For eyes with axial lengths between 22 and 25 mm, the two formulas converge. The clinically meaningful divergence occurs at the extremes — and the evidence tells a nuanced story.
SRK/T (Sanders-Retzlaff-Kraff/Theoretical), introduced in 1990 by Retzlaff, Sanders, and Kraff, is a hybrid theoretical-empirical formula that calculates IOL power from axial length (AL), keratometry (K), and an A-constant. It was a major advance over the purely empirical SRK II and remains one of the most widely used formulas globally.
Barrett Universal II, developed by Graham Barrett, uses a five-variable model incorporating AL, K, anterior chamber depth (ACD), lens thickness (LT), and corneal diameter. Unlike SRK/T, which predicts effective lens position (ELP) from AL and K alone, Barrett uses a theoretical model of the lens-eye system that more accurately accounts for the variability in ELP prediction — the primary source of error in third-generation formulas.
In the largest comparative study to date, Melles et al. (2018) analyzed 18,501 eyes across multiple formulas and found that for eyes with AL between 22.0 and 25.0 mm, the difference in mean absolute error (MAE) between SRK/T and Barrett Universal II was less than 0.1 D — not clinically meaningful for most patients.
A retrospective study by Ahmed et al. (2022) comparing SRK/T, Barrett Universal II, and Hill-RBF 2 across 100 eyes found no statistically significant difference in mean prediction error between SRK/T and Barrett Universal II (0.092 ± 0.041 D; p = 0.078), consistent with earlier findings.
Clinical takeaway: For eyes with AL 22–25 mm, formula choice has minimal impact on refractive outcomes when optimized constants are used. The focus should be on biometry accuracy and constant optimization.
Long eyes have a documented risk of postoperative hyperopic error with older formulas when axial-length adjustment and modern formula cross-checking are not used. SRK/T can perform reasonably in long eyes, but its ELP and axial-length assumptions become less dependable at extreme axial lengths. Multiple studies have confirmed that newer formulas and axial-length adjustment methods reduce this risk.
Abulafia et al. demonstrated that in eyes with AL greater than 26.0 mm, Barrett Universal II significantly outperformed SRK/T, with a lower MAE and fewer outliers beyond ±1.0 D. The Barrett formula's more sophisticated ELP model better accounts for the altered anterior segment geometry in myopic eyes.
For extreme axial myopia (AL ≥ 30 mm), a 2025 study from the Institute of Science Tokyo evaluated Kane, Hill-RBF, Barrett Universal II, and SRK/T in 80 eyes. SRK/T showed the highest MAE (0.96 D) and the highest rate of errors exceeding ±1.0 D (42.5%), compared to Kane (0.51 D MAE) and Barrett Universal II (0.66 D MAE). Both AI-based formulas significantly outperformed SRK/T in this cohort (p < 0.05).
For short eyes, the picture is more complex. A 2022 study by Turnbull et al. comparing 10 formulas in 172 eyes with AL ≤ 22 mm found that VRF-G (0.242 D MedAE), Haigis (0.247 D), and Kane (0.263 D) outperformed both SRK/T and Barrett Universal II. Barrett Universal II showed a myopic shift in the shortest eyes (AL < 21 mm), particularly in eyes with higher astigmatism and wider white-to-white measurements.
An earlier review by Moschos et al. (2014) of the 50-year published literature concluded that Haigis, Hoffer Q, and Holladay 2 are the best options for short eyes (<22 mm), with SRK/T underperforming in this range.
| Axial Length | SRK/T Performance | Barrett Universal II | Preferred Formula |
|---|---|---|---|
| <21 mm | Hyperopic shift common | Myopic shift reported | Haigis, Hoffer Q, Kane |
| 21–22 mm | Acceptable | Acceptable | Hoffer Q, Haigis, Barrett |
| 22–25 mm | Excellent | Excellent | Either; optimize constants |
| 25–26 mm | Slight myopic trend | Excellent | Barrett Universal II |
| >26 mm | Hyperopic error risk without adjustment | Better, but Kane/Hill-RBF may outperform | Kane, Hill-RBF, Barrett |
| ≥30 mm | MAE 0.96 D, 42.5% >±1D | MAE 0.66 D | Kane, Hill-RBF |
Formula comparison studies that use non-optimized constants systematically favor newer formulas, which are often published with pre-optimized constants. A key finding from the Melles et al. meta-analysis was that after constant optimization, the performance gap between formulas narrows substantially for average eyes. The User Group for Laser Interference Biometry (ULIB) maintains optimized constants for most major IOLs across all major formulas.
A surgeon using SRK/T with a personally optimized A-constant from their own outcomes database will likely outperform a surgeon using Barrett Universal II with a generic constant — particularly for average eyes.
Evidence for formula selection during combined cataract and glaucoma surgery is less settled than for routine cataract eyes or long-eye cohorts. Because anterior segment anatomy, intraocular pressure changes, and surgical planning may differ from standard cataract cases, compare more than one formula and review the calculation rationale rather than relying on a single formula result.
Based on the published evidence, the following approach is clinically defensible:
AL 22–25 mm: Either SRK/T or Barrett Universal II with optimized constants. The formula matters less than the biometry accuracy and constant quality.
AL >25 mm: Barrett Universal II, Kane, Hill-RBF, and adjusted SRK/T should be compared rather than relying on unadjusted SRK/T alone. For AL ≥ 28 mm, include more than one modern formula and consider axial-length adjustment.
AL <22 mm: Hoffer Q, Haigis, or Kane. Barrett Universal II should be used with caution in very short eyes, particularly those with additional biometric anomalies.
Combined glaucoma surgery: Treat formula choice as a review point rather than a fixed rule; compare formulas and consider ocular anatomy, target refraction, and surgeon outcomes.
Bottom line: The debate between SRK/T and Barrett Universal II is largely resolved for average eyes — both perform well with optimized constants. The clinical decision point is at axial length extremes, where Barrett Universal II has a consistent evidence base over SRK/T for long eyes, while neither is optimal for very short eyes.