A Prospective Study of Individuals at Risk of Multiple Sclerosis Informs the Design of Primary Prevention Studies

ABSTRACT

Objective

In multiple sclerosis, the optimal time for deploying a therapeutic intervention is before the central nervous system is damaged; given the success of trials treating the earliest stage of MS, the radiologically isolated syndrome, developing primary prevention strategies is an important next challenge. To inform this question, we examined the conversion of asymptomatic family members to a diagnosis of MS among Genes and Environment in Multiple Sclerosis (GEMS) project participants. We also evaluated an updated Genetic and Environmental risk score (GERS) and interest in participation in primary prevention trials.

Methods

GEMS is a nationwide prospective study of first‐degree relatives of persons with MS; we analyzed data from seven surveys completed over 11 years. 1903 participants completed all questionnaires at study entry. Biological samples were collected in a subset of participants.

Results

Among these 1903 first degree relatives, 141 participants had MS at the time of enrollment; 18 participants converted to MS; and 1744 participants did not acquire an MS diagnosis during 8526 person‐years of observation. The GERS was significantly higher among participants who had MS at the initial survey than those who did not have an MS diagnosis. The incidence rate of MS was estimated as 211 cases with a 95% confidence interval of 113 to 308 cases per 100,000 first‐degree family members per year, which is 100 times greater than the reported incidence of sporadic MS. 48% of the asymptomatic participants were willing to participate in a prevention trial, while 75% expressed willingness when presented with a scenario assuming an elevated risk (20%) of developing MS based on genetic tests performed with blood samples.

Interpretation

Primary prevention trials in MS are urgently needed. Our findings support the feasibility for such studies and inform the design of future primary prevention trials in individuals at risk for MS.

Trial Registration

ClinicalTrials.gov identifier: NCT01353547

1. Introduction

Multiple sclerosis (MS) is a chronic inflammatory and neurodegenerative disorder of the central nervous system [1]. Multiple studies have demonstrated that treatment with disease‐modifying therapies (DMT) earlier in the disease course leads to decreased disability over time [2, 3]. Further, patients with radiologically isolated syndrome (RIS) are at increased risk of being diagnosed with MS within 5–10 years [4, 5, 6, 7], and this risk is reduced with treatment [8, 9]. There is also a possible MS prodrome, during which the target organ, the central nervous system, is involved and affected individuals may experience a range of symptoms months before they meet clinical criteria for MS [10, 11]. Finally, there is increased discussion surrounding the earliest, presymptomatic, pre‐RIS phase of MS, highlighting the importance of early detection and intervention to potentially alter the disease trajectory and improve long term outcomes for individuals at risk of developing MS [12, 13].

Epidemiologic studies have consistently validated a number of different environmental risk factors for MS, such as smoking, adolescent obesity, vitamin D levels, latitude of residence, and a history of infectious mononucleosis or serological evidence of Epstein Barr Virus (EBV) infection [14, 15, 16]. Interest in the latter exposure has led to discussions of primary prevention studies using vaccines [17]. However, the feasibility of such studies rests on identifying target populations for trials and understanding the characteristics of such populations as well as the molecular/cellular dynamics of the presymptomatic, pre‐radiologic phase of the disease.

The Genes and Environment in Multiple Sclerosis (GEMS) project is a prospective cohort study of first‐degree relatives of people with MS. The project was launched in 2011 to investigate the events leading to MS in high‐risk individuals, with the goals of characterizing the incidence rate in this population and, ultimately, designing primary prevention trials [12, 13]. To date, the GEMS project has (1) demonstrated that asymptomatic individuals at higher risk of developing MS can be engaged in clinical research nationwide through social media platforms [13], (2) validated known risk factors for MS [13], (3) demonstrated the feasibility of developing and deploying at large scale an integrated Genetic and Environmental Risk Score (GERS) in a population at risk for MS [12, 13], and (4) illustrated that we can partition first‐degree relatives into subgroups with differential risk of MS using the GERS [12, 13]. Namely, the top quartile of the GERS distribution has an increased risk of MS that is approximately twice that of the average first‐degree relative [12, 13]. Thus, we can use the GERS to determine a high‐risk subset among first‐degree relatives. We have also reported that such high‐risk individuals display early signs of neurologic deficits, such as diminished vibration sensation in the great toe [12].

