Refractive Error Study in Children: Results From Shunyi District, China

JIALIANG ZHAO, MD, XIANGJUN PAN, OD, RUIFANG SUI, MD, SERGIO R. MUNOZ, PHD, ROBERT D. SPERDUTO, MD, AND LEON B. ELLWEIN, PHD

PURPOSE:

To assess the prevalence of refractive errors and vision impairment in school-age children in Shunyi District, northeast of Beijing, the Peoples Republic of China. METHODS: Random selection of village-based clusters was used to identify a sample of children 5 to 15 years of age. Resident registration books were used to enumerate eligible children in the selected villages and identify their
See also pp. 421 426, 436 444, 445 454, and 525527.

current school. Ophthalmic examinations were conducted in 132 schools on children from 29 clusters during May 1988 to July 1998, including visual acuity measurements, cycloplegic retinoscopy, cycloplegic autorefraction, ocular motility evaluation, and examination of the external eye, anterior segment, media, and fundus. Independent replicate measurements of all children with reduced vision and a sample of those with normal vision were done for quality assurance monitoring in three schools. RESULTS: A total of 6,134 children from 4,338 households were enumerated, and 5,884 children (95.9%) were examined. The prevalence of uncorrected, presenting, and best visual acuity 0.5 (20/40) or worse in at least one eye was 12.8%, 10.9%, and 1.8%, respectively; 0.4% had best visual acuity 0.5 or worse in both eyes. Refractive error was the cause in 89.5% of the 1,236 eyes with reduced vision, amblyopia in 5%, other causes in 1.5%, with unexplained causes in the remaining 4%.
Accepted for publication Nov 5, 1999. From the Peking Union Medical College Hospital, Beijing, China (Drs Zhao, Pan, and Sui); Unidad de Epidemiologia Clinica, Universidad de La Frontera, Temuco, Chile (Dr Munoz); and National Eye Institute, National Institutes of Health, Bethesda, Maryland (Drs Sperduto and Ellwein). This study was supported by the World Health Organization under National Institutes of Health Contract N01-EY-2103, Geneva, Switzerland. Preliminary data from this study were presented at the 17th Congress of the Asia Pacic Academy of Ophthalmology in Manila, the Phillipines, March 7 to 12, 1999. Reprint requests to Leon B. Ellwein, PhD, National Eye Institute, 31 Center Dr, Bethesda, MD 20892-2510; fax: (301) 496-9970.
0002-9394/00/$20.00 PII S0002-9394(99)00452-3

Myopia 0.5 diopter or less in either eye was essentially absent in 5-year-old children, but increased to 36.7% in males and 55.0% in females by age 15. Over this same age range, hyperopia 2 diopters or greater decreased from 8.8% in males and 19.6% in females to less than 2% in both. Females had a signicantly higher risk of both myopia and hyperopia. CONCLUSIONS: Reduced vision because of myopia is an important public health problem in school-age children in Shunyi District. More than 9% of children could benet from prescription glasses. Further studies are needed to determine whether the upward trend in the prevalence of myopia continues far beyond age 15 and whether the development of myopia is changing for more recent birth cohorts. (Am J Ophthalmol 2000;129: 427 435. 2000 by Elsevier Science Inc. All rights reserved.)

HE PRIMARY PURPOSE OF THE MULTICOUNTRY RE-

fractive Error Study in Children was to assess the prevalence of refractive error and vision impairment in school-age children in different ethnic and cultural settings. Population-based surveys were conducted in Chile, China, and Nepal using a common protocol.13 The study cohort in China was drawn from the rural population of Shunyi District, referred to as Shunyi County before August 1998. Shunyi District is a farming community located northeast of Beijing City (Peking). The topography of the area is mostly plains with a small remote hilly region. Compared with China as a whole, the socioeconomic status of the district ranks at an uppermiddle level. The population of Shunyi County in 1995 was 540,123 with approximately 21.7% of residents 5 to 15 years of age. Approximately 92% of the population resides in 27 rural townships, from which the study sample was drawn; there is one main town (urban area), which was excluded. Townships are made up of villages with an average population of 1,150. A unit in the District Bureau of Education is responsible for the health care of all students, including eye care. There are few qualied optometrists and ophthalmologists in the district. Both the Bureau of Education and the Bureau of Public
RIGHTS RESERVED.

