Predicting lifetime cardiovascular risk and benefits of preventive treatment in patients with established atherosclerotic cardiovascular disease: the SMART-REACH2 model

Model development

The SMART-REACH2 model was developed using a two-step approach. First, the base of the model was derived using data from patients aged 40–90 years with established atherosclerotic CV disease in the Utrecht Cardiovascular Cohort (UCC-SMART). The UCC-SMART cohort is an ongoing single-centre prospective cohort study at the University Medical Centre Utrecht, The Netherlands. Details have been described previously.⁹ Briefly, since 1996 patients newly referred to the University Medical Centre Utrecht with, or at high risk for, CV disease can be included. Baseline characteristics were determined using a standardized baseline assessment protocol consisting of questionnaires, physical examination, and laboratory testing. To negate the effects of the COVID-19 pandemic, follow-up time was truncated at 1 February 2020, and analyses were restricted to patients included before this date (n = 8708). Established atherosclerotic CV disease at baseline was defined as coronary artery disease (CAD), cerebrovascular disease (CeVD), peripheral artery disease (PAD), and/or abdominal aortic aneurysm (AAA) (see Supplementary data online, Table S1 for definitions).

The second step in model development was the incorporation of recalibration factors to account for regional variations in background CV risk, thus ensuring that the model can be applied across these regions. These regions where the ESC-defined European risk regions (low-, moderate-, high-, and very-high-risk), which are based on CV disease-related mortality rates from the World Health Organization (see Supplementary data online, Figure S1). In the absence of similar classifications, regions outside Europe were grouped based on geography. The recalibration factors for each of these regions are sex-specific expected-observed ratios for the separate endpoints. These were derived from the cohorts deemed most representative for reach risk region considering their size and broad patient inclusion. For the European high- and very-high-risk regions the cohorts within each respective risk region were combined, due to limited number of events in individual cohorts available for recalibration. The cohorts used for recalibration were CPRD for the European low-risk region, SWEDEHEART for the European moderate-risk region, the Estonian Biobank for the European high-risk region, and the combination of EUROASPIRE V, REACH and SURF CHD II for the European very-high-risk region. For other global regions, REACH was used for recalibration, except for the VA which was recalibrated independently. Methods for recalibration for future data sources are available in Supplementary data online, Methods.

External validation

Including these cohorts used for recalibration a total of 15 cohorts with patients aged 40–90 with established atherosclerotic CV disease were available for external validation. Cohorts spanning multiple risk regions were analysed separately by region. For the European low-risk region, the validation cohorts were the United Kingdom's Clinical Practice Research Datalink (CPRD, n = 147 566) the National Health Service in England Secure Data Environment accessed via the British Heart Foundation Data Science Centre's CVD-COVID-UK/COVID-IMPACT Consortium (https://bhfdatasciencecentre.org/areas/cvd-covid-uk-covid-impact/) (NHSE SDE, n = 931 460), the Norwegian Trøndelag Health Study (HUNT3, n = 5698 and HUNT4, n = 6010), the Norwegian COgnitive impairment After STroke (Nor-COAST, n = 654), and NORwegian CORonary Prevention (NORCOR, n = 1112) cohorts, the Spanish Biomarkers in Acute Coronary Syndrome and Biomarkers in Acute Myocardial Infarction (BACS/BAMI, n = 946), and data from low-risk countries in the international Reduction of Atherothrombosis for Continued Health (REACH, n = 6616) registry and European Action on Secondary and Primary Prevention by Intervention to Reduce Events (EUROASPIRE V, n = 1082) survey. The moderate-risk region validation cohorts included the Swedish Web-System for Enhancement and Development of Evidence-Based Care in Heart Disease Evaluated According to Recommended Therapies (SWEDEHEART, n = 87 847), the North Karelia Wellbeing Services County cohort from Finland (Karelia, n = 3303), and moderate-risk countries from REACH (n = 6266) and EUROASPIRE V (n = 1723). The high-risk region validation cohorts comprised the Estonian Biobank (n = 11 669), the Polish Bialystok PLUS/Polaspire (POLASPIRE, n = 243), and high-risk countries from REACH (n = 836) and EUROASPIRE V (n = 1529). For the very-high-risk region validation cohorts were the Serbian part of the Survey of Risk Factors in Coronary Heart Disease (SURF CHD) II registry (n = 203), and very-high-risk countries from REACH (n = 4382) and EUROASPIRE V (n = 2710). Validation cohorts outside of Europe included the American Veterans Health Administration (VA, n = 930 725) and non-European parts of REACH (n = 28 407) which were subdivided into North America (n = 15 857), Latin America (n = 1446), Australia (n = 1963), Middle Eastern countries (n = 654), and Asian countries (n = 4742) excluding Japan (n = 3745) due to differences in CV risk. In total the 15 external validation cohorts covered 54 countries.

