Article Text

THU0447 Calculation of the smallest detectable change (SDC) of radiographic progression in clinical trials with multiple time points: The mean SDC is a valid approximation of the overall SDC
  1. V. Navarro-Compán1,
  2. D. van der Heijde1,
  3. H. Ahmad2,
  4. C. Miller2,
  5. R. Landewé3
  1. 1Rheumatology, LUMC, Leiden, Netherlands
  2. 2Medical Affairs, BioClinica, Newtown, United States
  3. 3Rheumatology, AMC, Amsterdam, Netherlands


Background SDC is an important statistic in evaluating radiographic progression in clinical trials in Rheumatoid Arthritis (RA). Two methods to estimate the SDC are available: one is based on the standard deviation (SD) of the difference between change-scores obtained by two readers (according to Bland&Altman (B&A)) and the other is based on generalisability analysis (using ANOVA). ANOVA is required in the case of multiple time points and multiple readers, but is more complicated and time consuming.

Objectives To evaluate if the average SDC of all available time intervals using the B&A method is a valid surrogate for the ANOVA method to estimate the SDC of radiological progression in RA studies including multiple time points.

Methods Fifteen datasets from radiographic trials in RA comprised of 20,098 radiographic time points from 7,643 patients were scored by 13 readers as pairs. Scoring was performed by each reader independently according to the modified Sharp van der Heijde method (SHS)1 with the reader blinded to chronological sequence. The number of time points per study was 2 in 2 trials, 3 in 10 trials, and 4 in 3 trials. According to the B&A method, the datasets were evaluated to calculate the 95% limits of agreement SDC [(± 1.96*SDΔ (changescore))/($√ $2*$√ $K)], where K represents the number of readers, for all intervals between different time points per study and its average. Using the ANOVA method, calculations of the SDC per study were performed considering all time points. Statistical analysis was performed using SPSS software version 18.0.

Results Median (range) baseline- and progression score was 32 (18-48) and 0.9 (0.5-1.8) units, respectively. Table 1 shows the results for all SDCs using both methods. The average SDC of all intervals calculated with the B&A method was very similar to SDC obtained by ANOVA. The mean difference between both methods was -0.13 (SD 0.28, range -0.48-0.25) units.

Conclusions We propose to report the mean interval SDC as a good surrogate for the true SDC in trials with multiple time points.

  1. van der Heijde DM. Lancet. 1989 May 13;1(8646):1036-8.

Disclosure of Interest None Declared

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.