Visual acuity (VA) test is a key component of visual function

Transcription

1 ORIGINAL CLINICAL STUDY Factors Contributing to Discrepancy Between Visual Acuity Fractions Derived From a Snellen Chart and Letter Scores on the Early Treatment Diabetic Retinopathy Study Chart Fred K. Chen, MBBS (Hons), PhD, FRANZCO,*Þþ Lillian Evangelia Agelis, BSc (Hons),Þ Khaik K. Peh, MB ChB, MRCOphth,Þ Joanne Teong, BA, LLB, GDipLegalPrac, MBBS,* and Evan Norman Xi Ming Wong, MBBS*Þ Purpose: To report factors influencing the relationship between visual acuity (VA) fractions measured on Snellen chart and letter scores on the Early Treatment Diabetic Retinopathy Study (ETDRS) chart. Design: This was a retrospective review from a single ophthalmology outpatient clinic. Methods: All patients had routine consecutive VA testing in the right eye using a Snellen chart (1Y6 m) and the ETDRS chart (4 m and/or 1 m), by the same optometrist, using a standardized testing protocol for each chart. Both acuity fractions and letter scores were converted to their equivalent logarithm of minimum angle of resolution (logmar) for comparison. Multiple regression analysis was performed. Results: A total of 237 patients with a wide range of ocular disease and VAs were enrolled. Mean age was 63 years (range, 18Y95 years). Recorded VA (logmar) was better on Snellen chart by a mean (95% limits of agreement) of j0.07 (j0.33 to +0.18, P G 0.001). There was a trend for subjects with VA worse than 6/36 to read 3 to 6 letters more and those better than 6/36 to read 4 to 5 letters less on the ETDRS chart. Visual acuity level and amblyopia contributed to a small portion (17%Y26%) of the total variance in the difference between logmar equivalents derived from the 2 charts. Conclusions: The discrepancy in VA derived from Snellen and ETDRS charts was nonuniform across VA range. This has implications on interpretation of published studies converting Snellen fractions to logmar for analysis and reporting of VA outcomes. Key Words: visual acuity measurement, visual acuity chart, amblyopia, macular degeneration (Asia Pac J Ophthalmol 2014;3: 277Y285) Visual acuity (VA) test is a key component of visual function assessment. The 2 most common types of charts used to From the *Centre for Ophthalmology and Visual Science (Lions Eye Institute), The University of Western Australia, Nedlands; Department of Ophthalmology, Royal Perth Hospital, Perth; and Department of Ophthalmology, Princess Margaret Hospital for Children, Subiaco, Western Australia, Australia. Received for publication April 07, 2013; accepted August 30, F.K.C., L.E.A., and K.K.P. are considered as equal first authors. This work was presented at the Annual Meeting for the Association or Research in Ophthalmology and Visual Science F.K.C. is funded by the Viertel Charitable Foundation Investigatorship and the Retina Australia (WA) Grant. The other authors have no funding or conflicts of interest to declare. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal s Web site ( Reprints: Fred K. Chen, MBBS, PhD, FRANZCO, 2 Verdun St, Nedlands, Western Australia 6009 Australia. fredchen@lei.org.au. Copyright * 2013 by Asia Pacific Academy of Ophthalmology ISSN: DOI: /APO measure VA are the Snellen-style chart developed 150 years ago by the Dutch ophthalmologist Hermann Snellen 1 and the logarithm of minimum angle of resolution (logmar)ystyle chart popularized 35 years ago by the Australian optometrists Bailey and Lovie. 2 The logmar-style chart has been recommended as the preferred standardized method by the National Academy of Sciences, National Research Council in 1980, 3 adopted by the Early Treatment Diabetic Retinopathy Study (ETDRS) since 1982, 4 and has also been incorporated into the British Standard BS:4274 in 2003 replacing the 1968 edition. Despite these recommendations and the numerous advantages of the ETDRS chart over the various Snellen-style charts, 5 many eye clinics continue to use the Snellen chart because of familiarity, lower cost, shorter time to obtain measurement, and the greater number of letters per line in the lower portion of the chart (6/9 to 6/4 range of VA). Several journals have recommended reporting and analyzing VAs in logmar equivalents of (1) acuity fractions derived from a Snellen-style chart or (2) letter score derived from a logmar-style chart. Visual acuity conversion tables are also published regularly to help readers who are not familiar with logmar notations, aiding interpretation of published results. 