Background and aims The effectiveness of surveillance for hepatocellular carcinoma (HCC) in the USA is largely unknown. The objective of this study was to evaluate the effectiveness of HCC surveillance in a national Veterans Administration (VA) practice setting, using the national VA hepatitis C virus (HCV) Clinical Case Registry.
Method The cohort consisted of 1480 HCV-infected patients who developed HCC during 1998–2007. The timing and intensity of receiving α-fetoprotein (AFP) and abdominal ultrasound (US) for HCC surveillance were evaluated. Overall mortality risk was examined using Cox proportional hazards regression models adjusting for demographics, clinical features and receipt of HCC-specific treatment.
Results The mean survival was 1.8 years following the HCC diagnosis date. Surveillance AFP or US were recorded in 77.8% of patients within 2 years prior to HCC diagnosis. Annual surveillance with both AFP and US was observed in only 2% of patients. The presence of either AFP or US surveillance during both 0–6 month and 7–24 month periods before HCC diagnosis was associated with a lower mortality risk (HR 0.71, 95% CI 0.62 to 0.82) compared with no surveillance. Receipt of two or more surveillance tests in the 0–6 months (HR 0.76 95% CI 0.66 to 0.88) and to a lesser extent in the 7–12 months (HR 0.81 95% CI 0.1 to 0.99) prior to HCC diagnosis was also associated with reduced mortality risk.
Conclusions Most patients with HCV-related cirrhosis do not receive regular imaging-based surveillance. The effectiveness of HCC surveillance tests in current clinical practice is rather modest in reducing HCC-related mortality.
- Hepatocellular carcinoma
- hepatocellular carcinoma
- liver imaging
Statistics from Altmetric.com
- Hepatocellular carcinoma
- hepatocellular carcinoma
- liver imaging
Significance of this study
What is already known about this subject?
Hepatocellular carcinoma (HCC) related to hepatitis C virus (HCV) infection has been rising.
Practice guidelines recommend HCC surveillance for high-risk patients to detect HCC at an early stage, when critical treatment can be applied.
The efficacy of HCC surveillance has been shown in a randomised controlled trial, and several non-randomised trials from referral centers have found a survival benefit in those identified with small and early tumours.
The effectiveness of surveillance for HCC in clinical practice is largely unknown.
What are the new findings?
We evaluated the timing and intensity of receiving α-fetoprotein (AFP) and abdominal ultrasound for HCC surveillance in a cohort of 1480 HCV-infected patients who developed HCC.
The presence of either AFP or ultrasound surveillance during the 2 years preceding HCC diagnosis was associated with a lower mortality risk compared with no surveillance.
Annual surveillance with both AFP and ultrasound was observed in only 2% of patients.
How might it impact on clinical practice in the foreseeable future?
Most patients with HCV-related cirrhosis do not receive regular imaging-based surveillance.
Interventions are required to improve implementation of HCC surveillance in practice.
Hepatocellular carcinoma (HCC) has become the fastest growing cause of cancer-related death in the USA.1 The increase in HCC is mostly attributable to a rise in hepatitis C virus (HCV)-related chronic liver disease,2 3 and it has been associated with a shift of disease distribution towards individuals of younger ages.4 Most patients with HCC are diagnosed at an advanced stage of disease when survival is poor (5-year survival <5%); however, a considerable improvement in survival has been observed (5-year survival is between 40% and 70%) in patients who receive potentially curative treatment (liver transplant, surgical resection or ablation).5
Practice guidelines recommend HCC surveillance for high-risk patients6–8 to detect HCC at an early stage, when critical treatment can be applied. The efficacy of HCC surveillance has been shown in a randomised controlled trial of nearly 19 000 hepatitis B virus (HBV)-infected patients in China, where HCC surveillance with serum α-fetoprotein (AFP) and abdominal ultrasound performed every 6 months was associated with a 37% reduction in HCC-related mortality.9 In addition, several non-randomised trials from referral centres have found a survival benefit in those identified with small and early tumours.10–12
While there is some evidence to support the efficacy of HCC surveillance in high-risk patients, the effectiveness of surveillance in clinical practice has not been determined. ‘Efficacy’ reflects the degree to which an intervention such as HCC surveillance produces the expected result under controlled conditions or in specialised centres chosen to maximise the likelihood of observing an effect if it exists. In contrast, ‘effectiveness’ addresses the extent to which the same intervention is beneficial when deployed in medical practice settings and broader populations.13 Given the repetitive nature of HCC surveillance, the need for timely recall and diagnostic evaluation, and application of relatively expensive and not widely available treatments, the effectiveness of HCC surveillance may fall short of the reported efficacy. Information about effectiveness of HCC surveillance is relevant to understanding patient outcomes, and to healthcare decisions of providers and patients.
