Uniformity of Visual Acuity Measures in Published Studies

Abstract

purpose. To investigate the methods used in contemporary ophthalmic literature to designate visual acuity (VA).

methods. Papers in all 2005 editions of five ophthalmic journals were considered. Papers were included if (1) VA, vision, or visual function was mentioned in the abstract and (2) if the study involved age-related macular degeneration, cataract, or refractive surgery. If a paper was selected on the basis of its abstract, the full text of the paper was examined for information on the method of refractive correction during VA testing, type of chart used to measure VA, specifics concerning chart features, testing protocols, and data analysis and means of expressing VA in results.

results. One hundred twenty-eight papers were included. The most common type of charts used were described as logMAR-based. Although most (89.8%) of the studies reported on the method of refractive correction during VA testing, only 58.6% gave the chart design, and less than 12% gave any information whatsoever on chart features or measurement procedures used.

conclusions. The methods used and the approach to analysis were rarely described in sufficient detail to allow others to replicate the study being reported. Sufficient detail should be given on VA measurement to enable others to duplicate the research. The authors suggest that charts adhering to Bailey-Lovie design principles always be used to measure vision in prospective studies and their use encouraged in clinical settings. The distinction between the terms logMAR, an acuity notation, and Bailey-Lovie or ETDRS as chart types should be adhered to more strictly.

There is an increasing expectation that clinicians have an evidence base for decisions they make. ¹ In eye care professions, most interventions are designed to preserve or improve vision. There are many aspects to vision, including light sensitivity, target resolution and recognition, contrast sensitivity, color vision, and motion detection. ² The essence of central vision, however, is said to be resolution of detail. ³ Visual acuity (VA) is defined as “spatial resolving capacity”; an expression of the angular size of detail that can just be resolved by the observer. ⁴ The ability of the eye to resolve fine detail is essentially a measure of macular function. Although minimum detectable resolution and minimum separable resolution can be tested, minimum recognizable resolution is more commonly measured. ⁴ Minimum recognizable resolution is the term given to the measurement of acuity by using optotype symbols. ⁵ Optotypes are standardized symbols, usually letters, although they can be numbers or pictures. Alternative charts may incorporate symbols designed using similar specifications, such as Landolt rings or tumbling E’s: identification of these symbols does not depend on literacy or language, with the result that they can be applied internationally. ⁶  

VA is the primary measure of visual function in both clinical and research settings ^{7 8} and is measured and analyzed by a variety of professionals, including ophthalmologists, optometrists, and visual scientists. It can be used as an entry criterion, an outcome or efficacy measure, or a safety endpoint. Potential for variation exists in the way in which different studies approach VA. 

First, the method of refractive correction of an eye at the time of VA measurement can vary. For example, VA can be documented as best corrected, unaided, or habitual. Best corrected VA (BCVA) refers to the level of vision achieved with an up-to-date and accurate refractive correction in place. Habitual VA indicates the level of vision achieved under everyday conditions, either with or without optical correction. 

Second, the practice used to measure VA can vary. A variety of test charts are available. It has been said that “there are as many VA test charts as there are ophthalmological departments.” ⁹ Commonly used test charts include the Snellen and Early Treatment of Diabetic Retinopathy Study (ETDRS) charts. The Snellen chart was developed in the 19th century, ^{10 11} and the new-generation Bailey-Lovie and ETDRS charts were developed in 1976 and 1982, respectively. ^{12 13} These charts test both resolution and the higher processes that lead to optotype recognition. There are important differences between the various types of VA chart, and these can affect the clinical relevance of results and their statistical validity. 