6 The assumption in this recommendation is that a recorded VA of 6/60 (or 20/200) on the Snellen chart is equivalent to a letter score of 35 from the ETDRS chart, and both can be assigned the same logmar value of This recommendation has 2 important implications: (1) VA data derived from either Snellen or ETDRS charts can be pooled for statistical analysis after logmar equivalents are calculated, and then the results can be converted back to equivalent acuity fractions or letters scores for ease of interpretation, and (2) eye clinics can directly switch from any of the Snellen-style charts to logmarstyle charts without the need to calibrate one chart against another. Two studies have challenged the above assumption by demonstrating discrepancy in logmar equivalents of VAs derived from a Snellen-style chart and the ETDRS chart. 7,8 These studies showed overall differences of 0.1 to 0.4 in logmar equivalents derived from a Snellen-style chart as compared with that derived from the ETDRS chart depending on VA range. To investigate factors influencing the variance between these 2 charts, we retrospectively analyzed the VA measurements recorded during a transition phase of switching our vision testing protocol when both Snellen and ETDRS charts were used on every patient. We reported the overall bias between the 2 charts and the variations in actual ETDRS letters score for each Snellen fraction and explored the predictive factors for discrepancy in logmar equivalents derived from the Snellen and ETDRS charts. MATERIALS AND METHODS This was a retrospective chart review of the consecutive patients who underwent VA measurements using both the Snellen Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October

2 Chen et al Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October 2014 years (range, 18Y95 years). There were 2 overlapping cohorts among these 237 patients: those who read the ETDRS from 4 m (n = 223) and those who also or had to move to 1 m to read the ETDRS chart (n = 116). Ocular diagnoses of the tested eye were corneal and conjunctival diseases (23 eyes, including corneal graft, pterygium, keratoconus, keratitis, and conjunctivitis), cataract (18 eyes), posterior segment diseases (100 eyes, including macular degeneration, macular edema, macular traction, ocular inflammation, tumor, and trauma), optic neuropathy (35 eyes, including glaucoma, optic neuropathy, and optic neuritis), and amblyopia (10 eyes). The remaining 51 eyes had normal ocular examination. Only 96 patients (41%) were wearing spectacle correction. FIGURE 1. Vision testing lane in the ophthalmology outpatient clinic at study site. chart and the ETDRS chart in the right eye at the Eye Clinic of Royal Perth Hospital, Western Australia, from June 2011 to August During this period, only the right eye of each patient underwent VA test with both charts. This study conformed to the Declaration of Helsinki for clinical research involving human subjects and has been approved by the Hospital Audit Committee. Patients A total of 237 patients were identified to have had VA measurements using both Snellen and ETDRS charts. Among these, 116 (49%) were female. The preference for the order of testing was to use the Snellen chart first. However, 117 patients (49%) had to read the ETDRS chart first to facilitate clinic flow because the Snellen chart was used by other examiners for other patients at the time of vision test. The mean age was 63 T 18 Visual Acuity Charts and Vision Lane Visual acuity testing for both charts was performed in the same vision lane, under the same lighting conditions. The ambient illuminance of the vision lane was 290 lux at 6 m, 240 lux at 4 m, and 420 lux at 1 m from the charts (Unfors XI Base Unit; Inderlec Medical, Billdal, Sweden). A retroilluminated wall-mounted Snellen chart (LongLife TM Project-O-Chart; Reichert Inc, Depew, NY) was used with the patient standing at 6 m from the chart at the beginning of the test (Fig. 1; (Supplemental Digital Content 1, Similarly, a retroilluminated box on a standard light box holding chart 1 of the ETDRS charts (Lighthouse International, New York, NY) was used for measuring VA of the right eye with the patient standing at 4 m from the chart at the beginning of the test (Supplemental Digital Content 2, Luminance over the charts was measured using a photom (Unfors XI Base Unit; Inderlec Medical). For the Snellen chart, the mean luminance (cd/m 2 )over the chart was 2300 (2660 at the top, 1130 at the center, and 3130 at the bottom portions of the chart; coefficient of variation: 0.45). For the ETDRS chart, the mean luminance (cd/m 2 ) was 270 (270 at the top left corner, 290 at the top right corner, 260 at the bottom left corner, 280 at the bottom right corner, and 260 centrally; coefficient of variation: 0.05). Vertical optotype sizes were measured with a ruler TABLE 1. Expected Versus Actual ETDRS Letter Scores Snellen Fraction No. Eyes at Each Snellen Fraction 6m Snellen Chart No. Eyes Also Tested With 4m 1m Expected ETDRS Letter Scores 6m Snellen Chart 4m Actual ETDRS Letter Scores T SD 1m 1/60 3 V 3 j4 V 11.0 T 7.8 2/60 3 V 3 11 V 9.3 T 5.1 3/ T T 8.2 4/ T T / T T 8.0 6/60, T T / T T 8.1 6/ T T 7.0 6/ T T / T T 4.1 6/ T T 8.9 6/ T / V T 2.6 V 6/ V T 4.1 V Total Boldface represents similar letter sizes between Snellen and ETDRS charts * 2013 Asia Pacific Academy of Ophthalmology

