The Physiological Variation of the Retinal Nerve Fiber Layer Thickness and Macular Volume in Humans as Assessed by Spectral Domain–Optical Coherence Tomography

Abstract

Purpose.: With the introduction of spectral domain–optical coherence tomography (SD-OCT), changes in retinal nerve fiber layer (RNFL) thickness and macular volume (MV) can be detected with high precision. The aim of this study was to determine whether there is a physiological quantifiable degree of variation of these structures in humans.

Methods.: This study took place during a 10-km charity run at VU University Medical Center Amsterdam. Weight, height, hydration status, RNFL thickness (ring scan, 12° around the optic nerve head), and MV (20° × 20°) were assessed in 69 subjects (44 runners, 25 controls) using SD-OCT with eye-tracking function. The SD-OCT scans were assessed before running (normal status), after running (more dehydrated status), and 1 to 1.5 hours after finishing the run (rehydrated status). Controls were measured at the same time intervals as the runners but did not participate in the running event. Changes over time were assessed by general linear models, correcting for repeated measurements.

Results.: In runners, a significant increase in both RNFL thickness (94.4 μm [baseline] to 95.2 μm [rehydration], P = 0.04) and MV (288.9 μm [baseline] to 291.0 μm [rehydration], P < 0.001) over time was observed. Controls did not show significant changes over time. Anatomically, the physiological change of RNFL thickness was most marked in the nasal sectors.

Conclusions.: This prospective study demonstrated a significant physiological variation of the RNFL thickness and MV at a proportion that, on an individual patient level, may be relevant for longitudinal studies in neurodegenerative diseases.

Neurodegeneration is a major problem for a range of neurologic diseases, and there is a lack of techniques that can readily quantify the process. Recently, there has been wide interest in the hypothesis that analysis of the retinal nerve fiber layer (RNFL) could act as a marker for neurodegenerative processes in the central nervous system. ^{1 –4} Research in this field was largely driven by the use of time-domain–optical coherence tomography (TD-OCT) in patients with optic neuritis (ON) and multiple sclerosis (MS). ^5,6 A consistent finding is that loss of RNFL occurs in the range of 20 μm after ON and 7 μm in MS without ON. ⁶ The estimated annual loss of RNFL thickness in MS patients without ON is approximately 2 μm compared to 0.1 μm in controls. ⁷  

OCT is a noninvasive technique that can generate a cross-sectional image of the RNFL by scanning the retina with a beam of light. The interference patterns of the back-scattered light are used to generate a high resolution image. ^8,9 In addition to RNFL thickness, which measures unmyelinated retinal axons that are continuous with the optic nerve, OCT technology allows for volume scans, which provide information about both the axons and their ganglion cell bodies. ^2,10 In the macula, the ganglion cell bodies account for approximately 34% of the total macular volume (MV). Thus far, OCT has been successfully used to monitor retinal ganglion cell axon loss in a number of ophthalmological diseases, ^{11 –16} but recent reports emphasize the usefulness of this technique in neurology. ^{2,4,17 –21} This is due in part to some advantages OCT has in a routine clinical setting; it is relatively easy to perform, time-efficient, and requires fewer resources than magnetic resonance imaging (MRI) based assessments of brain atrophy. ²²  

High-resolution SD-OCT makes it possible to detect changes in RNFL thickness with high accuracy (1.14–2.39 μm). ²³ For comparison, the diameter of retinal ganglion cells is approximately 15 μm. ²⁴ It may, therefore, be possible to reliably detect changes in RNFL thickness related to physiological factors at a cellular level. 

Consequently, it is important to assess this effect because physiological variation in the range of the anticipated annual change of RNFL thickness (0.1–2 μm) may influence data interpretation from longitudinal studies on neurodegeneration. Hence, the objective of this exploratory study was to determine whether physiological variation can cause changes in RNFL thickness and MV detectable by SD-OCT. 

