Risk Factors for Incident Myopia in

Australian Schoolchildren
The Sydney Adolescent Vascular and Eye Study
Amanda N. French, BAppSci (Orth) Hons,1 Ian G. Morgan, PhD,2 Paul Mitchell, MD, PhD,3
Kathryn A. Rose, PhD1

Purpose: To examine the risk factors for incident myopia in Australian schoolchildren.
Design: Population-based, longitudinal cohort study.
Participants: The Sydney Adolescent Vascular and Eye Study (SAVES) was a 5- to 6-year follow-up of the
Sydney Myopia Study (SMS). At follow-up, 2103 children were reexamined: 892 (50.5%) from the younger cohort
and 1211 (51.5%) from the older cohort. Of these, 863 in the younger cohort and 1196 in the older cohort had
complete refraction data.
Methods: Cycloplegic autorefraction (cyclopentolate 1%; Canon RK-F1; Canon, Tokyo, Japan) was
measured at baseline and follow-up. Myopia was defined as a spherical equivalent refraction of
0.50 diopters
(D). Children were classified as having incident myopia if they were nonmyopic at baseline and myopic in either
eye at follow-up. A comprehensive questionnaire determined the amount of time children spent outdoors and
doing near work per week at baseline, as well as ethnicity, parental myopia, and socioeconomic status.
Main Outcome Measures: Incident myopia.
Results: Children who became myopic spent less time outdoors compared with children who remained
nonmyopic (younger cohort, 16.3 vs. 21.0 hours, respectively, P < 0.0001; older cohort, 17.2 vs. 19.6 hours,
respectively, P¼0.001). Children who became myopic performed significantly more near work (19.4 vs. 17.6 hours;
P¼0.02) in the younger cohort, but not in the older cohort (P¼0.06). Children with 1 or 2 parents who were myopic had
greater odds of incident myopia (1 parent: odds ratio [OR], 3.2, 95% confidence interval [CI], 1.9e5.2; both parents:
OR, 3.3, 95% CI, 1.6e6.8) in the younger but not the older cohort. Children of East Asian ethnicity had a higher
incidence of myopia compared with children of European Caucasian ethnicity (both P < 0.0001) and spent less time
outdoors (both P < 0.0001). A less hyperopic refraction at baseline was the most significant predictor of incident
myopia. The addition of time outdoors, near work, parental myopia, and ethnicity to the model significantly improved
the predictive power (P < 0.0001) in the younger cohort but had little effect in the older cohort.
Conclusions: Time spent outdoors was negatively associated with incident myopia in both age cohorts. Near
work and parental myopia were additional significant risk factors for myopia only in the younger cohort.
Financial Disclosure(s): The author(s) have no proprietary or commercial interest in any materials discussed
in this article. Ophthalmology 2013;120:2100-2108 ª 2013 by the American Academy of Ophthalmology.

The prevalence of myopia in school-aged children varies
significantly between populations living in different
locations,1e12 with particularly high prevalence rates in
urban areas of East Asia.1e4,6,7 This variation is not
accounted for by ethnicity, with large differences in preva-
lence also noted for children of the same ethnic background
living in different locations5,13,14 and high prevalences re-
ported for different ethnic groups living in the same loca-
tion, particularly in urban East Asia.5,15 Although there is
some evidence for a genetic contribution to myopia,
including increased risk of myopia in children whose
parents are myopic8,16,17 and high heritability estimates
from twin studies,18 the search for “myopia genes” has
yielded only limited positive results.19 In addition, large
and rapid increases in prevalence have been documented
in a number of populations over the past few decades,20,21
which is difficult to explain through genetic change,
implying that environmental risk factors play a major role in
the development of myopia.20,21
A number of environmental risk factors have been impli-
cated. The association with higher educational attainment has
been strong and consistent.15,22e24 Near work has been
proposed as a possible mediator of this association through
increased accommodative demand,8,25e27 but the association
between near work and myopia has been weak and incon-
sistent,8,25e27 with longitudinal data indicating that children
who become myopic do not perform significantly more near
work than those who do not develop myopia.27 There also have
been reports of an association between less outdoor sport and
physical activities and myopia, but these 2 risk factors were
not analyzed separately.17,27e29 In 2008, Rose et al30
reported on children in the Sydney Myopia Study (SMS) and
noted a protective association of time spent outdoors,
including both leisure and sport, on myopia prevalence, but

