Abbreviations: (AGE) advanced glycation end product, (IR) infrared, (DFA) 1-deoxyfructosyl glycation adduct, (LED) light-emitting diode, (UV) ultraviolet Keywords: advanced glycation end products, biomicroscope, diabetes mellitus, fluorescence ratio, lens autofluorescence

Lens autofluorescence is increased in patients with diabetes mellitus, but clinical application has been limited by the lack of an instrument suitable for routine clinical use. We investigate possible uses of a new scanning confocal biomicroscope (1) to identify subjects with undiagnosed type 2 diabetes and (2) as a marker for the progression of diabetes. One hundred seventyeight subjects self-reported as normal and 53 subjects physician-diagnosed with diabetes or prediabetes were recruited. Measurements were collected using a ClearPath DS-120 Lens Fluorescence Biomicroscope calibrated with standards traceable to National Institute of Standards and Technology (NIST). Fluorescence intensities were corrected for age by subtracting the value expected from a regression of intensity versus age for normal subjects. This "fluorescence deviation" showed progressively higher values for normal, prediabetes, type 2 diabetes, and type 1 diabetes and a high degree of predictability of diabetes diagnosis. A receiver operating characteristics curve was used to determine sensitivity and specificity for prediction of diabetes type 2. At a fluorescence deviation of 2500, a sensitivity of 67% at 94% specificity was observed detection of type 2 diabetes. The progressively higher fluorescence deviations are consistent with the physiological mechanisms of accumulation of fluorescent advanced glycation end products as the subject ages. The sensitivity and specificity performance of the lens autofluorescence test for type 2 diabetes is comparable to the performance of glucose threshold tests. The statistically significant difference between fluorescence deviations of normal and type 2 diabetes supports the feasibility of lens autofluorescence to screen subjects for undiagnosed type 2 diabetes. Ophthalmic practices are points of care at which there may be a public health benefit for screening patients for undiagnosed diabetes.

We have developed a new lens measurement system (LMS) that simultaneously measures the intensities of fluorescence and Rayleigh components at various distances into the lens along the optical axis. The noninvasive measurement is performed through an undilated pupil, and with the assistance of a pupil tracking system that facilitates maintaining the x and y positions of the sample volume (ca. 300 m in length, 600 m high, and 80 m wide) to within Ϯ100 m of any programmed ''lock'' position. The intensity of the Rayleigh component that is used to normalize the measured fluorescent signal serves to correct the attenuation effects due to absorption and lens light scatter (Mie-Tyndall type). This report, resulting from a SpectRx Site L clinical study using a refined instrumentation (SpecRx Scan V), presents analysis of fluorescence and Rayleigh data from the lenses of 923 controls and 239 diabetic subjects (45 Type I and 194 Type II) ranging from 23 to 75 years old. Fluorescence and Rayleigh data have been obtained via confocal mode from various locations nominally along the lens optical axis (the true trajectory of a sample volume could be shown in a figure) for controls and diabetics, at different ages, using three pairs of excitation and collection wavelengths: 364/495 nm, 434/495 nm, and 485/515 nm. For control subjects, there exists a strong, almost linear relationship between age and fluorescence, while diabetic subjects tend to deviate from this age-fluorescence relationship. Our data show that the lenses of diabetic patients are subject to an accelerated aging process, presumably due to an elevated level of brown and fluorescent protein adducts and crosslinks from nonenzymatic glycosylation (Maillard-Amadori reactions). We have also shown that by using the measured Rayleigh profiles to normalize the measured fluorescence, most of the absorption effects are removed and therefore the separation between the fluorescence of diabetics and controls is greatly improved. Thus, the device for measuring fluorescence/Rayleigh ratios can be used to noninvasively screen populations for possible undiagnosed diabetes.