Relative risk/absolute risk reduction

The most valid method of providing results of clinical trials or meta-analysis of clinical trials is to provide the rates of occurrence of important clinical outcomes for both the treated (Pe) and the untreated (Pc) groups, calculating the relative risk reduction, (Pe-Pc)/Pc, and the absolute reduction, Pe-Pc. Some authors express the reciprocal of the absolute risk reduction (1/Pe-Pc) or the number needed to treat (NNT) as a measure of the benefits of treatment.

Appropriately applying the NNT and absolute benefits of treatment from a clinical trial or systematic review to an individual patient requires comparable risk of disease in that patient compared with the study population, whereas the relative risk reduction should be robust even when the risk of disease is much different in that individual patient compared to the study population analyzed. For example, the rate of persistent ear pain at two to seven days in children younger than age two with bilateral disease may be greater than the 14 percent found in this meta-analysis in the untreated group—perhaps 28 percent. We will still expect a 34 percent reduction relative to this baseline level with the use of antibiotics if a causal relationship exists between pain and antibiotic use. However, the NNT for one patient to have benefit would be 10.5, not 20 as in the present Cochrane review.