2100  2013 by the American Academy of Ophthalmology ISSN 0161-6420/13/$ - see front matter
Published by Elsevier Inc. http://dx.doi.org/10.1016/j.ophtha.2013.02.035
 French et al 
Risk Factors for Incident Myopia
not for indoor sport. This indicated that the protective effect of
outdoor activity was related to being outdoors rather than to
physical activity per se. This finding has since been
confirmed by a number of other studies.31e33 A protective
effect for outdoor activity/sports also was established using
longitudinal data from the Collaborative Longitudinal Evalu-
ation of Ethnicity in Refractive Error study in the United
States.17,27 It has been suggested that this protective effect may
be mediated by the release of retinal dopamine in response to
the typically higher-intensity light outdoors,30 which has since
been shown in both chicks and primates to block the
development of form-deprivation myopia.34e36
The Sydney Adolescent Vascular and Eye Study (SAVES)
is a 5- to 6-year longitudinal follow-up of the SMS cohort. In
this report, we investigated whether risk factors previously
associated with prevalent myopia, including parental myopia,
near work, and time spent outdoors, also were associated with
incident myopia in a large population-based sample of
Australian schoolchildren.
Methods
Population
The SMS, a population-based cross-sectional study, was conducted
from 2003 to 2005. Children were examined in 2 age samples, 6
and 12 years, from 55 schools across the Sydney metropolitan area.
Detailed methodology for the SMS has been described previ-
ously.37 A 5- to 6-year longitudinal follow-up study of the SMS
cohort, the SAVES study, was conducted from 2009 to 2011. Of
the 34 original primary schools, 13 still had the original cohort
enrolled and were revisited to capture this group. Children from the
other 21 primary schools, who had moved to secondary schools,
were followed up at their secondary school or individually at an
eye clinic. Of the 21 secondary schools in the SMS, 20 were
revisited. One school was unable to be visited before the students
had completed their final year of schooling, but these students were
invited to participate individually. Of the original SMS sample,
2103 children were reexamined in SAVES: 892 (50.5%) from the
younger cohort and 1211 (51.5%) from the older cohort. The mean
time between baseline and follow-up examinations was 6.1 years
for the younger cohort and 4.6 years for the older cohort.
Demographics and refractive status were compared between
children who participated in both examinations (SMS and SAVES)
and children who were lost to follow-up. In the younger cohort,
children who were reexamined were more likely to be of European
Caucasian ethnicity (P ¼ 0.0003) and less likely to be from other
ethnic categories (P ¼ 0.0004) than those lost to follow-up.
Consequently, the younger follow-up cohort was marginally
more hyperopic at baseline (mean difference, 0.08 diopters [D])
than children lost to follow-up (P ¼ 0.04). There was no significant
differences in socioeconomic factors, including parental education
(P¼1.0) and parental employment (P ¼ 0.2), or in the frequency of
parental myopia (P ¼ 0.2). In the older cohort, children who were
followed were more often of East Asian ethnicity (P < 0.0001) and
less often of European Caucasian ethnicity (P ¼ 0.01) or other
ethnicities (P¼0.02). Correspondingly, children followed had
a slightly more myopic mean spherical equivalent refraction (SER)
at baseline (mean difference, 0.21 D; P < 0.0001). In addition,
children who were followed were more likely to be of higher
socioeconomic status (parental education: P < 0.0001; parental
employment: P¼0.03) and more likely to have parents who were
myopic (P¼0.002).
Procedures
Ethical approval for the study was provided by the Human
Research Ethics Committee, University of Sydney, the New South
Wales Department of Education and Training, and Catholic
Education Office. The study adhered to the tenets of the Declara-
tion of Helsinki. Informed, written consent was obtained from
parents or participants who were over the legal age of consent
(18 years) before participation. Verbal consent was obtained from
all participants before examination.
All children completed a comprehensive eye examination at
both baseline (SMS) and follow-up (SAVES) that included
cycloplegic autorefraction using the Canon RK-F1 (Canon, Tokyo,
Japan) at both examinations. Cycloplegia was induced by 1 drop
each of cyclopentolate 1% and tropicamide 1% administered in 2
cycles, 5 minutes apart, following corneal anesthesia with ame-
thocaine hydrochloride 1%. Autorefraction was performed 20 to 30
minutes after the last cycle of cycloplegic eye drops.
A detailed questionnaire was administered to obtain information
on children’s activities, ethnicity, parental myopia, and socioeconomic
status at both baseline and follow-up. Children’s activities included
hours spent in outdoor leisure activities, such as bike riding, picnics
and barbeques, and playing or walking outside; outdoor and indoor
sports; near-based activities, including reading for pleasure; and mid-
distance activities, such as computer use and television viewing.
Diopter hours of near work also were investigated, as proposed by
Mutti et al,8 using a modification of the original definition, as follows:
total diopter hours per week ¼ 3 homework þ 3 reading þ 3
handheld games þ 2 playing musical instruments þ 2 using
a computer þ 2 playing video games þ 2 playing board games.
Ethnicity was ascertained by the self-identified ethnic origin of both
parents using ethnic categories consistent with the Australian Standard
Classification of Cultural and Ethnic Groups.38 Parental myopia was
determined by spectacle prescription or analysis of spectacle-use
questions as previously described and validated.39 Socioeconomic
status was based on parental education, parental employment status
(employed or not employed), and home ownership.
Statistical Analyses
Data were analyzed using IBM SPSS Predictive Analytics Soft-
ware (version 19; IBM, New York, NY). Myopia was defined as
a SER of
0.50 D. Children were determined to have incident
myopia if classified as not myopic at baseline and myopic in either
eye at follow-up. Cumulative incidence of myopia was defined as
the proportion of children who were not myopic (at risk) at baseline
and who subsequently developed myopia in either eye during the
follow-up period. Annual incidence was calculated by dividing the
cumulative incidence percentage by the mean follow-up time in
years. To examine the potential causality of environmental expo-
sures in the development of myopia, baseline measures for activ-
ities were used. We used t tests to determine differences in time
spent on activities between children with incident myopia and
children who remained nonmyopic. Activity variables were further
categorized into low, moderate, and high levels (according to
population tertiles) of the average weekly hours spent in these
activities. To investigate trends across population tertiles for
activities and number of parents with myopia, chi-square analysis
and Ptrend values from linear-by-linear association were used.
Multivariate logistic regression analysis was used to calculate
odds ratios (ORs) and 95% confidence intervals (CIs), using inci-
dent myopia as the dependent variable and various activity and
parental factors as covariates and adjusting for sex, age, and
parental myopia where appropriate. Univariate and multivariate
logistic regression models were further used to investigate the
predictive ability of significant risk factors for incident myopia, and
2101
 Ophthalmology Volume 120, Number 10, October 2013
Figure 1. The proportion of children with incident myopia and the
proportion who remained nonmyopic according to baseline spherical equiv-
alent (SE) refraction of the refractive error (RE) in the (A) younger cohort
(age 6 years at baseline) and (B) older cohort (age 12 years at baseline).
receiver operating characteristic (ROC) curves were plotted. From
the ROC curves, the area under the curve (AUC) was used as
a measure of the predictive ability of each model. The AUC
denotes the probability (q) that any randomly matched pair of
normal (those who remained nonmyopic) and abnormal (those with
incident myopia) individuals would be correctly ranked according
to their likelihood of disease (incident myopia).40 This measure is
statistically compared with 0.50, which is the value determined to
relate to chance prediction, with an AUC >0.50 indicating a better
than chance prediction. The ORs were calculated as opposed to
relative risk because of logistic regression analysis being used to
produce ROC curves and because we could not determine the
time of myopia onset for survival analysis, given our 5- to
6-year follow-up time.
Results
Longitudinal data were available for the analysis of a total of 2059
children: 863 in the younger cohort and 1196 in the older cohort.
Both follow-up cohorts were predominately European Caucasian,
with East Asian ethnicity second (younger cohort; 67.5%, 15.8%
and older cohort; 57.0%, 19.6%, respectively). Female subjects
made up 48.0% of the younger cohort and 50.1% of the older
2102
cohort. The cumulative incidence of myopia was 14.8% for the
younger cohort and 17.3% for the older cohort, with an annual
incidence of 2.4% and 3.8%, respectively. There was no significant
difference in the incidence of myopia between male and female
subjects in the younger (P ¼ 0.3) or older follow-up cohorts (P ¼
0.1). Children of East Asian ethnicity had a significantly greater
incidence of myopia (younger cohort, 48.5%; older cohort, 35.8%)
than children of European Caucasian ethnicity (younger cohort,
8.7%; older cohort, 14.5%) in both age cohorts (P < 0.0001).
Baseline Spherical Equivalent Refraction and
Parental Factors
Figure 1 presents the proportion of children with incident myopia
according to baseline SER. For the younger cohort, there was
a greater proportion of children who became myopic in the
3 refractive categories that were
þ1.00 D at baseline (at mean
age of 6 years) compared with children with a more hyperopic
refraction. By taking a baseline refraction þ2.00 D as the
reference group, the odds for incident myopia increased with
decreasing refractive status at baseline (Ptrend < 0.0001), from an
OR of 36.5 (95% CI, 4.95e269.27) for a baseline
refraction þ1.00 to >þ0.50 D to an OR of 131.7 (95% CI,
16.8e1030.03) for baseline refraction þ0.50 to >0.00D and to
an OR of 395.0 (95% CI, 21.40e7290.54) for those with
refraction between 0.00 and >0.50 D at baseline. In the older
cohort, this trend with baseline refraction (at a mean age of 12
years) continued (Ptrend < 0.0001) but with lower odds: children
within þ0.50 to >0.00 D had an OR of 17.1 (95% CI,
2.30e127.92) for incident myopia, increasing to 153.3 (95% CI,
19.3e1219.26) for the lower baseline refraction 0.00 to >0.50 D.
In the younger cohort, the proportion of children with incident
myopia varied according to their number of parents with myopia
(Table 1), so that only 7.8% of children without a myopic parent
became myopic, whereas 21.4% of children with 1 myopic
parent and 22.0% of those with 2 myopic parents (P < 0.0001)
became myopic, with similar odds for incident myopia whether
one (OR, 3.16; 95% CI, 1.94e5.15) or both parents were
myopic (OR, 3.33; 95% CI, 1.63e6.82). By contrast, in the
older cohort there was no relationship between the proportion of
children with incident myopia and parental myopia (P ¼ 0.4).
There was no significant difference in incident myopia in either
cohort for children whose parents had higher educational
attainment (younger cohort, P ¼ 0.1; older cohort, P ¼ 0.3) or
were employed (younger, P ¼ 0.5; older, P ¼ 0.9). European
Caucasian children with 1 or both parents myopic had
a significantly higher incidence of myopia (younger cohort,
P < 0.0001; older cohort, P ¼ 0.01). However, this was not
evident for children of East Asian ethnicity in either cohort
(younger cohort, P ¼ 0.5; older cohort, P ¼ 0.7).
Environmental Risk Factors for Incident Myopia
In the younger cohort, children with incident myopia had spent
significantly less total time outdoors at baseline than children who
remained nonmyopic (P < 0.0001, Table 2). Children with incident
myopia also performed significantly more near work (P ¼ 0.02) and
did more D-hours of near work (P ¼ 0.02) per week than those who
remained nonmyopic. In the older cohort, children with incident
myopia spent significantly less total time outdoors (P ¼ 0.001)
than children who remained nonmyopic. There were no
significant differences in near work (P ¼ 0.06) or in D-hours of
near work (P ¼ 0.1) performed between children with incident
myopia and children who remained nonmyopic.
When stratified by ethnicity (Table 3), all children in the
younger cohort with incident myopia showed a tendency toward
 French et al 
Risk Factors for Incident Myopia