Here, we report updated results from this cohort study, providing an updated incidence rate of MS diagnosis in GEMS, an evaluation of additional genetic risk factors, and, critically, results of an investigation of the willingness of these family members to participate in primary prevention trials.

2. Methods

2.1. Participants and Study Design

The inclusion criteria for the GEMS project [13] include (1) age between 18 and 50 years at the time of enrollment and (2) having at least one first‐degree family member (i.e., parent, full sibling, or child) diagnosed as having MS. There were no exclusion criteria. In short, a solicitation for participation in the study was posted on the National MS Society website, and participants contacted the research team by phone or email. The participants who did not complete the baseline questionnaires were not considered further. The survey was collected between 2017 and 2022. Biological samples were collected in a subset of participants. The questionnaires can be found in the Appendix S1. All analyses are therefore performed on these 1903 participants with available baseline data and varying levels of follow‐up data. The institutional review boards of Partners HealthCare and Columbia University have approved the study. All participants provided informed consent.

2.2. MS Conversion

MS converters were defined as participants who reported no diagnosis of MS at enrollment and subsequently reported being diagnosed with MS on one or more of the follow‐up surveys. Thus, we have no information regarding whether these participants had an RIS diagnosis at study entry. Asymptomatic participants reported no neurologic symptoms and therefore did not have a clinically isolated syndrome. Confirmation of the incident diagnosis was attempted in all cases by study personnel based on the 2010 McDonald Criteria given that the study was initiated in 2011. Participants were asked to provide office notes and related diagnostics (magnetic resonance imaging). Available data were reviewed by a neurologist with subspecialty training in MS who confirmed the diagnosis of MS using the 2010 McDonald criteria of MS [18], following the protocol developed in 2011 at the beginning of the study. There were no discrepancies between a reported MS diagnosis and the diagnosis made after a review of submitted documents. In participants for which additional information could not be obtained, the subject was counted as an MS converter if either of the following criteria were met: (1) a date of diagnosis was provided in addition to at least one reported diagnosis of MS or (2) two or more consecutive follow‐up surveys reported a diagnosis of MS after the baseline questionnaire. Participants without a diagnosis of MS were counted as such if both of the following criteria were met: the participant (1) reported not having MS at study entry and (2) did not report a diagnosis of MS on any of the subsequent surveys.

2.3. Risk Score Calculation

We measured and compared 3 different risk scores among the participants with MS at study entry, MS converters, and asymptomatic family members: (1) a genetic risk score (GRS⁶⁴) calculated from a list of 64 validated and replicated single nucleotide polymorphisms (SNPs) that were associated with MS susceptibility prior to the onset of the GEMS study [13], (2) a genetic and environmental risk score (GERS) that integrates the GRS⁶⁴ score with scores from 3 other risk factors for MS (female sex, history of infectious mononucleosis, and current smoking status) [13], and (3) an updated genetic risk score (GRS²²⁴) that was calculated based on an updated list of genetic susceptibility loci associated with MS published in 2019, as described in a prior report [19] (224 genome‐wide significant loci including 31 independent loci in the major histocompatibility complex [MHC] region and 193 independent autosomal SNPs).

More specifically, The GERS was calculated using the same strategy that we used in our earlier manuscripts. Genotyping of MS risk variants and calculation of the Genetic and Environmental Risk Score (GERS) for Multiple Sclerosis Susceptibility in GEMS participants have been previously described [13]. In brief, a list of 64 single nucleotide polymorphisms (SNPs) that are significantly associated with the risk of developing MS based on published genome‐wide association studies of MS susceptibility [20, 21, 22], including five within the major histocompatibility complex (MHC) region, was the most up‐to‐date as of 2012 when targeted genotyping of GEMS participants first began. Each SNP was coded additively by the established risk allele and weighted by the natural log of the published odds ratio for MS susceptibility. For the validated non‐genetic risk factors, we counted the presence or absence of the factor, each weighted by the natural log of a published odds or risk ratio: sex (which confers different hormonal milieu between women and men) [23], infectious mononucleosis [24], current smoking status [25].

