Hepatocellular carcinoma (HCC) is the

HEPATOLOGY, VOL. 67, NO. 6, 2018 AMERICAN ASSOCIATION FOR THE STUDY OFLIVERD I S E ASES The Short-Term Incidence of Hepatocellular Carcinoma Is Not Increased After Hepatitis C Treatment With Direct-Acting Antivirals: An ERCHIVES Study Darrick K. Li, 1,6 Yanjie Ren, 2 Daniel S. Fierer, 3 Stephanie Rutledge, 1 Obaid S. Shaikh, 2 Vincent Lo Re III, 4 Tracey Simon, 1,6 Abdul-Badi Abou-Samra, 5,7 Raymond T. Chung, 1,6 and Adeel A. Butt 2,5,7 Recent studies have reported higher rates of hepatocellular carcinoma (HCC) in individuals treated with direct-acting antivirals (DAAs). However, making definitive conclusions has been challenging because of the heterogeneous populations and methodologies of these reports. We investigated whether DAA use is associated with higher rates of incident HCC compared to treatment with interferon (IFN)-based regimens. We performed a retrospective, population-based cohort study using the Electronically Retrieved Cohort of HCV Infected Veterans (ERCHIVES) database. In a cohort of 17,836 persons, sustained virological response (SVR) was achieved by 66.6% and 96.2% of the IFN and DAA groups, respectively. Among all treated persons, risk of HCC was not higher in the DAA group compared to the IFN group (hazard ratio, 1.07; 95% confidence interval, 0.55, 2.08). Among persons with cirrhosis who achieved SVR, neither the HCC incidence rate nor HCC-free survival were significantly different in the DAA group compared to the IFN group (21.2 vs. 22.8 per 1,000 person-years; P 5 0.78 and log-rank P 5 0.17, respectively). Untreated persons with cirrhosis had a significantly higher HCC incidence rate (45.3 per 1,000 person-years) compared to those treated with either IFN or DAAs (P 5 0.03). Both groups of treated persons had significantly lower probability of HCC development compared to untreated persons (log-rank, P 5 0.0004). Conclusion: DAA treatment is not associated with a higher risk of HCC in persons with cirrhosis with chronic HCV infection in the short term. Previously reported higher rates of HCC associated with DAA treatment may be explained by both the presence of relatively fewer baseline HCC risk factors in persons treated with IFN as well as selection bias, given that DAA regimens were used to treat persons at higher risk for developing HCC. (HEPATOLOGY 2018;67:2244-2253). Hepatocellular carcinoma (HCC) is the second-leading cause and fastest rising cause of cancer-related death worldwide. (1) One of the most common risk factors for HCC is cirrhosis attributed to chronic infection with the hepatitis C virus (HCV), which is estimated to infect over 150 million people globally. (2,3) For decades, the standard of care for HCV treatment had been interferon (IFN)- based regimens, with approximately 50%-60% of eligible patients achieving sustained virological response (SVR). IFN-induced SVR has been shown to result in an approximately 4-fold reduction of HCC risk regardless of stage of liver disease. (4,5) Thus, with the recent development of highly efficacious direct-acting Abbreviations: AFP, alpha-fetoprotein; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; cddd, cumulative defined daily dose; CI, confidence interval; DAA, direct-acting antiviral; DCL, daclatasvir; ERCHIVES, Electronically Retrieved Cohort of HCV Infected Veterans; FDA, U.S. Food and Drug Administration; FIB-4, Fibrosis-4; HBsAg, hepatitis B surface antigen; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HIV, human immunodeficiency virus; HR, hazard ratio; ICD, International Classification of Diseases; IFN, interferon-based regimen; LDV, ledipasvir; PEG, pegylated interferon; PPI, proton pump inhibitor; PRoD, paritaprevir/ritonavir/ombitasvir 1 dasabuvir; RBV, ribavirin; SMV, simeprevir; SOF, sofosbuvir; SVR, sustained virological response; SVR12, SVR at 12 or more weeks posttreatment; VA, Veterans Affairs. Received July 6, 2017; accepted November 20, 2017. Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.29707/suppinfo. 2244

HEPATOLOGY, Vol. 67, No. 6, 2018 LI ET AL. antiviral (DAA) agents against chronic HCV, in whom >90% of treated patients achieved SVR, expectations were that the use of DAAs would result in significantly lower rates of HCC in these patients. (6,7) However, several recent studies reported a paradoxically increased risk of HCC in patients treated with DAA therapy, particularly among those with previously treated HCC, whereas others have not supported this observation. (8-14) Moreover, given that the studies were performed in relatively small, heterogenous populations, and had inconsistent methodologies, definitive conclusions based on these studies is challenging. Thus, we sought to investigate whether DAA treatment is associated with higher rates of incident HCC using a large, well-established national cohort of HCV-infected U.S. veterans without a past diagnosis of HCC. Patients and Methods STUDY DESIGN AND DATA SOURCE We conducted a retrospective cohort study among HCV-infected persons in the