How AI is making clinical trials more accurate and affordable

pharmafile | February 5, 2020 | Feature | Business Services, Manufacturing and Production, Medical Communications, Research and Development, Sales and Marketing |  AI, AI Clinical Trials, Apple, Apple Watch., apple watch, clinical trials 

Clinical trials are one of the longest and most expensive steps of bringing a drug to market. Conor Kavanagh investigates how developments in AI are looking to offset this by improving patient recruitment for trials and their adherence during testing.

The pharmaceutical industry is one of the most profitable in the global economy. The research, manufacturing and marketing of drugs is expensive, but one of the biggest drains on finances when creating a new pharmaceutical product are the clinical trials. It takes on average 10-15 years to bring a new drug to market and half of this time and investment is consumed during the clinical trial phases of the drug development cycle.  The two main factors that cause these trials to be unsuccessful and costly are recruiting mechanisms and patient adherence.

While there are still numerous challenges, the largest pharmaceutical companies are looking towards development in AI to help stem the loss of funds.

Potential problems

AI possesses the potential to solve many problems in the clinical trial process. Two of the most notable are recruiting the right type of candidate, who will be the best fit for your trial, and making sure this patient is adherent throughout with the rules and regulations of the process.

Recruitment is key to every trial and even the best designed can be ruined by ineffective patients. Roughly 80% of clinical trials fail to meet enrolment times and approximately one-third of Phase 3 clinical study terminations are due to enrolment difficulties.  Phase 3 trials also carry 60% of the overall trial costs and the resulting loss per failed clinical trial lies between $800 million-$1.4billion.

Just recruiting the right candidate requires jumping over several hurdles. Pharmafocus spoke to Dr Stefan Harrer on the many issues surrounding clinical trials and how AI can help solve them. On patient recruitment Dr Harrer, a member of the IBM Academy of Technology, as well as an expert on AI and clinical trials who works at IBM Research Australia, says that: “Patients becoming aware of a trial they are eligible for is often an almost insurmountable hurdle. Information resources are scattered and even if they come across a trial announcement, understanding eligibility criteria, getting access to their own Electronic Health Records (EHR) and then contextualising both data sources against trial requirements is extremely difficult. Then once eligibility and suitability are established, empowering patients to actually participate in the trial can be time-consuming as well: is the trial site far from their home? How often do they need to physically show up at the trial site?”

When the right patient is accepted onto the trial process, getting patients to stay on the trial can often be an overwhelming and cumbersome task. Companies see, on average, 40% of patients becoming non-adherent after 150 days of the trial process.

Patient adherence does not just mean getting patients to take their medicine at the right time. They may also have to attend a clinic or hospital to have tests carried out, complete forms on how they feel between appointments and make sure to record any possible side effects.

Patients not adhering to the clinical trial’s rules and regulations can contribute to a huge financial cost to the sponsor funding the research. Missing data from the patients also means the results will be poor and not help with the developments in the medical field.

Dr Harrer believes that several factors combine to cause significant problems in adherence: “Age and cognitive capabilities of a patient can pose adherence challenges as well and self-administering medication following complex protocols can be confusing. And then there might be trial-immanent reasons such that the medication does not show any effect over a certain period of time, leading to a ‘it does not matter whether I follow protocol or not’ attitude in the patient, or patients do follow protocol but fail to document their adherence properly. Often a combination of the above factors leads to non-adherence.”

AI and Electronic Health Records

The efficiency of clinical trials could be much improved if AI was put to use in analysing EHR – it’s one of the most sought-after applications of AI pharmaceuticals. In particular, it would be extremely helpful in the patient recruitment phase. Instead of doctors assessing and choosing the right candidates, AI has the potential to tap into patients’ electronic medical records to cross-reference hundreds of available trials based on specific criteria. It can also use existing patient information to assess the probability of a good fit for a given trial and ultimate responsiveness and success rates.

Dr Harrer believes that: “The digitisation of health records is absolutely crucial for making clinical trials more efficient. Only digitised data allows us to correlate data from different sources, to analyse data using advanced analytical methods such as AI and for decision makers such as doctors, patients and trial organisers to react to insights from data analysis fast and efficiently.”

The US Federal Government have spent more than $28 billion to digitise EHR over the last decade. However, it is still hard for patients to access their own records from all the health institutions they have visited, and there is no centralised repository or standard format for patient medical data.

There is also the issue of sharing heath information easily across institutions and software systems. Different hospitals and providers may not use the same EHR software, and it may be hard to send the data over in the proper format, and this can pose a problem for developing AI technology.

Dr Harrer also recognises the difficulties in EHR due to a lack of standardisation: “EHR data often sits in compartmentalised data buckets with no compatible interfaces for moving data back and forth and connecting different data sources. Even if compatible EHR data exists in different sources, often data owners have no knowledge of such compatible data outside their ecosystem. And then of course on top of digitisation and standardisation, EHR data privacy, security and integrity have to be ensured at all times. Getting all these processes going hand in hand is no easy task.”

Dr Harrer feels more must be done to make uniform EHR a priority, and that “regulatory, technological, operative and monetary aspects all play equally important roles in creating an EHR data ecosystem that allows leverage in EHR data in a secure and efficient way. But creating such an EHR data ecosystem is absolutely essential for making clinical trials more efficient in the future.”

Some private companies have attempted to offer alternatives to the current system. Mendel.ai tried to solve the challenge of piecing together a patient’s medical history by allowing cancer patients to submit their medical records to its platform. Alternatively, patients could give Mendel.ai permission to collect all medical records from doctors on their behalf.

