Test | Question. |
1) Statistical significance | Could the association have been a result of the play of chance? This has come to be tested by the probability that the result is not random at the statistical level of 95%, 1 in 20 or p = 0.05. |
2) Strength of association | Is the association strong enough to define a concern? How many individuals will suffer if the effect is real? |
3) Consistency | Is the association found in different studies of the same exposure? However, they must be similar studies involving the same exposures. |
4) Specificity and reversibility | Is the effect specific to the putative cause? If the cause is removed, does the effect disappear? For effects which are the result of permanent damage, reversibility cannot occur. |
5) Relationship in Time | Does the cause precede the effect? |
6) Biological gradient | Does increasing the exposure increase the effect? This may only exist over a defined range, if high exposures cause death, gradient cannot be constant. |
7) Mechanism: biological plausibility | Is a mechanism known? Do cell studies or animal studies support a plausible mechanism? Hill made clear that this was not a necessary requirement since the true mechanism may not be known. |
8) Alternative explanation | Is there a confounding explanation? Could there be some other cause that is responsible for the effect? |