Predictions and extrapolation

What are called physiologically based pharmacokinetic (PBPK) and pharmacodynamic (PBPD) models are more mechanistically complex and often include more compartments, more parameters, and more detailed expressions of rates and fluxes and contain more mechanistic representation. This type of model is reviewed in more detail in Section 22.5. Here, we merely classify such models and note several characteristics. PBPK models have more parameters, are more mechanistic, can exploit a wider range of data, often represent the whole body, and can be used both to describe and interpolate as well as to predict and extrapolate. Complexity of such models ranges from moderate to high. They typically contain 10 or more compartments, and can range to hundreds. The increase in the number of flux relationships between compartments and the related parameters is often more than proportional to compartment count. 

If, through a proper landscape analysis, a realistic combination of multiple stressors is identified, food-web models may be used to predict and extrapolate their ecological effects to relevant ecosystems of the landscape unit of concern. An overview of models that can be used for the integrated assessment of eutrophication and organic contaminants in aquatic ecosystems is provided by Koelmans et al. (2001). Examples of aquatic food-web models that can be used or adapted to predict effects of multiple stressors are IFEM (Bartell et al. 1988), AQUATOX (Park 1999), and C-COSM (Traas 2004). 

To make predictions and extrapolations based on the mathematical expressions and... 

A brief review of the limited number of in situ measurements is then presented. This is followed by a review of the available correlations proposed for estimating these kinds of MTCs in Section 12.4. Section 12.5 is concerned with the subject of classical molecular diffusion in porous media at steady state. The presentation includes a brief description of the upper sediment layers, measurement techniques, laboratory measurement data of effective diffusion coefficients, and models for prediction and extrapolation. A guide appears in Section 12.6 to steer users to suggested procedures for estimating these two types of MTCs. The chapter ends with some example problems and their solutions in Section 12.7. 

There are many large molecules whose mteractions we have little hope of detemiining in detail. In these cases we turn to models based on simple mathematical representations of the interaction potential with empirically detemiined parameters. Even for smaller molecules where a detailed interaction potential has been obtained by an ab initio calculation or by a numerical inversion of experimental data, it is usefid to fit the calculated points to a functional fomi which then serves as a computationally inexpensive interpolation and extrapolation tool for use in fiirtlier work such as molecular simulation studies or predictive scattering computations. There are a very large number of such models in use, and only a small sample is considered here. The most frequently used simple spherical models are described in section Al.5.5.1 and some of the more common elaborate models are discussed in section A 1.5.5.2. section Al.5.5.3 and section Al.5.5.4. 

The sohd line in Figure 3 represents the potential vs the measured (or the appHed) current density. Measured or appHed current is the current actually measured in an external circuit ie, the amount of external current that must be appHed to the electrode in order to move the potential to each desired point. The corrosion potential and corrosion current density can also be deterrnined from the potential vs measured current behavior, which is referred to as polarization curve rather than an Evans diagram, by extrapolation of either or both the anodic or cathodic portion of the curve. This latter procedure does not require specific knowledge of the equiHbrium potentials, exchange current densities, and Tafel slope values of the specific reactions involved. Thus Evans diagrams, constmcted from information contained in the Hterature, and polarization curves, generated by experimentation, can be used to predict and analyze uniform and other forms of corrosion. Further treatment of these subjects can be found elsewhere (1—3,6,18). 

Herbst et al. [International J. Mineral Proce.ssing, 22, 273-296 (1988)] describe the software modules in an optimum controller for a grinding circuit. The process model can be an empirical model as some authors have used. A phenomenological model can give more accurate predictions, and can be extrapolated, for example from pilot-to full-scale apphcation, if scale-up rules are known. Normally the model is a variant of the popiilation balance equations given in the previous section. 

Consider first the boiling points of HI, HBr, HQ, and HF. The last, hydrogen fluoride, is far out of line, boiling at 19.9°C instead of below —95°C as would be predicted by extrapolation from the other three. There is an even larger discordancy between the boiling point of HjO and the value we would predict from the trend suggested by HjTe, HjSe, and H2S. 

With this focus on CYP and fiver metabolism, most companies have established high throughput assays to measure compound stability in the presence of human (or preclinical species) fiver microsomes [49]. Disappearance of starting compound from an incubation with microsomes is monitored. Measurement at a single time point enables a rank-ordering of compounds for stability based on percent of parent compound remaining acquisition of data at multiple time points allows determination of half-life, intrinsic clearance, and extrapolation to a predicted in vivo clearance [50]. 

The techniques used for prediction are also useful for the correlation, and extrapolation and interpolation, of experimental values. 

Accelerated testing depends critically on selecting a parameter whose effect on service life is so well understood that long lifetimes at low values of the parameter can be predicted from shorter lifetimes at higher values. The parameter may be the prime cause of degradation, such as in a stress-rupture test where longer lifetimes at lower loads are predicted by extrapolation from short lifetimes at higher loads. It can also be a secondary parameter, such as when temperature is increased to accelerate chemical attack while the principal factor, chemical concentration, is kept constant. This is because there is more confidence in the relation between rate of reaction and temperature than in the relation of rate of reaction to concentration. It is clearly essential that extrapolation rules from the test conditions to those of service are known and have been verified, such that they can be used with confidence. 

In systems where the liquid phase interaction between the solute and solvent is close to ideal, then Eq. 2 can be used successfully on it s own to fit and extrapolate solubility data with respect to temperature. The technique is valuable in an industrial setting, where time pressures are always present. Solubility data points are often available without any additional effort, from initial work on the process chemistiy. The relative volume of solvent that is required to dissolve a solute at the highest process temperature in the ciystallization is often known, together with the low temperature solubility by analysis of the filtrates. If these data points fit reasonably well to the ideal solubility equation then it can be used to extrapolate the data and predict the available crystallization yield and productivity. This quickly identifies if the process will be acceptable for long term manufacture, and if further solvent selection is necessary. 

Considering other families of similar compounds, the contributions given by Guillermet and Frisk (1992), Guillermet and Grimvall (1991) (cohesive and thermodynamic properties, atomic average volumes, etc. of nitrides, borides, etc. of transition metals) are other examples of systematic descriptions of selected groups of phases and of the use of special interpolation and extrapolation procedures to predict specific properties. 

We now turn briefly to the problem of peptide stability in the solid state [8] [88], First, we note that most - if not all - reactions discussed in the previous and subsequent sections can also occur in the solid state, although the kinetics and mechanisms of the reactions can be quite different from those observed in solution. Moisture content, the presence of excipients that act as catalysts, and surface phenomena are all factors whose roles are all-but-im-possible to predict. As a result, each formulation poses a new challenge to pharmaceutical scientists. As a rule, solution data cannot be used to predict the shelf-life of solid formulations, and extrapolating from one solid formulation to another can be misleading. 

---