Are residents of communities located around fields for petroleum extraction at increased risk of cancer? It is hard to answer this question, since statistical evidence is weak and gives contradictory results. Since affected areas are mostly in remote regions of developing countries, there are serious bstacles to carrying out good population studies. Demographic information is scarce, health registries are lacking, sickness and death are under reported, and population is often exposed to many other contaminants because of loose environmental regulation.
What should experts do with the scarce evidence available? Is one entitled to make decisions based on it? Experts disagree on this matter. Some think that it is not the business of epidemiology to infer risk from unreliable correlations and biased experimental designs. Other epidemiologists, instead, think that in some cases the discussion about p-values and potential confounders must be supported by an element of plausibility, for instance by referring to known mechanisms of harm.
In a new paper, ‘Why Causal Evidencing of Risk Fails, and Example from Oil Contamination’, we look at the philosophical Basic Implicit Assumptions (philosophical BIAS) behind these two opposing argument, focusing on assumptions about the nature of causation.