Methods

This study was approved by the medical ethics committee (protocol number 2010/336) and the scientific research committee (protocol number CWO/10-22E) of the VU University Medical Center in Amsterdam, The Netherlands. 

Study Design

The study design was prospective with longitudinal data assessment. The study took place at the VU University Medical Center Amsterdam on November 23, 2010. The setting was a 10-km charity run to support MS research. The more than 800 participants of the charity run were informed about the possibility to participate in this study at time of subscription. Included subjects were measured three times and were first scanned before the start of the charity run, which represented a “normal ” situation. Thereafter, the subject started his or her 10-km run, and within 5 to 10 minutes after the subject had finished the second scan was performed. All subjects were advised to refrain from drinking during the race and before the second scan to attain a more dehydrated status. Afterward, the subjects were advised to drink water and sports drinks to rehydrate. The third and last scan was performed 1.5 hours after rehydration. Control subjects were also measured three times, at the same time intervals as the runners, but did not participate in the running event and did not perform any other strenuous exercise on the day of scanning. 

Subjects

All subjects included were in good general health. There were 44 participants of the charity run and 25 controls included. Because of limited scanning time on the day of the event, only one eye of each runner and only six controls were scanned on the day of the charity run. Of the remaining 19 controls, one eye was scanned on subsequent days, with strict adherence to the predefined time intervals. Both runners and controls were eligible for inclusion in this study if they were 18 to 60 years of age. Exclusion criteria were any ophthalmologic or neurologic conditions or myopia of more than −6.0 diopters. The study adhered to the tenets of the Declaration of Helsinki, and informed written consent was obtained from all included runners and controls before study entry. 

Optical Coherence Tomography

OCT images were acquired with SD-OCT (Spectralis software version 1.1.6.3; Heidelberg Engineering Inc., Heidelberg, Germany) using dual-beam simultaneous imaging, with the eye-tracking function enabled. Baseline and follow-up scans were co-registered to allow for optimal anatomic alignment of the retina. 

The accuracy for repeated RNFL thickness measurements of the SD-OCT (Spectralis) used in this study is excellent (coefficient of variation [CV], 0.42%–1.45%; intraclass correlation coefficient [ICC], 0.95). ^3,23,25,26 Similarly, the reproducibility of MV measurements gives a CV below 0.5% and an ICC of 0.96. ^{27 –29}  

A circular scan with a 3.4-mm diameter circle (12°) was used. The ring scan was centered manually on the optic nerve head (ONH) after the eye-tracking function in each subject was activated. Similarly, the MV scan (20° × 20°, 25 sections) was manually centered over the macula after activation of the eye-tracking function. 

Quantitative OCT data collected were global mean RNFL thickness around the ONH and sector-specific RNFL thickness (temporal, temporal/inferior, temporal/superior, nasal, nasal/inferior, nasal/superior, and papillo-macula bundle; Fig. 1B). The volume scan provides volume data (mm³) and mean retinal thickness (μm) per quadrant of the macular area of the retina (Fig. 1F). In this study the global mean of the circle scan and the average thickness in micrometers of the inner area of the volume scan were included in the analyses, thus avoiding interpolation from the data. Fixation control during retinal OCT was achieved using an internal fixation point. All OCT scans at the charity run were performed by one trained observer (AP) to minimize analytical error. 

Figure 1.