Table 1. Effect of Significant Baseline Risk Factors on Incident Myopia

Younger Cohort Older Cohort

Incident Odds Ratio* Incident Odds Ratio*
Risk Factors Myopia, % (n) (95% CI) P Value (c2) Ptrend Myopia, % (n) (95% CI) P Value (c2) Ptrend
Parental Myopia
0 parents 7.8 (32) Reference 19.7 (91) Reference
1 parent 21.4 (46) 3.16 (1.94e5.15) 23.9 (60) 1.27 (0.88e1.84)
2 parents 22.0 (13) 3.33 (1.62e6.82) <0.0001 <0.0001 20.9 (9) 1.12 (0.52e2.43) 0.41 0.36
Time outdoorsy
High 8.2 (22) Reference 15.5 (52) Reference
Moderate 12.5 (33) 1.14 (0.59e2.21) 24.0 (76) 2.00 (1.28e3.14)
Low 23.3 (64) 2.84 (1.56e5.17) <0.0001 <0.0001 25.8 (77) 2.15 (1.35e3.42) 0.003 0.001
Near workz
Low 14.3 (34) Reference 18.9 (61) Reference
Moderate 11.4 (27) 1.68 (0.89e3.16) 22.4 (74) 1.43 (0.93e2.21)
High 20.5 (49) 2.35 (1.30e4.27) 0.11 0.036 23.8 (70) 1.31 (0.83e2.06) 0.32 0.14

Bold values indicate statistical significance (P <0.05).