3 Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October 2014 VA Fractions Measured on Snellen & s TABLE 2. Summary of logmar Equivalent From Snellen and s Snellen Chart Acuity Fraction (logmar) (n = 237) ETDRS 4-m Letter Score (logmar) (n = 223) ETDRS 1-m Letter Score (logmar) (n = 116) Minimum VA (logmar) 1/60 (1.78) 31 (1.08) 2 (1.66) Maximum VA (logmar) 6/4 (j0.18) 100 (j0.30) 70 (0.30) Mean VA (logmar) 6/12 (0.33) 68.1 (0.34) 70.0 (0.73) Median VA (logmar) 6/9 (0.18) 48.5 (0.30) 52.0 (0.66) FIGURE 2. Bland-Altman plots showing relationship between absolute differences and mean. logmar equivalents derived from Snellen and ETDRS charts at 4 m (A) and 1 m (B). Difference against mean with mean difference and 95% limits of agreement marked by dotted line for ETDRS charts at 4 m (C) and 1 m (D) versus Snellen acuity. * 2013 Asia Pacific Academy of Ophthalmology 279

4 Chen et al Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October 2014 FIGURE 3. Histograms showing distribution of actual ETDRS letters scores at 4 m (A) and 1 m (B) for each Snellen fraction. Expected ETDRS letter scores (equivalent to the Snellen fraction) are shown on the top of each histogram * 2013 Asia Pacific Academy of Ophthalmology

5 Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October 2014 VA Fractions Measured on Snellen & s and converted to angular angles in minutes of arc (Supplemental Digital Content 1 and 2, and for comparison between the 2 chart designs. Visual Acuity Measurement All VA measurements were performed in undilated eyes. The right eye of each patient was tested first, consecutively with the Snellen and then the ETDRS charts or vice versa at the discretion of the assessor. Vision test was performed by a single examiner (L.A.). Refraction was not performed during VA testing as this was not routine in our general ophthalmology clinic. Patients were tested with habitual spectacle prescription (if available) and then with pinhole correction. If worn, the pinhole occluder was placed over the spectacle. For the Snellen chart test, subjects were simply instructed to read from the top of the chart to as far down as they could see. Although encouraged to read as many letters as possible, patients were not asked to guess letters as required in a forcedchoice testing protocol. Vision test was completed when the patient decided he/she was not able to read any more letters. If the patient failed to read the top letter from 6 m, he/she was moved closer to the chart 1 m at a time until the top letter can be read correctly. Visual acuity was recorded using the lineassignment approach. Visual acuity was assigned to the lowest line with more than 50% of the letters (ie, 3 letters in a row with 4 letters or 4 letters in a row with 7 letters) read correctly, minus the number of letters read incorrectly. If 50% or less of the letters (ie, 2 letters in a row with 4 and 2 letters in a row with 5 letters) were read correctly, VA was assigned to the line immediately above plus the number of letters read correctly in that line. The best VA achieved with or without pin hole was recorded. For the ETDRS chart test, patients were asked to read the chart from the top to as far down as they could see starting from a viewing distance of 4 m from the chart. The lowest line at which the patient can correctly read all 5 letters was noted, and any errors made in the lines above were disregarded. Forced-choice protocol was not incorporated to reduce testing time. If the patient was unable to read more than 30 letters with or without a pinhole occluder at 4 m, he/she was moved to 1 m from the chart to repeat the same procedure. The final letter score was calculated by adding 30 to the total number of letters read correctly at 4 m (letter-by-letter scoring). Alternatively, the letter score was simply the sum of numbers of letters read correctly at 1 m. The maximum number of letters read at 1 and 4 m was recorded. Statistical Analysis The data are presented in mean, SD, median, and range. The logmar equivalent acuities were calculated using the following formulas: VA ð ETDRS into logmar equivalentþ ¼0:0 2xð 85Y letter scoreþ VA ð Snellen into logmar equivalentþ ¼ Log 10 ð acuity fractionþ The null hypothesis was tested by the Wilcoxon signed rank test comparing VA (ETDRS in logmar equivalent) at 1 and 4 m to VA (Snellen in logmar equivalent). The relationships between differences in VA versus mean VAs (logmar) were examined using nonparametric Randall correlation. Bland-Altman analysis was performed to determine the relationship between differences and mean. Ninety-five percent limits of agreement were calculated. 