Projektportal

Startsida Om denna webbplats

Logga in

Using Full Random Effects Models in different software

Terrmin

HT24

Beskrivning

Full random effects modelling (FREM) is an innovative covariate modelling approach. It is different from traditional approaches in that it treats covariates as observations and estimates the impact of them via their covariance to the structural parameters of the model. This means that FREM is insensitive to correlations between covariates and that it implicitly handles missing covariate values. It has been demonstrated that it generates unbiased parameter estimates with as much as 90% missing covariate under most missing data mechanisms. Since it includes the covariates on all parameters FREM is also robust to omission bias, which can be a problem for alternative covariate modelling approaches.

FREM models are mixed effects models and require that the estimation software supports the estimation of multiple response variables simultaneously and that the residual error term for some of the response variables can be fixed to a small value. So far, FREM has mainly been used with NONMEM.

Due to its nature, FREM models require both pre- and post-processing to be estimable and interpretable. For example, since the covariates are treated as observations it is necessary to make sure the data set structure reflects this. Once a FREM model has been fit to the data, it has to be converted to a regular fixed effects model (FFEM) for the generation of goodness of fit diagnostics and simulations. The pre-processing is typically supported by Perl-speaks-NONMEM (PsN) and the post-processing by the R package PMXFrem.

The aim of this project is to support a wider use of FREM by investigating the possibilities of implementing FREM models in other softwares than NONMEM, specifically Monolix, Pumas and nlmixr.

The primary goal is to investigate the possibility to implement the same FREM model in all the softwares and compare the results. A secondary and optional goal is to develop software specific pre-processing tools and implement support for post-processing of the output from the softwares in PMXFrem.

The work will be carried out in Uppsala, at the company Pharmetheus in an international environment. The working language may be be English.

Does this project sounds like something for you? Apply here or contact the supervisor (see below) or Maria Kjellsson (maria.kjellsson@farmaci.uu.se) if you have additional questions. We look forward to your application!

Huvudområde

Farmaceutisk vetenskap

Ämne

Farmakometri

Typ

Beräkningsstudie

Företag

Pharmetheus

Ort/Plats

Uppsala

Handledarens namn

Niclas Jonsson

Handledarens e-post

niclas.jonsson@pharmetheus.com

Institution

Institutionen för farmaci

Program

Masterprogram i läkemedelsmodellering

Kurs

Degree Project in Pharmaceutical Modeling within Pharmacometrics 45 c - 3FB029

Omfattning/hp

45hp