Impact of Drug Therapy Duration in Patients With Hepatitis C
Adherence to drug therapy in hepatitis C increased total costs in year 1 relative to nonadherent patients, then decreased total costs in year 2.
The hepatitis C virus (HCV) is a major medical and worldwide public health concern. The prevalence of chronic HCV infection in the United States is estimated between 2.7 and 3.9 million.1 Aggregate costs were estimated at $1.8 billion in 1997 and are expected to increase significantly2-4 as the cohorts of individuals infected with hepatitis C several decades ago begin to seek medical care. Until recently, the standard therapy for HCV consisted of a combination of pegylated interferon and ribavirin (P/R),5-8 which has been shown to be cost-effective.9-13 Despite the risk of liver failure, liver transplantation, and cancer, patients with chronic HCV frequently delay treatment because of the lack of associated symptoms in the early stages of the disease and the frequency of adverse effects with antiviral therapy.14-17 Recently approved therapies have demonstrated greater sustained virological response rates and a potentially shorter required duration of treatment when used in conjunction with pegylated interferon and ribavirin.18,19
Key parameters for estimating the economic implications of new therapies for hepatitis C include valid estimates of the rate of adequate duration of P/R therapy in real world practice and estimates of the incremental costs associated with duration of therapy. Butt and colleagues20 compared matched cohorts of HCV-infected and uninfected patients treated by the Veterans Administration (VA). While the presence of an HCV infection was estimated to increase the risk of death by 37% relative to non-HCV patients, the risk of death decreased by 29% for patients treated for up to 23 weeks (hazard ratio [HR] = 0.71); by 40% for patients treated for 24 to 47 weeks (HR = 0.60); and by 59% for patients treated for greater than 48 weeks (HR = 0.41). However, Butt and colleagues20 did not investigate the relationship between the duration of P/R treatment and healthcare costs or the risk of adverse HCV-related events, which have been documented to be increasing in prevalence. For example, Kanwal and colleagues21 documented that the prevalence of cirrhosis doubled in a cohort of VA patients with HCV, from 9% in 1996 to 18.5% in 2006. Similarly, the prevalence of hepatocellular carcinoma increased 20-fold (0.07% to 1.3%).
This study estimates the impact of duration of P/R therapy on healthcare costs and the risk of adverse clinical events over 2 years in a commercially insured population in the United States. Our hypothesis is that duration of P/R therapy will be associated with reductions in nondrug healthcare costs and adverse event risk, especially in the second year when P/R therapy has been completed.
Deidentified paid claims data were derived from a large health insurance company in the United States that provides a range of insurance plans to commercial, Medicare, and Medicaid patients. The vendor of the data used in this analysis provides these data to researchers in a format that meets all Health Insurance Portability and Accountability Act requirements and institutional review board approval at Bristol-Myers Squibb.
Patients treated with an HCV infection were identified using data for all HCV patients enrolled during an 8-year period (2003-2010) based on primary or secondary International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) HCV diagnoses recorded on paid claims (070.41, 070.44, 070.51, 070.54, 070.70, 070.71, or V02.62) and the use of P/R drug therapy. A 2½-year observation period was then defined for each patient treated for HCV as 6 months of pre-treatment and 2 years of post-treatment data. Patients with a diagnosis of cirrhosis (571.5, 571.2, 571.6), liver cancer (155.0), or liver transplantation (996.82, V42.7, current procedural terminology codes: 00796, 47135, 47136) in the period prior to their index treatment date were excluded.
Definition of Duration of P/R Therapy
Duration of uninterrupted P/R therapy was calculated based on the assumption that pegylated interferon was to be administered weekly using subcutaneous injections and ribavirin taken orally twice a day. A patient was assumed to have interrupted therapy if a gap of greater than 30 days appeared after the expected date of the next injection of pegylated interferon or purchase of ribavirin based on the days supply dispensed previously. There is no accepted standard for setting the gap definition of the calculation of days of continuous therapy and we used the days supply for a typical oral prescription. Under this gap assumption, the count of continuous days of P/R therapy was set equal to the number of days between the start of therapy and the end of the last injection period (date of injection + 14 days) or the end of the supply of the last prescription of ribavirin (fill date + days supply). Any period under 30 days during which the patient was off both therapies before restarting treatment would be included in the count of continuous days of treatment.
The calculated duration of therapy was used to define 2 levels of duration of treatment: 24 to 48 weeks and over 48 weeks of therapy based on the American Association for the Study of Liver Diseases (AASLD) treatment guidelines.22 The AASLD guidelines also differentiate the minimum duration of therapy required for in-treatment guidelines specified by genotype. Using the limited laboratory data available in this data set, we found that 76% of our HCV patients were viral genotype 1 or 4, requiring 48 weeks of therapy.
