Sustained Virologic Response and Costs Associated with Direct-Acting Antivirals for Chronic Hepatitis C Infection in Oklahoma Medicaid

BACKGROUND: Outcomes involving newer direct-acting antiviral (DAA) hepatitis C virus (HCV) regimens have not been studied extensively among the Medicaid population. OBJECTIVE: To assess clinical (treatment failure) and economic outcomes for chronic HCV-infected Oklahoma Medicaid members following treatment with DAAs and to measure associations with patient, treatment, and clinical characteristics. METHODS: This cross-sectional study used Oklahoma Medicaid pharmacy and medical claims data for adult members who used a newer DAA agent and had reported a successful or failed sustained virological response rate 12 weeks after therapy completion (SVR12) from January 1, 2014, to June 30, 2016. Multivariable logistic and gamma regressions assessed predictors of SVR12 failure and costs controlling for member demographics (i.e., age, sex, race, rural residence); type of DAA and adherence; clinical characteristics (e.g., comorbid conditions, advanced liver disease); and the implementation of changes to a prior authorization program. RESULTS: Of 934 Medicaid members eligible for treatment with DAAs between January 1, 2014, and June 30, 2016, 906 received DAA treatment, 40.6% (368/906) had reported SVR12 outcomes, and 59.4% (n = 538) did not have a reported SVR recorded. Of those with reported SVR12 outcomes, patients were 53.1 ± 9.7 years of age, 51.1% were male, 8.4% had SVR12 failure, and each member had mean costs of $140,283 ± $52,779. Multivariable analyses indicated higher odds of SVR12 failure was independently associated with cirrhosis (OR [decompensated] = 6.69 and OR [compensated] = 3.52, P < 0.001), while males had higher odds of failure than females (OR = 3.34, P < 0.010). No significant difference in SVR12 failure was noted, according to DAA type or a medication adherence threshold of > 95%. Ledipasvir/sofosbuvir was independently associated with lower costs (exp[b] = 0.81; P < 0.001) compared with sofosbuvir, while higher costs were associated with decompensated cirrhosis (exp[b] = 1.22; P < 0.001) and treatment failure (exp[b] = 1.18, P < 0.010). In an analysis including members without reported SVR12 outcomes, decompensated and compensated cirrhosis had lower odds (P < 0.001) of no reported SVR12 from ambulatory clinic settings. CONCLUSIONS: Almost 60% of Medicaid members receiving DAA treatment did not have a final reported SVR12 outcome. Among those with viral load measurements, treatment success was high and both decompensated and compensated cirrhosis were independently associated with significantly higher odds of treatment failure. Addressing a loss to follow-up among HCV patients and curtailing the development of cirrhosis to improve treatment success may warrant interventions that improve access to care and remove barriers that impede treatment initiation and completion.

The purpose of this study, which focused on beneficiaries within a state Medicaid program, was to determine the clinical and economic outcomes for persons with chronic HCV infection following treatment and to assess the association between these outcomes and demographics, clinical characteristics, and DAA-related medication use from January 1, 2014, to June 30, 2016. Specific analyses also involved assessments of HCVrelated liver sequelae, medication adherence, and SVR12.

■■ Methods
This investigation of a historical cohort used comprehensive, patient-level medical and pharmacy claims data from an administrative claims database provided by the Oklahoma Health Care Authority (OHCA) from January 1, 2014, to June 30, 2016, to Pharmacy Management Consultants (PMC). [18][19][20] PMC, maintained by the University of Oklahoma College of Pharmacy, provides operational, consultant, and educational services to support OHCA in pharmacy benefit administration. The administrative data (i.e., claims) included information concerning medication use (e.g., type, quantity filled, days supply, date of service, prescriber, reimbursement amount); member demographics (e.g., age, sex, race, and residence type); and medical services (e.g., diagnoses, inpatient/outpatient/ED visits, dates of service, procedures, physician identifier, reimbursement amount).
These data were merged with additional clinical information reported by providers as required by OHCA's HCV prior authorization (PA) program. A PMC pharmacist oversees the PA process and maintains the database containing the clinical information for the HCV program. Throughout the course of therapy, providers report treatment regimen, initial viral load, treatment completion or discontinuation, and SVR12 outcomes. Oklahoma Medicaid initiated amendments to the PA program in July 2014 concerning DAA HCV medications, which included collection of additional information about treatment length, genotype, prior treatments, and fibrosis score. Notably, this revised PA program also implemented stricter treatment criteria, including a requirement for a fibrosis score of ≥ F2 and no alcohol abuse or unauthorized injection drug use in the past 6 months. The Office of Human Research Participant Protection (institutional review board) at the University of Oklahoma Health Sciences Center and OHCA granted approval to this study as exempt.
