1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Background  Serum alpha fetoprotein (AFP), ultrasound, computerized tomography scanning, and magnetic resonance imaging are commonly used to screen for hepatocellular carcinoma (HCC) in patients with cirrhosis.

Aim  To assess the accuracy of screening in advanced cirrhosis.

Methods  The study group consisted of 239 patients with proven HCC in the explanted liver at the time of liver transplant. AFP and imaging were done at referral and serially until transplant.

Results  Hepatocellular carcinoma was detected before liver transplant in 78% and discovered incidentally in 22%. The cause of cirrhosis was hepatitis C (HCV) (55%), hepatitis B (HBV) (17%), alcohol (9%), and other/unknown (19%). Although AFP was elevated 62%, the median level was 15 ng/mL. Only 26%, 15% and 13% were more than 100, 400 and 1000 ng/mL, respectively. By comparison, AFP was elevated in 20% without HCC, but exceeded 100 ng/mL in only 3%. The overall accuracy of AFP was poor regardless of the cutoff. Magnetic resonance imaging was more accurate than computerized tomography or ultrasound in detecting tumour, particularly when performed within 3 months of transplant.

Conclusions  Magnetic resonance imaging is most sensitive for imaging HCC and best reflects actual tumour size. AFP is insensitive and adds little to screening strategies, but has prognostic value when extremely elevated.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Hepatocellular carcinoma (HCC) is becoming more common. The age-adjusted incidence increased 2-fold in the United States between 1985 and 1998.1, 2 It is currently estimated that there are more than 12 000 new cases of primary liver cancer, mostly HCC, in this country and more than 500 000 worldwide each year.3 Furthermore, it has been projected that the number of cases of HCC will continue to increase for the next three decades, primarily as a consequence of the prevalence of hepatitis C.4, 5

Surveillance for HCC in patients with cirrhosis or long-standing hepatitis B infection has been a common practice for many years. Serum alpha fetoprotein (AFP) testing and imaging studies are usually recommended at intervals or 6 or 12 months.6 This interval is based on the low incidence, typically 1–4% per year among cirrhotics, and slow growth of these tumours with a median doubling time of 117 to 150 days.7 Current recommendations support this practice, although the optimal interval and method for screening have been debated.8 Indeed, the cost-effectiveness of screening has been questioned, particularly in the years before therapeutic options such as ablation or transplantation became widely available. Now, however, HCC is often a curable tumour so early identification of lesions is desirable. Thus, the optimal strategy for HCC surveillance needs to be better characterized.

The goal of this retrospective study was to evaluate the ability of serum AFP and different imaging techniques to identify and accurately stage proven HCC in patients with advanced cirrhosis who were prospectively followed-up while waiting for liver transplantation.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Patient population

This was a retrospective study. The study was conducted at the Baylor Regional Transplant Center, including Baylor University Medical Center in Dallas, TX and Baylor All Saints Hospital in Fort Worth, TX and was reviewed and approved by the Baylor Institutional Review Board. The study group consisted of patients with cirrhosis who were discovered to have HCC, either before or at the time of orthotopic liver transplantation. Patients transplanted between January 1, 1988 and December 31, 2004 was included. Subjects without HCC who were transplanted during the same time period served as controls. The study group and controls had undergone prospective AFP testing and imaging during their initial evaluation and thereafter while waiting for transplant according to a standard institutional protocol. Information reported in this study was obtained from the Baylor Transplant Database and supplemented as necessary by review of medical records and computerized laboratory and imaging databases. Clinical and demographic information included age, gender, date of transplant, primary diagnosis, laboratory tests and calculated Model of Endstage Liver Disease (MELD) score at initial presentation, serum AFP, ultrasound and gadolinium magnetic resonance imaging (MRI). Subjects in whom HCC was suspected prior to transplantation also underwent computerized tomography (CT) scanning of the abdomen with and without intravenous contrast as part of a metastatic evaluation. All imaging studies were performed with a standard ‘tumour protocol’ employed for all potential liver transplant recipients. This ensured consistent methodology and reading by dedicated radiology personnel, although some technological changes in CT and MRI methodology have taken place over time since the study began (see text below). Imaging studies that were not performed according to protocol, e.g. without intravenous contrast, or were performed at other facilities are not included in this study. AFP and imaging tests were repeated during listing, although compliance with the recommended protocol depended in large part on patients’ ability to come to the transplant centre on a regular basis. A normal AFP in our laboratory is less than 8.9 ng/mL. The interval of testing varied during the study period. Patients with a known or suspected lesion had imaging every 3 months, while imaging was recommended every 6 to 12 months in others.