Electronically Retrieved Cohort of HCV Infected Veterans (ERCHIVES) database. ERCHIVES is a large, well-established national cohort of HCV-infected and uninfected Veterans and has been described. (15-18) All HCV-infected Veterans at any of the nation-wide Department of Veterans Affairs (VA) medical facilities with a positive HCV antibody test between 2002 and 2016 were identified. Demographic, clinical, and laboratory data were obtained from the National Patient Care Database and the Corporate Data Warehouse, and pharmacy information, including all prescriptions written, doses, duration, number of pills, number of refills, and date of refills, was retrieved from the Pharmacy Benefits Management database. Data were then merged based on established algorithms. The study was approved by the Institutional Review Board at the VA Pittsburgh Healthcare System. STUDY PARTICIPANTS Persons were included in the study cohort if they had positive serology for HCV antibody and were excluded if they had coinfection with human immunodeficiency virus (HIV), a positive test for hepatitis B surface antigen (HBsAg) or a diagnosis of hepatocellular carcinoma prior to baseline (defined below). This material is the result of work supported with resources and the use of facilities at the VA Pittsburgh Healthcare System and the central data repositories maintained by the VA Information Resource Center, including the National Patient Care Database, Decisions Support System Database, and Pharmacy Benefits Management Database. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs. Dr. Chung was supported by NIH DK078772 and the MGH Research Scholars Program. Copyright VC 2017 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI 10.1002/hep.29707 Potential conflict of interest: Dr. Butt received grants from Gilead and Merck. Dr. Fierer owns stock in and received grants from Gilead. Dr. Chung received grants from Gilead, AbbVie, Merck, Janssen, and Boehringer Ingelheim. ARTICLE INFORMATION: From the 1 Department of Medicine, Massachusetts General Hospital, Boston, MA; 2 VA Pittsburgh Healthcare System, Pittsburgh, PA; 3 Division of Infectious Diseases, Icahn School of Medicine at Mount Sinai, New York, NY; 4 Center for Clinical Epidemiology and Biostatistics, Department of Biostatistics and Epidemiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA; 5 Hamad Medical Corporation, Doha, Qatar; 6 Liver Center, Division of Gastroenterology, Department of Medicine, Massachusetts General Hospital, Boston, MA; 7 Weill Cornell Medical College, Doha, Qatar and New York, NY. ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO: Adeel A. Butt, M.D., M.S. VA Pittsburgh Healthcare System University Drive C Building 30 Mailstop 151 Pittsburgh, PA 15240 E-mail: aab2005@qatar-med.cornell.edu Tel: 1974 4439 6283; or Raymond T. Chung, M.D. Warren 1007 Massachusetts General Hospital 55 Fruit Street Boston, MA 02114 E-mail: rtchung@partners.org Tel: 11 617 724 7562 2245

LI ET AL. HEPATOLOGY, June 2018 Furthermore, persons were excluded if prior to baseline they: 1) were missing an HCV RNA or Fibrosis-4 (FIB-4) score, 2) did not have complete data to permit calculation of a repeat FIB-4 score at least 12 weeks after baseline, or 3) were missing HCV-RNA data to determine attainment of SVR at 12 or more weeks posttreatment (SVR12). This cohort was then divided into the following groups based on treatment received: Group A (IFN): Persons included if received treatment with pegylated interferon (PEG) and ribavirin (RBV) for 28 days. If the person received multiple courses of PEG1RBV, the first treatment course was used for establishing baseline. Persons treated with standard IFN were excluded from this study. Group B (DAA): Persons included if received treatment with one of the following regimens: sofosbuvir 1 simeprevir 6 RBV (SOF/SMV); sofosbuvir 1 ledipasvir 6 RBV (SOF/LDV); sofosbuvir 1 daclatasvir 6 RBV (SOF/DCL); or paritaprevir/ritonavir/ombitasvir 1 dasabuvir 6 RBV (PRoD) for 28 days. Other, newer DAA regimens were not included because small numbers of persons initiated those regimens. Group C: Persons who never received any HCV treatment for 28 days. DEFINITIONS For persons in the treatment group, baseline was the date of HCV treatment initiation. For the untreated group, we assigned the baseline to equal the duration of HCV infection before treatment initiation in the corresponding treated person. This was calculated by determining the average time between the first HCV antibody-positive date and treatment initiation date for the treated persons group, and adding the same duration to the time between the first HCV antibodypositive date in each untreated person and assigning that date as the baseline. Treatment completion was defined according to U.S. Food and Drug Administration (FDA)-approved labels for each drug regimen. Cirrhosis was defined as a FIB-4 score of >3.5, based on published work (16) and was calculated as follows: FIB-4 5 ðage ½yearsŠ 3 AST ½IU=LŠÞ= ðplatelet count½platelets 3 