Members identified for the study consisted of adults aged ≥ 18 years who had (a) at least 1 HCV diagnosis, including chronic or acute cases (International lower education levels and lower family income. The most common HCV transmission method is through percutaneous exposure to infected blood, including injection drug use, blood transfusion before the advent of anti-HCV tests in 1992, organ transplantation, needlestick injuries in health care settings, and birth to an HCV-infected mother. 3 Treatment success for HCV infection is defined as sustained virological response (SVR) or an undetectable HCV RNA at 12 weeks after treatment (i.e., SVR12). 8 Numerous treatment options for HCV have recently been introduced to the market. Before the advent of the direct-acting antivirals (DAAs), combination pegylated interferon and ribavirin were a mainstay of treatment, which achieved SVR in 54%-61% of patients but also high withdrawal rates due to debilitating adverse reactions. 9 Two DAA protease inhibitors, boceprevir and telaprevir, were introduced in 2011 to treat HCV genotype 1 infection, 10,11 with SVR rates of 66%-68% for boceprevir and 69%-75% for telaprevir. 12 Despite higher SVR rates, these medications had relatively higher rates of adverse effects, including anemia and neutropenia. 13 In 2013, newer second-generation DAAs were introduced with substantially higher SVR rates (i.e., > 90%) and fewer adverse events than older DAAs. 14, 15 Additional agents have been approved since then. 16,17 The economic burden of HCV to the patient and to payers is substantial. A systematic literature review estimated the direct and indirect costs to payers of liver transplants at $201,110, HCC at $23,766-$44,200, variceal hemorrhage at $25,595, compensated cirrhosis at $585-$1,110, refractory ascites at $24,755, hepatic encephalopathy at $16,430, sensitive ascites at $2,450, moderate chronic HCV at $155, and mild chronic HCV at $145 per year per person. 5 An analysis of national discharges from the Agency for Healthcare Research and Quality Nationwide Inpatient Sample estimated that charges for hospitalizations with HCV increased from $0.9 billion in 2004-2005 to $3.5 billion during 2010-2011. 6 Payer types for patients hospitalized for HCV were primarily Medicare and Medicaid, with an increasing trend for Medicare. 6 Research evaluating health care utilization for HCV infection found > 2.3 million outpatient, 73,000 emergency department (ED), and 475,000 inpatient visits annually in the United States from 2001 to 2010. 7 The cost was greater than $15 billion annually, and most visits were covered by Medicare and Medicaid. 7 Newer DAAs present a major challenge to payers, especially public ones. Recent state Medicaid strategies to treat HCV, some of which have provoked much debate, include limiting access to patients with advanced liver disease (e.g., fibrosis stage 3 and 4, cirrhosis) and restricting access based on past alcohol and injection drug use. 18 As public payers are under pressure to more equitably manage HCV, real-world examples of HCV management and its outcomes become valuable to guide patient, payer, and provider decision making.

Sustained Virologic Response and Costs Associated with Direct-Acting
Antivirals for Chronic Hepatitis C Infection in Oklahoma Medicaid ledipasvir/sofosbuvir, daclatasvir, or dasabuvir/ombitasvir/ paritaprevir and ritonavir) during the study period, and (c) reported a successful or failed SVR12. A successful SVR12 was defined as an undetectable HCV RNA at 12 weeks after treatment as reported by the provider, whereas a failed SVR12 was defined as detected HCV RNA at 12 weeks.
During the course of the study, the large number of unreported SVR12s prompted further investigation. These members were analyzed as an additional subgroup. Members receiving telaprevir, boceprevir, or interferon-based regimens (i.e., interferon and ribavirin without a DAA) were excluded. Pharmacy and medical costs were compared between baseline and followup periods. These periods were demarcated by an index date, which was defined as the date of DAA initiation (i.e., the first date of service on administrative pharmacy claims). The baseline period included the 12-month pre-index date, and the follow-up period included the 12-month post-index date.
Outcomes of this study included SVR12 failure and total direct per member (PM) and per member per month (PMPM) pharmacy and medical costs. Costs were defined as the total amounts reimbursed by Medicaid as recorded in administrative claims. Only costs from the Medicaid database were available. Costs were inflation-adjusted to 2017 U.S. dollars using the prescription drugs and medical care services components of the Consumer Price Index (Appendix A, available in online article).
