Prevalence of hyperopia of +3diopters or greater in white persons (A) and black and Hispanic persons (B).BES indicates Baltimore Eye Survey, Baltimore, Md; BDES, Beaver Dam Eye Study,Beaver Dam, Wis; BMES, Blue Mountains Eye Study, Sydney, New South Wales,Australia; RS, Rotterdam Study, Rotterdam, the Netherlands; Melbourne VIP,Melbourne Visual Impairment Project, Melbourne, Victoria, Australia; and ProjectoVER, Vision Evaluation and Research, Nogales and Tucson, Ariz.
Prevalence of myopia of −1diopter or less in white persons (A) and black and Hispanic persons (B). BESindicates Baltimore Eye Survey, Baltimore, Md; BDES, Beaver Dam Eye Study,Beaver Dam, Wis; BMES, Blue Mountains Eye Study, Sydney, New South Wales,Australia; RS, Rotterdam Study, Rotterdam, the Netherlands; Melbourne VIP,Melbourne Visual Impairment Project, Melbourne, Victoria, Australia; and ProjectoVER, Vision Evaluation and Research, Nogales and Tucson, Ariz.
Prevalence of myopia of −5diopters or less in white persons (A) and black and Hispanic persons (B).BES indicates Baltimore Eye Survey, Baltimore, Md; BDES, Beaver Dam Eye Study,Beaver Dam, Wis; BMES, Blue Mountains Eye Study, Sydney, New South Wales,Australia; RS, Rotterdam Study, Rotterdam, the Netherlands; Melbourne VIP,Melbourne Visual Impairment Project, Melbourne, Victoria, Australia; ProjectoVER, Vision Evaluation and Research, Nogales and Tucson, Ariz.
. The Prevalence of Refractive Errors Among Adults in the United States,Western Europe, and Australia. Arch Ophthalmol. 2004;122(4):495-505. doi:10.1001/archopht.122.4.495
Copyright 2004 American Medical Association. All Rights Reserved.Applicable FARS/DFARS Restrictions Apply to Government Use.2004
To estimate the prevalence of refractive errors in persons 40 yearsand older.
Counts of persons with phakic eyes with and without spherical equivalentrefractive error in the worse eye of +3 diopters (D) or greater, −1D or less, and −5 D or less were obtained from population-based eyesurveys in strata of gender, race/ethnicity, and 5-year age intervals. Pooledage-, gender-, and race/ethnicity–specific rates for each refractiveerror were applied to the corresponding stratum-specific US, Western European,and Australian populations (years 2000 and projected 2020).
Six studies provided data from 29 281 persons. In the US, WesternEuropean, and Australian year 2000 populations 40 years or older, the estimatedcrude prevalence for hyperopia of +3 D or greater was 9.9%, 11.6%, and 5.8%,respectively (11.8 million, 21.6 million, and 0.47 million persons). For myopiaof −1 D or less, the estimated crude prevalence was 25.4%, 26.6%, and16.4% (30.4 million, 49.6 million, and 1.3 million persons), respectively,of whom 4.5%, 4.6%, and 2.8% (5.3 million, 8.5 million, and 0.23 million persons),respectively, had myopia of −5 D or less. Projected prevalence ratesin 2020 were similar.
Refractive errors affect approximately one third of persons 40 yearsor older in the United States and Western Europe, and one fifth of Australiansin this age group.
A refractive error may be defined as a state in which the optical systemof the nonaccommodating eye fails to bring parallel rays of light to focuson the fovea. Myopia and hyperopia are the states of refractive error in whichthe optical system of the eye brings parallel rays of light into focus anteriorand posterior to the fovea, respectively, resulting in blurred vision. Mildto moderate hyperopia can be overcome by accommodation in youth and earlyadulthood, with the result that low degrees of hyperopia often are not noticeduntil the onset of presbyopia in midadulthood. Myopia results in blurred visionat all ages.