CI ¼ confidence interval.
*The odds ratios were calculated from multivariate logistic regression analysis and are adjusted for age, sex, and parental myopia. Parental myopia analysis
adjusted for age and sex only.
y
Tertile ranges for time spent outdoors per week are as follows: younger cohort: low (
16 hours), moderate (>16e
23 hours), and high (>23 hours); older
cohort: low (
13.5 hours), moderate (>13.5e
22.5 hours), and high (>22.5 hours).
z
Tertile ranges for time spent in near work per week are as follows: younger cohort: low (
13 hours), moderate (>13e
19.5 hours), and high (>19.5
hours); older cohort: low (
17 hours), moderate (>17e
25.5 hours), and high (>25.5 hours).

spending fewer hours per week outdoors at baseline in comparison
with children who remained nonmyopic in both the children of
European Caucasian (nonmyopic ¼ 22.1 hours; myopic ¼ 20.5
hours) and East Asian (nonmyopic ¼ 13.9 hours; myopic ¼ 12.9
hours) ancestry, although comparisons within ethnic groups were
not statistically significant (P ¼ 0.18 and 0.51, respectively). In
the older cohort, total time spent outdoors also was higher in the
children who remained nonmyopic compared with those with
incident myopia of both European Caucasian (nonmyopic ¼ 21.0
hours; myopic ¼ 20.3 hours) and East Asian ethnicity
(nonmyopic ¼ 14.1 hours; myopic ¼ 13.4 hours), but again
these differences were not statistically significant.
In the younger cohort, for both ethnic groups, there were no
significant differences in the amount of near work performed or
time spent on other activities between children who remained
nonmyopic and those with incident myopia (Table 3). However, in
the older cohort, European Caucasian children who became
myopic spent significantly more time on near work (22.4 vs.
20.3 hours, P ¼ 0.04) and total hours indoors per week (63.4 vs.
58.0 hours, P ¼ 0.04). This was not evident in the East Asian
children for near work (P ¼ 0.81) or time indoors (P ¼ 0.96).
Significant baseline activities (i.e., time outdoors and near
work) were further split into population tertiles for analysis in
Table 1. In the younger cohort, there was a trend toward an
increasing proportion of children with incident myopia spending
fewer hours of time outdoors per week (Ptrend < 0.0001) and
performing higher levels of near-work activity (Ptrend ¼ 0.04) at
baseline. Odds ratios for incident myopia were significantly
higher for the lowest tertile of time outdoors compared with the
highest and for the highest tertile of near work compared with
the lowest. In the older cohort, there was a significant trend
toward greater incident myopia in the children who had spent
less time outdoors (Ptrend ¼ 0.001), with increased odds for
incident myopia among children in the moderate and low tertiles
of time spent outdoors, but there were no significant trends in
relation to near work (Table 1).

Table 2. Baseline Activities per Week for Children with Incident Myopia Compared with Children Who Remained Nonmyopic in the
Younger (Aged 6 Years at Baseline) and Older (Aged 12 Years at Baseline) Cohorts

Younger Cohort Older Cohort

Remained Nonmyopic Incident Myopia Remained Nonmyopic Incident Myopia
Activities (Mean/Week) (Mean/Week) P Value (Mean/Week) (Mean/Week) P Value
Time outdoors (hr) 20.96 16.29 <0.0001 19.62 17.15 0.001
Outdoor leisure (hr) 18.27 14.08 <0.0001 15.85 13.66 0.001
Outdoor sport (hr) 3.60 2.76 0.049 4.69 4.28 0.27
Indoor sport (hr) 1.05 0.76 0.054 1.12 1.12 0.10
Near work (hr) 17.55 19.36 0.020 20.71 22.08 0.062
Near work (D-hr) 43.21 49.73 0.022 79.79 84.49 0.11
Books read (n) 4.35 4.19 0.57 2.94 3.13 0.52
Middle-distance (hr) 16.15 16.07 0.92 21.99 22.80 0.28
Time indoors (hr) 50.46 51.94 0.48 59.68 62.71 0.11

Bold values indicate statistical significance (P <0.05).

2103
 Ophthalmology Volume 120, Number 10, October 2013

Table 3. Baseline Activities per Week for Children with Incident Myopia Compared with Children Who Remained Nonmyopic in the
Younger (Aged 6 Years at Baseline) and Older (Aged 12 Years at Baseline) Cohorts, Stratified by Ethnicity

Younger Cohort
European Caucasian East Asian
Remained Nonmyopic Incident Myopia Remained Nonmyopic Incident Myopia
Activities (Mean/Week) (Mean/Week) P Value (Mean/Week) (Mean/Week) P Value
Time outdoors (hr) 22.1 20.5 0.18 13.9 12.9 0.51
Outdoor leisure (hr) 19.1 17.8 0.24 12.7 11.3 0.39
Outdoor sport (hr) 3.7 3.1 0.38 2.6 2.0 0.37
Indoor sport (hr) 1.0 0.7 0.14 1.3 0.8 0.07
Near work (hr) 17.3 18.3 0.34 19.1 19.8 0.67
Near work (D-hr) 40.5 42.8 0.51 55.3 54.9 0.95
Books read (n) 4.5 4.2 0.47 3.6 4.3 0.17
Middle-distance (hr) 15.6 15.1 0.61 17.9 17.0 0.60
Time indoors (hr) 49.8 49.3 0.84 53.7 53.5 0.96

Older Cohort
Remained Nonmyopic Incident Myopia Remained Nonmyopic Incident Myopia
Activities (Mean/Week) (Mean/Week) P Value (Mean/Week) (Mean/Week) P Value
Time outdoors (hr) 21.0 20.3 0.49 14.1 13.4 0.66
Outdoor leisure (hr) 17.1 16.4 0.48 10.9 10.3 0.65
Outdoor sport (hr) 4.8 4.4 0.38 3.8 3.7 0.83
Indoor sport (hr) 1.1 1.3 0.44 0.8 0.9 0.87
Near work (hr) 20.3 22.4 0.041 23.8 23.4 0.81
Near work (D-hr) 76.7 81.9 0.20 96.2 94.2 0.81
Books read (n) 2.7 2.5 0.65 3.5 3.5 0.95
Middle-distance (hr) 21.5 22.6 0.30 23.2 25.1 0.29
Time indoors (hr) 58.0 63.4 0.038 69.3 65.7 0.48

Bold values indicate statistical significance (P <0.05).

Combined Effects of Time Spent Outdoors and amount of near work performed did not significantly alter this level
Near Work of risk.

