Our results show that age screening loses little in screening performance compared with multiple risk factor measurement methods and, with appropriately priced preventive treatment, is less expensive. Offering preventive treatment to everyone above a specified age has the advantage of simplicity. It avoids needless worry that would be caused through selecting individuals on account of the results of a personal medical assessment.
Age screening avoids the costs and time spent in connection with the measurement and explanation of risk factor levels and avoids having to issue regular invitations for blood tests and medical examinations. With age screening people are not singled out as being at risk other than on account of their age, so those taking preventive treatment are less likely to feel “abnormal” or that they have become patients and possibly given a medical diagnosis. Age screening moves the emphasis from the assessment of risk to the reduction of risk.
In multiple risk factor screening, a 10-year CVD risk cut-off of 1 in 5 (20%) has been adopted by the UK government, and recommended by the National Institute for Health and Clinical Excellence 
. The cut-off is high for such devastating medical events as heart attacks and strokes, and does not offer preventive treatment to many people who would benefit. Over half the preventable CVD-free years of life lost would occur in people with a lower risk (see ). If this same risk were expressed as an individual having a 1 in 50 chance of a heart attack or stroke within the next 12 months, the seriousness of the situation would be more apparent and the individuals concerned would be better motivated to take steps to reduce the risk.
Preventive treatment has adverse effects but these are largely minor and reversible. With age screening a higher proportion of people are treated for the same number of cardiovascular events prevented but, given the extremely low incidence of serious adverse effects, the difference is not large enough to influence which screening method to adopt.
All methods of screening for CVD involve some people receiving preventive treatment without benefit because they die of another cause without having a CVD event, while others who would benefit do not receive treatment. For a given detection rate the proportions in these two categories are similar with Framingham screening and with screening using age alone. The prediction of CVD events using a Framingham risk assessment is relatively poor even though the Framingham equations were used to determine CVD events. This is because a Framingham risk only determines the probability of having a CVD event, not who actually has an event (eg. who, among 100 individuals with a 20% risk, are the 20 who have a CVD event).
The curves in show that it is similarly cost-effective to deliver a screening policy designed to achieve a person-years detection rate of about 75% (a detection rate of about 85%) as it is to deliver one designed to achieve a 45% person-years detection rate, in that the number of CVD-free years of life gained for a given expenditure is similar. There is little justification for screening using a Framingham-based 20% CVD risk cut-off that has a person-years detection rate of 45% and, provided the cost of treatment is not high, is less cost-effective than the alternative of age screening using a 50 or 55 year age cut-off (see ) which would achieve a detection rate of about 85%. Our results indicate that there is no practical justification for using different age cut-offs for men and women. The age cut-off, however, could be lower in people with diabetes; they have a high CVD risk and will already be aware of this.
The monetary costs we have used are illustrative and designed to provide an indication of the financial implications arising from the three methods of screening in relation to their efficacy. The costs will vary according to healthcare setting. Some costs associated with both methods have not been considered here, including the initial treatment consultation with a health professional, which does not affect the comparative costs. With Framingham screening, physicians may vary treatment according to the assessment results, and such treatment tailoring is likely to increase costs relative to age screening. Our analysis provides a reasonable indication of the relative cost-effectiveness of the two screening methods using illustrative unit costs.
We used the Framingham risk algorithm published in 1991 because it provides risk equations for the three cardiovascular outcomes we specified in this analysis (myocardial infarction, fatal coronary heart disease and stroke). Framingham risk equations published in 2008 
combine various cardiovascular outcomes, including, for example, angina and intermittent claudication 
. These added outcomes are less well predicted both by age screening and by multiple risk factor measurement and consequently the detection rate is about 10 percentage points less for a 20% false-positive rate (see figure S1 in appendix S1), but our conclusions regarding the similar screening performances of age, 5-yearly, and Framingham screening still apply. They are also likely to apply to other similar algorithms, for example the Reynolds risk score 
or QRISK2 
because screening performance with respect to the same clinical outcomes depends on the ranking of risk rather than the magnitude of risk. While the algorithms differ in estimating the magnitude of risk, there is little difference in the ranking of risk between individuals, mainly because in all the algorithms risk is dominated by age.
Screening performance was based on a population aged 0–89, but screening programmes would invite people aged about 40 or over for a risk assessment or simply for preventive treatment if about 50–55 or over. Using the whole population in estimating screening performance has several advantages. First, it standardizes the estimates of screening performance and avoids variation arising from the age range selected. Starting at age 40, as in gives similar detection rates at the same age or risk cut-off as but higher false-positive rates. Second, it means that all CVD events are included in the analysis to derive the estimates of screening performance, particularly CVD events in older people, in whom the disease is common and who stand to benefit considerably from preventive treatment.
A perceived limitation of this study is that it is based on statistical modelling and not on observed measurements from a cohort of individuals. The modelling is, however, based on observed data used to define the distributions of the risk factors in the population at large. The method is therefore no different from the modelling used in estimating the screening performance of, for example, Down's syndrome in pregnancy 
. Such data-derived modelling is the preferred method of estimating and comparing screening performance because it can be based on a large enough sample to provide the necessary precision, the sample genuinely represents the population at large, and there are no missing values, with complete ascertainment of clinical events. Nonetheless it would be desirable for the estimates to be independently validated against data from a cohort study.
Causal CVD risk factors, even in combination, are poor CVD screening tests 
. To achieve even a 50% detection rate for a 5% false-positive rate, a risk factor must have a relative risk across the top and bottom quintile groups of about 100 
. Combining the measurement of risk factors that individually have a poor screening performance has only a small effect in improving screening performance 
. Inappropriate emphasis on causal risk factors in CVD screening may have arisen from analyses of studies to identify causes of the disease, where the effect of age is deliberately minimized (eg. by age stratification) so that the effect of a causal risk factor is revealed. However, as we have shown, age may be, and in cardiovascuolar screening is, the dominant factor in determining risk so in assessing the value of a risk factor in screening the effect of age must be retained, and the impact of adding the risk factor to age in improving screening performance quantified.
European guidelines on the prevention of cardiovascular disease 
recommend that “global” risk of CVD should be used to determine who should receive preventive treatment. Age alone does this. Any age can be converted into a risk; for example in Britain, at age 50 the 10-year CVD risk is 2.8%, at age 55 it is 4.5% and at age 60 it is 7.1% 
; the risk doubles every 7.6 years, so a 90 year old person has a risk 240 times greater than a 30 year old.
In summary, CVD is common and serious. To have a major impact on its incidence a proactive cost-effective public health policy is needed. This should be designed to prevent most CVD events and should simplify access to preventive treatment without making people become patients. Age screening meets these objectives and warrants serious consideration, given its advantages over current methods of cardiovascular disease screening and prevention.