Cancer in People with and without Hepatitis C Virus Infection: Comparison of Risk Before and After Introduction of Direct-Acting Antivirals

Chronic hepatitis C virus (HCV) infection is associated with 50% of liver cancer cases in the United States and is a well-established cause of hepatocellular carcinoma (HCC; ref. 1). While HCV infection mainly affects the liver, it can also cause extrahepatic disease, including hematologic malignancies such as B-cell non–Hodgkin lymphoma (NHL; refs. 1–4). HCV may also be associated with extrahepatic solid tumors, including cancers of the pancreas, kidney, lung, and gastrointestinal tract, but evidence has been limited and inconsistent across studies (1, 3, 5–7).

Most HCV infections can be cured with direct-acting antiviral (DAA) agents. DAAs are better tolerated and more efficacious than traditional interferon (IFN)-based therapies, with more than 95% of treated patients achieving HCV clearance (8). Importantly, patients with HCV treated with DAAs have a significantly reduced risk of developing HCC (9). Whether DAA therapy also reduces the risk of extrahepatic cancers is not clear but plausible. Some studies report that DAA-treated patients experience reductions in cardiovascular, metabolic, renal, and hematologic complications of HCV infection, indicating that the beneficial effects of DAA therapy extend to extrahepatic conditions (2, 10–14). This is consistent with studies on the extrahepatic benefits of IFN-based therapies, which report that successfully treated patients have a decreased risk of NHL and gastrointestinal cancers when compared with people with untreated HCV infection (3, 15, 16).

Data on the risk of extrahepatic cancers following DAA therapy specifically have been sparse but thus far have not supported the hypothesis that DAA therapy protects against extrahepatic malignancies. In fact, a few small studies and case reports have documented the unexpected occurrence of extrahepatic malignancies post DAA treatment (17–20), leading to some speculation that DAA use may actually increase extrahepatic cancer risk due to downregulation of immune responses or dysregulation of immune surveillance (21). A large study of U.S. veterans who received DAA therapy found no significant difference in risk of NHL between those with and without HCV clearance, but did not compare this to NHL risk in people without HCV infection and did not evaluate non-NHL cancers (11). Another study in U.S. veterans found that while IFN-induced HCV cure was associated with significant reductions in hematologic malignancies, similar results were not observed among those with DAA-induced HCV cure (16). This suggests that evidence on cancer risk in the setting of IFN-based therapy may not extrapolate to people treated with DAAs.

An improved understanding of how DAA treatment may affect risk of extrahepatic cancers could critically inform HCV clinical care and cancer surveillance post-HCV cure, especially as more people with HCV will be cured within the next decade (22). This shift in the clinical profile of people with HCV is already reflected in recent ecologic data; HCV-related mortality rates have declined steeply during the DAA era compared with the pre-DAA era (23, 24) while cause-specific mortality rates from extrahepatic cancers in the DAA era have increased (25), suggesting that DAAs have prevented HCC-related deaths only to give way to death due to extrahepatic cancers.

In our healthcare system, IFN-free DAA therapies have been used to treat people with HCV since 2014, providing an opportunity to examine DAA use and incident cancers among patients within a single-payer healthcare system. We compared the incidence of hepatic and extrahepatic cancers in (i) people with and without HCV infection, both overall and by calendar eras before and after introduction of DAAs, and (ii) DAA-treated and -untreated HCV infection versus people without HCV infection.

We conducted an observational cohort study using data from January 1, 2007 through December 31, 2017 at Kaiser Permanente Northern California (KPNC), an integrated healthcare system that provides comprehensive medical services to over 4.3 million members (26). The study period was divided into 3 calendar eras: 2007 to 2010 (i.e., IFN treatment era), 2011 to 2013 (i.e., IFN prescribed in combination with DAAs following the FDA approval of telaprevir/boceprevir in 2011), and 2014 to 2017 (i.e., primarily IFN-free era following the FDA approval of ledipasvir/sofosbuvir in 2014). The start of the study in 2007 was chosen to have calendar eras of approximately equal duration for comparisons of cancer incidence pre- and post-DAA introduction.

