Background: Doubts have been expressed about the performance of the American College of Rheumatology (ACR) clinical classification criteria for osteoarthritis when applied in the general population.

Objective: To investigate whether the distribution of population subgroups and underlying disease severity might explain the performance of these criteria in the population setting.

Methods: Population-based cross-sectional study. 819 adults aged > or = 50 years reporting knee pain in the last 12 months were clinically assessed by research therapists using standardised protocols and blinded to radiographic status. All participants underwent plain radiography of the knees, scored by a single reader blinded to clinical status. The relationship between fulfilling the ACR clinical classification criteria for knee osteoarthritis and the presence of symptomatic radiographic knee osteoarthritis was summarised for the sample as a whole and within subgroups.

Results: Radiographic osteoarthritis was present in 539 participants (68%) and symptomatic radiographic knee osteoarthritis in 259 (33%). 238 participants (30%) fulfilled the ACR clinical criteria for knee osteoarthritis. Agreement between the ACR clinical criteria and symptomatic radiographic knee osteoarthritis was low (sensitivity 41%; specificity 75%; positive predictive value 44%; negative predictive value 72%). Sensitivity and specificity did not vary markedly between population subgroups, although they were influenced by the underlying severity of radiographic osteoarthritis.

Conclusion: The ACR clinical criteria seem to reflect later signs in advanced disease. Other approaches may be needed to identify early, mild osteoarthritis in the general population and primary care.