I now have everything needed for a comprehensive residency study guide. Here it is:
IOL Power Calculation - Complete Residency Study Notes
Fundamental Concept: What We're Calculating
The IOL power required to achieve the target refraction (usually emmetropia) after cataract surgery. All formulas revolve around accurately predicting the Effective Lens Position (ELP) - where the IOL will sit in the eye postoperatively. This is the single most important variable and the main source of error.
Core inputs (vary by generation):
- AL - Axial length (most critical measurement)
- K - Keratometry (corneal power)
- ACD - Anterior chamber depth
- LT - Lens thickness
- WTW - White-to-white corneal diameter
- Pre-op refraction, Age
IOL Constants (Per-Lens, Per-Formula)
| Constant | Used in | Meaning |
|---|
| A-constant | SRK I, SRK II, SRK/T | Empirical lens-specific constant (typically ~118-119) |
| Surgeon Factor (SF) | Holladay 1 | Distance from corneal principal plane to IOL |
| pACD (personalized ACD) | Hoffer Q | Predicted ACD for that specific IOL |
| a0, a1, a2 | Haigis | Three constants: a0 = offset, a1 = ACD-related, a2 = AL-related |
| Lens Factor | Barrett Universal II | Related to A-constant but accounts for principal plane of IOL |
| C-constant | Olsen | Ratio of IOL fixation in capsular bag (0 = anterior, 1 = posterior) |
Key point for exams: A-constant can be converted: SF = (A-constant / 0.5663) - 65.6; pACD = 0.62467 × A-constant - 68.747
Generations of IOL Formulas
FIRST GENERATION (1960s-1970s) - Pure Regression, Two Variables
Fyodorov (1967) - First ever IOL formula
- Theoretical optical formula
- Landmark historical significance
Binkhorst (1975)
- Early theoretical formula
- Introduced the concept of predicted postoperative ACD
SRK I (Sanders, Retzlaff, Kraff - 1980)
Formula:
P = A - 2.5 × AL - 0.9 × K
| Variable | Meaning |
|---|
| P | IOL power |
| A | A-constant (lens-specific, e.g. 118.4 for PMMA) |
| AL | Axial length (mm) |
| K | Average keratometry (D) |
Limitations:
- Simple linear regression - ignores the non-linear relationship between AL and lens position
- Overcorrects in very short and very long eyes
- No longer used clinically
SECOND GENERATION (1980s) - Regression with AL Correction
SRK II (1988)
Modification: Adjusts A-constant based on axial length
| Axial Length (mm) | A-constant adjustment |
|---|
| < 20.0 | A + 3 |
| 20.0 - 21.0 | A + 2 |
| 21.0 - 22.0 | A + 1 |
| 22.0 - 24.5 | A (no change) |
| > 24.5 | A - 0.5 |
Formula:
P = A1 - 2.5 × AL - 0.9 × K
(where A1 = corrected A-constant from table above)
Improvement over SRK I: Attempts to correct for extreme axial lengths
Still limited: Only adjusts one constant, does not model ELP geometrically
THIRD GENERATION (1988-1993) - Theoretical + Vergence, Three Variables
All third-generation formulas use AL, K, and one IOL constant to predict ELP using theoretical optical/vergence models. Much more accurate than 1st/2nd gen.
Holladay 1 (1988) - Jack Holladay
Variables: AL, K, Surgeon Factor (SF)
ELP prediction: Based on AL and K
Best for: Normal and long eyes; performs well in short eyes too
Key feature: Uses a theoretical formula with regression-derived constants. The SF is the only IOL-specific constant needed.
