Corneal autofluorescence has been lately studied as a predictor of ret
inopathy severity in diabetic patients. We measured corneal autofluore
scence in 138 eyes of 69 diabetic patients and 64 eyes of 32 healthy c
ontrols. Diabetic patients were subdivided by the severity of retinopa
thy according to the Modified Airlie House Classification (stage 1: no
or minimal retinopathy; stage 2: minimal background retinopathy; stag
e 3: background retinopathy; stage 4: (pre-) proliferative retinopathy
. The fluorescence peak value and the area underlying the corneal auto
fluorescence curve were measured with a scanning fluorophotometer (Flu
orotron Master, Coherent Radiation Pale Alto CA) Healthy controls' val
ues of corneal autofluorescence (peak value: mean 11.03 +/- 3.77 ng. e
quivalent fluorescein/ml; area: mean 163.85 +/- 61.40 scan-point x ng.
equivalent fluorescein/ml) resulted similar (peak value: p = 0.83; ar
ea: p = 0.61) to those of diabetic patients without retinopathy (peak
value: mean 11.2 +/- 3.4 ng.eq/ml; area: 170.07 +/- 57.23 scan-pnt.ng.
eq/ml). A statistically significant difference was found between diabe
tic patients without retinopathy and patients with stage 2, 3, 4 retin
opathy. No statistically significant difference was found both for the
peak value (p = 0.50) and for the area (p = 0.61) between stage 3 and
stage 4 retinopathy. The sensitivity and specificity of corneal autof
luorescence as a screening test for diabetic retinopathy were 82% and
62% for the peak value, 87% and 60% for the area; the positive predict
ive value for the presence of diabetic retinopathy was 65% for the pea
k and 63% for the area.