Simon Tilley, global lead for healthcare and life sciences at SAS, breaks down the benefits and drawbacks of decentralised clinical trials (DCT). He also explains how AI, machine learning and advanced analytics can refine all stages from participant recruitment to compiling results to ensure the best outcomes are achieved.
Key insights:
- Decentralised trial designs can provide a high testing and monitoring frequency of its participants, as well as more inclusive and representative of the people who will use the drug.
- DCTs is a double edge sword as it can also deliver a large volume and variety of data which can become overwhelming for manual data entry.
- AI and machine learning can overcome limitations of DCTs by processing large amounts of data quickly, and predicting participant drop out probability.
Remote data capture, through using wearable technology for instance, opens the door for large-scale studies to be undertaken without the need for participants to always be based in a testing facility.
The NHS rolled out a nationwide service during the pandemic, where pulse oximeters were supplied to patients at home who had been diagnosed with Covid, and were most at risk of becoming seriously unwell. Those on the service were regularly alerted to record their blood oxygen levels and report back, all via text, allowing them to be monitored regularly at a time when capacity in hospitals was at breaking point.
While not strictly a DCT, this service provided a high frequency of testing and monitoring for its participants. It did so while only taking up a fraction of NHS resources compared to the traditional inpatient equivalent and was rolled out over a diverse geographical area – which makes DCTs so effective.
The benefits of decentralisation
Decentralised trial designs can be much more inclusive, and representative of the people who will ultimately use the drug.
The time and financial requirement needed for patient management services is greatly reduced, due to remote monitoring and telehealth calls. Even in cases where home visits are required to distribute or train on equipment use, the number of participants is not limited by physical facility constraints.
Many of the tools needed for reporting by participants in DCTs are becoming increasingly available. Using suitably certified smart devices, participants can report image data, oxygen levels, and submit diary entries of their thoughts and condition.
With more people from different backgrounds able to take part, researchers are able to build bigger and more accurate datasets. The possibility of round the clock reporting from wearable technology means results are more reflective of the real world. For example, time of day is especially important within trials where the daily internal changes of the body can have a subtle impact on efficacy.
Overcoming current limitations
The increased breadth and depth of data that is generated by DCTs is a double-edged sword, offering incredible insight but as a trade-off are hard to analyse and extract valuable information from.
Clinical data is often scattered across a number of different proprietary systems that are independent and incompatible. With DCT datasets comprising even more types of data this challenge is exacerbated.
Getting helpful insights from a combination of image, text and numerical data manually takes a lot of time, and risks error or oversight. The volume and variety of data available in modern DCTs quickly becomes overwhelming for manual data entry.
Remote patient management has a number of benefits in terms of scale, but it also creates issues for retention. Trials can be an unnerving time for those involved. A lack of human interaction and reassurance can lead to uncertainty or dissatisfaction in the trial, which are leading factors for study drop-outs. Any delay in getting a drug to market has repercussions for the patients who could benefit from it, and is commercially damaging for pharma companies.
AI for data management
AI and machine learning can be applied to counter the current limitations of DCTs and also prevent issues currently faced across all trial types.
AI has the capability to process far more data than humans ever could in quick time, and can be trained to assess and sort the data based on any combination of different variables. On top of this, information fed into the data set is continuously managed throughout the trial, as opposed to being assessed manually in batches.
Similarly, AI and machine learning can overcome the hurdles created by having multiple file types. Automatic extraction of important information can be sent to a single, usable source. Strictly-governed AI models can make sense of vast amounts of information quickly, guided by humans at every step to ensure outcomes are fair and equitable.
Analytics’ role in participant management
Cloud analytics and AI can overcome the issues faced in managing large numbers of participants too. It can predict the participants who are likely to drop out or become unable to complete the study, right from the patient selection and early trial stages. By assessing this risk and acting accordingly, drop- outs are minimised, increasing trial efficiency and minimising costs.
AI models can also assess the risk to participants at all stages of the trial, including applicants. This enables researchers to consider far more variables simultaneously, meaning that a more accurate assessment can be made. It helps to reduce risk in clinical trials by identifying typical indications of negative effects and simultaneously comparing results across a whole cohort to check for outliers.
Signs of potential side effects or abnormal results can be instantly logged and reported to medical professionals to take action; much faster than manual check-ups could allow. If there has been a negative impact the algorithm can act preventatively to safeguard other participants by analysing retrospectively for any early warning signs.
We are only beginning to see the full potential of what DCTs have to offer. As the application of technologies is developed with increasing success there will be significant improvements in the outcomes and learning from these trials. Analytics will be leveraged to build on participant safety, whilst machine learning and AI will help in tackling datasets from which it would otherwise have been impossible to draw meaningful findings in a reasonable time. Managed in the right way, these tools will be a driving force for progress in pharmaceuticals and medicine… seeing clearer, seeing further, thinking smarter and acting faster.