Blurred vision from refractive error can be relieved—in most cases—byneutralizing the refractive error with spectacles, contact lenses, or refractivesurgery. Nevertheless, the high prevalence of refractive errors and the costsof refractive correction make these conditions a substantial public healthand economic problem in many parts of the world. The extent of the problemof refractive error in the United States has not been evaluated, except inselect populations, since the 1971-1972 National Health and Nutrition ExaminationSurvey found that 25% of the 12- to 54-year-old US population was myopic.1
The Eye Diseases Prevalence Research Group, an initiative jointly sponsoredby the National Eye Institute of the National Institutes of Health (Bethesda,Md) and Prevent Blindness America (Schaumburg, Ill), seeks to estimate theprevalence rates for major eye disorders in older adults by combining datafrom large, high-quality, population-based eye surveys. Population-based surveysprovide the optimal method of estimating the population prevalence of medicalconditions. In recent years, several locally representative population-basedeye surveys have assessed the prevalence of refractive error in older adults.To estimate the current extent of the problem of refractive error in olderadults in the United States, Western Europe, and Australia, the Eye DiseasesPrevalence Research Group applied pooled data from these population-basedeye surveys to obtain population prevalence estimates for these regions.
For refractive error, eligible studies were those conducted since 1985in the United States, Western Europe, and Australia that used a method ofrefractive error measurement deemed to have an acceptable degree of reproducibilityand comparability between studies. Participating studies were the BaltimoreEye Survey in Maryland,2 the Beaver Dam EyeStudy in Wisconsin,3 the Proyecto VER (VisionEvaluation and Research) study in Arizona,4 theRotterdam Study in the Netherlands,5 the BlueMountains Eye Study in Australia,6 and theMelbourne Visual Impairment Project in Australia7 (Table 1). All eligible studies agreed toparticipate except the Salisbury Eye Evaluation Project.8 Theresearch activity was conducted in accordance with the principles of the Declarationof Helsinki and was approved by the institutional review board of The JohnsHopkins University School of Medicine, Baltimore.
Refractive error measurement varied only slightly between studies. Allstudies measured noncycloplegic refractive error. The Baltimore,2 BeaverDam,3 Proyecto VER,4 andRotterdam5 studies all used autorefractionfollowed by subjective refinement. The Blue Mountains study6 andthe Melbourne Visual Impairment Project7 scoredrefractive error as zero if uncorrected visual acuity was 54 letters or moreon an Early Treatment Diabetic Retinopathy Study chart, equal to the currentspectacles if visual acuity with current correction was 54 letters or more,or equal to an autorefraction with subjective refinement if neither of thesecriteria were met. These methods of ascertaining refractive status were judgedto be comparable for purposes of our outcome definitions (see "Comment" section).In addition to refractive error status, each of the population-based eye surveysprovided data on age, gender, and race/ethnicity based on self-report.
Refractive error was categorized on the basis of the consensus of ourresearch group. Hyperopia was defined as a refraction of +3 diopters (D) ormore positive, myopia was defined as a refraction of −1 D or more negative,and high myopia was defined as the subset of myopia of −1 D or lesswith a refraction of −5 D or more negative. For purposes of this report,refractive error results are reported by person, rather than by eye. The eyewith the larger absolute value of the measured spherical equivalent refractionwas used to categorize each subject's refractive state. Subjects for whomthe refraction of the more ametropic eye was between −1 D and +3 D,not inclusive, were considered not to have refractive errors.
Counts of the number of persons with phakic eyes with and without hyperopiaof +3 D or more, myopia of −1 D or less, and myopia of −5 D orless, as well as the number of persons with phakic eyes evaluated, were providedby the participating studies (Table 1)in strata of gender, race/ethnicity, and 5-year age bands (40-44 years, 45-49years, etc, through ≥80 years). Thus, persons with pseudophakic and aphakiceyes, who made up 3.7% to 8.3% of the populations studied, were not consideredin either the numerator or the denominator for refractive error prevalenceestimations, because their refraction had been altered at the time of cataractsurgery. The 2000 US Census data were used for the US population in that year.9 Year 2000 population data for Western Europe and Australia,as well as the anticipated population of all 3 regions in the year 2020, alsowere obtained from the US Department of the Census.10 (WesternEurope was defined as including Andorra, Austria, Belgium, Denmark, the FaroeIslands, Finland, France, Germany, Gibraltar, Greece, Guernsey, Iceland, Ireland,Italy, Jersey, Liechtenstein, Luxembourg, Malta, the Isle of Man, Monaco,the Netherlands, Norway, Portugal, San Marino, Spain, Sweden, Switzerland,and the United Kingdom.)
