Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Differences in hepatocellular carcinoma risk, predictors and trends over time according to etiology of cirrhosis

  • George N. Ioannou ,

    Roles Conceptualization, Formal analysis, Funding acquisition, Methodology, Project administration, Writing – original draft, Writing – review & editing

    georgei@medicine.washington.edu

    Current address: Department of Medicine, Veterans Affairs Puget Sound Healthcare System, Seattle, Washington, United States of America

    Affiliations Division of Gastroenterology, Veterans Affairs Puget Sound Healthcare System and University of Washington, Seattle, Washington, United States of America, Department of Medicine, Veterans Affairs Puget Sound Healthcare System and University of Washington, Seattle, Washington, United States of America, Health Services Research and Development, Veterans Affairs Puget Sound Healthcare System, Seattle, Washington, United States of America

  • Pamela Green,

    Roles Data curation, Formal analysis, Methodology

    Affiliation Health Services Research and Development, Veterans Affairs Puget Sound Healthcare System, Seattle, Washington, United States of America

  • Elliott Lowy,

    Roles Data curation, Formal analysis, Methodology

    Affiliation Health Services Research and Development, Veterans Affairs Puget Sound Healthcare System, Seattle, Washington, United States of America

  • Elijah J. Mun,

    Roles Writing – original draft, Writing – review & editing

    Current address: Department of Medicine, University of Washington, Seattle, Washington, United States of America

    Affiliation Department of Internal Medicine, University of Washington, Seattle, Washington, United States of America

  • Kristin Berry

    Roles Formal analysis, Investigation, Methodology, Writing – review & editing

    Affiliation Health Services Research and Development, Veterans Affairs Puget Sound Healthcare System, Seattle, Washington, United States of America

Abstract

Background and aims

Hepatocellular carcinoma (HCC) risk is high in cirrhosis. We sought to describe differences in HCC risk, predictors and trends over time according to etiology of cirrhosis.

Methods

We identified 116,404 patients with cirrhosis diagnosed between 2001–2014 in the VA healthcare system and determined incident HCC cases occurring from the date of cirrhosis diagnosis until 01/31/2017. Patients were divided by cirrhosis etiology into hepatitis C virus (HCV, n = 52,671), alcoholic liver disease (ALD, n = 35,730), nonalcoholic fatty liver disease (NAFLD, n = 17,354), or OTHER (n = 10,649).

Results

During a mean follow-up of 4.3 years, 10,042 new HCC cases were diagnosed. Patients with HCV had >3 times higher incidence of HCC (3.3 per 100 patient-years) than patients with ALD (0.86/100 patient-years), NAFLD (0.90/100 patient-years) or OTHER (1.0/100 patient-years), an association that persisted after adjusting for baseline characteristics. HCC incidence was 1.6 times higher in patients with cirrhosis diagnosed in 2008–2014 (2.47/100 patient-years) than in 2001–2007 (1.55/100 patient-years).

Independent predictors of HCC among all cirrhosis etiologies included: age, male sex, Hispanic ethnicity, high serum alpha fetoprotein, alkaline phosphatase and AST/√ALT ratio and low serum albumin and platelet count. Diabetes was associated with HCC in ALD-cirrhosis and NAFLD-cirrhosis, and BMI in ALD-cirrhosis.

Conclusions

HCC risk is 3 times greater in cirrhotic patients with HCV than ALD or NAFLD. HCC risk continues to increase over time in analyses extending to 2017 in cirrhosis of all etiologies. Multiple readily available risk factors for HCC were identified that were influenced by cirrhosis etiology and could be used to develop HCC risk estimation models.

Introduction

Patients with cirrhosis have significantly increased risk of hepatocellular carcinoma (HCC) ranging from 1 to 8% per year[1]. The three most common etiologies of cirrhosis in the United States, which also account for the majority of HCC cases, are hepatitis C virus (HCV), alcoholic liver disease (ALD) and nonalcoholic fatty liver disease (NAFLD).

The risk of HCC appears to be higher in patients with HCV-related cirrhosis than cirrhosis related to ALD, NAFLD or other etiologies[24]. It is unclear whether this difference persists after adjustment for baseline characteristics that are risk factors for HCC. Furthermore, some HCC risk factors may be different in magnitude depending on the underlying liver disease responsible of cirrhosis, such as HCC risk factors that are particularly linked to a specific liver disease (e.g. obesity and diabetes, which are strongly associated with NAFLD) or risk factors that are only relevant for a specific liver disease (e.g. HCV genotype). Describing the most important HCC risk factors by cirrhosis etiology is important for the future development of HCC risk prediction models that can be used in clinical practice.

The distribution of cirrhosis etiologies is expected to change dramatically as HCV-cirrhosis begins to decline and NAFLD-cirrhosis continues to increase. Although HCC incidence appears to continue to increase over time[5], it is unclear whether HCC risk is still increasing among patients with cirrhosis caused by different underlying liver diseases.

We aimed to compare different liver diseases (HCV, ALD and NAFLD) with respect to HCC risk in cirrhotic patients and to determine the most important risk factors for HCC separately for patients with HCV, ALD or NAFLD-cirrhosis. We also aimed to compare trends over time in the incidence of HCC among patients with HCV, ALD or NAFLD-cirrhosis.

Methods

Data source and study population

The Veterans Health Administration is the largest integrated healthcare system in the United States, providing care at 168 VA Medical Centers and 1053 outpatient clinics and serving more than 8.9 million Veterans each year as of 2016[6]. We identified all patients whose cirrhosis was diagnosed for the first time during the 14-year period from 01/01/2001 to 12/31/2014 in the VA national healthcare system (n = 136,050) and retrospectively followed these patients until 01/31/2017 for incident HCC. Patients with a history of HCC before or within 90 days of the diagnosis of cirrhosis (n = 6531) and patients with a diagnosis of cirrhosis only after liver transplantation (n = 934) were excluded. Also, patients who died or underwent liver transplantation within 90 days of the diagnosis of cirrhosis or who had fewer than 90 days of follow-up (n = 12,181) were excluded leaving 116,404 patients in the current analysis. Data extended backward to October 1999 in order to allow determination of baseline characteristics and comorbidities and forward to 01/31/2017 so that all patients would have at least 25 months of potential follow-up for the development of HCC.