Following the amendments to the PA program on July 1, 2014, a treatment gap of > 3 days per month was considered a criterion for subsequent denial of therapy and was therefore included as a predictor variable. The calculation for the treatment gap was based on the regimen's specific treatment length and the length of time in days between fill dates plus days supply. Members with a length of time between fill dates plus days supply > 3 days were considered to have a gap. Members were determined to have discontinued treatment early if they did not have a days supply sufficient to fill their specific regimen's treatment length. Treatment length was either imputed or taken from their PA forms. Treatment lengths for members who initiated treatment before PA program amendments were imputed based on American Association for the Study of Liver Diseases-Infectious Diseases Society of America guidelines for treatment length during the preamendment period (12 or 24 weeks) and length of time between index date and last fill date plus days supply.
If administrative pharmacy claims had a combined days supply showing ≤ 12 weeks of therapy, then 12 weeks was assigned as the length. If the combined days supply was > 12 weeks, then 24 weeks was assigned. In this time period, genotype, fibrosis score, and treatment experience were not gathered on the PA form, precluding any clinical assessment.
To determine adherence for each member, the medication possession ratio (MPR) was calculated as the combined days supply from DAA therapy divided by the assigned treatment length in days. In general, MPR is expressed as a ratio of the days of the medication supply to the number of days in an assigned time interval, yielding a percentage value. 23 Any MPR > 1 was truncated to a value of 1. An investigator manually determined combined days supply for a single course by summing all days supply for claims from and including the index date to probable end of therapy. Any new DAA claim with a date of service conceivably after the full length of therapy (i.e., gap from last claim > 120 days) was considered a restart. Restarts were only included in descriptive analysis.
Comorbidities were defined according to ICD-9-CM and ICD-10-CM codes. They included the conditions in the Deyo-Charlson Comorbidity Index and the following conditions for liver sequelae: 13,24 alcoholic cirrhosis, chronic hepatitis, nonalcoholic cirrhosis, biliary cirrhosis, hepatocellular carcinoma, esophageal varices, hepatic encephalopathy, portal hypertension, hepatorenal syndrome, other sequelae of chronic liver disease, gastrointestinal bleed, ascites, and liver transplant (Appendix B, available in online article). 6,25,26 As clinical cirrhosis measures were not available in the database, ICD codes were used to create proxy definitions. Mild or compensated cirrhosis was defined as alcoholic, nonalcoholic, or biliary cirrhosis without advanced liver sequelae. Decompensated cirrhosis was defined as cirrhosis with advanced liver sequelae.
Descriptive statistics analyzed member, disease, and economic characteristics. Patients who experienced treatment success versus treatment failure and those who reported an SVR12 result versus those who did not were compared using Wilcoxon-Mann-Whitney tests for continuous variables, Pearson chi-square tests for categorical variables, and Fisher's exact test for categorical variables with small cell counts. Differences in means between patients' paired (and therefore dependent) pre-index and post-index utilization and costs were compared using Wilcoxon signed rank sum tests. Logistic regression with a Huber-White robust standard error estimation was used to model odds ratios (ORs) and 95% confidence intervals (CIs) for SVR12 failure between groups while adjusting for important covariates. An additional logistic regression was conducted to model estimates for SVR12 nonreport. Loglinked gamma regression with Huber-White robust standard errors was used to compare groups on the ratios of total postinitiation expenditures. 27 Regression diagnostics (e.g., deviance, residuals) were conducted to identify influence and Sustained Virologic Response and Costs Associated with Direct-Acting Antivirals for Chronic Hepatitis C Infection in Oklahoma Medicaid to assess model fit. Sensitivity analyses were conducted using alternate definitions of cirrhosis (e.g., dichotomized yes/no cirrhosis variable; decompensated cirrhosis defined as any presence of advanced liver disease with or without cirrhosis diagnoses) and adherence (MPR thresholds ≥ 0.80, ≥ 0.95, and 1.00) in order to assess the robustness and determine the sources of uncertainty in the models. 28 Table 1. Also presented within Table 1, characteristics of the 538 members without an SVR12 reported indicated that nonreporters had fewer mild liver sequelae diagnoses (43.1% vs. 52.7%, P < 0.010), fewer moderate to severe liver sequelae (24.5% vs. 31.5%, P < 0.050), and less cirrhosis (34.0% vs. 44.3%, P < 0.010), mainly driven by decompensated cirrhosis (18.0% vs. 27.2%, P < 0.010). Additionally, nonreporters had higher proportions of early treatment discontinuation (18.8% vs. 6.0%, P < 0.001).
