It is necessary to distinguish between relative risk (RR) and the risk difference (RD), because the relative risk is typically much larger than then the risk difference so that when you read results in the form of a RR it can be rather misleading. Think about reading a statement that says something has reduced the patient’s risk by 50%. This sounds quite good. However, if you think about it, it might mean nothing more than you reduced the risk from, say, 2% to 1%.
More clinically, imagine that there are three subpopulations of patients, each getting the same treatment, and for each group we see a decrease in risk of 1/3 (that is, RRR= 0.33; RR= 0.67). If this was administered to a subpopulation that has a 30% risk of dying, getting the treatment reduces that risk to 20%. When it is given to a subpopulation that has a 10% risk of dying, the treatment reduces the risk to 6.7%. And when given to a subpopulation that was just a 1% risk of dying, the treatment reduces the risk to 0.67%.
In each group, the treatment reduces the risk of dying by a third, but it really fails to adequately express the full impact of treatment. What if this treatment had a high rate of serious adverse events? Would it make much sense to offer it to the patients in the lowest risk group (where the RD is only 0.3%)? For the middle group, we’d certainly want to let them know about the risks and benefits of treatment, inasmuch as the as the absolute reduction in risk is about 3%. In the high risk group, it makes complete sense to offer this treatment, because there is an absolute benefit of 10%.
The authors of the Users’ Guide suggest that the relative risk reduction be examined in light of your assessment of your patient’s baseline risk. Obviously, you need to have information on the risk reduction of a therapy, which is not always available for chiropractic interventions but could be calculated as per our last blog post. If your know your therapy offers a relative risk reduction of 30% in a general group of patients, you might view it differently if one patient is 40 years old and in generally good health compared to one who is 70 years old and has significant co-morbidity.
This is not the easiest concept to grasp. However, the Users’ Guide offers a great resource in understanding this (1). Further information can be found at http://www.annalsofian.org/temp/AnnIndianAcadNeurol104225-401063_110826.pdf
NOTE: As I was home over the weekend, I came across this article in the Washington Post: Howell D. Making sense of science reporting (2). In it, the Washington Post Ombudsman, Deborah Howell, defends the nature of science reporting in her newspaper, but offers the following comment regarding one such article: “FIAR Director George M. Carter's chief complaint was that stories emphasized a change in "relative risk" -- a 44 percent fall in the number of heart attacks, strokes and surgical procedures among people taking the statin, compared with those in the placebo group. He said the fact that everyone in the study had an extremely low "absolute risk" for heart problems should have been emphasized more. About 1.36 percent of people taking the placebo suffered a heart attack or stroke; that fell to 0.8 percent among those taking the statin. That means that nearly 97 percent of the people using the drug would not see any benefit, he said.”
The article goes on to note the author of the news report “quoted a skeptic in the ninth paragraph and noted near the story's end that ‘the actual risk reduction for an individual would be very small, given the relatively low risk for most middle-age people, so that the benefits easily could be outweighed by the costs of thousands more people taking tests, drugs and being monitored by doctors.’”
Finally, the author said "While I would have liked to have explored many of the nuances of this study more fully, I feel confident we struck a responsible balance. I think it's crucial to provide readers with both the evidence supporting new claims and enough context and interpretation to help them gauge its significance." Independent experts, he said, concluded the study was "a very well done, very convincing piece of research."
Which, of course, was not the point. It is only that the average reader cannot draw this distinction between the absolute and relative risk reductions. That much remains clear, and points to why we need to better understand such concepts.
1. Guyatt G, Rennie D, Meade MO, Cook DJ. Users’ guides to the medical literature: a manual for evidence-based practice. New York, NY; McGraw Hill;208:90-91
2. Howell. D. Making sense of science reporting. http://www.washingtonpost.com/wp-dyn/content/article/2008/12/05/AR2008120502959.html?hpid=opinionsbox1