10 9 =LŠ 3 ALT 1=2 ½IU=LŠÞ Laboratory data were obtained at yearly intervals and the FIB-4 score was recalculated at each interval. An average of two values of the components of score (aspartate aminotransferase [AST], alanine aminotransferase [ALT] and platelet count) closest to the selected time point of interest was used for calculation of FIB-4 scores. SVR was defined as undetectable HCV RNA in all follow-up HCV-RNA tests after the end of treatment, with one test at least 12 weeks after end of initial treatment. Persons were defined as having diabetes using a combination of blood glucose measurement, use of anti-diabetic agents, and International Classification of Diseases (ICD)-9 codes, as previously described. (19) Mean ALT, AST, and gamma glutamyl transferase were defined as the average of two values closest to and before baseline. Alpha-fetoprotein (AFP) level was defined as the most recent available value but preceding the first diagnosis of HCC. History of alcohol/ drug abuse or dependence was based on the presence of at least one inpatient or two outpatient ICD-9 diagnoses any time from 12 months before to 3 months after baseline. For diabetes, alcohol and drug abuse, and other independent variables, ICD-10 codes were used for data after October 1, 2015. Drug exposure to proton pump inhibitors (PPIs) and statins was defined as >30 cumulative defined daily doses (cddds) prescribed during the study period. The use of cddd is recommended by the World Health Organization and is used widely for comparison of drug exposure. Briefly, information regarding the dates of prescriptions ordered, number of days prescribed, number of pills per prescription, and number of refills ordered were collected. Using this information, the cddd of PPIs and statins was calculated for each subject as described previously. (19) PRIMARY OUTCOME Our primary outcome of interest was the development of incident HCC, defined as the presence of at least one inpatient or two outpatient ICD-9 or ICD- 10 codes for HCC made 3 months after baseline. This approach has been recently validated within VA data compared to medical record review. (20) Time to development of HCC was determined relative from baseline (i.e., treatment initiation). Given that advanced fibrosis is one of the strongest predictors of HCC development, we performed our primary analysis on persons in our cohort with baseline cirrhosis (FIB-4 >3.5). 2246

HEPATOLOGY, Vol. 67, No. 6, 2018 LI ET AL. STATISTICAL ANALYSIS Baseline demographic and clinical factors were compared between the DAA, IFN, and untreated groups using chi-square tests for categorical variables and Wilcoxon rank-sum or t tests, as appropriate, for continuous variables. We calculated the unadjusted incidence rate of HCC per 1,000 person-years for all three treatment groups. Kaplan-Meier curves were generated to demonstrate time to development of HCC by SVR status for each treatment regimen, stratified by the presence of HCC risk factors at baseline, including cirrhosis. For our primary analysis, we performed incidence rate calculations and Cox regression among persons with baseline cirrhosis (FIB-4 >3.5) to examine the risk of HCC in the DAA group compared to the IFN group. In a secondary analysis, all treated persons were included, regardless of FIB-4 score. We also performed a sensitivity analysis in which we performed incidence rate calculations after only including those with FIB-4 scores that were calculated from values collected with 12 months of baseline. We performed an additional secondary analysis using multivariable Cox regression analysis to identify factors that were independently associated with HCC. We also performed a sensitivity analysis in which we performed Kaplan-Meier analysis to assess the probability of HCC development in individuals with specific risk factors for HCC, including age and baseline AFP levels. SAS (version 9.4; SAS Institute Inc., Cary, NC) and STATA software (version 11; StataCorp LP, College Station, TX) were used for statistical analyses. Results BASELINE CHARACTERISTICS From the ERCHIVES database, we identified 17,836 persons that met criteria for inclusion in this study (Fig. 1). We divided this cohort into three groups based on treatment exposure. A comparison of the baseline demographics and clinical characteristics of these populations is shown in Table 1. Overall, all three groups were predominantly male and had clinically similar median HCV-RNA levels and body mass index (BMI). Notably, DAA-treated persons were significantly older (62 vs. 54 years; P < 0.01) and more commonly had cirrhosis (19.9% vs. 13.1%; P < 0.01) compared to persons treated with IFN. In addition, DAA-treated and untreated persons more frequently had diabetes, higher median AFP levels, and higher PPI and statin use compared to IFN-treated persons. SVR was attained in 66.6% and 96.2% of persons in the IFN and DAA groups, respectively. Mean followup time was 2,719.2 days for IFN-treated persons and 396.4 days for DAA-treated persons. There