Portal de Periódicos da CAPES

Retinal nerve fibre layer thickness in full-term children assessed with Heidelberg retinal tomography and optical coherence tomography: normal values and interocular asymmetry

2009; Wiley; Volume: 89; Issue: 2 Linguagem: Inglês

10.1111/j.1755-3768.2009.01680.x

1755-3768

Eva Larsson, Urban Eriksson, Albert Alm,

Retinal Imaging and Analysis

Purpose: This study aimed to investigate normal values and interocular differences in retinal nerve fibre layer (RNFL) thickness, using optical coherence tomography (OCT) and Heidelberg retinal tomography (HRT), in 5–16-year-old children born at full-term with normal birthweights. Methods: Fifty-six children with normal visual acuity and refraction were examined with Stratus OCT and HRT. Three examinations were performed in each eye. One eye in each child was randomized for analyses of normal values. Findings in 54 eyes were evaluated. Mean values of RNFL thickness were calculated. Coefficients of variance and intraclass correlations were calculated. The correlation between right and left eyes and the limits of difference were determined for both methods. Results: Mean RNFL thickness was 98.4 μm (standard deviation [SD] 7.88 μm) assessed with OCT and 213.0 μm (SD 54.0 μm) assessed with HRT. No correlations between age or gender and RNFL thickness were found. The coefficients of variance were 2.9% and 5.6% for OCT and HRT, respectively, and intraclass correlations were 0.85 and 0.88, respectively. The limits of difference between the two eyes ranged from −9 μm to 9 μm with OCT and from −109 μm to 87 μm with HRT. Conclusions: Both OCT and HRT can be used in children aged 5–16 years, but OCT provides less variability in determinations of RNFL thickness, both in repeated examinations of the same eye and in comparisons between the two eyes. The present study provides values for normal RNFL thickness in healthy children which can be used to make comparisons with values in children with optic nerve diseases. Diagnosing the causes of unexplained reductions in vision in children is often a clinical challenge. Several factors are used to evaluate visual function, such as visual acuity (VA), colour vision and visual fields. However, these examinations require interaction and participation and in some children this can be complicated and results may be difficult to interpret. Determining the magnitude of structural aberrations requires substantial clinical experience and easily obtained objective measures would be of great value. Recent technical developments have resulted in two techniques for imaging the optic disc and the retinal nerve fibre layer (RNFL): scanning laser tomography (Heidelberg retinal tomography [HRT]) and optical coherence tomography (OCT). Both techniques can be used in children as they are easily performed and non-invasive (Eriksson et al. 2004; Martin & Lundvall-Nilsson 2007). They provide objective information of optic disc configuration and/or RNFL thickness and may represent valuable tools in the evaluation of children with subnormal vision and in children with known diseases that may affect the optic nerve, such as craniopharyngeoma and glioma. In children with binocular subnormal vision, a normal database is necessary for purposes of comparison. Similarly, in order to assess children with monocular subnormal vision, we need to know the normal range of differences between the two eyes. Normal databases have been established for both HRT and OCT, but these have been determined in adults and may not be applicable to children. Other authors have recently reported RNFL thickness values using OCT and HRT in normal children (Huynh et al. 2006; Ruberto et al. 2006; Salchow et al. 2006), but no reports have compared the two techniques in children. Recently, we measured macular thickness with OCT in a population-based group of 5–16-year-old children, all of whom had been born at term and with normal birthweights (Eriksson et al. 2008). The aims of the present study were to determine, in the same population of children, normal values of RNFL thickness, assessed with OCT and HRT, the intrasession repeatability of each of these methods and the intraocular difference in RNFL thickness in the individual child. Finally, we aimed to investigate which method would be of most value for RNFL assessments in children. Children aged 5–16 years, born in Uppsala County, Sweden, were randomly selected from the Swedish National Board of Health and Social Welfare database. All of them had been born at term (> 37 weeks gestational age, birthweights>2500 g). The children were located using the 10-digit personal numbers assigned to each member of the population in Sweden and their caregivers were asked by letter whether they wished their child to participate in this study. Best corrected VA was assessed monocularly with linear logMAR charts. Cycloplegic retinoscopy was performed after dilating the pupil with a mixture of phenylephrine 1.5% and cyclopentholate 0.85%. The spherical equivalent was calculated and astigmatism recorded as a negative cylinder. Children with VA ≤ 0.65 (≥ 0.2 logMAR) and/or spherical equivalent > +3 D or < −3 D and/or astigmatism>2 D were excluded. Ophthalmoscopy was performed. The measurements were made through dilated pupils with Stratus OCT 3 (Version 4.0.1; Carl Zeiss Meditec, Inc., Dublin, CA, USA) and HRT-II/III (Heidelberg