The VA uses a single, nationwide, comprehensive electronic healthcare information network (known as the Veterans Information Systems and Technology Architecture or VistA), which consists of nearly 180 applications of clinical, financial, administrative and infrastructure needs integrated into a single, common database. We derived electronic data from the VA Corporate Data Warehouse, a national, continually updated repository of data from VistA developed specifically to facilitate research[7]. We extracted patient demographics, inpatient and outpatient visits, problem lists, procedures, vital signs, diagnostic tests, laboratory tests, and pharmacy prescriptions.

The study was approved by the Institutional Review Board of the Veterans Affairs Puget Sound Healthcare System and the requirement to obtain informed consent was waived.

Definition of cirrhosis

The diagnosis of cirrhosis was based on the presence of the ICD-9 codes for cirrhosis or complications of cirrhosis (gastroesophageal varices, encephalopathy, nonmalignant ascites) listed in S1 Table, recorded at least twice in any inpatient or outpatient encounter. This approach has been validated and widely used in VA-based studies by us[2, 5, 814] and others[1517]. The diagnosis of cirrhosis using a single ICD-9 code in VA data has been shown to have a 90% positive predictive value (probability that cirrhosis is present among those with a code) compared to chart extraction[18]. By requiring the relevant ICD-9 codes to be recorded at least twice we have found that the positive predictive value increased to 97% in a random sample of 250 patients included in the current study.

Cause of liver disease

Among patients with cirrhosis, we defined the following four etiologies of cirrhosis based on our previously published studies[5]:

  1. HCV: Patients with a positive serum HCV RNA were categorized as HCV regardless of any additional etiologies
  2. ALD: Patients with ICD-9 codes for alcohol use disorders (S1 Table) in the absence of serological markers of chronic HCV or HBV infection and in the absence if ICD9 codes for hemochromatosis, primary biliary cirrhosis, primary sclerosing cholangitis, and autoimmune hepatitis (S1 Table).
  3. NAFLD: was defined in patients with diabetes (ICD-9 code 250–250.92, recorded at least twice[19]) or body mass index (BMI) ≥30 kg/m2 prior to the diagnosis of cirrhosis, who did not have HCV, ALD (defined as above) or ICD9 codes for hemochromatosis, primary biliary cirrhosis, primary sclerosing cholangitis, and autoimmune hepatitis. NAFLD-related cirrhosis does not have pathognomonic serological, radiological, or histological features—even hepatic steatosis is frequently absent after cirrhosis develops. Hence we adapted a clinical definition of NAFLD based on previous work[5, 20] that reflects the diagnostic process used in clinical practice, in which NAFLD is suspected in the presence of risk factors such as obesity and diabetes after exclusion of other etiologies.
  4. OTHER: All other patients not meeting criteria above for HCV, ALD or NAFLD.

These definitions were mutually exclusive by design in order to allow comparisons by cirrhosis etiology. Since patients categorized under HCV may also have concomitant ALD, we performed secondary analyses subdividing HCV into those with and without concomitant ALD.

Baseline patient characteristics

We extracted age, sex, race/ethnicity and all the baseline laboratory tests shown in Table 1, ascertained within 90 days of the diagnosis of cirrhosis. Body mass index (BMI) was calculated as the measured weight in kilograms divided by the square of the measured height in meters recorded within 6 months prior to cirrhosis diagnosis. We also determined the presence of signs of decompensated cirrhosis (defined as the presence of ascites, spontaneous bacterial peritonitis, encephalopathy, gastroesophageal varices or hepatorenal syndrome), type 2 diabetes mellitus, alcohol use disorders, substance use disorders and HIV infection based on appropriate ICD-9 codes recorded at least twice prior to the date of cirrhosis diagnosis in any inpatient or outpatient encounter (S1 Table). These ICD9-based definitions of cirrhosis, decompensated cirrhosis and other comorbidities have been widely used and validated in studies using VA medical records[5, 1417, 19, 21].

thumbnail
Table 1. Baseline characteristics at the time of diagnosis of cirrhosis for all patients diagnosed with cirrhosis in the VA healthcare system from 2001–2014 (n = 116,404), presented according to cirrhosis etiology.

https://doi.org/10.1371/journal.pone.0204412.t001

Diagnosis of hepatocellular carcinoma

The diagnosis of HCC was based on the presence of ICD-9 code 155.0 and ICD-10 code C22.0 (the VA switched to ICD-10 codes on 10/1/2015) recorded at least twice. The ICD-9 code-based definition of HCC using VA records has been shown to have a positive predictive value of 84–94% compared to chart extraction[17, 22, 23] and has been widely used by us[2, 5, 13, 24] and other investigators[2527].

Statistical analysis

We used multivariable Cox proportional hazards regression to compare cirrhotic patients with HCV versus those with ALD or NAFLD with respect to the risk of developing HCC after cirrhosis diagnosis, adjusting for baseline characteristics. We also used multivariable Cox proportional hazards regression to determine the association between baseline characteristics at the time of cirrhosis diagnosis and the risk of developing HCC, separately for each cirrhosis etiology (HCV, ALD or NAFLD). Finally, we used Cox proportional hazards regression to compare the risk of HCC in patients with cirrhosis diagnosed in 2001–2007 versus those diagnosed in 2008–2014, separately by cirrhosis etiology limiting maximum follow-up to 5 years.