Mendel.ai was developing machine learning algorithms to extract information from digital records and to match patients with ongoing trials that best suited their needs. The start-up charged patients on a subscription basis, starting at $99 for the first 3 months to process all data from medical records. But Mendel.ai has since gone quiet. The company has not announced any further expansion plans, nor raised additional funding rounds.

Apple, EHR and making patient data more accessible

Apple is the second most profitable company in the world and developments in its iPhone and, in particular, Apple Watch technology has changed the way many people track their health. It is using its health collection technology and allowing third parties to use their AI applications in clinical trials. It is also making sure that patient data during a trial can be monitored effectively, mitigating human error in the reporting of such data.

In January 2018, Apple announced that iPhone users will now have access to all their EHR from participating institutions on their iPhone’s Health app. This came after Apple bought AI start-up Gliimpse, in 2016, with Apple’s Health Records API being based on their work. Their technology helped turn medical documents into data, easily allowing searches, filtering graphs and dashboards that show the data that matters most. Apple’s interface is also easy to use, and can record and display information on allergies, lab results, medications, procedures and vitals.

Apple is working with popular EHR vendors like Cerner and Epic to solve problems of interoperability. This partnership could see a two-way data flow where vendors are incentivised to integrate patient recorded data into Apple. It will also help doctors and researchers using Apple’s ResearchKit and CareKit software tool for clinical studies.

There are already numerous other examples of organisations using the technology of the iPhone. Researchers at Duke University developed an Autism & Beyond app that uses the iPhone’s front camera and facial recognition algorithms to screen children for autism.

mPower is another app that is used by nearly 10,000 people. It uses exercises like finger tapping and gait analysis to study patients with Parkinson’s disease who have consented to share their data with the research community. The Android version of the app uses machine learning to process this data collected from smartphones to create a severity score for patients. 

More accurate patient adherence

The push from companies like Apple to make their smart products cater more to the health-conscious has already shown great potential for the future of clinical trials. It is clear more companies in the health industry will use the platform that has been designed for them on phones and smart watches.

Dr Harrer believes that wearables running AI models have the potential to boost adherence: “Wearables have the potential to play a key role for improving adherence, probably not so much for improving recruitment. Wearables are empowered to run AI models that have the potential to automatically detect and log patient activities for digital disease diaries. Such intelligent patient monitoring systems – we call them Cognitive Sensors – may help improve adherence, retention and end-point detection during the active phase of clinical trials.”

In terms of AI more generally, Dr Harrer believes the increasing use of it in adherence will be invaluable: “AI has been shown in studies to potentially help impact all of the above described issues: the creation of intelligent patient monitoring systems, so called digital disease diaries, has the potential to automatically detect and log disease episodes. Such monitoring systems can also be used to monitor patient activities.

“Deep Learning plays a particularly important role for digital disease diaries as it is capable of automatically analysing large noisy data in real time and detecting episodes of relevance in such data automatically. In another application of AI, human-machine interfaces have the potential to be used to pro-actively coach a patient through complicated parts of the trial, for example guiding them through medication administration processes or alerting them automatically of a pending action that needs to be taken to adhere to trial protocol.

“Such systems could also be used to monitor patient behaviour over time and assess drop-out and non-adherence risk, allowing intervention before a patient drops out of a trial.”

The Big Pharma companies have already started using AI tech to help with patient adherence. Pfizer and Novartis already text their patients’ phones to remind them to take their medication but are also investing in “ingestible trackers” to track patients’ drug intake.

In 2017, Merck Ventures gave funding to Medisafe, who were developing wireless pill bottles. Named iCap, the Bluetooth-equipped cap fits most standard pill bottles dispensed with prescriptions in the US. It has a companion app, iSort, which is a Bluetooth-enabled weekly pill organiser for those taking multiple medications.

The two together allow users to have their medication intake automatically logged as the iConnect devices transmit signals when they are opened directly into the Medisafe software. This information is then stored in the cloud and available to the user via the Medisafe mobile app.

Applications and equipment like this can help mitigate human error and encourage taking your pills at the right time and recording it, as well as streamlining complicated medicinal regimes.

Smaller AI startups are also using their technology to improve AI adherence. AiCure is company that has designed a program that uses facial recognition algorithms to track adherence. This occurs by the user recording a video of them taking the pill, and the program confirms the right person took it. This eventually raised $27 million in funding.

Another area that AI startups are looking as is artificial assistants similar to Siri, Cortana or Alexa. Catalia Health is developing a healthcare companion and coach. They hope to help create behavioural changes in patients by setting reminders, asking specific questions and tailoring conversations for each individual patient.  They aim to understand why patients miss dosages.

Can AI solve all the problems in clinical trials?

Developments in AI are set to drastically trim the financial cost of, and improve the medical data from, clinical trials from problems surrounding recruitment and adherence. But Dr Harrer sees AI as a tool to assist researchers and doctors, rather than replace them.

“AI has the potential to play a role in all steps of patient recruiting: first suitability and eligibility assessment, then empowerment and motivation of patients to participate in a trial. However, AI will never fully solve or take over any of these steps: patients and doctors will always be the decision makers, but AI may help them be more informed and make decisions faster.”

The digitisation of health records is one of the key areas that need to be improved for AI to flourish, and governments and private companies must collaborate to standardise them. Innovative wearables like the Apple Watch are making patient data easier to collect while creating a database for many third parties to give patients their data.

In terms of adherence, new AI projects look to create more feasible methods to improve patient adherence across the board. Items like Medisafe’s iCap and iSort look set to make taking and registering a patient’s intake of medicine more accurate.

While human error can never totally be mitigated, developments in AI will help cut the amount of financial and data costs of clinical trials.