All cases of HCC identified by imaging before transplant were confirmed by pathologic examination of the explanted liver and these cases were defined as known. Cases only identified in the explant were labelled as incidental. Explanted livers were cut into sequentially oriented slices of 1 cm thickness. Nodules that were distinct from surrounding parenchyma by virtue of size, colour, or consistency were sampled for histology. The size, number and location of tumours were documented. Cholangiocarcinoma, fibrolamellar carcinoma and other types of liver tumours were not included in this analysis.

Statistical methods

Summary statistics are reported as mean ± standard deviation. Median and range are provided for skewed variables. Sensitivity, specificity, predictive values and accuracy were calculated using conventional methods. Continuous variables were compared using Students t-test and dichotomous variables were compared using test Chi-square test with Yate’s correction (XL-Stat 2006). Comparison of tumour size by different imaging studies and pathologic examination utilized the paired t-test and linear regression. Not all patients had each imaging study. Thus, the denominators vary for different imaging studies as we censored empty fields to avoid biasing interpretation of correlations.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Demographic data

Four-thousand, nine-hundred and ninety-nine patients were referred for evaluation for liver transplantation. Of these, HCC was the reason for referral in only 267 (5.3%). Twenty-three hundred and seventy-two were approved for listing and underwent transplant at Baylor between January 1, 1988 and December 31, 2004. HCC was present in 239 (10.1%) patients who underwent transplantation. HCC was known prior to transplantation in 186 (77.8%) and was incidental in 53 (22.2%). HCC accounted for 154 of 1874 transplants (8.2%) performed before and 85 of 498 transplants (17.1%) performed after implementation of the MELD allocation system in 2002.

Demographic characteristics of patients transplanted for HCC are summarized in Table 1. Patients were predominantly male (73.2%), Caucasian (82.0%), and middle aged (52.7 ± 9.4 years). The cause of liver disease was HCV [132, 55.2%; 35/132 (26.5%) of these also had a history of excessive alcohol use], HBV (40, 16.7%; five of these had concurrent HCV infection), alcohol alone (22, 9.2%), cryptogenic (16, 6.7%), other (15, 6.3%) and unknown (14, 5.9%). Most (226 of 239; 94.6%) had cirrhosis. Ten of the 13 without cirrhosis had no apparent liver disease, while two had chronic hepatitis B and one had chronic hepatitis C. Among patients transplanted for HCC, the proportion due to non-A, non-B hepatitis or HCV increased from 13.3% in 1988–1990 to 44.3% thereafter, reaching 60.9% in 2004. During the same period, non-A, non-B hepatitis or HCV accounted for 1.6% of all patients transplanted in 1989 and increased to a peak of 49.7% in 2002 before appearing to level off at 40.6% in 2004.

Table 1.   Baseline characteristics of 239 patients transplanted with hepatocellular carcinoma
  1. NASH, non-alcoholic steatohepatitis.

  2. * 35 of 132 (26.5%) with chronic hepatitis C had heavy alcohol intake; † 5 also had HCV and 2 had HDV (included in HBV category only); ‡ 35 patients with HCV and alcohol not included in this number.

Age (mean ± s.d.)52.7 ± 9.4 (range 17–72)
Gender (male)175 (73.2%)
 Caucasian196 (82.0%)
 African-American18 (7.5%)
 Asian12 (5.0%)
 Other13 (5.4%)
Cirrhosis226 (94.6%)
Aetiology of underlying liver disease
 Hepatitis C132 (55.2%)*
 Hepatitis B40 (16.7%)†
 Alcohol22 (9.2%)‡
 Cryptogenic cirrhosis16 (6.7%)
 Hemochromatosis4 (1.7%)
 Alpha-1 antitrypsin deficiency3 (1.3%)
 Primary biliary cirrhosis3 (1.3%)
 Primary sclerosing cholangitis2 (0.8%)
 Autoimmune hepatitis2 (0.8%)
 NASH1 (0.4%)
 No Diagnosis14 (5.9%)
Incidental tumours53 (22.2%)