 

The two scan sequences used in this study. (A) A circular scan centered on the ONH. The ring scan starts at the arrowhead seen to the right of the circle (nasal sector). For orientation, a small black line is inserted (asterisk). (B) The data are expressed in micrometers of RNFL thickness for each retinal sector and the papillo-macula bundle (PMB). (C) Cross-sectional image of RNFL (B-scan, averaged from approximately 200 A-scans). The vertical gray line corresponds to the position of the asterisk shown in (A), next to the superior temporal retinal artery and vein. (D) RNFL thickness is measured across 768 locations. Baseline data are shown by a closed black line. Normative data are shown in shades of gray, where the whitish layer in the center corresponds to the 95% normal range, and the dark gray and lighter gray areas represent values outside the 99% confidence interval of the normal distribution, indicating outside normal limits. The vertical gray line corresponds to the position of the asterisks shown in (A) and (C). (E) A volume scan, centered on the macula. The scan consists of multiple aligned scans of which the black arrow indicates one. For orientation, a small black line is inserted (asterisk). (F) Circles on the plots represent 1-, 3-, and 6-mm scan diameters. The innermost circle defines the fovea, an area with few retinal ganglion cells. Data in the inner sector and in the inner and outer quadrants are expressed in macular thickness (μm) and MV (mm³). (G) A cross-sectional image of macula in which the vertical gray line corresponds to the position of the asterisk shown in (E). (H) Three-dimensional image of the macula, visualizing the retinal layers. The black line corresponds to the position of the asterisk shown in (E) and (G).

Quality Control

All OCT scans underwent consensus quality control by two trained observers (LB, AP). Criteria for scan rejection were as published ³⁰ : poor quality (signal strength, <15 dB), algorithm line failures, decentration artifacts, and boundary line errors. Scans failing quality control were excluded from the analyses. 

Physiological Parameters

Demographic data (ethnicity, age, gender) and physiological data (weight, height, hydration status) were collected. For ethical reasons, each subject's hydration status was assessed noninvasively (AP) using validated clinical signs (dryness of the tongue or mucosa, skin turgor, capillary refill time) ³¹ and by calculating the change in weight (Δ weight = weight before run − weight after run). All runners were asked to go to the toilet before the first weight assessment to ensure that any weight loss before or after the run was caused by perspiration. 

Statistical Analysis

Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m²). It was meant to categorize the subjects into normally hydrated and dehydrated based on clinical signs (dryness of the tongue or mucosa, skin turgor, capillary refill time) ³¹ and weight loss. 

Baseline characteristics were compared between runners and controls using independent-samples t-tests (continuous variables) and χ² tests (dichotomous variables). Given that RNFL thickness and MV were measured three times in every subject, longitudinal data analyses methods were used because these methods take into account the fact that repeated observations of each subject are correlated. To assess whether there was a difference in RNFL thickness and MV among the three longitudinal measurements and to determine whether these changes over time were different between runners and controls, a general linear model for repeated measures was used. In figures, the data for the runners and controls was normalized to baseline (100%) for ease of visual comparison. The mean relative changes of the groups were based on means of individual changes that were, for the circular scan, calculated for each of the 768 A-scans. 

In addition to the analyses in which the global mean of the circular scan was used to describe changes over time, analyses were performed with the extended data of the circular scan. Instead of sector-specific RNFL thickness values, all individual scan points were used. A circular scan consisted of 768 consecutive scan points, together forming a circle around the ONH, as shown in Figure 1A (gray circle). Analyzing all scan points might give a more detailed depiction of the anatomic location at which changes in RNFL thickness occur and might make it possible to relate these changes to the underlying anatomy. All 768 scan points were equally divided in 12 clockwise sectors. Subsequently, linear multilevel analyses were performed, including interactions on sector level, to determine differences between runners and controls for every retinal sector. 

Moreover, assuming weight loss as a measure of hydration status, the effect of hydration status was explored in greater detail. To determine the impact of weight loss from perspiration and evaporation, sensitivity analyses were performed. Therefore, all analyses were repeated after excluding subjects in the lowest tertile of weight change to determine whether this changed the results. 

Statistical analysis was performed using statistical software (SPSS 17.0; SPSS Inc., Chicago, IL), but multilevel analyses were performed using a different program (MLwiN 2.22; University of Bristol, Center for Multilevel Modeling, Bristol, UK). A significance level of 5% (two-sided) was assumed. 

