Long-term cardiovascular risks and the impact of statin treatment on socioeconomic inequalities: a microsimulation model

Background UK cardiovascular disease (CVD) incidence and mortality have declined in recent decades but socioeconomic inequalities persist. Aim To present a new CVD model, and project health outcomes and the impact of guideline-recommended statin treatment across quintiles of socioeconomic deprivation in the UK. Design and setting A lifetime microsimulation model was developed using 117 896 participants in 16 statin trials, 501 854 UK Biobank (UKB) participants, and quality-of-life data from national health surveys. Method A CVD microsimulation model was developed using risk equations for myocardial infarction, stroke, coronary revascularisation, cancer, and vascular and non-vascular death, estimated using trial data. The authors calibrated and further developed this model in the UKB cohort, including further characteristics and a diabetes risk equation, and validated the model in UKB and Whitehall II cohorts. The model was used to predict CVD incidence, life expectancy, quality-adjusted life years (QALYs), and the impact of UK guideline-recommended statin treatment across socioeconomic deprivation quintiles. Results Age, sex, socioeconomic deprivation, smoking, hypertension, diabetes, and cardiovascular events were key CVD risk determinants. Model-predicted event rates corresponded well to observed rates across participant categories. The model projected strong gradients in remaining life expectancy, with 4–5-year (5–8 QALYs) gaps between the least and most socioeconomically deprived quintiles. Guideline-recommended statin treatment was projected to increase QALYs, with larger gains in quintiles of higher deprivation. Conclusion The study demonstrated the potential of guideline-recommended statin treatment to reduce socioeconomic inequalities. This CVD model is a novel resource for individualised long-term projections of health outcomes of CVD treatments.


Introduction
Cardiovascular diseases (CVDs) are the leading cause of morbidity and mortality globally. 1In the UK, despite declines in age-standardised CVD mortality and morbidity over the last few decades, CVD remains a significant burden for the health system. 2 Regional and socioeconomic inequalities persist, with slower improvements in the most deprived areas 3,4 and increasing gaps in life expectancy at birth between the least and the most deprived quintiles from 5-7 to 6-8 years from 2001 to 2017. 3 Disease policy models use epidemiological evidence to model key disease stages and relationships, and project the health status of individuals and populations. 5Unlike CVD risk equations, such as QRISK or SCORE, 6,7 which estimate risk of a particular CVD outcome over a fixed period (typically 5 or 10 years), policy models can project long-term risks, accounting for competing risks, and enable estimation of long-term quality-of-life (QoL)-adjusted survival and costs, providing useful predictions for clinical and policy decision making.Policy models are often developed using summary data, 8,9 which do not allow reliable assessment of outcomes in distinct patients.1][12][13][14][15] These models, however, have often used older data, [13][14][15] limiting their usefulness for contemporary policy analysis.Although the use of newer data to recalibrate CVD risk equations has been shown to improve their performance in contemporary populations, 16,17 the authors of the current study are not aware of efforts to update policy models, which are structurally more complex, using newer data.

Aim
To present a new CVD model, and project health outcomes and the impact of guideline-recommended statin treatment across quintiles of socioeconomic deprivation in the UK.

Design and setting
A lifetime microsimulation model was developed using 117 896 participants in 16 statin trials, 501 854 UK Biobank (UKB) participants, and quality-of-life data from national health surveys.

Method
A CVD microsimulation model was developed using risk equations for myocardial infarction, stroke, coronary revascularisation, cancer, and vascular and non-vascular death, estimated using trial data.The authors calibrated and further developed this model in the UKB cohort, including further characteristics and a diabetes risk equation, and validated the model in UKB and Whitehall II cohorts.The model was used to predict CVD incidence, life expectancy, quality-adjusted life years (QALYs), and the impact of UK guideline-recommended statin treatment across socioeconomic deprivation quintiles.

