Use of the prevalence ratio v the prevalence odds ratio in view of confounding in cross sectional studies.