Changes in utilization and costs for the 368 members from pre-to post-index are presented in Table 2. No differences were seen with inpatient hospitalizations or ED visits. Physician office visits increased for members with SVR12 failures (10.9 ± 5.6 visits PM pre-index to 13.1 ± 6.6 visits PM post-index). As expected, due to increased medication use, overall expenditures increased ($19,872 ± $28,442 PM preindex to $140,283 ± $52,779 PM post-index, P < 0.001). HCV treatment cost increased from $1,156 ± $8,134 PM before DAA use to $122,283 ± $46,819 PM following use (P < 0.001     Table 3 presents the results of the multivariable analyses of SVR12 failure that controlled for demographics, clinical attributes, comorbidities, HCV regimens, adherence, and health resource utilization. The presence of cirrhosis was associated with markedly higher odds of treatment failure, independent of other factors: The adjusted odds of treatment failure for members with decompensated cirrhosis was 6.69 times higher (95% CI = 2.62-17.10) than for those with no decompensated cirrhosis. The adjusted odds of treatment failure for those with compensated cirrhosis was 3.52 times higher (95% CI = 1.11-11.14) than for those with no cirrhosis. Independent of the presence of cirrhosis, the adjusted odds of treatment failure was 3.34 times higher (95% CI = 1.31-8.55) among men than among women. No statistically significant differences were seen among treatment regimens when compared with sofosbuvir. Analysis for the odds of SVR12 nonreport is also presented in Table 3. Controlling for previously mentioned variables, those with decompensated cirrhosis had -52% lower odds (95% CI = 0.32-0.72) of SVR12 nonreport compared with those without decompensated cirrhosis, and those with compensated cirrhosis had -30% lower odds (95% CI = 0.39-0.89) of SVR12 nonreport compared with those without compensated cirrhosis.
Sensitivity analyses that varied the definitions of cirrhosis and MPR thresholds showed no differences in results.

■■ Discussion
This investigation of Medicaid beneficiaries with HCV assessed DAA treatment success, SVR12 reporting, and direct medical cost within a real-world setting. Although a majority of individuals who received DAAs did not have a reported SVR (538/906, 59.4%), a high percentage of those with reported viral loads achieved treatment success (337/368, 91.6%). Independent of other factors, a 6.69-times higher odds of treatment failure was associated with decompensated cirrhosis and a 3.52-times higher odds was observed for compensated types (P < 0.050). No difference was observed between either DAA treatments or medication adherence < 95%, and males were associated with a higher odds of treatment failure (P < 0.050). Adjusted costs were lower among ledipasvir/sofosbuvir regimens, and both treatment failure and decompensated cirrhosis cases incurred higher adjusted costs (P < 0.050).  The relatively large percentage of members (59.4%) without any reported SVR12 is an important observation of this study. These individuals may have been lost to follow-up, discontinued treatment early, failed to achieve SVR12, or involved a combination of the above. Therefore, it may be anticipated that the current study is likely to have underestimated the SVR12 failure rate. Analysis of those without SVR12 reports showed that those with advanced liver sequelae (i.e., cirrhosis) were being retained in the HCV PA program, while those with milder or no liver sequelae were dropping out more often.
