In eyes longer than 32 mm, drop Hoffer QST and use Cooke K6, EVO, Pearl-DGS or Zhu-Lu

Design

Retrospective observational formula comparison, prediction error zeroed per formula and lens model

Population

330 Asian cataract eyes with axial length 32.00 mm or more (307 at 32 to 35 mm, 23 at 35 mm or more)

Primary outcome

Root mean square absolute error, median absolute error and proportion within 0.50 D

Effect

Main group: Zhu-Lu 0.562 D and Cooke K6 0.563 D root mean square error, Pearl-DGS 75.24 per cent within 0.50 D, Hoffer QST worst at 0.682 D and 61.89 per cent

Three hundred and thirty Asian eyes with axial length of 32.00 mm or more were used to compare eight modern intraocular lens power formulas, with prediction errors arithmetically zeroed per formula and per lens model so that systematic bias could not flatter any of them. Because Hill-RBF and Kane refuse an axial length above 35.00 mm, the analysis was split into a main group of 307 eyes between 32.00 and 35.00 mm and an ultra-long group of 23 eyes at 35.00 mm or more.

In the main group Zhu-Lu had the lowest root mean square absolute error at 0.562 D with Cooke K6 immediately behind at 0.563 D; Cooke K6 had the lowest median absolute error at 0.271 D, with EVO 2.0 and Pearl-DGS at 0.272 D. Pearl-DGS put the most eyes within 0.50 D at 75.24 per cent, and Zhu-Lu the most within 1.00 D at 93.49 per cent. Hoffer QST was last on every metric: 0.682 D root mean square error, 0.369 D median error, 61.89 per cent within 0.50 D, and a hyperopic drift that grew with axial length. In the 23 ultra-long eyes Zhu-Lu led with 0.438 D and 78.26 per cent within 0.50 D.

The practical change is small and immediate: for a long eye, use one of the four formulas that performed, and check which one your biometer defaults to. Long eyes are common in Indian cataract lists, and all four of these calculators are freely available online, so this costs nothing but the habit.

• Set a second formula on the biometer for eyes above 32 mm rather than accepting the machine default.
• Avoid Hoffer QST in this range; its error grew as the eye got longer.
• Above 35 mm, remember Hill-RBF and Kane will not accept the input at all.
• Warn the patient that roughly one eye in four still lands outside 0.50 D of target at this axial length.
• Audit your own long-eye refractive outcomes before adopting any single formula wholesale.

The statistics, in plain English

Root mean square absolute error punishes large misses, median absolute error describes the typical eye, and the percentage within 0.50 D is what the patient experiences; the four leading formulas trade places depending on which you weight, which is why no single winner is named. The ultra-long group holds only 23 eyes, so its ranking is fragile and should be read as a hint rather than a result. This is a retrospective single-population study in Asian eyes, and prediction errors were adjusted to zero mean, so it compares consistency rather than out-of-the-box calibration.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.