It can be exciting to find a significant association between your primary predictor and your study outcome. The coefficient is in the right direction, the biological plausibility is all there, it’s indicative of a true effect! Publish! Wait – is it a real association or indicative of reverse causality? This is the step that some of us can forget to check, and indeed can be all our ‘finding’ is showing. While particularly a potential concern for cross-sectional or case-control studies, even studies in which data from a longitudinal cohort study are analysed should take into account the possibility of reverse causality, and rebut this possibility as an explanation for their findings. Using the model of multiple sclerosis and environmental and biologic predictors of clinical disability, we present some cases where a promising association may simply reflect reverse causality. We also present some analytical methods whereby reverse causality can be assessed, and the utility of which can make you – and the reviewers – more confident your results mean what you think they do.

Item Type:	Non Refereed Article
Keywords:	epidemiology, reverse causality
Research Division:	Health Sciences
Research Group:	Epidemiology
Research Field:	Epidemiology not elsewhere classified
Objective Division:	Health
Objective Group:	Public health (excl. specific population health)
Objective Field:	Disease distribution and transmission (incl. surveillance and response)
UTAS Author:	Simpson Jr, S (Dr Steve Simpson JR)
UTAS Author:	Blizzard, L (Professor Leigh Blizzard)
UTAS Author:	Taylor, B (Professor Bruce Taylor)
UTAS Author:	Tettey, P (Mr Prudence Tettey)
UTAS Author:	van der Mei, I (Professor Ingrid van der Mei)
ID Code:	93679
Year Published:	2013
Deposited By:	Menzies Institute for Medical Research
Deposited On:	2014-08-13
Last Modified:	2014-08-28
Downloads:	0