Streamlining clinical data management - part 2: Wearable devices and automation

Concluding our mini-series on clinical trial data management ahead of Pharma IQ Live: Clinical Data Management Excellence, Fausto Artico, Global R&D Tech Head and Director of Innovation and Data Science at GlaxoSmithKline (GSK), explains how to enable data sharing through automation and other streamlined procedures.

Pharma IQ: Wearable devices enable rich datasets, however doctors must still check and input this data into digital systems before it is aggregated by data brokers and made available to the companies running clinical trials. Is there anything that can be done to reduce this hurdle?

Fausto Artico: Data reliability is checked downstream and not by doctors. This is because, due to the amount and variety of the data generated by wearable devices, it simply would not be feasible for doctors to check the reliability of the data.

Downstream, on the other hand, data engineers can design automated procedures and embed “checking rules and heuristics” in the data pipelines that they need to create to clean, link, contextualize and harmonize the data. If data anomalies are discovered by the data pipelines, the engineers and/or the automated procedures that have been statistically validated, can tag some data as suspicious, eliminate the data from the final dataset that will be made available to the data scientists and analysts, or bring the anomalies to the attention of the patients’ doctors to determine what should be done. This could be discarding the data, correcting the data or confirming that the data can reasonably be considered correct even if it is anomalous.

As for the inputting of the data, the wearable or under-the-skin devices can easily communicate with wireless points connected to the internet and have their own Application Program Interfaces (APIs) that can be accessed and used. Furthermore, such devices can seamlessly and automatically transfer data to the wireless points and input the data through the network directly into the IT systems of the pharma companies involved in the clinical trial study.

In future the need for middlemen data brokers will disappear or will diminish. This is because wearable or under-the-skin devices can eliminate or reduce their need, enable closer interactions among the pharma companies, doctors and patients, and reduce the need for doctors to input a lot of data into digital systems – although there will always be some data that will still have to be created and inputted by doctors.

Pharma IQ: What is your experience concerning the general reliability of clinical trial data?

FA: Errors are inevitable. The fact that we think some data is correct and reliable does not mean that it is. For example, there have probably been undiscovered cases where data seemed to be perfectly fine and was considered correct and reliable, but maybe the data was simply of a person other than the one intended.

For a long time, doctors have been taking notes and writing results on paper-based documents before transferring the information they contained to digital systems. The probability that somewhere someone simply swapped two documents or the people’s names while inputting data into a digital system is greater than zero.

I am especially worried about all the procedures that require manual intervention as well as the number of activities that today third parties need to execute to transfer data to pharma companies. The probability of something going wrong increases in proportion to the number of necessary activities and today it is especially difficult to track the origins of problems and very time-consuming to fix them, if possible, after discovery. The long lead times required to deliver data to pharma companies makes it almost impossible to correct and adjust clinical trial execution due to the months necessary to just gather the data.

We can greatly reduce the possibility of errors by automating and digitalizing many of the activities that need to be executed in the highly fragmented clinical trial supply chain, as well as reduce the number of steps necessary to transfer the data from the original sources generating the data to the final scientists and analysts using the data.

Pharma IQ: Clinical trial data contains personal information that is not supposed to be disclosed to third parties or even to people preparing the reports required for submission to regulatory organizations. Is it easy to provide clinical trial data to the right people at the right time?

FA: A big challenge today is that many descriptions related to the benefits and adverse effects experienced by patients are long, handwritten texts from documents that not only require time to be inputted into digital systems but are also difficult to redact to mask patients’ personal information that must not be disclosed. For these reasons, often data that could be useful for the discovery of future drugs and important to design future clinical trials is removed from the final dataset because it is intermingled with personal information. Today, a “solution” is to simply have somebody manually redact the personal information after it has been inputted into the digital systems; but as you can imagine, this is not scalable and is very time consuming.

Advances in artificial intelligence (AI) and machine learning (ML) are promising for helping to redact data on the fly and to simplify at least part of the problem related to the preparation and modification of clinical trial data for sharing.

• Read part one of this article, where Fausto Artico describes the challenges of collecting clinical trial data.

Clinical Data Management Excellence featured experts from Janssen, Pentavere, GSK and Rgenix Inc discussing how to enhance clinical trial outcomes through efficient integration, standardization and compliance of data. You can watch the event on-demand by visiting Clinical Data Management Excellence.