Engineering GmbH, Heidelberg, Germany). In OCT, tomographic images of the retina are constructed from axial scans by using back-scattered, low-coherence light. The computer algorithm can identify the posterior and anterior borders of the RNFL (Schumann et al. 1995). In the optic nerve head (ONH) program, the peripapillary RNFL thickness was assessed by a circular scan with a diameter of 3.4 mm, centred on the ONH (Fig. 1). Three scans were performed in each eye according to the standard protocol; scans with a signal strength < 5 were excluded. In each eye the RNFL thickness was measured in the superior, temporal, inferior and nasal quadrants and the average RNFL thickness for 360 ° was calculated (Fig. 2). The mean values of the three examinations were calculated by the OCT software (Carl Zeiss Meditec AG 2004). A circular scan with a diameter of 3.4 mm around the optic nerve head was used in the optical coherence tomography program. Optical coherence tomography output for retinal nerve fibre layer (RNFL) thickness in one study subject. The HRT is a confocal scanning laser ophthalmoscope, which uses a diode laser, in which consecutive two-dimensional sections are integrated to produce a three-dimensional image of the ONH (Miglior et al. 2003). A topographic image is automatically obtained from three scans centred on the ONH. The optic disc margin is defined by a contour line drawn by the operator. When the contour line is set, the HRT software calculates a series of variables that describe the morphology of the ONH, including the mean thickness of the RNFL along the contour line (Fig. 3). We made three separate measurements in each eye. The HRT-II was used during the examinations, but the data were transferred for reanalysis by HRT-III software. Only good quality images (standard deviation [SD] < 40) were accepted. Heidelberg retinal tomography output for retinal nerve fibre layer (RNFL) thickness in one study subject. The right eye was examined first with both techniques, applying HRT before OCT. All OCT scans were performed by one of the authors (UE), except for a few that were performed by the experienced technician who carried out all the HRT examinations. In the HRT assessments, the contour line of the optic disc was drawn by one operator, who was very experienced in the procedure. One of the two eyes in each subject was randomized for inclusion in the analyses for our normal database. In order to determine the interocular asymmetry of RNFL thickness, we compared the right and left eyes and calculated the limits of difference. Written consent was obtained from all caregivers before enrolment. The study was approved by the local ethics committee in Uppsala, Sweden. Microsoft excel was used to randomize the eyes for analysis. spss Version 13.0 (SPSS, Inc., Chicago, IL, USA) was used for all other analyses. Kolmogorov–Smirnov test was used to test normal distribution. Pearson's correlation was used for bivariate correlations and one-way anova for comparisons of continuous data. The intrasession coefficient of variance (CV) was calculated by dividing each SD for the repeated measurements by its mean. Intraclass correlations (ICC) were calculated using spss. Repeatability is considered to be good when the CV is close to 0% and the ICC close to 1.0. The difference between right and left eyes was analysed with the paired t-test. The limit of difference was defined as the mean difference between the eyes (right–left eye) ± 1.96 SD. The differences were illustrated in Bland–Altman plots (Bland & Altman 1986). Fifty-six White children, consisting of 28 girls and 28 boys, with a mean age of 10.1 years (SD 3.0, range 5–16 years) were examined. Mean VA was − 0.06 (SD 0.08, range −0.02 to 0.2) logMAR in right eyes and − 0.06 (SD 0.07, range −0.02 to 0.1) logMAR in left eyes. No eye was excluded for poor VA. We excluded one child because of refractive exclusion criteria, a 7-year-old girl who did not co-operate in the OCT examination and a 5-year-old girl who did not co-operate in the HRT assessment. Therefore, after randomization, 54 eyes were examined with each method, of which 53 were analysed with both OCT and HRT. On ophthalmoscopy, all eyes had clear media and normal fundi, except one which had a small retinal naevus. In one child, one of three measurements with OCT was excluded because of low signal strength (< 5). In another child, two of three measurements were excluded as a result of artefacts, although the signal strength was good. Of the three measurements with HRT, one of three examinations was excluded as a result of poor fixation in two children. Five children co-operated in only two of three examinations and one child in only one HRT examination. Three measurements had to be excluded because of low quality (SD>40). Therefore, of the 54 eyes assessed with HRT, we obtained three measurements in 43 eyes, two in nine eyes and only one in two eyes. Normal values, based on the average of three scans (see Materials and Methods), of RNFL thickness in the temporal, superior, nasal and inferior quadrants assessed with OCT are shown in Fig. 4. Mean RNFL thickness was 98.40 μm (SD 7.88, range 85–120 μm). Retinal nerve fibre layer thickness was normally distributed in the four quadrants as well in as the average value. The temporal and nasal RNFL were significantly thinner than the superior and inferior RNFL (p < 