Cost-effectiveness analysis of highly concentrated n-3 polyunsaturated fatty acids in secondary prevention after myocardial infarction

A Markov model was developed in Excel version 2013, starting with the event-free MI (post-MI). During each 1-year cycle of the model, some patients entered one of the four health states (cardiovascular mortality, non-fatal MI, non-fatal stroke, and heart failure (HF)), while the others remained in the event-free state (Figure 1).The rate at which people moved through the model was regulated by transition probabilities, which described the likelihood of moving between states over each model cycle. The transition between each state and event was determined by probabilities and adjusted factors obtained from randomized control trials and the local registry database. The model was run until the patients were 100 years old or died.

Figure 1

The main assumptions made in this model were as follows; 1) patients in the acute health state (MI, stroke, HF) would not move to other acute health state; 2) the effectiveness of n-3 PUFAs would be maintained over life-time horizon; 3) the duration in each cycle was 1 year.

The discount rate of 3% was applied for both costs and outcomes. The time horizon of the model and the effects in life expectancy were life-long calculation. The provider perspective was chosen for this study; thus only direct health care costs were used in the analysis.

We carried out a meta-analysis to estimate the effectiveness of n-3 PUFAs for the base case analysis. A comprehensive literature search of MEDLINE, EMBASE, and the Cochrane Library (up to September 2013) was conducted. All trials that met the following criteria were included: 1) the trial enrolled adult patients (aged at least 18 years) with a history of myocardial infarction; 2) n-3 PUFAs were consumed in dose of at least 1 gram/day; 3) the followup period was at least 1 year, which would be long enough for n-3 PUFAs to demonstrate efficacy; 4) the outcomes included CV diseases or events, cardiovascular death, sudden death, all-cause mortality, coronary heart disease, heart failure, stroke, fatal or nonfatal MI. The exclusion criteria included studies with less than 1-year duration, studies with multiple interventions from which the effects of n-3 PUFAs could not be separated, and studies that did not report CV-related endpoints. If two or more articles reported results from the same group of patients at different follow-up periods, only the article with longer follow-up period was retained in order to prevent data duplication.

Titles and abstracts obtained from the electronic and bibliographic searches were rejected on initial screening if it could be determined that the articles did not meet the eligibility criteria. In case the reviewer was uncertain to reject an article based on its title/ abstract, the full text of the article was further evaluated. Data extraction was performed by 2 reviewers independently. In case there were discrepancies, the reviewers discussed or consulted the third reviewer. If the same studies were presented in more than 1 article, we included only the larger or first published study.

From 4,042 articles identified from the database search, 3 studies which met the inclusion criteria and did not meet the exclusion criteria were included in the meta-analysis. The heterogeneity of studies was not statistically significant only in the cardiovascular mortality endpoint, while the endpoint of total mortality and sudden death showed significant heterogeneity. Therefore, the cardiovascular mortality endpoint was selected, which was reported in 2 studies, i.e. GISSI-P (Gruppo Italiano per lo Studio della Sopravvivenzanell’ Infarto Miocardico-Prevenzione) study [5, 18], and OMEGA (Effect of Omega 3-Fatty Acids on the Reduction of Sudden Cardiac Death after Myocardial Infarction) study [7]. In total 15,174 patients after MI were included. The results of included studies were pooled using fixed effect model, weighted by inverse of variance. The summary of studies included in the meta-analysis has been shown in Table l.The meta-analysis results in Figure 2 demonstrated the significant reduction in cardiovascular mortality in patients using n-3 PUFAs compared to those using standard treatment only.

Figure 2

Table 1

Summary of studies included in the meta-analysis

Effectiveness data of n-3 PUFAs on MI and stroke were obtained from the GISSI-P study (5, 18). There was only 1 study using high dose of n-3 PUFAs in 6,975 patients with chronic heart failure, 41% of whom had a history of MI; so the result of this study was used to estimate the effectiveness in HF (19). These effectiveness data are shown in Table 2.

Table 2

Parameters used in the Markov model

As the provider perspective is used for this study, cost data in the study included only direct health care costs, which were defined as the costs of goods and services directly provided by the service provider such as medication cost, hospital cost, etc. All costs were inflated to 2013 values using Consumer Price Index (CPI) of medical care group from the Bureau of Trade and Economic Indices, Ministry of Commerce [20]. The MI costs were taken from a published literature using data collected from patients in the Thai ACS Registry who were admitted to Ramathibodi Hospital [21]. The stroke costs were obtained from a prospective observational study collecting data in 2 regional hospitals [22]. Since the costs of stroke in subsequent years were not collected, it was assumed to be the same as in the first year except for hospitalization cost. The costs of heart failure were collected from 2 sets of the National Health Security Office (NHSO) of Thailand database, i.e. hospitalization cost of inpatient from government hospitals and the cost of patients visiting the out-patient departments of hospitals under Bangkok Metropolitan Administration. As Omacor (Abbott Laboratories) is the only highly concentrated formulation of n-3 PUFAs in Thailand, the price in the Drug and Medical Supply Information Center (DMSIC) website was used in this study. All costs are shown in Table 2.

The utility weights of acute state of each condition were lower than those of post state. The utility weight data of MI, stroke and HF were obtained from literature review [23-27], as shown in Table 2.

Data from the Thai ACS Registry collected from 1,525 patients who experienced MI were used to estimate the probability of cardiovascular mortality using survival analysis. The probabilities of transition between other health states were derived from published articles.

The health outcome measures were quality-adjusted life year (QALY) and life-year gained (LYG). QALY was used to measure overall health-related quality of life, which is the preferred approach in economic evaluation of health intervention. LYG was a modified mortality measure in which remaining life expectancy was taken into account. Life years were calculated as the remaining life expectancy at the point of each averted death.

Univariate sensitivity analyses were performed to identify the variables which had significant impacts on the results. The key individual variables which influence the results of cost-effectiveness have been shown in a Tornado diagram. To examine the uncertainty of inputs in the model, the probabilistic sensitivity analysis using a second-order Monte Carlo simulation was conducted. All parameters used were assigned a probability distribution to take into account the uncertainty in all random quantities. The simulation of 1000 times could provide a range of possible values capturing the specified probability distribution of parameters used in the analysis.

The base case scenario was done in patients aged 63 years old which was the mean age of patients who experienced MI in the Thai ACS Registry. The total life years obtained were 17.82 years for the supplementation of n-3 PUFAs and 15.48 years for no supplementation. The QALYs gained were 15.06 and 13.05 years for n-3 PUFAs supplementation and no supplementation, respectively. Thus incremental effectiveness values for supplementation of n-3 PUFAs were 2.34 life-years gained (LYG) and 2.01 QALYs. The lifetime cost from 1,000 simulations was estimated as 2,600,349 THB in the n-3 PUFAs supplementation cohort, and 2,001,572 THB in the no supplementation group, resulting in the incremental cost of 598,777 THB for the addition of n-3 PUFAs. The incremental cost-effectiveness ratio (ICER) was 297,193 THB per QALY and 256,199 THB per LYG (Table 3). ICER/QALY and ICER/LYG reduced when the patients had MI at the older age, indicating that the addition of n-3 PUFAs is more cost-effective in elderly patients than in younger patients (Figure 3). The probability of cost-effectiveness at different levels of willingness-to-pay in different ages of the patients is shown in Figure 4.

Figure 3

Figure 4

Table 3

Lifetime costs, quality-adjusted life years, life-year gained, and incremental costeffectiveness ratio of adding n-3 polyunsaturated fatty acids compared to standard therapy in post-MI patients