were 196 incident HCC events among the IFN-treated persons and 50 incident HCC events among the DAA-treated persons (Table 1). EFFECT OF TREATMENT REGIMEN ON INCIDENT HCC To help account for the effect of baseline HCC risk factors (i.e., cirrhosis) in the risk of HCC development in both treated populations, we performed our primary analysis on persons with cirrhosis. For persons with cirrhosis who achieved SVR, we found no significant difference in HCC incidence rate between the DAAand IFN-treated groups (22.8 vs. 21.2 per 1,000 person-years; P 5 0.78; Table 2). In contrast, untreated persons with cirrhosis had a significantly higher HCC incidence rate than either of the two treatment groups (45.31 per 1,000 person-years; P 5 0.03). Among persons with cirrhosis who did not achieve SVR, the incidence of HCC was higher than those who did achieve SVR, although there was no significant difference between the rates in the DAA- and IFN-treated groups (62.81 vs. 48.90 per 1,000 personyears; P 5 0.59). We next performed Kaplan-Meier analyses to determine the probability of HCC development in persons with cirrhosis who achieved SVR in each of the treatment groups. Persons with cirrhosis who achieved SVR with either DAA- or IFN-based therapy had lower probability of developing HCC compared to the untreated group (log-rank, P 5 0.0004; Fig. 2). We also performed a secondary analysis in which we assessed HCC incidence in the entire cohort, which included persons without cirrhosis. Interestingly, among all persons who achieved SVR, the DAAtreated group had a significantly higher HCC incidence rate compared to the IFN-treated group (7.41 vs. 3.48 per 1,000 person-years; P < 0.01; Supporting Table S1). However, among individuals who did not achieve SVR, the HCC incidence rates between the two groups were not significantly different (P 5 0.44). We next performed a sensitivity analysis in which we calculated HCC incidence rates after limiting our definition of baseline FIB-4 to within 12 months before baseline. Overall, 1,933 values for FIB-4 were 2247

LI ET AL. HEPATOLOGY, June 2018 TABLE 1. Baseline Cohort Characteristics IFN Group DAA Group Group A Group B Untreated Group C P Value N 5 3,534 N 5 5,834 N 5 8,468 A vs. B A vs. C Age in years, median (IQR) 54 (50,57) 62 (58,65) 58 (54,62) <0.01 <0.01 Sex, % male 95.9 96.6 97.2 0.12 <0.01 Race, % White 67.4 51.4 49.8 <0.01 <0.01 Black 16.6 30.7 35.3 Hispanic 5.5 2.7 6.3 Other/missing 10.5 15.2 8.6 Diabetes, % 17.6 30.9 28.9 <0.01 <0.01 BMI, kg/m 2, median 28.37 (24.96, 32.48) 28.02 (24.36, 32.52) 26.93 (23.24,31.09) 0.02 <0.01 Alcohol abuse history, % 24.8 20.8 32.8 <0.01 <0.01 Smoking history, % <0.01 <0.01 Current smoker 60.8 49.7 68.1 Former smoker 22.9 32.4 18.2 Never smoker 16.3 17.9 13.7 HCV genotype, % 1 28.5 84.5 51.1 <0.01 <0.01 2 15.8 0.2 9.3 3 9.3 3.3 4.5 4, 5, 6 0.6 0.5 0.6 Missing 45.8 11.5 34.5 Baseline median HCV RNA (log 10 ) 6.11 (5.50, 6.60) 6.27 (5.77, 6.65) 6.20 (5.59, 6.64) <0.01 <0.01 PPI use during treatment 38.2 37.1 54.0 <0.01 <0.01 Statin use (ever before to baseline) 17.7 33.0 28.9 <0.01 <0.01 FIB-4 at baseline, % <1.45 (no fibrosis) 45.7 36.9 48.7 <0.01 <0.01 >3.5 (cirrhosis) 13.1 19.9 14.6 1.45-3.50 41.2 43.2 36.7 Median AFP, IU/mL, IQR 3.20 (2.20, 5.30) 4.30 (2.78, 6.90) 4.01 (2.52, 7.18) <0.01 <0.01 Ribavirin use, % 100 37.3 0 <0.01 <0.01 Treatment duration, weeks (days) <0.01 <8 (29-49) 7.8 0.6 8, % (50-70) 7.0 21.5 12, % (71-126) 19.2 74.1 24, % (>126) 66.0 3.8 Attainment of SVR, % 66.6 96.2 <0.01 Incident cases of HCC, N (%) 196 (5.55) 50 (0.86) 436 (5.04) <0.01 0.25 excluded in this sensitivity analysis to meet this criterion (209 from the IFN group, 82 from the DAA group, and 1,642 from the untreated group). Overall, 13.3% of IFN-treated persons were classified as with cirrhosis compared to 20% of DAA-treated persons, similar to what was observed in our primary cohort (Supporting Table S2). Among those included in the sensitivity group, there were 183 incident cases of HCC in IFN-treated persons and 48 incident cases among DAA-treated persons. We also limited our definition of baseline AFP to only those values collected within 12 months before baseline. A significant number of persons (n 5 14,960) in our cohort did not have an AFP within 12 months before baseline, but baseline AFP after these exclusions among the three groups was also quite similar to our primary cohort. In this sensitivity analysis, DAA treatment was associated with a significantly lower incidence rate of HCC compared to IFN treatment (17.90 vs. 31.99 per 1,000 person-years; P < 0.01) among all persons with cirrhosis. This apparent difference by treatment group disappears, however, when comparing only persons with cirrhosis who achieved SVR (16.76 vs. 19.77 per 1,000 person-years, respectively; P 5 0.51; Supporting Table S3). Furthermore, persons with cirrhosis who did not achieve SVR had a higher incidence of HCC than those who did achieve SVR, regardless of which type of therapy they failed. This was significant for persons treated with IFN (46.65 vs. 19.77 per 1,000 person-years), but not quite significant for DAAs (38.03 vs. 16.76 per 1,000 personyears), likely because of the very small number of DAA-treated persons who did not achieve SVR. Similar to treated persons who did not achieve SVR, persons with cirrhosis who did not receive any 2248

