Description of the Economic Model

The 2017 PBAC recommendation to list glecaprevir/pibrentasvir on the PBS was not achieved on cost-effectiveness grounds versus ‘no treatment’. The PBAC recommended glecaprevir/pibrentasvir under the condition that the cost per course be identical (cost-minimized) to sofosbuvir/velpatasvir [14]. The model structure used to retrospectively assess the cost-effectiveness of DAAs in 2016 and to determine whether DAAs remain cost-effective in 2023, is similar to the model presented as part of the reimbursement submission to achieve a PBAC recommendation for Viekira PAK® in 2015 [15]. However, where appropriate, amendments have been made to the Viekira PAK® model to better reflect newly available evidence (e.g., the model now incorporates transition from cirrhosis to liver-related death, and retreatment).

Only direct medical care costs incurred using an Australian government perspective were included. In accordance with Australian PBAC guidelines, 5% discount rates and half-cycle corrections were applied. The model was developed in Microsoft Excel (Microsoft Corporation, Redmond, WA, USA).

A ten-state Markov model captures disease progression and death over a 30-year time horizon (Fig. 1). In the first year (in which treatment occurs) a four-weekly cycle length is used to capture the different treatment durations for glecaprevir/pibrentasvir. From the second year, annual Markov cycles are used.

Fig. 1

Schematic of Markov model. CHC chronic hepatitis C, SVR sustained virological response. Background (all-cause) mortality not shown in figure; however, background mortality is applied to all health states in the model, including health states relating to retreatment

Consistent with PBS reimbursement of antivirals in Australia, chronic HCV patients with either mild fibrosis, moderate fibrosis, or compensated cirrhosis are eligible for DAA treatment. Baseline patient demographics and disease characteristics are reported in Table 2. Initially, for this study, we assumed a dramatic change over time of the fibrosis distribution of HCV-infected individuals in Australia. However, data presented in Table 2 suggest that patients living with HCV in 2020 (latest available data) have a similar fibrosis distribution to those living with HCV prior to the introduction of DAAs. At the time of initial DAA reimbursement consideration (2015–2016), there were limited data on the fibrosis distribution of individuals who might be likely to seek treatment. Therefore, cost-effectiveness models were informed by clinical trial participants or published Australian prevalence estimates. The fibrosis distribution of treated patients in the last 2–3 years is not yet known. Therefore, for consistency, the model uses the fibrosis distribution for those living with HCV. A sensitivity analysis has been conducted in which the fibrosis distribution of treated patients in 2020 (latest available data) has been used. A further sensitivity analysis assumes a substantially less severe patient population (in which all patients have mild/moderate fibrosis at the time of treatment initiation) to reflect the potentially lesser severity of treated patients in future years.

Changes to the Model Regarding Reinfection and Retreatment

In the earlier PBAC assessments, reinfection following SVR12 could occur (based on an annual probability of 0.9% taken from a meta-analysis of three trials [16]). However, in these earlier models, reinfected patients returned to the fibrosis health state that they occupied prior to achieving their SVR12 and were not subsequently retreated. Similarly, first-line patients who do not achieve SVR12 (first-line failures) were not retreated in the 2015/2016 analyses. Given second-line treatment is permitted on the PBS, the 2023 analysis incorporates additional Markov tunnel states to allow reinfected individuals to be retreated infinitely over the 30-year time horizon. As such, the model allows for multiple reinfections (and subsequent treatment each time). In the model, first-line virologic failures (2.6%; equivalent to 100% minus the SVR12 rate of 97.4% described below) only receive one subsequent course of treatment (unless they achieve SVR12 during their second treatment course, and subsequently become reinfected, and then they are permitted treatment for this second infection). While the first-line failure rate is higher than the annual reinfection rate (2.60 vs. 2.23%), first-line failures only enter the second-line treatment model once at the time at which they do not achieve SVR12, whereas the annual reinfection rate of 2.23% is applied over the entirety of the modeled time horizon to those who achieve SVR12. Thus, over the entire modeled time horizon, more reinfected individuals are retreated than first-line failures.

