NONMEM 7.6 is the latest release of ICON’s industry-standard nonlinear mixed-effects modeling software for population pharmacokinetic and pharmacodynamic (PK/PD) analysis . It is developed and licensed by ICON Early Clinical (versions up to VI were property of the Regents of the University of California) .

Key New Features in Version 7.6

Advanced Estimation & Bayesian Methods

  • NUTS (Hamiltonian No-U-Turn Sampling): A new Markov-Chain Monte Carlo Bayesian method alongside existing Gibbs sampling, with the ability to store random samples of individual parameters or any user-defined variable .

  • SAEM Enhancements: Stochastic Approximation Expectation-Maximization now allows storage of individual parameter samples or user-defined variables for deeper analysis .

  • FAST Algorithm: A Monte Carlo search algorithm introduced to improve FOCE estimation efficiency .

Delay Differential Equation (DDE) Solvers

  • ADVAN16: Uses RADAR5 by Guglielmi and Hairer (implicit 4th-order Runge-Kutta) for stiff delay equations .

  • ADVAN17: For stiff delay differential-algebraic equations .

  • Distributed delay convolution algorithms are also included for modeling complex delay systems .

Performance & Efficiency

  • Parallel Computing: A single problem can be distributed across multiple cores or computers for estimation, covariance assessment, simulation, nonparametric analysis, and posthoc evaluation, significantly reducing completion time .

  • Dynamic Memory Allocation: Improved to handle very large problems without recompiling the NONMEM program for unusually large models .

Optimal Design & Robust Analysis

  • Clinical Trial Design Evaluation: Includes robust design analysis and optimal design evaluation tools .

  • WFIMS Module: New user-defined module for adjusting objective function weights in robust design .

Bug Fixes & Stability

  • Fixed memory access conflicts in ADVAN5/7/8 during $TABLE steps .

  • Resolved parallelization conflicts with FPI (File Passing Interface) and importance sampling .

  • Fixed COVSIRparallelizationissuesandTABLE ETAS() output errors .

  • Addressed empty subject data handling in parallel computing .