"The future for Clinical Research is AI"! It's commonplace to hear this these times, but what exactly is it? We've all heard of the ways in which AI is used for basic research, in identifying chemical molecules, in identifying the patterns of disease in patients, as well as in Virtual Trials. In this article, I'll briefly discuss the many widely known and less-known uses that use AI and Automation for clinical trials procedure.
Machine Learning (ML) is a subset of AI that focuses on using algorithms for data that allow the system to "learn" and improve. ML lets users analyze large amounts of data, make intelligent inferences, and also predict the outcome. These insights can be utilized to automate various parts of the system. This can result in the speediest and most effective clinical trial system. Automation allows ML forecasts to be returned to the system, and for specific actions to be implemented which reduces the requirement for human intervention, as well as increases the speed and quality. Automation and ML are utilized at every step of the process.
Study Design
Machine Learning is a powerful tool for the design of protocols and translation. By utilizing existing protocol data and health library databases for specific areas of treatment the protocol for a brand new study could be created by the machine. The ML algorithms could be able of designing an optimal protocol based on the knowledge base. This leads to a reduction in design time and changes to protocols and disruptions to studies. Language translation can also be completed quickly and quickly and with a higher degree of accuracy than conventional methods because the ML model will have the domain-specific language knowledge base to study.
Study Setup
ML can be utilized to automate the designing and setting up of the form for a case report or study database. Utilizing a CRF library that is specific to particular therapies and study plans using the method, the ML model is able to be trained to create the best CRF and edit check. Automated output can be converted into the actual study design and validation, which allows database designers to alter the design whenever necessary. This method results in the most efficient design, which includes edit checks that would be overlooked when designed by a person. Automation allows for the ML-created study to be created and verified. The validation report offers the required information to designers who can apply the final elements before going live. ML can be utilized in order to streamline SDTM mapping or to create SDTM annotations for studies.
Trial Management
Automation that involves machine learning can be achieved in the area of trial management. The most obvious scenarios include site selection, enrollment of patients as well as Risk Based Monitoring (RBM), and chatbots.
Site Selection: The most optimal site selection can be achieved by using models that are based on machine learning. They can be trained to analyze parameters for sites like Enrolment Safety, Enrolment Compliance as well as Data Quality and predict which sites are suitable for a new study within the particular area of study. The order in which these parameters are evaluated is based on the kind of trial being conducted and the Sponsor/CRO. The algorithm can be developed based on data from previous studies and be able to determine the performance of a site for a new study.
Patient Enrollment: Predictive Analytics to aid in the enrollment of patients is a common usage case. It makes use of variables like the area of treatment study duration, study duration and disease burden (from Health Economics), study complexity and adverse events, as well as multi-centric, randomization, etc. The ML algorithm would examine the above variables and pick those with the greatest influence (relevant). The model that is developed could be utilized in future studies to determine the likelihood of a patient's enrollment. Although this is a common usage case, the vast number of external variables makes this prediction extremely difficult and has a low likelihood.
Risk-Based Monitoring: The concept of Risk-Based Monitoring, also known as RBM in clinical trials may be utilized at different phases of the trial in order to detect and reduce risks that could affect the clinical trial. One kind of RBM uses a few elements of selecting a site, such as enrollment, safety, compliance, and Data Quality as well as other variables such as therapeutic area and multi-centricity of trials. to determine the performance of a site during the course of a clinical trial. These predictors can be utilized to reduce the risk of a trial by identifying risks beforehand and attempting to reduce the risk, such as closing some sites and reopening others or focusing on those sites that are performing very well.
Chatbots: Chatbots make up some of the easiest and simplest to build examples of the potential in machine learning. Chatbots are available to cater to different kinds of users, such as website users, CRO personnel and patients, as well as patients. Users can communicate with them through voice and text. chatbots can recognize the natural language and context and can provide exact responses. This enhances the users' experience and lessens the workload on the support team.
Data Management
Data Management offers tremendous scope for AI automated automation. A few of them are as follows:
Smart queries: In smart querying, the machine-learning algorithm scans the entered trial data and then identifies possible queries that can be requested for different items in the field. This can be done by an amalgamation of previous studies and data from the therapeutic region. The algorithm identifies possible values of a data point in relation to a specific therapeutic area and then raises a query when it detects an unintentional deviation. The query is then reviewed by a data manager, and declared a valid query or rejected. This ML algorithm learns from the process and can improve its classification in the future forward.
Medical Coding: Terms used in medical coding made using WHODD and MedDRA dictionaries can be automated coded with regular programming, but only up to a specific percentage, over which a medical coder is required to look over the information in order to manually encode the remainder of the terms. Machine learning algorithms can learn from code libraries that cover a variety of therapeutic fields. They then match the transcript of the study to the appropriate dictionary word for the specific area. Machine Learning can accomplish this by coding with a high degree of precision.
Query Management: Many thousands of queries are asked for in every clinical trial. A large portion of the time is dedicated to these questions. Many of these questions are not needed and are raised due to misconfiguration of edit checks within the EDC system. These queries can be identified with machine learning and then managed in bulk, or the appropriate edit checks can be set up in mid-study to fix the issue moving forward. Machine learning makes use of clustering to detect groups of queries that are grouped, and issues discovered. The clusters could also be addressed in massive.
Smart SDV: Companies spend a significant amount of time and money on monitoring trials. The CRAs are required to travel to the site to oversee the study, as well as perform Source Data Verification (SDV). Machine learning could greatly cut down the manual work involved. Site staff can create pictures of the document sources and upload them to the server. Machine learning algorithms are able to extract content from the images, and transmit these to EDC. The EDC will then match this data with the input data and labels them as authentic source data if there's the same. If not, it will raise an inquiry that needs to be verified manually.
Data Analysis
Machine Learning can provide many insights into the clinical data collected throughout as well as after the test. Prediction, classification, and clustering are just a few methods that can be utilized in data analysis to draw important insights into large data sets. Patient behavior, adverse events, etc. can be predicted by machine learning.
Regulatory Submission
Clinical trials that are regulated involve a significant amount of documentation. They can be formatted and automated with machine learning.
CSR Automation: in CSR Automation CSR Automation, CSR Automation, the Clinical Study Report can be produced automatically through machine learning after taking a look at the Study Protocol and the Study Analysis Report (SAR). By following ICH GCP templates, most of the CSR could be produced. Natural Language Processing (NLP) algorithms can be utilized to alter the language used in CSR and also utilized to create narratives. The narratives can then be evaluated by the medical writer, and altered to create the ultimate CSR. This can be done in just a few days. This procedure reduces the submission process for regulatory purposes dramatically and increases the quality of submissions.
About Author
Name: Manuj vangipurapu
Gmail: manuj@quadone.com
Designation: Founder & CEO
Company: Clinion
Company Url: https://www.clinion.com/
Linkedin profile: https://www.linkedin.com/in/manujv/