Quantitative exposure data in epidemiologic studies are frequently analyzed by ordered categories. Categorization by quantiles can be based on the distribution of (1) cases, (2) noncases, and (3) all subjects. The advantages of the three schemes in determining quantiles were compared. They were found to give the same power regarding a test for trend over a wide range of study situations. Considerations on the representativeness and the ease of the analysis implementation could dictate the choice of the categorization scheme.