18.104.22.168. LD50 determination
The study design is clearly defined in EPPO (2010b), OECD (1998a), or CEB (2011) guidelines. Well known statistical regression analysis from BLISS and LITCHFIELD and WILCOXON (Siegel and Castellan, 1988) and more recent publications lead to perform regressions with dose transformation as logarithm and probit or logit transformation on the response rate.
Dose-response curves at each recommended observation time should be plotted and the slopes of the curves and the median lethal doses (LD50) with 95% confidence limits are calculated (Abbott, 1925). The LD50 is determined by the equation of the linear regression. Raw data provide dispersed values which need to be corrected by the control (see section 8.4.1.), then the 50% mortality is calculated with the equation type y = ax + b.
In some cases a lack of fit can be observed due to no dose related response. It depends on S shape component or an asymptotic data trend (Winer et at, 1991). Non-linear standard or modified GOMPERTZ regression may give a better fit on experimental data.
Generally for the LDs calculation, different statistical softwares (both commercial and open source) are used. The computer-aided procedure performs the calculations automatically, thus helping to prevent errors.