| About the Guideline and Calculator
How the Calculator was developed
Development of the Aus CVD risk calculator
The Aus CVD risk calculator is based on the NZ PREDICT-1° equation, which was developed from a large, contemporary New Zealand primary care cohort study.17 The equation has been recalibrated to the Australian population and modified for the Australian healthcare system. See Appendix 4 for details of the recalibration procedure.
A literature review and targeted audit of CVD risk equations recommended for use in major international CVD risk management guidelines was conducted. The NZ PREDICT-1° equation was selected based on pre-specified criteria including:
- use of contemporary data sources
- inclusion of established CVD risk factors such as smoking, cholesterol, blood pressure and diabetes
- measures of ethnicity and social deprivation (to improve health equity)
- global CVD events and deaths as outcomes
- representation of the general or primary care population
- excellent model performance
- external validation in populations similar to the Australian population
- the ability to be recalibrated and modified.
Executive Summary: updating of the cardiovascular disease risk equation and risk treatment thresholds used in Australia
Assessing and treating according to a cardiovascular disease (CVD) risk estimated using a validated risk equation is considered international best practice. Under the 2012 Australian CVD risk management guidelines, CVD risk was assessed using an algorithm which included both an assessment of clinically determined high risk criteria and the application of the Framingham Risk Equation to quantify the likelihood of a CVD event in the next five years. Those who were determined to be at clinically determined high risk or those assessed as having a >15% chance of having a major CVD event in the next five years were recommended lifestyle changes and commencement of lipid- and blood pressure-lowering medications, unless contraindicated. The Framingham risk equation is known to generally overestimate risk in the general Australian population and underestimate risk in Aboriginal and Torres Strait Islander populations. Although the Framingham risk equation was the first and one of the best-known equations for estimating CVD risk, there are more than 350 risk equations, scores or tools published that estimate a person’s risk of developing CVD.
Many countries, including the United States, United Kingdom and New Zealand have developed their own CVD risk equations using contemporary linked datasets. To develop a country-specific CVD risk prediction equation, large-scale data on risk factors and CVD outcomes with sufficient follow-up time are needed from one or more cohorts of people from the target country. These data should ideally be representative of the primary care population that the risk equation is being developed for and validated using a separate dataset for the same country. Although Australia does have some high-quality cohorts that capture information on CVD risk factors and outcomes, these are broadly not representative of the general adult Australian population or the primary care population. Instead, an existing validated risk equation developed for a different country could be applied in Australia if it is statistically updated so that the risks predicted by the equation match the CVD event rates in Australia. The update of the Australia’s CVD risk assessment and treatment guidelines provides an opportunity to systematically evaluate which existing international risk equation is best suited to Australia.
Since CVD risk equations are used to guide treatment in primary care, it is vital they quantify risk as accurately as possible. Risk equations do not always perform well when applied in settings different to those in which they were developed. However, risk equations can be statistically updating to match the CVD rates in the country to which they are applied, through a process known as recalibration. In addition to updating a new risk equation, consideration should be given to whether the risk thresholds for recommending commencing treatment should also be changed.
A team led by the Australian National University undertook a program of work to: 1) compare existing international CVD risk equations to a set of agreed selection criteria to identify which risk equation might be best suited for adaptation to Australia; 2) generate the data needed to update the chosen risk equation for use in Australia; and 3) systematically review the published, peer- reviewed global evidence in relation to different risk treatment thresholds for the primary prevention of CVD. The key findings of this work are outlined below and the full reports that these are taken from are provided at the end of this document.
Evidence underpinning the choice of the new CVD risk equation
A systematic approach was used to evaluate the appropriateness of existing international CVD risk equations for use in Australia. First, a set of selection criteria was developed to guide the selection of the most appropriate risk equation. Second, an audit of existing international CVD risk equations used for the primary prevention of CVD was undertaken, focusing on risk equations currently recommended in international guidelines. Third, the existing international risk equations were compared against the selection criteria to identify those most appropriate for Australia.
A set of a priori selection criteria were defined at a National Stakeholder Roundtable on CVD risk in November 2019 and were agreed by the Heart Foundation Absolute CVD Risk Guideline Expert Steering Group in March 2021. The following selection criteria were agreed to be considered when selecting a new CVD risk equation for Australia:
- Use of contemporary data sources
- Inclusion of established CVD risk factors such as cholesterol, blood pressure, diabetes and smoking
- Consideration of measures of ethnicity and social deprivation (to improve health equity)
- Global CVD events and deaths as outcomes
- Population representativeness, either of the general population or the primary care population
- Excellent model performance
- External validation in populations similar to Australia
- The ability to be recalibrated and modified.
Ten international CVD risk equations were reviewed, and of these, the PREDICT equation developed for New Zealand and the QRISK3 equation developed for the United Kingdom most closely matched the selection criteria, meeting 7 and 6 of the 8 selection criteria, respectively.
