Studies of patients with cirrhosis uncover limitations in liver cancer screening
Two studies (Full Text Available On Both Studies Below) available in the December issue of Hepatology, a journal of the American Association for the Study of Liver Diseases, have uncovered limitations in screening for primary liver cancer, also known as hepatocellular carcinoma (HCC).
The first study found that, if given the choice during a clinical trial, most patients with cirrhosis prefer surveillance over the possibility of non-screening, therefore making a randomized study of HCC screening not feasible.
A second study determined that ultrasonographic screening at three monthly versus six monthly intervals did not improve the detection of small liver cancers. Medical evidence reports HCC to be the sixth most common cancer and the third most common cause of cancer death worldwide, with 90% of all cases in western countries attributed to chronic liver diseases, typically at the cirrhosis stage.
The National Cancer Institute estimates that more than 26,000 cases and close to 20,000 deaths from liver and bile duct cancer occurred in the U.S. in 2011. Clinical guidelines recommend routine screening for HCC, but the efficacy and optimal intervals for testing are strongly debated by experts. In the first study, researchers led by Professor Jacob George from the University of Sydney and Westmead Hospital in Australia, examined the feasibility of undertaking a randomized controlled trial of HCC surveillance in patients with cirrhosis.
The screening program included ultrasonography every six months and alpha-fetoprotein testing every three months. Of the 205 participants with cirrhosis who received information outlining the risks and benefits of surveillance for liver cancer, 99.5% declined randomization, with 88% electing for a non-randomized screening program.
"While a randomized controlled trial is ideal to assess the success of a cancer surveillance program, we found that patients with cirrhosis declined randomization due to possible allocation to a non-screening group," explains Professor George. "Since HCC screening in cirrhotic patients is routine practice for a majority of clinicians, it is impossible to assign patients to a genuine control group.
However, further prospective studies that compare individual screening strategies are warranted." In a survey of 40 gastroenterologists of the Sydney Liver Group, the authors found that 74% routinely screen cirrhotic patients despite believing that screening did not increase patient survival (37%) or that the surveillance was cost-effective (66%). One such liver cancer screening strategy was investigated by a team of French and Belgian researchers led by Professor Jean-Claude Trinchet with the Hôpital Jean Verdier in Bondy, France. The team conducted a multicentre trial with 1278 cirrhotic patients who received ultrasonographic screening at either three-month or six-month intervals.
Their results indicated that cirrhosis resulted from excessive alcohol use in 39% of participants, 44% from hepatitis C virus (HCV), and 13% from hepatitis B virus (HBV). During the study period from July 2000 to July 2009, researchers detected at least one focal lesion in 28% of patients, but confirmed small HCC (less than 30 mm) in only 10% of participants. Dr. Trinchet said, "Our study found that ultrasonographic surveillance performed every three months detects more small focal lesions than screening at six-month intervals. However, more frequent screening did not improve the detection of liver cancer at an earlier stage."
The authors note that detection of small tumors were more likely missed in patients with HCV or who abuse alcohol and suggest the limitations of current diagnostic procedures may explain their negative findings. Again, further investigations of screening methods and diagnostic procedures are needed to improve the outcomes in those at risk for developing liver cancer.
Feasibility of conducting a randomized control trial for liver cancer screening: Is a randomized controlled trial for liver cancer screening feasible or still needed? 'informed patients prefer surveillance'
Hepatology December 2011
Hossein Poustchi,1,2 Geoffrey C. Farrell,3 Simone I. Strasser,4 Alice U. Lee,5 Geoffrey W. McCaughan,4 and Jacob George2
http://onlinelibrary.wiley.com/doi/10.1002/hep.24581/abstract.
Abstract
Screening for hepatocellular carcinoma (HCC) is commonly practiced and recommended in published guidelines, but evidence for its efficacy has been controversial. We tested the feasibility of conducting a randomized controlled trial (RCT) of HCC surveillance in patients with cirrhosis and followed up those offered screening to detect clinical outcomes. Participation was offered to patients with cirrhosis attending liver clinics at three university hospitals. Following discussion, patients received a decision aid (DA) that outlined the risks and benefits of surveillance. The proposed screening program comprised ultrasonography 6-monthly and serum alpha-fetoprotein every 3 months. We envisaged five groups of patients: those who agreed to randomization, those choosing nonrandomized screening, those wanting continuation of usual care, those who were undecided, and those refusing participation. Among 205 patients, 204 (99.5%) declined randomization. Of these, 181 (88%) elected for a nonrandomized screening program, 10% chose usual care (which typically included ad hoc screening), and two were undecided. Among 176 patients fluent in English communication skills, 160 (91%) preferred nonrandomized screening compared with 22/29 (76%) patients needing an interpreter (P < 0.026). Of 173 patients in nonrandomized screening followed up for a mean 13.5 ± 6.04 months, three developed HCC, two died from nonliver-related causes, and one underwent liver transplantation for liver failure. Eighteen of 21 patients in "usual care" received ad hoc screening. A simultaneous survey on the quality of the DA showed that the majority of participants believed that the information provided was unbiased. Conclusion: Although an RCT is theoretically ideal for determining the efficacy, efficiency, and cost-effectiveness of HCC screening, informed patients prefer surveillance. A randomized study of HCC screening is not feasible when informed consent is imparted. (HEPATOLOGY 2011;)
On a global scale, hepatocellular carcinoma (HCC) is the third commonest cause of cancer death.1 In the United States the greatest increase in cancer death rate over the last decade has been from HCC, the incidence of which has risen faster than all cancers except for cancers of the lung.2 Early detection, made possible through the use of imaging or serum markers, is desirable because of its dismal prognosis. At the same time, HCC fulfils several criteria that make it suitable for a surveillance program, most notably the fact that small lesions identified early may benefit from potentially curative therapy.3 Other criteria include the relatively high disease burden in selected populations and the availability of reasonably accurate diagnostic tests. For these reasons, surveillance has been advocated in order to identify those with small tumors.4 Several reports suggest an improved survival rate from liver cancer among patients who participate in a screening program.5-9
However, in the absence of a randomized controlled trial (RCT), these results are open to various types of bias, notably lead-time bias in determining quality years of life gained and cost-efficacy. Although RCTs offer the best design for comparing the effectiveness of an intervention,10 to our knowledge only two RCTs have been conducted on screening for HCC.11, 12 Both studies were conducted in China, which has a high prevalence of chronic hepatitis B virus (HBV) infection and HCC. In both reports those with chronic HBV infection with11 or without12 evidence of chronic hepatitis were randomly assigned to either surveillance or to a control group. In neither study were patients offered the option of choosing nonrandomized screening and no information on individual informed consent or contemporary local clinical practice was available.
An important consideration in RCT design is that of patient willingness to participate. From the researcher's point of view, RCTs provide the best evidence for the efficacy of an intervention. However, the critical issue is whether this is also important to patients. Many reports indicate that patients are often unwilling to participate in RCTs when they are aware of the fact that chance determines their treatment allocation. McQuellon et al.13 noted that 90% of breast cancer patients considering a hypothetical trial scenario would not allow the toss of a coin to determine their treatment arm.
Patient preference and the relative lack of intensity of screening programs already in routine clinical practice are two obstacles to having a control group in an RCT related to surveillance. Although RCTs seem justified when there is uncertainty about the effectiveness of two drugs for a particular condition,14 it is difficult to convince patients to accept participation in a control group in the context of a cancer surveillance program.
To date, there is no study to systematically document whether an RCT for liver cancer screening is practical in an at-risk population in a developed country in the modern era of readily available hepatic imaging and serological testing, and whether patients with advanced liver disease are willing to participate in such a trial. In the present study we attempted to determine if conducting an RCT for liver cancer surveillance was feasible and to determine the willingness of patients with cirrhosis to participate in such a study, as well as the outcomes if they did not.
ALD, alcoholic liver disease; CHB, chronic hepatitis B; CHC, chronic hepatitis C; DA, decision aid; HCC, hepatocellular carcinoma; RCT, randomized controlled trial.
Patients and Methods
The study was undertaken in the liver clinics of three university-affiliated teaching hospitals (Westmead, Royal Prince Alfred, and Concord Hospitals), all in Sydney, Australia. The study and all documents were approved by the respective Human Research Ethics Committees of the hospitals and that of the University of Sydney. Patients with cirrhosis and Childs-Pugh A or B status attending the liver clinics between March 2004 and August 2005 were invited to participate in an RCT that compared screening with a nonscreening approach for the detection of primary liver cancer.
The surveillance protocol comprised estimations of alpha-fetoprotein every 3 months and hepatic ultrasonography every 6 months. To ensure the tests were performed based on scheduled time, patients were reminded by investigators if their tests were past due. Patients with Childs-Pugh C cirrhosis were excluded, as it was considered that a high rate of non-HCC related endpoints including death and liver transplantation may result in an inadequate number of incident cases. Cirrhosis was confirmed by liver biopsy. In the absence of histology, cirrhosis was defined by the presence of at least one of the following clinical stigmata: ascites, esophageal varices or splenomegaly, and laboratory findings of a low serum albumin, a prolonged prothrombin time, or thrombocytopenia at enrolment.
Random allocation to either the screening or the nonscreening group was offered to all participants. Patients not consenting to random allocation were asked to choose the group they wished to join (or to decline participation). To enable participants to make an informed choice, a written decision aid (DA) (online Supporting Material) was developed and provided in addition to the participant information and consent form. Patients were given the opportunity to discuss issues around their participation with the investigators, their local general practitioner, family members, and significant others. The information and consent form included a summary of the study and details about the screening protocol.
The DA provided detailed simple information to ensure patients fully understood the implications of undergoing surveillance for HCC. Topics addressed included risk factors for liver cancer and its natural history, the results of previous studies of HCC screening, and the probable advantages and disadvantages of surveillance. The DA emphasized that outcomes of HCC surveillance programs varied in different parts of the world and that there were no data demonstrating a long-term survival benefit for patients subjected to screening. More practical considerations such as the inconvenience of undergoing regular surveillance and clinic visits were discussed. The DA was provided to patients and their families in the presence of an interpreter for those not fluent in English. After 2 weeks, patients were reinterviewed and consented if appropriate.
We envisaged five groups of patients following this process: those agreeing to randomization to screening or to nonscreening (group 1), those choosing nonrandomized screening (group 2), those wanting continuation of usual care (which may or may not have included an element of screening) (group 3), those who were undecided (group 4), and those who refused participation (group 5). A follow-up questionnaire was developed that addressed patients' attitudes and their involvement in the decision-making process. This was provided to all consenting participants.
To determine what was likely to be "usual care" for individuals refusing study entry, we asked 35 gastroenterologists and gastroenterology trainees attending meetings of the Sydney Liver Group to complete a questionnaire about their attitudes to screening, and routine practice in relation to patients with known cirrhosis.
Results
Characteristics of Patients
In all, 212 patients with cirrhosis (Child-Pugh A and B) were approached to participate. Of these, 7 (3%) declined, whereas 205 (97%) accepted. Reasons given by patients for nonparticipation included time constraints (n = 3), lack of interest (n = 2), and already being screened (n = 2). The demographic characteristics of the remaining 205 consenting participants are shown in Table 1. The majority (75%) were men with a mean age of 54.5 years (21-78 years), largely reflecting the known gender differences in chronic HBV and HCV infections. Most were Caucasian (62.5%), followed by participants of Asian (22.5%) and Middle Eastern (14%) ethnicity. A total of 176 (86%) spoke English fluently, whereas 29 (14%) required the assistance of an interpreter. The most common cause of liver disease was chronic hepatitis C (CHC) (n = 101, 49%) followed by chronic hepatitis B (CHB) (n = 56, 27.5%), and alcoholic cirrhosis (n = 18, 8.5%) (Table 2).
