Details
| Original language | English |
|---|---|
| Pages (from-to) | 97-103 |
| Number of pages | 7 |
| Journal | Alternatives to laboratory animals |
| Volume | 31 |
| Issue number | SUPPL. 1 |
| Publication status | Published - Nov 2003 |
Abstract
The analysis of dose-response relationships is an important objective in toxicology, and one in which both modelling and testing approaches are used. One particular question is whether a threshold exists at low doses. The concept of a pragmatic threshold is used, i.e. low doses with biologically unimportant effects are assumed to be threshold doses. "Biologically unimportant" means, in statistical terms, a lower effect than the effect of the negative control, or at least a just-tolerable margin δ higher than the effect of the negative control. Therefore, threshold doses can be tested in terms of a one-sided hypothesis of equivalence. A new approach is proposed, assuming, at the least, that the low dose is a threshold dose, and the highest dose is superior to the negative control. By analogy to the k-fold rule commonly used in mutagenicity studies, tests on ratio-to-control are used. The a priori definition of the threshold margin is inherently needed. A further approach proposes the analysis of dose-response relationships by means of order-restricted inference (the so-called trend test). A modification of a multiple-contrast test is used, in which only those contrasts are included that are sensitive for no effects at low doses. A further modification treats the complicated, but real, problem of simultaneous existence of a threshold, a monotonic increase, and a downturn effect at high dose(s). A parametric procedure is considered, together with an extension for proportions. The important problem of a priori sample size definition is discussed. The approaches are demonstrated by means of examples based on real data.
Keywords
- Contrast test, Dose-response analysis, In vitro toxicology, Intersection-union test, Test on equivalence, Test on ratio-to-control, Threshold dose
ASJC Scopus subject areas
- Medicine(all)
- General Medicine
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: Alternatives to laboratory animals, Vol. 31, No. SUPPL. 1, 11.2003, p. 97-103.
Research output: Contribution to journal › Review article › Research › peer review
}
TY - JOUR
T1 - Dose-response and thresholds in mutagenicity studies
T2 - A statistical testing approach
AU - Hothorn, Ludwig A.
AU - Bretz, Frank
PY - 2003/11
Y1 - 2003/11
N2 - The analysis of dose-response relationships is an important objective in toxicology, and one in which both modelling and testing approaches are used. One particular question is whether a threshold exists at low doses. The concept of a pragmatic threshold is used, i.e. low doses with biologically unimportant effects are assumed to be threshold doses. "Biologically unimportant" means, in statistical terms, a lower effect than the effect of the negative control, or at least a just-tolerable margin δ higher than the effect of the negative control. Therefore, threshold doses can be tested in terms of a one-sided hypothesis of equivalence. A new approach is proposed, assuming, at the least, that the low dose is a threshold dose, and the highest dose is superior to the negative control. By analogy to the k-fold rule commonly used in mutagenicity studies, tests on ratio-to-control are used. The a priori definition of the threshold margin is inherently needed. A further approach proposes the analysis of dose-response relationships by means of order-restricted inference (the so-called trend test). A modification of a multiple-contrast test is used, in which only those contrasts are included that are sensitive for no effects at low doses. A further modification treats the complicated, but real, problem of simultaneous existence of a threshold, a monotonic increase, and a downturn effect at high dose(s). A parametric procedure is considered, together with an extension for proportions. The important problem of a priori sample size definition is discussed. The approaches are demonstrated by means of examples based on real data.
AB - The analysis of dose-response relationships is an important objective in toxicology, and one in which both modelling and testing approaches are used. One particular question is whether a threshold exists at low doses. The concept of a pragmatic threshold is used, i.e. low doses with biologically unimportant effects are assumed to be threshold doses. "Biologically unimportant" means, in statistical terms, a lower effect than the effect of the negative control, or at least a just-tolerable margin δ higher than the effect of the negative control. Therefore, threshold doses can be tested in terms of a one-sided hypothesis of equivalence. A new approach is proposed, assuming, at the least, that the low dose is a threshold dose, and the highest dose is superior to the negative control. By analogy to the k-fold rule commonly used in mutagenicity studies, tests on ratio-to-control are used. The a priori definition of the threshold margin is inherently needed. A further approach proposes the analysis of dose-response relationships by means of order-restricted inference (the so-called trend test). A modification of a multiple-contrast test is used, in which only those contrasts are included that are sensitive for no effects at low doses. A further modification treats the complicated, but real, problem of simultaneous existence of a threshold, a monotonic increase, and a downturn effect at high dose(s). A parametric procedure is considered, together with an extension for proportions. The important problem of a priori sample size definition is discussed. The approaches are demonstrated by means of examples based on real data.
KW - Contrast test
KW - Dose-response analysis
KW - In vitro toxicology
KW - Intersection-union test
KW - Test on equivalence
KW - Test on ratio-to-control
KW - Threshold dose
UR - http://www.scopus.com/inward/record.url?scp=0041695386&partnerID=8YFLogxK
U2 - 10.1177/026119290303101s07
DO - 10.1177/026119290303101s07
M3 - Review article
C2 - 15595904
AN - SCOPUS:0041695386
VL - 31
SP - 97
EP - 103
JO - Alternatives to laboratory animals
JF - Alternatives to laboratory animals
SN - 0261-1929
IS - SUPPL. 1
ER -