Early evidence of multiple sclerosis in a prospective study of high-risk family members

Abstract

Importance

Subclinical inflammatory demyelination and neurodegeneration likely precede symptom onset in multiple sclerosis (MS).

Objective

To investigate the prevalence of brain magnetic resonance imaging (MRI) and subclinical abnormalities in asymptomatic individuals at risk for MS.

Design, Setting and Participants

The Genes and Environment in Multiple Sclerosis (GEMS) is a prospective cohort study of MS first-degree family members. We assessed each participant’s risk for MS using a weighted score comprising genetic burden and environmental exposures.

Main Outcomes and Measures

Participants in the top and bottom 10% of the risk distribution underwent standard and quantitative neurological examination (including disability scale, visual, cognitive, motor, and sensory testing) as well as qualitative and quantitative neuroimaging with 3-tesla brain MRI and optical coherence tomography.

Results

This study included 100 participants: higher-risk (n=41, 98% women) and lower-risk (n=59, 42% women); mean age of 35.1 years. Given the unequal sex distribution between the two groups, we restricted analyses to women (n=65). When considering all measured outcomes, higher-risk women differed from lower-risk women (omnibus p-value=0.0125). We found the greatest difference in vibration sensation between the two groups: higher-risk women exhibited more impaired vibration perception in distal lower extremities as detected by the Vibratron-II device (p=0.0078, adjusting for age, height, and testing date). Further, five women (8%) met the primary neuroimaging outcome of having T2-hyperintense brain lesions consistent with the 2010 MS MRI criteria for dissemination in space (4 higher-risk; 1 lower-risk). A subset of participants harbor many different neuroimaging features associated with MS pathology, including perivenous T2-hyperintense lesions and focal leptomeningeal enhancement, consistent with the hypothesis that these individuals are at higher risk of developing clinical symptoms of MS than the general population.

Conclusion and Relevance

Higher-risk asymptomatic family members of MS patients are more likely to have early manifestations of MS, which may constitute evidence of myelitis, abnormal myelin structure/function, and/or neurodegeneration. These findings underscore the importance of early detection in high-risk individuals.

Trial Registration

NCT01353547

Introduction

Subclinical inflammatory demyelination and neurodegeneration have been postulated to precede symptom onset in multiple sclerosis (MS)¹. This concept is supported in part by the observation that 34% of the neurologically asymptomatic individuals with incidentally discovered brain lesions consistent with MS (or “Radiologically Isolated Syndrome”) develop clinical MS within five years^2,3. The time frame of the underlying disease process that precedes symptom onset is unknown. Further, the lack of primary prevention strategies for MS in high-risk populations such as family members¹, remains an unmet challenge⁴.

We set out to investigate the sequence of events leading to the onset of MS. To accomplish the long-term goal of identifying targets for prevention strategies and the optimal timing for their deployment, we launched a multi-centered, prospective cohort study of individuals who are first-degree relatives of people with MS: the Genes and Environment in Multiple Sclerosis (GEMS) study⁵. Within MS family members, we showed that an aggregate estimate of MS risk (Genetic and Environmental Risk Score: GERS_MS) that incorporates an individual’s genetic burden^6–8 and environmental exposures^9,10 is informative of MS risk beyond family history. Further, first-degree relatives in the upper strata of the GERS_MS distribution had the greatest probability of having clinical MS⁵.

In the present study, we investigated early evidence of the disease in family members whose risk profiles for MS susceptibility are at the extreme ends of the spectrum. We hypothesized that neurologically asymptomatic higher-risk individuals are more likely to exhibit subtle manifestations that are consistent with inflammatory demyelination and/or neurodegeneration, the two cardinal pathological features of MS. Given the wide range of possible manifestations in MS, we performed standard and quantitative clinical assessments (including visual, cognitive, motor, and sensory function) as well as qualitative and quantitative neuroimaging measures such as brain magnetic resonance imaging (MRI) and optical coherence tomography (OCT).

