We conducted a retrospective study of all HCC patients with PVTT who were treated with RT at Samsung Medical Center between January 1998 and December 2008. During this period, total 742 patients were received RT because of primary liver cancer in our institution. Of these patients, 710 patients had been diagnosed as HCC. And 320 patients (45.1%) had vascular invasion from HCC. Without PVTT, inferior vena cava or hepatic vein was involved in 19 (2.7%). And 20 patients (2.8%) had PVTT in small branches of portal vein (segmental/section branches except main, left or right hemiliver portal vein). If the PVTT is confined to small branches only, the other local modality (surgery, radiofrequency ablation, TACE, etc) should be considered. In the current study, we analyzed 281 patients (39.6%) who had tumor thrombosis in the main, right or left hemiliver portal vein.

The diagnosis of HCC was made based on the guidelines proposed by the Korea Liver Cancer Study Group (9). Using these criteria, a patient is diagnosed as HCC if who has one or more risk factors (hepatitis B or C virus infection, and/or cirrhosis) and one of the following: a serum α-fetoprotein (AFP) level of > 400 ng/mL and a positive result with at least one of the three typical imaging techniques (triple phase computerized tomography [CT], contrast enhanced dynamic magnetic resonance imaging [MRI] or hepatic angiography); or a serum AFP level of < 400 ng/mL and positive findings with at least two of three imaging techniques. A positive finding for typical HCC with dynamic CT or MRI is indicative of arterial enhancement followed by venous washout in the delayed portal/venous phase.

PVTT was demonstrated using helical CT scans with contrast enhancement, MRI scans, or angiography. On contrast-enhanced CT scans, PVTT was identified by the presence of a low-attenuation intra-luminal filling defect adjacent to the primary tumor. If thrombi were located in both hemiliver branches and the main trunk, we categorized the thrombi as main PVTT because of the retrograde invasion of portal vein thrombosis. Differentiating malignant PVTT and benign PVTT may be difficult. Intrathrombus vascularity, observed in the arterial phase of imaging studies after the administration of contrast, has been reported to be a sign that is specific for PVTT on both CT and MRI (10). We assessed malignant PVTT based on these enhancing and/or growing pattern, response to treatment, and relationship with primary tumor.

The following clinical data were collected from the medical records: patient demographics; complete blood count; chemistry profiles; liver function test; AFP; size of the tumor and PVTT; number of the tumor; involved side of the liver by the tumor; involved site of the portal vein by the PVTT; length of the PVTT; completeness of the portal vein occlusion; presence of lymph nodes or distant metastasis; Eastern Cooperative Oncology Group (ECOG) performance status; Child-Pugh score and class; date of diagnosis; type of treatment; treatment response; date of last follow-up; live status; and cause of death.

Two hundred forty two patients of 281 in the current series had received one or more treatments with surgical resection (19, 6.8%), radiofrequency ablation (35, 12.5%), percutaneous ethanol injection (3, 1.1%), TACE (232, 82.6%), RT (0.7%), chemotherapy (0.4%) before they were diagnosed as HCC with PVTT.

Before simulation, shallow-breathing during the procedure was asked and all patients got education and training. Then, respirespiratory motion of the liver was checked by fluoroscopy, and these data were used for determination of the margin. After that, all patients underwent contrast enhanced CT scans in a supine position for RT planning, with both arms raised above the head to facilitate use of lateral RT portals. CT scan data were transferred to a 3D-CRT treatment planning system (from 1998 to 2003, PROWESS, Alliant Medical Technology, Chico, CA, USA; from 2004 to 2008, PINNACLE, The Philips Medical System, Best, Netherland).

The PVTT, main mass, the normal liver, kidneys, duodenum and spinal cord were contoured and reconstructed to form a 3D representation. The RT beam angles were designed to minimize critical organ injury. A dose-volume histogram (DVH) was also generated.

