Table 2. Multivariable Analysis of Primary Outcome: DR Worsening by 2 or More Steps Within the ETDRS Fields on UWF-Color Images or Receipt of DR Treatment Through 4 Years by Baseline NPI, FA PPL, and Participant Factors.
| Factor | Multivariable analysisa | |
|---|---|---|
| Hazard ratio (95% CI) | P value | |
| NPI with adjustment for participant risk factors | ||
| Overall NPI: per 0.1-unit increase | 1.14 (1.04-1.24) | .003 |
| DRSS level within ETDRS fields | NA | .15 |
| Age: per 10-y increase | 0.87 (0.73-1.04) | .13 |
| Diabetes type: type 2 vs type 1 | 1.36 (0.67-2.74) | .39 |
| Duration of diabetes: per 10-y increase | 0.90 (0.76-1.07) | .22 |
| Hemoglobin A1c: per 1% increase | 1.12 (1.02-1.23) | .01 |
| Albumin to creatinine ratio: albuminuria vs no albuminuriab | 1.88 (1.16-3.04) | .01 |
| NPI and FA PPL with adjustment for participant risk factors | ||
| Overall NPI: per 0.1-unit increase | 1.11 (1.02-1.22) | .02 |
| FA PPL present vs FA PPL absent | 1.89 (1.35-2.63) | <.001 |
| DRSS score within ETDRS fields | NA | .43 |
| Age: per 10-y increase | 0.86 (0.72-1.02) | .09 |
| Diabetes type: type 2 vs type 1 | 1.54 (0.75-3.15) | .24 |
| Duration of diabetes: per 10-y increase | 0.86 (0.72-1.03) | .09 |
| Hemoglobin A1c: per 1% increase | 1.13 (1.03-1.24) | .008 |
| Albumin to creatinine ratio: albuminuria vs no albuminuriab | 1.80 (1.11-2.92) | .02 |
Abbreviations: DR, diabetic retinopathy; DRSS, Diabetic Retinopathy Severity Scale; ETDRS, Early Treatment Diabetic Retinopathy Study; FA, fluorescein angiography; NA, not applicable; NPI, nonperfusion index; PPL, predominantly peripheral lesion; UWF, ultra-widefield.
Eyes receiving treatment with anti–vascular endothelial growth factor, steroid, or vitrectomy for conditions other than DR were censored after treatment initiation. Participant-level risk factors that either were significant in the univariable analysis (in eTable 6 in Supplement 1) or had ≥10% difference in the Kaplan-Meier estimates of the outcome were included as additional risk factors in the multivariable analysis to account for potential confounding. Hazard ratios and P values were obtained from a Cox proportional hazards model with additional adjustment for percentage of total imaged area on UWF-FA and the correlation between the 2 study eyes from the same participant.
The unknown group was included in the analysis, but no comparison was performed of this subgroup to the subgroups with known values. Hazard ratio and P value were provided only for comparing the 2 groups with known values (ie, albuminuria vs no albuminuria).