Diabetes · Diabetes Care · 2021
Only 2 deaths, yet a margin of error nearly 4× too narrow makes transplants look protective.
Deaths per 1,000 patients a year; bars span rates the data can’t rule out
The paper says
Of 49 type 1 diabetes patients given insulin-making cell transplants, 2 died. The paper gives 3.28 deaths per 1,000 patients a year (margin of error 2.12 to 5.05), lower than in three earlier groups of type 1 patients, and suggests transplants may protect against death.
Its own data show
The rate matches its own counts: 2 deaths in 610.1 years of follow-up, all patients combined. But 2 deaths leave lots of room for luck: the standard calculation for small counts gives 0.40 to 11.84, nearly 4 times as wide, covering all three earlier rates (5.31 to 10.79). The paper’s margin fits about 20 deaths.
A margin of error is the span of death rates the data can’t rule out, and the fewer the deaths, the wider it is. If these patients truly died at the lowest of the three earlier rates, 5.31, 2 or fewer deaths would still happen more than 1 time in 3. So the data can’t show transplants protect against death.