Alpha fetoprotein

The mean AFP level at initial evaluation for all patients transplanted for HCC was 1246 ± 5550 ng/mL, but the median was 14.8 ng/mL. More than half of these patients, 148/239 (61.9%), had an AFP level above the upper limit of the normal range and of these, 75 (31.4%), 61 (25.9%), 48 (20.1), 35 (14.6%) and 32 (13.4%) were greater than 50, 100, 200, 400 and 1000 ng/mL, respectively. In solitary tumours less than or equal to 2 cm, an elevated AFP was present in 38.1%, but none had an AFP more than 200 ng/mL. In patients transplanted without HCC, the AFP was elevated in 345/1696 (19.6%). Among patients without HCC, 61 (3.6%), 27 (1.6%), 13 (0.8%), eight (0.5%) and two (0.1%) were above 50, 100, 200, 400 and 1000 ng/mL (< 0.0001 vs. HCC patients). There was no apparent cause for the elevated AFP in these patients and levels returned to normal after transplant. The sensitivity, specificity and predictive values of AFP at these levels are detailed in Table 2. AFP was a poor predictor of HCC regardless of the cutoff used with statistical accuracy of only 0.33 to 0.48. Interestingly, AFP levels were higher in the 13 patients without cirrhosis (8863 + 15 354 ng/mL, five of 13 > 1000 ng/mL) than in those with cirrhosis (831 + 4031 ng/mL), but this difference was not significant (= 0.11) because of the wide variability of values in both groups.

Table 2.   Test characteristics of alpha fetoprotein
Cut-offSensitivitySpecificityPPVNPVPositive DLR*
  1. DLR, diagnostic likelihood ratio; NPV, negative predictive value; PPV, positive predictive value.

  2. * Positive DLR is defined as: sensitivity/(1-specificity) and it represents the odds ratio that an elevated alpha fetoprotein result will be observed in an hepatocellular carcinoma (HCC) patient compared to a patient without HCC. Thus, tests with higher positive DLR values are more useful.

8.9 ng/mL (normal) 0.620.800.300.943.04
50 ng/mL0.310.960.550.918.72
100 ng/mL0.260.980.690.9016.03
200 ng/mL0.20>0.990.790.9026.20
400 ng/mL0.15>0.990.810.8931.05
1000 ng/mL0.13>0.990.940.89113.54

Hepatocellular carcinoma patients with an AFP equal to or greater than 200 ng/mL were not more likely than those with lower values to have tumour detectable by ultrasound (54.3% vs. 46.6%, = 0.48), CT scan (60.9% vs. 44.0%, = 0.06), or MRI (71.7% vs. 62.7%, = 0.33). The aetiology of liver disease or prior treatment of viral hepatitis was also not different between these two groups. Likewise, the mean AFP in HCC patients with or without HCV was not different (1210.8 ± 5679.6 vs. 1304.1 ± 5364.9 ng/mL; =0.90). Patients with a single lesion were less likely to have an AFP of 200 ng/mL or higher (39.1% vs. 58.0%, = 0.03). Patients with diffuse tumour noted in the explanted organ were more likely to have an AFP of 200 ng/mL or more (23.9% vs. 9.8%, = 0.02). AFP level was not predictive of microscopic vascular invasion in the explant (= 0.11).

Imaging studies


Ultrasound examinations were performed in 199 of the 239 patients (83.4%) with HCC and identified the tumour in 115 (57.8%). Ultrasound was able to detect only 23 of 66 (34.9%) lesions less than or equal to 2 cm in largest diameter compared to 92 of 133 (69.2%) of larger lesions (< 0.0001) (Table 3). The estimated size of tumours identified by ultrasound was 3.4 ± 2.0 cm and this correlated closely with the measured size of the lesion in the explant (3.6 ± 2.0 cm; < 0.0001). The average diameter of tumours that were not seen by ultrasound was 2.4 + 1.5 cm (range, 0.3–7.0 cm). The concordance between the number of lesions seen on ultrasound and in the explant was 45.2%. Concordance was 54.0% for single lesions and 30.0% for multifocal or diffuse tumour. Although ultrasound examinations performed within 3 months of the time of transplantation were more likely to demonstrate tumour when present (64.3%), the benefit over scanning performed within 6 months (59.5%) or 12 months (58.2%) was not significant (= 0.42).

Table 3.   Characteristics of hepatocellular carcinoma (HCC) identified by different imaging techniques
 UltrasoundComputerized tomography (CT)Magnetic resonance imaging (MRI)
  1. * MRI vs. Ultrasound, < 0.00001; MRI vs. CT, = 0.08; CT vs. Ultrasound, = 0.04.

