Home / Resources / Blog / Virtual Twin Patients Can Bridge Data Gaps for COVID-19 Halted Clinical Trials

Virtual Twin Patients Can Bridge Data Gaps for COVID-19 Halted Clinical Trials

By: Dr. Hugo Geerts and Dr. Piet van der Graaf

COVID-19 has had a massive impact on clinical research globally. Researchers have had to pivot quickly to test on-market drugs for activity against SARS-CoV-2, while also investigating novel compounds, drug combinations and potential vaccines. Most non-pandemic-related clinical research was put on hold so that staff could focus on providing care for the rapidly growing numbers of COVID-19 patients and to reduce the risk of infection for trial participants. Those studies that did continue were predominantly for life-threatening conditions.

Alzheimer’s Disease Trials

For many patients with Alzheimer’s disease, the clinical trial halt meant an involuntary drug holiday, changes in medications for addressing mental health issues, and missed site visits that could be only partially replaced by remote monitoring. There were also concerns that the drug holiday would allow the pathological mechanism targeted by the drug to resume, and that some members of this vulnerable patient population would succumb to COVID-19.

Compensating for Missing Data

As trials restart, researchers need to determine how best to combine data from subjects who completed the study, with those who are returning but have experienced protocol deviations or amendments, and others who have dropped out or passed away.

Statistical methods, such as Last Observation Carried Forward, which are traditionally used to account for missing data, will likely be inadequate to handle such complex, fragmented datasets. Furthermore, the number of patients who have completed the trial may be too small to serve as a training set to allow machine learning to generalize their responses to other trial participants.

In a paper published in CPT: Pharmacometrics & Systems Pharmacology in May,1 we described the virtual twin patient computational modeling approach, which combines limited clinical data with underlying biological principles to generate predictions at the individual patient level and could prove invaluable in helping to recover halted trials.

Creating Virtual Twin Patients  

This approach harnesses mechanistic physiologically based pharmacokinetic (PBPK) and quantitative systems pharmacology (QSP) modeling and allows the readouts from complex, fragmented clinical datasets to be harmonized in a biologically relevant way.

PBPK modeling uses in vitro and in vivo data to predict drug performance across all the body’s organs. QSP modeling integrates quantitative drug data with knowledge of its mechanism of action. QSP can also predict how drugs modify cellular and neuronal networks in space and time and how they are impacted by human pathophysiology and therapeutic interventions. QSP is a promising technology for Alzheimer’s drug development; it integrates knowledge of the disease’s complex pathology and the relationships between the many mechanisms involved in the pathophysiology and progress of the disease.

First, a computer simulated PBPK model is created for each patient. This model replicates the patient’s unique attributes that affect a drug’s fate in their body and, hence, its effects. Those include the patient’s age, weight, sex, ethnicity, and genetics for drug metabolizing enzymes and drug transporters. The model also reflects the patient’s current drug dosage, fed or fasted state, comorbid conditions, comedications, and their level of organ function which affects the activity of certain metabolic enzymes and transporters. The PBPK model can account for potential drug-drug interactions both from a PK and a pharmacodynamic standpoint. This polypharmacy capability is especially important because elderly patients with Alzheimer’s disease take an average of eight medications in clinical practice.

Then, a virtual twin QSP model is developed for each patient using the same comedications, genotypes, biomarkers, and β‐amyloid and tau load as in the trial. The main neuronal circuitry involved in cognition is also incorporated. Fortunately, the pharmacology and target exposure for many comedications used to treat Alzheimer’s disease in clinical practice are well documented. Some of the common genotype variants involved, such as APOE, 5‐HTTLPR, rs23351, and COMTVal158Met, can also be derived from imaging studies.

The pharmacology and PK profile of the investigative drug is then layered on top of each virtual twin model, allowing the cognitive trajectory of each patient to be simulated. Adding imaging and biological fluid biomarker data further defines the levels of target engagement for the investigational drug and the disease state. As each virtual twin represents a specific patient, it can also be adapted as necessary to account for changes in their medication.

Model Validation

Once virtual twin patients have been created for all the participants who completed the trial, their model-generated expected clinical outcomes can then be compared with their actual outcomes. When this validation process has been completed satisfactorily, the virtual twin patient models can be used to predict the clinical progression of patients that have left the study. It can also “correct” for protocol amendments when the trial restarts, allowing those incomplete results to be pooled with complete ones to salvage that valuable information.

While this virtual twin patient approach is new, PBPK modeling has been widely accepted by pharmaceutical companies and global regulatory agencies for many years and several proof-of-concept studies have been published for the newer QSP models.2-5

Conclusion

Virtual twin patients represent a powerful, objective method for extrapolating and bridging data gaps in clinical trials and ensuring that no scientific knowledge is lost when a study is halted temporarily.

Dr. Hugo Geerts is head of QSP, neuroscience, and Dr. Piet van der Graaf is senior vice president of QSP at Certara.


References

  1. Geerts H, van der Graaf PH. Salvaging CNS Clinical Trials Halted Due to COVID‐19. CPT: Pharmacometrics & Systems Pharmacology. 28 May 2020. https://ascpt.onlinelibrary.wiley.com/doi/10.1002/psp4.12535
  2. Timothy Nicholas, S.D. et al. Systems pharmacology modeling in neuroscience: Prediction and outcome of PF-04995274, a 5-HT4 partial agonist, in a clinical scopolamine impairment trial. Adv. Alzheimer Dis. 2, 83–98 (2013).
  3. Geerts, H., Spiros, A. & Roberts, P. Impact of amyloid-beta changes on cognitive outcomes in Alzheimer’s disease: Analysis of clinical trials using a QSP model. Alzheimer Res. Ther. 10, 14 (2018).
  4. Geerts, H. & Spiros, A. Learning from amyloid trials in Alzheimer’s disease. A virtual patient analysis using a QSP approach. Alzheimers Dement. 07 April 2020. https://doi.org/10.1002/alz.12082.
  5. Roberts, P., Spiros, A. & Geerts, H. A humanized clinically calibrated QSP model for hypokinetic motor symptoms in Parkinson’s disease. Front. Pharmacol. 7, 6 (2016)
Back to top