The birth of a digital biomarker

How to create new digital biomarkers — using MS Sherpa as a showcase

Biomarkers are indicators for a biological condition or state. In medicine, biomarkers are frequently limited to traceable substances such as molecules in blood or proteins in urine. Typical examples are the elevated levels troponin in blood as an indicator of myocardial infarction or PSA (Prostate Specific Antigen) as a marker of prostate cancer. However other measurable quantities can also be considered as a biomarker, such as body temperature as a clinical marker for fever. Currently, the gathering of clinical data and biomarkers is mostly done in healthcare settings such as hospitals. Additionally, many assessments rely on subjective outcome measures as self-reported scores or questionnaires. This common practice of using brief “snapshots” and limited objectivity to guide clinical decision making is currently being challenged by modern technology.

With the current increase of connected and smart technology (internet of things) relevant health information can be obtained from many more sources. At this moment you probably wear a smartphone that is equipped with a wealth of sensors. Those sensors are dragged along with you when you perform everyday tasks. Sensors can also be found in your smartwatch, in your light bulbs, your car navigation system, your thermostat, and in many other places.

This wealth of available personal data is underutilized in healthcare practice. We believe healthcare can be a lot smarter by using these data sources and creating information from them as indicators of personal health. These digital indicators for health care status or condition are called digital biomarkers. They are useful in clinical practice and can guide treatment decisions. But the difficult question is how to go from a sensors data to a digital biomarker?

Creating a digital biomarker

Let’s follow the CTTI guidelines for the development of novel endpoints and see how you can apply these guidelines! These guidelines consist of six steps which are detailed below by the help of the path we have followed in developing our MS sherpa product.

1. Affected health aspect

The first and most important step is identifying what aspect of health is affected by the disease. Which symptoms bother the healthcare professionals most? Which complications limit the patients in their daily activities? Before developing our MS sherpa product we interviewed Multiple Sclerosis (MS) patients and figured out that they were mostly bothered by disease-related fatigue.

2. Scope of the assessment

The second step is to define the scope of the assessment. The assessment should directly measure an unmet need that if relieved, improved, or prevented, is meaningful to patients. Fatigue is a very common symptom in MS patients, but not always properly understood by family, friends or colleagues, who sometimes assume that the patient is just depressed or not trying hard enough. Having an objective measure of fatigue may help patients to get proper recognition of their feelings. Additionally, while the actual cause of MS fatigue is unknown, it may be related to disease activity. If so, measuring fatigue is of importance since disease activity is a target to treat on.

3. Specific measurement

The third step is to find an appropriate and exact measurement. The patients' partners told us that they could tell the fatigue from the patients facial expressions. Obviously, this is difficult to objectively capture in an exact measurement. Therefore we have scanned the scientific literature and consulted key experts in the field. We found out that reaction time to a stimulus was affected by MS fatigue. We, therefore, decided to monitor the eye movements of MS patients. When patients are fatigued, their reaction to a stimulus is significantly delayed (see for example Finke et al.).

4. Suitable mobile technology

As a fourth step, the technology should be suitable to execute the desired measurement. This means that the assessment should be very convenient to do for the user. Patients should be able to easily do a measurement whenever they feel fatigued or believe they are getting fatigued. Since current mobile phones are equipped with all necessary sensors, and people typically have their phones readily available at elbow distance, we picked a mobile phone measurement as the most suitable option.

5. Standards

Measuring in a laboratory environment is one thing, assessments in the field is quite something else. In a laboratory environment, one can more easily control the conditions and environment to optimally determine the parameters of interest. In the field, no-one is there to make sure the laboratory requirements are met. Under what lighting conditions can the patient’s eyes reliably be detected and monitored? How much movement is allowed? Are the mobile sensors sensitive enough to do the job? It is critical to determine the environmental condition thresholds for a reliable measurement. Therefore we created guidelines to measure with certain lightning and distance of the phone to the eyes. In the development of these standards, it is also important to use standards for collection and reporting if possible and be able to show transparency around the workings of the algorithms.

6. Validation

After these steps, it is time to test. To recapitulate: we are interested in assessing disease activity. Since disease activity can be expressed quite differently between persons, we decided to measure within subject: over time we measure the same person. This allows us to correlate the measure not only with the subjective feeling of fatigue but also with relapses of persons with MS and with subclinical detection of lesions in the brain by measurement of MRIs. This immediately sets the stage for the population to measure: Relapsing-remitting patients are the focus.

Of course, the measurements need to be done in different conditions, to validate robustness, sensitivity, specificity, accuracy, precision and other relevant performance characteristics. Also, tolerability and acceptability of the measurement by the participants should be tested. In this testing, all key stakeholders should be involved: patients, clinicians, and regulators.

For MS Sherpa we are currently in the process of validating the digital biomarker as a patient-centered measurement for fatigue and disease activity. We have set up a large study where 100 MS patients are followed for one year and multiple MRIs are taken during this year.

Conclusion

Creation of digital biomarkers is by no means easy. The text above should shed some light on the different steps involved in digital biomarker development. Until now we have learned the following important lessons during our voyage with MS Sherpa:

- Novel measures for the assessment of an unmet need can add real value to patients, caregivers, or the healthcare system in general. In particular, if they help to better understand disease condition or conceptualize treatment benefit.

- The path towards a widely accepted digital biomarker is not easy. The steps above provide guidance and help to make sure the new digital measure ticks all boxes.

- Make sure to work with all key stakeholders. For a new digital biomarker to be successful all stakeholders need to benefit.

- Select the technology after selecting the outcome assessment. In other words: do not push technology but let the user or patient be your guidance.

With special thanks to Hans Weda — Senior Data Scientist @ Orikami data science boutique- for reviewing and editing. Want to read more? Stay tuned for Hans’ next blog: “Eyes are the mirror of the soul - saccadic latency as a digital biomarker for fatigue”