Validation Benchmark

Automated NCA computation validation using the Theophylline Subject 1 standard dataset. All tests compare computed parameters against published reference values from established pharmacokinetic textbooks.

Methodology

This benchmark validates the PK-Swift computation engine against established pharmacokinetic reference values. The validation suite includes 18 tests covering observed parameters, AUC metrics, terminal phase analysis, MRT, and dose-dependent PK parameters:

  • Dataset: Theophylline Subject 1 — an 11-point oral PK profile (Dose = 319.992 mg), commonly used in pharmacokinetic textbooks and software validation.
  • Integration Method: Linear-up Log-down trapezoidal rule. Ascending segments (Cᵢ₊₁ ≥ Cᵢ) use the linear trapezoidal formula; descending segments (Cᵢ₊₁ < Cᵢ, both > 0) use the log-trapezoidal formula. This is the most widely implemented method in commercial NCA software (Gabrielsson & Weiner, 2016).
  • AUMC Calculation: For log-linear segments, AUMC is computed via the exact integral: AUMC_i = Δt·(t₁C₁ − t₂C₂)/ln(C₁/C₂) + Δt·(C₁ − C₂)/[ln(C₁/C₂)]². Linear segments use the standard trapezoidal formula for the first-moment curve.
  • λz Selection: Automated "Best Fit" algorithm: considers terminal subsets (≥ 3 points, post-Tmax, C > 0), requires negative slope, selects subset with highest adjusted R² (ties broken by largest number of points). This approach follows the convention used in Phoenix WinNonlin and other NCA software.
  • Scoring: Each test parameter receives an accuracy score: Accuracy = max(0, 1 − |Δ|/|Expected|) × 100%. The overall benchmark score is the geometric mean of individual accuracies.
  • Tolerance-based pass/fail: Each parameter has an absolute tolerance threshold. Directly observed values (Cmax, Tmax, Clast, Tlast) have near-zero tolerance. Computed parameters have tolerances reflecting expected inter-software variability and propagated numerical differences.

References

  1. Gabrielsson J, Weiner D. Pharmacokinetic & Pharmacodynamic Data Analysis: Concepts and Applications. 5th ed. Stockholm: Swedish Pharmaceutical Press; 2016. — Primary reference for trapezoidal integration (§10.1–10.2), AUMC calculation, and MRT derivation.
  2. Rowland M, Tozer TN. Clinical Pharmacokinetics and Pharmacodynamics: Concepts and Applications. 4th ed. Philadelphia: Lippincott Williams & Wilkins; 2011. — Standard NCA parameter definitions (CL/F, Vz/F, t½, AUC).
  3. Gibaldi M, Perrier D. Pharmacokinetics. 2nd ed. New York: Marcel Dekker; 1982. — Classical derivations of trapezoidal and log-trapezoidal rules for AUC and AUMC.
  4. Wagner JG. Fundamentals of Clinical Pharmacokinetics. Hamilton, IL: Drug Intelligence Publications; 1975. — Foundations of terminal elimination phase analysis and regression methodology.
  5. Zhang Y, Huo M, Zhou J, Xie S. PKSolver: An add-in program for pharmacokinetic and pharmacodynamic data analysis in Microsoft Excel. Comput Methods Programs Biomed. 2010;99(3):306-314. — Independent NCA implementation for cross-validation reference.