Interocular Asymmetry of OCT Retinal Nerve Fiber Layer Values in a Normative Population: The Framingham Heart Study

Abstract

Purpose: To explore interocular asymmetry in retinal nerve fiber layer (RNFL) thickness using optical coherence tomography (OCT), assess factors that predict this asymmetry, and quantify the 95% central range for OCT-defined average cup-disc ratio (CDR).

Methods: Participants from the Framingham Heart Study were included. Interocular differences in OCT parameters were calculated by subtracting left eye values from the right eye. To quantify the range for interocular differences, the percentile distributions were described, with normal ranges established as 2.5th to 97.5th percentiles. Multivariable linear regression models were employed to explore predictors of interocular differences in RNFL thickness.

Results: In total, 522 participants were studied with a mean (standard deviation [SD]) age of 74.5 (6.9) years, and most were female (59.4%) and white (88.3%). The mean (SD) peripapillary RNFL thickness was 88.1 (9.7) in the right eye and 87.7 (9.7) in the left eye, with a nonstatistically significant difference (mean [SD], 0.4 [6.1]; P = 0.18). The 2.5th and 97.5th percentiles for the interocular difference in average RNFL were −12.7 µm and 12.7 µm, while that of average CDR was −0.19 and 0.21, respectively. In multivariable models, only differences in rim area (β = 8.06/mm²; P < 0.001) and differences in signal strength (β = 1.37/unit; P < 0.001) were significantly and positively associated with interocular differences in average RNFL thickness.

Conclusions: The 95% limits for average RNFL and CDR were within 12.7 microns and 0.2 units, respectively, between eyes.

Translational Relevance: Comparing OCT RNFL of both eyes may aid in detecting early cases of unilateral/asymmetric glaucoma or other optic nerve–related pathology when used in conjunction with other clinically or biologically relevant findings.

Introduction

Glaucoma is a major cause of irreversible visual impairment and blindness in the United States, leading to a substantial economic and social burden.¹ Early detection and proper treatment significantly reduce the risk of a patient progressing to visual disability and blindness. Since structural damage to the optic nerve head can precede clinically detectable loss of visual function using perimetry,² assessing the optic nerve head and retinal nerve fiber layer (RNFL) using optical coherence tomography (OCT) is crucial to identifying and treating cases of early disease and ruling out disease in those not needing treatment. 

RNFL thickness, however, can vary widely among healthy subjects, making it more difficult to distinguish glaucomatous eyes from healthy ones. We previously reported on factors associated with participant-level and scan-level RNFL variability in a population-based sample.³ Further, RNFL and other OCT metrics may differ between the eyes of the same subject.^4,5 In fact, an interocular difference in the cup-to-disc ratio (CDR) greater than 0.2, as measured from photographic readings, is a commonly used, but imperfect, diagnostic criterion for glaucoma.^6–11 Previous work showed that OCT measurements correlate well with photographic readings, although less known is how a cutoff of 0.2 translates to OCT, including spectral domain (SD)–OCT.¹² Further, as described by Cameron and colleagues,⁶ there are limited studies exploring interocular asymmetry utilizing the more precise SD-OCT in a population-based sample of older Americans. 

Given the widespread adoption of SD-OCT, it is essential to quantify normative interocular differences in representative populations. These differences are of clinical relevance, as large intereye differences can be a sign of optic nerve disease (which is often asymmetric). Thus, it is important to know the range of asymmetry typically seen in the population and which individuals may have intereye differences for structural or scan-related reasons (suggesting that differences are not from pathology). The population-based, largely normal cohort of older adults in the Framingham Heart Study, combined with imaging from newer OCT devices (Cirrus 6000; Carl Zeiss Meditec, Dublin, CA, USA), provides a unique opportunity to examine interocular differences in RNFL thickness measurements within this well-characterized population. 

This study aims to (1) explore interocular asymmetry in RNFL thickness using SD-OCT in a normative sample of older Americans; (2) assess participant-level, eye-level, and scan-level factors that predict this asymmetry; and (3) quantify the 95% central range for OCT-derived average CDR. 

Methods

Study Design and Participants

This cross-sectional study included participants from the Framingham Heart Study Offspring and Omni Cohorts. The Framingham Heart Study is an ongoing prospective, longitudinal, population-based cohort study that recently incorporated eye exams, including the 200 × 200 Optic Disc Cube SD-OCT. Eye exams were offered to participants who attended their 10th/5th clinic examination cycle (2019–2022). Further details about sampling and participants’ characteristics have been described previously.^3,13,14 

The study was conducted in accordance with the tenets of the Declaration of Helsinki and was approved by the institutional review board of Boston University Medical Center. Informed consent was obtained from all participants. 

