To evaluate the association of the serum factors with the severity of intraretinal diabetic retinopathy, patients were divided into two groups, depending on the severity. The patients with severe nonproliferative diabetic retinopathy (level 53), as assessed with the ETDRS-modified Airlie House grading system, had to have at least one of the following: four stereo fundus photographic fields with severe hemorrhages and microaneurysms, two fields with at least definite venous beading, or one field with at least moderate intraretinal microvascular abnormalities. The first group consisted of patients with retinopathy of this severity or worse. They were compared to patients with less severe diabetic retinopathy and those with none to mild or moderate nonproliferative changes caused by diabetic retinopathy.
Analyses were performed with logistic regression, to evaluate the association between the serum factors and the severity of the retinopathy. We adjusted for the following variables by including them in the statistical model: age, sex, total cholesterol, hemoglobin A1C, creatinine, duration of diabetes, type of diabetes, and the presence or absence of macular edema. Age, total cholesterol, hemoglobin A1C, creatinine, and duration of diabetes were treated as continuous variables. Macular edema was defined as retinal thickening affecting or threatening the center of the fovea and/or presence of focal laser photocoagulation. Several measurements of chemokine and cytokine levels were imputed, because their measured level was listed as below the predefined minimum level for that test (ENA-78, RANTES, IL-8, MIP-1α, and IL-6). For analysis purposes, such levels were imputed tobe half the predefined minimum, rather than zero. Some of the chemokine–cytokine analyses contained a high number of these censored levels (IL-8, 67%; MIP-1α, 77%; and IL-6, 88%) and rather than analyzing their actual levels they were dichotomized. The analyses were performed by computer (SAS System, 8.2; SAS, Cary, NC).