Patient Election to the Study Arms.
When offered participation into the RCT of HCC screening versus nonscreening (Gp 1), 204 of the 205 (99.5%) patients declined entry. Of these, 181/204 (88.3%) elected for nonrandomized screening (Gp 2), 21 (10.2%) chose nonrandomized "usual care" (Gp 3) and two (1%) were undecided (Gp 4).
We next determined whether demographic variables influenced patient selection of study arm (Gp1-5). By univariate analysis, effective English communication skills influenced choice: 160/176 (91%) of those who were fluent in English preferred nonrandomized screening compared with 22/29 (76%) of patients who needed an interpreter (P < 0.03). Likewise, patients who could attend the liver clinic independently were more likely to choose nonrandomized screening than those who relied on others to come to the hospital (90% versus 79%, respectively), but this difference did not attain statistical significance. Other variables such as gender, age, and ethnicity did not influence the decision to select screening (Gp 2) versus usual care (Gp 3) (Table 3). We performed multiple logistic regression analysis that included all variables with an initial P < 0.25. English proficiency was the only independent predictor of patient choice (odds ratio [OR] 0.387; 95% confidence interval [CI] 0.150-0.952, P = 0.04).
Outcomes for Screened Patients
The mean follow-up for patients was 13.5 ± 6.04 months when data were censored for the analysis in this report. During this period, 173 (95%) patients who chose nonrandomized screening (Gp 2) continued to receive active follow-up according to the protocol. Of these, three (1.5%) had developed HCC, two (1%) died from nonliver-related causes, and one (0.5%) underwent liver transplantation for liver failure. Nine (5%) patients withdrew and were subsequently lost to follow-up. The majority of patients in usual care (Gp 3) (18 of 21) continued to receive ad hoc screening as part of their clinical care.
Patient Involvement in Decision Making and Quality of the Decision Aid.
Patient involvement in the decision to participate in screening was assessed by their responses to a questionnaire. Of 205 patients provided with the questionnaire, 110 responses were received; 56 (51%) patients determined their surveillance arm allocation on their own, or after discussion with their doctor, whereas 16 (14%) made the decision jointly with their doctor. In only one (1%) case did the doctor solely make the final decision on behalf of the patient. About one-third (32%) of respondents did not discuss the program with their doctor. Two participants (2%) did not respond to this question.
A univariate analysis of factors associated with a patients' decision to consider the liver cancer screening program is presented in Table 4. Only the level of education (high school or more versus less education) influenced patient attitudes toward screening. Thus, those with greater education were more likely to make the decision alone (47 [57.3%]) than those who had no education or who had only completed primary school education (10 [38.5%]; P = 0.034).
As part of the study design patients were questioned about the quality of the DA and whether the aid was biased in favor of or against participation in a RCT of screening. Fifty-nine (53.5%) patients believed that all information in the DA was clear, whereas 41 (37.5%) thought most of the information was clear. The majority of patients (62 [56.5%]) believed that the information provided was unbiased; 16 (14.5%) considered the DA to be very biased in favor of screening, whereas 27 (24%) considered it was slightly biased in favor of screening. One patient considered that the DA did not favor screening. The majority of respondents (79%) considered that the amount of information in the DA was adequate, although 19 (17.5%) would have liked more information. When asked whether the DA would be helpful for other patients in the same scenario, 72 respondents (65.5%) agreed that it would be very helpful, 33 (30%) considered it somewhat helpful, and four (3.5) patients stated that the DA would be of little help.
Physician Attitudes Toward Screening for Liver Cancer
Finally, we undertook a survey among gastroenterologists and gastroenterology trainees attending meetings of the Sydney Liver Group to ascertain their views about screening for HCC in cirrhotic patients; 35 of 40 attendees completed the questionnaire. Most respondents (20 [57%]) cared for patients in hospital, 19 (54.5%) saw more than five cirrhotic patients each week, and 15 (43%) had more than 10 years experience. The characteristics of the respondents are summarized in Table 5.
Thirteen respondents (37%) believed that screening of cirrhotic patients did not increase patient survival, whereas four (11.5%) were unsure. Twenty-three (65.5%) believed there was no evidence that screening was cost-effective. Despite these concerns, the majority of respondents (26 [74%]) routinely screened all cirrhotic patients. Only seven (20%) discussed the options available with their patients before undertaking screening. Thirty (86%) participants screened all cirrhotic patients, two (5.5%) screened patients with cirrhosis caused by CHC, CHB, or alcohol, whereas one (3%) screened HBV- and HCV-infected cirrhotic patients only. One respondent (3%) screened all patients with abnormal liver tests.
Discussion
To our knowledge, the present report is the first attempt to systematically test the feasibility of conducting an RCT of surveillance for liver cancer in a clinic-based population of cirrhotic patients at high risk for developing primary liver tumors. Entry into this study was informed by a DA developed specifically for this purpose that outlined the risks and benefits of screening. This approach was deemed ethically necessary, as screening has been recommended in published guidelines, whereas clinical surveys15-18 suggest that screening is frequently undertaken, despite borderline and controversial evidence of its benefits.
This study demonstrates that (1) RCTs in cirrhotic patients in developed nations is not possible and should not be further considered, and (2) given the responsibility to decide whether to accept randomization or not, the approach and concerns of patients differed radically from that of researchers. Despite the fact that there is no convincing data on the cost-effectiveness of HCC screening, almost all participants rejected randomization and preferred surveillance. One reason for declining randomization is fear of the arbitrary nature of the process. Consistent with this notion, the results of an earlier study demonstrated that 63% of patients refused entry based purely on an aversion to randomization. In this regard, emphasis given to chance in the explanation of the concept of randomization is known to increase patient unease,19 whereas a literature review to assess factors that influence an individual's willingness or not to participate in a clinical trial noted that the patient's degree of uncertainty, random allocation to treatment, and the use of a placebo were the three factors that caused the greatest concern and led patients to decline study entry.20
A further reason for the overwhelming lack of interest in randomization by our study participants might be the adequacy of information about the study process. We supplemented the standard participant information sheet and consent form with a decision aid in order to ensure that patients were well informed about the purposes of the study and the methods by which they were to be allocated to a study group, if they chose randomization. This assertion is supported by the results of published reports that suggest that individuals are unaware of being "randomized," despite a participant information sheet.16-18, 21 Thus, if more information is provided, patients may be less likely to agree to randomization.12-14 For example, in a study among patients with various types of cancer, the overall attitudes to participation in medical research was positive, with 69% of respondents agreeing to take part in a protocol comparing two treatments. However, this figure dropped to 34% when the treatment arm was chosen by random allocation.22 Similar results have been reported by others.23-25
Reluctance of patients to participate in RCTs may also stem from their desire to have a more active role in medical decision-making. In the present study, 56 of 112 (50%) respondents made the final decision to join the screening program alone. Indeed, a third did not discuss the program with their family physician. Several publications have emphasized this aspect, noting that patients usually refuse participation in RCTs because of a preference by either the treating physician or themselves to make the decision about treatment choices.26, 27
We observed that the majority of patients chose nonrandomized surveillance in the belief that screening helps doctors to detect cancer earlier. This points to a general misconception about, and unrealistic expectations of, the benefits of screening in the general population that has been confirmed in other reports. For example, in a study of 4,140 women surveyed on the benefits of breast cancer surveillance, 68% believed that screening prevented or reduced the risk of breast cancer, whereas 62% believed that screening halved breast cancer mortality.28 In a further publication, of women over 40 and men over 50 with no known history of cancer, 87% believed that routine cancer surveillance is always a good idea, whereas 74% considered that finding cancers earlier saved lives.29
It should be noted that some participants who chose the control arm (usual care), were already in an ad hoc screening program. When we examined doctors' attitudes towards HCC screening, our results suggested that although the benefits of surveillance for patients are not clear to doctors, it is currently routine practice among the majority of gastroenterologists in Sydney. The overwhelming majority believed that all cirrhotic patients, irrespective of their underlying liver disease, would benefit from screening, a result that is consistent with a report among members of the American Association for the Study of Liver Diseases (AASLD) that revealed that 84% routinely screen patients with cirrhosis for primary liver cancer.15 Based on our findings, it is clearly impractical to use random allocation to assign cirrhotic participants to HCC surveillance and impossible to have a control cohort in this high-risk population, despite the lack of strong efficacy data.
Although most of the information required for decision-making on an individual basis was provided for participants, it is possible that they were not aware of all the potential harms of a surveillance program, including the risks involved in work-up of potentially benign incidental lesions, including that of biopsy and radiation exposure, not to mention mental anxiety and community costs of a surveillance program.
An RCT is the ideal method to assess the efficacy of a cancer surveillance program. In practice, we found that when patients with cirrhosis are asked to make an informed choice about participation in a randomized clinical trial, the vast majority declined randomization and preferred to undergo surveillance rather than to accept possible allocation to nonscreening. Further, because screening for liver cancer in cirrhotic patients is routine practice for the majority of clinicians, even if patients show no interest in such a program, they are highly likely to be "screened," thus making it impossible to allocate to a genuine control group. Hence, RCTs of screening for HCC is not ethically feasible in current clinical practice. However, while this is the case, carefully conducted prospective studies to compare individual HCC screening strategies and modalities are needed to improve early diagnosis and hopefully to improve the outcomes of liver cancer.
Two studies (Full Text Available On Both Studies Below) available in the December issue of Hepatology, a journal of the American Association for the Study of Liver Diseases, have uncovered limitations in screening for primary liver cancer, also known as hepatocellular carcinoma (HCC).
The first study found that, if given the choice during a clinical trial, most patients with cirrhosis prefer surveillance over the possibility of non-screening, therefore making a randomized study of HCC screening not feasible.
A second study determined that ultrasonographic screening at three monthly versus six monthly intervals did not improve the detection of small liver cancers. Medical evidence reports HCC to be the sixth most common cancer and the third most common cause of cancer death worldwide, with 90% of all cases in western countries attributed to chronic liver diseases, typically at the cirrhosis stage.
The National Cancer Institute estimates that more than 26,000 cases and close to 20,000 deaths from liver and bile duct cancer occurred in the U.S. in 2011. Clinical guidelines recommend routine screening for HCC, but the efficacy and optimal intervals for testing are strongly debated by experts. In the first study, researchers led by Professor Jacob George from the University of Sydney and Westmead Hospital in Australia, examined the feasibility of undertaking a randomized controlled trial of HCC surveillance in patients with cirrhosis.
The screening program included ultrasonography every six months and alpha-fetoprotein testing every three months. Of the 205 participants with cirrhosis who received information outlining the risks and benefits of surveillance for liver cancer, 99.5% declined randomization, with 88% electing for a non-randomized screening program.
"While a randomized controlled trial is ideal to assess the success of a cancer surveillance program, we found that patients with cirrhosis declined randomization due to possible allocation to a non-screening group," explains Professor George. "Since HCC screening in cirrhotic patients is routine practice for a majority of clinicians, it is impossible to assign patients to a genuine control group.
However, further prospective studies that compare individual screening strategies are warranted." In a survey of 40 gastroenterologists of the Sydney Liver Group, the authors found that 74% routinely screen cirrhotic patients despite believing that screening did not increase patient survival (37%) or that the surveillance was cost-effective (66%). One such liver cancer screening strategy was investigated by a team of French and Belgian researchers led by Professor Jean-Claude Trinchet with the Hôpital Jean Verdier in Bondy, France. The team conducted a multicentre trial with 1278 cirrhotic patients who received ultrasonographic screening at either three-month or six-month intervals.
