Distribution of spherical equivalent refraction by age. Data for those aged 12 through 19 years are not shown because of the possible effects of accommodation on noncycloplegic refractions. D indicates diopters.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Vitale S, Ellwein L, Cotch MF, Ferris FL, Sperduto R. Prevalence of Refractive Error in the United States, 1999-2004. Arch Ophthalmol. 2008;126(8):1111–1119. doi:10.1001/archopht.126.8.1111
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
To describe the prevalence of refractive error in the United States.
The 1999-2004 National Health and Nutrition Examination Survey (NHANES) used an autorefractor to obtain refractive error data on a nationally representative sample of the US noninstitutionalized, civilian population 12 years and older. Using data from the eye with a greater absolute spherical equivalent (SphEq) value, we defined clinically important refractive error as follows: hyperopia, SphEq value of 3.0 diopters (D) or greater; myopia, SphEq value of −1.0 D or less; and astigmatism, cylinder of 1.0 D or greater in either eye.
Of 14 213 participants 20 years or older who completed the NHANES, refractive error data were obtained for 12 010 (84.5%). The age-standardized prevalences of hyperopia, myopia, and astigmatism were 3.6% (95% confidence interval [CI], 3.2%-4.0%), 33.1% (95% CI, 31.5%-34.7%), and 36.2% (95% CI, 34.9%-37.5%), respectively. Myopia was more prevalent in women (39.9%) than in men (32.6%) (P < .001) among 20- to 39-year-old participants. Persons 60 years or older were less likely to have myopia and more likely to have hyperopia and/or astigmatism than younger persons. Myopia was more common in non-Hispanic whites (35.2%) than in non-Hispanic blacks (28.6%) or Mexican Americans (25.1%) (P < .001 for both).
Estimates based on the 1999-2004 NHANES vision examination data indicate that clinically important refractive error affects half of the US population 20 years or older.
Refractive error is recognized as one of the most important causes of correctable visual impairment,1-4accounting for nearly 80% of the visual impairment in persons 12 years and older in the United States.5Providing eye care services to the many persons who use or need refractive correction involves substantial expense: the direct annual cost of refractive correction for distance visual impairment is estimated to be between $3.8 and $7.2 billion6for persons 12 years and older (based on an estimated annual direct cost per person of $35-$56) and, in a separate study7of persons 40 years and older, $5.5 billion. We used the 1999-2004 National Health and Nutrition Examination Survey (NHANES) data to estimate the population prevalence of refractive error and to describe the refractive characteristics of the US population in greater detail.
The NHANES is an ongoing nationally representative survey of the US civilian, noninstitutionalized population conducted by the National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention.8,9Participants in NHANES undergo a home interview and subsequent comprehensive physical examination and functional assessment in a mobile examination center (MEC).8The 1999-2004 NHANES protocol was reviewed and approved by the NCHS Research Ethics Review Board. All participants (and parents or guardians of minors) gave their written informed consent after receiving a description of the study and the potential risks of participation.
Demographic characteristics were collected at the household interview. Participants reported their ethnicity and race.10The NCHS later categorized these classifications into non-Hispanic black, non-Hispanic white, Mexican American, or other (including Asian [various countries], Pacific Islander, Native American, non–Mexican American Hispanic, and multiracial).
The vision examination was conducted on participants 12 years and older. An autorefractor (NIDEK ARK-760; Nidek Co Ltd, Tokyo, Japan) was used to measure the refractive error of each eye after corrective lenses had been removed. Three separate measurements of sphere, cylinder, and axis were acquired and averaged by the autorefractor. For myopia and hyperopia, we categorized participants based on the refractive error measurements from the eye with the larger absolute spherical equivalent (SphEq) value. If one eye had missing refractive error data, data from the eye with available data were used. Refractions were recorded in plus cylinder notation, and the SphEq value was computed as the sphere measurement plus half the cylinder measurement. Clinically important myopiawas defined as a myopic SphEq value of at least −1.0 diopter (D); clinically important hyperopiawas defined as a hyperopic SphEq value of 3.0 D or more. Severe myopiawas defined as a myopic SphEq value of at least −5.0 D.11,12We also present data with myopia defined as a SphEq value of at least −0.5 D. Astigmatismwas defined as a cylinder of 1.0 D or greater in the eye with the larger cylinder value.