0.001), and the temporal RNFL was significantly thinner than the nasal (p = 0.008), but there was no difference in RNFL thickness between the superior and inferior quadrants. We found no correlations between age and RNFL thickness in the four quadrants, or between age and average RNFL thickness. Further, no difference in RNFL thickness between boys and girls was revealed. The repeatability of the three circular OCT scans, expressed as CVs and ICCs, is given in Table 1. Normal values of retinal nerve fibre layer thickness in the temporal, superior, nasal and inferior quadrants in 54 eyes, assessed with optical coherence tomography. The rectangles include 50% of the values; the horizontal line represents the median, and the whiskers minimum and maximum values. The correlation between right and left eyes was 0.83 (p < 0.01) and was analysed in 53 pair of eyes. The mean values for right (98.5 μm [SD 7.9]) and left (98.3 μm [SD 7.6]) eyes did not differ statistically (p = 0.7). The limits of differences between the eyes are illustrated in Fig. 5. Bland–Altman plot of the interocular difference in mean retinal nerve fibre layer thickness assessed with optical coherence tomography. Mean RNFL thickness assessed with HRT was 213 μm (SD 54, range 130–340 μm) and was normally distributed. No correlation between age and RNFL thickness was noted and no difference between the genders was found. The repeatability of HRT examinations was 5.6% (SD 8.2) expressed as CV and 0.88 (95% confidence interval [CI] 0.82–0.93) expressed as ICC. The correlation between right and left eyes was 0.50 (p < 0.01) and was analysed in 50 pair of eyes. The difference between right (203 μm, SD 56) and left (214 μm, SD 44) eyes was not statistically significant (p = 0.12). The limits of differences between the eyes are illustrated in Fig. 6. Bland–Altman plot of the interocular difference in mean retinal nerve fibre layer thickness assessed with Heidelberg retinal tomography. A weak but statistically significant correlation between average RNFL, assessed with OCT, and mean RNFL, assessed with HRT, was found (r2 = 0.1, p = 0.02) (Fig. 7). The correlation between retinal nerve fibre layer (RNFL) thickness assessed with optical coherence tomography (OCT) and Heidelberg retinal tomography (HRT) in 53 eyes. In the present study, RNFL thickness was determined with OCT and HRT in a population of children, aged 5–16 years, who had been born at term and with normal birthweight. Average RNFL thickness, assessed with OCT, was 98.4 μm and mean RNFL thickness, assessed with HRT, was 213 μm. We found no correlations between age and RNFL thickness or between gender and RNFL thickness using either of these methods. The intrasession repeatability for measuring RNFL thickness, expressed as CV, was 3% with OCT and 6% with HRT, and ICCs were 0.85 and 0.88, respectively. Interocular differences were between −9 μm and 9 μm in OCT and between −109 μm and 87 μm in HRT. Most studies on RNFL thickness assessed with OCT and HRT have been performed in adults. Recently, other authors have also used OCT to measure RNFL thickness in children (Table 2). Although all of these studies were performed in normal children, variations in methodology and epidemiology may have affected the results. For example, different versions of the OCT software were used, and some authors used the standard scan protocol (Repka et al. 2006; Salchow et al. 2006), whereas others used the fast scan protocol (Hess et al. 2005; Huynh et al. 2006; El-Dairi et al. 2007). Most studies were hospital-based (Ahn et al. 2005; Altinas et al. 2005; Hess et al. 2005; Yoon et al. 2005; Kee et al. 2006; Repka et al. 2006; Salchow et al. 2006; El-Dairi et al. 2007). Huynh et al. (2006) performed a large, population-based study in 6-year-old Australian children. Although the present study was population-based, it has some differences with the Australian study, which was carried out in a population with a mixed ethnic composition, whereas the children in our study were all Caucasian, born at full-term, had normal refraction and were aged 5–16 years. These factors may affect the results because RNFL thickness varies in different ethnic groups (Poinoosawmy et al. 1997; Huynh et al. 2006) and a correlation has been shown between RNFL thickness and birthweight (Wang et al. 2006). In accordance with studies in both adults and children, superior and inferior RNFLs were thicker than nasal and temporal RNFLs (Schumann et al. 1995; Huynh et al. 2006; Salchow et al. 2006). This is presumably because a larger number of nerve fibres converge into the optic nerve from the superior and inferior arcuate bundles than from the nasal and temporal retina. The nasal quadrant had a thicker RNFL than the temporal area, as reported by Kee et al. (2006) and Huynh et al. (2006). Several authors have stated that RNFL thickness declines with age in adults (Alamouti & Funk 2003; Parikh et al. 2007; Nagai-Kusuhara et al. 2008). However, we found no correlation between age and RNFL thickness, as assessed with OCT, which agrees with findings in other studies in children (Ahn et al. 2005; Salchow et al. 2006), and suggests that age-related neural loss may not be detectable in children. Therefore, age is not a consideration in comparisons in children. The effects of refraction on RNFL thickness have been debated; some authors have found no difference (Mrugacz