SAFETY ENGINE

Every plan — no exceptions — passes through binding safety floors before it reaches a signature. This page describes the actual numbers, because a safety claim without numbers is a slogan.

THE FLOORS, WITH THEIR VALUES

• PTA — percent tissue altered,^[1] ceiling dependent on Kmax: 40% when Kmax ≥ 46 D, 45% below; binding in the LASIK family, informational (with honest labelling) in procedures where the threshold was never validated
• RSB — residual stromal bed:^[2] 300 µm for LASIK/iLASIK/Q-LASIK; 280 µm for lenticule procedures with a 470 µm pre-op minimum; graded 300–350 µm for surface ablation depending on biomechanical evidence
• RCT — residual corneal thickness ≥ 400 µm, controlled independently: no flap choice can repair it, only reducing ablation can
• Post-op K window (Min–Max K) — a pass/fail test, from which each cornea's maximum treatable correction is computed
• Age & stability — 18 minimum (18–21 flagged for surgeon's discretion), 12-month refractive stability, medication and chronic-disease screening on every case

DISCIPLINE IN THE COUNTING

Ectasia indices are counted with correlation awareness — derived indices never masquerade as independent evidence. Missing data is never assumed normal; it is named in the report. Device floors are respected but never relied on when RLES AI's own floors are stricter: a laser that warns at 250 µm of stroma does not loosen a 300 µm rule.

WHAT THE ENGINE NEVER DOES

It never rules that surgery "cannot be done". It states which floor a plan violates and by how much, shows what parameter change would repair it, and offers the alternatives — a smaller zone, a different procedure, an ICL pathway. The verdict belongs to the surgeon; the engine's job is to make sure the verdict is informed.

WHAT ONE PLAN ACTUALLY REQUIRES

A safety floor only protects a plan if every value it depends on has been collected, read and calculated. The table below counts what a single case demands of the surgeon in each procedure family: questions asked at history-taking, headings checked at the slit lamp, values read from device printouts, and calculations performed before the plan is fixed. Figures are for both eyes and are counted from the RLES AI rule set — this is not a literature claim but the inventory of items the software audits in every case. "Recall" counts the thresholds, rules and device nomogram rows a surgeon must hold in memory to interpret those items.

History and exam columns are shown as "headings / items". Total items = history items + exam values + test values + calculations; recall rows are not included in the total. In the laser row, opening the optical zone and flap alternatives raises the calculation count into the several hundreds.

EVERY UNMEASURED OR UNCALCULATED VALUE IS A RISK

Some of these items are not measured but derived; a value that is never calculated is invisible, and it steers the plan silently. The evidence in refractive surgery is explicit. In the first case-control series of post-LASIK ectasia, not one ectasia case lacked an identifiable pre-operative risk factor: 88% met forme fruste keratoconus criteria and 70% had a residual stromal bed below 250 µm.^[3] Abnormal pre-operative topography was present in 35.7% of ectasia eyes and in none of the controls, and a weighted risk score built from topography, residual bed, age and pachymetry separated at-risk eyes with 96% sensitivity and 91% specificity.^[2] In eyes whose topography was normal, the strongest single marker was a number no device displays — percent tissue altered: PTA ≥ 40% was present in 97% of eyes that developed ectasia, with an odds ratio of 223.^[1]

The surface being measured must be questioned too: 12% of contact-lens-wearing refractive surgery candidates showed topographic warpage, with refractive shifts up to 3.0 D and keratometric shifts up to 2.5 D, and stabilisation took 1 to 20 weeks.^[4] A measurement accepted without that question feeds a correct calculation with a wrong input.

Lens surgery is no different. About 35% of the error in intraocular lens power comes from predicting effective lens position and 17% from axial length measurement;^[5] ignoring posterior corneal astigmatism produces an error above 0.50 D in 5% of eyes and biases toric selection toward overcorrection in with-the-rule and undercorrection in against-the-rule astigmatism.^[6] None of these is a rare event — each is a single item that can be skipped in a routine case.

The value of systematic checking has been quantified in surgery at large: a 19-item surgical safety checklist reduced complications from 11.0% to 7.0% and death from 1.5% to 0.8% across eight hospitals in eight countries.^[7] Its power came not from new knowledge but from every known item being verified in every patient. That is precisely what the safety engine does: the items above are audited in the same order and against the same thresholds in every case — what was not measured is written as "not reported", what must be calculated is calculated, and no item is left to memory on a busy operating day.

FREQUENTLY ASKED QUESTIONS