[Evaluation of the diagnostic capabilities of nailfold capillaroscopy in diabetic retinopathy].

PURPOSE

This study evaluates the diagnostic capabilities and the prognostic value of nailfold capillaroscopy data of patients with diabetic retinopathy (DR) to develop an algorithm of monitoring patients with type 2 diabetes mellitus.

MATERIAL AND METHODS

The study involved 90 patients (mean age 67 years), among them 31 with nonproliferative diabetic retinopathy, 29 patients with proliferative DR and 30 patients without retinopathy. In addition to conventional ophthalmological examination, optical coherence tomography angiography (OCTA) on the Optovue RTVue-100 device (USA) was performed using en face vessel density protocol to examine the state of the microvasculature of the superficial and deep layers of the vascular plexus of the central retinal zone, as well as nailfold capillaroscopy using computerized capillaroscope KK-01 (ZAO Centr Analiz veshhestv, Russia).

RESULTS

The cut-off points for detecting the presence of non-proliferative DR (capillary network density below 38.4%, arterial velocity below 512 mm/s and venous blood flow below 585 mm/s), and the presence of proliferative DR (capillary network density below 30.4%, the arterial velocity below 451 mm/s and the venous blood flow below 441 mm/s) were identified according to ROC-analysis of nailfold capillaroscopy data. In the diagnosis of proliferative DR the capillary network density parameter has a slightly higher diagnostic information value (AUC=0.963) than arterial blood flow velocity (AUC=0.941) or venous blood flow velocity (AUC=0.909). Using the identified critical parameters for predicting the initial and proliferative DR, we created a diagnostic algorithm involving a comprehensive assessment of all characteristics.

CONCLUSION

The study revealed that nailfold capillaroscopy indicators (capillary network density, velocity of arterial and venous blood flow) have high diagnostic information value for detecting both non-proliferative and proliferative retinopathy. We constructed mathematical models for predicting DR with an accuracy of predicting the presence of a non-proliferative stage in 92.2% of cases and a proliferative stage in 94.4% of cases. For practical use in clinical environment, we created a computer program calculating the results of DR predictions according to nailfold capillaroscopy data.