Although Snellen charts are commonly used to measure VA, criticisms have been made of their use ¹⁴ even within a few years of Snellen’s first description of the chart, ¹⁵ particularly in those cases in which vision is impaired. Among the variables affecting VA measurements recorded with Snellen-type charts are optotype selection, font style, the numbers of symbols used per row and row progression. ^{5 12} Optotype selection matters, as letters vary in their legibility. ⁶ Adjacent rows do not always include the same balance of easy and difficult letters. Some letters (C, D, E, G, O) are said to be easier to recognize than others (A, J, L) ² although the relative legibility of different letters and also of fonts has been the subject of much debate. ¹⁶ Also rows include progressively more letters, as the sizes of the letters decrease and spacing between letters, and indeed between rows, is not proportional to letter size. Several problems arise as a result. First, crowding occurs to a variable degree in different rows. Crowding refers to an increase in difficulty with closely packed letters, as this arrangement requires more accurate eye movements. ⁴ Thus “uncrowded acuity” is measured at the top of most charts, and “crowded acuity” at the bottom. Second, contour interaction, the effect of neighboring contours on the discriminability of detail, varies on many charts. Third, as the number of letters per row varies from row to row, missing one letter on the bottom row does not have the same significance as missing one letter on the second row. Fourth, the ratio of optotype size from one row to the next is not constant and on most Snellen charts mimics a decimal progression sequence. ⁵ This irregular progression of letter sizes varying from row to row has statistical consequences, as data based on an irregular measurement scale creates problems in any subsequent parametric analysis. Although these criticisms are common to any Snellen chart, the potential for diversity is compounded by the fact that “Snellen chart” no longer refers to a specific chart design. The criteria for a chart to be called a Snellen type are not rigid and in practice amount to charts generally similar to the original Snellen charts, from which they historically descended. Snellen charts from different manufacturers vary in their use of different fonts and optotypes and indeed have different spacing ratios. The means by which charts are illuminated or projected also vary. All of these factors may affect performance. 

The ETDRS chart has become the standard used in many multicenter clinical trials sponsored by the National Institutes of Health. ⁷ On ETDRS charts, which are calibrated for use at 4 m, the optotypes used are Sloan letters. ETDRS is an example of a chart based on Bailey-Lovie design principles. Charts based on these principles are designed to overcome the problems inherent in Snellen charts and have certain key advantages: optotype size is the principal factor that varies in different rows of the chart, and there is a logarithmic progression of optotype size from one row to the next. ¹² On a chart designed on Bailey-Lovie principles, there are the same number of optotypes at each size level, and spacing between optotypes and rows is proportional to optotype size. The result is that changes in vision by a certain number of letters represent the same percentage change, irrespective of where on that chart that change occurs, which allows for the finer grading of vision. ¹⁷ This is particularly useful for patients or subjects in whom a progressive change in vision is occurring and in particular for those with low vision. When the VA is measured in subjects with low vision, each subject should be able to read at least four optotypes at the largest size and none of the smallest optotypes on the chart, at the selected test distance. ¹³ Alterations in the patient-to-chart distance or in the magnification used to view the chart, as for example with a telescope, may be needed in those cases in which vision is severely impaired, and, with charts based on Bailey-Lovie principles, these alterations should give predictable changes in the result. Discrepancies can arise when a subject is moved closer to the chart than 2 m, as factors other than angle subtended by the optotype come into play. ^{18 19}  

LogMAR (logarithm of the minimum angle of resolution) is frequently and often inappropriately used as a synonym for Bailey-Lovie principles. The term logMAR, however, should not be used to describe a chart type, but rather is a generic term, referring to a geometric notation that can be used to express VA. The logMAR notation was introduced along with the Bailey-Lovie principles in 1976. ¹² Minimum angle of resolution (MAR) expresses the angular size of critical detail within a just-resolvable optotype. ⁵ Using the common logarithm of the MAR allows equal credit (0.02 log units per optotype) to be given for each letter read. Charts referred to as logMAR charts may in fact vary in certain aspects, such as optotype selection, recommended test distance or the method of illumination used. The term ETDRS chart is more specific: in the latest version, an attempt has been made to standardize optotype difficulty in each row, although ETDRS charts may still vary in the optotype sets used. 

The procedure used to measure VA can vary in many ways other than chart design. For example the level on the chart at which testing begins and the point at which testing ends can vary. A subject’s motivation to read optotypes depends on myriad factors, such as their underlying competitiveness or determination, their level of attention and fatigue, and even their respect for the test or the tester. It is impossible to control fully for psychological or physiological factors, but it may be possible to go some way toward such control by offering standardized encouragement and feedback during VA testing, to maximize the subject’s performance. However, feedback may also result in subjects’ adoption of strategies that could bias results. ²⁰ The means of dealing with off-chart values, such as counting fingers, can vary. The strictures of a testing protocol are more likely to be adhered to if the status or training of the tester is consistent. 

