TABLE 2.
Characteristics and key findings of studies included in this review.
| References | Study design | Sample size (patients/ eyes) | Imaging device | Primary outcome(s) | Key finding with effect size | Level of evidence* |
|---|---|---|---|---|---|---|
| Silva et al. (9) | Prospective cohort | 206 eyes | Optos® | Distribution of PPL and DR severity | Approximately one-third of H/Ma, IRMA, and NVE were predominantly outside ETDRS fields; UWF -CFP suggested more severe DR in 10% of eyes. | II |
| Silva et al. (37) | Prospective cohort | 200 eyes | Optos® | PPL and risk of DR progression/PDR development | Eyes with PPL had a 3.2-fold higher risk of DR progression and a 4.7-fold higher risk of progression to PDR; approximately half of DR eyes had PPL outside ETDRS 7-field | II |
| Aiello et al. (11) | Multicenter cross-sectional | 764 eyes | Optos® | Agreement between UWF-CFP and ETDRS 7-field | High agreement in DR grading; UWF-CFP detected more peripheral lesions: PPL present in 308 of 751 eyes (41.0%), suggesting increased DR severity by ≥2 steps in 34 eyes (11.0%). | III |
| Marcus et al. (29) | Prospective cohort | 544 eyes | Optos® | PPL and risk of DR progression | PPL associated with 1.7-fold increased risk of DR worsening over 4 years | II |
| Verma et al. (10) | Retrospective study | 1406 eyes | Optos® | Distribution of DR lesions (PPL vs. PCL) across ICDR severity grades in an Indian population | PPL distribution in 37% of eyes vs. PCL in 63% (P < 0.003); PPL frequency varied by ICDR grade: mild NPDR 30.9%, moderate NPDR 40.3%, severe NPDR 38.5%, PDR 34.9% (P = 0.005); temporal fields showed greatest PPL frequency across all grades | II |
| Silva et al. (38) | Retrospective study | 544 eyes | Optos® | Effect of DR lesion type, severity, and distribution on disease worsening (≥2-step ETDRS-SS worsening or DR treatment over 4 years) | More severe lesion grades outside ETDRS fields associated with greater risk of disease worsening: on UWF-CFP, H/Ma HR 1.74 (1.28–2.36); on UWFFA, H/Ma HR 1.90 (1.38–2.61), IRMA HR 1.68 (1.13–2.49), NVE HR 1.99 (1.36–2.93) | III |
| Stino et al. (30) | Cross-sectional | 75 eyes | Optos® + UWFFA | Correlation between IRMA and NPA | Number of IRMA regions highly correlated with NPA area; rs = 0.86 (P < 0.01) | III |
| Li et al. (34) | Prospective cohort | 153 eyes | Optos® + UWF SS-OCTA | UWF-CFP + UWF SS-OCTA vs. UWF-CFP + FFA for DR grading | Very good agreement in DR severity; κ = 0.869; combined non-invasive approach showed similar lesion detection rates to FFA (p > 0.05 for all, except NPA p = 0.039) | II |
| Cui et al. (35) | Prospective observational study | 152 eyes | Optos® + UWF SS-OCTA | UWF-CFP + UWF SS-OCTA vs. UWFFA for DR lesion detection rate | Combined UWF SS-OCTA + UWF-CFP achieved equivalent DR lesion detection rates to UWFFA (p > 0.05), with high DR grading agreement (κ = 0.916) | II |
| Zhao et al. (40) | Multicenter diagnostic | 59,475 images | Optos® | Deep learning model (WARM) for 25 fundus diseases | WARM superior to region-restricted model for peripheral/diffuse lesions | II |
| Deng et al. (43) | Cross-sectional/AI | 339 patients | Optos® | Quantitative vascular analysis (fractal dimension, density) | Vascular density and fractal dimension decreased with increasing DR severity; fractal dimension: 1.33 ± 0.05 vs. 1.41 ± 0.03; vascular density: 1.12 ± 0.44 vs. 2.09 ± 0.36 | III |
| Rezaei et al. (20) | Cross-sectional/AI | 8,524 images from BRSET + 5,164 images from mBRSET | Zeiss Clarus® | UWF-CFP + clinical data vs. images alone | Multimodal model improved accuracy by 2.1% and Kappa by 0.022 on primary dataset; external validation Kappa improvements of 0.065 and 0.048 | III |
| Meng et al. (54) | Cross-sectional/AI | 734,084 images | Conventional fundus camera | DeepDKD model for DKD prediction | DeepDKD AUC 0.842–0.906; image model significantly superior to metadata model (AUC 0.680); combined model AUC 0.847; prospective validation sensitivity: combined model 86.7% vs. image-only 89.8% vs. metadata-only 66.3%enabled non-invasive DKD screening | II |
| Vujosevic et al. (55) | Multicenter cross-sectional | 516 eyes | UWF-CFP (Optos®) + OCT + OCTA | DR severity and history of stroke | Each 1-step increase in DR severity associated with 47% higher stroke risk; OR 1.47 (95% CI 1.02–2.11, p = 0.044); middle capillary plexus perfusion density (p = 0.017) and vessel density (p = 0.010) most significantly associated with stroke history each 1-step increase in DR severity: OR for stroke = 1.47 (95% CI: 1.02–2.11) | III |
| Xu et al. (41) | Generative AI/cross-sectional | 270 paired UWF-CFP and UWF-FA images | Optos® | GAN-synthesized UWF-FA from UWF-CFP | UWFDR-GAN outperformed all comparison models: MS-SSIM 0.7214; PSNR 20.00; FID 77.48; IS 1.0123. Qualitatively superior reconstruction of neovascularization and non-perfusion areas, with preserved global vascular architecture. Demonstrated potential to reduce reliance on invasive fluorescein imaging synthetic FA aided DR grading without intravenous dye | III |
| Abitbol et al. (12) | Retrospective study | 224 images | Optos® | Deep learning for retinal vascular disease classification | Per-class performance: DR AUC 90.5%, accuracy 85.2%. Cross-device generalizability not validated high performance but cross-device generalizability not fully validated | III |
| Ploumi et al. (56) | Cross-sectional | 182 eyes | UWF-CFP | Association between PPL and systemic complications | PPL significantly associated with peripheral arterial disease (OR 11.36, p = 0.033), coronary artery disease (OR 5.86, p < 0.001), stroke (OR 10.11, p = 0.003), and diabetic nephropathy (OR 2.97, p = 0.016); eGFR decreased with PPL presence (β = −12.66, p = 0.037) | III |
*Level of evidence: II = prospective cohort/diagnostic accuracy study/prospective observational study; III = cross-sectional/retrospective/case-control study (non-randomized). Only original studies are listed; reviews and commentaries are excluded. This classification is intended as a qualitative guide and does not imply formal risk-of-bias assessment. OR, odds ratio; HR, hazard ratio; κ, Cohen’s kappa; r_s, Spearman’s rank correlation coefficient; AUC, area under the curve; CI, confidence interval. H/Ma, hemorrhages and/or microaneurysms; IRMA, intraretinal microvascular abnormalities; NVE, neovascularization elsewhere; NPA, non-perfusion area; PPL, predominantly peripheral lesions; PDR, proliferative diabetic retinopathy; NPDR, non-proliferative diabetic retinopathy; DKD, diabetic kidney disease; CAD, coronary artery disease; PAD, peripheral arterial disease; eGFR, estimated glomerular filtration rate.