Influence of confounding factors on designs for dose-effect relationship estimates.