Why do quit-smoking results at three and twelve months differ?

01

Put the assessment date and the question's time window on separate clocks

The SRNT methods review distinguishes a study's time origin from its abstinence duration. Follow-up may run from a target quit day or intervention initiation; randomisation can be another reported reference. Twelve months from one event is not twelve months from another. Locate the actual anchor instead of supplying it from the headline.

Then find the period being assessed. A seven-day snapshot collected at twelve months is a late short window, not twelve months of continuous abstinence. Also note when support was delivered or ended: an assessment while support continues and one after it ends place the outcome in different contexts. A service with ongoing access may have no single intervention-end date.

[1][2]

02

One cohort, two visits, the same short-window question

This is an invented arithmetic exercise, not a trial or a treatment result. The same original 100 people are assessed three and twelve months after randomisation. At both visits, the endpoint is no cigarette smoking in the preceding seven days. Assume every available status is correctly recorded.

At month three, 24 meet the endpoint, 66 do not and 10 are unknown. At month twelve the counts are 18, 62 and 20. With unknown outcomes coded as non-abstinent and the denominator kept at 100, the reported proportions are 24% and 18%: a decrease of six percentage points. The later unknown count is larger in this exercise; that is not a rule for all studies.

03

Why six percentage points cannot be renamed ‘six relapses’

The 18 later people might all belong to the earlier 24. Alternatively, only eight might belong to that earlier set, with ten other people meeting the later endpoint. Both possibilities fit the same totals. In the first, six earlier meeters are outside the later meeting set; in the second, sixteen are outside it.

Being outside that set is still not proof of resumed smoking: some later outcomes are unknown. Nor do two seven-day windows show what happened between visits. Someone could smoke between them and meet the final window again. These are possible arrangements of fictional counts, not verified histories or estimates of transition probabilities.

A defensible sentence is: ‘The invented reported proportion fell from 24% to 18% under this coding rule; the totals do not identify the number who resumed smoking.’ No causal effect, confidence interval or personal forecast is provided.

04

Is the longest follow-up always the best result to quote?

Later assessments can inform what is observed after more time, including after an intervention ends, but their usefulness depends on the research question and endpoint. A late short-window result still has short-window limits. Reaching participants may be harder later; this possibility is not evidence that every later result has more missing data or more bias.

For snapshots, people can newly meet the endpoint at a later visit, so the proportion need not fall. Under a fixed-start, no-smoking continuous definition, extending the interval adds a requirement instead; that is a different endpoint logic, not a reason to treat snapshots as continuous histories. Neither logic alone dictates how a between-group effect must change.

Locate the prespecified primary assessment and show relevant other time points in their proper roles. Do not replace it with whichever visit looks most impressive, or choose twelve months merely because the label sounds more rigorous. Counts, precision, actual support arrangements and missing-data assumptions still matter.

[1][2][3]

05

Read across time without inventing a trajectory

At each reported visit, keep the time origin, endpoint window, group-specific numerator and analysis denominator together with the numbers whose outcomes were available or missing. Check reasons and handling of missing data at that particular time. A person missing one visit need not have no usable data at every other visit; a flow diagram alone may not answer each outcome-time question.

If the report claims individual transitions, look for linked assessments or event-history data and the stated relapse definition. If those are absent, limit your summary to the reported visit-specific proportions. For a comparison of studies, differing anchors, windows or support periods prevent the month labels alone from establishing comparability. No search for participants' identities is needed.

[1][2][3]

06

Study clocks are not deadlines for your own support

In England, NHS Better Health points to local Stop Smoking Services for personal support. Ask a qualified professional about your circumstances and confirm local arrangements. A study's three- or twelve-month visit is not a deadline by which a reader must succeed.

[4]

07

What to keep in mind

• A twelve-month visit may ask only about the previous week.
• A change in totals is not a count of individual relapses.
• Preserve the intended assessment, available data and support context.