From The Editor | January 29, 2021

AI And Machine Learning Prepare Pharma For The Data Onslaught

Ed Miseta

By Ed Miseta, Chief Editor, Clinical Leader

Badhri Srinivasan, Patrick Nadolny, and David Horn Solomon

When I reached out to several clinical executives in my network for their thoughts on trends that will impact clinical trials in the coming year, the two mentioned most frequently were artificial intelligence (AI) and decentralized trials. The COVID pandemic of 2020 has most sponsor companies scrambling for new technologies that will increase the efficiency of trials and make the experience easier on patients.

“When I look at the industry and what will make a difference, the first thing that comes to mind is how we can perform trials in a virtual manner,” says Badhri Srinivasan, head of global development operations at Novartis. “The industry needs to figure out how to perform trials in a patient’s home rather than at an investigator’s site. Since March 2020, many companies have been doing that. But a related trend is the increase in clinical data and the data collection approach companies are using. With the advent of decentralized trials, we now have data streaming in from many new sources. Companies now require the capability to mediate between those data sources and the ultimate destination of the data. That will be a challenge for most of them.”

Badhri Srinivasan
Srinivasan notes this is an area in which even regulators are struggling. As a result, many questions remain. How do we deal with wearable devices? What data and systems are considered to be 21 CFR part 11 compliant? Is the Apple watch a GxP system? What devices are clinical grade? What blood collection instruments are approved for use in a patient’s home? These are fundamental questions that must be answered. The industry will have to work closely with regulators to get these questions answered.


A few years ago, collecting data from devices and sensors was something most people were just talking about. Today those devices are being used in greater numbers. The streaming of real-time data also enables sponsor companies to perform real-time monitoring and real-time signal detection on those data. This is where AI and machine learning enter the picture. Srinivasan believes one of the most significant impacts of those trends will be a reduction in cycle times.

David Horn Solomon
“Cycle times need to be shorter,” states Srinivasan. “The industry now expects that. There should not be a gap of months or years between the time data are collected and when they are interpreted and feedback is provided. That is no longer acceptable. AI and machine learning will step in to make sense of the data, interpret them, and pull wisdom from them even as the data are being collected.

Moving to this new model of conducting trials will certainly require an investment in new technologies and processes. Srinivasan believes this will pose a burden for companies. In many sponsor companies, processes are still tailored toward manual, in-person work. This old way of collecting data has not undergone many revisions.

“For some companies, there will be a massive effort to understand the new sources of data and how they will fit into the existing ecosystem,” notes Srinivasan. “But I think the mindset is far more important than the technology. Technology, processes, and digital devices are not barriers. Companies have proven they can work, and any challenges that arise are surmountable. What worries me is that many companies are not instilling the proper mindset in their workforce. Going forward, sponsor companies will need an empowered workforce with a mindset that things are going to be done differently. Employees cannot remain ingrained and embedded in the old ways of working. My real concern is how quickly we can change that way of thinking.”


Patrick Nadolny, former AVP, global head, clinical data management and programming for Allergan, notes the main purpose of machine learning is pattern detection. He believes the impact AI and machine learning can have on clinical trials is almost limitless. However, he cautions you will not benefit from the technology if you do not have a clear plan for employing it.

Patrick Nadolny
“Many organizations will waste their energy by going in many different directions without a clear vision,” he cautions. “Many do not fully understand the technology or how to leverage it. Unfortunately, for those companies, the end result will be proof of concepts that fail to show value. Companies wanting to implement the technology must first understand its potential and then articulate a clear implementation strategy.”

Virtual trials and patient sensors are creating a deluge of data for sponsors. According to Nadolny, there are not enough data scientists in the world to generate real-time actionable insights or knowledge from that data. The Big Data problem requires a reliable and large-scale solution to mine the data. He believes that solution is machine learning. The big question is whether pharma is ready to implement the technology.


The first benefit of AI will be process automation, which will optimize existing processes. Nadolny believes large organizations with many manual and labor-intensive tasks will realize a fast ROI with minimum investment. Those firms will also realize increased quality via the reduction of manual handoffs. However, Nadolny notes that a truly intelligent and transformative solution will require a great deal of historical data, which many small companies may not have. This again makes the technology a better fit for large companies. Although the needed data can be gathered, it will not be an easy process.

