Variations of OCT measurements corrected for the magni ̄ cation e ® ect according to axial length and refractive error in children

Inmaculada Bueno-Gimeno*, Enrique España-Gregori, Andres Gene-Sampedro*, Juan Carlos Ondategui-Parra and Carlos J. Zapata-Rodriguez* *Department of Optics and Optometry and Vision Science University of Valencia, Spain Department of Surgery, University of Valencia University Hospital La Fe, Spain Faculty of Optics and Optometry, University Vision Centre Polytechnic University of Catalonia, Spain inmaculada.bueno@uv.es


Introduction
Optical coherence tomography (OCT) is a noninvasive, noncontact technology that provides crosssectional images of the retina.The instrument uses speci¯c algorithms to automatically measure retinal thickness (RT) and retinal nerve ¯ber layer (RNFL), to quantify optic nerve head (ONH) morphology and to assess changes in retinal diseases. 1This is especially important in children because visual loss associated with retinal disease can negatively in-°uence their visual development. 2ang et al. 3 revealed that magni¯cation attributable to axial length (AL) and refractive error has minimal impact on measurements of macular and RNFL thickness in 120 children, but they asserted that transverse measurements such as optic disc diameter should be corrected by magni¯cation effect.However, other studies conducted in adults 4,5 and children 6,7 found that AL [4][5][6][7] and refractive error 5,6,8 in°uence the measurements of OCT.The studies mentioned earlier [4][5][6][7][8] referred that AL-related ocular magni¯cation should be taken into account to ensure better accuracy in the measurements made with OCT.Previous studies used time domain OCT (Stratus OCT); however, spectral domain (SD-OCT) such as Cirrus HD-OCT o®ers a higher axial resolution and scanning speed than conventional time-domain techniques. 9These OCT devices (Stratus and Cirrus OCT) did not apply the correction for magni¯cation e®ect.As referred Ctori et al., 10 ocular magni¯cation of retinal images is a®ected by refractive error, corneal curvature, refractive index, AL and anterior chamber depth as well as the distance from the eye.Spectralis SD-OCT employs an automatic modi¯cation process in order to neutralize the e®ect of ocular magni¯cation. 10Ocular magni¯cation is a factor of disagreement between both devices (Spectralis SD-OCT and Cirrus OCT) because of the scan path used.Spectralis SD-OCT uses a circular scan path, and Cirrus OCT interpolates the scan from a 200 Â 200 A-scan cube centered on the optic nerve. 6im and Chun 8 compared, in a sample of 35 children younger than 10 years, the peripapillary RNFL thickness, macular thickness and total macular thickness of high myopic eyes with a mean spherical equivalent (SE) of À7.93 AE 1.46 D with those of low myopic eyes with a mean SE of À1.41 AE 1.32 D, concluding that high myopic children had signi¯cantly lower values of overall peripapillary RNFL thickness, macular thickness and macular volumes than low myopic children.In this study, 8 the authors corrected for ocular mag-ni¯cation e®ect.Taş et al. 6 divided a sample of 164 hyperopic children into three groups according to their SE as low, moderate and high hyperopia group and found that the mean RNFL and the RNFL of inferior and nasal quadrants were thicker in children with high hyperopia than in children with low hyperopia.However, these di®erences disappeared after correction of magni¯cation. 6Savini et al. 4 used the magni¯cation correction for evaluating RNFL thickness and ONH parameters (optic-disc area and rim area) in 45 healthy adults (mean age: 39.4 AE 7.2 years), dividing their sample into three groups according to AL as short, medium and long eyes.The authors revealed that AL in°uences measurements of both RNFL thickness and ONH parameters and suggested caution when measurements of myopic and hyperopic eyes were compared with the normative database of the instrument. 4uynh et al. 2 corrected in 1309 6-year-old children transverse disc OCT measurements for mag-ni¯cation and concluded that AL appears to have a stronger e®ect on disc and rim area than the refraction.All of the aforementioned studies utilized Stratus-OCT to take the measurements.As previously reported by Savini et al., 4 the Cirrus HD-OCT and Stratus-OCT should behave similarly, when AL-induced ocular magni¯cation is accounted for.
The aim of this study was to evaluate the distribution of macular and RNFL thickness and ONH parameters of myopic and hyperopic eyes in comparison with emmetropic control eyes and to investigate the variations of these parameters according to AL and refractive error in healthy Caucasian children.Measurements were performed with SD-OCT and corrected for magni¯cation.To our knowledge, there are a small number of studies, which analyze the impact of both AL and refraction on OCT measurements in Caucasian children.
both eyes, and subjects were divided into three groups: 99 emmetropic eyes as a control group (45 boys and 54 girls), 100 myopic eyes (46 boys and 54 girls) and 94 hyperopic eyes (54 boys and 40 girls).Emmetropia was de¯ned as a cycloplegic SE between þ0.75 and À0.25 D. 12 The myopic group was further categorized into three subgroups according to the SE: low myopia (SE between À0.50 and À3.00 D), moderate myopia (SE between À3.25 and À6.00 D) and high myopia (SE greater than À6.00 D). 12,13 The hyperopic group was similarly categorized into low (SE between þ1.00 and þ3.00 D), moderate (SE between þ3.25 and þ6.00 D) and high (SE greater than þ6.00 D) hyperopia.As the e®ect of AL on macular and RNFL thickness and optic disc parameters was investigated, our sample was also divided into three groups according to the magnitude of AL: short (<22.00mm, 68 eyes, 33 boys and 35 girls), medium (from !22.00 mm to 25.00 mm, 189 eyes, 88 boys and 101 girls) and long eyes groups (>25.00 mm, 36 eyes, 19 boys and 17 girls). 14xclusion criteria for the study were a corrected distance visual acuity worse than 20/25 in either eye and a refractive cylinder of more than 2.00 D because the degree of corneal astigmatism in°uences the ONH parameters and peripapillary RNFL thickness measurements by the Cirrus HD-OCT. 15,16Subjects with a previous history of ocular surgery, trauma, pathology or ocular medication and those with tilted optic disc, anisometropia more than 1.00 D or strabismus were also excluded from the study.