  2. Helical CT technology and advanced MRI gradient hardware and processing were introduced in 2000.

  3. † Size by ultrasound, CT or MRI vs. actual size (explant), NS.

  4. ‡ Size of detected HCC vs. not detected by ultrasound (< 0.00001), CT (= 0.18), and MRI (= 0.002).

  5. § Number of detected HCC vs. not detected by ultrasound (< 0.002), CT (= 0.13), and MRI (= 0.28).

Number of imaging studies199164197
Sensitivity (# detected/# imagings)*
 Overall115 (57.8%)113 (68.9%)153 (77.7%)
 Sensitivity before 2000N/A43/71 (60.5%)59/84 (70.2%)
 Sensitivity after 2000N/A70/93 (75.3%)94/113 (83.2%)
Tumour size
 Largest lesion on imaging (cm)†,3.4 + 2.03.5 + 2.23.3 + 2.5
 Size of same lesion in explant (cm)3.6 + 2.03.6 + 2.33.6 + 2.7
 Size of lesions not detected (cm) 2.4 + 1.53.1 + 2.02.3 + 1.6
Tumour number
 Number of lesions by imaging§1.4 + 0.81.6 + 0.91.5 + 0.8
 Number of lesions in explant2.0 + 1.21.7 + 1.21.7 + 1.0
 Concordance of imaging/explant  number45.2%46.0%49.0%
Sensitivity according to size and scanning interval
 Tumours <2 cm maximum diameter23/66 (34.9%)36/44 (59.1%)41/65 (63.1%)
 Tumours >2 cm maximum diameter92/133 (69.2%)87/120 (72.5%)112/132 (84.8%)
 Imaging if <3 vs. >3 months before  transplant 64.3%/53.0% (= 0.15)76.2%/63.8% (= 0.12)86.6%/60.3% (< 0.0001)
 Imaging if <6 vs. >6 months before  transplant59.5%/54.4% (= 0.59)75.9%/56.3% (< 0.0001)81.4%/56.7% (= 0.006)
 Imaging if <12 vs. >12 months before  transplant58.2%/54.6% (= 0.92)73.5%/41.2% (< 0.0001)80.0%/14.3% (= 0.0002)
Computer tomography

There were 164 patients with HCC who had at least one CT scan during their evaluation or while listed. The tumour was identified in 113 of these (68.9%). CT was more sensitive than ultrasound for small lesions. CT detected 26 of 44 (59.1%) of lesions with a maximum diameter less than or equal to 2 cm, while it detected 87 of 120 (72.5%) of larger lesions (= 0.15) (Table 3). The estimated size of tumours identified by CT scan was 3.6 ± 2.2 cm, which was nearly identical to the actual size of the lesions in the explant (3.6 ± 2.3 cm; < 0.0001). The average diameter of tumours that were not seen by CT was 3.1 + 2.1 cm (range, 0.5–12.0 cm). The concordance between number of lesions seen on CT and in the explanted organ was 46.0%. Concordance was 81.1% for single lesions and 50.0% for diffuse tumours. CT scans were only slightly more likely to demonstrate the tumour if performed within 3 months of the time of transplantation (76.2% vs. 63.8% for more than 3 months), but the difference was not significant. However, scans performed within 6 or 12 months of transplant were significantly more likely to demonstrate tumour than those done at greater intervals (75.9% vs. 56.3% for 6 months and 73.5% vs. 41.2% for 12 months; both < 0.0001).

Newer helical computerized scanning technology was introduced at our institution in 2000. This improved the sensitivity of CT scans from 60.5% (43 of 71) before 2000 to 75.3% (70 of 93) thereafter (= 0.06) (Table 3). However, the sensitivity of CT for 2 cm or smaller lesions only increased from 54.6% to 60.6% (= 0.74).