Their results indicated that cirrhosis resulted from excessive alcohol use in 39% of participants, 44% from hepatitis C virus (HCV), and 13% from hepatitis B virus (HBV). During the study period from July 2000 to July 2009, researchers detected at least one focal lesion in 28% of patients, but confirmed small HCC (less than 30 mm) in only 10% of participants. Dr. Trinchet said, "Our study found that ultrasonographic surveillance performed every three months detects more small focal lesions than screening at six-month intervals. However, more frequent screening did not improve the detection of liver cancer at an earlier stage."
The authors note that detection of small tumors were more likely missed in patients with HCV or who abuse alcohol and suggest the limitations of current diagnostic procedures may explain their negative findings. Again, further investigations of screening methods and diagnostic procedures are needed to improve the outcomes in those at risk for developing liver cancer.
Feasibility of conducting a randomized control trial for liver cancer screening: Is a randomized controlled trial for liver cancer screening feasible or still needed? 'informed patients prefer surveillance'
Hepatology December 2011
Hossein Poustchi,1,2 Geoffrey C. Farrell,3 Simone I. Strasser,4 Alice U. Lee,5 Geoffrey W. McCaughan,4 and Jacob George2
http://onlinelibrary.wiley.com/doi/10.1002/hep.24581/abstract.
Abstract
Screening for hepatocellular carcinoma (HCC) is commonly practiced and recommended in published guidelines, but evidence for its efficacy has been controversial. We tested the feasibility of conducting a randomized controlled trial (RCT) of HCC surveillance in patients with cirrhosis and followed up those offered screening to detect clinical outcomes. Participation was offered to patients with cirrhosis attending liver clinics at three university hospitals. Following discussion, patients received a decision aid (DA) that outlined the risks and benefits of surveillance. The proposed screening program comprised ultrasonography 6-monthly and serum alpha-fetoprotein every 3 months. We envisaged five groups of patients: those who agreed to randomization, those choosing nonrandomized screening, those wanting continuation of usual care, those who were undecided, and those refusing participation. Among 205 patients, 204 (99.5%) declined randomization. Of these, 181 (88%) elected for a nonrandomized screening program, 10% chose usual care (which typically included ad hoc screening), and two were undecided. Among 176 patients fluent in English communication skills, 160 (91%) preferred nonrandomized screening compared with 22/29 (76%) patients needing an interpreter (P < 0.026). Of 173 patients in nonrandomized screening followed up for a mean 13.5 ± 6.04 months, three developed HCC, two died from nonliver-related causes, and one underwent liver transplantation for liver failure. Eighteen of 21 patients in "usual care" received ad hoc screening. A simultaneous survey on the quality of the DA showed that the majority of participants believed that the information provided was unbiased. Conclusion: Although an RCT is theoretically ideal for determining the efficacy, efficiency, and cost-effectiveness of HCC screening, informed patients prefer surveillance. A randomized study of HCC screening is not feasible when informed consent is imparted. (HEPATOLOGY 2011;)
On a global scale, hepatocellular carcinoma (HCC) is the third commonest cause of cancer death.1 In the United States the greatest increase in cancer death rate over the last decade has been from HCC, the incidence of which has risen faster than all cancers except for cancers of the lung.2 Early detection, made possible through the use of imaging or serum markers, is desirable because of its dismal prognosis. At the same time, HCC fulfils several criteria that make it suitable for a surveillance program, most notably the fact that small lesions identified early may benefit from potentially curative therapy.3 Other criteria include the relatively high disease burden in selected populations and the availability of reasonably accurate diagnostic tests. For these reasons, surveillance has been advocated in order to identify those with small tumors.4 Several reports suggest an improved survival rate from liver cancer among patients who participate in a screening program.5-9
However, in the absence of a randomized controlled trial (RCT), these results are open to various types of bias, notably lead-time bias in determining quality years of life gained and cost-efficacy. Although RCTs offer the best design for comparing the effectiveness of an intervention,10 to our knowledge only two RCTs have been conducted on screening for HCC.11, 12 Both studies were conducted in China, which has a high prevalence of chronic hepatitis B virus (HBV) infection and HCC. In both reports those with chronic HBV infection with11 or without12 evidence of chronic hepatitis were randomly assigned to either surveillance or to a control group. In neither study were patients offered the option of choosing nonrandomized screening and no information on individual informed consent or contemporary local clinical practice was available.
An important consideration in RCT design is that of patient willingness to participate. From the researcher's point of view, RCTs provide the best evidence for the efficacy of an intervention. However, the critical issue is whether this is also important to patients. Many reports indicate that patients are often unwilling to participate in RCTs when they are aware of the fact that chance determines their treatment allocation. McQuellon et al.13 noted that 90% of breast cancer patients considering a hypothetical trial scenario would not allow the toss of a coin to determine their treatment arm.
Patient preference and the relative lack of intensity of screening programs already in routine clinical practice are two obstacles to having a control group in an RCT related to surveillance. Although RCTs seem justified when there is uncertainty about the effectiveness of two drugs for a particular condition,14 it is difficult to convince patients to accept participation in a control group in the context of a cancer surveillance program.
To date, there is no study to systematically document whether an RCT for liver cancer screening is practical in an at-risk population in a developed country in the modern era of readily available hepatic imaging and serological testing, and whether patients with advanced liver disease are willing to participate in such a trial. In the present study we attempted to determine if conducting an RCT for liver cancer surveillance was feasible and to determine the willingness of patients with cirrhosis to participate in such a study, as well as the outcomes if they did not.
ALD, alcoholic liver disease; CHB, chronic hepatitis B; CHC, chronic hepatitis C; DA, decision aid; HCC, hepatocellular carcinoma; RCT, randomized controlled trial.
Patients and Methods
The study was undertaken in the liver clinics of three university-affiliated teaching hospitals (Westmead, Royal Prince Alfred, and Concord Hospitals), all in Sydney, Australia. The study and all documents were approved by the respective Human Research Ethics Committees of the hospitals and that of the University of Sydney. Patients with cirrhosis and Childs-Pugh A or B status attending the liver clinics between March 2004 and August 2005 were invited to participate in an RCT that compared screening with a nonscreening approach for the detection of primary liver cancer.
The surveillance protocol comprised estimations of alpha-fetoprotein every 3 months and hepatic ultrasonography every 6 months. To ensure the tests were performed based on scheduled time, patients were reminded by investigators if their tests were past due. Patients with Childs-Pugh C cirrhosis were excluded, as it was considered that a high rate of non-HCC related endpoints including death and liver transplantation may result in an inadequate number of incident cases. Cirrhosis was confirmed by liver biopsy. In the absence of histology, cirrhosis was defined by the presence of at least one of the following clinical stigmata: ascites, esophageal varices or splenomegaly, and laboratory findings of a low serum albumin, a prolonged prothrombin time, or thrombocytopenia at enrolment.
Random allocation to either the screening or the nonscreening group was offered to all participants. Patients not consenting to random allocation were asked to choose the group they wished to join (or to decline participation). To enable participants to make an informed choice, a written decision aid (DA) (online Supporting Material) was developed and provided in addition to the participant information and consent form. Patients were given the opportunity to discuss issues around their participation with the investigators, their local general practitioner, family members, and significant others. The information and consent form included a summary of the study and details about the screening protocol.
The DA provided detailed simple information to ensure patients fully understood the implications of undergoing surveillance for HCC. Topics addressed included risk factors for liver cancer and its natural history, the results of previous studies of HCC screening, and the probable advantages and disadvantages of surveillance. The DA emphasized that outcomes of HCC surveillance programs varied in different parts of the world and that there were no data demonstrating a long-term survival benefit for patients subjected to screening. More practical considerations such as the inconvenience of undergoing regular surveillance and clinic visits were discussed. The DA was provided to patients and their families in the presence of an interpreter for those not fluent in English. After 2 weeks, patients were reinterviewed and consented if appropriate.
We envisaged five groups of patients following this process: those agreeing to randomization to screening or to nonscreening (group 1), those choosing nonrandomized screening (group 2), those wanting continuation of usual care (which may or may not have included an element of screening) (group 3), those who were undecided (group 4), and those who refused participation (group 5). A follow-up questionnaire was developed that addressed patients' attitudes and their involvement in the decision-making process. This was provided to all consenting participants.
To determine what was likely to be "usual care" for individuals refusing study entry, we asked 35 gastroenterologists and gastroenterology trainees attending meetings of the Sydney Liver Group to complete a questionnaire about their attitudes to screening, and routine practice in relation to patients with known cirrhosis.
Results
Characteristics of Patients
In all, 212 patients with cirrhosis (Child-Pugh A and B) were approached to participate. Of these, 7 (3%) declined, whereas 205 (97%) accepted. Reasons given by patients for nonparticipation included time constraints (n = 3), lack of interest (n = 2), and already being screened (n = 2). The demographic characteristics of the remaining 205 consenting participants are shown in Table 1. The majority (75%) were men with a mean age of 54.5 years (21-78 years), largely reflecting the known gender differences in chronic HBV and HCV infections. Most were Caucasian (62.5%), followed by participants of Asian (22.5%) and Middle Eastern (14%) ethnicity. A total of 176 (86%) spoke English fluently, whereas 29 (14%) required the assistance of an interpreter. The most common cause of liver disease was chronic hepatitis C (CHC) (n = 101, 49%) followed by chronic hepatitis B (CHB) (n = 56, 27.5%), and alcoholic cirrhosis (n = 18, 8.5%) (Table 2).
Patient Election to the Study Arms.
When offered participation into the RCT of HCC screening versus nonscreening (Gp 1), 204 of the 205 (99.5%) patients declined entry. Of these, 181/204 (88.3%) elected for nonrandomized screening (Gp 2), 21 (10.2%) chose nonrandomized "usual care" (Gp 3) and two (1%) were undecided (Gp 4).
We next determined whether demographic variables influenced patient selection of study arm (Gp1-5). By univariate analysis, effective English communication skills influenced choice: 160/176 (91%) of those who were fluent in English preferred nonrandomized screening compared with 22/29 (76%) of patients who needed an interpreter (P < 0.03). Likewise, patients who could attend the liver clinic independently were more likely to choose nonrandomized screening than those who relied on others to come to the hospital (90% versus 79%, respectively), but this difference did not attain statistical significance. Other variables such as gender, age, and ethnicity did not influence the decision to select screening (Gp 2) versus usual care (Gp 3) (Table 3). We performed multiple logistic regression analysis that included all variables with an initial P < 0.25. English proficiency was the only independent predictor of patient choice (odds ratio [OR] 0.387; 95% confidence interval [CI] 0.150-0.952, P = 0.04).
Outcomes for Screened Patients
The mean follow-up for patients was 13.5 ± 6.04 months when data were censored for the analysis in this report. During this period, 173 (95%) patients who chose nonrandomized screening (Gp 2) continued to receive active follow-up according to the protocol. Of these, three (1.5%) had developed HCC, two (1%) died from nonliver-related causes, and one (0.5%) underwent liver transplantation for liver failure. Nine (5%) patients withdrew and were subsequently lost to follow-up. The majority of patients in usual care (Gp 3) (18 of 21) continued to receive ad hoc screening as part of their clinical care.
Patient Involvement in Decision Making and Quality of the Decision Aid.
Patient involvement in the decision to participate in screening was assessed by their responses to a questionnaire. Of 205 patients provided with the questionnaire, 110 responses were received; 56 (51%) patients determined their surveillance arm allocation on their own, or after discussion with their doctor, whereas 16 (14%) made the decision jointly with their doctor. In only one (1%) case did the doctor solely make the final decision on behalf of the patient. About one-third (32%) of respondents did not discuss the program with their doctor. Two participants (2%) did not respond to this question.