HEPATOLOGY, Vol. 67, No. 6, 2018 LI ET AL. FIG. 1. Flowchart of cohort construction. treatment had a higher HCC incidence rate compared to either of the two treatment groups. EFFECT OF DAA TREATMENT REGIMENS ON INCIDENT HCC We next performed a separate subgroup analysis investigating the incidence of HCC in persons treated with specific IFN-free DAA treatment regimens. These included an early DAA regimen, SOF/SMV, and the more recently approved SOF/LDV. SOF/ DCL was not included because of low patient numbers, and PRoD was not included because of insufficient follow-up time. The baseline characteristics for this subgroup are shown in Supporting Table S4. Persons treated with SOF/SMV had significantly higher baseline rates of cirrhosis (39.6%) compared to those treated with IFN (13.1%; P < 0.01) and SOF/LDV (18.9%; P < 0.01). Among persons with cirrhosis who achieved SVR, there was a trend toward a higher HCC rate in the SOF/ SMV group compared to the SOF/LDV group (33.26 vs. 15.49 per 1,000 person-years; P 5 0.07; Table 3) and a weaker trend toward a higher HCC rate in the SOF/ SMV group compared to the IFN group (33.26 vs. 21.20 per 1,000 person-years; P 5 0.18). Consistent with these analyses, Kaplan-Meier analysis revealed that SVR after SOF/SMV may be associated with higher probability of HCC development compared to SVR after SOF/LDV and or IFN (Supporting Fig. S1). These results were robust to a sensitivity analysis with a more restrictive definition of baseline FIB-4 as described before (Supporting Table S5). RISK FACTORS FOR HCC DEVELOPMENT We also identified risk factors for the development of HCC using a multivariate Cox proportional hazards 2249

LI ET AL. HEPATOLOGY, June 2018 TABLE 2. Incidence Rate of HCC in Persons With Cirrhosis, by Treatment Group and SVR Status Treatment Group Incidence Rate, per 1,000 Patient-Years 95% CI P Value Using IFN as Control All persons with cirrhosis IFN 34.70 28.50, 42.25 DAA regimens, overall 25.16 17.98, 35.21 0.11 Untreated controls 45.31 39.42, 52.08 0.03 SVR12 subgroup IFN 21.20 14.91, 30.14 DAA regimens, overall 22.80 15.84, 32.81 0.78 Non-SVR12 subgroup IFN 48.90 38.56, 62.02 DAA regimens, overall 62.81 26.14, 150.9 0.59 TABLE 3. Incidence Rate of HCC in Persons With Cirrhosis, by DAA Treatment Regimen and SVR Status Treatment Group Incidence Rate, per 1,000 Patient-Years 95% CI P Value Using IFN as Comparator Persons with cirrhosis IFN 34.70 28.50, 42.25 SOF/SMV 34.84 20.23, 60.00 0.99 SOF/LDV 19.55 11.79, 32.43 0.04 SVR12 subgroup IFN 21.20 14.91, 30.14 SOF/SMV 33.26 18.89, 58.56 0.18 SOF/LDV 15.49 8.58, 27.97 0.37 Non-SVR12 subgroup IFN 48.90 38.56, 62.02 SOF/SMV 81.24 11.44, 576.7 0.61 SOF/LDV 69.87 26.22, 186.2 0.49 FIG. 2. Comparison of probability of HCC development in persons with cirrhosis who achieve SVR between treatment groups by Kaplan Meier analysis. model. Among cirrhotic persons with cirrhosis, DAA treatment was not associated with a higher risk of HCC compared to IFN treatment (hazard ratio [HR], 1.07; 95% confidence interval [CI], 0.55, 2.08; Table 4). The risk of incident HCC was higher, however, among those with known HCC risk factors, including older age (HR, per 10-year increase, 1.76; 95% CI, 1.26, 2.46) and AFP >20 (HR, 4.10; 95% CI, 2.75, 6.10) at baseline. Statin use was associated with a lower risk of HCC (HR, 0.50; 95% CI, 0.31, 0.80), (16) and PPI use was associated with a higher risk of HCC (HR, 1.65; 95% CI, 1.07, 2.55). We also performed supplemental Kaplan-Meier analyses of HCC-free survival of other groups of DAA or IFN SVR persons at inherently higher risk of HCC, including those with older age and those with higher baseline AFP values. In each of these analyses, there was no difference in TABLE 4. Predictors for the Development of HCC in Persons With Cirrhosis Based on Multivariate Cox Proportional Hazards Analysis HR 95% CI Age, per 10-year increase 1.76 1.26, 2.46 Male sex 1.34 0.32, 5.62 Race, White (comparator) 1 Black 0.65 0.37, 1.13 Hispanic 0.94 0.45, 1.95 Other/missing 1.27 0.77, 2.10 Diabetes 1.01 0.67, 1.50 BMI, per 1-unit increase 0.98 0.95, 1.01 Alcohol abuse history 1.27 0.85, 1.89 Smoking history No smoker (comparator) 1 Current smoker 1.44 0.79, 2.60 Former smoker 1.76 0.96, 3.23 Missing 1.65 0.82, 3.33 HCV genotype, % 1 (comparator) 1 2 0.39 0.16, 0.95 3 0.90 0.47, 1.72 4, 5, 6 No events N/A Missing 0.81 0.53, 1.24 HCV RNA, per 1 log 10 increase 0.96 0.83, 1.11 PPI use (baseline onward) 1.65 1.07, 2.55 Statin use (baseline onward) 0.50 0.31, 0.80 AFP >20 (vs. < 5 20) 4.10 2.75, 6.10 Treatment regimen PEG/RBV(comparator) 1 Any DAA 1.07 0.55, 2.08 Attainment of SVR 0.66 0.42, 1.03 Abbreviation: N/A, not applicable. HCC-free survival between the two treatment groups (Supporting Fig. S2A,B). Though achievement of SVR was not associated with a significantly lower HCC rate in our original analysis, when we performed a multivariable analysis of the data after limiting the 2250

