Definition of Causation


The chicken or the egg causality dilemma: which one came first?

The chicken or the egg causality dilemma:
which one came first?


Causation describes the association between two events, such that one causes the other.  One aim of epidemiological research, amongst others, is to examine whether there exists an association between any risk factor and a health-related outcome (e.g. a specific disease).


In contrast to a controlled experiment, in which the intervention group and the control group are allocated randomly, epidemiological research is generally observational.  In such a study, the researcher gathers information on exposure/s and health outcome/s in real life health conditions, and analyzes it to obtain answers to a research hypothesis.  Before it can be claimed that an observed association between exposure and outcome is real, it must be analyzed, because such an association may be caused by chance (incidental error), through bias (systematic error) or it may be due to confounding factors.


One way to evaluate the association is through the statistical significance between the potential risk factor and the studied disease.  But even statistical significance may not necessarily lead to a conclusion that causation exists, and conversely absence of statistical significance does not necessarily imply absence of causation.


The development of chronic diseases is generally influenced by a number of factors, and there are no clear criteria for unequivocally declaring an observed association to be causal.  In order to determine whether an observed association is causal, the following criteria, outlined by Bradford Hill as early as 1965 (Hill's criteria for causation), are generally applied:

  1. Temporality – The exposure must precede the outcome.
  2. Biological gradient (Dose-response relationship)– An association between the amount/level of exposure and the magnitude of its effect reinforces the possibility of a causal association.
  3. Strength– The stronger the association between a risk factor and an outcome, the greater the probability of a causal association.  A strong association is expressed by a relatively high risk of developing a disease following exposure to the risk factor in question.
  4. Consistency– If similar findings are obtained in additional studies conducted in different populations, using various research methods and at different times, the probability of a causal association is strengthened.
  5. Plausibility – The existence of a potential biological mechanism (an explanation in accordance with current scientific knowledge) strengthens the probability of a causal association.  However, in certain cases, current knowledge about a potential biological mechanism is limited, and it should be noted that some associations were found and confirmed many years before the biological cause for them was known.
  6. Coherence – Consistency between experimental laboratory results and epidemiological research strengthen the likelihood of causation in the observed association.
  7. Specificity– If the observed association is specific to the risk factor and the outcome, the likelihood of a causal association is strengthened.  However, it should be noted that some risk factors, such as smoking, may be causally associated with a variety of health outcomes.
  8. Experiment – If removal of the risk factor changes the outcome, a causal association between the factor and the outcome may be assumed.
  9. Analogy – Similar risk factors may be tested to see whether they produce similar effects on outcome.

Of the criteria mentioned above, only the temporality criterion, i.e., that the exposure must precede the outcome, is necessary for proving causation. Scientific causation is usually determined by expert committees that discuss the overall scientific evidence whilst considering Hill's criteria for causation.   


  • Hill AB. The Environment and Disease: Association or Causation? J Roy Soc Med. 1965; 58:295-300