AI in Cannabis Medicine: Personalised Dosing and Clinical Decision Support

Artificial Intelligence in Cannabinoid Therapeutics

• Machine learning models can analyse large patient datasets to identify dosing patterns
• AI can process multi-variable data including genetics, condition, medications, and prior responses
• Natural language processing tools can extract insights from unstructured patient-reported data
• AI-assisted clinical decision support is being developed for several areas of medicine including cannabis

Artificial intelligence offers substantial potential to address one of the central challenges in medical cannabis: the absence of standardised dosing protocols. The variability in individual response to cannabis preparations is driven by genetics, pharmacology, comorbidities, and product-related factors — a complexity that is well-suited to machine learning approaches capable of identifying patterns in high-dimensional data. AI in cannabis medicine is an emerging but rapidly developing field.

Predictive Dosing Models

• AI dosing models can use patient characteristics to predict likely effective dose ranges
• Models trained on large real-world datasets can identify subgroups that respond differently to products
• Continuous learning from patient outcomes refines model accuracy over time
• Predictive dosing tools must be validated in clinical populations before clinical deployment

The application of AI to cannabis dosing represents one of the most clinically promising near-term applications of machine learning in this space. Models that can predict, with meaningful accuracy, which patients will respond well to specific preparations and at what doses could significantly reduce the duration of the titration phase and improve the proportion of patients achieving good outcomes. Developing these models requires large, well-characterised patient datasets — creating a compelling case for multi-clinic data sharing initiatives.

AI in Patient Monitoring and Safety

• AI can identify early warning signs of adverse events or treatment failure in monitoring data
• Anomaly detection algorithms can flag unusual patterns in patient-reported outcomes
• AI-assisted triage can help clinics prioritise which patients need urgent clinical review
• Automated safety monitoring at scale could identify population-level signals not visible at individual level

Patient safety monitoring is an area where AI can provide substantial value at relatively low risk. The application of AI to identify unusual patterns in patient-reported monitoring data — patterns that might presage a serious adverse event or indicate treatment failure — does not require the AI to make clinical decisions; it simply helps clinicians prioritise where to focus their attention. This kind of AI-assisted triage is likely to become standard in high-volume medical cannabis practice.

Regulatory and Ethical Considerations

• AI clinical decision support tools used in prescribing may be classified as medical devices
• Algorithm bias is a real risk if training datasets are not representative of the prescribing population
• Transparency in AI decision-making is required by both regulators and ethical principles
• Patients must understand when AI is contributing to their treatment decisions

The deployment of AI in clinical settings raises important regulatory and ethical questions that the medical cannabis sector must navigate carefully. The MHRA’s evolving framework for AI as a medical device, the risk of algorithmic bias from non-representative training data, and patients’ rights to understand the basis of clinical decisions all require careful consideration. Companies developing AI tools for cannabis medicine should engage with regulators early and build transparency and explainability into their systems from the outset.