Background The design of clinical trials for osteoarthritis is challenging; structural changes in tissues are quantitatively small and proceed very slowly. No clear guidance exists on how to optimise recruitment. KL grade is a poor recruitment criterion as centres interpret KL differently. Quantitative measures should be better, and metric radiographic joint space width (rJSW) is related to subsequent risk of radiographic progression. Although new MRI measures provide increased responsiveness in DMOAD trials, it is unknown whether selecting for recruitment based on radiographic criteria are well suited for responsiveness of these new measures.

Objectives (1) To determine which baseline rJSW values are associated with most subsequent progression for rJSW, MRI cartilage and bone outcomes (2) Explore baseline covariates that influence progression rates (3) Estimate the trial numbers needed using the criteria determined by steps (1) and (2).

Methods We used all knees from the Osteoarthritis Initiative which had all 3 measures recorded (rJSW – Duryea method; MRI cartilage thickness & bone area, Imorphics) at baseline, 1 and 2 years. We categorised knees into bins of 1mm rJSW, and assessed the 2 year changes of each bin, and characterised the distribution of rJSW in KL 0 knees. We used ANCOVA models to consider which covariates (including gender, height, weight, alignment, age, pain severity) affected 2-year slope of change, and responsiveness using SRM. For the final optimised recruitment groups, we calculated SRMs (CIs assessed using the bootstrap method of Efron) and derived the number of patients per arm in a putative trial.

Results 4796 knees were included (2789 females, mean age 61.45). The lower 95th percentile values for rJSW in women and men were 3.9 and 4.5mm respectively. The mean changes at 2-years for all 3 outcomes were greatest for the categories of 2–3 and 3–4mm baseline rJSW (Figure 1A) with notably little change in knees with rJSW<2mm. Of the covariates, only pain improved responsiveness. Using a total WOMAC pain criterion; ≥3/20 reduced numbers from 726 knees to 331 knees, and improved 1 year SRM (95% CI) from 0.27 (0.17,0.34) to 0.41 (0.29,0.51) in rJSW, from 0.45 (0.37,0.51) to 0.55 (0.45,0.65) in MRI cartilage and from 0.60 (0.52,0.66) to 0.73 (0.62,0.83) in MRI bone. Figure 1B shows the relative SRMs for the 3 outcomes based on 2 inclusion criteria (rJSW 2–4mm and pain ≥3/20; n=331) and demonstrates the required trial numbers (with confidence intervals) based on the SRMs.

• Download figure
• Open in new tab
• Download powerpoint

Conclusions Selecting patients based on 2 simple criteria will improve responsiveness in clinical trials for all 3 imaging outcomes using standard imaging outcomes. Selecting for rJSW of 2–4mm is most important while adding a pain criteria further improves responsiveness; no other covariates improved this. Caution should be applied when using SRM to power a study because of the inherent difficulties in calculating standard deviations; Fig 1B shows for example that the confidence limits for rJSW at 12 months vary from 302 to 1489. This analysis also confirms the advantages of MRI outcomes over rJSW in terms of study size and duration: a 12-month study with cartilage thickness or bone area endpoints needs no more than 238 or 137 patients (including the upper 95th percentile confidence limit).

Disclosure of Interest M. Bowes Employee of: Imorphics Ltd, G. Guillard Employee of: Imorphics Ltd, A. Brett Employee of: Imorphics Ltd, G. Vincent Employee of: Imorphics Ltd, P. Conaghan: None declared