Influence of cataract surgery on corneal endothelial cell density estimation.

PURPOSE: To assess the effect of cataract surgery on semiautomated human corneal endothelial cell density (ECD) estimation using noncontact specular microscopy.

METHODS

In this prospective clinical study, 62 consecutive patients undergoing cataract surgery were studied. To evaluate possible variations of accuracy and agreement under clinical circumstances, the corneal ECD was determined before and after cataract surgery. The parameter ECD (cells/mm) in the central and paracentral cornea was consecutively determined with 2 algorithms available in a noncontact specular microscope [fixed-frame method (FFM) and automatic center method (ACM)]. The postoperative evaluation was performed at 1 day, after 4 weeks, and after 6 months. The accuracy, the relative error, and the 95% limits of agreement (LoA) were determined for both counting methods.

RESULTS

The overall group-averaged accuracy was -19.4 cells/mm (0.86%) centrally and -17.3 cells/mm (0.76%) paracentrally. The LoA (95% CI) were within 234 cells/mm (10.4%) and 250 cells/mm (10.9%), respectively. After cataract surgery, a slight decrease in accuracy was noted in the central (2.09%) and paracentral areas (1.76%). The relative error increased from 3.66% to 6.02% centrally and from 4.96% to 6.55% paracentrally. The LoA (95% CI) increased from +/-194 cells/mm to +/-304 cells/mm centrally and from +/-275 cells/mm to +/-322 cells/mm paracentrally. In the later postoperative period, endothelial stabilization improved the accuracy and agreement in ECD estimation.

CONCLUSIONS

The estimation of ECD after cataract surgery employing current algorithms was achieved with a clinically acceptable level of accuracy and agreement. However, the analysis of images in the early postoperative period as well as paracentral corneal areas revealed larger variabilities. This revealed that the FFM and ACM counting methods cannot be used interchangeably under all circumstances. The ACM seemed preferable when only low-quality images were available and permitted determination of additional qualitative endothelial cell parameters.