Bats, bugs, and babies: why stringent guidelines are needed for cause-and-effect interpretation in epidemiological studies.

In a recently published study (Frank, 2024), the author claimed that farmers in the USA compensated for bat decline by increasing their insecticide use by 31.1%, which allegedly caused a 7.9% increase in human infant mortality. The author concluded that his "result highlights that real-world use levels of insecticides have a detrimental impact on health, even when used within regulatory limits". The study is a prime example of how statistics should not be used in a design based on a classical logical fallacy. That is, in short, factor A (detection of a lethal bat disease) correlates with factor B (increased insecticide use) and factor A also correlates with factor C (increased infant mortality). Based on a flawed logic, the author concludes that consequently a causal relationship exists between factor B (pesticide use) and factor C (infant mortality). Remarkably, the causal relationship between increased pesticide use (factor B) and increased infant mortality (factor C) was claimed despite not investigating a statistical correlation between these two factors. The study also contains numerous important but less obvious flaws, including the use of aggregated data to make inferences about individual outcomes; inadequate consideration of maternal characteristics; a suboptimal proxy for bat population declines; inadequate timing of 'treatment'; assumption of quasi-randomness; questionable binning of coefficients and a lack of data documentation. Scientifically, the article highlights the need for more stringent statistical and scientific quality controls. However, given the implications such cavalier claims have in the context of public health, it also shows that the scientific community urgently needs a better understanding of the robustness of statistical conclusions in the context of epidemiological and ecological studies.