A univariate analysis of factors associated with a patients' decision to consider the liver cancer screening program is presented in Table 4. Only the level of education (high school or more versus less education) influenced patient attitudes toward screening. Thus, those with greater education were more likely to make the decision alone (47 [57.3%]) than those who had no education or who had only completed primary school education (10 [38.5%]; P = 0.034).
As part of the study design patients were questioned about the quality of the DA and whether the aid was biased in favor of or against participation in a RCT of screening. Fifty-nine (53.5%) patients believed that all information in the DA was clear, whereas 41 (37.5%) thought most of the information was clear. The majority of patients (62 [56.5%]) believed that the information provided was unbiased; 16 (14.5%) considered the DA to be very biased in favor of screening, whereas 27 (24%) considered it was slightly biased in favor of screening. One patient considered that the DA did not favor screening. The majority of respondents (79%) considered that the amount of information in the DA was adequate, although 19 (17.5%) would have liked more information. When asked whether the DA would be helpful for other patients in the same scenario, 72 respondents (65.5%) agreed that it would be very helpful, 33 (30%) considered it somewhat helpful, and four (3.5) patients stated that the DA would be of little help.
Physician Attitudes Toward Screening for Liver Cancer
Finally, we undertook a survey among gastroenterologists and gastroenterology trainees attending meetings of the Sydney Liver Group to ascertain their views about screening for HCC in cirrhotic patients; 35 of 40 attendees completed the questionnaire. Most respondents (20 [57%]) cared for patients in hospital, 19 (54.5%) saw more than five cirrhotic patients each week, and 15 (43%) had more than 10 years experience. The characteristics of the respondents are summarized in Table 5.
Thirteen respondents (37%) believed that screening of cirrhotic patients did not increase patient survival, whereas four (11.5%) were unsure. Twenty-three (65.5%) believed there was no evidence that screening was cost-effective. Despite these concerns, the majority of respondents (26 [74%]) routinely screened all cirrhotic patients. Only seven (20%) discussed the options available with their patients before undertaking screening. Thirty (86%) participants screened all cirrhotic patients, two (5.5%) screened patients with cirrhosis caused by CHC, CHB, or alcohol, whereas one (3%) screened HBV- and HCV-infected cirrhotic patients only. One respondent (3%) screened all patients with abnormal liver tests.
Discussion
To our knowledge, the present report is the first attempt to systematically test the feasibility of conducting an RCT of surveillance for liver cancer in a clinic-based population of cirrhotic patients at high risk for developing primary liver tumors. Entry into this study was informed by a DA developed specifically for this purpose that outlined the risks and benefits of screening. This approach was deemed ethically necessary, as screening has been recommended in published guidelines, whereas clinical surveys15-18 suggest that screening is frequently undertaken, despite borderline and controversial evidence of its benefits.
This study demonstrates that (1) RCTs in cirrhotic patients in developed nations is not possible and should not be further considered, and (2) given the responsibility to decide whether to accept randomization or not, the approach and concerns of patients differed radically from that of researchers. Despite the fact that there is no convincing data on the cost-effectiveness of HCC screening, almost all participants rejected randomization and preferred surveillance. One reason for declining randomization is fear of the arbitrary nature of the process. Consistent with this notion, the results of an earlier study demonstrated that 63% of patients refused entry based purely on an aversion to randomization. In this regard, emphasis given to chance in the explanation of the concept of randomization is known to increase patient unease,19 whereas a literature review to assess factors that influence an individual's willingness or not to participate in a clinical trial noted that the patient's degree of uncertainty, random allocation to treatment, and the use of a placebo were the three factors that caused the greatest concern and led patients to decline study entry.20
A further reason for the overwhelming lack of interest in randomization by our study participants might be the adequacy of information about the study process. We supplemented the standard participant information sheet and consent form with a decision aid in order to ensure that patients were well informed about the purposes of the study and the methods by which they were to be allocated to a study group, if they chose randomization. This assertion is supported by the results of published reports that suggest that individuals are unaware of being "randomized," despite a participant information sheet.16-18, 21 Thus, if more information is provided, patients may be less likely to agree to randomization.12-14 For example, in a study among patients with various types of cancer, the overall attitudes to participation in medical research was positive, with 69% of respondents agreeing to take part in a protocol comparing two treatments. However, this figure dropped to 34% when the treatment arm was chosen by random allocation.22 Similar results have been reported by others.23-25
Reluctance of patients to participate in RCTs may also stem from their desire to have a more active role in medical decision-making. In the present study, 56 of 112 (50%) respondents made the final decision to join the screening program alone. Indeed, a third did not discuss the program with their family physician. Several publications have emphasized this aspect, noting that patients usually refuse participation in RCTs because of a preference by either the treating physician or themselves to make the decision about treatment choices.26, 27
We observed that the majority of patients chose nonrandomized surveillance in the belief that screening helps doctors to detect cancer earlier. This points to a general misconception about, and unrealistic expectations of, the benefits of screening in the general population that has been confirmed in other reports. For example, in a study of 4,140 women surveyed on the benefits of breast cancer surveillance, 68% believed that screening prevented or reduced the risk of breast cancer, whereas 62% believed that screening halved breast cancer mortality.28 In a further publication, of women over 40 and men over 50 with no known history of cancer, 87% believed that routine cancer surveillance is always a good idea, whereas 74% considered that finding cancers earlier saved lives.29
It should be noted that some participants who chose the control arm (usual care), were already in an ad hoc screening program. When we examined doctors' attitudes towards HCC screening, our results suggested that although the benefits of surveillance for patients are not clear to doctors, it is currently routine practice among the majority of gastroenterologists in Sydney. The overwhelming majority believed that all cirrhotic patients, irrespective of their underlying liver disease, would benefit from screening, a result that is consistent with a report among members of the American Association for the Study of Liver Diseases (AASLD) that revealed that 84% routinely screen patients with cirrhosis for primary liver cancer.15 Based on our findings, it is clearly impractical to use random allocation to assign cirrhotic participants to HCC surveillance and impossible to have a control cohort in this high-risk population, despite the lack of strong efficacy data.
Although most of the information required for decision-making on an individual basis was provided for participants, it is possible that they were not aware of all the potential harms of a surveillance program, including the risks involved in work-up of potentially benign incidental lesions, including that of biopsy and radiation exposure, not to mention mental anxiety and community costs of a surveillance program.
An RCT is the ideal method to assess the efficacy of a cancer surveillance program. In practice, we found that when patients with cirrhosis are asked to make an informed choice about participation in a randomized clinical trial, the vast majority declined randomization and preferred to undergo surveillance rather than to accept possible allocation to nonscreening. Further, because screening for liver cancer in cirrhotic patients is routine practice for the majority of clinicians, even if patients show no interest in such a program, they are highly likely to be "screened," thus making it impossible to allocate to a genuine control group. Hence, RCTs of screening for HCC is not ethically feasible in current clinical practice. However, while this is the case, carefully conducted prospective studies to compare individual HCC screening strategies and modalities are needed to improve early diagnosis and hopefully to improve the outcomes of liver cancer.
Hepatology December 2011
Ultrasonographic surveillance of hepatocellular carcinoma in cirrhosis: A randomized trial comparing 3- and 6-month periodicities
Jean-Claude Trinchet,1,2 Cendrine Chaffaut,3,4 Vale rie Bourcier,1,2 Francoise Degos,5,6 Jean Henrion,7He lene Fontaine,8,9 Dominique Roulot,2,10 Ariane Mallat,11,12 Sophie Hillaire,13 Paul Cales,14Isabelle Ollivier,15 Jean-Pierre Vinel,16 Philippe Mathurin,17 Jean-Pierre Bronowicki,18 Vale rie Vilgrain,6,19,20Gise le N'Kontchou,1,2 Michel Beaugrand,1,2 Sylvie Chevret,3,4and for the Groupe d'Etude et de Traitement du Carcinome He patocellulaire (GRETCH)
View Full Article with Supporting Information (HTML
Hepatology
Volume 54, Issue 6, pages 1987–1997, December 2011
Abstract
Detection of small hepatocellular carcinoma (HCC) eligible for curative treatment is increased by surveillance, but its optimal periodicity is still debated. Thus, this randomized trial compared two ultrasonographic (US) periodicities: 3 months versus 6 months. A multicenter randomized trial was conducted in France and Belgium (43 sites). Patients with histologically proven compensated cirrhosis were randomized into two groups: US every 6 months (Gr6M) or 3 months (Gr3M). For each focal lesion detected, diagnostic procedures were performed according to European Association for the Study of the Liver guidelines. Cumulative incidence of events was estimated, then compared using Gray's test. The prevalence of HCC ≤30 mm in diameter was the main endpoint. A sample size of 1,200 patients was required. A total of 1,278 patients were randomized (Gr3M, n = 640; Gr6M, n = 638; alcohol 39.2%, hepatitis C virus 44.1%, hepatitis B virus 12.5%). At least one focal lesion was detected in 358 patients (28%) but HCC was confirmed in only 123 (9.6%) (uninodular 58.5%, ≤30 mm in diameter 74%). Focal-lesion incidence was not different between Gr3M and Gr6M groups (2-year estimates, 20.4% versus 13.2%, P = 0.067) but incidence of lesions ≤10 mm was increased (41% in Gr3M versus 28% in Gr6M, P = 0.002). No difference in either HCC incidence (P = 0.13) or in prevalence of tumors ≤30 mm in diameter (79% versus 70%, P = 0.30) was observed between the randomized groups. Conclusion: US surveillance, performed every 3 months, detects more small focal lesions than US every 6 months, but does not improve detection of small HCC, probably because of limitations in recall procedures. (HEPATOLOGY 2011;)
In Western countries, hepatocellular carcinoma (HCC) occurs in more than 90% of cases in patients with chronic liver diseases, most often at the cirrhosis stage. Prognosis remains very poor due to late diagnosis and the associated cirrhosis, often precluding curative treatment.1 Currently, a major goal is to detect HCC at an early stage, when curative treatments can apply. Curable HCC is usually defined as either one tumor measuring ≤50 mm in diameter, or 2-3 tumors ≤30 mm in diameter without vascular extension or metastasis (Milan criteria),1 even though these criteria can be controversial.2 The most favorable results in terms of tumor destruction and local recurrence, by far, are observed for single tumors ≤30 mm in diameter, especially in patients treated by percutaneous ablation.3 Patients with small HCC tumors are usually asymptomatic and early detection needs active surveillance. Patients with cirrhosis are the main target population as recommended by international guidelines,1, 4, 5 even though surveillance is also recommended for patients with chronic liver disease without cirrhosis, such as hepatitis B virus (HBV) chronic hepatitis.6
Clinical effectiveness of the surveillance policy in cirrhotic patients has not been demonstrated. A randomized trial, performed in China, which included almost 20,000 patients (mainly with chronic HBV infection), found a significant survival benefit from biannual surveillance (mortality decreased by 37%), although compliance was relatively low (58.2%).7 It is unlikely that further randomized trials that compare surveillance versus no surveillance can be performed in the future due to obvious ethical considerations. However, some data indirectly suggest that surveillance is effective in patients with cirrhosis. In the most recent studies, HCC was detected at an early stage in up to 70% of patients submitted to regular surveillance.8 Several recent cost-effectiveness studies have concluded that surveillance is a cost-effective procedure in high-risk patients.9, 10 Additionally, a recent retrospective study found that surveillance performed between 1998 and 2004 was more effective than during the period 1991-1997, and resulted in better survival, probably due to the increased performance of curative treatments.11
The modalities of surveillance in cirrhotic patients are still controversial. In 2000 international guidelines recommended performing periodic ultrasonography (US) as well as a serum alpha-fetoprotein (AFP) assay, even if doubts concerning usefulness of this latter biomarker were clearly expressed.1, 4, 5 US is probably the most appropriate imaging procedure, as it is noninvasive and cheap, even though its sensitivity is considered relatively low.8 When US use is not technically valid (often due to obesity), there is no consensus on the best substitution: i.e., computed tomography (CT) scan or magnetic resonance imaging (MRI).12 Although a serum AFP assay is routinely used, this test is considered to have a low surveillance value due to the high rates of false-positive and -negative results.5, 13 The best period of periodicity for surveillance is also controversial, ranging from every 3 months to every 12 months. In 2000, international guidelines recommended surveillance performed every 6 months on an empirical basis.4 A recent study (not available when the trial was designed) suggests that a 12-month interval between each examination results in lower survival and HCC detection than a 6-month period.14
The main objective of periodic surveillance in cirrhotic patients is to detect HCC at an early stage when it is possible to offer a curative treatment option.15 It could be postulated that shortening the interval between each surveillance assessment could result in better detection of small HCC tumors, permit more curative treatments, and, consequently, improve survival. Accordingly, this multicenter randomized trial aimed to compare a 3-month periodicity of US versus a 6-month period, which is considered the benchmark interval.
