Applying generalisability and transportability analyses to realistic synthetic data in situations where no real-world data (RWD) are available (eg, because a drug has only recently been marketed) can help to inform which patient can benefit from the new drug. Generalisability and transportability analyses analyses should be considered as a statistical technique to provide valuable insights for clinical decision-making and to guide future trials, and under no circumstances as a replacement for randomised controlled trials (RCTs) with a more diverse trial population. It will be interesting to see how the proposed approach works in practice in the future, whether in the chosen example of a new antibody-drug conjugate for breast cancer treatment or in other conceivable situations where knowing the patient benefit in routine care is particularly important.
This pilot case shows how a repurposed QSP model could contribute to informed decision-making in everyday clinical practice. With increasing knowledge in the actual patient course, the model updates itself in a Bayesian approach to predict, in our case, the expected potassium course for the next 24 hours, which also takes planned drug administrations into account. Thus, the model prediction could give reason to preemptively modify potassium supplementation, to modify comedication affecting potassium concentrations, to reduce the spironolactone dose or, for safety, to arrange for additional laboratory measurements. Our use case presented here shows a proof-of-principle that this is also conceptually possible with mechanistic QSP models after being operationalized for this purpose.