Breast cancer · Clin Cancer Res · 2024
Summary: “significantly” longer survival. Its own test: about 1 in 18, short of its 1-in-20 rule.
How often chance alone gives this survival gap if tumor type doesn’t matter
The paper says
In patients with one kind of breast cancer, a test of which genes are active in each tumor sorted most into two types, named “basal” (64 patients) and “HER2-enriched” (12). The summary’s only survival result says the 12 lived “significantly” longer than the 64.
Its own data show
Its only test of this gives p = 0.055, about 1 time in 18. That is more often than 1 in 20, so it fails the paper’s own rule for “significant”. Elsewhere it calls another result, p = 0.056, “nonsignificant”. We reran the test on the patient-by-patient data published with the paper: 0.055 again.
A p-value is how often chance alone would produce a gap this big if tumor type made no difference. The paper’s rule: “significant” means rarer than 1 time in 20. With only 12 patients, the longer survival (half had died by 1,258 days, versus 510 for basal) is a hint, not proof. Its main findings don’t rest on it.