Definitions of established atherosclerotic CV disease at baseline in the external validation cohorts were aligned as closely as possible with the definitions in the derivation data. Baseline presence of atherosclerotic CV disease was ascertained through either (i) questionnaires/records or (ii) linkage to national databases (CPRD, NHSE SDE, HUNT3, HUNT4, SWEDEHEART, Karelia, Estonian Biobank). International Classification of Diseases (ICD) codes for disease history are available in Supplementary data online, Table S2. Outcomes were ascertained by (i) linkage to primary care records, hospital records, or disease/mortality registries, (ii) contacting patients (SURF CHD, POLASPIRE), annual questionnaires (REACH registry), or combination of these (EUROASPIRE V). More detailed descriptions of these cohorts are provided in Supplementary data online, Methods, with additional information available in prior publications.^10–26 For the main analyses, follow-up and inclusion were restricted to a period prior to the COVID-19 pandemic if applicable. A sensitivity analysis regarding performance of SMART-REACH2 during the COVID-19 pandemic was performed in a second NHSE SDE cohort spanning including patients with established CV disease alive on 1 January 2020 and follow-up until 1 March 2022 (n = 878 660).

Outcome

The primary outcome of SMART-REACH2 was the same as in the original SMART-REACH model: recurrent CV events, defined as a composite of myocardial infarction, CeVD, or CV death.⁷ Death from causes not included in CV death (i.e. non-CV death) was considered the competing endpoint. The ICD codes used for endpoint definitions in the external validation cohorts are available in Supplementary data online, Table S3.

Statistical analyses

Sex-stratified, cause-specific Cox proportional hazard models were developed separately for recurrent CV events and non-CV death. Age was used as the underlying timescale, with age at entry into the cohort as the start of follow-up (i.e. left truncation) up to age of censoring (end of follow-up) or (primary or competing) event (i.e. right censoring). To adjust for any effect of age of entry in the cohort, age was added as a predictor.²⁷ All models included the following prespecified predictors: age at entry, current smoking, separate indicators for history of CAD, CeVD, PAD, and AAA, diabetes mellitus (any type), (office) systolic blood pressure (SBP), non-high-density lipoprotein (HDL; mmol/L) cholesterol, years since first atherosclerotic CV disease diagnosis at enrolment (those defining the study population), estimated glomerular filtration rate (eGFR) as estimated by the Chronic Kidney Disease Epidemiology Collaboration Epidemiology Collaboration equations and C-reactive protein (CRP; mg/L). For the model predicting recurrent CV events, use of any type of antithrombotic medication was included as an offset term with a hazard ratio (HR) of 0.81.^28,29 This offset term promotes both transferability to other settings by limiting the influence of local prescription practices, and clinical utility by resulting in risk estimates that can inform future treatment decisions on antithrombotic treatment (e.g. initiation of dual-pathway inhibition).²⁸

To limit influence of outliers on model parameters, continuous predictors were winsorized at their sex-specific 1st and 99th percentiles. The need for non-linear transformations was assessed for each continuous predictor by comparing Akaike's Information Criterion between models with and without squared polynomials, cubic polynomials, logarithmic transformations, and square root transformations. Proportional hazards were assessed through visual inspection of Schoenfeld residuals. If applicable, interactions with age were considered. No time-varying covariates were incorporated. To decrease the odds of chance findings, non-linear transformations were initially assessed in the combined population before proceeding with the sex-specific development. Ridge penalization using penalized maximum likelihood estimation was applied to reduce optimism. Optimal shrinkage was determined using five-fold cross-validation with identical folds for both endpoints to ensure proportional congruence. Continuous predictors were centred and scaled (using means and standard deviations, respectively) to enable proportionate penalization (see Supplementary data online, Methods).

