The cost‐effectiveness of coronary calcium score‐guided statin therapy initiation for Australians with family histories of premature coronary artery disease

The 2012 Australian National Vascular Disease Prevention Alliance cardiovascular disease primary prevention guidelines1 differ markedly from overseas guidelines in several respects. The American College of Cardiology/American Heart Association (ACC/AHA) guidelines2 assess risk with the pooled cohort equation (PCE), predicting hard cardiovascular outcomes only (death, myocardial infarction, stroke, resuscitated cardiac arrest), and may be less prone to overestimating risk than the Australian absolute cardiovascular disease risk (ACVDR) tool. Second, the current threshold for initiating statin therapy in Australia is ACVDR 5‐year CVD risk of 10%;1 the ACC/AHA guidelines recommend statin therapy for people with 10‐year PCE risk of 7.5%.2 Finally, the ACC/AHA guidelines, but not the Australian guidelines, recognise the value of coronary artery calcium (CAC) scores for guiding treatment of people at intermediate risk of cardiovascular events.1,2

CAC scores are correlated with atherosclerotic plaque burden and risk of cardiovascular disease (CVD) and, together with standard risk factors, can improve risk stratification; a consensus is emerging that statin therapy is warranted for people with CAC scores of 100 or more.3,4 Moreover, by identifying subclinical disease and the potential benefit of reducing low density lipoprotein cholesterol (LDL‐C) levels, CAC scoring may motivate clinicians to initiate and patients to adhere to statin therapy.5,6

Recent position statements by the Cardiac Society of Australia and New Zealand (CSANZ)7 and the National Heart Foundation (NHF)8 reflect CAC recommendations in overseas guidelines, but note that local information on outcomes and cost‐effectiveness is limited. A United States study found that statin treatment guided by CAC scores may be more cost‐effective than standard predicted risk thresholds, and that adherence to therapy is an important contributor to its benefit.9

We therefore evaluated the cost‐effectiveness of CAC‐guided and risk score‐based statin treatment thresholds by analysing treatment outcomes for Australians with family histories of premature coronary artery disease.

Methods

We analysed data from the Coronary Artery calcium score: Use to Guide management of HerediTary Coronary Artery Disease (CAUGHT‐CAD) trial10 (Australian New Zealand Clinical Trials Registry ACTRN12614001294640). This randomised controlled trial assessed the utility of CAC scoring for guiding risk evaluation and statin therapy as primary prevention in people with family histories of early onset coronary artery disease. In brief, CAC scores were calculated for 1083 men and women aged 40–70 years at low to intermediate CVD risk (5‐year ACVDR: 2–15%), but free of clinical cardiovascular disease at baseline (Supporting Information, table 1). We compared the cost‐effectiveness of following the Australian guidelines for statin therapy (ACVDR 5‐year risk ≥10%) with three sets of less restrictive eligibility criteria:

• people with ACVDR 5‐year risk ≥2% and CAC scores >0;
• people with ACVDR 5‐year risk ≥2% and CAC scores ≥100;
• people with 10‐year PCE risk ≥7.5% (ACC/AHA guidelines2).

Economic model

We evaluated economic and clinical outcomes over fifteen years from the perspective of the Australian health care system (Medicare, Pharmaceutical Benefits Scheme, public hospitals) in a Markov microsimulation state transition model (TreeAge Healthcare Pro 2018) (Box 1). In the model, all people were initially symptom‐free, receiving or not receiving statin therapy according to eligibility and the probability of statins being prescribed. Each year, statin therapy could be maintained or discontinued, or the person could experience a relevant event, including death from non‐cardiovascular causes. In the case of a non‐fatal cardiovascular event, the person was deemed to have entered a symptomatic, secondary prevention state. Health care costs and changes to health status, measured as quality‐adjusted life years (QALYs), were associated with time in each state and the transitions between states (Supporting Information, table 2).

Model assumptions and outcomes

The annual risk of coronary events was determined on the basis of the 10‐year Multi‐Ethnic Study of Atherosclerosis (MESA) risk,11 that of stroke events on the basis of 10‐year Framingham Stroke Risk12 (Supporting Information,

table 2). Numbers of deaths from non‐cardiovascular causes were derived from age and sex‐dependent mortality in the General Record of Incidence of Mortality dataset for 201713 (Supporting Information, table 3).

