Keratometry and IOL power formulae

Keratometry & Corneal Power

Standard manual (Javal-Schiøtz) and automated keratometers measure the radius of curvature of the anterior corneal surface ($r_a$) across a central $3.2\text{ mm}$ ring by evaluating reflected light mires. To calculate total corneal refractive power ($K$) from anterior curvature alone, instruments use the standard keratometric index of refraction ($n_k = 1.3375$):

$$K = \frac{n_k - 1}{r_a} = \frac{1.3375 - 1}{r_a}$$

This index assumes a fixed, physiological ratio between the anterior radius ($r_a = 7.7\text{ mm}$) and posterior radius ($r_p = 6.8\text{ mm}$), assigning the posterior cornea a net negative divergent power of approximately $-5.88\text{ D}$ to account for the aqueous-corneal interface.

Posterior Corneal Astigmatism & The Baylor Nomogram

Modern tomography (Scheimpflug Pentacam, AS-OCT) measures anterior and posterior corneal surfaces independently, demonstrating that the posterior cornea is not a spherical surface:

• The posterior corneal surface acts as a divergent lens with negative refractive power (mean $-6.0\text{ D}$).
• A vertically steep posterior corneal meridian commonly contributes against-the-rule astigmatic power because the posterior surface has negative power. Its magnitude and orientation vary; use a validated measured or modelled posterior-cornea approach rather than assuming an identical vector at every age.
• Consequently, standard anterior keratometry:
  • Overestimates With-The-Rule (WTR) astigmatism (steep anterior meridian at $90^\circ$) by $\sim 0.5-0.6\text{ D}$.
  • Underestimates Against-The-Rule (ATR) astigmatism (steep anterior meridian at $180^\circ$) by $\sim 0.2-0.4\text{ D}$.
• Modern toric nomograms (Koch-Wang / Baylor nomogram) and formulas (Barrett Toric, Kane Toric) integrate measured or modeled posterior corneal astigmatism to prevent undercorrection of ATR and overcorrection of WTR cylinder.

---

Evolution of Intraocular Lens Power Formulas

IOL power calculation formulas have evolved across five distinct generations, primarily differentiated by how they predict the Effective Lens Position (ELP)—the axial distance from the anterior corneal vertex to the principal optical plane of the IOL in the postoperative state.

1st & 2nd Generation Formulas (Historical Regression)

• SRK I Formula (1980): A simple linear regression equation derived from retrospective clinical data: $P = A - 2.5L - 0.9K$ where $P$ is emmetropic IOL power, $A$ is the manufacturer A-constant, $L$ is axial length (mm), and $K$ is average keratometry (D). Highly inaccurate outside normal axial lengths ($22.0-24.5\text{ mm}$).
• SRK II Formula (1988): Attempted to adjust for short and long eyes by modifying the A-constant in stepwise fashion based on axial length brackets ($L < 20\text{ mm} \to A+3; 20-21\text{ mm} \to A+2; 21-22\text{ mm} \to A+1; 22-24.5\text{ mm} \to A; >24.5\text{ mm} \to A-1$). Obsolete and unacceptable in modern practice.

3rd Generation Formulas (Two-Variable Theoretical Vergence)

Third-generation formulas use Gaussian theoretical vergence optics based on Gullstrand's model eye. They predict ELP based on two preoperative variables: Axial Length ($L$) and Corneal Curvature ($K$):

1. Hoffer Q, Holladay 1 and SRK/T are established vergence formulas with differing effective-lens-position models.
2. Older preferences by axial-length band are historical simplifications. Newer formulas with optimised constants often perform well across extreme lengths.
3. Compare outputs, measurement quality and local audited outcomes; no formula guarantees the best answer in every eye.

4th Generation Formulas (Multivariable Theoretical)

Recognizing that corneal curvature does not reliably predict anterior chamber depth (e.g., a flat cornea does not necessarily indicate a shallow anterior chamber), fourth-generation formulas incorporate additional measured anatomical variables:

• Haigis Formula: Uncouples ELP from keratometry. Uses three input variables: Axial Length ($L$), measured anterior chamber depth ($ACD$), and Keratometry ($K$). ELP ($d$) is determined by three custom regression constants ($a_0, a_1, a_2$): $d = a_0 + (a_1 \times ACD) + (a_2 \times L)$ When user-optimized with triple constants, Haigis provides outstanding predictive accuracy across all axial lengths.
• Holladay 2 Formula: Utilizes 7 variables: AL, K, ACD, Lens Thickness (LT), White-to-White corneal diameter (WTW), preoperative refraction, and patient age.

Modern Advanced Formulas (5th Generation & AI)

• Barrett Universal II: A theoretical vergence formula that models the crystalline lens as a thick lens, utilizing a virtual internal lens model to predict the true physical position and optical plane of the IOL. Inputs: AL, K, ACD, LT, and WTW.
• Hill-RBF (Radial Basis Function): An artificial intelligence / machine learning model entirely free of theoretical model-eye assumptions. Trained on multidimensional clinical datasets using pattern recognition. Features a unique boundary check indicator that alerts the surgeon if an eye falls outside the AI model's validated data envelope.
• Kane formula: Combines theoretical optics and data-driven methods. Its performance, like other modern formulas, depends on measurement quality, lens constants and the population; avoid an unconditional claim of superiority.

---