Paperonium

Diabetes · Diabetes Care · 2009

Aspirin’s heart-attack benefit in diabetes: a range ending at 1.17 leaves out its best guess, 1.19

The summaryends before its best guessIts own numbersholds the best guess1.19×0.82 to 1.17should be inside0.821.730.8×1×1.2×1.4×1.6×2×equal benefit; to the right, less with diabetes
The summaryends before itsbest guessIts own numbersholds thebest guess1.19×0.821.17should be inside0.821.730.8×1×1.2×1.5×2×equal benefit; to the right,less with diabetes

Aspirin’s heart-attack benefit with diabetes versus without. Dots: best guesses; bars: ranges.

The paper says

With diabetes, aspirin cut heart attacks 14% (0.86 times the risk); without, 28% (0.72 times). Comparing them, 0.86 ÷ 0.72 = 1.19; above 1 means aspirin helps less with diabetes. The summary gives 1.19 a range of 0.82 to 1.17 and calls the benefit similar.

Its own data show

The standard formula the paper says it used for comparing two risk numbers, applied to its own table’s 0.86 and 0.72 and their ranges, gives 1.19 with a range of 0.82 to 1.73. It also reproduces the paper’s printed ranges for deaths and for strokes from blocked arteries, so only this top end is wrong.

A range holds the values the data can’t rule out, so it must include the best guess. The top end, times 0.72 (aspirin’s figure without diabetes), gives the worst case with diabetes: 1.17 gives 0.84, still a 16% cut; about 1.73 gives 1.25, no cut at all. “Similar benefit” is still possible, just far less sure than it looks.

“The summary” means the paper’s abstract: the short summary printed at its top, which is the part most readers see.

Read the paper →Full technical write-up →

Aspirin for the Primary Prevention of Cardiovascular Events · Diabetes Care (2009) · doi:10.2337/dc09-1297