The dichotomous variables for levels of duration of therapy were entered into the analysis using 2 approaches. First, the study sample was divided into 3 mutually exclusive subsets based on duration of therapy, and a dummy variable was entered into the analysis to estimate the marginal impact of 24 to 48 and 48 weeks or more of treatment relative to treated patients with less than 24 weeks of treatment. The second approach divided the patient sample into 2 groups with and without 24 weeks of treatment. A second dichotomous variable for 48 weeks or more of treatment was also created and entered into the multivariable models as an interaction variable with the dichotomous variables for 24 weeks or more of treatment. This specification measured the impact of achieving 24 weeks of treatment and the incremental impact of duration of treatment beyond 48 weeks. That is, this specification measures the extent to which duration of treatment beyond 48 weeks results in better treatment outcomes and lower costs.
Patient Outcome Measures
Costs were measured over each of the 2 years posttreatment broken down by type of service (physician services, hospital care, prescription drugs, etc). Costs were calculated based on the actual paid amount with no adjustments for infl ation. The incremental health risks were measured using dichotomous variables indicating if a diagnosis for liver cancer, liver transplantation, or cirrhosis appeared for the first time in the patient’s paid claims data following the initiation of P/R therapy. We also analyzed whether or not the patient experienced an episode of depression (296.2, 296.3, 300.4, 311) in the post-treatment period, though it need not have been a new episode. The presence of a depression diagnosis in the pre-treatment period was used as an independent variable in all analyses.
Multivariable statistical methods were used to control possible differences in 3 categories of duration of therapy. Independent variables were defi ned using data from the 6-month pre-treatment period. These variables included age, geographic region, insurance source (private, Medicare, Medicaid), insurance type, use and cost of healthcare services broken down by service type, 21 comorbidity classes based on ICD-9-CM codes, and 15 drug profi le variables based on prescription drug claims. Ordinary least squares (OLS) models were estimated for costs. Generalized linear models with a gamma distribution and log link function were estimated to test the sensitivity of the OLS results for costs to alternative estimation procedures. Logistic regression models were estimated for event risk associated with duration of treatment. The sensitivity of the OLS results compared with alternative estimation procedures was tested using a generalized linear model with a gamma distribution and log link function.
A subsample of patients treated for HCV with laboratory data on genotype was identified for a sensitivity analysis of the impact of duration of P/R therapy on patient outcomes. In this analysis, duration of therapy was defined depending on genotype. Specifically, duration of treatment by patients with genotype 1 or 4 required more than 48 weeks of P/R therapy while patients with genotype 2 or 3 were classified as adherent if they achieved 24 weeks or more of P/R therapy. A P value below .05 was considered statistically significant.
A total of 148,176 patients infected with HCV were identified using the 8 years of paid claims data of whom only 7840 (5.3%) received P/R treatment. A total of 1856 patients exhibited 6 months of eligibility prior to treatment and 2 full years of eligibility following treatment. A subsample of 519 treated patients had data on genotype.
The descriptive statistics for the study population are presented in
and are broken down by adherence groups. The age and gender distributions of the study sample correspond well with the national data on the characteristics of the hepatitis C patient who are predominately male and over 45 years of age. Most of our study sample resides in the South, which is also consistent with national data. The remaining data document the values for those variables likely to impact the cost of treating hepatitis C patients: insurance source and type, prior healthcare costs, and comorbidity and prescription drug profile.
Only 665 (35.8%) of treated HCV patients completed 24 to 48 weeks of P/R treatment while an additional 402 patients (21.7%) completed 48 weeks or more of treatment. Just over 76% of patients treated for HCV were viral genotype 1 or 4, which implies that most patients did not complete an adequate course of therapy as defined by AASLD guidelines.22 This is confirmed in the sample of patients with genotype data in which only 37.2% of patients completed a course of P/R therapy consistent with treatment guidelines for their genotype. Only 4 of the characteristics listed in Table 1 were statistically different across the 3 classes of the duration of therapy used in the analysis: the prevalence of infectious diseases and digestive disorders, and the use of beta-blockers and calcium channel blockers in the 6 months prior to treatment.