Follow-up started 90 days after the diagnosis of cirrhosis (since cases of HCC diagnosed within 90 days were almost certainly present at the time of cirrhosis diagnosis) and continued until the development of HCC or the date of liver transplantation or the date of death or the date of last follow-up in the VA, whichever occurred first. Follow-up could potentially extend up to 01/31/2017 such that all patients had a minimum potential follow-up of 25 months. Patients who did not develop HCC were censored at the date of death, liver transplantation or last follow-up. Patients with HCV who were cured as a result of antiviral treatment were censored at the date they achieved sustained virologic response.

Mortality is high in patients with cirrhosis and death can be a competing risk for HCC, that is, patients who are more likely to develop HCC may also be more likely to die beforehand thus masking the development of HCC. Liver transplantation, although much less common than death, can also be a competing risk for HCC. For this reason, we additionally performed competing risks[28] proportional hazards analysis to model the risk of HCC (which yields “sub-hazard” ratios instead of hazard ratios) while simultaneously accounting for the competing risk of death or liver transplantation.

Results

Comparison of patients according to cirrhosis etiology

The most common etiology of cirrhosis was HCV (45%) followed by ALD (31%), NAFLD (15%) and other etiologies (9%). Among patients with HCV, 51% also had a history of alcohol use disorders. Patients with NAFLD were substantially older, more likely to be white and, by definition, more likely to have diabetes and high BMI than other etiologies (Table 1). Conversely, patients with HCV were the youngest and most likely to be non-white.

Clinical manifestations of decompensated cirrhosis (ascites, encephalopathy or varices) were most common in NAFLD and OTHER etiologies and least common in HCV at the time of cirrhosis diagnosis. However, mean platelet count was lowest and mean serum alpha fetoprotein (AFP) level and AST/√ALT ratio were highest in patients with HCV.

Compared to patients who did not develop HCC, those who developed HCC were slightly older, had lower platelet count, higher serum AFP and were more likely to be male, Hispanic or diabetic (among ALD and NAFLD groups)–Table 2.

thumbnail
Table 2. Baseline characteristics of patients with or without a diagnosis of HCC during follow-up, presented according to cirrhosis etiology*.

https://doi.org/10.1371/journal.pone.0204412.t002

Risk of HCC by cirrhosis etiology

During a mean follow-up of 4.3 years, 10,042 new cases of HCC were diagnosed in 116,404 patients (incidence = 2.0 per 100 patient-years). Patients with HCV had more than 3 times higher incidence of HCC (3.3 per 100 patient-years) than patients with ALD (0.86 per 100 patient-years), NAFLD (0.90 per 100 patient-years) or OTHER (1.0 per 100 patient-years) etiologies of cirrhosis (Table 3 and Fig 1A). Multivariable Cox proportional hazards analyses confirmed that compared to patients with HCV-cirrhosis, those with ALD-cirrhosis (adjusted hazard ratio [AHR] 0.30, 95% CI 0.28–0.32), NAFLD-cirrhosis (AHR 0.34, 95% CI 0.31–0.37), or OTHER-cirrhosis (AHR 0.40, 95% CI 0.36–0.44) had significantly lower risk of HCC after adjusting for baseline characteristics (Table 3). Competing risks analysis yielded very similar adjusted sub-hazard ratios (Table 3).

thumbnail
Fig 1. Kaplan-Meier curves showing the probability of being free of HCC after the diagnosis of cirrhosis, plotted by a.

Etiology of cirrhosis. b. Age. c. Serum AFP level. d. Blood platelet count. e. Serum AST/√ALT ratio. f. Serum albumin level.

https://doi.org/10.1371/journal.pone.0204412.g001

thumbnail
Table 3. Comparison of HCC risk after the diagnosis of cirrhosis, according to etiology of cirrhosis.

https://doi.org/10.1371/journal.pone.0204412.t003

Among patients with HCV, there was no difference in HCC incidence or risk between those with and without a history of alcohol use disorders (Table 3), supporting our decision to consider patients with HCV infection as a single category irrespective of history of alcohol use.

When we divided patients into those diagnosed with cirrhosis in 2001–2007 and those diagnosed in 2008–2014, patients with cirrhosis due to ALD, NAFLD or Other etiologies had significantly lower HCC risk than those with HCV in both time periods (Table 3).

Trends in HCC risk over time by cirrhosis etiology

HCC incidence was higher in patients with cirrhosis diagnosed in 2008–2014 (2.47 per 100 patient-years) compared to those diagnosed in 2001–2007 (1.55 per 100 patient-years) (Table 4 and Fig 2). An increase in HCC incidence was observed among all etiologies of cirrhosis. After adjustment for baseline characteristics, the 2008–2014 cohort had higher HCC risk than the 2001–2007 cohort (AHR 1.38, 95% CI 1.32–1.45), an association that persisted among all cirrhosis etiologies. Competing risks analysis yielded very similar adjusted sub-hazard ratios (Table 4).

thumbnail
Fig 2. Kaplan-Meier curves showing the probability of being free of HCC after the diagnosis of cirrhosis, comparing patients diagnosed with cirrhosis in 2001–2007 to those diagnosed in 2008–2014.

a. All patients. b. HCV. c. ALD. d. NAFLD.

https://doi.org/10.1371/journal.pone.0204412.g002

thumbnail
Table 4. Comparison of HCC risk among patients with cirrhosis diagnosed in 2008–2014 versus those diagnosed in 2001–2007.

https://doi.org/10.1371/journal.pone.0204412.t004

The greatest increase over time occurred in patients with NAFLD-cirrhosis in relative terms (AHR 1.44, 95% CI 1.18–1.77) and patients with HCV-cirrhosis in absolute terms (from 2.62 to 3.92 per 100 patient-years).

