Using latent variables in logistic regression to reduce multicollinearity, A case-control example: breast cancer risk factors