Age-specific prevalence rates for white, black, and Hispanic men andwomen were derived from the contributing studies in 2 steps. First, pooledprevalence proportions were estimated for each stratum of age, gender, andrace/ethnicity by means of minimum variance linear estimation. Stratum-specificproportions from each study were transformed by means of a log odds transformation;proportion variances were calculated assuming the binomial distribution. TheCochran test for homogeneity was used to evaluate variation between studiesfor the pooled age- and gender-specific rates. Second, logistic regressionmodels were fitted to the pooled prevalence proportions using the midpointof each age interval as the independent variable. Models for prevalence estimatesfor black and Hispanic persons each were based on the data from a single study,Baltimore and Proyecto VER (a study of primarily Mexican-American Hispanics),respectively.4 Population-based data meetingentry criteria on the prevalence of refractive errors in persons not identifyingthemselves as black, white, or Hispanic were not available. To make nationalprojections, rates for these other racial/ethnic groups were assumed to equalthe average of the pooled stratum-specific rates for white, black, and Hispanicpersons. Age and race/ethnicity effects in the models were tested by the Wald χ2 test statistic. Tests for gender differences, based on the observeddata from all contributing studies, were performed separately by race/ethnicitywith the Mantel-Haenszel χ2 test, controlling for both ageand study.
The estimated number of cases in the United States in each race/ethnicity,gender, and age category was generated by applying the modeled prevalencerate for each year of age to the US census population for the year 2000 andsumming over the age range for each 5-year age category. Projected estimateswere derived in the same manner with US Census middle series projections forthe year 2020, assuming constant age-, gender-, and race/ethnicity–specificrefractive error rates over time. Stratum-specific US prevalence rates werecomputed by dividing the total number of estimated cases for each stratumby the stratum-specific US population.
Prevalence estimates for Western Europe were based on the modeled age-specificrates for white persons from the Baltimore, Beaver Dam, and Rotterdam studies.Prevalence estimates for Australia were based on the pooled age- and sex-specificrates from the Blue Mountains and Melbourne Visual Impairment Project studies.
Six population-based eye surveys contributed data from a total of 29 281persons for the pooled analysis. Age- and gender-specific prevalence ratesof myopia and hyperopia were found to vary by geographic locale of the study(Figure 1, Figure 2, and Figure 3),with studies from the United States and Rotterdam having similar results,and Australian studies having lower prevalence rates for myopia and hyperopia.Therefore, for purposes of estimation of US and Western European prevalencerates for refractive errors—and for reports of association with age,race/ethnicity, and gender—only data from the Baltimore, Beaver Dam,Proyecto VER, and Rotterdam studies were used. The 20 957 subjects inthese studies (14 414 from US populations and 6543 from Western Europe[Rotterdam]) included 2348 black and 4507 Hispanic persons. For estimationof the Australian prevalence of refractive errors, only Blue Mountains andMelbourne Visual Impairment Project data were used, including results from8324 white subjects.
Estimates of age-, gender-, and race/ethnicity–specific prevalencerates based on the pooled data from the studies used to estimate the US andWestern European prevalence rates of refractive errors are given in Table 2. The prevalence rates for hyperopia,myopia of −1 D or less, and myopia of −5 D or less varied substantiallywith age and, to a lesser degree, with gender and race/ethnicity (Figure 1, Figure 2, and Figure 3).
The prevalence of hyperopia was observed to be progressively higherwith increasing age. The linear association of increasing hyperopia prevalencewith age was statistically significant (P ≤.01)in every gender- and race/ethnicity–specific fitted model except thatfor black men (P = .76), and was highly significantamong both white men (P <.001) and white women(P <.001). Among the gender- and race/ethnicity–specificgroups other than black men, the prevalence of hyperopia was 4.2- to 7.4-foldhigher in the oldest (≥80 years) with respect to the youngest (40-49 years)age stratum. Among black men, the prevalence of hyperopia was approximatelyconstant for all age groups.