As reported previously, 7% of the patients who have received PBS-subsidized DAA therapies have received more than one line of therapy [9]. Thus, the current annual rate of re-infection is likely to be higher than that originally modeled (0.9%). The 2023 analysis uses an annual probability of reinfection in those who previously achieved SVR12 of 2.23%, taken from high-risk individuals in Simmons 2016 [17]. The reinfection rate reported by Simmons 2016 was comparable to that reported by the Australian Kirby Institute (2.5%) [18].

Not all first-line failures and reinfected individuals actively seek additional lines of treatment. An Australian observational study of 10,843 individuals reported that 52% of first-line virologic failures were retreated [19]. Thus, in the 2023 analysis in which retreatment is permitted, 52% of first-line failures and reinfected individuals are retreated. This percentage is assessed in sensitivity analyses. Patients in the first-line setting are treated with either glecaprevir/pibrentasvir or sofosbuvir/velpatasvir, and PBS script data were used to determine the proportional use of these two brands. Given that publicly available PBS script data do not distinguish between first- and subsequent-line use, the model assumes that reinfected individuals (i.e., patients who achieved an SVR12 but subsequently acquired a new infection) are retreated with their first-line regimen (i.e., weighted average of PBS script data for glecaprevir/pibrentasvir and sofosbuvir/velpatasvir). However, consistent with Australian clinical practice, first-line failures (i.e., patients that did not achieve an SVR12) will switch to a different DAA in the second-line treatment setting (sofosbuvir/velpatasvir/voxilaprevir as this is the only regimen recommended for retreatment in first-line treatment failures [20]).

Efficacy of DAAs in Chronic HCV Infection (Using Glecaprevir/Pibrentasvir as an Example)

Given that PBS reimbursement is restricted to chronic HCV (rather than acute HCV), the model does not incorporate spontaneous remission, as this usually occurs within 6 months of diagnosis [21]. Treated individuals either achieve a SVR12 or remain infected and potentially develop progressive liver disease. Efficacy and safety data in the model are taken from the glecaprevir/pibrentasvir trials, and are considered broadly representative of the efficacy of all DAAs used in the first-line treatment setting. A pooling of seven glecaprevir/pibrentasvir trials (covering 1262 treatment-naive and treatment-experienced individuals from all six genotypes) results in an SVR12 rate of 97.4% [22,23,24]. With regard to reinfected individuals, as each re-infection behaves as a new infection [25], it is assumed that SVR12 rates in these patients are also comparable to the pooled estimate of 97.4%. With regard to first-line DAA failures, it is assumed that patients will switch to a different DAA in the second-line treatment setting. Evidence suggests that SVR12 rates in first-line failures (when second-line treatment occurs with a different DAA) remain high [26]. This is confirmed by several of the glecaprevir/pibrentasvir trials, which permitted enrolment of DAA-experienced individuals [22] and yielded SVR12 rates > 90%. Therefore, the model assumes that the SVR12 rate in first-line DAA failures is also comparable to the first-line setting. Thus, in the 2023 analysis, which incorporates the costs and outcomes associated with retreatment, an SVR12 rate of 97.4% has been assigned to any retreated individual (irrespective of whether their second-line treatment is due to first-line failure or reinfection). This assumption is tested in sensitivity analyses.

Markov Transition Probabilities

Markov transition probabilities (Table 3) were primarily obtained from Australian-specific sources, and where possible, meta-analyses were preferred over results from single studies. Rates of background, all-cause mortality are sourced from Australian life tables [27] and have been increased by 40% to account for the potentially higher risk profile of this patient group [28, 29].

Table 3 Annual transition probabilitiesCosts and Utilities

In Australia, published prices for pharmaceutical therapies are presented in the Schedule of Pharmaceutical Benefits [30]. However, many pharmaceuticals have confidential rebates implying that the effective price (paid by the Australian government) is below the published price. Given that effective prices are confidential, this analysis uses the published General Schedule prices (DPMQ) of glecaprevir/pibrentasvir and sofosbuvir/velpatasvir to inform the cost per course of first-line treatment. It should be noted that while the published per-pack prices for glecaprevir/pibrentasvir and sofosbuvir/velpatasvir differ on the PBS schedule due to differences in treatment duration and the public prices agreed between Commonwealth and the sponsor, these DAA therapies are cost-minimized to each other, and therefore, the confidential, post-rebate, effective cost per course are identical across both brands [14]. It should further be noted that use of the published prices overestimates the incremental cost-effectiveness ratio (ICER) as the true cost per course paid by the Australian government is based on the lower post-rebate, effective prices.

