Preoperative Predictors of Survival After Resection of Small Hepatocellular Carcinomas

Abstract

Objective

To determine preoperative predictors of survival that can guide the choice of treatment for patients with small hepatocellular cancers (HCCs).

Summary Background Data

The treatment of patients with small (≤5 cm in diameter) HCCs is controversial.

Methods

A cohort of 249 patients (69 women, 180 men; median age 62 years) who underwent resection with curative intent for small HCC was identified from a multiinstitutional database. For each patient, the clinical data and pathology slides were reviewed. Six clinical factors (age, gender, preoperative α-fetoprotein level, hepatitis serology, number of tumors [single vs. multiple], and Child-Pugh score) and three pathologic factors (hepatitis activity score, fibrosis score, and Edmondson-Steiner tumor grade) that can be determined before surgery were correlated with survival. Log-rank tests and Cox proportional hazards modeling were used to determine factors influencing survival.

Results

The median overall survival for the entire cohort was 4.2 years. The estimated overall 5- and 8-year survival rates were 41.1% and 19.8%, respectively. Multivariate Cox analysis indicated that fibrosis score, Edmondson-Steiner grade, and Child-Pugh score were simultaneously significant predictors of survival after resection. A prognostic scoring system based on these covariates was derived and applied to the entire cohort. Patients lacking all three risk factors were assigned a score of 1, patients with one risk factor were assigned a score of 2, and patients with two or three risk factors were assigned a score of 3. Pairwise log-rank tests indicated significant differences in survival between scores 1 and 2, scores 2 and 3, and scores 1 and 3. This scoring system retained its prognostic significance when a subset of 98 patients with positive hepatitis C serology was analyzed separately.

Conclusions

Patients with small HCCs who will derive the least benefit from resection can be identified before surgery using a score based on tumor grade and the severity of underlying liver disease. In these patients, transplantation and/or ablation should be considered as possible alternative therapies.

Although the incidence and death rate of most cancers have stabilized or decreased during the past two decades, the incidence of hepatocellular carcinoma (HCC) has steadily increased. ^1–3 In most instances, HCC arises in patients with chronic underlying liver disease secondary to hepatitis B or C infection. ^4,5 Increasingly, patients with chronic liver disease undergo laboratory and imaging studies that lead to the early detection of a small HCC (≤5 cm in diameter).

Resection, ablation, and orthotopic liver transplantation (OLT) have all been proposed for the treatment of small HCCs, but the subsets of patients who benefit most from each modality remain controversial. ^6–10 In the past, OLT was performed in patients with HCC who were not candidates for resection because of tumor size and inadequate hepatic reserve. This approach led to high rates of early tumor recurrence. However, it was incidentally noted that certain subsets of patients with small tumors benefited from OLT; in these patients, survival rates with OLT were close to those in patients who underwent OLT for nonmalignant disease. ^11,12

Advances in intraoperative techniques and postoperative care have led to a marked decline in perioperative rates of death and complications after liver resection for HCC. Thus, at tertiary hepatobiliary centers, resection has remained the treatment of choice for HCC even in patients with chronic liver disease. ^13–15 This favorable experience, as well as a shortage of transplantable organs, ¹⁶ has led to an increasingly aggressive approach to the resection of small HCCs. A large single-institution study recently reported a 5-year survival rate of 57% for a subset of patients with tumors less than 5 cm in diameter who underwent curative resection. ¹⁷ Other authors have reported similar results in patients with small HCCs treated with partial hepatectomy. ^10,18,19

To determine which patients benefit most from resection, we reviewed a multiinstitutional database to determine the factors that predicted survival after hepatic resection for small HCCs. We examined the clinical, pathologic, and survival data of 249 patients who underwent resection for HCC 5 cm or smaller and analyzed the clinical and pathologic factors that could be determined before surgery to determine hepatic and tumor variables predictive of survival for this subset of patients.

METHODS

Patients included in this study were identified using a multiinstitutional database that contained records of 591 patients who underwent curative resection for HCC between 1980 and 1998 at one of four major hepatobiliary centers: the University of Texas M. D. Anderson Cancer Center (Houston, TX), Mayo Clinic (Rochester, MN), Hôpital Beaujon (Paris, France), and Kyoto University Graduate School of Medicine (Kyoto, Japan). Among the 591 patients, 303 patients were identified who underwent complete resection for single or multiple tumors, all 5 cm or less in greatest diameter. All 303 patients were considered in the development of our prognostic scoring system; however, only the 249 patients with complete data on the clinicopathologic predictors that were found to be significantly associated with survival were retained as the study population.

