The World Health Organization (WHO) recently completed an impressive global review of a large number of surveys on visual impairment, and estimated that there were 161 million persons worldwide with visual impairment in the year 2002, including 37 million with blindness [1, 2]. This estimate is now commonly quoted, including by VISION 2020 – The Right to Sight, the global initiative launched jointly by the WHO and the International Agency for the Prevention of Blindness, which aims to help eliminate avoidable blindness globally by the year 2020 [3]. This estimate was based on the definitions of visual impairment in the International Statistical Classification of Diseases (ICD), which define blindness as best-corrected visual acuity less than 3/60 or central visual field no greater than 10 degrees in the better eye, and low vision (visual impairment less severe than blindness) as best-corrected visual acuity less than 6/18 to 3/60 [4]. These definitions of visual impairment using best-corrected visual acuity exclude uncorrected refractive error as a cause of visual impairment, thereby leading to underestimation of the total burden of visual impairment [5, 6]. We therefore attempted to estimate the global burden of visual impairment, including that caused by uncorrected refractive error, by reviewing data from published population-based surveys of visual impairment that included presenting visual acuity.

The WHO estimate of visual impairment in the different Global Burden of Disease (GBD) sub-regions, classified according to the GBD 2000 Project [7], for the year 2002 was used as the base estimate of visual impairment due to causes other than uncorrected refractive error, as best-corrected vision was used for this estimate [1]. To assess the additional contribution of uncorrected refractive error to global visual impairment from published population-based surveys, we followed the guidelines for reporting meta-analysis suggested by the Meta-analysis of Observational Studies in Epidemiology Group [8].

The PubMed literature database [9] was searched in early August 2005 and again in early December 2005. The terms "blindness AND population-based survey", "blindness AND population-based study", "visual impairment AND population-based survey", "visual impairment AND population-based study", "low vision AND population-based survey" and "low vision AND population-based study" were used to locate papers on population-based surveys of visual impairment in any language published 1996 onwards, covering about a decade up to the present. This search yielded 271 publications. The abstract of each of these publications was reviewed, and the papers that were actually population-based surveys of visual impairment and documented its causes were obtained from the journals and through contact with authors if these were in the English language. If additional papers on population-based surveys of visual impairment and its causes published 1996 onwards were found in the references of these papers located through the PubMed search, these papers were also obtained, resulting in a total of 283 publications for review. The five papers from the PubMed search that were in a language other than English were assessed on the basis of their English abstracts. The aim of this literature search was to locate publications that included presenting visual acuity to define visual impairment, which would allow us to ascertain the contribution of uncorrected refractive error to visual impairment in one or both of the categories used by WHO for reporting, i.e. visual acuity less than 3/60 in the better eye and visual acuity less than 6/18 to 3/60 in the better eye.

If one or more qualifying surveys were available for a GBD sub-region, the data from those were utilised to arrive at the best estimates for the proportion of blindness (presenting visual acuity less than 3/60 in the better eye) and less severe visual impairment (presenting visual acuity less than 6/18 to 3/60 in the better eye) caused by uncorrected refractive error. For the GBD sub-regions for which no qualifying survey was available, the most closely matching sub-region based on mortality strata as classified by GBD [7] was selected for which a qualifying survey was available. Data on the proportional contribution of uncorrected refractive error to blindness and less severe visual impairment from this matching selected sub-region were extrapolated to the sub-region for which no qualifying survey was available. This extrapolation either used data from the matching sub-region directly, or introduced adjustments if the population characteristics were somewhat different, as explained in the results section below.

The WHO base estimates for blindness and less severe visual impairment were then adjusted by adding the contribution of uncorrected refractive error, to arrive at estimates indicating the total number of persons with blindness and less severe visual impairment including that caused by uncorrected refractive error. As clear data regarding the contribution of uncorrected refractive error to blindness and less severe visual impairment were available from only a small number of countries, and extrapolations were used to arrive at the best estimates, sensitivity analysis was done to arrive at plausible ranges for the overall estimates. This was done by assuming that the actual contribution of uncorrected refractive error could be less or more than the proportional contribution of uncorrected refractive error to blindness and less severe visual impairment estimated from the limited available data.

Sensitivity analysis was done assuming that the proportional contribution of uncorrected refractive error to blindness could be 20% less or more than our estimate of 12.3% (i.e. 9.8% to 14.7%), and that the corresponding contribution to less severe visual impairment could be 10% less or more than our estimate of 42.8% (i.e. 38.5% to 47.1%). A higher percentage variation in the plausible range for blindness was assumed, as the proportional contribution of uncorrected refractive error to blindness was smaller than that for less severe visual impairment, and smaller proportions may be associated with larger percentage variations. This sensitivity analysis revealed that the total number of persons worldwide who were visually impaired because of uncorrected refractive error could range from 82 million (4 million blind and 78 million with less severe visual impairment) to 117 million (6 million blind and 111 million with less severe visual impairment).

The obvious limitation of our estimates for the contributions of uncorrected refractive error to blindness and less severe visual impairment is the sparseness of relevant published data from around the world. Although we did an extensive PubMed search and examined cross-references from the papers located, we did not search non-English-language databases, which might have led us to overlook some relevant papers. Faced with the scanty data from the literature search, we had the option of aborting this exercise or making the best estimates using the most reasonable assumptions. As estimates of visual impairment worldwide without inclusion of uncorrected refractive error are obviously underestimates, and because frequent reference to such underestimates is not only inaccurate but can be inadvertently misleading, we felt that it was better to make initial indicative estimates of visual impairment worldwide including uncorrected error with whatever published data were available.