HEPATOLOGY, Vol. 67, No. 6, 2018 LI ET AL. baseline FIB-4 to within a 12-month window before baseline, SVR was statistically significant (HR 5 0.60; 95% CI, 0.38, 0.93; Supporting Table S6). Discussion The availability of highly effective DAAs has revolutionized HCV treatment and has been anticipated to further reduce the incidence of HCC given extremely high rates of SVR. However, recent conflicting data regarding whether the risk of HCC after DAA treatment is higher has made it critically important to determine whether this is, in fact, true given the millions of patients who could potentially benefit from DAA therapies. In this study, we analyzed the association between DAA and IFN treatment and HCC incidence among individuals with chronic HCV while controlling for other clinically relevant risk factors for HCC that we believed were not sufficiently controlled for in previous analyses that suggested the HCC rate was higher after DAA treatment. In our cohort, we found that DAA-treated persons did have a significantly higher HCC rate than IFNtreated persons (Supporting Table S1), but they also had a significantly higher rate of known risk factors for HCC, including cirrhosis, older age, and higher baseline AFP level. When we performed our primary analysis in persons with cirrhosis alone, we found no difference in HCC-free survival between the DAA and IFN groups. This finding strongly suggests that the pretreatment risk of HCC is what drives the posttreatment risk of HCC. Observed differences in the incidence of HCC between DAA- and IFN-treated groups are therefore very unlikely to be related to the treatment itself. To explain the higher prevalence of underlying HCC risks among DAA-treated individuals in our cohort, we evaluated for an effect of the common practice of warehousing of patients, that is, awaiting curative treatment for those who could not be cured with existing (i.e., IFN) treatment, (21-23) on prevalence of HCC risk factors. We analyzed the treatment trends of our cohort using a bar plot (Supporting Fig. S3). This plot shows that although IFN was the primary treatment of those who entered our cohort before the FDA approval of SOF, SOF-based treatments quickly overtook IFN as the preferred treatment modality in 2014, as soon as both SOF and SMV became available in the VA system. We therefore hypothesized that treatment with the earliest DAA regimens, specifically SOF/SMV, which was used to treat those who had previously failed treatment with IFN, may function as a surrogate for greater baseline risk for HCC, therefore confounding the association between the earliest DAA regimens and HCC incidence in historical patient cohorts. We further hypothesized that the next-wave DAA regimens were used to treat patients with comparatively lesser degrees of underlying liver disease, and would therefore be associated with a lower risk of HCC compared to the first wave of DAA. Indeed, in our analysis, we found that persons treated with SOF/ SMV, the first combination DAA that was FDA approved, had higher rates of cirrhosis compared not only to IFN treated persons, but also to those treated with more recent DAA-based therapies, including SOF/LDV, which became available a year after SOF/ SMV. Furthermore, SOF/SMV-treated persons had the highest risk of HCC compared not only to IFNtreated persons, but also to those treated with SOF/ LDV. Similar to our findings, Conti et al. also found that treatment of persons with cirrhosis with the earliest DAA regimens, including SOF/RBV, was associated with higher rates of HCC (10 of 52; 19.2%) compared to those treated with the subsequently approved regimens SOF/LDV (1 of 24; 4.2%) and PRoD (2 of 52; 3.8%). (9) However, Conti et al. did not present detailed baseline characteristics for each of the DAA treatment subgroups. Similarly, several other studies that have reported a higher rate of HCC after DAA treatment have not provided baseline characteristics of subgroups treated with various DAA combinations, likely attributed to the overall small numbers in these studies. (8,10-12) Treatment with early DAA regimens may therefore represent a novel confounding factor contributing to selection bias that has not been reported nor rigorously adjusted for in past studies. Thus, in contrast to reports in which a higher rate of HCC was found among DAA-treated persons, we clearly show that DAA treatment is not associated with a higher risk of HCC development compared to IFN treatment when controlling for differences in baseline HCC risk factors between the treatment groups. Importantly, our results support the concept that treatment and achievement of SVR with any antiviral regimen is the most important determinant of a lower HCC risk. Among all persons with cirrhosis, untreated controls had significantly higher rates of HCC compared to both DAA- and IFN-treated groups (Table 3). Moreover, the lower HCC risk is most likely attributable primarily to the attainment of SVR. Compared 2251