This study included all KPNC members meeting the inclusion criteria: adult age (≥18 years) and having KPNC membership for at least 3 consecutive months during the study period. People were considered to have HCV infection if they had detectable HCV RNA or an HCV genotype documented in the KPNC electronic health record (EHR) at any time prior to meeting the other inclusion criteria. Person-time contributed by people with HCV infection began on the date the inclusion criteria were met and continued until: date of incident cancer, end of KPNC health plan membership, death, or end of the study. Person-time contributed by people without HCV infection was estimated by first determining the total membership duration of all adult KPNC members based on combined annual mid-year membership (up to and including the year of incident cancer, KPNC health plan disenrollment, or death), and then subtracting the exact person-time contributed by all people with HCV infection combined. In other words, follow-up time for people with HCV infection was determined from individual-level dates of study entry and exit whereas follow-up time for people without HCV infection was estimated based on total years of active KPNC membership up to and including the year of study exit. Given the large size of the KPNC member population, this method allowed for inclusion of all KPNC members without HCV infection as the reference population in a computationally efficient manner.

For analyses comparing cancer incidence by HCV status and calendar era, we restricted the study population to people with active KPNC membership during the calendar era of interest, regardless of type of HCV treatment previously received. People could contribute person-time to multiple calendar eras.

For analyses comparing cancer incidence in DAA-treated and -untreated HCV infection, we restricted the study to people with active KPNC membership at any time between October 1, 2014 (when all-DAA therapies became widely available within KPNC) and the end of the study on December 31, 2017. People with HCV were further restricted to those who were treatment eligible, defined as having no evidence of prior HCV treatment, cure, spontaneous clearance, or HCV suppression between the most recent measure of HCV RNA and the date inclusion criteria were met. For these analyses, follow-up time was divided into DAA-untreated and DAA-treated follow-up time using the date of first DAA prescription fill. A study population flow chart is shown in Fig. 1. This study was approved by the KPNC institutional review board.

For people with HCV infection, data on DAA prescription fills were obtained from KPNC pharmacy databases.

Cancer diagnoses were ascertained using the KPNC Cancer Registry, a contributing site to the Surveillance, Epidemiology, and End Results (SEER) program registry that uses standardized methods for verifying and coding incident reportable cancers (27). The following types of cancer were identified for analysis: (i) liver cancers, including HCC and intrahepatic bile duct cancer, (ii) extrahepatic cancers, including hematologic [i.e., Hodgkin lymphoma and NHL, multiple myeloma, leukemia, myelodysplastic syndrome, and diffuse large B-cell lymphoma (DLBCL)], gastrointestinal (including esophageal, colorectal, stomach, and small intestine), anal, female breast, lung, melanoma, oral/oropharyngeal, pancreatic, prostate, renal, thyroid, and urinary bladder cancers, and (iii) ill-defined/unknown primary-site cancers.

Data were gathered from the EHR on: (i) sociodemographic factors, including age, sex, and race/ethnicity, (ii) substance use, including history of smoking, alcohol use disorder, or substance use disorder, and (iii) clinical factors, including hepatitis B virus (HBV) infection and human immunodeficiency virus (HIV) infection, as previously described (28). For people with HCV infection, all of these covariates were obtained at baseline (i.e., start of follow up) and age groups were updated annually. Because follow up for people without HCV was based on mid-year membership estimates and not exact dates, their baseline characteristics were estimated based on those who had active membership in 2007 (i.e., the first year of the study) and age groups were updated annually by subtracting the age-updated person-time for people with HCV from the age strata-specific counts of people without HCV who had active membership in the year of interest.

Analyses of cancer incidence first compared rates between people with and without HCV, regardless of DAA treatment status, for the entire study period (2007–2017). Then, ecologic analyses compared cancer incidence between people with and without HCV across 3 calendar eras (2007–2010, 2011–2013, and 2014–2017). Lastly, using data from the all-DAA treatment era only (October 2014–December 2017), cancer incidence for DAA-treated and -untreated HCV infection was compared with cancer incidence among people without HCV.

