For the pharma industry real world evidence (RWE) is the next frontier (Read about in our last blog entry). Data collection is only one half of the prize though. The other is analysing the data and generating valuable insights.
Which is something machines are pretty good at. But they can be picky too. Not any old data jungle will do. As Daniel Keys Moran famously said: “You can have data without information, but you cannot have information without data.”
The evidence source which pharma relies on most are clinical trials (Phase I-III). Of course the number, variety and comorbidities of people studied is limited. Now more and more attention is being paid to what happens when the product reaches the market. This rethinking led to observational and non-interventional studies. Then pharma started to collect RWE (1).
In the digital age, data sources from digital applications and mobile phones are exploding. More data has been created since 2003 than in all of previous recorded history (2). This includes RWE. In 2011, there was still limited use of RWE, mostly focused on safety and post-marketing surveillance. By 2015, RWE was more integrated and involved throughout the product life cycle. In the R&D department its use rose from 30% (in 2011) to 66%. At the same time, disease/treatment understanding rose by 35% due to recognized RWE impact. In 2016, RWE was additionally adopted in regulatory decision making, reimbursement decisions and clinical guidelines (3).
So far, so good. But data warehouses at pharmaceutical companies were originally designed for housing structured clinical data. RWE data is less structured, including for example qualitative or even narrative patient reports on symptoms, adverse events and adherence. Some of the most interesting data stems from online patient chat groups, communities such as patients like me or even consumer apps.
In order to make best use of this type of data, manage and measure successful outcomes from it and ultimately argue for value based reimbursement pharma needs the tools to analyse large and unstructured data sets. Big data and machine learning can help. This means the centralizing and structuring of anonymised data, followed by a labeling process to correctly analyse and use the data (4).
Roche started using smartphone sensors in clinical trials for Parkinson disease patients in 2014. Daily gait and mobility performance of early-stay Parkinson patients were monitored. The goal was to get a deeper insight on how the disease affects these patterns over time. Over 30,000 hours were objectively recorded, generating data free from reporting bias. Data collected by digital sensors in the patients’ smartphones was analyzed by machine learning (Human Activity Recognition (HAR) model using Deep Neutral Networks (DNN)) (5).
“When we saw the first machine learning based analysis from our digital biomarker studies, it became clear that we had made a big step forward, as it opened a window to a more comprehensive understanding of Parkinson’s and the patients’ daily life with the disease,” says F. Lipsmeier, Digital Biomarker Data Analysis Lead from Roche (6).
Preview: Leverage digital solutions from professional partners.
Stay connected, stay healthy!
P.S.: Ariana can provide you with real time patient data – get in touch to learn how she does it.