AI can be used for target identification, in-silico drug design, validation and optimization of drug candidate, and repurposing of existing drugs. NLP-based literature search and ML algorithms can help explore favorable pharmacology profiles and predict physiochemical and pharmacological properties of drugs. This will assist in identifying novel biomarkers, and understanding therapeutic efficacy on multiple disease pathways by screening thousands of potential compound libraries. Predictive analytics and ML can be used to identify new indications for known drugs and match existing drugs with rare diseases.
AI helps researchers discover insights on cell interactions from large datasets through automatic selection, manipulation and analysis of cells for target validation, toxicology and phenotypic screening to evaluate drug safety and effectiveness. This can help optimize time, cost and uncertainty in planning experiments.
NLP and ML can be used on Real World Data to enable patient-trial matching by identifying predictive biomarkers to select patients with high probability of treatment outcomes and prioritize prognostic biomarkers to identify disease drivers. Also, these techniques help in measuring drug responses, and designing protocols.
Predictive analytics and ML help screen sites and investigators, predict site performance, and enable centralized risk-based trial monitoring. It can predict and tailor recruitment strategy before launch of clinical trials and assist in dropout risks forecasting and intervention. AI and ML can help process and classify Trial Master File documents providing algorithm-based ability to normalize and validate data.
Chat bots can help sites with initial logical questionnaire-based assessment of patients for Patient Recruitment / Pre-Screening. Site selection can be expedited with chat bots to gather site information like staff qualifications, past performance etc. Chat bots can also collect participants’ data remotely without visiting site directly on eCRFs. Patient diaries and surveys can be replaced with chat bots for electronic patient-reported outcome (ePRO) and electronic clinical outcome assessment (eCOA). Also, chat bots can be explored for eConsent by providing patients clear and easy-to-understand clinical trial information to make informed decisions.
Thus, building knowledge-based AI ecosystem to accelerate operational data management and clinical development can enable site-less/virtual trials.
NLP and ML can help analyze medical records to identify and select patients for clinical trials to accelerate 'right' patient recruitment to complete clinical trials faster. Predictive analytics can be used to monitor patient’s health and analyze patient’s behavior and to educate them on use of treatment, thereby improving medication adherence during trials. AI can be used for clinical endpoint detection by assessing potential impact of treatment on the disease by specific symptom/s by analyzing various forms of databases including speech samples, image data etc. Predictive analytics can be used to determine patient dropout risks, potential non-compliant patient and subsequent interventions needed. Chat bots can be used for patient engagement, personal coaching, also by sending automatic personalized messages and notifications to encourage them to continue participating in the trial.
Pharmacovigilance /safety, product complaints
NLP can automatically extract adverse event and product complaints data from various sources such as social media and publications to generate critical insights and to detect safety signals. AI can help simplify Adverse Events, product complaints process by replacing internal monitoring and tracking activities with intelligent risk management, signal detection, aggregate reporting and single case processing. ML can be used to automatically code Adverse Events, improving coding accuracy and speed. This will accelerate safety operations and provide higher quality data and improved compliance. This helps companies refocus time and resources on more strategic endeavors and improve patient safety.
NLP, ML, and predictive analytics can improve efficiency, compliance and transparency for research sites and sponsors, by helping researchers monitor performance and risk indicators for audits, change controls, non-conformance and complaints. AI can automate identification of operational lapses, FDA findings, root causes, risks, reminders, schedulers, actions and validation, automated review and approval, reporting and detection of trends and patterns. This can significantly lower trial cost, time and risk to deliver milestones on time.
AI can help bring products faster to the market and reduce cost of development by leveraging multiple evolving datasets for better insights and decision-making, and to enhance speed, operational efficiency, accuracy and compliance.
The way forward
Although, Clinical Research industry has started experimenting with AI, there have been challenges in adoption of AI including algorithm transparency, data privacy, data integrity, data security and data interoperability issues. Validation of algorithms used for clinical development is vital to understand AI-powered conclusions. Use of patient data is highly sensitive and requires appropriate regulatory compliance measures to protect data privacy. Data integrity and interoperability controls should be in place for data exchange. Data security measures including authorization and authentication are essential. Failure to meet compliance requirements may lead to reputational and financial consequences for manufacturers.
FDA is evolving a strategy to regulate the use of AI in clinical development, emphasizing the use of AI as a confirmatory tool. AI will lead this era as most manufacturers invest in AI to reap benefits of optimized time and cost, leading to faster time to market and competitive advantage.
- 2019 CIO Agenda: Life Science Industry Insights, Published: 15 October 2018, ID: G00367588, Analyst(s): Jeff Smith, Life Sciences CIOs, Accelerate Clinical Development With New Applications of Artificial Intelligence, Published: 17 January 2019 ID: G00367582, Analyst(s): Jeff Smith, Michael Shanler