et al. 2004) and others have described thinning of the RNFL with greater negative refraction (Choi & Lee 2006; Huynh et al. 2006). This was not examined in the present study because all children had normal refraction. In the present study, RNFL thickness was also assessed with HRT. Ruberto et al. (2006) reported normal values for HRT in children in a hospital-based study and our findings closely resemble theirs: RNFL thickness was 210 μm in right eyes and 240 μm in left eyes in Ruberto et al. (2006), compared with 213 μm in the present study. Recently, Tong et al. (2007) presented a study of HRT in myopic 11–12-year-old Singaporean children, in which the value for untilted discs (mean value 290 μm) was more similar to ours than for the tilted discs. In the present study there were no tilted discs. Furthermore, comparisons with measurements in adults showed RNFL thickness to be about the same (Nakamura et al. 1999). It should be noted that normal values for mean RNFL thickness ranged widely in all these studies, reflecting a large interindividual variation. We found no correlation between age and RNFL thickness assessed with HRT. This was not studied by Ruberto et al. (2006), but several authors have shown such a correlation in adults (Nakamura et al. 1999; Bowd et al. 2002; Nagai-Kusuhara et al. 2008). Nakamura et al. (1999) reported a decrease of 10 μm and Nagai-Kusuhara et al. (2008) a decrease of 20 μm per decade in adults. Such small differences probably indicate that, as in OCT measurements, age need not be considered in comparisons in children. In our study, mean RNFL thicknesses were similar in boys and girls, in accordance with the study by Tong et al. (2007) and with studies in adults (Bowd et al. 2002). This was not analysed by Ruberto et al. (2006). Some studies have shown a correlation with refraction of the eye (Nakamura et al. 1999), whereas others have not (Bowd et al. 2002; Tong et al. 2007). In accordance with OCT assessments, this was not analysed in the present study because all children had normal refraction (mean spherical equivalent 0.6). Both OCT and HRT examinations can be performed quickly, require no contact and are non-invasive. The children co-operated well during the examinations and good quality images were obtained in almost all subjects. Repeatability is defined as the variation in measurements performed by the same person on the same subject under the same conditions and has previously been reported in one study on OCT assessments in children (Wang et al. 2007). To the best of our knowledge, this is the first study of repeatability of HRT measurements in children. In our study, repeatability was good. However, the CV was almost twice as high in HRT as in OCT, reflecting a higher variability in HRT assessments of RNFL thickness. This may be explained by the fact that the observer is required to draw a contour line in HRT. Although we found a correlation between RNFL thickness, assessed with OCT and HRT, it was weak. It should be remembered that OCT and HRT measure RNFL thickness with different techniques and at different locations. Therefore, RNFL values obtained with OCT and HRT cannot be interchanged (Hoffmann et al. 2005; Barakana et al. 2006). One method should be selected and maintained for follow-up purposes. When evaluating diseases that may affect the two eyes differently or one eye alone, it is important to know the normal extent of asymmetry between the eyes. We cannot assume that the two eyes are exactly alike (Brenton et al. 1986; Ong et al. 1999). Therefore, we determined the limits of difference between right and left eyes and found that the interocular difference in RNFL thickness varied from − 8.6 μm to 9.0 μm when assessed with OCT, which is slightly better than the limits reported by Huynh et al. (2007). The limits of difference assessed with HRT varied from −109 μm to 87 μm, representing a much broader interval relative to the mean value of RNFL thickness compared with OCT measurements and thus indicating that a difference in RNFL thickness between the two eyes of about 10% may be clinically significant in OCT, whereas a difference of as much as 50% may be seen in normal eyes with HRT. Gherghel et al. (2000) used HRT and reported significantly thicker RNFLs in left eyes in adults. In children, we found an interocular difference of 0.2 μm using OCT. Using HRT, we found the RNFL in left eyes to be 11 μm thicker than in right eyes, but this was not statistically significant (p = 0.12). In conclusion, both techniques are easily performed in children and have allowed us to provide normal values of RNFL thickness as assessed with the two methods, respectively. Repeatability was slightly worse in HRT assessments, but the main difference between the techniques concerned the much higher variation between the eyes with HRT. Therefore, we conclude that HRT is less useful than OCT for detecting a clinically significant difference in RNFL thickness between the two eyes in one child. It should be stressed, however, that the main use of HRT is to determine the topography of the optic disc. Such data will be reported separately. We thank Eva Nuija and Börje Nordh for their skilful and valuable help. This study was supported by the Crown Princess Margareta Foundation for the Visually Impaired, the Swedish Society for Medicine and the Mayflower Charity Foundation for Children.