The risk difference is the difference between the observed risks (proportions of individuals with the outcome of interest) in the two groups (see Box 9.2.a). The risk difference can be calculated for any study, even when there are no events in either group. The risk difference is straightforward to interpret: it describes the actual difference in the observed risk of events between experimental and control interventions; for an individual it describes the estimated difference in the probability of experiencing the event. However, the clinical importance of a risk difference may depend on the underlying risk of events. For example, a risk difference of 0.02 (or 2%) may represent a small, clinically insignificant change from a risk of 58% to 60% or a proportionally much larger and potentially important change from 1% to 3%. Although the risk difference provides more directly relevant information than relative measures (Laupacis 1988, Sackett 1997) it is still important to be aware of the underlying risk of events and consequences of the events when interpreting a risk difference. Absolute measures, such as the risk difference, are particularly useful when considering trade-offs between likely benefits and likely harms of an intervention.
The risk difference is naturally constrained (like the risk ratio), which may create difficulties when applying results to other patient groups and settings. For example, if a study or meta-analysis estimates a risk difference of –0.1 (or –10%), then for a group with an initial risk of, say, 7% the outcome will have an impossible estimated negative probability of –3%. Similar scenarios for increases in risk occur at the other end of the scale. Such problems can arise only when the results are applied to patients with different risks from those observed in the studies.
The number needed to treat is obtained from the risk difference. Although it is often used to summarize results of clinical trials, NNTs cannot be combined in a meta-analysis (see Section 9.4.4.4). However, odds ratios, risk ratios and risk differences may be usefully converted to NNTs and used when interpreting the results of a meta-analysis as discussed in Chapter 12 (Section 12.5).