The Veterans Administration (VA) healthcare system is the largest integrated healthcare system in the USA. We previously described the rates and determinants of HCC surveillance among HCV-infected patients.14–16 In this report, we have conducted a retrospective cohort study of all eligible HCV-infected patients who developed HCC to evaluate the association between HCC surveillance and survival in a large nationwide VA clinical practice setting.
Data sources and study population
Data for this study were obtained from the VA HCV Clinical Case Registry (CCR), which contains information collected from 128 VA healthcare facilities nationwide. Additional details of the CCR data are published elsewhere.17 The date of death was obtained from the VA vital status file. We identified HCV-infected patients with incident HCC between 1 October 1998 and 1 January 2007. All patients had HCV infection defined by one positive HCV antibody test combined with at least one HCV ICD-9-CM (International Classification of Diseases, Ninth Revision, Clinical Modification) code. We identified all patients who developed HCC using a previously validated algorithm based on the presence of ICD-9-CM code 155.0 and the absence of code 155.1.14 Patients who developed HCC or died within 12 months following their HCV index date were excluded. We only included patients in the study cohort if they had at least one inpatient or outpatient encounter at any VA facility within the 2 years prior to and 1 year following the date of HCC diagnosis.
We identified all AFP and abdominal ultrasound tests performed prior to HCC diagnosis. AFP tests were identified using laboratory data and abdominal ultrasounds were identified using CPT codes. Among these tests, we ascertained AFP or ultrasounds that were performed for HCC surveillance by applying a previously developed and validated algorithm that utilises laboratory data and ICD-9-CM diagnostic codes available in the VA HCV CCR database.18
The timing and intensity of HCC surveillance were examined among HCV-infected patients who developed HCC. The timing of HCC surveillance prior to the HCC diagnosis date was categorised as 0–6 months, 7–12 months, 13–18 months, 19–24 months and >24 months. The intensity of HCC surveillance was measured as 0, 1 or ≥2 surveillance tests during the time period of interest.
Information was collected on year of HCC diagnosis, demographics (age at HCC diagnosis, race and gender), liver-related characteristics (Model for End Stage Liver Disease (MELD) score and indicators of liver disease severity (cirrhosis, ascites, varices and encephalopathy)), number of VA inpatient/outpatient encounters, and the presence of co-existing medical conditions including alcohol use, cocaine use, HIV, HBV, anxiety, depression/ bipolar, psychosis, coronary artery disease, congestive heart failure, respiratory failure, diabetes, hypertension and co-morbidity summary score, as well as receipt of HCC-specific treatment.
Laboratory data were used to determine HBV surface antigen status, alcohol use (positive serum alcohol), cocaine and cannabis use (positive lab tests), and serum levels of creatinine, bilirubin and international normalised ratio (INR) within 6 months prior to or following the HCV (or cirrhosis) index date for the MELD.19 ICD-9 codes were used to define the rest of the variables. Cirrhosis was identified by one of several previously validated ICD-9-CM codes (571.2, 571.5 or 571.6) and patients were assigned a cirrhosis index date based on the first appearance of a code.20 19 The burden of co-morbidity was measured by the adapted Charlson co-morbidity index score.21 HCC treatment was ascertained by ICD-9 procedure and CPT codes for liver transplant, surgical resection, local ablation with alcohol or radiofrequency, as well as transarterial chemoembolisation.
