Date of Award


Document Type


Degree Name

Doctor of Philosophy in Pharmaceutical Sciences


Pharmaceutical Sciences

First Advisor

David Taft

Committee Chair and Members

David Taft, Chair

Jaclyn Cusumano

Rutesh Dave

Vamshi Jogiraju


Antimicrobial pharmacodynamics, Clinical pharmacology, In-vitro time-kill assays, Pharmacometrics, Simcyp, Translational modeling


Pharmacokinetic-pharmacodynamic (PKPD) modeling and simulation have emerged as pivotal tools in drug development and usage. Such models characterize typical trends in data and quantify the variability in relationships among dose, concentration, and desired effects. For antibacterial applications, models characterizing bacterial growth and antibiotic-induced bacterial killing offer insight into interactions between antibiotics, bacteria, and the host. Simulations from these models predict outcomes for untested scenarios, refine study designs, and optimize dosing regimens.

Enterococcus faecalis, a significant opportunistic bacterial pathogen with increasing clinical relevance, is commonly found in the gastrointestinal tract but can lead to severe infection, such as endocarditis. Treatments for E. faecalis endocarditis involves combination antibiotic therapy, such as beta-lactam antibiotics and aminoglycosides. However, due to the toxicity of aminoglycosides, the primary treatment is typically double beta-lactam therapy—ampicillin and ceftriaxone. Eradicating an E. faecalis infection typically requires a lengthy six-week course of antibiotic treatment. However, keeping patients in hospitals for such an extended duration is impractical. Therefore, the objective of this thesis project is to explore the extension of double beta-lactam therapy to outpatient antibiotic treatment (OPAT). This approach is gaining importance due to the rising risks of hospital-acquired infections and escalating healthcare expenses. Leveraging the stability of penicillin G, which can be stored at room temperature for extended periods, makes it a promising candidate for OPAT, offering potential benefits in terms of both efficacy and cost-effectiveness. Despite limited evidence for penicillin G plus ceftriaxone, this research successfully bridges the gap through in-vitro time-kill assays and the subsequent development of a semi-mechanistic model for this antibiotic combination against E. faecalis isolates.

This dissertation research evaluated 21 clinical strains of E. faecalis isolated from infected patients' blood, sourced from Mount Sinai Health System and a hospital in Detroit as part of Dr. Jaclyn Cusumano’s American Association of Pharmacists (AACP) new investigator award research project. The first aim was to conduct susceptibility testing on these isolates. This testing played a pivotal role in guiding antibiotic therapy by determining a drug's minimum inhibitory concentration (MIC) for a specific bacterial strain, offering insight into its effectiveness. The project highlights the importance of knowing a patient's strain susceptibility since it influences the dosing regimen or treatment strategy. After susceptibility testing using broth microdilution techniques, strains were categorized as highly susceptible (MIC ≤ 2 μg/ml) or less susceptible (MIC = 4 μg/ml) to penicillin G.

The next phase of the project involved in-vitro time-kill assays—a gold standard method for testing antibiotic concentrations and synergy in combination therapies. All 21 patient isolates were tested with penicillin G monotherapy and in combination with ceftriaxone, along with testing ampicillin and ceftriaxone combination therapies for comparison. It was noted that both combinations showed efficacy for strains highly susceptible to penicillin G (MIC ≤ 2 μg/ml), exhibiting bactericidal and synergistic activity. However, both treatments demonstrated poor performance for the less susceptible strains (MIC = 4 μg/ml). This observation focuses on the importance of in-vitro pharmacodynamic studies in understanding antibiotic action dynamics, forming the basis for the semi-mechanistic model. These 24-hour time-kill assays strongly suggested further investigation into the penicillin G and ceftriaxone combination, while considering the differential effects of the combination on more and less susceptible strains. Semi-mechanistic models were created for two out of the twenty-one tested strains, one with high susceptibility and another with lower susceptibility, with the goal of understanding the bacterial growth and drug kill effect in greater detail along with testing different dosing regimens.

Following the typical progression of constructing a semi-mechanistic PK-PD model, a bacterial sub-model was created by employing intensive sampling during time-kill assays. This approach enabled the comprehension of the complete bacterial growth dynamics for both strains. By employing non-linear least squares regression within RStudio, the predictive model was effectively fitted to the observed data, providing estimates of essential bacterial growth parameters. The utilization of the Gompertz growth model yielded a remarkably close match between predicted and observed data, giving confidence in the accuracy of the estimated growth parameters.

