It is this simple - say you were measuring a drug’s ability to lower blood pressure.
You would NOT take the blood pressure of all the study participants in each group, sum that number and divide it by the number of participants and compare it by using the same methodology after the intervention/placebo.
Instead, you would measure the change in blood pressure (absolute or %) of each participant and perform a statistical test comparing that change in each group.
When listing entry demographics, you would normally list all initial participants.
When analyzing data, you would only include those that you had both pre- and post- intervention/placebo measurements - anything else is invalid.
That is why entry demographics and study results differ in ANY study that does not retain 100% participation (whether by loss to follow-up, death, exclusion of genetic variants [common in my field], or intolerance of the intervention/placebo), of which almost never occurs.
Entry demographics are included just to show that randomization did not have a statistical bias (outside SD) that might show one group was “different” than the other - all entry demographics here fell easily within SD.
Not to be condescending, but this is all medical stats 101. Now perhaps the presentation could have been clearer, but the numbers are valid as they stand.
Ponder this simple question - if a patient unfortunately died during any study, would it be valid to include their pre- intervention data in your analysis if you did not have their post- analysis data? Well, of course the answer is “no”, unless death itself was a primary outcome measure.
Plain and simple,
EB
