Purchase this article with an account.
Rui Wu, Ruth E Manny, Michele Melia, Elizabeth Lazar, Jonathan M Holmes, Eileen E Birch, Susan A Cotter, Raymond Kraker, David K Wallace; Classifying No Improvement in Visual Acuity in Children with Amblyopia Using Simulations. Invest. Ophthalmol. Vis. Sci. 2018;59(9):178.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Classification of “no improvement” of visual acuity (VA) is important for treating children with amblyopia. Current practice may result in misclassification impacting treatment decisions because it does not fully account for measurement error. We used simulations to evaluate the performance of rules classifying “no improvement” in VA.
Monte Carlo simulation generated a true amblyopic-eye VA (20/40 to 20/400) and 3 observed VA (initial test, retest, and 2nd retest) at each of 3 visits for 10,000 hypothetical subjects. The observed VA was calculated by adding random measurement error (based on established test-retest error) to the true VA. Eighteen rules for classifying no improvement based on observed VA were evaluated where true VA did not change over time or improved by a constant or decreasing amount between successive visits. For each rule, sensitivity and specificity were calculated.
The typical clinical practice of comparing a single test of VA at 2 consecutive visits had a sensitivity of 70% and a specificity (for subjects with 1 line true improvement) of 72%. Adding both a same-day retest (at each visit) and a subsequent visit to confirm “no improvement” and comparing best VA tested at each visit improved the performance (sensitivity: 80%, specificity: 80%). Comparing the average of test and retest VA at each visit instead improved specificity at the expense of sensitivity (sensitivity: 73%, specificity: 95%). Adding 2 same-day retests (at each visit) and a subsequent visit to confirm “no improvement” and comparing average of 3 tests at each visit also improved performance (sensitivity 86%, specificity: 93%).Generally, rules that required 3 successive visits to demonstrate “no improvement” had considerably higher specificities (25% to 61% increase) but lower sensitivities (14% to 60% decrease) compared with rules that required only 2 visits. Specificity decreased when the amount of true VA improvement decreased and both sensitivity and specificity decreased when measurement error increased.
Adding 1 or 2 same-day retest(s) at each visit and a 3rd visit to confirm “no improvement” improved accuracy in classification of “no improvement” of VA among children with amblyopia. Although requiring more time and visits, such strategies are worth considering, depending on the specific treatment.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.
This PDF is available to Subscribers Only