Imaging and Quality Assessment

The 200 × 200 Optic Disc Cube Cirrus SD-OCT scans were acquired for both eyes (without dilation) of each participant, with the right eyes imaged first. Image quality was assessed by A.S. and A.C.M. two trained graders not involved in image acquisition and masked to participant demographics. Any differences in quality assessment were adjudicated by A.H.K a senior grader. Images were graded as “Excellent,” “Good,” “Suboptimal,” or “Not Acquired” based on (1) signal strength, (2) motion artifacts, (3) media opacities, (4) focus, and (5) decentration. Specifically, scans rated as “Excellent” had a signal strength ≥7, decentration less than or equal to half of the optic disc diameter, minimal or no motion artifacts that displaced the large vessels by more than one vessel width, and minimal or no media opacities. Scans rated as “Good” had a signal strength ≥7, decentration <1/2 optic disc diameter, few motion artifacts that displaced the large vessels by more than one vessel width, and/or some media opacities. Scans rated as “Suboptimal” had a signal strength <7, were decentered by >1/2 optic disc diameter, had ≥5 motion artifacts that displaced the large vessels by more than one vessel width, and had large media opacities. The following metrics were reported: average peripapillary RNFL; temporal, superior, nasal, and inferior (TSNI) quadrants; clock hours; average and vertical CDR; disc area; rim area; cup volume; and signal strength. 

Further, axial length was measured using the commercially available IOL Master 500 (Carl Zeiss Meditec, Dublin, CA, USA). All measurements showed a signal-to-noise ratio (SNR) >10 and were confirmed by a green signal quality indicator, indicating good SNR, as previously done.¹⁵ Participants with both eyes having an OCT scan graded as “Excellent” or “Good” and complete axial length data were included in the study. 

Covariates

Demographic variables included age, gender, race, and ethnicity. All demographic variables were based on self-report. Further, participants were asked whether they had ever been diagnosed with glaucoma. Participants who answered yes were classified as self-reported glaucoma. 

Statistical Analysis

Descriptive statistics were used to compare the cohort by self-reported glaucoma status. Comparisons were performed using the Pearson χ² test for categorical variables and a t-test for continuous variables. We compared the distribution of interocular differences in RNFL between individuals with and without self-reported glaucoma as well as RNFL thickness values in the worse eye, defined as the lower RNFL thickness value. 

To explore interocular differences in OCT parameters in a normative sample, we excluded adults with self-reported glaucoma. A paired t-test was used to compare the means of OCT parameters (average RNFL, TSNI quadrants, average and vertical CDR, disc area, rim area, and cup volume) between right and left eyes. Interocular differences in those parameters were calculated by subtracting left eye values from the right eye. To quantify the range of interocular differences, the percentile distributions in OCT parameters were described, with normal ranges established as 2.5th and 97.5th percentiles, as previously done.⁴ Pearson correlation coefficients were calculated to assess the correlation between right and left eye values. 

Next, univariable linear regression models were employed to estimate the effect of participant-level (age, gender, race, ethnicity), eye-level (axial length), and scan-level (signal strength, average RNFL thickness, average and vertical CDR, cup volume, disc area, and rim area) variables on interocular differences in RNFL thickness. Variables with P < 0.20 in the univariable models were included in the multivariable analyses. The final multivariable model was refined using backward selection until all variables showed P < 0.05. In sensitivity analyses, we excluded eyes with an average RNFL <70 microns in both eyes as a proxy to exclude those with glaucoma who did not self-report their diagnosis and repeated all analyses. Since axial length measurements influence RNFL values, we repeated the analyses using axial length–corrected RNFL values, calculated using the Littman formula as previously defined.^16–18 Further, we utilized absolute values of the difference across eyes and repeated all analyses. For comparison, we repeated the analyses in adults with self-reported glaucoma as well. The statistical significance level was defined at α = 0.05. All P values were two-sided but not adjusted for multiple analyses. All analyses were conducted using R software (R Foundation for Statistical Computing, Vienna, Austria). 

