SPRINT-2 study design

SPRINT-2 (ClinicalTrials.gov number, NCT00705432) was a Phase 3, international, randomized, double-blinded placebo-controlled study comparing the safety and efficacy of therapy with peginterferon alfa-2b and ribavirin (PegIntron and Rebetol, respectively; Merck) with two treatment regimens that added boceprevir (Victrelis, Merck) after a 4-week lead-in treatment period with peginterferon–ribavirin alone [17]. SPRINT-2 was conducted in accordance with the principles of Good Clinical Practice. The study protocol and study design were approved by each of the sites institutional review board and regulatory agencies, and each participant provided written informed consent before undergoing any study-related procedure. A list of the institutional review boards that approved the study protocol and study design is provided in Additional file 1: Table S1. The full study protocol is available at http://www.nejm.org/action/showSupplements?doi=10.1056%2FNEJMoa1010494&viewType=Popup&viewClass=Suppl. Previously untreated patients (N = 1097) ≥18 years of age with genotype 1 chronic HCV and plasma HCV-RNA level ≥10,000 IU/mL were eligible. Because of the known marked difference in SVR rates with peginterferon–ribavirin between black and non-black patients [16], self-identified blacks and non-blacks were enrolled separately into two cohorts. Exclusion criteria included liver disease of other etiology, decompensated cirrhosis, renal insufficiency, HIV or hepatitis B, pregnant/breast feeding women, or active malignancy. Liver biopsies were assigned METAVIR fibrosis and steatosis scores by a single pathologist who was unaware of treatment assignment.

Peginterferon alfa-2b was administered subcutaneously at 1.5 μg/kg once weekly. Ribavirin was administered using weight-based dosing of 600–1400 mg/day (divided daily dose). Boceprevir was administered orally at a dose of 800 mg three times daily (to be taken with food and with an interval of 7 to 9 hours between doses) in four capsules of 200 mg each. Placebo was matched to boceprevir. The study was double-blinded regarding the administration of boceprevir.

All patients received peginterferon–ribavirin during the 4-week lead-in period. Patients randomized to control received peginterferon–ribavirin treatment for 44 weeks after the lead-in period, as well as placebo three times daily beginning at week 5 (PR48). The overall SVR rate in the PR48 arm of SPRINT-2 was 38% (137/366) for both cohorts, 40% (125/311) for the non-black, and 23% (12/52) for the black cohort [17]. Patients randomized to the response-guided therapy (BOC/RGT) regimen received peginterferon–ribavirin plus boceprevir for a total of 24 weeks after the lead-in period; if HCV-RNA levels were undetectable from week 8 through week 24, treatment was considered complete, but if HCV-RNA levels were detectable at any visit from week 8 up to but not including week 24, peginterferon–ribavirin was continued, and placebo was administered at week 28 through week 48. The overall SVR rate in the BOC/RGT arm of SPRINT-2 was 63% for both cohorts (233/368), 67% (211/316) for the non-black, and 42% (22/52) for the black cohort [17]. Patients randomized to the third regimen received peginterferon–ribavirin plus oral boceprevir for 44 weeks after the lead-in period (BOC/PR48). The overall SVR rate in the BOC/PR48 arm of SPRINT-2 was 66% for both cohorts (242/366), 68% (213/311) for the non-black, and 53% (29/55) for the black cohort [17].

In each arm, patients with detectable HCV-RNA at week 24 discontinued treatment as a standard futility rule. Boceprevir was given for 24 weeks in the BOC/RGT arm and 44 weeks in the BOC/PR48 arm. All patients were followed through week 72.

Model structure

We created an Excel-based (Microsoft Corp., Redmond, Washington) Markov model to project health-related outcomes and to estimate the expected costs and quality adjusted life-years (QALYs) associated with the three treatment strategies studied in SPRINT-2. The structure of the model was based on other published health economic models of HCV disease [22–27]. The model consists of two phases: the first phase corresponds to the treatment strategies and follow-up period and the second phase corresponds to post-treatment, which includes the natural history of HCV of cured or uncured patients (Figure 1).

Patients entered the model with chronic HCV and immediately began AV drug therapy. The treatment phase of the model includes a weekly cycle length in which patients can stop therapy early for a variety of reason (i.e. discontinued due to standard futility rule, adverse events, or other non-medical reasons). Patients with undetectable HCV-RNA at the end of treatment were followed for an additional 24 weeks. Patients with undetectable HCV-RNA after 24 weeks of follow-up achieved an SVR. Relapse was defined as the occurrence of undetectable HCV-RNA at the end of treatment, but detectable HCV-RNA after the 24 week follow-up period. In the model, patients who experienced relapse returned to the chronic HCV health states. Patients who failed to pass a futility rule or who had detectable HCV-RNA at the end of treatment were considered treatment failures and also returned to the chronic HCV health states.