9 To facilitate direct comparison between Snellen chart acuity fractions and ETDRS chart letter scores, the following formula was used to convert VA fractions in the letter score equivalents, as suggested by Gregori et al 10 : VA ð Snellen intoetdrs letter score equivalentþ ¼ 85þ½ 50 x Log 10 ðacuity fractionþš Patients were divided into 14 subgroups representing each of the 14 possible Snellen fractions: 1/60, 2/60, 3/60, 4/60, 5/60, 6/60, 6/36, 6/24, 6/18, 6/12, 6/9, 6/6, 6/5, and 6/4. The mean, SD, and range of actual ETDRS letter score for each Snellen fraction allocation were calculated. Separate analysis was performed for the 4- and 1-m test scores because they are not directly comparable. 11,12 One-sample t test was used to compare actual ETDRS letter score to expected letter score equivalent for selected Snellen lines (3/60, 6/60, 6/24, 6/12, and 6/6) that have the same font sizes as the ETDRS chart at 4 m (lines 1, 5, 8, and 11; Supplemental Digital Content 1, Hierarchical multiple regression was used to assess whether ocular diagnostic category (normal, anterior segment abnormality, posterior segment abnormality, and amblyopia) predicts discrepancy between Snellen and ETDRS charts after controlling for 5 independent variables including; age, sex, order of chart used for vision test, presence of habitual refractive correction, and VA on Snellen chart (in logmar equivalent). These 5 variables were entered in the first block, and then diagnostic category was entered stepwise in the second block. Criterion for removal of a variable was set at an F ratio probability of 0.1. All effects are reported as significant at P G 0.05 incorporating Bonferroni correction within multiple regression analysis and analysis of variance post hoc tests. Statistical analysis was performed using SPSS (SPSS Inc, Chicago, Ill). RESULTS Visual Acuity Distribution on Snellen and ETDRS Charts Visual acuity, as measured on the Snellen chart, ranged from 1/60 to 6/4 (logmar of 1.78 to j0.18). There were 30 eyes with VA of worse than 6/36. A total of 10 and 13 eyes had acuities of 6/60 and 6/36, respectively (Table 1). The mean and median VAs of the entire cohort based on the Snellen chart (logmar) were 0.33 and 0.18, respectively. At 4 m, 223 (94%) of 237 eyes were able to read 1 (recorded as letter score of 31, logmar of 1.08) to 70 (letter score of 100, logmar of j0.30) letters on the ETDRS chart. The remaining 14 eyes and 102 of the 223 eyes (a total of 116 eyes, 49%) read the ETDRS chart from 1 m. The range of letter scores at 1 m was 2 (logmar of 1.66) to 70 (logmar of 0.30; Table 2). Visual Acuity Comparison After Conversion Into logmar Equivalents For the 223 eyes that were able to read the ETDRS chart from 4 m, their logmar equivalent derived from the Snellen chart was significantly better than the logmar equivalent derived from ETDRS chart (null hypothesis was rejected with a mean difference of j0.07 T 0.13 logmar, Wilcoxon signed rank test, Z = j0.847, P G 0.001). Similarly, the logmar equivalent derived from Snellen chart was significantly better than the logmar equivalent derived from ETDRS chart in those 116 eyes, which were tested at 1 m on the ETDRS chart (null hypothesis was rejected * 2013 Asia Pacific Academy of Ophthalmology 281

6 Chen et al Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October 2014 with a mean difference of j0.08 T 0.19 logmar, Z = j0.405, P G 0.001). There was no statistically significant relationship between absolute difference and the mean for Snellen and ETDRS acuities at 1 m (r = 0.12, P = 0.07). A very weak positive correlation exists between absolute difference and the mean for Snellen and ETDRS acuities at 4 m (r = 0.14, P = 0.002). The 95% limits of agreement between the logmar equivalent derived from the Snellen chart and the ETDRS chart at 4 and 1 m were j0.33 to +0.18, and j0.47 to +0.31, respectively. Figure 2 shows Bland-Altman plots of differences against mean after conversion of Snellen acuity fractions and ETDRS letter scores to logmar equivalents. A Comparison Between Actual (Measured on ) and Expected (Calculated From Snellen Chart Measurement) Letter Scores To facilitate direct clinical application of the results, the mean of actual ETDRS letter scores were compared with the expected ETDRS scores derived from Snellen fractions (Table 1). One-sample t tests was performed to compare the mean actual ETDRS letter score with expected letter score equivalent for eyes allocated Snellen fractions of 3/60, 6/60, 6/24, 6/12, and 6/6 corresponding to 20-, 35-, 55-, 70-, and 85-letter score equivalent. There was a trend for higher mean actual ETDRS letter scores (1 m) for eyes with 3/60 and 6/60 VA compared with their expected letter