Methods

Overall Study Design

The inclusion criteria for the GEMS study are: (1) being 18 to 50 years of age at enrollment, (2) having at least one first-degree family member (i.e., parent, full-sibling, or child) with a diagnosis of MS (Figure 1A). The GEMS study design, including calculation of the Genetic and Environmental Risk Score for Multiple Sclerosis Susceptibility (GERS_MS) for individualized risk stratification, has been described earlier⁵ (also Supplementary Methods). The institutional review boards of Partners Health, National Institute of Health (NIH), and University of Pittsburgh approved the study.

Figure 1. Study Design and Quantitle-quantile (qq) plot.

(A) Detailed characterization of neurologically asymptomatic first-degree family members from the Genes and Environment in Multiple Sclerosis (GEMS) cohort who are at the extremes of the risk profile for multiple sclerosis susceptibility. GERS_MS, genetic and environmental risk score; MRI, magnetic resonance imaging; OCT, optic coherence tomography. (B) For a given outcome, the expected p-values (−log10 [p-value]) are shown on the x-axis, and the observed p-values (−log10 [p-value]) are shown on the y-axis. The expected p-values assume a null distribution, with no difference between higher and lower risk participants. The corresponding dark and light grey areas indicate the extreme ranges of the qq-plots as generated by chance, at a threshold of p=0.10 and at p=0.05, respectively. The 90% and 95% confidence intervals were derived empirically by randomly assigning participants to the higher or lower risk group and repeating the analysis 10,000 times. When the distribution of the observed p-values for measured outcomes was taken as a whole, the overall difference between the higher and lower risk participants was unlikely to have occurred by chance (omnibus p=0.0125). For phenotypes where we took measurements from both sides of each participant (e.g., vibratron-II), we used the measurement from the worse side to avoid inclusion of highly correlated phenotypes. For a given outcome, we expect the difference between higher and lower risk participants to be beyond chance if the observed probability is outside the confidence interval and the further to the right. This is the case for Vibratron-II measurements of vibration sensitivity from the worse side (*), which is the great toe on the side with the higher measurement or worse vibration sensitivity. (This is also the case for vibratron-II measurement in left great toe, in the right toe, and the average of the measurement in the left and right great toe; not shown in this plot.) For the vibratron-II measurement from the worse side, we also compared the observed p-value (univariate: 0.0002; multivariate: 0.0078) to the set of minimum p-values, as taken across all measured outcomes, generated from the 10,000 randomly sampled permutations (10,000 minimum p-values), and found an empirical p-value of 0.0016, suggesting that the p-value for this phenotype remains significant after adjusting for multiple testing.

Study Design for Deep Phenotyping

Between August 2012 and July 2015, one hundred neurologically asymptomatic GEMS participants traveled from various locations across the United States to NIH in Bethesda, MD, for detailed neuroimaging, laboratory, and neurological examination (Figure 1A, see Supplementary Methods for details of subject selection criteria, the battery of standard and quantitative neurological evaluation, including vibration sensitivity testing with Vibratron-II device, as well as OCT and qualitative and quantitative brain MRI). For a given participant, all measures were collected during a single study visit. Examiners were blinded to each participant’s risk profile.

Statistical Analysis

For unadjusted comparison between higher and lower risk participants, we used χ² and Fisher’s exact tests for categorical variables, independent sample two-tailed t-tests for continuous variables with normal distribution, and a Wilcoxon rank-sum test for continuous variables with non-normal distribution. The continuous Vibratron-II outcomes were analyzed using linear regression models both without adjustment and while adjusting for age, height, and date of exam. Please see Supplementary Methods for details on the omnibus test (Figure 1B), which tested the hypothesis that the chi-squared statistic as calculated from a Fisher’s combined probability test (combining the p-values of all phenotypes) was higher in the observed data than chance. All analyses were performed using SAS 9.3 or JMP Pro 12.2.0 (SAS Institute, Cary, NC) or R ^11,12.

