GUIDELINES FOR THE ACCEPTANCE OF CAUSAL ASSOCIATIONS (After A.B. Hill)
1. Temporal relationship
It is clear that if a factor is to be the cause of a disease, exposure to the factor must have occurred before the disease has developed. This is easier to verify in concurrent cohort studies than in retrospective cohort or case control studies. Also important is the determination of the lag or delay between exposure and onset, as this is characteristic for most diseases.
2. Strength of the association (SOA)
The stronger the association, the more likely it is causal. SOA is measured by such statistical tools as the odds ratio, or the regression coefficient.
3. Dose-response relationship
If the risk of disease increases as the dose (exposure) increases, then the association is more likely to be causal. But an association my still be causal in the absence of a dose-response, due to a threshold effects.
4. Replication of the findings
We should see the same results of an exposure under the same conditions.
5. Biological plausibility
Although epidemiology sometimes precedes biology,* this guideline requires that the proposed association to be compatible with, or at least not counter to, current biological knowledge.
6. Consideration of alternative explanations
The investigators are required to have searched exhaustively for alternative ways to explain the findings, and to have dealt satisfactorily with them. This would include all forms of bias, confounders, and design effects.
7. Cessation of exposure
We would expect the risk of a disease to reduce when the hypothesized exposure is reduced or removed. While this is usually the case, instances can be found in which damage or disruption has reached an irreversible stage, and removal of exposure would have no effect.
8. Specificity of the association
This suggests that a specific exposure is associated with only one disease. It is the weakest of the guidelines, and probably should be removed from the list. Cigarette smoking, as an example, has been linked to dozens of separate diseases and conditions; this may be due to a single Aexposure@ actually being an exposure to hundreds of separate chemicals. Similarly, a single substance may damage DNA in a specific manner, but the DNA could be in any tissue in the body, resulting in birth defects, or several types of cancer.
9. Consistency with other knowledge
Wider in scope than #5, this requires that a proposed association should be compatible or consistent with other data or research results.
An example: Assessment of the Evidence Suggesting Helicobacter pylori as a Causative Agent of Duodenal Ulcers (using Hillís criteria)
1. Temporal relationship.
H. pylori is clearly linked to chronic gastritis. About 11% of chronic gastritis patients will go on to develop duodenal ulcers over a 10‑year period.
! In one study of 454 patients who underwent endoscopy 10 years earlier, 34 of 321 patients who had been positive for H. pylori (11%) developed duodenal ulcer compared with 1 of 133 H. pylori ‑negative patients (0.8%).
2. Strength of the relationship.
H. pylori is found in at least 90% of patients with duodenal ulcer. In at least one population reported to lack duodenal ulcers, a northern Australian aboriginal tribe that is isolated from other people, it has never been found.
3. Dose‑response relationship.
Density of H. pylori per square millimetre of gastric mucosa is higher in patients with duodenal ulcer than in patients without duodenal ulcer. Also see item 2 above.
4. Replication of the findings.
Many of the observations regarding H. pylori have been replicated repeatedly.
5. Biologic plausibility.
Although originally it was difficult to envision a bacterium that infects the stomach antrum causing ulcers in the duodenum, it is now recognized that H. pylori has binding sites on antral cells and can follow these cells into the duodenum.
H. pylori also induces mediators of inflammation.
H. pylori-infected mucosa is weakened and is susceptible to the damaging effects of acid.
6. Consideration of alternate explanations.
Data suggest that smoking can increase the risk of duodenal ulcer in H. pylori infected patients but is not a risk factor in patients in whom H. pylori has been eradicated.
7. Cessation of exposure.
Eradication of H. pylori heals duodenal ulcers at the same rate as histamine receptor antagonists.
Long‑term ulcer recurrence rates were zero after H. pylori was eradicated using triple antimicrobial therapy, compared with a 60% to 80% relapse rate often found in patients with duodenal ulcers treated with histamine receptor antagonists.
8. Specificity of the association.
Prevalence of H. pylori in patients with duodenal ulcers is 90% to 100%. However, it is found in some patients with gastric ulcer and even in asymptomatic individuals.
9. Consistency with other knowledge.
Prevalence of H. pylori infection is the same in men as in women. The incidence of duodenal ulcer, which in earlier years was believed to be higher in men than in women, has been equal in recent years.
Prevalence of ulcer disease is believed to have peaked in latter part of the 19th century, and prevalence of H. pylori may have been much higher at that time because of poor living conditions. This reasoning is also based on observations today that H. pylori prevalence is much higher in developing countries.
This excellent example courtesy of Megraud F, Lamouliatte H: Helicobacter pylori and duodenal ulcer: Evidence suggesting causation. Dig Dis Sci 37:769‑772, 1992; and DeCross AJ, Marshall BJ: The role of Helicobacter pylori in acid‑peptic disease. Am J Med Sci 306:381‑391, 1993.
The usual sequence from initial, non-controlled observation, to the cautious conclusion that the exposure may actually cause the outcome, is as follows...
Field or clinical observations
First examination of data
Evaluation of the evidence ( 9-point Guidelines for determination of causation )
/ (or) \
Association is Association is
probably probably not
*Gregg's observation that rubella during pregnancy appeared to increase the risk of congenital cataracts in the baby preceded any knowledge of teratogenicity in viruses.