We have examined the reliability of computed tomography (CT) in the assessment of patients with advanced ovarian cancer. Three radiologists independently read the same set of 140 CT scans. Fifty percent of the films were reread by the same radiologist to assess intraobserver agreement. In addition we assessed variability among radiologists in reporting signs of disease such as a mass and ascites and the variability in reporting a change in signs in response to therapy. Intraobserver agreement on the identification of signs such as a mass and ascites and on overall response to therapy was moderate to excellent (Kappa 0.52-0.84). Agreement among observers on the same signs was not as good (Kappa 0.36-0.79).