Heart rhythm · Nat Med · 2025
Three patients harmed in each group, yet the summary’s statistic signals a real difference
How often chance alone would split the harmed patients at least this unevenly
The paper says
A 342-patient trial tested a heart-rhythm procedure with or without an extra step. Its summary, the part most readers see, reports “no difference” in serious harm (1.7% versus 1.8%) yet gives a p-value below 0.0001, which marks a gap chance alone would almost never produce.
Its own data show
Its own safety table gives the counts behind those rates, 3 of 170 patients versus 3 of 172, and p above 0.999, which re-running the paper’s stated test confirms. Even with all six harmed patients in one group, p could not fall below 0.014 (about 1 time in 70), so p below 0.0001 is impossible.
A p-value is how often chance alone, if both versions were equally safe, would split the harmed patients at least this unevenly. Three and three is as even as it gets, so the answer is every time. The words “no difference” are right; the number beside them says the opposite, and the summary can’t say which is wrong.