Even if the method of refractive correction and the way VA is measured are uniform in different studies, variation may still be introduced by different approaches to analysis of VA data. The approach to assigning an acuity score if the subject fails to identify all the letters in a row correctly can vary, and the method used may not be clear. The approach depends on whether the number of optotypes on each row varies. ¹⁷ Also, some investigators may choose to transform VA data into a form more amenable to statistical analysis, or more familiar to the reader, and thus variation may be introduced as the MAR achieved can be expressed in various ways. It can, for example, be expressed as a fraction or as a decimal. Snellen acuities are usually expressed as a fraction. In Europe, the fraction is metric, all reference distances being expressed in meters, whereas in North America reference distances are expressed in feet. Irrespective of whether in European or North American form, the fraction can be reduced to a decimal (so 6/6 or 20/20 would become 1.0). Decimal reduction is common throughout Europe. Many charts are in fact produced with the size progressing in even steps along the decimal scale, and the size labels are printed in decimal notation adjacent to the optotypes. 

There is therefore a need for clear and consistent approaches to measuring VA to allow meaningful comparison of outcomes of ophthalmic interventions. Attempts at standardizing VA charts were first made in 1868, ¹⁵ and have been repeated at intervals since. ¹⁶ In 1965 it was stated that “existing charts no longer provide a comparable basis for estimating visual acuity.” ¹⁶ In 1996 procedures were recommended for use in clinical research to help ensure “accurate and reproducible assessment of VA”. ²¹ In 2003 the desirability of “standard VA testing and scoring” protocols in multicenter clinical trials was discussed. ⁷  

The purpose of this study was to investigate the methods used in contemporary ophthalmic literature to designate VA—specifically, to survey (1) the method of refractive correction used (i.e., best corrected, habitual, or unaided); (2) what details on VA measurement procedure are given (i.e., chart design, chart features, testing protocol, and approach to analysis); and (3) the means used to express VA in Results sections. 

Methods

Five ophthalmic journal titles were selected. They were chosen because they all have impact factors among the highest of clinically oriented ophthalmology journals. The journals were the American Journal of Ophthalmology, Archives of Ophthalmology, the British Journal of Ophthalmology, the Journal of Cataract and Refractive Surgery, and Ophthalmology. All the 2005 issues of these five journals were considered. Papers were included if (1) VA, vision, or visual function was mentioned in the abstract, either in the Methods section as an outcome variable or in the Results or Conclusions, and if (2) the study related to age-related macular degeneration (AMD), cataract, or refractive surgery. Papers were excluded if (1) they described studies that included subjects less than 16 years of age or (2) if there were fewer than 10 subjects. If a paper was selected on the basis of its abstract, the full text of the paper was studied. 

The Methods sections of each paper were examined for detail on (1) the method of refractive correction used during VA testing (best corrected, unaided, or habitual); (2) information given on measurement of VA—specifically, chart design, chart features, testing protocol, and data analysis; and (3) within the Results section, the method chosen to express VA. 

The presence of information on the following chart features was noted: (1) the manufacturer or supplier; (2) the luminance level; (3) the display method (i.e., means of chart illumination or optotype projection); and (4) the number of optotypes per row. With regard to luminance level, in the case of internally illuminated charts, information on the level of chart luminance was sought, whereas in the case of externally illuminated charts or projected charts, details on chart illuminance were required. 

The presence of description of the following aspects of the testing protocol was noted: (1) where on the chart testing began; (2) whether feedback or encouragement was given; (3) whether detail was given on termination of testing; (4) the means of dealing with off-chart acuities, such as counting fingers; and (5) who measured the acuity. 

Finally, mention of the following aspects of analysis was noted: (1) scoring criteria and (2) transformation of measures for statistical reasons. 

If the paper gave a reference for the method of measuring VA and the referenced paper contained the required details, the information was considered to be given. The sample size of the study was recorded. Each journal’s editorial policy on how vision should be measured and presented was sought, both by sending an inquiry by e-mail to the editor and by examining online instructions for authors. 

Results

A total of 128 abstracts met the inclusion criteria (Table 1) . The data in this survey were all derived from an examination of the full texts. 

Thirty-two of the studies evaluated were retrospective and 96 were prospective. In 15 papers, references were given relating to measurement of VA. All referenced papers were retrieved and checked. In four papers, manuals of procedures were referenced, but these were not readily available to us and were not checked. 

