3/2007Statistical risk and risk of the real patient

MICHAŁ JAŚKIEWICZ (Katedra i Zakład Medycyny Rodzinnej Akademii Medycznej w Gdańsku; Kierownik: prof. ndzw. dr hab. med. Janusz Siebert)

Background
Based on prospective studies, some mathematical models were evolved, that describe relationships between risk factors of coronary heart disease and the risk of developing coronary incident. Although these models are supposed to describe the same phenomenon, the number and kind of used independent variables in each model is different. The credibility of these models is questioned.

Material and methods
SCORE, Framingham and UKPDS models were analyzed and used to calculate risk in virtual male population of 1000 subjects. The values of independent variables defining each patient were randomly generated within physiological and pathological range.

Results
Mean risk: SCORE 25.49%, Framingham 40.75%, UKPDS 51.17%. Mean difference between SCORE and Framingham 15.3 ± 18.6%; between UKPDS and SCORE 25.7 ± 24.2%; between UKPDS and Framingham 10.4 ± 17.1%. Results are incoherent. The maximum difference between Framingham and SCORE 72.4%; UKPDS and SCORE 90.4%; UKPDS and Framingham 68.8%.

Key words: risk, coronary disease, prediction