“The real limiting factor on implementing AI and machine learning will not be the technology itself,” he says, “it will be human understanding. People need to understand the technology and the data that will be required. They also will need to know what questions to ask, what data will need to be collected, and how those data should be stored. It’s important to understand these requirements before embarking on any installation.”

Many companies will need to begin the process of aggregating the data prior to implementing an AI solution. Data and business process expertise also will have to be acquired, which may force some sponsors to change their talent acquisition and retention strategies.


For companies thinking about implementing AI, Nadolny recommends putting clear roadmaps in place. For example, the roadmap for enabling the next generation of data reviews powered by AI may consist of four stages. Stage 1 is automated reviews, or the automation of labor-intensive activities. This will reduce operating costs and free up process experts. Stage 2 is actionable reviews, or the use of machine learning to detect complex data trends and anomalies. Stage 3 is guided reviews, which are recommended corrective actions based on the lessons from Stage 2. Finally, Stage 4 is supervised reviews, which will detect complex data trends and anomalies along with corrective actions based on lessons from Stage 2 and Stage 3.

“Moving forward, scaling up the number of clinical trials will not require a proportionate increase in resources,” says Nadolny. “All that will be required is an incremental investment in computing power. Several leading pharma companies have already entered Stage 2 and are discovering the viability of AI-based solutions. Intelligent applications are going to change the way we do clinical research. The prospects are vast and considerable, but the strategy to get there must include people, processes, partnerships, and technology.”


David Horn Solomon, CEO of Pharnext and chairman of the board of Advicenne Pharma and Rexgenero, likes the technological advancements he is seeing in clinical trials, especially the progress being made in virtual trials, AI, and digital devices. But despite the advances being made by technology providers, Solomon notes companies often must create their own technologies. This is done by developing new tools or adapting apps to meet sophisticated endpoints that have not been used in the past.

“We are pioneers of something new,” he says. “Today we may be sending a nurse practitioner to someone’s house for a blood draw. The future will be about how we measure and input data and how patients will control a lot of that input. This is exciting, and we should look at it as a challenge. Right now, we are determining how we will reinvent clinical trials for the future.”

Vendors might be designing new technologies, and regulators may be providing the guidance, but Solomon states that sponsor companies must understand their patients and the goals they need to achieve. Those companies are the ones that need to determine how we get there from here. They need to determine what type of measurement is needed and what type of device is required. Sponsors then need to share that knowledge with regulators and technology companies to develop the tools needed.

“This is where patient-centricity really comes into play,” says Solomon. “We can’t look at available devices and determine how to force them into our trials. We must be the developers of the new tools. We need to determine the needs of patients, work with technology firms to develop the tools we need, and pressure test them with regulators. We cannot look at the trial process as simply extracting data from patients — we have to understand their condition, the purposes of the study, and how technology will help us get to those endpoints.”


According to Solomon, now is the time for executives to be leaders. In the past, the industry could sit back and wait for a consulting firm or an app company to come up with a new tool. But, with the changes taking place with technology solutions, sponsor companies need to be the leaders. Innovation creates value, and the companies conducting trials need to be the innovators.

“The clinical trial system is changing,” he states. “We have to adapt to that, and the best way to do so is to take the lead. Do not sit back and wait for the next new technology to be developed. That could take a year, force our trial to be put on hold, and disappoint our patients. We need to determine what our trials and our patients need. In 1901, if you had asked people what they wanted, they would have said a faster buggy. Henry Ford recognized that what they really needed were cars that were affordable. We must do the same thing. Sponsor companies will lead by talking to patients, making assumptions about what is going to work in this new environment, and becoming the innovators of the technologies.”

Even when new technologies are available, there is no guarantee they will be adopted by sponsor companies. One positive impact of the COVID pandemic is that it has hastened technology adoption in clinical trials. The big question is whether these changes will be permanent. When the pandemic has passed, will patients return to in-person visits to a clinic? Srinivasan doesn’t think so.

“Expectations of how a trial will be conducted have gone up,” he says. “Regulators got involved this year and released new guidance that promotes virtualization, remote visits, and remote monitoring. In the future, I believe regulators will point to that guidance and say, ‘This is how trials should be performed.’ The industry has shown this model can work during a pandemic. If that’s the case, then why can’t it work when there is not a pandemic? The option of going back to in-person visits for assessments that can be done at home is no longer on the table. The question now is how quickly can we scale and move forward.”