Simulation model design

We constructed a microsimulation decision tree analytic model to estimate the incidence of ASCVD events, health-care costs, and quality-adjusted life years (QALYs) over a 10-year time horizon. We defined ASCVD as acute coronary events (i.e., myocardial infarction, cardiac arrest, and coronary death (International Classification of Diseases, 10th Revision [ICD-10] codes: I20–I25 and I46) and ischemic stroke (ICD-10 codes: I63 and I69.3). Costs were estimated from both a societal perspective (including direct and indirect costs) and from a health-care sector perspective (including only direct costs). Future costs and QALYs were discounted 3% annually by applying an average weight of 0.875 for cost and QALYs for the 10-year simulation in the decision tree model [11].

As shown in the model diagram (Figure 1), people without a history of ASCVD are at risk of developing ASCVD over the 10-year simulation, conditional on exposure to ASCVD risk factors. Details concerning the model structure can be found in the Supplementary Appendix and Online Figures 1 and 2. In the model, total non-ASCVD and ASCVD health-care costs and indirect costs were calculated for each patient by summing the product of the duration of time spent in each health state (i.e., no ASCVD, acute ASCVD, and chronic ASCVD) and the corresponding annual cost. Similarly, total QALYs were calculated by summing the product of the duration of each health state by the corresponding utility (i.e., quality of life weights ranging from zero, death, to one, perfect health). The incremental cost-effectiveness ratio (ICER) was calculated by dividing the change in mean costs by the change in mean QALYs.

Persons without ASCVD at baseline enter the model and are then eligible to experience one of several outcomes: no ASCVD event, fatal or non-fatal incident ASCVD, or non-ASCVD death (Online Figure 1). Every person aged 35–64 years old at baseline could survive for up to 10 years. Each person was first assigned a 10-year risk of ASCVD calculated using the Chinese 10-year ASCVD risk assessment equation, which was developed based on the China Multi-provincial Cohort Study (CMCS) data, conditional on their base year risk factors [15]. Then, predicted ASCVD events were categorized into either first acute coronary event or first ischemic stroke event based on the sex and age group-specific allocation proportions of these two types of disease observed in the CMCS. ASCVD fatalities could occur either within or after 30 days, and in or out of hospital. Both non-fatal and fatal hospitalized ASCVD patients were assumed to receive standard hospital treatments. Patients with non-fatal ASCVD additionally received outpatient secondary prevention treatment including statin treatment in the base case. Hospital and outpatient secondary prevention treatments were held constant across primary prevention scenarios we evaluated. We assumed people could have a maximum of two ASCVD events during 10-year simulation period. A proportion of individuals taking statins discontinued treatment due to statin intolerance or statin-related adverse events (Table 1) [16].

Simulated treatment strategies

We explored the effect and cost-effectiveness of two doses of statins treatment implemented for eligible people by ASCVD risk level for China’s ASCVD primary prevention (Table 2). One was low-dose statins (i.e., simvastatin 20 mg/day, atorvastatin 10 mg/day, or rosuvastatin 5 mg/day) which lowered LDL-C by 30.0% (Table 1) [17, 23]. The second was ‘moderate-dose’ statins, that is, double the dose of statins used in the first scenario, which decreased LDL-C by 49.0%. The two dose levels corresponded to Chinese and international statin guideline recommendations [3, 17].

In the statin treatment arms, the risk of ASCVD events was decreased per mmol/L of LDL-C lowering using disease-specific β coefficients based on the meta-analysis from Cholesterol Treatment Trialists’ Collaboration (Table 1) [18]. In model simulations, statin treatment was also associated with a risk of medication-related adverse events, including diabetes, rhabdomyolysis, and myopathy, which incurred costs and utility decrements. Other potential statin adverse effects, including elevated liver enzymes, were assumed not to be clinically meaningful but could result in medicine discontinuation (Supplementary Appendix). We assumed those who discontinued statin treatment did so in the middle of the simulation time horizon, and no longer had a reduction in LDL-C or ASCVD risk.

