چارچوب ساختاری و کلیدی پارامترهای مدل در تجزیه و تحلیل هزینه اثربخشی برای حال حاضر و درمان آینده از مزمن هپاتیت C

Objectives Published economic evaluations have reported available treatments for chronic hepatitis C to be cost-effective as part of the current approach to disease management, but as standards of care evolve, their approach to modeling should be reconsidered. This study aimed to review structural frameworks and key model parameters as reported in current economic evaluations for treatments for chronic hepatitis C, and model the impact of variability across parameters on results. Methods A systematic review of studies published from 2000 to 2011 was performed. Studies were retrieved from five electronic databases using relevant search strategies. Model structures, disease progression rates, utilities, and costs were extracted from included studies, and were qualitatively reviewed and incorporated into a cost-utility model. Results Thirty-four studies were appropriate for data extraction. A common pathway of six disease states was identified. In some studies the early disease stages and/or the decompensated cirrhosis state were further subdivided. Large variability in values used for disease progression rates, utilities, and costs were identified. When incorporated into a model, incremental cost-effectiveness ratios (ICERs) varied: in the least favorable scenario, peginterferon plus ribavirin was dominated by interferon plus ribavirin; and in the most favorable scenario, peginterferon plus ribavirin dominated interferon plus ribavirin ($8,544 per quality-adjusted life year [QALY]; costs are given in 2008 US dollar amounts). Using mean values the ICER was $15,198 per QALY. Conclusions Current models use a simplistic structure resulting from the lack of available data reflecting patient heterogeneity. Key model parameters are currently based on a small number of studies and the variability across these values can affect the interpretation of results.

Chronic hepatitis C is a major cause of progressive liver disease and represents a significant and increasing burden in terms of morbidity, mortality, and costs in both developed and developing countries [1] and [2]. Factors such as prevalence, cumulative time of exposure to the hepatitis C virus (HCV), and genetic and environmental aspects mean that the disease burden differs across geographic regions. For example, the average time elapsed since exposure in Japan, Italy, and Spain means that a high proportion of patients have already progressed to chronic hepatitis, cirrhosis and, occasionally, hepatocellular carcinoma (HCC); while in the US, the prevalence of the complications of disease progression associated with HCV, such as HCC, are predicted to increase over the next 20 to 30 years as the average time since exposure increases [3]. Chronic hepatitis C is unique among chronic viral infections in that it is considered to be curable, and thus effective treatment, in terms of sustained virological response (SVR), has substantial long-term benefits. In many countries, combination antiviral therapy with peginterferon alfa plus ribavirin has become the standard of care [4] and [5] and is considered cost-effective over a patient's lifetime for patients with chronic hepatitis C achieving SVRs of 40% in those with genotype 1 HCV, and 75% in those with genotype 2 and 3 [6]. A recent review of cost-effectiveness analyses reported that the majority of published incremental cost-effectiveness ratios (ICERs) fall within published acceptability thresholds [7]. The approach to disease management and treatment continues to evolve, as means of addressing the current and substantial unmet medical needs in hepatitis C are sought. Indeed, the introduction of the first direct-acting antiviral (DAA) therapy is anticipated in 2011. Also, research now indicates that certain biomarkers and differential responses to treatment at earlier periods of measurement (e.g., rapid viral response at 4 weeks) may also influence long-term treatment success across this heterogeneous patient group [8], [9], [10], [11] and [12]. As treatment benefits are maximized in those who respond, and exposure to adverse events and treatment costs are minimized in those who do not respond, so it can be expected that cost-effectiveness will be maintained or improved. To assess the true value of current and future standards of care, the relevant benefits and costs must be considered over the lifetime of the patient using appropriate modeling techniques. To ensure that models are fit-for-purpose and that their interpretation by payers and society is accurate and appropriate, it is essential that the design, methodological assumptions, and data input parameters are relevant to the research question posed. To date, no one has reviewed the current model frameworks or underlying data used by published economic evaluations to assess the variability across analyses or quantify its affect on results. In anticipation of the advent of new approaches to treatment and new agents, it is timely to review the currently available analyses and describe their underlying data sources to aid future modeling activities. The objective of this study is to review the model structural frameworks and key parameters reported in published economic evaluations of treatments for chronic hepatitis C, and to use one such model to explore the impact of variability in key parameters on resulting ICERs.

