Aims: To produce a viral kinetic (VK) model of Cytomegalovirus (CMV) that predicts changes in viral load within an individual as the infection responds to antiviral therapy and the immune response, based on data from an immunocompromised paediatric population.
Methods: Complete clinical datasets were taken from 335 bone marrow transplant (BMT) patients receiving care at Great Ormond Street Hospital (GOSH) between January 2010 and December 2014. 86 of these patients exhibited a serious active CMV infection following their BMT-associated immunosuppressive treatment regime. A total of 1,598 CMV viral load observations were included from these patients. For each of the patients studied, we had access to clinical history, treatment outcomes, and the results of all clinical tests undertaken. These tests include regular white cell counts, white cell subset counts, full blood counts, and a selection of viral load PCRs. Alongside this rich clinical data, we have access to full drug administration datasets for each patient during their stay at the hospital.
Viral load data was fitted to a viral growth model using NONMEM V7 3.0 , a growth inhibition term related to antiviral treatment was incorporated into the VK model alongside a growth limiting V MAX term that scales the rate of viral growth against the maximum amount of virus that is possible within an individual. Total lymphocyte count (TLC) was chosen to represent the efficacy of a patient's immune response against the virus due to the rich data available on this. From the 86 patients with active CMV infections, 13,933 TLCs were available from the course of their BMT recovery, and were utilised as a scalar on an immune kill parameter. Virions are also assumed to have a fast turnover rate, represented by a constant clearance parameter in the model. Viral load was used as the primary predictor of treatment outcome as high viral loads (>1 million copies/mL) are typically associated with higher levels of mortality.
Results: The viral kinetic model was shown to appropriately fit the data from the BMT patients, accurately estimating the viral growth and decay during key moments in the patient's disease progression cycle. Model parameter estimates generated during the estimation were appropriate, given the biological context, with viral doubling time agreeing with the faster ~1.4 day CMV doubling time previously shown in BMT patients with CMV infections . Antiviral efficacy was estimated to be highly variable within the patient population, with the inclusion of immune parameters improving the fit, and accounting for large portions of the variation in viral growth. The model can provide realistic disease trajectory simulations that can be fitted to data from newly infected patients, as they present.
Conclusion: A dynamic VK model has now been developed for CMV. Future work will include the estimation of whole body CMV content, by scaling viral load up to match total predicted blood volume for each patient, and the inclusion of individual antiviral PK information. Following this, the use of rich CMV sequencing data currently being collected from BMT patients may allow us to monitor and predict the emergence of drug resistance, and model the relationship between it, estimated total viral load, and antiviral efficacy.
References:  Beal SL, Sheiner LB, Boeckmann AJ & Bauer RJ (Eds.) NONMEM Users Guides. 1989-2013. Icon Development Solutions, Ellicott City, Maryland, USA.
 Mueller, Nicolas J. "Cytomegalovirus: Why Viral Dynamics Matter." EBioMedicine 2.7 (2015): 631.