Of note, targeted genotyping was used to identify allelic information for the 64 SNPs used in the calculation of GRS⁶⁴ [13]. In short, the MassArray iPLEX platform (Sequenom, San Diego, CA) was used for this targeted genotyping, and the resulting genotypes were pre‐processed using standard quality control: minor allele frequency > 0.01, genotype call rate > 0.90, Hardy–Weinberg equilibrium p > 0.001. On the other hand, GRS²²⁴ was calculated using data extracted from genome‐wide genotyping data available from a subset of the participants in a prior study [19]. The list of the SNPs is displayed in Table S1a,b. Briefly, genotyping was performed using the Illumina MEGA‐EX array. Preprocessing was performed using PLINK 2.0. Genome‐wide imputation was performed using the Michigan Imputation Server and the Haplotype Reference Consortium (HRC) panel v1.1. Imputation of the MHC region was separately performed using the SNP2HLA package and the Type 1 Diabetes Genetics Consortium HLA reference panel. The GRS²²⁴ was then calculated as the sum of the imputed dosages of the MS‐associated SNPs multiplied by their effect sizes (log odds ratio).

2.4. Incidence Rate

The incidence rate of MS was calculated among the participants who did not have MS diagnosis at the initial survey (n = 1762, 1903 participants minus 141 participants who had MS at initial survey). We note that GEMS recruits first degree relatives of persons with MS; hence, a small subset of GEMS particpants have MS themselves at study entry. The maximum follow‐up time was 11 years and the follow‐up time was different among the participants based on date of enrollment. The incidence rate was calculated as the number of MS converters (n = 18) over person‐years of follow‐up. To estimate the 95% confidence interval (CI) for the incidence rate, we used a Poisson distribution approach. The CI was calculated using $\frac{\lambda -1.96x\sqrt{\lambda }}{\mathrm{PY}}\times 100000$ as the lower limit and $\frac{\lambda +1.96x\sqrt{\lambda }}{\mathrm{PY}}\times 100000$ as the upper limit where $\lambda$ represents the number of new MS cases, and PY is the total person‐years at risk. This method provides an estimate of the precision of the incidence rate, indicating the range within which the true incidence rate is likely to fall with 95% confidence.

2.5. Statistical Analyses

The continuous demographic variables were compared between the MS groups (MS at enrollment, MS converter between initial survey and Survey 6, and asymptomatic family member) using a Student's t‐test. Chi‐squared tests were performed to assess the independency between the categorical variables and participant groups. The correlation of each pair of risk scores was assessed using Spearman's correlation. The Spearman's correlation coefficient (r) was given to assess the amplitude of the correlation. The tests used in each analysis are reported in each figure or table legend. Results were considered as significant when p < 0.05. Overall, data from the initial survey and subsequent update surveys were analyzed using R software version 4.1.2.

3. Results

3.1. Description of Participants

Of the 2686 individuals who signed a consent form, 1903 (70.8%) completed the initial baseline survey deployed at study entry and were considered to have been fully enrolled into the study. Of the 1903 individuals who signed a consent form, 1117 participants responded to the question regarding participation in a trial to test a treatment to prevent the onset of MS (see Figure 1 and Table S2 for the number and percentage of the completed survey responses on the initial and follow‐up surveys). We note that, of the 1903 participants who completed study entry, 1453 participants completed two or more of the surveys and 1106 participants (58.1%) completed all questions on Survey 6 (Table 2) which relate to participation in prevention trials.

FIGURE 1.

Flow chart of the study.

TABLE 2.

Interest in participation in prevention trial based on risk of developing MS.

Assuming 3% lifetime risk of developing MS (based on having a first‐degree relative with MS)				Assuming 20% lifetime risk of developing MS
	Yes	Undecided	No		Yes	Undecided	No
Participate? (n = 1057)	48% (n = 511)	37% (n = 390)	15% (n = 156)	Participate? (n = 949)	75% (n = 710)	23% (n = 216)	2% (n = 23)
If Yes, willingness to participate by different trial agents (n = 511)				If Yes, willingness to participate by different trial agents (n = 710)
	Novel vaccine	Yes No Undecided	59% (n = 303) 4% (n = 20) 37% (n = 188)		Novel vaccine	Yes No Undecided	64% (n = 456) 2% (n = 15) 34% (n = 239)
	Known vaccine	Yes No Undecided	89% (n = 454) 1% (n = 6) 11% (n = 51)		Known vaccine	Yes No Undecided	93% (n = 660) 0.4% (n = 3) 6.6% (n = 47)
	Novel daily oral pill	Yes No Undecided	49% (n = 248) 9% (n = 44) 42% (n = 219)		Novel daily oral pill	Yes No Undecided	58% (n = 414) 5% (n = 34) 37% (n = 262)
	Known daily pill	Yes No Undecided	67% (n = 344) 5% (n = 27) 28% (n = 140)		Known daily pill	Yes No Undecided	75% (n = 534) 3% (n = 21) 22% (n = 155)

