Cancer treatment · BMC Med · 2024
A blood-cell ratio predicts death, says the paper. Its own range says it could be chance.
How often chance alone would give a link this strong, if none existed
The paper says
In 173 cancer patients with hormone-gland side effects from immunotherapy, the summary says a higher ratio of platelets (clotting cells) to lymphocytes (immune cells) predicts death: 0.4% higher odds per point, p = 0.006, passing its 0.05 cutoff. The ratio is typically 145 (Table 3).
Its own data show
The same summary sentence, the main text and Table 5 also print a range of likely values: 0.7 to 1.4 times the odds per point (30% lower to 40% higher). It includes no change, implying p = 0.98. A p of 0.006 needs a range of 0.1% to 0.7% higher. Rounding can’t explain the gap.
A p-value is how often chance alone would produce a link this strong if there were none; the paper’s cutoff is 0.05 (1 time in 20). The range comes from the same calculation, so a range spanning no change always misses that cutoff. At least one printed number is wrong, so readers can’t tell if the ratio predicts death.