Based on the findings of the Swedish VIPVIZA evaluation of changes in CVD risk after ultrasound assessment of asymptomatic carotid plaque,6 our modelling of CAC‐guided selection assumed higher rates of statin therapy initiation, greater LDL‐C level reduction by statin treatment (1.0mmol/L with CAC screening [relative risk, 0.76]; 0.5mmol/L without CAC screening [relative risk, 0.88]), and lower rates of statin therapy discontinuation than for non‐CAC‐guided strategies (Australian1 and American guidelines2). Our assumptions regarding statin initiation (90% of eligible patients with CAC‐guided selection, 70% without CAC‐guided selection) and persistence were based on data from the CAUGHT‐CAD study14 and meta‐analysis of trials examining the influence of CAC scoring on primary prevention therapy.5

Assumptions about health status changes were based on the findings of Australian and American studies (Supporting Information, table 2). Statin therapy‐related harm (disutility) was deemed equivalent to losing two weeks of health over ten years (0.0031 QALYs). Total CAC screening costs included the computed tomography scan ($198) and costs related to adverse outcomes, incidental findings, and risk of cardiac stress testing. Treatment costs were based on the National Hospital Cost Data, and economic evaluations15,16 and the Medicare Benefits Schedule (MBS; http://www.mbsonline.gov.au). Medication costs were based on the Pharmaceutical Benefits Scheme (PBS; https://www.pbs.gov.au) dispensed price for the maximum quantity of atorvastatin (40mg), adjusted for lower treatment adherence associated with eligibility criteria not taking CAC scores into account. All costs were deflated to 2019 prices according to the Australian Institute of Health and Welfare Total Health Price Index17 and the Australian Bureau of Statistics consumer price index for health goods.18 Costs and outcomes were discounted by 3% per annum.

Statistical analysis

The statistical significance of differences between strategies or subgroups in baseline characteristics, costs, and QALYs were assessed in t tests (continuous variables) or χ² tests (categorical variables) without adjustment for multiple comparisons. Incremental cost‐effectiveness ratios (ICERs) were calculated by dividing mean incremental cost differences by incremental effectiveness (QALYs gained). The 95% confidence interval (CI) for the ICER of a CAC‐guided strategy compared with standard care was derived from 1000 microsimulations. Cost‐effectiveness was defined by standard and strict (highly cost‐effective) willingness‐to‐pay thresholds of $50000 and $35000 per QALY gained respectively.19 Subgroup analyses assessed cost‐effectiveness by age, sex, socio‐economic status, and LDL‐C level.

We assessed the impact on cost‐effectiveness of varying individual parameter values (deterministic sensitivity analyses), including statin therapy initiation, statin adherence, time horizon, global discount rate (0%, 5%), and the costs and utility consequences for CVD and non‐CVD events, including death. Multi‐parameter probabilistic sensitivity analyses with 500 second‐order Monte Carlo simulations incorporated almost all model parameters (value distributions: Supporting Information, table 2). Probabilistic sensitivity analyses were used to derive confidence intervals and evaluate the uncertainty of our ICER estimates. Probabilistic sensitivity analyses randomly select parameter values according to their probability distributions; the aim is to assess whether a strategy is likely to be cost‐effective in practice, where parameter values may differ from those in our base model.

To assess the economic benefits of treatment at lower risk thresholds than recommended in Australian guidelines when risk is more accurately predicted, we also assessed the relative cost‐effectiveness of initiating statin therapy for people with 10‐year MESA risk of 7.5% or more (equivalent to a CVD event rate of 10% or more).

Ethics approval

Our study was approved by the Tasmanian Human Research Ethics Committee (14‐281) and by all participating institutions. All participants provided written informed consent for inclusion in the CAUGHT‐CAD trial.

