PURPOSE: To assess the accuracy of intraocular lens (IOL) power formulas modified specifically for patients with keratoconus (Holladay 2 with keratoconus adjustment and Kane keratoconus formula) compared with normal IOL power formulas (Barrett Universal 2, Haigis, Hoffer Q, Holladay 1, Holladay 2, Kane, and SRK/T). DESIGN: Retrospective consecutive case series. PARTICIPANTS: A total of 147 eyes of 147 patients with keratoconus. METHODS: Data from patients with keratoconus who had preoperative IOLMaster biometry were included. A single eye per qualifying patient was randomly selected. The predicted refraction was calculated for each of the formulas and compared with the actual refractive outcome to give the prediction error. Subgroup analysis based on the steepest corneal power measured by biometry (stage 1: ≤48 diopters [D], stage 2: >48 D and ≤53 D, and stage 3: >53 D) was performed. MAIN OUTCOME MEASURE: Prediction error. RESULTS: On the basis of the mean absolute prediction error (MAE), the formulas were ranked as follows: Kane keratoconus formula (0.81 D), SRK/T (1.00 D), Barrett Universal 2 (1.03 D), unmodified Kane (1.05 D), Holladay 1 (1.18 D), unmodified Holladay 2 (1.19 D), Haigis (1.22 D), Hoffer Q (1.30 D), and Holladay 2 with keratoconus adjustment (1.32 D). The Kane keratoconus formula had a statistically significant lower MAE compared with all formulas (P < 0.01). In stage 3 keratoconus, all nonmodified formulas had a hyperopic mean prediction error ranging from 1.72 to 3.02 D. CONCLUSIONS: The Kane keratoconus formula was the most accurate formula in this series. The SRK/T was the most accurate of the traditional IOL formulas. All normal IOL formulas resulted in hyperopic refractive outcomes that worsened as the corneal power increased. Suggestions for target refractive aims in each stage of keratoconus are given.