Paperonium

Headache medicine · BMJ · 2010

The summary says headache relief grows 0.11 a week, then gives a range that excludes 0.11.

Results section:tension headachesinside its rangeResults section: migraineinside its rangeThe paper’s summaryresult falls outside its own rangeOur calculation: bothtypes combinedclose to 0.110.110.15 to 0.63should be inside0.0760.15600.20.40.60.8no change with timeadvantage grows
Results section:tension headachesinside its rangeResults section:migraineinside its rangeThe paper’ssummaryresult falls outsideits own rangeOur calculation:both typescombinedclose to 0.110.110.15 to 0.63should be inside0.076 to 0.15600.20.40.60.8no change with timeadvantage grows

How much the drug’s advantage over dummy pills grows per week of treatment

The paper says

Tricyclics, a type of antidepressant, cut headaches more than dummy pills. The paper’s summary says this edge widens by 0.11 for each week of treatment, on a common scale that lets different studies be compared, with a range of 0.15 to 0.63.

Its own data show

The results section gives the weekly figure only for each headache type, and each sits inside its range: tension headaches 0.16 (0.10 to 0.23), migraine 0.09 (0.04 to 0.14). Neither range goes past 0.23, yet the summary’s runs to 0.63. Combining the two ourselves gives 0.116, with a range of 0.076 to 0.156.

A range lists the values the data can’t rule out; it is built around the result, so the result always sits inside it. The 0.11 fits the paper’s own data, but the range printed beside it cannot be right, and the paper never gives the correct one. The conclusion still holds: both headache types show the benefit growing over time.

“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 →

Tricyclic antidepressants and headaches: systematic review and meta-analysis · BMJ (2010) · doi:10.1136/bmj.c5222