Method of Refractive Correction

Most papers (n = 115/128) identified the method of refractive correction of the subjects. The most common method of refractive correction tested was BCVA (85.9%, n = 110/128). In 46 of these papers, unaided acuity was measured as well as BCVA (34 were on refractive intervention), and in 2 of 110 cases (both on AMD), habitual VA was measured, as well as BCVA. Unaided acuity was the only form of VA documented in four papers (three on refractive intervention) and habitual VA was the only form in one case (on AMD). However in 13 papers (10.2%) no information was given on the method of refractive correction. 

Details of the Measurement Process as Given in Methods Sections

Chart Design.

Of the 128 papers, only 75 (58.6%) identified the VA test chart used (Table 2) . The most common type of chart referred to in Methods sections was logMAR (n = 41). Thirty-four of the 41 papers using logMAR reported that the chart used was an ETDRS chart; the other 7 papers did not specify which logMAR chart was used. The next most common chart used was the Snellen chart (n = 32). 

Thirty-six papers in total specified the optotype set used. The means of referring to VA measurements in the Methods sections varied with the specialty topic of the paper (cataract, refractive surgery, or AMD). More papers within the AMD group referred specifically to the use of ETDRS charts (68.2%, n = 15/22) in Methods than mentioned Snellen charts (27.3%, n = 6/22). Correspondingly, comparing specialties, fewer papers on cataract (32.4%, n = 11/34) and refractive surgery (23.5%, n = 8/34) referred specifically to ETDRS in the Methods sections than did papers on AMD (44.1%, n = 15/34). Twenty-two of the 23 papers on AMD gave a chart type. 

Chart Features.

All papers on retrospective and at least 85% of papers on prospective studies failed to give information on the following chart features: manufacturer of the chart (which could affect the susceptibility of the chart to ageing); luminance; the display method, i.e., means of illumination of the chart or the method of optotype projection used; and the number of optotypes per row (Table 3) . 

Information that was given on luminance (in only 11 papers) varied and included luminance levels on and adjacent to the chart in lux, room illumination levels in candelas per square meter, chart luminance on an LCD monitor in candelas per square meter, or a statement that acuity tests were performed at a particular luminance level. In the rare cases when information was given on the display method (14 papers), 6 referred to the ETDRS protocol or ETDRS Fast protocol (both of which state that “a light box accommodates the charts”) ^{13 22} ; 5 papers specified that retroillumination or rear lighting was used; 2 named a Snellen projection system; and 1 specified use of an LCD monitor. 

Measurement Procedure.

All papers on retrospective studies and most on prospective studies failed to provide, or adequately reference, sufficient details on the testing procedure to allow replication by others (Table 4) . The specific areas of omission included an indication of the level on the chart on which testing began. For example normally sighted subjects may be expected to start from the top of the chart, or may be allowed to start a few rows suprathreshold. All six papers that gave information on the level of the chart on which testing began referred to the ETDRS protocol or ETDRS Fast protocol (both of which state that “testing begins with the first letter on the top row”). ^{13 22} The nature of any feedback given to the subject was not mentioned in most papers. Only eight gave an indication that encouragement was given: six of these referred to the ETDRS protocol (which states, “When a subject has difficulty reading a letter, he or she is encouraged to guess”) ¹³ ; one referred to the TAP Report 1 (which states that the subject was “encouraged to achieve the best identification of each letter”) ²³ ; and one stated that “patients were encouraged to read each letter until the majority of the line was read incorrectly.” ²⁴ Information on termination rules or termination criteria was given in a mere 10 cases in total. In four papers, a row-by-row approach was taken (i.e., testing was stopped when the majority of letters in a row were read incorrectly). In five papers, subjects were asked to read the whole chart. ¹³ One paper referred to the ETDRS Manual of Operations. How off-chart acuities were dealt with was omitted in most cases; for example, what would happen if the subject could not read the first row of the chart? Fifteen papers gave information on who measured the VA—for example, naming the examiner as one of the authors or stating that a protocol-trained or certified optometrist measured acuity. 

Approach to Analysis.

Most papers did not give details on scoring criteria (Table 5) . The scoring method was outlined in only 20 studies: in 17, letter-by-letter scoring, whereas in 3, row-by-row scoring was used (in all 3, Snellen charts were used to measure VA). 