2000 BY

ELSEVIER SCIENCE INC. ALL

427

Health in Shunyi District gave their approval of the Refractive Error Study in Children survey and cooperated fully with its implementation. Human subject research approval for the study protocol was obtained from the World Health Organization Secretariat Committee on Research Involving Human Subjects. Earlier surveys have suggested that refractive error is an important problem in China.4 However, most of the surveys dealt with children attending schools or patients attending eye clinics and thus, were not necessarily representative of the general population. Furthermore, the methods used to assess refraction were of questionable reliability. In some instances refractive status was based only on visual acuity measurement; others used refractive examinations, but without cycloplegia; some assessed refractive status only in individuals with abnormal vision. The Refractive Error Study in Children was an attempt to address the lack of reliable data on the prevalence of refractive error.

METHODS
A RANDOM SAMPLE OF CHILDREN LIVING IN RURAL SHUNYI

County was obtained through cluster sampling of villages. Clusters were dened by grouping villages with an estimated population less than 500 and subdividing those with population greater than 1,500. This resulted in 440 clusters with an average population of 1,003. The most recent published census data at the village level was from 1989; thus, these data were used in creating the sampling frame. Twenty-nine clusters were selected randomly with equal probability. The selected clusters ranged from 645 to 1,356 in estimated population, with an aggregate population of 29,488, including an estimated 6,450 children 5 to 15 years of age. Before the enumeration of children in the selected clusters, an announcement describing the Refractive Error Study in Children was sent to all townships and schools by the county government. For the 39 villages and village segments comprising the 29 selected clusters, the ofcial resident register book of each village was used to enumerate all households and children age 5 to 15 years, including identication of the school they were attending. Five to 6 year olds are in kindergarten, 7 to 13 year olds attend preliminary schools, and 14 to 15 year olds attend juniormiddle schools. Once all children in the selected clusters were enumerated, a list of children in the study sample was prepared for each school. The District Bureau of Education held a meeting with the heads of each school to introduce the Refractive Error Study in Children survey. Visual acuity assessment was practiced on kindergarten children before the survey. The entire enumeration process was completed in 10 days by two enumerators, who then accompanied two examination teams to 132 schools to assist with the 428 AMERICAN JOURNAL
OF

identication of the study subjects. Informed consent for the examination was obtained from the childs parent through a written notice sent to each home by the school head. Visual acuity measurements were performed by one of the two ophthalmic assistants on each team. Binocular motor function evaluation, anterior segment examination, cycloplegic autorefraction, and media and fundus examinations were done by the team ophthalmologist. Pupillary dilation and cycloplegia were performed by the second ophthalmic assistant, and cycloplegic retinoscopy and subjective refraction by the team optometrist. Field operations took place between May 1998 and July 1998. Prepilot training and a pilot exercise took place during the preceding 2 weeks. The pilot study involved the examination in two schools of 314 children, 5 to 13 years of age, who were not part of the study cohort. Quality assurance was monitored using replicate assessments in three preliminary and junior-middle schools with children from 11 of the 29 study clusters. Children with uncorrected visual acuity 0.625 (20/32) or worse in either eye and 10% of those with normal vision were selected for repeat testing. The 440 eyes of the 220 quality assurance subjects had the following visual acuity distribution: 17.5% had uncorrected visual acuity 0.25 (20/80) or worse, 25.5% were 0.32 (20/63) to 0.50 (20/40), and 57% were 0.625 or better; 36.1% were 1.00 (20/20). Half the subjects were 10 to 12 years of age, and 45.5% were male. Reproducibility for uncorrected visual acuity with two independent examiners was good: unweighted kappa statistics of 0.82 for right eye (RE) measurements and 0.81 for the left eye (LE) were obtained. Sixty-seven of the 440 measurements differed by at least 1 line; two of the right eye measurements and two of the left eye measurements differed by 2 or more lines. Mean differences for repeat cycloplegic retinoscopy were not signicant: 0.007 diopter (SD, 0.277) for RE measurements and 0.030 diopter (SD, 0.399) for LE measurements. The 95% limits of agreement were 0.551 to 0.536 diopter for RE measurements and 0.812 to 0.752 diopter for LE measurements. Repeat cycloplegic autorefraction readings had a mean RE difference of 0.010 (SD, 0.262) and an LE difference of 0.011 (SD, 0.273), also not statistically signicant. The 95% limits of agreement were 0.524 to 0.504 diopter and 0.518 to 0.554 diopter for RE and LE measurements, respectively.