AFP, alpha-fetoprotein; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; US: ultrasonography.
Patients and Methods
The promoter of this trial was the Assistance Publique-Hopitaux de Paris. The trial was funded by the French Ministry of Health (PHRC 1998 and 2003) and the French Ligue de Recherche contre le Cancer. The protocol obtained approval from the Ethics Committee (CCPPRB, Aulnay-sous-Bois, France). All patients gave written informed consent to participate in the trial. The trial was performed according to Consort recommendations16 and registered on the ClinicalTrials.gov website ( http://clinicaltrials.gov/ct2/show/NCT00190385).
Selection of Patients.
Patients were recruited from clinical centers belonging to a cooperative group (Supporting Appendix), which included 43 specialist liver disease centers in France and Belgium. Preinclusion assessment included the usual clinical and biological parameters; a US Doppler examination was also undertaken to check inclusion and noninclusion criteria.
Patients with all the following criteria were selected for inclusion in the trial: (1) age older than 18 years; (2) histologically proven cirrhosis, whatever the time of biopsy; (3) cirrhosis related to either excessive alcohol consumption (80 g per day in males and 60 g per day in females for at least 10 years), chronic infection with hepatitis C virus (HCV) (serum anti-HCV antibodies-positive) or hepatitis B virus (HBV) (serum hepatitis B surface antigen (HBsAg)-positive), or hereditary hemochromatosis (liver-iron overload and C282Y homozygosity); (4) absence of previous complications of cirrhosis (particularly ascites, gastrointestinal hemorrhage or HCC); (5) patients belonging to Child-Pugh class A or B and without a focal liver lesion at inclusion; and (6) written informed consent.
Patients with at least one of the following criteria were not included in the study: (1) patients belonging to Child-Pugh class C; (2) severe uncontrolled extrahepatic disease resulting in estimated life expectancy of less than 1 year; and (3) coinfection with human immunodeficiency virus (HIV), even if controlled by an antiviral treatment.
Design
As stated in the protocol (http://clinicaltrials.gov/ct2/show/NCT00190385), this was a multicenter, stratified (according to cirrhosis etiology and center), randomized clinical trial conducted in France and Belgium (43 sites), based on a two-by-two factorial design with balanced randomization, to compare two US periodicities (3 months versus 6 months) simultaneously, and to assess the value of the serum AFP assay (no assay versus assay every 6 months). After checking selection criteria and written consents, patients were randomized into one of four groups: US and a serum AFP assay every 6 months; US every 3 months and a serum AFP assay every 6 months; US every 6 months and no serum AFP assay; and US every 3 months and no serum AFP assay.
Randomization
Randomization was computer-generated, with allocation concealed using a centralized phone procedure to the data-management center (DBIM, Saint-Louis Hospital, Paris, France).
Randomization sequence used a permuted block design with fixed block sizes of four (with trialists unaware of the block size), and a 1:1 allocation ratio. Randomization was stratified by recruitment site and by the main etiology of the cirrhosis, which distinguished three strata: excessive alcohol consumption (more than 80 g/d in men and 60 g/d in women for at least 10 years; negative serum HBsAg and HCV-antibodies; no hemochromatosis); HCV chronic infection (negative serum HBsAg and positive HCV antibodies; no hemochromatosis) whatever the alcohol consumption; and other situations: HBV chronic infection (positive serum HBsAg) or hemochromatosis.
Follow-Up
Patients were seen by physicians at regular intervals, as established by randomization for US surveillance. The usual clinical and biological data were recorded at least once a year. Regular endoscopic surveillance was performed to detect esophageal varices and other portal hypertension-related lesions. In cases of esophageal varices, preventive therapy was recommended either by beta-blockers or endoscopic ligation, according to international recommendations.17
All events occurring during follow-up were recorded. Their management was performed according to international recommendations. In case of death the circumstances and likely cause(s) were recorded.
US Surveillance
Examination by Doppler US was performed every 6 months or 3 months according to randomization. For a given patient it was recommended to perform US in the same center by the same experienced operator. A standardized report was completed by each operator, mentioning the presence or not of focal liver lesions. In cases of focal lesions, echogenicity, number and diameter of nodules (classified as ≤10 mm, 11-20 mm, 21-30 mm, 31-50 mm, or ≥51 mm), and anatomic localization according to Couinaud were reported. Portal vasculature (main trunk and branches), hepatic veins, and vena cava were systematically examined.
HCC Diagnosis and Treatment
In cases of focal liver lesions a diagnostic procedure using contrast-enhanced imaging, a serum AFP assay, and/or a guided biopsy was performed according to the European Association for the Study of the Liver (EASL) guidelines, published in 2001.4 HCC diagnosis was established in the following situations: (1) histological proof of HCC; and (2) when a focal lesion was >2 cm in diameter, assessed by early arterial hypervascularization, using two contrast-enhanced methods (CT-scan, MRI, arteriography), or when there was an association between serum AFP level of >400 ng/mL plus early arterial hypervascularization, assessed by one contrast-enhanced method. In case of an increase in serum AFP level without liver focal lesion at US, a CT scan was performed according to recommendations.4 Subsequent modification of recommendations from the American Association for the Study of Liver Diseases (AASLD), published in 20051 and 2011,5 were not taken into account in this trial.
When an HCC diagnosis was established treatment was determined using a multidisciplinary approach at each medical center, by the physicians in charge of the patient. It was recommended to perform curative treatment (percutaneous ablation, resection, or transplantation) whenever possible.
Statistical Analyses: Sample Size Computation
The main objective of the trial focused on comparing differences between the US groups using a two-step procedure. The first step was based on the expected prevalence of the primary endpoint (HCC ≤30 mm in diameter) being 50% in the control group. From this we calculated that we would need 158 primary endpoint events to give 95% power to detect a significant difference between randomized groups, which corresponds to a 25% increased prevalence of HCC (with a one-sided type 1 error of 5%). Based on a 5% expected yearly incidence of HCC,18-20 within 3 years of follow-up, a sample size of at least 1,200 patients was computed to be needed. Inclusion was scheduled to continue into a second step if a significant benefit was found, on the basis of survival outcomes.
Statistical Methods
A modified intention-to-screen analysis was performed; that is, all patients were analyzed in the randomized groups, whether it applied or not, after excluding those with a focal hepatic lesion at inclusion. The date of the final analysis was set at 1 April 2008.
Comparison of the incidence of HCC tumors ≤30 mm in diameter in the randomized groups was based on Fisher's exact test. Cumulative-incidence curves were estimated using a competing-risk setting because of deaths that precluded the occurrence of focal lesions that included HCC. These were compared using Gray's test, whereas cause-specific Cox models, stratified according to randomization strata (cirrhosis etiology), allowed estimation of a hazard ratio (HR) with a 95% confidence interval (95% CI) as a measure of surveillance effect. Adjusted HRs were computed where the set of prognostic variables were first selected by a stepwise selection procedure in a multivariate model. Survival curves were estimated by the Kaplan-Meier method and then compared by the log-rank test.
Statistical analyses were performed using SAS 9.2 (Cary, NC) and R 2.10.1 (http://www.R-project.org) software. All tests were two-sided, with P ≤ 0.05 denoting statistical significance.
Results
Inclusion Period
The flow chart of the trial is presented on Fig. 1. Inclusion of patients started in June 2000 in the 43 participating clinical centers (Supporting Appendix). The minimal number of patients to include in the trial (n = 1,200) was reached in May 2005, allowing us to perform comparison between rates of HCC ≤30 mm in diameter for each group. At the first analysis (see below), it was decided by the steering committee to stop further inclusions into the trial by March 2006. At that time, 1,340 patients were included.
Among the 1,340 randomized patients, 62 were subsequently excluded from analysis after revision of individual data due to either immediate loss to follow-up (n = 12) or to the presence of a focal liver lesion at inclusion (n = 50). The focal lesions corresponded to HCC (n = 8), intrahepatic cholangiocarcinoma (n = 1), hemangioma (n = 15), and regenerative or indeterminate nodules (n = 26). Consequently, the final analyses were performed on 1,278 patients (Fig. 1).
Randomization
The 1,278 patients included in the final analyses were randomized into four groups: US plus an AFP assay every 6 months (n = 326), US every 3 months plus an AFP assay every 6 months (n = 328), US every 6 months but no AFP assay (n = 312), and US every 3 months but no AFP assay (n = 312). After data analyses, high rates of serum AFP assays were actually observed in the two latter groups (60.5% and 54.8%, respectively), which precluded reliable interpretation based on serum AFP assay randomization. Consequently, the steering committee decided to restrict the final analysis to US randomization only.
Accordingly, the final analysis considered only US randomization as follows: US every 3 months (n = 640, Gr3M) or US every 6 months (n = 638, Gr6M).
Baseline Characteristics of Patients
The main characteristics of patients at inclusion, according to US randomization, are reported in Table 1. Overall, patients were mainly males (69.1%), mean 55 years old, and belonged to Child-Pugh classes A, B, and C, at 87%, 12%, and 1%, respectively. The main causes of cirrhosis were excessive alcohol consumption, HCV infection, HBV infection, or hemochromatosis in 39.2%, 44.1%, 13.2%, and 1.6% of patients, respectively. Thirty-two (2.5%) patients had cirrhosis related to other etiologies, namely nonalcoholic steatohepatitis (n = 15), primary biliary cirrhosis (n = 2), autoimmune hepatitis (n = 5), and cryptogenetic cirrhosis (n = 10).
Follow-Up and Compliance with the Protocol
Mean follow-up was 47.1 months in Gr3M and 46.8 months in Gr6M (Table 2). Median time intervals between each US examination were in agreement with those scheduled for the randomization, 3 months for Gr3M and 6 months for Gr6M, with no significant time variations during the first 6 years of the trial in either group (data not shown). However, compliance was estimated as inadequate in 143 (11.9%) patients: 86 (14.6%) of Gr6M and 57 (9.4%) of Gr3M patients.
Focal Liver Lesions
Overall, a first focal lesion was observed in 358 patients (28%) during the trial: 192 in Gr3M and 166 in Gr6M patients (Table 2). The 5-year cumulated incidence was estimated as 34.1% (95% CI: 34.06-34.24). This was not significantly affected by randomization (35.5% in Gr3M compared to 32.8% in Gr6M; P = 0.067; Fig. 2). Similarly, the cumulative incidence of focal lesions ≤30 mm in diameter was not modified by 5-year estimates, at 30.1% in Gr3M versus 27.5% in Gr6M (P = 0.06; Fig. 2). An increased number of focal lesions ≤10 mm in diameter was observed in Gr3M compared to Gr6M (5-year cumulative incidence of 41% versus 28%, respectively; P = 0.002; Table 2, Fig. 2).