Method
The CVD microsimulation policy model structure This CVD policy model is a microsimulation model of the progression of CVD and key competing events.The model (see schematic in Supplementary Figure S1) includes seven events, namely first occurrences of myocardial infarction (MI), stroke, coronary revascularisation (including percutaneous coronary intervention and coronary artery bypass grafting), incident cancer, incident diabetes, and vascular and non-vascular death.Parametric proportional hazards risk equations inform the risk of these events.The model inputs are individuals' characteristics of age; sex; ethnicity; body mass index (BMI); smoking status; blood pressure; lipids; haemoglobin A1C (HbA1c) and creatinine levels; previous CVD history; treated hypertension; diabetes and cancer history; mental illness; physical activity; diet quality; and socioeconomic deprivation (based on Townsend Score) (see details in Supplementary Table S1 and Supplementary Information S1).The model projects disease events and health-related QoL annually until death or 110 years of age using each individual's characteristics.In each annual cycle, the occurrences of events are simulated in a random order.Individuals' age and history of events are updated annually and inform the subsequent events' risks in a time-dependent manner.

Estimation and calibration of model risk equations
The risk equations underlying the CVD policy model were developed using Cholesterol Treatment Trialists' (CTT) Collaboration and UK Biobank (UKB) IPD.First, Cox proportional hazards risk equations were estimated for model endpoints except incident diabetes (unavailable in CTT Collaboration data provided for this study) using 117 896 individuals from 16 randomised statin trials from the CTT Collaboration. 18econd, data from the 501 854 participants aged 40-70 years, recruited in the UKB between 2006 and 2010 throughout the UK, and their follow-up information until 31 March 2017 19 was used to calibrate the risk equations and develop a de novo incident diabetes risk equation.Separate parametric risk equations were fitted for individuals without and with CVD history, with the exception of the incident diabetes and incident cancer equations that were fitted across all participants.Parameter uncertainty was assessed using bootstrapping. 20See Supplementary Information S2 for further details.

Model validation
The model was validated by comparing the model-simulated cumulative incidences of each endpoint with the observed cumulative incidences overall and in participant categories.This included validating the UKB-calibrated model in the UKB cohort and among Whitehall II participants (n = 6761; 10 years' follow-up; external validation) (see Supplementary Information S2 and S3).

Health-related QoL
Health-related QoL associated with participant characteristics and disease histories was estimated using a linear regression model of EuroQoL-5 Dimension (EQ-5D) utility using pooled 2006, 2011, and 2017 Health Survey for England (HSE) participant data.QoL utility ranges from −0.594 for the worst health state to 1 for full health where 0 is equivalent to death and values <0 represent health states worse than death in the EQ-5D-3L used in 2006 and 2011. 21The EQ-5D-5L, used in HSE 2017, was mapped into EQ-5D-3L 22 before pooling.The QoL model was integrated into the CVD model to annually predict individuals' QoL.

Model applications
The CVD microsimulation model was used to perform lifelong projections for all UKB participants using their baseline characteristics.The authors executed 500 microsimulations for each individual to minimise the Monte Carlo uncertainty, and probabilistic sensitivity analyses using 500 and 1000 bootstrap coefficient sets for individuals without and with CVD history at entry, respectively.The results were summarised in categories of UKB participants by CVD history, age, sex, and estimated 10-year CVD risk (QRISK 3 score). 6is model was used to assess the remaining life years, quality-adjusted life years (QALYs), and the effects of statin treatment, as recommended by the UK National Institute for Health and Care Excellence guidance, 8 across quintiles  23

Stakeholder involvement
The project was guided by multidisciplinary project management and steering groups, including primary care and specialist clinicians, lay people, trialists, statisticians, and health economists.Three lay people were involved as members of these groups, helping refine methodology and approaches to presenting study findings.

Summary of the CTT Collaboration and UKB data
In CTT Collaboration trials, 68 018 participants without and 49 878 with CVD history were followed over 3.9 and 4.6 years on average, respectively.In UKB, 444 576 participants without and 57 278 with CVD history were followed for 8.1 and 7.9 years on average, respectively.Their characteristics are summarised in Table 1.The details of contributing trials and the numbers of events during follow-up are summarised in Supplementary Tables S2 and S3.

