4.1. Study design
When testing the effect of plant protection products or insect growth regulators on immature honey bees, the design of the study (i.e. the number of replicates and larvae) is crucial. First, clarity regarding the minimum number of replicates is needed. A replicate is a repetition within a treatment group, commonly made up of a series of microtiter plates or petri dishes, where each larva can be regarded as a single statistical event. We recommend a minimum of three replicates with a minimum of 30 larvae per replicate. In toxicity testing, a replicate typically consists of a control plate without solvent, one control with solvent (if necessary) and several plates with the doses or concentrations of pesticides to be tested. As a reference, one treatment with a substance of known toxicity (e.g. dimethoate, Aupinel et al., 2007a, 2007b) must be conducted (see the BEEBOOK paper on toxicology methods (Medrzycki et al., 2013)).
Decisions about study design should carefully consider the aim of the investigation. Depending on the study design, it is possible to study the influence of the genetic background of the bees (Fig. 3a) or of the season (Fig. 3b) on the susceptibility of larvae to a certain substance or pathogen. Mixing larvae from different colonies (Fig. 3c) and testing them even over the entire season (Fig. 3d) will reveal the effect of a substance / pathogen that is not influenced by the genetic background of the bees or by the season. To fully understand the effect of a substance / pathogen, ideally all four approaches should be carried out. If the variance due to an effect of the genetic background needs to be reduced or controlled for, the usage of haploid drone larvae instead of diploid worker offspring of different patrilines may be desirable. This approach however requires that the phenotypes of the drones resemble those of the workers or that differences have been investigated. If usage of diploid workers is inevitable the genetic variance of the study population can be reduced by taking the worker offspring of a single drone inseminated queen. In any case, the minimum required sample size must be calculated and for this we recommend consulting a statistician.
Fig. 3a-d. Examples of four different study designs to conduct three replicates of an experiment with different levels of genetic and seasonal variability. X indicates date of grafting and origin of larvae used for experiments.