score equivalents (25 vs 20 [P = 0.14] and 36 vs 35 [P = 0.88], respectively). The actual ETDRS letter score was greater than the expected letter score (from Snellen chart) by 3 letters (P = 0.195) and 6 letters (P = 0.003) for 1- and 4-m charts. Conversely, there was a significant difference between mean actual and expected letter score equivalents for eyes with 6/24, 6/12, and 6/6 acuity fractions on the Snellen charts (50 vs 55 [P = 0.002], 66 vs 70 [P = 0.001], and 79 vs 85 [P = 0.001], respectively). Figure 3 shows the distribution of actual mean ETDRS letter scores for each Snellen acuity fraction. The correlation between logmar equivalent of ETDRS letter score versus logmar equivalent of Snellen acuity fraction is shown in Figure 4. A sensitivity analysis showed similar results when we excluded eyes that could not read all the letters or read additional letters below the allocated Snellen fraction (data not shown). Prediction Model for Discrepancy in logmar Equivalents Derived From Snellen and s The most significant predictors for the differences between the logmar equivalents derived from the Snellen chart and the ETDRS chart at 4 m were VA on Snellen chart, age, vision test using ETDRS letter chart first, and diagnosis of amblyopia. The 5 variables in block 1 contributed to 12.7% of the variance, F 5,217 =6.29,P G 0.001, and ocular diagnosis of amblyopia only contributed to an additional 3.8%, F 6,216 =5.457P G The trend for Snellen chart to overestimate VA was reduced in those with poorer VA (A = 0.434, P G 0.001) and increased with increasing age (A = j0.154, P = 0.022), use of the ETDRS chart first (A = j0.152, P = 0.026), and in eyes with amblyopia (A = j0.200, P = 0.002; Table 3). A similar outcome was found in the differences between the logmar equivalents derived from Snellen chart and the ETDRS chart at 1 m. Again, amblyopia (A = j0.202, P = 0.017) was a significant predictor of overestimation of VA by Snellen charts compared with ETDRS chart, and this was reduced in eyes with poorer VA (A =0.438,P G 0.001; Table 3). Block 1 variables explained 21.6% of the variance, F 5,110 = 6.047, P G 0.001, and amblyopia only contributed to an additional 4%, F 6,109 = 6.247, P G DISCUSSION In this retrospective study, we compared the logmar equivalents of Snellen acuity fractions and ETDRS letter scores. Three important findings were as follows: one, our Snellen chart gave an overall better VA (logmar) by 0.07 to 0.08 (3.5Y4.0 letters) compared with the ETDRS chart. Two, for eyes assigned to a particular Snellen fraction, there was a range of actual ETDRS letter score, and this discrepancy was not uniform throughout the range of VA. For example, the mean of actual ETDRS letter score was higher than the expected letter score in eyes with worse than 6/36 vision, whereas in eye with better than 6/36, the actual score was lower than expected score. Three, 4% of our study population had amblyopia, and they contributed to only 4% of the variance in VA measurement between the charts, after controlling for VA, age, sex, test sequence, and history of habitual refractive correction. The majority (75%Y84%) of the variance in VA measures between the Snellen and ETDRS charts could not be predicted by our regression model. FIGURE 4. Correlation between VA measured on the ETDRS chart at 4 m (A) or 1 m (B) and the Snellen chart. The x axis and y axis are the logmar equivalent of Snellen fractions and VA letter scores, respectively. The yellow line represents the equivalence line if ETDRS letter score and Snellen acuity fraction were truly interchangeable. The blue line represents the regression line based on raw data. The brown dotted line represents floor effect at 1.10 logmar at 4 m (A) and ceiling effect at 0.30 logmar at 1 m (B) on the ETDRS chart * 2013 Asia Pacific Academy of Ophthalmology

7 Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October 2014 VA Fractions Measured on Snellen & s TABLE 3. Multiple Regression for Predictors of Difference Between Snellen Versus 4 m ETDRS and Snellen Versus 1 m ETDRS in logmar Unstandardized Coefficients (95% Confidence Interval) Predictors in Final Model Difference in logmar SnellenVETDRS 4 m Difference in logmar SnellenVETDRS 1 m Constant j0.020 (j0.081, 0.041) j0.220 (j0.375, j0.066)* Snellen VA (logmar) (0.105, 0.206) (0.139, 0.310) Age, y j0.001 (j0.002, 0.000) (j0.001, 0.002) Female vs male (j0.019, 0.043) (j0.051, 0.080) Using ETDRS chart first (1 vs 0) j0.039 (j0.073, j0.005) j0.021 (j0.091, 0.048) Habitual correction (1 vs 0) j0.022 (j0.055, 0.010) j0.052 (j0.118, 0.014) Amblyopia versus normal (1 vs 0) j0.129 (j0.211, j0.048)* j0.140 (j0.254, j0.026) Block 1 (5 variables entered) R Adjusted R F ratio Block 2 (diagnosis stepwise) R Adjusted R F ratio *P G P G P G Discrepancy between the various types of Snellen charts and the standardized ETDRS chart has been reported. 