The Dependency of LogMAR Visual Acuity Measurements on Chart Design and Scoring Rule

The study entitled “The Dependency of logMAR Acuity Measurements on Chart Design and Scoring Rule” by Charlotte A. Hazel and David J. Elliott 1 raises some important issues. For example, their comparison between psychometric function and letter by letter scoring, may have been compromised by their use of an 8 letters per line (Regan) chart. In another study 2 it was found that even 10 letters per line were not sufficient for reliably estimating the threshold and slope of psychometric function. Hazel and Elliott 1 demonstrated the influence of different chart designs on acuity scores. However, even different versions of the same chart design can give rise to significant differences in letter 3 and line difficulty. These differences result in the loss of chart equivalence of the same design. 4 It is possible that a lack of chart equivalence occurs in most letter chart designs when the frequency of appearance of each subset letter is varied from one line to the next. The exception appears to be a 10 letters per line design 4-7 that gives the advantage of chart equivalence for different versions providing the same 10 letter subset is used in each version. A 10 letters per line design reduces the influence of the different degrees of readability and legibility across any 10 letter subset and creates equal scaling intervals between lines. 4, 8 Hazel and Elliott 1 refer to the importance of chart equivalence so that comparisons can be made in clinical trials of ophthalmic interventions. The increased precision of the 10 letter design 8 not only achieves chart equivalence for different versions but also offers the potential for significantly reducing the sample size needed in such clinical trials, with associated reduced funding requirements. However, reliability of acuity assessment is no less important for the detection and management of disease, as well as other routine clinical applications. Hazel and Elliott 1 draw attention to the importance of chart design, and there are many other variables that can reduce the reliability of acuity assessment. 9 More of them need to be considered if reliability is to be raised to acceptable standards. For example, when frequent assessment is required, reliability of acuity assessment also depends on the potential for chart memory to improve scores. 9 In some research and clinical applications, control of chart memory 10 may contribute to greater reliability of measurement. The cumulative effects of small improvements in reliability by multiple means appear to be necessary for statistically significant changes in acuity to become a useful adjunct to clinical judgment in detecting significant change. While confidence limits for significant change remain at the level of 5 letters, 11 one line on many charts, clinical judgment will substitute for statistical significance, 9 and clinical trials will need to be larger to achieve significant results. An improvement in reliability of only one-half of a letter gives a reduction in the confidence limits for significant change of 20%. Another one-half of a letter improvement will reduce confidence limits by an additional 25%. One of the core issues for reliability appears to be the variable legibility/readability of letters in any subset. To achieve acceptable reliability it may be necessary to grade errors according to their varying letter legibility/readability and to recognize reasonable substitutes as partially correct responses. 12 Charles W. McMonnies