All frameworks and cheat sheets

Key figure

Only 8 of 359 trials reported number needed to treat

Stat card: 8 of 359 randomised drug trials in five leading medical journals reported the number needed to treat, and 18 reported absolute risk reduction.
If the top medical journals rarely report it, a product review deck almost never will. Source: Nuovo, Melnikow and Chang, JAMA 287:2813-2814, 2002
Download PNGRead the full guide

A stat card from Nuovo, Melnikow and Chang, who reviewed four years of Annals of Internal Medicine, the BMJ, JAMA, The Lancet and the New England Journal of Medicine. Of 359 randomised drug trials with a significant effect, 8 reported the number needed to treat and 18 the absolute risk reduction. It is a useful reminder for product teams that effects are almost always shown as relative lifts.

Where this appears

What's in this key figure

8 of 359 randomised drug trials in five leading medical journals reported the number needed to treatAbsolute risk reduction was reported in 18. These were trials with a significant treatment effect, reviewed across four years.

Use this image

Free to use in your docs, decks and posts under CC BY 4.0, as long as you credit Koji with a link to this page. To embed it, paste this into your page:

html
<a href="https://www.koji.so/visuals/trials-reporting-number-needed-to-treat-stat"><img src="https://www.koji.so/visuals/trials-reporting-number-needed-to-treat-stat/trials-reporting-number-needed-to-treat-stat.png" alt="Stat card: 8 of 359 randomised drug trials in five leading medical journals reported the number needed to treat, and 18 reported absolute risk reduction." width="1200" height="669" /></a>
<p>Source: <a href="https://www.koji.so/visuals/trials-reporting-number-needed-to-treat-stat">Only 8 of 359 trials reported number needed to treat</a>, Koji</p>

Related

Checklist of seven steps for reporting a product effect as a number needed to treat, from computing the absolute reduction to interviewing non-responders.

Number needed to treat: a reporting checklist

Table showing one onboarding result, churn down from 12% to 9%, as a relative reduction, an absolute reduction, a number needed to treat and an annual count.

One churn result, four ways to report it

Stat card: 47% of unique usability problems were found by only one of four evaluators who analysed the same unmoderated study.

The evaluator effect: 47% of problems found by one evaluator