User:ChyranandChloe/Workshop 15

Inference

Causation implies that the burden of disease may be avoided through prevention or cessation. This definition is essential to studies moving from an associative relationship to a causal relationship. The definition of causation rests upon a "counterfactual" state; which builds upon the notion that if a person was observed with and without the causal factor (tobacco), and without changing any other characteristics: a difference would observed between the two states. Since a perfect counterfactual state is impossible, randomization is often used to distribute potential confounders and allow results to be describable using probability.

Randomization requires that both the selection of the individuals to be observed the allocation of treatments be randomized. Randomly selecting people into two groups, and administering a hazardous substance to one group, cannot be ethically done. Therefore most evidence is observational. In absence of the randomization in treatments, causal inferences is difficult. A nine item criteria is therefore applied to an already established statistical association:

• Consistency refers to the persistent finding of an association between exposure and outcome. It serves two purposes. The first is to make unmeasured confounders an unlikely explanation for the association. The second is to make the chance effect of the random selection of individuals unlikely, by increasing the statistical strength by accumulating large bodies of data.
• Strength of Association Strength of association makes alternative explanations by either confounders or chance unlikely. Weak association are more likely to reflect chance, modest bias, or unmeasured weak confounding.
• Specificity is the interpretation that a single effect follows a single cause. When available, specificity can increase a causal claim, but its absence does not weaken it. Most cancers have multiple causes, because of this, relative risk is used to measure the development of cancer.
• Temporality, exposure does not immediately lead to the purported disease. Failing to establish the sequence from which the purported disease develops severely weakens a causal claim, but establishing temporality is not a strong claim in favor of causality.
• Coherence, Plausibility, and Analogy entails that the biological understanding remains consistent with statistical evidence. Biological understanding is constantly evolving, as the exact mechanisms—whether through a genetic or a molecular basis—develop better explanations for disease pathogenesis and cancer carcinogenisis. Since the early 1950s, the association remains biologically plausable.
• Biologic Gradient or dose-response, entails that increase consumption is associated with increased risks. Dose-response increase the causal claim, providing a predictable cause-effect model and makes non-causal explanations unlikely.
• Experiment Nautral processes can be reproduced in controlled experiments. This process to control the exposure in individuals in a way that does not depend on the individual's characteristics.

Evaluating causal conclusions has lead to disparities within the language used to classify causation. The Surgeon General revising to the approached used by the Institute of Medicine (IOM) and the International Agency for Research on Cancer (IARC) use a four-level hierarchy for classifying the strength of causation:

“

1. Evidence is sufficient to infer a causal relationship.
2. Evidence is suggestive but not sufficient to infer a causal relationship.
3. Evidence is inadequate to infer the presence or absence of a causal relationship (which encompasses evidence that is sparse, of poor quality, or conflicting).
4. Evidence is suggestive of no causal relationship.

”