Participants' characteristics with and without DR, DME, and CSME were compared by using the χ2 statistic for proportions, and a t-test or Mann–Whitney U test for means or median (Intercooled STATA version 10.1 for Windows; StataCorp., College Station, TX). Serum lipids were assessed categorically (in quintiles) and continuously (per standard deviation change). The four measures of macular thickness were assessed as continuous variables. Normality was checked, and data were transformed as appropriate. Person data were used for DR, DME, and CSME based on the worse eye. Multinomial and ordinal logistic regression models were performed to assess associations between serum lipids and severity of DR (mild, moderate, severe NPDR, and PDR) or DME (mild, moderate, and CSME). A binomial logistic regression model was used to assess the association between serum lipids and CSME. Association between serum lipids and macular thickness was examined using generalized linear models. We initially adjusted for age and sex (model 1) and additionally for diabetes duration, GHb, SBP, BMI, use of diabetes medications, use of lipid-lowering medications, and insulin usage (model 2). Covariates included in the models were either categorical or continuous (per standard deviation changes for age, SBP, and BMI; per year for diabetes duration; and per percent for GHb level).