The effect of interobserver differences in post-implant prostate CT image interpretation on dosimetric parameters

The purpose of this study was to clarify where observers differ in their interpretation of CT scans, and to relate those differences to clinically relevant dosimetric parameters. Twenty unselected patients treated with I-125 or Pd-103 brachytherapy at the Veterans Affairs Puget Sound Health Care System (VAPSHCS) in 2001 were studied. Patients were implanted with I-125 (7 patients, 0.87 mCi/source) or Pd-103 (13 patients, 2.54 U/source). The number of I-125 sources implanted ranged from 52 to 78. The number of Pd-103 sources implanted ranged from 58-144. Post-implant 3 mm CT images were imported into a laptop running Varian Variseed trade mark sign and sent to the four physician investigators, who outlined the prostate independently. Investigators were not coached specifically for this study, beyond their having read prior reports regarding prostate volume determinations. There was moderate interobserver variability in CT volume determination, with the standard deviations as a percent of the mean ranging from 9% to 29% (median: 17%). An average of 14% of implants (range: 5%-20%) would have been judged inadequate based on a minimum V100 of 80%, versus 24% of implants (range: 5%-45%) being judged inadequate based on a minimum D90 of 90% of prescription dose. The greatest variability was seen in prostate length (median standard deviation: 0.57 cm), due to vagaries in base and apical localization. However, the prostatic width and thickness also varied substantially between observers, with median standard deviations of 0.24 and 0.32 cm, respectively. Treatment margin variability was greatest at the anterior border, with a median standard deviation of 0.21 cm±0.10. We believe that CT-based dosimetry, while influenced by CT interpretation, still provides useful general dosimetric calculations, that are likely to be reproducible enough to provide clinically useful information between institutions. The V100 and TMs are less influenced by interobserver CT interpretation variability than is the D90, and may be better suited as interinstitutional quality indices