Risk factors for HCC by cirrhosis etiology (Table 5 and Table 6)

In univariable analyses (Table 5), the most important predictors of HCC among all liver disease etiologies were older age, male sex, Hispanic ethnicity, high serum AFP level, alkaline phosphatase level and AST/√ALT ratio, and low platelet count and serum albumin level (Fig 1B–1F).

thumbnail
Table 5. Univariable (unadjusted) associations between baseline characteristics and the risk of HCC in patients with cirrhosis, presented according to etiology of cirrhosis (HCV, ALD and NAFLD).

https://doi.org/10.1371/journal.pone.0204412.t005

thumbnail
Table 6. Independent risk factors for HCC in patients with cirrhosis derived from multivariable Cox proportional-hazards regression, presented according to etiology of cirrhosis (HCV, ALD and NAFLD).

https://doi.org/10.1371/journal.pone.0204412.t006

Table 6 shows only the characteristics that remained statistically significant after adjusting for the most important potential confounders (age, sex, race/ethnicity, diabetes, BMI, albumin, platelet count and AST/√ALT ratio) and categorizing continuous variables. The strongest predictors among all liver disease etiologies were: older age, male sex, Hispanic ethnicity, high serum AFP level, alkaline phosphatase level and AST/√ALT ratio, and low platelet count and serum albumin level.

Diabetes was significantly associated with HCC in patients with ALD and NAFLD-cirrhosis but not in patients with HCV-cirrhosis. Also, Hispanic ethnicity was much more strongly associated with HCC in patients with ALD and NAFLD-cirrhosis than in patients with HCV-cirrhosis. High BMI was associated with HCC in patients with ALD but not in patients with HCV or NAFLD. HCV genotype 3 infection was associated with significantly higher risk while genotype 2 infection was associated with lower risk, compared to genotype 1 infection. Certain ranges of high GGT were associated with HCC in patients with HCV (61–586 U/L) and NAFLD (286–586 U/L), but not in patients with ALD-cirrhosis.

Noteworthy but counterintuitive results included the association of HIV co-infection with lower HCC risk in patients with HCV-cirrhosis and the strong association of high serum creatinine with lower HCC risk among all cirrhosis etiologies.

In the analyses shown in Table 6, patients with missing data in one or more laboratory tests dropped out of the multivariable analyses–hence the sample size is slightly lower. We repeated the multivariable analyses including a missing category instead of dropping those with missing predictors and found almost identical results.

Discussion

In a national cohort of cirrhotic patients, those with HCV-cirrhosis had approximately 3 times greater risk of HCC than those with NAFLD-cirrhosis or ALD-cirrhosis, after adjusting for baseline characteristics. HCC risk was 1.38 times higher in patients diagnosed with cirrhosis in 2008–2014 than those diagnosed in 2001–2007, an association that persisted among all cirrhosis etiologies. The most important risk factors for HCC irrespective of underlying cirrhosis etiology were older age, male sex, Hispanic ethnicity, high serum AFP level, alkaline phosphatase level and AST/√ALT ratio, and low platelet count and serum albumin level. Additionally, diabetes was a risk factor in patients with ALD and NAFLD and increased BMI in patients with ALD; neither diabetes nor increased BMI were risk factors for HCC in patients with HCV.

There was a remarkable increase in HCC incidence among all cirrhotic patients in the 7-year interval between the two cohorts that we compared (2001–2007 versus 2008–2014) from 1.55 to 2.47 cases per 100 patient-years. This increase was more pronounced in patients with HCV-cirrhosis and NAFLD-cirrhosis than in ALD-cirrhosis. Although this association was attenuated slightly by adjustment for baseline characteristics, it remained significant (AHR 1.38, 95% 1.32–1.45) suggesting that changes over time in baseline characteristics that we adjusted for–such as the “aging” of the HCV cohort[5], or the increasing prevalence of obesity and diabetes–do not entirely explain this trend. This suggests that unknown factors, other than the ones we adjusted for, might be causing this increase. However, it is impossible to exclude that increasing awareness of HCC over time or the introduction of the liver imaging reporting and data system (LI-RADS) in 2011 might have led to an increase in the rate of diagnosis without a true increase in incidence.

Our finding that HCV-related cirrhosis is independently associated with a 3-fold greater risk of HCC than ALD or NAFLD-related cirrhosis after adjusting for known risk factors suggests that the hepatitis C virus itself may have a direct carcinogenic effect[2931]. HCV is an RNA virus without a DNA intermediate and, hence, does not integrate into the host genome. However, HCV viral proteins have been implicated in various oncogenic processes. In particular, the HCV core protein has been shown to play a key role in the downregulation of tumor suppressor genes, promoter activation of oncogenes, dysregulation of apoptosis, reactive oxidation species (ROS) formation, and immune modulation[32]. HCV may also play a role in inducing epigenetic changes associated with HCC. Specifically, hypermethylation of tumor suppressor genes and alteration of micro-RNA profiles implicated in hepatocarcinogenesis have been observed at higher frequencies in HCV-positive HCC than HCV-negative HCC[33]. In addition, the fact that genotype 3 HCV is associated with a greater risk of HCC than other genotypes suggests a direct carcinogenic effect of the virus.

Other mechanisms besides a direct viral carcinogenic effect could also explain the association between HCV-cirrhosis and HCC. Chronic inflammation and increased cell turnover are thought to drive HCC development in cirrhosis[30, 34]. There are important differences in the pathogenesis of necroinflammation between HCV, NAFLD and ALD, which may result in differences in HCC risk. Furthermore, in patients with HCV-related cirrhosis, the primary necro-inflammatory stimulus (i.e. the hepatitis C virus) is still present, whereas in ALD or NAFLD-related cirrhosis, the primary necro-inflammatory stimulus (i.e. alcohol use in ALD or a hypothesized lipotoxic molecule in NAFLD) may or may not still be present after the development of cirrhosis.