Crude and adjusted cancer incidence for people with and without HCV were calculated by dividing the number of cancer cases by total person-years of follow-up. Crude and adjusted incidence rate ratios were obtained from Poisson regression models. A χ² goodness-of-fit test was used to confirm reasonable fit to a Poisson distribution for all cancers evaluated. Covariates known to be associated with HCV infection and cancer risk were included in adjusted models. Demographics-adjusted models included terms for age, sex, race/ethnicity, and calendar year of follow up. Fully adjusted models additionally included terms for smoking, substance use disorder, and ever HBV or HIV infection.

In analyses of all types of cancers combined, we selected the first diagnosis of any cancer. In cancer-specific analyses, we selected the first diagnosis of the cancer of interest without regard to any prior diagnosis of cancer(s) at other sites. Individuals with a prevalent diagnosis of the cancer of interest were excluded from the model of that cancer type but included in models of other cancer types. Ill-defined cancers and cancers with an unknown primary site were included only in analyses of all cancer types combined and not in cancer type-specific analyses. Statistical analyses were conducted in SAS, version 9.3. A two-tailed P < 0.05 was considered statistically significant.

The study included 156,806 person-years of follow-up for people with HCV and 29,012,134 person-years of follow-up for people without HCV. The mean age at baseline was 53.4 years (SD = 10.2) for people with HCV and 46.8 years (SD = 17.4) for people without HCV. People with HCV infection were 62% male, 54% non-Hispanic White, 14% non-Hispanic Black, 16% Hispanic, 7% non-Hispanic Asian, and 5% other race/ethnicity (Table 1). People with HCV infection were more likely to have a history of smoking (75% vs. 41%), alcohol use disorder (14% vs. 3%), and substance use disorder (19% vs. 9%). They were also more likely to have a history of HBV infection (3% vs. 0.3%) and HIV coinfection (4% vs. 0.7%).

During follow up, 2,451 people with HCV and 173,548 people without HCV were diagnosed with at least one type of cancer (Table 2). We observed 1,070 cases of any liver cancer, 1,399 cases of extrahepatic cancer, and 71 cases of ill-defined/unknown cancers among people with HCV, and 2,670 cases of any liver cancer, 165,289 cases of extrahepatic cancer, and 6,602 cases of ill-defined/unknown cancers among people without HCV (Table 2).

Overall cancer incidence was higher among people with HCV than people without HCV [fully adjusted incidence rate ratio (aIRR) = 1.7, 95% confidence interval (CI) = 1.6–1.7; Table 2]. The incidence of liver cancer (aIRR = 31.4, 95% CI = 28.9–34.0) and specifically HCC (aIRR = 38.3, 95% CI = 35.1–41.7) was also higher among people with HCV.

The overall incidence of extrahepatic cancers (i.e., all types combined) did not differ by HCV status (aIRR = 1.0, 95% CI-0.9–1.0). However, we observed significant positive associations between HCV infection and hematologic cancers (aIRR = 1.3, 95% CI = 1.1–1.5), particularly NHL (aIRR = 1.7, 95% CI = 1.4–2.0) and DLBCL (aIRR = 2.2, 95% CI = 1.7–2.9), but not leukemia (aIRR = 1.0, 95% CI = 0.7–1.4).

In terms of nonhematologic extrahepatic cancers, HCV infection was associated with an increased risk of cancers of the lung (aIRR = 1.3, 95% CI = 1.2–1.5), pancreas (aIRR = 2.0, 95% CI = 1.6–2.5), oral cavity/oropharynx (aIRR = 1.4, 95% CI = 1.1–1.8), and anus (aIRR = 1.6, 95% CI = 1.1–2.4). HCV infection was associated with a decreased risk of prostate cancer (aIRR = 0.6, 95% CI = 0.6–0.7), breast cancer (aIRR = 0.7, 95% CI = 0.6–0.8), and melanoma (aIRR = 0.6, 95% CI = 0.4–0.8). The association of HCV infection with an increased risk of kidney and gastrointestinal cancers was significant in models adjusted only for demographics (1.4, 95% CI = 1.1–1.7 and 1.5, 95% CI = 1.2–1.9, respectively), but not in fully adjusted models that additionally accounted for substance use, HBV infection, and HIV infection (1.3, 95% CI = 1.0–1.6 and 1.3, 95% CI = 1.0–1.7, respectively). The associations between HCV infection and all other cancer types were similar in both demographics-only and in fully adjusted models (Table 2).

