Reply: Cost-effectiveness of combination daclatasvir-sofosbuvir for treatment of genotype 3 chronic hepatitis C infection in the United States

Reply to: Patel J, Zhao X, Shah D, Alhussain K, Kamal K. Cost-effectiveness of combination daclatasvir-sofosbuvir for treatment of genotype 3 chronic hepatitis C infection in the United States. https://doi.org/10.1080/13696998.2017.1360310

Original Article: Saint-Laurent Thibault C, Moorjaney D, Ganz ML, et al. Cost-effectiveness of combination daclatasvir-sofosbuvir for treatment of genotype 3 chronic hepatitis C infection in the United States. J Med Econ 2017;20:692-702

We thank the authors of the letter for their careful review of our article¹ and the thoughtful comments. We also thank the Editor of JME for the opportunity to respond to those comments.

Baseline characteristics and initial distribution of fibrosis and cirrhosis stages

We acknowledge that older data were used in our model, which can be explained by the fact that we finalized the model inputs before the 2014 CDC statistics² were published. The updated age and sex distribution data are not materially different from the values we used in our analyses and, hence, the results based on the updated population characteristics are similar (i.e. daclatasvir-sofosbuvir dominates sofosbuvir-ribavirin). The baseline distribution of fibrosis and cirrhosis stages that we used, and that are presented in Table 1 of our paper, came from the McGarry et al.³ article (for all ages). We, however, rounded those values to whole numbers before using them in our model. We acknowledge the confusion caused by the inappropriate accuracy implied by the “.00” digits reported in Table 1 of our article.

Choice of therapies evaluated in our model

We acknowledge that current guidelines recommend the sofosbuvir-velpatasvir combination or the daclatasvir-sofosbuvir combination plus ribavirin. However, at the time we conducted our analyses, the sofosbuvir-ribavirin combination for 24 weeks was the recommended regimen for patients with HCV genotype 3 infection according to 2014 American Association for the Study of Liver Disease (AASLD) guidelines⁴. We recognize this limitation, which we pointed out in the Discussion section of our article. Furthermore, the results from the ASTRAL-3 phase III trial⁵ evaluating the sofosbuvir-velpatasvir combination in this population of patients were not yet available. For these reasons, we compared daclatasvir-sofosbuvir with sofosbuvir-ribavirin in our study.

Sustained virological response and post-treatment relapse assumptions

The structure of our economic model, which does not include the concept of post-treatment relapse, is consistent with the opinions we have received from hepatologists in the United States (US) that patients with HCV who are successfully treated (i.e. achieved SVR) are not at risk of experiencing disease progression. In a literature review, it was reported that the incidence of late relapse, defined as reappearance of serum HCV RNA, is extremely low (with the majority of studies showing incidence between 0%–1%) and, therefore, the evidence of SVR being a durable and clinically meaningful end point of successful antiviral therapy⁶ is supported. In addition, a large body of research exists that has relied on the general structure that we used, the “MOdeling the NAtural histoRy and Cost effectiveness of Hepatitis C” (MONARCH) model^7–15, without, to our knowledge, criticism about it not accounting for post-treatment relapse. The lack of post-treatment relapse is reflected in our model by the absence of a pathway from the post-SVR state to DC and HCC (i.e. the corresponding transition probabilities were assumed to be zero). We, however, indirectly assessed the implications of relapses by varying the probabilities of SVR in additional sensitivity analyses. Our conclusions are similar to those in our published article if we lower the SVR probabilities of both arms by 10%. However, the ICER is ∼ $234,000 per QALY if we only lower the SVR probability for the daclatasvir-sofosbuvir arm by 10%, and daclatasvir-sofosbuvir is dominant over sofosbuvir-ribavirin if we only lower the SVR probability for the sofosbuvir-ribavirin arm by 10%.