Results

One hundred participants of self-reported European ancestry from the GEMS cohort who were neurologically asymptomatic at the time of testing traveled from across the United States to the NIH for detailed examination: 41 higher-risk participants (98% women) from the top 10% and 59 lower-risk participants (42% women) from the bottom 10% of the GERS distribution (eTable 1A). Because women have a higher participation rate in the GEMS study overall and female sex is a risk factor in the GERS⁵, there was a significantly greater representation of women in the higher-risk subgroup (p=1.5×10⁻⁹). To meaningfully compare the two risk groups without attributing any potential difference primarily to the role of sex, we restricted subsequent analyses to the 65 female participants: 40 higher-risk women and 25 lower-risk women (eTable 1B). In comparing the demographic characteristics and clinical history of the women, we observed a trend toward higher body mass index (p=0.046) and greater weight (p=0.011) in the higher-risk subgroup when compared to the lower-risk subgroup, but the difference is not statistically significant after adjusting for multiple testing. Otherwise, higher-risk and lower-risk women had no difference in age, height, cigarette smoking exposure, 25-OH Vitamin D level, history of infectious mononucleosis or migraine. No participant reported a history of diabetes.

Qualitative and Quantitative Neuroimaging Measures

To investigate neuroimaging evidence of MS in neurologically asymptomatic individuals (Table 1), we predefined the primary outcome measure of the study as the presence of T2-hyperintense lesions on brain MRI that met the 2010 McDonald criteria for dissemination in space (DIS)¹³. Among the 65 women, five (8%) met the primary outcome, consistent with prior MRI findings of MS-type lesions in asymptomatic family members^14,15: four higher-risk women and one lower-risk woman met this criterion (p=0.64).

Table 1.

Neuroimaging and clinical evaluation of the neurologically asymptomatic women from the Genes and Environment in Multiple Sclerosis cohort of first-degree family members

Brain Magnetic Resonance Imaging
Parameter	All Women	Lower Risk Subgroup	Higher Risk Subgroup	p-value *
Number of Participants	65	25	40	N/A
Primary Outcome
McDonald 2010 A, N (%)#	5 (8%)	1 (4%)	4 (10%)	0.64
Secondary Outcome
Proposed 2016 Criteria for MS B, N (%)#	3 (5%)	1 (4%)	2 (5%)	>0.99
RIS C, N (%)#	2 (3%)	1 (4%)	1 (3%)	>0.99
Total Number of T2 Lesions D, Median (Q1, Q3)	1 (0, 3)	0 (0, 1)	1 (0, 4)	0.18
Leptomeningeal Enhancement E, N (%)#	4 (6%)	1 (4%)	3 (8%)	>0.99
≥40% T2 lesions Perivenous F, N (%)#	14 (22%)	2 (8%)	12 (43%)	0.061
McDonald 2010 A + ≥40% T2 lesions Perivenous F, N (%)#	3 (5%)	0 (0%)	3 (8%)	0.28
Volumetric Analysis
Normalized G Thalamus Volume, Mean (SD)	0.0103 (0.0014)	0.0106 (0.0016)	0.0101 (0.0013)	0.23
Normalized G Total Gray Matter Volume, Mean (SD)	0.482 (0.019)	0.478 (0.020)	0.485 (0.018)	0.16
Normalized G Total White Matter Volume, Mean (SD)	0.321 (0.019)	0.321 (0.020)	0.322(0.018)	0.78
Normalized G Total Brain Volume, Mean (SD)	0.821 (0.012)	0.816 (0.012)	0.824 (0.011)	0.012
Thalamus Volume/Total Brain Volume, Mean (SD)	0.0125 (0.0018)	0.0130 (0.0019)	0.0123(0.0016)	0.14
Optic Coherence Tomography
Parameter	All Women	Lower Risk Subgroup	Higher Risk Subgroup	p-value *
Number of Participants	65	25	40	N/A
Retinal Nerve Fiber Layer Thickness, Right, μm, Mean (SD)	99.8 (10.0)	99.8 (11.8)	99.7 (8.9)	>0.99
Retinal Nerve Fiber Layer Thickness, Left, μm, Mean (SD)	99.1 (10.0)	99.0 (11.0)	99.1 (9.5)	>0.99
Macular Volume, Right, mm3Mean (SD)	8.69 (0.36)	8.75 (0.31)	8.66 (0.39)	0.33
Macular Volume, Left, mm3Mean (SD)	8.69 (0.36)	8.76 (0.32)	8.65 (0.39)	0.11
Quantitative Neurological Evaluation
Parameter	All Women	Lower Risk Subgroup	Higher Risk Subgroup	p-value *
Number of Participants	65	25	40	N/A
Expanded Disability Status Scale, Median (IQR)/[range]	0 (0)[0–1.5]	0 (0)[0–1.0]	0 (0)[0–1.5]	0.72
Timed 25-foot Walk H, seconds, Mean (SD)	3.82 (0.61)	3.69 (0.12)	3.91 (0.10)	0.18
9-Hole Peg Test Dominant Hand H, seconds, Mean (SD)	17.2 (2.2)	17.0 (2.3)	17.3 (2.2)	0.67
9-Hole Peg Test Non-Dominant Hand H, seconds, Mean (SD)	17.9 (2.2)	18.1 (2.1)	17.8 (2.3)	0.53
Timed Up-and-Go H, seconds, Mean (SD)	5.20 (0.96)	4.92 (0.67)	5.38 (1.07)	0.036
PASAT I, % correct, Mean (SD)	84.6 (13.1)	81.2 (13.8)	86.8 (12.2)	0.10
SDMT J, % correct, Mean (SD)	99.4 (1.2)	99.3 (1.1)	99.5 (1.3)	0.57