For all cancer groups, there was no significant difference in aIRRs across calendar eras (Table 3). For liver cancers, there was a nonsignificant decrease in aIRR from 33.9 (95% CI = 29.6–38.8) in 2007 to 2010 to 33.5 (95% CI = 29.1–38.6) in 2011 to 2013 and then to 28.6 (95% CI = 25.5–32.1) in 2014 to 2017 (p for difference in aIRR by era = 0.08). For HCC, there was an increase in aIRR from 39.6 (95% CI = 34.4–45.7) in 2007 to 2010 to 40.6 (95% CI = 35.0–47.1) in 2011 to 2013, followed by a decrease in aIRR to 36.0 (95% CI = 31.8–40.7) in 2014 to 2017. The difference in aIRR between these calendar eras was not statistically significant (P = 0.37). For the other cancer groups, there was no notable decrease in aIRR across calendar eras.

Compared with people without HCV, the aIRR for any type of incident cancer was 1.7 (95% CI = 1.6–1.9) among the DAA-untreated and 1.3 (95% CI = 1.1–1.5) among the DAA-treated (Fig. 2). While DAA treatment was associated with a significant reduction in overall cancer incidence (i.e., from aIRR 1.7 to 1.3, P = 0.001), cancer risk remained 30% above that of people without HCV infection. Similarly, compared with people without HCV, the aIRR for any liver cancer was 31.9 (95% CI = 27.9–36.4) among the DAA-untreated and 21.2 (95% CI = 16.8–26.6) among the DAA-treated, and the aIRR for HCC was 38.4 (95% CI = 33.4–44.1) among the DAA-untreated and 25.5 (95% CI = 20.0–32.4) among the DAA-treated. While DAA treatment was associated with significant reductions in liver cancers overall and HCC specifically (P = 0.001 and P = 0.002, respectively), risk for these cancers remained 21 and 25 times above that of people without HCV, respectively. For all extrahepatic cancers combined, the aIRR was similar in DAA-untreated (vs. people without HCV, 0.9; 95% CI = 0.8–1.0) and DAA-treated (vs. people without HCV, 0.8; 95% CI = 0.7–1.0). For extrahepatic hematologic cancers, the aIRR was reduced from 1.5 among the DAA-untreated (vs. people without HCV; 95% CI = 1.1–2.0) to 0.6 among the DAA-treated (vs. people without HCV; 95% CI = 0.3–1.2). For NHL specifically, the aIRR was reduced from 1.8 among the DAA-untreated (vs. people without HCV; 95% CI = 1.3–2.7) to 0.3 among the DAA-treated (vs. people without HCV; 95% CI = 0.1–1.2). For hematologic cancers overall and NHL specifically, DAA treatment was associated with a significantly reduced cancer risk (P = 0.02 and P = 0.01, respectively) to levels comparable with those among people without HCV.

HCV infection is recognized to be a systemic disease, but few studies have evaluated its association with extrahepatic cancers, particularly in the DAA treatment era. Using EHR data spanning pre- and post-DAA introduction in a large cohort, we examined the risk of incident cancer among people with HCV infection and the extent to which DAA therapy might modify cancer risk. We found that people with HCV infection were at increased risk for liver cancer and some extrahepatic cancers, including hematologic cancers. While DAA treatment was associated with a considerable reduction in the risk of liver cancer, including HCC, the risk of liver cancer among DAA-treated people remained significantly elevated compared with people without HCV. On the other hand, risk of hematologic cancers following DAA treatment was substantially reduced, reaching that among people without HCV. We did not find strong evidence of the impact of DAAs on cancer incidence in ecologic analyses by calendar era. However, the reduction in cancer when examining individual-level cancer risk underscores the benefit of DAA therapy for people with HCV.