Table 3 reports the results of the prognostic analyses. Factors associated with outcome at the 5% level (alcoholic etiology of cirrhosis, age, body-mass index, platelet count, serum AST and ALT, and prothrombin activity) were introduced into a multivariate model. Only two variables were selected by the multivariate model, age and prothrombin activity. Adjusted HR of the focal lesion, stratified according to cirrhosis etiology, was estimated at 0.77 (95% CI: 0.62-0.96) in the Gr6M group compared to the Gr3M group (P = 0.02).
Overall, after the diagnostic procedures, most focal liver lesions detected during surveillance remained indeterminate (44.1%) or were considered regenerative (benign) nodules (8%) at the end of the trial (Table 2). A precise diagnosis was established in 152 patients (42.5%): HCC (n = 123), intrahepatic cholangiocarcinoma (n = 3), metastasis (n = 1), and hemangioma (n = 25) (Table 2). At the end of follow-up, only 19% of nodules ≤10 mm in diameter were confirmed as HCC, without a significant difference between the two groups (16 [22%] versus 6 [14%] for the Gr3M and Gr6M groups, respectively).
Hepatocellular Carcinoma
HCC was diagnosed in 123 patients (9.6%) during the trial: 53 in Gr3M and 70 in Gr6M (Table 2). The prevalence of HCC ≤30 mm in diameter was estimated at 79% (95% CI: 69-90%) in Gr3M and 70% in Gr6M (95% CI: 59-81%) (P = 0.30). The 5-year cumulative incidence of HCC was 11.9% (95% CI: 11.85-11.97), and was 10.0% in Gr3M versus 12.3% in Gr6M (P = 0.13) (Fig. 3). Similarly, there was no difference in the cumulative incidence of HCC ≤30 mm in diameter between the Gr3M and Gr6M groups (7.8% versus 9.1%, P = 0.48; Fig. 3). Additionally, no differences in the cumulative incidences of HCC ≤20 mm in diameter were observed between the two groups (Fig. 3).
The characteristics of HCC at diagnosis are reported in Table 4. Most tumors were uninodular (58.5%) and ≤30 mm in diameter (74%). In accordance with these results, portal obstruction and serum AFP levels >200 ng/mL at diagnosis were only observed in a small subset of patients (11.4% and 3.3%, respectively). Overall, 74.8% of patients with HCC were within the Milan criteria, and curative treatments were performed in 61% (Table 4). Only five patients had HCC ≤10 mm in diameter at diagnosis (Table 4).
Predictive factors for the occurrence of HCC were the alcoholic and HCV etiologies of cirrhosis, age, platelet count, serum bilirubin, AST, ALT, alkaline phosphatase, gamma-glutamyl-transpeptidase, albumin, prothrombin activity, and serum AFP (Table 5). When considered jointly in a multivariate model, three variables remained associated with the outcome: age, platelet count, and serum bilirubin. Adjusted HR, stratified according to the etiology of cirrhosis, in the Gr6M versus Gr3M groups, was estimated at 1.18 (95% CI: 0.82-1.72; P = 0.37).
Survival
Overall, 154 patients (12%) died during the trial: 72 (11.3%) in the Gr3M group and 82 (12.1%) in the Gr6M group (Table 2). No evidence of difference in survival between the randomized groups was observed regarding 5-year estimated survival at 84.9% versus 85.8% for the Gr3M and Gr6M groups, respectively (P = 0.38; Fig. 4). The main causes of deaths were HCC or cholangiocarcinoma (18.8%), liver failure (37.6%), extrahepatic cancer (8.9%), and severe bacterial infection (8.2%).
Discussion
The main goal of our trial was to compare the effectiveness of US surveillance according to the time interval between two examinations: 3 months versus 6 months. A second goal was to assess the importance of serum AFP in this surveillance.13 The latter part of this study was rapidly abandoned, as serum AFP assays were inadequately prescribed in more than half of the patients within the nonsurveillance group. Therefore, the steering committee considered such a high rate as an intolerable deviation to the protocol, and this precluded any reliable analysis.
Conversely, the compliance of patients toward US surveillance was generally adequate, as shown in Table 2, and the observed periodicities of US examinations were close to those scheduled. This allowed us to conclude that, in our population of cirrhotic patients, US surveillance performed every 3 months did not improve either the rate of detection of small HCCs eligible for curative treatment or the overall survival rates compared to patients undergoing US surveillance every 6 months. When the trial was designed in 1998-2000, HCC below 30 mm in diameter was widely considered an adequate limit for small HCC and therefore was chosen as the main criterion for the trial. It is currently recognized that 20 mm in diameter is a more reliable limit for small HCC,5 but again the incidence of such nodules was not increased in the 3-month group (Fig. 3).
A further result from our trial was that US surveillance every 3 months increased the cumulative incidence of detected focal lesions, although not significantly, thereby increasing the cost of recall procedures. At 2 years, focal lesions were detected in more than 20% of the 3-month group versus =13% of the 6-month group; most lesions proved nonmalignant during the follow-up. Moreover, this increase was mainly related to a significantly higher number of lesions ≤10 mm in diameter (Table 2, Fig. 2). Such nodules represented 41% of focal lesions in the 3-month group versus 28% in the 6-month group. Interestingly, the number of detected nodules sized ≤20 mm in diameter was similar between the two groups: 81% in the 3-month group versus 78% in the 6-month group (Table 2). This suggests that performing US at shorter intervals than 6 months allowed us to only detect a higher rate of very small nodules (≤10 mm in diameter), for which recall policies according to current guidelines usually fail to achieve a definite diagnosis and are considered not indicated.5 This might be expected owing to the lead-time bias that incurs in the shorter interval.
Most of the detected focal lesions were followed according to the EASL recommendations4 but, even when malignant, their earlier detection did not lead to earlier diagnosis or treatment. It is currently admitted that, for very small nodules, the sensitivity and specificity of elevated serum AFP is low,5, 13 that contrast-enhanced imaging only demonstrates a typical HCC pattern in a minority of cases,21, 22 and that a US-guided biopsy provides a high rate of false-negative results.1, 23 Therefore, in many cases the putative lesions are kept under imaging surveillance for several months, which precludes the potential benefit of early detection. It is noteworthy that, in our trial, a very low rate of HCCs were diagnosed that were ≤10 mm in diameter (4%, Table 4) in contrast to the high number of nodules detected below this size (Table 2). Interestingly, at the end of the trial, and despite a long follow-up, about 45% of the detected nodules either disappeared or were considered to be of indeterminate nature (Table 2).
Our conclusions apply only to those conditions in which the US surveillance was tested in our study:
inclusion of French and Belgian patients with cirrhosis caused mostly by HCV or alcohol, and application of the guidelines of the first Barcelona conference endorsed by EASL.4 Although other international guidelines have been proposed that allow noninvasive diagnosis of HCC by radiological means, for nodules between 10 and 20 mm in diameter1, 5 it is likely that application of these new guidelines would not have modified the results of our trial, as only a minority of nodules between 10 and 20 mm diameter exhibited a typical vascular pattern using two different imaging techniques.21 Conversely, the cause of the underlying liver disease could have influenced the results. The data from our double multivariate analysis of predictive factors (Tables 3, 5) show that more focal lesions were discovered in patients with alcoholic cirrhosis in contrast to a higher rate of HCC in patients with HCV cirrhosis. This suggests that "false lesions" could be more common in patients with alcoholic liver disease. Our data do not allow the likely explanation that irregular steatosis sparing of some cirrhotic nodules can be seen by contrast imaging with hypoechoic.24 It is likely that the significance of a small nodule (particularly those ≤10 mm in diameter) is not similar according to the cause of cirrhosis. This allows us to speculate that, in countries where HCV etiology of cirrhosis is predominant, such as Japan, results could have been different with a lower rate of "false positive" lesions.
The fact that focal lesions that were not eventually characterized as HCC were more numerous in the 3-month group, although not significantly, is clearly a disadvantage, as it enhances the rate of recall procedures, which leads to increased costs, increased stress for patients, and a lassitude that could lead to demotivation for surveillance.
Our study also shows the limitations of the current surveillance policy. At diagnosis, a significant proportion of patients had an infiltrative tumor (10%), more than three nodules (9%), or vascular involvement (11%), and less than 75% had a well-limited nodule ≤30 mm in diameter. In addition, about 25% had a tumor burden beyond the Milan criteria. This emphasizes the high prevalence of multicentric hepatocarcinogenesis, but also the need for improving diagnostic procedures and surveillance methods. Moreover, knowing the main predictive factors for HCC in patients with cirrhosis, such as age, gender, body-mass index, platelet count, and basal serum AFP level, as well as the etiology of cirrhosis,25 it is tempting to interpret the significance of a newly seen echographic nodule according to these easily recordable criteria. Therefore, we need to refine the current probabilistic approach, which, up to now, has relied mainly on radiological means and remains poorly sensitive to small nodules.
In conclusion, US surveillance performed every 3 months in patients with cirrhosis, mainly caused by HCV or alcohol abuse, fails to improve the detection rate of HCCs ≤30 mm in diameter that are eligible for curative treatment, although it detects more focal lesions than US performed every 6 months. This negative result is probably linked to the limitations of the recommended diagnostic procedures for small focal lesions in current practice.