Examination protocol
All subjects underwent a comprehensive ocular examination that included visual acuity measurement, stereopsis assessment, motility exam, cycloplegic refraction, anterior segment and dilated fundus examination.Cycloplegia was induced with three drops of cyclopentolate 1% separated by 5 min, to achieve adequate mydriasis (!6 mm). 2 At least 30 min after the last drop, autorefraction was performed with an autorefractometer (Topcon KR-8100P), followed by a subjective refractive re¯nement.
The same experienced examiner (I.B.-G.) performed all of the measurements in both eyes in a random order 15 days after the initial ocular examination in order to avoid any e®ect induced by cycloplegia.Only data from one eye randomly selected in each patient were included in the study.The AL was measured using the IOL-Master system (Version 5.2.1, Carl Zeiss Meditec, Jena, Germany). 17Three consecutive AL readings were taken and averaged.Only the AL measurements with a signal-to-noise ratio greater than 2 were included in the database. 13All measurements were performed without pupil dilation.Intraocular pressure was determined and analyzed in a previous study, 12 and all children included in the analysis were selfreported healthy.
The study followed the tenets of the Declaration of Helsinki, and informed written consent was obtained from the children's legal guardians.

OCT (Cirrus TM HD-OCT) measurements
Measurements of the RNFL thickness, ONH parameters and macular thickness were obtained using an SD Cirrus TM HD-OCT system (Version 5.0.0.326,Carl Zeiss Meditec, Dublin, CA, USA), without cycloplegia.All of the scans were performed by a single experienced examiner (I.B.-G.).Optic disc cube 200 Â 200 protocol was utilized to assess RNFL thickness (average and all four quadrants: superior, nasal inferior and temporal) and ONH measurements.This protocol generates 200 Â 200 cube images with 200 linear scans that are performed by 200 A-scans.Cirrus TM HD-OCT software algorithms automatically detect the center of the optic disc and place a calculation circle of 3.46 mm diameter evenly around it.In each series of scans, average RNFL thickness and RNFL thickness in each quadrant (superior, inferior, temporal and nasal) were analyzed. 4,18The macular images were obtained using a macular cube 512 Â 128 protocol.This protocol produces 128 horizontal scans at high resolution (512 A-scans per B-scan).RT was calculated using the built-in Macular Analysis software on the Cirrus device, which is automatically determined by taking the di®erence between the inner limiting membrane and the retinal pigment epithelium and provides average RT of nine zones including a 1 mm central zone and average macular thickness over a 6 mm scan diameter.Macular volume is determined on the basis of the radius of the circle subtended by the scan lines.The total macular volume corresponds to the sum of the volumes of the neural retina in the central 6 mm of the macula. 19The technique of OCT, 2,20,21 as well as its reproducibility in children, 18 has already been described in several reports.For image acquisition, an internal ¯xation target was used to ensure proper alignment of the eye.Three sequential measurements were taken, and the best centered one with signal strength of !7 was chosen for analysis.Scan quality was checked for every OCT image, and manual correction of the boundary detection was enabled if segmentation errors were present.If an involuntary saccadic eye movement was detected during the scan, it was discarded and repeated. 18,21