The second phase of the model uses cycles with a length of one year and describes patient outcome post-treatment. After the trial follow-up period (72 weeks), patients either attained an SVR or returned to the natural health states of HCV. The severity of chronic HCV infection was defined by the degree of fibrosis using the METAVIR scoring system: no fibrosis (F0), portal fibrosis without septa (F1), portal fibrosis with few septa (F2), numerous septa without fibrosis (F3), and cirrhosis (F4). Patients with mild or moderate chronic HCV at baseline, described by a METAVIR fibrosis score of F0-F3, who attained an SVR were considered permanently cured; cirrhotic patients who attained an SVR were considered partially cured. Patients who were permanently cured were considered to be neither at risk for developing further HCV-related liver complications nor for reinfection. Patients with cirrhosis at baseline continued to be at risk for developing advanced stages of liver complications associated with cirrhosis even if they achieved SVR; however, for cirrhotic patients who attained an SVR the probability of developing advanced stages of liver complications was less than that of an untreated cirrhotic patient [28].

Patients who failed to achieve an SVR returned to the natural HCV health states and were at risk for developing serious liver disease and could receive a liver transplant at the same rate as patients who did not receive treatment. The model simulates the natural history of chronic HCV and advanced liver-related diseases and its treatment, consistent with the current understanding of the biology of chronic HCV-related liver disease and its treatment (e.g., liver transplant). The progressive disease model assumed that during one cycle, a person with a given fibrosis score could progress to the next fibrosis level of severity or could remain in that current health state. In the absence of successful treatment, regression to less severe health states, or spontaneous clearance of the virus, was not permitted. Patients with compensated cirrhosis were at risk for developing the sequelae of cirrhosis - decompensated cirrhosis (DC) and hepatocellular carcinoma (HCC).

Patients who developed DC and/or HCC could receive a liver transplant. Because of differential mortality, costs and quality of life between the first year and subsequent years of these health states, they were divided into two states: DC, first year and DC, subsequent years; HCC, first year and HCC, subsequent years; and the Liver Transplant and Post-Liver Transplant states. The Liver Transplant health state lasted a total of 1 year. If a patient was alive at the end of 1 year, then the patient transitioned to the Post-Liver Transplant state. Patients who received a liver transplant were assumed to be at no risk of reactivation and progression to liver disease.

The mortality risk of the general population was applied to all states in the model. In addition, an excess mortality rate was applied to patients with DC, HCC, or who received a liver transplant.

Model inputs

Patient characteristics

As patient characteristics impact the efficacy of the treatment regimens and the annual mortality rate, the analyses were conducted on a cohort of persons with chronic HCV genotype 1 who were representative of participants in SPRINT-2. In the model, a series of 20 cohorts progressed through each treatment regimen. The cohorts represent all possible combinations of gender, race cohort, and baseline METAVIR fibrosis score (2 × 2 × 5 = 20). The average age of the overall SPRINT-2 study cohort was applied to all patients in the analysis. The reported distributions of gender, baseline fibrosis level, and race cohort from SPRINT-2 were assumed for the treatment population (17, Table 1).

Treatment characteristics

Treatment characteristics were obtained from reported data in SPRINT-2 (Table 2). We assumed that patients experienced the same treatment efficacy, discontinuation rates, treatment-related anemia, and utilized erythropoietin (EPO) to treat anemia as observed in SPRINT-2.

Clinical inputs

Clinical inputs described the rate of HCV progression, the probability of receiving a liver transplant, and all-cause and liver-related mortality rates and were used in the model to determine the amount of time patients spend in each HCV health state (Table 3).

Progression of HCV infection

The literature search resulted in a wide range of values for progression rates [52, 53]. We used the progression rates from Thein et al. [29], a recent meta-analysis of published progression rates from 111 studies of individuals with chronic HCV infection. They provided stage-specific progression rates by fibrosis level, as described using the METAVIR scoring system. The meta-analysis demonstrated that the progression rates are not linear and are generally higher in the initial stage F0 to F1 than the transitions between the stages with more fibrosis (i.e. F1 to F2). The estimates were also adjusted for biases attributable to study design and selection factors associated with study population and clinical characteristics.

