When your physician writes you a prescription, they do so knowing that the dose prescribed will (on average) be safe and beneficial to you, but that it is likely certainly not optimal for you as an individual (as has been the topic of a previous blog post).
The balance between absorption, distribution and elimination of a medicine differs between individuals. As such, dosing should be tailored to the individual. A principal objective of the Simcyp PBPK Simulator is to determine optimal dosing regimens within ‘virtual’ populations to predict safe and effective doses in real humans.
However, predicting removal or elimination of a medicine depends on many factors including liver volume. Because of variation in liver size, people of different ethnicities, genders, ages and disease status will eliminate or remove these medicines at different rates. We undertook a two-phase study to collect and analyze data on liver size in healthy individuals of different ages and ethnicities respectively (Fig). The objective was producing a model that could adequately explain an individuals’ liver size in terms of their age, ethnicity and anatomy.
A precedent meta-analysis published in 2005 by Dr. Trevor Johnson et al  provided a framework for collecting data and a liver volume model based on an individuals’ body surface area. One principal benefit of extending this meta-analysis was taking an objective approach to obtaining and evaluating publication records containing liver size data from PubMed®. Briefly, we utilized medical subject heading terms (MeSH) that had been used to index the Johnson et al, Pomposelli et al manuscripts and all publications cited by these two (64 articles in total) (see Figure). This allowed us to develop and tailor a PubMed search query using terms that had been previously employed for indexing PubMed articles regarding the determination of liver size. After we obtained publication records, pre-defined exclusion criteria were used to delineate a relevant data-set. Missing parameters (eg, age, body-weight and height) were simulated using the Simcyp Simulator, this allowed compiling the final data-set.
However, the problem remained of how to evaluate a new model that would better predict liver volume. Fortunately, with Certara’s Phoenix group, we had the opportunity to apply state-of-the-art Non-Linear-Mixed-Effects modeling methods (NLME) to this challenge. Briefly, implementing new and published liver volume models based on any combination of age, body-weight, height, and body surface area allowed us to compare goodness of fits (GOF) between models and the data. We quickly found that a body surface area (BSA) model provided some of the best goodness of fit estimates. Using the Phoenix covariate search step-wise function, we took this BSA based model and assessed the effects of including and excluding a number of covariates and combinations thereof (eg, age, ethnicity, height, weight, etc.) on GOF. After incorporating covariates, we re-evaluated GOF estimates for models. Finding a model that accounted for individuals of Japanese ethnicity and age that also gave good GOF estimates, we re-sampled (bootstrapped) the data-set and also compared the model to the original Johnson dataset  to ensure a robust model performance. We found the BSA based model that accounted for Japanese ethnicity and age to be most appropriate for accounting for an individual’s liver volume.
Dr. Masoud Jamei (Vice President of Simcyp R&D) a senior author on the paper commented “These days, lots of data (physiological, biological, epidemiological, etc.) are generated and collated. An NLME approach is often needed to model these data. There are limited tools for such analysis, and Phoenix is a user-friendly, powerful platform for such analysis. This approach also reinforced and validated a model that Trevor had previously developed on predicting liver size.
Dr. Trevor Johnson (Deputy Head of Systems Pharmacology) the senior author on the paper commented “The expert modeling of physiological parameters with relevant covariates such as age, sex, study methodology and ethnicity is a key part of our strategy to ensure the most robust parameters are included in the simulation platform. This project built on my previous work to model liver size from newborns to adults and included more data from different ethnic groups and older subjects. I was pleased that we were able to leverage expertise across Certara for the project. This collaboration certainly improved the outcome. The result was a new model to predict individual liver volumes and also publication in Biopharmaceutics and Drug Disposition .”
Dr. Bernd Wendt (Director, Training and Support, Certara) a co-author of the manuscript commented: ”We have been exposed to extraordinarily rich liver volume data assembled by Ben Small. This presented a unique opportunity to apply non-linear-mixed-effects methods (NLME) to this data-set. Finally, we uncovered new covariate relationships (e.g. Japanese ethnicity and age) and confirmed Trevor’s choice for BSA as an independent variable for a model that appears to be the best model to predict liver volumes in silico.”
In summary, we’ve provided an improved way of determining liver volume in individuals. This will contribute to and support simulations from the Simcyp PBPK Simulator, allowing improved predictions of dose individualization to recommend a safe and efficacious dose for the patient.
 Johnson TN, Tucker GT, Tanner MS and Rostami-Hodjegan A. Changes in liver volume from birth to adulthood: a meta-analysis. Liver transplantation: official publication of the American Association for the Study of Liver Diseases and the International Liver Transplantation Society 2005; 11: 1481-93. DOI: 10.1002/lt.20519
 Small BG, Wendt B, Jamei M and Johnson TN. Prediction of Liver Volume – a population-based approach to meta-analysis of pediatric, adult and geriatric populations – an update. Biopharmaceutics & Drug Disposition 2017: n/a-n/a. DOI: 10.1002/bdd.2063
To learn more about how the Simcyp Simulator can be used to predict drug exposure in patients with renal or hepatic impairment, read this case study.