Baseline risks were derived by estimating 1-year interval endpoint-specific survival from the cause-specific Cox models over the complete age range (40–90 years), using a reference individual with sex-specific mean values for all risk factors and absence of binary predictors and keeping the risk factor profile stable. These intervals were smoothed using an exponential model weighted by the number of patients in each age group and bounded at 100% survival probability to ensure realistic probabilities (see Supplementary data online, Methods).

Missing data were handled using the mice package, with predictive mean matching for continuous predictors and logistic regression for the binary predictors (15 datasets and 50 iterations). Both endpoints were incorporated using Nelson-Aalen estimates.³⁰ Convergence of the imputation model was assessed using trace plots. Estimates were pooled using Rubin's rules where applicable. Missingness per predictor variable in the derivation data is available in Supplementary data online, Table S4. Missing values in the external validation cohorts were handled within each cohort separately (see Supplementary data online, Methods).³¹

Prediction of individual and lifetime risk

Lifetime risk was defined as a patient's cumulative risk of experiencing a recurrent CV event from their current age up to age 90. Risks were estimated using the lifetable approach which has been validated for prediction horizons up to at least 17 years.³¹ The 1-year risks of recurrent CV events and non-CV death were calculated based on a patient's risk factors for each remaining life year up to age 90. The probability of surviving each year event-free (interval survival probability) was estimated, accounting for the competing risks between CV events and non-CV death. These interval survival probabilities were multiplied to calculate the cumulative probability of reaching a specific age free of any event. CV disease-free life expectancy was defined as the age at which this cumulative survival probability drops below 50%. A detailed example of the lifetable calculations is provided in Supplementary data online, Methods. An interactive online version of the SMART-REACH2 will be made available until future implementation of the model for clinical use in U-prevent.com.

Model performance

Model performance was assessed in terms of discrimination and calibration. Discrimination was assessed using Harrel's concordance (C)-statistic and calibration was assessed through calibration plots. Competing risk-adjusted C-statistics were pooled using a random effects model with inverse variance weighting to account for differences such as follow-up and event rates impacting precision.³² Calibration was evaluated visually through calibration plots of predicted versus observed risks, which were adjusted for competing risks (see Supplementary data online, Methods). Age-specific calibration plots were constructed to assess model performance across observed ages and provide insight into the risk accumulation approach underlying lifetime estimates. Cumulative incidence curves were used to determine the maximum follow-up duration at which model performance could be evaluated in each cohort, with a maximum of 10 years of follow-up. The follow-up times at which model performance was assessed were 2 years for Nor-COAST, HUNT4, both NHSE SDE cohorts, the combined high- and very-high-risk region cohorts, and all REACH regional sub cohorts; 3 years for the VA and the Karelia Health Service cohort; 6 years for BACS/BAMI; 7 years for NORCOR; 8 years for the Estonian Biobank, and CPRD; and 10 years for SWEDEHEART and HUNT3. REACH data from Middle Eastern and Asian countries (excluding Japan) were combined due to comparable event rates and limited cohort sizes, and performance was assessed at 2-year follow-up. The SMART-REACH2 model was compared to the original SMART-REACH model in the HUNT3 cohort and the SWEDHEART registry to assess the impact of the regional recalibration. Performance was assessed through C-statistics, calibration, and clinical utility assessed through decision curve analyses.

