Intra- and interobserver variability of MRI prostate volume measurements

The Early Intervention Study, a placebo-controlled clinical trial, enrolled approximately 200 patients between age 45-65 years who had mild symptoms associated with early benign hyperplasia (BPH). Prostate volume was measured by magnetic resonance imaging (MRI) and read by the local radiologist, who was blinded to the treatment group but not to the sequence of MRI scans. In order to ascertain whether knowledge of the sequence of the MRI scans by the local radiologist was introducing bias in prostate volume changes, a single radiologist was selected to reread all MRI scans at the end of the study.

Each film was masked as to patient identity, study drug, and date. A new randomization schedule was prepared to blind the single reader to both sequence and treatment. The two sets of readings were compared.

Accuracy was dramatically improved when defined as percentage of patients with a 20% or more decrease in prostate volume, from 14% to 0% in placebo patients, and from 38% to 29% in finasteride patients. The variability was significantly reduced (by 50%) when read by a single observer [1].

A single reader, blinded to time sequence as well as to therapy, improves the accuracy and precision of the measurements when compared to multiple readers, blinded only to drug therapy.