In this manuscript, the 1903 participants who completed study entry were used for all downstream analyses (see Figure S2 for the geographical distribution of the participants to the initial survey). Participants were mostly female (n = 1515, 80%) and had a mean age of 33 years (SD 8.45 years). The maximum number of affected relatives was 9 family members with MS by one participant (Table S3). Of the 1903 participants (first degree relative of a person with MS) who completed the baseline data collection, 8% (n = 141) had an existing diagnosis of MS at the time of enrollment, 0.9% (n = 18) converted to MS between the initial and last follow‐up surveys, and 91% (n = 1744) remained without a diagnosis (Table 1 and Figure 1 for the detailed flow chart). We refer to participants without an MS diagnosis as asymptomatic family members. Given the nationwide nature of the study, MRI were not obtained; thus, it is possible that a small proportion of family members could have had RIS at study entry or fulfilled RIS criteria at some point during the study. The analysis here focuses on participants who are symptomatic and received an MS diagnosis.

TABLE 1.

Comparison of the demographic parameters: Existing diagnosis of MS, MS converters, and asymptomatic first‐degree family members at GEMS study enrollment.

Variables	MS at enrollment (n = 141)	MS converter between initial survey 6 (n = 18)	No MS at seventh survey (n = 1744)	p (MS v. No MS)	p (MS v. Converter)	p (Converter v. No MS)
Age at enrollment	40 [34, 45]	37 [32.50, 40.75] a	33 [27, 40], 17 NR	0.001	0.47	0.01
BMI at study enrollment	25.33 [22.14, 29.98]	25.07 [21.57, 29.99]	24.97 [22.14, 28.9], 6 NR	0.15	0.84	0.59
BMI at age 18	21.91 [19.76, 24.27]	22.21 [19.57, 25.36]	21.79 [19.97, 24.27], 6 NR	0.48	0.42	0.33
# family members with MS, mean (SD)	1.74 (1.10), 11 NR	1.61 (1.09)	1.36 (0.78), 35 NR	< 0.001	0.62	0.35
GERS	11.69 [10.83, 12.25], 34 NA	11.21 [10.59, 12.07], 1 NA	11.14 [10.41, 11.84], 707 NA	< 0.001	0.63	0.41
Gender
Women	123 (87%)	16 (89%)	1376 (79%)	0.02	0.99	0.45
Men	18 (13%)	2 (11%)	368 (21%)
Race
White	135 (96%)	16 (88%)	1661 (95.3%)	0.92	0.02	0.003
Black b	3 (2%)	0 (0%)	30 (1.7%)
Multi‐race	3 (2%)	1 (6%)	31 (1.8%)
Asian c	—	—	5 (0.3%)
Native American d	—	—	2 (0.1%)
Not sure	—	1 (6%)	4 (0.2%)
Other	—	—	11 (0.6%)
Current smoker
Yes	22 (16%)	3 (17%)	150 (8.5%)	0.003	0.86	0.21
No	32 (22%)	5 (28%)	313 (18%)
NR	87 (62%)	10 (55%)	1281 (73.5%)
History of infectious mononucleosis
Yes	42 (30%)	4 (22%)	421 (24%)	0.16	0.69	0.99
No	99 (70%)	14 (78%)	1323 (76%)
Age at infectious mononucleosis diagnosis	16 [14, 19], 100 NR	14 [10.25, 19.75], 14 NR	17 [15, 19.75], 1326 NR	0.26	0.86	0.42

Note: The median and [1st quartile, 3rd quartile] are represented for all continuous variables except the number of family members with MS which is represented using mean (standard deviation). NA: indicates the metric was not available for some participants; NR: participants did not respond to the specific survey question. Age, BMI, the number of family members with MS, age at infectious mononucleosis diagnosis, and GERS scores are compared between each pair of MS groups (MS at enrollment, MS converter between initial Survey 6, and no MS at seventh year survey) using Student's t‐test as these variables follow normal distribution. Chi‐squared test was performed to test the independence between each pair of MS groups (MS at enrollment, MS converter between initial Survey 6, and no MS at seventh year survey) and categorical variables such as sex, race, whether having a family member with MS, smoking status, and history of infectious mononucleosis.