In patients who have HCV infection as well as a substantial history of alcohol use, it may be impossible to distinguish the etiology of cirrhosis, which is frequently labeled as “HCV plus alcohol”. However, our results clearly demonstrate that in cirrhotic patients with HCV infection, HCC risk is not influenced by a history of alcohol use disorders (Table 3). Stated differently, in cirrhotic patients with both HCV infection and alcohol use disorders, it is the HCV that determines HCC risk. Therefore, for the purposes of HCC risk estimation, cirrhotic patients with HCV can be categorized in a single group irrespective of prior alcohol use. This should be distinguished from ongoing alcohol use after the development of cirrhosis which may affect HCC risk but was not specifically studied in this report.

The most important risk factors for HCC among all etiologies of liver disease were older age, male gender, Hispanic ethnicity, high serum AFP, alkaline phosphatase, and AST/√ALT ratio, and low platelet count and serum albumin. Although some of these risk factors have been previously described [2, 24, 25, 27, 35, 36], others merit further discussion. Serum AFP has mostly been investigated as a screening test for the detection of HCC, the idea being that some HCCs secrete AFP. Serum AFP has only modest accuracy as a screening or diagnostic test for HCC and thresholds >20 ng/mL are most commonly recommended for screening[1, 37, 38]. However, we demonstrated that serum AFP level is a strong predictor of the future development of HCC and that even low levels of serum AFP, within the “normal range”, were strongly predictive. Compared to the baseline category of serum AFP ≤3.0 ng/mL (which was the lowest quartile), relatively low serum AFP levels in the second quartile (>3.0–5.0 ng/mL), the third quartile (>5.0–10.2), the 75th-90th percentile (>10.2–28.9), as well as higher levels >90th percentile (>28.9) were associated with high and progressively increasing HCC risk. It is very unlikely that these associations were caused by occult cancers that were present at baseline but only diagnosed later because we excluded all cancers diagnosed within the first 90 days and because the Kaplan-Meier curves of AFP categories continue to diverge from each other for many years of follow-up. Our results suggest that cirrhotic livers that produce higher baseline levels of AFP are at greater risk of developing HCC, for reasons that remain to be elucidated.

Low platelet count is one of the most consistently reported risk factors for HCC in cirrhotic patients[2, 3941]. Low platelet count is a marker of more advanced cirrhosis. However, it is unlikely that this explains the association between thrombocytopenia and HCC because other markers of more advanced cirrhosis, such as hyperbilirubinemia or increased INR were not associated with HCC. Perhaps thrombocytopenia predicts HCC because it is a marker of portal hypertension and correlates with the hepatic venous portal gradient[42]. However, other signs of portal hypertension such as ascites and gastroesophageal varices were not significant predictors. Platelets, which cannot synthesize 5-hydroxytryptamine (5-HT), absorb 5-HT efficiently from the plasma pool and store it in their dense-body granules. It has been suggested that this platelet-derived 5-HT is important in liver regeneration[43]. It is tempting to speculate that a low platelet count results in a low level of platelet derived 5-HT which attenuates the ability of the liver to regenerate and thereby leads to increased risk of HCC.

AST/√ALT ratio, a component of the fibrosis-4 score, is considered to be a measure of fibrosis. Fibrosis development continues even after the development of cirrhosis and likely leads to worsening liver dysfunction. Therefore, even among cirrhotic patients, a high AST/√ALT ratio likely captures those with even more advanced fibrosis. The most plausible explanation for the strong association that we described between AST/√ALT ratio and HCC is that increasing fibrosis increases the risk of HCC. High AST/√ALT was associated with HCC risk in all etiologies of cirrhosis, though in ALD, its greatest risk peaked within the third quartile (8.7–12.4). This nonlinear association is likely explained by the fact that alcohol use acutely elevates AST. Thus, at very high levels (>12.4), the AST/√ALT ratio may reflect recent alcohol use and consequently become a less accurate predictor of fibrosis or HCC.

Some important differences between HCV, ALD and NAFLD in HCC risk factors may be particularly informative. Diabetes was associated with HCC risk in patients with ALD and NAFLD, but not in patients with HCV. High BMI was associated with HCC risk in patients with ALD, but not in patients with HCV (BMI is difficult to interpret in patients with NAFLD because, by definition, these patients had to have a BMI ≥30 kg/m2 or diabetes). Furthermore, Hispanic ethnicity was a much stronger risk factor for HCC in patients with ALD (AHR 1.73) or NAFLD (AHR 2.02) than in patients with HCV (AHR 1.2). Metabolic risk factors are common in Hispanics and may mediate the association between Hispanic ethnicity and HCC[44]. Collectively, these findings suggest that metabolic factors are much more important risk factors for HCC in NAFLD and ALD than in HCV-related cirrhosis.

Although HCV co-infection undoubtedly increases HCC risk in patients with HIV infection, whether HIV co-infection increases HCC risk in patients with HCV infection and cirrhosis is unclear. It has been hypothesized that HIV-related immune suppression might increase the risk of HCC[45]. However, a critical review of the literature did not reveal any clinical or epidemiological studies to support that HIV co-infection increases HCC risk in patients with HCV and cirrhosis after adjusting for potential confounders[46, 47]. We actually found the opposite association: HIV co-infection was associated with a lower risk of HCC in patients with HCV-cirrhosis. This could be due to unmeasured confounding such as differences in ongoing alcohol use or in the mode of transmission of HCV. Future studies need to confirm this controversial finding.

Serum creatinine concentration is a direct reflection of skeletal muscle mass, assuming constant glomerular filtration rate[48]. Sarcopenia is very common in patients with cirrhosis and strongly affects outcomes including survival in patients with cirrhosis and HCC[49]. It is plausible that the association that we found between high serum creatinine and low HCC risk reflects an association between low muscle mass and HCC risk. An association between sarcopenia and HCC risk has not previously been established and will require further confirmation.

