Abstract
The correlation between 12 variables retinal findings and 9 variables clinical laboratory findings in 223 diabetics was studied by multivariate analysis such as correlation analysis, factor analysis (varimax method) and discriminant function. Hammer-Smith Hospital standards set by Oakley et al. were adopted for the statistical analysis of retinal findings and claculations for multivariate analysis were carried out with a model NE AC 2200 computer.
Factor analysis revealed that diabetic retinopathy was divided into “simple retinopathy” composed of deep haemorrhages and hard exudates, and “malignant retinopathy” composed of new vessel formations and fibrous proliferation. Superficial haemorrhages were closely related to the latter.
i) Stages of diabetic retinopathy by Scott's classification were closely related to systolic blood pressure, duration of diabetes mellitus and fasting blood sugar.
ii) Deep haemorrgages were closely related to systolic blood pressure, fasting blood sugar and urinary protein; superficial haemorrhages to fasting blood sugar, BUN and urinary protein; hard exudates to serum cholesterol; and new vessel formations to fasting blood sugar, serm cholesterol, systolic blood pressure and urinary protein.
iii) First component of factor analysis representing malignant retinopathy was closely related to fasting blood sugar and urinary protein. The second representing simple retinopathy was closely related to serum cholesterol, fasting blood sugar and systolic blood pressure.
Discriminant functions for predicting malignant retinopathy applying retinal and clinical laboratory findings were composed of a linear combinating formula. Probability by this function was 97.6%. Multivariate analysis is also useful in determing the prognoses of diabetics with retinopathies.
