Evidence supporting the choice of a new cardiovascular risk equation for Australia

Summary This article reviews the risk equations recommended for use in international cardiovascular disease (CVD) primary prevention guidelines and assesses their suitability for use in Australia against a set of a priori defined selection criteria. The review and assessment were commissioned by the National Heart Foundation of Australia on behalf of the Australian Chronic Disease Prevention Alliance to inform recommendations on CVD risk estimation as part of the 2023 update of the Australian CVD risk assessment and management guidelines. Selected international risk equations were assessed against eight selection criteria: development using contemporary data; inclusion of established cardiovascular risk factors; inclusion of ethnicity and deprivation measures; prediction of a broad selection of fatal and non‐fatal CVD outcomes; population representativeness; model performance; external validation in an Australian dataset; and the ability to be recalibrated or modified. Of the ten risk prediction equations reviewed, the New Zealand PREDICT equation met seven of the eight selection criteria, and met additional usability criteria aimed at assessing the ability to apply the risk equation in practice in Australia.


Narrative reviews
Evidence supporting the choice of a new cardiovascular risk equation for Australia Sinan Brown 1 , Emily Banks 1 , Mark Woodward 2,3 , Natalie Raffoul 4 , Garry Jennings 4,5 , Ellie Paige 1,6 C ardiovascular disease (CVD) is the leading cause of death globally, responsible for an estimated 17.9 million deaths in 2019, 1 and accounting for 25% of all deaths in Australia in 2021. 2 CVD is highly preventable through health risk behaviour modification and pharmacotherapy. Assessment and treatment of cardiovascular risk using validated risk equations is considered international best practice and is the cornerstone of primary CVD prevention.
In the 2012 version of the Australian guidelines on CVD risk assessment and management, 3 the Framingham risk equation 4 was recommended as part of an overall algorithm that included a first stage assessment of an individual's medical history. 3 The United States Framingham equation was the first and is still the best known equation for assessing absolute CVD risk. 5 Although pioneering at the time, its development in a predominantly white US cohort recruited from 1948 limits its applicability in contemporary Australia. On average, it overestimates risk for the general Australian population 6 and underestimates risk for Aboriginal and Torres Strait Islander peoples in both remote and urban settings. 7,8 Many countries, including New Zealand, the United Kingdom, the US and several European countries have developed countryor region-specific risk equations based on local data and tailored to local CVD event rates. Australia currently lacks the large scale contemporary linked datasets with population-based data on CVD risk factors and event rates to generate an Australianspecific CVD risk equation. An alternative is to use an existing risk equation and recalibrate it to align with the CVD event rates observed in Australia.
There is little published guidance on how to choose an existing CVD risk equation for a country. In 2020, the World Health Organization (WHO) published a technical package (HEARTS) for risk-based CVD management in primary health care, which included a process for selecting a risk equation. 9 However, this primarily focused on the use of either population-specific risk equations or the WHO global CVD risk charts and does not consider the use of other existing risk equations.
The Australian CVD risk assessment and management guideline, including the risk equation, was updated in 2023 (https://www. cvdch eck.org.au). This article outlines the process and evidence informing the decision on the CVD risk equation recommended for Australia in the 2023 guideline update.

Methods
We used a systematic approach to evaluate the appropriateness of existing international CVD risk equations for use in primary care in Australia. First, selection criteria were developed to guide the choice of the most appropriate risk equation. Second, we undertook a review of existing international CVD risk equations, focusing on risk equations currently recommended in major international guidelines. Third, the existing international risk equations identified through the review were assessed against each selection criterion. Evidence from this review and comparison to selection criteria were presented to the Guideline Expert Steering Group to inform decisions on the risk equation recommended for Australia in the updated CVD risk assessment and management guidelines.
A set of a priori selection criteria were defined at a National Stakeholder Roundtable on CVD risk, hosted by the National Heart Foundation of Australia and the Australian National University. The Roundtable was held on 19 November 2019 and included 24 attendees from 13 organisations: the National Heart Foundation of Australia, the Australian National University, Monash University, the University of Sydney, the University of Queensland, Northern Hearts WA, Sunshine Coast University, Bond University, the University of Tasmania, The George Institute for Global Health, the University of Adelaide, the Commonwealth Scientific and Industrial Research Organisation, and the University of Auckland. The Roundtable agreed on the following criteria for the selection of a new CVD risk equation for Australia: • use of contemporary data -because the prevalence of risk factors and the relationship of predictors to CVD outcomes can change over time, risk equations based on more contemporary datasets are likely to be better at predicting risk for that population;