Performance was assessed in the overall population and separately in men and women. An additional analysis was performed for the VA, restricting to those eligible only for healthcare through the VA (n = 103 439) to assess the potential for missing outcomes. Two sensitivity analyses were performed. First, to assess the performance across CV disease subtypes (i.e. CAD, CeVD, and PAD/AAA), and their respective (heterogeneous) aetiologies, discrimination and calibration were assessed in two scenarios: (i) among all patients with a specific CV disease subtype regardless of concomitant CV subtypes (e.g. all patients with CeVD, including those with additional CV subtypes) and (ii) among patients with only a single CV disease subtype (e.g. patients with only CeVD, without CAD or PAD/AAA). For the first scenario, discrimination was assessed in all cohorts, the second scenario was only assessed in HUNT3 where most CV subtypes were represented, and follow-up was long. Calibration for both scenarios was assessed in HUNT3 only. The sensitivity analyses regarding performance during the COVID-19 pandemic was assessed in the NHSE SDE pandemic-specific cohort, and these results were not included in the pooled performance estimates.

Estimation of treatment effects

One potential application of the SMART-REACH2 model is the estimation of individual benefits from initiation or intensification of preventive treatment, similar to the original SMART-REACH model.^7,31 Such benefits can be estimated by combining the risk predictions from the SMART-REACH2 model with relative treatment effects derived from clinical trials or meta-analyses, as detailed in Supplementary data online, Methods.³³ This approach parallels other proposed methods of integrating treatment effects with risk prediction to estimate individual benefits.³⁴ As an illustrative example, the SMART-REACH2 model was combined with the effect of lowering of SBP (HR 0.80 per 10 mmHg reduction) and low-density lipoprotein (LDL) cholesterol (HR 0.78 per 1 mmol/L reduction) as obtained from meta-analyses.^35,36 These effects were applied to four identical example patients from all four European risk regions in order to estimate the benefits of these preventive treatments. These benefits can be expressed as absolute risk reductions and gain in CV disease-free life expectancy as a result of treatment. Assumptions for this approach are listed in Supplementary data online, Methods. All analyses were performed with R-statistic programming Version (4.4.1). Results are reported in accordance with TRIPOD (see Supplementary data online, Checklist).

Model development

The derivation cohort consisted of 8708 patients with established atherosclerotic CV disease from UCC-SMART. Mean age at baseline was 61 ± 9 years and 26% (n = 2257) were women. CAD was the most prevalent atherosclerotic CV disease subtype (62%; n = 5412), followed by CeVD (29%; n = 2527), PAD (18%; n = 1534), and AAA (8%; n = 729). Overall, 15% (n = 1318) had two or more atherosclerotic CV disease subtypes. Baseline characteristics stratified by sex are available in Table 1.

Over a median follow-up of 8.5 [25th–75th percentile range; 4.3–13.0] years, 2057 recurrent CV events and 1498 non-CV deaths occurred. HRs for each predictor per endpoint are provided in Supplementary data online, Table S5. No violations of the proportional hazards were observed and no interactions with age were required. The unsmoothed 1-year interval survivals and fitted exponential models are available in Supplementary data online, Figures S3 and S4. For the base model, the C-statistic for recurrent CV events upon internal validation was 0.72 [0.68–0.71] and for non-CV death 0.72 [0.70–0.73]. Calibration at 10 years was adequate for both endpoints (see Supplementary data online, Figure S5). The regional recalibration factors incorporated onto the base model are available in Supplementary data online, Table S9. All parameters required for risk estimation using SMART-REACH2 are available in Supplementary data online, Tables S6–9, including a worked-out example (see Supplementary data online, Figure S2). R code for the SMART-REACH2 model with illustration of use are available on GitHub (https://github.com/JorisHolt/SMARTREACH2). An online interactive calculator for the SMART-REACH2 model is available here (https://holtropjoris.shinyapps.io/SMART-REACH2/).

