PREVALENCE ODDS RATIO OR PREVALENCE RATIO IN THE ANALYSIS OF CROSS-SECTIONAL DATA - WHAT IS TO BE DONE

Objectives-To review the appropriateness of the prevalence odds ratio (POR) and the prevalence ratio (PR) as effect measures in the analysis of cross sectional data and to evaluate different models for the mult ivariate estimation of the PR. Methods-A system of linear differential equations corresponding to a dynamic model of a cohort with a chronic disease was developed. At any point in time, a cross sectional analys is of the people then in the cohort provided a prevalence based measur e of the effect of exposure on disease. This formed the basis for expl oring the relations between the FOR, the PR, and the incidence rate ra tio (IRR). Examples illustrate relations for various IRRs, prevalences , and differential exodus rates. Multivariate point and interval estim ation of the PR by logistic regression is illustrated and compared wit h the results from proportional hazards regression (PH) and generalise d linear modelling (GLM). Results-The FOR is difficult to interpret wi thout making restrictive assumptions and the FOR and PR may lead to di fferent conclusions with regard to confounding and effect modification . The PR is always conservative relative to the IRR and, if PR>1, the POR is always >PR. In a fixed cohort and with an adverse exposure, the FOR is always greater than or equal to IRR, but in a dynamic cohort w ith sufficient underlying follow up the FOR may overestimate or undere stimate the IRR, depending on the duration of follow up. Logistic regr ession models provide point and interval estimates of the PR land FOR) but may be intractable in the presence of many covariates. Proportion al hazards and generalised linear models provide statistical methods d irected specifically at the PR, but the interval estimation in the cas e of PH is conservative and the GLM. procedure may require constrained estimation. Conclusions-The PR is conservative, consistent, and inter pretable relative to the IRR and should be used in preference to the F OR. Multivariate estimation of the PR should be executed by means of g eneralised Linear models or, conservatively, by proportional hazards r egression.