Between December 1998 and December 1999 twenty-one patients with painful knee arthroplasties underwent assessment by radionuclide arthrography. There were eleven female and ten male patients, with an average age of 60 years at the time of the index procedure. The index procedure was a primary total knee arthroplasty, primary medial unicompartmental knee arthroplasty and revision medial UKA in seventeen, three and one patients respectively. Nineteen arthroplasties were uncemented and two were hybrids. All patients had previously been investigated by clinical examination, serological testing, fluoroscopic AP and lateral radiographs, and Tc99 bone scan with equivocal results. The presence of radionuclide about the tibial stem was considered diagnostic of tibial loosening. Nine patients underwent revision knee arthroplasty. The presence of radionuclide about the tibial stem correctly predicted a loose tibial component in four of five cases. Similarly, the absence of radionuclide about the tibial stem correctly predicted a stable tibial component in four of four cases. In the single misdiagnosed case the tibial component did not have a large central stem, had focal osteolysis about several screws, but remained stable. Radionuclide arthrogram is a useful tool in the investigation of painful knee arthroplasty. Radionuclide about the tibial stem is the key predictor of tibial component loosening.