2.1.

Human Subjects

Participant recruitment occurred between February 2013 and September 2013, and MS was diagnosed based on 2010 McDonald criteria.¹⁷ We distributed flyers to participants from previous studies in our laboratory who had expressed interest in future research opportunities and to participants within the North American Research Committee on MS registry. MS patients outside of the age range of 18 to 64 years and Expanded Disability Status Scale (EDSS)¹⁸ score $>8.0$ were excluded. Participants with MS ($n=58$) and HCs without any history of ocular or neurologic disease ($n=13$) were recruited for the study. The analysis includes 24 eyes from 13 HC subjects, 38 eyes from 20 MS-mild patients, 31 eyes from 19 MS-moderate patients, and 36 eyes from 19 MS-severe patients. Further, 24 eyes, all in MS patients, had a history of ON. The MS study subjects were divided into three disability groups based on the EDSS score: mild (EDSS 1.0 to 3.5), moderate (EDSS 4.0 to 5.5), and severe (EDSS 6.0 to 7.5). It is to be noted that the number of eyes is not double the number of patients or HCs. We excluded eyes that exhibited inadequate OCT signal strength on baseline scan and excluded subjects that had other ophthalmologic or neurologic disorders including glaucoma, diabetic and/or hypertensive retinopathy, and/or a refractive error greater than six diopters, from the study. Further, images with poor OCT scan centration were also excluded, as error in centration may artificially decrease or increase the measured retinal layer thicknesses.¹⁹

The protocol was reviewed and approved by the Institutional Review Board (IRB) at the University of Illinois, Urbana-Champaign, and all participants signed an informed consent prior to participating in the study. Table 1 shows the subject demographics and group characteristics.

Table 1

Subject demographics and group characteristics.

2.2.

Optical Coherence Tomography Imaging and Segmentation

Retinal imaging in this study was performed using a commercial OCT system (Spectralis HRA-OCT, Heidelberg Engineering, Heidelberg, Germany). The OCT scanning protocol consisted of a 3.9-mm-diameter circumferential scan centered on the optic nerve head (total 1536 A-scans). All the OCT scans were performed by several people, all with the same level of training. After converting the images from the proprietary E2E format to tiff using the commercial viewing software (Heidelberg Eye Explorer, version 1.7.1.0), OCT images were flattened by manually fitting a line to the border between Bruch’s membrane (BM) and choroid [i.e., the outer border of BM (OBBM)] in order to make all the OCT scans uniform. Figure 1 shows representative circumferential peripapillary OCT images and corresponding en face IR-SLO images, respectively, for the three study groups: (a) HC, (b) MS non-ON, and (c) MS ON subjects. The green arrows show the scan direction and were centered over the optic disc.

Fig. 1

Representative circumferential peripapillary OCT images and corresponding en face IR-SLO images, respectively, for (a) HC, (b) MS non-ON subjects, and (c) MS ON subjects. The green arrows show the scan direction.

Six retinal layer lines: (i) inner border of internal limiting membrane (IBILM), (ii) outer (posterior) border of nerve fiber layer (OBNFL), (iii) inner border of inner plexiform layer (IBIPL), (iv) outer border of outer plexiform layer (OBOPL), (v) junction of inner and outer photoreceptor segments (ISOS), and (vi) OBBM were automatically segmented using freely available open source code [OCT Segmentation and Evaluation Graphical User Interface (OCTSEG), Pattern Recognition Lab, Friedrich-Alexander University, Erlangen-Nuremberg, Germany] developed based on research work on OCT.²⁰ We rarely corrected the segmented images manually, but did so when there was an obvious discontinuity in the automatic segmentation. Figure 2 shows representative segmented OCT images for (a) HC, (b) MS non-ON, and (c) MS ON subjects shown in Fig. 1. The six segmenting lines in Figs. 2(a)–(c), from top to bottom, represent IBILM, OBNFL, IBIPL, OBOPL, ISOS, and OBBM, respectively. The OCTSEG software generates a metadata file containing the retinal layer thicknesses in pixels, which was then read and processed in MATLAB™ to prepare the tables for analysis of variance (ANOVA).

Fig. 2

Representative segmented OCT images for (a) HC, (b) MS non-ON subjects, and (c) MS ON subjects shown in Fig. 1. The six segmenting lines in (a–c), from top-to-bottom, represent IBILM, OBNFL, IBIPL, OBOPL, ISOS, and OBBM, respectively.

Based on the segmentation results, 15 different retinal layer thickness combinations (IBILM–OBNFL, IBILM–IBIPL, IBILM–OBOPL, IBILM–ISOS, IBILM–OBBM, OBNFL–IBIPL, OBNFL–OBOPL, OBNFL–ISOS, OBNFL–OBBM, IBIPL–OBOPL, IBIPL–ISOS, IBIPL–OBBM, OBOPL–ISOS, OBOPL–OBBM, and ISOS–OBBM) from the six retinal layer borders were possible. These thicknesses were calculated by using the global average of the thicknesses. In a similar way, there were 105 possible retinal layer thickness ratios from the 15 different thicknesses. It is noteworthy to mention that we first renamed all the OCT images with random numbers and then segmented them blindly to avoid any kind of bias. After segmentation, we regrouped the segmented images based on the key for randomization. All segmentations were performed by the same person to avoid variability.