7,8 However, several ophthalmology journals continue to recommend direct mathematical transformation of Snellen fractions or decimals into a logmar scale so that it can be interpreted alongside measurements derived from the ETDRS charts. 4,13 More recently, Gregori and colleagues 10 also recommended direct conversion of acuity fraction derived from Snellen charts to equivalent ETDRS letter scores for easier statistical manipulations and more intuitive interpretation of data, using the formula: approximate ETDRS letter score = log(snellen fraction). A number of smaller studies comparing Snellen, ETDRS, or compact reduced logmar charts in a general clinical setting reported no statistically significant differences between VAs derived from these charts. 14,15 However, these studies were primarily designed to examine testretest variability, and hence the sample sizes were relatively small (around 40 patients). Falkenstein et al 8 reported a difference in VA measurements obtained from Snellen and ETDRS charts in 190 eyes of 104 participants. In contrast to our finding, they reported that, on average, the Snellen chart underestimated VA by 2.5 lines compared with that obtained from the ETDRS chart at 2 m. However, their cohort of patients was from a retinal practice, in which 77% of the patients had age-related macular degeneration. They also utilized a line-assignment method for stopping and scoring of the ETDRS letter scores. Kaiser 7 also demonstrated that the Snellen chart gave an equivalent VA score that was 6.5 letters lower than the ETDRS chart at 4 m. His study cohort of 163 eyes of 163 patients encompassed a wider range of pathology, although 56% were patients with age-related macular degeneration. The study also used a forced-choice paradigm in ETDRS test. 7 The Snellen chart used in our department has a design common to that used in many other eye clinics around the world. The chart is wall mounted close to the ceiling so that it does not occupy floor space, but the consequence is that it is located above the eye level of the patient, and the distance between the eye and 6/60 letter of the chart is effectively increased. However, this effect cannot explain the overestimation of VA in the range of 6/24 to 6/4. The 10-fold greater luminance of the Snellen chart light box is also unlikely to cause significant difference because there is negligible improvement of VA between 250 and 2500 cd/m We also measured the actual size of the letters on the charts and confirmed that the 6/60, 6/24, 6/12, and 6/6 lines on the Snellen chart have similar letter sizes compared with lines 1, 5, 8, and 11 on the ETDRS chart from 4 m (Supplemental Digital Content 1, A fourth possibility is the way in which VA fraction was allocated. For example, if an individual can read only 4 of the 5 letters on the 6/12 line, 5 of 6 letters on the 6/9 line, and 5 of 7 letters on the 6/6 line and none of the letters on the 6/5 line, this individual will be assigned to a Snellen fraction of 6/6 (j2) disregarding the errors made in the 6/12 and 6/9 lines. On the other hand, the actual score from the ETDRS chart will take into account of the errors made in any of the lines above the 11th row (letters on this row has the same angular size as the 6/6 line on Snellen chart). Therefore, the score will be lower than the 85 as predicted by direct conversion of the assigned Snellen fraction of 6/6. However, sensitivity analysis (data not shown), including only the eyes which could read every letter on the lowest line seen, suggested that the method of assigning VA fractions was not an important factor in explaining the discrepancy between the two charts. The fifth possibility is that we did not use forced-choice paradigm, which may have led to an overall underestimation of the actual ETDRS letter score. To explore whether there was a uniform bias between the Snellen and ETDRS charts, we examined the relationship between the discrepancy in expected and actual letter scores at the various levels of VA. We noted that the overall trend for better acuity derived from the Snellen chart seemed to be driven by eyes with VA of better than 6/36. In contrast, in the subgroup with VA of worse than 6/36, the Snellen chart underestimated vision by 3 to 6 letters. This finding of poorer acuity derived * 2013 Asia Pacific Academy of Ophthalmology 283