LI ET AL. HEPATOLOGY, June 2018 to treated persons with cirrhosis who attained SVR, lack of SVR achievement was associated with significantly higher incidence rates of HCC in IFN-treated persons (48.90 vs. 21.20 per 1,000 person-years; Table 2). In DAA-treated persons, there was a trend toward statistical significance (Table 2), but the very small number of treatment failures (n 5 222; 3.8%) strongly limited the statistical power of this comparison. The strength of our work lies in the large national cohort used along with long length of follow-up as well as serial measurements of many detailed baseline characteristics. Moreover, our adjustment for baseline risk factors for HCC (cirrhosis, Child-Pugh stage, age, diabetes mellitus, history of alcohol abuse, and baseline AFP level) was not performed in many of the studies that reported an increased risk of HCC. Our study has a number of limitations. A key limitation of our study is the relatively short follow-up time in the DAA cohort, a result of the recent introduction of DAAs into the HCV treatment model. Future studies should be performed as the cohort of DAA-treated persons matures to ensure that our findings are durable. Our study may also be limited by its generalizability, given that our cohort was comprised of veterans who are almost all male and have higher rates of smoking and alcohol use than the general population. Our study is also limited by its design as a retrospective cohort study; however, prospective studies comparing DAA therapy with IFN would be unethical because DAAs have essentially replaced IFN in the treatment of HCV. In addition, we used a clinical score (the FIB-4 score), rather than liver biopsy, to identify persons with cirrhosis. However, the FIB-4 is a validated and widely used marker of liver fibrosis progression in chronic HCV (24) and has been frequently used in epidemiological studies as a surrogate for a diagnosis of cirrhosis. (25,26) We also could not address the contribution of surveillance practices to the perceived elevated risk of HCC following DAA treatment because these data were not available to us. We also note that treatment assignment in this cohort was confounded by calendar time, because IFN was rarely used as a treatment after the availability of DAAs, but there is no available evidence to suggest that there is a biological interaction between calendar time and treatment regimen. Finally, we excluded persons with a past diagnosis of HCC and so could not address the question of whether there is a higher recurrence rates of HCC after treatment with DAA, which has been reported in some, (8,10,11) but not all studies. (13,27,28) We took this approach because we believe that de novo occurrence is the more clinically relevant issue, and the risk of recurrence following HCC will be higher than the de novo rate regardless of HCV treatment type. In conclusion, in persons with cirrhosis with chronic HCV infection, we found no association between HCC incidence rates and DAA treatment compared to IFN treatment in this large, retrospective cohort study. Furthermore, our secondary analyses suggested that both insufficient controlling for baseline risks for HCC and selection bias likely drove the reported associations between DAA treatment and increased HCC risk. Acknowledgments: Guarantor of the article: Drs. Butt and Chung had complete access to data at all times and accept responsibility for the integrity of this article. All authors approved the final version of the manuscript. REFERENCES 1) Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo M, et al. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer 2015;136:E359-E386. 2) Li DK, Chung RT. Impact of hepatitis C virus eradication on hepatocellular carcinogenesis. Cancer 2015;121:2874-2882. 3) Baumert TF, J uhling F, Ono A, Hoshida Y. Hepatitis C-related hepatocellular carcinoma in the era of new generation antivirals. BMC Med 2017;15:52. 4) Morgan RL, Baack B, Smith BD, Yartel A, Pitasi M, Falck- Ytter Y. Eradication of hepatitis C virus infection and the development of hepatocellular carcinoma: a meta-analysis of observational studies. Ann Intern Med 2013;158(5 Pt 1):329-337. 5) Nahon P, Bourcier V, Layese R, Audureau E, Cagnot C, Marcellin P, et al. Eradication of hepatitis C virus infection in patients with cirrhosis reduces risk of liver and non-liver complications. Gastroenterology 2017;152:142-156.e2. 6) Cramp ME, Rosenberg WM, Ryder SD, Blach S, Parkes J. Modelling the impact of improving screening and treatment of chronic hepatitis C virus infection on future hepatocellular carcinoma rates and liver-related mortality. BMC Gastroenterol 2014; 14:137. 7) Harris RJ, Thomas B, Griffiths J, Costella A, Chapman R, Ramsay M, et al. Increased uptake and new therapies are needed to avert rising hepatitis C-related end stage liver disease in England: modelling the predicted impact of treatment under different scenarios. J Hepatol 2014;61:530-537. 8) Reig M, Mari~no Z, Perello C, I~narrairaegui M, Ribeiro A, Lens S, et al. Unexpected high rate of early tumor recurrence in patients with HCV-related HCC undergoing interferon-free therapy. J Hepatol 2016;65:719-726. 9) Conti F, Buonfiglioli F, Scuteri A, Crespi C, Bolondi L, Caraceni P, et al. Early occurrence and recurrence of hepatocellular carcinoma in HCV-related cirrhosis treated with direct-acting antivirals. J Hepatol 2016;65:727-733. 2252

