According to the National Institutes of Health (NIH, USA), biomarkers are objectively measured indicators of physiologic processes, pathologic processes or pharmacologic responses to a therapeutic intervention ^[1][2][6,7]. In MS, they can be subdivided into diagnostic (help to differentiate between different diseases, e.g., anti-aquaporin-4 antibodies, oligoclonal bands, etc.), prognostic (enable physicians to estimate how a disease might develop once it has been diagnosed, e.g., neurofilaments, oligoclonal bands, etc.), predictive (predict the treatment response and thus help to decide which patient is most likely to benefit from a certain treatment), disease activity (measure the inflammatory/neurodegenerative components of the disease, e.g., MRI, clinical parameters, etc.) and treatment response (responders versus non-responders of a certain treatment) biomarkers ^[1][6]. Especially with the focus on personalized medicine in pwMS, treatment response biomarkers can enable neurologists to differentiate patients regarding efficacy (e.g., neurofilament light chains, neutralizing antibodies against interferon-ß or natalizumab) or potential side effects (e.g., anti-varicella zoster virus antibodies, anti-John Cunningham virus antibodies) of a certain treatment ^[1][6]. The collection of such data is crucial to adapt the treatment of each patient individually to his/her results. However, it is also time-consuming if these data have to be gathered by physicians or other healthcare staff. With the increasing digitalization of healthcare, medicine now gains access to a new type of biomarker. So-called digital biomarkers enable the translation of up-to-date new data sources into informative, actionable knowledge. They can be used by healthcare professionals (HCPs) by implementing digital devices in their assessment (e.g., MRI, optical coherence tomography (OCT) and tablet-based neurostatus); they also enable data collection directly from the patient. They can collect such data directly as part of disease management on a regular basis, and thus ensure good monitoring and a prompt reaction to the progression of MS and the worsening of symptoms. Digital biomarkers mean objective, quantifiable physiological and behavioral data that are measured and collected by digital devices. The data collected by, e.g., portables, wearables, implantables or digestibles are typically used to generate, influence and/or predict health-related outcomes, and thus represent deep digital phenotyping, collecting clinically meaningful and objective digital data ^[3][8]. As digital technologies are usually less expensive than the process of collecting these data face to face, and as some of these data can be collected even without patients being actively involved (passive monitoring, e.g., by the use of wearables) data can also be collected more frequently and longitudinally. Health-related outcomes can vary, from explaining health and disease states, predicting drug responses or influencing health behaviors. In addition to this rather strict definition of digital biomarkers, digitalization in medicine also includes patient-reported measures (e.g., survey data), genetic information and other data that now can be collected by digital infrastructure. These data can complement the mentioned digital biomarkers, creating a digital multidimensional dataset.

Due to the technological transformation of healthcare, new technologies are leveraged to generate, track and collect new data. With the wealth of novel data, the responsibility is on the system to turn them into promising information that helps clinicians, researchers, patients and entrepreneurs to better understand states of disease and health ^[4][9].

The path to implementing digital biomarkers in the clinic is complex, because the benefits that can be achieved by the use of digital biomarkers come with significant challenges (Table 1).

Digital biomarkers will, at least, face the same regulatory requirements as traditional biomarkers, and need to be tested for feasibility and reliability. The knowledge on how to establish and validate digital biomarkers is still limited. It can be challenging to identify relevant data and analyze them, and especially difficult in terms of how to use accurate baselines to relate this data for evaluation ^[5][10]. On the other hand, collecting continuous real-time data out of the patient’s everyday life closes the data gap between visits, and thus can reveal changes in the disease course as soon as they occur. A continuous dataflow from patients to their treating physician could generate a big dataset that shows real-world evidence, therefore being more meaningful and enabling faster decision making. This is only possible with patients who are carefully educated about the need for such sensitive data and demonstrate appropriate adherence. To avoid patients getting obsessed with even minor, non-significant changes, as to decrease the potential of over-reactions and increased anxiety, networking between physician and patient is crucial to evaluate and discuss the significance of these biomarkers. Besides necessary reflections on data security and the possibility to store these data over a long period of time, a huge dataset arises through the use of digital devices, which requires complex analyses.

Digital biomarkers have great potential for medical domains that are not well-understood, especially if digital biomarkers lead the way to phenotypic signatures. Challenges around infrastructure, evidence generation, consistent data collection and workflow remain.

To be seen less as a challenge than as an aspect to be considered is the distribution and availability of digital devices for data collection. Not every patient can afford to buy wearables or a smartphone to collect their data during their everyday life. In addition, some patients will have difficulties with their usage, due to age-related reasons or impairments that prevent the handling of digital devices.

Digital biomarkers are basically collected by digital tools. A way to classify these measures focuses on what has been measured, and the added clinical value derived from that data. At this, measurements can be familiar, such as the measurement of blood pressure, or innovative, such as the continuous measurement of blood pressure. A known clinical value is one that is well-understood and has previously been validated., e.g., blood pressure can be used as an indicator of cardiovascular risk. Alternatively, the known measurement can additionally be used to detect a new finding, linking blood pressure to, e.g., major depression. These different digital biomarker categories will influence the level of evidence required for regulatory approval, validation and clinical implementation ^[2][4][7,9].

Due to the increasing digitalization of health, a growing amount of patient data can be collected digitally in the care of pwMS (Figure 1). This not only refers to digital assessment results during clinical visits, but also daily patient-driven data collection, e.g., via the usage of smart devices, such as motion sensors, that arouse great interest in characterizing lifelong MS disease in a more granular way.

Figure 1 shows the five steps in digital clinical assessment from where we are now to where the future of digital clinical assessments could be. The typical clinical examination is still for the most part paper-based (except MRI, which is already digital), with, at best, subsequent digital storage of scanned documents in the hospital information system (step one). Digital clinical evaluations of, e.g., gait, patients’ perception regarding symptoms (patient-reported outcomes) or the digital version of the Multiple Sclerosis Functional Composite (MSPT; Section 3.4) are not available for every neurologic practice or hospital for use in clinical routine, but are available mostly as part of clinical trials (step two). Digital biomarkers cannot only be collected actively. Additionally, passive monitoring and data collection are possible using, e.g., voice analysis during calls with patients (step three). As step two relates to digital data collection at given points in time during patients’ visits, step three is already the transition towards data collection outside the clinical setting (e.g., passive collection of mobility via smartphones). Symptoms can vary over time, and disease progression may therefore be detected too late. For this reason, real-life monitoring is crucial (step four). Future devices could be smart applications, such as mirrors that automatically recognize body temperature and mood (step five).

Increasing evidence supports a forward-thinking chance of treatment decisions due to inter-individual highly variable clinical presentation, the extent of disease progression and a growing amount of defining biomarkers and surrogate endpoints, which personalize each disease presentation and favor our objective of a tailored treatment approach ^[6][7][8][9][1,11,12,13].

In the subsequent chapters, we will focus on digital biomarkers collected to investigate the involved functional systems or subdomains that are affected by different topographic lesions that occur during the course of MS. As MS is such a multidimensional disease, affecting different functional systems, collecting digital biomarkers capturing changes in those systems can offer insights into a comprehensively personalized disease.