A comparison of quantitative and semi-quantitative methods for assessing cartilage status and change over time; data from the osteoarthritis initiative.

introductionThe aim of this study was to compare the cross-sectional relationship and longitudinal responsiveness of the semi-quantitative MRI Osteoarthritis Knee Score (MOAKS) with automated quantitative cartilage thickness measures.

methodsImages and MOAKS scores from 297 participants with evidence of radiographic progression (groups 1 and 2) from the OAI FNIH sub-cohort were included. To facilitate direct comparison, novel quantitative measures of cartilage loss (termed Q-MOAKS) were matched to MOAKS regions. Mean normative cartilage thickness was computed for each subregion using OAI non-OA controls. The Q-MOAKS thickness loss score was based on the proportion of cartilage thickness over a subregion that was < 95% normative thickness, the denudation score was based on < 5% normative thickness. Q-MOAKS area proportions were categorised into scores as for MOAKS. Quantitative cartilage thickness (ThCtAB) was also measured in MOAKS subregions. We compared MOAKS against Q-MOAKS and ThCtAB cross-sectionally using Spearman's rank correlation and descriptive statistics including proportions and boxplots. Longitudinally, responsiveness was assessed at 1 and 2 years using standardised response means (SRM).

resultsCross-sectionally, there was a poor correlation between MOAKS and Q-MOAKS thickness loss and denudation scores in all regions except central medial femur (cMF) and tibia (cMT) with moderate correlation for thickness loss scores: cMF, ρ = 0.59, (95%CI:0.51, 0.66) cMT, ρ = 0.58, (0.50, 0.65). In cMF, despite the broad range for the MOAKS thickness loss score = 2 (10-75% region surface area), only 56% (89/159) of knees were Q-MOAKS = 2 and 23% of MOAKS denudation = 2 were represented in Q-MOAKS = 2. In cMT, the results for similar comparisons were 61% and 66% respectively. MOAKS appeared to overestimate grades 2 and 3. Over 2-year follow-up MOAKS thickness loss and denudation scores were less responsive than Q-MOAKS in most subregions. MOAKS thickness loss was most responsive in cMT (SRM = 0.47, (95%CI:0.41, 0.54)). ThCtAB was substantially more responsive: SRM=-0.84, (-0.96, -0.73) in this region.

conclusionThough MOAKS status scores showed reasonable correlation with quantitative measures of thickness in medial compartments, concordance between MOAKS and quantitative cartilage area loss was poor. Quantitative measures of thickness loss were substantially more responsive then MOAKS scores over a 1 and 2-year period.