Impact of repeated measures of joint space width on the sample size calculation: An application to hip osteoarthritis

The aims of the study were to investigate the impact of aggregation of repeated readings (a) on minimising variability of joint space width (JSW) measurements based on calculation of the smallest detectable difference (SDD) and (b) on sample size calculation in a hip OA randomized controlled trial (RCT).

(a) Post-hoc analysis of 50 radiographs from a hip OA RCT (ECHODIAH). JSW reliability was calculated by the SDD through the aggregation of repeated readings of hip OA radiographs by an experienced rheumatologist. (b) The observed SDDs were applied to the real ECHODIAH data, to determine the post-hoc theoretical sample sizes.

(a) Although the intra-reader reliability did not improve, SDDs decreased from 0.75 mm to 0.27 mm when aggregating two or more consecutive readings. (b) A significant decrease of sample size was noticed until three consecutive paired JSW measurements, with a sample size ranging from 6588 patients per group (SDD=0.75 mm) to 377 patients (SDD=0.45 mm). However, an increase of the sample size was observed for smaller SDDs.

In hip OA, the aggregation of repeated radiograph readings leads to a significant decrease in SDD, although the intra-reader reliability of the experienced reader remains stable. The decrease in SDD allowed to reduce significantly the post-hoc calculation of sample size until a SDD value of 0.45 mm. For smaller cut-offs of radiological progression, the sample size increased again indicating that gain in sensitivity does not automatically lead to gain in sample size, particularly if the treatment effect is limited.