UCSF

Objective

Large studies of knee osteoarthritis (KOA) require well-characterized efficient methods to assess progression. We previously developed the local-area cartilage segmentation (LACS) software method, to measure cartilage volume on magnetic resonance imaging (MRI) scans. The present study further validates this method in a larger patient cohort and assesses predictive validity in a case-control study.

Method

The OA Biomarkers Consortium FNIH Project, a case-control study of KOA progression nested within the Osteoarthritis Initiative (OAI), includes 600 subjects in four subgroups based on radiographic and pain progression. Our software tool measured change in medial femoral cartilage volume in a central weight-bearing region. Different sized regions of cartilage were assessed to explore their sensitivity to change. The readings were performed on MRI scans at the baseline and 24-month visits. We used standardized response means (SRMs) for responsiveness and logistic regression for predictive validity.

Results

Cartilage volume change was associated strongly with radiographic progression (odds ratios (OR) = 4.66; 95% confidence intervals (CI) = 2.85-7.62). OR were significant but of lesser magnitude for the combined radiographic and pain progression outcome (OR = 1.70; 95% CI = 1.40-2.07). For the full 600 subjects, theSRM was -0.51 for the largest segmented area. Smaller areas of cartilage segmentation were also able to predict the case-control status. The average reader time for the largest area was less than 20 min per scan. Smaller areas could be assessed with less reader time.

Conclusion

We demonstrated that the LACS method is fast, responsive, and associated with radiographic and pain progression, and is appropriate for existing and future large studies of KOA.