The PREDICT and QRISK3 equations were both developed using datasets that are broadly representative of the primary care populations that they were developed for, use a range of established CVD and clinical risk predictors, include measures for ethnicity and social deprivation, and estimate risk for a wide range of CVD events and deaths.
The performance of neither the PREDICT nor the QRISK3 equations have been examined in Australia, although both have been shown to outperform Framingham risk equations within the populations they were developed for.
PREDICT was the only equation of the ten reviewed that was able to be readily adapted for use in Australia. Major practical constraints to implementation of the QRISK3 equation in Australian primary care were identified, including that over 30 variables are used to calculate risk.
Based on the evidence, the New Zealand PREDICT equation was recommended by the Heart Foundation Algorithm Working Group as the most suitable primary tool for assessing cardiovascular disease risk in Australia in the short-term. Longer-term Australia should move towards developing a contemporary large-scale data system for developing and regularly updating an Australian-specific CVD risk equation.
Updating the PREDICT risk equation for Australia
The PREDICT equation was developed for the contemporary New Zealand population and is used in primary care to estimate the likelihood of a patient having a first-time CVD event in the next five years. Statistically updating a risk equation usually involves using information on the distribution of risk factors and the CVD event rates in the target country to replace the values in the original risk equation.
However, Australia does not have any suitable representative primary care datasets with information on all the risk factors used in the PREDICT equation to generate average risk factor levels for Australia. Since risk factor levels are less important than baseline CVD event rates in determining CVD risk, a pragmatic approach is to update the PREDICT equation for Australia by aligning the CVD risks predicted by the equation to CVD event rates observed in Australia.
Since the PREDICT equation is already calibrated for the New Zealand population, updating this for the Australian population can be done using the relative differences in CVD event rates between the two populations. Using this approach, CVD event rates in age- and sex-groups would be calculated for the Australian and New Zealand populations separately and multipliers would be calculated as the ratio of CVD event rates between the two countries. The multipliers would be applied to a person’s estimated CVD risk calculated using the PREDICT equation, moving the risk estimate up or down depending on age and sex. So, for example, if CVD event rates were 0.9 times lower in Australian women aged 60-64 years than New Zealand women the same age, the CVD risk score estimated by the PREDICT equation for an Australian woman in this age group would be multiplied by 0.9 to get an updated risk score. This statistical adjustment would happen in the “back end” of the computer program so only a single risk score would be presented to the patient.
The data on CVD event rates in Australia and New Zealand can come from either data on CVD death rates or on first-time CVD events estimated using linked hospital and death data. Since CVD risk equations estimate risk of a first-time CVD event, updating the risk equation using the difference between Australia and New Zealand in first-time CVD events would be ideal. However, in practice, Australia does not yet have national level-linked data on hospitalisations linked to death data to estimate first-time CVD events. Differences between Australia and New Zealand in coding or admissions practices in hospitals also make comparisons difficult. Instead, high quality data on CVD deaths was used to estimate the difference in CVD event rates between Australia and New Zealand to update the PREDICT risk equation for use in Australia. This approach assumes that between-country relative differences in CVD death rates are comparable to relative differences in first-time CVD events (fatal plus non-fatal) in primary care.
Data on CVD death rates was estimated in five-year age- and sex groups using national data for Australia and New Zealand. For Australia, age- and sex-specific CVD death rates were obtained from the Australian Bureau of Statistics. CVD death rates for New Zealand were derived from the New Zealand Ministry of Health Mortality Collection. Average five-year age- and sex-specific deaths rates were calculated using data from 2014-2018, inclusive (most recent data available for both countries). Average five-year CVD death rates were calculated since the PREDICT equation predicts CVD events over a five-year period. CVD deaths were identified using the same set of ICD-10 codes for both countries with outcomes matching those predicted by the PREDICT equation including: heart attack, unstable angina, other coronary heart disease, stroke, transient ischaemic attack, peripheral vascular disease, congestive heart failure, and other CVD-related deaths.
Calculating the adjustment factors
Since it was decided that ethnicity would be removed from the PREDICT risk equation used in Australia, five-year age- and sex-specific CVD death rates for calculated for the full Australian population (regardless of ethnicity) and these were compared to CVD death rates for the New Zealand population excluding Māori and Pacific Islander people. Māori and Pacific Islander people were excluded from the New Zealand population compaison as they account for a substantial proportion (~25%) of the total New Zealand population.
Age- and sex-specific rate ratios for the differences in CVD death between Australia and New Zealand were then calculated by dividing the CVD death rates (per 100,000 people) in Australia by the CVD deaths rates in New Zealand for each age- and sex-group. These rate ratios become the multipliers that can be used to adjust the CVD risks predicted by the PREDICT equation up or down in different age- and sex-groups in Australia.