Note:

^AMeeting the 2010 McDonald criteria for dissemination in space ¹³

^BMeeting the 2016 proposed revised criteria for dissemination in space ¹⁶

^CMeeting Okuda’s criteria for radiologically isolated syndrome ²

^DTotal number of T2-hyperintense lesions that are greater than 3mm

^EHaving evidence of focal leptomeningeal enhancement

^FHaving >=40% of T2-hyperintense lesions that exhibit perivenous appearance ^18,19

^GNormalized by intracranial volume (total brain volume + total cerebrospinal fluid volume): each of the normalized measures is without a unit

^HAverage of two trials per participant

^IPaced Auditory Serial Addition Test (PASAT): % of answers that are correct (out of a total of 60)

^JSymbol Digit Modality Test (SDMT): % of answers that are correct

^*Comparing the lower-risk and higher-risk subgroup

^#Percentage of the number of participants in the column

N/A Not applicable

We examined additional brain MRI measures as secondary outcomes (Table 1). Among the women, three (5%) met the recently proposed MAGNIMS (Magnetic Resonance Imaging in Multiple Sclerosis) consensus MRI criteria for MS diagnosis¹⁶, two (3%) met the criteria for Radiologically Isolated Syndrome ², and four (6%) exhibited evidence of leptomeningeal enhancement as a single focus¹⁷. Interestingly, 14 participants (22% of the women: 12 higher-risk, 2 lower-risk) had ≥40% of T2 brain lesions with a perivenous appearance^18,19; three of these participants also met the 2010 McDonald MRI Criteria for DIS (Figure 2). Overall, a subset of the MS first-degree family members harbor a variety of neuroimaging features that are associated with MS pathology, consistent with the hypothesis that these individuals are at higher risk of developing clinical manifestations of MS than the general population.

Figure 2. Representative 3-tesla brain magnetic resonance imaging scans.

of a neurologically asymptomatic participant from the Genes and Environment in Multiple Sclerosis cohort meeting the 2010 McDonald Criteria for dissemination in space. (A) Lesions typical of multiple sclerosis are seen in the periventricular region on the sagittal T2-weighted FLAIR sequence. (B) The lesions are hypointense on the corresponding T1-weighted sequence. (C–D) A lesion’s central blood vessel is appreciated on the T2*-weighted sequence.

After adjusting for multiple testing, there was no difference between the two subgroups in terms of the secondary neuroimaging outcomes as well as the normalized whole and regional (thalamus, total gray matter, total white matter) brain volume on brain MRI (Table 1). The OCT analysis likewise showed no significant retinal nerve fiber layer thickness and total macular volume differences between the two subgroups (Table 1).

Quantitative Neurological Measures

To investigate subtle subclinical evidence of dysfunction in neurologically asymptomatic participants, we measured the following quantitative outcomes that are known to correlate with physical and cognitive disability in MS: EDSS²⁰, timed 25-foot walk²¹, 9-hole peg test²¹, PASAT²¹, SDMT^22,23, and timed up-and-go or TUG^24,25 (Table 1) as well as high-contrast and low-contrast visual acuity (eTable 2). Using these outcomes and after consideration for multiple testing, we did not identify any statistically significant impairment in asymptomatic higher-risk women when compared to the lower-risk women. However, it is worth noting that higher-risk women were slower in completing the TUG test than lower-risk women (p=0.036).