^a Age at diagnosis.

^b African or African American.

^c Asian (e.g., Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese).

^d Native (e.g., American Indian or Alaskan Native or Native Hawaiian).

Participants with MS at the time of enrollment were significantly older, more likely to be a smoker, and more likely to be female (Table 1, p < 0.05). Participants with MS at enrollment also had a significantly higher number of family members with MS and a higher GERS score compared to those without MS (p < 0.001) (Table 1). Table 1 also summarizes evaluations of known MS risk factors in our cohort including adolescent obesity, smoking, infectious mononucleosis, age at infectious mononucleosis diagnosis, and GERS score.

3.2. MS Converters

Over the course of the study, 18 participants without MS at the initial survey were diagnosed with incident MS. This group will be referred to as MS converters, 44% (n = 8) of whom have a confirmed diagnosis by our study team, 28% (n = 5) of whom did not share documentation but provided a date of diagnosis, and 28% (n = 5) without documentation or date of diagnosis but provided two or more consecutive and consistent reports of an MS diagnosis in follow‐up questionnaires. The median duration of time to conversion was 2 years with a median age of 37 at the time of diagnosis. Most were female (89.5%) and non‐Hispanic (94.7%). During the year of diagnosis, 5.5% (n = 1) of MS converters reported an additional neurological disorder (1 viral encephalitis) and 33.3% (n = 6) of MS converters reported an additional non‐neurological disorder (Table S4). Most of the participants were treated after diagnosis: Dimethyl Fumarate (n = 6), Interferon beta (n = 3), and Glatiramer Acetate (n = 2).

3.3. Incidence Rate

The incidence rate of MS was calculated among the participants who did not have an MS diagnosis at the initial survey (n = 1762, 1903 participants minus 141 participants who reported an MS diagnosis on the initial survey at study entry). The maximum follow‐up time was 11 years, and the follow‐up time from the date of enrollment varied among participants. Overall, we report 8526 person‐years of observation for the 1762 participants who did not have a diagnosis of MS at the initial survey and an incidence rate of 211 cases of incident MS—with a 95% confidence interval of 113 to 308 cases—per 100,000 asymptomatic first‐degree family members per year. Thus, the rate of clinical conversion among all of the asymptomatic first‐degree relatives was approximately 1/500 participants per year with an average age of 33 at conversion. Additionally, we performed a power analysis to identify the minimum sample size needed for MS prevention studies (similar in design to ours) with different age range groups. Based on the power calculation, at least 1034 participants are needed to achieve 80% power. This required sample size for 80% power is lower than the number of participants in the current study. Since the average age of our GEMS participants is very similar to the average age of onset of MS, we have likely missed many incident cases among younger individuals. If the incidence rate were to double when recruitment targets a younger cohort, a sample size of 400 participants would provide 80% power.

3.4. Risk Score Comparison

Using our prior GERS that integrated major identified genetic (64 MS susceptibility variants) and environmental risk factors (female sex, history of infectious mononucleosis, and current smoking status) [13], we compared the distribution of GERS between the three participant groups (Table 1; n = 1161 with available GERS score). Participants with MS at enrollment had significantly higher GERS compared to those family members without MS (p < 0.01), consistent with expectation for the distribution of MS risk alleles discovered in the general population: the subset of family members with MS is more likely to harbor genetic risk alleles and to have been exposed to risk‐associated exposures. Further, the distribution of GERS was not statistically different between MS converters and those without MS (p = 0.41) or those participants with MS at study entry (p = 0.63). The small number of converters (n = 18) provides inadequate statistical power to answer this question effectively.