Eighteen papers stated that VA data were transformed into another notation to facilitate statistical analysis. In 14 of these studies, Snellen charts were used and the data converted to logMAR form. Two papers referred to no chart type in the Methods section, but stated that logMAR form was used for statistical purposes. In one study, a decimal chart was used to measure acuity and the acuity converted to logMAR form, and in another logMAR acuity was measured and converted to decimal form. 

Five papers (3.9% of total) did not make any mention of vision in the Methods section, neither giving the refractive correction used nor any aspects of the measurement process, despite the presentation of acuity measurements in the results. 

Means of Expressing VA in the Results Section.

All 128 papers presented acuity data in their Results sections. The commonest ways in which acuity measurements were presented were as Snellen fractional (46.9%, n = 60/128), logMAR (21.9%, n = 28/128), Snellen fractional and logMAR (6.3%, n = 8/128), or Snellen decimal acuities (10.2%, n = 13/132). Nine papers presented acuity results as the number of letters read; in seven of these an ETDRS chart and in two a Snellen chart was used to measure VA. Three papers described results in terms of rows: in one the use of Snellen was reported and in the other two no chart type was given. Two papers presented VA results as Snellen fractional and decimal: one as logMAR and decimal, one as logMAR and rows. Three did not present raw or summary VA data: one of these reports presented correlations between VA and an ocular outcome of interest, whereas two presented the number of outcomes that showed a change in BCVA. The scoring method used to assign a VA level during testing therefore did not necessarily correlate with the way VA was expressed in the Results sections. 

Two factors influenced how VA was expressed in Results sections: the method used to measure VA and the specialty topic of the paper (cataract, refractive surgery, or AMD). 

The method used to measure VA influenced how acuity was expressed in the results. For example, when no chart was named in the Methods section (n = 53), the results were expressed as Snellen fractions in 56.6% (n = 30/53) of cases, as Snellen decimal in 13.2% (n = 7/53), as logMAR in 11.3% (n = 6/53), and as Snellen fractional and logMAR in 7.5% (n = 4/53). In contrast, of 34 papers that stated that VA was measured with an ETDRS chart, 38.2% (13/34) expressed results as Snellen fractions, 26.5% (9/34) as logMAR, and 8.8% (n = 3/34) as Snellen fractional and logMAR in the results. In all 13 studies in which VA was measured with an ETDRS chart and the results expressed as Snellen fractions, a numerator of either 20 (feet) or 6 (meters) was used. When a Snellen chart was used to measure VA (n = 32), the results were expressed as Snellen fractions in 46.9% (n = 15/32), as logMAR in 21.9% (n = 7/32), and as Snellen fractional and logMAR in 3.1% (n = 1/32). 

VA was expressed as a Snellen fraction in 51.9% (n = 28/54) of papers on cataract, 49.0% (n = 25/51) of papers on refractive surgery, and 30.4% (n = 7/23) of papers on AMD. In contrast, 24.1% (n = 13/54) of cataract, 17.6% (n = 9/51) of refractive surgery, and 26.1% (n = 6/23) of AMD papers presented VA results in logMAR form. Letter-by-letter scoring was used to express VA in eight (34.8%) of the AMD papers, one refractive surgery paper, and none of the cataract papers. 

All five journals replied to the e-mail inquiry. None of the five journals had set rules or requirements about how vision should be measured and presented, but some had recommendations (Appendix). 

Discussion

We found that of the 128 papers analyzed, the reporting of the measurement of VA was, in nearly all cases, wholly inadequate. The methods used and the approach to analysis were rarely described in sufficient detail to allow others to replicate the study being reported, even though VA was an important outcome variable in these studies. Although most (89.8%) of the studies reported on the method of refractive correction during VA testing, only 58.6% gave the chart design, and less than 12% gave any information whatsoever on chart features or measurement procedures used. 

Method of Refractive Correction

Of the 128 papers in our sample, 115 (89.8%) gave information on the method of refractive correction of the eye at the time of VA testing. Such information can be used to distinguish safety endpoints (BCVA) from efficacy endpoints (uncorrected VA) (for example, Oral et al. ²⁵ ). However, in these 115 papers, further information on VA measurement was frequently absent: 29 of the 115 papers did not give the chart type used. Information on the method of refractive correction alone fails to provide the reader with detail that allows replication of the measurement, and method of refractive correction is not an alternative to chart type, as clarified by the following excerpt: “The best corrected visual acuity was determined using the Early Treatment of Diabetic Retinopathy Study chart… .” ²⁶  

Chart Design

Seventy-five papers gave information on the design of chart used, with reference to logMAR-based charts being the most common. As mentioned earlier, however, the use of the term logMAR-based charts is potentially misleading. It cannot be assumed that the presentation of data in logMAR notation means that the chart used follows Bailey-Lovie design principles. 

