In mydriatic imaging, sensitivity was also higher when ungradable images were excluded (92%, 87–95) compared to treating them as positive (85%, 70–93), while specificity was slightly lower (89 vs 95%), with a borderline significant difference (p = 0.05).
Mydriatic imaging (STDR) In a meta-analysis of six studies (623 cases among 2891 screened individuals, overall prevalence 21.5%; Figure 4 , Supplemental Figure 3 , supplementary Table 2 ), direct comparisons between non-mydriatic and mydriatic fundus imaging for the detection of sight-threatening diabetic retinopathy (STDR) showed a trend toward higher sensitivity with mydriatic imaging (70% [95% CI: 49–86] vs. 78% [95% CI: 58–90]) and significantly higher specificity (90% [95% CI: 79–95] vs. 94% [95% CI: 87–97]).
Trained graders demonstrated slightly higher sensitivity, but lower specificity compared to ophthalmologists, though these differences did not reach statistical significance.