The cost-effectiveness of antiviral treatments for chronic hepatitis C is an area of interest, reflected in the large numbers of studies identified by this review and those by other reviews [7] and [13]. A common framework for modeling chronic hepatitis C has been identified, which is currently populated with underlying data from a limited number of sources and associated with a high level of variability for key model parameters. This variability has the potential to significantly affect resulting ICERs. The findings of this review suggest that the data sources on which many economic evaluations are based may not adequately reflect the patient population or its heterogeneity, and may not be sufficient for the needs of current and future research questions. The structure of the studies identified by this review was based around six core health states, with the two most common deviations being the expansion of the chronic hepatitis C and decompensated cirrhosis health states. This review found that although some studies have addressed these points by presenting a structure that expands one or more of these states. The underlying data reported for one or more of the model parameters are the same, thus reducing the value of the perceived expansion. This is despite the availability of data for each of the possible model parameters for the expanded states. The value of the expansion of the early disease states is through the assessment of the cost-effectiveness of early treatment with antivirals at a time when the response rate is higher, but the patient may be asymptomatic and many years away from disease states associated with costly complications. However, in those studies that addressed this issue, the only distinction made with regards to utilities and costs was between those patients considered to be experiencing “mild” or “moderate” chronic hepatitis C, even in those studies that presented a model structure categorized by fibrosis score such as F0, F1, F2, F3, and F4. Progression rates for patients with early stage disease have been calculated through a meta-analysis of published prognostic studies [64]. The value of the expansion of the decompensated cirrhosis health state is a consequence of the anticipated differences in utility and cost associated with individual complications of decompensated disease, and over time (i.e., in the first year or subsequent years). In some studies in which expanded model structures are presented, the same values for utility and cost have been used for each of the possible health states, essentially rendering the visible expansion of the health states meaningless in the analysis. In other cases, the studies may report different values for one of the aspects of utility or cost, but not the other (e.g., reporting different costs for those experiencing ascites or refractory ascites), but reporting the same utility for both health states. It seems unlikely that if the resources required to manage refractory ascites are significantly less than ascites, that the quality of life of those patients experiencing refractory ascites wouldn't be improved compared with those experiencing ascites. We have reviewed the underlying data for key model parameters in published cost-effectiveness models, namely disease progression rates, utilities, and costs. A consistent finding is that there is a high level of variation in the values used for these model parameters and, following qualitative review, no apparent pattern was found regarding how data sources for these parameters have been chosen or how the variation is distributed across the studies. For example, for each parameter, one value has not evolved over time as the most frequently used and it has not been possible to identify a consensus regarding the use of one value for each model parameter. With regards to disease progression rates and utilities, the underlying data can be traced to a small subset of studies published prior to 1997. In Bennett et al. [53], the authors state that data on the natural progression of the disease are limited; rates of disease progression in the study were taken from published studies, reviewed by a panel of experts, and modified where appropriate. In the discussion, Bennett et al. recognize and describe the limitations of the data used for rates of progression in their analysis and state that these may need to be adjusted as more is learnt about the disease [53]. As a direct result of small number of data sources on which the studies included in our review are based, the data frequently do not directly reflect the country-specific perspective of the research question posed. Because the ICERs achieved are significantly influenced by these values and in particular by disease progression rates, their validity is fundamental to the interpretation of the results. One common assumption used was that once patients had achieved SVR, their utility was 1 (10 of the studies), although in four studies an SVR-specific utility was assigned following treatment success. Patients who enter the SVR state commonly remain in it for the remainder of the model and, therefore, its associated costs and utilities have the potential to significantly affect ICERs and the interpretation of results. The age and comorbidities of a typical population with hepatitis C render a utility of 1 unlikely, and may be artificially high, and thus the utility for the SVR state requires further consideration. For example, a recent analysis demonstrated that if the utility of the SVR state was age dependent (i.e., utility decreased with increasing patient age), while all other parameters remained equal, the resulting ICERs would be higher [65]. Hence misleading and optimistic results may be associated with the assumption that those achieving SVR have a utility of 1, particularly in older patients. It is noteworthy that 20 studies did not explicitly comment on whether or not the patient may experience subsequent progression and complications following SVR. Current studies suggest that a number of factors can impact response to treatment and disease progression, including: age, gender, aminotransferase (ALT) levels, disease severity, genotype, comorbidities and coinfections (e.g., HIV), drug use, alcohol abuse, prior treatment and response to that treatment, and disease progression [7] and [66]. All of the models identified by this review adopted a simple Markov-based framework, many of which were run with a single cohort based on patients with average baseline characteristics, and thus they lack the capacity to adequately