Correction for ocular magni¯cation
Cirrus HD-OCT provides printout values of retinal features by taking default AL and refraction, which are set to x 0 ¼ 24:46 mm and 0 D, respectively.However, deviations of those previously de¯ned parameters in the eye that is being scanned can lead to signi¯cant variations in the magni¯cation.Littmann 22 showed that the true length of a transverse retinal measurement can be derived as t ¼ p Á q Á s, where p is the camera constant related to the OCT imaging system, q is a factor that depends on the opto-geometrical characteristics of the given eye and s is the size of the retinal feature taken from the fundus image.Speci¯cally, the factor q may be estimated by the formula q ¼ ðx À 1:82 mmÞ Á 0:01306 /mm 2 , which depends on the AL (x), but neglects further recti¯cations caused by refractive power. 23,24As a consequence, the OCT reading t 0 will be corrected in order to achieve the true size t ¼ M Á t 0 , where the correction M ¼ ðx À 1:82 mm)/(24.46mm-1.82mm).As on-axis measurements of the optical instrument will not be corrected, areas taken from B-scans follow the same correction factor M. Finally, areas inferred from en-face images undergo a two-fold correction that is applied by means of the factor M 2 .The latter also occurs for volumes.
The correction for magni¯cation e®ects on the transverse scale of the OCT data, which is based on the Littman's formula, uses a factor q taken from schematic eyes that are based on prototypal adult ocular dimensions.However, the accurate application of the q factor to scale retinal image sizes on a child population should be subject to a reevaluation that, to the authors' knowledge, has not been reported in the literature so far.In fact, such analysis deserves a comprehensive treatment in virtue of its own relevance that is out of the scope of the present study.Accordingly, the results presented here, founded on the invariance of the factor q, should be considered as an advantageous approach to the unconditionally prerequisite to rectify pediatric OCT data.

Statistical analysis
Statistical analysis was carried out with the commercially available statistical package SPSS version 22.0 (SPSS, Chicago).Kolmogorov-Smirnov tests were used to assess sample distribution, and all variables were normally distributed.Analysis of variance (ANOVA) was used to investigate di®erences in mean values among the groups of children classi¯ed by AL and SE.To determine pairwise di®erences, a Bonferroni post hoc analysis was conducted when the groups were of equal sizes, and the Games-Howell post hoc test was used when the group sizes were unequal.Multiple linear regression model was constructed with RNFL thickness, RNFL thickness in the four quadrants, rim area, disc area and average cup-to-disc ratio as the dependent variables, and AL, SE and age as covariates.We also built multiple linear regression model with average macular thickness, central macular thickness and macular volume as the dependent variables and AL and SE as covariates.AL and SE were analyzed separately as they were highly correlated (r ¼ 0:86).All linear regression models were examined for the presence of multicollinearity (variance in°ation factor and tolerance statistics).Likewise, the independence of error assumption was assessed with the Durbin-Watson test.Residuals were ensured to be normally distributed.All data were found to exhibit homoscedasticity by assessing the plots of standardized predicted values vs studentized residuals.The presence of any sig-ni¯cantly in°uential case was assessed by Cook and Mahalanobis distances.A value of p < 0:05 was considered statistically signi¯cant.