RESULTS
WITHIN THE 29 CLUSTERS, 4,338 HOUSEHOLDS WITH CHIL-

dren were identied: 2,701 (62.3%) with one child, 1,493 (34.4%) with two, 132 (3.0%) with three, and 12 (0.3%) with four or more children. A total of 6,134 children were enumerated (Table 1). The skewed age distribution of enumerated children reects the one-child policy that was OPHTHALMOLOGY APRIL 2000

TABLE 1. Enumerated and Examined Population by Age and Sex

No. (%) of Males Age Enumerated Examined % Exam Enumerated No. (%) of Females Examined % Exam Numerated No. (%) All Examined % Exam

5 6 7 8 9 10 11 12 13 14 15 All

66 (2.1) 90 (2.8) 187 (5.9) 314 (9.9) 467 (14.8) 447 (14.1) 420 (13.3) 351 (11.1) 303 (9.6) 268 (8.5) 250 (7.9) 3,163 (100.0)

57 (1.9) 85 (2.8) 177 (5.9) 300 (10.0) 461 (15.3) 436 (14.5) 413 (13.7) 347 (11.5) 288 (9.6) 244 (8.1) 199 (6.6) 3,007 (100.0)

86.4 94.4 94.6 95.5 98.7 97.5 98.3 98.9 95.1 91.0 79.6 95.1

48 (1.6) 118 (4.0) 188 (6.3) 305 (10.3) 399 (13.4) 396 (13.3) 402 (13.5) 359 (12.2) 249 (8.4) 259 (8.7) 248 (8.3) 2,971 (100.0)

46 (1.6) 114 (4.0) 181 (6.3) 295 (10.3) 394 (13.7) 388 (13.5) 401 (13.9) 357 (12.4) 239 (8.3) 242 (8.4) 220 (7.7) 2,877 (100.0)

95.8 96.6 96.3 96.7 98.8 98.0 99.7 99.4 96.0 93.4 88.7 96.8

114 (1.9) 208 (3.4) 375 (6.1) 619 (10.1) 866 (14.1) 843 (13.7) 822 (13.4) 710 (11.6) 552 (9.0) 527 (8.6) 498 (8.1) 6,134 (100.0)

103 (1.8) 199 (3.4) 358 (6.1) 595 (10.1) 855 (14.5) 824 (14.0) 814 (13.8) 704 (12.0) 527 (9.0) 486 (8.3) 419 (7.1) 5,884 (100.0)

90.4 95.7 95.5 96.1 98.7 97.8 99.0 99.2 95.5 92.2 84.1 95.9

TABLE 2. Distribution (Percent With 95% Condence Intervals) of Uncorrected, Presenting, and Best Visual Acuity
No. (%) With Uncorrected Visual Acuity No. (%) Wearing Glasses* No. (%) With Presenting Visual Acuity No. (%) With Best Visual Acuity

Visual Acuity Category

0.625 or better Both eyes 0.625 or better One eye 0.32 or better to 0.50 or worse Better eye 0.125 or better to 0.25 or worse Better eye 0.10 or worse Better eye

5,128 (87.2) 272 (4.62) (4.045.21) 291 (4.95) (4.095.80) 180 (3.06) (2.363.76) 11 (0.19) (0.090.33) 5,882 (100.0)

4 (0.08) 16 (5.88) 52 (17.9) 80 (44.4) 7 (63.6) 159 (2.7)

5,239 (89.1) 283 (4.81) (4.135.49) 255 (4.34) (3.585.09) 101 (1.72) (1.142.29) 4 (0.07) (0.020.17) 5,882 (100.0)

5,779 (98.3) 78 (1.33) (1.001.65) 20 (0.34) (0.140.54) 4 (0.07) (0.020.17) 1 (0.02) (0.000.09) 5,882 (100.0)

