Eyes with axial length below 22 mm present the highest formula-dependent variability in cataract surgery. The published evidence consistently identifies Hoffer Q, Haigis, and modern AI-based formulas as the most accurate options — but the picture is nuanced by ACD and the degree of shortness.
Standard IOL power formulas were developed and validated using datasets dominated by eyes with axial lengths in the normal range (22–25 mm). Short eyes (<22 mm) represent approximately 5–8% of cataract surgery candidates and exhibit several anatomical characteristics that challenge formula accuracy:
Crowded anterior segment: Short eyes typically have shallower anterior chambers and thicker lenses relative to axial length. This relationship between AL and ACD is less predictable than in average eyes, making ELP prediction — the primary source of formula error — less reliable.
Higher IOL powers: Short eyes require IOL powers of 25–35 D or higher. Small errors in keratometry or AL measurement translate into larger dioptic errors at higher powers. A 0.1 mm error in AL measurement produces approximately 0.27 D of refractive error in an average eye but up to 0.45 D in a short eye.
Steep corneas: Short eyes often have steeper corneas (K >46 D), which further complicates ELP prediction in formulas that use K as a surrogate for ACD.
Hoffer Q was specifically designed to improve ELP prediction in short eyes. Unlike SRK/T, which uses a linear relationship between AL and ACD, Hoffer Q uses a trigonometric model that accounts for the non-linear relationship between these variables in short eyes. The formula was published in 1993 by Kenneth Hoffer and has accumulated the largest body of short-eye validation data of any traditional formula.
A 2007 study by MacLaren et al. in 41 eyes with AL <22 mm (range 20.29–21.96 mm) found Hoffer Q outperformed SRK/T with a mean prediction error of 0.61 D vs 0.87 D respectively. The study used partial coherence interferometry biometry without customized ACD constants, making it a conservative comparison.
A 50-year literature review by Moschos et al. (2014) concluded that Haigis, Hoffer Q, and Holladay 2 are the best formula options for short eyes (<22 mm), while SRK/T is specifically not recommended in this AL range.
A prospective study analyzing 62 short eyes (AL 20.58–21.97 mm) found Hoffer Q achieved the lowest mean absolute error (MAE = 0.09 ± 0.08 D) among six formulas tested (Holladay 1, SRK/T, Hoffer Q, Haigis, Holladay 2, Barrett Universal II) when only Hoffer Q-calculated lenses were implanted — though this introduces selection bias favoring Hoffer Q.
A comprehensive 2022 study by Turnbull et al. evaluated 10 formulas in 172 eyes with AL ≤ 22 mm, providing the most rigorous modern comparison available. Key findings:
| Formula | MedAE (D) | % within ±0.50 D | Generation |
|---|---|---|---|
| VRF-G | 0.242 | 75.67% | 5th (modern) |
| Haigis | 0.247 | 73.84% | 4th |
| Kane | 0.263 | 73.26% | AI-based |
| Holladay 2 | 0.279 | 71.51% | 4th |
| Barrett Universal II | 0.301 | 68.02% | 4th |
| Hoffer Q | 0.313 | 66.28% | 3rd |
| SRK/T | 0.382 | 58.72% | 3rd |
| Holladay 1 | 0.344 | 62.79% | 3rd |
This study reveals that while Hoffer Q outperforms SRK/T and Holladay 1, it is no longer the top-performing formula for short eyes when compared to Haigis with optimized constants, Kane, and newer options like VRF-G. Barrett Universal II underperformed in this cohort relative to its reputation for average eyes.
A 2025 study by Chen et al. in 184 short eyes (AL <22 mm) demonstrated that formula accuracy in short eyes depends significantly on anterior chamber depth. Subgroup analysis showed that EVO 2.0 and Kane performed best in eyes with shallower ACDs (<2.5 mm), while Haigis and Hoffer QST performed well across the ACD range. The study found that BUII, Hoffer QST, Hoffer Q, Holladay 1, and SRK/T all showed myopic shifts (−0.49 to −0.18 D; p < 0.05) in this cohort, while K6 displayed a hyperopic shift.
This subgroup finding suggests that for the shortest eyes with the most crowded anterior segments (AL <21 mm, ACD <2.5 mm), neither Hoffer Q alone nor any single formula is sufficient — a multi-formula consensus approach is warranted.
Nanophthalmos (AL <20 mm) is a separate clinical entity requiring special consideration. These eyes have markedly thickened sclera and a disproportionately small anterior segment relative to total axial length. Standard biometry with partial coherence interferometry may be unreliable due to dense lens or posterior pole abnormalities.
In nanophthalmic eyes, traditional formula performance is less predictable and the evidence base is thinner. Published comparisons support checking Haigis with optimized constants if available, Kane or other modern formulas, and a multi-formula consensus rather than relying on a single calculation. Target refraction in nanophthalmos should account for the risk of postoperative uveal effusion and choroidal detachment, which can shift the final refraction unpredictably.
The fundamental reason Hoffer Q and Haigis outperform SRK/T in short eyes lies in their ELP prediction models. SRK/T uses a simplified linear relationship between AL and ACD constant. Hoffer Q uses a trigonometric correction that better models the geometry of short eyes. Haigis uses three individualized constants (a0, a1, a2) that separately weight ACD and AL, providing the most granular ELP prediction of the traditional formulas — but only when properly optimized.
Modern AI-based formulas (Kane, Hill-RBF) use large training datasets to learn the non-linear relationships between biometric variables and ELP without explicitly modeling the geometry. Their strong performance in short eyes likely reflects this flexibility in capturing non-linear patterns that explicit geometric models approximate less well.
Clinical recommendation for short eyes:
AL 21–22 mm: Haigis (optimized), Kane, or Hoffer Q — any of these is defensible
AL 20–21 mm: Haigis (optimized), Kane, or multi-formula consensus
AL <20 mm (nanophthalmos): Haigis (optimized) with awareness of measurement reliability issues; consider intraoperative aberrometry
Avoid relying on SRK/T alone in short eyes, especially below 22 mm. Consider a mild myopic target when the formula set or eye anatomy suggests risk of hyperopic surprise.