Results

The baseline characteristics of the 1083 CAUGHT‐CAD participants included in our analysis are summarised in Box 2, their 10‐year predicted MESA risk by treatment strategy and statin therapy eligibility criteria in the Supporting Information, table 4. According to Australian guidelines,1 statin therapy would be recommended for 77 participants (7.1%). With ACVDR 5‐year risk ≥2% and CAC score>0 as the criteria, 496 people would be eligible (46%); with ACVDR 5‐year risk ≥2% and CAC score≥100, 155 would be eligible (14%). According to ACC/AHA guidelines,2 266 people were eligible for statin therapy (25%) (Supporting Information, table 4).

Costs and cost‐effectiveness

Compared with applying the standard Australian eligibility criteria,1 CAC‐guided strategies increased the mean cost to the health care system by $493 (95% CI, $466–521) per person with CAC score>0 as the criterion and by $289 (95% CI, $270–308) with CAC score≥100 as the criterion. With CAC score>0 as the criterion, 267 deaths and 795 symptomatic cardiovascular events were averted, achieving a mean increase of 0.0093 (95% CI, 0.0074–0.0112) QALYs per person at a mean ICER of $53028 per QALY gained. With CAC score≥100 as the criterion, 163 deaths and 330 symptomatic cardiovascular events were averted, achieving a mean increase of 0.0087 QALYs per person at an ICER of $33108 per QALY gained. The mean cost of applying the American guidelines2 was also higher than applying the Australian guidelines (+$92 per person; 95% CI, $81–103 per person), but did not increase utility. With a 10‐year MESA CVD risk threshold of 7.5% as the criterion for statin therapy, the number of eligible people was 209 (19.7%), with an estimated ICER of $37919 per QALY gained (Box 3).

Optimal risk threshold for coronary artery calcium scanning

The cost‐effectiveness of CAC score‐guided selection relative to Australian or American guidelines increased with baseline risk (Box 4; Supporting Information, tables 5 and 6). A strategy based on CAC scores greater than 0 was cost‐effective and dominated the current Australian strategy (ie, was cost‐effective and was less expensive) when the baseline 5‐year ACVDR risk was 10% or more, but was not cost‐effective if the baseline risk was less than 8% (Supporting Information, table 5). A strategy based on CAC scores ≥100 was cost‐effective if the baseline 5‐year ACVDR risk was at least 5% (Supporting Information, table 7).

Subgroup, sensitivity, and uncertainty analyses

CAC‐guided selection (CAC score>0) was cost‐effective for men, and for people over 60 years of age, with LDL‐C levels exceeding 4.0mmol/L, or from lower socio‐economic status areas (Box 5). The American strategy was more cost‐effective than the Australian approach for people with LDL‐C levels of 4.1mmol/L or more (Supporting Information, table 8).

Reducing CAC screening costs by $1 reduced the ICER for CAC‐guided selection (CAC score>0) compared with the standard Australian approach by $99 per QALY gained. Disregarding statin therapy disutility reduced the ICER (by $24832 per QALY gained); if disutility was deemed equivalent to losing four weeks of health over ten years (ie, twice the base estimate), the CAC‐guided strategy (CAC score>0) was no longer cost‐effective (mean change in ICER, $412251 per QALY gained) (Supporting Information, table 9). The costs and utility values associated with acute symptomatic CVD events had small effects on the cost‐effectiveness of CAC‐guided selection (CAC score>0) (Supporting Information, table 10).

Deterministic sensitivity analyses

Changes in relative risk reduction attributable to statin therapy for the Australian strategy (0.75–0.95, as a marker of statin adherence) had a smaller influence on the ICER for the CAC‐guided strategy (CAC score>0) than changes in relative risk reduction attributable to statin therapy with the CAC‐guided strategy (Supporting Information, table 11). With the Australian strategy, statin initiation in at least 80% of eligible people was required to render the CAC‐guided strategy (CAC score>0) highly cost‐effective (ICER < $35000/QALY gained) even at low (20%) initiation rates (Supporting Information, figure 1). This increased to >90% when the statin initiation rate in the Australian strategy was 80%.

Applying a shorter time horizon (ten years) increased the ICER for the CAC‐guided strategy by $133367 per QALY gained, a higher global discount rate (5%) by $15410 per QALY gained (Supporting Information, table 11).