Results

Of all subjects assessed for eligibility (both runners and controls, n = 91), 18 were excluded before the first scan because of eye disease (n = 3), age older than 60 (n = 9), or showing up after the run had already started (n = 6). All included runners finished the 10-km charity run. None of the runners was, however, dehydrated. In fact, skin turgor and capillary refill time were suggestive of a hyperemic state. Runners lost on average 1.03 kg during running (range, 0.4–2.4 kg). 

Of all scanned subjects (n = 73), four were excluded; their scans failed quality control because of decentration artifacts (n = 1) or algorithm line failure (n = 3). Figure 2 summarizes the flow of subject inclusion and exclusion. 

Figure 2.

 

Flowchart of subjects recruited in this study. Left: inclusion, exclusion, and scan data of runners. Right: inclusion, exclusion, and scan data of controls. In total, 44 runners and 25 controls were included for statistical analysis.

Changes over Time and Difference between Runners and Controls

Table 1 shows the baseline characteristics of the runners and controls. The groups were matched for age, height, weight, and BMI (P > 0.05 for all comparisons). There were more men in the runners group than in the control group (P = 0.01). Mean RNFL thickness and MV between both groups were comparable at baseline (P > 0.05 for both comparisons). 

Table 1.

 

Physiological Characteristics of All Subjects at Baseline

	Runners (n = 44)	Controls (n = 25)	P
Male, %	77	44	0.01
Age, y	42.3 (10.6)	38.1 (13.3)	0.19
Height, m	1.80 (0.1)	1.74 (0.2)	0.07
Weight, kg	78.5 (12.7)	75.3 (11.2)	0.30
BMI	24.0 (2.8)	23.8 (3.5)	0.77
Mean RNFL thickness baseline scan, μm	94.4 (10.2)	96.9 (9.7)	0.40
Inner macular volume baseline scan, μm	288.9 (20.1)	279.6 (17.1)	0.06

 Except for gender (percent), values shown are mean (SD).

There was a significant change from baseline RNFL thickness (P = 0.04) and MV (P = 0.00,001) in the runners but not in the controls (Table 2). Of note, the significant physiological change in MV in the runners occurred mainly after oral rehydration. 

Table 2.

 

Physiological Change in RNFL Thickness and MV after Exercise and Oral Rehydration

	Scan 1 (before run)	Scan 2 (after run)	Scan 3 (after hydration)	P *
Runners
Mean RNFL thickness, μm	94.4 (10.2)	94.9 (10.5)	95.2 (10.4)	0.04
Inner MV, μm	288.9 (20.1)	289.1 (19.9)	291.0 (20.3)	0.00001
Controls
Mean RNFL thickness, μm	96.9 (9.7)	97.1 (9.9)	97.1 (9.7)	0.56
Inner MV, μm	279.6 (17.1)	279.4 (16.5)	279.4 (16.8)	0.94

 Values shown are mean (SD).

*  General linear model correcting for repeated measures.

In addition to the group-specific changes over time on absolute data presented in Table 2, comparisons in relative development over time between runners and controls are shown in Figures 3A (circle scan) and 3B (volume scan). To this purpose, individual scans were normalized to baseline (100%), with physiological deviation from baseline expressed as relative values. Each subject was taken as his or her own control. 

Figure 3.

 

(A) Relative physiological change in RNFL thickness over time. The individual scans were normalized to the baseline value (100%). Changes over time were presented as relative to the individual baseline value (for absolute data and standard deviations, see Tables 1 and 2). The range of expected noise between repeated OCT measurements was expressed as the measurement accuracy (also referred to as the CV) above and below the 100% mark (gray shaded area). The vertical black and gray bars represent the (scan-specific) 95% confidence intervals. In the control group, repeated measurements remain within the noise level (gray shaded area) and did not change significantly over time (NS; P = 0.56). In contrast, for the runners there was a significant (P = 0.04) increase in RNFL thickness over time that was clearly above the level of measurement noise. (B) Relative physiological change of macular volume over time. Individual scans were normalized to the baseline value (100%). Changes over time were presented as relative to the individual baseline value (for absolute data and standard deviations, see Tables 1 and 2). The range of expected noise between repeated OCT measurements was expressed as the measurement accuracy (also referred to as the CV) above and below the 100% mark (gray shaded area). The vertical black and gray bars represent the (scan-specific) 95% confidence intervals. In the runners group, a significant increase in RNFL thickness over time (P = 0.00001) was observed, clearly above the noise level at scan 3. In contrast, for the controls there was no significant change over time (P = 0.94), and values remained within the noise level.

