Prediction of total corneal power in keratoconus using anterior surface data.

CLINICAL RELEVANCE

Keratoconus results in an increase in anterior and posterior curvatures and a reduction in corneal thickness. Anterior corneal ectasia is partially compensated by remodelling the corneal epithelium. Therefore, there is an alteration in the relationship between corneal surfaces and variation in corneal power. The variation in corneal power is one of the sources that induces errors in IOL power calculation.

BACKGROUND

This study aimed to assess a method for predicting total corneal power in keratoconus using several anterior surface parameters at 3 mm and 4 mm.

METHODS

Tomographic data obtained using Pentacam (Oculus, Germany) were analysed from 280 eyes of 140 patients with keratoconus using anterior and posterior keratometry, anterior Q-value at 8 mm, central corneal thickness, Kmax location and value, and true net power at 4 mm (TNP). Calculated total corneal power (TCPc) at 3 mm was obtained using the Gauss formula. Predicted total corneal power at 3 mm (TCPp3) and 4 mm (TCPp4) was obtained from univariate (TCPp3u and TCPp4u) and multivariate linear regression formulae (TCPp3m and TCPp4m). SimK, anterior Q-value, vertical location, and Kmax value were used in the multivariate formulae. Mean absolute error (MAE) and median absolute error (MedAE) were also calculated. Absolute frequencies within dioptric ranges of all formulas divided for keratoconus grading were evaluated.

RESULTS

TCPc and TNP exhibited a good correlation (R2 = 0.58, p < 0.05) with a higher dispersion above 50 D of corneal power. Highly significant correlations were observed between TCPp3u and TCPc (R2 = 0.978, p < 0.05) and TCPp3m and TCPc (R2 = 0.989, p < 0.05). Lower but significant correlations were observed between TCPp4u and TNP (R2 = 0.692, p < 0.05) and between TCPp4m and TNP (R2 = 0.887, p < 0.05). The best results for TCP prediction at 3 and 4 mm were obtained with TCPp3m and TCPp4m as follows: MAE of TCPp3m was 0.24 ± 0.20 (SD) D with MedAE of 0.20 D, while MAE of TCPp4m was 0.96 ± 0.77 D with MedAE of 0.80 D. The 3 mm multivariate regression formula results in higher absolute frequencies of prediction errors in the total eyes within 0.5 D (93%) than the univariate formula (81%). At 4mm, the multivariate regression formula has a lower percentage within 0.5 D (32%) than the univariate formula (41%), but the percentage of the multivariate formula is higher within 1 D (63%) than the univariate formula (56%).

CONCLUSION

All formulas show a decrease in accuracy with increasing grades of keratoconus. Multivariate linear regression formulae using only anterior surface data can predict TCP with good approximation in eyes with keratoconus in cases where posterior surface parameters are unavailable. The vertical location of Kmax and the anterior asphericity could play a relevant role in the prediction of total corneal power in keratoconus.