Permission was obtained from the NSW Department of Education and the NSW Catholic Education Office to approach all schools in the eastern region of Sydney (some thirty coeducational primary schools) to send entire classes to the VEC. A flyer was sent describing the VEC science excursion and age-appropriate eye examination, inviting Years 1, 3 and 5 particularly to participate.

The group of 1,936 children examined came from the eastern suburbs along the southern beaches of Sydney, and may be thought of as randomly selected with little likelihood of bias to the data as individual classes were free to respond. Children were drawn from twelve government and non-government primary schools and one pre-school and attended the clinic only once. During the 1996 Australian Bureau of Statistics census 14,785 children aged 4 to 12 years were recorded in this region (Randwick and Waverley precincts of Eastern Sydney) who came from a very broad range of ethnic and socio-economic backgrounds present, where 37 different languages might be spoken in the home 32. This was reflected in the children attending VEC. Census data indicate approximately 9% of the children in the current study were likely to be of Asian origin 32, a figure supported by our interpretation of family name for each child 30. Participation in the eye examinations was typically well over 90% for each class, with teachers reporting non-participation to be predominantly due to illness on the day. Less than 3% of parents intentionally prevented participation, even if eye care had previously been sought. This particularly high participation rate was largely due to the attraction of a an age-appropriate student-centred hands-on science lesson about eyes and vision 31 delivered alongside the eye examination.

The comprehensive optometric examination by experienced paediatric practitioners included all age-appropriate tests meeting Australian Optometric Competency Standards, except that parents/guardians were not present to ratify history. Refractive error was determined by non-cycloplegic retinoscopy with optical fogging while the child maintained fixation on a distant non-accommodative (6 metre) target. In most cases refractive status was confirmed by subjective refraction. Other tests included letter visual acuity at 6 m and 33 cm, cover test for strabismus, motilities, saccades, pupil reactions, near point of convergence, heterophoria, stereopsis, accommodative facility, colour vision and ophthalmoscopy.

Cycloplegic retinoscopy was not undertaken for many reasons including the fact that VEC studies started prior to the 2000 convention suggesting use of cycloplegic retinoscopy for studies of refractive error prevalence 11. Secondly, the VEC visit was meant as an excursion and the children had to return to normal classes with near work demands after the morning outing. Thirdly, it was important for comparison purposes to use refractive data procured under the same conditions as that used for the earlier groups of children. Fourthly, an initial evaluation without cycloplegia is necessary in order to understand daily function. Fifthly, non-cycloplegic retinoscopy was only one component of the exam. Outcomes regarding a decision to refer would not alter for most children had a cycloplegic refraction been carried out, as several other near function tests that would also indicate the possible existence of latent hyperopia or pseudo-myopia were included. Lastly, the degree of refractive error may in fact be influenced by cycloplegia (see Discussion for elaboration 333435363738).

Autorefractors were not employed as hand-held versions were unavailable when the first cohort was seen. Equally as important, there is no convincing evidence that the proportion of myopes identified in the sample would have changed 39.

To compare the estimated prevalence of myopia in this urban population of 'Australian children' over the last decade, this more recent 2000s data set was analysed against data from an earlier cohort of 2,322 children with similar demographics seen in the early 1990's, using the same testing protocols and seen at the same venue 30. The optometric results of that earlier cohort have previously been reported 40, and it was noted that 7.1% of those children were already wearing spectacles 30, indicating that our recruitment procedure did not preclude children already under the care elsewhere. The data for any child examined in both cohorts was deleted from the earlier data set to avoid bias in the analysis. The mean date of assessment for this last 2000s cohort was September 2000, and for the early 1990s cohort was June 1992. Thus, the average gap between assessments of children from the two cohorts was 8 years and 3 months.

Data was analysed by Analysis of Variance ANOVA (StatView software). Only refractive data from right eyes was used for the current refractive class analysis, as the correlation between right and left eye refractions was extremely high (p < 0.0005). The preferred criterion to define myopia in this study is that used clinically in Australia: a spherical equivalent equal to or more minus than -0.50 D. However, as myopia more minus than -0.50 D has occasionally been used to define myopia in epidemiological studies 131941, analyses using the criterion 'myopia more minus than -0.50 D' were also performed for comparison. Hyperopia was defined as spherical equivalents greater than +0.50 D. Thus, emmetropia for this study was defined as refractions in the range -0.25 to +0.50 dioptres spherical equivalence inclusive. Means are quoted with the associated standard error.

The number of children per age group for the two cohorts is shown in Fig. 3, and notably, the age profile differs slightly between cohorts, though the mean age of 8.37 years for the earlier cohort was similar to that of this later cohort. The mean and standard error for right eye spherical equivalent refractive error by age is shown for each cohort in Fig. 4.