model patient heterogeneity. As with other complex and long-term disease and as our understanding of the disease evolves and the approach to treatment becomes more complex, this approach may be considered over-simplistic. The assumption that mean transition probabilities, utilities, and costs apply to all individuals within a cohort may mask the contributory effects of some patient sub-groups, for example, those with favorable ICERs in whom therapy may be most cost-effective, or those with poorer ICERs in whom therapy may not be considered cost-effective. For example, a recent analysis demonstrated that using static or age-independent rates of disease progression has the potential to provide higher ICERs and misleading and pessimistic results, particularly in older patients [67]. Accommodating patient heterogeneity and the dynamic nature of an individual's progression through the disease may be required to accurately model the future standards of care and address future research questions. Indeed the incorporation of a more dynamic approach to modeling heterogeneous patient populations has already been adopted as standard practice in other therapy areas such as diabetes [68]. Such modeling approaches will only be worthwhile if sufficient data exist to populate all of the required input parameters. The ICERs obtained from this analysis are for illustrative purposes, solely to demonstrate the potential impact of the variability in underlying data for key model parameters seen across published analyses. We have found that the variability in the values incorporated has the ability to change the resulting ICERs for new treatments from dominant to dominated, when all other considerations remain equal. More complex interpretation of these results is limited by our simplistic approach. We have run the model using SVR rates for patients with HCV genotype 1 only, and input values for disease progression, utility, and cost from all identified studies (i.e., from heterogeneous patient populations including both those with genotype 1 and other genotypes). The analysis is still considered insightful because the variability across the extracted values for these model parameters was independent of study country perspective, year of analysis, and cohort characteristics. This review was limited in the aspects of published cost-effectiveness analyses that could be considered. Some of the older publications do not report key information required for this assessment, thus the analysis may be biased towards newer studies with more complete reporting. In addition, many small differences are seen between individual models, and it was beyond the scope of this study to assess all of these. For example, the possible transitions to and from transplantation varied across the identified studies, and in some studies transplantation was not included at all, largely as a consequence of local disease management strategies. Furthermore, it's possible that differences between studies, such as country perspective and year, may affect results, and by using simply the maximum, minimum, and mean we may have masked these differences. We did not deem it necessary to control for possible country-specific variation because the underlying data used for transitions and utilities were rarely directly relevant to the country perspective, and as a result of a lack of good quality data, were commonly taken from a small pool of studies and utilized in good faith irrespective of their country of origin. Hence, conducting the analysis by country would not provide a more valuable result than those presented from this analysis. For similar reasons, it was not deemed necessary to account for the age of the study; the underlying data sources are the same for more recent studies as they are for older studies. Other approaches to weighting the mean were not considered because of inconsistencies in the approach to reporting data and underlying data sources. Our review excluded a small number of studies that recruited patients specifically because of their comorbidities and coinfections in order to permit comparability across the greatest number of studies. However, it is likely that in the future, to achieve optimal effectiveness and cost-effectiveness, antiviral treatment will be tailored to specific patient groups based on their individual characteristics. The prognostic factors associated with these groups should be specifically addressed in any modeling activities intended to assess the cost-effectiveness of treating such subgroups. As the treatment landscape continues to evolve, we anticipate that the research questions of the future will address the heterogeneity that exists in a normal HCV-infected population, and will incorporate a new set of early markers of treatment effectiveness, in addition to rapid viral response that may emerge to help clinicians optimize patient outcomes, and mechanisms for screening to identify those patients most likely to benefit from treatment before antiviral therapy is initiated [69]. With this in mind, the future modeling framework will require increased flexibility and the ability to tailor analyses for individual patients. This may necessitate a move away from a Markov modeling approach and introduce the use of discrete event simulation. In addition, greater emphasis should be placed on justifying the choice of underlying data and ensuring that data used are the best available that are relevant to the research question. It is recognized that to obtain such data may require long-term observational studies or prospective data collection, conducted over possibly decades. In conclusion, economic evaluations to date have been based on a common and well-accepted model framework that is simplistic as a consequence of the lack of available data reflecting patient heterogeneity. Despite the paucity of data for incorporation into these studies, there is high variability across the input parameters used, which can have a significant impact on the results obtained. For future modeling activities is it important that the model framework and underlying data sources are appropriate to address the research question posed. As our knowledge of hepatitis C and its management continues to evolve, further research regarding improved characterization of chronic hepatitis C, its progression within the context of the current and future patient populations and treatment options, and the accuracy of the health-related quality of life and costs associated with each health state is required to improve the accuracy of such analyses.