Additionally, even though the adherence rate was high among those with report ed SVR12s (0.98 ± 0.10), the MPR among those without reported SVR12 was significantly lower at 0.91 ± 0.21 (P < 0.001). Although not systematically measured, anecdotal observations within the HCV PA program appeared to attribute nonreported SVR12s due to loss to patient follow-up at the clinic setting (e.g., Medicaid members did not return to their physicians for final viral load testing). As such, findings from the current study suggest that future research should elicit reasons from patients for this loss to follow-up or early discontinuation, and that interventions may be especially important to retain members with less severe liver sequelae. An interdisciplinary approach may be suggested for all patients to improve adherence, to minimize barriers for successful treatment, and to track final viral loads to assess clinical outcomes. 29 The relatively high treatment success found in the current work (i.e., 91.6% achieving SVR12) parallels the overall percentages of 86%-100% reported in the ION, COSMOS, OPTIMIST, and TURQUOISE clinical trials. [30][31][32][33][34] Several other studies have examined SVR in real-world settings, with overall SVRs ranging from 82.4% to 92.8%. 35 37 Backus et al. (2016) also examined DAA use (i.e., ledipasvir/sofosbuvir and ombitasvir/paritaprevir/ritonavir + dasabuvir) among 21,142 genotype 1 VA patients and reported SVR12 ranges of 87.8%-91.7% depending on treatment regimen. 38 These authors also noted that decompensated liver disease was a negative predictor of SVR12 based on multivariable analyses. Reddy et al. (2017) examined DAA treatments in 220 patients with advanced liver disease (Model for End-Stage Liver Disease [MELD] score of ≥ 10) from medical centers in the United States, Germany, Israel, and Canada, reporting SVR12 from 35% to 84% depending on genotype, MELD score, and treatment regimen. 40 While Reddy et al. reported lower SVR12s in cirrhotic cases than in the current study, these authors reported findings based on sofosbuvir or simeprevir therapies only. 38 Shin et al. (2017) reported SVR12s of 75.0%-92.2% among 343 patients with genotype 1 based in a Seattle, Washington, medical center, also noting that SVR12s among patients with cirrhosis were lower, and that female sex was a positive predictor of SVR12. 33 Dalgard et al. (2017) found an overall SVR12 of 91% among patients in Scandinavia and reported that cirrhosis and male sex were negative predictors of treatment success. 39 The present study reinforces these findings, while also highlighting HCV treatment among Medicaid members, a notably vulnerable population.

■■ Conclusions
The present study provides real-world evidence of the clinical and economic impact of the second-generation DAAs in a state Medicaid program. High loss to follow-up was observed. Among those not lost to follow-up, high SVR12 achievement rates were seen even among those with severe liver disease, although cirrhotic disease was associated with large and significant odds of SVR12 treatment failure and increased costs. Continued work should be directed at addressing loss to follow-up and minimizing those HCV cases that manifest with advanced sequelae, potentially including a removal of barriers that impede either continued follow-up or treatment completion among already vulnerable populations.
While the relationship between failure to achieve SVR12 and cirrhosis has been long observed, limited data exist concerning decompensated cirrhosis and DAA treatmentrelated outcomes. [41][42][43] Historically, HCV patients with cirrhosis incurred lower SVR12 rates, higher risks of serious adverse events, and greater risks of sepsis, liver failure, and death. 42 In an analysis of trial data for Child Class C cirrhosis patients treated with DAAs, Guarino et al. (2017) reported that SVR12s were 50.0%-96.0%, with an average SVR12 across 10 studies of 74.9% (95% CI = 65.6%-82.4%). 43 In the current study, while SVR12 among cirrhotic patients was relatively high at 85.3%, a need to ascertain the severity of liver disease appears warranted, as more assertive overall clinical management may be required in these cases. From a broader policy or public health stance, expansion of access to treatment should be considered for HCV patients without advanced liver sequelae, in order to prevent progression and possible deleterious effects on SVR12 achievement.
A significantly higher odds of treatment failure was observed among males (OR = 3.34), independent of other factors in the current work (P < 0.050). Similar findings have been observed in other studies, including Dalgard et al. among Scandinavian countries for genotype 3, Shin et al. in genotype 1 patients based in Seattle, and the ALLY-3 study conducted by Nelson et al. (2015). 35,39,44 Limited work has focused on the possible factors surrounding this association, although biological differences have been implicated as playing a role (i.e., a higher rate of spontaneous clearance of HCV may occur in women) as have behavioral factors (e.g., men are more likely to smoke and drink alcohol more frequently, which may negatively affect the course of disease). [43][44][45][46] Limitations Despite findings of this investigation of DAAs using both administrative claims and clinical data among members of Medicaid with HCV, certain limitations should be considered. Administrative claims data do not include information on MELD scores or Child Turcotte Pugh classifications. Proxy definitions of cirrhosis were determined through ICD-9-CM and ICD-10-CM codes, which may have underestimated the true prevalence of severe liver disease. No data on indirect costs or costs borne by the member were available or estimated; expenditures in claims data reflect amounts reimbursed by Medicaid. Treatment-related rebates were also not included, and final costs to the payer may be substantially lower than those reported. While researcher imputation of treatment length for pre-PA members may have overstated MPRs, only Sustained Virologic Response and Costs Associated with Direct-Acting Antivirals for Chronic Hepatitis C Infection in Oklahoma Medicaid