Analysis strati¯ed by SE
The mean (AE SD) of descriptive and OCT parameters obtained from the entire sample and in the control, myopic and hyperopic groups are shown in Table 1(a), as well as the comparisons among them.Table 1(b) shows post hoc analysis to determine pairwise di®erences among groups classi¯ed according to SE refraction.
Statistically signi¯cant di®erences in average RNFL were found between control and myopic groups (p ¼ 0:002, Bonferroni post hoc test) and between myopic and hyperopic groups (p < 0:001).We also obtained statistically signi¯cant di®erences in inferior RNFL thickness between control and myopic groups (p < 0:001) and between myopic and hyperopic groups (p < 0:001) as well as in nasal RNFL thickness between control and myopic groups (p ¼ 0.003) and between myopic and hyperopic groups (p < 0:001).Regarding the macular parameters, statistically sig-ni¯cant di®erences in the three parameters evaluated among the three groups assessed (average macular thickness, central macular thickness and macular volume) were found, as shown in Table 1(a).
Table 2(a) summarizes descriptive and OCT parameters in the control group and in the myopic and hyperopic groups subdivided into three refractive subgroups and the comparisons among them.Tables 2(b)-2(e) show post hoc analysis to determine pairwise di®erences among groups classi-¯ed according to SE refraction subgroups (Games-Howell post hoc analysis).It should be observed that we did not include rim area, disc area and RNFL in temporal quadrant in these tables, because we did not ¯nd pairwise di®erences among them.
Statistically signi¯cant di®erences in average RNFL thickness were found between high myopic subgroup compared with low myopic subgroup, control group and low, moderate and high hyperopic subgroups (p < 0:05 for all comparisons, Games-Howell post hoc test).No signi¯cant di®erences between high and moderate myopic subgroups were found.We also found statistically signi¯cant di®erences in superior and inferior RNFL thickness between high myopic subgroup compared with all subgroups analyzed (p < 0:05 for all comparisons), except for the moderate myopic one.Concerning the nasal RNFL thickness, statistically signi¯cant differences were found between high myopic subgroup compared with control group and moderate and high hyperopic subgroups (p < 0:05 for all comparisons).Statistically signi¯cant di®erences in cup-to-disc ratio (C/D) between high myopic subgroup compared with high hyperopic subgroup (p < 0:05) were obtained.No signi¯cant di®erences between other diagnostic subgroups were found (Tables 2(b)-2(e)).
The results also revealed statistically signi¯cant di®erences in average macular thickness between high myopic subgroup compared with moderate and high hyperopic subgroups (p < 0:05 for all comparisons, Games-Howell post hoc test), as well as in central macular thickness between high myopic subgroup compared with moderate and high hyperopic subgroups (p < 0:05 for all comparisons).Statistically signi¯cant di®erences in macular volume between high myopic subgroup compared with all subgroups analyzed were found (p < 0:05 for all comparisons), except for the moderate myopic one (Tables 2(b)-2(e)).Statistically signi¯cant di®erences in average RNFL thickness and in the thicknesses in inferior and nasal quadrants were found between short eyes group compared with medium and long eyes groups and between medium and long eyes groups (p < 0:05 for all comparisons, Games-Howell post hoc test).There were statistically signi¯cant differences in average macular thickness, central macular thickness and macular volume among short eyes group compared with medium and long eyes groups (p < 0:05 for all comparisons).