Probabilistic sensitivity analysis

The mean cost for the CAC‐guided strategy (CAC score>0) was $462 per person (95% CI, $455–468) higher than for current Australian practice but was more effective (+0.0085 [95% CI, +0.0081–0.0090] QALY gained per person), yielding a mean ICER of $54055 (95% CI, $22 847–203834) per QALY gained. The mean incremental cost of the American strategy was $129 (95% CI, $126–132) higher than current Australian practice, but was not more effective (mean difference in utility, –0.0001 QALYs gained per person; 95% CI, –0.0002 to 0.0000 QALYs gained per person). Applying a willingness‐to‐pay threshold of $50000, the CAC score‐guided approach was cost‐effective in 50.5% of model iterations; applying a threshold of $35000, it was cost‐effective in 17.8% of iterations (Supporting Information, figure 2).

Discussion

Our major findings were that CAC‐guided statin treatment selection, but not simply reducing the risk threshold for commencing statin therapy (American guidelines), can be cost‐effective compared with current Australian recommended practice; that the quality of risk assessment and adherence to statin therapy influenced cost‐effectiveness; that CAC‐guided selection was more cost‐effective with a CAC score threshold of 100 than with a threshold of 0; and that CAC‐guided selection was not cost‐effective when baseline ACVDR risk was less than 5% (CAC ≥100) or 8% (CAC>0).

Initiating statin therapy for people with family histories of early onset coronary artery disease and CAC scores greater than 0 was more expensive but more cost‐effective than applying Australian guideline recommendations.1 However, at a willingness‐to‐pay threshold of $50000 per QALY gained, the probability that it was not cost‐effective for an individual patient was almost 50%. The ICER could be reduced by screening only people at higher baseline CVD risk: those aged 60 years or more (217 of 356 [61%] had positive CAC scans; Supporting Information, table 12) or with 5‐year ACVDR risk of at least 6%, men, and people from lower socio‐economic status areas. Our findings also support CSANZ7 and NHF8 statements that a CAC score of 100 is a suitable eligibility threshold for statin treatment, as is an ACVDR threshold of 5% for CAC assessment.7 Nevertheless, traditional risk factors retain their independent predictive value, and statin treatment is appropriate for people with low CAC scores but at higher baseline risk.

PCE scores seem better correlated with subclinical disease than the ACVDR,21 and the 10‐year MESA risk (which takes both CAC scores and traditional risk factors into account) outperforms the ACVDR with respect to absolute risk prediction and reclassification.4,12 Optimal risk reclassification and not treating people with low CVD risk is important for the cost‐effectiveness of CAC score‐guided selection. Lower treatment thresholds (American guidelines) increase the number of people treated but do not increase mean benefit, probably because of statin disutility and the low value of extending treatment to people with low CAC scores. In contrast, extending treatment to people with an equivalent MESA 10‐year risk of 7.5%, or to anyone with non‐zero CAC scores (about 45% of the study sample), were each more cost‐effective than the American guideline recommendations when compared with the Australian guidelines. Lower treatment thresholds may therefore be economically justifiable if more accurate risk assessment tools are used.

Moreover, if we assumed that statin therapy reduces CVD risk by 25% (equivalent to LDL‐C reduction of 1.0mmol/L) for all selection strategies, CAC score‐guided selection was not markedly more cost‐effective than the Australian approach. Our findings that lower treatment thresholds alone may not be cost‐effective are consistent with those of studies in similar health care systems (eg, the United Kingdom), where acute health care costs are lower than in the United States.22 The limits of CAC score‐guided reclassification of risk was evident for people with baseline risk ACVDR of 5% or lower, for whom the proportion of positive scans is small and 10‐year CVD risk can be low, despite non‐zero CAC scores.3

Assumptions about statin initiation, adherence, and persistence influenced the estimated cost‐effectiveness of CAC score‐guided selection, and therefore the generalisability of our findings to all people with family histories of early onset coronary artery disease. Our model assumed an initiation rate of 90% for the CAC score‐based strategy, but in one prospective intervention study only 77% of people with CAC scores of 100 or more subsequently commenced cholesterol‐ or blood pressure‐lowering therapy.23 We found that a statin initiation rate of at least 90% was required for the CAC‐guided strategy (CAC score>0) to be highly cost‐effective. This is optimistic; reported initiation rates in the community are generally no higher than 50% and may be as low as 15%.14,23