External validation

External validation was performed in 2 085 780 patients, among whom 396 902 women, with established atherosclerotic CV disease from 15 cohorts spanning 54 countries. Mean age at baseline was 69 ± 10 years. Most data (58%, n = 1 206 351) originated from cohorts in Europe. Median follow-up ranged from 1.3 [1.1–1.6] years in EUROASPIRE-V to 12.0 [5.8–13.8] years in HUNT3. In total, 307 706 recurrent CV events and 91 967 non-CV deaths occurred during follow-up. The incidence rate for recurrent CV events varied from 24 [95% confidence interval: 19–29] to 100 [100–100] per 1000 person-years between cohorts (Table 2). The incidence rate of non-CV death varied from 3 [0–9] to 29 [21–40] per 1000 person-years between cohorts (Table 2).

The regional recalibration factors incorporated onto the base model are available in Supplementary data online, Table S9. The final model showed adequate calibration across risk regions for both endpoints (Figures 1–3 and Supplementary data online, Figures S6–8). Age-specific calibration plots showed good agreements between the average predicted risk and observed incidences across all ages from 40 and older (see Supplementary data online, Figures S9–11). In additional analysis, good agreement between predicted and observed risks in those eligible only for healthcare through the VA was demonstrated, suggesting the miscalibration to be the result of events registered outside of the VA (see Supplementary data online, Figure S12). The pooled C-statistic across all cohorts for the risk of recurrent CV events was 0.68 [95% confidence interval 0.66–0.69] (Figure 4). The C-statistics for recurrent CV events within the four European risk regions were 0.66 [0.64–0.69] for the low-risk region, 0.69 [0.67–0.70] for the moderate-risk region, 0.70 [0.64–0.75] for the high-risk region, and 0.68 [0.63–0.73] for the very-high-risk region (see Supplementary data online, Figures S13–15). For non-European regions, the pooled C-statistic was 0.62 [0.56–0.69] and ranged from 0.62 [0.62–0.62] for the Veterans Affairs cohort to 0.72 [0.66–0.78] for Latin America (see Supplementary data online, Figure S16). The C-statistics for non-CV death were 0.73 [0.68–0.78] for the low-risk region, 0.69 [0.65–0.73] for the moderate-risk region, 0.68 [0.65–0.72] for the high-risk region and 0.66 [0.62–0.70] for the very-high-risk region, with a pooled estimate across all cohorts of 0.69 [0.66–0.71] (see Supplementary data online, Figure S17). Calibration and C-statistics for both endpoints were comparable in subgroups of sex (see Supplementary data online, Figures S18–23). Comparison with the original SMART-REACH model demonstrated comparable calibration and clinical utility in the low-risk region. However, in the moderate-risk region, SMART-REACH2 maintained adequate calibration whilst the original model underestimated risk, resulting in higher clinical utility for the updated model (see Supplementary data online, Figure S24).

Sensitivity analyses

Discrimination for both recurrent CV events and non-CV death was similar across specific CV disease subtypes (see Supplementary data online, Figures S25 and S26). Restricting the analyses to patients with a single CV disease subtype did not change the results. Adequate calibration was observed across different CV subtypes (see Supplementary data online, Figure S27). Model performance in the NHSE SDE COVID-19 pandemic-specific cohort showed underprediction for non-CV death and mild underprediction for recurrent CV events (see Supplementary data online, Figure S28). This analysis was performed according to a pre-specified analysis plan published on GitHub (https://umcutrecht-my.sharepoint.com/personal/j_holtrop-3_umcutrecht_nl/Documents/Documenten/GitHub), along with the phenotyping definitions, and data curation and analysis code.

Example of estimation of treatment effects

A clinical example of the possible use of SMART-REACH2 for estimation of long-term gain from SBP and LDL cholesterol lowering is available in Figure 5. The modelled gain in CV disease-free life expectancy reflects for the combined effect of lowering SBP by 15 mmHg and LDL cholesterol by 1.0 mmol/L for the same clinical vignette for each of the four European risk regions. These example patients are males aged 50 with a history of CAD for 1 year with the following risk factor levels: SBP 150 mmHg, CRP 2 mg/L, eGFR 75 mL/min/1.73 m², non-HDL cholesterol of 2.8 mmol/L with an LDL cholesterol of 1.8 mmol/L. The gain in CV disease-free life expectancy for this intensification of preventive treatment was 2.0 years for the patient from the low-risk region, 3.1 for the moderate-risk region, 4.0 years for the high-risk region, and 4.4 years for the very-high-risk region.