The annual probability of developing advanced stages of liver complications associated with cirrhosis was derived from published studies. The baseline likelihood of developing DC from compensated cirrhosis was estimated using a weighted average of the annual incidence rates reported in five natural history studies of 1,276 cirrhotic patients [30–34]. Similarly, the annual transition rate from compensated cirrhosis to HCC was estimated using a weighted average of the annual incidence rates reported in nine natural history studies of 1,905 cirrhotic patients [30–38]. The baseline likelihood of developing HCC from DC was estimated from a study by Planas et al. [39] that followed 200 patients with DC. The estimates for the transition rates to DC and HCC in cirrhotic patients who achieved SVR were obtained from a study by Cardoso et al. [28].

Probability of receiving a liver transplant

Previously published U.S. based cost-effectiveness models estimated the probability of receiving a liver transplant from DC using Bennett et al. [54], which estimated the prevalence of DC using mortality rates from a 1987 study by Gines et al. [55] and 1994 data from United Network of Organ Sharing (UNOS) and the Division of Organ Transplantation [40]. In our model, this estimate was updated to take into account the increase in survival rates, prevalence of patients with liver decompensation, and changes in liver-transplant practice that have occurred since 1994. According to the analysis of the Scientific Registry of Liver Transplant Recipients (SRTR) data, from 1999 to 2007, the number of recipients with HCV increased to a peak of 2,481 in 2006 and remained relatively stable at around 2,400 transplants annually thereafter [41]. In addition, HCV-related DC became more common after 1995 with the prevalence in 2010 estimated as 103,117 [42]. We estimated the annual probability of receiving a liver transplant from DC by dividing the most recent data for the number of HCV-related liver-transplants with the prevalence of HCV-related DC, i.e. 2400/103117 = 2.3%. Although the number of liver transplants has increased, our estimate is lower than that assumed in previous studies (3.1%) primarily because of a substantial increase in the prevalence of DC. The annual probability of patients with HCC receiving a liver transplant was estimated as 4.0% using a study by Lang et al. [43].

Mortality

We used gender and age-specific all-cause mortality rates from the 2006 U.S. life tables to describe the risk of mortality associated with all states in the model [44]. In addition, an excess mortality rate associated with decompensation of the liver was estimated from Planas et al. [39] and probability of death from HCC was obtained from Fattovich et al. [30]. Liver-transplant related mortality was estimated from the most recent data available on liver transplants using the study by Wolfe et al. [45].

Cost inputs

The model was developed from the payer perspective. We included the cost of AV therapy and management of HCV disease in patients who did not achieve SVR. All costs were expressed in terms of 2010 US dollars. AV therapy costs include drug costs, the cost of managing treatment-related anemia, and monitoring costs for patients on therapy. AV drug costs were calculated using the weekly costs of peginterferon and the generic version of ribavirin, assuming a daily dose of 1000 mg [46]. The weekly cost of boceprevir was assumed to be $1100. The drug costs accounted for the discontinuation of treatment due to the standard futility rule, adverse events, or other non-medical reasons. As ribavirin is administered using weight-based dosing, the average weight of patients was used to determine the cost of ribavirin.

The average costs applied to each natural HCV health state were derived from published studies. The costs of treating chronic HCV and compensated cirrhosis were based on a retrospective, matched cohort claims database study [49]. The reported costs were modified by subtracting the AV therapy costs reported by patients with a METAVIR score of F0–F4 and by adjusting the inpatient hospitalization costs using the national hospital cost-to-charge ratio. McAdam-Marx et al. [49] only reported an aggregated cost for the F0–F3 health states. The disaggregated costs of ''HCV without liver disease" were estimated by taking into account the relative contribution of mild (F0, F1), moderate (F2), and severe (F3) to the total HCV-related cost reported in Davis et al. [50]. Future costs were discounted at 3% per year.

