Economic evaluation of pegylated interferon plus ribavirin for treatment of chronic hepatitis C in Thailand: genotype 1 and 6

This study was a model-based economic evaluation. A Markov model was developed from a societal perspective to estimate costs and health outcomes of CHC patients treated with a combination of pegylated interferon alpha 2a or alpha 2b and ribavirin versus a usual palliative care. The study population was a hypothetical cohort of 1000 CHC genotypes 1 or 6 patients. Although, the Thai treatment guidelines (THASL) [12] recommends pegylated interferon and ribavirin combination to an 18-year-old HCV patient, this model cohort assume the patients received treatment at 45 years which is likely to be an average age of current patients in Thailand and other countries [16, 17]. At the beginning of the study, according to THASL guidelines [12], CHC patients genotypes 1 or 6 were 45 years old, a positive RNA test and METAVIR score ≥ 2. The treatment regimens were designed based on THASL guidelines [12] as follows: 1) pegylated interferon alpha-2a 180 μg once a week plus ribavirin 1000 mg/day for 48 weeks, or 2) pegylated interferon alpha-2b 1.5 μg/kg weekly plus ribavirin 800 mg/day for 48 weeks, or 3) a palliative care. SVR rate was assessed 24 weeks after treatment discontinuation, as the achievement. If HCV RNA was undetectable (less than 50 IU/ml) at week 4, defined as a rapid virological response (RVR), the treatment stopped at 24 weeks. If HCV RNA was undetectable at week 12, defined as an early virological response (EVR), the treatment stopped at 48 weeks. If HCV RNA was reduced by less than 2 log at week 12 compared with the baseline level, defined as a null response (NR), the treatment was stopped at week 12. For the remaining cases, we assumed the treatment results as a partial nonresponse (PR), which the treatment was stopped at week 24. Figure 1 presents the study treatment guidelines. In the base-case analysis, costs and health outcomes were discounted at 3 % annually [21].

Figure 2 illustrates the schematic diagram of the Markov model composing of six health states, including CHC, compensated cirrhosis, decompensated cirrhosis, HCC, healthy person, and death. The patients were moved through health states based on transition probabilities. The model started at CHC patients who meet the inclusion criteria for a treatment. If the CHC patients achieved a treatment, they could proceed to the healthy person state, if not, they could continue to the compensated cirrhosis or HCC or death states. The patients in the compensated cirrhosis, decompensated cirrhosis, and HCC could not reverse to CHC or healthy person states. This model assumptions were: 1) the CHC patients weigh sixty kilograms, 2) the patients failure to the treatment are not re-treated, 3) the patients respond to the treatment were not re-infected from HCV, and 4) the patients are 100 % compliant with the treatment. This model was simulated throughout the patients’ lifetime with a 1-year cycle length.

The parameters used in the model were health state transition probabilities, efficacy data, utility data, and cost data. Health state transition probabilities and utility data were based on the study by Werayingyong, P. and Teerawattananon, Y [13], which conducted a systematic review from published literatures and a meta-analysis of CHC patients’ utility according to their health states (Table 1). Efficacy data of treatment with a combination of pegylated interferon alpha 2a or alpha 2b and ribavirin versus a usual palliative care for CHC genotype 1 were retrieved from a meta-analysis study [22] which included randomized controlled trial studies. According to the expert meeting, the inclusion criteria of RCT studies recruited in the study were publishing after 2008, and reporting SVR and RVR. Finally, the SVR and RVR from three RCT studies, and NR from two RCT studies were pooled to calculated for SVR, RVR and NR. Additionally, Thai general population death rates at each age were used in the analysis [23]. For a 45 year old patient, probability of death from other causes was 0.0044 and varied throughout the patient’s lifetime. All input parameters used in the model presented in Table 1.

The costs of CHC treatment composed of direct medical costs and direct non-medical costs. To avoid double counting for utility outcomes, indirect costs were excluded. Direct medical costs included drugs (pegylated interferon and ribavirin), laboratory tests (investigation and monitoring) and complication treatment. The prices of pegylated interferon and ribavirin were obtained from the prices that the drug companies presented to the NLEM committee. The types and the number of laboratory tests used in the model were estimated by the expert panel. The tests for investigation included complete blood count (CBC), liver function test, prothrombin time, genotype, viral load, HIV, creatinin, thyroid-stimulating hormone, antinuclear antibodies, abdominal ultrasound, pregnancy test, alanine aminotransferase (ALT), and aspartate aminotransferase (AST). For monitoring, CBC, creatinin, ALT and viral load were included.

Direct non-medical costs included transportation and food expenditure of patients. The number of OPD visits, IPD admissions and length of stay were retrieved from the existing study conducted in Thailand [33]. The unit costs of transportation and food expenditure were derived from Thai Standard Cost List for Health Technology Assessment [34]. All costs were converted in 2013 values by using consumer price index (CPI) and discounted at a rate of 3 %. The average exchange rate of Thai baht (THB) to 1 $US was 35 Baht (Table 2).

Model structure and all parameters were approved by the experts during the expert consultation meeting.

A one-way sensitivity analysis and a probabilistic sensitivity analysis (PSA) were performed to determine the uncertainty of model parameters. For one-way sensitivity analysis, each parameter was varied at a time across the plausible range and shown graphically as a tornado diagram. In addition, PSA were carried out by varying all parameters randomly within the plausible range. The Monte Carlo Simulation was generated in order to randomly select a value of each parameter for 1000 times and calculated for expected cost and outcome. The results of PSA were presented by a cost-effectiveness plans and acceptability curves.