8 Chen et al Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October 2014 from the Snellen chart in eye with VA of less than 6/36 is consistent with the 2 previous reports. 7,8 Falkenstein et al 8 showed that in eyes with VA of worse than 20/100, Snellen fraction underestimated VA (logmar) by 0.40 (4 lines or 20 letters) as compared with ETDRS letter scoring. Similarly, Kaiser 7 reported that in eyes with VA of worse than 20/200, the logmar equivalent of Snellen fraction was lower than that of ETDRS letter score by 0.20 (2 lines or 10 letters). The relatively smaller discrepancy seen at the lower end of VA in our study may be related to the small sample size (only 28 eyes) in this subgroup. The implication of this reversal of bias across VA range is that changes in VA will be overestimated when Snellen chart acuity fractions are directly converted to logmar or letter score equivalents for calculation. For example, an eye with 6/60 vision improving to 6/24 after treatment will have a theoretical gain of 55 j 35 = 20 letters. However, based on our data, the mean actual letter score change will only be 50.5 j 39.0 = 11.5 (at 4 m) or 50.3 j 35.7 = 14.6 (at 1 m) letters. Hence, retrospective studies using logmar equivalents of Snellen-derived VAs for calculation will tend to overestimate treatment effects compared with studies that used the ETDRS charts. We attempted to explore various factors that may contribute to the discrepancy between acuities derived from the Snellen and ETDRS charts by multiple regression analysis. We confirmed the importance of the level of vision in determining the direction of bias in the discrepancy between the 2 charts (ie, patients performed better on the ETDRS chart when VA was poor and vice versa). For comparison between Snellen chart and ETDRS chart at 1 m, there was also a trend for older individuals and those who read the ETDRS chart first to perform better on the Snellen chart (negative A value, smaller logmar on Snellen chart compared with ETDRS chartyderived logmar). One could postulate that this is due to fatigue in older individuals because there are 5 letters per line in the ETDRS chart where there is only 1 letter per line for VAs between 1/60 and 6/60. As for the effect of order of charts used for testing VA, this may be related to learning effect because there was a trend for better VA derived from the second chart read. Our cohort had a wide range of ocular diagnoses. After controlling for these factors, only amblyopia was found to contribute significantly to the variance between the 2 charts. There is uniform crowding effect across the entire range of letter sizes on the ETDRS chart, whereas this is lacking for upper portion of the Snellen chart. However, even after controlling for VA range, amblyopic eyes performed better on the Snellen chart compared with ETDRS charts. Crowding phenomenon in amblyopia is well described, and it can also vary depending on contrast. 17 This finding suggests that a customized formula is required to convert Snellen fractions into equivalent ETDRS letter scores in eyes with amblyopia. This study has several limitations. First, the study was a retrospective analysis of data collected from a heterogeneous group of patients during routine vision testing. However, it was important to analyze this cohort as the results reflect reallife situation in our clinic. Second, we also did not perform refraction before VA testing to compare best-corrected VAs. Sun et al 18 showed that the median absolute difference between VA obtained following manual refraction and with pinhole testing was 5 letters. In addition, they also found that agreement between the scores from both examination methods was poor, especially at better VA levels. 18 We also found that pin-hole testing gave a median of 3- to 5-letter improvement over and above the acuity score recorded on ETDRS letter chart using habitual spectacle correction or unaided, respectively. However, refraction before VA testing is not routine practice in our clinical setting, whereby VA testing is usually performed by personnel who are not trained in refraction and/or are casual staff who are not ophthalmic trained. Our multiple regression modeling showed that presence or absence of habitual spectacle correction was not a significant independent variable. Furthermore, linear regression showed that VA discrepancy reduced as the improvement in VA from pinhole testing increased (r 2 = 0.02). This would suggest that we may have underestimated the discrepancy between the charts by including eyes with refractive errors that have not been fully corrected. Third, we performed only 1 measurement from each chart, and therefore we could not take test-retest variability into account when comparing VAs. Greater variability in VA measurement has been shown in (a) the use of Snellen chart as compared with ETDRS charts, 14,15 (b) patients with ocular pathology of varying type and severity, 19,20 and (c) the use of various stopping and scoring rules. 