Generating Virtual Patients by Multivariate and Discrete Re-Sampling Techniques

Abstract

PurposeClinical Trial Simulations (CTS) are a valuable tool for decision-making during drug development. However, to obtain realistic simulation scenarios, the patients included in the CTS must be representative of the target population. This is particularly important when covariate effects exist that may affect the outcome of a trial. The objective of our investigation was to evaluate and compare CTS results using re-sampling from a population pool and multivariate distributions to simulate patient covariates.MethodsCOPD was selected as paradigm disease for the purposes of our analysis, FEV1 was used as response measure and the effects of a hypothetical intervention were evaluated in different populations in order to assess the predictive performance of the two methods.ResultsOur results show that the multivariate distribution method produces realistic covariate correlations, comparable to the real population. Moreover, it allows simulation of patient characteristics beyond the limits of inclusion and exclusion criteria in historical protocols.ConclusionBoth methods, discrete resampling and multivariate distribution generate realistic pools of virtual patients. However the use of a multivariate distribution enable more flexible simulation scenarios since it is not necessarily bound to the existing covariate combinations in the available clinical data sets.Electronic supplementary materialThe online version of this article (doi:10.1007/s11095-015-1699-x) contains supplementary material, which is available to authorized users.

Evaluating optimal therapy robustness by virtual expansion of a sample population, with a case study in cancer immunotherapy

“…To explore robustness, the data obtained from the bootstrap replicates were used to construct virtual populations, similar to the nonparametric approach described in ref. 29. In particular, herein, 1,000 virtual populations were created by pseudorandomly picking the six fit parameters (r , β, cA, cT , c kill , and χD ) from their posterior distributions approximated using the bootstrap replicates and the hierarchical fitting process.…”

“…Others have given consideration to further preserve the covariance structure among all variables; see, for instance, refs. 16,23,25,26,29, and 30. The second constraint imposed is that only virtual populations whose parameter values are all within their respective 95% credible intervals were considered; this can be thought of as an "inclusion-exclusion" criterion that refines the virtual population pool to be statistically similar to the experimental population (29), including mirroring its heterogeneity.…”

“…16,23,25,26,29, and 30. The second constraint imposed is that only virtual populations whose parameter values are all within their respective 95% credible intervals were considered; this can be thought of as an "inclusion-exclusion" criterion that refines the virtual population pool to be statistically similar to the experimental population (29), including mirroring its heterogeneity. That said, this approach does constrain the virtual populations to statistically resemble the experimental data, which, because of the small sample size in our experiments, runs the risk of not resembling the population data (29).…”

“…Virtual populations have been used by others, particularly in the field of quantitative systems pharmacology (16,(22)(23)(24)(25)(26)(27)(28)(29)(30). Overwhelmingly (see, for instance, refs.…”

“…These virtual population methods require dozens to thousands of samples to create virtual populations that are representative of the target population (23,26,(28)(29)(30). On the other hand, the VEPART procedure proposed herein works with very small amounts of data, and "expands" the available data in an effort to model the heterogeneity expected in the full population.…”