Reciprocally, the prevalence of myopia of −1 D or less tendedto be substantially lower for older than younger age groups, as demonstratedin every gender- and race/ethnicity–specific model. However the associationof myopia and age was not linear, in that all gender- and race/ethnicity–specificgroups (except black men) demonstrated an increase in myopia among the oldestage groups. Among black men, who had very low rates of myopia in the oldestage groups, the prevalence of myopia was 8.0-fold higher in the youngest (40-49years) with respect to the oldest (≥80 years) age stratum. In other gender-and race/ethnicity–specific groups, the youngest age groups had a 1.3-to 2.7-fold higher prevalence of myopia of −1 D or less than the oldestage groups, whereas the youngest age groups had a 1.7- to 3.1-fold higherprevalence of myopia of −1 D or less than the age group with the lowestmyopia prevalence. Quadratic models that allowed for decreasing myopia prevalencewith age and also allowed for an increase in prevalence in the oldest agestrata were supported by the data for white women (P <.001),white men (P <.001), black women (P <.001), and Hispanic women (P = .03).For Hispanic men (P = .06), the quadratic model wasless well supported. For black men, the pattern of decreasing myopia prevalencewith age was strongly linear (P = .001), with almostno improvement in model fit by adding a quadratic term. For most gender- andrace/ethnicity–specific groups, the inflection point in the quadraticcurve appeared to lie close to age 70 years.
Similarly, myopia of −5 D or less was strongly associated withage among white persons (P = .007) and Hispanic persons(P = .008), in a nearly linear pattern, with thehighest prevalence in the youngest strata. Among white and Hispanic persons,the prevalence of myopia of −5 D or less was 2-fold and 3.3-fold higher,respectively, in the youngest with respect to the oldest stratum. However,no clear association between age and myopia of −5 D or less was evidentamong black persons (P = .71).
The general patterns of higher prevalence rates of hyperopia and lowerprevalence rates of myopia with increasing age were consistent across allstudies.
Overall, women were found to have a higher prevalence of hyperopia (oddsratio [OR], 1.28; P <.001) than men, after adjustingfor age and race/ethnicity. However, the overall age- and race/ethnicity–adjustedprevalence of myopia of −1 D or less was similar in men and women (OR,1.01; P = .89). The age-adjusted prevalence of myopiaof −5 D or less tended to be higher in women than men (OR, 1.16; P = .07). Although the latter association was nonsignificantin the pooled data set from the US-European studies, in the pooled data setfrom the Australian studies white women had significantly higher age-adjustedrates of myopia of −5 D or less than men (OR, 1.61; P = .002).
Adjusting for age and gender, white persons had significantly higherprevalence rates for hyperopia (OR, 1.22; P <.001),myopia of −1 D or less (OR, 1.25; P <.001),and myopia of −5 D or less (OR, 1.41; P <.001)than did Hispanic persons. Hispanic persons in turn had significantly higherage- and gender-adjusted prevalence rates of hyperopia (OR, 1.69; P <.001) and myopia of −1 D or less (OR, 1.52; P = .001) than did black persons. For myopia of −5 D or less,the age- and gender-adjusted prevalence was higher among Hispanic personsthan among black persons, but the difference was not statistically significant(OR, 1.32; P = .09). The age- and gender-adjustedprevalence rates of hyperopia (OR, 2.50; P <.001),myopia of −1 D or less (OR, 2.43; P <.001),and myopia of −5 D or less (OR, 2.52; P <.001)all were substantially higher among white persons than black persons.
The estimated overall prevalence rates for refractive errors in theUS population 40 years or older in the year 2000 are given in Table 3, Table 4, and Table 5. The estimated prevalence of hyperopiawas 9.9% (95% confidence interval, 9.7%-10.1%), and the estimated prevalenceof myopia of −1 D or less was 25.4% (95% confidence interval, 24.5%-26.4%).The estimated prevalence of myopia of −5 D or less was 4.5% (95% confidenceinterval, 4.2%-4.7%). Thus, persons with high myopia, according to this definition,made up 17.4% of all persons with myopia. These prevalence rates correspondto a burden of 11.8 million persons with hyperopia and 30.4 million personswith myopia of −1 D or less (5.3 million of whom have myopia ≤−5D) in the year 2000 US population 40 years or older. A total of 42.2 millionpersons (35.3%) in this age group are estimated to have either hyperopia ormyopia, according to our definitions. Applying age-, gender-, and race/ethnicity–specificrates to the anticipated population structure of the United States in theyear 2020, the projected prevalence rates for hyperopia, myopia of −1D or less, and myopia of −5 D or less are 10.8% (16.6 million persons),22.5% (34.7 million persons), and 4.0% (6.2 million persons), respectively.
