STUDY DESIGN

Between March 1989 and November 1991, a cohort of patients with LC and without HCC, hospitalised at the Department of Internal Medicine, S Orsola-Malpighi University Hospital, or referred as outpatients to the liver unit of the department were enrolled in the study. The diagnosis of LC was histologically proved or made on unequivocal clinical, biochemical, ultrasonographic, and endoscopic findings. Exclusion criteria were: (1) Child-Pugh C class16 in patients older than 60 years; (2) a previous diagnosis of focal liver lesion at US; and (3) a serum AFP level >200 ng/dl. Patients were withdrawn from further surveillance when they were >60 years old and belonged to Child-Pugh C class, developed other neoplasms, or underwent orthotopic liver transplantation (OLT).

The study protocol conformed to the ethical guidelines of the 1975 declaration of Helsinki. All patients gave informed consent to enter the prospective surveillance programme in conformity with the modalities reported below. The initial evaluation included clinical examination, routine biochemical tests, determination of serum AFP, upper gastrointestinal tract endoscopy, and abdominal US.

During the subsequent surveillance period, serum AFP determinations and abdominal US, together with physical examination and routine biochemical tests, were repeated every six months.

DIAGNOSIS AND STAGING STRATEGY

The diagnostic protocol for detection of a nodular liver lesion at US was based on contrast enhanced computed tomography (CT) and echo guided biopsy (when feasible, according to location of the nodule and bleeding risk). When a negative result was obtained after CT and echo guided biopsy, a strict follow up procedure was followed (three month intervals) and the nodule was rebiopsied when an increase in size was detected at US. Morphological staging (number of nodules, location, and vascular involvement) of HCC was based on US and CT findings. Treatment selection was based on this morphological stage and other parameters related to the severity of the underlying cirrhosis and patient characteristics, such as Child-Pugh class, age of the patient, and coexisting diseases (see “treatment schedule” below). Angiography and lipiodol CT were used only for therapeutic purposes and prior to surgical treatments. All diagnoses of HCC were further confirmed on surgical specimens or by follow up.

The incidence of HCC during follow up of the study population was first assessed and individual risk factors for developing HCC were investigated using multivariate analysis.

For the control group, we evaluated a consecutive series of HCC incidentally detected outside any specific surveillance programme in the same period (March 1989 to November 1991) and referred to our centre for definite diagnosis and treatment (contemporaneous non-randomised controls). A comparative analysis was made between characteristics and outcome of HCC detected during the prospective surveillance programme and those which arose in the unsurveilled population. Follow up of patients ended in December 1997 when the data were analysed.

TREATMENT SCHEDULE

When a diagnosis of HCC was established, the following therapeutic options were first considered both for the surveilled and unsurveilled patients:

(a)

hepatic resection for patients of Child-Pugh A class with unifocal nodules located in peripheral segments;

(b)

percutaneous ethanol injection (PEI) and/or transarterial chemoembolisation (TACE) for all Child-Pugh B patients over 60 years with unifocal nodules, for Child-Pugh A patients with unifocal nodules located centrally in the liver and over 60 years, for patients who refused surgery, and for patients with 2–3 nodules. The choice between the two procedures was done case by case according to the characteristics of the HCC (size, location) and of the underlying cirrhosis;

(c)

TACE was the only option for cases with more than three nodules or diffuse HCC involving less than 30% of the liver parenchyma;

(d)

OLT was considered for all patients with unifocal nodules <60 years old, except for those Child-Pugh A subjects with resectable nodules. A single session of TACE was performed in Child-Pugh A and B patients prior to assignment to the waiting list for OLT; further sessions were subsequently performed at four month intervals until transplantation.

Exclusion criteria for treatment were neoplastic portal vein thrombosis, diffuse HCC (>30% of the liver involved), other life threatening diseases, and Child-Pugh C class in patients >60 years old. Patients >75 years were considered only for PEI.

