On causal inferences from retrospective and observational studies and their implications for neuraxial labor analgesia: the CHRISTMAS* study.

Abstract

Background: Observational studies should not be used to infer causation as they are prone to confounding factors, selection bias, and reverse causality. Many observational studies of labor analgesia treated epidurals as an independent exposure and concluded that "epidurals" cause dystocia, despite multiple randomized controlled trials showing no effect. We highlight this problem using reductio ad absurdum. We explore whether people request reading glasses when their progressively increasing focal length equals or exceeds their fixed arm length.

Methods: We designed a cross-sectional retrospective and prospective observational study to assess whether there is an association between arm length and age when first requesting reading glasses in presbyopia. We evaluated individuals aged 38-55 receiving their first reading glasses for presbyopia (either currently or within the past year). We recorded age at first request for reading glasses, the refractive correction (additions) in each eye, and we measured arm length.

Results: Seventy subjects were included in the study. No association was found between arm length and the age at request for reading glasses or the severity of presbyopia at presentation.

Conclusions: Even if this observational study had demonstrated a strong correlation between age at request for reading glasses and arm length, it would have been absurd to conclude that spectacles somehow cause our arms to shrink. Similarly, women in obstructed labor with a narrow pelvis are more likely to request neuraxial labor analgesia, but "epidurals" do not make their pelvis shrink. Making far-reaching causal inferences based on retrospective or observational data is very shortsighted.