The clinical data were reviewed on site by the principal investigators (D.M.N., J.N.V.). The clinical factors investigated were age, gender, preoperative α-fetoprotein level, hepatitis serology, number of tumors (single vs. multiple), and hepatic reserve, which was rated using the Child-Pugh score. ²⁰ The Child-Pugh score was assigned to all patients regardless of fibrosis severity as used in previous studies to provide uniform assessment of hepatic reserve in patients with HCC and possible underlying liver disease. ²¹ The serologic presence of any hepatitis B antigen or antibody was considered positive evidence of hepatitis B exposure, because both increase the risk of HCC. ²² The serologic presence of hepatitis C antibody was considered positive evidence of hepatitis C exposure. All pathology slides were reviewed by one hepatobiliary pathologist (G.Y.L.). The pathologic variables investigated were hepatitis activity score, fibrosis score, microscopic vascular invasion, and the grade of the primary tumor as defined by Edmondson and Steiner. ²³ Well-differentiated and moderately differentiated tumors were defined as grade 1 and 2; poorly differentiated and undifferentiated tumors were defined as grade 3 and 4. Hepatitis activity and fibrosis stage were scored separately according to the classification of Ishak et al. ²⁴ Necrosis and inflammatory changes characteristic of hepatitis were scored as mild (0–5), moderate (6–12), or severe (13–18). Fibrosis was scored as no or minimal fibrosis (0–2), moderate fibrosis (3 or 4), or severe fibrosis and cirrhosis (5 or 6). Vascular invasion was defined as the presence of tumor emboli within the central veins or portal or capsular vessels.

The primary endpoint of the analysis was survival time after liver resection. Survival probabilities were constructed using Kaplan-Meier survival estimates, ²⁵ and the actuarial survival estimates were also obtained to permit comparisons between our study and other studies in which actuarial estimates were reported. Log-rank tests and Cox proportional hazards modeling were used to determine factors influencing survival. ²⁶ Martingale residuals were used to evaluate the proportional hazards assumption in the Cox regression models. Once the simultaneously significant predictors of survival were identified with Cox regression model, ²⁷ a prognostic scoring system was constructed that distinguished between patient groups on the basis of survival duration after resection. Because the study included patients who underwent resection before hepatitis C serology became available, subset analyses of patients with known hepatitis C serology and positive hepatitis C serology were also performed. The significance level of P < .05 was used in all hypothesis tests. Data were analyzed using S-plus statistical software. ²⁸

RESULTS

The study cohort included 69 women and 180 men (72%). The median age was 62 years (range 28–84). Patient clinical and pathologic characteristics are listed in Table 1. Sixty-six patients (27%) underwent a major resection, defined as resection of a lobe or more, and 172 patients (73%) underwent a minor resection, defined as resection of less than a lobe, including segmental and nonanatomic resections. Fifteen patients (6.1%) died in the perioperative period, nine with Child-Pugh A, five with Child-Pugh B, and one with Child-Pugh C.

Table 1. CLINICOPATHOLOGIC FEATURES

* Because data on certain clinicopathologic characteristics were not available for some patients, totals in some categories are less than 249.

† Performed after 1991.

‡ Includes 67 patients with fibrosis score 0–4.

The median follow-up was 3.6 years (range 0.01–11.9). The median overall survival was 4.2 years (95% confidence interval [CI], 3.7–4.7). The estimated Kaplan-Meier overall 5- and 8-year survival rates were 41.1% (95% CI, 35.1–48.1%) and 19.8% (95% CI, 14.5–27.0%), respectively (Fig. 1). There were no significant survival differences related to age, gender, type of resection, tumor-free margin, hepatitis activity score, or α-fetoprotein levels.

Figure 1. Overall survival.

Positive hepatitis B serology was not associated with survival, but patients with positive hepatitis C serology had marginally longer survival than patients with negative hepatitis C serology (5.1 vs. 4.0 years, P = .055). Due to marginally better survival in the patient subset with positive hepatitis C serology, the prognostic scoring system was also investigated in 145 (59%) of the 249 patients who had hepatitis C serology performed.

