Printed in UJS.A. LETTER TO THE EDITOR DEFINITIONS OF ATTRIBUTABLE RISK

tion of attributable risk which is in essence: the proportion of the disease (in the entire population) attributable to exposure to the risk factor (2). Let / i = incidence in total population Jo = incidence in the nonexposed L = incidence in the exposed r = relative risk = 7e/70 6 = proportion of total population that has been exposed As Lilienfeld pointed out, Levin's definition, b(r- 1) b(r- 1) + 1 takes into account the proportion of the population exposed, whereas the definition employed by MacMahon and Pugh (3) does not. MacMahon and Pugh have defined attributable risk as the rate of disease in exposed individuals that can be attributed to the exposure (i.e., I, — Jo). On the other hand, the term population attributable risk ased by MacMahon and Pugh is the rate of disease in the entire population that can be attributed to the exposure (i.e., It — 70). The MacMahon and Pugh population at-tributable risk, when expressed as a propor-tion [(/ ( — /o)//J, however, is identical to what Levin meant by attributable risk. The following is intended to demonstrate this identity. The MacMahon and Pugh population at-tributable risk proportion is h-