Box 3. Calculating measures of “effect” in the GATE frame
| Risk ratio or relative risk [RR] | ||
|---|---|---|
| *Incidence or prevalence. | ||
| †The GATE checklists4 calculate 95% CIs for effect estimates. | ||
| ‡NNE is the generic form of NNT (numbers needed to treat). | ||
| §When the CI for the risk difference crosses 0, the CI for an NNT (or NNE) passes through infinity.7 | ||
| = the ratio of occurrence* (risk) in the exposure group to the occurrence in the comparison group = EGO ÷ CGO (or EER ÷ CER in intervention studies) | ||
| Example: The occurrence (risk) of CHD death in the HRT group relative to the risk in the placebo group at the conclusion of HERS | ||
| RR | = 12.5 ÷ 10.2 events per 1000 women per year (from box 1) = 1.23 (95% CI 0.87 to 1.74)† | |
| Risk difference [RD] or absolute risk difference [ARD] | ||
| = the difference in occurrence* (or risk) between the exposure group and the comparison group = EGO – CGO or = EER – CER | ||
| Example: The difference in occurrence between the HRT treated group and the comparison group in HERS | ||
| RD | = 12.5 – 10.2 events per 1000 women per year (from box 1) = 2.3 events per 1000 women per year (95% CI −1.61 to 6.25)† | |
| Number needed to expose [NNE] (or number needed to treat [NNT]) to reduce (or increase) events by 1 during time (T) | ||
| = 1 ÷ (EGO – CGO) or = 1 ÷ (EER – CER) | ||
| Example: The number needed to be treated with HRT for 1 year to cause 1 additional CHD death in HERS | ||
| NNE or NNT‡ | ||
| = 1 ÷ 2.3 events per 1000 women per year = 435 women for 1 year (95% CI 160 to ∞ to −620) | ||
| ie, 435 women in HERS needed to be treated with HRT for 1 year to cause 1 additional CHD death | ||