Linear regression analysis
Average RNFL thickness (Fig. 1) and the thickness in superior, inferior and nasal quadrants were shown to decrease as the AL increases.Linear regression revealed signi¯cant negative correlations of AL with average RNFL thickness and thicknesses in superior, inferior and nasal quadrants.Consistent relationships of SE with average RNFL thickness and thicknesses in superior, inferior and nasal quadrants were also found.These associations are described in Table 4. Table 5 displays the results of the linear regression analysis of the relationship of AL and SE with macular parameters.AL correlated negatively with average macular thickness (Fig. 2) and macular volume and positively with central macular thickness.In contrast, negative SE was also negatively associated with average macular thickness and macular volume and positively with central macular thickness.depth is governed by the source coherence, which is unaltered by the geometrical magni¯cation inherent to the formation of fundus images.
Our ¯ndings showed that average RNFL thickness of all quadrants, except for the temporal one, were thinner in myopic eyes when compared with hyperopic and emmetropic eyes.On the other hand, we did not ¯nd statistically signi¯cant di®erences among the three groups evaluated in the rim area and disc area.However, we found statistically sig-ni¯cant di®erences between control and myopic groups and between myopic and hyperopic groups in cup-to-disc ratio.When we compared pairwise di®erences among the groups classi¯ed according to AL, we did not ¯nd statistical signi¯cance in rim and disc area parameters.
We also categorized the myopic and hyperopic groups into subgroups according to SE and obtained that the average RNFL thickness and the RNFL thickness in the superior, inferior and nasal quadrants were thinner in high myopic subgroup compared with low myopic subgroup, control group and low, moderate and high hyperopic subgroups.With reference to rim and disc area, our results did not show statistically signi¯cant di®erences in post hoc analysis.These results are in agreement with the results of earlier studies carried out in both adults 26 and children, 8 which revealed signi¯cantly lower values of average peripapillary RNFL thickness in high myopic subjects.In Caucasian children, Barrio-Barrio et al. 27 found a signi¯cant association between RNFL (without correction for magni¯cation) and SE and a tendency toward signi¯cance between AL and RNFL.In our series, there were no signi¯cant di®erences in temporal quadrant thickness among the groups and subgroups evaluated.This would be in agreement with that given by Lim and Chun, 8 who found that temporal RNFL thickness remained una®ected by increasing myopia.
Taş et al. 6 reported signi¯cant di®erences between low and high hyperopia groups regarding the average RNFL thickness and the RNFL thicknesses of inferior and nasal quadrants in hyperopic children.These ¯ndings also are consistent with our data.However, the authors 6 found that di®erences among hyperopia subgroups in RNFL thickness disappeared when the magni¯cation attributable to SE/AL was taken into account.We also investigated the e®ect of AL on OCT measurements.We found that the average RNFL thickness and the thickness in the inferior and nasal quadrants decreased as AL increased.This ¯nding is in agreement with the previous report. 4However, in the present study, associations among ONH parameters and AL or SE after applying the correction for ocular magni¯cation, as other researchers reported, 2,4 were not found.Huynh et al. 2 referred an increasing optic disc area and a decreasing rim area with longer eyes in a sample predominantly hyperopic.Likewise, Savini et al. 4 observed an inverse association between AL and OHN parameters (disc and rim areas).It should be highlighted that before applying the correction for ocular magni¯cation, negative correlations of AL and negative SE with disc and rim areas were found.These associations disappeared when the correction for ocular magni-¯cation was implemented.Our results did not show correlation among disc and rim areas with AL, SE and age, as we reported in Tables 4 and 5.
The results of our study di®er from those obtained in previous studies.We did not apply the Littmann formula to the RNFL thicknesses as it has been justi¯ed above.This could explain the disagreements between our results and those given by other investigators in studies in which the Littmann formula has been applied or has not been correctly implemented. 4,27As other authors suggested, when this formula is applied, the relationships of OCT measurements with other refractive or anatomical parameters might change, except for age. 4,6Other factors that may have accounted for di®erences between our study and previous one would be the diverse races analyzed, the number of subjects, the range of refractive error included in the study and the di®erent devices used to take the measurements.
To our knowledge, there are no studies evaluating the e®ect of both AL and SE on OCT measurements, especially in Caucasian children.To this date, only two previous studies have been conducted in Spanish children evaluating OCT parameters. 25,27There are some di®erences between our study and these previous ones, 25,27 such as the range of age, the amount of refractive error and the parameters included in the analysis.In spite of this, we obtained some results similar to those given by these authors. 25,27Although AL and SE are significantly correlated, the analysis of the e®ect of these parameters on OCT measurements was justi¯ed as the refractive error is the result of the balance among the refractive power of the optical elements of the eye and AL.Therefore, both parameters are not representing exactly the same.
With reference to macular parameters, the longer the eye, the thinner was the average macular thickness and the smaller macular volume.We also found thinner average macular thickness and smaller macular volume in myopic eyes compared with hyperopic and emmetropic eyes.However, central macular thickness was thinner in hyperopic eyes in comparison to myopic and emmetropic eyes.When comparing the refractive subgroups, our results showed that average macular thickness and macular volume decreased as the level of myopia increased.In contrast, we found thinner central macular thickness in shorter and more hyperopic eyes.These results are in agreement with those provided by other researchers in both adults 28 and children. 8Wu et al. 28 reported that RT in individuals with high myopia and long eyes was thicker in the foveola and fovea, but thinner in the inner and the outer macular regions.These authors also showed smaller macular volume in subjects with high myopia. 28Lim and Chun 8 also obtained thinner macular thickness and lower macular volumes in high myopic children.It has been suggested that these relations would be owing to the mechanical globe elongation linked with myopia, resulting in retinal stretching and thinning. 8,29This could explain the decrease in RNFL and macular thickness.Ostrin et al., 30 in a study performed in adults, reported that high myopes tend to have a thinner subfoveal choroid and larger scleral canal.They referred that these relationships might help to explain the increased risk in glaucoma.Our outcomes showing RNFL thinning, average macular thinning, smaller macular volume and central macular thickening in longer and more myopic eyes support this suggestion.It should be noted that there were few highly myopic children in our sample as the population with high myopia is much reduced, especially in white children and therefore less population is required to have a relevant sample. 12Note that in our sample the mean age (AE SD) was 10.8 AE 31 as Foster and Jiang 31 referred; the prevalence rates of high myopia at this age in Caucasian children are less common.
We did not ¯nd age-related changes of the RNFL thicknesses, ONH parameters and macular thickness.In contrast, a positive association was found between central macular thickness and macular volume with age.Barrio et al. 27 reported positive associations between average macular thickness and macular volume with age, but they did not ¯nd correlation between RNFL and age.Huynh et al. 29 also reported thinner central macular thickness in younger children.6][27][28][29][30][31][32][33] It seems that the onset of decrease in the RFNL with age does not appear to be signi¯cant until the age of 15 years. 33We consider, as Huynh et al. 34 suggested, that the growth of macular, RNFL and optic disc parameters is almost complete at birth or soon after it.Therefore, we believe that these correlations are likely to be clinically insigni¯cant.
In the Cirrus HD-OCT, the normative database is not available for patients under 18 years of age.Furthermore, when the database was built, the manufacturers did not take into account the e®ect of AL and refractive error, although several studies have shown the signi¯cant e®ect of both parameters on OCT measurements, 2,4,8,26,28,32 regardless of age.In addition, a magni¯cation correction is not provided by the Cirrus HD-OCT manufacturers.Kang et al. 9 proposed that the e®ect of ocular magni¯cation should be taken into consideration for myopic eyes greater than À4.00 D. Myopia is the most prevalent refractive error in the world, and the number of people with myopia continues to increase worldwide. 35A lot of research is currently being conducted to investigate causes and development of myopia.High levels of myopia are linked with ocular pathologies such as retinal detachment and chorioretinal degeneration risk of developing glaucoma.It is considered very important to follow up changes in retinal parameters in myopic and in glaucoma patients.This is feasible with OCT devices, but it is reported that each instrument results in variability in RT measurements, due to the algorithms used by each OCT.Therefore, we consider that the magni¯cation on lateral measurements 10 such as disc, rim area and macular volume must be corrected for establishing diagnosis and treatment protocols, specially in myopic and glaucoma patients.
Therefore, from our point of view in future improvements of the instrument or when the software is updated, the manufacturers might introduce corrections to take into account the e®ect of related ocular magni¯cation and set up a new database including subjects below 18 years old.
In summary, the average RFNL thickness and the RFNL thickness in the superior, inferior and nasal quadrants as well as the average macular thickness and macular volume decrease as the AL and the level of myopia increase, even considering a correction of the magni¯cation e®ect.Furthermore, the RFNL in the temporal quadrant is thicker in longer eyes and the central macular thickness is thicker in longer and more myopic eyes.No correlations between AL/SE and ONH parameters are present after correction for magni¯cation e®ect.Clinicians should be cautious when eyes with shorter or longer AL are measured and should take into account the e®ect of ocular magni¯cation in order to avoid errors in the interpretation of the data from Cirrus TM HD-OCT, regardless of age.