Evidence to support the choice of CVD risk treatment thresholds
A systematic review was undertaken to identify studies published since 2012 (date of the last guideline update) that reported on the effects of initiating or using pharmacological therapy at different risk treatment thresholds on major CVD events and deaths, and reported findings from randomised controlled trials (RCTs) or modelling studies where treatment effect estimates were taken from RCTs.
A total of 13 studies were identified for review, including three RCTs, seven modelling studies, and three meta-analyses of RCT data. Six studies reported on data relating to initiating or using blood pressure-lowering medication and seven studies reported on lipid-lowering medications. No studies statistically compared the effects of treatment between different absolute risk categories and only two studies aimed to identify the most appropriate risk threshold for commencing treatment. The other 11 studies looked at the effects of treatment on CVD outcomes, stratified by different levels of baseline absolute CVD risk. Only two studies reported data specific to Australia.
The overall evidence on risk treatment thresholds for blood pressure-lowering therapy generally suggests a risk treatment threshold lower than the currently recommended 15% would prevent more major CVD events while requiring a similar number of people to be treated to avert a single event. Data from an individual participant data meta-analysis of trial data showed that blood pressure treatment lowers the risk of major CVD by around 18-22%, even at low levels of absolute risk. This study also suggests that treating those with a 5-year absolute risk >10% is similar in terms of numbers needed to treat to avert one event as recommending treating all those with a blood pressure ≥160/100 mmHg. Treating people with a consistently elevated blood pressure level ≥160/100 mmHg is recommended in the current Australian guidelines where lifestyles changes have not achieved the desired results. From the trial data, applying a five-year risk treatment threshold of >5% would mean treating around 50 people to prevent one CVD event; >10% would mean treating around 35 people over five years to prevent one CVD event and around 30 people would need to be treated at >15% five-year risk to avoid one CVD event. Results from modelling data from China and India showed similar results. The evidence indicates that the number needed to treat to prevent one CVD event increases gradually as the risk treatment threshold is increased.
The overall evidence on risk treatment thresholds for lipid-lowering therapy is also supportive of a lower risk treatment threshold than the currently recommended 15% five-year level. RCT data shows that statin use reduces the relative risk of a CVD event by around 25%, with little evidence that this differs by risk category. Those in the lowest and highest absolute CVD risk categories experience similar relative reductions in CVD events. A meta-analysis of individual participant data from RCTs found that at a five-year absolute risk of <10%, each mmol/L reduction in LDL cholesterol would prevent around 11 major vascular events per 1000 people treated over five years. These benefits of statins were estimated to exceed potential adverse events, even at these lower risk levels. A modelling study which used the inputs from RCTs, including this meta-analysis, aimed to identify 10-year CVD risk thresholds for a range of countries, including Australia. Ten- year risk thresholds ranging from 11 to 18% (approximately equivalent to a five-year risk threshold of around 6-9%) were identified for Australia, varying by age and sex. This was the risk treatment threshold beyond which the estimated benefits of statins outweighed the estimated adverse effects of statins in at least 60% of the simulations.
Decisions on appropriate risk treatment thresholds need to consider the potential benefits and adverse events of treatment, and preferences of patients and doctors. Both blood pressure- and lipid-lowering therapies have been found to be generally safe and are effective across all levels of absolute CVD risk. The international evidence suggests that a risk treatment threshold below the currently recommended 5-year 15% treatment threshold and closer to 6-10% risk could increase the number of major CVD events prevented, while limiting potential adverse effects of treatment. This is consistent with major international guidelines which generally recommend treating at a five- year equivalent risk threshold of around 5-10% in most guidelines. Preventive medications reduce the relative risk of CVD at all levels of absolute risk and have few side effects, so treating individuals at lower levels of absolute risk might be appropriate in certain circumstances. The numbers needed to treat to prevent one CVD event become less favourable at lower levels of risk ( .g. ≤5% f -year risk and lower), based on the current available evidence.
- Paige E, Brown S, Agostino J, Wong D, Timothy A, Nguyen M and Banks E. Evaluation of international cardiovascular disease risk equations for use in Australia: overview and recommendations. Report commissioned by the National Heart Foundation of Australia, 2021.
- Paige E, Jackson R, Patel A, Woodward M, Korda R.J., Wong D, Timothy A, Agostino J, and Banks E. Updating Australia’s caridovascular disease risk prediction equation: a proposed strategy for recalibration. Report commissioned by the National Heart Foundation of Australia, 2021.
- Paige E, Welsh J, Zhang Y, Nedkoff L, Katzenellenbogen J, Lum On M, Brown S, Agostino J, Jackson R, and Banks E. Updating Australia’s cardiovascular disease risk prediction equation: Evidence, methods and data for recalibration. Report for the National Heart Foundation of Australia, 2022.
- Brown S, Nguyen M, Zhang Y, Banks E and Paige E. Evidence on the effects of initiating cardiovascular preventative treatments at different risk treatment thresholds: a systematic review. Report commissioned by the National Heart Foundation of Australia, 2021.