OUTCOME MEASURES AND ECONOMIC ISSUES

Outcome measures analysed in patients with HCC were: (1) number and size of HCC nodules detected by US; (2) eligibility for surgical or local treatment; (3) survival of patients with HCC; and (4) survival of all cirrhotic patients. The economic issues were calculated according to the charges at our university hospital for the various diagnostic and therapeutic procedures. They were: (1) overall cost of the surveillance programme (derived from charges for AFP, US, and other diagnostic procedures performed in patients with nodular lesions); (2) cost (derived from charges of diagnostic and therapeutic procedures) per treatable HCC; and (3) cost per year of life saved (if any).

PATIENTS

A total of 313 patients with LC screened by US and AFP were enrolled in the surveillance programme and had a follow up of at least six months. The characteristics of the study population are summarised in table 1. Contemporaneous controls numbered 104.

STATISTICAL ANALYSIS

The cumulative rate of HCC detected during the surveillance programme and survival curves for HCC detected during follow up and those in the unsurveilled population were calculated by life tables according to the Kaplan-Maier method18 and differences were tested by log rank test. The incidence of HCC per year of follow up was calculated according to a person time unit method, to fully utilise the whole period of follow up of each person. The Wilcoxon test was applied for comparison between tumour masses. The χ²test was used for comparisons between proportions in different groups.

The following variables were first investigated by univariate analysis in the group of 313 patients undergoing the surveillance programme: age, sex, hepatitis B virus, hepatitis C virus, antibodies to hepatitis B virus, ethanol abuse, primary biliary cirrhosis, cryptogenic cirrhosis, Child-Pugh class at entry, and serum AFP level at entry. The stepwise Cox proportional hazard model19 was used to identify those variables significantly and independently associated with the risk of HCC. All assumptions of the Cox model were tested and met. The relative importance of each variable identified was estimated on the basis of the ratio between the variable coefficient and its standard error. Patients with LC who developed other tumours or hepatic decompensation aged >60 years, those who underwent OLT, those lost to follow up, and those who died of causes unrelated to LC were regarded as censored cases.

To calculate the number of months of life possibly saved in surveilled patients in comparison with those unsurveilled, median survival times obtained from life tables were used, according to the recommendations of Peto and colleagues,20 which were applicable to our series.

The incremental cost per incremental unit of clinical outcome21 was estimated by calculating the incremental cost (if any) for treatable HCC in the surveilled population. The cost per year of life saved was consequently measured by dividing the incremental cost for the number of months of life gained in the surveilled group.

A multivariate analysis was performed to identify variables associated with survival. In this analysis, sex, Child-Pugh class, tumour staging, and inclusion in the surveillance programme were considered.

INCIDENCE OF HCC DURING THE SURVEILLANCE PROGRAMME

The mean follow up period of the cohort of 313 patients in the surveillance programme was 56 (SD 31) months (range 6–100 months). During this follow up period nodular liver lesions were detected in 74 cases by US (23.6%). In 13 cases (4.1%) the diagnosis of HCC could not be proved at the time of data analysis by other imaging techniques or by US guided biopsy and the final diagnosis was: macroregenerative nodule in six cases, haemangioma in three cases, and undefined in the remainder. All lesions were <2 cm in diameter. In 61 cases (19.5%) the diagnosis of HCC was finally proved by US guided biopsy (in 49 cases) and/or lipiodol CT and was confirmed at follow up. Four of these HCC were first diagnosed as macroregenerative nodules and subsequently met the criteria for a diagnosis of HCC. The sensitivity and specificity of the AFP assay at entry at the cut off level of 20 ng/dl were 41% and 82%, respectively, with a positive predictive value of 46% and negative predictive value of 85%. In none of the cases in the present series was a diagnosis of HCC made on the basis of an increase in AFP above 200 ng/dl in the absence of US visualisation of a liver mass. Figure 1 shows the rate of development of HCC over the period of observation. The incidence of HCC per year of follow up was 4.1%.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Cumulative rate of patients free of hepatocellular carcinoma (HCC) in the cohort of 313 patients with liver cirrhosis.