In the Western European population for the year 2000, an estimated 11.6%of persons 40 years and older had hyperopia (21.6 million persons), 26.6%had myopia of −1 D or less (49.6 million persons), and 17.1% of thelatter group (4.6% of the general population, 8.5 million persons) had myopiaof −5 D or less. In the year 2020, projected prevalence rates for hyperopia,myopia of −1 D or less, and myopia of −5 D or less are 12.8% (27.8million persons), 26.5% (57.4 million persons), and 4.6% (10.0 million persons),respectively.
In Australia, which had lower age-specific prevalence rates for refractiveerrors and has a younger general population age structure than Western Europeand the United States, the estimated population prevalence rates for hyperopia,myopia of −1 D or less, and myopia of −5 D or less among the year2000 population 40 years and older were 5.8% (471 000 persons), 16.4%(1.3 million persons), and 2.8% (231 000 persons), respectively. In theyear 2020, projected prevalence rates are 6.4% (721 000 persons), 15.7%(1.8 million persons), and 2.6% (292 000 persons), respectively.
Pooled data from the participating population-based eye studies, conductedin persons 40 years or older, indicate that the crude prevalence of myopiais the highest of any eye disorder in this age group, affecting about 1 in4 persons in the United States and Western Europe, and about 1 in 6 Australians.Approximately 1 of 6 persons with myopia—1 of every 24 persons in thegeneral US and Western European population 40 years or older—has myopiaof −5 D or less and may be at risk of pathologic complications of highmyopia. Even with our conservative definition of hyperopia, its estimatedprevalence is substantial, affecting about 1 in 10 persons in this age group.Prevalence rates would be higher if a definition including lower degrees ofhyperopia were used. Refractive errors, in aggregate, affect approximately1 in 3 older adults in the United States and Western Europe, and about 1 in5 older Australians.
Because refractive error typically can be neutralized with spectacles,contact lenses, or refractive surgery, persons with refractive error who accesstreatment generally are not disabled. However, population-based studies havedemonstrated that refractive errors often are not adequately corrected inthe general population; rather, they generally are the leading cause of mildvisual impairment observed in population-based studies in developed nations.4,6,8,11- 13 Inaddition, vision loss uncorrectable by refraction may occur as a complicationof treatments for refractive error. Bacterial keratitis, the primary vision-threateningcomplication of contact lens use, has been estimated to occur at a rate of1 case per 2500 person-years of daily-wear contact lens use, and 1 case per500 person-years of extended-wear contact lens use.14 Visionloss due to complications of refractive surgery also is uncommon but appearsto be more frequent than with contact lens use.15- 17 Also,persons with a high degree of myopia appear to have higher risk of associatedocular diseases that may lead to vision loss, including peripheral retinallesions that may predispose to retinal detachment,18,19 glaucoma,20 cataract,21,22 andmyopic degeneration23 (sometimes complicatedby choroidal neovascularization24). In theRotterdam study, 6% of visual impairment was attributed to myopic degeneration,which was 1 of the 2 leading causes of visual impairment in persons youngerthan 75 years.25 Our data suggest that a largenumber of persons are in this at-risk category. Thus, while the per-case riskof loss of best-corrected visual acuity in persons with refractive error isless than that of the other eye conditions studied by our group, the highprevalence of refractive errors suggests that the absolute number of personslosing vision as a result of complications of contact lens use, refractivesurgery, or high myopia–associated ocular diseases is substantial.
Similarly, while the per-person costs of using spectacles and contactlenses are modest (by developed-country standards), in aggregate these costsare substantial. An estimated $12.8 billion (1990 US dollars) was spent tocorrect refractive errors of Americans in 1990.26 Refractivesurgery incurs a higher up-front cost than spectacles or contact lenses, butthese costs may be recovered to a large extent over the years,26 providedthat subsequent shifts in refractive state that require refractive correctionare unusual—an assumption that may not be warranted (discussed later).
