KERATOCONUS NATURAL PROGRESSION: A TIME SERIES APPROACH

To predict the visual and topographic outcomes of untreated keratoconus using a time series approach.

Department of Ophthalmology, Hospital Universitario Central de Asturias, Oviedo, Spain.

Time series model was created using IBM SPSS Statistics version 21 for Windows (Microsoft Corp.), including all patients under 30 years of age diagnosed with keratoconus with a topographic follow-up period ≥ 12 months (35 patients, 57 eyes). Operated cases (corneal segments, cross-linking, keratoplasty) or those with low-quality topographies were ruled out. The best-corrected visual acuity (BCVA) and topographic parameters related to anterior and posterior curvature and elevation, thickness and corneal aberrations on Scheimpflug–Placido topography (Sirius, CSO, Italy) were noted at all visits at 6-month intervals. The stationary R-squared was used to evaluate the model. Semester variation rates and 95% confidence intervals (CIs) from 84 to 96 months were predicted.

Mean follow-up was 43.12 ± 19.15 months (12-76). The expert modeler was selected to determine the best-fitting model for each parameter. The predicted values were consistent with the observed ones from 6 to 76 months. The variables maximum keratometry (Kmax), mean anterior keratometry (Kmean), anterior and posterior asphericity in the central 8 mm (QA and QP, respectively), surface asymmetry index of back surface (SIb) and Baiocchi Calossi Versaci back index (BCVb), showed a trend better adjusted to the Brown model, whereas corneal volume, longitudinal spherical aberration (LSA), surface asymmetry index of front surface (Slf) and Baiocchi Calossi Versaci front index (BCVf), better fit the Holt model. The BCVA, root mean square per area of front and back surface (RMS / A_ant and (RMS / A_post, respectively), mean pupil power (MPP), elevation at keratoconus vertex front and back (KVf and KVb, respectively), thinnest corneal thickness (TCT), root mean square of coma and high-order aberrations (RMS coma and RMS HOA, respectively), as well as total astigmatism (CYL total), however, they did not present a trend and were explained using a simple model. At 96 months, an increase in Kmax was predicted by 0.71% (95% CI -1.81 to 3.22); of Kmean of 0.77% (95% CI -0.87 to 2.41), of QA of -12.74% (95% CI -42.83 to 17.33), of QP of -16.59% (95% CI -50.93 to 17.75), of Slf of 15.08% ( 95% CI 10.87 to 90.91), of Slb of 15.08% (95% CI -21.40 to 51.57), of BCVf of 29.35% (95% CI 7.87 to 54.84) and of BCVb of 16.80% (95% CI -23.23 to 56.84). A decrease in corneal volume of -4.26% (95% CI -7.03 to -1.50) and of LSA of -888.17% (95% CI -2097.89 to 321.36) was also predicted.

Time series is an interesting statistical method that can predict the natural progression of keratoconus. Topographic variables such as asphericity, Slf and SIb, BCV, corneal volume and LSA may change earlier than BCVA, Kmax or TCT, currently used to indicate a cross-linking surgery, and can be a prompt indicator of ectasia progression.