To determine the influence of biological variation on the reliability of data from different types of urine specimens, we measured nine analytes in first-morning, randomly collected, and 24-h samples of urine from 53 healthy individuals (14 men and 39 women). The urines were collected once a week for 10 weeks. The data obtained were used as a basis for specimen collection and to gain insight into the influence of urine quantities in the diagnosis, screening, and monitoring of patients. We found that 24-h urine expressed in output rather than concentration units is the most reliable specimen for diagnosis and monitoring for most of the analytes studied. On the basis of the ratio between estimated within- and between-subject variation, the tests with greatest medical usefulness for diagnosis and screening of specific pathologies are those measuring protein and sodium. Moreover, the results indicate that urine creatinine may be a poor test for diagnosis, monitoring, and screening.