Our results demonstrate a strong association of refractive error withage, wherein older persons tended to have higher rates of hyperopia and lowerrates of myopia. It has been debated whether this pattern of association isdue to changes in the optical system with age, or differences in the environmentexperienced during life by persons born during different calendar-year periods(cohort effects). Graphic analysis of cross-sectional data from 3 US eye surveyssuggested that the observed pattern is more consistent with age effects thancohort effects.27 However, the Beaver Dam EyeStudy, which has provided the only available report of longitudinal changesin refractive error with age during a 10-year period in a population-basedsample,28 found that both aging and cohorteffects on refractive error appeared to be important in determining changesin age-specific prevalence of refractive error over time. Cohort effects weredemonstrated by substantial changes in mean refractive error between birthintervals, nearly −0.5 D per 5-year birth interval from 1918 through1942, suggesting that environmental factors changed in a manner that promotedmyopia for persons born during this period in rural Wisconsin. Further longitudinaldata are needed to evaluate whether similar cohort effects occurred in otherpopulations during this period and subsequently. Aging effects also were demonstratedin the Beaver Dam study, in that participants aged 43 to 59 years at enrollmenthad an average 10-year change in refraction of +0.54 D, those initially aged70 years and older had an average 10-year change in refraction of −0.41D, and those aged 60 to 69 years at enrollment had little change. A similarpattern has been observed in 5-year longitudinal data from the Blue MountainsEye Study.29 These patterns are in close agreementwith the pattern observed in our pooled cross-sectional myopia data, in whichthe prevalence of myopia of −1 D or less declined with age to a nadirnear the age of 70 years, after which an increase in the prevalence of myopiaof −1 D or less was observed.
A hyperopic shift of this magnitude in the population's average refractiveerror during middle age would seem to have considerable importance in evaluatingthe appropriate use of and goals for refractive surgery in myopic adults.Potential explanations for a hyperopic shift in measured refractive errorduring middle age include loss of accommodative tone with progressive presbyopia,and/or actual biometric changes (decreasing axial length and/or increasingcorneal power) during this period of aging. These theories could be addressed,at least partially, by performing cycloplegic refractions and biometry infuture longitudinal population-based eye studies. Under the accommodativetone theory, studies using cycloplegic refraction would be expected to finda hyperopic shift of the refractive error distribution with respect to ourresults (based on noncycloplegic refraction) in younger age groups, with higherrates of hyperopia and lower rates of myopia.
The theory that nuclear sclerosis contributes to a myopic refractiveshift in the elderly is well supported by longitudinal data.28,29 However,actual biometric changes with age also could have contributed to a myopicchange in age-specific prevalence in elderly adults. It is worth noting thatif elderly persons with nuclear sclerosis–induced myopia in the participatingstudies were more likely to undergo cataract surgery than other study participantsbefore participating in the surveys—resulting in censoring from thepooled analysis—the myopic shift in the elderly may be even greaterthan is reported herein.
Other important observations of our study were that hyperopia, but notmyopia (except perhaps high degrees of myopia), was more common in women thanmen, and that race/ethnicity–specific differences in the prevalenceof refractive errors appear to exist. It is interesting to note that bothhyperopic and myopic refractive errors occurred more frequently among whitepersons than Hispanic persons, and among Hispanic persons than black persons.Although only one study, the Baltimore Eye Survey, evaluated more than oneracial group simultaneously, the pattern we observed in our pooled analysiswas observed within that study. Thus, the observed differences between racial/ethnicgroups appear to exist in the variability aspect of the refractive error distribution,suggesting that the Hispanic and especially black racial/ethnic groups, onaverage, may more frequently have successful emmetropization during life thanthe white group. Research evaluating the genetic and/or environmental factorsunderlying these differences between racial/ethnic groups is needed. WesternEuropean, US, and Australian data were sparse regarding Asian persons, a groupthought to have a different pattern of refractive error prevalence than theracial/ethnic groups studied herein,30,31 withtoo few persons studied to make population prevalence estimates. Researchassessing the prevalence of refractive errors in Asian persons living in theUnited States, Western Europe, and Australia is needed.
We interpreted differences between the US and Western European studieswith respect to the Australian studies as reflecting regional differencesin the prevalence of refractive errors. An alternative explanation of theobserved differences is the possibility that the populations included in theestimation process for the United States and Western Europe and/or the 2 Australianstudies were nonrepresentative to a substantial degree. However, because thestudies were all population based and showed agreement with each other withinthese geographic groupings, this possibility seems unlikely. Another alternativeexplanation is that because the Australian studies did not perform autorefractionon participants who saw well with their current refraction, they underascertainedrefractive errors. However, it is difficult to believe that a substantialnumber of participants with hyperopia of +3 D or greater and/or myopia of−1 D or less would have had normal visual acuity on rigorous testing,as done in these studies. Therefore, our results suggest that a real differenceexists in the prevalence of refractive errors between Australia and the UnitedStates–Western Europe. Further research as to what environmental differencesmight underlie regional differences in the age-, gender-, and race/ethnicity–specificprevalence of refractive errors potentially could provide important informationregarding the etiology of refractive error and means of prevention.