The gross tumor volumes (GTV) were defined as the radiographically abnormal areas (main mass and PVTT, usually) noted on the CT images refer to the diagnostic CT and/or MRI. But, in several cases (large tumor [more than 2/3 of the liver] with severe liver cirrhosis, Child-Pugh class C, huge bilateral intrahepatic tumor with main PVTT, numerous intrahepatic metastasis, etc), only PVTT was delineated as GTV. The clinical target volumes (CTV) were regarded as same as GTV. In designing the PTV, the margins were individualized by observing liver position as well as liver movement at the time of simulation. In determining the cranial-caudal margins, the distance of diaphragmatic excursion by respiration, which was observed fluoroscopically, was added to the cranial-caudal margins (minimum 1.5 cm + motion). In the same way, anterio-posterior, and lateral margins (minimum 1.0 cm + motion) were decided.

A median daily dose of 3 gray (Gy, range 1.8 to 4.5 Gy) was administered using 6, 10 or 15 MV X-rays, at 5 fractions per week, to deliver a total dose of 30-54 Gy, which translates to a biologic effective dose (BED) of 39-70.2 Gy₁₀ as the α/β = 10. BED was calculated using a linear quadratic model to be equivalent to 3 Gy fraction treatments in respect of acute effects on normal tissues and tumors. 3D-CRT planning was designed under tentative guidelines so that the normal liver volume irradiated with more than one-half of the prescription dose should not be 50% of the total liver volume.

Four-dimensional gated CT with RPM signal and gated verification in the treatment room were conducted after March 2008, and 44 patients were enrolled in this study.

Additional treatment, mostly TACE, was planned and administered after the RT without considering of RT response after 2004. In 202 patients who received RT after 2004, TACE in 122 patients (60.4%), re-irradiation in 7 patients, radiofrequency ablation in 4 patients, chemotherapy (sorafenib) in 4 patients, and hepatectomy in 2 patients was conducted. By contrast, before 2004, TACE was considered mainly in patients who showed good response to RT. In 79 patients, 23 patients (29.1%) received TACE.

PVTT and/or tumor response (according to RT target) was assessed using World Health Organization criteria by serial CT scans, 4-8 weeks after completion of treatment and then every 2-3 months. Complete disappearance of PVTT and tumor was defined as a complete response (CR), a more than 50% reduction of thrombus in the greatest cross-sectional area and/or perfusion normalization of the area involved by tumor thrombosis was defined as a partial response (PR), a less than 50% reduction of thrombus and tumor (including no change) was defined as stable disease (SD), and tumor growth was defined as progressive disease (PD). Responders were patients with CR or PR, whereas non-responders were patients with SD or PD.

Acute morbidity was evaluated weekly during treatment and 1 month after the treatment. Late morbidity was defined as occurring after 3 months. RT induced liver disease (RILD) was defined as anicteric ascites and elevation of alkaline phosphatase (ALP) levels > 2 times the pre-treatment values in the absence of PD.

If patients complained of melena or upper abdominal pain persisting for longer than 2 weeks during the follow-up period, fiberoptic gastroduodenoscopy was conducted for clarifying the gastrointestinal (GI) toxicity. Toxicity was scored using the Common Terminology Criteria for Adverse Events (CTCAE; version 3.0) at every visit.

Overall survival (OS) and PVTT progression free survival (PVTT PFS) were estimated using Kaplan-Meier product-limit method. OS was measured from the RT start date to the date of death or the last follow-up visit, and PVTT PFS from the RT start date to the date of PVTT progression or the last follow-up visit.

Univariate logistic regression analysis was used to evaluate the association between OS and various parameters. For multivariate analysis to evaluate the relation between the OS and various parameters, the stepwise procedure was performed using a logistic regression model containing all variables that attained or had a trend toward statistical significance on univariate analysis. P < 0.05 was considered statistically significant. All calculations were performed with PASW 17.0 for Windows (SPSS, Chicago, IL, USA).