People with HCV infection had a 31-fold increased risk of any liver cancer and a 38-fold increased risk of HCC, even after adjusting for demographics, substance use, and HBV and HIV coinfection. HCV infection was also associated with an increased risk of hematologic cancers (1.3-fold), including NHL (1.7-fold) and DLBCL (2.2-fold), consistent with prior studies (1–3, 29). Our study also contributes evidence that people with HCV infection are at significantly higher risk for lung (1.3-fold), pancreatic (2-fold), oral/oropharyngeal (1.4-fold), and anal (1.6-fold) cancers compared with people without HCV infection. These findings are somewhat consistent with those in a large U.S. population-based case–control study of HCV infection and extrahepatic cancers among Medicare recipients, which found that HCV infection was significantly associated with pancreatic cancer and anal cancer, but not lung cancer, oral cavity cancer, or oropharyngeal cancer (30). These findings also align with a meta-analysis of 5 observational studies from the pre-DAA era, which found that HCV infection was associated with increased pancreatic cancer risk (31), although this elevated risk has not been consistently observed across all study populations (5, 32). Compared with models adjusted only for demographics, fully adjusted models that accounted for substance use and coinfections attenuated the incidence rate ratios (IRR) for cancers of the liver, lung, oral cavity/pharynx, and anus. This is not surprising given the strong association of HBV infection and alcohol use with liver cancer, smoking with oral and lung cancer, and HIV infection with anal and oropharyngeal cancer. The IRRs for pancreatic cancer were the same in demographics-only and fully adjusted models, perhaps because of its stronger pathophysiological link to HCV infection. Given the close proximity of the pancreas to the liver, it has been postulated that hepatitis C virions may replicate in pancreatic cells or that the pancreas may be a reservoir for hepatitis viruses (5, 31).

This study contributes evidence that HCV infection increases risk for some extrahepatic cancers, which adds to knowledge of total HCV-related morbidity. Additional evidence from studies with extended follow-up time and in other diverse cohorts would be important to understand whether guidelines for extrahepatic cancer surveillance may need to be tailored to patients with HCV infection. One intriguing observation was the lower risk of prostate cancer, female breast cancer, and melanoma among people with HCV, which is consistent with other work and may be in part explained by disparities in screening rates for these cancers across population subgroups (33–37). For example, lower screening rates have been observed among people of lower socioeconomic status, which may disproportionately include people with HCV infection. Another potential explanation is the competing risk of mortality due to these aging-related cancers, which may not have been fully ameliorated despite analytic adjustment for age.

As observed in other studies (38, 39), people with HCV infection in this study remained at elevated risk for developing HCC despite DAA treatment. This supports current clinical recommendations to continue HCC surveillance among DAA-treated patients regardless of treatment completion and HCV cure (40–42).

DAA treatment reduced risk of hematologic cancers, including NHL, to levels comparable with that among people without HCV infection. While a study in primarily male U.S. veterans found that DAA treatment was not associated with significant reductions in NHL in those who achieved HCV clearance compared with those who did not, it did not include a comparison with those without HCV infection (11). We are not aware of any other studies that have compared the incidence of hematologic cancers or NHL by DAA treatment status and with people without HCV, but such a study could help confirm that DAA treatment normalizes hematologic cancer risk.

Our finding that DAA treatment normalizes the risk of hematologic cancer is indirectly supported by prior studies that have reported the favorable impact of DAA treatment on NHL-related morbidity and mortality. For example, not only are DAAs safe and effective when given to patients with an NHL diagnosis (43–45), they also improve chemotherapy and cancer outcomes (46), as well as five-year overall survival (47). The observed reduction in NHL incidence associated with DAA treatment may reflect the immune-modulating effects of DAAs on NHL-related pathogenic processes (48). This may explain our observation that DAAs were associated with a larger relative reduction in NHL incidence as compared with HCC incidence, which is more strongly tied to nonimmune-related risk factors.

Our study captured only the initial years of DAA treatment of HCV infection and had relatively short follow-up of cured patients. Broader DAA use and longer lifespans of HCV patients will provide opportunities to further evaluate the influence of DAA treatment on cancer risk (49). In particular, additional follow-up time would be required to confirm the lower risk of liver cancer and HCC among people with HCV in the DAA era compared with prior treatment eras, which was suggested in our ecologic analyses but was nonsignificant, possibly due to limited DAA era follow-up time. Other potential explanations include slow uptake of DAAs during the initial years of DAA availability or increased screening of the “baby boomer” population starting in 2014, which may have resulted in more diagnoses of cancers in the DAA era, offsetting some of the population-level decrease in cancer incidence driven by DAA-induced HCV cure. We interpret the ecologic analyses with caution as they could be subject to ecological fallacy, whereby inferences about DAAs and cancer risk from aggregated data might not be meaningfully extrapolated to the individual.