SRK/T (1990) - Sanders, Retzlaff, Kraff (T = Theoretical)
Variables: AL, K, A-constant
ELP prediction: Based on AL and corneal height (a function of K and AL)
Simplified expression:
P = A - 2.5(AL) - 0.9(K) (same form as SRK I, but ELP predicted theoretically, not by simple regression)
Key modifications over SRK I:
- Postoperative ACD is predicted using a theoretically derived corneal height model
- Retinal thickness correction factor incorporated
- Corneal refractive index adjustment
Best for: Long eyes (AL > 26 mm) - historically the most validated formula for high myopia
Still used widely as a benchmark formula
Hoffer Q (1993) - Kenneth Hoffer
Variables: AL, K, pACD (personalized ACD)
ELP prediction: Unique tangent function incorporating AL
Key formula feature: Uses a Q factor (a personal modifier of the pACD) based on AL - creates a non-linear relationship between AL and predicted ACD.
Best for: Short eyes (AL < 22.0 mm) - historically the reference standard for nanophthalmos/hyperopic eyes
- Equally accurate with Holladay 1 for AL 21.0-21.49 mm
Quick 3rd-Gen Comparison
| Formula | Best AL range | IOL Constant | ELP Predictor Variables |
|---|
| Holladay 1 | Normal/long | SF | AL, K |
| SRK/T | Long (>26 mm) | A-constant | AL, K |
| Hoffer Q | Short (<22 mm) | pACD | AL, K |
FOURTH GENERATION (1996-2000) - Multi-Variable, Personalized ELP Prediction
Key advance: Add measured ACD (and other variables) to improve ELP prediction, because ACD does NOT scale linearly with AL.
Holladay 2 (1996) - Jack Holladay
Variables (7 parameters):
- AL
- Average K
- Horizontal white-to-white (WTW)
- Pre-operative ACD
- Lens thickness (LT)
- Age
- Pre-operative refraction
IOL constant: Modified A-constant / SF
Strength: Most biometric data of any traditional formula; designed for all eye types
Limitation: Algorithm is proprietary (not published); requires purchase of software
Haigis (2000) - Wolfgang Haigis
Variables: AL, K, pre-operative ACD
Three constants: a0, a1, a2
How it works:
ELP = a0 + (a1 × ACD) + (a2 × AL)
- a0 = basic offset (analogous to ACD constant)
- a1 = scales with measured pre-op ACD
- a2 = scales with measured AL
Key advantage: Decouples ACD from AL - recognizes that ACD and AL can vary independently (e.g., deep ACD in a short eye).
Limitation: Requires optimized triple constants (a0, a1, a2) from a large dataset for full accuracy; if only a0 is optimized, loses advantage over 3rd gen.
Best for: Eyes with unusual ACD/AL relationships; post-refractive surgery eyes (Haigis-L variant)
Olsen Formula (2006) - Thomas Olsen
Variables: ACD, LT (primary); AL, K, WTW, refraction, age, gender (secondary)
IOL constant: C-constant (ratio of capsular bag fixation)
Unique concept - C-constant:
IOL position = ACD + C × LT
(C = 0 means IOL sits at anterior capsule; C = 1 means IOL sits at posterior capsule; typical value ~0.42)
Method: Full ray-tracing optical model - most physically accurate approach
Strength: No reliance on regression; works well post-refractive surgery without special modifications
Best for: Complex eyes, post-refractive, exact anatomical approach
FIFTH GENERATION / NEW-GENERATION FORMULAS (2010s-present)
These use machine learning (AI), big data, and/or advanced ray-tracing. They outperform all previous generations across most AL ranges in current comparative studies.
Barrett Universal II (BU II) - Graham Barrett
Variables: AL, K; optional: ACD, LT, WTW
IOL constant: Lens Factor (derived from A-constant)
Key innovation: Retains the principal plane of refraction of the IOL as a relevant variable (other formulas treat the IOL as a thin lens). Uses a theoretical model eye where ACD is related to AL and K in a non-proportional way.
Performance: One of the most widely recommended formulas worldwide; excellent across short-normal-long AL range.
