Sources of variation within and between litters in developmental preclinical models, and how to deal with them

Maternal effects are the most important aspect of the early environment and influence the physiology, development and behaviour of offspring. Maternal effects result in both between-litter variation (VarB), where litters reared by different dams exhibit marked differences in outcomes, and within-litter variation (VarW), where differences between pups of the same litters results in outcome heterogeneity. In addition to this maternally-caused between and within-litter variability in outcomes measures, experimental procedures can give rise to VarB and VarW through pre- and postnatal manipulations. Unless sufficiently and appropriately mitigated, experimental outcomes can be misinterpreted. Approaches to mitigating VarB, such as the use of single offspring within a litter, often overlooks VarW and can result in erroneous conclusions through sampling error. In contrast, approaches that aim to account for VarW, such as analysing offspring as independent samples, could result in VarB being overlooked. Recently, the use of random models in statistical analysis, whereby dam is considered a random source of variation and offspring a pseudo-replication, can afford insight into both sources of variation. We provide an example of these sources of variation through maternal immune activation (MIA), commonly used as a preclinical model of neurodevelopmental disorders. MIA involves inducing gestational inflammation, resulting in VarB from the dam-specific immune response, and VarW from the differential response within the litter from factors such as intrauterine position. In this review, we identify key sources of variation between and within litters and suggest considerations for how to deal with VarB and VarW in designing experiments and data analysis.