Page 72 - 66_04

ARMIN WOLF ANAL REAL ACAD. FARM.

Statistics

A two-way analysis of variance was performed (group and animal;
each of them were regarded as qualitative). If the effect of the animal
number was not significant it was omitted. A quantile plot was used to
visually judge the normality of the residuals. If the residuals were not
normally distributed, we tried to achieve (approximate) normality by
transforming the response or by omission of outliners.

A multiple comparison method was applied using the three
methods of Turkey (Hayter 1989), Sidak (Sidak 1967) and Dunnett
(Dunnett 1964). The Dunnett test compares every treated group with the
control group, while the other two methods can be used to compare each
group with each other group.

The multiple comparison method delivers an estimation of the
difference in the response expected between the two groups compared,
the standard error of the response, and a lower and an upper confidence
limit for the difference. If the two limits do not include zero, the
difference is significantly different from zero, on the level of 5 %. By
repeating the method for 1 % and 0.1 %, we could analyze how big the
significance was.

The S-Plus software (version 5) was used for the computations;
the three methods of Turkey (Hayter 1989), Sidak (Sidak 1967) and
Dunnett (Dunnett 1964) were used adaptively, i.e. in every case, the most
sensitive method was used.

RESULTS

Cytotoxicity

Primary hepatocytes were incubated with CsA at concentrations of
0, 10, 25 and 50 µM for 4 and 20 hours. LDH release was determined as a
parameter of cytotoxicity. CsA treatment resulted in a dose- and time-
dependent induction of LDH release. While CsA was not cytotoxic at any
concentration after 4 hours, CsA was statistically significantly cytotoxic
at 25 and 50 µM after 20 hours (Fig.1)

10