At the end of the study, 111 (35.5%) patients were still in follow up and 61 (19.5%) had developed HCC. The remaining patients were censored for the following reasons: 65 (20.8%) had died of liver failure or variceal haemorrhage, 24 (7.7%) had undergone transplantation, 24 (7.7%) were lost to follow up, 13 (4.3%) developed a non-hepatic malignant neoplasm, 10 (3.2%) died of non-liver related causes (mainly cardiovascular diseases), and five (1.6%) were over 60 years old in Child-Pugh class C.

ANALYSIS OF RISK FACTORS FOR HCC

Child-Pugh class at entry into the study of the 61 cirrhotic patients who developed HCC during the surveillance programme were as follows: 35 class A (57.4%), 21 class B (34.4%), and five class C (8.2%). Serum AFP levels at entry into the study were normal (<20 ng/dl) in 258 patients (82.4%) and >20 ng/dl in 55 (17.6%).Throughout the follow up period AFP levels remained less than 20 ng/dl in 236/313 patients (75.5%) and were persistently elevated in 37/313 (11.9%). In 24/313 patients (7.6%) (18 from the group with elevated AFP and six with normal AFP levels at entry), AFP levels fluctuated around the upper normal level, and in 16/313 (5.0%) they showed an increasing pattern above the normal range.

Univariate analysis of risk factors for HCC proved significant for sex, AFP >20 ng/dl, and Child-Pugh B/C class (table 2). The Cox model identified serum AFP >20 ng/dl at entry (coefficient/SE 4.03), Child-Pugh B/C class (coefficient/SE 3.06) and male sex (coefficient/SE 2.06) as independent significant risk factors for the development of HCC (table 2). In this model each variable was categorised binomially as 1 or 2 if, respectively, values were below/⩾20 ng/dl for AFP, A or B/C for Child-Pugh class, and female or male for sex. Figure 2 shows the differences in the rates of developing HCC among the different risk classes of cirrhotic patients stratified with respect to the most powerful risk factors (AFP and Child-Pugh class at entry) identified by the Cox model.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Cumulative rates of patients free of hepatocellular carcinoma (HCC) according to Child-Pugh class and α fetoprotein (AFP) level at entry into the study. 1, Child-Pugh class B/C and AFP >20 ng/ml (n=26); 2,Child-Pugh class A and AFP <20 ng/ml (n=173); 3, Child-Pugh class A and AFP >20 ng/ml (n=27); 4, Child-Pugh class B/C and AFP <20 ng/ml (n=87).

NUMBER AND SIZE OF THE DETECTED LESIONS

Table 3 reports the clinical pattern and morphological features of incidental HCC in the surveilled population. Sixty one HCC were unifocal at US in 49 cases (80.4%), multifocal in six cases (9.8%), and diffuse in six (9.8%). In 7/49 (14.3%) HCC that were unifocal at US the subsequent staging and therapeutic procedures revealed a multifocal hepatic disease. The mean (SD) volume of the tumour mass, calculated only in unifocal HCCs, was 12.5 (18.9) cm³. The ultrasonographic stage of HCC at the time of detection did not significantly change during follow up. This implies that HCCs detected in the last period of the study were similar to those detected at the beginning. Only 55 HCC of the unsurveilled population were unifocal (52.9% v 80.4% of HCC in surveilled patients; p<0.001) and the mean (SD) volume of their tumour mass was 20.1 (42.2) cm³ (NS v unifocal HCC in surveilled patients) (table4).

