Simulo: a new PK-PD-Disease model simulator
Ruben Faelens, Quentin Leirens, Philippe Jacqmin
SGS Exprimo NV
What is Simulo?
Simulo is a new PK-PD-Disease model simulator, providing a user-friendly interface available on the web. It was developed to be a shared and user-friendly platform, running simulations on a private dedicated computing cluster. It provides the ability to simulate and subsequently analyse clinical studies using public, published or custom-developed nonlinear mixed-effects models.
Why was it developed?
Many simulation software solutions are available for model and simulation experts within the pharmaceutical industry. However, they often require extensive and complex scripting, especially when implementing mixed effect models and designing clinical trials. This complexity and this large choice of tools do not facilitate communication between M&S consultants and in-house M&S experts. In addition, they may be difficult to install as well as validate and they need scientific clusters to run, which have to be administrated.
Simulo solves these problems by providing a user-friendly web interface. Complex scripting is replaced by a straight-forward interface where you can define treatments, dose adjustment, observations, inclusion/exclusion criteria… Models are shared between colleagues, which enables collaboration.
Ability to organize and implement complex trial designs in a user-friendly interface
Simulo is comprised of three major parts allowing to build and understand a model easily while conserving a very high flexibility thanks to its use of R code.
The drug model editor - Structural equations defining the core of the drug model can be introduced through algebraic, ordinary differential and/or delayed differential equations. Model parameters are computed from variability generated at different levels (population parameters, covariates, IIV & IOV). This variability can be sampled from non-correlated parametric distributions (e.g. constant, uniform, normal, lognormal, logit, logistic, Poisson, negative binomial), correlated parametric distributions and discrete distributions. It can even be bootstrapped from existing databases (sequentially or at random, with or without replacement, independently or jointly). Error models can be easily built in order to reproduce residual variability. Conditional events brings the ability to intervene during the simulated trial (e.g. dropout, adverse event, TTE).
The protocol editor - Enrolment, inclusion criteria, treatments (route & mode of administration, dose and dosing schedule), observations (type and schedule), lead-in phase or a complex protocol (e.g. parallel, cross-over Latin-square designs) can all be defined in a user-friendly interface. Simulo can be used to simulate complex dose adjustment schemes or special drug effects with ease.
The analysis editor - Analysis methods can automatically be applied on the in-memory results at the end of each replicate through customized R scripts. It allows you to only save final output graphs and tables.
Once these elements are defined, a clinical trial can be simulated. All element definitions are converted into R code and executed on a high performance cluster. The raw and/or analysis results can be downloaded.
On the top of that, Simulo has some tools valuable during the process of implementation. First, a live validation module checks constantly the implementation and shows problems. Secondly, for easily seeing how a drug model behaves, a live simulation was created. It simulates a limited number of subjects and displays the results graphically, allowing you to quickly verify how your model behaves in different situations. Finally, hypotheses for each population characteristic, model parameter and study design attributes can be configured in a scenario view.
Technically
Simulo is web-based, which means it doesn’t require a specific installation: a recent web browser is sufficient to connect. It was built by professional software engineers at Altran under the close supervision of domain experts at Exprimo and F. Hoffman-La Roche. The principle is a Java-based application running on an R backend. Simulations are done by first translating the study model to R code, and then efficiently executing this code. Moreover, Simulo is installed on Exprimo’s servers and therefore the maintenance, validation and version control are kept centralized.
Service
Simulo has already faced an extensive use within Exprimo for client projects and has been validated on more than 80 models. It is now offered as an integrated part of the consultancy service that Exprimo delivers to clients during a project. The sponsor has the opportunity to reproduce the simulations performed at Exprimo and/or to independently run advanced simulation scenarios that he would like to test. The simulated data can then be remotely visualized and analyzed through customized R scripts and/or can be downloaded as .csv files for further local processing.
We will provide a live demonstration but you are also invited to visit the website www.simulo.eu.