Measures of Vibration Sensitivity

As part of the standard neurological exam, we observed impaired vibratory sensation in the distal lower extremities as measured by the 128 Hz tuning fork test. This clinical observation during the early phase of the study led to the subsequent incorporation of two additional measures of vibration sensitivity into the study protocol: the Rydel-Seiffert graduated tuning fork^26,27 and the Vibratron-II device^28–31. In total, 47 (27 higher-risk and 20 lower-risk) out of the 65 women participants completed vibration testing using all three modalities. This subset of the women had demographic and clinical characteristics similar to the overall cohort of women (eTable 1B).

Of the three testing modalities for vibration sensitivity, we used the Vibratron-II measurement as the primary outcome since the Vibratron-II device provides the most objective quantitative measurement of vibration sensation in MS^29–31. Given that we measured vibration sensitivity in the left and the right distal lower extremity (specifically, the great toe), we compared the vibration sensitivity threshold between higher and lower-risk participants using both the vibration unit value from the worse side³² (i.e., higher value) and the average value of the two sides.

Using the more objective quantification with the Vibratron-II device, we found evidence of relatively impaired vibration sensitivity in the distal lower extremity of the higher-risk female participants. Specifically, higher-risk women had a higher vibration sensitivity threshold (with higher mean vibration units) than lower-risk women, when comparing either the worse side of the distal lower extremity (p=2×10⁻⁴) or the average value between the two sides (p=2×10⁻⁴) in a univariate analysis (Table 2; Figure 3).

Table 2.

Evaluation of vibration sensitivity in a subset of neurologically asymptomatic women from the Genes and Environment in Multiple Sclerosis Cohort of first-degree family members

Parameter	Women	Lower Risk Subgroup	Higher Risk Subgroup	p-value *	p-value ** Adjusted
Number of Participants	47	20	27	N/A	N/A
Primary Outcomes
Vibratron-II, Worse Side A, vu, Mean (SD)	2.21 (0.66)	1.83 (0.54)	2.48 (0.60)	2.0 × 10−4	0.0078
Vibratron-II, Average A, vu, Mean (SD)	2.07 (0.63)	1.71 (0.53)	2.34 (0.57)	2.0 × 10−4	0.018

Note:

^AVibratron-II measurement of vibration sensitivity in the great toe in vibration unit (vu), which is the amplitude of the vibration (proportional to the square of applied voltage) and a lower unit of measurement indicates better vibratory sensation: Worse Side, the great toe on the side with the higher measurement or worse vibration sensitivity ³²; Average, average of the measurement in the right great toe and left great toe

^*Comparing the lower-risk and higher-risk subgroup

^**Comparing the lower-risk and higher-risk subgroup, after adjusting for age (at the time of evaluation), height, and testing date. Current smoking status was included in the Genetic and Environmental Risk Score (GERS), which categorized a participant as higher versus lower risk. There was no smoker in the lower risk group of women who completed the vibration sensitivity studies.

N/A Not applicable

Figure 3. Plots of the vibration sensitivity as quantified by Vibratron-II.

comparing the higher-risk (n=20) and lower-risk (n=27) neurologically asymptomatic women from the Genes and Environment in Multiple Sclerosis cohort of first-degree family members. (A) The x-axis indicates the higher and lower risk subgroups according to the genetic and environmental risk score (GERS): women with GERS at the top 10% (High) and bottom 10% (Low) from the Genes and Environment in Multiple Sclerosis Cohort. The y-axis indicates vibration sensitivity in vibration units (vu), which are the amplitudes of vibration and are proportional to the square of applied voltage. Vibration sensitivity threshold was quantified in the right and left great toe separately, and an average measure was calculated. A higher vibration unit indicates worse vibration sensitivity. The p-value represents the univariate comparison between the higher and lower-risk group. (B) A participant with the highest GERS in the study who was diagnosed with MS after the testing is shown in the larger red bubble. The GERS is shown in the x-axis. The y-axis indicates the vibration sensitivity as the average measurement of both great toes. The z-axis indicates the vibration sensitivity as the measurement from the worse side. Of note, the two vibratron measurements do not perfectly correlate in this participant. Higher-risk (H) women are shown in the red bubbles. Lower-risk (L) women are shown in the green bubbles.

