Recent evidence suggests that pharmacological treatment may alter the rate of progression of cartilage damage in osteoarthritis (OA). However, a lack of accurate and precise noninvasive assessments of cartilage structure makes it difficult to answer this question directly with prospective clinical trials, prevents early diagnosis of OA and restricts assessment of treatment to evaluation of symptoms or joint function. It is important, therefore, to develop precise, noninvasive methods both for diagnosis of early OA before damage is extensive and irreversible and for evaluation of therapeutic effectiveness. Magnetic resonance imaging (MRI) allows for noninvasive, multiplanar body imaging which depends on proton density, flow, and the T1 and T2 relaxation times. Because these variables differ markedly among joint tissues, cartilage erosions are visible with MRI and it should be possible to quantify them. Our objective was to compare MRI with arthroscopy for assessing the depth of lesions in the articular cartilage of human knees to help develop and validate MRI for use in clinical trials designed to assess the effect of therapy on cartilage structure. In the first part of our study, the effect of the MRI pulse sequence variables on the images was evaluated by varying them systematically and comparing the anatomy seen with MRI with that seen at arthroscopy or arthrotomy and with the histology. In the second part, 31 patients were assessed with MRI before arthroscopy. The MRI were graded on a 4-point ordinal scale by 2 observers who were unaware of the clinical diagnosis and compared with findings at arthroscopy which were graded using the same scale.(ABSTRACT TRUNCATED AT 250 WORDS)