Simulation cohort

The simulation cohort was derived from the CMCS study, a nationwide, multi-center cohort study on the determinants of ASCVD incidence in China. Participants free of cardiovascular disease were enrolled at study baseline from 1992 to 1996 [24]. Informed consent was obtained from all participants. The baseline LDL-C levels and other risk factors of the 21,265 individual Chinese participants aged 35–64 years in the CMCS study were imported into the model to calculate the individual 10-year risk of ASCVD. The individual risk formed the incidence number of ASCVD and was also a criterion for selecting participants eligible for statin intervention under contemporary guidelines [3]. Details of the CMCS cohort can be found in the Supplementary Appendix (Online Table 1). This study was approved by the ethics committee of Beijing Anzhen Hospital, Capital Medical University (Approval Number: KS2016005).

Risk stratification

High ASCVD risk was defined according to the recommendations of the Chinese Guidelines for the Management of Dyslipidemia in Adults as follows: 1) LDL-C ≥ 4.9 mmol/L or total cholesterol (TC) ≥ 7.2 mmol/L; 2) diabetes and LDL-C in the range of 1.8–4.8 mmol/L (or TC 3.1–7.2 mmol/L) and age ≥ 40 years; 3) 10-year ASCVD risk ≥ 10%. Otherwise, ASCVD risk was based only on a 10-year ASCVD risk score: 5%–9.9% was considered moderate risk and <5% was considered low risk [3].

Costs

Health-care and ASCVD treatment costs were derived from the Fifth National Health Services Survey [19], China’s Health and Family Planning Statistical Yearbook [20], and other related studies (Table 1) [21, 22]. For new ASCVD events, we included non-ASCVD health expenditures, ambulance costs, hospitalization cost, post-event outpatient ASCVD management costs, and indirect costs (Online Figure 2). Indirect costs, which estimated the financial losses in current and future years due to disability and death from ASCVD for the patients and their caregivers, included patient-time costs, unpaid caregiver-time costs, transportation costs, labor market earnings lost due to disease (Online Table 2) [25]. The average wage used in calculating indirect costs was based on wages of people working in non-private and private institute in 31 provinces (autonomous regions, municipalities) of China. More details concerning the indirect cost estimation method are provided in Online Table 3.

For statin treatment, we also included medication costs, routine lipid panel test costs, the cost of ASCVD risk assessment, medication-related adverse event monitoring, and adverse event treatment, if such events occurred. The yearly cost of statins was calculated by multiplying days in a year by the dose per day (mg/day), and the average cost per mg across the types of statins available in China (Int$/mg).

Prices for statins prior to implementation of the new centralized medicine procurement policy were obtained from the website of the Integrated Management Platform of Beijing Medicine Sunshine Purchase, which lists the prices of medicine in hospitals in Beijing prior to implementation of the new policy [13]. The lower statin prices that could be achieved after the implementation of the new policy in China were taken from the national centralized procurement policy bid-winning announcement file [14]. Annual statin treatment costs are listed in Table 1. All costs were inflated to 2019 using the rate of inflation in China published by Trading Economics and then converted into international dollars (Int$) according to the purchasing power parity exchange rate published by the Organization for Economic Co-Operation and Development (1 Int$ = 3.539 Chinese Yuan (RMB) in 2019; accessed 2/6/2020) [26, 27]. Purchasing power parity exchange rates are based on each country’s prices for a package of standard goods and, unlike currency exchange rates, are not subject to central bank policy effects.

Utilities

Every health state was assigned a utility which was used to estimate the benefits and harms of statin use on health in the model (Online Table 4 and Supplementary Appendix), including utilities for chronic and acute stage of ASCVD and adverse events [6, 7, 28, 29].

Model calibration and validation

The model was calibrated to achieve the best fit (minimal residual difference) between the simulation model’s predicted rate and the observed rate, including the person-year incidence of ASCVD and non-ASCVD mortality. Details were shown in the Supplementary Appendix. After calibration, the ratios of the simulated to observed rates were all close to one (Online Figure 3).

