Keratometric indices for detecting the type of keratoconus: a combined discriminant analysis

Abstract

BackgroundThis study sought to determine the diagnostic ability of anterior, posterior and total keratometric indices in discriminating keratoconus (KCN) stages 1 to 4 from normal corneas.MethodsAnterior and posterior simulated keratometry (sim‐Ksteep and sim‐Kflat), maximum keratometry (Kmax), Kmax y co‐ordinate, anterior and posterior radii of curvature centred on the thinnest point (ARC‐3-mm and PRC‐3-mm), and anterior, posterior, and total asphericity (Q‐value) measured by Pentacam were extracted from electronic medical records of 200 KCN and 200 normal cases. Stepwise leave‐one‐out cross‐validation and areas under receiver operating characteristic curves (AUROC) were used to detect the best set of indices for differentiating normal from low‐grade (stages 1–2, Kmax ≤ 53.00 D) and high‐grade (stages 3–4, Kmax > 53.00 D) KCN.ResultsMean age in KCN and normal groups was 33.10 ± 7.48 and 32.24 ± 9.00-years (p = 0.332), respectively. In the low‐grade KCN group, PRC‐3-mm (AUROC = 0.986), Kmax (AUROC = 0.979), and Kmax y co‐ordinate (AUROC = 0.824) performed best, and the difference in AUROC between PRC‐3-mm and Kmax was not significant (p = 0.153). In the high‐grade KCN group, Kmax (AUROC = 1.000), PRC‐3-mm (AUROC = 0.998), posterior Ksteep (AUROC = 0.970), posterior Q‐value (AUROC = 0.940), and posterior Kflat (AUROC = 0.894) performed best, and there was no significant difference in AUROC values between Kmax and PRC‐3-mm (p = 0.307) or between posterior Ksteep and posterior Q‐value (p = 0.113).ConclusionsAmong the studied keratometric indices, PRC‐3-mm and Kmax appear to have the best ability for detecting mild to severe KCN. The next best factors which showed good discrimination ability were Kmax y co‐ordinate in low‐grade cases and posterior Ksteep and Q‐value in high‐grade cases.