With the exception of microscopic vascular invasion (that cannot be reliably determined by preoperative needle biopsy and therefore not evaluated), all clinicopathologic characteristics were investigated in the multivariate Cox analysis for their associations with survival. Among the factors considered, fibrosis score, Edmondson-Steiner grade, and Child-Pugh score were found to be simultaneously significant predictors of survival after resection. The covariate-adjusted relative risk of death associated with severe fibrosis and cirrhosis (score 5 or 6) compared with mild or moderate fibrosis (score 0–4) was 1.51 (95% CI, 1.05–2.18;P = .013). Compared with patients with Edmondson-Steiner grades 1 or 2, patients with Edmondson-Steiner grades 3 or 4 had a relative risk of death of 1.42 (95% CI, 1.05–1.92;P = .022). Compared with patients with Child-Pugh A, patients with Child-Pugh B or C had a relative risk of death of 1.56 (95% CI, 1.10–2.22;P = .013). The evaluation of fibrosis provided additional important prognostic information in patients who were not Child-Pugh B or C. Among these patients with preserved liverfunction, one third presented with a fibrosis score of 0 to 4 (Tables 2 and 3), and this low fibrosis score was associated with a better prognosis (Fig. 2).

Table 2. DISTRIBUTION OF ALL PATIENTS ACCORDING TO THE PROGNOSTIC SCORING SYSTEM

Table 3. DISTRIBUTION OF PATIENTS WITH POSITIVE HEPATITIS C SEROLOGY ACCORDING TO THE PROGNOSTIC SCORING SYSTEM

Figure 2. Survival stratified by (A) Child-Pugh score, (B) Edmondson-Steiner grade, and (C) fibrosis score.

A prognostic scoring system was derived on the basis of these simultaneously significant predictors of survival. All possible combinations of these predictors (i.e., fibrosis grade 0–4 and 5 or 6, Child-Pugh A and B or C, and Edmondson-Steiner grades 1 plus 2 and 3 plus 4) were considered as predictors of survival, and the differences in survival between patient groups presenting with different predictors were examined using log-rank tests. Patients who presented with a fibrosis score 0 to 4, Child-Pugh A, and Edmondson-Steiner grade 1 or 2 (median survival after resection, 7.5 years) had the best prognosis. Patients with one risk factor (e.g., fibrosis grade 5 or 6, Child-Pugh B or C, or Edmondson-Steiner grade 3 or 4) had an intermediate prognosis (median survival, 4.5 years). Patients who presented with two or three risk factors had the least favorable prognosis (median survival, 2.9 years).

Patients lacking all three risk factors were assigned a score of 1, patients with one risk factor were assigned a score of 2, and patients with two or three risk factors were assigned a score of 3 (Table 4). The differences in survival between these three patient groups were statistically significant (log-rank, P < .001) (Fig. 3), and pairwise log-rank tests indicated significant differences in survival between scores 1 and 2 (P = .029), scores 2 and 3 (P < .001), and scores 1 and 3 (P = .007). This scoring system retained its prognostic significance in a subset analysis of 145 patients with known hepatitis C serology (log-rank, P < .001) and in a subset analysis of 98 patients with positive hepatitis C serology (log-rank, P = .020) (Fig. 4). In the 145 patients with known hepatitis C serology, median survival times for patients assigned scores 1, 2, and 3 were 7.5, 5.1, and 2.9 years, respectively. In the 98 patients with positive hepatitis C serology, the median survival times for scores 1, 2, and 3 were 7.5, 6.0, and 2.9 years, respectively.

Table 4. PREOPERATIVE PROGNOSTIC SCORING SYSTEM

Figure 3. Survival stratified by prognostic score.

Figure 4. Survival stratified by prognostic score in (A) 145 patients with known hepatitis C serology and (B) 98 patients with positive hepatitis C serology.

DISCUSSION

In the current study, a large cohort of patients who underwent hepatic resection for small HCCs was identified from a multiinstitutional database. In multivariate analysis, two liver factors (Child-Pugh and fibrosis score) and one tumor factor (Edmondson-Steiner grade) that can be determined before surgery were found to be simultaneously preoperative predictors of survival after resection. The importance of chronic underlying liver disease in the prognosis of patients undergoing resection for small HCCs is underscored by the statistical weight of the Child-Pugh score and fibrosis score in our analysis.

The current series confirms the worse survival associated with resection of small HCC in Child-Pugh B or C patients indicated in previous series. ^15,19 In a recent study, the Clinical Stage of liver disease (a Japanese classification similar to Child-Pugh score) predicted survival in 12,888 patients with small HCC undergoing resection, percutaneous alcohol injection, or hepatic artery embolization. ²¹ This finding not only confirms the value of a preoperative clinical assessment of liver function but also indicates equivalent prognostic value after nonsurgical therapies in patients with small HCCs.

