Race affects SVR12 in a large and ethnically diverse hepatitis C‐infected patient population following treatment with direct‐acting antivirals: Analysis of a single‐center Department of Veterans Affairs cohort

Abstract Hepatitis C virus (HCV) infection is a major cause of chronic liver disease. HCV cure has been linked to improved patient outcomes. In the era of direct‐acting antivirals (DAAs), HCV cure has become the goal, as defined by sustained virological response 12 weeks (SVR12) after completion of therapy. Historically, African‐Americans have had lower SVR12 rates compared to White people in the interferon era, which had been attributed to the high prevalence of non‐CC interleukin 28B (IL28B) type. Less is known about the association between race/ethnicity and SVR12 in DAA‐treated era. The aim of the study is to evaluate the predictors of SVR12 in a diverse, single‐center Veterans Affairs population. We conducted a retrospective study of patients undergoing HCV therapy with DAAs from 2014 to 2016 at the VA Greater Los Angeles Healthcare System. We performed a multivariable logistic regression analysis to determine predictors of SVR12, adjusting for age, HCV genotype, DAA regimen and duration, human immunodeficiency virus (HIV) status, fibrosis, nonalcoholic fatty liver disease (NAFLD) fibrosis score, homelessness, mental health, and adherence. Our cohort included 1068 patients, out of which 401 (37.5%) were White people and 400 (37.5%) were African‐American. Genotype 1 was the most common genotype (83.9%, N = 896). In the adjusted models, race/ethnicity and the presence of fibrosis were statistically significant predictors of non‐SVR. African‐Americans had 57% lower odds for reaching SVR12 (adj.OR = 0.43, 95% CI = 1.5‐4.1) compared to White people. Advanced fibrosis (adj.OR = 0.40, 95% CI = 0.26‐0.68) was also a significant predictor of non‐SVR. In a single‐center VA population on DAAs, African‐Americans were less likely than White people to reach SVR12 when adjusting for covariates.

African-Americans were less likely than White people to reach SVR12 when adjusting for covariates.

K E Y W O R D S
Adherence, direct-acting antivirals, direct drug acting, drug metabolism, ethnicity, Hepatitis C, medication procession ratio, polymorphisms, race, racial disparity, sustained virological response 12, Veterans Affairs

| INTRODUCTION
Hepatitis C remains a burden with at least 5.2 million infected people worldwide. 1 Long-term infection can lead to complications including cirrhosis, hepatocellular carcinoma, and death. 2 Eradication of HCV has become a focus given that sustained virological response (SVR) has been associated with reversal of hepatic fibrosis and decreased rates of HCC. 3,4 Historically, race has played a major role in chronic hepatitis C rates and treatment responses with African-Americans being disproportionately affected. 5,6 Host immune responses largely explained these differences as African-Americans were more likely to harbor the non-CC interleukin IL28B genotype which affected their response to pegylated interferon. 7,8 More recently, new direct-acting antivirals (DAAs) have revolutionized hepatitis C therapy given their ease of administration, tolerability and reported sustained virological response at 12 weeks (SVR12) with rates in the 90s, depending on the extent of liver fibrosis and genotype. 9 Despite these strides, racial differences in treatment outcomes remain. In large cohorts of patients with genotype 1 treated with sofosbuvir/ledipasvir with or without ribavirin, Africans-Americans were shown to have higher rates of relapse, although these differences disappeared after extending therapy from 8 to 12 weeks. 10 These data were corroborated in a larger Veterans Affairs cohort where African-American patients treated with 12 weeks of therapy instead of 8 reached SVR12 rates similar to their White counterparts. 11 More recently, Su and colleagues demonstrated that African-Americans in the VA treated with DAAs have lower SVR rates than White Veterans. In a multivariable regression model, adjusting for factors that can affect SVR, the adjusted odds ratio for African-Americans was 0.77 for reaching SVR compared to White people (95% CI = 0.69-0.87), where 29% of the population was African-American (N = 6171). 12 Whether these race-based differences in SVR, particularly with abbreviated regimens, both in the interferon and DAA eras are due to genetic or other socioeconomic factors remains to be elucidated.
The disproportionate HCV infection rates in African-Americans and higher risk of developing HCC independently of fibrosis underscores the importance of understanding why these differences exist. 13

| Study population
The study population consisted of consecutive HCV-infected patients who received DAA therapy at VA Greater Los Angeles Healthcare System. Patients without SVR data available 12 or more weeks following antiviral therapy were excluded from analysis. All genotypes 1-6 were included. Choice of DAA regimen was at the discretion of the provider. On-treatment and posttreatment monitoring followed an established protocol that included serum SVR evaluation every 2-4 weeks.

| Sustained virological response
The primary outcome of our study was SVR12, which was defined as an undetectable HCV RNA (<15 IU/mL) 12 weeks or beyond the conclusion of treatment. 15

| Baseline characteristics
Baseline demographic variables obtained at the initiation of therapy included: age, self-reported race and ethnicity; HCV genotype; nonalcoholic fatty liver (NAFLD) fibrosis score (<À1.455 being unlikely to have advanced NAFLD fibrosis vs >0.676 being predictive of advanced NAFLD-associated fibrosis) 16 ; the Fibrosis-4 (Fib4; which will be referred to as advanced fibrosis from here forward), as a marker of advanced liver disease using the formula (age x aspartate aminotransferase)/(platelets X alanine aminotransferase 1/2 ) 17 ; body mass index (BMI) (≥30 kg/m 2 and <30 kg/m 2 ); HIV status; and HCV treatment status (na€ ıve or experienced). For race and ethnicity, we used a single variable that combines concepts of race and ethnicity into five mutually exclusive categories for race/ethnicity: non-Hispanic White, non-Hispanic Black (African-Americans), Hispanics, Asians, and Unknown/Other. The presence of diabetes and its complications were determined by ICD-9 (250.00-250.92) and ICD-10 codes, as were diagnoses of hypertension, hyperlipidemia, HIV, and AIDS with their complications. Psychiatric disorders, both organic and nonorganic (associated with substance abuse or not) were included in our analysis as was substance abuse and homelessness.