Results

Out of 1698 offspring/OMNI participants who attended their 10th/5th clinic examination, 1000 underwent OCT testing. Among them, 825 participants had OCT scans graded as “Excellent” or “Good” quality with complete axial length data, and 522 participants had OCT scans acquired from both eyes. As such, 522 participants (1044 eyes) were included (Table 1), with a mean (standard deviation [SD]) age of 74.5 (6.9) years; 310 (59.4%) were female, 461 (88.3%) were white, and 44 (8.4%) self-reported glaucoma. Age and gender did not significantly differ by self-reported glaucoma status (P > 0.05). Race/ethnicity was not compared statistically across glaucoma status due to the limited number of participants in the self-reported glaucoma group. Adults with self-reported glaucoma showed lower average RNFL thickness values in the worse eye (mean [SD], 74.0 [11.5] µm vs. 85.6 [9.3] µm) and a larger interocular RNFL difference compared to those without self-reported glaucoma (Fig.). 

Table 1.

 

Demographics and Characteristics of Included Study Participants by Self-Reported Glaucoma Status

 

Figure.

 

Distribution of the difference in interocular average RNFL thickness by self-reported glaucoma status. Figure includes participants from the Framingham Heart Study who had OCT scans for both eyes and complete axial length data.

Next, we limited our sample to adults without self-reported glaucoma (n = 453) to explore interocular differences in OCT parameters in a normative sample. The mean (SD) of the average RNFL was 88.06 (9.65) in the right eye and 87.67 (9.74) in the left eye. The mean (SD) intraocular difference was not significantly different across eyes (0.39 [6.12]; P = 0.18). The magnitude of interocular differences in quadrants and clock hours was larger and more variable (wider SD) compared to the average RNFL (Table 2). The differences in average/vertical CDR, disc area, cup volume, and rim area between eyes were small, ranging from 0 to 0.02. 

Table 2.

 

Means of OCT Parameter Values of Right and Left Eyes, Interocular Differences (Right Eye Minus Left Eye), and Intereye Correlation

 

Among the OCT metrics, cup volume showed the highest correlation coefficient (r = 0.87; P < 0.001). Average RNFL thickness was also significantly correlated in both eyes (r = 0.80; P < 0.001), with the remaining OCT parameters showing significant (P < 0.001 for all) correlations between 0.48 and 0.84 (Table 2). 

The 2.5th and 97.5th percentiles for the interocular difference in average RNFL thickness were −12.7 µm and 12.7 µm, while that of average CDR was −0.19 and 0.21, respectively. The percentile distributions of interocular differences for the remaining OCT parameters are further described in Table 3. The normal ranges for interocular differences in quadrants and clock hours were wider than those for average RNFL. 

Table 3.

 

Percentile Distribution of Interocular Differences (Right Eye Minus Left Eye) in OCT Parameter Values

 

In separate univariable models (Table 4), interocular differences in average CDR (β = −7.36/unit; P = 0.01), differences in vertical CDR (β = −7.07/unit; P = 0.01), differences in rim area (β = 7.59/mm²; P <.001), differences in axial length (β = −1.53/mm; P = 0.04), and differences in signal strength (β = 1.30/unit; P < 0.001) were significantly associated with interocular differences in average RNFL thickness. In multivariable models, only differences in rim area (β = 8.06/mm²; P < 0.001) and differences in signal strength (β = 1.37/unit; P < 0.001) were significantly and positively associated with interocular differences in average RNFL thickness (R² = 0.08). 

Table 4.

 

Univariable Linear Regression Models Exploring Factors Associated with Interocular Difference (Right Eye Minus Left Eye) in Average RNFL Thickness

 

When examining asymmetry in adults with self-reported glaucoma, we included 88 eyes from 44 participants. Adults with self-reported glaucoma showed a mean (SD) interocular difference of 2.36 (9.47) µm, with a correlation of 0.68 and 95% of the values between −11.9 µm and 25.4 µm. In multivariable models, only differences in rim area (β = 17.22/mm²; P = 0.001) and Hispanic ethnicity (β = 13.05 vs. not Hispanic; P = 0.009) were significantly and positively associated with interocular differences in average RNFL thickness in adults with self-reported glaucoma (R² = 0.30). 

