Flood risk of levee protected areas stems from the possibility of levee failure. Regarding the basic product formula for the quantification of risk, the reliability of a levee stretch or system of levees, which is dependent on all failure relevant conditions and loadings, must be expressed numerically. Some of the properties (i.e. biological factors) believed to affect a levee's probability of failure significantly, cannot be measured on a metric scale, but can be easily expressed qualitatively. As logistic regression models are able to make use of quantitative as well as of qualitative parameters, the application of such multivariate, statistical models for the reliability analysis of river levees is presented and discussed in this paper. Focusing on the practical issues for the specific purpose, the basic approach is explained only briefly at the beginning. Illustrated by example applications, the requirements, advantages and disadvantages of the approach are presented in more detail. Considering the rapid developments in remote sensing and levee monitoring possibilities during the last years, an increasing application potential of this approach is expected in the future.