When assessing changes in glomerular filtration rate (GFR) it is important to differentiate pathological change from intrinsic biological and analytical variation. GFR is measured using complex reference methods (e.g., iohexol clearance). In clinical practice measurement of creatinine and cystatin C are used in the Modification of Diet in Renal Disease [MDRD] or Chronic Kidney Disease Epidemiology Collaboration [CKD-EPI] equations to provide estimated GFR. Here we studied the biological variability of measured and estimated GFR in twenty nephrology outpatients (10 male, 10 female; median age 71, range 50-80 years) with moderate CKD (GFR 30-59 ml/min per 1.73 m2). Patients underwent weekly GFR measurement by iohexol clearance over four consecutive weeks. Simultaneously, GFR was estimated using the MDRD, CKD-EPIcreatinine, CKD-EPIcystatinC and CKD-EPIcreatinine+cystatinC equations. Within-subject biological variation expressed as a percentage [95% confidence interval] for the MDRD (5.0% [4.3-6.1]), CKD-EPIcreatinine (5.3% [4.5-6.4]), CKD-EPIcystatinC (5.3% [4.5-6.5]), and CKD-EPIcreatinine+cystatinC (5.0% [4.3-6.2]) equations were broadly equivalent. The within-subject biological variation for MDRD and CKD- EPIcreatinine+cystatinC estimated GFR were each significantly lower than that of the measured GFR (6.7% [5.6-8.2]). Reference change values, the point at which a true change in a biomarker in an individual can be inferred to have occurred with 95% probability were calculated. By the MDRD equation, positive and negative reference change values were 15.1% and 13.1% respectively. If an individual's baseline MDRD estimated GFR (ml/min per 1.73 m2) was 59, significant increases or decreases would be to values over 68 or under 51 respectively. Within-subject variability of estimated GFR was lower than measured GFR. Reference change values can be used to understand GFR changes in clinical practice. Thus, estimates of GFR are at least as reliable as measured GFR for monitoring patients over time.