Purpose: To evaluate the utility of remote digital fundus imaging as compared to indirect ophthalmoscopy to screen for retinopathy of prematurity (ROP).
Methods: This was a prospective, multicenter, masked clinical trial. Infants <31 weeks gestational age and <1000 g at birth were eligible for enrollment. Eligible enrolled infants were screened for ROP employing serial fundus imaging followed by indirect ophthalmoscopy. The main outcome measures were diagnostic sensitivity, specificity, positive and negative predictive values, and accuracy of image interpretation compared to ophthalmoscopy.
Results: Fifty-one infants (102 eyes) meeting eligibility criteria were enrolled between February 2001 and February 2002. Mean weekly examinations per infant (+/-SD) were 5.73 +/- 3.22 (median 7; range 2-10). For the purposes of this study, the reading center established a definition of ROP seen on digital fundus images deemed sufficiently severe (termed clinically significant ROP, or CSROP) to warrant on-site examination by an ophthalmologist experienced in ROP. CSROP developed in 59 of 102 eyes (57.8%; 31 right eyes and 28 left eyes). Of the eyes with CSROP, 22% (13/59; 7 right eyes and 6 left eyes) progressed to ROP severe enough to require treatment according to the criteria of the Early Treatment for ROP Randomized Trial. Using onsite indirect ophthalmoscopic diagnosis as the reference standard, CSROP was identified by digital images with a sensitivity of 92% (94% right eyes and 89% left eyes) and specificity of 37.21% (40% right eyes and 35% left eyes), and Early Treatment for Retinopathy of Prematurity (ETROP) prethreshold Type I with a sensitivity of 92% (86% right eyes and 100% left eyes) and specificity of 67.39% (67% right eyes and 68% left eyes).
Conclusions: Remote interpretation of digital fundus images is a useful adjunct to conventional bedside ROP screening by indirect ophthalmoscopy. Diagnostic sensitivity in this study was excellent. It was highly unlikely that severe ROP would be missed when image quality was high. Differences between the two screening approaches in timing of diagnosis of CSROP and ETROP were not statistically significant. Remote digital fundus imaging as deployed in this study is unlikely to supplant bedside ophthalmoscopic examination due to limitations in diagnostic sensitivity, specificity, and accuracy when image quality is poor.