Real‐PET and CONSORT

The study presented by Whone and colleagues is methodologically and ethically flawed and should not have been published in a journal upholding CONSORT guidelines. [F]dopa PET scans were carried out at six different centres and the data analysed locally before central analysis. Prior to central analysis, the study result did not reach significance. Central analysis included the removal of 11% of the results because the first PET scan was normal. The result then reached significance. The timing of the decision to remove these results is not clear. We read that “ a decision was taken, prospectively, to include only patients with PET evidence of PD” but the published study design fails to mention this. The ethics of treating and rescanning 11% of patients, yet prospectively excluding their data from the trial, is questionable. There is, furthermore, no scientific validity in the decision. [F]dopa PET does not completely discriminate PD from normality at symptom onset . Figures for symptomatic threshold (75% of normal) and distribution in normal range (SD 14%) from work with 2D PET 6 give the positive and negative predictive value of PET scans. If the study population has 90% PD and 10% normal subjects, the negative predictive value (the chance of being disease free after a normal scan) is only 16%. 3D PET might be different but the information needed to make that calculation has not been provided. I question the decision to publish the paper in this form. Schulz, in his introduction to CONSORT, stated that analysis should include “all properly randomised participants in the originally assigned groups”. The data could have been published with and without the removal of the cohort of PD patients so that its effect could clearly be seen.