Of note, our previous GRS (GRS⁶⁴) was calculated based on the significant genetic loci identified in major genome‐wide meta‐analyses published prior to the onset of the GEMS study. Taking advantage of the most recent update in the list of genetic susceptibility loci associated with MS [26], we calculated an updated GRS using 224 SNPs (GRS²²⁴) [19] for a subset of our individuals who had available genome‐wide genotyping data (n = 111 among 1161 who had previously undergone targeted genotyping for the 64 SNPs in the prior version of GRS). Similar to the previous GRS (GRS⁶⁴), the updated GRS (GRS²²⁴) included susceptibility loci from both the Major Histocompatibility Complex (MHC) and non‐MHC regions of the genome. However, the updated list included a larger number of susceptibility loci (5 MHC and 59 non‐MHC loci in GRS⁶⁴ vs. 31 MHC and 193 non‐MHC loci in GRS²²⁴). Evaluating the updated GRS (Figure 2), we found a significant and positive correlation between GRS²²⁴, GRS⁶⁴, and GERS⁶⁴ (r = 0.85 between GRS²²⁴ and GRS⁶⁴, r = 0.81 between GERS⁶⁴ and GRS⁶⁴, and r = 0.80 between GERS⁶⁴ and GRS²²⁴, p < 0.001). This suggests that the updated GRS behaves similarly to the previously published GERS and GRS. Additionally, Figure 1 shows that some participants had a strong correlation between GRS⁶⁴ and GERS⁶⁴, which might be due to the limited number of environmental risk factors used to calculate the GERS⁶⁴ score. Finally, we calculated the area under the ROC curve (AUC) for GRS and GERS scores to identify their accuracy in classifying the participants as converters vs. non‐converters. The AUC was moderate and similar for GRS and GERS (0.5688 for GRS and 0.5561 for GERS).

FIGURE 2.

The scatter plot of the updated and original genetic risk scores (GRS²²⁴ and GRS⁶⁴) and genetic and environmental risk score (GERS⁶⁴) plotted for the participants who commonly had at least two of these risk scores. The Spearman's correlation was performed to correlate each pair of risk scores. The Spearman's correlation coefficient (r) was given in the text to assess the amplitude of the correlation.

3.5. Consideration for the Design or Future MS Prevention Trials

Survey 6 included new questions related to the willingness to participate in a future prevention trial given that the participant has a first degree relative with MS. 1117 participants responded to the question regarding participating in a trial to test a treatment to prevent MS (See Figure 1 for the detailed flow chart). 60 of these participants had MS at the initial survey or converted to MS between the initial survey and Survey 6. Of the 1057 participants who responded to this question and remain asymptomatic, 48% (n = 511) were willing to participate, 37% (n = 390) were undecided, and 15% (n = 156) were not interested (Table 2). Of the 1057 participants, 90% (n = 949) were willing to take a genetic test via blood samples to determine if they are at high risk of developing MS, while 8% (n = 83) were undecided and 2% (n = 25) were unwilling. Of the 949 participants that responded to the hypothetical scenario where they were found to have elevated risk (20%) of developing MS based on the genetic tests performed with blood samples, interest in participating in trials increased to 75% (n = 710), while 23% (n = 216) were undecided and 2% (n = 23) were not interested. Of note, 199 participants who previously were undecided or not interested subsequently were willing to participate if they were found to have a higher risk of developing MS with the theoretical biomarker test.

Of the 511 participants who were interested in participating in a clinical trial testing the efficacy of a medication, the highest willingness was with an existing, approved vaccine (89%; n = 454) followed by an existing, approved daily oral pill (67%; n = 344). When presented with the hypothetical scenario of having a greater chance of developing MS (> 20%), most participants were willing to join a study with a known vaccine (93%; n = 660) or a known daily oral pill (75%; n = 534). Although willingness with novel agents was lower than known agents, the majority of participants were still willing to participate with novel vaccines (64%; n = 456) and novel oral pills (58%; n = 414).