*Refers to the number (%) of those within each uncorrected visual acuity category wearing glasses. Condence intervals calculated using the exact binomial distribution instead of the normal approximation. Cluster design effects ranged from 1.101 to 2.762 for the condence intervals estimated with the normal approximation. Design effects, which ranged from 0.812 to 1.020, are not reected in the exact binomial estimates.

introduced in the mid-1980s: the number of births initially increased but then dropped off sharply once the policy was more fully implemented. Except for 5 and 15 year olds, the percent examined (response rate) was above 91% for each year of age in males and, except for 15 year olds, above 93% in females. Some 15 year olds, males in particular, were not available for the examination because they had already left school for work outside of the area. Response within each of the 29 clusters exceeded 94%. Because age-specic and sexspecic response rates were generally uniform, the age and sex distribution of those examined were not signicantly different from those of the enumerated population (chisquare goodness of t, P .28). The 5,884 examined children came from 4,126 families: 2,519 (61.1%) from single-child households, 1,469 (35.6%) from two-child VOL. 129, NO. 4

households, 128 (3.1%) from three-child households, and 10 (0.2%) from households with four or more children. The results from visual acuity measurements are shown in Table 2. Visual acuity was not available for two children. As shown, 754 (12.8%) of children had visual acuity 0.50 or worse in one or both eyes without correction. Differences in the distribution of uncorrected visual acuity between males and females was statistically signicant (Kolmogorov-Smirnov test, P .001), with females having more impairment (data not shown). The sexspecic distributions did not differ, however, when children with visual acuity 0.625 or better in both eyes were excluded, that is, when only those with reduced vision in at least one eye were considered. Of the 482 children with uncorrected visual acuity 0.50 or worse in both eyes, only 139 (28.8%) were wearing
IN

CHILDHOOD REFRACTIVE ERROR

CHINA

429

glasses. Another 63 children indicated that they had glasses but were not wearing them on the day of the examination. Presenting visual acuity was 0.50 or worse in at least one eye in 643 children (10.9%) and in both eyes in 360 (6.12%). For best visual acuity, the numbers decrease to 103 (1.75%) and 25 (0.43%), respectively. Accordingly, another 540 children (9.18%) would achieve visual acuity 0.625 or better in both eyes with prescription glasses. Of the 360 with reduced presenting visual acuity in both eyes, 335 (93.1%) would achieve visual acuity 0.625 or better in at least one eye with glasses. Of the 11 children who were bilaterally blind with uncorrected visual acuity 0.10 or worse, seven were wearing glasses that improved vision, leaving four who presented blind. Except for one child with bilateral retinal disease, the remaining three improved somewhat with correction of refractive error. Thirty-seven presented unilaterally blind, which decreased to 21 with best visual acuity. Pupillary dilation was successful (6 mm or greater and absent papillary reaction to light) in all but two children ages 9 and 12: one child had uncorrected visual acuity of 0.25 in one eye and worse than 0.10 in the fellow eye; the other child had right eye and left eye acuities of .40 and .80, respectively. These two cases are excluded from the subsequent refractive error analyses. As shown in Figure 1, the mean retinoscopic refractive error changed from a positive value to a negative one with increasing age, from 1.06 diopters in the 5-year-old male cohort to 0.23 diopter in the 15-year-old cohort and from 1.35 to 0.84 in the corresponding female cohorts. The same age-associated and sex-associated trend held for refractive error in left eyes (not shown). Although 5-yearold females were slightly more hyperopic than males in the 5-, 6-, and 7-year-old cohorts, from 8 years onward they were less hyperopic, with a negative mean refractive error starting with the 13-year-old cohort versus 14 years in males. As evidenced by the relatively large standard deviations, refractive error varied considerably within each age cohort. Figure 2 presents the distribution of refractive error in right eyes of children from multiyear age intervals for males and females combined. With advancing age, the distribution of refractive error shifted toward less positive values. The percentage of eyes with hyperopia 2 diopters or greater decreased from 8.5% in the 5-year to 7-year age group to 1.1% in the 14-year to 15-year age group, whereas eyes with myopia 0.5 diopter or less increased from 1.2% to 38.8%. For all ages, the prevalence of RE hyperopia was 2.6% and RE myopia was 14.9%. The relationship between retinoscopic ametropia (in either eye) and age in male and female children is shown in Figure 3. For children age 5 years, the prevalence of hyperopia was 8.8% (95% condence interval [CI], 2.37% to 15.2%) in males and 19.6% (95% CI, 8.10% to 31.0%) in females. The cluster design effects were 0.759 and 0.999, 430 AMERICAN JOURNAL
OF