In sensitivity analyses, we examined 858 eyes from 429 participants, which had an average RNFL thickness ≥70 microns in both eyes and without self-reported glaucoma status (Supplementary Table S1). The 2.5th and 97.5th percentiles for the interocular difference in average RNFL thickness were −13.0 and 12.3 microns, while that of average CDR was −0.18 and 0.21 units, respectively (Supplementary Table S2). The OCT parameters showed similar correlations between eyes. The multivariable analysis in this subpopulation also showed that only differences in rim area (β = 7.90/mm²; P < 0.001) and differences in signal strength (β = 1.35/unit; P < 0.001) were significant predictors of interocular difference in average RNFL thickness (Supplementary Table S3). When using axial length–corrected RNFL, the 95% limits were −12.5 and 12.2 microns, with a correlation coefficient of 0.81. In multivariable models, greater differences in rim area (β = 7.42/mm²; P < 0.001), axial length (β = 2.64/mm; P < 0.001), and signal strength (β = 1.29/unit; P < 0.001) were significant predictors of interocular difference in axial length–corrected RNFL thickness. Using absolute values, the mean (SD) absolute difference between eyes was 4.45 (4.21), with 95% of the values between 0 and 16.7 µm. In multivariable models using absolute values, absolute difference in rim area (β = 4.22/mm²; P = 0.02) was significantly associated with absolute interocular difference in RNFL. Among participants with self-reported glaucoma, the mean (SD) absolute difference between eyes was 7.14 (6.58), with 95% of the values between 1 and 25.6 µm. In multivariable models using absolute values, absolute difference in rim area (β = 13.05/mm²; P = 0.02) and Hispanic ethnicity (β = 9.40 vs. not Hispanic; P = 0.01) were significantly associated with absolute interocular difference in RNFL in participants with self-reported glaucoma. 

Discussion

In this cross-sectional study of older Americans, we define norms for interocular differences on peripapillary OCT findings. For example, the mean (SD) difference in peripapillary RNFL between eyes was 0.39 (6.12) microns, with 95% of the values within 12.7 microns, while the mean (SD) difference in CDR between the two eyes was 0.01 (0.10), with the 95% central range for the difference in OCT-defined CDR between −0.19 and 0.21. Interocular differences in average RNFL were not related to age, gender, race, or axial length, although they were greater with larger differences in rim area (β = 7.59/ mm²; P < 0.001). Given that glaucoma may manifest asymmetrically, these values may provide valuable clinical reference to alert clinicians when the presence of unilateral/asymmetric disease should or should not be suspected. 

We quantified the 95% limits for the interocular difference in average RNFL to be within 12.7 microns, while those for the average CDR ranged between −0.19 and 0.21. This suggests that, in a normal individual, RNFL thickness and CDR should not differ by more than 12.7 microns and 0.2 units, respectively, between eyes. This finding agrees with previous work by Budenz,¹⁹ who studied the interocular difference in RNFL measured with time-domain OCT in 108 participants, with a mean (SD) age of 46 (15) years, and concluded that average RNFL thickness should not differ by more than 9 to 12 microns between eyes of normal individuals. Similarly, an interocular difference in the CDR greater than 0.2 has been commonly used as a diagnostic criterion for glaucoma, and our findings suggest that this cutoff may translate to SD-OCT. Overall, our findings suggest that these values may be generalized to a population of older Americans (mean [SD] age of 74 [7] years) and may aid in detecting early cases of glaucoma when used in conjunction with other suspicious findings. For example, eyes with OCT RNFL thickness differences greater than 13 microns may have glaucoma in one eye, especially if there is other corroborating evidence (family history of glaucoma, or signs of glaucoma or higher IOP in an eye with lower RNFL thickness). 

Our study also showed that the average RNFL thickness and all other OCT metrics were significantly correlated between both eyes, with the correlation decreasing when analyzed by quadrant and clock hours. Our findings align with previous work by Budenz,¹⁹ which suggested that the symmetry of RNFL thickness measurements in quadrants and clock hours is relatively low and varies by location. Their lower symmetry may make them less useful clinically as compared to the average RNFL thickness. Still, since asymmetry could be a marker for disease, it is important to know what a normal sectoral difference is. 