Limitations

Eradication of HCV reduces the risk of HCC[8]. For this reason, we censored patients with HCV-cirrhosis at the time they eradicated HCV. Therefore, our study reflects the incidence and risk factors of HCC in patients with HCV-cirrhosis who still have HCV infection. We are conducting separate studies specifically addressing the incidence and risk factors of HCC after eradication of HCV by antiviral treatment. The diagnoses of cirrhosis and HCC were based on ICD-9 codes recorded by the patients’ providers in their electronic medical records. These definitions of HCC and cirrhosis extracted from national VA datasets have been extensively validated and used in research. Prospective ascertainment of cirrhosis and HCC using a priori selected criteria is clearly not feasible in large, national studies with very long follow-up such as the one we conducted. Our results apply primarily to male patients with cirrhosis who constituted 97.5% of our study population–although the study population was so large that is still included a very large number of women (n = 2972) and the association between sex and HCC could be robustly ascertained. Substantial strengths of the study include the large sample size, large number of incident HCCs and long follow-up time. Data were available for most of the important potential risk factors for HCC. All patients were derived from a single, national healthcare system with fairly uniform practices and guidelines across its facilities.

In conclusion, our findings demonstrate the need for a more nuanced approach to HCC risk assessment among patients with cirrhosis. Current AASLD guidelines recommend a one-size-fits-all screening strategy for HCC (abdominal ultrasound with or without serum AFP every six months) in patients with cirrhosis[38], regardless of etiology or the presence or absence of various risk factors. However, HCC risk is very heterogeneous and critically depends on the predictors that we described. Our results also suggest that AFP can be a powerful predictor of HCC when used in combination with other risk factors, even if it is not adequate as a sole screening test. Investigators have recently attempted to develop algorithm-based risk models to estimate HCC risk and target high-risk patient populations for screening[26, 27, 35, 50]. Many of these models include AFP and other variables described in this study, and have shown early promise, though they have yet to be validated in clinical practice. Our study may help inform the future development of risk assessment models. Our results suggest that models that estimate HCC risk should be developed separately for patients with HCV, ALD and NAFLD-cirrhosis since they have different baseline risks as well as different predictors. Also, our results suggest that models developed using older data (e.g. data before 2008), will likely underestimate current HCC risk given the dramatic increase in incidence over time.

Supporting information

S1 Table. Definition of patient characteristics based on diagnostic ICD9 codes recorded at least twice in inpatient or outpatient records.

https://doi.org/10.1371/journal.pone.0204412.s001

(DOCX)

