burden, higher serum α-foetoprotein (AFP) level, poorer PS, more advanced BCLC stages and more often underwent non-curative anti-cancer treatments (all P < .05). The most common site of metastasis was lung (n = 175), followed by lymph node (n = 136), bone (n = 63), adrenal gland (n = 12) and omentum/peritoneum (n = 6). Forty-eight patients had at least 2 metastatic sites/organs at the time of diagnosis. Patients with metastasis were more likely to receive best supportive care, follow by palliative transarterial chemoembolization (TACE) and targeted therapy (P < .05).

Associated Factors of Metastasis Stratified by TTV

Patients were further stratified by TTV to investigate the associated factors of metastasis. Vascular invasion was consistently linked with metastasis independent of the status of TTV (Table 2 and Figure 1; all P < .05). For patients with TTV <250 cm3, higher AFP level was significantly associated with metastasis (P < .05). Multinodular tumours were related to metastasis only in patients with TTV <50 cm3 (P < .05). The largest tumour dimension with a cut-off of either 3 or 5 cm was not significantly associated with metastasis after patients were stratified by TTV (all P > .05).

Figure 1.

Association between hepatocellular carcinoma (HCC)-related variables and metastasis stratified by total tumour volume (TTV). Vascular invasion is associated with metastasis regardless of tumour burden (all P < .05). Multiple tumours and higher serum α-foetoprotein level are significantly related to metastasis in patients with smaller tumour burden (P < .05)

Univariate Survival Analysis

After an average of 31 ± 33 months or 8820 person-years follow-up, 2101 patients died. As shown in Table 3, male, age of 65 years or older, alcoholism, larger tumour burden, advanced cirrhosis, hyponatremia, renal insufficiency, poorer PS, higher serum AFP and diabetes mellitus were significantly associated with decreased overall survival in univariate analysis (all P < .05). Patients with metastasis had a significantly shortened survival compared to patients without metastasis (Figure 2A; P < .001), but overall survival was not significantly influenced by the sites of metastasis among patients with metastasis (P > .1). Patients with vascular invasion had a poorer prognosis compared to patients without vascular invasion (Figure 2B; P < .001). Among 4 levels reflecting vascular invasion and metastasis including negative for both, vascular invasion alone, metastasis alone and positive for both, patients with a combination of vascular invasion and metastasis had the worst overall survival (Figure 2C; P < .05).

Figure 2.

Metastasis and vascular invasion are associated with significantly reduced survival of hepatocellular carcinoma (HCC) patients (panels A and B; both P < .001). HCC patients with the co-existence of vascular invasion and metastasis have significantly shortened survival compared to patients with vascular invasion alone or metastasis alone (panel C; both P < .05)

Multivariate Survival Analysis

In the multivariate Cox regression model (Table 3), male (HR: 1.131, 95% CI: 1.014–1.262, P = .027), age ≥65 years (HR: 1.204, 95% CI: 1.098–1.321, P < .001), albumin <3.7 g/dL (HR: 1.685, 95% CI: 1.526–1.86, P < .001), bilirubin ≥0.9 mg/dL (HR: 1.267, 95% CI: 1.148–1.399, P < .001), international normalized ratio of prothrombin time ≥1.1 (HR: 1.169, 95% CI: 1.055–1.296, P = .003), sodium <139 mmol/L (HR:1.187, 95% CI: 1.082–1.302, P < .001), eGFR <60 mL/min/1.73 m2 (HR: 1.237, 95% CI: 1.118–1.368, P < .001), AFP ≥400 ng/dL (HR: 1.635, 95% CI: 1.478–1.809, P < .001), multiple tumours (HR: 1.307, 95% CI: 1.196–1.428, P < .001), TTV ≥47 cm3 (HR: 1.551, 95% CI: 1.396–1.723, P < .001), ascites (HR: 1.426, 95% CI: 1.274–1.596, P < .001), PS 1–2 (HR: 1.412, 95% CI: 1.268–1.573, P < .001), PS 3–4 (HR: 2.084, 95% CI: 1.78–2.441, P < .001), metastasis alone (HR: 2.548, 95% CI: 2.035–3.191, P < .001), vascular invasion alone (HR: 2.633, 95% CI: 2.335–2.969, P < .001) and co-existence of metastasis and vascular invasion (HR: 2.799, 95% CI: 2.348–3.338, P < .001) were identified as independent predictors of poor prognosis.

Comparison of Prognostic Accuracy of Original and Modified BCLC Systems

The modified BCLC staging system was designed by re-allocating 181 patients with the co-existence of metastasis and vascular invasion from BCLC stage C to stage D. Survival distributions are shown in Figure 3; pairwise comparison shows significant survival differences across different BCLC stages for both systems, and the modified BCLC model had lower AIC value compared to the original BCLC system.

Figure 3.

Re-allocating 181 hepatocellular carcinoma (HCC) patients with a combination of vascular invasion and metastasis from Barcelona Clinic Liver Cancer (BCLC) stage C to D improves the predictive accuracy of BCLC system (lower Akaike information criterion value)