Introduction Attributable Fraction is the commonest method of describing the proportion of a health outcome attributable to an exposure in an exposed group. It applies to binary variables. Many variables are continuous; changing them to binary variables results in loss of information. Using traditional analyses we compared the importance of cardiovascular disease (CVD) risk factors in a continuous form.

Methods A population based cohort study including 1802 men and 2301 women aged ≥40 years in Tehran. We considered modifiable continuous exposures at baseline and CVD events during 8.5 years of follow-up. Using factor analysis we extracted some uncorrelated and standardised factors, each related to a cluster of continuous variables with the same general feature (eg, systolic and diastolic blood pressure as blood pressure factors or body mass index and waist circumference as anthropometric factor); then, a Cox regression including these factors as scores was conducted to estimate the RR of the last quintile to the first for each factor. Finally we compared these similar RRs in the model using the Wald test.

Results Anthropometric, blood glucose, blood pressure and cholesterol factors were extracted. The total variance explained by factors was 88.6% in men and 87.3% in women. In men all factors had the nearly the same RRs ranging from 1.7 to 2.2 but in women the RR of cholesterol was significantly higher than the others (3.4 vs 1.7–2.5).

Conclusion To prevent CVD, all clusters of risk factors should be considered in control programs. Hypercholesterolaemia maybe more important in women.