Analysis

We compared the simulated incidence of ASCVD with no treatment to each statin treatment arm (i.e., the low- and moderate-dose statins) and calculated the number needed to treat to prevent one ASCVD event in 10 years (NNT₁₀) and the QALYs gained. We used the projected incremental costs and QALYs to calculate the ICER, defined as the incremental cost to gain one additional QALY. In this study, the treatment effect in terms of preventing ASCVD and the budgetary impact of the statins price change stemming from the new policy were estimated on the basis that all eligible people aged 35–64 years old in China received statins for primary prevention. Details are shown in the Supplementary Appendix.

We used willingness-to-pay thresholds recommended by the World Health Organization’s CHOosing Interventions that are Cost Effective (WHO-CHOICE) initiative to determine the cost-effectiveness of statins in China. WHO-CHOICE defines an ICER that is less than the gross domestic product (GDP) per capita as highly cost-effective; an ICER that is 1–3 times GDP per capita as moderately cost-effective; and an ICER more than three times GDP per capita as not cost-effective. The GDP per capita of China in 2018 was Int$ 18,266 (64,644 Chinese Yuan) [30].

One-way sensitivity analyses explored the impact of statin prices, utilities, and adverse event incidence rates on the cost-effectiveness estimates. Monte Carlo probabilistic sensitivity analyses randomly sampled the uncertainty distributions of three key sets of inputs simultaneously: the relative risk of ASCVD events per mmol/L LDL-C change, all direct and indirect cost inputs, and the incidence rate of three types of adverse events. In the analysis, the distributions of these inputs were randomly sampled 1,000 times, resulting in 1,000 pairs of total costs and QALYs that were used to calculate the proportion of ICERs falling under our predetermined willingness-to-pay threshold.

The China statin decision ASCVD microsimulation model was developed in SAS software (version 9.4, SAS Institute, Cary, NC, USA). The SAS CALL STREAMINIT function was used to generate log normal distributions of treatment effects and adverse event parameters for probabilistic sensitivity analysis. The SAS MCMC function was used to create gamma distributions for costs. This study was designed and reported according to the CHEERS statement [31].

ASCVD incidence

The 10-year cumulative incidence rates of ASCVD among low, moderate, and high risk people were 1.8%, 6.9%, and 10.7%, respectively, in CMCS cohort population. Compared with no statin treatment, the low-dose statin treatment resulted in relative reductions of the ASCVD incidence by 4.1% (95% CI: 3.2%–5.2%), 6.4% (95% CI: 5.1%–8.3%), and 15.5% (95% CI: 12.2%–20.2%) among people with baseline low, moderate, and high 10-year ASCVD risk, respectively (Table 3 and Online Table 5). These reductions could be translated into reductions in 10-year ASCVD incidence among the low, moderate, and high risk populations in China of 323,000 (95% CI: 253,000–413,000), 399,000 (95% CI: 317,000–521,000), and 1,350,000 (95% CI: 1,083,000–1,748,000), respectively (Table 4). With the moderate-dose statin treatment, a larger relative reduction in ASCVD incidence would be achieved among all ASCVD risk groups.

Number needed to treat for ten years to prevent one ACVD event

When eligible people were treated with statins for primary prevention, the NNT₁₀ decreased with the increase of ASCVD risk. The NNT₁₀ was 193.0 (95%CI: 150.6–246.0), 67.3 (95%CI: 51.7–84.4), and 42.8 (95%CI: 33–53.4) for treating low to high ASCVD risk populations, respectively (Table 3). When switching low-dose to moderate-dose statins treatment, the NNT₁₀ would decrease significantly in every risk groups.

