According to van Belle, there is no plausible mechanism to identify one or more factors as causal. Therefore, statistical models we come up with to explain the cause of a disease that is multifactorial may be inadequate. This is true for those diseases with an environmental component. He suggests using sensitivity analyses to address this.

Other problems to watch out for is selection bias and measurement error. Publication bias has been found to reduce excess risk from 24% to 15%, thereby exaggerating risk.

G. van Belle makes a case for developing a public health definition for “effect size.” For example, low levels of air pollution have small effects, but the number of people at risk is hugh. In essence, the overall effect can be large. Additionally, studies of small effects require large sample sizes, which can be costly.