Survival rates were estimated using a Kaplan–Meier approach. Patients who were alive at 30 September 2009 were censored. We examined the association between the exposure to HCC surveillance and time to death using a Cox proportional hazards regression model. The two main exposures of interest were timing and intensity of HCC surveillance. The demographic and clinical variables listed above were tested as potential determinants of survival. Variables with p<0.10 in univariable Cox regression model were further evaluated using a stepwise regression analysis to identify independent prognostic factors. HCC treatment was then added to this model to evaluate the explanatory effect of HCC treatment on potential effects of HCC surveillance.
To adjust for lead time bias, two approaches were used. First, the survival time of patients who received surveillance was analysed from the date of HCV diagnosis rather than the HCC diagnosis date. Secondly, we applied a parametric model proposed by Duffy, assuming an exponential distribution of the sojourn time. The expected additional follow-up time, s, due to lead time, that is for a patient with surveillance known to be dead at time t, and for a patient with surveillance known to be alive at time t.
The lead time was corrected by subtracting E(s) from the observed survival time. We assumed an HCC sojourn time (1/λ) of 70 or 140 days; the latter estimate approximates those in previous reports.22–24
To examine the robustness of the findings, we conducted three a priori sensitivity analyses: (1) survival analysis using Cox proportional hazards regression to examine HCC surveillance in a subgroup of patients with a diagnosis of cirrhosis recorded at least 1 year prior to HCC diagnosis6; (2) survival analysis using Cox proportional hazards regression to examine receipt of any AFP or abdominal ultrasound irrespective of the purpose in the 2 years that preceded HCC diagnosis; and (3) in a convenience subsample of Medicare-eligible patients in the 1999–2002 VA HCV CCR cohort, we calculated the proportions of HCC patients with additional AFP or abdominal ultrasound tests recorded in Medicare Part A and B claims in the 2 years prior to the HCC diagnosis date.25
The study protocol was approved by the Institutional Review Board of Baylor College of Medicine and the office of Human Subjects Research of the National Institutes of Health.
From 128 505 patients diagnosed with HCV in the VA HCV CCR during fiscal years 1998–2006, we identified 2128 patients who developed HCC during a follow-up period that ended in 1 January 2007. We excluded 498 patients because the HCC diagnosis was recorded within the first 12 months following HCV, and 13 patients who received transplant or resection before the recorded HCC diagnosis date.
The final study cohort comprised 1480 patients with HCC who fulfilled the inclusion and exclusion criteria, of whom 599 (40.5%) had a diagnosis of cirrhosis recorded at least 1 year prior to HCC diagnosis. Of the final study cohort, 1278 patients were tested for HCV RNA and 1209 (94.6%) were positive for HCV RNA. The mean age of patients was 58.1 years (SD±8.6) at the HCC diagnosis date, 54.2 years (SD±8.6) at the HCV index date and 53.4 years (SD±7.9) at the cirrhosis index date (if present). Most patients were men (99.3%), and the racial distribution was 55.6% white, 27.1% black, 11.6% other races and 5.7% unknown race. The mean duration of time between the HCV index date and the HCC diagnosis date was 1410.7 days (SD±717.0), and time between the cirrhosis index date and the HCC diagnosis date was 1292.0 days (SD±891.1).
