|
|
 |
|
|
| What should be done, when the standard test produces variable results? |
|
 |
|
The standard test, on which all the new methods are measured against, has therefore a high %CV value*. How does this effect the ability to make comparisons between the alternative methods?
The image shown, is a purely mathematical game of numbers. What is compared, are mathematically generated redistributions of hypothetical results, of the two methods.
It is assumed, that both methods describe an effect in exactly the same manner; nevertheless both methods have a certain element of variability. In the left hand graphic a variability of 5%CV has been assumed for both examples, in the right hand one of 50%; with these guidelines the distribution of the measurement results has been simulated.
The diagrams show clearly, that the high variability of a reference method (which can be observed in the Draize Test ) greatly hinders good correlation with other methods.
This could also be a reason why, in a large validation study, no alternative methods have shown a sufficiently good enough correlation with the Draize test, to be able to replace it. (Balls et al. 1995).
A great number of in-vitro methods have been written about, which all have particular advantages and disadvantages, but manage to cover well a particular aspect relating to eye irritation. The following methods are the ones which are probably most commonly used.
-BCOP (bovine cornea opacity and permeability test)
-HET-CAM, CAMVA (hen‘s egg test - chorioallantoic membrane)
-Isolated enucleated rabbit / chicken eyes
-NRU (neutral red uptake) in cell cultures (main entry for corrosive substances and photo-toxicity)
*Results from the Draize Test follow a normal distribution, with high dispersion and, as a consequence, with a high standard deviation. From the standard deviations and the means the coefficient of variation is calculated (CV = Coefficient of Variation). Normal distributions with high dispersion and standard deviation result in high CV-values.
|
|
|
|
|
| References, links and documents |
|
|
|
|
|
|
| previous next |
|
|