Disclosure of Interest

J.H., M.N.G., S.V.D., C.-E.L., H.D., M.O, I.M.F., I.S., M.C., M.T., M.G.M., Y.M.R., K.K., I.B., D.B., J.W.M., F.L.J.V., S.H.J.H., and J.A.N.D. have no potential conflicts of interests. S.B.L., S.K., A.W., and E.D.A. are supported by core funding from the British Heart Foundation (RG/18/13/33946: RG/F/23/110103), NIHR Cambridge Biomedical Research Centre (NIHR203312) [*], BHF Chair Award (CH/12/2/29428), Cambridge BHF Centre of Research Excellence (RE/18/1/34212), and by Health Data Research UK, which is funded by the UK Medical Research Council, Engineering and Physical Sciences Research Council, Economic and Social Research Council, Department of Health and Social Care (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health and Social Care Research and Development Division (Welsh Government), Public Health Agency (Northern Ireland), British Heart Foundation, and the Wellcome Trust. A.W. is supported by the BHF Data Science Centre (HDRUK2023.0239), Health Data Research UK (Big Data for Complex Disease-HDR-23012), and as an NIHR Research Professor (NIHR303137). J.M. has received modest lecture fees from Novartis, Bohringer Ingelheim outside the submitted work. This work was supported by the Ministry of Education and Research Centres of Excellence grant TK214 name of CoE. This project has received funding from the European Union's Horizon Europe research and innovation programme under grant agreement No 101060011. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. U.V. was supported by the Estonian Research Council grant PUT (PRG1291). The Project leading to this article has received funding from the European Union's Horizon Europe research and innovation programme under grant agreement No 101137278. M.O. is an employee of the ESiOR Oy company. ESiOR Oy applied for data access and provided the calibration results for the North Karelia Wellbeing Services County cohort in Finland. J.M. is a founding partner of ESiOR Oy. P.G.S. received research grants from Amarin, Sanofi, has received honoraria from Amarin, Amgen, AstraZeneca, Bayer, Bristol-Myers Squibb, Idorsia, Janssen, Lexicon, Lilly, Merck, Novartis, Novo-Nordisk, Pfizer, Sanofi. He is Chief Medical Officer for Bioquantis and Senior Associate Editor at Circulation. T.T. was supported by the Estonian Research Council grant PSG809, the Estonian Ministry of Education and Research (project TEM-TA72; and grant TK218 establishing the Estonian Centre of Excellence for Well-Being Sciences (EstWell)). K.K.R. has received research grants from Amarin, Amgen, Daiichi Sankyo, Merck Sharp & Dohme, Pfizer, Regeneron, and Sanofi, and is consultant for Abbott, Amarin, Amgen, AstraZeneca, Bayer, Biologix, Boehringer Ingelheim, Cargene Therapeutics, CRISPR, CSL Behring, Eli Lilly and Company, Esperion, Kowa Pharmaceuticals, NewAmsterdam Pharma, Novartis, Novo Nordisk, Pfizer, Regeneron, Resverlogix, Sanofi, Scribe Therapeutics, Silence Therapeutics, Vaxxinity, and Viatris, honoraria for lectures from Amarin, Amgen, Astra Zeneca, Biologix Pharma, Boehringer Ingelheim, Daiichi Sankyo, Esperion, Macleod, Novartis, Novo Nordisk, Sanofi, and Viatris, and stock options from New Amsterdam Pharma, Pemi 31, and SCRIBE Therapeutics. D.L.B. discloses the following relationships—Advisory Board: Angiowave, Bayer, Boehringer Ingelheim, CellProthera, Cereno Scientific, E-Star Biotech, High Enroll, Janssen, Level Ex, McKinsey, Medscape Cardiology, Merck, NirvaMed, Novo Nordisk, Stasys; Tourmaline Bio; Board of Directors: American Heart Association New York City, Angiowave (stock options), Bristol