Magnetic resonance imaging

One-hundred and ninety-seven patients with HCC had an MRI performed at evaluation and, in many cases, on multiple occasions during listing. Gadolinium was the standard contrast agent for liver MRI scans, but feridex was sometimes used to clarify the cause of lesions discovered on gadolinium studies. Tumour was identified in 153 of 197 (77.7%) cases. MRI was more sensitive than ultrasound or CT for small lesions (Table 3). MRI detected 41 of 65 (63.1%) of lesions with a maximum diameter less than or equal to 2 cm, while it detected 112 of 132 (84.8%) of larger lesions (= 0.001). The estimated size of tumours identified by MRI scan was 3.3 ± 2.5 cm and correlated closely with the measured size of the lesion in the explant (3.6 ± 2.7 cm; < 0.0001). The average diameter of tumours that were not seen by MRI was 2.3 + 1.6 cm (range, 0.4–9.5 cm). The concordance between number of lesions seen on MRI and in the explant was 49.0%. Concordance was 79.5% for single lesions, but only 20.0% for multifocal or diffuse tumours. MRI scans were most sensitive when performed within 3 months of transplantation and demonstrated 86.6% of tumours (vs. 60.3% for MRIs done more than 3 months before transplant, < 0.0001). Similarly, scans performed within 6 or 12 were significantly more likely to demonstrate tumour than those done at greater intervals (81.4% vs. 56.7%, = 0.006 and 80.0% vs. 14.3%, = 0.0002). However, scans performed within 3 months of transplant were not statistically more sensitive than those performed within 6 or 12 months (= 0.27 and = 0.15, respectively).

Advances in gradient hardware and processing strategies that improved echoplanar spectroscopic MRI imaging were introduced at Baylor in 2000. These changes resulted in a significant improvement in MRI sensitivity from 70.2% (59 of 84) in patients scanned before 2000 to 83.2% (94 of 113) in those done thereafter (= 0.047) (Table 3). The sensitivity for 2 cm or smaller lesions also increased from 40.0% to 73.3% (= 0.01).

Impact of MELD

After the Milan criteria became the basis for the MELD allocation priority in 20029, 10 the mean AFP level fell from 3245.3 ± 7120.8 ng/mL (median, 13.0 ng/mL) to 30.1 ± 47.9 (median, 3.5 ng/mL) (= 0.06). This was coincident with the fall in average number of lesions (2.0 ± 1.1 to 1.3 ± 0.8, = 0.03) and maximum tumour diameter (4.4 ± 4.2 to 2.5 ± 1.1 cm, < 0.0001), as would be predicted from the Milan restrictions. Accordingly, as the range of AFP values decreased, the ability of AFP to predict HCC became even less.

We divided all 239 HCC patients according to whether or not the number and size of tumours in the explanted organ met MELD guidelines for transplantation.9, 10 Those who exceeded MELD tumour limits had higher AFP levels (3036.5 ± 8790 ng/mL vs. 211.1 ± 1160.9 ng/mL, = 0.004; 33.7% vs. 11.3%≥200 ng/mL, < 0.00001), were less likely to have hepatitis C (44.9% vs. 61.3%, = 0.02) or single tumours (21.3% vs. 74.0%, < 0.00001), and were more likely to have diffuse tumour (33.7% vs. 0%, < 0.00001) or vascular invasion (34.8% vs. 9.3%, < 0.00001) on pathological examination of the explanted liver.

Incidental tumours

The number of incidental tumours varied from year to year but has trended downward over time. The proportion of total tumours represented by incidental tumours fell slightly after introduction of more sensitive CT and MRI imaging technology in 2000 (25.4% to 18.8%, = 0.28).

Alpha fetoprotein levels were lower in incidental tumours than in tumours identified prior to transplantation [17.7 ± 36.6 (range, 1.5–240.0) vs. 1592.4 ± 6244.4 ng/mL (range, 0.9–48 444); < 0.001]. However, while patients with incidental and known tumours had the same number of tumour nodules (2.0 ± 1.2 vs. 1.7 ± 1.0, NS), the former group had smaller maximum tumour diameters (2.2 ± 1.2 vs. 4.1 ± 3.9, < 0.0001) and were more likely to be within the Milan criteria based on examination of the explanted liver (79.3% vs. 58.1%, < 0.001).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Hepatocellular carcinoma is common and increasing in patients with cirrhosis, particularly those with chronic viral hepatitis.1–5, 11 Our study found that HCC has also become a more common reason for referral for transplantation. This parallels a 3-fold increase in the proportion of cases with chronic hepatitis C among those referred to our centre for liver transplantation since 1990. While cases undergoing liver transplantation may reflect a referral bias, similar increases in the incidence of HCC and HCC-related deaths have been seen throughout the world.3, 12, 13 It is anticipated that this trend may continue in coming decades.4, 5

