Answers are grounded in this paper and cite their sources.
The Shead et al. article about an improvement to the SRK/T formula,1 describes a method to remove the “cusp characteristic”2 from the calculated intraocular lens (IOL) power curve as well as reduce an apparent small linear bias in the corneal height as a function of the mean corneal power. Figures 4 and 7 in the article nicely identify these issues. Based on a large sample of cases, Shead et al. state that with the SRK/T formula, 98% of eyes were within ±1.00 D and 78.7% were within ±0.50 D of the predicted outcome compared with 98.5% and 80.9%, respectively, with the T2 modification. They also state that the mean absolute prediction error was 0.3229 D with the SRK/T formula compared with 0.3064 D with the T2 modification. The claim that prediction error with the T2 modification was 9.7% less than the error with the SRK/T is based on the slope of a best-fit line to scattered data, showing the SRK/T prediction error on 1 axis and the T2 prediction error on the other. Casual inspection of the data and best-fit line in Figure 5 indicate that a rotation to a slope closer to 1 would better fit the data in the region between 1.00 D and 2.00 D and that a single outlier point appears to be the reason a slope closer to 0.9 was obtained in the line fit. Overall, the authors do show a modest improvement in the performance of the T2 modification over the standard SRK/T, but it appears the claim of a 10% improvement in outcomes is not justified. In any event, given the current step size of IOL powers, a reduction in 0.0165 D in mean absolute prediction error is interesting but not a large enough improvement to be clinically significant.
No takes yet. Share an insight, caveat, or question.
Edwin J. Sarver (2011) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: