Longitudinal analysis of 5-year refractive changes in a large Japanese population

Refractive changes are reportedly affected by age, sex, and current refractive error. To clarify the pattern of refractive changes in a Japanese population, we conducted a 5-year follow-up longitudinal analysis of spherical equivalent (SE) refractive changes with stratification by sex, age, and SE in 593,273 eyes from Japanese individuals ages 3–91 years. The 5-year SE change with myopic shift dramatically increased over time after age 4 years, and the largest change was observed in both males and females who were age 8 years at baseline [males: − 2.654 ± 0.048 diopters (D); females: − 3.110 ± 0.038 D]. During school age, the 5-year myopic change was greater in females than in males, and emmetropic and low-to-moderate myopic eyes underwent larger myopic changes than hyperopic and high-to-severe myopic eyes. After the peak at age 8 years, the 5-year myopic change gradually declined with age and fell below − 0.25 D at age 27 in males and age 26 years in females. The 5-year SE changes transitioned from a myopic to a hyperopic shift at age 51 in both sexes, and hyperopization advanced more quickly in hyperopic eyes. Our findings highlight the importance of myopia prevention in school-aged children.

www.nature.com/scientificreports/ D in hyperopic and − 0.55 D in myopic eyes, suggesting that children with myopia experience faster myopic progression than children with hyperopia 15 . In 124 Danish school children ages 9-12 years, Jensen followed up SE refractive errors for 8 years and found that the earlier myopia started, the more it progressed 23 . Xiang et al. conducted a 4-year follow-up analysis of refraction tests in 607 Chinese twins ages 7-15 years and reported that the annual SE refractive changes in younger children (ages 7-11 years) were larger than in older children (ages 12-15 years) 24 . In studies of SE refractive changes over 2 years in 2053 new medical students in China 25 , with a mean age of 18.27 ± 1.83 years, and in 166 clinical microscopy technicians in Britain, ages 21-55 years 26 , the higher the myopic level at study start, the faster the myopic shift. Few previous studies have investigated SE refractive changes in a Japanese population. In one study evaluating the effectiveness of progressive addition lenses and single vision lenses on myopia progression over 3 years in 86 Japanese myopic children ages 6-12 years, greater myopia progression was observed in the more myopic eyes and at younger ages 27 . These longitudinal studies have investigated refractive changes over time, but the sample sizes did not allow for sufficient statistical power to fully elucidate the pattern of changes. Moreover, because myopia progression at different life stages depends on age and current refractive error, refractive changes should be assessed in a large cohort finely stratified by age, sex, and refractive error, but no such reports have been published. We conducted the largest longitudinal study of SE refractive changes in a Japanese cohort to date, stratified by age, sex, and SE. The findings from this longitudinal study covering a wide range of ages will aid understanding of SE refractive changes throughout the lifetime for each baseline SE.

Results
Five-year refractive changes. We retrospectively followed 593,273 eyes from Japanese individuals ages 3-91 years for 5 years (Supplementary Table S1). Of these eyes, 228,074 (38.4%) were from males and 365,199 (61.6%) were from females. We investigated 5-year SE refractive changes across age in years and by sex ( Fig. 1 and Supplementary Table S2).
In females at age 3 years, the mean change (± standard error) in SE in the following 5 years was − 0.750 ± 0.395 D, although this myopic change did not reach significance because of the small sample size. For those age 4 years at baseline, the 5-year mean SE change in males was − 0.479 ± 0.162 D, whereas females showed no mean change in SE (0.000 ± 0.279 D). Afterward, the 5-year mean SE change with myopic shift dramatically increased with age, exceeding − 2.00 D at the age of 6 years and peaking at 8 years in both males and females. The largest 5-year SE mean change was observed in both males and females who were age 8 years at baseline (males: − 2.654 ± 0.048 D; females: − 3.110 ± 0.038 D). After the peak at age 8 years at baseline, the 5-year mean SE change began to decrease, and 5-year mean SE changes exceeding − 2.00 D up to age 11 years and − 1.00 D up to age 14 years were observed in both males and females.
After the late teens, the rate of the 5-year mean SE change with myopic shift decreased slowly with increasing age, falling below − 0.50 D at age 19 years in males and age 18 years in females. It declined to − 0.25 D, the minimum value for correction with eyeglasses, at age 27 years in males and age 26 years in females, and to − 0.10 D at age 34 years in males and age 32 years in females. At age 51 years, the 5-year mean SE change transitioned 5-year mean change in spherical equivalent (SE) refractive error for each age at baseline. The x-axis represents ages at baseline, in years. The y-axis represents the magnitude of 5-year mean SE refractive changes at each baseline age in males (black circles) and females (white circles). www.nature.com/scientificreports/ from a negative to a positive value in both males and females, indicating a shift from myopia to hyperopia. After that, the 5-year mean SE change with a hyperopic shift tended to slowly increase with age up to around the late sixties in both males and females. Among those ages 51 to 71 years, the largest 5-year SE mean change with a hyperopic shift was observed at 64 years in males (+ 0.394 ± 0.040 D) and 67 years in females (+ 0.202 ± 0.039 D). From the age of 72 years, the direction of the 5-year mean SE change occasionally showed a myopic shift.
The statistical analysis showed that SE significantly changed from baseline age to the 5-year follow-up in males ages 4-27 years and females ages 5-27 years (P < 0.05 after Bonferroni correction) (Supplementary Table S2). When comparing the rate of the 5-year mean SE change between males and females, myopic shift was greater in females at ages 5 to 14 years (except for age 11 years), which was associated with a large 5-year mean SE change, especially in females compared with males from ages 5 to 8 years, with significant sex differences in 5-year myopic change at ages 7 and 8 years. In contrast, for those ages 15 to 45 years, the 5-year myopic change was slightly greater in males than in females, and sex differences in the 5-year myopic change for ages 15 to 39 and at age 41 were significant. At ages 54 to 79 years, the 5-year hyperopic change was greater in males than in females, except at ages 67, 73, 75, and 76 years, and sex differences in the 5-year hyperopic change were significant at most ages (Supplementary Table S2).
Five-year refractive changes stratified by age group and refractive error level. In contrast, the smallest myopic change in males was observed at SE − 6.75 to − 8.50 D (− 1.000 ± 1.189 D), followed by + 2.50 to + 0.75 D and + 2.75 D or more, and at + 2.50 to + 0.75 D (− 1.169 ± 0.234 D) in females, followed by − 6.75 to − 8.50 D and + 2.75 D or more. Of these, myopic changes in SE groups from " + 2.75 D or more" to "− 4.75 to − 6.50 D" were significant in both males and females.
In the 10-14-year age group, the largest 5-year myopic change in males was observed at SE + 0.

Discussion
To investigate patterns of SE refractive changes in a Japanese population, we conducted this large-scale population-based study and analysed longitudinal data for 593,273 eyes from individuals whose SE refractive errors were followed up for 5 years. Because age and current refractive error affect myopia progression at different life stages, we stratified the data by age and refractive error.
The results show that school-aged children underwent a large myopic change. The 5-year mean SE change exceeded − 2.00 D at the baseline age range of 6 to 11 years in both males and females. The largest 5-year change was observed in those who were age 8 years at baseline (males: − 2.654 ± 0.048 D; females: − 3.110 ± 0.038 D). In addition, stratification by age groups showed that those ages 3-9 years had the largest 5-year myopic change (males: − 2.460 ± 0.059 D; females: − 2.872 ± 0.056 D), followed by those ages 10-14 years (males: − 1.478 ± 0.010 D; females: − 1.491 ± 0.008 D). A previous study with Chinese children ages 6-15 years identified 5-year mean SE refractive changes of − 2.21 ± 1.87 D and showed that myopic progression was associated with younger age 28 . Another study in India of people ages 1-30 years showed an age effect on myopia progression and identified the largest annual change in SE in children ages 6-10 years, followed by those ages 0-5 and 11-15 years 29 . In a longitudinal multi-ethnic study in children ages 7 to 13 years, the peak progression in the first year of myopia onset www.nature.com/scientificreports/ was − 0.58 D at age 7 years, with decreasing annual SE refractive change as the myopia onset age increased 30 . Thus, our findings in school-aged children of the Japanese population are similar to previous results in other countries. We also found that myopic progression in Japanese children ages 5 to 14 years was faster in females than in males, with much larger sex-based differences in those who were ages 5 to 8 years at baseline. We observed 5-year mean SE changes exceeding − 3.00 D only in girls ages 6, 7, and 8 years but not in boys. Previous studies also have shown a significantly faster myopic progression in school-aged females compared to males in China, Singapore, and the United States 28,30-33 , but studies from Australia and the UK have shown only slightly faster progression in female than in male school-aged children 34,35 . Based on the current and previous findings, we confirmed that females have a faster myopic shift than males during school age. This faster myopic shift could result from average lifestyle differences between males and females, with girls spending more time on close-up visual work such as reading and doing homework [36][37][38] , which are thought to be risk factors for myopia 39,40 , and boys spending more time on outdoor activities, which are thought to be protective factors against myopia 40,41 . In addition, differences between males and females in sex hormone levels and physical structural changes in height and weight may also underlie sex differences in SE refractive changes 42 .
Our stratification analysis of data for the Japanese school-aged children by baseline SE levels revealed that the emmetropic, low myopic, mild myopic, and moderate myopic groups with a baseline SE ranging from + 0.50 to − 6.50 D had a larger 5-year myopic change than the hyperopic, high myopic, and severe myopic groups with a baseline SE of + 0.75 D or more or − 6.75 D or less. Many previous studies have reported a relationship between myopia progression and baseline SE in school-aged children (Supplementary Table S3 31 . Based on these findings, baseline SE certainly affects myopia progression in school-aged children, but the reported degree of this effect is not necessarily the same among the current and previous studies. One reason for these divergences may be differences in age ranges and genetic, environmental, and lifestyle backgrounds among the cohorts. Considering the importance of myopia prevention in school-aged children, further detailed study is required to clarify the relationship between baseline SE and myopia progression. In the current study, the 5-year myopic change peaked in childhood and decreased with age regardless of baseline SE values. Because the minimum value for eyeglass correction is generally 0.25 D, a 5-year SE change falling below − 0.25 D was defined as cessation of clinical myopic changes. We found that the 5-year myopic change was slightly larger in males than in females at ages 15 to 45 years and that clinical myopic changes stopped at age 27 years in males (− 0.217 ± 0.006 D) and age 26 years in females (− 0.220 ± 0.005 D). These findings indicate that clinical myopic changes cease earlier in females than in males in a Japanese population. When we stratified the participants by SE levels, however, the 5-year mean SE change fell below − 0.25 D at all SE levels after ages 30-34 years in males and 35-39 years in females, demonstrating that baseline SE level affects myopic change even in adults and suggesting a difference by baseline SE in the age at which myopia progression stops.
The 5-year mean SE changes transitioned from a myopic to a hyperopic shift at age 51 years in both males and females in our Japanese population. Among the SE groups, hyperopization for both males and females in the + 2.50 to + 0.75 D and + 2.75 D or more groups was faster than in other SE groups, and the 5-year SE changes in these groups became hyperopic at ages 30-39 years. These findings indicate that hyperopization progresses with age and advances more quickly in hyperopic eyes. These hyperopization trends can be attributed to latent hyperopia in hyperopic eyes that gradually becomes manifest over time because of an age-related decrease in accommodation 43 . The 5-year SE change remained in hyperopic shift to age 71 years, but at or after age 72 years, the changes varied with age, and some eyes showed a slight myopic shift in our Japanese population. The hyperopic shift after middle age and the myopic shift in old age have been reported previously in several ethnic groups (Supplemental Table S3). A 5-year follow-up study of SE changes in an elderly Chinese population showed 5-year mean SE changes of − 0.21 D, + 0.14 D, + 0.40 D, + 0.08 D, and − 0.34 D for those ages 30-39, 40-49, 50-59, 60-69, and 70 years or older, respectively 44 . A 10-year follow-up study of U.S. adults showed that younger people tended to shift toward being hyperopic, whereas older people shifted toward being myopic (mean SE change: + 0.48 D, + 0.03 D, and − 0.19 D for those ages 43-59, 60-69, and 70 years or older, respectively) 45 . In addition, a 9-year follow-up study of adults of African ancestry revealed a hyperopic shift in those ages 40-49 years at baseline (mean SE change: + 0.47 D) and a myopic shift in those ages 50 years or older (mean SE change: − 0.14 D, − 0.90 D, and − 0.78 D for ages 50-59, 60-69, and 70 years or older, respectively) 46 . Regarding the mechanism of SE changes after middle age, an aging-related decrease in lens refractive power has been suggested as a factor in hyperopia after middle age 44 . In contrast, the myopic shift in older age is thought to be caused by progression of nuclear cataracts and changes in ocular structure [47][48][49] . In the present study, SE changes after age 72 years varied with individual ages. Considering the effects of cataracts and changes in ocular structure on SE [47][48][49] , this difference in tendency to SE change at older ages may have resulted from individual differences in ocular conditions in our elderly population, and not solely because of small numbers in the older age group. www.nature.com/scientificreports/ A potential weakness of the current study is that the participants were all ophthalmic patients who visited one eye clinic in Yokohama. Our findings may not be generalizable to the Japanese population as a whole because the distribution of age or illness in our patients may not reflect those of the general population. However, compared with findings from a recent epidemiological study of myopia prevalence in elementary and junior high school students in Tokyo 50 , the distribution of SE was similar among schoolchildren in the current study, indicating that our study cohort could reflect the general population in urban Japan.
Another weakness is that cycloplegics were not used in all participants but only for children whom we suspected to have refractive hyperopia or amblyopia. Cycloplegic autorefraction is important for obtaining accurate SE in all children to avoid the influence of active accommodation. This weakness is the most significant limitation of our study. Furthermore, we did not consider as covariates in these analyses a family history of refractive errors or environmental, lifestyle, or genetic factors, which all carry possible risk for refractive changes. In addition, the current study did not include ocular biometric parameters such as axial length, corneal curvature radius, and anterior chamber depth for the evaluation of SE change, which are reportedly related to the progression of myopia or hyperopia 51,52 . Although the large sample size (more than 500,000 eyes) is likely to reduce the effect of these limitations on public health statistical analyses, to overcome these limitations and understand refractive changes in detail, a population-based multicentre longitudinal study is needed that includes consideration of various risk and correlation factors for refractive changes in individuals universally receiving cycloplegics.
In conclusion, in this study in the Japanese population, the largest myopic change occurred at age 8 years in males and females, with a steeper change in females. Clinical myopization ceased at age 27 years in males and age 26 years in females. SE refractive changes transitioned from a myopic to a hyperopic shift at age 51 years in both males and females. Our findings can provide useful information for prediction of future SE change, and we have constructed an algorithm for predicting SE refractive changes after 5 years using baseline SE, baseline age, and sex as variables based on the data obtained from 593,273 eyes in the current study (manuscript in preparation). The prediction of future SE changes can guide uptake of preventive behaviours to prevent myopia progression, such as less time spent on work activities, especially in school-aged children. We expect our data and algorithm to contribute to prevention of myopia worsening.

Methods
Participants. We retrospectively investigated 5-year SE refractive changes in 593,273 eyes of patients ages 3-91 years who visited Okada Eye Clinic (Yokohama, Japan) for refractive error or suspected refractive error between January 2000 and December 2012. All participants were of Japanese ethnicity and living in Yokohama and its vicinity. Those with corneal diseases, a history of eye surgery, or with SE changes greater than ± 5.0 D within a year were excluded because of possible interference with refraction. Those with a history of any treatment with low-dose atropine eye drops, orthokeratology lens, or bifocal or progressive addition lens for prevention of myopia progression were also excluded. In terms of age stratification, patients were divided into groups at an interval of 5 years, except for a group of patients ≤ 9 years and those age ≥ 60 years (Supplementary Table S1). An Auto Refractometer (KR-3000; TOPCON, Tokyo, Japan) was used to measure SE refractive errors. Patients were divided into eight SE groups: + 2.75 D or more, + 2.  Table S1). This study followed the tenets of the Declaration of Helsinki and was approved by the Ethics Committee of Okada Eye Clinic Institute. The Ethics Committee waived the need for informed consent due to the retrospective, observational, and anonymous nature of this study, according to the Ethical Guideline for Clinical Research issued by the Ministry of Health, Labour and Welfare, Japan. Statistical analysis. We investigated changes in SE refraction from baseline age to 5-year follow-up in each patient and analysed the 5-year mean SE change in each category classified by sex, baseline age, and/or baseline SE level using analysis of covariance (ANCOVA) 53 . When 5-year mean SE changes in each baseline age were analysed, the adjusted mean levels of 5-year SE changes were calculated from ANCOVA, with baseline SE as an independent variable. Similarly, when we analysed 5-year mean SE changes stratified by baseline age groups and baseline SE levels, we calculated the adjusted mean levels of 5-year SE changes using ANCOVA with baseline age and baseline SE as independent variables. P values to assess the significance of 5-year mean SE changes were calculated using ANCOVA, with Bonferroni correction for multiple testing and a corrected P < 0.05 was considered significant. The comparison of 5-year mean SE change between males and females was done using Welch's t test. We excluded the categories with fewer than 10 participants from the statistical analysis. Age-standardized prevalences in SE groups were calculated using the direct method 54 , based on a 1985 Japanese model population 55 . All analyses were conducted using the SAS statistical software package (SAS Institute Inc., Cary, NC, USA).