Ultrasonographic surveillance of hepatocellular carcinoma in cirrhosis: A randomized trial comparing 3- and 6-month periodicities
Jean-Claude Trinchet,1,2 Cendrine Chaffaut,3,4 Vale rie Bourcier,1,2 Francoise Degos,5,6 Jean Henrion,7He lene Fontaine,8,9 Dominique Roulot,2,10 Ariane Mallat,11,12 Sophie Hillaire,13 Paul Cales,14Isabelle Ollivier,15 Jean-Pierre Vinel,16 Philippe Mathurin,17 Jean-Pierre Bronowicki,18 Vale rie Vilgrain,6,19,20Gise le N'Kontchou,1,2 Michel Beaugrand,1,2 Sylvie Chevret,3,4and for the Groupe d'Etude et de Traitement du Carcinome He patocellulaire (GRETCH)
View Full Article with Supporting Information (HTML
Hepatology
Volume 54, Issue 6, pages 1987–1997, December 2011
Abstract
Detection of small hepatocellular carcinoma (HCC) eligible for curative treatment is increased by surveillance, but its optimal periodicity is still debated. Thus, this randomized trial compared two ultrasonographic (US) periodicities: 3 months versus 6 months. A multicenter randomized trial was conducted in France and Belgium (43 sites). Patients with histologically proven compensated cirrhosis were randomized into two groups: US every 6 months (Gr6M) or 3 months (Gr3M). For each focal lesion detected, diagnostic procedures were performed according to European Association for the Study of the Liver guidelines. Cumulative incidence of events was estimated, then compared using Gray's test. The prevalence of HCC ≤30 mm in diameter was the main endpoint. A sample size of 1,200 patients was required. A total of 1,278 patients were randomized (Gr3M, n = 640; Gr6M, n = 638; alcohol 39.2%, hepatitis C virus 44.1%, hepatitis B virus 12.5%). At least one focal lesion was detected in 358 patients (28%) but HCC was confirmed in only 123 (9.6%) (uninodular 58.5%, ≤30 mm in diameter 74%). Focal-lesion incidence was not different between Gr3M and Gr6M groups (2-year estimates, 20.4% versus 13.2%, P = 0.067) but incidence of lesions ≤10 mm was increased (41% in Gr3M versus 28% in Gr6M, P = 0.002). No difference in either HCC incidence (P = 0.13) or in prevalence of tumors ≤30 mm in diameter (79% versus 70%, P = 0.30) was observed between the randomized groups. Conclusion: US surveillance, performed every 3 months, detects more small focal lesions than US every 6 months, but does not improve detection of small HCC, probably because of limitations in recall procedures. (HEPATOLOGY 2011;)
In Western countries, hepatocellular carcinoma (HCC) occurs in more than 90% of cases in patients with chronic liver diseases, most often at the cirrhosis stage. Prognosis remains very poor due to late diagnosis and the associated cirrhosis, often precluding curative treatment.1 Currently, a major goal is to detect HCC at an early stage, when curative treatments can apply. Curable HCC is usually defined as either one tumor measuring ≤50 mm in diameter, or 2-3 tumors ≤30 mm in diameter without vascular extension or metastasis (Milan criteria),1 even though these criteria can be controversial.2 The most favorable results in terms of tumor destruction and local recurrence, by far, are observed for single tumors ≤30 mm in diameter, especially in patients treated by percutaneous ablation.3 Patients with small HCC tumors are usually asymptomatic and early detection needs active surveillance. Patients with cirrhosis are the main target population as recommended by international guidelines,1, 4, 5 even though surveillance is also recommended for patients with chronic liver disease without cirrhosis, such as hepatitis B virus (HBV) chronic hepatitis.6
Clinical effectiveness of the surveillance policy in cirrhotic patients has not been demonstrated. A randomized trial, performed in China, which included almost 20,000 patients (mainly with chronic HBV infection), found a significant survival benefit from biannual surveillance (mortality decreased by 37%), although compliance was relatively low (58.2%).7 It is unlikely that further randomized trials that compare surveillance versus no surveillance can be performed in the future due to obvious ethical considerations. However, some data indirectly suggest that surveillance is effective in patients with cirrhosis. In the most recent studies, HCC was detected at an early stage in up to 70% of patients submitted to regular surveillance.8 Several recent cost-effectiveness studies have concluded that surveillance is a cost-effective procedure in high-risk patients.9, 10 Additionally, a recent retrospective study found that surveillance performed between 1998 and 2004 was more effective than during the period 1991-1997, and resulted in better survival, probably due to the increased performance of curative treatments.11
The modalities of surveillance in cirrhotic patients are still controversial. In 2000 international guidelines recommended performing periodic ultrasonography (US) as well as a serum alpha-fetoprotein (AFP) assay, even if doubts concerning usefulness of this latter biomarker were clearly expressed.1, 4, 5 US is probably the most appropriate imaging procedure, as it is noninvasive and cheap, even though its sensitivity is considered relatively low.8 When US use is not technically valid (often due to obesity), there is no consensus on the best substitution: i.e., computed tomography (CT) scan or magnetic resonance imaging (MRI).12 Although a serum AFP assay is routinely used, this test is considered to have a low surveillance value due to the high rates of false-positive and -negative results.5, 13 The best period of periodicity for surveillance is also controversial, ranging from every 3 months to every 12 months. In 2000, international guidelines recommended surveillance performed every 6 months on an empirical basis.4 A recent study (not available when the trial was designed) suggests that a 12-month interval between each examination results in lower survival and HCC detection than a 6-month period.14
The main objective of periodic surveillance in cirrhotic patients is to detect HCC at an early stage when it is possible to offer a curative treatment option.15 It could be postulated that shortening the interval between each surveillance assessment could result in better detection of small HCC tumors, permit more curative treatments, and, consequently, improve survival. Accordingly, this multicenter randomized trial aimed to compare a 3-month periodicity of US versus a 6-month period, which is considered the benchmark interval.
AFP, alpha-fetoprotein; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; US: ultrasonography.
Patients and Methods
The promoter of this trial was the Assistance Publique-Hopitaux de Paris. The trial was funded by the French Ministry of Health (PHRC 1998 and 2003) and the French Ligue de Recherche contre le Cancer. The protocol obtained approval from the Ethics Committee (CCPPRB, Aulnay-sous-Bois, France). All patients gave written informed consent to participate in the trial. The trial was performed according to Consort recommendations16 and registered on the ClinicalTrials.gov website ( http://clinicaltrials.gov/ct2/show/NCT00190385).
Selection of Patients.
Patients were recruited from clinical centers belonging to a cooperative group (Supporting Appendix), which included 43 specialist liver disease centers in France and Belgium. Preinclusion assessment included the usual clinical and biological parameters; a US Doppler examination was also undertaken to check inclusion and noninclusion criteria.
Patients with all the following criteria were selected for inclusion in the trial: (1) age older than 18 years; (2) histologically proven cirrhosis, whatever the time of biopsy; (3) cirrhosis related to either excessive alcohol consumption (80 g per day in males and 60 g per day in females for at least 10 years), chronic infection with hepatitis C virus (HCV) (serum anti-HCV antibodies-positive) or hepatitis B virus (HBV) (serum hepatitis B surface antigen (HBsAg)-positive), or hereditary hemochromatosis (liver-iron overload and C282Y homozygosity); (4) absence of previous complications of cirrhosis (particularly ascites, gastrointestinal hemorrhage or HCC); (5) patients belonging to Child-Pugh class A or B and without a focal liver lesion at inclusion; and (6) written informed consent.
Patients with at least one of the following criteria were not included in the study: (1) patients belonging to Child-Pugh class C; (2) severe uncontrolled extrahepatic disease resulting in estimated life expectancy of less than 1 year; and (3) coinfection with human immunodeficiency virus (HIV), even if controlled by an antiviral treatment.
Design
As stated in the protocol (http://clinicaltrials.gov/ct2/show/NCT00190385), this was a multicenter, stratified (according to cirrhosis etiology and center), randomized clinical trial conducted in France and Belgium (43 sites), based on a two-by-two factorial design with balanced randomization, to compare two US periodicities (3 months versus 6 months) simultaneously, and to assess the value of the serum AFP assay (no assay versus assay every 6 months). After checking selection criteria and written consents, patients were randomized into one of four groups: US and a serum AFP assay every 6 months; US every 3 months and a serum AFP assay every 6 months; US every 6 months and no serum AFP assay; and US every 3 months and no serum AFP assay.
Randomization
Randomization was computer-generated, with allocation concealed using a centralized phone procedure to the data-management center (DBIM, Saint-Louis Hospital, Paris, France).
Randomization sequence used a permuted block design with fixed block sizes of four (with trialists unaware of the block size), and a 1:1 allocation ratio. Randomization was stratified by recruitment site and by the main etiology of the cirrhosis, which distinguished three strata: excessive alcohol consumption (more than 80 g/d in men and 60 g/d in women for at least 10 years; negative serum HBsAg and HCV-antibodies; no hemochromatosis); HCV chronic infection (negative serum HBsAg and positive HCV antibodies; no hemochromatosis) whatever the alcohol consumption; and other situations: HBV chronic infection (positive serum HBsAg) or hemochromatosis.
Follow-Up
Patients were seen by physicians at regular intervals, as established by randomization for US surveillance. The usual clinical and biological data were recorded at least once a year. Regular endoscopic surveillance was performed to detect esophageal varices and other portal hypertension-related lesions. In cases of esophageal varices, preventive therapy was recommended either by beta-blockers or endoscopic ligation, according to international recommendations.17
All events occurring during follow-up were recorded. Their management was performed according to international recommendations. In case of death the circumstances and likely cause(s) were recorded.
US Surveillance
Examination by Doppler US was performed every 6 months or 3 months according to randomization. For a given patient it was recommended to perform US in the same center by the same experienced operator. A standardized report was completed by each operator, mentioning the presence or not of focal liver lesions. In cases of focal lesions, echogenicity, number and diameter of nodules (classified as ≤10 mm, 11-20 mm, 21-30 mm, 31-50 mm, or ≥51 mm), and anatomic localization according to Couinaud were reported. Portal vasculature (main trunk and branches), hepatic veins, and vena cava were systematically examined.
HCC Diagnosis and Treatment
In cases of focal liver lesions a diagnostic procedure using contrast-enhanced imaging, a serum AFP assay, and/or a guided biopsy was performed according to the European Association for the Study of the Liver (EASL) guidelines, published in 2001.4 HCC diagnosis was established in the following situations: (1) histological proof of HCC; and (2) when a focal lesion was >2 cm in diameter, assessed by early arterial hypervascularization, using two contrast-enhanced methods (CT-scan, MRI, arteriography), or when there was an association between serum AFP level of >400 ng/mL plus early arterial hypervascularization, assessed by one contrast-enhanced method. In case of an increase in serum AFP level without liver focal lesion at US, a CT scan was performed according to recommendations.4 Subsequent modification of recommendations from the American Association for the Study of Liver Diseases (AASLD), published in 20051 and 2011,5 were not taken into account in this trial.
When an HCC diagnosis was established treatment was determined using a multidisciplinary approach at each medical center, by the physicians in charge of the patient. It was recommended to perform curative treatment (percutaneous ablation, resection, or transplantation) whenever possible.
Statistical Analyses: Sample Size Computation
The main objective of the trial focused on comparing differences between the US groups using a two-step procedure. The first step was based on the expected prevalence of the primary endpoint (HCC ≤30 mm in diameter) being 50% in the control group. From this we calculated that we would need 158 primary endpoint events to give 95% power to detect a significant difference between randomized groups, which corresponds to a 25% increased prevalence of HCC (with a one-sided type 1 error of 5%). Based on a 5% expected yearly incidence of HCC,18-20 within 3 years of follow-up, a sample size of at least 1,200 patients was computed to be needed. Inclusion was scheduled to continue into a second step if a significant benefit was found, on the basis of survival outcomes.
Statistical Methods
A modified intention-to-screen analysis was performed; that is, all patients were analyzed in the randomized groups, whether it applied or not, after excluding those with a focal hepatic lesion at inclusion. The date of the final analysis was set at 1 April 2008.
Comparison of the incidence of HCC tumors ≤30 mm in diameter in the randomized groups was based on Fisher's exact test. Cumulative-incidence curves were estimated using a competing-risk setting because of deaths that precluded the occurrence of focal lesions that included HCC. These were compared using Gray's test, whereas cause-specific Cox models, stratified according to randomization strata (cirrhosis etiology), allowed estimation of a hazard ratio (HR) with a 95% confidence interval (95% CI) as a measure of surveillance effect. Adjusted HRs were computed where the set of prognostic variables were first selected by a stepwise selection procedure in a multivariate model. Survival curves were estimated by the Kaplan-Meier method and then compared by the log-rank test.
Statistical analyses were performed using SAS 9.2 (Cary, NC) and R 2.10.1 (http://www.R-project.org) software. All tests were two-sided, with P ≤ 0.05 denoting statistical significance.
Results
Inclusion Period
The flow chart of the trial is presented on Fig. 1. Inclusion of patients started in June 2000 in the 43 participating clinical centers (Supporting Appendix). The minimal number of patients to include in the trial (n = 1,200) was reached in May 2005, allowing us to perform comparison between rates of HCC ≤30 mm in diameter for each group. At the first analysis (see below), it was decided by the steering committee to stop further inclusions into the trial by March 2006. At that time, 1,340 patients were included.
Among the 1,340 randomized patients, 62 were subsequently excluded from analysis after revision of individual data due to either immediate loss to follow-up (n = 12) or to the presence of a focal liver lesion at inclusion (n = 50). The focal lesions corresponded to HCC (n = 8), intrahepatic cholangiocarcinoma (n = 1), hemangioma (n = 15), and regenerative or indeterminate nodules (n = 26). Consequently, the final analyses were performed on 1,278 patients (Fig. 1).