Costs

When the lower prices of statins resulting from the new centralized medicine procurement policy were used in the model, statins treatment-related costs decrease greatly compared with the pre-policy prices (Table 3 and Online Table 6). Costs for treating high ASCVD risk people with low-dose statins could decrease from cost Int$ 3,000 (95% CI: 1,500–5,300) to Int$ 1,100 (95% CI: 900–1,400) per person over 10 years from a health-care sector perspective. If eligible people aged 35–64 year old in China were all treated for 10 years for ASCVD primary prevention under the new policy, it could save 126.4 (95% CI: 35.7–244.1), 53.6 (95% CI: 16.5–110.3), and 114.0 (95% CI: 35.7–232.2) billion Int$ when treating low to high ASCVD risk populations. The results show that approximately a quarter of primary prevention costs went on statins drug costs when treating with lose-dose statins (Figure 2). The costs related to moderate-dose statins treatment were shown in Table 3 and Online Table 6.

Cost-effectiveness

From the health-care sector perspective, both low- and moderate-dose statin treatments were cost-effective for high ASCVD risk people when statin prices were derived from the new centralized medicine procurement policy (Figure 3, Table 3, and Online Figure 4). Monte Carlo probabilistic sensitivity analyses show treatment of high ASCVD risk people with either low- or moderate-dose statins was cost-effective in 64.9% and 90.8% of probabilistic simulations, respectively, from the health-care sector perspective. The cost-effectiveness acceptability curves are shown in Figure 4. The ICERs and probabilities of being cost-effective would be more favorable in the analysis from the societal perspective because of the saving of indirect costs (Figure 3 and Online Table 5). By contrast, for low ASCVD risk people, neither low- nor moderate-dose statins treatments were cost-effective, even when statin prices were based on the new policy (Table 3). Assuming the higher prices in place before the 2019 pricing policy, statins treatment for primary prevention would not meet standard cost-effectiveness thresholds regardless of risk level, choice of therapy, or inclusion of indirect costs (Figure 3 and Online Table 5).

One-way sensitivity analysis

One-way sensitivity analysis results are presented in Online Table 7. When the lowest price of statins arising from the new policy was used, treating moderate risk people with moderate-dose statins could be cost-effective from the societal perspective (48,800 Int$/QALY). One-way sensitivity analyses show that the ICER was sensitive to changes to the incidence rate and the utility of statin-induced diabetes. Among low ASCVD risk people, the incidence of statin-induced diabetes might exceed the number of prevented ASCVD cases when low- or moderate-dose statin treatments were used (Table 4 and Online Figure 5).

Limitations

The current study was constrained by the baseline age range of the CMCS study and therefore did not include people older than 64 years old. Older individuals comprise a growing portion of the Chinese population and future research should strive to include them. CMCS baseline cholesterol and other risk factor levels observed in 1990s and may be lower than the current levels representative of Chinese adults. The CMCS cohort may not be representative of contemporary Chinese adult population. Comparing to the risk factor level in 2012 nation-wide survey and published in the Report on Chinese Residents’ Chronic Diseases and Nutrition [36], the average level of diastolic blood pressure, TC, and fasting blood glucose of the CMCS population used in this study were similar to those in the national survey (Online Table 9), but the average systolic blood pressure was lower and HDL-C was higher than those in the national survey. Past randomized, controlled clinical trials that established the effect of statins on lowering ASCVD risk did not enroll many Chinese participants. We applied the average effects estimated in clinical trials meta-analyses of these trials, which may not exactly be generalizable to China [18]. Future research should explore the effect of statin use on LDL-C and ASCVD risk in more diverse populations, including Chinese adults.

Higher statin prices reflecting typical prices seen before the 2019 centralized procurement policy were obtained from the website that indicates statin prices for hospitals located in Beijing [13]. The national representativeness of these statin prices is not known, as the Chinese government has continued trying to lower prices and every province has its own project and bidding system [37]. To calculate indirect costs, we used an average wage estimated based on statistical data from 31 provinces (autonomous regions, municipalities) of China for the simulation population, meaning that our results do not reflect the effect of income diversity on the results.

Conclusion

By lowering statin prices substantially, China’s new centralized medicine procurement policy will improve the cost-effectiveness of statins for primary ASCVD prevention in adults aged 35–64 years. After 10 years of treatment, new lower prices make statin treatment cost-effective among high ASCVD risk people from both a health-care sector and a societal perspective.