Premium
INFLUENCE OF CONFOUNDING FACTORS ON DESIGNS FOR DOSE–EFFECT RELATIONSHIP ESTIMATES
Author(s) -
Girard Pascal,
LaporteSimitsidis Silvy,
Mismetti Patrick,
Decousus Hervé,
Boissel JeanPierre
Publication year - 1995
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.4780140918
Subject(s) - confounding , statistics , monotonic function , econometrics , pharmacodynamics , regression , mathematics , linear regression , medicine , computer science , pharmacokinetics , mathematical analysis
Three types of designs can be used to estimate the drug dose–effect relationship during phase II clinical trials: parallel‐dose designs (//); cross‐over designs (X), and dose‐escalation designs (↗). Despite the use of non‐linear mixed effect models, the potential influence of confounding factors on ↗ designs has not been previously fully elucidated; we undertook simulations to investigate this for all three experimental designs. We found that: (i) monotonic spontaneous evolution of the effect (EV) did not affect the maximum effect estimation ( E max ) and the dose giving 50 per cent of this ( ED 50 ); (ii) EV similar to a regression to the mean gave rise to biases for ↗ designs; (ii) the introduction of a pharmacodynamic carry‐over generates important biases and imprecision for ↗ designs, even when the carry‐over is adjusted for; (iv) the introduction of non‐responders resulted in bias and imprecision for both E max and ED 50 in all three designs.