Randomization
The 1,278 patients included in the final analyses were randomized into four groups: US plus an AFP assay every 6 months (n = 326), US every 3 months plus an AFP assay every 6 months (n = 328), US every 6 months but no AFP assay (n = 312), and US every 3 months but no AFP assay (n = 312). After data analyses, high rates of serum AFP assays were actually observed in the two latter groups (60.5% and 54.8%, respectively), which precluded reliable interpretation based on serum AFP assay randomization. Consequently, the steering committee decided to restrict the final analysis to US randomization only.
Accordingly, the final analysis considered only US randomization as follows: US every 3 months (n = 640, Gr3M) or US every 6 months (n = 638, Gr6M).
Baseline Characteristics of Patients
The main characteristics of patients at inclusion, according to US randomization, are reported in Table 1. Overall, patients were mainly males (69.1%), mean 55 years old, and belonged to Child-Pugh classes A, B, and C, at 87%, 12%, and 1%, respectively. The main causes of cirrhosis were excessive alcohol consumption, HCV infection, HBV infection, or hemochromatosis in 39.2%, 44.1%, 13.2%, and 1.6% of patients, respectively. Thirty-two (2.5%) patients had cirrhosis related to other etiologies, namely nonalcoholic steatohepatitis (n = 15), primary biliary cirrhosis (n = 2), autoimmune hepatitis (n = 5), and cryptogenetic cirrhosis (n = 10).
Follow-Up and Compliance with the Protocol
Mean follow-up was 47.1 months in Gr3M and 46.8 months in Gr6M (Table 2). Median time intervals between each US examination were in agreement with those scheduled for the randomization, 3 months for Gr3M and 6 months for Gr6M, with no significant time variations during the first 6 years of the trial in either group (data not shown). However, compliance was estimated as inadequate in 143 (11.9%) patients: 86 (14.6%) of Gr6M and 57 (9.4%) of Gr3M patients.
Focal Liver Lesions
Overall, a first focal lesion was observed in 358 patients (28%) during the trial: 192 in Gr3M and 166 in Gr6M patients (Table 2). The 5-year cumulated incidence was estimated as 34.1% (95% CI: 34.06-34.24). This was not significantly affected by randomization (35.5% in Gr3M compared to 32.8% in Gr6M; P = 0.067; Fig. 2). Similarly, the cumulative incidence of focal lesions ≤30 mm in diameter was not modified by 5-year estimates, at 30.1% in Gr3M versus 27.5% in Gr6M (P = 0.06; Fig. 2). An increased number of focal lesions ≤10 mm in diameter was observed in Gr3M compared to Gr6M (5-year cumulative incidence of 41% versus 28%, respectively; P = 0.002; Table 2, Fig. 2).
Table 3 reports the results of the prognostic analyses. Factors associated with outcome at the 5% level (alcoholic etiology of cirrhosis, age, body-mass index, platelet count, serum AST and ALT, and prothrombin activity) were introduced into a multivariate model. Only two variables were selected by the multivariate model, age and prothrombin activity. Adjusted HR of the focal lesion, stratified according to cirrhosis etiology, was estimated at 0.77 (95% CI: 0.62-0.96) in the Gr6M group compared to the Gr3M group (P = 0.02).
Overall, after the diagnostic procedures, most focal liver lesions detected during surveillance remained indeterminate (44.1%) or were considered regenerative (benign) nodules (8%) at the end of the trial (Table 2). A precise diagnosis was established in 152 patients (42.5%): HCC (n = 123), intrahepatic cholangiocarcinoma (n = 3), metastasis (n = 1), and hemangioma (n = 25) (Table 2). At the end of follow-up, only 19% of nodules ≤10 mm in diameter were confirmed as HCC, without a significant difference between the two groups (16 [22%] versus 6 [14%] for the Gr3M and Gr6M groups, respectively).
Hepatocellular Carcinoma
HCC was diagnosed in 123 patients (9.6%) during the trial: 53 in Gr3M and 70 in Gr6M (Table 2). The prevalence of HCC ≤30 mm in diameter was estimated at 79% (95% CI: 69-90%) in Gr3M and 70% in Gr6M (95% CI: 59-81%) (P = 0.30). The 5-year cumulative incidence of HCC was 11.9% (95% CI: 11.85-11.97), and was 10.0% in Gr3M versus 12.3% in Gr6M (P = 0.13) (Fig. 3). Similarly, there was no difference in the cumulative incidence of HCC ≤30 mm in diameter between the Gr3M and Gr6M groups (7.8% versus 9.1%, P = 0.48; Fig. 3). Additionally, no differences in the cumulative incidences of HCC ≤20 mm in diameter were observed between the two groups (Fig. 3).
The characteristics of HCC at diagnosis are reported in Table 4. Most tumors were uninodular (58.5%) and ≤30 mm in diameter (74%). In accordance with these results, portal obstruction and serum AFP levels >200 ng/mL at diagnosis were only observed in a small subset of patients (11.4% and 3.3%, respectively). Overall, 74.8% of patients with HCC were within the Milan criteria, and curative treatments were performed in 61% (Table 4). Only five patients had HCC ≤10 mm in diameter at diagnosis (Table 4).
Predictive factors for the occurrence of HCC were the alcoholic and HCV etiologies of cirrhosis, age, platelet count, serum bilirubin, AST, ALT, alkaline phosphatase, gamma-glutamyl-transpeptidase, albumin, prothrombin activity, and serum AFP (Table 5). When considered jointly in a multivariate model, three variables remained associated with the outcome: age, platelet count, and serum bilirubin. Adjusted HR, stratified according to the etiology of cirrhosis, in the Gr6M versus Gr3M groups, was estimated at 1.18 (95% CI: 0.82-1.72; P = 0.37).
Survival
Overall, 154 patients (12%) died during the trial: 72 (11.3%) in the Gr3M group and 82 (12.1%) in the Gr6M group (Table 2). No evidence of difference in survival between the randomized groups was observed regarding 5-year estimated survival at 84.9% versus 85.8% for the Gr3M and Gr6M groups, respectively (P = 0.38; Fig. 4). The main causes of deaths were HCC or cholangiocarcinoma (18.8%), liver failure (37.6%), extrahepatic cancer (8.9%), and severe bacterial infection (8.2%).
Discussion
The main goal of our trial was to compare the effectiveness of US surveillance according to the time interval between two examinations: 3 months versus 6 months. A second goal was to assess the importance of serum AFP in this surveillance.13 The latter part of this study was rapidly abandoned, as serum AFP assays were inadequately prescribed in more than half of the patients within the nonsurveillance group. Therefore, the steering committee considered such a high rate as an intolerable deviation to the protocol, and this precluded any reliable analysis.
Conversely, the compliance of patients toward US surveillance was generally adequate, as shown in Table 2, and the observed periodicities of US examinations were close to those scheduled. This allowed us to conclude that, in our population of cirrhotic patients, US surveillance performed every 3 months did not improve either the rate of detection of small HCCs eligible for curative treatment or the overall survival rates compared to patients undergoing US surveillance every 6 months. When the trial was designed in 1998-2000, HCC below 30 mm in diameter was widely considered an adequate limit for small HCC and therefore was chosen as the main criterion for the trial. It is currently recognized that 20 mm in diameter is a more reliable limit for small HCC,5 but again the incidence of such nodules was not increased in the 3-month group (Fig. 3).
A further result from our trial was that US surveillance every 3 months increased the cumulative incidence of detected focal lesions, although not significantly, thereby increasing the cost of recall procedures. At 2 years, focal lesions were detected in more than 20% of the 3-month group versus =13% of the 6-month group; most lesions proved nonmalignant during the follow-up. Moreover, this increase was mainly related to a significantly higher number of lesions ≤10 mm in diameter (Table 2, Fig. 2). Such nodules represented 41% of focal lesions in the 3-month group versus 28% in the 6-month group. Interestingly, the number of detected nodules sized ≤20 mm in diameter was similar between the two groups: 81% in the 3-month group versus 78% in the 6-month group (Table 2). This suggests that performing US at shorter intervals than 6 months allowed us to only detect a higher rate of very small nodules (≤10 mm in diameter), for which recall policies according to current guidelines usually fail to achieve a definite diagnosis and are considered not indicated.5 This might be expected owing to the lead-time bias that incurs in the shorter interval.
Most of the detected focal lesions were followed according to the EASL recommendations4 but, even when malignant, their earlier detection did not lead to earlier diagnosis or treatment. It is currently admitted that, for very small nodules, the sensitivity and specificity of elevated serum AFP is low,5, 13 that contrast-enhanced imaging only demonstrates a typical HCC pattern in a minority of cases,21, 22 and that a US-guided biopsy provides a high rate of false-negative results.1, 23 Therefore, in many cases the putative lesions are kept under imaging surveillance for several months, which precludes the potential benefit of early detection. It is noteworthy that, in our trial, a very low rate of HCCs were diagnosed that were ≤10 mm in diameter (4%, Table 4) in contrast to the high number of nodules detected below this size (Table 2). Interestingly, at the end of the trial, and despite a long follow-up, about 45% of the detected nodules either disappeared or were considered to be of indeterminate nature (Table 2).
Our conclusions apply only to those conditions in which the US surveillance was tested in our study:
inclusion of French and Belgian patients with cirrhosis caused mostly by HCV or alcohol, and application of the guidelines of the first Barcelona conference endorsed by EASL.4 Although other international guidelines have been proposed that allow noninvasive diagnosis of HCC by radiological means, for nodules between 10 and 20 mm in diameter1, 5 it is likely that application of these new guidelines would not have modified the results of our trial, as only a minority of nodules between 10 and 20 mm diameter exhibited a typical vascular pattern using two different imaging techniques.21 Conversely, the cause of the underlying liver disease could have influenced the results. The data from our double multivariate analysis of predictive factors (Tables 3, 5) show that more focal lesions were discovered in patients with alcoholic cirrhosis in contrast to a higher rate of HCC in patients with HCV cirrhosis. This suggests that "false lesions" could be more common in patients with alcoholic liver disease. Our data do not allow the likely explanation that irregular steatosis sparing of some cirrhotic nodules can be seen by contrast imaging with hypoechoic.24 It is likely that the significance of a small nodule (particularly those ≤10 mm in diameter) is not similar according to the cause of cirrhosis. This allows us to speculate that, in countries where HCV etiology of cirrhosis is predominant, such as Japan, results could have been different with a lower rate of "false positive" lesions.
The fact that focal lesions that were not eventually characterized as HCC were more numerous in the 3-month group, although not significantly, is clearly a disadvantage, as it enhances the rate of recall procedures, which leads to increased costs, increased stress for patients, and a lassitude that could lead to demotivation for surveillance.
Our study also shows the limitations of the current surveillance policy. At diagnosis, a significant proportion of patients had an infiltrative tumor (10%), more than three nodules (9%), or vascular involvement (11%), and less than 75% had a well-limited nodule ≤30 mm in diameter. In addition, about 25% had a tumor burden beyond the Milan criteria. This emphasizes the high prevalence of multicentric hepatocarcinogenesis, but also the need for improving diagnostic procedures and surveillance methods. Moreover, knowing the main predictive factors for HCC in patients with cirrhosis, such as age, gender, body-mass index, platelet count, and basal serum AFP level, as well as the etiology of cirrhosis,25 it is tempting to interpret the significance of a newly seen echographic nodule according to these easily recordable criteria. Therefore, we need to refine the current probabilistic approach, which, up to now, has relied mainly on radiological means and remains poorly sensitive to small nodules.
In conclusion, US surveillance performed every 3 months in patients with cirrhosis, mainly caused by HCV or alcohol abuse, fails to improve the detection rate of HCCs ≤30 mm in diameter that are eligible for curative treatment, although it detects more focal lesions than US performed every 6 months. This negative result is probably linked to the limitations of the recommended diagnostic procedures for small focal lesions in current practice.
