TY - JOUR

T1 - Estimating the population attributable risk for multiple risk factors using case-control data

AU - Bruzzi, P.

AU - Green, S. B.

AU - Byar, D.

AU - Brinton, L. A.

AU - Schairer, C.

PY - 1985

Y1 - 1985

N2 - A straightforward and unified approach is presented for the calculation of the population attributable risk per cent (etiologic fraction) in the general multivariate setting, with emphasis on using data from case-control studies. The summary attributable risk for multiple factors can be estimated, with or without adjustment for other (confounding) risk factors. The relation of this approach to procedures in the literature is discussed. Given values of the relative risks for various combinations of factors, all that is required is the distribution of these factors among the cases only. The required information can often be estimated solely from case-control data, and in some situations relative risk estimates from one population can be applied to calculation of attributable risk for another population. The authors emphasize the benefits to be obtained from logistic regression models, so that risks need not be estimated separately in a large number of strata, some of which may contain inadequate numbers of individuals. This approach allows incorporation of important interactions between factors, but does not require that all possible interactions be included. The approach is illustrated with data on four risk factors from a pair-matched case-control study of participants in a multicenter breast cancer screening project.

AB - A straightforward and unified approach is presented for the calculation of the population attributable risk per cent (etiologic fraction) in the general multivariate setting, with emphasis on using data from case-control studies. The summary attributable risk for multiple factors can be estimated, with or without adjustment for other (confounding) risk factors. The relation of this approach to procedures in the literature is discussed. Given values of the relative risks for various combinations of factors, all that is required is the distribution of these factors among the cases only. The required information can often be estimated solely from case-control data, and in some situations relative risk estimates from one population can be applied to calculation of attributable risk for another population. The authors emphasize the benefits to be obtained from logistic regression models, so that risks need not be estimated separately in a large number of strata, some of which may contain inadequate numbers of individuals. This approach allows incorporation of important interactions between factors, but does not require that all possible interactions be included. The approach is illustrated with data on four risk factors from a pair-matched case-control study of participants in a multicenter breast cancer screening project.

UR - http://www.scopus.com/inward/record.url?scp=0022353142&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0022353142&partnerID=8YFLogxK

M3 - Article

C2 - 4050778

AN - SCOPUS:0022353142

VL - 122

SP - 904

EP - 914

JO - American Journal of Epidemiology

JF - American Journal of Epidemiology

SN - 0002-9262

IS - 5

ER -