4. Discussion

In this prospective cohort study of first‐degree relatives of persons with MS, we have collected 7 surveys since 2011. In 2016, we reported an initial estimate of the incidence rate of MS in our study population [13]. We now offer an extension of that study, with good long‐term participant retention of 54% over 11 years despite limited engagement longitudinally with the study team given limited resources. 1020 participants completed the latest survey in 2022. With the longer follow‐up time and greater number of converters, we provide an updated incidence rate for a clinical diagnosis of MS with symptomatic manifestations of 211 cases with a 95% confidence interval of 113 to 308 cases per 100,000 first‐degree family members annually, which is 100 times the average incidence rate of the general population [27]. Further, we can use the GERS approach leveraging genetic and environmental risk factors to further stratify family members, with the top 28% of relatives having twice the risk of the average family member [13]. Thus, the annual incidence rate for a clinical diagnosis of MS in stratified, high‐risk first‐degree family members could be approximately 1 in 250 participants, making prevention studies in this population feasible. We note that our participants are, on average, the same age at study entry as the average age of MS onset (33 years) [28], so over‐sampling younger individuals and older children in future studies should further enhance the rate of incidence. Importantly, 90% of participants showed interest in having a blood test to produce a more precise estimate of risk based on genetic information, and nearly one‐half of the respondents were interested in participating in a future MS prevention clinical trial. This proportion increased substantially if participants are presented with information that they have an elevated risk of developing MS based on the genetic tests performed with blood samples, over their enhanced baseline risk as first‐degree relatives of a person with MS.

Multiple clinical trials have demonstrated that early therapeutic intervention reduces the risk of a future relapse and disability in MS [29, 30, 31, 32, 33], and the successful trials for RIS have recently extended the narrative that earlier intervention in the trajectory to a clinical diagnosis (with symptomatic manifestations) of MS is clinically meaningful [8, 9]. Thus, extending interventions to the presymptomatic, pre‐radiologic phase of the disease—where the involvement of the central nervous system is not yet detected—is the logical next step. Our study further supports the use of GERS as a tool to generate an individualized risk estimate for the development of MS among family members to identify the subset that is at the highest risk of developing MS. However, genetic and environmental risk factors may not be sufficient to drive clinical decision making, and more work is needed to better delineate the early molecular events in the transition to MS—such as RNA measures of altered T cell reactivity [19] or the presence of plasma cells [34] or oligoclonal bands in the CSF. Serum markers of axonal damage, such as neurofilament light chain, may be elevated in presymptomatic patients who eventually develop MS [35]. From an imaging standpoint, central vein sign, paramagnetic rim lesion [36], slowly expanding lesions have emerged as biomarkers for compartmentalized smoldering disease, and our earlier GEMS study suggested that MRI measures offer promise in this area [12]. Conceptually, a strict, fixed causal chain of events leading to the onset of MS is unlikely given our understanding of risk factors today: none of them are deterministic. There are many people who develop MS without having been obese, having smoked, or having all of the risk alleles for the genetic variants associated with MS. Thus, consistent with the etiology of most common diseases (inflammatory or otherwise), it is more likely that the onset of MS is a probabilistic event, with many different types of factors contributing to an additive model that may include some synergistic interactions. The goal of primary prevention is to reduce the probability that the CNS is damaged by an inflammatory response; the earliest stage of the latter process is RIS, in which disruption of the CNS parenchyma is evident on MRI.

For consistency, we have implemented the 2010 MS diagnostic criteria which was stipulated in the study protocol at its launch in 2011. In light of the 2024 MS criteria, which recognize asymptomatic MS (RIS), our proposed approach to early intervention and prevention becomes even more relevant. The inclusion of asymptomatic MS is likely to broaden access to treatment worldwide, providing an opportunity to target individuals at high risk before the onset of clinical symptoms. This aligns with our goal of intervening early to prevent or delay the biological progression of MS, including any tissue damage seen in the CNS by MRI. By utilizing genetic and environmental risk factors to identify these high‐risk individuals, we can initiate timely interventions that may significantly alter the disease trajectory or prevent the onset of MS. As such, our findings support the need for primary prevention strategies and underscore the importance of developing targeted trials that leverage these evolving diagnostic criteria to address MS before it manifests clinically or radiologically.

Ultimately, our goal is to design trials of primary prevention of MS in high‐risk individuals. Our questionnaire suggests that nearly half of GEMS participants are interested in preventive trials and that this proportion is increased substantially if a risk‐assessment tool is deployed (Table 2). Not surprisingly, enthusiasm is higher for approved therapeutic agents and for agents such as a vaccine that do not require continuous drug exposure to prevent disease onset, compared to investigational drugs. This survey is therefore very timely in informing study design for the group that is most likely to power a trial in individuals at risk of MS, as discordant monozygotic twins are too small in number to provide a robust study population.