Subsequently, the focus shifted to obtaining the most suitable pharmacodynamic (PD) parameters to accurately encapsulate the drug's antibacterial effects. This necessitated the use of a mathematical model. A widely employed model for this purpose is the Sigmoidal Emax model—an empirical model that is widely published. This model emerged as a valuable tool for formalizing the interpretation of experimental data and understanding the influence of altering penicillin G concentrations, both individually and in conjunction with ceftriaxone.

Leveraging the data analysis capacity of RStudio, nonlinear least squares regression analysis was used to intricately fit the sigmoidal Emax equation to the observed data. This led to obtaining vital parameters, including Emax (maximum effect), EC50 (half-maximal effective concentration), and the sigmoidicity factor. Subsequent evaluation of goodness of fit based visual predictive checks and low standard errors in estimated parameters confirmed the favorable alignment between the predicted model and observed data.

Physiologically based pharmacokinetic (PBPK) modeling and simulation stands as a well-established approach that bridges insights from preclinical studies to clinical outcomes. By combining drug-specific information with a comprehensive understanding of physiological and biological processes at the organism level, PBPK models mechanistically depict the behavior of drugs within biological systems. This enables the a priori simulation of drug concentration-time profiles. What distinguishes PBPK modeling is its unique capability to account for physiological variations within specific populations, offering predictive insights into pharmacokinetics tailored to those groups. This thesis project ventured into two vital applications of PBPK models: extrapolating novel clinical scenarios and exploring pharmacokinetics in special populations, particularly the geriatric demographic.

With the aim of comprehending the pharmacokinetics of penicillin G and ceftriaxone, the project leveraged the Simcyp® Simulator, a modeling and simulation tool that is widely used in drug development. This platform pools the anatomical, physiological, drug-related, and trial design parameters to generate plasma drug concentration profiles. The simulated concentrations were compared against published data, with the fold error—a ratio of simulated to observed values—serving as a benchmark for model accuracy. Typically, predictions within a fold error range of 0.5 to 2 are deemed acceptable.

Upon verification within the healthy population, the models were extended to geriatric subjects utilizing the Simcyp® population library. The same fold error criteria were applied, and the models adeptly predicted concentrations across both young and elderly populations. Remarkable differences in pharmacokinetics were seen in the geriatric cohort compared to a young adult population. Notably, for penicillin G, the AUC increased by 46% in the elderly due to an almost 47% decline in total clearance, stemming from a 49% reduction in glomerular filtration rate (GFR).

Further expanding the PBPK model for penicillin G, the inclusion of a pharmacodynamic (PD) component led to the final goal of this project. Lua scripting in Simcyp® was utilized to build the PD model. This model used an equation that combined the bacterial growth model with the drug's inhibitory effect via the Emax model. The impacts of monotherapy and combination were explored through the modulation of PD parameters. Consequently, when co-administered with ceftriaxone, kill rates for penicillin G increased, and IC50 values decreased, indicative of ceftriaxone's augmentative effect. The free (unbound) plasma concentration-time profile from the developed PBPK model was linked as input to the PD model, facilitating testing and simulation of diverse penicillin G dosing regimens.

Notably, penicillin G, a time-dependent beta-lactam antibiotic, exhibited a strong correlation with the PK/PD index %fT>MIC (% of the dosing interval with a free concentration above MIC). This was especially pertinent for high-susceptibility strains, wherein continuous infusion of penicillin G led to the most significant reduction in bacterial density, irrespective of combination therapy or monotherapy. However, for low-susceptibility strains, the scenario differed, revealing that reliance on a single PK/PD index is not all-encompassing.

For the geriatric population, the PBPK-PD model aligned with literature-backed dosing modifications for penicillin G. For highly susceptible strains, increasing the dosing interval or reducing the dose resulted in comparable reductions in bacterial density. Conversely, in low7 susceptibility strains, even an increase in AUC within the geriatric demographic failed to eradicate the bacteria.

In summary, this comprehensive thesis journey navigates through the in-vitro bacterial studies and pharmacokinetic-pharmacodynamic (PKPD) modeling and simulation. This project sheds light on the ability to integrate in-vitro data with PBPK models which not only predict untested scenarios but also help dosing strategies. Overall, by addressing the clinical challenge of E. faecalis infections, the project showcased the extension of double beta-lactam therapy to penicillin G and ceftriaxone combination through a stepwise development of semi-mechanistic PK/PD model.