There was no significant difference in relative ring scan data between the two groups (Fig. 3A; P = 0.17), despite the significant increase over time in runners (Fig. 3A; P = 0.04). The significant physiological change over time in the runners was further highlighted by the clear increase in RNFL thickness above the published measurement noise (CV) of 0.42% ³ (Fig. 3A, gray shaded area). 

The physiological changes became even more noticeable when relative volume data were analyzed. There were significant differences between runners and controls in MV data (Fig. 3B; P = 0.003). As discussed, the physiological changes of the runners' MV was clearly above the published measurement noise level (0.45% ^{27 –29} ; Fig. 3B, gray shaded area). 

Anatomic Presentation of Physiological RNFL Thickness Changes at ONH

Changes in RNFL thickness of the 768 A-scans around the ONH were related to the underlying anatomy. Multilevel analyses over the entire observation period (±3 hours) demonstrated that the differences between runners and controls were not the same for all 12 sectors (Fig. 4A). Significant differences between the two groups were observed only for the nasal sectors (Fig. 4A, sectors 3–5) and for sector 12. The relationship with the underlying anatomy is illustrated in Figure 4B. As can be appreciated from the standard deviations in Figures 3A and 3B, these changes were not significant at short time intervals but only over the entire observation period. 

Figure 4.

 

(A) The absolute difference between runners and controls (μm), averaged over the entire observation period, is shown across 12 retinal sectors around the ONH. *P < 0.05, significant differences. A positive value means higher RNFL thickness for controls. (B) For orientation, the 12 positions of a clock are projected on the ONH. The sectors are labeled clockwise. Sectors in which the difference between runners and controls was significant are shaded in gray. The sectors 3 to 5 (P < 0.05) correspond to the well-vascularized nasal area of the retina. Sector 12 (P < 0.05) is located superiorly, adjacent to the superior temporal retinal artery and vein.

Sensitivity Analysis

Assuming that subjects who lost more weight had more changes in hydration status, subjects in the lowest tertile of weight loss were excluded from the following analyses. Consequently, 30 runners remained. 

Both previously reported differences between runners and controls (Figs. 3A, 3B) slightly changed (RNFL thickness, P = 0.63; MV, P = 0.02), whereas the significance of the results of the extended circular scan data did not change (data not shown). 

Discussion

Changes in RNFL thickness and MV over time are firmly established as a robust measure of disease progression for glaucoma, macular degeneration, and an increasing number of other ophthalmological disorders. Subsequently, using RNFL thickness around the ONH and MV has been suggested as a biomarker for neuro-axonal degeneration in neurologic disorders in general and for MS in particular. ^4,32,33 Because an increasing number of clinical trials include RNFL thickness and inner MV as a secondary outcome measure, it seemed timely to investigate the range of physiological variation in these eloquent areas in vivo. The present study examined whether physiological variation could cause detectable changes in RNFL thickness and MV in healthy human subjects as measured by SD-OCT. 

