Reproducibility of in vivo corneal confocal microscopy as a novel screening test for early diabetic sensorimotor polyneuropathy

Aim With the goal of identifying a valid biomarker of early diabetic sensorimotor polyneuropathy, we aimed to identify the most reliable in vivo corneal confocal microscopy (CCM) parameter for detection of abnormality of small nerve fibre morphology. Methods Cross-sectional examination of 46 subjects (26 with Type 1 diabetes and 20 healthy volunteers) examined by corneal confocal microscopy for intra-and interobserver reproducibility by the intraclass correlation coefficient method. Corneal nerve fibre density, nerve branch density, nerve fibre length and tortuosity were measured on the same day that subjects underwent clinical and electrophysiological examination. Results The 26 subjects with Type 1 diabetes had mean age and diabetes duration 42.8 +/- 16.9 and 22.7 +/- 16.4 years, respectively. Twelve of those subjects (46%) did not meet criteria for diabetic sensorimotor polyneuropathy, while five (19%) had mild, three (12%) had moderate and six (23%) had severe diabetic sensorimotor polyneuropathy. None of the healthy volunteers (mean age 41.4 +/- 17.3 years) had polyneuropathy. Re-examination of selected corneal confocal microscopy images or sets of 40 images yielded very good to excellent intraclass correlation coefficients for all parameters. However, only one parameter (corneal nerve fibre length) emerged with consistently very good reproducibility using a clinically relevant 'study-level' protocol of subject re-examination (intra-observer intraclass correlation coefficient 0.72; interobserver intraclass correlation coefficient 0.73). Despite no differences in intraclass correlation coefficient between subgroups, corneal nerve fibre length was significantly lower (14.76 vs. 16.15 mm/mm(2), P = 0.04) in those with diabetes. Conclusions Development of corneal confocal microscopy may need to focus on the measurement of corneal nerve fibre length, as it appears to have superior reliability in comparison with other parameters, and as evidence exists for its potential as a clinical biomarker of early diabetic sensorimotor polyneuropathy.