Patients with cirrhosis or chronic hepatitis B require regular surveillance for HCC.6 Although serum AFP and ultrasonography every 6 to 12 months have been the preferred method of screening for many years, the optimal method and interval for surveillance are debated.8 Until recently, these were only academic concerns as there were few options for managing tumours that were identified. Surgical resection or liver transplantation was possible in a relatively small proportion of these cases. Recently, however, transplant has become more widely available and effective ablative therapies such as radiofrequency ablation, transarterial chemoembolization, and image-guided radiation therapy offer further options to most patients with non-metastatic HCC.14–17 Thus, early diagnosis is critical for optimal management. Our study examined the accuracy of screening tests in detecting and staging HCC in the subset of patients with advanced cirrhosis, many of whom already had known HCC. It is important to recognize that this was not a prospective study to examine efficacy of a particular surveillance strategy. Thus, our results speak only of the accuracy of AFP and imaging tests in detection and staging of HCC in this subset of patients and do not specifically address the effectiveness of surveillance methods in the general population of patients with cirrhosis.

Elevation of the AFP level was the way HCC was diagnosed prior to the availability of sensitive abdominal imaging techniques in the late 1970’s.18 However, it soon became apparent with the advent of ultrasonography and CT scanning that many tumours could be detected in patients without an elevated AFP. Indeed, our study confirms the low sensitivity of approximately 60% previously reported by others and suggests that the clinical utility of AFP as a marker for HCC has perhaps declined below a level of usefulness as the sensitivity of imaging techniques has improved.8, 18, 19 While AFP was more specific, if a higher threshold for concern was used, the overall accuracy of AFP remained poor regardless of the cutoff employed. Thus, the role of AFP for HCC surveillance needs to be reconsidered and a more sensitive and specific replacement must be sought. On the other hand, marked elevations of the AFP may imply a greater tumour burden that is likely to exceed the Milan restrictions for transplantation.

Abdominal imaging has improved dramatically over the last two decades and has replaced angiography, exploratory laparotomy and percutaneous biopsy as the preferred tools to identify hepatic tumours. Despite this progress, however, ultrasound, CT and MRI remain relatively insensitive for detecting HCC, particularly with tumours less than 2 cm in diameter.20, 21 Our study confirms that new hardware and software technology that was introduced for CT and MRI in 2000 improved the sensitivity of both modalities and partly explains the recent decrease in incidental tumours discovered only at the time of transplant. MRI is the most sensitive modality and identifies almost 80% of tumours, including 63% of tumours less than 2 cm. These findings may reflect not only the accuracy of the imaging modality, but also local expertise and practice preference. This, as well as local availability of different technologies, must be considered when selecting the optimal test for detecting and staging tumours a particular centre.

Not surprisingly, imaging within 3 or 6 months of transplantation was generally more likely to detect tumour HCC and this was statistically more sensitive than longer intervals for CT or MRI scanning. Nearly 90% of tumours were detected when an MRI was performed within 3 months of transplantation. Thus, it appears that CT or MRI performed every 6 months would be the most sensitive method for detecting HCC. Based on the sensitivity of testing in our analysis, MRI exams every 3 to 6 months would have increased detection from 67.7% to 81.4–86.6% (88.2% with addition of AFP), compared to AFP (≥100 ng/mL) and ultrasound (= 0.004). Although we did not perform cost-benefit analyses, the cost of frequent MRI examinations might be prohibitive unless limited to patients with advanced cirrhosis who are candidates for transplant or ablation. However, our findings should be interpreted with caution. Our study group included many patients with known tumour. It is not known whether shorter intervals between imaging tests would improve detection of HCC in the general population of patients with cirrhosis who undergo surveillance. Furthermore, it remains to be determined whether earlier identification of small lesions by frequent MRI imaging would improve tumour-free survival. Thus, these should not be considered general recommendations for surveillance in the general population of patients with cirrhosis. Rather, they might serve as a basis for prospective studies in that group. Future recommendations for cancer surveillance strategies also need to consider cost, impact of tumour-free survival, and institutional availability of and proficiency with imaging methods.22