Journals’ preferences about chart type and VA notation vary (see Appendix). The British Journal of Ophthalmology suggests (in an e-mailed reply to the inquiry) that “primary data should be captured as ETDRS if possible, and expressed as LogMAR, as per most trials.” The American Journal of Ophthalmology’s author instructions specify that “the Snellen equivalent in feet be presented in parentheses next to each vision that is not in the Snellen format to aid readers in the United States.” A trend is detectable in the past decade toward the use of and reference to logMAR-based charts in published studies. ²⁷ We suggest that charts produced according to Bailey-Lovie–based principles should always be used to measure vision in planned prospective studies in research settings, given the disadvantages of Snellen charts. VA should be expressed in a logMAR format, as intended by those who developed the ETDRS chart. This requirement cannot be demanded for studies that are retrospective, because investigators have to rely on case notes from an earlier time. Thus, there should be some tolerance of variance in the VA data in retrospective studies. Standardized measurement conditions may also be difficult to maintain in multicenter studies, particularly in epidemiologic studies involving the assessment of vision in rural, domiciliary, or developing-world settings. However, in clinical practice, the use of charts obeying Bailey-Lovie principles and expression of acuities as logMAR units should still be encouraged. Furthermore the distinction between the terms logMAR, as an acuity notation, and charts conforming to Bailey-Lovie principles, as chart types, should be adhered to more strictly. 

Chart Features and Measurement Procedure

Most papers gave no details on the process of measuring VA relating to chart features or measurement procedure. In a similar study on measures of acuity in ophthalmic literature, the method of presenting VA was found not to be stated in 17.4% of papers. ²⁸  

In particular most papers gave no information on rules or criteria for termination of testing. The use of termination rules is said to reduce the variance of VA measures in almost all cases. ²⁰ Termination rules should be distinguished from scoring methods. For example, although a subject’s acuity may be scored letter by letter, the protocol may call for testing to be stopped when the majority of characters in a row are read incorrectly. Four papers described this rule for termination of testing. Five papers referred to Ferris, ¹³ in which a description is given of a method of scoring by which 0.02 units are given “for each letter read on the entire chart.” The termination practice in these five studies was therefore presumed to be to force subjects to attempt all optotypes on the entire chart. The termination rules, and the termination criteria set, can affect both the acuity score and the variance of scores. ²⁰ There is no consensus as to best practice to reduce variance, although recommendations on optimal termination rules have been made. ²⁰  

Uniformity of VA measurement between studies is desirable, but it is impossible to comment meaningfully on uniformity of VA measurement in our sample, given that most papers gave insufficient pertinent information. It may be that the potential sources of variation in the measurement of VA that we surveyed are not considered to be of sufficient clinical magnitude to justify either controlling or reporting, especially if acuity is a secondary outcome measure. However, we feel that it is not pedantic to insist on such details. Distance VA is measured in almost every ophthalmologic consultation and is often a major determinant of treatment decisions in clinical practice. Similarly, clinical trials in ophthalmology often have VA as the main or primary outcome measure, and a prerequisite to a successful grant application is the inclusion of formalized methodology concerning method of refractive correction and acuity measurement. ²⁹ In the instructions for authors in Investigative Ophthalmology and Visual Science, the advice is given to “Provide sufficient detail to enable others to duplicate the research”, and this should apply equally rigorously to measurement of vision as to any scientific measurement. An example of good practice is in Broman et al. ³⁰ : “Distance acuity was tested with the Early Treatment Diabetic Retinopathy Study chart at 3 m, illuminated at 130 candela/m². Participants who failed to read the largest letters at 3 m were retested at 1.5 m, then at 1 m. VA was scored as the total number of letters read correctly, transformed to LogMAR units. Failure to read any letters was assigned an acuity of 1.7 LogMAR units, which is equivalent to acuity of 20/1000. An E chart was used for participants who were illiterate.” 