An important finding of our study is the worse prognosis associated with severe fibrosis or cirrhosis. It is interesting to note that although patients with Child-Pugh score B or C presented with severe fibrosis (score 5 or 6), the evaluation of fibrosis provided important prognostic information in patients with preserved liver function. In this study, one third presented with a fibrosis score of 0 to 4, and this low fibrosis score was associated with a better prognosis. In a recent study, Bilimoria et al ²⁹ showed that patients with normal underlying liver or minimal fibrosis had a risk of death from HCC of 7% beyond 5 years after resection, compared with 58% for patients with fibrosis. The same study suggested that fibrosis of the underlying liver could be used as a marker of a field of cancerization associated with the heightened risk of metachronous HCC.

In 1954, Edmondson and Steiner ²³ described a grading system based on nuclear pleomorphism in 100 patients with HCC. Although grade is controversial as a prognostic factor after resection of HCCs unselected for size, ^30,31 poorly differentiated and undifferentiated tumor grades were recently associated with a sixfold increase in the risk of vascular invasion in small HCCs. ³² In our multivariate models of prognostic factors for survival after resection of HCC, vascular invasion and grade appear as complementary factors. ³³ It is therefore not surprising that grade remained the single significant prognostic tumor factor after vascular invasion was eliminated from the current regression model. Vascular invasion is a well-known prognostic factor after resection in patients with small HCCs ³⁴ and HCCs unselected for size, ^33,35 and after OLT. ^36,37 Unfortunately, vascular invasion is often difficult to determine before surgery unless gross vascular invasion is present. Chapel et al ³⁸ previously showed that Edmondson-Steiner grade as assessed by percutaneous biopsy was strongly correlated with survival by univariate and multivariate analysis in patients with unresectable HCC. Our findings, taken together with the other studies described here, indicate that high-grade tumors could theoretically be used as a surrogate for microscopic vascular invasion in predicting prognosis in patients with small HCCs.

In selected patients with small HCCs, survival after OLT approximates that of patients with small HCCs who undergo OLT for benign disease (Table 5). Mazzaferro et al ¹² reported a 4-year actuarial survival rate after OLT of 75% in patients with up to three tumors all 3 cm or less in diameter or single tumor 5 cm or less in diameter without microscopic vascular invasion. Similarly, Bismuth et al ³⁹ reported 3- and 5-year survival rates of 76% and 74% in 45 patients using selection criteria including up to three tumors 3 cm or less without involvement of the portal vein. Llovet et al ¹⁰ reported a 69% 5-year survival rate after OLT when patients were analyzed on an “intent-to-treat” basis.

Table 5. SURVIVAL RATES

Reference numbers are given in parentheses.

* Includes patients with and without cirrhosis.

† Includes only patients with cirrhosis.

‡ Perioperative deaths excluded.

§ Criteria for transplantation < 3 tumors, with the largest ≤ 3 cm in size without portal vein invasion.

¶ Includes 2 patients with > 5-cm tumors.

# “Intent-to-treat” analysis of transplanted patients.

Although data regarding the prognosis and selection of patients with HCC after OLT exist, data on preoperative indicators that could be used to identify patients who would most benefit from resection are scarce. A recent Western study ¹⁷ reported a 57% 5-year survival rate in 38 patients with HCCs smaller than 5 cm who underwent resection. Another study ¹⁰ reported a 74% 5-year survival rate in 35 patients who underwent resection for small HCCs and were selected for this treatment modality on the basis of the absence of clinical portal hypertension and a bilirubin level less than 1 mg%. In the current study, a score of 1 was associated with a 68% 5-year actuarial survival rate in 45 patients, whereas a score of 3 was associated with a 28% 5-year actuarial survival rate.

Recent studies have attempted to correlate the prognosis of patients with HCC with hepatitis serology status. Correlation between hepatitis serology and outcome has been inconsistent. ^40–45 Yamanaka et al ⁴² reported that patients with hepatitis B virus-associated HCC had improved 5-year survival rates compared with patients with hepatitis C virus-associated HCC (54% vs. 42%). Haratake et al ⁴⁰ found the opposite to be true. In their study, patients with hepatitis C-associated HCC had improved survival rates compared with patients with hepatitis B-associated HCC. In the current study, the univariate analysis suggested possibly improved survival in the small subset of patients with positive hepatitis C serology. Liver disease and tumor factors in our prognostic scoring system were distributed similarly in patients with positive hepatitis C serology (see Table 3) and negative hepatitis C serology (data not shown). Our finding that the long-term prognosis after resection of small HCC is influenced by factors relating to tumor and liver disease stage rather than hepatitis serology is consistent with the findings of other investigations on prognosis after resection of HCC regardless of size. 41,43,44,46 These data also suggest that serologic evidence of hepatitis B or C infection should not be used to exclude patients from consideration for resection.