We acknowledge several limitations of our study. First, without data on conditions such as liver steatosis, nonalcoholic fatty liver disease, alcohol-related liver disease, fibrosis, or cirrhosis, we were unable to report the proportion of patients with underlying liver disease who may have been at greater risk for cancer. However, we have no reason to believe that the prevalence of these liver conditions would be substantially different in our study population compared with other populations. Also, successful treatment with DAAs is typically less influenced by clinical factors than were IFN-based treatments (50). While a preexisting liver condition may increase the likelihood of HCC despite treatment, recent data show that the relative reduction in risk of HCC with DAA treatment is similar in patients with and without cirrhosis (40). Given that DAAs were more likely used in people with advanced liver disease with a higher risk of cancer, our analyses may have underestimated the reduction in cancer risk. Second, we did not have data on the proportion of patients in our study who achieved sustained virologic response with treatment, but prior studies have found DAA cure rates of 94% to 95% in the KPNC population (28, 51, 52). Third, we could not evaluate potential differences in cancer incidence by HCV genotype. Fourth, observed results may be due to residual confounding from factors we were unable to fully capture using EHR data. And lastly, results may not generalize to uninsured populations or to other healthcare settings with differences in HCV surveillance, patient risk stratification, or prioritization for treatment. When all-DAA therapies first became available in 2014, patients in KPNC with more advanced liver disease were prioritized for treatment, but there were no restrictions on treatment, and patients were evaluated for treatment by their practitioners in a manner consistent with standard clinical practice.

Strengths of this study included the large study sample, which was closely representative of the underlying community; inclusion of a comparison group of people without HCV infection; multiple cancer types evaluated; and robust EHR data that allowed for adjustment for cancer-related risk factors. While routine HCV screening of all patients was not done within KPNC, and individuals with undiagnosed HCV infection may have been misclassified as not having HCV, the strong HCV surveillance program and integrated design of our health system likely maximized HCV case capture. Additionally, given the high cost of DAAs, it was unlikely that people with HCV infection were treated outside of KPNC, likely resulting in complete capture of DAA use.

HCV-related cancer incidence remains high in the era of DAAs. HCV infection is associated with increased risk of liver cancer and extrahepatic cancers, including NHL. HCV infection also appears to be associated with some extrahepatic solid tumors, including lung, pancreatic, oral/oropharyngeal, and anal cancers, which could have important implications for enhanced cancer surveillance in people cured of HCV infection. DAA treatment is associated with reduced risk of liver cancers and hematologic cancers, which underscores the importance of prioritizing broad and timely access to DAA treatment to reduce cancer burden among people with HCV infection.

J.O. Lam reports grants from Gilead Sciences Inc. outside the submitted work. E.Y. Chiao reports grants from NIH during the conduct of the study as well as grants from NIH outside the submitted work. J.L. Marcus previously consulted for Kaiser Permanente Northern California on a research grant from Gilead Sciences, Inc. outside the submitted work. M.J. Silverberg reports grants from Gilead Sciences Inc. outside the submitted work. No disclosures were reported by the other authors.

J.O. Lam: Writing–original draft, writing–review and editing. L.B. Hurley: Formal analysis, writing–review and editing. J.B. Lai: Writing–review and editing. V. Saxena: Writing–review and editing. S. Seo: Writing–review and editing. S. Chamberland: Writing–review and editing. C.P. Quesenberry Jr: Writing–review and editing. J.H. Champsi: Writing–review and editing. J. Ready: Writing–review and editing. E.Y. Chiao: Funding acquisition, writing–review and editing. J.L. Marcus: Conceptualization, supervision, funding acquisition, methodology, writing–review and editing. M.J. Silverberg: Conceptualization, supervision, funding acquisition, methodology, writing–review and editing.

This work was supported by grants from the Kaiser Permanente Northern California Delivery Science Grants Program (PIs: J.L. Marcus and M.J. Silverberg), the National Institute of Allergy and Infectious Diseases (K01AI122853, PI: J.L. Marcus), and the National Cancer Institute (R01CA206476, PI: E.Y. Chiao).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.