Several potential limitations of our study should be considered in interpretingthese results. First, we have assumed that the study populations of the contributingstudies are representative of the general populations to which we have appliedtheir results. Clearly, results from our locally representative studies cannotbe generalized to the US population with as much confidence as results froma nationally representative study, if such a study were available. In particular,only results from white persons were used for Western European and Australianprevalence calculations, which may have inadequately represented the prevalencerates in other racial/ethnic groups in these regions.
Second, our data do not address the prevalence of refractive errorsfor some groups of interest. For children and younger adults, no recent datawere available. As noted previously, the National Health and Nutrition ExaminationSurvey in 1971-1972 found that persons aged 12 to 54 years had high ratesof myopia1; therefore the lack of data on thisgroup is an important gap in our knowledge of the prevalence of refractiveerrors. Population-based studies of younger Americans are needed to allowsuch estimates to be made. In addition, data for black and Hispanic personswere obtained from single studies. Because the particular populations studiedmay not have given an average representation of the broader black and Hispanicpopulations, our prevalence estimates for these groups may be less reliablethan are the estimates for white persons, which are based on pooled data frommultiple population-based studies. Because no data were available regardingthe prevalence rates for refractive errors in racial/ethnic groups other thanwhite, black, and Hispanic persons, we had to rely on arbitrary methods toestimate the prevalence of refractive errors in these groups for US estimatesand projections. Further research evaluating the prevalence of eye diseasesin these groups would be of interest, particularly among persons of Chineseand Indian descent, who, respectively, have been reported to have remarkablyhigh30 and low31 ratesof refractive errors in population-based studies in Asia. Findings of no increasedrisk of myopia among Asians in the Melbourne Visual Impairment Project7 suggest that results from Asian persons living inAsia may not be generalizable to a non-Asian setting.
Third, data are reported according to categories determined by consensusof the Eye Diseases Prevalence Research Group. Categories that were more inclusive(eg, −0.50 D or more negative, +1 D or more positive) would have resultedin even higher estimated prevalence rates. We elected to choose these categoriesto take a conservative approach in reporting the prevalence of refractiveerrors. However, this conservative approach may have underestimated the numberof persons in the populations who would benefit from refractive correction,particularly among older persons with lesser degrees of hyperopia than +3D. Similarly, our use of the threshold of −5 D for "high" myopia probablyled to higher prevalence estimates than the other commonly used thresholdof −6 D would have. Sensitivity analyses of how prevalence rates wouldhave differed with different outcome definitions were not possible, becausethe contributing studies only provided data regarding the cutoffs agreed onby the study group in advance of the pooled analysis.
Finally, future projections of the prevalence rates for refractive errorsare difficult to make because longitudinal information from population-basedcohort studies is limited to 10-year results from the Beaver Dam study, andlesser degrees of follow-up from other studies. Our future projections areprobably the least reliable estimates produced by this study, because cohorteffects may alter prevalence rates in years to come in ways that we cannotpredict. Use of these future projections for policy decision making shouldtake their uncertainty into account. However, because changes in the prevalenceof myopia and hyperopia are not driven by age to the same extent as the prevalenceof age-related eye diseases, the dramatic changes in prevalence expected withpopulation aging for cataract, age-related macular degeneration, and glaucomaare not as likely to occur for refractive errors in aggregate.
A relative strength of this study, with respect to most pooled analyses,is that the outcomes of interest were ascertained by highly comparable, reliablemethods. All studies included in the estimation of US and Western Europeanprevalence rates used autorefraction followed by subject refinement to measurerefractive error. Direct provision of data according to our specificationsby each contributing study team allowed use of identical cutoffs for the diagnosesof hyperopia and myopia.