FIGURE 1. Mean RE spherical equivalent refractive error by age in males and females. Standard deviations are represented by the vertical bars.

respectively. Hyperopia decreased to less than 2% in both male and female 15-year-old cohorts. Myopia was essentially absent among 5 year olds, but increased to 36.7% (95% CI, 29.9% to 43.4%) in males and 55.0% (95% CI, 49.4% to 60.6%) in females in the 15-year-old cohorts. Design effects were 1.015 and 0.720, respectively. For the study population as a whole, children of all ages and both sexes, the prevalence of hyperopia was 3.5% (204 of 5,882) and myopia was 16.2% (955 of 5,882). With autorefraction the prevalences were 2.7% (159 of 5,882) and 21.6% (1,271 of 5,882), respectively. In a multivariate logistic regression model to assess the association of age and gender with hyperopia prevalence, age had an odds ratio of 0.75 (95% CI, 0.71 to 0.79), reecting the decreased risk of hyperopia with advancing age. The odds ratio for gender, 1.51 (95% CI, 1.08 to 2.13), reecting a higher risk in females, was also statistically signicant. In a regression model for myopia, age had an odds ratio of 1.51 (95% CI, 1.46 to 1.57), indicating that the risk of myopia increased over 50% with each additional year of age. The odds ratio for gender was 1.79 (95% CI, 1.49 to 2.15), reecting the signicantly higher risk of myopia in females. Astigmatism 0.75 diopter or greater was present in OPHTHALMOLOGY APRIL 2000

FIGURE 2. Distribution of spherical equivalent refractive error in right eyes of children ages 5 to 7, 8 to 10, 11 to 13, and 14 to 15 years. Data points represent a half-diopter interval (for example, those associated with 2 on the x-axis represent greater than 1.75 diopters to less than or equal to 2.25 diopters). Those at the extreme ends represent less than or equal to 4.25 diopters and greater than 4.25 diopters.

nearly 15% of eyes as measured by cycloplegic retinoscopic refraction and in approximately 10% of eyes with cycloplegic autorefraction (Table 3). The prevalence of the more severe level of astigmatism (2 diopters or greater) was not lower with autorefraction. With retinoscopy, astigmatism was more common in the left eye than the right eye (chi-square, P .001). In multivariate logistic regression models, astigmatism as measured by autorefraction was associated with older age ( P .025) and female gender ( P .015) for both right and left eyes. With retinoscopy, astigmatism was associated with younger age ( P .015) but not gender. Although spherical equivalent refractive error measurements from retinoscopy and autorefraction were highly correlated (Pearson correlations of 0.95 for both right and left eyes), autorefraction measurements were consistently more negative. There was a mean difference of 0.222 diopter (SD, 0.412) for RE measurements and 0.198 (SD, 0.409) for LE measurements, both statistically significant differences (paired t test, P .001). The 95% limits of agreement were 1.029 to 0.584 diopter for RE VOL. 129, NO. 4

measurements and 1.000 to 0.603 diopter for LE measurements (Figure 4). The systematic difference between retinoscopy and autorefraction was present across both negative and positive retinoscopy measurements. Tropias, mostly exotropia 15 degrees or less, were found in 165 children (2.8%), generally at both 0.5 and 4.0 meters of xation. Two-hundred forty children (4.1%) had abnormalities of the external eye and anterior segment, involving 396 eyes: inammation of the conjunctiva was observed in 245 eyes of 130 children, eyelid abnormalities in 116 eyes of 87 children, and corneal, pupillary, or other anterior segment abnormalities in 61 eyes of 42 children. Twenty children (0.3%) had media or fundus abnormalities, affecting a total of 26 eyes: lens-related abnormalities were observed in six eyes of four children, vitreous opacities in three eyes of three children, and anatomic retinal or optic disk abnormalities in 22 eyes of 17 children. Reduced vision was overwhelmingly the result of refractive error rather than other functional or organic abnormalities. Among the 754 children with visual acuity 0.50 or worse in one or both eyes, 651 attained visual acuity 0.625 or better in
IN

CHILDHOOD REFRACTIVE ERROR

CHINA

431

FIGURE 3. Prevalence of ametropia in male and female individuals by age. Myopia is dened as 0.5 spherical equivalent diopter or less in either eye. Hyperopia is dened as 2 spherical equivalent diopters or greater in either eye, as long as neither eye is myopic. Four myopia cases had hyperopia in the fellow eye.

TABLE 3. Prevalence of Astigmatism in Eyes Examined by Cycloplegic Retinoscopic Refraction and Cycloplegic Autorefraction
Retinoscopy: No. (%) of Eyes Cylinder (diopters) Right Eye Left Eye Autorefraction: No. (%) of Eyes Right Eye Left Eye

0.75 0.75 to 2.0 2.0 Total

5,086 (86.5) 700 (11.9) 92 (1.6) 5,878 (100.0)

4,954 (84.2) 823 (14.0) 103 (1.8) 5,880 (100.0)

5,317 (90.5) 436 (7.4) 124 (2.1) 5,877 (100.0)

5,318 (90.5) 440 (7.5) 120 (2.0) 5,878 (100.0)

both eyes with corrective refraction (Table 2). Another 11 children had correctable refractive error in one eye but an uncorrectable cause in the fellow eye, for a total of 662 children with refractive error as the cause in one or both eyes (Table 4). Of the 54 children without an organic cause of impairment and who satised one or more of the explicit amblyopia criteria, 15 (28%) had a tropia, 36 (67%) anisometropia 2 spherical equivalent diopters or greater, and 20 (37%) hyperopia 6 spherical equivalent diopters or greater. 432 AMERICAN JOURNAL
OF

Eight of these 54 children had the cause of impairment in both eyes attributed to amblyopia. Other causes of vision loss were uncommon. Fifty eyes of 38 children had no explanation for RE reduced vision. The relationship between RE refractive error, using spherical equivalent retinoscopic measurements, and uncorrected distance visual acuity is shown in Table 5. Although eyes with reduced vision possibly caused by something other than refractive error were excluded to OPHTHALMOLOGY APRIL 2000

DISCUSSION
WITHIN THE STUDY POPULATION, 12.8% HAD UNCOR-

FIGURE 4. Agreement between spherical equivalent cycloplegic autorefraction and cycloplegic retinoscopy in right and left eyes. The average of the two measurements is represented by the x-axis, and the difference between them (autorefraction minus retinoscopy) by the y-axis. Ninety-ve percent of the data points fall within the horizontal lines above and below the central one, which represents the mean of the differences (0.222 diopter for RE measurements and 0.198 diopter for LE measurements). The box plots along the axes represent data frequencies. The box extends from the 25th percentile to the 75th percentile, the interquartile range, with the line in the middle representing the median. The lines emerging from the box extend to the upper and lower adjacent values, dened as the 75th percentile plus 1.5 interquartile range and the 25th percentile minus 1.5 interquartile range.

avoid the inclusion of irrelevant data in the analysis, the relationship remains imprecise. Eyes with refractive error had no reduction in vision, and vice versa. Spearman rank correlation between refractive error and visual acuity was 0.86 ( P .001) for the 843 myopic right eyes. (The correlation between myopia and visual acuity in left eyes was 0.85). All but 19 of the 235 myopic right eyes with uncorrected visual acuity 0.80 or better had refractive error 0.50 to 0.75 diopter. Among the 131 emmetropic eyes with reduced vision, 18 had negative refractive error measurements and 61 had positive measurements 0.625 to less than 2 diopters. Of the 118 hyperopic eyes 2 diopters or greater, 107 (90.7%) had uncorrected visual acuity 0.80 or better. VOL. 129, NO. 4

rected visual acuity 0.50 or worse in at least one eye. This percentage decreased to 1.8% with best visual acuity. Females were disproportionately affected. Few children were wearing glasses, leaving 9.2% of the population who could benet from correction of refractive error. With prescription glasses, only 103 children (1.8%) would remain with some degree of uncorrectable vision loss in at least one eye, including 25 with visual acuity 0.50 or worse in both eyes. This 0.4% (25 of 5,882) prevalence of best-corrected visual acuity 0.50 or worse in the better eye is consistent with that reported elsewhere.5 Myopia was the principal type of refractive error, with prevalences increasing steadily beginning with the 8-yearold male and female cohorts. The increase in females was more rapid, with a prevalence of 55.0% in 15 year olds vs 36.7% in males. Whether the higher prevalence in females persists or whether males eventually catch up with or even surpass females is unknown. Temporal differences in development between males and females may be partly responsible for these differences in the prevalence of myopia. There is no reason to suspect, however, that different visual experiences in daily life are responsible. As noted, the relationship between refractive error and visual acuity is not clear cut. One consequence of this is that the prevalence of refractive error inferred from correctable visual acuity decits can differ substantially from that obtained through direct measurement. For example, the prevalence of RE refractive error would be 13% (750 of 5,764) based on uncorrected visual acuity 0.625 or worse correctable to visual acuity 0.80 or better (Table 5). However, with spherical equivalent retinoscopic measurement, the prevalence of myopia 0.5 diopter or less was 14.6% and the prevalence of hyperopia 2 diopters or greater was 2%. Because uncorrected visual acuity was assessed before cycloplegia, accommodation might explain the normal measurements in many of the hyperopic children. The extent to which squinting improved the vision of myopic eyes is also unknown. Although children were to be observed during testing, some squinting may have gone undetected. Emmetropic eyes with reduced, but correctable, vision created an offsetting effect: they contributed to the visual acuityinferred prevalence but not to that measured directly. The discrepancy in prevalence results using these two methods could, of course, be reduced by drawing more severe refractive error thresholds for the denitions of myopia and hyperopia. There should be little concern that study ndings were biased because of sampling deciencies. Although children were examined in schools, they were selected directly from the rural Shunyi County population with a sampling plan that ensured that each child 5 to 15 years of age had an essentially equal chance of being included in the survey. Furthermore, because enumeration of the study population
IN

CHILDHOOD REFRACTIVE ERROR

CHINA

433

TABLE 4. Causes of Uncorrected Visual Acuity 0.50 or Worse

No. (%) of Eyes With Visual Acuity 0.5 or Worse Cause Right Eye Left Eye

No. (%) of Children With Visual Acuity 0.5 or Worse in One or Both Eyes*

Prevalence Percentage in Population (One or Both Eyes Affected)*

Refractive error Amblyopia Corneal opacity Cataract Retinal disorders Other causes Unexplained causes Any cause

560 (91.1) 23 (3.74) 0 (0.0) 0 (0.0) 2 (0.33) 5 (0.81) 25 (4.07) 615 (100.0)

546 (87.9) 39 (6.28) 1 (0.16) 1 (0.16) 4 (0.64) 5 (0.81) 25 (4.03) 621 (100.0)

662 (87.8) 54 (7.16) 1 (0.13) 1 (0.13) 5 (0.66) 8 (1.06) 37 (4.91) 754 (100.0)

11.3 0.92 0.02 0.02 0.09 0.14 0.63 12.8

*Children with visual acuity 0.5 or worse in both eyes may represent two different causes of reduced vision. Thus, the total for all causes exceeds the any-cause number. Refractive error was assigned as the cause of reduced vision for eyes correcting to 0.625 or better with subjective refraction. Includes only cases meeting explicit tropia, anisometropia, or bilateral hyperopia criteria.

TABLE 5. Refractive Error and Uncorrected Visual Acuity in Right Eyes With Best-corrected Visual Acuity 0.8 or Better
Spherical Equivalent Refractive Error: No. (%) of Eyes 0.5 0.5 to 2 2 to 4 4 All

Uncorrected Visual Acuity

0.8 0.5 to 0.625 0.32 to 0.4 0.25 All

235 (4.69) (27.9) 203 (61.3) (24.1) 171 (94.0) (20.3) 234 (98.7) (27.8) 843 (14.6) (100.0)

4672 (93.2) (97.3) 117 (35.3) (2.43) 11 (6.04) (0.23) 3 (1.27) (0.06) 4803 (83.3) (100.0)

100 (1.99) (90.9) 10 (3.02) (9.09) 0 (0.0) (0.0) 0 (0.0) (0.0) 110 (1.91) (100.0)

7 (0.14) (87.5) 1 (0.30) (12.5) 0 (0.0) (0.0) 0 (0.0) (0.0) 8 (0.14) (100.0)