Table 1a .
Descriptive and OCT parameters in the entire group and in the three groups examined: emmetropic, myopic and hyperopic groups.

Table 3 (
.00 mm to 25.00 mm, 88 boys and 101 girls) and long eyes (>25.00 mm, 19 boys and 17 girls) and the comparisons between them.Table3(b) shows post hoc analysis to determine pairwise di®erences among groups classi¯ed according to AL.
a) shows descriptive and OCT parameters in the three groups examined according to AL: short (<22.00mm,33 boys and 35 girls), medium (from!22

Table 1b .
Post hoc analysis to determine pairwise di®erences among groups classi¯ed according to SE refraction (Bonferroni post hoc analysis).

Table 2a .
Descriptive and OCT parameters in the control group and in the myopic and hyperopic groups subdivided into three refractive subgroups.

Table 2d .
Post hoc analysis to determine pairwise di®erences among groups classi¯ed according to SE refraction subgroups (Games-Howell post hoc

Table 2e .
Post hoc analysis to determine pairwise di®erences among groups classi¯ed according to SE refraction subgroups (Games-Howell post hoc

Table 3a .
Descriptive and OCT parameters in the three groups examined according to AL: short, medium and long eyes.

Table 3b .
Post hoc analysis to determine pairwise di®erences among groups classi¯ed according to AL (Games-Howell post hoc analysis).

Table 4 .
Results of multivariate-mixed model analysis of the association between AL, SE and age and RNFL thickness, RNFL thickness in the four quadrants, rim area, disc area and average cup-to disc-ratio.