MEASURES OF BENEFIT + TREATMENT Flashcards
Define number needed to treat
Measure of benefit of intervention when compared with control in trials. Indicates how many participants need to receive intervention to prevent one adverse outcome (that is, for one more participant to benefit) than if those same people had received control treatment
In the study about financial incentives to stop women smoking in weeks 34-38 gestation, the NNT was 7.2. What does this mean?
On average, if 7.2 pregnant smokers received financial incentives, one more woman would be abstinent at 34-38 weeks’ gestation than if those same women received control treatment
What is the ideal value of NNT?
One
In the context of the financial incentives for pregnant women, what would a NNT of 1 mean?
Would imply that every woman offered financial incentives would be abstinent at 34-38 wks’ gestation and none would be abstinent if offered control treatment
If no treatment effect existed i.e. no difference between intervention + control groups in proportion of women abstinent at 34-38 wks gestation, what would absolute risk difference be and what would NNT be?
Absolute risk difference would be around zero and NNT would approach infinity
The proportion of women who were abstinent at 34-38 wks gestation was 22.55% in financial incentives group and 8.58% in control. What is the absolute risk difference? How can you work out NNT from this?
0.14 (14.0%).
Therefore, NNT was 1÷0.14 (that is, 7.2)
Define absolute risk difference
Risk difference, excess risk, or attributable risk is the difference between risk of an outcome in exposed group and unexposed group
Why is it essential that the comparator treatment is detailed when interpreting the NNT value?
Because different treatments would provide different values of NNT. For example, in the study of financial incentives in pregnant smokers if the control treatment had been no intervention, rather than usual care, the NNT would have been very different. When not specifying the comparator treatment it is uninformative and potentially misleading
What is the interpretation of a 95% confidence interval?
With probability of 0.95, the true population parameter of NNT will be contained by the given interval. In other words, as few as 5.1 or as many as 12.2 pregnant smokers may have needed to be treated for one additional woman to benefit from intervention vs if she had received control treatment
To enable treatment effects to be fully assessed, it is important that in addition to NNT, what other figues should be reviewed?
Absolute risks, relative risks, odds ratios, and hazard ratios (where appropriate) are also provided
Define the number needed to harm
Estimated number of people that need to be treated with intervention (new treatment) to result in one additional patient to be harmed, compared to if those same people had been treated with control
Number needed to harm normally refers to one of which 2 oucomes?
Adverse events, or intervention less effective than control
Can you compare NNT and NNH between trials?
No, very circumstantial
Sedgewick’s opinion on statistical significance
Just because it is statistically significant, it doesn’t make it important or vice verse, especially when scaling up to a population
