Monday, September 30, 2013

How to Read an Article about Harm

Consider this question: how would you test the idea that smoking causes cancer? The best way to do so would be to develop a clinical trial in which half of the participants were required to smoke three packs of cigarettes per day for five years while the other half did not smoke at all. You could then see if there were differences in the rate of lung cancer between the two groups. But, of course, this is not ethical.

Given that we cannot use clinical trials to answer such questions, we can use other designs, including both a case-control study design and a cohort design.  In a case-control study we would look in the past medical records and exposures of cases (those with lung cancer) and controls (those without), and would find that some in each group were heavy smokers while others were not. We could then calculate the difference in rates of cancer between the two groups. In such case, we would end up with an odds ratio; that is, the odds that exposure to smoking leads to lung cancer. In a cohort study, we would follow people forward in time while allowing them to live their life. None have cancer at the beginning of the study. We would find, years down the road, that some in both groups were heavy smokers while others were not, and again we could calculate the difference in rates of cancer between the two groups. In this case, we would end up with a risk ratio.
Risk is associated with disease incidence; that is, the rate of newly diagnosed conditions in a population. In a case-control study, we are starting with people who already have the condition of interest; therefore, we cannot calculate risk (which requires us to newly diagnose a disease), and instead we look at odds.

When we read an article about harm, we need to understand the specific study design being used. While clinical trials are best, they often cannot be conducted. Cohort studies are stronger than case-control studies, since they allow us to calculate the true disease rate in a group. But in studies of harm, other questions to look at while reading are to ensure that the exposures and outcomes in both groups were measured the same way, that follow-up was long enough, that the exposure precedes the adverse outcome, and that the association between exposure and outcome is strong.
It is important to understand that risk and odds ratios do not tell you how frequently a problem occurs, only that the effect occurs more or less often in the exposed group compared to the unexposed group. This can then tell you whether or not to recommend the patient stop the exposure. Once we know, for example, that smoking is associated with a higher rate of cancer, we can advise patients to stop smoking.

For additional information on harm, please see

No comments: