If I wanted to create a new characterization factor for use in USEtox, I would need to know the EC50 and/or ED50 (among other environmental mechanism data). How many EC50 (or ED50) do I need in order to have a robust, solid average number? USEtox says to get as many as you can, but how many is many? Thank you.
The standard error of the mean log EC50 is inversely proportional to the square root of n and proportional to t(n-1). This makes a number of data points of n=2 certainly too small for a serious consideration (for calculating a mean), whereas n >20 would be more more appropriate.
However, useually one has no choice, but to accept however small n is, because often the number of available data is very limited. The key is to estimate the confidence interval of the characterization factor and to report it along with the characterization factor to give an impression of the underlying data quality.
Thank you, USETox Team, for your reply to my question about the number of EC50 or ED50 needed to create a new CF. I have a follow-on question for the Forum:
Ho many different species should be represented by the EC50 or ED50 value used in the new CF for a new chemical? For example, is it better to have EC50s for multiple fish, amphibian, and bird species or 10 EC50s from a single species?
Thank you.
Thank you for your follow-up question.
Species selection for calculation of EC50s should in general aim for the highest physiological variability, for as many species as possible, representing as many taxonomic groups as possible. The sources used in USEtox require at least three EC50s from three different phyla to reflect the variability of the physiology and ensure a minimum diversity of biological responses. Further details are provided in Henderson et al. (2011).