When exploring predictors of interocular difference, only differences in rim area (β = 7.59 microns/mm²; P < 0.001) were significantly and positively associated with interocular differences in average RNFL thickness. Also, as expected, RNFL decreased with age in both right and left eyes, although interocular difference in RNFL was not associated with age (i.e., age-related RNFL loss is the same between eyes). These findings are somewhat at odds with previous studies. Mwanza et al.⁴ assessed the interocular difference in RNFL measured with the Cirrus HD-OCT (Carl Zeiss Meditec) in 284 normal participants, with a mean (SD) age of 46 (17) years, and showed that differences in axial length (β = −0.21 µm/mm; P = 0.001) and disc area (β = 0.17 µm/mm²; P < 0.001) predicted interocular differences in RNFL thickness. Jacobsen et al.²⁰ explored asymmetry in RNFL measured with the Spectralis SD-OCT (Heidelberg Engineering, Heidelberg, Germany) in 105 healthy Caucasian adults, with a mean (SD) age of 29 (8) years, and found that older age (β = 0.04 µm/year) and male sex (β = 0.54 µm; vs. female) were significantly associated with greater asymmetry. This association with age may be attributed to glaucoma, as the disease (and its asymmetry) is associated with older age. The discrepancy in results, as compared to prior studies, could be due to the different participant-level variables (such as age or ethnicity), different OCT devices used, correction of ocular magnification, or different imaging protocols. Overall, it is important to consider variables that could result in physiologic versus pathologic RNFL asymmetry. 

Further, adults with self-reported glaucoma displayed larger interocular RNFL differences compared to those without self-reported glaucoma, with a mean (SD) difference of 2.36 (9.47) µm and 95% of the values between −11.9 µm and 25.4 µm. Greater differences in rim area and Hispanic ethnicity were significantly associated with greater interocular differences in average RNFL thickness in adults with self-reported glaucoma. Previous studies have shown that Hispanics tend to have thicker RNFL values compared to other populations.^21,22 In this context, any glaucomatous damage may produce more noticeable asymmetry, making Hispanics more likely to exhibit greater differences in RNFL thickness. However, due to the relatively small numbers of Hispanic participants, these findings should be interpreted with caution. 

Our findings have several implications. First, the 95% limit for asymmetry in RNFL and CDR may aid in detecting early cases of glaucoma when used in conjunction with other suspicious findings, after accounting for factors that could result in physiologic asymmetry. Given our unique population-based cohort, we demonstrated that these previously suggested criteria could be effectively applied to older Americans. Notably, RNFL asymmetry may differ by laterality and glaucoma severity. For example, the utility of the cutoff values described earlier may decrease in patients with bilateral disease or those with more severe stages of glaucoma (due to the floor effect). Second, we described factors that could result in RNFL asymmetry in older Americans. Understanding this variation in asymmetry is important to distinguish physiologic from pathologic changes. Third, several studies randomly select a single eye or measure only one eye, assuming it to be representative of both eyes. As described by Cameron et al.⁶ and as shown in our findings, this approach may lead to misleading results, especially when utilizing sectoral RNFL data. 

Our findings should be interpreted within the context of several limitations. First, most participants in our cohort were non-Hispanic whites, limiting our ability to compare asymmetry by ethnicity. Second, we did not have ophthalmic clinical data, such as intraocular pressure or visual field. Fundus photos were not available for this cohort, and we were unable to rule out other ocular morbidities that may affect RNFL or the optic disc. It is possible that some participants had optic nerve disease but did not self-report it.²³ However, we ran sensitivity analyses excluding participants with an average RNFL thickness below 70 microns, and results were similar. Third, we acknowledge the possibility of residual confounding that might affect our findings, including measurement and segmentation errors and the presence of nonocular diseases. 

Conclusions

In a population-based sample of older Americans, the 95% limits for average RNFL and CDR were within 12.7 microns and 0.2 units, respectively, between eyes. Interocular differences in average RNFL were not related to age, gender, race, axial length, or signal strength, although they were greater with larger differences in rim area. These values may aid in detecting early cases of glaucoma when used in conjunction with other suspicious findings. 

Acknowledgments

The authors thank Jared Zucker for his contribution to the data collection of the study. 

Supported by R01AG066524 (AHK, SS), FHS contract 75N92019D00031, and P30AG066546. The sponsor or funding organization had no role in the design or conduct of this research. 

Funding by A.H. Kashani, Boone Pickens Professorship in Ophthalmology, Pooled Professors Fund and Unrestricted Departmental Funding from Research to Prevent Blindness. 

Disclosure: L. Almidani, None; J. Sabharwal, None; A. Shahidzadeh, None; A.C. Martinez, None; S.J. Ting, None; B. Vaidya, None; X. Jiang, None; T. Kowalczyk, None; A. Beiser, None; L. Sobrin, None; S. Seshadri, None; P. Ramulu, None; A.H. Kashani, Carl Zeiss Meditec (F)