21,22 We had a large sample size showing statistically significant difference in acuities derived from Snellen and ETDRS charts at both 1 and 4 m. However, a prospective study will be required to dissect the contribution of test-retest variability to the discrepancy between charts, and an even larger sample size is needed to study the impact of the various stopping and scoring rules on this discrepancy in VA measurements. Nevertheless, despite the above limitations, we believe that this study adds to the body of evidence supporting the observation that Snellen VA scores expressed as ETDRS letter score or logmar units are not interchangeable with VA scores obtained from the ETDRS chart. We showed that there was an overall better performance on VA test using the Snellen chart compared with ETDRS chart. However, this difference is driven by preponderance of patients with better acuity, and the relationship is reversed in eyes with acuity of worse than 6/36. In addition to VA range, we showed that amblyopia had a predictable but small contribution to the direction of bias. In the light of our data and those from the 2 previous reports, 7,8 we recommend careful reevaluation of the large body of published literature reporting VA outcomes based on direct mathematical transformation of Snellen fractions into logmar equivalents. Further work is required for investigating the impact of repeatability and scoring and stopping rules on the discrepancy between various VA charts. REFERENCES 1. Snellen H. Optotypi ad visum determinandum. Utrecht: PW van de Weijer; Bailey IL, Lovie JE. New design principles for visual acuity letter charts. Am J Optom Physiol Opt. 1976;53:740Y Recommended standard procedures for the clinical measurement and specification of visual acuity. Report of working group 39. Committee on vision. Assembly of Behavioral and Social Sciences, National Research Council, National Academy of Sciences, Washington, DC. Adv Ophthalmol. 1980;41:103Y Ferris FL III, Kassoff A, Bresnick GH, et al. New visual acuity charts for clinical research. Am J Ophthalmol. 1982;94:91Y Hussain B, Saleh GM, Sivaprasad S, et al. Changing from Snellen to logmar: debate or delay? Clin Exp Ophthalmol. 2006;34:6Y8. 6. Holladay JT. Visual acuity measurements. J Cataract Refract Surg. 2004;30:287Y Kaiser PK. Prospective evaluation of visual acuity assessment: a comparison of Snellen versus ETDRS charts in clinical practice (an AOS thesis). Trans Am Ophthalmol Soc. 2009;107:311Y * 2013 Asia Pacific Academy of Ophthalmology

9 Asia-Pacific Journal of Ophthalmology & Volume 3, Number 5, September/October 2014 VA Fractions Measured on Snellen & s 8. Falkenstein IA, Cochran DE, Azen SP, et al. Comparison of visual acuity in macular degeneration patients measured with Snellen and early treatment diabetic retinopathy study charts. Ophthalmology. 2008;115:319Y Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307Y Gregori NZ, Feuer W, Rosenfeld PJ. Novel method for analyzing Snellen visual acuity measurements. Retina. 2010;30:1046Y Dong LM, Hawkins BS, Marsh MJ. Consistency between visual acuity scores obtained at different test distances: theory vs observations in multiple studies. Arch Ophthalmol. 2002;120:1523Y Bullimore MA, Zadnik K. Consistency between visual acuity scores obtained at different test distances. Arch Ophthalmol. 2004;122:1729Y Reporting visual acuities. Am J Ophthalmol. 2013;155: Rosser DA, Laidlaw DA, Murdoch IE. The development of a reduced logmar visual acuity chart for use in routine clinical practice. Br J Ophthalmol. 2001;85:432Y Lim LA, Frost NA, Powell RJ, et al. Comparison of the ETDRS logmar, compact reduced logmar and Snellen charts in routine clinical practice. Eye (Lond). 2010;24:673Y Sheedy JE, Bailey IL, Raasch TW. Visual acuity and chart luminance. Am J Optom Physiol Opt. 1984;61:595Y Giaschi DE, Regan D, Kraft SP, et al. Crowding and contrast in amblyopia. Optom Vis Sci. 1993;70:192Y Sun JK, Aiello LP, Cavallerano JD, et al. Visual acuity testing using autorefraction or pinhole occluder compared with a manual protocol refraction in individuals with diabetes. Ophthalmology. 2011;118:537Y Kiser AK, Mladenovich D, Eshraghi F, et al. Reliability and consistency of visual acuity and contrast sensitivity measures in advanced eye disease. Optom Vis Sci. 2005;82:946Y Patel PJ, Chen FK, Rubin GS, et al. Intersession repeatability of visual acuity scores in age-related macular degeneration. Invest Ophthalmol Vis Sci. 2008;49:4347Y Camparini M, Cassinari P, Ferrigno L, et al. ETDRS-fast: implementing psychophysical adaptive methods to standardized visual acuity measurement with ETDRS charts. Invest Ophthalmol Vis Sci. 2001;42:1226Y Carkeet A. Modeling logmar visual acuity scores: effects of termination rules and alternative forced-choice options. Optom Vis Sci. 2001;78:529Y538. "As a surgeon, you have to control your emotions because they can overcome your intelligence and your decision making." VArthur SM Lim * 2013 Asia Pacific Academy of Ophthalmology 285