Cardiovascular risk factors: are they useful screening tests? Malcolm Law Wolfson Institute of Preventive Medicine Barts and The London School of Medicine
Mortality in Britain Ischaemic heart disease 130,000 deaths per year Stroke 63,000 deaths per year Together one third of all deaths
Average values of risk factors: present day and prehistoric values at age 60 Present Western Prehistoric Proportion of present Western below prehistoric average Systolic BP (mmHg) 145 110 <1% Serum cholesterol (mmol/l) 6.0 3.2 <1% Plasma homocysteine (mmol/l) 13.4 9.7 <10% Body mass index (kg/m2) 27 22 <10% Law & Wald BMJ 2002;324:1570-6 Ubbink J Nutr 1996;126:1254S-7S
Blood pressure and IHD in a meta-analysis of cohort studies Relative risk Usual diastolic BP (mmHg) MacMahon et al Lancet 1990
Blood pressure and stroke in a meta-analysis of cohort studies Relative risk Diastolic BP (mmHg) MacMahon et al Lancet 1990
Serum cholesterol and ischaemic heart disease Did not die of IHD Died of IHD DR = 15% FPR = 5% Serum cholesterol (mmol/l) BUPA cohort study
Blood pressure and ischaemic heart disease Did not die of IHD Died of IHD DR = 13% FPR = 5% Diastolic blood pressure (mmHg) BUPA cohort study
Serum homocysteine and ischaemic heart disease Did not die of IHD Died of IHD DR = 13% FPR = 5% Serum homocysteine (mmol/L) (log scale) BUPA cohort study
Blood pressure and stroke Did not die of stroke Died of stroke DR = 24% FPR = 5% Diastolic blood pressure (mmHg) BUPA cohort study
Systolic blood pressure 28% Ischaemic heart disease Disease concentration in the 10% of the population with the most extreme risk factor values Proportion of cases in the most extreme 10% Stroke Systolic blood pressure 28% Ischaemic heart disease Systolic blood pressure 21% Serum cholesterol 21% Plasma homocysteine 20% Body mass index 22% Law & Wald BMJ 2002;324:1570-6
Proportion of events detected for a 5% FPR Combining risk factors to screen for ischaemic heart disease events Proportion of events detected for a 5% FPR apoB (or LDL cholesterol) 17% apoB and systolic blood pressure (SBP) 22% apoB, SBP, apoA1, apo(a) 24% apoB, SBP, apoA1, apo(a), smoking 27% 3 measures of apoB, SBP, apoA, apo(a) + smoking 28% 3 measures of apoB, SBP, apoA1, apo(a) + smoking, family history 29% Wald et al Lancet 1994;343:75-9
Why combining risk factors is less effective than one would think? LDL cholesterol detects 17% of cases for 5% FPR Systolic BP also detects 17% of cases for 5% FPR
Why combining risk factors is less effective than one would think? LDL cholesterol detects 17% of cases for 5% FPR Systolic BP also detects 17% of cases for 5% FPR What proportion would be detected if we used both?
Why combining risk factors is less effective than one would think? LDL cholesterol detects 17% of cases for 5% FPR Systolic BP also detects 17% of cases for 5% FPR What proportion would be detected if we used both? 17% + (17% of 83%) = 31%
Why combining risk factors is less effective than one would think? LDL cholesterol detects 17% of cases for 5% FPR Systolic BP also detects 17% of cases for 5% FPR What proportion would be detected if we used both? 17% + (17% of 83%) = 31% But the FPR would no longer be 5% - it would be nearly 10%
Why combining risk factors is less effective than one would think? LDL cholesterol detects 17% of cases for 5% FPR Systolic BP also detects 17% of cases for 5% FPR What proportion would be detected if we used both? 17% + (17% of 83%) = 31% But the FPR would no longer be 5% - it would be nearly 10% This is what is often overlooked Keeping FPR at 5%, detect 22% instead of 17% for one alone
If you found a group with a very high risk few people are in it The paradox of risk If you found a group with a very high risk few people are in it so one misses most of the cases
Example Healthy man age 60 who: smokes cigarettes cholesterol in top 5% (7.6mmo/L) blood pressure in top 5% (175 mmHg systolic) Absolute risk of MI or stroke = 9% per year (12 fold increase) But prevalence of such men is 6 per 10,000 [Five GPs between them would have one]
The paradox of risk Annual risk of IHD event or stroke in men aged 60 Risk Detection Positive cut-off rate rate 0.25% 99% 98%
The paradox of risk Annual risk of IHD event or stroke in men aged 60 Risk Detection Positive cut-off rate rate 0.25% 99% 98% 0.5% 88% 73%
The paradox of risk Annual risk of IHD event or stroke in men aged 60 Risk Detection Positive cut-off rate rate 0.25% 99% 98% 0.5% 88% 73% 1% 52% 28%
The paradox of risk Annual risk of IHD event or stroke in men aged 60 Risk Detection Positive cut-off rate rate 0.25% 99% 98% 0.5% 88% 73% 1% 52% 28% 2% 16% 5%
The paradox of risk Annual risk of IHD event or stroke in men aged 60 Risk Detection Positive cut-off rate rate 0.25% 99% 98% 0.5% 88% 73% 1% 52% 28% 2% 16% 5% 4% 1% 0%
Groups with an annual risk of heart attack or stroke of 5% or more event rate (fatal or not) annual death rate Previous myocardial infarction 5% 10% Previous stroke 5% 10% Angina without MI 3% 6% Transient ischaemic attacks without stroke 2% 5% Law at al Arch Intern Med 2002;162:2405-10
Asking for a history of any of these disorders is a screening enquiry It has a detection rate of about 50% ond a low FPR
Proportion with such a history About half of deaths from heart disease and stroke occur in people with a history of a first non-fatal event Proportion with such a history MI stroke Men aged 60 7% 3% 70 15% 7% Women aged 60 4% 2% 70 5% 3% Health Survey for England
In people with no history of disease age alone is the best screening test age cut-off 55 years - detection rate 94% positive rate = 23%
Conclusions Cardiovascular risk factors are important aetiologically Lowering them can greatly reduce risk But they are poor screening tests - they do not usefully discriminate between people who will and will not have an ischaemic heart disease event or stroke The best screening test is the presence of existing disease. In people without existing disease it is age