Local staging of prostate cancer using magnetic resonance imaging: a meta-analysis

Our objective was to determine the influence of patient-, study design-, and imaging protocol characteristics on staging performance of MR imaging in prostate cancer. In an electronic literature search and review of bibliographies (January 1984 to May 2000) the articles selected included data on sensitivity and specificity for local staging. Subgroup analyses examined the influence of age, prostate specific antigen, tumor grade, hormonal pre-treatment, stage distribution, publication year, department of origin, verification bias, time between biopsy and MR imaging; consensus reading, study design, consecutive patients, sample size, histology preparation, imaging planes, fast spin echo, fat suppression, endorectal coil, field strength, resolution, glucagon, contrast agents, MR spectroscopy, and dynamic contrast-enhanced MRI. Seventy-one articles and five abstracts were included, yielding 146 studies. Missing values were highly prevalent for patient characteristics and study design. Publication year, sample size, histologic gold standard, number of imaging planes, turbo spin echo, endorectal coil, and contrast agents influenced staging performance (p=0.05). Due to poor reporting it was not possible to fully explain the heterogeneity of performance presented in the literature. Our results suggest that turbo spin echo, endorectal coil, and multiple imaging planes improve staging performance. Studies with small sample sizes may result in higher staging performance.