Myers Squibb (stock), DRS.LINQ (stock options), High Enroll (stock); Consultant: Broadview Ventures, Corcept Therapeutics, GlaxoSmithKline, Hims, SFJ, Summa Therapeutics, Youngene; Data Monitoring Committees: Acesion Pharma, Assistance Publique-Hôpitaux de Paris, Baim Institute for Clinical Research (formerly Harvard Clinical Research Institute, for the PORTICO trial, funded by St. Jude Medical, now Abbott), Boston Scientific (Chair, PEITHO trial), Cleveland Clinic, Contego Medical (Chair, PERFORMANCE 2), Duke Clinical Research Institute, Mayo Clinic, Mount Sinai School of Medicine (for the ENVISAGE trial, funded by Daiichi Sankyo; for the ABILITY-DM trial, funded by Concept Medical; for ALLAY-HF, funded by Alleviant Medical), Novartis, Population Health Research Institute; Rutgers University (for the NIH-funded MINT Trial); Honoraria: American College of Cardiology (Senior Associate Editor, Clinical Trials and News, ACC.org; Chair, ACC Accreditation Oversight Committee), Arnold and Porter law firm (work related to Sanofi/Bristol-Myers Squibb clopidogrel litigation), Baim Institute for Clinical Research (formerly Harvard Clinical Research Institute; AEGIS-II executive committee funded by CSL Behring), Belvoir Publications (Editor in Chief, Harvard Heart Letter), Canadian Medical and Surgical Knowledge Translation Research Group (clinical trial steering committees), CSL Behring (AHA lecture), Cowen and Company, Duke Clinical Research Institute (clinical trial steering committees, including for the PRONOUNCE trial, funded by Ferring Pharmaceuticals), HMP Global (Editor in Chief, Journal of Invasive Cardiology), Journal of the American College of Cardiology (Guest Editor; Associate Editor), Level Ex, Medtelligence/ReachMD (CME steering committees), MJH Life Sciences, Oakstone CME (Course Director, Comprehensive Review of Interventional Cardiology), Piper Sandler, Population Health Research Institute (for the COMPASS operations committee, publications committee, steering committee, and USA national co-leader, funded by Bayer), WebMD (CME steering committees), Wiley (steering committee); Other: Clinical Cardiology (Deputy Editor); Patent: Sotagliflozin (named on a patent for sotagliflozin assigned to Brigham and Women's Hospital who assigned to Lexicon; neither I nor Brigham and Women's Hospital receive any income from this patent); Research Funding: Abbott, Acesion Pharma, Afimmune, Aker Biomarine, Alnylam, Amarin, Amgen, AstraZeneca, Bayer, Beren, Boehringer Ingelheim, Boston Scientific, Bristol-Myers Squibb, Cardax, CellProthera, Cereno Scientific, Chiesi, CinCor, Cleerly, CSL Behring, Faraday Pharmaceuticals, Ferring Pharmaceuticals, Fractyl, Garmin, HLS Therapeutics, Idorsia, Ironwood, Ischemix, Janssen, Javelin, Lexicon, Lilly, Medtronic, Merck, Moderna, MyoKardia, NirvaMed, Novartis, Novo Nordisk, Otsuka, Owkin, Pfizer, PhaseBio, PLx Pharma, Recardio, Regeneron, Reid Hoffman Foundation, Roche, Sanofi, Stasys, Synaptic, The Medicines Company, Youngene, 89Bio; Royalties: Elsevier (Editor, Braunwald's Heart Disease); Site Co-Investigator: Cleerly.

Data Availability

The data used for derivation of the SMART-REACH2 model can be applied for here (https://www.umcutrecht.nl/en/ucc-smart). The North East-Newcastle and North Tyneside 2 research ethics committee provided ethical approval for the CVD-COVID-UK/COVID-IMPACT research programme (REC No 20/NE/0161) to access, within secure trusted research environments, unconsented, whole-population, anonymized data from electronic health records collected as part of patients’ routine healthcare.