HEPATOLOGY, Vol. 67, No. 6, 2018 LI ET AL. 10) Kozbial K, Moser S, Schwarzer R, Laferl H, Al-Zoairy R, Stauber R, et al. Unexpected high incidence of hepatocellular carcinoma in cirrhotic patients with sustained virologic response following interferon-free direct-acting antiviral treatment. J Hepatol 2016;65:856-858. 11) Yang JD, Aqel BA, Pungpapong S, Gores GJ, Roberts LR, Leise MD. Direct acting antiviral therapy and tumor recurrence after liver transplantation for hepatitis C-associated hepatocellular carcinoma. J Hepatol 2016;65:859-860. 12) Ravi S, Axley P, Jones D, Kodali S, Simpson H, McGuire BM, et al. Unusually high rates of hepatocellular carcinoma after treatment with direct-acting antiviral therapy for hepatitis C related cirrhosis. Gastroenterology 2017;152:911-912. 13) ANRS Collaborative Study Group on Hepatocellular Carcinoma (ANRS CO22 HEPATHER, CO12 CirVir and CO23 CUPILT cohorts). Lack of evidence of an effect of direct-acting antivirals on the recurrence of hepatocellular carcinoma: data from three ANRS cohorts. J Hepatol 2016;65:734-740. 14) Romano A, Capra F, Piovesan S, Chemello L, Cavalletto L, Anastassopoulos G, et al. Incidence and pattern of de novo hepatocellular carcinoma in HCV patients treated with oral DAAs. HEPATOLOGY 2016;64(S1):10A. 15) Butt AA, Yan P, Shaikh OS, Chung RT, Sherman KE; ERCHIVES study. Treatment adherence and virological response rates in hepatitis C virus infected persons treated with sofosbuvir-based regimens: results from ERCHIVES. Liver Int 2016;36:1275-1283. 16) Butt AA, Yan P, Bonilla H, Abou-Samra AB, Shaikh OS, Simon TG, et al. Effect of addition of statins to antiviral therapy in hepatitis C virus-infected persons: Results from ERCHIVES. HEPATOLOGY 2015;62:365-374. 17) Butt AA, Ren Y, Marks K, Shaikh OS, Sherman KE; ERCHIVES study. Do directly acting antiviral agents for HCV increase the risk of hepatic decompensation and decline in renal function? Results from ERCHIVES. Aliment Pharmacol Ther 2017;45:150-159. 18) Butt AA, Yan P, Lo Re V, Rimland D, Goetz MB, Leaf D, et al. Liver Fibrosis progression in hepatitis C virus infection after seroconversion. JAMA Intern Med 2015;175:178. 19) Simon TG, Bonilla H, Yan P, Chung RT, Butt AA. Atorvastatin and fluvastatin are associated with dose-dependent reductions in cirrhosis and hepatocellular carcinoma, among patients with hepatitis C virus: results from ERCHIVES. HEPATOLOGY 2016;64:47-57. 20) Omino R, Mittal S, Kramer JR, Chayanupatkul M, Richardson P, Kanwal F. The Validity of HCC Diagnosis Codes in Chronic Hepatitis B Patients in the Veterans Health Administration. Dig Dis Sci 2017;62:1180-1185. 21) Alberti A, Colombo M, Craxı A, Rizzetto M. The dilemma for patients with chronic hepatitis C: treat now or warehouse? Dig Liver Dis 2014;46:27-29. 22) Suda KJ, Halbur DJ, Hunkler RJ, Matusiak LM, Schumock GT. Spending on Hepatitis C Antivirals in the United States, 2009-2015. Pharmacotherapy 2017;37:65-70. 23) Tambourine BM, Sadeghi A, Yang J, Stockl KM, Lew HC, Solow BK, et al. Patient Characteristics and Prescribing Patterns Associated with Sofosbuvir Treatment for Chronic HCV Infection in a Commercially Insured Population. Am Heal Drug Benefits 2016;9:74-82. 24) Vallet-Pichard A, Mallet V, Nalpas B, Verkarre V, Nalpas A, Dhalluin-Venier V, et al. FIB-4: an inexpensive and accurate marker of fibrosis in HCV infection. comparison with liver biopsy and fibrotest. HEPATOLOGY 2007;46:32-36. 25) Berenguer J, Zamora FX, Aldamiz-Echevarria T, Von Wichmann MA, Crespo M, Lopez-Aldeguer J, et al. Comparison of the Prognostic Value of Liver Biopsy and FIB-4 Index in Patients Coinfected With HIV and Hepatitis C Virus. Clin Infect Dis 2015;60:950-958. 26) Klevens RM, Canary L, Huang X, Denniston MM, Yeo AE, Pesano RL, et al. The Burden of Hepatitis C Infection Related Liver Fibrosis in the United States. Clin Infect Dis 2016;63: 1049-1055. 27) Torres HA, Vauthey JN, Economides MP, Mahale P, Kaseb A. Hepatocellular carcinoma recurrence after treatment with directacting antivirals: First, do no harm by withdrawing treatment. J Hepatol 2016;65:862-864. 28) Virlogeux V, Pradat P, Hartig-Lavie K, Bailly F, Maynard M, Ouziel G, et al. Direct-acting antiviral therapy decreases hepatocellular carcinoma recurrence rate in cirrhotic patients with chronic hepatitis C. Liver Int 2017;37:1122-1127. Supporting Information Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.29707/suppinfo. 2253