CBD Drug-Interaction Models: A Predicted Exposure Ratio Is Not a Patient Result
CBD interaction models can identify combinations worth investigating, but a predicted exposure increase is not a measured adverse event or an individual patient’s drug concentration. A modeling study illustrates why assumptions and external validation determine how far predictions can travel.
The study combined several approaches
Researchers integrated enzyme-inhibition information with basic screening models, mechanistic static models and physiologically based pharmacokinetic models. The work considered CBD and 7-hydroxy-CBD alongside sensitive substrates of multiple drug-metabolizing enzymes.
These are in-silico predictions informed by experimental and clinical data. The paper was not a new randomized patient trial of every predicted drug combination.
Model categories describe different questions
A conservative screening model identifies potential risk. More detailed models incorporate assumptions about organ exposure, inhibition, metabolism and absorption. Their outputs should not be treated as interchangeable certificates of clinical harm.
The model used estimated inhibitor concentrations at sites that cannot readily be measured directly. A prediction for a sensitive probe substrate therefore does not necessarily give the same magnitude of interaction for every medicine using that enzyme.
Food and repeated dosing changed the inputs
The investigators adjusted absorbed fractions to describe less-than-dose-proportional fasted exposure and higher exposure with food or repeated dosing. Predicted CBD peaks still tended to be high under some fasted conditions.
External validation was available for selected regimens, not independently for every dose used during development. Reproducing a dataset used to tune a model offers a different test from predicting a separate dataset.
How to apply the prediction
Identify the actual drug, dose, schedule, food conditions and clinical population. Ask which inputs were measured, which were assumed and whether the relevant scenario was externally tested.
A predicted area-under-the-curve ratio describes systemic exposure, not automatically sedation, toxicity or driving performance. Clinical consequences depend on the affected medicine and patient. This study supports more informed interaction research and monitoring; it does not supply a universal CBD dose adjustment or establish causation in an individual adverse-event case.
For context, see our discussion of THC concentrations and the separate question of impairment.
Primary research
Sulafa Al Sahlawi and colleagues. Integrating Static and Dynamic Models to Predict Cannabidiol‐Mediated Metabolic Drug–Drug Interactions. Clinical Pharmacology and Therapeutics (2026). DOI: 10.1002/cpt.70509.
