Predicting MI - Risk factors or Disease detection?

Most patients presenting to the ED with an MI have at least 1 cardiovascular risk factor, contrary to the popular opinion that MIs come out of nowhere. This has been demonstrated in the MRFIT study (Greenland et al, JAMA, 2003), where >87% of those with fatal MI had exposure to at least one risk factor.

The Interheart study (Yusuf et al, Lancet, 2004) showed that this was the case in countries around the globe, in both sexes and across age groups. Collectively, the big nine risk factors accounted for 90% of the MI risk in men and 94% in women.

On the contrary, the predictive power of risk factors alone to determine future events in asymptomatic people is low. The likelihood ratio for CHD death with >1 risk factor is only 1.31 in males and 1.39 in females, based on the 3 studies cited by Greenland in the 2004 paper above (Weissler JAMA 2004 vol. 291 (3) pp. 299-300; author reply 300-1).

It's easy to see why - most people with risk factors don't end up with an MI.

For example, take cholesterol : MRFIT (Circulation 1991) showed a strong correlation between total cholesterol level and 10 year CHD death rate. Trouble is, 60% of all CHD deaths occur in people with total cholesterol levels below 240mg/dL (that's about 75% of the population).

From the AFCAPS/TexCAPS trial we saw that statin therapy in those with low CRP and LDLs under 150mg/dL did nothing, whereas treating those with LDLs under 150mg/dL and hsCRP >2mg/dL had a benefit. JUPITER was designed to explore this further, testing whether statin therapy would lower event rates in this patient cohort. This was a very positive trial in favor of lowering inflammation, raising the question of whether or not we should be routinely checking hsCRP in our patients.

Similar story exists for high blood pressure : in a meta-analysis of >1 million adults without vascular disease, published in the Lancet in 2002 by the Prospective Studies Collaboration, there was a strong and direct correlation between both systolic and diastolic blood pressure and CHD death. But same issue with prediction as with lipids : 80% overlap in diastolic BP distributions of those who die and don t die of IHD (Wald, Law BMJ 2003; 326:1419-23).

Where does this leave us?

  1. Add more risk factors into our scoring systems in an attempt to improve them? CRP is proving controversial ( in this respect.

Most of us likely don't actually use these algorithms anyway (Fuster, V editorial, Nature Reviews Cardiology August 2009).

  1. How about testing for subclinical disease? Calcium scoring, carotid IMT measurement, vascular reactivity etc. The SHAPE guidelines suggest this is more cost-effective than population risk factor modification because it's targeting the therapy to those most at risk (

The US Preventative Services Task Force report thinks not : (

  1. How about we just identify and treat the conventional risk factorsand not search for disease in those who are asymptomatic?

This debate has been raging for years - what's your view? Feel free to leave a comment!