The combined effects of time spent outdoors and in near work at
baseline are explored in Figure 2. Children with high levels of time
outdoors and low levels of near work had the lowest odds of
incident myopia and were used as the reference group (OR, 1).
In the younger cohort (Fig 2A), children who spent low amounts
of time outdoors and performed high levels of near work when
at a mean age of 6 years had significantly increased odds of
incident myopia by age 12 years (OR, 15.9; 95% CI,
3.45e73.40). This was also true for those who spent less time on
near work (moderate near work: OR, 7.9; 95% CI, 1.69e37.34;
low near work: OR, 5.3; 95% CI, 1.11e25.38). However, the
differences between those performing different levels of near
work combined with low amounts of time outdoors were not
statistically significant.
In the older cohort, a similar pattern emerged (Fig 2B), with less
time outdoors in combination with high levels of near work at
a mean age of 12 years conferring a lesser but still significant
OR of 5.1 (95% CI, 1.91e13.45) for the development of
incident myopia by 17 years. However, in this cohort, only those
spending greater amounts of time outdoors at age 12 years
seemed to be protected from developing myopia. Increasing time
spent in near work combined with a high level of time outdoors
was associated with a 2-fold higher odds of incident myopia than
the reference group, but this difference also was not statistically
significant (moderate near work: OR, 2.45; 95% CI, 0.95e6.32;
high near work: OR, 2.27; 95% CI, 0.88e5.85). Conversely,
moderate to low amounts of time spent outdoors at baseline
significantly increased the OR for incident myopia by more than 3-
fold compared with the reference group, but again, varying the
Predictive Analysis of Risk Factors for
Incident Myopia
Risk factors associated with significantly greater odds of incident
myopia were examined in univariate logistic regression analyses
from which ROC curves were plotted. In the younger cohort, all of
the variables examined, namely, baseline SER, time spent outdoors,
near work, parental myopia, and ethnicity, were significant predic-
tors of incident myopia. Baseline SER was the strongest predictor of
incident myopia, with an AUC in ROC analysis of 0.84 (P < 0.0001).
Ethnicity had an AUC of 0.67 (P < 0.0001), parental myopia had an
AUC of 0.65 (P < 0.0001), time outdoors had an AUC of 0.64
(P < 0.0001), and near work had an AUC of 0.61 (P ¼ 0.001).
Various multivariate models were constructed to determine the
best combination of factors for predicting incident myopia. In the
younger cohort, models that excluded baseline SER did not
significantly increase predictive ability by comparison with
univariate analysis of baseline SER alone. Those models that
included baseline SER with the addition of other variables that
improved the AUC for the younger cohort are shown in Figure 3A.
Of these variables, the addition of time spent outdoors and parental
myopia slightly improved the AUC from 0.84 to 0.86
(P < 0.0001); however, parental myopia became nonsignificant.
The addition of near work to this model slightly increased the
AUC to 0.87 (P < 0.0001). Furthermore, the addition of ethnicity
to the model increased the AUC to 0.89 (P < 0.0001), with all
variables, other than parental myopia, remaining significant.
However, removing parental myopia from this model resulted in
a significantly lower AUC (0.86).

2104
 French et al 
Risk Factors for Incident Myopia

Figure 2. Multivariate odds ratios (ORs) of incident myopia by tertiles of

baseline time outdoors and near work in the (A) younger cohort (age
6 years at baseline) and (B) older cohort (age 12 years at baseline). The
ORs were calculated from multivariate logistic regression analysis and are
adjusted for parental myopia, age, and sex.

In the older cohort, on univariate analysis only, the baseline

SER (AUC ¼ 0.89; P < 0.0001), time spent outdoors (AUC ¼
0.58; P ¼ 0.001), and ethnicity (AUC ¼ 0.65; P < 0.0001)
produced a significant AUC for predicting incident myopia. Again,
multivariate models were constructed using univariate baseline
SER as the benchmark (Fig 3B). The addition of time outdoors or
ethnicity to baseline SER did not improve the predictive ability of
the model (all AUC ¼ 0.88; P < 0.0001).
Discussion
In this longitudinal follow-up analysis, having a less
hyperopic refraction at baseline was the most significant
predictor of incident myopia in the younger and older
cohorts. Environmental risk factors, in particular, less time
spent outdoors and greater levels of near work, were
significantly associated with incident myopia in the younger
cohort (age 6 years at baseline), with less time outdoors the
most consistently associated. However, in the older cohort
(age 12 years at baseline), only less time spent outdoors was
significantly associated with incident myopia, and the effects
Figure 3. The receiver operating characteristic curves for prediction of
incident myopia in the (A) younger cohort (age 6 years at baseline) and (B)
older cohort (age 12 years at baseline). Results of the univariate model for
baseline spherical equivalent refraction (SER) are plotted for comparison with
multivariate models. The multivariate models for each cohort were con-
structed on the basis of variables that were significant on univariate analysis.
were not as strong. Parental myopia was associated with
incident myopia in the younger cohort but not the older
cohort. In both cohorts, children of East Asian ethnicity
spent significantly less time outdoors than children of
European Caucasian ethnicity, which may in part contribute
to the significantly higher incidence of myopia in this group.
Our previous analysis of these cohorts has shown that a more
hyperopic SER at baseline was associated with greater time
spent outdoors30 and more weakly with lower levels of time

2105
 Ophthalmology Volume 120, Number 10, October 2013
spent on near work.25 Thus, less time outdoors seems to be
related to both prevalent and incident myopia, particularly
in younger children. Whether these factors also affect the
rate of myopia progression remains to be determined, but
significant seasonal differences in myopia progression have
been reported,41,42 which are potentially explicable in
terms of the same risk factors. However, some recent
evidence does not support this idea.43
The refraction of children at baseline was the strongest
individual predictor of incident myopia in both cohorts. In
our younger cohort, children with a baseline refraction of

þ1.00 D had a significantly greater risk of developing
myopia than children with a more hyperopic refraction. In
the older cohort, children with a refraction at baseline

