Paperonium

Diabetes · BMJ · 2020

Prediabetes: the summary’s 7.36 extra deaths is the low end, not the best guess (9.57).

The summarybest guess 7.36, outside itsown rangeIts own method: 73.6 × 0.137.36 is the low end7.369.59 to 12.51should be inside9.5712.517 deaths per 10,000 a year91113
The summarybest guess 7.36,outside its ownrangeIts own method:73.6 × 0.137.36 is the low end7.369.59 to 12.51should be inside9.5712.51791113deathsper 10,000 a year

Extra deaths a year per 10,000 people with prediabetes. Dot: best guess; bar: range.

The paper says

People with prediabetes (blood sugar above normal, below diabetes) die at 1.13 times the normal rate in the general public (heart patients studied separately). The summary turns that into extra deaths: 7.36 a year per 10,000 of them (best guess), likely range 9.59 to 12.51.

Its own data show

Its stated method: extra deaths = death rate with normal blood sugar (73.6 per 10,000 a year, its extra tables) × extra risk. 1.13 times is 0.13 extra (its range, 1.10 to 1.17, is 0.10 to 0.17). 73.6 × 0.13 = 9.57, range 7.36 to 12.51: best guess and low end traded places.

A range spans the values the data allow, so it must contain the best guess. Swapped back, the summary would read 9.59, range 7.36 to 12.51; the small gap from 9.57 fits rounding of 1.13. The same multiplication reproduces the paper’s seven other such figures. The summary’s extra deaths are 23% too low; the 1.13 finding stands.

“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 →

Association between prediabetes and risk of all cause mortality and cardiovascular disease: updated meta-analysis · BMJ (2020) · doi:10.1136/bmj.m2297