In summary, refractive errors of a magnitude expected to require refractivecorrection are estimated to be present in one third of the population 40 yearsor older in the United States and Western Europe, and one fifth of this populationin Australia. The prevalence of refractive errors is strongly related to ageand varies with gender and race/ethnicity. More research is needed in theUnited States, Western Europe, and Australia regarding the prevalence ratesfor refractive errors in younger age groups and in minority populations suchas persons of Asian origin to more accurately estimate the extent of the problemin these groups. Also, further research regarding changes in the refractionof individuals over time seems warranted, to evaluate whether refractive surgeryis really likely to provide a long-term solution to the problem of refractiveerror, a concept that seems uncertain on the basis of our results and thoseof studies with longitudinal follow-up of individuals.28,29 Althoughthe morbidity (if refractive correction is available and is used) and costassociated with refractive errors are lower on a per-case basis than for someother ocular disorders, the aggregate morbidity and cost are substantial enoughthat these conditions should be considered a priority subject for health policydecision making and research.
Corresponding author and reprints: John H. Kempen, MD, PhD, 550 NBroadway, Suite 700, Baltimore, MD 21205 (e-mail: firstname.lastname@example.org).
Submitted for publication May 12, 2003; final revision received November19, 2003; accepted November 19, 2003.
From the Department of Ophthalmology, The Johns Hopkins University Schoolof Medicine, Baltimore, Md (Drs Kempen, Tielsch, Congdon, and Friedman andMs Broman); Departments of Epidemiology (Drs Kempen and Tielsch) and InternationalHealth (Drs Tielsch, Congdon, and Friedman), The Johns Hopkins UniversityBloomberg School of Public Health, Baltimore; Department of Ophthalmology,University of Sydney, Sydney, New South Wales (Dr Mitchell); Centre for VisionResearch, Westmead Hospital, Westmead, New South Wales (Dr Mitchell); Departmentof Ophthalmology and Visual Sciences, University of Wisconsin Medical School,Madison (Ms Lee); Centre for Eye Research Australia, University of Melbourne,Melbourne, Victoria (Dr Taylor); Departments of Epidemiology and Biostatistics,Erasmus Medical Center, Rotterdam, the Netherlands (Dr Ikram), and Schoolof Public Health and Health Services, George Washington University MedicalCenter, Washington, DC (Ms O'Colmain); and Macro International, Inc, Calverton,Md (Ms O'Colmain).
This study was supported by contract NO1-EY-8-2108 from the NationalEye Institute, Bethesda, Md. Additional support was provided by grant EY00386(Dr Kempen) from the National Eye Institute.
Data from the 2000 US Census is in the public domain. The research groupgratefully acknowledges the work of the many contributors to create thesedata.
The members of the Eye Diseases Prevalence Research Group are as follows:
Baltimore Eye Survey, Baltimore, Md: JamesM. Tielsch, Alfred Sommer, Joanne Katz, Harry A. Quigley. Beaver Dam Eye Study, Beaver Dam, Wis: Barbara E. K. Klein, RonaldKlein, Scot E. Moss, Kristine E. Lee, Sandra C. Tomany. Blue Mountains Eye Study, Sydney, New South Wales, Australia: PaulMitchell, Jie Jin Wang, Elena Rochtchina, Wayne Smith, Robert G. Cumming,Karin Attebo, Jai Panchapakesan, Suriya Foran. MelbourneVisual Impairment Project, Melbourne, Victoria, Australia: Hugh R.Taylor, Cathy McCarty, Bickol Mukesh, LeAnn M. Weih, Patricia M. Livingston,Mylan Van Newkirk, Cara L. Fu, Peter Dimitrov, Matthew Wensor, Yury Stanislavsky. Proyecto VER (Vision Evaluation Research), Nogales and Tucson,Ariz: Sheila K. West, Jorge Rodriguez (deceased), Aimee Broman, BeatrizMuñoz, Robert Snyder, Ronald Klein, Harry A. Quigley. Rotterdam Study, Rotterdam, the Netherlands: Paulus T. V. M. de Jong,Johannes R. Vingerling, Roger C. W. Wolfs, Caroline C. W. Klaver, Albert Hofman,Redmer van Leeuwen, M. Kamran Ikram, Simone de Voogd.
Coordinating Center: John H. Kempen, NathanG. Congdon, David S. Friedman, Benita J. O'Colmain; NationalEye Institute: Frederick L. Ferris III.