Model values for absolute LDL‐C reduction with the CAC score‐based strategy (1.0mmol/L) and for the difference between CAC score‐ and non‐CAC‐guided strategies (about 0.5mmol/L) were derived from the VIPVIZA study.6,24 Greater reductions in LDL‐C when subclinical atherosclerosis is assessed have been reported,24,25 providing robust evidence that taking it into account can help reduce CVD risk to a greater extent than standard care, perhaps because of improved medication adherence and lifestyle changes, or more aggressive management of risk factors.5,24 An important assumption of our model was that LDL‐C reduction would be greater with the CAC‐guided strategy over fifteen years. CVD risk was reduced for three years in the VIPVIZA study, but whether this benefit is sustained in the longer term has not been established.26

Statin therapy discontinuation (assumed rates: CAC‐guided selection: 35%; Australian and American guideline‐guided selection: 45%) moderately influenced cost‐effectiveness. The persistence of lipid level reduction in people who underwent CAC scanning range between 69% and 90%,5 but one meta‐analysis found that long term statin therapy persistence in the general community was lower than 60%.27 Our findings suggest that CAC screening may play a role beyond reclassifying risk in people at intermediate risk of CVD. A CAC‐guided strategy for people with higher baseline CVD risk was less expensive than the Australian guidelines, probably because of greater on‐treatment LDL‐C reduction and treatment persistence that result in lower treatment costs for symptomatic CVD. Recommendations for statin therapy for this group do not markedly differ by strategy, as most participants with high baseline risk had positive CAC scores and were eligible for statin therapy using any strategy.

Limitations

Our study included people with family histories of premature coronary artery disease, but generally at low CVD risk (low rates of diabetes and smoking); the role for CAC assessment in de‐escalating therapy in people at high risk is, in any case, limited. Second, MESA and PCE risk algorithms have not been validated in an Australian population; systematic overestimation of 10‐year risk using MESA would bias our results in favour of CAC screening. We did not model improvements in other manifestations of atherosclerosis, such as peripheral vascular disease, nor did we model the effect of statin selection strategy on anti‐hypertensive treatment, as guideline‐recommended pharmacotherapy does not take CAC into account.28 However, identifying CAC may also encourage better blood pressure control (not modelled), which would increase the cost‐effectiveness of CAC‐guided selection.24 Third, we did not assess indirect costs (eg, productivity losses), perhaps leading to underestimation of the cost‐effectiveness of CAC‐guided selection in younger people. Fourth, we did not model statin treatment after risk re‐assessment (with any selection strategy) over 15 years. Finally, values for the probability of specialist referral and stress testing by strategy were based on the authors’ opinions because relevant publicly available data are not available; varying the values in sensitivity analyses yielded similar results to the main analysis. We did not include invasive coronary angiography referrals in our model because the participants did not have symptomatic CVD; further, CAC screening does not increase downstream medical testing costs in the medium term.25

Conclusion

A CAC‐guided strategy for selecting people with family histories of premature coronary artery disease for statin therapy (CAC score>0) was more cost‐effective than selection according to Australian CVD prevention guidelines if treatment adherence rates are taken into account. Systematic CAC assessment was cost‐effective if statin therapy was initiated in people with baseline ACVDR risk of at least 5% and CAC scores of 100 or more. Incorporating non‐traditional risk factors and refining criteria for statin treatment should be considered when revising primary CVD prevention guidelines in Australia.

Box 1 – The Markov microsimulation state transition model used in our analysis*

---

* For further details, see Supporting Information, table 2.

Box 4 – Incremental cost and incremental effectiveness of the coronary artery calcium score‐guided statin therapy eligibility (compared with Australian guidelines), by baseline 5‐year absolute cardiovascular risk (ACVDR)

---

Points indicate 5‐year ACVDR; dashed line indicates willingness‐to‐pay threshold of $50000 per quality‐adjusted life year (QALY) gained.