Results of this prospective, longitudinal study clearly demonstrated that in the runners, both RNFL thickness at the ONH and the inner MV changed under physiological conditions (Tables 1, 2). In controls, neither the circular nor the volume scan showed significant changes over time. To add further weight to this argument, the observed physiological changes in the runners were clearly greater than the published measurement noise level (Figs. 3A, 3B, gray shaded area). This strongly suggests that the findings of this study were not only statistically significant but were also biologically relevant. Results of the analyses with extended RNFL thickness data give an indication of the anatomic presentation of the observed changes in RNFL thickness. These results should, however, be interpreted with caution because the analyses were performed using absolute data, and there is considerable intersubject variability in RNFL thickness. 

The significant change in RNFL thickness in runners, as a consequence of physiological variation, was on average 1.6 μm. This change is small but is more than 10 times larger than the estimated annual RNFL thickness loss in healthy controls (0.1 μm). ⁷ Given the high accuracy of the SD-OCT device (CV, 0.42%), ³ the observed change was unlikely to have been caused by measurement error. Furthermore, the data of a 2-year longitudinal MS cohort study of RNFL thickness showed a change over time well within the range of physiological variation reported in the present study. ³⁴ It is possible that the paradoxical finding of an increase in RNFL thickness over time in patients with secondary progressive MS who do not have optic neuritis ³⁴ may be attributed to physiological variation. 

How can the observed physiological change in RNFL thickness and MV be explained? Because this study was designed within the constraints of a public charity run, there are no hard physiological data relying on invasive physiological or laboratory measures. It is therefore only possible to speculate on feasible mechanisms, such as change in rehydration-related change of cellular volume, which may be dependent on each runner's electrolyte status. ³⁵ Alternatively, the exercise-induced hyperemic state may be relevant. Future studies, including blood and urine laboratory measures, and experiments designed to induce controlled hypercapnia may be informative. 

Of note, Kinkelder et al. ³⁶ have reported a measurement artifact that was not controlled for in this study. Heartbeat-induced axial motion during OCT volume scanning of the retina can affect RNFL measurements. This artifact is likely to increase noise rather than to introduce a bias and is therefore an unlikely explanation for our reported physiological changes. Additionally, the detailed anatomic analyses in the present study did not show that the change in RNFL thickness was confined to the retinal branch arteries and veins around the ONH. Finally, it may be that the measurement accuracy of the SD-OCT ring scan was lowest where the start and end-point of the scan met, which in this experiment was nasally. This issue may be addressed by programming the OCT software such that it permits for selection of a ring-scan start point for future studies. 

The reproducibility of the SD-OCT device used in this study (Spectralis; Heidelberg Engineering Inc.) is excellent. ^3,23,29 This is most likely because of the system's eye-tracking function (TruTrack) during the scanning process and its automatic recognition of the exact scan location for follow-up examination. Using these features minimizes extrinsic factors, such as the patient's ability to fixate, and measurements are therefore less sensitive to subjective operator judgment. ²⁹ Moreover, Jo et al. ³⁷ observed diurnal variation in retinal thickness, measured with TD-OCT but not with SD-OCT. ³⁷ They stated that the difference was likely attributed to the good reproducibility of SD-OCT, in contrast to that of TD-OCT. Consequently, the observed changes in RNFL thickness and MV in the present study are unlikely to represent measurement error and strongly suggest relevant physiological variation. 

In conclusion, this study showed for the first time that the physiological variation in RNFL thickness and MV not only exceeded the estimated measurement error of the SD-OCT device but, more important, occurred at a proportion that, on an individual patient level, was approximately 10 times greater than the estimated annual change. These findings are relevant for the design and data interpretation of longitudinal studies that use RNFL and MV as biomarkers in humans. 

Footnotes

 Supported in part by a program grant from the Dutch MS Research Foundation (MS Center, VU University Medical Center).

Footnotes

 Disclosure: L.J. Balk, None; J.M. Sonder, None; E.M.M. Strijbis, None; J.W.R. Twisk, None; J. Killestein, None; B.M.J. Uitdehaag, None; C.H. Polman, None; A. Petzold, None

The authors thank all participants of the 6th VU Medical Centre Building Run and the controls who participated in our study.