Patients presenting with Child-Pugh B or C have a score of 3 as a result of coexisting severe fibrosis or cirrhosis (fibrosis score 5 or 6) (see Tables 2 and 3). These patients are possible candidates for ablation or OLT as appropriate. Determining tumor grade before considering the patient for OLT may be appropriate given the strong association between vascular invasion and grade and the possibility of an increased risk of recurrence after OLT. ^32,37,47 Patients with a fibrosis score of 5 or 6 and an Edmondson-Steiner grade 3 or 4 (score 3) likewise are candidates for nonresectional therapy. Although reporting on patient subsets that are somewhat different from patients undergoing resection, several series indicate that long-term survival is possible after percutaneous ethanol ablation of small HCCs (see Table 5). ^48–51

Patients with a score of 2 as a result of a fibrosis score of 5 or 6 with a low Edmondson-Steiner grade (1 or 2) have an intermediate prognosis after resection, and decisions regarding therapy should be individualized. These patients are possible candidates for OLT because of the long-term risk of recurrence associated with underlying liver disease. ²⁹ Unfortunately, because of the shortage of donors and limitation in resources, only a few can actually receive transplants. Resection therefore remains the only option that can be considered for most patients. In this situation, the decision regarding resection should be based on tumor location and the predicted perioperative risk of complications. These patients have apparently preserved liver function based on their Child-Pugh score A, but they remain at risk for liver failure after resection. Previous algorithms and recent guidelines have been suggested for patients who require major hepatectomy in the setting of severe fibrosis or cirrhosis. ^34,52 In the absence of severe liver disease (i.e., when the fibrosis score is 0–4), patients with either high- or low-grade tumors (scores 1 and 2) should be prioritized for resection, with recognition of the worse prognosis expected for high-grade tumors.

A potential criticism of this scoring system is that it requires a tumor biopsy for determination of grade, which may be associated with a risk of needle tract seeding. In a recent series, the risk of needle tract implantation was reported to be 1.6% in 134 patients who underwent tumor biopsies before resection or OLT for HCC. ⁵³ This risk needs to be balanced against the information gained regarding diagnosis (HCC vs. regenerating nodule) and tumor biology (grade), and the 10% 1-year death rate associated with OLT. Currently, many specialized centers already perform biopsies for tumors smaller than 2 to 3 cm in patients with cirrhosis before ablation or OLT because the differentiation between HCC and regenerating nodules is not possible in more than 30% of patients. ^50,51

Although the application of each modality should remain individualized and the choice of therapy may be obvious for many patients, we believe the proposed scoring system provides additional information about the best choice of therapy in patients with small HCCs. The prognostic scoring also affords background data that will assist in the design and stratification of future studies investigating the treatment of small HCCs.

Discussion

Dr. Frederick L. Greene (Charlotte, NC): As Dr. Pollock knows, the American Joint Committee on Cancer (AJCC) is evaluating the TNM staging system to see what changes should occur in the staging of solid tumors. I would like to ask whether these reported prognostic factors are robust enough to be included in current TNM staging or whether other factors that you have discussed should affect the TNM staging strategy of early tumors of the liver.

Dr. Alan W. Hemming (Gainesville, FL): Presumably some of the patients in this study would have met the criteria for transplantation. Although I realize this is a retrospective study, could you comment on how resection was chosen over transplantation, especially in the 49 patients that were either Child’s B or C grade?

Second, with tumors 2 cm in size or smaller, the Japanese have demonstrated equivalent results with either local ablation or resection. We have been resecting such lesions. Could you comment on your current approach?

Lastly, one of the determinants of how well patients will do after resection is the status of their underlying liver disease. Child’s status is a relatively blunt instrument, and fibrosis doesn’t necessarily correspond with function. Do you use any further assessment of liver function, such as indocyanine green clearance, preoperatively?

Dr. John S. Bolton (New Orleans, LA): A couple of questions.

We have only had hepatitis C serology available for 9 years. And if I understand your presentation correctly, hepatitis C was not an adverse prognostic factor. But this kind of is at odds with other reports that suggest a relapse rate of 3% to 5% from new hepatocellular carcinoma formation in the remnant liver over time, 3% to 5% per year. And I just would like you to comment, if you would: do you think that the long-term risk of carcinogenesis from hepatitis C has been overestimated? It didn’t seem to come through in your report.

