Using Virtual Twin™ Technology for Model-informed Precision Dosing

In a recent Expert Review of Clinical Pharmacology article, I assessed the status and future direction of precision dosing in clinical medicine.1 Model-informed precision dosing (MIPD) is a modeling and simulation (M&S) approach in healthcare that is used to predict the most effective and/or least toxic drug dose for a patient. MIPD could revolutionize healthcare by reducing the incidence of adverse drug reactions (ADRs), improving drug efficacy, and increasing patient adherence. Precision dosing in clinical medicine is needed for drugs with a narrow therapeutic index (eg, anti-arrhythmics, anti-coagulants, anti-epileptics, anti-neoplastics, aminoglycoside antibiotics, immunosuppressants) and drugs with very wide interpatient pharmacokinetic/pharmacodynamic (PK/PD) variability. Populations at increased risk of medication-related harm also benefit from MIPD, including pediatrics, infants and neonates, geriatrics, pregnant women, those with rare diseases, oncology patients, and patients with impaired organ function.

The need for model-informed precision dosing

Physicians know about the obvious patient factors that influence drug exposure and possibly response, such as age, weight, renal function, liver function, frailty, and some drug-drug interactions (DDIs). Consider the case of an elderly female patient with renal impairment. Her dosing would normally follow the “start low, go slow” paradigm. Although possibly not realized, clinicians therefore already employ “in cerebro” M&S—they combine experience and patient factors to inform prescribing decisions.

However, the ability to leverage key “hidden” factors that alter drug exposure and/or response distinguishes MIPD from current attempts in clinical practice to do precision dosing. The hidden factors that can drive intra-subject variability include drug metabolizing enzymes and transporter (DMET) genotype and/or phenotype, organ sizes and blood flows, inflammatory status, and DDIs. Unlike in cerebro modeling, MIPD can consider these factors simultaneously to support a quantitative approach to selecting the right drug at the right dose at the right time for an individual patient. Let’s revisit the elderly female patient with renal impairment. MIPD would highlight that she might be on the 99th percentile for liver size and thus the “start low, go slow” approach for a drug cleared primarily by the liver will not work, although this might be the way physicians today would start treatment. Importantly, MIPD can simulate the multiple factors that determine clinical outcomes to identify a better tactic to individualize treatment, eg, poor metabolizer phenotype in a patient concurrently taking multiple CYP inhibitors with renal impairment (we call this a “dress rehearsal” for what may happen in real life; in other words, a “try before you buy” approach).

Current approaches to MIPD

Any precision dosing strategy must know the factors driving between-subject variability in exposure and/or response. Today, these factors can’t be integrated fully and quickly for point-of-care dosing decisions because they are simply too much for in cerebro modeling. Quantitative M&S methods provide a more predictive approach to precision dosing and include the use of basic nomograms, population pharmacokinetics (PopPK), and physiologically-based PK (PBPK). The most commonly used MIPD method in clinical practice today is PopPK linked to therapeutic drug monitoring (TDM) of plasma drug concentrations. This Bayesian method predicts subsequent doses for particular individuals based on their unique PK parameters. This is a very powerful way to interpret TDM data as it becomes available.

Using Virtual Twin technology to provide point-of-care dosing

The Simcyp® Simulator, the basis for the Virtual Twin technology, includes extensive demographic, physiologic, molecular, and genomic databases. The Simulator links in vitro data to in vivo ADME (absorption, distribution, metabolism, and excretion) and PK/PD outcomes to help explore potential clinical complexities prior to human studies and support decision-making in drug development. This enables the user to predict drug behavior in virtual populations instead of waiting for clinical studies which are costly, especially identifying the types of individuals who could be at extreme risk of ADRs. PBPK has been tremendously successful in drug development and has transitioned from an academic curiosity to a regulatory necessity.2 Its main use has been in predicting the clinical relevance of metabolic DDIs, but it is being increasingly used for prediction of PK in special populations, such as pediatrics.

Virtual Twin technology incorporates knowledge of biological and physiological functions and creates a computer-simulated model of each patient, replicating the patient’s various attributes that affect drug exposure and/or response. These attributes include the patient’s age, weight, height, gender, ethnicity, and genetics of DMETs. The Virtual Twin model has the potential to account for the patient’s fed or fasted state, co-morbid conditions and co-medications that affect the activity of DMETs, and their level of organ function. The Virtual Twin concept simply adapts some of the biological and physiological  data used to build a base PBPK model representative of the population, with individual information about these parameters that better matches the patient, eg, specific renal, liver, and cardiac function, hematocrit, and DMET genotype and phenotype. Monte Carlo simulation can be used to understand the extent of variability attributed to parameters that are not known for that patient (eg, any missing information that might be important in predicting dose). Recent publications have demonstrated the use of this technology to (1) predict olanzapine systemic exposure predicted in Virtual Twins when compared to actual drug concentrations in the corresponding patients,3 and (2) to develop an in silico quantitative systems toxicology (QST) model for citalopram to predict the likely occurrence of cardiotoxic events in real patients under different clinical conditions.4

Advantages to using the Virtual Twin approach for MIPD

The Virtual Twin approach can be applied to novel clinical scenarios and situations where no guidance is available regarding the dose for a particular patient. A good example of this is in pediatrics, where physicians are often left to guess the dose for their patients, leaving it to trial and error—this is concerning for all involved. Other advantages include flexibility to change with new additional information (eg, novel in vitro data for a disposition process), and the ability to simulate factors outside the range of clinical data that is used to inform other types of dose recommendations, such as the prescribing information. Thus, the Virtual Twin approach is flexible and supports assessing difficult scenarios occurring in clinical practice that require guidance about dose.

Examples of potential high-impact MIPD scenarios where the evidence is being collected include the treatment of resistant schizophrenia (clozapine initiation), getting the anti-coagulant dose right in patients with atrial fibrillation, atomoxetine dosing in children with ADHD, and treatment of solid tumors and hematological malignancies (kinase inhibitor initiation and titration).

Although there are challenges to implementing MIPD, we should focus on select areas of clinical medicine where Virtual Twin technology could lead to better patient outcomes, reduced hospital stays, and less time for patients in out-patient clinics.5 A really exciting prospect of Virtual Twin is its potential to help physicians and their patients in primary healthcare, including those who live far away from big hospitals.


[1] Polasek TM, Shakib S, & Rostami-Hodjegan A. (2018). Precision dosing in clinical medicine: Present and future. Expert Rev. Clin. Pharmacol. 11(8), 743–746.

[2] Jamei M. (2016). Recent advances in development and application of physiologically-based pharmacokinetic (PBPK) models: A transition from academic curiosity to regulatory acceptance. Pharmacol. Rep. 2, 161–169.

[3] Polasek TM, Tucker GT, Sorich MJ, Wiese MD, Mohan T, Rostami-Hodjegan A, Korprasertthaworn P, Perera V, & Rowland A. (2018). Prediction of olanzapine exposure in individual patients using physiologically-based pharmacokinetic modelling and simulation. Br J Clin Pharmacol. 84(3), 462–476.

[4] Patel N, Wiśniowska B, Jamei M, & Polak S. (2018). Real patient and its virtual twin: Application of quantitative systems toxicology modeling in the cardiac safety assessment of citalopram. AAPS J 20(6), 1–10.

[5] Darwich A et al. (2017). Why has model‐informed precision dosing not yet become common clinical reality? Lessons from the past and a roadmap for the future. Pharm. Therap. 101 (5), 646–658.

To get an overview on how Virtual Twin technology can be used to predict drug exposure in individual patients, particularly in the areas of psychiatry and oncology, watch this webinar.