Because of time constraints imposed by the comprehensive NHANES, it was not feasible to obtain cycloplegic refractions. Accommodation may, therefore, have affected the measurement of the refractive error, particularly in younger participants, despite the autofogging used by the autorefractor to minimize accommodation. Consequently, for persons aged 12 through 19 years, the prevalence of myopia is likely to be overestimated and the prevalence of hyperopia is likely to be underestimated by the autorefractor measurements. We therefore have not used refractive error data from the 1999-2004 NHANES for estimating the prevalence of refractive error among persons aged 12 through 19 years.
Refractive error data obtained from eyes with a history of cataract surgery or refractive surgery, or in which contact lenses were worn during objective refraction, were treated as missing values. Participants excluded from analyses (n = 2203) were those who reported a history of cataract surgery (n = 791) or refractive surgery (n = 142) in both eyes, those who wore contact lenses during objective refraction (n = 19), and those who were missing autorefractor data for both eyes (n = 1251, because of lack of time, inability to cooperate with the protocol, or equipment malfunction). We included 615 participants with refractive error data for only 1 eye (the other eye had cataract extraction [n = 264], refractive surgery [n = 21], or missing autorefractor [n = 330] data).
NHANES uses a complex, multistage probability sample design with oversampling within selected demographic subgroups.13Sampling weights derived by NCHS statisticians reflect the probability of selection into the sample and incorporate adjustments for differential nonresponse and poststratification,14,15and their incorporation into analyses16(using SUDAAN statistical software, version 9.0.0; Research Triangle Institute, Research Triangle Park, North Carolina) ensures that the weighted estimates reflect the US population sizes for specified demographic categories and that standard errors are unbiased.17,18We age standardized prevalence estimates to the 2000 US Census population.19
A total of 14 213 participants 20 years and older participated in the 1999-2004 NHANES MEC examination. Of these, 12 010 (84.5%) had complete refractive error data (Table 1). Participants with incomplete refractive error data were more likely to be older and female, to report a lower annual income, and to have fewer years of formal education.
The distribution of SphEq refraction by age is shown in the Figure. Participants 60 years and older were less likely to have myopic refractive error and more likely to have hyperopic refractive error than younger individuals.
The prevalence of hyperopia (SphEq value ≥3.0 D) (Table 2) for participants aged 20 through 39 years, 40 through 59 years, and 60 years and older was 1.0% (95% confidence interval [CI], 0.6%-1.4%), 2.4% (95% CI, 1.7%-3.0%), and 10.0% (95% CI, 9.1%-10.9%), respectively. In persons 60 years and older, hyperopia was more common in women (12.9%) than in men (6.6%) (P < .001).
The overall pattern of myopia prevalence by age and sex was similar, regardless of the definition of myopia (≤−1.0 D, ≤−0.5 D, or ≤−5.0 D): prevalence estimates were approximately equal for the 20- through 39-year-old and 40- through 59-year-old groups and markedly lower for those 60 years and older; for those 20 through 39 years old, women had a higher prevalence than men. For myopia defined as a SphEq value of −1.0 D or less (Table 3), prevalence estimates for participants aged 20 through 39 years, 40 through 59 years, and 60 years and older were 36.2% (95% CI, 34.2%-38.3%), 37.6% (95% CI, 35.1%-40.1%), and 20.5% (95% CI, 18.3%-22.8%), respectively. Myopia of −1.0 D or less was more prevalent in non-Hispanic whites (35.2%) than in non-Hispanic blacks (28.6%) or Mexican Americans (25.1%) (P < .001 for both). For those 60 years and older, no differences in prevalence were observed among race/ethnicity categories or between men and women. For participants aged 20 through 39 years and 40 through 59 years, the prevalence of myopia of −1.0 D or less was higher for non-Hispanic whites (38.7% and 40.6%, respectively) than for non-Hispanic blacks (32.0% and 32.1%, respectively; P < .001 for both) or Mexican Americans (27.1% and 26.4%, respectively; P < .001 for both). For those aged 20 through 39 years, women had a slightly higher prevalence of myopia of −1.0 D or less (39.9%) than did men (32.6%) (P < .001).