The cost per course in the first-line setting is based on a weighted average of the (published) cost per courses for glecaprevir/pibrentasvir and sofosbuvir/velpatasvir. PBS script data were used to determine the proportional use of glecaprevir/pibrentasvir versus sofosbuvir/velpatasvir. Based on the Australian prescribing rules and the rates of discontinuation observed in the clinical trials, the average duration of first-line therapy is estimated to be 7.95 weeks for glecaprevir/pibrentasvir (where the approved duration of therapy is 8 weeks [31]) and 11.98 for sofosbuvir/velpatasvir (where the approved duration of therapy is 12 weeks [32]). After the incorporation of trial-based discontinuation rates and weighting between the two brands, the (published) cost per course of first-line treatment in DAA-naïve individuals, as of July 2023, is calculated to be $35,897. As reinfections in patients who achieved an SVR12 behave as a new infection [25], it is assumed that the treatment duration for reinfected individuals is comparable to the DAA-naïve setting (weighted average of 7.95 weeks of glecaprevir/pibrentasvir and 11.98 weeks of sofosbuvir/velpatasvir).

While first-line DAA failures (those that did not achieve an SVR12 during first-line treatment) would switch to a different agent in the second-line setting (namely sofosbuvir/velpatasvir/voxilaprevir), DAA therapies are all cost-minimized to each other, and therefore, the confidential, effective cost per course would be identical across brands. However, this analysis uses published prices, and therefore, the cost per course varies between the first-line and second-line settings. For this analysis, the cost of second-line treatment in first-line failures is based on 12 weeks of sofosbuvir/velpatasvir/voxilaprevir. The cost of second-line treatment in first-line DAA failures is based on sofosbuvir/velpatasvir/voxilaprevir at $36,111 per 12-week course.

The model incorporated the cost of treatment-induced adverse events (headache, rash, and nausea). The incidences of these adverse events were taken from a pooled analysis of the glecaprevir/pibrentasvir trials [22].

Resource use for the Markov health states was collected from medical records for a stratified random sample of 276 GT1 patients first attending two Australian liver clinics between January 2011 and December 2013, and supplemented by 112 GT1 patients attending one liver transplant clinic in the same period; patients were followed to June 30, 2014 [4]. The sample included in the medical resource utilization (MRU) audit was restricted to patients attending treatment centers. Given that the majority of patients who achieve SVR12 may no longer require management at a treatment center (and can be managed by their GP), it was not possible to use the audit to inform the long-term MRU used by patients who achieve SVR12. Expert opinion has been used to inform the medical resource items used by patients who have achieved SVR12. Patients without cirrhosis (i.e., patients with F0–F3 prior to achieving their SVR12) who achieved SVR12 following treatment are assumed to visit their GP every 2 years. As per current Australian guidelines in HCV management [20], patients with cirrhosis who achieved SVR12 following treatment are assumed to visit their specialist every year, and undergo an ultrasound to detect HCC. Furthermore, the 5.3% of patients who have returned to injecting drug use [33] are assumed to undergo a confirmatory HCV RNA qualitative assay every year. Many of the published models assumed that SVR12 is equivalent to a cure (and thus SVR12 patients incur no further costs). However, this model conservatively assumes that SVR12 patients incur the costs shown in Table 4 every year of the 30-year model.

Table 4 Utility values and costs for each Markov state

Health state utility values were taken from a time-trade-off (TTO) utility study conducted on 100 members of the Australian general public (Table 4) [34]. The PBAC has previously accepted a utility gain of 0.041 for patients achieving SVR12 [35]. To adjust for treatment-induced adverse events, an annualized disutility of 0.0037 was applied while on DAA treatment (average of glecaprevir/pibrentasvir trials [36]).

Ethical Approval

Given the economic analysis was based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors, ethics approval was not required.