by Dr Geoffrey Modest
In prior blogs I commented in a not particularly quantitative fashion that it made sense to have a gestalt about how aggressive to be in coronary artery disease prevention (see https://blogs.bmj.com/bmjebmspotlight/2014/05/09/primary-care-corner-with-geoffrey-modest-md-aha-lipid-guidelines-again/ or https://blogs.bmj.com/bmjebmspotlight/2016/07/01/primary-care-corner-with-geoffrey-modest-md-migraine-and-heart-disease-in-women/ for example). The point was that the available atherosclerotic risk calculators did not include many conditions that confer increased risk, including such things as peripheral artery disease, chronic kidney disease, microalbuminuria, chronic inflammatory conditions, A1c, migraine, etc. (Eg, for someone with a borderline cardiovascular risk in terms of initiating statin or other more aggressive therapies, perhaps the decision to initiate such therapy should include these conditions). A recent article from the UK included many of these conditions in a new quantitative risk calculator QRISK3 (see doi.org/10.1136/bmj.j2099 ).
— in the UK, clinicians have been using a 10-year cardiovascular risk tool called QRISK since 2007, which was soon updated to QRISK2 with frequent subsequent updates. This latter risk tool evaluates risk on all people aged 25-84, and includes age, ethnicity, deprivation (the Townsend score, which reflects social class by measuring unemployment levels, non-car ownership, non-home ownership, household overcrowding), systolic blood pressure, BMI, total cholesterol/HDL ratio, smoking, family history of coronary artery disease in a first-degree relative aged less than 60, type I diabetes, type II diabetes, treated hypertension, rheumatoid arthritis, atrial fibrillation, and chronic kidney disease (stage 4 or 5).
— They are now proposing updating this calculator (QRISK3) to include chronic kidney disease (stage 3, 4, or 5), a measure of systolic blood pressure variability (standard deviation of repeated measures), migraine, corticosteroid use, systemic lupus erythematosus, atypical psychotics, severe mental illness, a erectile dysfunction in men, and HIV/AIDS.
— 981 clinical practices we used to derive the prediction scores based on these new parameters, with 7.89 million patients aged 25-84; and 328 clinical practices with 2.67 million patients were used in the validation cohort. All patients were free of cardiovascular disease at baseline and were not on statins.
— Baseline characteristics of these cohorts were similar, mean age 43, Townsend score 0.5, BMI 26, cholesterol/HDL ratio 3.9, systolic blood pressure 126 mmHg, systolic blood pressure variability 9.5 mmHg, non-smoker 45%/former smoker 15%/light smoker 12%/moderate smoker 6%/heavy smoker 4%, family history of heart disease 10%, type I diabetes 0.3%, type II diabetes 2%, treated hypertension 5%, rheumatoid arthritis 1%, atrial fibrillation 0.5%, chronic kidney disease (stage 3, 4, or 5) 0.5%, migraine 2%, steroid use 2%, HIV-AIDS 0.2%, SLE 0.1%, atypical psych antipsychotic use 0.5%, severe mental illness 6%, erectile dysfunction 2%
— The outcome measured was cardiovascular disease defined as the composite outcome of coronary heart disease, ischemic stroke, and TIA.
— in the derivation cohort, for the mulivariate-adjusted model, the hazard ratios were: 5.62 for type 1 diabetes, 4.92 for atrial fibrillation, 2.91 for type 2 diabetes, 2.31 for chronic kidney disease, 2.14 for SLE, 1.81 for steroid use, 1.58 for family history of CHD, 1.35 for migraine, 1.29 for antipsychotic use, 1.24 for RA, 1.13 for severe mental illness, and 1.08 for SBP variability
— there were 363, 565 incident cases of cardiovascular disease from 50.8 million person-years of observation to assess the value of these newly proposed risk factors, and all met the model inclusion criteria (at least a 10% increased or decreased risk) except for HIV/AIDS.
— In women, QRISK3 explained 59.6% of the variation in time to diagnosis of cardiovascular disease
— in men, 54.8% of the risk was explained
— overall performance of the QRISK3 algorithm was similar to the QRISK2 one.
— it is not surprising that adding these risk factors did not significantly change the predictive model, since the overall prevalence of these risk factors was quite small
— In fact, looking at the baseline demographics, this population was pretty young and remarkably devoid of medical comorbidities (quite an apparently healthy population, at least as compared to any I’ve ever worked with) and therefore had a pretty low likelihood of a clinical event in the next 10 years. and this might affect its generalizability to older/sicker populations.
— It is notable that the basic risk calculator initially used in the UK since 2007 already included several items not in the US risk calculators, such as markers of social class, chronic kidney disease, atrial fibrillation, rheumatoid arthritis.
— The new risk calculator is available online at https://qrisk.org/three/ . It is important to recognize that many patients do not have information on all of the components of this risk calculator (and the risk calculator does not require them). In fact both the derivation and validation cohorts above were often missing pieces of data, and the researchers used the statistically valid technique of multiple imputation whereby the missing data is imputed by assuming different possible scenarios (eg assuming that the missing data was random and not different from the observed data, or perhaps incorporating various models where there are systematic differences between the missing data and the observed data: for fuller description, see Sterne JAC. BMJ 2009; 338: b2393)
— this study did find that all of their proposed conditions that might change the risk of atherosclerotic events, except HIV/AIDS, did in fact confer a significant increased risk on their multivariate-adjusted model
–however, overall this new risk model QRISK3 did not add anything to their prior risk model, which I think is simply because the prevalence of these conditions in their population was so small. It would have been interesting to look at a population with, for example, high levels of RA or on lots of antipsychotic meds to see if their calculator were superior to models without these issues (and even high prevalence of HIV/AIDS, to see if that mattered)
–but, I think they are on the right track overall. Studies have shown that these additional risk factors do confer increased risk, controlling for the traditional ones. And the applicability of the older risk calculators is limited: does the Framingham Risk Score really apply to non-white, non-working class patients? (we do know that the Framingham Risk Score does systematically overestimate risk in Japanese people, for example). How about the disabled poor Nicaraguan woman in front of me who has chronic kidney disease from the widespread indiscriminate use of the weed-killer glyphosate on the crops???? My guess is that including as many of the documented risk factors above (social class/poverty, renal failure,…) do add to the predictive model, even if not tested in the specific subgroup of such Nicaraguan women.
–and, another issue is the real utility of a 10-year risk calculator (which QRISK-3 does not challenge). one of the positives in the 2013 AHA/ACC risk calculator is that they did include a longterm risk calculator for 30-year/lifetime risk (not based on rigorous data, though). the obvious reality is that a prime target for hyperlipidemia intervention is people in their 40s, who have a pretty low likelihood of a cardiovascular event in the next 10 years.
–to me, the real positive of QRISK3 is that it challenges the traditional risk calculators, reinforcing the need to look at the individual patient in a more global way than a few medical components. And it would be great to have more studies which validate this in lots of different populations of different ages and prevailing comorbidities.