Approach to Analysis

Information on the approach to analysis of VA data was similarly lacking in most cases. For example, approaches to scoring were mentioned in 20 papers and varied. Although some used letter-by-letter scoring, as charts of Bailey-Lovie design principles allow, others defined acuity as the smallest row on which the subject could identify most letters correctly. In all three studies in which row-by-row scoring was used, Snellen charts were employed to measure VA. Caution must be exercised: parametric analysis of data acquired by row-by-row scoring is inappropriate for a chart with an irregular progression of letter sizes, as Snellen acuities are not interval-type data. In general, letter-by-letter scoring with a chart of Bailey-Lovie design is recommended; otherwise, a change in acuity of one letter can result in a change of one row. ⁵  

Eighteen papers stated that VA data were transformed to another form for analysis; such a practice, however, is hazardous. VA results obtained with a Snellen chart are not amenable to parametric analysis, even if converted to another form such as a decimal notation, as they are based on an irregular measurement scale. Furthermore, decimal acuities can be so easily, and yet so inappropriately, confused with logMAR data, particularly when lifted from papers for more general reference. Transformation of results obtained with an ETDRS chart can also be done inappropriately. In 13 studies, an ETDRS chart was used to measure vision and the results expressed as Snellen fractions. The numerator of a Snellen fraction is the working distance at which VA is measured, and ETDRS charts are calibrated for use at a working distance of 4 m. However, in all 13 papers, the numerator of the Snellen fraction in the results was either 6 or 20, depending on whether meters or feet were being used, respectively. In 8 of the 13 papers, no working distance was specified in the Methods section, and it was therefore not clear whether the working distance actually used was 6 m (or 20 ft). In the other papers, the working distances for measuring vision were specified. Working distances were 2 m (two papers) and 3, 4, and 6 m (one paper each). In four of these five papers, the Snellen fraction presented in the Results section did not therefore correctly report the working distance used. 

Expression of VA Results

VA data were expressed in various ways in Results sections. Whether the paper was on cataract, refractive surgery, or AMD affected the way vision was referred to in both the Methods and the Results sections. ETDRS was the stated method in a higher proportion of papers on AMD than in those on cataract or refractive surgery (68.2%, n = 15/22; 36.7%, n = 11/30; and 34.8%, n = 8/23, respectively; Table 2 ). This finding reflects the scientific inadequacies of Snellen, which become more significant when the charts, which have very few large letters in comparison to Bailey-Lovie design principle–based equivalents are used to assess the visual status of those with moderate to severe visual impairment. Charts designed on Bailey-Lovie principles are a more valid way of assessing low vision and therefore are more appropriate for studies on AMD. It also reflects a proliferation of sponsored clinical trials on the treatment of AMD, in which detailed manufacturer-driven protocols are typical. 

Conclusions

In this report, we sought to assess the uniformity of VA measurement practices, but information for a meaningful assessment was lacking. We, like many others, have suggested standardizing VA protocols. ^{7 21 27} We therefore recommend that journal editors and research regulatory committees insist, when examining Methods sections of studies, that the following details be given (at least as an addendum if word counts are prohibitive):

1.  
   The method of refractive correction of eyes tested (BCVA, unaided, or habitual, or a combination of these).
2.  
   Information about the chart: design; manufacturer chart; luminance level; display method, i.e., means of chart illumination or optotype projection; and specification and arrangement of optotypes.
3.  
   Measurement procedure: test distance; where on the chart testing began; what level of feedback or coaxing was given; what termination rules and criteria were applied; how off-chart acuities were dealt with; and the status or training of the tester.
4.  
   Approach to analysis: scoring criteria and means of analysis of VA data, including information on transformation of data, if transformed data are reported.

Only when such details are given can a debate be conducted on what the standards should be. Computer-based technology will facilitate manipulation of variables affecting VA performance, but standards should be set before a variety of different computer-based VA measurement techniques proliferate. 

Appendix 1

These are the full responses of journals to an e-mailed inquiry that asked whether the journal had any requirements, preferences, or recommendations regarding the measurement or presentation of VA in studies submitted for publication. 

American Journal of Ophthalmology

Thank-you for writing. Our visual acuity standards are below in an excerpt from our Instructions for Authors: 

The AJO encourages authors to report the visual acuity in the manuscript using the same nomenclature that was used in the study, provided the data was recorded in the Snellen system (using either meters or feet), decimal fraction or LogMAR systems. The same visual acuity nomenclature should be presented throughout the manuscript. A Tableof Equivalent Visual Acuity Measurements will be provided for the reader in each print issue. Standardized reporting of visual acuity would be a better option but has not yet been accepted by all research groups. It is requested that the Snellen equivalent in feet be presented in parentheses next to each vision that is not in the Snellen format to aid readers in the United States. 

