Epidemiology assignment due on 2/21/15, What is a cause? Please answer questions

Excelsior College PBH 321

1

CAUSAL "G UIDE LINES" SUGGE STED BY SI R AUSTIN BRA DFORD H ILL (1965)

Sir Bradford Hill developed a set of nine criteria for assessing causality, which have been extensively applied in epidemiology and public health since that time. These criteria should be used as guidelines to help determine if associations are causal, but should not be used as rigid criteria to be followed slavishly. They are not “hard and fast rules”, but instead a tool by which we may approach the process of determining if the association observed between an exposure and disease is causal. In fact, some of these criteria may be too subjective, or too easily refuted to serve as evidence of causality. We’ll explore each of these “guidelines” and note particular instances where they may not be applicable.

• Strength of the association • Consistency • Specificity • Temporality • Biological gradient • Plausibility • Coherence • Experiment • Analogy

1. Strength of the association: The larger the association, the more likely the exposure is causing the disease. Example: Relative risk of lung cancer in smokers vs. non-smokers = 9.0 Relative risk of lung cancer in heavy smokers vs. non-smokers = 20.0 How strong is strong? This is a difficult question to answer, and may depend on the nature of the exposure and disease of interest. Here is an example of a typical epidemiologist’s interpretation:

Relative risk Interpretation 1.1-1.3 Weak 1.4-1.7 Modest 1.8-3.0 Moderate 3-8 Strong 8-16 Very strong 16-40 Dramatic 40+ Overwhelming

Strong associations are more likely to be causal because they are unlikely to be due entirely to bias and confounding. An RR or OR of 10.0 may be biased, but it is unlikely that bias completely explains all of the association. In comparison, a weaker association, such as an RR or OR of 1.1, may be causal, but it is harder to rule out bias and confounding as explanations for the relationship. A famous classical epidemiological study, which examined of smoking and risk of lung cancer conducted by Doll and Hill in the 1950s found ORs (comparing smokers to non-smokers) in excess of 30!

Excelsior College PBH 321

2

2. Consistency: The association is observed repeatedly in different persons, places, times, and circumstances. Replicating the association in different populations, with different study designs, and different investigators gives evidence of causation. Example: Smoking has been associated with lung cancer in at least 29 retrospective and 7 prospective cohort studies. Note: You have learned to evaluate studies somewhat critically with respect to the populations studied, sources of bias, and other characteristics. Sometimes there are good reasons why study results differ. For example, one study may have looked at low level exposures while another looked at high level exposures. 3. Specificity: A single exposure should cause a single disease. This is a hold-over from the concepts of causation that were developed for infectious diseases. There are many exceptions to this. Example: Smoking is associated with lung cancer as well as many other diseases. In addition, lung cancer results from smoking as well as other exposures. When present, specificity does provide evidence of causality, but its absence does not eliminate the possibility of causation. 4. Temporality: The causal factor must precede the disease in time. This is the only one of Hill's criteria that always applies. Prospective studies do a good job establishing the correct temporal relationship between an exposure and a disease. Example: A prospective cohort study of smokers and non-smokers starts with the two groups when they are healthy and follows them to determine the occurrence of subsequent lung cancer. 5. Biological Gradient: A “dose-response” relationship between exposure and disease. This means that people who have increasingly higher exposure levels have increasingly higher risks of disease. Usually, confounding and bias would not produce a dose-response relationship, so evidence of one provides support for causality. Example: Lung cancer death rates increase with the number of cigarettes smoked.

Some exposures might not have a "dose-response" effect but rather a "threshold effect", below which no adverse events arise.

Lung cancer deaths

(incidence)

Number of cigarettes smoked (dose)

0

Excelsior College PBH 321

3

Example: Investigators in one study found no increased risk of miscarriage among pregnant women exposed to trihalomethanes (a water disinfection byproduct) in drinking water, until levels reached 75 ug per liter.

6 & 7: Plausibility/Coherence: Either a biological or social model exists to explain the association. The association does not conflict with current knowledge of natural history and biology of disease. Example: Cigarettes contain many carcinogenic substances. Thus, a relationship between cigarette smoking and lung cancer is plausible. Many epidemiologic studies have identified cause-effect relationships before biological mechanisms were identified. For example, the carcinogenic substances in cigarette smoke were discovered after the initial epidemiologic studies linking smoking to cancer. 8. Experiment: As you learned in Module 3, experimental studies provide the strongest evidence of causality because of their ability to control for bias. Investigator-initiated intervention that modifies the exposure through prevention, treatment, or removal should result in less disease. Example: Smoking cessation programs result in lower lung cancer rates. Experimental studies provide strong evidence for causation, but most epidemiologic studies are observational. It is often not ethical or feasible to manipulate the risk factor to determine whether it will lead to change in disease. 9. Analogy: Refers to the idea that an observed association is more likely to be causal if similar to other associations believed to be casual. We determine this by asking the question, “Has a similar relationship been observed with another exposure and/ or disease?” Example: Effects of thalidomide on the fetus provide analogy for effects of similar substances on the fetus. (Thalidomide, a drug prescribed in the 1950s for treatment of morning sickness, was ultimately determined to cause birth defects). Hill concludes: “Here then are nine different viewpoints from all of which we should study association before we cry causation.... None of my nine viewpoints can bring indisputable evidence for or against the cause-and-effect

Miscarriage

Maternal exposure to trihalomethanes in drinking water

Threshold

0

Excelsior College PBH 321

4

hypothesis…. What they can do, with greater or lesser strength, is to help us make up our minds on the fundamental question --is there any other way of explaining the set of facts before us, is there any other answer equally, or more, likely than cause and effect?” Though they do not establish causality directly, Hill’s criteria are useful guides for:

• Remembering distinctions between association and causation in epidemiologic research; • Critically reading epidemiologic studies; • Designing epidemiologic studies; • Interpreting the results of your own study

Fundamentally, causal inference is a philosophical activity akin to asking, “What is truth?” There are no simple and direct set of rules or procedures to answer this question. We can’t prove causality with statistical tests or checklists (like Hill’s criteria). Causal inference involves judgments and social processes that change over time as our knowledge base expands. The epidemiologist must use all the tools at his/her disposal to come to thoughtful conclusions about causality.