Since vibration perception is under the potential influence of several factors, we included age and height at the time of testing as covariates in a multivariate analysis. Current cigarette smoking status was a component of the risk score (GERS) that we used to stratify participants such that there was no current smoker among the lower risk participants who underwent vibratron testing (eTable 1B). Given that the study was completed over the course of three years, we further included test date (in ordinal format) as a covariate and as a control for batch effect. In the multivariate analysis, the evidence of worse vibration sensitivity (or greater vibration sensitivity threshold detected by the Vibratron-II device) in higher-risk women than lower-risk women from the GEMS cohort persisted when comparing the worse vibration value (p=0.0078) and the average value (p=0.018) (Table 2). We observed the same finding when current smoking status was additionally included in the multivariate analysis (worse vibratron value, p=0.0031; average vibratron value, p=0.0082).

We observed a consistent trend when comparing the Vibratron-II results with the other modalities: 128 Hz tuning fork (proportion of participants reporting duration that lasts ≤25 seconds) and Rydel-Seiffert graduated tuning fork (median duration) (eTable 2). Specifically, among the GEMS women participants, we found that a higher vibration value (or worse vibration sensitivity threshold) using the Vibratron-II measurement was correlated with a shorter duration on the 128 Hz tuning fork test (Spearman coefficient = −0.332, p=0.022), and the 128 Hz and the Rydel-Seiffert tuning fork measures were correlated (Spearman coefficient = 0.383, p=0.011) (eFigure 1).

Interestingly, one female participant from the GEMS cohort who was neurologically asymptomatic at the study visit subsequently developed symptoms of tingling in the fingers and received a diagnosis of MS 14 months after enrollment into the GEMS study and two months after completing the NIH study visit. This individual had the highest GERS within the entire GEMS cohort. Notably, she exhibited evidence of impaired vibration sensitivity as quantified by all three modalities: 128 Hz tuning fork (14 seconds bilaterally); Rydel (7 out 8 bilaterally); Vibratron-II (2.28 vibration units on the right and 3.25 vibration units on the left) (Figure 3B). In addition, her brain MRI met the primary MRI outcome, consistent with the 2010 McDonald MRI criteria for DIS (Figure 2A–B) and all secondary MRI outcomes, including having T2-hyperintense lesions localized to the perivenous region (Figure 2C–D).

Global Assessment of the Burden of Potential Neurological Dysfunction

Given the wide spectrum of neurological dysfunction seen in MS patients at symptom onset, we investigated a large number of clinical and neuroimaging outcomes in this cohort at risk individuals. To globally assess the burden of neurological dysfunction in these family members, we performed an omnibus test given the number of tests performed. Omnibus test examined whether there was a statistically significant difference between the higher and lower risk participants when considering all of the measured outcomes, many of which have suggestive associations (Figure 1B). When the distribution of the observed p-values for these outcomes was taken as a whole, the overall difference between the two groups of participants was greater than what one would expect by chance (p=0.0125), as calculated with permutation tests.

Discussion

Here, we report the detailed investigation of early evidence of MS-related dysfunction in a subset of participants in the GEMS study, a large prospective cohort study of first-degree family members. Our major finding is that neurologically asymptomatic women deemed to be at higher risk of developing MS by our risk algorithm (GERS) exhibited a greater burden of potential neurological dysfunction than women at lower risk, with impaired vibration perception in the distal lower extremities being the most extreme result. In addition, 8% of all the women studied in detail met the primary neuroimaging outcome of having brain T2-hyperintense lesions consistent with the 2010 MS MRI criteria for DIS, higher than one would expect to find in the general population. The presence of these findings and of other MS-type changes, including the presence of central veins in T2-hyperintense lesions and of focal leptomeningeal enhancement, suggests that a subset of these family members have asymptomatic inflammation and deserve further monitoring.