Impact of Duration of Therapy in Year 1
The impact of duration of P/R therapy on healthcare costs, service use, and the risk of adverse events in the first post-treatment year are presented in
Duration of P/R treatment for 24 to 48 weeks did not significantly reduce medical costs (—$3247) but did increase drug costs by $10,671 (P <.0001). The corresponding firstyear results for duration of treatment 48 weeks or more were a —$6194 reduction in medical costs (P <.01) and an increase in drug costs of $27,116 (P <.0001). About half of the savings in medical costs associated with duration of treatment was due to reductions in hospital costs and reduced risk of a hospital admission. However, duration of P/R therapy was not found to impact the likelihood of ED use or the risk of post-treatment depression in year 1. Duration of treatment in excess of 48 weeks did reduce the risk of cirrhosis (odds ratio [OR] = 0.514, P <.05).
The results for the number of outpatient visits indicates that duration of P/R therapy significantly increased the number of outpatient visits consumed by patients using 24 to 48 weeks (+2.14 visits, P <.0001) and 48 weeks or more of treatment (+3.32, P <.0001). Hospitalizations were reduced by 0.22 and 0.37 admissions, respectively (P <.0001 for both estimates), while visits to the ED during year 1 were not associated with duration of P/R treatment.
Results from the logistic regression analyses of the likelihood of hospital admissions and ED use confirm the results for visit counts. Persistence with pegylated interferon and ribavirin therapy of 24 to 48 weeks reduces the likelihood of a hospital admission by 24% relative to nonpersistent patients (OR = 0.764, P >.05) while persistence of 48 weeks or more reduces the likelihood of a hospital admission by 58% (OR = 0.42, P <.001). The results from model 2 measure the impact of extending duration of therapy beyond 24 to 48 weeks and are consistent with the results reported for model 1.
The estimated impacts of duration of treatment on cost and event risk using data for those patients with genotype data are consistent with the results for all patients. In the first year, geno-specific adequate duration of therapy is associated with lower medical costs which are not sufficient to offset the significantly higher drug costs associated with an adequate course of therapy ($13,085, P <.0001). Duration of therapy was also associated with a significant reduction in the risk of hospitalization during the first post-treatment year of 60% (OR = 0.398, P <.01).
Impact of Duration of Therapy in Year 2
Persistence with P/R therapy was found to reduce healthcare costs in year 2 for all types of service (
). Persistence of 24 to 48 weeks was associated with lower medical costs (—$3069; P >.05), prescription drug costs (—$3182; P <.0001), and total cost of —$6171 (P <.05). Persistence of 48 weeks or more was associated with a reduction in total cost in year 2 of —$8130 (P <.01) relative to treated patients who consumed less than 24 weeks of P/R therapy.
Duration of P/R therapy in year 1 reduced use of outpatient visits in year 2, reversing the corresponding results found in year 1, and significantly reducing the number of admissions in year 2. Duration of treatment was not found to affect the risk of emergent depression or hepatic cancer in year 2 but did significantly reduce the risk of cirrhosis. Model 2 results for year 2 found that extending P/R therapy beyond 48 weeks resulted in an additional $1952 (P <.05) in drug costs but also reduced total costs by an additional —$1741. Most of the remaining estimated marginal effects of extended treatment were not statistically significant except for a significant reduction in hospital admissions of –0.15 admissions (P <.0001) relative to patients treated for 24 to 48 weeks.
The estimated impacts of genotype-specific definitions of an adequate duration of therapy on cost and event risk in year 2 were again consistent with the results from the analysis of all patients. Achieving an adequate level of genotype-specific duration of therapy was associated with lower medical costs and drug costs, which result in a significant reduction in total costs of —$6295 (P <.05). Duration of therapy was also associated with a significant reduction in the incidence of cirrhosis in year 2 of 89% (OR = 0.107, P <.01).
Sensitivity Analysis Using Generalized Linear Models of Cost
The sensitivity of the OLS results for ambulatory care, medical costs, prescription drug costs, and total costs to alternative estimation procedures were tested and reported in
. The estimated impact of duration of treatment using a generalized linear model with a gamma distribution and log link function are consistent with the corresponding OLS results. While the estimated effects using generalized linear models are smaller in magnitude, the qualitative results do not change. Duration of P/R therapy increases costs in the first post-treatment year due to the cost of drug therapy, which is sensitive to duration of treatment. In year 2, all costs are lower for patients who were adherent with P/R therapy in year 1.