This study was approved of permission by the institutional review board of Samsung Medical Center, Sungkyunkwan University (IRB No. 2011-03-033). Informed consent was waived by the board.

Table 1 describes the tumor characteristics of 281 patients. The median age of patients was 54 yr (range, 26-80 yr). Among the patients, 249 (88.6%) were male, and the male-to-female ratio was 7.8:1. The median size of tumor plus PVTT was 7 cm (range, 2-18 cm). Serum hepatitis B virus antigen markers were positive in 258 patients (91.8%) and hepatitis C virus antigen in 13 patients (4.6%). Specific sites of PVTT were as follows: left hemiliver portal vein (n = 50, 17.8%), right hemiliver portal vein (n = 100, 35.6%), bilateral hemiliver branch portal vein (n = 4, 1.4%), main portal vein (n = 114, 40.6%), extended to superior mesenteric vein (n = 13, 4.6%).

Most patients had Child-Pugh classification A (84.0%) or B (14.0%), and only 2.0% had Child-Pugh class C. Thirty two patients (11.4%) presented poor performance status (ECOG 2). Eighty-four patients (29.9%) who showed one or more of these characteristics (large tumor [more than 2/3 of the liver] with severe liver cirrhosis, Child-Pugh class C, huge bilateral intrahepatic tumor with main PVTT, numerous intrahepatic metastasis, etc.) received RT only on PVTT.

Four to eight weeks after the completion of the RT, an objective response was observed in 151 patients (53.8%). Of these patients, CR was observed in 10 patients (3.6%), PR in 141 patients (50.2%), and SD in 72 patients (25.6%), and PD in 37 patients (13.2%). There were 21 unevaluable patients (7.5%), who were lost to follow-up. AFP decrement was observed in 163 patients (58.0%) at 1 month follow up after RT.

Table 2 shows the complication profile. During the RT, the most common adverse effects were anorexia and nausea (152 patients, 54.1%), but only 1 patient showed grade III nausea, the others were grade I or II. Elevation of liver enzymes was also common, but generally it was less than a five-fold of the upper normal limit. And these elevations were self-limited in most cases. After the RT, 260 patients (92.5%) were evaluable for toxicities, except 21 patients who were lost to follow up. In 15 patients (5.3%), grade II or III clinical hepatic dysfunction was observed after the RT. Among them, PD was observed in 9 of 15 patients (3.2%), and there was no classic RILD. But, in 2 patients (0.7%), transaminases (aspatate aminotransferase [AST], alanine aminotransferase [ALT]) were elevated more than five times the upper limit of the normal or of pre-treatment level. For gastrointestinal complications, 8 patients (2.8%) were scored as grade II and 1 patient (0.4%) as grade III resulting from gastric or duodenal ulcers inside the RT field.

The follow-up period ranged from 0 to 103.2 (median 8.0, mean 11.1) months and 111 of 281 (39.5%) patients were still alive at the time of last follow-up. The median survival was 11.6 (range 1.0 to 103.2) months and the actuarial 6-month, 12-month, 18-month, and 24-month overall survival rates were 65.2%, 48.1%, 35.1%, and 26.9%, respectively (Fig. 1A). In the cases of PVTT, the median PVTT PFS was 11.2 (range 0.2 to 103.2) months (Fig. 1B).

In responders, the median survival duration was 22.0 months and the 6-month, 12-month, 18-month, and 24-month overall survival rates were 85.3%, 70.0%, 53.9% and 41.4%, respectively. In non-responders, the median survival duration was 5.0 months and the 6-month, 12-month, 18-month, and 24-month overall survival rates were 34.3%, 13.8%, 2.9%, and 0.0%, respectively (Fig. 2A). The responders had a significantly higher overall survival rate than the non-responders (P < 0.001).