The prevalence of myopia defined as an SphEq value of −0.5 D or less (Table 4) for participants aged 20through 39 years, 40 through 59 years, and 60 years and older was 50.2% (95% CI, 47.8%-52.7%), 50.1% (95% CI, 47.8%-52.4%), and 26.5% (95% CI, 24.0%-29.0%), respectively. Myopia of −0.5 D or less was most prevalent in those categorized as non-Hispanic white (46.0%) compared with those categorized as non-Hispanic black (41.5%, P = .003) or Mexican American (40.8%, P < .001). The prevalence of myopia of −0.5 D or less did not differ significantly between men and women, except in those 20 through 39 years, for whom the prevalence in women (53.9%) was significantly higher than in men (46.6%) (P < .001).
The prevalence of severe myopia (SphEq value ≤−5.0 D) (Table 5) for participants aged 20 through 39 years, 40 through 59 years, and 60 years and older was 7.4% (95% CI, 6.5%-8.3%), 7.8% (95% CI, 6.4%-9.1%), and 3.1% (95% CI, 2.2%-3.9%), respectively. Severe myopia was most prevalent in those categorized as non-Hispanic white (7.0%) compared with non-Hispanic blacks (4.7%, P = .001) and Mexican Americans (3.6%, P < .001). The prevalence of severe myopia did not differ significantly between men and women, except in those aged 20 through 39 years, for whom the prevalence in women (9.2%) was significantly higher than in men (5.6%) (P < .001).
The prevalence of astigmatism (Table 6) increased with increasing age: for ages 20 through 39 years, 40 through 59 years, and 60 years and older, prevalence estimates were 23.1% (95% CI, 21.6%-24.5%), 27.6% (95% CI, 25.8%-29.3%), and 50.1% (95% CI, 48.2%-52.0%), respectively. The prevalence of astigmatism varied little by race/ethnicity category. In those 60 years and older, astigmatism was more prevalent among men (54.9%) than among women (46.1%) (P < .001).
The prevalence of any clinically important refractive error (myopia, hyperopia, and/or astigmatism) (eTable) increased with increasing age. Prevalence estimates for participants aged 20 through 39 years, 40 through 59 years, and 60 years and older were 46.3% (95% CI, 44.5%-48.0%), 50.6% (95% CI, 48.1%-53.0%), and 62.7% (95% CI, 60.3%-65.1%), respectively. The prevalence of any clinically important refractive error was lower for Mexican Americans (44.4%) than for non-Hispanic whites (53.4%, P < .001) or non-Hispanic blacks (49.3%, P = .002). For those 60 years and older, the prevalence of any clinically important refractive error was higher in men (66.8%) than in women (59.2%) (P < .001).
We found that refractive error was common in the US population 20 years and older: the prevalence of myopia was 33%; of severe myopia, 6.5%; of hyperopia, 3.6%; and of astigmatism, 36%. Our estimated prevalence of myopia was higher than the 25% reported in previous US studies11,20and similar (in persons ≥40 years) to that of ethnic Chinese persons in Singapore.12General overall statements regarding the distribution of myopia among demographic subgroups cannot be made because we found statistically significant interactions between age and sex (P < .001) and between age and race/ethnicity (P < .001), as well as a borderline statistically significant interaction between sex and race/ethnicity (P = .07). Overall, it appears that the pattern of myopia prevalence by age is similar, regardless of the definition of myopia, with nearly identical prevalence estimates within the 20- to 39-year and 40- to 59-year age groups and a markedly lower prevalence for those 60 years and older (approximately half that for the younger ages). Sex differences in myopia prevalence were observed within the 20- to 39-year age group, in whom myopia (SphEq value ≤−1.0 D) was more prevalent in women (40%) than in men (33%) (P < .001); the same pattern was observed for myopia defined as an SphEq value of −0.5 D or less (P < .001) and for myopia defined as an SphEq value of −5.0 D or less (P < .001). The prevalence of myopia of −0.5 D or less was similar to that found in studies of Asian populations.12Sex differences in refractive error prevalence were also observed within the 60 years or older group, in whom hyperopia was more common (13% vs 7%; P < .001) and astigmatism less common (46% vs 66%; P < .001) among women than among men. The prevalence of refractive error varied by age, with those 60 years and older being less likely to have myopia (P < .001 for all myopia definitions) and more likely to have hyperopia (P < .001) and/or astigmatism (P < .001) than younger persons. Although hyperopia and astigmatism prevalence did not vary among race/ethnicity categories, myopia prevalence was higher in non-Hispanic whites than in non-Hispanic blacks (38.7% vs 28.6%; P < .001) or Mexican Americans (38.7% vs 25.1%; P < .001) (although this race/ethnicity difference was not apparent in those ≥60 years).