Quality of life inputs

Utility weights for each of the health states and liver disease conditions were applied to the utilities of the general population. The utility weights were estimated from a previously published study of patients with chronic HCV [51] and adjusted to the U.S. population norm by using age- and gender-specific utility weights of the U.S. general population [48]. There are limited data on the impact of treatment on health-related quality of life. Side effects associated with pegylated interferon and ribavirin are well-documented and include IFN-induced bone marrow depression, flu‒like symptoms, neuropsychiatric disorders, autoimmune syndromes, and anemia [56]. In addition to these side effects, the boceprevir-based regimens were also associated with a higher probability of anemia and dysguesia [17]. The magnitude of the decrement in quality of life associated with the side effects of boceprevir has not been empirically quantified. In our analysis, we assumed that patients who do not experience side effects experienced the same decrement in quality of life regardless of their treatment strategy (PR48 vs. BOC/RGT and BOC/PR48). Similarly, we assumed that patients with incident anemia experienced the same decrement in quality of life regardless of their treatment strategy. We used previously published estimates to quantify the impact of AV therapy on the quality of life for patients with chronic HCV without incident anemia [24]. The impact of incident anemia on the quality of life for patients who are receiving AV treatment for HCV was estimated from a study of the impact of anemia on the quality of life of patients with cancer [47]. Differences in the average quality of life amongst the regimens for patients on AV therapy are due to differences in the proportion of patients experienced anemia.

For patients who are cured by treatment, we assumed that an SVR following treatment eliminates all decrements in health-related quality of life associated with living in the chronic HCV infection state. Future QALYs were discounted at 3% per year.

Model outcomes

The model was run for each of the specified 20 patient profiles. An overall weighted average of the results was generated based on the distribution of the patient characteristics assumed for a given analysis. The model projected the lifetime incidence of serious liver complications, total costs, and QALYs associated with each treatment strategy as well as no treatment. In addition, the incremental cost-effectiveness ratios (ICERs) of boceprevir-based regimens compared with treatment with peginterferon and ribavirin alone as well as for PR48 compared with no treatment were estimated.

Sensitivity and subset analyses

Extensive sensitivity analyses on inputs were performed to estimate their impact on the total costs and QALYs of the boceprevir-based regimens and the dual therapy regimen. Specifically, we examined the effect of varying the values of inputs related to SVR rates, progression rates, cost, and quality of life weights in both one-way and multivariate sensitivity analyses. In the sensitivity analyses, the lower and upper bounds of the transition rates and QALYs were obtained from published studies; the costs were varied by increasing and decreasing the base case values by 25%; and the lower and upper bounds of the SVR rates were obtained from the bounds of their respective 95% confidence intervals on the reported values. In the multivariate sensitivity analyses, we applied the lower and upper bounds of all similar variables (i.e. the lower bounds of all progression rates). In addition, a probabilistic sensitivity analysis (PSA) was conducted to examine the impact of varying these covariates simultaneously. The baseline values, range of values examined in sensitivity analyses, distribution of parameters assumed for the PSA, and references for clinical, economic, and utility parameters are included in Table 3. The baseline values, range of values examined in sensitivity analyses, and distribution of parameters assumed for the PSA for the treatment efficacy rates are included in Table 2. The results of the PSA were based on 10,000 Monte Carlo simulation runs.

Patients in SPRINT-2 were randomized by race cohort to each of the three treatment groups and the data was analyzed separately in the primary efficacy analysis. Hence, subgroup analyses were generated for each race cohort (non-black vs. black). Because the recommended treatment regimen for boceprevir use in the U.S. is not exactly as was studied in the trials, we conducted a supplemental analysis depicting the Food and Drug Administration (FDA)-approved label [57] recommendations by modifying the treatment strategy phase in the model and reanalyzing the data from SPRINT-2. The label includes different treatment strategies for patients without cirrhosis and patients with cirrhosis. The recommended treatment strategy for patients without cirrhosis is similar to the BOC/RGT arm in SPRINT-2 and the recommended treatment strategy for patients with cirrhosis is similar to the BOC/PR48 arm in SPRINT-2. Model modifications include estimating the treatment characteristics for patients based on their baseline cirrhosis status (METAVIR score of F0–F3 vs. F4), and accounting for an additional futility rule at week 12.

No treatment compared with dual therapy

Over the lifetime of this cohort, our model predicted that treatment with PR48 will result in relative decreases in the cumulative incidence by 37% in DC, 38% in HCC, 38% in liver transplants, and 38% in liver-related deaths compared to no treatment. The total discounted lifetime costs and QALYs associated with no treatment are $37,230 and 13.67, respectively. The total discounted lifetime costs and QALYs associated with PR48 treatment are $58,761 and 14.55, respectively. The corresponding ICER comparing PR48 treatment with no treatment is $24,435/QALY.

