In this study, we report good correlation of the patient-reported MSRS measure developed by PLM with standard clinical assessment measures used in MS, such as the EDSS. The availability of this online instrument and of the large PLM membership provides a useful, complementary platform for MS clinical research. It does not replace rigorous, prospective cohort studies or targeted clinical research projects performed in academic MS centers; instead, it offers a different option, that has an opportunity to leverage the frequent sampling of data from a large subject population.
While we found small demographic differences between PLM members with recent profile updates and patients at our MS center, they were modest in absolute terms. It should be noted that both sets of subjects represent somewhat biased subsets of the overall MS patient population; this limitation is somewhat mitigated by their large sample size but needs to be clearly stated. Interestingly, the differences that we describe between the patient populations we compared – including that PLM members had a younger age, higher education, larger proportion of women and lower proportion of self-reported white non-Hispanic origin - are slightly at odds with other demographic surveys of users of online tools, that have noted greater use and more frequent use by individuals who are white, female, older, with higher incomes, with good internet skills, and who are not employed 
. We limited our comparative analysis to the subset of PLM members with recent profile updates, in order to examine active members who are likely to participate in ongoing online investigations; however, it is possible that these individuals differ in demographic terms from the general PLM population. The greater proportion of interferon beta-treated PLM members may partially reflect the fact that they had more relapsing onset disease. We also compared the PLM population to the 31,232 respondents from NARCOMS, a large and well-described registry-based cohort 
). Here, we found statistically significant but small absolute differences in demographic characteristics, further suggesting that the PLM members, offer a reasonable alternative platform for MS clinical research that could yield insights that are generalizable to other MS populations after appropriate control for confounders.
In this study, we explored the use of PLM to answer a timely clinical question in which the current literature offers conflicting results: the relationship between excess body weight and MS disease course in the North American population that is becoming increasingly heavier. Obesity in an individual’s mother 
or during adolescence of the individual 
may increase the risk for MS and perhaps for a relapsing onset 
. However, in adulthood, insofar as it is possible to disentangle the effects of age, disability and comorbidities, individuals with MS may have lower body mass index (BMI) than age-matched controls 
. Only one prior study, by the North American Research Committee on Multiple Sclerosis (NARCOMS), examined longitudinal data and suggested that fluctuations in BMI are not associated with changes in disease severity as measured by Patient-determined disease steps 
. Here, using a large sample size and prospectively collected data, we report consistent results suggesting that there is no evidence for increased BMI having a strong effect on disability in MS: we noted a very modest correlation between BMI and MSRS in the cross-sectional analysis (rho
1336) and only a nominally significant association in the analysis of the longitudinal MSRS data (estimate (95% CI)
0.0133 (−0.00013, 0.0267)), suggesting that there is no clinically meaningful role for BMI in disability for MS in most individuals. Individuals do tend to under-estimate their BMI in self-reports 
; however, because we examined BMI as a continuous variable this under-estimation should not influence the statistical estimate of this relationship on longitudinal course. However, we cannot exclude the possibility that BMI may have a very weak effect on the accumulation of disability, that it may influence other measures of disability in MS or that it may have a role in a subset of individuals. Further, the adverse health consequences of high BMI are well documented and doubtlessly impact the life course of MS patients 
, irrespective of a direct effect of BMI on MS disability.
Our findings and those of others relating BMI and disease severity in adults 
, contrast with the increased risk of MS associated with higher BMI during adolescence noted in prior studies 
, possibly through immunologic modulation by hormones such as leptin and adiponectin 
. This disjunction of a risk factor’s role in susceptibility and disease course has been suggested for other susceptibility factors such as EBV infection 
and genetic risk factors 
, but not for cigarette smoking 
and vitamin D levels 
, which may influence both disease susceptibility and disease course. Additionally, because high BMI is associated with low vitamin D status, the association between BMI and risk of MS may in fact be confounded by vitamin D status 
. As we better understand the functional consequences of these various risk factors, we may begin to see how they influence downstream molecular events leading to MS onset and/or disability.
There are challenges and limitations to online research platforms, including biases arising from the digital divide, issues of privacy, autonomy and data storage, variable and inconsistent data sampling schedules, and lack of objective validation of reported data such as height and weight 
. Also, in the absence of a gold standard for patient reported outcomes, it is not feasible to determine whether the physician or patient is “right”, only to understand the broader areas of agreement and discordance. For example, a patient may be more aware of sensory disturbances, but a clinician may better appreciate the broader range of gait disturbance seen in clinical practice. In this study a bias towards greater MSRS agreement between provider and patient may have existed, as the provider relied on the patient during the clinical interview to answer some questions, such as those pertaining to swallowing. Also, as with any clinical instrument, test-retest reliability is critical to enable powerful analysis; while we did not assess this feature of the MSRS as part of our study, the reliability of the MSRS has been previously evaluated and revealed one-week test-retest correlations of r
0.91 (p<0.001) 
. While not as comprehensive as a clinical examination or as biomedically grounded as an MRI scan, a patient reported outcome of disability such as the MSRS has the advantage of being rapid and free to administer; it also has the potential to enable patients themselves to learn more about the progression of their illness through self-monitoring. Further, missing data and incomplete participation by many members are important limitations that require rigorous quality control measures such as the ones that we implemented. In this study, from an initial 28,025 members reporting at least some information pertaining to MS, limiting the analysis to subjects with at least two separate dates of data entry in the study period and implementing the other quality measures reduced the analysis to a final N of 10,255 subjects. Finally, because online platforms are relatively new, the follow up time may not yet be sufficient long to detect clinically meaningful changes in the MSRS.
In conclusion, online research remains in its infancy and continues to pose ethical and methodological challenges. Nonetheless, the opportunity of capturing an assessment of disability on a very frequent basis opens exciting new possibilities that are not possible using more episodic data collection strategies such as those collected by NARCOMS. Prospective tracking of symptoms in “real time” may eventually allow researchers to assess whether certain symptoms such as increases in fatigue or fluctuations in “brain fog” can serve as harbingers for clinical attacks or be useful in assessing a patient’s clinical trajectory or response to disease-modifying therapy. The PLM member population therefore offers an intriguing new platform for MS investigations. Its results, after appropriate consideration of potential demographic and other confounders, may be generalizable to and replicable in the larger patient population seen in large MS centers.