The data used in this study are available in NHS England's Secure Data Environment (SDE) service for England, but as restrictions apply they are not publicly available (https://digital.nhs.uk/services/secure-data-environment-service). The CVD-COVID-UK/COVID-IMPACT research programme, led by the BHF Data Science Centre (https://bhfdatasciencecentre.org/), received approval to access data in NHS England's SDE service for England from the Advisory Group for Data (AGD) (https://digital.nhs.uk/about-nhs-digital/corporate-information-and-documents/advisory-group-for-data)—formerly the Independent Group Advising on the Release of Data (IGARD)—via an application made in the Data Access Request Service (DARS) Online system (ref. DARS-NIC-381078-Y9C5K) (https://digital.nhs.uk/services/data-access-request-service-dars/dars-products-and-services). The CVD-COVID-UK/COVID-IMPACT Approvals & Oversight Board (https://bhfdatasciencecentre.org/areas/cvd-covid-uk-covid-impact/) subsequently granted approval to this project (CCU004) to access the data within NHS England's SDE service for England. The anonymized data used in this study were made available to accredited researchers only. Those wishing to gain access to the data should follow the application process of the relevant national data custodian.

Funding

S.H.J. Hageman was funded by the Dutch Heart Foundation 03-006-2023-009. The UCC-SMART study was supported by a grant from the University Medical Centre Utrecht, The Netherlands. This work was supported by core funding from the British Heart Foundation (RG/18/13/33946: RG/F/23/110103), NIHR Cambridge Biomedical Research Centre (NIHR203312) [*], BHF Chair Award (CH/12/2/29428), Cambridge BHF Centre of Research Excellence (RE/18/1/34212), and by Health Data Research UK, which is funded by the UK Medical Research Council, Engineering and Physical Sciences Research Council, Economic and Social Research Council, Department of Health and Social Care (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health and Social Care Research and Development Division (Welsh Government), Public Health Agency (Northern Ireland), British Heart Foundation and the Wellcome Trust. D.A.W. is supported by the BHF Data Science Centre (HDRUK2023.0239), Health Data Research UK (Big Data for Complex Disease-HDR-23012), and as an NIHR Research Professor (NIHR303137). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. BACS-BAMI studies were funded by Spanish Society of Cardiology and Fondo de Investigaciones Sanitarias: Projects PI14/01567, PI20/00923, and PI24/00978 funded by Instituto de Salud Carlos III (ISCIII) and co-funded by the European Union (http://www.isciii.es/). The REACH Registry was supported by Sanofi-Aventis and Bristol-Myers Squibb and is endorsed by the World Heart Federation. The EUROASPIRE V survey was carried out under the auspices of the ESC –EORP. Since the start of EORP, the following companies have supported the programme: Amgen, Eli Lilly, Pfizer, Sanofi, Ferrer, and Novo Nordisk. The research was conducted using the Estonian Centre of Genomics/Roadmap II funded by the Estonian Research Council (project number TT17). Białystok PLUS Polaspire were supported by statutory funds of the Medical University of Bialystok. The funding sources of the original studies were not involved in the design and conduct of the current study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

The British Heart Foundation Data Science Centre (grant No SP/19/3/34678, awarded to Health Data Research (HDR) UK) funded co-development (with NHS England) of the Secure Data Environment service for England, provision of linked datasets, data access, user software licences, computational usage, and data management and wrangling support, to coordinate national COVID-19 priority research. Consortium partner organizations funded data access, user software licences and computational usage and the time of contributing data analysts, biostatisticians, epidemiologists, and clinicians.

Ethical Approval

Separate ethical approval was not required for this study. The UCC-SMART cohort involves human participants and was approved by the Medical Research Ethics Committee (MREC) NedMec of the UMC Utrecht (reference number 22-088). Participants gave informed consent to participate in the study before taking part.

Pre-registered Clinical Trial Number

Not applicable.