2.3.

Statistical Analysis

All statistical analyses were performed using SPSS (Version 22.0, IBM, Chicago, Illinois) and MATLAB™. Values are presented in the text as mean (standard deviation), unless otherwise noted. A set of ANOVA was run to perform comparisons between HC subjects and MS non-ON subjects, and between the MS non-ON and ON groups for each of the average thicknesses and ratios. Such ANOVA was also performed to compare the MS-disability groups. Statistical significance was set at $p<0.05$. We included omega-squared (${\omega }^2$) as an effect size estimate for understanding the magnitude of differences between groups and interpreted the values based on the guidelines mentioned by Field.²¹

3.1.

Comparison of Average Retinal Thicknesses

We first compared all the retinal thicknesses (global average) among the study groups. Some of these are presented in Tables 2 and 3, which show the mean values of thicknesses along with the standard deviations in parentheses, and the $F$ and $p$ values from the one-way ANOVA as well as the ${\omega }^2$ value as effect size. Table 2 details some of the ANOVA results for the MS non-ON group with respect to HC subjects and for the MS ON group with respect to MS non-ON subjects. The statistically significant p-values (p<0.05) are shown in bold text in Table 2 and all the other tables. It can be seen from Table 3 that the ${\mathrm{RNFL}}_T$ (i.e., the layer thickness between the IBILM and OBNFL) and the combination of ${\mathrm{RNFL}}_T$ and ${\mathrm{GCL}}_T$ (i.e., the region between IBILM and IBIPL) are different in MS non-ON subjects compared to HC. The GCL and the thicknesses involving the RNFL are different in subjects with a history of ON, compared to MS non-ON subjects. Table 3 reports such ANOVA results among the MS-disability groups: mild, moderate, and severe. From the comparison, it appears that the retinal pigment epithelium (i.e., the region between ISOS and OBBM) layer and the thicknesses involving the RNFL are different among the MS-disability groups.

Table 2

ANOVA results on some retinal layer thicknesses (global average) from the comparisons: MS non-ON with HCs and MS-ON with MS non-ON.

Table 3

ANOVA results on some retinal layer thicknesses (global average) from the comparisons: the MS-disability groups with HCs and among the disability groups.

3.2.

Comparison of Retinal Ratios

Different ratios were generated from the retinal layer thicknesses (global average) and compared among the study groups. We again performed ANOVA for this with respect to HC for the MS non-ON group, with respect to MS non-ON for the MS-ON group, and also among the MS-disability groups. Since there were a total of 105 possible ratios, we divided them into two categories: those involving ${\mathrm{RNFL}}_T$ (60) or those that did not (45), since ${\mathrm{RNFL}}_T$ is the only single layer that is significantly different among the study groups. We subsequently identified ratios in both categories that were significantly different among the study groups.

Table 4 shows comparisons of some significant ratios involving the RFNL layer. For the MS non-ON group, ANOVA was performed with respect to HC, whereas for MS-ON, ANOVA was performed with respect to MS non-ON. There were a total of 52 such significant ratios for both MS non-ON and MS ON groups. In contrast, Table 5 shows some significant ratios (out of a total of 23) that do not involve the RFNL layer for the MS-ON group when compared to MS non-ON. Interestingly, there were no such significant ratios for the MS non-ON group.

Table 4

Significant ratios involving the RFNL in both MS non-ON and MS-ON groups.

Table 5

Ratios not involving the RFNL in both MS non-ON and ON groups, but significant only for the MS-ON group.

The ANOVA results of the ${\mathrm{RNFL}}_T$ and a few ratios among the disability groups, compared to HC subjects, are shown in Table 6. MS-mild eyes show several significant ratios involving the RNFL layer, even though ${\mathrm{RNFL}}_T$ itself in these eyes does not change significantly with respect to HC. The total number of such significant ratios is 14 (only three ratios are listed in Table 6). In contrast, there were 28 and 43 different ratios in MS-moderate and MS-severe groups, respectively, compared to HC subjects along with a different ${\mathrm{RNFL}}_T$. The ANOVA among the MS-disability groups revealed 15 different ratios (only the ratio RNFL/IBIPL-ISOS is shown in Table 6, last column and the last row) other than a different ${\mathrm{RNFL}}_T$.

Table 6

ANOVA results from the comparison of RNFL thickness and ratios: the MS-disability groups with HCs and among the disability groups.

Finally, Table 7 shows the ANOVA results for the cohort of MS non-ON eyes ($n=24$) with ${\mathrm{RNFL}}_T$ (global average) in the range of 96 to $106\mu \mathrm{m}$, when compared to HC subjects ($101\pm 5\mu \mathrm{m}$). The table shows the ${\mathrm{RNFL}}_T$ and the few significant ratios involving the RNFL layer. Statistically significant thickness ratios ($p<0.05$) are shown in bold numbers in Table 7. It is important to mention that none of the other thicknesses are different in this cohort.

Table 7

ANOVA results on the cohort of MS non-ON eyes with RNFL thickness (global average) in the range of 96 to 106  μm.