1 Department of Biology, Faculty of Science, SDU2 Department of Biology, Faculty of Science, SDU
Since 2012, European experts work towards the development and validation of an OECD test guideline for mollusc reproductive toxicity with the freshwater gastropod Lymnaea stagnalis. A ring-test involving six laboratories allowed studying reproducibility of results, based on survival and reproduction data of snails monitored over 56 days exposure to cadmium. A classical statistical analysis of data was initially conducted by hypothesis tests and fit of parametric concentrationresponse models. However, as mortality occurred in exposed snails, these analyses require to be refined, particularly in avoiding bias that exists when the number of clutches/eggs is analysed without accounting for mortality, or when replicates where mortality occurred are excluded; in the latter case, a number of organisms are discarded and valuable data can be lost. In this context, the purpose of our statistical study was twofold. First, we refined the statistical analyses of reproduction data accounting for mortality all along the test period. The variable “number of clutches/eggs produced per individual-day” was used for EC x modelling, as classically done in epidemiology in order to account for the time-contribution of each individual to the measured response. Furthermore, the combination of a Gamma-Poisson stochastic part with a Weibull concentration-response model allowed accounting for the inter-replicate variability. Second, we checked for the possibility of optimizing the initial experimental design through the reduction of exposure duration and/or number of replicates. Based on the six datasets, we show that using the ‘per individual-day’ unit in ecotoxicology avoids the exclusion of data (as a consequence the ECx could be not estimated from remaining data) and ensures an unbiased reproduction data analysis when mortality occurs in exposed animals. We also show that the experimental design may be optimized, depending on what should be prioritized. Even if further studies would be necessary with other kinds of compounds, we illustrate the fact that, in the case of cadmium, and if 6 replicates are kept, a 35-day exposure duration would be sufficient to characterize the toxicity. In the same way, 3-4 replicates appear sufficient if the exposure duration stays at 56 days. However, before the reproduction test with L. stagnalis can be standardized, other works are necessary to furtherrefine the experimental design, e.g., by regarding exposure duration and replication simultaneously.