The life science industry is no stranger to the benefits of artificial intelligence (AI). Clinical research organizations (CROs) have been developing and deploying AI software to clean and analyze clinical data for years but now AI is being leveraged in new ways to impact additional aspects of clinical development and operations.

What’s all the buzz about AI?

AI exploded into the industry when we came to a place of cloud computing and distributed computing. As software experts adopted and experienced the benefits of smarter data processing and storage, they began to turn their attention to tools capable of analyzing huge volumes of data. 
 
A subtype of AI that has received a lot of attention lately is machine learning (ML), which focuses on creating algorithms that can problem solve much like a human would. Unlike standard rules-based programming logic that has a set path and functionality, ML has allowed data scientists to build algorithms that can extract patterns from an amount of data that would be nearly impossible for humans to interrogate. Much like how people gather information and make their best-educated decisions, ML attempts to do the same while often outperforming human cognition.

Non-traditional applications of AI emerging

ML’s ability to effectively process a wide variety of clinical data makes it a cross-functional solution. However, differences across data sets require developers to create custom programs to categorize information in their given context and parameters. Data forms vary and can be quantitative or qualitative. Adjusting for the range of data is where natural language processing (NLP) can help.
 
NLP focuses on extracting insights from unstructured data – specifically text data in our industry – and categorizing it faster than humans can do manually. Much like ML is designed to problem solve like a human, NLP is designed to understand language and be able to apply the same acute understanding that a human would. NLP can be utilized in multiple ways in a clinical trial to save both time and costs, and the PPD clinical research business of Thermo Fisher Scientific is using it to tackle one of the most labor- and time-intensive tasks – reviewing contracts between sites and sponsors.

A common pain point for sponsors

A site’s ability to recruit patients is the key to completing a clinical trial. A single trial can have dozens to hundreds of sites that all require onboarding and contracting with the sponsor or CRO before they can bring the first patient through the door. The amount of time spent negotiating contracts and budgets to progress the site through startup can add months to the overall timeline. 
 
These negotiations are repeated with each site selected for a trial. Without fail, there are always adjustments and edits made by the parties to reach a satisfactory agreement, resulting in an abundance of work hours invested to review and carry out those negotiations. With hundreds of sites running dozens of clinical trials at once, it’s easy to see how the workload can stack up for a single trial.

Expediting the site activation process

Historically, there have been few tools to accelerate the contract review process, but recently, with the help of ML and NLP, there is the potential to reduce working hours and enable sites to activate faster. This means saving sponsors time and costs on their trials and bringing new drugs to market more quickly for patients.
 
NLP-based tools help legal teams break down document information, collate it in a standardized fashion and compare the content of two contracts side by side even if the documents are developed from different templates. Consider the most common legal documents that people interact with – application user agreements, purchase agreements, insurance paperwork, etc. All these documents tend to follow a standard template and organization, making information easier to find, digest and compare. However, this often is not the case with clinical trial research agreements. Sponsors and sites tend to use their own templates, making it difficult to quickly locate and compare specific information in each document. NLP may solve this problem by classifying legal content, even between documents of variable structure.
 
Other than saving time from contract reviews, bespoke technology that combines ML and NLP provides valuable insight into identifying key areas of sponsor contracts that tend to generate the most feedback and can analyze them for meaningful improvements. The insights provided by innovative AI solutions allow sponsors to develop contracts that predict the ever-changing needs of sites and ultimately reduce negotiation timelines.
 
The site contracting world and clinical settings are constantly evolving. ML and NLP provide us with foresight into new trends in negotiations. One example is the increasing number of stricter site access agreements since the start of the COVID-19 pandemic. These types of insights have enabled teams to be better prepared, greatly reducing the impact on research timelines. 

A bright future for AI and clinical research

AI functions have come a long way since their ascent into clinical data processing. ML and NLP will continue to make huge impacts across nonclinical functions. This developing technology creates the ability to gather similar insights into more complicated documents like informed consent forms or even generate entire contract templates – eliminating the need for a sponsor or site to formulate documents themselves. 

The potential time and cost savings are tremendous. We are optimistic that ML-based tools will become a significant component in future clinical research as we continue to seek ways to organize non-traditional data.

---