Summary
• This article reviews the risk equations recommended for use in international cardiovascular disease (C VD) primary prevention guidelines and assesses their suitability for use in Australia against a set of a priori defined selection criteria.
• The review and assessment were commissioned by the National Heart Foundation of Australia on behalf of the Australian Chronic Disease Prevention Alliance to inform recommendations on C VD risk estimation as part of the 2023 update of the Australian C VD risk assessment and management guidelines.
• Selected international risk equations were assessed against eight selection criteria: development using contemporary data; inclusion of established cardiovascular risk factors; inclusion of ethnicity and deprivation measures; prediction of a broad selection of fatal and non-fatal C VD outcomes; population representativeness; model performance; external validation in an Australian dataset; and the ability to be recalibrated or modified.
• Of the ten risk prediction equations reviewed, the New Zealand PREDIC T equation met seven of the eight selection criteria, and met additional usability criteria aimed at assessing the ability to apply the risk equation in practice in Australia.

Narrative reviews
• inclusion of established CVD risk factors -age, sex, cholesterol, blood pressure, diabetes and smoking; • consideration of measures of ethnicity and social deprivation (to improve health equity); • inclusion of a broad range of CVD events and deaths as outcomes, such as myocardial infarction, stroke and coronary heart disease; • population representativeness, either of the general population or the primary care population; • excellent model performance including discrimination (Cstatistic > 0.7) and calibration in the population in which the equation was tested; • external validation in an Australian population; and • ability to be recalibrated and modified for a different population.
A targeted narrative review of major international CVD risk management guidelines and the risk equations recommended by the Roundtable was conducted in February 2020, supplemented by a previous literature review on CVD  Each of the risk equations recommended in the guidelines was then assessed against the selection criteria, noting whether each criterion had been met. For CVD risk equations that met most of the selection criteria, practical implications of applying the risk equation in Australia were considered, guided by two of the usability considerations outlined in the WHO HEARTS technical package, 9 specifically whether: (i) the format (online risk calculator, charts) available for the risk equation was appropriate for the target population; and (ii) the risk factors could be feasibly measured in the target population.

Review of existing international CVD guidelines and risk equations
Eleven international guidelines on CVD risk management were identified, five of which were from Europe (Box 1). There was wide variation in which underlying CVD risk equations were recommended, with a total of 12 unique CVD risk equations recommended for use, 10 of which estimated risk over a specific time period (Box 2). All but two estimated risk over a 10-year period, most were derived from general or cohort population samples, and all but one had been validated in one or more external datasets (Box 3). Although the QRISK2 risk equation is recommended for use in UK guidelines, we also assessed the QRISK3 risk equation which supersedes QRISK2 and is currently available for use in primary care practices in the UK.

Evaluation of CVD risk equations against the selection criteria
Overall, the identified equations varied in the extent to which they met the selection criteria (Box 4 Global CVD events and deaths outcomes. All the CVD risk equations predict, at a minimum, coronary heart disease (CHD) and stroke events and deaths, except for the Suita score, SCORE and NORRISK 2 equations (Box 3 and Box 4). The Suita score predicts CHD events and deaths but not stroke; SCORE predicts fatal CVD outcomes only; and NORRISK 2 only predicts acute myocardial infarction (MI) and stroke (Box 3 Recalibration potential. Of the 10 CVD risk equations evaluated, five could potentially be recalibrated: Framingham-D'Agostino (2008), PREDICT, SCORE, Suita score, and the WHO risk charts (Box 4). All these equations, except PREDICT, could be recalibrated using CVD event rate and risk factor data from national health surveys.

