Investigating the effect of independent, blinded digital image assessment on the STOP GAP trial

Emily Patsko, Peter J. Godolphin, Kim S. Thomas, Trish Hepburn, Eleanor J. Mitchell, Fiona E. Craig, Philip M. Bath, Alan A. Montgomery

Blinding is the process of keeping treatment assignment hidden and is used to minimise the possibility of bias. Trials at high risk of bias have been shown to report larger treatment effects than low-risk studies. In dermatology, one popular method of blinding is to have independent outcome assessors who are unaware of treatment allocation assessing the endpoint using digital photographs. However, this can be complex, expensive and time-consuming. The objective of this study was to compare the effect of blinded and unblinded outcome assessment on the results of the STOP GAP trial. The STOP GAP trial compared prednisolone to ciclosporin in treating pyoderma gangrenosum. Participants' lesions were measured at baseline and at 6 weeks to calculate the primary outcome, speed of healing. Independent blinded assessors obtained measurements from digital photographs using specialist software. In addition, unblinded treating clinicians estimated lesion area by measuring length and width. The primary outcome was determined using blinded measurements where available, otherwise unblinded measurements were used (method referred to as trial measurements). In this study, agreement between the trial and unblinded measurements was determined using the intraclass correlation coefficient (ICC). The STOP GAP trial's primary analysis was repeated using unblinded measurements only. We introduced differential and nondifferential error in unblinded measurements and investigated the effect on the STOP GAP trial's primary analysis. Eighty-six (80%) of the 108 patients were assessed using digital images. Agreement between trial and unblinded measurements was excellent (ICC = 0.92 at baseline; 0.83 at 6 weeks). There was no evidence that the results of the trial primary analysis differed according to how the primary outcome was assessed (p value for homogeneity = 1.00). Blinded digital image assessment in the STOP GAP trial did not meaningfully alter trial conclusions compared with unblinded assessment. However, as the process brought added accuracy and credibility to the trial it was considered worthwhile. These findings question the usefulness of digital image assessment in a trial with an objective outcome and where bias is not expected to be excessive. Further research should investigate if there are alternative, less complex ways of incorporating blinding in clinical trials. Current Controlled Trials, www.isrctn.com ISRCTN35898459. Registered on 26 May 2009.