Treating Hepatitis C in an Era of Interferon-free Therapies

109 31
Treating Hepatitis C in an Era of Interferon-free Therapies


General Model Overview

To forecast population-level outcomes according to alternative treatment strategies, we created a dynamic Markov model. The model has a compartmental structure, with each compartment representing a pertinent health state (as per the known natural history of HCV infection and disease progression; see online supplementary eFigure 1 Scotland was chosen as the setting for this model given her rich data sources; in particular, Scotland has a national HCV diagnosis database, electronically linked to a national HCV treatment database, and conducts regular epidemiological surveys of her PWID population (to note, throughout this paper, the 'PWID' acronym refers specifically to on-going injecting drug use). We drew upon these data sources in order to estimate the number of Scottish persons residing in each model compartment in the year 2009 (see Table 1 and online supplementary appendix A Then, we used intercompartment transfer rates obtained from the scientific literature (outlined in Table 2), to project liver-related disease/deaths and incident transmission over a 20-year timeframe (i.e. 2010–2030), in the context of IFN-free/sparing treatment regimens, available from 2015 onwards.

Detailed Model Overview

Incident Infection. The model incorporates an inflow of new injecting initiates each year (entry via PWID states; see online supplementary eFigure 1 The number of persons entering PWID states equals the number exiting (where exit occurs through death or through ceasing injecting); hence, an implicit assumption underpinning this model is a stable PWID population size (i.e. ~15 300 persons), in Scotland, over the timeframe of this model. The hallmark of PWID states is that uninfected subjects risk acquiring chronic infection with genotypes 1–3 HCV (note, genotypes 4–6 account for <1% of infections in Scotland, and thus, for simplification, were omitted from this model). We assumed that PWID mix randomly, regardless of their HCV infection status, and so we model incident chronic infection using a mass-action transmission function. We calibrated this transmission function to an estimate of 600 incident chronic infections (95% CI 400 to 800)—equating to an incidence of 10 (7–14) per 100 person years—occurring in the year 2009 among all PWID in Scotland; see appendix B for further details. The average injecting 'career' is 9.1 years (i.e. an 11% per annum chance of cessation). Upon ceasing injecting drug use, model individuals transfer permanently to former/never PWID states where incident HCV transmission no longer occurs (see online supplementary eFigure 1

Liver Disease Progression

Model individuals differ according to the extent of fibrosis incurred to their liver (as per the Ishak score: Ishak 0–1=mild fibrosis; Ishak 2–5=moderate; Ishak 6=compensated cirrhosis). Under chronic infection, fibrosis advances from a mild to moderate severity at a rate of 1.8% per annum; then from moderate severity to compensated cirrhosis at a rate of 2.7% per annum (see online supplementary appendix C.1 for derivation). On average, these progression rates equate to 7% of persons developing cirrhosis within 20 years of initial infection (consistent with community-level observational data; as is appropriate in a population-level model). Subjects with compensated cirrhosis are at risk of developing decompensated cirrhosis (defined as ascites, bleeding varices, jaundice or encephalopathy) and liver cancer. These two severe liver morbidity (SLM) states carry a marked risk of a liver-related death (43% per annum with liver cancer; 13% per annum with decompensated cirrhosis). Although this bleak prognosis can be improved through liver transplantation, suitable donors are scarce (hence, our model incorporates only a 2% chance, per annum, of transplantation). Of those that do receive a transplant, 85.6% survive the first year; thereafter, the risk of a liver death falls to 4.4%. Patients with chronic infection and compensated liver disease are eligible for antiviral treatment. Those attaining the optimal treatment outcome (a sustained viral response (SVR), defined as testing undetectable for viral RNA at least 6 months after terminating treatment) move into 'Treatment-induced viral clearance' states, and exhibit an improved prognosis vis-à-vis risk of subsequent liver disease progression (see online supplementary eFigure 1 In post-SVR Ishak-6 states, decompensated cirrhosis and liver cancer occur at a diminished rate (the per annum risk of decompensated cirrhosis falls from 6.5% to 0.8%; the risk of liver cancer falls from 1.4% to 0.4%). Persons in post-SVR Ishak 0–5 states can still advance through to cirrhosis (Ishak 6), but at a far reduced rate relative to their chronic counterparts (see online supplementary appendix C.2 Of note, we assume SVR does not benefit PWID apropos risk of reinfection and the subsequent chance of spontaneous clearance. Finally, at every stage of the model, death from non-liver-related causes occurs at a rate of 1.8% and 1.4% per annum, in PWID and former/never PWID states, respectively (see online supplementary appendix D regarding estimation of the latter mortality rate).

