Before the ICH Harmonized Tripartite Guideline Q9—“Quality risk management”—was introduced in 2005, the pharmaceutical industry was evolving but lacked a structured, scientific, and systematic approach. Various stakeholders, including the industry, regulators, and patient rights groups, recognized the need for a methodical risk assessment process. However, clear guidance was missing.

Previous to the ICH Q9 framework, risk assessment approaches were characterized in three ways: 1) they were reactive, addressing issues only as they emerged rather than proactively identifying and mitigating risks; 2) they were variable, with significant differences in practices across companies and regions leading to inconsistencies in drug safety and quality standards; and 3) they had limited formalization. Although some risk management activities were present, there was no comprehensive, formalized approach that fully integrated risk management into all aspects of pharmaceutical development and manufacturing.

From reactive to proactive: A paradigm shift

The International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH), established in 1990, brings together regulatory authorities from the United States, European Union, and Japan. Its primary mission is to create a unified system of pharmaceutical requirements, ensuring the registration of safe, effective, and high-quality medicines worldwide in a resource-efficient manner. The implementation of ICH Q9 has significantly improved the industry’s ability to identify, assess, and manage risks throughout the life cycle of pharmaceutical products.

The ICH Q9 guideline was developed to help pharmaceutical companies implement quality risk management (QRM) in their manufacturing processes, encouraging a more adaptable approach to regulatory compliance while promoting continual improvement.

Along with ICH Q8 (R2)—“Pharmaceutical development—Scientific guideline” and ICH Q10—“Pharmaceutical quality system—Scientific guideline,” ICH Q9 is one of the ICH Q-topics that encourages the development of science-based and risk-based approaches to quality. It aims to focus the behaviors of both industry and regulatory authorities on the two primary principles of QRM:

Science-based: Risk evaluation should be grounded in scientific knowledge and ultimately linked to patient safety.

Proportional: The level of effort, formality, and documentation in the QRM process should align with the level of risk involved.

Overview of a typical quality management process (ICH Q9 Briefing Pack)

For two decades, the pharmaceutical and biopharmaceutical industries have employed these principles to streamline their procedures, processes, and systems. However, new technologies such as artificial intelligence (AI) present opportunities to reimagine risk assessments in R&D (the “left” side of the business) as well as quality by design (QbD) and commercial operations (the “right” side).

This article focuses on the development side of risk assessment and the possible uses of AI.

How AI is revolutionizing pharma risk assessment

AI-enabled risk assessment for QbD in drug product development provides a sophisticated approach to managing and mitigating risks throughout the development life cycle. Starting by identifying the quality target product profile (QTPP) and defining the control strategy, AI can streamline operations, reduce bias, and enhance the efficiency of risk assessments.

Identifying the QTPP

AI can analyze historical data from an organization’s database or global databases of similar products to predict potential QTPPs for an active ingredient or drug product. It can also assign criticality scores to QTTPs based on data patterns, reducing operator bias and improving efficiency. Scientists can use this information to make informed decisions about which QTTPs are critical, which are noncritical, and which require further evaluation.

Generating process maps or process flow diagrams

AI can analyze data to automatically populate process flow stages based on patterns from similar products. It can also auto-populate inputs and outputs from the database, reducing bias. AI assigns criticality scores to inputs and outputs, and scientists can fine-tune these elements to enhance work efficiency.

Identifying critical quality attributes (CQA) and critical material attributes (CMA)

AI can use data from previous stages or similar products to auto-populate CQAs and CMAs. Advanced algorithms assign severity and likelihood scores to these attributes and parameters. Predictive analytics can assess potential failures and their effect on a product. Scientists should validate or adjust these identifications as needed.

Generating an Ishikawa or a cause-and-effect matrix

AI can enhance efficiency in building Ishikawa diagrams or generating cause-and-effect matrices based on CMAs and critical process parameters. It elucidates complex matrices and cause-effect relationships, making decision-making more comprehensive and unbiased.

Design of experiments

For attributes and parameters without predictions, AI can design and optimize experiments to explore relationships between CQAs, CMAs, and process parameters. It identifies significant factors affecting product quality and quantifies associated risks.

Generating risk assessments via failure mode and effects analysis (FMEA)

With its advanced algorithms, AI can auto-populate information from previous stages and generate FMEA-based risk assessments. It reduces bias by suggesting action plans based on historical data and recommending design, operational, procedural, or systemic controls.

Defining action plans and recommending control strategies

AI-enabled risk assessments can populate current controls and propose mitigation strategies. It can route these strategies to responsible groups, assign target dates based on priority, monitor completion, and send reminders or escalations as needed. AI also checks the effectiveness of control strategies against similar products.

Ongoing processes where AI adds value to QbD

Real-time data analysis: AI can analyze real-time data from ongoing experiments, enabling researchers to make informed decisions more quickly. Additionally, AI can update risk assessments whenever new information is added to the existing knowledge base.

Communication: AI can be configured to follow predefined workflows to keep stakeholders informed about relevant risks, promoting or demoting risk information in real time.

Generative AI in risk assessment: AI automatically generates risk reports, ensuring accuracy and reducing administrative burden, particularly important for regulatory submissions with strict timelines.

How AI unlocks new dimensions of risk assessment in development

Traditional risk assessment in development projects can be time-consuming and prone to human error. AI-powered tools transform this process by analyzing vast amounts of data and uncovering hidden patterns. This enhances development efforts by improving efficiency, accuracy, and the ability to predict and mitigate emerging challenges.

Here are the ways that AI simplifies the development process.

Enhanced decision-making: AI provides data-driven insights that improve the accuracy and reliability of decision-making throughout the drug development process.

Better process understanding: AI uses data from all available knowledge bases, including historical data, trends, and past events, collectively referred to as prior knowledge.

Increased efficiency: AI automates time-consuming tasks such as data analysis, documentation, and monitoring, thereby accelerating the drug development timeline.

Improved compliance: AI ensures continuous compliance with regulatory requirements, reducing the risk of noncompliance.

Better patient outcomes: AI enhances the safety and efficacy of drugs by identifying and mitigating risks early, leading to more successful therapies reaching the market.

Cost optimization: AI identifies potential risks early, preventing costly late-stage failures and allowing for more effective resource allocation.

Charting a sensible course in risk assessment with AI

Emerging standards and guidelines are shaping the use of AI in risk assessment within the drug development process. To support its mission of protecting, promoting, and advancing public health, the FDA Center for Drug Evaluation and Research (CDER), in collaboration with the Center for Biologics Evaluation and Research (CBER) and the Center for Devices and Radiological Health (CDRH), including the Digital Health Center of Excellence (DHCoE), has published a document to guide discussions with stakeholders on the use of AI and machine learning in drug development, including medical devices intended to be used with drugs.

The guidelines emphasize several crucial aspects to ensure the effective use of AI in drug development.

First, they stress the importance of explainability, which means AI models must be understandable and interpretable by humans. This helps stakeholders grasp how decisions are made by the AI systems.

Additionally, reliability and robustness are key, ensuring that AI systems perform consistently across various conditions.

The guidelines also focus on privacy and security, emphasizing the need to protect patient data and handle information securely.

Finally, the guidelines address bias reduction to minimize biases in AI models, aiming to ensure fair and equitable outcomes for all users.