Most patients with HCC (87.3%) in the study cohort died during a total follow-up of 2711.3 patient years (1.8 year mean duration). The overall median survival after HCC diagnosis was 262 days, and the 1-, 3- and 5-year survival rates were 43.4, 19.4 and 12.2%, respectively. Approximately 93% of patients with HCC received at least one AFP (89%) or ultrasound (78%) between HCV diagnosis and HCC diagnosis dates. However, only 21.2% ultrasound tests were deemed as surveillance tests. Therefore, we estimated that 83% of patients received at least one AFP or ultrasound test for HCC surveillance during a time period of up to 6 years before their HCC diagnosis. Most of these patients (77.6%) received at least one surveillance test in the 2 years preceding their HCC diagnosis. Among those who received any test, only 509 (34.4% of 1480) patients received annual testing in these 2 years, and among patients in the latter group only 29 (2.0% of 1480) patients received annual testing with both ultrasound and AFP (figure 1). In the 2 years before HCC diagnosis, 57.7% received surveillance with AFP only, 19.4% received both AFP and ultrasound and 0.5% received ultrasound only. Only 7.8% (116 of 1480) of patients with HCC had an abdominal CT scan recorded within 3 months to 2 years prior to HCC diagnosis, and most were performed during hospitalisations, and therefore we did not consider CT scans as part of HCC surveillance in the analyses.
We examined the association between survival and the timing and intensity of HCC surveillance tests received between the HCV index date and the HCC diagnosis date. We found no association with timing of surveillance tests received >2 years before HCC diagnosis irrespective of intensity (data not shown). Furthermore, there were very few patients with surveillance tests in all four 6-month periods (85, 5.7%), (figure 1), or even in three of the four 6-month periods (223, 15.1%). Therefore, the timing of surveillance was examined in groups of patients based on receiving tests in two consecutive periods before HCC diagnosis, 0–6 months only, 7–24 months only, both periods or no tests in the two periods (table 1). The longest survival was observed in those who received surveillance tests in both time periods, while the shortest survival was observed in patients who received no surveillance tests in either period. The differences were most striking for the 1-year survival rates (50.3% vs 31.9%), less pronounced for the 3-year (21.9% vs 15.4%) and not significant for the 5-year (13.0% vs 11.4%) survival rates.
The significant association between timing of surveillance tests and mortality risk was observed in unadjusted as well as adjusted survival analyses (table 2). There was a significant 29% reduction in mortality risk among those who received surveillance tests in both time periods (0–6 and 7–24 months) and 20% risk reduction in those who received surveillance tests in the 0–6 month period only compared with no tests in either period. There were no significant associations between mortality and surveillance tests recorded only during the 7–24 month period before HCC diagnosis. In multivariable analyses that adjusted for demographic and clinical features (table 2), the receipt of any surveillance test during both 0–6 months and 7–24 months remained significantly associated with a 16% reduction in mortality risk.
The intensity of HCC surveillance tests (receipt of ≥2 tests as compared with one or none) was also associated with a reduced risk of mortality during the 0–6 months prior to HCC diagnosis and to a lesser extent during the 7–12 months, but not in the 13–18 or 19–24 month periods. Receipt of only one surveillance test in any 6-month period was not significantly associated with survival (table 3). Receipt of ≥2 surveillance tests during the 0–6 month period prior to HCC diagnosis remained significantly associated with reduced mortality risk after adjustment for several demographic and clinical factors (table 3). Both timing and intensity of HCC surveillance became no longer significant when HCC treatment variables were added to the final models (tables 2 and 3).
Surveillance during both 0–6 months and 7–24 months remained independently associated with a reduced mortality risk in analyses with correction of lead time by calculating survival from the HCV diagnosis date or using the parametric model with sojourn time of 70 days. If sojourn time in the parametric model is assumed to be 140 days, the association between surveillance and mortality becomes non-significant (table 4).
As expected, older age, higher MELD scores and the presence of varices, encephalopathy or ascites were associated with a significant increase in mortality risk. There were no significant differences related to year of HCC diagnosis, race or gender (data not shown).
