Before recommending a course of action, doctors weigh the potential risk of harm from a treatment against its potential benefit.
Sometimes the benefit of treatment is to decrease symptomsfor example, reduce pain. Benefit may also be improved function, for example, being able to walk farther. At times, benefit is cure of a disease. At other times, a treatment reduces the future likelihood of undesirable events, such as complications of a disease. Then, the benefit is a decrease in the risk of an event that people want to avoid.
For example, doctors may consider recommending a particular drug to reduce the risk of a stroke. They evaluate the results of a controlled clinical trial in which 2,000 people were studied. The results show that of 1,000 people given the drug, 20 have a stroke despite taking the drug. The study also shows that of the other 1,000 people in the study who are not given the drug (are given a placebo), 40 have a stroke. The results of the study could be expressed as showing that the drug cut the risk of stroke in half because 20 is half of 40 (50% decrease in relative risk). However, the results could also be expressed as showing that only 20 people of 1,000 benefited (2% decrease in absolute risk). It sounds much more impressive to say the risk of stroke is cut in half than to say there was a 2% decrease in risk.
Risk is the likelihood that a harmful outcome will occur. When describing risk of harm, absolute versus relative risk should be evaluated. In the example above, perhaps the drug that prevents stroke caused severe bleeding in 3% of people. Although 3% does not sound like much of a risk, it means that 30 people of every 1,000 who took the drug had severe bleeding.
People and their doctors must carefully choose which statistics they use to help make decisions. Obviously, contrasting “50% improvement” against “3% chance of serious harm” makes treatment sound like a good option. However, the same numbers also mean that the treatment benefited 20 of 1,000 people but harmed 30. Stated this way, the treatment does not sound like a good option. In this case, people must weigh the severity of the harm of treatment against the severity of the illness. For example, if a stroke is severe, leaving people unable to talk or care for themselves, and the drug-induced bleeding requires only some blood transfusions but no surgery and is not fatal, people may accept the higher risk of bleeding to avoid the lower but more serious risk of stroke.
Research studies provide information only about the average risk of harm and benefit. But average effects do not always tell doctors how a person will respond to a treatment. Because of this uncertainty, many scientific studies try to determine characteristics of people (such as age, other disorders they have, and blood test results) that can better identify those who are more likely to benefit from or be harmed by a treatment.
Last full review/revision June 2007 by Thomas V. Jones, MD, MPH