Background Radiographic damage is one of the key outcomes for research in rheumatoid arthritis (RA). The optimal method to read radiographs, i.e. blind or unblind to sequence, is unknown. Sequence-blinded reading may be less efficient, but unblinded reading may cause bias.
Objectives This study evaluated the difference between blinded and unblinded sequence reading in a series of radiographs with 11 years follow up. In addition, the influence of the starting point of the series was evaluated.
Methods Two experienced readers independently and repeatedly read digitized radiographs of 62 patients at time points 0, 2, 5, 8 and 11 years of follow-up from the COBRA follow-up database according to the Sharp/van der Heijde method (SvH). These films were selected based on quality and distribution of damage progression (results from previous readings) to optimally represent low or high damage at baseline, and low or high damage progression. Readers were aware of the study question, but unaware of the number of repeated readings; the order of the films was randomized for every repeated reading. A statistical linear mixed model was fitted to the data, including all main effects and all first order interactions between method of reading, length of a series (2-3 years or 8-9 years), incorporation of baseline, baseline damage, progression, time point, as well as two second order interactions. Efficiency was calculated with standard errors (SE) of progression rates between unblinded(1) and blinded(2) reading as: (SE1/SE2)2.
Results In this dataset mean yearly progression rate (averaged over all iterations) was 3,1. Compared to blinded sequence reading, unblinded reading resulted in a slightly but significantly increased progression rate of 0.43 points per year (p=0.008) and a lower standard error of the mean total progression rate of 0.67 (compared to 0.77 for blinded reading). Over 11 years, this results in a small difference in progression of about 5 points, but a highly relevant difference of 25% of patients needed in a study to find a difference in radiological outcome between treatment groups. Sensitivity analyses in the subset of short series that more closely resembles trial datasets (follow up 2-3 years) reveals a range of 10-55% increased efficiency for unblinded vs blinded reading. Knowledge of the baseline radiograph influences the results of the other readings. When the baseline radiograph is read, the estimation of the progression rate is significantly higher (2 points per year, p<0.001) than when the baseline radiograph is not read.
Conclusions Unless the assessment of repair is the aim of the study, unblinded reading of radiographs should be preferred above blinded reading, due to the high similarity of the numerical results even over a long follow up period, but decreased variability around the estimation of the progression rate. Increased efficiency translates into smaller sample sizes, or increased power to detect small differences. In addition, unblinded reading more closely resembles the routine clinical situation. For studies with long term follow up the same two readers should read all radiographs including baseline.
Disclosure of Interest None Declared