Our study presents some limitations, chief of which is participant retention. Participant response rates declined gradually over time, perhaps due to diminished study engagement or changes in contact information. More frequent and enhanced interaction with participants will be necessary to ensure a slower rate of attrition. A second limitation is assumption bias of MS diagnosis, especially for MS converters: some MS converters were unable to be reached via multiple communication modalities by study personnel to gather clinical notes to validate the diagnosis. However, all participants who provided documentation have a confirmed diagnosis. Third, the number of MS converters remains small, limiting the statistical power in our current analyses. Fourth, our genotype data for most individuals is limited to 64 targeted SNPs; however, in the subset of participants that we genotyped genome‐wide (n = 111), we found that the original GRS (or GRS⁶⁴) is highly correlated to the updated GRS (GRS²²⁴). Thus, while the GRS⁶⁴ may be less precise, the interpretation of its results remains relevant. The GRS/GERS are likely to be improved by the incorporation of polygenic scores which leverage a larger fraction of the human genome. There is also a limitation in generalizability with regards to gender, self‐reported race, and ethnicity. These limitations may be driven by the original call for participation that was performed over a decade ago through a national foundation that supports MS research; since then, efforts towards diversifying the engagement of participants with MS will have improved the yield of future efforts of recruitment. The over‐representation of women most likely stems from the general skew of greater participation by women in human research studies and of the well‐publicized excess susceptibility of women to MS, which may increase interest in participation in our study by first‐degree relatives who are women. Higher rate of women participants might have resulted in a higher incidence rate as women are more likely to develop MS compared to men. The GEMS project will address this challenge as a new wave of recruitment is initiated by having focused recruitment and education efforts among men and non‐white populations to accurately reflect the demographics of the United States. For now, the generalizability of our results should be considered carefully. We also have to consider the possibility that participation in our study may have had an impact on the patient's behavior and may have somewhat increased the likelihood that a participant seeks evaluation for a complaint and is diagnosed with MS over the course of the study. Another limitation is the reliance on self‐reported data to identify asymptomatic individuals, given the size of the study, its national scope and limited resources; without MRI evidence to detect subclinical evidence of diseases, such as cases of RIS, our study is likely to underestimate the incidence of clinically relevant inflammatory demyelination in the at‐risk population. The lack of MRI in this nationwide study also introduces the limitation that some of our participants could have had an RIS at study entry. The non‐MS converters were also significantly younger, which introduces a potential time bias as they may not have enough time to show disease progression. This difference in age could contribute to an overestimation of the non‐conversion rates in multiple sclerosis and underestimation of incidence rate. The total number of MS converters was small; therefore, we could not provide an age‐adjusted incidence rate. Future studies with a higher number of MS converters are needed to provide an incidence rate for different age groups.

Overall, our analyses provide important information with which to design the next generation of studies for individuals at risk of MS: we need to (1) oversample the subset of family members with a high GERS score, (2) increase the study size, (3) include biomarkers (such as MRI) in addition to a clinical diagnosis as the outcome measure, and (4) focus on younger participants such as adolescents and young adults under 25, as the current study recruited participants > 18 and < 50 years of age. The latter observation is also informed by the fact that a small proportion of MS cases present symptomatically in childhood and that several risk factors such as menarche, exposure to EBV and adolescent obesity operate before the age of 18 years [28]; it is likely that the earliest molecular and cellular stages of MS start in childhood. Thus, the next iteration of GEMS will focus on recruiting older children, adolescents and young adults to capture as many family members as possible before microscopic and eventually radiologically and clinically manifest CNS disease emerges. Our findings support the feasibility of primary prevention trials in individuals at risk for MS and pave the way for the first generation of such studies to be designed.

Author Contributions

A.L., C.T., and P.L.D. contributed to the conception and design of the study, analysis of data, drafting the text, and preparing the figures. T.R., F.D., K.O., and J.S. contributed to the acquisition of data. Z.X. contributed to the drafting the text. P.L.D. obtained funding for the study.

Funding

This work was supported by the National Multiple Sclerosis Society, FG‐2008‐36976, R1708‐28972, TA‐2204‐39428.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Appendix S1: acn370340‐sup‐0001‐AppendixS1.docx.

Appendix S2: acn370340‐sup‐0002‐AppendixS2.pdf.

Appendix S3: acn370340‐sup‐0003‐AppendixS3.pdf.

Funding Statement

This work was funded by National Multiple Sclerosis Society grants FG‐2008‐36976, R1708‐28972, and TA‐2204‐39428.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.