Who is included in a quit-vaping study’s percentage?

Find the people behind the fraction

A research summary may describe enrollment in one sentence and outcomes among respondents in another. Those numbers can refer to different sets of people. Look for a participant flow diagram, methods section, or table explaining who contributed to the specific analysis.

Cochrane explains why missing outcomes can introduce bias and why an analysis label alone is insufficient. “Intention to treat” should not be treated as a magic phrase that answers every question about missing data. Read what the authors actually did and retain uncertainty if that detail is unavailable.

Cochrane: Cochrane Handbook, Chapter 8: Assessing risk of bias in a randomized trial

Practice with a made-up arithmetic sheet

The following numbers illustrate denominators only; they are not a study or a success statistic. A sheet begins with 60 rows. Thirty rows have a follow-up answer, and 12 of those answers meet the sheet’s outcome definition. Twelve divided by 30 is 40 percent among the answered rows.

Twelve divided by all 60 rows is 20 percent, but that calculation alone does not identify the outcomes of the other 30. Calling every missing row a success or a failure would add an assumption. A real report needs to explain its chosen analysis and why it is appropriate; you cannot repair it just by selecting whichever fraction looks better.

Ask how unknown outcomes entered the analysis

Check whether the authors excluded missing outcomes, made an explicit assumption, used a statistical method, or tested alternative assumptions. Note the explanation in ordinary language. If you cannot understand its implications, keep the method’s name and ask for qualified interpretation instead of claiming the result is reliable or invalid.

Also look at whether missingness differed across comparison groups and whether reasons were reported. Equal amounts of missing data do not by themselves prove absence of bias. Do not apply an invented acceptable-dropout threshold, or assume that every person who stopped answering had resumed vaping. A missing observation and a known behavior remain different facts.

Cochrane: Cochrane Handbook, Chapter 8: Assessing risk of bias in a randomized trial

Keep the claim tied to the reported analysis

When you summarize a result, state its population and follow-up as well as its denominator. For example: “The quoted proportion is among follow-up respondents; I have not verified the analysis for those without outcomes.” That description is more precise than changing the number into a general probability for anyone who uses a program.

Missing-data questions do not require you to calculate a new treatment effect or decide care yourself. They can become questions for a professional when a health decision depends on the paper. For your own record, the same distinction is useful: lack of an entry does not establish lack of vaping. Keep research interpretation and personal tracking honest without pretending they are equivalent measurement systems.