Available free: Online at
barrettformula.iolcalc.org
Hill-RBF (Radial Basis Function - AI) - Warren Hill
Method: Pure machine learning (artificial neural network / radial basis function interpolation)
Training data: Large database of post-op outcomes
Variables: AL, K, ACD; LT and WTW in newer versions
Version: Hill-RBF 3.0 (current)
Special feature: Has a confidence interval output - if the patient's parameters fall outside the training data range, it flags "out of bounds" (don't use the result - use another formula)
Strength: No assumptions about ocular geometry; learns from outcomes directly
Kane Formula (2017) - Jack Kane
Method: Hybrid - theoretical optics + thin lens + AI/regression elements
Variables: AL, K, ACD, sex; optional: LT, CCT
IOL constant: A-constant (optimized)
Performance: Currently among the top performers in most large comparative studies (MAE ~0.30-0.35 D). Particularly strong at extreme ALs.
Available free: iolformula.com
EVO Formula (2020) - Tun Kuan Yeo
Method: Vergence-based theoretical formula with empirical refinements
Variables: AL, K, ACD, LT
Good for: All AL ranges; particularly noted for strong performance in short eyes
PEARL-DGS
Method: AI (deep learning regression)
Performance: Excellent in short eyes; among the best in recent meta-analyses
ESCRS iTrace / Ladas Super Formula
Hybrid formulas that select the best traditional formula for a given set of measurements automatically.
Formula Selection by Axial Length - Quick Reference
| Axial Length | Recommended Formulas (2025 Evidence) |
|---|
| Short (<22.0 mm) | Hoffer Q (traditional), Kane, Hill-RBF, PEARL-DGS, EVO, Barrett UII |
| Very short (<21.0 mm) | Hoffer Q, Kane, AI formulas; SRK/T most inaccurate |
| Normal (22.0-26.0 mm) | Any 3rd/4th/5th gen; Barrett UII, Kane, Holladay 1 all excellent |
| Long (>26.0 mm) | SRK/T (classic), Wang-Koch adjusted SRK/T, Barrett UII, Kane, EVO 2.0 |
| Very long (>30.0 mm) | Wang-Koch SRK/T, Barrett UII, Zhu-Lu formula; segment AL carefully |
(Kanski 10th ed., p. 331-332; The Ophthalmologist 2026)
Special Situations
Post-Refractive Surgery (LASIK/PRK/RK) - HIGH YIELD
Why standard formulas fail:
- Keratometry error - Standard K readings are inaccurate because they assume normal anterior/posterior corneal power ratio (6:1), which is altered after ablation
- ELP error - Formulas predict ELP from K value; abnormally low K gives wrong ELP prediction
- Double K problem - ELP estimated using pre-op K; power calculated using post-op K
Methods to calculate IOL power post-refractive surgery:
| Method | What it requires | Formula |
|---|
| Clinical History Method | Pre-op K, pre-op refraction, post-op refraction | Calculate true K change, insert into standard formula |
| Contact Lens Method (Rigid) | Flat CL base curve, over-refraction | K = CL BC + CL power + over-refraction |
| Haigis-L | Post-op K only | Regression formula with statistical correction; no history needed |
| Shammas-PL | Post-op K only | No-history regression formula |
| Double-K method (Aramberri) | Pre-op K AND post-op K | Uses pre-op K for ELP, post-op K for power |
| Masket formula | Pre/post refraction change | Corrects IOL power based on amount of refractive correction |
| Barrett True-K | Post-op K, optional history | Uses total corneal power; best current option |
Rule of thumb: Always use multiple methods and pick the highest (most myopic) IOL power to avoid post-op hyperopia, which is poorly tolerated.
