Estimation of prevalence rate ratios for cross sectional data: an example in occupational epidemiology.

Sir,-A cross sectional or prevalence study is often used in an occupational setting to assess whether an association exists between exposure in the workplace and some physiological state where information on exposure and physiological state are obtained contemporaneously. If the physiological state is dichotomised as "normal" or "pathological" the data can either be analysed by stratification, standardisation,' or by multiple logistic regression.2 The last is an especially valuable statistical tool in that it allows statistical adjustment of several confounders as well as assessment of effect modification based on modest sample size. The drawback with logistic regression for cross sectional data is that the model estimates the prevalence odds ratio (POR) as effect measure. Under certain restricted conditions the POR approximates the incidence density ratio,34 which makes it (arguably) a useful effect measure for causal inference. Nevertheless, because prevalence data lack time dimension-they do not establish that cause antecedes effect5-the usefulness of POR as an indicator of aetiology may be illusory. In aetiological research, especially if the latent period (interval between exposure and occurrence of disease) is protracted and ill defined, a cross sectional study can only be used to