Reading One Trial's Numbers Three Ways

In practice

A trial reports that a scheduled repositioning protocol reduced pressure injury incidence from six percent to three percent. Set out the two-by-two table first, so the four cells are visible rather than inferred. Absolute risk in the control group is six percent, and in the treatment group three percent. The absolute risk reduction is three percentage points. The relative risk is the ratio of the two, which is one half, so the relative risk reduction is fifty percent. The same data have now produced two impressions: a halving, and three fewer patients in a hundred. Neither is wrong and neither is complete. The number needed to treat is the reciprocal of the absolute reduction, which is thirty-three — thirty-three patients repositioned on schedule for one injury prevented. Now the baseline does its work. On a unit whose own incidence is one percent rather than six, that same relative reduction produces an absolute reduction of half a percentage point and a number needed to treat near two hundred. The relative figure travelled between settings unchanged and meant something entirely different in each. Adverse effects enter the same judgement: disturbed sleep across thirty-three patients is the cost of the one injury prevented.