This study consists of a two-part model. First, a one-year decision tree model was used to estimate the relevant costs and efficacy outcomes in each treatment strategy for first year when patients receive either clopidogrel or ticagrelor. In second part, a Markov model in Microsoft Excel and TreeAge was used to simulate a cohort of 1,000 patients with ACS with an average 62 years and long-term health and cost outcomes could be extrapolated. The clinical efficacy and safety data were extracted from international evidences due to lack of local data. However, we used local costs based on official tariffs in public sectors.
Population of interest:
In the present study, the characteristics of patients of our hypothetical cohort were assumed to be the same as acute coronary syndrome patients in PLATO trial with or without ST segment elevation, or unstable angina. According to PLATO study, the mean age of patients was 62 and 25% of them were women. These characteristics were comparable with Iranian ACS patients whose mean age was 60.5 and 62.9, respectively in NSTEMI and STEMI groups; and 27% of them were females according to registry data of Iran heart center hospitals (29).
Treatment strategies:
The ACS patients could receive either ticagrelor 180 mg as loading dose and 90 mg BID afterward or clopidogrel 300 mg as loading dose in medical approach or 600 mg in patients whom are candidate for PCI followed by 75 mg once daily. Both medicines were assumed to be used only for 1 year.
1-year decision tree model
In this decision tree model, ACS patients could end up with one of four health states including «No event», «Post MI», «Post Stroke», and «Death» at the end of one-year treatment (Figure 1a.)
The results of PLATO trial were applied to extract clinical efficacy and safety data of ticagrelor, and clopidogrel in ACS patients, given it was head-to-head design (30).
Long term Markov model
A Markov model (figure 1 b) was used to simulate long term medical and cost consequences of using each of these treatments on remaining years of life. This model includes six states of «No event», «Non-fatal MI», «Non-fatal Stroke», «Post MI», «Post Stroke», and «Death», therefore hypothetic patients resulted in each of four states of one-year decision tree model are transferred to first cycle of Markov model. The cycles of Markov model are 1 year and the time horizon of model is 20 years. Since, it is assumed that patients receive either clopidogrel or ticagrelor only in first year, the probabilities of transition in Markov model are identical in both groups (31), therefore the distribution of patients in each state after 1 year of decision tree model (input of Markov model) is the determinant factor. The same approach is used by other cost effectiveness studies (32-35). In case of mortality in Markov model, 2019 life table of Iran (36) were used for developing an age dependent mortality risk, adjusted by hazard ratio (HR) of death in each of no event, non-fatal MI, non-fatal stroke, post MI and post stroke states (37). More details about probabilities of decision tree and Markov model are provided in table 1.
Utility scores:
Health utility of each state were extracted from international studies that were calculated based on prospective EQ-5D data collection in PLATO trial and extrapolation methods (32,37). 3% annual discount rate was considered for including time preferences in the model. Details are provided in table 1.
Cost analysis:
In the current study, only direct medical costs were included based on payer perspective. To estimate the cost incurred by each treatment strategy, the local cost of procedures and cares for each event was included in the analysis. Since the prices of pharmaceutical products are quite the same in all centers in Iran, they were extracted from Iran FDA’s list of pharmaceutical prices. Medical tariffs of public sector were used based on patient discharge documents collected from 5 public hospitals and after double checking them with latest officially published tariff list. According to these data, a cost per package of care were estimated for different states of model (Table 1). 7.2% as annual discount rate was used to include the value of time preference in costs.
Cost effectiveness threshold:
The result of analysis was reported based on ICER. The measurement of effectiveness was according to QALYs, and the costs were calculated in 2022 US Dollars (exchange rate: USD $1 = IRR 249,359 [Iranian Rial]). To assess cost effectiveness of healthcare intervention in Iran, CEA threshold of 1604 USD $ was used, based on latest announced acceptable CEA threshold by HTA committee of Iran FDA.
Sensitivity Analysis:
The impact of uncertainties of model inputs on the final results of the study was addressed using deterministic and probabilistic sensitivity analysis. For deterministic sensitivity analysis, ±5% variation in some important inputs including the average cost of MI care, the average cost of stroke care, ticagrelor and clopidogrel prices, probability of MI, stroke and death in first year were assigned. For probabilistic sensitivity analysis, a Monte Carlo simulation were used with 5000 iterations in order to produce scatter plot and acceptability curve.