References

  1. 1. Bruix J, Sherman M. Management of hepatocellular carcinoma: an update. Hepatology. 2011;53(3):1020–2. Epub 2011/03/05. pmid:21374666.
  2. 2. Ioannou GN, Splan MF, Weiss NS, McDonald GB, Beretta L, Lee SP. Incidence and Predictors of Hepatocellular Carcinoma in Patients With Cirrhosis. Clin Gastroenterol Hepatol. 2007;5(8):938–45. pmid:17509946.
  3. 3. Labonte ED, Camarota LM, Rojas JC, Jandacek RJ, Gilham DE, Davies JP, et al. Reduced absorption of saturated fatty acids and resistance to diet-induced obesity and diabetes by ezetimibe-treated and Npc1l1-/- mice. American journal of physiology. 2008;295(4):G776–83. Epub 2008/08/23. doi: ajpgi.90275.2008 [pii] pmid:18718999; PubMed Central PMCID: PMC2575916.
  4. 4. West J, Card TR, Aithal GP, Fleming KM. Risk of hepatocellular carcinoma among individuals with different aetiologies of cirrhosis: a population-based cohort study. Aliment Pharmacol Ther. 2017;45(7):983–90. pmid:28144999.
  5. 5. Beste LA, Leipertz SL, Green PK, Dominitz JA, Ross D, Ioannou GN. Trends in Burden of Cirrhosis and Hepatocellular Carcinoma by Underlying Liver Disease in US Veterans, 2001–2013. Gastroenterology. 2015;149(6):1471–82 e5. pmid:26255044.
  6. 6. Veterans Health Administration. Available at: http://www.va.gov/health/findcare.asp Last accessed 12/19/2016.
  7. 7. Veterans Affairs Corporate Data Warehouse. Available at: http://www.hsrd.research.va.gov/for_researchers/vinci/cdw.cfm Last accessed on 12/19/16.
  8. 8. Ioannou GN, Green PK, Berry K. HCV eradication induced by direct-acting antiviral agents reduces the risk of hepatocellular carcinoma. J Hepatol. 2017. pmid:28887168.
  9. 9. Moon AM, Dominitz JA, Ioannou GN, Lowy E, Beste L. Use of Antibiotics Among Patients With Cirrhosis and Upper Gastrointestinal Bleeding is Associated With Reduced Mortality. Clin Gastroenterol Hepatol. 2016. pmid:27311621.
  10. 10. Lerrigo R, Beste LA, Leipertz SL, Green PK, Lok AS, Kogut MJ, et al. Characteristics and outcomes of transjugular intrahepatic portosystemic shunt recipients in the VA Healthcare System. Eur J Gastroenterol Hepatol. 2016;28(6):667–75. pmid:26886386.
  11. 11. Ioannou GN, Beste LA, Chang MF, Green PK, Lowy E, Tsui JI, et al. Effectiveness of Sofosbuvir, Ledipasvir/Sofosbuvir, or Paritaprevir/Ritonavir/Ombitasvir and Dasabuvir Regimens for Treatment of Patients With Hepatitis C in the Veterans Affairs National Health Care System. Gastroenterology. 2016;151(3):457–71 e5. pmid:27267053.
  12. 12. Ioannou GN, Bryson CL, Weiss NS, Boyko EJ. Associations between lipodystrophy or antiretroviral medications and cirrhosis in patients with HIV infection or HIV/HCV coinfection. Eur J Gastroenterol Hepatol. 2015;27(5):577–84. pmid:25769096.
  13. 13. Ioannou GN, Bryson CL, Weiss NS, Miller R, Scott JD, Boyko EJ. The prevalence of cirrhosis and hepatocellular carcinoma in patients with human immunodeficiency virus infection. Hepatology. 2013;57(1):249–57. Epub 2012/04/26. pmid:22532055.
  14. 14. Beste LA, Ioannou GN, Larson MS, Chapko M, Dominitz JA. Predictors of early treatment discontinuation among patients with genotype 1 hepatitis C and implications for viral eradication. Clin Gastroenterol Hepatol. 2010;8(11):972–8. Epub 2010/08/10. doi: S1542-3565(10)00735-4 [pii] pmid:20692371.
  15. 15. Backus LI, Boothroyd DB, Phillips BR, Mole LA. Predictors of response of US veterans to treatment for the hepatitis C virus. Hepatology. 2007;46(1):37–47. Epub 2007/06/15. pmid:17567830.
  16. 16. Davila JA, Henderson L, Kramer JR, Kanwal F, Richardson PA, Duan Z, et al. Utilization of surveillance for hepatocellular carcinoma among hepatitis C virus-infected veterans in the United States. Ann Intern Med. 2011;154(2):85–93. Epub 2011/01/19. doi: 154/2/85 [pii] pmid:21242365.
  17. 17. Kanwal F, Hoang T, Kramer JR, Asch SM, Goetz MB, Zeringue A, et al. Increasing Prevalence of HCC and Cirrhosis in Patients With Chronic Hepatitis C Virus Infection. Gastroenterology. 2011;140(4):1182–8 e1. Epub 2010/12/28. doi: S0016-5085(10)01877-9 [pii] pmid:21184757.
  18. 18. Kramer JR, Davila JA, Miller ED, Richardson P, Giordano TP, El-Serag HB. The validity of viral hepatitis and chronic liver disease diagnoses in Veterans Affairs administrative databases. Aliment Pharmacol Ther. 2008;27(3):274–82. pmid:17996017.
  19. 19. Miller DR, Safford MM, Pogach LM. Who has diabetes? Best estimates of diabetes prevalence in the Department of Veterans Affairs based on computerized patient data. Diabetes Care. 2004;27 Suppl 2:B10–21. pmid:15113777.
  20. 20. Ascha MS, Hanouneh IA, Lopez R, Tamimi TA, Feldstein AF, Zein NN. The incidence and risk factors of hepatocellular carcinoma in patients with nonalcoholic steatohepatitis. Hepatology. 2010;51(6):1972–8. pmid:20209604.
  21. 21. Kramer JR, Kanwal F, Richardson P, Giordano TP, Petersen LA, El-Serag HB. Importance of patient, provider, and facility predictors of hepatitis C virus treatment in veterans: a national study. Am J Gastroenterol. 2011;106(3):483–91. Epub 2010/11/11. doi: ajg2010430 [pii] pmid:21063393.
  22. 22. Kramer JR, Giordano TP, Souchek J, Richardson P, Hwang LY, El-Serag HB. The effect of HIV coinfection on the risk of cirrhosis and hepatocellular carcinoma in U.S. veterans with hepatitis C. Am J Gastroenterol. 2005;100(1):56–63. pmid:15654781.
  23. 23. Davila JA, Weston A, Smalley W, El-Serag HB. Utilization of screening for hepatocellular carcinoma in the United States. J Clin Gastroenterol. 2007;41(8):777–82. pmid:17700427.
  24. 24. Beste LA, Green PK, Berry K, Kogut MJ, Allison SK, Ioannou GN. Effectiveness of hepatitis C antiviral treatment in a USA cohort of veteran patients with hepatocellular carcinoma. J Hepatol. 2017;67(1):32–9. pmid:28267622.
  25. 25. El-Serag HB, Kanwal F, Richardson P, Kramer J. Risk of hepatocellular carcinoma after sustained virological response in Veterans with hepatitis C virus infection. Hepatology. 2016;64(1):130–7. pmid:26946190; PubMed Central PMCID: PMC4917456.