Impact of non-structural protein 5A resistance-associated variants

Our model does not account for non-structural protein 5A (NS5A) resistance-associated variants (RAV) of HCV. It is possible, had the model included parameters for NS5A RAV, that we could have accounted for its impact on treatment failure, but not on relapses as previously explained. Any possible impact of NS5A RAV on treatment failure could be assessed by varying the SVR and/or treatment discontinuation probabilities, which we did in the published deterministic and probabilistic sensitivity analyses. At the time we conducted our analyses, AASLD⁴ was not recommending RAV testing for genotype 3 HCV patients. It is worth mentioning that analyses of real-world data from the GenBank HCV genome database - comprised of the DNA DataBank of Japan (DDBJ), the European Nucleotide Archive (ENA), and the GenBank at NCBI - found that no prevalent cases resistant to daclatasvir-sofosbuvir among genotype 3 patients have been reported (Figure 5b in the Chen et al.¹⁶ article). Based on this fact, we can conclude that NS5A RAV are not major concerns for performing economic evaluations of daclatasvir-sofosbuvir.

Sources of transition probabilities for end stage liver disease and complications

The Rein et al.¹⁷ article was published after we finalized our analyses that were based on a published Markov model that was adapted to reflect US demographic characteristics, treatment patterns, costs (drug acquisition, monitoring, disease, and adverse event management), and mortality risks. The original model was designed to simulate the natural history of hepatitis C of all genotypes (GT 1, 2, 3, or 4) and its complications. In that context, we relied on the Martin et al.¹⁸ article as being the best and most up-to date source of transition probabilities for end-stage liver disease and complications. Our clinical advisers in the US confirmed that those transition probabilities are still applicable to the US settings for GT 3 HCV patients. We did, however, compare the transition probabilities for end-stage liver disease and complications in our study to those used by Rein et al.¹⁷. Many of the probabilities are very similar except for the transitions from CC to HCC (lower in our study) and from DC to HCC (higher in our study) as well as for the mortality risk due to liver transplant (higher in our study). We have applied the same transitions from Rein et al.¹⁷ to our model, and the conclusions were not substantially different: treatment with daclatasvir-sofosbuvir resulted in lower costs ($161,725 vs $190,460) and more QALYs gained (10.53 vs 10.43) over 20 years than sofosbuvir-ribavirin in the overall GT 3 HCV patient population in the US.

Treatment-related monitoring and post-treatment monitoring costs

We derived treatment-related and post-treatment monitoring costs from expected healthcare resource use patterns based on the 2014 recommendations for testing, managing, and treating hepatitis C published by the AASLD, which was cited in our article as Reference 33. Total costs were derived by applying unit costs to the individual healthcare services. The unit costs, which were obtained from the Centers for Medicare and Medicaid Services, were cited in Table 1 of our published article as References 47–49.

Treatment adherence

Our model does not explicitly account for medication adherence, but it does indirectly account for the effects of less than perfect adherence observed in the relevant clinical trials through the efficacy inputs (i.e. the SVR probabilities) and by accounting for the possibility of treatment discontinuation. Updating the model to include parameters for adherence would not likely substantially change the effectiveness results, as adherence would be treated as its own parameter instead of part of the SVR probabilities, but would lower the total costs via less medication-related expenditures. Finally, it is worth noting that the Nelson et al.¹⁹ article does not report mean adherence observed in the ALLY-3 trial, separately by treatment arm, and that overall adherence pooled for both arms was quite high (≥90%), except for one patient due to pregnancy.

Transparency

Declaration of funding

There is no funding to declare for this letter.

Declaration of financial/other interests

CST and MLG are employees of Evidera, Inc, an independent research organization that received consulting fees from Bristol-Myers Squibb for the development of the study. BS, SH, and YY are employees of Bristol-Myers Squibb, the manufacturer of daclatasvir. At the time of the research project, DM and BG were employees of Evidera and Bristol-Myers Squibb, respectively, but are currently employed elsewhere.

Acknowledgments

None reported.