Risk equations
MI was strongly associated with an increased risk of stroke, and MI and stroke were associated with an increased risk of vascular death, with increases greatest in the years of these events.Coronary revascularisation was associated with a reduced risk of vascular death.Duration since diabetes diagnosis was associated with increased risks of all cardiovascular events and, in those without diagnosed diabetes, higher HbA1c levels were associated with increased risks of all cardiovascular events.These patterns were similar between people without and with CVD history although magnitudes differed (Figure 1 and Supplementary Tables S4-S6).
Age, male sex, smoking, treated hypertension, unhealthy diet, and lower physical activity were strongly associated with an increased risk of cardiovascular events.Greater socioeconomic deprivation was associated with a higher risk of stroke, incident diabetes, and vascular and non-vascular death (Supplementary Tables S4-S6).

Model validation
In the internal validation of the model based on CTT Collaboration risk equations, the model-predicted cumulative incidence rates closely matched the observed rates (Supplementary Figure S2).External validation of the model based on CTT Collaboration risk equations in the UKB cohort indicated the need for calibration to improve the accuracy of predictions.After calibration, there was good correspondence between cumulative incidence rates predicted by the model and the observed rates in the UKB for all endpoints across follow-up years in categories of participants (Figure 2 and Supplementary Figures S3 and S4).The UKB-calibrated model demonstrated a good overall performance in the external Whitehall II cohort (Figure 2).
Effects of CVD events on QoL CVD was a key determinant of QoL.MI was associated with a decrement in EQ-5D utility of 0.10 (95% confidence interval [CI] = 0.03 to 0.16) in the year of event and 0.07 (95% CI = 0.04 to 0.10) in subsequent years.Stroke was associated with a decrement of 0.09 (95% CI = 0.04 to 0.13) in the year of event and 0.13 (95% CI = 0.11 to 0.16) in subsequent years.Diabetes was associated with a decrement of 0.04 (95% CI = 0.03 to 0.06) in the first 10 years from diagnosis and 0.08 (95% CI = 0.06 to 0.10) in subsequent years.Cancer affecting daily life was associated with a decrement in QoL of 0.13 (95% CI = 0.11 to 0.14) (Supplementary Table S7).For use in the CVD microsimulation model, the cancer-related utility decrement was revised to 0.03 informed by the literature [24][25][26] to reflect any cancer history.

Long-term projections of survival and QALYs
The differences in CVD risk between individuals of different age, sex, and cardiovascular risk were reflected in their model-predicted survival and QALYs (Figure 3).For individuals in the same age and sex category, shorter life expectancy and fewer QALYs were predicted for people with CVD history or higher 10-year CVD risk.Males had shorter life expectancy but more QALYs as a proportion of their life expectancy than females.The projected remaining life expectancy (from age at entry) ranged between 19.5 (95% CI = 18.7 to 20.4) and 38.9 (95% CI = 37. 3   S5).

Health outcomes across quintiles of socioeconomic deprivation
There were moderate gradients at 10 years in the predicted years of life and QALYs across the quintiles of socioeconomic deprivation (Figure 4).Over a longer duration, the gradients in the predicted survival and QALYs increased.Over a lifetime, individuals in the most socioeconomically deprived quintile had 4-5 years' shorter life expectancy (5-8 less QALYs) than those in the least deprived quintile.
Benefits from statin therapy across quintiles of socioeconomic deprivation UKB participants in more socioeconomically deprived quintiles were more likely to be at higher CVD risk and, therefore, meet statin treatment criteria, and gained more benefit.For instance, compared with no use of statin therapy, if all people in their 50s recommended statin therapy were initiating and taking it over their lifetime, 703 and 360 life years (399 and 170 QALYs) per 1000 males and females were projected to be gained in the most deprived quintile, respectively, compared with 406 and 94 life years (277 and 55 QALYs) projected to be gained in the least deprived quintile (Figure 5).
In scenario analysis, the real-world use of statin treatment among eligible

Research
participants ranged from about 40% in the least deprived to 45% in the most deprived quintile and, although treatment benefits were proportionately reduced, larger benefits were projected in quintiles with higher socioeconomic deprivation (Supplementary Figure S6).

