J. PKPD Online First

Our conclusion is that the Laplace approximation may be unreliable in certain circumstances.   The data analyst would be well advised to use adaptive Gaussian quadrature rather than the Laplace method to approximate the likelihood if the numbre of observations per subject is small.

the (PK/PD) analyses were further tested and reviewed by the FDA, who reached similar conclusions:-   The PK/PD results confirmed evidence of efficacy across the three studied doses, so that additional trials would not be needed for approval of the product.

  This paper is based on a lecture given as a memorial to my friend and colleague Lewis B. Sheiner (see acknowledgements)

    • See ACKNOWLEDGEMENTS.