Our indicative estimates suggest that the number of visually impaired persons in the world is about 259 million, including 42 million blind with presenting visual acuity less than 3/60 in the better eye. This estimate includes 98 million persons with visual impairment due to uncorrected refractive error, who are not included in the WHO estimate of 161 million visually impaired persons worldwide based on the ICD visual impairment definitions that use best-corrected visual acuity [1]. Because the contribution of uncorrected refractive error to visual impairment was uncertain due to the limited available data, we estimated the plausible range of the number of persons worldwide who were visually impaired due to uncorrected refractive error as 82 to 117 million. The original WHO estimate of visual impairment due to causes other than uncorrected refractive error does not report a plausible range because of the uncertainty that may be associated with the available data [1]. Ideally, all worldwide estimates of visual impairment should include a plausible range based on the degree of uncertainty of the data, as is done by UNAIDS for HIV estimates [26].

We estimated from the available data that the proportional contribution of uncorrected refractive error to visual impairment was highest in the GBD sub-region that includes China, and lowest in the GBD sub-regions of Africa. This is consistent with the highest rates of refractive error reported for Chinese populations [23–25], and a review suggesting a relatively low contribution of uncorrected refractive error to blindness in Africa [27].

We have used the generic term 'uncorrected refractive error' in this paper to cover both refractive error that is not corrected at all and that which is inadequately corrected if a person is using refractive correction but is still visually impaired. It is important to point this out, as several studies from different parts of the world have suggested that in addition to completely uncorrected refractive error, inadequately corrected or under-corrected refractive error is also a significant problem [28–33].

It is important to note that refractive error is the most easily treatable cause of visual impairment, in most cases by simple spectacles. In addition, since the onset of visual impairment due to natural refractive error sets in at a younger age than the other major causes, it is responsible for a much larger number of blind years lived by a person than most other causes if left uncorrected [5, 14]. It was estimated in an Indian state that blindness due to uncorrected natural refractive errors resulted on average in over 30 years of blindness for each person as compared with 5 years of blindness due to untreated cataract for each person [14]. Not only do our estimates indicate that uncorrected refractive error is the most common cause of visual impairment in the world, the burden it causes in the more productive younger years of life has a potentially serious adverse socio-economic impact on society. This underscores the point that visual impairment due to uncorrected refractive error cannot be overlooked in worldwide estimates, even though the currently available data regarding it are not extensive. Our indicative estimates can obviously be refined as more data become available from around the world. For now, we suggest that the indicative estimate of 259 million persons with visual impairment worldwide, which includes 42 million blind with presenting visual acuity less than 3/60 in the better eye, is more appropriate for further planning of the VISION 2020 initiative than the commonly quoted 161 million estimate including 37 million blind, which excludes uncorrected refractive error.

Even for countries for which data on the contribution of uncorrected refractive error to visual impairment were available in this assessment, clear description of the method of attributing visual impairment to uncorrected refractive error and the types of refractive error leading to visual impairment was often missing. This could have led to under- or over-estimation of the contribution of uncorrected refractive error to the visual impairment burden. For example, under-estimation could have occurred in studies that did not include proper refraction, and over-estimation in studies that included substantial index myopia (induced by development of nuclear cataract) as a refractive error cause of visual impairment – the correct cause in such cases would be cataract. This situation points to the need to develop a standardised system for assessing the contribution of uncorrected refractive error to visual impairment in population-based studies, including distinguishing index myopia from natural refractive error and documentation of refractive error-related amblyopia [5].

It has to be kept in mind that the visual impairment definitions are currently based on distance visual acuity. A subset of the persons who have poor distance vision due to uncorrected refractive error, and qualify as visually impaired, may have good near vision. The difference between the impact on quality of life due to visual impairment caused by uncorrected refractive error that is associated with good near vision, and that due to visual impairment causing poor vision both at distance and near, needs to be better understood. On the other hand, it should also be noted that uncorrected presbyopia, refractive error due to aging that causes difficulty in seeing at near which usually starts progressing around 40 years of age, also causes disability. However, at this stage, the data available are not adequate to enable presbyopia-related visual impairment to be included in the definitions of visual impairment. It would be useful for such data to become available over a period of time.

We recommend that all population-based assessments of blindness and less severe visual impairment be based on presenting visual acuity so that uncorrected refractive error as a cause is not missed. In addition, if standardised requirements for reporting of visual impairment and its causes from population-based surveys were developed and implemented through a combined effort of journal editors, this would facilitate more efficient utilisation of future data for systematic tracking of visual impairment around the world.

Although data on the contribution of uncorrected refractive error to visual impairment worldwide are scanty, our indicative estimate based on the available data suggests that the total number of persons with visual impairment in the world, who have presenting visual acuity less than 6/18 in the better eye, is about 259 million. This includes 42 million blind persons with presenting visual acuity less than 3/60 in the better eye and 217 million with less severe visual impairment of presenting visual acuity less than 6/18 to 3/60 in the better eye. This estimate of visually impaired persons is 61% higher than the 161 million estimate recently made by WHO on the basis of best-corrected visual acuity, which excluded visual impairment caused by uncorrected refractive error. Our estimate of 98 million persons in the world with visual impairment due to uncorrected refractive error makes this the largest cause of visual impairment. Estimates of visual impairment that include uncorrected refractive error must be used to avoid misleading underestimates from becoming the basis of planning for visual impairment reduction globally. We therefore suggest that the indicative estimate of 259 million persons in the world with visual impairment, which includes 42 million blind with visual acuity less than 3/60 in the better eye, be used for further planning of the VISION 2020 initiative instead of the commonly quoted recent WHO estimate of 161 million, which includes 37 million blind.