  26. 26. White DL, Richardson P, Tayoub N, Davila JA, Kanwal F, El-Serag HB. The Updated Model: An Adjusted Serum Alpha-Fetoprotein-Based Algorithm for Hepatocellular Carcinoma Detection With Hepatitis C Virus-Related Cirrhosis. Gastroenterology. 2015;149(7):1986–7. pmid:26519622.
  27. 27. El-Serag HB, Kanwal F, Davila JA, Kramer J, Richardson P. A new laboratory-based algorithm to predict development of hepatocellular carcinoma in patients with hepatitis C and cirrhosis. Gastroenterology. 2014;146(5):1249–55 e1. pmid:24462733; PubMed Central PMCID: PMC3992177.
  28. 28. Bob Gray (2011). cmprsk: Subdistribution Analysis of Competing Risks. R package version 2.2–2. http://CRAN.R-project.org/package=cmprsk
  29. 29. Lemon SM, McGivern DR. Is hepatitis C virus carcinogenic? Gastroenterology. 2012;142(6):1274–8. pmid:22537433; PubMed Central PMCID: PMC4422399.
  30. 30. Kamegaya Y, Hiasa Y, Zukerberg L, Fowler N, Blackard JT, Lin W, et al. Hepatitis C virus acts as a tumor accelerator by blocking apoptosis in a mouse model of hepatocarcinogenesis. Hepatology. 2005;41(3):660–7. Epub 2005/02/22. pmid:15723444.
  31. 31. Chen CL, Tsukamoto H, Machida K. Oncogenic signaling pathways and origins of tumor-initiating stem-like cells of hepatocellular carcinomas induced by hepatitis C virus, alcohol and/or obesity. Hepatol Int. 2014;8(3):330–8. Epub 2015/07/24. pmid:26202636.
  32. 32. Jeong SW, Jang JY, Chung RT. Hepatitis C virus and hepatocarcinogenesis. Clinical and molecular hepatology. 2012;18(4):347–56. Epub 2013/01/17. pmid:23323249; PubMed Central PMCID: PMCPMC3540370.
  33. 33. Zhang Y. Detection of epigenetic aberrations in the development of hepatocellular carcinoma. Methods Mol Biol. 2015;1238:709–31. Epub 2014/11/26. pmid:25421688.
  34. 34. Budhu A, Wang XW. The role of cytokines in hepatocellular carcinoma. Journal of leukocyte biology. 2006;80(6):1197–213. Epub 2006/09/02. pmid:16946019.
  35. 35. Ganne-Carrie N, Layese R, Bourcier V, Cagnot C, Marcellin P, Guyader D, et al. Nomogram for individualized prediction of hepatocellular carcinoma occurrence in hepatitis C virus cirrhosis (ANRS CO12 CirVir). Hepatology. 2016;64(4):1136–47. Epub 2016/06/28. pmid:27348075.
  36. 36. Singal AG, Mukherjee A, Elmunzer BJ, Higgins PD, Lok AS, Zhu J, et al. Machine learning algorithms outperform conventional regression models in predicting development of hepatocellular carcinoma. Am J Gastroenterol. 2013;108(11):1723–30. pmid:24169273; PubMed Central PMCID: PMC4610387.
  37. 37. Bruix J, Sherman M. Management of hepatocellular carcinoma. Hepatology. 2005;42(5):1208–36. pmid:16250051.
  38. 38. Heimbach J, Kulik LM, Finn R, Sirlin CB, Abecassis M, Roberts LR, et al. Aasld guidelines for the treatment of hepatocellular carcinoma. Hepatology. 2017. pmid:28130846.
  39. 39. Velazquez RF, Rodriguez M, Navascues CA, Linares A, Perez R, Sotorrios NG, et al. Prospective analysis of risk factors for hepatocellular carcinoma in patients with liver cirrhosis. Hepatology. 2003;37(3):520–7. pmid:12601348.
  40. 40. Lok AS, Seeff LB, Morgan TR, di Bisceglie AM, Sterling RK, Curto TM, et al. Incidence of hepatocellular carcinoma and associated risk factors in hepatitis C-related advanced liver disease. Gastroenterology. 2009;136(1):138–48. pmid:18848939; PubMed Central PMCID: PMC3749922.
  41. 41. Lu SN, Wang JH, Liu SL, Hung CH, Chen CH, Tung HD, et al. Thrombocytopenia as a surrogate for cirrhosis and a marker for the identification of patients at high-risk for hepatocellular carcinoma. Cancer. 2006;107(9):2212–22. pmid:17019738.
  42. 42. Garcia-Tsao G, Abraldes JG, Berzigotti A, Bosch J. Portal hypertensive bleeding in cirrhosis: Risk stratification, diagnosis, and management: 2016 practice guidance by the American Association for the study of liver diseases. Hepatology. 2017;65(1):310–35. pmid:27786365.
  43. 43. Starlinger P, Assinger A, Haegele S, Wanek D, Zikeli S, Schauer D, et al. Evidence for serotonin as a relevant inducer of liver regeneration after liver resection in humans. Hepatology. 2014;60(1):257–66. pmid:24277679.
  44. 44. Ashktorab H, Kupfer SS, Brim H, Carethers JM. Racial Disparity in Gastrointestinal Cancer Risk. Gastroenterology. 2017;153(4):910–23. pmid:28807841; PubMed Central PMCID: PMC5623134.
  45. 45. Clifford GM, Rickenbach M, Polesel J, Dal Maso L, Steffen I, Ledergerber B, et al. Influence of HIV-related immunodeficiency on the risk of hepatocellular carcinoma. AIDS. 2008;22(16):2135–41. Epub 2008/10/04. 00002030-200810180-00011 [pii]. pmid:18832877.
  46. 46. Deng LP, Gui XE, Zhang YX, Gao SC, Yang RR. Impact of human immunodeficiency virus infection on the course of hepatitis C virus infection: a meta-analysis. World J Gastroenterol. 2009;15(8):996–1003. pmid:19248201; PubMed Central PMCID: PMC2653408.
  47. 47. Di Benedetto N, Peralta M, Alvarez E, Schroder MT, Estepo C, Paz S, et al. Incidence of hepatocellular carcinoma in hepatitis C cirrhotic patients with and without HIV infection: a cohort study, 1999–2011. Ann Hepatol. 2013;13(1):38–44. pmid:24378264.
  48. 48. Proctor DN, O'Brien PC, Atkinson EJ, Nair KS. Comparison of techniques to estimate total body skeletal muscle mass in people of different age groups. Am J Physiol. 1999;277(3 Pt 1):E489–95. pmid:10484361.
  49. 49. Fujiwara N, Nakagawa H, Kudo Y, Tateishi R, Taguri M, Watadani T, et al. Sarcopenia, intramuscular fat deposition, and visceral adiposity independently predict the outcomes of hepatocellular carcinoma. J Hepatol. 2015;63(1):131–40. pmid:25724366.
  50. 50. Sharma SA, Kowgier M, Hansen BE, Brouwer WP, Maan R, Wong D, et al. Toronto HCC risk index: A validated scoring system to predict 10-year risk of HCC in patients with cirrhosis. J Hepatol. 2017. pmid:28844936.