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Declaration of personal interests: Drs. Snowberger, Chinnakotla, Lepe, and Goldstein have no interests to declare. Ms. Peattie has no outside interests to declare. Dr. Klintmalm receives clinical research funding from AbSorber, Astellas, Genxyme, Isotechnika, Novartis, Pfizer, Roche, and Y’s Therapeutics. Dr. Davis receives clinical research funding from Roche, Schering-Plough, Human Genome Science, and Vertex. Declaration of funding interests: This study was funded entirely by the Baylor Regional Transplant Institute.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    El-Serag HB, Mason AC. Rising incidence of hepatocellular carcinoma in the United States. N Engl J Med 1999; 340: 74550.
  • 2
    El-Serag HB, Davila JA, Petersen NJ, McGlynn KA. The continuing increase in the incidence of hepatocellular carcinoma in the United States: an update. Ann Intern Med 2003; 139: 81723.
  • 3
    Bosch FX, Ribes J, Diaz M, Cleries R. Primary liver cancer: worldwide incidence and trends. Gastroenterol 2004; 127: S516.
  • 4
    Davis GL, Albright JE, Cook SF, Rosenberg DM. Projecting the future healthcare burden from hepatitis C in the United States. Liver Transpl 2003; 9: 3318.
  • 5
    Armstrong GL, Alter MJ, McQuillan GM, Margolis HS. The past incidence of hepatitis C virus infection: implications for the future burden of chronic liver disease in the United States. Hepatology 2000; 31: 77782.
  • 6
    Bruix J, Sherman M. AASLD Practice Guideline: management of hepatocellular carcinoma. Hepatology 2005; 42: 120836.
  • 7
    Sheu J-C, Sung J-L, Chen D-S, et al. Growth rate of asymptomatic hepatocellular carcinoma and its clinical implications. Gastroenterology 1985; 89: 25966.
  • 8
    Di Bisceglie AM. Issues in screening and surveillance for hepatocellular carcinoma. Gastroenterology 2004; 127: S1047.
  • 9
    Mazzaferro V, Regalia E, Doci R, et al. Liver transplantation for the treatment of small hepatocellular carcinomas in patients with cirrhosis. N Engl J Med 1996; 334: 6939.
  • 10
    Sharma P, Balan V, Hernandez JL, et al. Liver transplantation for hepatocellular carcinoma: the MELD impact. Liver Transpl 2004; 10: 3641.
  • 11
    Di Bisceglie AM, Order SE, Klein JL, et al. The role of chronic viral hepatitis in hepatocellular carcinoma in the United States. Am J Gastroenterol 1991; 86: 3358.
  • 12
    Kiyosawa K, Umemura T, Ichijo T, et al. Hepatocellular carcinoma: recent trends in Japan. Gastroenterology 2004; 127: S1726.
  • 13
    Fottovich G, Stroffolini T, Zagni I, Donato F. Hepatocellular carcinoma in cirrhosis: incidence and risk factors. Gastroenterology 2004; 127: S3550.
  • 14
    Kudo M. Early detection and curative treatment of early stage hepatocellular carcinoma. Clin Gastroenterol Hepatol 2005; 3: S1448.
  • 15
    Sutherland LM, Williams JA, Padbury RT, Gotley DC, Stokes B, Maddern GJ. Radiofrequency ablation of liver tumors: a systematic review. Arch Surg 2006; 141: 18190.
  • 16
    Tateishi R, Shiina S, Teratani T, et al. Percutaneous radiofrequency ablation for hepatocellular carcinoma. An analysis of 1000 cases. Cancer 2005; 103: 12019.
  • 17
    Sangiovanni A, Del Ninno E, Fassani P, et al. Increased survival of cirrhotic patients with a hepatocellular carcinoma. Gastroenterology 2004; 126: 100514.
  • 18
    Sherman M. Alpha fetoprotein: an obituary. J Hepatol 2001; 34: 6035.
  • 19
    Trevisani F, D’Intino PE, Morselli-Labate AM, et al. Serum alpha-fetoprotein for diagnosis of hepatocellular carcinoma in patients with chronic liver disease: influence of HBsAg and anti-HCV status. J Hepatol 2001; 34: 5705.
  • 20
    Baron RL, Brancatelli G. Computed tomography imaging of hepatocellular carcinoma. Gastroenterology 2004; 127: S13343.
  • 21
    Taouli B, Losada M, Holland A, Krinsky G. Magnetic resonance imaging of hepatocellular carcinoma. Gastroenterology 2004; 127: S14452.
  • 22
    Patel D, Terrault NA, Yao FY, Bass NM, Ladabaum U. Cost-effectiveness of hepatocellular carcinoma surveillance in patients with hepatitis C virus-related cirrhosis. Clin Gastroenterol Hepatol 2005; 3: 7584.