Archives of Ophthalmology

Although there isn’t a uniform requirement, the acuity should be consistent throughout the manuscript. Snellen acuity is the most commonly used. 

British Journal of Ophthalmology

Thank-you for your e-mail. There is no rule. However primary data should be captured as EDTRS (sic) as possible and expressed as LogMAR, as per most trials. 

Journal of Cataract and Refractive Surgery

We do not have a strict requirement for visual acuity measurements as we publish a visual acuity equivalency in each issue. Snellen fractionals are probably used the most, but we also publish papers with measurement in Snellen decimals and LogMAR. 

Ophthalmology

We generally prefer Snellen—I’ve included a copy of Appendix V that shows conversions in case people use other preferences. 

 Presented in part at the annual meeting of the European Association for Vision and Eye Research (EVER), Vilamoura, Portugal, October 2006 and at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2008.

 Submitted for publication April 27, 2007; revised September 23, 2007, and January 22 and March 28, 2008; accepted August 21, 2008.

 Disclosure: M.A. Williams, None; T.N. Moutray, None; A.J. Jackson, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Number of Eligible Papers in Each Subject Area and Journal

	Cataract	Refractive Surgery	AMD	Total
Am J Ophthalmol	6	3	3	12
Arch Ophthalmol	1	1	7	9
Br J Ophthalmol	5	2	5	12
J Cataract Ref Surg	29	32	0	61
Ophthalmology	13	13	8	34
Total	54	51	23	128
Median subjects, n (range)	50.5 (10–36,072)	36 (10–16,111)	70 (15–4,774)	50 (10–36,072)

Table 2.

 

Chart Design Referred to in the Methods Sections

	ETDRS*	LogMAR*	Snellen Fractional	Snellen Decimal	Landolt	Total n (%)
Cataract (n)	11	1	16	1	1	30 (40.0)
Refractive (n)	8	5	10	0	0	23 (30.7)
AMD (n)	15	1	6	0	0	22 (29.3)
Total, n (%)	34 (45.3)	7 (9.3)	32 (42.7)	1 (1.3)	1 (1.3)	75 (100)

 Data are the number of papers.

*  Thirty-four papers specified that the logMAR chart used was ETDRS; seven merely referred to using a logMAR chart.

Table 3.

 

Number of Papers Giving Details on Chart Features in the Methods Section

	Manufacturer/Supplier of Chart	Luminance (Background or External)	Display Method (illumination/Projection Method)	Number of Optotypes per Row
Retrospective (32 papers)	0	0	0	0
Prospective (96 papers)	13	11	14	10
Chart or method used to measure VA	10 used ETDRS; 3 used Snellen	7 specified ETDRS; 2 used logMAR; 2 used Snellen	10 specified ETDRS; 2 used logMAR; 2 used Snellen	All 10 used ETDRS

 Data are the number of journals in each category.

Table 4.

 

Number of Papers Giving Details on Measurement Procedure in the Methods Section

	Where on Chart Testing Began	Whether Encouragement or Feedback Was Given	Whether Termination Rules or Criteria Were Given	How Off-Chart Acuities Were Dealt With	Whether the Status or Level or Training of the Examiner Was Mentioned
Retrospective (32 papers)	0	0	1 (Snellen)	0	0
Prospective (96 papers)	6	8	9	10	15
Chart or method used to measure VA	All 6 used ETDRS	7 specified ETDRS; 1 used logMAR	6 specified ETDRS; 2 used logMAR; 1 used Snellen	All 10 used ETDRS	9 specified ETDRS; 3 used Snellen; 1 used logMAR; 2 gave no chart

Table 5.

 

Number of Papers Giving Details on Analysis of VA Data

	Scoring Method	Whether a Statement Was Made That VA Data Was Transformed
Retrospective (32 papers)	2	3
	Both used Snellen	All used Snellen
Prospective (96 papers)	18	15
Chart used	15 specified ETDRS; 2 used logMAR; 1 used Snellen	See text

The authors thank Verica Paunović for help in translating a referenced article.