And the second question. Almost half of your patients were in the prognostic score 3 group. Most of those patients were Child-Pugh A and were in that prognostic score 3 by virtue of a high fibrosis score and high grade. Still, if I understood you right, you had a 29% 5-year survival in that group. Would you not resect that group? Or in that group with a prognostic score of 3, what would be your indications for resection?

Dr. William C. Wood (Atlanta, GA): The question that always comes when you go through a large series, a spectacular series like that, and come up with cut points, is, would they be validated in a prospective series? You had the opportunity of applying them not only to the whole group but to the individual institutions. And were these consistent among the institutions when applied in that way?

Dr. Michael A. Choti (Baltimore, MD): In patients who then it is obvious that they have hepatocellular cancer, are you recommending that every patient undergo a preoperative needle biopsy of the tumor and preoperative needle biopsy of the surrounding liver in order to determine both the fibrosis and histologic grade? There are certainly associated risks involved with that. And in a patient who does not have evident cirrhosis, when is it necessary? When would you recommend that biopsies be done on both tumor and liver in the preoperative setting?

Dr. Ralph R. Bollinger (Durham, NC): The authors repeatedly pointed out that transplantation is a better alternative if you only had the liver available. The fact is, many of these patients have a willing, living donor, some family member who would like to give them part of their liver. Should living-related donor transplantation be used for patients with small hepatocellular carcinomas?

Dr. J. Nicolas Vauthey (Houston, TX): I would like to thank the discussants for their insightful questions.

To answer Dr. Greene’s first question regarding the new TNM staging. As you know, we have now established a new TNM staging based on this database, and it will appear in the sixth edition of the Cancer Staging Manual in April or May. The main driving force, if you will, in the prognosis of hepatocellular carcinoma is vascular invasion. So, Dr. Greene, we could not use microscopic vascular invasion in the scoring. Since the scoring is devised as a preoperative tool, microscopic vascular invasion could not be used. However, we found out that high grade in this analysis of data could be used as a surrogate marker of microscopic vascular invasion.

Dr. Hemming, how was resection chosen? These resections were undertaken at major hepatobiliary centers with a large experience with resection in hepatocellular carcinoma. Fifty-five percent of the patients were from Japan and 45% were from the West. In Japan, ICG clearance was used; in the West it was not. I think ICG clearance can add to your preoperative evaluation in patients with Child-Pugh A class. The way the Japanese look at ICG clearance, they look at it to determine the extent of resection in their patients who have normal or near-normal liver function tests. If the ICG clearance is less than 10% at 15 minutes, they consider major resection. At our institution, we are very cautious about major resection in cirrhosis, and we do volumetric measurement and look at the amount that remains rather than the amount resected.

Dr. Bolton mentioned the increased relapse rate in patients with hepatitis C. This answer will be also the answer I will give to Dr. Wood. It is clear that the risk of relapsing has been higher in longitudinal studies of patients with hepatitis C and advanced chronic liver disease. In the current study, the patients with hepatitis C were mainly from Japan and had a broad spectrum of liver disease background. They had significantly less microscopic vascular invasion. We included center to look for a center effect in our multivariate analysis. Japanese patients had a slightly better survival, but the three elements of the prognosis score remained after controlling for center. Our data suggest that tumor and liver factors determine prognosis rather than hepatitis serology.

This is the algorithm, and this is how we approach it now at our institution [slide]. Most patients with small HCC have chronic liver disease. The first thing we do in clinic, we evaluate the patient’s Child-Pugh class. Patients with Child-Pugh B class are immediately in a score 3 because, based on our analysis, all of them had severe fibrosis or cirrhosis. We recommend transplantation or ablation. Score 3 with Child-Pugh A are also candidates for ablation or transplantation. Score 2 patients without severe fibrosis or cirrhosis are prioritized for resection.

Dr. Choti, as you know, the issue of biopsy of hepatocellular carcinoma is controversial. The risk of needle tract implantation after percutaneous biopsy was recently reported to be 1.6% in 134 patients resected or transplanted at a specialized center (J Hepatol 2001; 35:254). This risk has to be balanced against the information gained regarding diagnosis and prognosis and the 10% 1-year mortality associated with transplantation. Patients with tumors up to 3 cm already get biopsied at many centers for diagnosis. The scoring proposed does not require routine tumor biopsy.