Risk prediction, risk management, and risk avoidance are in fashion and invading medical practice, but current terminology is unfortunate and tendentious in a clinical context. Gambling and life insurance led to the first mathematics of risk. Individual outcomes were unpredictable, but if the circumstances and rules of the game did not change, the long run odds could be estimated from previous experience.

Longitudinal studies, of which the Framingham study was a pioneer, identified human risk factors which, when measured at the start of observation, correlated with the subsequently recorded risk. Risk is recorded as an event rate having numerator, denominator, and time period, but problems arose in how to portray different levels of risk at different levels of risk factors.

For half a century relative risk was king. Dividing high risk by low produced a simple multiple with no units—relative risk of smoking = 2.5. Relative risk had the advantage of consistency across populations differing in their overall levels of risk. Being a ratio, however, it can be paradoxical, highest when actual risk is very low—for example, in young versus older smokers. The alternative to division was subtraction. The risk difference (18 per thousand a year) was no longer relative.

Measured in real units of incidence it was therefore labelled “attributable” or “absolute” risk because in the language of logic “absolute” is the opposite of “relative.” Absolute risk varied so much by age, sex, and other risk factors that it did not generalise well across populations. Risk scores from one population predicted relative risk correctly in another while under or overestimating absolute risks.

Measures of absolute risk were so unfamiliar to doctors and others as incidence rates that many were incapable of recognising a small versus a large risk. Even high incidence rates are considerably smaller than one. Relative risk is more impressive as, for a harmful risk factor, it must be significantly greater than one. For medical education and patient motivation relative risk therefore reigned supreme, although modified in one scoring system to inform patients of their ranking within their own age and sex group. Relative risk was applied to single risk factors, but also to multifactorial risk scores where several were combined.

Relative risk, used in clinical practice to motivate patients to change their lifestyle, is now being supplanted by absolute risk, because the latter should determine clinical action. Estimates of real risk should be used in deciding on treatments which have fixed costs and risks associated with them such as antihypertensive and lipid lowering drugs. Age, sex, and the presence or level of relevant factors are used in multifactorial or global estimates of real risk to see whether the risk threshold for intervention has been crossed. Reality has replaced relativity, but at the cost of adopting the term “absolute,” with inappropriate connotations for practice.

Risk can be measured with hindsight and precision in large populations and arguably, therefore, “absolutely” in a retrospective context. Prospective prediction of future risk, however, using historical data, can be only an estimate because populations, circumstances, and rules of the game are continuously changing.

For an individual there are always unmeasured and unknown factors and the outcome anyway can be only yes or no, so predictions cannot be individually validated. Risk varies far more across different populations than can be accounted for by the risk factors incorporated into conventional scores such as that from the Framingham study, so that this score must be wrong much of the time, as it is in characterising risk by age and sex outside middle age.

Such criticisms do not matter if the Framingham score is the best on offer, and its use improves the efficiency of prescribing and helps to prevent cardiovascular disease. They do matter, however, if the term “absolute” risk is misunderstood. It suggests absolute power of prediction and a single, infallible answer, although “absolute” risk varies with what factors are included, what score is used, and what outcomes are being predicted. It therefore has misleading medicolegal and ethical implications.

For these reasons “absolute” risk is a misnomer when used for clinical prediction and patient management calculated from multifactorial scores. It should not be used prospectively.

It can be renamed simply as “the” risk, or, if risk is to be qualified, “quantitative” risk. Calculation of an individual's predicted or projected risk could be called “quantitative risk estimation” and the individual result a “quantitative risk estimate.” The associated acronym, QRE, nearly homonymous with “query,” may remind those using it that, although prognosis is relevant to management, it is hazardous to prophesy and foolhardy to prophesy absolutely.