| Medication adherence
Patient adherence was assessed by calculating the medication procession ratio (MPR) since this has been a validated method to determine adherence. 18 MPR was defined by the total number of pills supplied over the total number of pills expected to be dispensed by the pharmacy department based on length of treatment regimen and genotype. For purposes of simplicity, "MPR" will be defined and described as "adherence" from here onward.

| Statistical analysis
Demographic data which included sex, race/ethnicity, DAA regimen, body mass index (BMI), advanced fibrosis, and genotype were summarized with frequencies and chi-square tests for comparisons. We conducted multivariable logistic regression analysis to model predictors of SVR12. A priori covariates selected for the model were age, race/ethnicity, genotype, treatment regimen, treatment length, being treatment na€ ıve, HIV status, advanced fibrosis, NAFLD fibrosis score, and adherence as defined by adherence. Age and adherence were continuous variables in the analysis. To analyze the adherence data further, we stratified adherence into the following groups: ≥90%, 80%-89%, 60%-79%, and <60%. In addition to using adherence as a continuous variable we also performed regression models with MRP with the defined groups as a categorical covariate. Prior to regression analyses, we tested for multiple collinearity, and no covariates were collinear as defined as a variance inflation factor of less than 10. Separate logistic multivariable regressions models were used to model SVR predictors in the African-American subgroup by assessing treatment length of 8-versus 12-week treatment, removing what is now considered suboptimal therapies (sofosbuvir/simeprevir AE ribavirin and sofosbuvir/ribavirin). We also performed regression models by race/ethnicity, genotype, and adherence on SVR12.  Table 1. Males comprised 97% of the study cohort, consistent with gender demographics within the VA system. The mean age was 61.8 (SE AE 0.2). White people and African-Americans were equally represented at 37.8% of the population. Genotype 1 (1a and 1b) was the most common genotype in 83.9% of patients (N = 896), followed by genotype 2 (7.9%, N = 84) and genotype 3 (6.9%, N = 74). None of the patients had DAA treatment regimen allocations for all patients and the corresponding SVR12 for each genotype are summarized in Table 2.
The most common regimen was sofosbuvir/ledipasvir AE ribavirin, which occurred for 47.8% of the population. Since the study cohort

| Predictors of SVR12
Predictors of SVR12 from adjusted regression models are summarized in Table 3. There were two clinically significant negative pre-

| SVR12 in genotype 1 patients
Given the heterogeneity of our patient population in genotype and therapies, as well as the observation that genotype 3 is more difficult to eradicate, 19

| DISCUSSION
This study demonstrates that in a large ethnically diverse community-based Veterans Affairs practice, SVR12 rates for chronic hepatitis C are influenced by race/ethnicity and advanced liver disease, which corroborates previously published data. 10,12,20 In our cohort, the lower SVR rates observed in African-Americans relative to White people is persistent despite at least 12 weeks of therapy even when only using current "optimal" therapies. One important consideration in our analysis was the effect of adherence by measures of MPR in an ethnically diverse population treated with direct-acting antivirals, which has not been evaluated in detail previously. We find that adherence explains some of these differences. These data suggest potential underlying biological differences between White people and African-Americans in medication response.
While adherence and medication tolerability are not a concern in well-resourced large clinical trials, they can be more difficult to measure in real-world effectiveness data for chronic hepatitis C treat- One potential biological explanation for these observations is differences in drug metabolism, driven by the patients' genetic background. Drug metabolism differences have been described to exist between African-Americans and White people. Although some of these differences have been attributed to environmental factors such as diet and concomitant medications, intrinsic host factors such as genetic variability and gene polymorphisms in drug metabolizers such as CYP2D6 and CYP2C19 have also been described. 22,23 Understanding genetic polymorphisms in African-Americans and elucidating their mechanism of action, namely in the IL28B gene also offered great advances in understanding the underlying genetic differences between African-Americans and White people in the interferon era. 7,24 Other genetic differences such as the variant in HAVCR1 gene variant (rs6880859) were also subsequently identified. 25 Although the polymorphisms and genetic differences listed above have not been shown to affect SVR12 with DAAs, such undiscovered genetic polymorphisms could explain these findings.
The effects of genetic variants on SVR in the DAA era have not been addressed systematically other than to assess patterns of resistance. 26 The previous findings that extending therapy in African- Another possibility for the large MRP effects in African-Americans is medication tolerability/side effects, which could also be related to the intrinsic metabolism of these medications. Ideally, metabolite levels would have been checked to further assess this; however, given the retrospective nature of our study, this was not addressed.
Examining the true associations between socioeconomic status and adherence can be more challenging and would require a root-cause analysis.
In conclusion, we present a large, ethnically and medically heterogeneous population within the Veterans Affairs System and their SVRs rates in the DAA era. Although our data demonstrate the importance of racial/ethnic differences, the true etiology of these differences remains unclear, which can be further explored in prospective studies where drug levels and patient genetics are taken into account.