Usability and practical implications
Due to the number of risk factors included in the PREDICT and QRISK3 risk equations, neither can practically be applied through colour risk charts, which are useful in situations where appropriate electronic resources are not available. Both PREDICT and QRISK3 were generated using data from primary care encounters and can only be implemented in practice using an online risk calculator, or risk calculators embedded within general practice software. All the risk factors in PREDICT could feasibly be measured and recorded during normal primary care consultations in Australia. However, the number of risk factors and the detailed information needed on other medical conditions and medications in QRISK3 means that it would be difficult to apply in practice in Australia. QRISK3 was designed specifically for use in the UK and is integrated within electronic health records such that risk factor data can be prefilled automatically using information recorded in a patient's electronic record.
The UK also uses a standard classification system for disease diagnoses, which means that these are captured in the same way across general practice systems. This is often not the case in Australia. These differences mean that it is currently not possible to practically implement QRISK3 in the Australian primary care system.

Discussion
The      A key limitation of adopting the PREDICT equation in Australia is that its performance has not been, and currently cannot be, directly examined in an Australian dataset owing to a lack of datasets containing all the necessary risk factors and outcomes data. However, PREDICT has been directly compared with previous Framingham equations in NZ and found to perform better. As Australia's population is more like the NZ population than a selected 1970s predominantly white US population (used to develop the Framingham equations), it is very likely that CVD risk prediction would be improved by using PREDICT rather than the Framingham equation, and recalibrating it for use in the Australian population. Priority should be given to validating the PREDICT equation within Australia when appropriate data become available.
CVD risk equations integrated into general practice software are more likely to be taken up in practice. In Australia, there are several software vendors supplying decision support tools to primary care. Ideally the CVD risk equation used in Australia will be integrated consistently into existing general practice software, with fields being pre-populated from clinical records.
The CVD risk equations assessed in our review were all from high income nations, and other than Australia, Japan, NZ and the global WHO risk charts, were solely from North America and Europe. There is an under-representation of equations from other global regions, such as Asia, reflecting in part a lack of major guidelines or, in the case of Chinese guidelines for CVD prevention, 81 a lack of accessible information on the risk equation used.
Since undertaking this review, updated versions of both the Canadian 11 and European 22 guidelines have been published (both in 2021). We did not assess these latest guidelines as both were published after completion of the current review and were not available in time to inform the updating of the Australian CVD risk assessment and management guidelines. There were no changes to the risk prediction equation recommended in the Canadian guidelines; however, the European guidelines now recommend the use of SCORE2, 82 an updated version of SCORE. SCORE2 predicts both fatal CVD and non-fatal MI and stroke, and now includes diabetes in the equation. SCORE2 would have met three of the selection criteria, thereby not changing the overall conclusions of this review.
In Australia, efforts should be made to develop the infrastructure needed to assimilate data from different primary care software platforms and link with hospital and deaths data, such that an Australian-specific CVD risk equation can be developed.

Conclusion
We used a systematic approach to review evidence on existing international CVD risk equations and compare them to a set of a priori defined selection criteria developed for Australia. The NZ PREDICT equation met the greatest number of selection criteria and was considered most feasible to implement in Australian primary care. The PREDICT equation, which was developed in a contemporary, diverse primary care population and includes measures of social deprivation, is likely to offer better detection of CVD risk in Australia compared with the currently used Framingham risk equation. There needs to be careful consideration of how to implement the PREDICT equation in Australian primary care to ensure use is in accordance with guideline recommendations across different general practice software platforms. Ultimately, Australia should be working towards capturing and linking the data necessary to develop Australian-specific CVD risk equations that can be refined and updated over time.