5014 (100.0) (87.0) 331 (100.0) (5.74) 182 (100.0) (3.16) 237 (100.0) (4.11) 5764 (100.0) (100.0)

within each of the randomly selected village clusters was based on the current village register, the accounting for children in each village should have been accurate. The ofcial village register is updated by the village secretary on a periodic basis with adjustments for births, deaths, immigration, and emigration and, thus, was considered a reliable and complete source of census information. Because 1998 projections were not available for the Shunyi County population as a whole, it was not possible to assess the age-sex representativeness of the enumerated population, that is, whether the randomized selection process resulted in an accurate representation of the general population. Also, the nearly 96% examination response rate with essentially no missing data should have minimized opportunities for bias in the testing and examination phases of the study. Visual acuity and examination methods were shown to have produced reliable results. Test/retest reproducibility was good for visual acuity testing with kappa statistics greater than 0.80. For the 67 of 440 measurements that 434 AMERICAN JOURNAL
OF

differed by at least 1 line, the second measurement was somewhat more likely to be better than the rst one, evidence of a possible learning effect among respondents. Reproducibility for cycloplegic autorefraction was within one-half diopter for 95% of both RE and LE measurements, with no tendency for the rst measurement to be consistently greater or less than the second one. Reproducibility for cycloplegic retinoscopy was similar for RE measurements but increased to three-fourths diopter for LE measurements, again with no systematic bias between the two measurements. The slightly poorer LE reproducibility may result from the examiner using his/her right eye for taking measurements in both of the subjects eyes. The alignment for the LE measurement, not being as favorable, could produce less accurate readings. This alignment bias was not seen with autorefraction. The agreement between retinoscopy and autorefraction was within three-fourths of a diopter for 95% of measurements. Autorefraction generally produced more negative measurements, however, approximately one-fourth diopter OPHTHALMOLOGY APRIL 2000

across the entire refractive error range for both right and left eyes. The tendency toward more negative measurements with autorefraction was not seen in the companion studies in Nepal and Chile.2,3 Refractive error in school-age children, and myopia in particular, was found to be common in Shunyi District. It was by far the major cause of reduced uncorrected vision. Although blindness is not a signicant problem in this age group, lesser degrees of impairment are. Few of the children with vision-reducing refractive error were wearing glasses; an additional 9.2% of all children examined could experience essentially normal vision in both eyes with prescription glasses. Obstacles to obtaining glasses were not studied, but vision problem awareness and cost may be implicated. Further study is also needed to determine whether the age-related upward trend in myopia prevalence continues far beyond age 15, and whether the natural history of myopia is changing for more recent birth cohorts.
ACKNOWLEDGMENT

and the Peking Union Medical College Hospital administration for their cooperation in facilitating the implementation of this survey. The authors also acknowledge the clerical assistance of Jeanne King, National Eye Institute, in the preparation of this manuscript.

REFERENCES
1. Negrel AD, Maul E, Pokharel GP, Zhao J, Ellwein LB. Refractive error study in children: sampling and measurement methods for a multi-county survey. Am J Ophthalmol 2000; 129:421 426. 2. Pokharel GP, Negrel AD, Munoz SR, Ellwein LB. Refractive error study in children: results from Mechi Zone, Nepal. Am J Ophthalmol 2000;129:436 444. 3. Maul E, Barroso S, Munoz SR, Sperduto RD, Ellwein LB. Refractive error study in children: results from La Florida, Chile. Am J Ophthalmol 2000;129:445 454. 4. Wu X. Ocular refraction and refractive error. In: Li Fengming, editor. System of ophthalmology, vol. 8 (Chinese). Beijing: Peoples Hygiene Press, 1996:2539 2558. 5. Gilbert CE, Anderson L, Dandona L, Foster A. Prevalence of visual impairment in children: a review of available data. Ophthalmol Epidemiol 1999;6:73 82.

Appreciation is expressed to the Shunyi District Bureau of Public Health, the Shunyi District Bureau of Education,

The full-text of AJO is now available online at www.ajo.com. Authors Interactive, currently available in limited form, is undergoing an upgrade.

VOL. 129, NO. 4

CHILDHOOD REFRACTIVE ERROR

IN

CHINA

435