The potential of AI, which unites innovative initiatives with a pragmatic perspective, has attracted the attention of a growing number of industry players. It is important to recognize that the implementation of these technologies, whose development used to be limited to the academic world, has posed many challenges. These include obtaining high-quality data to feed these systems, validating and regulating inherently changing predictive models, and navigating through ethical dilemmas that highlight the importance of interpretability and detection of potential biases in these systems. Nevertheless, the positive impact that AI can have on clinical research justifies and encourages efforts to overcome these obstacles in a rigorous and responsible manner.

Numerous organizations within the Clinical Research industry are investing in applied AI, aware of its potential to improve both the efficiency and accuracy of Clinical Research, critical elements in today’s era of personalized medicine.

There are many examples of the application of AI in clinical research, which are undoubtedly revolutionizing the sector, improving the effectiveness and efficiency of clinical trials:

  1. Improving patient recruitment processes in clinical trials is an area where AI is already proving to be a particularly useful tool. Historical data from previous clinical trials make it possible to train models capable of predicting which healthcare sites are the most promising when it comes to initiating a specific clinical trial. Thus, there are systems that make use of AI models to make the best match between sponsor and hospital, reducing the time required to successfully execute a given protocol by selecting the optimal sites for each clinical trial. These systems are typically capable of extracting the inclusion and exclusion criteria from the research protocols, sometimes using natural language processing techniques, and once this information has been extracted, identifying each hospital, and thanks to access to the information in the clinical records, the expected recruitment rate of patients who will meet the criteria indicated in the protocol.
  1. In the collection and analysis of health data, AI opens up a new scenario that represents a turning point compared to the use of traditional methods and can contribute enormously to automating certain processes that traditionally consumed an enormous amount of time and could also introduce human error. For example, it is possible to use AI models trained to identify relevant information from electronic medical records, whether this information is recorded in structured or unstructured form, in the form of clinical notes expressed in natural language. These models not only speed up the collection of this information but, when properly applied, they are capable of increasing the accuracy of the data compared to methodologies based solely on human review. The key here is the ability of AI models to streamline and automate the data collection process in clinical studies, reduce the mental overload of human reviewers who must perform these tasks manually, minimize errors, and speed up the data collection process. Using machine learning algorithms, AI can analyze large volumes of data, identify candidates, and more accurately characterize patients. This capability provides invaluable support for classical inferential statistical analysis aimed at understanding and confirming cause-effect relationships. In addition to enhancing and facilitating traditional statistical data analysis, which will undoubtedly remain a cornerstone of clinical research, AI opens up new opportunities in the generation of predictive models, the purpose of which is complementary to inferential statistical methods and should not be confused with them. The development of these predictive models, capable of supporting clinical decisions, must be carried out while ensuring the protection of privacy and confidentiality of patient data. Moreover, it should always be designed to support decision-making by health professionals and not as a substitute for them.
  1. In the field of pre-clinical research and so-called “Drug Discovery“, AI-based applications have started to emerge that are able to integrate natural language analysis of scientific literature, together with clinical and experimental data. This large-scale integration makes it easier to find patterns that help to find new indications for a given drug, more efficient drug combinations, better characterize mechanisms of action, and suggest new lines of research. One of the main challenges in this area lies in the integration of often heterogeneous data sources through the use of controlled biomedical vocabularies. The effective combination of Large Language Models and Knowledge Graphs is one of the most developed areas in recent months in this application area.
  1. The application of machine learning and Artificial Intelligence techniques has a certain tradition in the specific field of pharmacoepidemiology and more specifically in Pharmacovigilance, one of the pioneering areas in the use of Real World Data extracted from secondary sources such as electronic medical records. AI is an invaluable aid in the detection and assessment of potential adverse events, making this process more efficient and accurate. As a complement to classical statistical methods of proven efficacy in the detection of medicinal product safety signals, it is nowadays possible to use different AI models capable not only of identifying adverse events mentioned in natural language in electronic medical records and not captured by classical reporting methods, but also of finding patterns present in a structured way in medical record databases of population coverage by means of data mining techniques. In addition to registers in medical records, and given that a large proportion of adverse events can occur in over-the-counter medicinal products, some of these systems use alternative sources of data such as posts on social media or internet forums by patients themselves. The ultimate aim of these systems is to identify the correlation between exposure to certain medicinal products and an increased risk of adverse events, which helps to proactively improve the safety of medicinal products during the post-marketing phase.

An aspect of particular importance in all these initiatives is data governance, a concept that refers to the policies, practices, and processes in place to manage and control an organization’s data. Data governance focuses on ensuring the quality, integrity, privacy, and security of data, as well as establishing responsibilities and roles for its management. Data governance allows data traceability to ensure that it meets the so-called FAIR criteria: Findable, Accessible, Interoperable, and Reusable.

In short, the possibilities offered by Artificial Intelligence applied to Clinical Research are numerous and its potential for optimizing time and resources, together with its extraordinary capacity for analysis, should allow us to improve the quality of life of patients. However, these systems must be designed and deployed in a responsible, rigorous, and ethical manner, which makes it necessary to have multidisciplinary teams that include healthcare professionals and experts in Clinical Research.

By Julio Bonis

Artificial Intelligence & Big Data Manager at Sermes CRO

Subscribe to Directory
Write an Article

Recent News

Exposure to Heat and Cold During Pregnan...

The research team observed changes in head circumf...

Using mobile RNAs to improve Nitrogen a...

AtCDF3 gene induced greater production of sugars a...

El diagnóstico genético neonatal mejor...

Un estudio con datos de los últimos 35 años, ind...

Highlight

Eosinófilos. ¿Qué significa tener val...

by Labo'Life

​En nuestro post hablamos sobre este interesante tipo de célula del...

Seven key genetic profiles associated wi...

by Fundació Clínic per a la Recerca Biomèdica

A study published in Nature Communications has identified different ri...

Photos Stream