First Meeting, 15 June 1992
The PAGE92 meeting (at that time called the First Meeting of the NONMEM European Users Group) was held in Basel, Switzerland, and was sponsored by Sandoz Pharma Drug Safety Assessment.
Basel river front (Photograph: Brigitte Tranchand)
List of Participants
Organising and Scientific Committee:
P. Girard and B. Tranchand (AP2POP)
J.L. Steimer and J. van Bree (Sandoz Pharma, Basel)
The following report was compiled by M.E. Ebelin, J.L. Steimer and J. Van Bree
The meeting has regrouped both experienced and young scientists with strong interest for the "population approach". A fraction were NONMEM users (of different expertise level) ; other people had no experience with the software yet. The number of participants was larger than initially expected : 63 participants from all over Europe including France, Germany, Scandinavia, Belgium, Spain,… There were about fifteen scientists (primarily pharmacokineticists) from the pharmaceutical industry in Basle. Within the audience were specialists of NONMEM (S. VOZEH, A. THOMSON) and of other population analysis methodologies like Gibbs sampling (A. RACINE POON). The objective of the meeting and probably the expectation of most participants was to exchange information about NONMEN methodology, software, hardware and applications. In addition to presentation of results, most of the communications have also emphasized methodological, practical and technical problems concerning the use of NONMEM (installation, running, analysis and validation of the results), and also raised issues in the application of the population approach to kinetic data analysis.
The meeting in SANDOZ ended up with a discussion about the opportunity for creation and the need for organization of a NONMEM European Users Group.
Overall, there was a strong pressure towards future meetings of this kind. It was decided that a next meeting will take place in June 93, with the goal to establish it once a year. The exact location of the next meeting was not agreed upon yet (Rhône-Poulenc Paris, Hoffmann-La Roche Basle, or again Sandoz Basle?). It was also decided that the name of the group should be associated to a broader area than NONMEM since the objective of these meetings are to discuss general aspects of the population approach. The logo "PAGE" for "Population Approach Group in Europe" has been chosen.
The conference was then followed by a demonstration of the program PPM (an interface between RS/1 and NONMEM) at Hoffmann-La Roche (H. WELKER) which has regrouped about fifteen people. One problem raised by the participants was the lack of availability of RS/1 in their universities or companies.
Comments On Specific Presentations.
Participants were welcomed by Dr Eckert (Head of Drug Safety Assessment within Sandoz Pharma). His short prensentation showed the awareness of Sandoz management of the ongoing interest for the population approach within regulatory authorities and the potential consequences for industry.
The meeting has begun with presentation of features likely to improve or facilitate the use of the software NONMEM :
- The recent developments of the software with, among other things :
First-Order Conditional Estimate method, subject specific parameter estimates, mixture of models for statistical distribution, different pharmacokinetic models for different individuals (report by JL STEIMER).
- PPM : an interface with RS/1 intended to manage the input and outputs of NONMEM (H. WELKER). The handling of the outputs (plots, smoothing,…) is the most interesting improvement that PPM offers. Each group seems to have its own solution in this respect. One participant has mentionned his own development of a SAS interface (including a change in the NONMEM source code!). Another reported the development of a data input tool on a macintosh.
- The next presentation was devoted to the installation of NONMEM under UNIX on a workstation, which revealed problem-free (P. GIRARD). One participant indicated to have detected mathematical problems with the last version (5.4) of VMS when he has installed NONMEM on his VAXstation.
Applications of NONMEM.
During the oral communications thereafter, the following practical and methodological issues were addressed :
- Choice of a compartmental pharmacokinetic model for population analysis of "extensive" and "screening" data after intravenous infusion (ME EBELIN).
- Dosage regimen of an anticancer agent and optimal (population) sampling strategy for Phase II trials with respect to clinical considerations and limited collection of plasma samples per individual, based on Phase I data (R. BRUNO).
- The dramatic consequence of the lack of control in data collection on the quality of the data base. In a multi-center study with mexiletine, about one-half of the plasma samples had to be discarded for population analysis. In this particular case, a linear regression of concentration vs covariates (without consideration for time of sampling within the dosing interval) revealed influential covariates, later detected by NONMEM analysis also (P. BURTIN)
- Methodological difficulties regarding the construction of the full population model including patient related covariates (ascending or descending strategy for inclusion/deletion of covariates ?). The clinical relevance of some empirical relationships (e.g. the power function Cl= θ1*GFRθ2) was questioned (how is θ2 interpretable in a physiological context?) (P. BURTIN).
- The capacity of NONMEM to handle complex dosing administration patterns, e.g, as they may arise in treatment/anaesthesia with computer-driven infusion pumps. In this example, NONMEM had seemingly be used as a tool for analysis of individual studies in small groups of volunteers with extensive blood sampling. For reasons that did not appear clearly (perhaps imbalance in the covariate data?), no global meta-analysis of all data had been carried out. (H. IHMSEN)
- Validation (of results) is a very important step of a population analysis.The results of a NONMEM analysis of vancomycin data from a group of paediatric patients of Glasgow (UK) were evaluated prospectively with data from children from Barcelona (Spain). The Scottish population analysis had shown a significant relation of clearance with the covariate PCA (postconceptional age) in the final model. The validation analysis of data from Barcelona babies having different demographic characteristics proved that prediction of plasma levels was adequate. In fact, it was even better without inclusion of PCA. The difference in both data sets regarding the values of PCA could be one reason (A. THOMSON).
- NONMEM analysis of data restricted to "peak/trough" sampling seems likely to produce high (possibly artefactual) correlations between Cl and V when the population covariance Omega is chosen non-diagonal. This was shown with gentamicin data. (W. WEBER).
- The use of patient population data (both parent and metabolite) to deal with issues which usually address extensive controlled volunteer studies, e.g. dose-dependency. Plasma level data from 599 (!) patients allowed to investigate the pharmacokinetics and linearity of risperidone in a 1-16 mg range. Such a study was unacceptable in volunteers due to limited tolerability of the compound. (A van PEER)
The presentation by Dr. Racine on Gibbs Sampling was at a high technical (statistical) level, challenging for most of the audience. Dr. Racine also brought many novel ideas and an original approach on mixed-effect modeling with an example raising very tough issues, the analysis of truncated categorical data arising from a trial with a pain-relieving compound. Such a problem cannot be handled routinely by NONMEM Version III, by the way…
NONMEM Version IV might allow to address it.
The meeting was a success, with good presentations, nice data, meaningful approaches and extensive discussion by the audience. It seems that the concepts, ideas and methods associated with the population approach have gained increased audience within the last few years, as well in industry as in academic clinical pharmacology groups. Impressive data sets showing long-term commitment of some groups to this idea were shown also. The field is expanding and moving fast.