Model web interface and prediction for a typical individual
The CVD microsimulation model interface and user guide are available

Strengths and limitations
The main strengths of this CVD model are in its use of rich contemporary IPD, including a larger set of characteristics and disease endpoints than other previous studies.First, the model has the advantage of reflecting contemporary CVD trends in the UK, being derived from a large current    10 10 10 9.9 9.9 9.9 9.9 9.9 9.9 9.8 9. 9.9 9.9 9.9 9.9 9.9 9.9 9.8 9.8 9.7 9.6 9.6 9.5 9.5 9.

Figure 2 .
Figure 2. Validation of the CVD model in UK Biobank and Whitehall II Phase 9 participants.Validation covers 12 years in UK Biobank (8 years for incidence diabetes because of stopping follow-up earlier) and 10 years in Whitehall II Phase 9 data (7 years for incidence cancer because of stopping follow-up earlier).CRV = coronary revascularisation.CVD = cardiovascular disease.MI = myocardial infarction.

Figure 3 .
Figure 3. Predicted remaining life expectancy (years) and QALYs for UK Biobank participants.Predicted outcomes presented by CVD history, sex, age, and, for people without CVD history, by 10-year CVD risk (QRISK3).CVD = cardiovascular disease.QALY = quality-adjusted life year.NA = not available.

Figure 4 .
Figure 4. Predicted life years and QALYs in 10 years, 20 years, and over lifetime, by sex, age, and quintile of socioeconomic deprivation in the UK.Predicted remaining life years and QALYs by age, sex, and socioeconomic deprivation quintiles (using Townsend Score) at entry into UK Biobank were standardised to mid-2020 UK population distribution by age, sex, and quintile of socioeconomic deprivation (using Index of Multiple Deprivation quintiles).The bars represent remaining life years or QALYs over lifetime and the areas under the black-dotted lines and red-dotted lines represent QALYs in 10 years and QALYs in 20 years, respectively.CVD = cardiovascular disease.QALY = quality-adjusted life year.

Table 1 . Baseline characteristics of study participants
Risk of vascular endpoints associated with disease-event histories.Adjusted for other individual characteristics at entry and current age.See Supplementary Tables S4-S6 for full details of the risk equations.CRV = coronary revascularisation.CVD = cardiovascular disease.HbA1c = haemoglobin A1C.MI = myocardial infarction.
at https://livedataoxford.shinyapps.io/shiny_ctt_ukb_model.The model interface enables users to project outcome for one or a group of patients.To illustrate its use, the model predicted 5.2% 10-year and 40% lifetime cumulative incidence of major vascular event (MI, stroke, coronary revascularisation, or vascular death), 28 years' further lifespan, and 23 QALYs over the lifetime for a 60-year-old White male, non-smoker, overweight (BMI 25-30 kg/m 2 ), in quintile 3 of socioeconomic deprivation, with moderate physical activities, a Figure 1.

3
Figure5.Predicted lifetime benefit from UK guideline-recommended statin therapy, by sex, age, and quintile of socioeconomic deprivation in UK.Predicted life years and QALYs gained with full implementation of UK National Institute for Health and Care Excellence guideline-recommended statin therapy at entry into UK Biobank were standardised to mid-2020 UK population distribution by age, sex, and quintile of socioeconomic deprivation (using Index of Multiple Deprivation quintiles).Atorvastatin 20 mg/day was used for individuals without cardiovascular (CVD) history but with a 10-year CVD risk ≥10% and/or type 1 diabetes, an estimated glomerular filtration rate (eGFR) <60 mL/ min/1.73 2 , or albuminuria; and atorvastatin 80 mg/day for individuals with CVD history (20 mg in those with eGFR <60 mL/min/1.73 2 ).QALY = quality-adjusted life year.