Commentary on selected evidence

With advice from topic experts we selected 1 study for further commentary.

Identifying and assessing CVD risk: risk tools for formal risk assessment

We selected the cohort study by Hippisley-Cox et al. (2017) for a full commentary because it is directly relevant to the existing recommended risk tool, QRISK2, has a potential impact on the guideline and included a large and representative NHS sample.

What the guideline recommends

NICE's guideline on cardiovascular disease (CVD) recommends using the QRISK2 risk assessment tool to assess CVD risk for the primary prevention of CVD in people up to and including age 84 years. It advises against the use of a risk assessment tool to assess CVD risk in people:

  • with type 1 diabetes

  • with an estimated glomerular filtration rate (eGFR) less than 60 ml/min/1.73 m2 and/or albuminuria

  • who are at high risk of developing CVD because of familial hypercholesterolaemia

  • who have pre-existing CVD.

The guideline also recommends that people with an estimated 10-year risk of CVD of 10% or more should be prioritised for a full formal risk assessment, and should be offered atorvastatin 20 mg for primary prevention of CVD.

Methods

The prospective cohort study by Hippisley-Cox (2017) aimed to develop and validate an updated version of the QRISK tool, QRISK3, to enable estimation of 10-year CVD risk for men and women. General practices in England that had been using the EMIS computer system for at least one year were included. Patients were excluded if they:

  • were not aged between 25 and 84 years

  • had no postcode related Townsend score (indicating absence of deprivation data, homelessness or no permanent residence)

  • had pre-existing CVD

  • were using prescribed statins at study recruitment.

A total of 981 practices with 7.89 million patients in England were used to develop the risk scores and 328 practices with 2.67 million patients were used to validate the scores. To establish separate 10-year risk equations for men and women, Cox proportional hazards models were used in the derivation cohort. CVD, defined as 'a composite outcome of coronary heart disease, ischaemic stroke, or transient ischaemic attack', was the primary outcome.

Risk factors in the new algorithm included those already in QRISK2 and several additional new ones:

  • chronic kidney disease

  • a measure of systolic blood pressure variability

  • migraine

  • corticosteroid use

  • systemic lupus erythematosus

  • second generation 'atypical' antipsychotic use

  • diagnosis of severe mental illness

  • diagnosis of HIV/AIDS

  • diagnosis or treatment of erectile dysfunction in men.

Three models were developed; model A contained the same variables as the 2017 version of QRISK2. Models B and C included the additional variables that met the inclusion criteria but differed in that model B did not include the standard deviation of serial systolic blood pressure values, whereas model C did.

Patients were classified as being at high risk of CVD if their 10-year risk was 10% or greater, as recommended by NICE's guideline on CVD.

Results

In total, there were 363,565 incident cases of CVD arising from 50.8 million person years of observation, during a median follow up of 4.4 years.

In terms of variation in time to diagnosis of CVD, the QRISK3 algorithm accounted for 54.8% of the variation in men and 59.6% of the variation in women for model A. Models B and C performed similarly.

In terms of calibration, defined by comparing the mean predicted risks at 10-years with the observed risk by 10th of predicted risk, the results showed that:

  • The mean predicted risk was 6.4% in men, with an observed 10-year risk of 7.5% (95% confidence interval [CI] 7.5% to 7.6%).

  • The mean predicted risk was 4.7% in women, with an observed 10-year risk of 5.8% (95% CI 5.8% to 5.9%).

  • The models appeared to be well calibrated, with the mean predicted risks and the observed risks corresponding closely within each model and in each age group, except in those aged 25–39 where mean observed risks were marginally lower than the predicted risks.

HIV/AIDS was the only additional new risk factor which didn't meet the inclusion criteria, as it was not statistically significant.

Overall performance of the updated QRISK3 algorithms was found to be non-inferior to the QRISK2 algorithms.

Strengths and limitations

Strengths
  • Since the cohort in the study was population based, it was representative of the NHS population. This minimised the risk of selection bias.

  • The study included a very large sample size and used an established approach for analysing large data sets, by randomly splitting data at the general practice rather than the individual level.

  • The study is directly relevant and applicable to the QRISK2 assessment tool that is currently recommended by the guideline, and incorporates additional variables highlighted by the guideline as important risk factors for CVD.

Limitations
  • The authors acknowledged the limitation of lack of formal adjudication of diagnoses, which may have led to classification bias.

  • The study was limited by the potential for bias owing to missing data. The authors used multiple imputations to replace missing values for body mass index, systolic blood pressure and its standard deviation, serum cholesterol, high density lipoprotein cholesterol, and smoking status.

  • The validation study was dependent on data provided by the authors, and an independent study would be a more stringent test of the risk score.

Impact on guideline

The guideline highlighted numerous conditions associated with increased CVD risk that may not be fully captured by QRISK2, including stage 3 kidney disease, systemic lupus erythematosus, severe mental illness, and use of atypical antipsychotics or corticosteroids. The collective new evidence and topic expert feedback indicates that the inclusion of these additional variables in QRISK3 has the potential to enable more accurate assessment of subgroups of patients with specific conditions. Incorporating additional data from the electronic health record may also improve CVD risk stratification.

The new evidence suggests that QRISK3 performs well for people with type 1 diabetes and chronic kidney disease, and may help some people with these conditions to make an informed choice on whether to take statins. There is therefore a potential impact on recommendations 1.1.9 and 1.1.11 to review the advice against using risk tools for people with type 1 diabetes and chronic kidney disease, respectively. This may also have a consequential impact on recommendations for primary prevention (see recommendations 1.3.23, 1.3.24, and 1.3.27) for people with these conditions, although some topic expert feedback indicated that recommendations for people with chronic kidney disease should remain unchanged.

There is a potential need to amend the guideline recommendations (1.1.8 and 1.1.10) to advise the use of QRISK3 in place of QRISK2 because QRISK2 is due to be superseded by QRISK3 in 2018.


This page was last updated: 25 January 2018