þ0.50 D were at greater risk. This is in agreement with
findings from the Orinda Longitudinal Study of Myopia in
third-grade children, which also found refraction at baseline
to be a strong predictor of future myopia, more so than
ocular components, such as axial length and corneal
radius,44 or risk factors, including parental myopia and
outdoor sport.17 This makes sense in the context of the
natural history of refractive development. A less hyperopic
refraction would put children at significant risk of moving
beyond emmetropia and into the myopic range with age
and axial elongation. This may be important because
children with a baseline refraction of close to þ1.00 D at
age 6 years or þ0.50 D at age 12 years could be targeted
as an at-risk group in prevention programs for myopia,
with the goal of maintaining a slightly hyperopic refraction.
In the younger cohort, although baseline refraction was the
strongest predictor overall, there was also a significant asso-
ciation between incident myopia and other risk factors
examined, in particular, time spent outdoors, near work, and
parental myopia. There was a significantly increased risk of
incident myopia for children in the lowest tertile of time
outdoors and the highest tertile of near work. However,
analyzing the combined effect of time outdoors and near work
showed that time outdoors was more consistently associated
with incident myopia in a stepwise fashion. Of note, children
who performed higher amounts of near work, but also spent
large amounts of time outdoors, were still protected from the
development of myopia. This is consistent with the cross-
sectional findings from the SMS30 and confirms the impact
of time spent outdoors before the onset of myopia in this
age group. Likewise, there was an increased risk of incident
myopia with parental myopia, although there was little
difference in risk associated with having 1 or 2 myopic
parents. This is also consistent with other reports.17,45
In multivariate models for the younger cohort, the addition of
these significant risk factors to the baseline refraction signifi-
cantly increased the predictive ability of the model. This
suggests that the risk factors have a significant impact on the
development of myopia between the ages of 6 and 12 years. This
is important because time outdoors is potentially modifiable,
and increasing the time that children in this age group spend
outdoors may help to prevent the development of myopia. This
requires further investigation through intervention studies.
Because baseline SER was already a strong predictor of incident
myopia in this age group, influences on refractive development
before age 6 years also warrant further investigation. One study
2106
has investigated time spent outdoors in a cross-sectional sample
of preschool-aged children in Singapore.46 Although this study
found no impact of time spent outdoors on myopia in this age
group, it is possible that the effect of time spent outdoors at
this age may not be above threshold myopia, but more on the
shape of the relatively hyperopic distribution of refractive
error and an individual child’s position within the distribution.
Longitudinal cohort studies of preschool-aged children are
required to determine the impact of time spent outdoors on the
eventual development of myopia.
An interesting aspect of these data is the differences in
the effect of risk factors between age cohorts. In the older
cohort, less time spent outdoors was the only factor associ-
ated with incident myopia, with no significant effect of
parental myopia or near work. Although, adding time spent
outdoors to baseline refraction did not improve prediction of
incident myopia in adolescence. Although, baseline refraction
remained a strong predictor of incident myopia, and our
previous analysis of this cohort has shown that a more
hyperopic baseline refraction was strongly associated with
greater time spent outdoors.30 This suggests that baseline
refraction at age 12 years may have been determined by
earlier lifestyle, such as more time outdoors when younger,
and that the future development of refractive error has been
somewhat constrained by this age. This is also relevant for
parental myopia, which was a significant determinant of
refraction in the younger cohort, but not in our older cohort,
implying that the effects of parental myopia are probably
predominantly seen in the onset of myopia up to age 12
years. Therefore, intervention studies based on increasing
time spent outdoors or changes to other modifiable risk
factors may be more effective at younger ages.
There were a number of ethnic differences noted, which
were consistent for both cohorts. Children of East Asian
ethnicity had a significantly higher incidence of myopia and
spent significantly less time outdoors and more time on near
work than European Caucasian children. This indicates that
differential risk factor exposures between these ethnic groups
in Australia may explain some of the differences in refractive
status. Further support for this hypothesis was presented by
Rose et al,13 in 2008 by comparing myopia prevalence and
environmental risk factors for children of Chinese ancestry
growing up in Australia and Singapore. Children from
Australia had a significantly lower prevalence of myopia
compared with children from Singapore, with a large
difference in time spent outdoors. Such comparison of
different ethnic groups within the same location, and
between the same ethnic groups in different locations,
indicates the strength of the effect of environmental factors,
such as time spent outdoors, on refractive development. In
addition, although parental myopia was a significant risk
factor for incident myopia among European Caucasian
children, there was no effect on children of East Asian
ethnicity. A likely explanation for this is that the distinct
myopiogenic pattern of activities of children with East
Asian ethnicity may largely override the influence of
hereditary factors.
Consistently throughout our multivariate analyses for
incident myopia, when both time spent outdoors and
ethnicity were included, the effect of time spent outdoors
 French et al 
Risk Factors for Incident Myopia
was no longer significant, despite trends toward time spent
outdoors providing a protective effect in other analyses.
