It is a fact often noted that life has changed beyond recognition from the dawn of large-scale societies 10,000 years ago. These changes have been driven by the uniquely human capability to build “technologies,” a broad concept that simply means the practical application of scientific knowledge. This knowledge generates its own virtuous cycle of new technologies building on the old, accelerating rates of change. Transistors were invented 70 years ago and now more than two-thirds of the world’s population carry an easily searchable Library of Alexandria and global communication device in their pocket.
Encapsulated by the concept of Moore's Law, this exponential growth has meant electronic technology has increasingly permeated all aspects of our lives, as hardware costs have dropped and technology has gotten "better" at what it does, whether through power, accuracy, ease of use, etc. We tend to think of electronic technologies as tools for improving our lives and serving data to us, but they are equally good at capturing data on those interacting with them, to the extent that it forms the basis of the business models of the world’s most successful companies. This has meant data are increasingly cheap to collect, whether defined by simple economic cost or more hazy notions of "burden" on users.
The pharmaceutical industry has not been sheltered from this evolution, and arguably finds itself in the interesting position of having more data than it knows what to do with. However, it is also facing the challenge of deciding what data to measure. Gaining insight into the patient experience of a disease or treatment is a vital element of assessing healthcare and gaining approval for new drugs. Technology has had a central role in the growth of patient reported outcomes (PRO) in the clinical trials space; while estimates vary, the use of electronic versions of clinical outcome assessments (eCOA) in clinical trials has been on a significant upwards trend for the last decade.
However, traditional eCOA has largely been about taking a paper version of a questionnaire and implementing it on a touch-screen device. While this has brought great advantages in terms of, among other things, improved data quality, it has not fundamentally changed the kind of information we’re capturing from patients. In the last few years there has been great interest in the use of wearable devices and sensors in clinical trials to capture “objective” data from patients.
It is currently an area of intense discussion within the industry as to how the data from these kinds of devices will be used in support of trial endpoints. Broadly, there are four immediate targets:
- Conducting defined performance tasks outside of a clinic setting (e.g. 6 minute walk test);
- Using sensor data itself as an endpoint (e.g. change in activity levels);
- Using a sensor measure in conjunction with a PRO measure as an endpoint (e.g. change in activity levels associated with a change in self-reported symptoms);
- Using a sensor measure to trigger more targeted PRO assessment (e.g. a measured increase in activity levels triggering a questionnaire asking about any associated change in symptoms)
These targets are relatively simple and fit within our current framework for endpoint development (to date, only 1 and 2 are reasonably well established), but they form a basis for exploring more advanced sensor-driven outcomes. For example, the use of “passive” data - that is, data captured from people with no need for them to actively engage with a device - holds great promise. This kind of data can provide a shocking level of insight into one's life, something brought to the public's attention through the Edward Snowden revelations regarding the power of metadata. More pragmatically, basic passive data captured from smartphones (for example, phone calls and SMS exchanges, timing of communication, diversity of communication and physical proximity with others) has been used to identify people's mood, the spread of ideas and weight-gain.
Until Google can simply extract phenomenological states directly from our brains there will be a need to ask patients, “how are you?” Indeed, regulators have been clear that technology cannot simply replace this more human, humane touch in trials. However there is significant room for more nuanced, dynamic insights into patient experience, which can decrease the burden on patients by extrapolating from objective data sources and reducing the number of questionnaires being asked, or asking them at the most appropriate time. The relentless drive of Moore’s Law will ensure these data become increasingly abundant and accurate, and analytic techniques and platforms for bringing together divergent sources of data are essential to bring meaning to this data-heavy world. It remains for the industry, working with patients, to meaningfully define what these data tell us.