For most diseases and illnesses, the effi cacy of treatment is related to the duration of P/R therapy. Antiviral therapy for chronic hepatitis C infection is no exception. The AASLD guidelines22 recommend a minimum of 24 weeks of pegylated interferon and ribavirin therapy for patients with HCV with genotypes 2 and 3, and 48 weeks of P/R treatment for patients with genotypes 1 and 4. Indeed, McHutchison and colleagues23 demonstrated that the sustained viral response (SVR) rates using P/R therapy were decreased if patients were unable to achieve minimum treatment duration thresholds. Equally important, the ability to persist on P/R therapy is associated with improved survival.20 Patient survival was incremental as related to the viral response to antiviral treatment. However, the impact of duration of treatment on healthcare costs and event risk has not been systematically evaluated.
The results of our study demonstrate a direct relationship between duration of therapy and healthcare costs. While approximately 76% of patients with HCV in our study sample were estimated to be genotype 1 or 4, less than 22% of patients in the study sample achieved the required 48 weeks of P/R therapy. The rate of adequate duration of therapy adjusted for genotype was found to be a little over 37% in the subsample of 519 patients with data on genotype. Even so, our results provide evidence that duration of P/R therapy beyond 24 weeks in the first post-treatment year will lead to signifi cant reductions in healthcare costs in year 2, including prescription drug costs. While year 2 savings associated with adequate duration of therapy were not sufficiently large to offset the $10,000 to $27,000 drug bill associated with duration of therapy in year 1, it is anticipated that duration of pegylated interferon and ribavirin therapy can lead to reductions in healthcare costs in years 3 and beyond.
The results presented here provide some of the key elements for projecting the cost impact of new therapies for treating HCV, which are already widely accepted despite the lack of cost-effective studies. The results of our study suggest that duration of antiviral therapy may be an important factor for efforts to contain healthcare costs associated with HCV. New therapies are expected to increase the likelihood of successful treatment by increasing the likelihood of achieving SVR and/or reducing the duration of therapy needed to achieve response.
This study has several important limitations. First, the number of treated patients with 2 years of post-treatment data is limited (N = 1856). Second, we do not have data on year 3 and beyond mainly due to the requirement that this level of data availability would have significantly reduced our already limited sample. Nevertheless, our study remains one of the largest to assess the impact of duration of P/R therapy on future costs for patients with HCV. Additional studies are needed to examine if our results are generalizable beyond 2 years.
This study used a 30-day gap definition to define when a patient interrupted or stopped treatment and terminate the count of continuous days of treatment. Arguments for shorter gap definitions can be made, especially in light of the 1-week duration of pegylated interferon injections. It is possible that our results would change if a different definition were used. Unfortunately, toggling between alternative gap defi nitions to test the sensitivity of study results was not undertaken, as this would have required that a completely new analytic fi le be created rather than simply changing a few parameters in an analysis.
An adequate course of therapy depends on the genotype of HCV, with type 2 and 3 genotypes requiring significantly less than 48 weeks of treatment. To test the sensitivity of our results to genotype, this study defined adequate duration of treatment based on the patient’s genotype for the small subsample of 519 patients for which these data were available. However, we still believe our results are accurate since the sensitivity analysis was consistent with our core estimates.
Finally, our results cannot estimate the impact of SVR on costs, which would be a better measure of treatment effectiveness often achieved before the duration of therapy specified in treatment guidelines. Data on the impact of SVR on cost would be particularly useful for modeling the potential cost-effectiveness of new therapies which may directly increase the rate of response while decreasing the duration of P/R therapy needed.
Our result may not be generalizable to all patient populations. The data for this study were derived from a commercial health insurance company in the United States and our results may not correspond well to the treatment duration achieved in other countries. The direct healthcare costs associated with duration of therapy are likely to be very specific to the US healthcare system. Even within the United States, patients with HCV are much more likely to be uninsured.24
The standard of care for treatment of chronic hepatitis C has currently evolved to include the newly approved protease inhibitors telaprevir and boceprevir. Moreover, hepatitis C treatment strategies will continue to evolve as an increasing number of therapies with different modes of action reach the market. However, to be cost-effective, these new adjunctive therapies must either improve duration of therapy with pegylated interferon and ribavirin therapy or improve patient outcomes associated with shorter treatment duration. Our results provide a baseline against which future studies of combination therapy consisting of pegylated interferon, ribavirin, and a protease inhibitor can be evaluated.
The results of our study demonstrate that duration of pegylated interferon and ribavirin therapy is associated with greater costs during the first year of treatment. However, total costs were greater in patients that were not able to complete the duration of therapy at year 2. The implications for managed care decision makers are 2-fold. First, strategies for containing the cost of hepatitis C need to incorporate drug programs to improve duration of therapy. Second, formulary evaluations of emerging therapies for hepatitis C should include the impact of these therapies on duration of the core P/R drug regimen in addition to other factors, such as improved viral response.