The results of univariate and multivariate analyses to evaluate the relation between OS and various parameters are summarized in Table 3. Of the pre-treatment parameters, significant prognostic factors for OS in univariate analysis were performance status (ECOG), Child-Pugh class, tumor size, multiple tumor, main PVTT involvement, bilateral involvement of PVTT, size of the PVTT, completeness of obstruction, lymph node metastasis, distant metastasis, pre-treatment AST/ALT/ALP/PIVKA-II level, and so on, while age, sex, hemoglobin, and pre-treatment AFP level (P = 0.053) were not. On the multivariate analysis, performance status (ECOG, P = 0.001), Child-Pugh class (P = 0.002), tumor size (P = 0.029), multiple tumors (more than 2, P = 0.016), main PVTT involvement (P = 0.022), completeness of obstruction (P = 0.001), and lymph node metastasis (P = 0.001) showed statistically significant effects on OS.

Most significant prognostic factor after treatment was radiation response as discussed above. AFP decrement after RT was also important post-treatment prognostic factor. The patients who achieved AFP decrement (163 patients, 58.0%) showed superior median survival duration than the others (14.7 months vs 5.5 months, P = 0.009). And the other significant prognostic factors in post-treatment parameters were decrement of the PIVKA-II level (P < 0.001) and total dose of the radiation delivered (median survival duration; less than 40 Gy₁₀ 4.2 months, 40 Gy₁₀ to 50 Gy₁₀ 10.4 months, more than 50 Gy₁₀ 13.8 months, P = 0.031) in univariate analysis. Additional treatments before and/or after RT showed a statistically significant effect on OS. But, those differences might be related with selection bias of the patients who live longer could receive additional treatments. It is biased on that additional treatments showed no differences to OS in multivariate analysis and closely related with RT response (P < 0.001).

Based on clinical relevance and availability of the information, we used seven pre-treatment factors that showed statistical significance in both univariate and multivariate analysis to construct a Predictive Index for PVTT of the HCC (PITH) and calculate a PITH score. The included factors were grade II or more of performance status (ECOG), Child-Pugh classification B or C, 10 cm or more of tumor size, multiplicity of the tumor, main PVTT involvement, complete occlusion of portal flow and lymph node metastasis. PITH score is defined as the number of the above factors. We estimated OS from the PITH score, and statistically significant differences were detected in each group (P < 0.001, Fig. 3A). We divided the patients into low, low intermediate, high intermediate and high risk groups according to the PITH score. The low risk group with score 0 or 1 consisted of 54 patients (19.2%) and had 82.5% OS at one year, 27.2 months median survival. The low intermediate risk group with score 2 or 3 consisted of 133 patients (47.3%) and had 45.9% OS at one year, 11.4 months median survival. The high intermediate risk group with score 4 or 5 consisted of 76 patients (27.0%) and high risk group who had 6 or more score are made up of 18 patients (6.4%), and 33.0%, 0.0% OS at one year, and 6.4 months, 3.3 months median survival, respectively (P < 0.001, Fig. 3B). And also, when divided them into two groups by the score 0 to 2 and 3 or more (low risk group 110 patients, high risk group 171 patients), OS showed difference (70.1% vs 33.0% at one year, P < 0.001, Fig. 3C).

The prognostic ability of the PITH score was compared with Barcelona Clinic Liver Cancer (BCLC), Cancer of the Liver Italian Program (CLIP), Okuda, Japan Integrated Staging (JIS), and Chinese University Prognostic Index (CUPI) staging systems. For each score, this performance was evaluated by comparing by log rank test the survival curves of the single categories, calculated using the Kaplan-Meier method. In Table 4, median values and interquartile ranges of the OS for the six prognostic systems are reported. In spite of there was a significant correlation between OS and stage assigned using all six systems (BCLC P = 0.001, CLIP P = 0.007, Okuda P < 0.001; JIS P < 0.001; CUPI P < 0.001), and trend in), the PITH score performs better than the other staging systems, allowing a better characterization of the extremely good and intermediate categories of patients.