These intriguing neuroimaging findings from this study are consistent with prior reports concerning the frequency of MS-like white matter lesions on brain MRI among asymptomatic first-degree relatives (ranging from 4–10%)¹⁵ and the frequency of RIS among healthy family members (3%)¹⁴. Of note, there were substantial differences in study design and population characteristics between our study and the studies by De Stefano et al. and Gabelic et al. as well as differences in the MRI protocol and predefined outcome. In addition, determining optimal management of first-degree relatives with MS-type findings was not the aim of the present study. The ongoing RIS trial (NCT02739542) will provide information to guide the decision of whether treatment should be considered in these participants.

Our results further point to a possible sequence of events leading to MS in which changes in vibration sensitivity may precede the appearance of demyelinating lesions in the brain. Such changes could derive from small inflammatory demyelinating lesions in the dorsal columns of the spinal cord, but we cannot rule out the possibility of abnormal myelin structure or early degeneration of long axonal tracts or subcortical structures such as the thalami. Indeed, the symmetrical vibration changes observed in most of the participants would suggest the latter three explanations over the former. Regardless of etiology, these results underscore the importance of deploying sensitive tools to detect subtle neurological changes early in the MS disease process so that an intervention could be deployed when it is most effective at preserving the CNS.

Our results also suggest that the current study was underpowered to detect a difference in neuroimaging outcomes between higher-risk and lower-risk women. As the GEMS study is the first prospective study of populations at risk for MS, and as this is the first detailed examination of higher and lower risk family members, there was limited information regarding the sample size required to detect early or subtle evidence of MS when we launched the study in 2010. The sample size calculation for the primary brain MRI outcome was initially estimated by extrapolating results from MS patients and healthy control subjects³³. Given the limited information available for power calculation, we planned to investigate 100 participants for this study. Our findings from this study demonstrated that we had excellent statistical power (99%) to detect a difference in vibration sensitivity threshold given the difference between the higher-risk and lower-risk group but only limited power (12%) to detect a difference in the primary MRI outcome of having brain MRI lesions consistent with the 2010 McDonald criteria for DIS (eTable 4). With our results in hand, we can now estimate that, to attain statistical significance, a study comparing 283 higher risk and 283 lower risk women participants will be needed. Thus, these family members with carefully characterized baseline phenotypes provide crucial insights for the future direction of the GEMS study and other similar studies. Specifically, we plan to confirm the finding of change in vibration sensitivity with a follow-up study.

Our findings have several limitations. First, we restricted analyses to women due to the inclusion of sex in the individualized risk score for MS susceptibility and the resulting unequal representation of women in the higher-risk group compared to the lower-risk group. The excess of women participating in the GEMS study was consistent with the sex discrepancy in clinical study participation, possibly heightened by public understanding of the increased risk of MS for women. We are actively working to increase men’s participation in the GEMS study. More important, the findings that emerge from this study are informative as we move forward with the GEMS study. Second, the cross-sectional assessment does not allow for definitive conclusions regarding the sequence of events over time. To address this issue, we plan to perform longitudinal assessment in individuals in whom we have the baseline measurements. Third, the vibration sensitivity thresholds, as quantified by the Vibratron-II device in both the higher-risk and lower-risk groups, were within the “normal” range (as presented in the package insert of the device), but large-scale normative data relevant to the demographics of the study population are not available. We have begun to quantify vibration sensitivity using Vibratron-II in healthy volunteers with similar demographic characteristics as the GEMS study participants. Fourth, we selected a slightly older cohort of participants for this study (mean age 35 years old) as we initially aimed to identify participants with RIS. With findings from this study that inform the risk model, we plan to pursue a prospective investigation in a younger cohort of participants in part to determine the mean age at which RIS develops. Finally, we selected the qualitative and quantitative clinical and neuroimaging measures that were deployed in this study primarily based on their utility in studies of MS patients. Although we expected low sensitivity in detecting early evidence of MS in neurologically asymptomatic first-degree family members using these measures, they inform important baseline information in the larger prospective study. Thus, our study highlights the important need to develop and test more sensitive measures, particularly with biometric devices, to detect subtle subclinical changes early in the disease process.