Base case analysis

The model projected the lifetime cumulative risk of developing HCV-related liver complications associated with each of the treatment strategies studied in SPRINT-2 over time (Figure 2). Over the lifetime of this cohort, our model predicted that treatment with BOC/RGT will result in relative decreases in the cumulative incidence by 38% in DC, 39% in HCC, 38% in liver transplants, and 38% in liver-related deaths compared to treatment with PR48. This implies that treating 21 patients with BOC/RGT instead of PR48 will avoid 1 case of DC; treating 17 patients will avoid 1 case of HCC; treating 116 patients will avoid 1 liver transplant; and treating 13 patients will avoid 1 liver-related death. Similarly, our model predicted treatment with BOC/PR48 will result in relative decreases in the cumulative incidence by 42% in DC, 43% in HCC, 42% in liver transplants, and 42% in liver-related deaths compared to treatment with PR48. This implies that treating 19 patients with BOC/PR48 instead of PR48 will avoid 1 case of DC; treating 15 patients will avoid 1 case of HCC; treating 104 patients will avoid 1 liver transplant; and treating 12 patients will avoid 1 liver-related death. In addition, treatment with BOC/RGT and treatment with BOC/PR48 are associated with overall increases in life expectancy of 0.97 and 1.07 years, respectively, when compared with PR48 treatment.

Sensitivity analyses

The ICERs compared with PR48 from the one-way sensitivity analyses of chronic disease progression rates, rate of developing advanced liver disease, all health state costs, and most utility values were within $6 K/QALY and $11 K/QALY of the BOC/RGT (range: $1,747 to $42,983/QALY) and BOC/PR48 (range: $21,016 to $88,789/QALY) base-case ICERs, respectively (See Additional file 2: Table S2 online). The ICERs that fell out of these ranges were obtained when the lower bound of the quality of life of the SVR state for patients who had a baseline METAVIR score of F1 was assumed (BOC/RGT: $25,685 and BOC/PR48: $87,264) and when assumptions concerning treatment efficacies were varied. When the efficacy of PR48 was assumed to be 45.4%, the upper limit of the 95% confidence bound, and the efficacies for BOC/RGT and BOC/PR48 remained the base case values, the ICERs of BOC/RGT and BOC/PR48 increased to $29,369 and $81,237, respectively. Conversely, when the efficacies of the boceprevir-based regimens were assumed to be the upper limits of the confidence bounds, and the efficacy of PR48 was assumed to be the base case value, both BOC/RGT and BOC/PR48 became cost-saving compared to dual therapy.

Compared to treatment with PR48, the ICERs from the multivariate sensitivity analyses ranged from $2,338 to $33,511 for BOC/RGT and from $21,016 to $117,395 for BOC/PR48 (Table 4). The ICERs are most sensitive to assumptions concerning the quality of life of the HCV health states (range: $10,906–$31,124/QALY for BOC/RGT vs. PR48; range: $34,927–$108,965/QALY for BOC/PR48 vs. PR48) and least sensitive to assumptions concerning the quality of life of patients on treatment for those who receive BOC/RGT (range: $16,724–$16,819/QALY) and quality of life of the general population for patients who receive BOC/PR48 (range: $54,133–$56,228/QALY).

The results of the corresponding probabilistic sensitivity analysis are described in the cost-effectiveness acceptability curves (Figure 3). Compared to treatment with PR48, and using an incremental cost-effectiveness ratio of $50,000 per QALY as a threshold, treatment with BOC/RGT was cost-effective in 99.9% of the simulations and treatment with BOC/PR48 was cost-effective in 51.9% of the simulations. Using an incremental cost-effectiveness ratio of $100,000 per QALY as a threshold, treatment with BOC/RGT was cost-effective in 100% of the simulations and treatment with BOC/PR48 was cost-effective in 99.5% of the simulations.

Subset analyses

In the non-black cohort, treatment with both boceprevir-based regimens were projected to result in a gain of approximately 0.64 QALYs over those obtained with PR48 treatment (Figure 4, Table 5). This corresponds to ICERs of $15,067/QALY and $56,013/QALY for the BOC/RGT and BOC/PR48 treatment regimens compared to the PR48 regimen, respectively. In the black cohort, treatment with BOC/RGT was projected to result in a gain of 0.47 QALYs and treatment with BOC/PR48 was projected to result in a gain of 0.68 QALYs over those obtained with PR48 treatment. This corresponds to ICERs of $30,627/QALY and $50,423/QALY for the BOC/RGT and BOC/PR48 treatment regimens, respectively, when compared to treatment with PR48. The ICER of BOC/PR48 compared with BOC/RGT in treating black patients is $94,610/QALY.