Most previous epidemiologic studies that provide prevalence estimates for refractive error were limited to a specific geographic location or age group. The 1971-1972 NHANES provided the first US-wide population-based prevalence estimates for many eye conditions, including refractive error. Sperduto et al20found that the prevalence of myopia in the United States in persons aged 12 to 54 years was approximately 25% (right eye, based on lensometry [for those with visual acuity of 20/40 or better] or on retinoscopy [for those with worse visual acuity]) compared with 33.1% in the current study (in persons ≥20 years). The prevalence estimates for myopia are substantially higher for both non-Hispanic black and non-Hispanic white participants in 1999-2004 (28.6% and 35.2%, respectively) than in 1971-1972 (13.0% and 26.3%, respectively). It is possible that myopia prevalence in the United States has increased during the 30-year period between the 2 NHANES studies. A previous study21documented an increased prevalence of myopia during a 12-year period in young Israeli adults and attributed it to a possible cohort effect. However, because of considerable differences in the methods used to define myopia in the 2 NHANES studies, additional research would be required to conclude that prevalence of myopia had increased and, if so, to explore possible reasons for such an increase.
Recently, the Eye Disease Prevalence Research Group (EDPRG) performed a meta-analysis of existing population-based studies in persons 40 years or older (conducted from 1985 to 2000) to derive robust population estimates for the prevalence of refractive error in the United States and other countries.11Definitions of hyperopia, myopia, and severe myopia were the same as in our study. Our NHANES-based estimate of prevalence of myopia in the United States for persons 40 years and older was 31.0% (95% CI, 29.1%-32.9%), which is substantially higher than the EDPRG estimate for the United States (25.4%; 95% CI, 24.5%-26.4%). The prevalence of myopia was higher in the recent NHANES study than in the EDPRG for both non-Hispanic blacks and non-Hispanic whites.
The prevalence of myopia of −0.5 D or less in the current study was 41.0% for persons 40 years and older. Wong et al12found a similar prevalence of myopia (38.7%) in ethnic Chinese persons living in Singapore. However, the prevalence of severe myopia in the current study was lower (6.0%) than that found by Wong et al (9.1%), possibly attributable to differences in the age structure of the 2 populations because the United States may have more older individuals (who have a relatively lower prevalence of myopia) than Singapore. Our estimated prevalence of myopia of −0.5 D or less for persons 60 years and older was 26.5%, greater than the 19.5% estimated from the Rotterdam Study22(persons ≥55 years). The prevalence of myopia in the Blue Mountains Eye Study23was 15%, which is strikingly lower than the prevalence of myopia of −0.5 D or less found in the current NHANES and that reported from the Beaver Dam Eye Study.24The Blue Mountains Eye Study and Rotterdam Study populations may have a different distribution of underlying risk factors for myopia that could explain their lower prevalence estimates, but we are unable to address this issue within the scope of the current study.
Our myopia prevalence estimates for non-Hispanic black participants aged 40 through 59 years (32.1%; 95% CI, 28.1%-36.1%) were also higher than those reported by the Barbados Eye Study25(conducted from 1987 through 1992) for participants aged 40 through 59 years (9%). Our prevalence estimates for myopia in those 60 years and older appear comparable to those reported from the Barbados Eye Study.
The EDPRG results for Hispanic individuals were based on a single study4in a Mexican American population (Proyecto Ver [Nogales, Arizona]). Recently, the Los Angeles Latino Eye Study (LALES)26reported prevalence estimates of myopia for Hispanic persons 40 years and older (>90% of whom reported Mexican ancestry27). For those aged 40 through 59 years, the prevalence of myopia among NHANES Mexican American participants was higher than in their LALES and EDPRG counterparts. For those 60 years and older, NHANES and LALES prevalences were similar and higher than those from the EDPRG.
It is not possible for us to compare results for “other” race/ethnicity with previous studies because this category includes participants with a wide range of ancestries and NHANES does not provide data to allow further differentiation within this subgroup. However, because the NHANES sample as a whole is designed to represent the entire US population, it is important to include the other subgroup in calculating estimates for the entire US population.
