Objective: To determine in the knee which individual radiographic feat
ure or combination of features in the patellofemoral and tibiofemoral
joints correlate best with a nonradiographic definition of clinical os
teoarthritis in order to recommend a definition of radiographic osteoa
rthritis for use in studies. Methods: Using data from the Framingham O
steoarthritis Study, we tested the correlation of clinical OA, defined
as frequent knee pain plus crepitus, with a variety of definitions of
radiographic OA including those based on individual radiographic feat
ures, e.g. greater than or equal to grade 2 osteophyte 0-3 scale, and
new definitions that included alternative combinations of features, [e
.g. either greater than or equal to grade 2 osteophyte or joint space
narrowing greater than or equal to grade 2 (0-3 scale) with a bony fea
ture (such as cyst, sclerosis, or grade 1 osteophyte)]. We performed a
nalyses looking at participants who had obtained both weight-bearing a
nteroposterior (AP) and lateral radiographs of both knees. Results: In
519 participants, we found that the definitions of radiographic osteo
arthritis best correlated with clinical OA were 'definite osteophyte g
reater than or equal to grade 2' (efficiency 62.4-67.1%) and an 'alter
nate definition' of either osteophytes greater than or equal to grade
2 or joint space narrowing greater than or equal to grade 2 with a bon
y feature of OA (efficiency 62.8-68.1%). A recursive partitioning anal
ysis selected the 'alternate definition' as best. Also, we found that
adding lateral views to the AP view improved the diagnostic test perfo
rmance of the best performing radiographic definitions. Conclusion: We
suggest that a knee should be characterized as having radiographic OA
if there is either an osteophyte of grade 2 or greater severity (0-3
scale) present or the presence of moderate to severe joint space narro
wing (greater than or equal to 2 on a 0-3 scale) with co-occurrence of
a bony feature in the compartment affected.