Although this result could be interpreted as ethnicity having
a stronger impact on incident myopia, there is another
plausible explanation. Because children with incident
myopia spend less time outdoors and tend to be of East
Asian ethnicity, it is possible that within an ethnic group in
one location there may not always be sufficient variation in
time spent outdoors to produce a significant result.
However, comparing across locations with a wider range of
variation13 may reveal significant differences. A further
possibility, given the strong correlation between these
variables and the subsequent loss of significance for time
spent outdoors, is that this could represent an effect of
collinearity within the regression model.47
A major strength of this study is that it represents a large
population-based cohort of school-aged children from
whom detailed cycloplegic refraction plus detailed ques-
tionnaire data were collected, providing a detailed assess-
ment of associations between incident myopia and risk
factors. A limitation of the study is that approximately half
the baseline study participants were lost to follow-up for
both age cohorts because of the 5- to 6-year follow-up time
and movement away from the original schools. In the
younger cohort, a higher proportion of children of European
Caucasian ethnicity were retained, whereas in the older
cohort, a higher proportion of children of East Asian origin
were retained. However, stratification by ethnicity allowed
us to resolve ambiguities in the analyses. In the older cohort,
but not younger cohort, children who were retained in the
study tended to have higher socioeconomic status. This is an
inevitable consequence of reexamining children in the final
2 years of schooling because a significant proportion of
children leave secondary school just before these final
school years. The children who leave tend to be of lower
socioeconomic status, whereas the retained children have
had extra years of schooling. This may have somewhat
increased the proportion of children with myopia within our
older cohort and should be taken into consideration.
The use of a subjective questionnaire to measure time spent
in various activities is a potential source of error through
inaccurate reporting or recall bias of participants. However,
this approach is cost-effective and feasible in large
population-based studies of this kind. In smaller, more
focused studies, the use of objective measures of light expo-
sure may be more accurate. These also may be used to test the
validity of the questionnaires. A strength of our questionnaire
was the detailed measures that allowed us to discriminate
between the effect of total time spent outdoors, outdoor sport,
outdoor leisure, and indoor sport. From our cross-sectional
data, it was evident that the protective effect of time spent
outdoors on myopia was a consequence of the outdoor envi-
ronment rather than participation in activities such as sport per
se,30 and this has been confirmed in other studies.28,31
In conclusion, having a less hyperopic refraction at
baseline was a significant predictor of incident myopia in
both younger and older cohorts of school children. In addi-
tion, less time spent outdoors, greater near work, and
parental myopia predicted a greater risk of incident myopia
among children in the younger cohort. Less time spent
outdoors also was associated with incident myopia among
children in the older cohort, but its effect was not as strong.
Thus, time spent outdoors seems to be associated with both
prevalent myopia at baseline and incident myopia, particu-
larly in younger children, indicating that public health
interventions based on promoting time outdoors to reduce
myopia should target children at younger ages.
References
1. Lin LL, Shih YF, Hsiao CK, Chen CJ. Prevalence of myopia
in Taiwanese schoolchildren: 1983 to 2000. Ann Acad Med
Singapore 2004;33:27–33.
2. He M, Zeng J, Liu Y, et al. Refractive error and visual
impairment in urban children in southern China. Invest Oph-
thalmol Vis Sci 2004;45:793–9.
3. Fan DS, Lam DS, Lam RF, et al. Prevalence, incidence, and
progression of myopia of school children in Hong Kong.
Invest Ophthalmol Vis Sci 2004;45:1071–5.
4. Quek TP, Chua CG, Chong CS, et al. Prevalence of refractive
errors in teenage high school students in Singapore.
Ophthalmic Physiol Opt 2004;24:47–55.
5. Saw SM, Goh PP, Cheng A, et al. Ethnicity-specific prevalences
of refractive errors vary in Asian children in neighbouring
Malaysia and Singapore. Br J Ophthalmol 2006;90:1230–5.
6. Zhao J, Pan X, Sui R, et al. Refractive Error Study in Children:
results from Shunyi District, China. Am J Ophthalmol
2000;129:427–35.
7. He M, Huang W, Zheng Y, et al. Refractive error and visual
impairment in school children in rural southern China.
Ophthalmology 2007;114:374–82.
8. Mutti DO, Mitchell GL, Moeschberger ML, et al. Parental
myopia, near work, school achievement, and children’s
refractive error. Invest Ophthalmol Vis Sci 2002;43:3633–40.
9. O’Donoghue L, McClelland JF, Logan NS, et al. Refractive
error and visual impairment in school children in Northern
Ireland. Br J Ophthalmol 2010;94:1155–9.
10. Villarreal MG, Ohlsson J, Abrahamsson M, et al. Myopisation:
the refractive tendency in teenagers. Prevalence of myopia
among young teenagers in Sweden. Acta Ophthalmol Scand
2000;78:177–81.
11. Logan NS, Shah P, Rudnicka AR, et al. Childhood ethnic
differences in ametropia and ocular biometry: the Aston Eye
Study. Ophthalmic Physiol Opt 2011;31:550–8.
12. Czepita D, Mojsa A, Zejmo M. Prevalence of myopia and
hyperopia among urban and rural schoolchildren in Poland.
Ann Acad Med Stetin 2008;54:17–21.
13. Rose KA, Morgan IG, Smith W, et al. Myopia, lifestyle, and
schooling in students of Chinese ethnicity in Singapore and
Sydney. Arch Ophthalmol 2008;126:527–30.
14. French AN, O’Donoghue L, Morgan IG, et al. Comparison of
refraction and ocular biometry in European Caucasian children
living in Northern Ireland and Sydney, Australia. Invest
Ophthalmol Vis Sci 2012;53:4021–31.
15. Tay MT, Au Eong KG, Ng CY, et al. Myopia and educational
attainment in 421,116 young Singaporean males. Ann Acad
Med Singapore 1992;21:785–91.
16. Ip JM, Huynh SC, Robaei D, et al. Ethnic differences in the
impact of parental myopia: findings from a population-based
study of 12-year-old Australian children. Invest Ophthalmol
Vis Sci 2007;48:2520–8.
17. Jones LA, Sinnott LT, Mutti DO, et al. Parental history of
myopia, sports and outdoor activities, and future myopia.
Invest Ophthalmol Vis Sci 2007;48:3524–32.
2107
 Ophthalmology Volume 120, Number 10, October 2013
18. Baird PN, Schäche M, Dirani M. The GEnes in Myopia
(GEM) study in understanding the aetiology of refractive
errors. Prog Retin Eye Res 2010;29:520–42.
19. Wojciechowski R. Nature and nurture: the complex genetics of
myopia and refractive error. Clin Genet 2011;79:301–20.
20. Morgan I, Rose K. How genetic is school myopia? Prog Retin
Eye Res 2005;24:1–38.
21. Morgan IG, Ohno-Matsui K, Saw SM. Myopia. Lancet
2012;379:1739–48.
22. Wong L, Coggon D, Cruddas M, Hwang CH. Education,
reading, and familial tendency as risk factors for myopia in
Hong Kong fishermen. J Epidemiol Community Health
1993;47:50–3.
23. Wu HM, Seet B, Yap EP, et al. Does education explain ethnic
differences in myopia prevalence? A population-based study of
young adult males in Singapore. Optom Vis Sci 2001;78:234–9.
24. Zylbermann R, Landau D, Berson D. The influence of study
habits on myopia in Jewish teenagers. J Pediatr Ophthalmol
Strabismus 1993;30:319–22.
25. Ip JM, Saw SM, Rose KA, et al. Role of near work in myopia:
findings in a sample of Australian school children. Invest
Ophthalmol Vis Sci 2008;49:2903–10.
26. Saw SM, Chua WH, Hong CY, et al. Nearwork in early-onset
myopia. Invest Ophthalmol Vis Sci 2002;43:332–9.
27. Jones-Jordan LA, Mitchell GL, Cotter SA, et al; CLEERE
Study Group. Visual activity before and after the onset of
juvenile myopia. Invest Ophthalmol Vis Sci 2011;52:1841–50.
28. Jacobsen N, Jensen H, Goldschmidt E. Does the level of
physical activity in university students influence development
and progression of myopia?ea 2-year prospective cohort
study. Invest Ophthalmol Vis Sci 2008;49:1322–7.
29. Pärssinen O, Lyyra AL. Myopia and myopic progression
among schoolchildren: a three-year follow-up study. Invest
Ophthalmol Vis Sci 1993;34:2794–802.
30. Rose KA, Morgan IG, Ip J, et al. Outdoor activity reduces the
prevalence of myopia in children. Ophthalmology 2008;115:
1279–85.
31. Dirani M, Tong L, Gazzard G, et al. Outdoor activity and
myopia in Singapore teenage children. Br J Ophthalmol
2009;93:997–1000.
32. Wu PC, Tsai CL, Hu CH, Yang YH. Effects of outdoor
activities on myopia among rural school children in Taiwan.
Ophthalmic Epidemiol 2010;17:338–42.
33. Guggenheim JA, Northstone K, McMahon G, et al. Time
outdoors and physical activity as predictors of incident myopia
in childhood: a prospective cohort study. Invest Ophthalmol
Vis Sci 2012;53:2856–65.
Footnotes and Financial Disclosures
Originally received: November 7, 2012.
Final revision: January 29, 2013.
Accepted: February 26, 2013.
Available online: May 11, 2013. Manuscript no. 2012-1688.
1
Discipline of Orthoptics, Faculty of Health Sciences, University of Syd-
ney, Sydney, NSW, Australia.
2
ARC Centre of Excellence in Vision Science, Research School of
Biology, The Australian National University, Canberra, ACT, Australia.
3
Department of Ophthalmology, Centre for Vision Research, Westmead
Millennium Institute, University of Sydney, Sydney, NSW, Australia.
Financial Disclosure(s):
The author(s) have no proprietary or commercial interest in any materials
discussed in this article.
2108
34. Smith EL III, Hung LF, Huang J. Protective effects of high
ambient lighting on the development of form-deprivation
myopia in rhesus monkeys. Invest Ophthalmol Vis Sci
2012;53:421–8.
35. Ashby R, Ohlendorf A, Schaeffel F. The effect of ambient
illuminance on the development of deprivation myopia in
chicks. Invest Ophthalmol Vis Sci 2009;50:5348–54.
36. Ashby RS, Schaeffel F. The effect of bright light on lens
compensation in chicks. Invest Ophthalmol Vis Sci 2010;51:
5247–53.
37. Ojaimi E, Rose KA, Smith W, et al. Methods for a population-
based study of myopia and other eye conditions in school
children: the Sydney Myopia Study. Ophthalmic Epidemiol
2005;12:59–69.
38. Australian Bureau of Statistics. 1249.0-Australian Standard
Classification of Cultural and Ethnic Groups (ASCCEG),
2005-06. Available at: http://abs.gov.au/AUSSTATS/abs@.nsf/
DetailsPage/1249.02005-06?OpenDocument. Accessed Feb-
ruary 25, 2013.
39. Ip J, Robaei D, Rochtchina E, et al. Can information on the
purpose of spectacle use and age at first use predict refractive
error type? Ophthalmic Epidemiol 2007;14:88–92.
40. Hanley J, McNeil B. The meaning and use of the area under
a receiver operating characteristic (ROC) curve. Radiology
1982;143:29–36.
41. Fulk GW, Cyert LA, Parker DA. Seasonal variation in myopia
progression and ocular elongation. Optom Vis Sci 2002;79:46–51.
42. Jiang BC, Schatz S, Seger K. Myopic progression and dark
focus variation in optometric students during the first academic
year. Clin Exp Optom 2005;88:153–9.
43. Jones-Jordan LA, Sinnott LT, Cotter SA, et al, CLEERE Study
Group. Time outdoors, visual activity, and myopia progression in
juvenile-onset myopes. Invest Ophthalmol Vis Sci 2012;53:
7169–75.
44. Zadnik K, Mutti DO, Friedman NE, et al. Ocular predictors of
the onset of juvenile myopia. Invest Ophthalmol Vis Sci
1999;40:1936–43.
45. Saw SM, Shankar A, Tan SB, et al. A cohort study of incident
myopia in Singaporean children. Invest Ophthalmol Vis Sci
2006;47:1839–44.
46. Low W, Dirani M, Gazzard G, et al. Family history, near work,
outdoor activity, and myopia in Singapore Chinese preschool
children. Br J Ophthalmol 2010;94:1012–6.
47. Bagley SC, White H, Golomb BA. Logistic regression in the
medical literature: standards for use and reporting, with
particular attention to one medical domain. J Clin Epidemiol
2001;54:979–85.
Funding:
The SMS was supported by Grant 253732 and the SAVES was supported
by Grant 512530 from the Australian National Health & Medical Research
Council and the Westmead Millennium Institute, University of Sydney. The
funding organization had no role in the design or conduct of this research.
Correspondence:
Kathryn A. Rose, PhD, Discipline of Orthoptics, Faculty of Health
Sciences, University of Sydney, PO Box 170, Lidcombe, NSW,
Australia, 1815. E-mail: kathryn.rose@sydney.edu.au.