Background: dehydration is common in elderly patients, but difficult to detect, by either an initial assessment of fluid balance or by monitoring fluid balance over time. In clinical chemistry, population-based reference ranges arc only of value for monitoring individual patients if within-subject variability in serial test results over time is larger than between-subject variability. This may have important implications in monitoring fluid balance. Ainu to assess the within-and between-subject variability of serial laboratory test results in euvolaemia and their responsiveness to dehydration in elderly patients. Methods: over 16 months, 218 patients were admitted to the geriatric department and 53 consented to participate. Fluid balance was assessed twice a week by physical examination, laboratory tests and weighing. Changes in fluid balance were quantified by measuring total body and extracellular water applying deuterium-and bromide-dilution techniques. Within-and between-subject variability in euvolaemia and responsiveness indexes (RI) for dehydration were calculated for haematocrit, serum sodium, urea and creatinine concentrations and for the urea/creatinine ratio. Results: during hospitalization 14 patients suffered from dehydration and 27 remained euvolaemic. Data from 12 overhydrated patients were excluded. In a mean study period of 30 days, each patient's fluid balance was assessed 6.3 (1.9) times. This resulted in 1084 laboratory tests and 271 assessments of fluid balance. In all quantities withinsubject variability was much smaller than between-subject variability in euvolaemia. Responsiveness of creatinine (mean RI = 2.5) was best and similar to the RI of serial weights (mean RI = 2.9). Conclusion: population-based reference ranges are of limited value in monitoring fluid balance. Repeatedly measuring plasma creatinine, combined with physical examination and weighing is the best way to monitor fluid balance in elderly patients.