Risk and Numbers Needed to Treat

Revised and reviewed 26 August 2015

OVERVIEW

• Risk is a nebulous term in evidence-based medicine, as it may refer to either relative or absolute risks
• Overall, absolute risks are more meaningful and can be converted in ‘numbers need to treat’ (NNT), which are useful in the clinical setting

RISK

Absolute Risk

• the actual event rate in the placebo or treatment group

Relative Risk

• the ratio of the incidence of disease among exposed to the incidence among non-exposed
• a measure of the strength of an association between groups
• prospective studies (RCTs and cohort studies)
• also called the incidence risk

RR = risk of disease in the exposed (a/a+b) / risk of disease in the non-exposed (c/c+d)

• RR of 3 -> there is three times the risk
• RR of 0.5 -> the risk is halved
• RR of 1 -> there is no association
• if the RR is reported with a CI that includes 1 -> then the RR is not significant.

Relative Risk Reduction

• expressed as a percentage reduction in events in treated vs untreated groups
• = 1 – (incidence in exposed / incidence in unexposed)

Attributable Risk

• a measure of association that provides information about the absolute effect of the exposure or excess risk of disease of those exposed compared to unexposed

Absolute Risk Reduction (ARR)

• incidence in exposed – incidence in unexposed
• a measure of treatment effect
• reverse of attributable risk

NUMBER NEEDED TO TREAT

• NNT is the number of patients who need to be treated in order to avoid one adverse event, which is the reciprocal of the absolute risk reduction
• NNT gives the RR some relevance in terms of the magnitude of clinical effect
•  ie. if the incidence of an adverse event = 0.06% & relative risk reduction = 0.33 the absolute risk reduction -> 0.02% and thus numbers needed to treat = 1/0.0002 = 5000
• however, if the incidence of an adverse event was 6% the absolute risk reduction -> 2% and thus the NNT = 1/0.02 = 50