A three-way analysis of variance was carried out for cohort, age and gender. There was a main effect for age (p < 0.0001), but no main effect for either cohort (p = 0.134) or gender (p = 0.61). However, there were age/gender/cohort interactions that indicate a trend towards an increasing shift away from the hyperopic refraction in the later cohort.

As alluded to in the introduction, the proportion of Sydney children with myopia is dramatically less than in Asia. Indeed, the proportion appears significantly lower than in the USA 41 and Canada 42 (4% and 6% of 6 year olds respectively, or 20% of 12 year olds in USA), but higher than urban India with only 4.4% of all school children under 16 years myopic 13 and higher particularly than in other less developed countries 10.

In the past, a lack of internationally accepted definitions for 'myopia' has hampered valid comparisons across the various studies 10. Commonly the criteria 'greater than -0.50 DS' or 'at least -0.50 DS' are employed. However, our separate analyses using both of these criteria only resulted in a difference of 1.5% of all children included as myopic, in keeping with other dual analyses 1341, and is low either way when compared with Asia or North America.

Comparison across studies is also difficult when only an 'overall' mean refraction is presented covering all children in a study, due to the well known increasing prevalence of myopia with age. Indeed, the comparison of data from our own two data sets is confounded to some extent by the slightly different age profiles for each cohort. However, in neither cohort was the age range nor mean significantly different, so the similar proportion of myopes is not unexpected.

The question of optimal ocular conditions for comparison of the prevalence of refractive errors remains controversial. A cycloplegic agent is typically proposed as the gold standard 34344 in the belief that it will eliminate ciliary muscle action or spasm, and thus unmask latent hyperopia or pseudomyopia. Thus, the use of a cycloplegic would be firstly predicted to lead to a decrease in the prevalence of myopia, and an increase in the prevalence of hyperopia. However, as a cycloplegic also leads to associated mydriasis and the introduction of unpredictable spherical aberrations, it is arguable that cycloplegia will induce unpredictable errors. In fact, Gao et al 38 in 2002 reported significant changes in the refractive components of children's eyes under conditions of deep cycloplegia and mydriasis that were greatest in hyperopic eyes and smallest in myopic eyes, adding no definitive evidence as to the relative efficacy of cycloplegia.

Thus there appears to be no scientific concurrence regarding the efficacy of cycloplegia for studies on the prevalence of myopia 353637, with several major studies electing to use cycloplegia (see review in 10911) and others not 18192021234245. Presumably this design variability exists because there is no decisive evidence indicating a difference between refractions determined with and without a cycloplegic agent in eyes that have a myopic refraction. In general, a more positive retinoscopic finding is reported under cycloplegia, though considerable individual variation is seen including a myopic shift in some 3335363746. Not surprisingly, the differences noted decreased both with age and with less positive refraction.

As our refractive data was derived from non-cycloplegic retinoscopy we readily concede that mean refractive error may be less hyperopic than if a cycloplegic had been used. However, we suggest that as the influence of a cycloplegic is uncertain and is of least concern for myopes, the estimated prevalence of myopia will not be significantly altered by our decision to not use a cycloplegic. In support of this notion are new conference data from Rose et al 4748 reporting refractive status ascertained by cycloplegic autorefraction in over 1,000 children aged 6–7 years from across the same city of Sydney. They reported values of 'around 3%' for the prevalence of myopia of at least 0.50D 47, and then the value of 1.5% for myopia of 'approximately 0.50D' 48 with a participation rate between 73 and 80%. From Table 2 it can be seen that 2.4% of our 6 year olds in the current study were at least 0.50D myopic – a value that is strikingly similar.

Worldwide patterns of the prevalence of myopia suggest significant differences are likely to be due to the different demographics and lifestyles 11049. Zadnik 41 concedes that the increase in numbers of myopic children in the US Orinda study may be due to changing ethnic demographics. The apparent slight increase in myopia in Australia reported in the current study may also be in part accounted for by our changing ethnic demographics in urban areas. However demographics and ethnic compositions are unlikely to be responsible for the large changes reported in Asian and some other western countries 150.

Whatever way it is argued, our results indicate little evidence for an epidemic of myopia although there is a developmental trend towards an earlier decrease in hyperopia to the point of myopia. Thus, the question of whether it is a matter of lifestyle, or perhaps familial environmental stress, or more, remains. Certainly, the education system and housing has changed little in Australia the last 30 years. By comparison, most Asian children participating in myopia epidemiological studies reportedly are more likely to live in high-rise residential blocks 17 and have strong demands at school to memorize along with parental and peer pressure to do well, and for some, a competitive entrance examination to enter school 1951.