ELIGIBILITY FOR TREATMENT

In 19 liver cancers (31.1%) of the 61 detected during the surveillance period, the possibility of treatment was ruled out and no specific therapy was performed for the following reasons: diffuse HCC (six cases), refusal by the patient (five cases), liver failure (two cases), severe portal hypertension with previous bleeding episodes (two cases), and other causes (coexistent systemic disease or congenital splanchnic arterial abnormality preventing TACE; four cases). In the remaining 42 cases (68.9%) the following treatments were carried out (table 3): TACE was performed as the sole treatment in 13 elective patients and in five patients who died while on the waiting list (n=3) or while still awaiting OLT (n=2) at the time of analysis. The mean number of TACE sessions was 2.2 (range 1–3) in this group. TACE+PEI was performed in four patients, with a single session of TACE preceding PEI. PEI was the sole treatment in 10 cases. Hepatic resection was performed in four cases. Eleven patients were selected or considered candidates for OLT; six received OLT 6–12 months after the diagnosis of HCC, three died awaiting OLT, and two are still on the list. All but three patients (who had decompensated LC) in this group received TACE prior to assignment. Eligibility for treatment was not significantly different in cases arising during the first three years of the surveillance programme compared with those arising in the last period (table 3). Sixty one of the unsurveilled HCC patients were treated (58.6% v 68.8% surveilled; NS) (TACE n=28; PEI n=14; hepatic resection n=5; TACE+PEI n=6; OLT n=8).

SURVIVAL ANALYSIS OF PATIENTS WITH HCC

Survival of patients with HCC detected during the surveillance programme proved significantly longer (p=0.02) than that of unsurveilled patients (fig 3). The three year survival rates were, respectively, 45% and 31.7% (median survival 30v 15 months).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Cumulative survival rates of hepatocellular carcinoma (HCC) detected during follow up and HCC detected in unscreened patients. 1, HCC detected in surveilled patients (36 dead, 25 censored); 2, HCC detected in unsurveilled patients (95 dead, nine censored).

SURVIVAL ANALYSIS OF PATIENTS ENROLLED IN THE PROSPECTIVE STUDY

A total of 112 (35.8%) of 313 patients who entered the surveillance programme died; 37 (11.8%) had HCC. The cumulative survival rate of cirrhotic patients and patients with HCC which arose during the follow up, stratified according to Child-Pugh class, is shown in fig 4. The three year survival rates of cirrhotic patients in Child-Pugh A and B/C classes compared with that of patients with HCC in the corresponding Child-Pugh classes were, respectively, 92% versus 66% (p<0.05) and 71% versus 34% (p<0.05). Multivariate analysis identified tumour staging (p<0.001) and Child-Pugh score (p<0.01) as the variables independently associated with survival. However, if tumour staging was removed from the analysis, inclusion in the surveillance programme became significantly associated (p<0.01) with survival, indicating that these two variables, tumour staging and surveillance, are strictly related, but with the former prevailing on the latter.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Cumulative survival rates of patients with liver cirrhosis (LC) enrolled in the study and patients with incidental hepatocellular carcinoma (HCC) stratified according to Child-Pugh class. 1, LC and Child-Pugh A class; 2, LC and Child-Pugh B/C class; 3, HCC and Child-Pugh A class; 4, HCC and Child-Pugh B/C class.

COST OF THE SURVEILLANCE PROGRAMME

To carry out the present surveillance programme we performed 2874 abdominal US examinations and AFP determinations. The cost of such a programme, based on charges for US examinations and serum AFP assays performed twice a year during a mean follow up period of 56 months in 313 patients, was US$175 314. Additional charges for the definite diagnosis and staging of the 74 nodular lesions detected during the surveillance programme involving CT scans (n=74), echo guided biopsy (n=60), and charges for treatments—PEI (n=14), hepatic resection (n=4), OLT (n=6) and both diagnostic and therapeutic TACE (total n=51)—resulted in a total charge for the surveillance programme of US$753 226 (table 5).

The cost of treatable HCC was US$17 934 for the surveilled and US$14 555 for the unsurveilled patients (Δ=US$3379) (table 5), resulting in an incremental cost in the surveilled group of US$141 918. Comparison of survival of patients with HCC detected during the surveillance programme (median survival 30 months) with that of patients with HCC detected incidentally in the unsurveilled population (median survival 15 months) showed a cost per year of life saved of US$112 996. This estimate, however, does not take into account other incidental costs such as clinic visits, doctors' time, other costs related to hospitalisation, time lost in employment, and travel costs, which further increase the cost of surveillance. Furthermore, it does not take into account the unknown cost in the unsurveilled population (where US and other diagnostic tests, apart from those performed at the time of detection, were randomly performed prior to our observation) whose inclusion in the analysis would improve the cost effectiveness of the programme.