Robustness and generalizability of smallest detectable difference in radiological progression

Abstract

The smallest detectable difference (SDD) reflects that component of a measure statistically attributable to error from the measurement process itself. As such it is an irreducible component of the inherent variability in measurements in clinical trials and will affect their design, whether randomized or observational. Even though the application of the SDD concept to assaying radiographs in rheumatoid arthritis is relatively new and not well understood, systematic work on the influences of radiographic SDD can be done. This report describes the effects of a number of clinical aspects of the disease and operational aspects of trials on the values of the SDD of radiographic progression data. We show that if conditions affecting SDD are known and kept constant across datasets, the SDD of radiological progression from one study may be generalizable to other studies. However, if any one condition varies, the SDD is distinctly unrobust and cannot be generalized to other studies.