In the Foscarnet-Ganciclovir Cytomegalovirus (CMV) Retinitis Trial, time to first progression of newly diagnosed CMV retinitis was similar in the 2 treatment groups but was shorter when assessed by grading of fundus photographs at a central reading center than when assessed at the participating clinical centers. This report describes the extent and causes of this disagreement and considers the implications of the findings for clinical practice and future research.
Clinical findings and photographic gradings were compared for extent and activity of retinitis at baseline and during follow-up. In selected cases of disagreement, the photographs and summaries of gradings and clinical findings were reviewed concurrently to determine the cause of disagreement.
Movement of the border of retinitis was observed sooner and activity of the border was considered to have increased more often at the reading center than at the clinical centers. Disagreements on time to first progression were more frequent when degree of border movement was small (odds ratios [ORs] for several comparisons ranged from 1.7 to 5.2), when border activity was judged to have decreased or remained the same since the preceding visit (OR, 2.0-193), and when retinitis at baseline did not involve zone 1 (the area within 1 disc diameter of the disc or within 2 disc diameters of the center of the macula [OR, 1.4-3.6]). There were 2 important causes of disagreement between clinical center and reading center. First, difficulty was encountered clinically in recognizing retinitis border movement in the absence of an obvious increase in border activity. Second, the reading center used a threshold for border movement small enough to be crossed by an initial expansion of retinitis borders occurring within 2 to 5 weeks of enrollment in some patients who were responding favorably to treatment (in that retinitis was becoming inactive and showed no further progression for many weeks).
Comparisons of photographs from the current visit with those from several previous visits may increase clinicians' abilities to detect progression promptly. The use of additional outcome measures by reading centers, such as border movement of 1500 μm or more and change in area of retina involved by retinitis, may provide more accurate and useful comparisons of treatments. In making such comparisons, centralized photographic grading has the advantages of greater reproducibility and lesser risk of observer bias.