PDF of presentation

Unambiguously defining 'innovation' in modelling and simulation of biomedical systems, especially pharmacokinetic and pharmacodynamic systems, has been a challenge since the beginning of the field. Technology that appears new in one context is not necessarily new in another, and many of the methods that are routinely used today in practical clinical data analysis and simulation are at least two decades old. No truly novel methods have become broadly available during this time, and those that have, like full Bayesian approaches, have not so far lived up to their initial promise.

The 'innovation challenge' is compounded by several factors, here cited in no particular order: the chasm between the undeniable complexity of the technology and the average sophistication of the community of users, for which methodological underpinnings are often of secondary importance, and frequently time or clinical impact is of the essence; the lag between technology availability through ready-to-use software and the establishment of broad, rigorous educational approaches; and lastly, the lack of proportionally large numbers of savvy practitioners. It can be perhaps said that the last two are tightly related, since biomedical funding trends in the last two decades show a strong emphasis on molecular biology, and have indirectly accelerated the disappearance of traditional quantitative disciplines at the clinical interface.

We have thus a contradiction: on one hand, clinical modelling and simulation is a proven technology, widely used in the private sector and of increasing impact, while on the other hand it is ignored by a growing number of educational institutions and it lies on the fringe of governmental biomedical funding trends. For a vibrant and ultimately very useful discipline, this background amounts to a recipe for long-term stagnation.

This talk will present some of the challenges and opportunities awaiting biomodelling and simulation, and especially the 'population approach', as it enters the 21st century and a new generation of scientists is preparing to take its helm. It will be argued, with examples, that innovation partly lies in the aggressive application of existing approaches to new problems in many areas of biomedicine, for example through promotion of the broad acceptance by the biomedical community of 'model-based biomarkers' and model-based hypothesis generation. While this is reasonable, it likely will be insufficient by itself to generate technological quantum leaps, unless unsolved problems proposed by other disciplines motivate non-trivial methodological advances. Truly novel methodological areas that show potential promise for non-incremental change will be also discussed, among them unsupervised model selection, ubiquitous computing, stochastic models and Bayesian model averaging.