We found that the prevalence of myopia was lowest in participants 60 years and older, consistent with other population-based studies.28,29Although it is possible that the lower prevalence of myopia in those 60 years and older is because of a cohort effect, as reported in other, younger populations,21,30this observation may also be because of intrinsic, age-related changes in the refractive components of the eye.31,32
The observed lower prevalence of myopia in persons 60 years and older may also in part be because of an association of cataract surgery (a criterion for exclusion) with myopic refractive status.33Because we have no information on the refractive status of participants before cataract surgery, we are unable to explore this hypothesis.
Our study has several important limitations. Because of time constraints and potential for participant fatigue, we could not perform a full ophthalmic examination on all participants in the MEC. Our study protocol, therefore, included the use of an autorefractor, operated according to a standard protocol by trained operators. Our noncycloplegic refraction values were based on the mean of 3 measurements by the autorefractor and did not include subjective refinement. This may have resulted in overestimates of the true number with myopia,34especially among younger participants, who have a higher amplitude of accommodation. Because of this concern, we did not use the 1999-2004 NHANES data to provide prevalence estimates of refractive error for those aged 12 through 19 years.
As in the EDPRG, we categorized participants based on the refractive characteristics of the eye with a greater absolute SphEq value, resulting in an estimated myopia prevalence of 32.99% (95% CI, 31.47%-34.52% [based on 5709 individuals]). If we had considered an individual as myopic if either eye was myopic,35we would have classified an additional 15 participants as myopic, resulting in a prevalence of 33.05% (95% CI, 31.52%-34.58%).
Although the NHANES sampling weights correct for nonresponse within population subgroups, it is possible that NHANES participants who had vision data had different refractive characteristics than those who did not have vision data. Response rates to the home interview and MEC examination were lower in older individuals.36Participants who reported for the MEC examination may have had missing vision data because of equipment malfunction or inability to cooperate with the examination.
Because refractive error's effect on visual acuity can be mitigated relatively easily, it has sometimes been overlooked as an important cause of visual impairment.37,38Many previous epidemiologic studies39based definitions of visual impairment on best-corrected rather than presenting (ie, with habitual correction) visual acuity. However, newer research40,41has focused attention on the benefits of correcting refractive error. A recent analysis7(p1760)found that much of the economic burden posed by vision disorders is because of refractive error, concluding that “interventions to diagnose and treat uncorrected refractive error have the potential to be highly cost-effective based on the improvements in patient quality of life they generate.”
In summary, the 1999-2004 NHANES data show that half of the US population 20 years and older has some type of clinically important refractive error. Refractive error is, therefore, the most common condition affecting the ocular health of the US population, involving young adults, middle-aged persons, and older adults of all ethnicities. In a previous study,6we estimated the annual direct cost of providing refractive correction to the 100 million people who need it to achieve good vision as exceeding $3.5 billion (not including the costs of identifying those who need refractive correction). Others estimated the economic burden (including indirect costs) of refractive error in those 40 years and older to be $5.5 billion.7Accurate, current estimates of the prevalence of refractive error are essential for projecting vision care needs and planning for provision of vision care services to the many people affected.
Correspondence:Susan Vitale, PhD, MHS, 5635 Fishers Ln, Ste 1100, Bethesda, MD 20892-9301 (firstname.lastname@example.org).
Submitted for Publication:August 20, 2007; final revision received January 2, 2008; accepted January 3, 2008.
Author Contributions:Dr Vitale had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosure:None reported.
Funding/Support:The NHANES is sponsored by the NCHS, Centers for Disease Control and Prevention. Additional funding for the NHANES Vision Component was provided by the National Eye Institute, National Institutes of Health, Intramural Research Program grant Z01EY000402.
Role of the Sponsors:The NCHS provided funding support for NHANES and was involved in the design, management, and conduct of the study and in data collection but was not involved in the analysis or interpretation of the vision examination study results or in the preparation, review, or approval of the manuscript. The National Eye Institute provided funding support for the vision examination and was involved in the design and conduct of the vision component, in the collection, analysis, and interpretation of the vision data, and in the preparation, review, and approval of this article before submission.
Additional Contributions:The staff of the NCHS at the National Centers for Disease Control and Prevention, especially Brenda G. Lewis, BS, ASCP, MPH, provided expert assistance with all technical aspects of the vision component; the staff at Westat Inc, especially Kay Apodaca, BA, MSW, and Beryl D. Carew, MPH, BN, helped with data collection activities; and Susan K. Corwin, CO, COMT, trained field staff. We thank the participants in the NHANES, without whose time and dedication this study would not have been possible.
Create a personal account or sign in to: