Oncology drug developers face a distinct set of challenges. Oncology drugs are often very toxic which precludes conducting clinical trials in healthy volunteers. In addition, cancer patients differ from healthy people in terms of their demographics and physiology. These changes mean that the pharmacokinetics of drugs may be altered in this population compared to healthy volunteers.
Furthermore, cancer patients often take multiple medications concurrently to treat comorbidities and treatment-associated side effects. Therefore, this already fragile population faces an increased risk of drug-drug interactions (DDIs). For all these reasons, there is significant interest in developing a virtual population to aid physiologically-based pharmacokinetic (PBPK) modeling of cancer drugs. In this blog post, I’ll discuss how we developed the Sim-Cancer population which is available in version 17 of the Simcyp Simulator.
Simcyp Simulator population databases
The Simulator allows users to change the system parameters to simulate different virtual populations. Currently, 19 populations are available comprising different ethnic groups, life stages (geriatric, pediatric, pregnancy, and pre-term babies), and disease states (renal/hepatic impairment, obesity, inflammation, psoriasis, and cancer). These population databases enable building models to predict drug exposure in different populations. For example, people from different ethnicities have demographic, physiological, and genetic differences that can impact pharmacokinetics. These PK differences can then impact the optimal dosage strategy for different populations.
Why develop a PBPK cancer population?
To accelerate the approval of oncology drugs, we need to understand the mechanisms of drug disposition in a cancer patients. A PBPK cancer population is a useful platform for investigating tumor disposition and therefore, impact on treatment regimens. It also enables conducting virtual DDI trials to assess the potential for safety concerns.
Our consortium members—Sanofi and Genentech—developed PBPK cancer populations in the Simcyp Simulator.1,2 They also generously provided data to us for developing our own PBPK cancer population. This highlights the strength of our consortium’s research collaborations. And, we used this data along with other published references for not only building the population, but also for independent verification.
Sim-Cancer population development
Several considerations informed developing the Sim-Cancer population. First, our focus was on solid tumor types. In terms of comorbidities, we tried to exclude the data for those subjects where possible. However, the data for treatment-naïve cancer patients is very limited. Hence, data from treated patients was often included. The base population was the Sim-North European Caucasian population. This target virtual population therefore represents cancer in Caucasian patients. It was also designed to be a “generic population” representing the general features of cancer. Users can tailor the generic Sim-Cancer population to model specific cancer types.
Now, I’ll discuss the physiological inputs and some of the performance verification that we did with the Sim-Cancer population. Cancer is predominantly a disease of older age. Therefore, we added a new user-defined age distributions feature to enable capturing the age distribution of cancer patients. These are fully customizable and allow greater flexibility for defining specific population age demographics. The North European Caucasian age/height relationship was used as the base for our target population, which adequately described both the mean and individual male and female data from cancer patients.
A key feature in cancer is weight loss in patients. Therefore, we used a cancer-specific relationship between height and body weight.1 Much of the weight loss in cancer is due to reduction in muscle, which in turn leads to increases in serum creatine levels. Our simulations in the Sim-Cancer population were able to reasonably simulate clinically observed data for serum creatine levels in cancer patients.
One caveat of the cancer population is that the traditional glomerular filtration rate (GFR) prediction methods—the Cockcroft-Gault and the Modification of Diet in Renal Disease (MDRD)—are less accurate in cancer patients. So, we created a new function to allow Lua scripting to incorporate a Wright formula3 to predict GFR. This approach enabled adequately predicting GFR for cancer patients and also allows for greater flexibility for applying different GFR models.
Cancer patients also have alterations in blood plasma volume and proteins. The Sim-Cancer model has a correction to account for increased plasma volume. Cancer patients also have reduced levels of albumin and increased levels of α1-acid glycoprotein. We collected data from multiple sources on how these protein concentrations varied by age and sex in cancer patients. When we performed stimulations with our model, we were able to capture the individual data.
Drug transporters play a vital role in governing drug concentrations in the blood, liver, brain, intestine, lung, and kidney. In addition, transporter protein-mediated drug-drug interactions (DDIs) can cause loss of drug effectiveness and toxicity. Thus, we wanted to examine whether we would need to incorporate alternations in hepatic transporters in our Sim-Cancer population.
In 2016, a database of hepatic transporter abundances were compiled for healthy Caucasians and cancer patients.4 We performed a meta-analysis of this literature data to examine relationships between cancer patients and healthy volunteers for different hepatic transporters. Compared to healthy volunteers, cancer patients showed an increase in some transporters’ abundance (OATP1B1), reductions in others (OATP2B1, MATE1, OCT1, and OATP1B3), and others which did not change. These data are all incorporated within the Sim-Cancer population.
Sim-Cancer performance verification
This performance verification was done using default Simcyp substrates. In total, we used eight substrates covering a range of CYP-mediated metabolism as well as a P-gp substrate. We performed simulations in the version 17 Sim-Cancer population to compare to clinical trial data. Importantly, study designs were matched including age and sex. We assessed the prediction accuracy for PK parameters to observed clinical data. In addition, we made comparisons to healthy volunteer trial observations as well as simulations.
Cancer patient vs. healthy volunteer PK
Cancer patients have significant PK differences from healthy volunteers. Several studies have shown a decrease in clearance in cancer. However, when we conducted performance verification for the Sim-Cancer population, no change in CYP abundance was required. This discrepancy may be due partly to the fact that healthy volunteer studies are often conducted in people much younger than the typical cancer patient. When we incorporated the physiological changes characteristic of older populations—principally decreased liver volumes—you already obtain the decreases in CYP abundance between the two populations.
Finally, we compared the observed and predicted area under the curve (AUC) and clearance ratios for cancer patients and healthy volunteers. The Sim-Cancer population did a reasonable job at predicting these ratios. Some compounds studied have highly variable PK. Thus, which clinically observed studies are chosen can impact the fold-difference in clearance between cancer patients and healthy volunteers.
To summarize, we’ve established a generic Sim-Cancer population for simulating drug disposition in cancer patients. We conducted performance verification of the physiological parameters unique to cancer patients and clinical pharmacokinetic verification for a range of Simcyp Simulator library compounds. In the future, we’d like to model tumor distribution of both small and large molecule drugs. The Sim-Cancer population will form the foundation for this research.
To learn more about this work and other exciting updates to the Simcyp Simulator, please watch this webinar that I gave with my colleagues, Nikunjkumar Patel and Matthew Harwood.
 Cheeti S, Budha NR, Rajan S, Dresser MJ, & Jin JY. (2013). A physiologically based pharmacokinetic (PBPK) approach to evaluate pharmacokinetics in patients with cancer. Biopharm Drug Dispos. 34(3):141-154. doi: 10.1002/bdd.1830 [doi].
 Thai HT, Mazuir F, Cartot-Cotton S, & Veyrat-Follet C. (2015). Optimizing pharmacokinetic bridging studies in paediatric oncology using physiologically-based pharmacokinetic modelling: Application to docetaxel. Br J Clin Pharmacol. 80(3):534-547. doi: 10.1111/bcp.12702 [doi].
 Wright JG, Boddy AV, Highley M, Fenwick J, McGill A, & Calvert. (2001). Estimation of glomerular filtration rate in cancer patients. Br J Cancer. 84(4):452-459. doi: 10.1054/bjoc.2000.1643 [doi].
 Burt HJ, Riedmaier AE, Harwood MD, Crewe HK, Gill KL, & Neuhoff S. (2016). Abundance of Hepatic Transporters in Caucasians: A Meta-Analysis. DMD. 44(10):1550-61. doi: 10.1124/dmd.116.071183 [doi].