Silicone Oil in Vitreous
- Sound velocity in silicone oil differs from vitreous (980 m/s vs 1532 m/s for ultrasound)
- A-scan underestimates AL significantly
- Correction: Add ~2.5 mm to A-scan AL measurement (exact factor depends on oil viscosity: 1000 cSt vs 5000 cSt)
- Optical biometry (IOLMaster) is preferred - less affected
Staphyloma (High Myopia)
- Posterior staphyloma causes AL measurement to the staphyloma floor instead of the fovea
- Use immersion B-scan + vector A-scan superimposed on the staphyloma for accurate measurement
Toric IOL Power Calculation
- Residual corneal astigmatism target = (pre-op corneal cylinder) - (IOL cylinder × Effective Lens Position factor)
- Barrett Toric Calculator and Kane Toric are the most accurate current tools
- Surgically induced astigmatism (SIA) must be entered as a constant per surgeon
Dense Cataract / Poor Cooperation
- Optical biometry (IOLMaster) fails if media too opaque
- Use immersion A-scan (more accurate than contact A-scan)
- If axial length of fellow eye is known, use it as a guide
A-Constants for Common IOLs (Approximate Reference)
| IOL Type | A-constant | pACD (Hoffer Q) | SF (Holladay) |
|---|
| PMMA (historical) | 118.4 | 5.41 | 1.22 |
| Alcon SN60WF (AcrySof) | 119.1 | 5.68 | 1.82 |
| AMO ZCB00 (Tecnis) | 119.3 | 5.81 | 1.95 |
| B&L enVista | 119.1 | 5.68 | 1.82 |
Use ULIB (User Group for Laser Interference Biometry) or manufacturer's optimized constants for your specific IOL and biometer combination.
Personalized A-Constant / Constant Optimization
If postoperative outcomes consistently deviate in one direction (e.g., always +0.25 D hyperopic), adjust the A-constant:
- Systematic hyperopia → decrease A-constant
- Systematic myopia → increase A-constant
- Each 1 D change in IOL power ≈ 0.7 D change in refraction
- Need at least 50-100 cases for meaningful constant optimization
Accuracy Targets (Residency Benchmarks)
| Target | Standard |
|---|
| Within ±0.5 D of predicted refraction | ≥ 55% of eyes (older standard), now >70% with new-gen formulas |
| Within ±1.0 D | ≥ 85-90% |
| Mean absolute error (MAE) | ~0.35 D (best current formulas) |
| Post-op hyperopia > +1.0 D | Considered a poor outcome |
(Kanski 10th ed., p. 332 - "within 1.0 D in ~90% of patients, within 0.5 D in two-thirds")
Memory Framework: Generation Mnemonics
| Generation | Mnemonic | Formulas |
|---|
| 1st | "First Biologists Study" | Fyodorov, Binkhorst, SRK I |
| 2nd | "SRK 2 adjusts" | SRK II |
| 3rd | "3 = Three Horses Share" | Hoffer Q, Holladay 1, SRK/T |
| 4th | "Four Has Old Haigis" | Holladay 2, Haigis, Olsen |
| 5th (New-gen) | "Barrett Helped Kids Excel" | Barrett UII, Hill-RBF, Kane, EVO |
One-Line Summary Per Formula (for Quick Revision)
| Formula | One-liner |
|---|
| SRK I | A - 2.5(AL) - 0.9(K); simple regression; avoid |
| SRK II | SRK I + A-constant corrected for AL extremes |
| Holladay 1 | 3rd gen; AL+K+SF; best for normal/long |
| SRK/T | 3rd gen; theoretical regression hybrid; gold standard for AL>26 |
| Hoffer Q | 3rd gen; tangent function; gold standard for AL<22 |
| Holladay 2 | 4th gen; 7 variables; proprietary; best pre-AI broad formula |
| Haigis | 4th gen; 3 constants (a0,a1,a2); ACD independent of AL |
| Olsen | 4th gen; ray-tracing; C-constant; best theoretical accuracy |
| Barrett UII | New-gen; principal plane variable; excellent all AL; most popular |
| Hill-RBF | New-gen; pure AI; out-of-bounds warning; 3.0 current |
| Kane | New-gen; hybrid AI; currently top performer in large studies |
| EVO | New-gen; vergence + empirical; strong in short eyes |