In a subgroup of 599 patients with cirrhosis diagnosis recorded at least 1 year prior to HCC diagnosis, 512 (85.5%) died during the study period, with a mean survival following HCC diagnosis of 681.5 days during 1117.6 person year follow-up (mean 1.9 years). The timing and intensity of HCC surveillance variables were generally associated with mortality risk of a similar magnitude and direction as those observed for the overall cohort. For example, receiving surveillance tests during 0–6 months and 7–24 months was associated with 24% risk reduction in unadjusted models (HR 0.76, 95% CI 0.60 to 0.95, p=0.02), which became no longer significant in the adjusted models without (HR 0.98, 95% CI 0.76 to 1.27, p=0.88) as well as with HCC treatment variables (HR 1.06, 95% CI 0.82 to 1.38, p=0.64). Intensity with ≥2 surveillance tests during the 0–6 month period was associated with 29% (HR 0.71, 95% CI 0.57 to 0.89, p<0.01) risk reduction in mortality in an unadjusted model, which became no longer significant in the adjusted models without (HR 0.83, 95% CI 0.65 to 1.06, p=0.14) and with HCC treatment (HR 0.86, 95% CI 0.67 to 1.11, p=0.25).
We examined the association between any AFP or ultrasound tests (irrespective of purpose) and mortality risk, and found that only performance of AFP or ultrasound during both 0–6 and 7–24 month periods was associated with a significant reduction in mortality risk (unadjusted HR 0.78, 95% CI 0.65 to 0.93, p<0.01). However, unlike the analyses for surveillance tests, receipt of any AFP or ultrasound either at 0–6 months (unadjusted HR 0.99; 95% CI 0.82 to 1.20, p=0.97) alone or at 7–24 months alone (unadjusted HR 0.94, 95% CI 0.47 to 1.20, p=0.62) was not associated with a significant change in mortality risk.
Lastly, among 1480 HCC patients, 262 (17.7%) patients had at least one record in Medicare inpatient, outpatient or physician files from 1999 to 2002. However, during the 2 years before HCC diagnosis, AFP or ultrasound claims were found in only 34 patients. In the 0–6 months before HCC, additional AFP and ultrasound tests were found in only 12 and 14 patients, respectively. In the 7–24 months before HCC diagnosis additional AFP and ultrasound tests were found in 4 and 9 patients, respectively.
This study examined the association between HCC surveillance and survival among newly diagnosed HCC cases in HCV-infected patients in VA clinical practice. Of those who received any HCC surveillance during the 2 years preceding HCC diagnosis, only 2.0% received annual testing with both AFP and ultrasound, while most received inconsistent surveillance, with the singular use of AFP as the most frequent modality. Nevertheless, repeated and frequent performance of surveillance tests (AFP or abdominal ultrasound) in the 2 years preceding HCC diagnosis was associated with a modest (20–30%) reduction in mortality risk following HCC diagnosis. The possible survival benefit related to surveillance was explained only by increased receipt of potentially palliative treatments in patients who received surveillance. The findings persisted in a subgroup of patients in whom cirrhosis was recorded at least 1 year prior to HCC diagnosis.
The findings were sensitive to the assumed sojourn time, which is the time interval between the onset of the detectable preclinical state and the point of progress to the clinical state detectable by routine methods of diagnosis. Lead time is defined as the time gained by diagnosing the disease using special detection modalities before the patient experiences symptoms, and it may vary from a portion of time, up to the full sojourn time. In this study, with an assumed sojourn time of 140 days (3.5 months), the association between surveillance and mortality risk reduction became no longer significant. Thus one cannot be completely confident that the observed prolonged survival is not at least partly related to lead time bias. Only randomised controlled trials can achieve this level of confidence. However, surveillance was also associated with increased receipt of HCC treatment, which was found in the multivariable analyses to be the main explanation for the survival benefit in the surveillance group.
