The detection of simulated retinoblastoma by using red-reflex testing

To determine the ability to detect simulated retinoblastoma by using the red-reflex test.

Discs that simulated retinoblastoma lesions were affixed to the retina of model eyes with an 8- or 3-mm pupil. The diameter, height, and location of the discs varied. Five examiners evaluated the red reflex with direct ophthalmoscopy by using straight-on and oblique viewing. The generalized estimating equation was used to assess the effects of pupil dilation and observer viewing orientation on tumor detection.

Significant 3-way interactions between pupil dilation, observer orientation, and tumor diameter (P < .004) or height (P < .02) were detected; these relationships depended on tumor diameter and height. A similar 3-way interaction was found between pupil dilation, observer orientation, and tumor location in degrees from the fovea (P < .001). Oblique viewing and pupillary dilation improved the tumor-detection rate. With straight-on viewing, the degree of detection was <48% (95% confidence interval [CI]: 39%-57%) for even the largest lesions, compared with 96% (95% CI: 93%-98%) for oblique viewing. For peripheral lesions, the percentage detection for straight-on viewing was 35% (95% CI: 21%-50%) for 30 degrees from the fovea and 16% (95% CI: 2%-31%) for 60 degrees from the fovea; these detection rates significantly improved with oblique viewing to 70% or higher (P < .001).

Detection of simulated retinoblastoma was better when lesions were large and when oblique viewing and dilation were used. Peripheral location was negatively associated with detection. Red-reflex testing to detect leukocoria may be improved with oblique viewing and pharmacologic dilation.