Modelled Treatment Strategies

On average, 1000 patients have commenced a course of antiviral therapy each year in Scotland, between 2010 and 2013. Of these treatment initiates, an estimated 12% were PWID (10% with mild fibrosis and 2% with moderate-advanced fibrosis); 49% were former/never PIWD with mild fibrosis and 38% were former/never PWID with moderate or advanced fibrosis (see online supplementary appendix E We assume treatment uptake and patient composition remain unchanged in 2014. From 2015 onwards, we model eight distinct treatment strategies. The first strategy is simply a continuation of the status quo (i.e. treating the same number of patients, with the same case mix, as per 2010–2014). The remaining seven are alternative treatment strategies, appropriate and feasible in the context of the impending IFN-free/sparing era. Each alternative strategy differs in terms of treatment uptake intensity and patient case mix; see Table 3 and online supplementary eTable 9

Modelled Treatment Efficacy

The status quo SVR rates (via current standard-of-care—pegylated- IFN, ribavirin±telaprevir/bocepreivr) are taken to continue until the end of 2014 (see online supplementary appendix F Thereafter, we anticipate the availability of IFN-free/IFN-sparing regimens. For patients infected with genotypes 1–2, we surmise that these regimens will deliver SVR rates of 95% regardless of fibrosis stage. Patients with genotype-3 infection will see marginally lower efficacy levels (90% for Ishak 0–5 and 80% for Ishak 6). Of note, in our base case, persons who fail a course of therapy in 2010–2014 are eligible for retreatment, but only with a post-2015 IFN-free/sparing regimen (and we make the simplifying assumption that the re-treatment patient has the same chance of an SVR as their treatment-naive counterpart). However, persons that fail a post-2015 course of therapy are not considered re-treatable (although we remove this latter assumption in sensitivity analysis 4).

Performance Outcomes

We considered the performance of each treatment strategy in terms of the following outcomes occurring over a short-term (defined as 2015–2020), medium-term (2015–2025) and long-term (2015–2030) time horizons.

  • Number of incident cases of SLM (i.e. decompensated cirrhosis and liver cancer) among persons with past/current chronic HCV infection.

  • Number of liver-related deaths among persons with past/current chronic HCV infection.

  • Number of incident chronic HCV infections among PWID.

Uncertainty Analysis

Probabilistic Sensitivity Analysis. We performed a probabilistic sensitivity analysis, to gauge total uncertainty attributable to the sampling error, inherent in our Table 1 parameters. This involved: (A) assigning, to each parameter an appropriate uncertainty distribution (see Table 1); (B) selecting, for each parameter, a random value from this distribution; (C) generating all specified outcomes under this unique set of parameter selections and (D)repeating this process 10 000 times. Hence, the 95% credible interval represents the range within which the central 95% of these 10 000 data points lie.

One-way Sensitivity Analyses

We performed various one-way sensitivity analyses to test our conclusions against uncertain assumptions. These are outlined in detail in online supplementary appendix J

Subscribe to our newsletter
Sign up here to get the latest news, updates and special offers delivered directly to your inbox.
You can unsubscribe at any time

Leave A Reply

Your email address will not be published.