This observational non-randomised study evaluated the effectiveness (rather than efficacy) of HCC surveillance. Effectiveness takes into account the benefits and harms of an intervention, and therefore can be more relevant to healthcare decisions of providers and patients as well as policy evaluation. It takes into account external factors such as individual patient characteristics, health system features and societal influences in a real-world clinical practice.26 However, the testing observed in this study does not mirror almost to any degree the HCC surveillance guidelines recently updated by the American Association for the Study of Liver Diseases (AASLD); notably these guidelines drop the use of AFP (in lieu of ultrasound only) and also drop the option for yearly frequency (in lieu of 6 months), the latter choice supported by recent observational studies.23 The effectiveness of implementing such recommendations needs to be tested in clinical practice, and cannot be deduced from this study.
Only a very small number of patients underwent a regular 6 monthly surveillance with either AFP or ultrasound (5–7%), or annual testing (2%) with both AFP and ultrasound. Therefore, the analyses in this study focused on comparing patients who received any type of surveillance with those who did not receive surveillance, and the main findings pertained to either AFP or ultrasound during two unequal periods (0–6 months and 7–24 months); the latter group constituted only 39.2% of the study population. The finding that only a minimal proportion of at-risk VA patients was kept under an actual surveillance (provided that a precise recall policy was followed) shows that, despite the recommendations of practice guidelines and what was declared by gastroenterologists/hepatologists,27 surveillance has not yet gained enough credit to be regularly applied.
There are several likely obstacles in the face of an effective HCC surveillance programme. These include the repetitive nature of testing over relatively short periods of time coupled with the need for recall strategies, somewhat complicated diagnostic evaluation and the limited availability of potentially curative treatment. While all of these factors might be at play in this study, the striking deficit seems to be in the implementation of regular repetitive testing using a combination of ultrasound and AFP.
Factors other than poor adoption by healthcare providers may explain the relatively poor effectiveness of HCC surveillance. The presence of medical and psychological co-morbid disorders could have affected receipt of surveillance, receipt of HCC treatment, as well as survival. We relied on diagnostic codes to adjust for several conditions, but the severity or reversibility of these co-morbidities could not be accurately ascertained. Additional unmeasured patient characteristics (eg, adherence, education) could also have affected or explained part of the findings.
In addition to being used for HCC surveillance, AFP and abdominal ultrasound tests could have been performed for diagnosing HCC. HCC surveillance tests cannot be directly identified from administrative data. We developed and validated an algorithm with good predictive value to identify both AFP and ultrasound tests performed for surveillance purposes.18 In addition, we also presented the results of all AFP and ultrasound performed irrespective of purpose, and the associations were of lower magnitude than those with surveillance tests, thus lending some credence to the designations employed in this study.
The study examined HCV-infected veterans who use the VA healthcare systems, and therefore the generalisability of the findings to patients with other risk factors for HCC, non-veterans or women is unknown. In addition, patients who use the VA healthcare system may receive care outside the VA using supplemental insurance or Medicare benefits. However, we examined a subset of VA–Medicare data and determined that a small percentage of additional tests were performed in these patients and therefore would probably not affect our findings.
In conclusion, the effectiveness of HCC surveillance as practised during this study period in reducing mortality in patients with HCC in clinical practice is low. Low rates of regular surveillance coupled with inconsistent implementation of regular imaging-based surveillance may be the main explanation. The very small proportion of patients who received regular surveillance seem to have had a modest gain in survival.
Funding This work was supported in part by the Houston VA HSR&D Center of Excellence (HFP90-020), National Institute of Diabetes and Digestive and Kidney Disease, Center grant DK56338, and the National Cancer Institute (R01-CA-125487). In addition, JRK is the recipient of a VA Health Services and Development MREP award (MRP05-305). The content is solely the responsibility of the authors and does not necessarily represent the official views of the Department of Veterans Affairs, Baylor College of Medicine, The University of Texas School of Public Health or the National Cancer Institute.
Competing interests None.
Ethics approval This study was conducted with the approval of the Baylor College of Medicine IRB.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.