How to read adverse events in a quit-smoking trial

01

Twelve events can come from fewer people than eight

Imagine two fictional groups, A and B, each with 100 participants, all assessed weekly with the same questionnaire during weeks 1–4. Six people in A report 12 episodes of coded event X; eight people in B report eight episodes. The report counts each episode separately and each affected person once.

A has more recorded episodes, but fewer affected people: 6/100 versus 8/100. Twelve episodes are not 12 people or a 12% participant risk. These small invented counts do not establish that either intervention is safer; they simply reveal why the table heading matters.

02

Serious, severe and related answer different questions

ICH E2A distinguishes intensity, described as severity, from seriousness criteria such as an event requiring inpatient hospitalization. It also defines an adverse event without requiring a causal connection to the study medicine. These are reporting definitions, not instructions for judging your own experience.

Return to the fictional record: X has an intensity grade, a separate seriousness field and a separate attribution field. A high-intensity grade alone does not fill in the other two. If the summary says ‘two related events’, request who assessed the relationship, what rule they used and whether they knew the group assignment. Do not change ‘possibly related’ into ‘proven caused by’.

[2]

03

Before comparing, inspect how the records were obtained

CONSORT Harms asks reports to explain how events were sought and defined, repeated-event counting, assessment times, analysis populations and attribution. Apply those questions to the actual methods and appendix, not to the word ‘safe’ in an abstract.

Now change only the fictional B record: it covers 12 weeks and accepts spontaneous reports instead of weekly questionnaires. The original side-by-side comparison no longer has matched collection and time. If only 90 of A's 100 participants supplied records, ask which denominator the analysis used, what is missing and how it was handled; don't silently count the remaining ten as event-free.

[1]

04

A blank cell is not an observed zero

A paper might report no serious events while omitting other assessed events, or list only events above a frequency threshold. CONSORT Harms calls for reporting assessed and detected harms, including serious events and events judged unrelated, with an explanation and access route for omitted outcomes.

Look for an explicit zero, the event definition, assessed population and window. ‘Not reported’, ‘not collected’ and ‘zero observed during this assessment’ are different statements. Even an observed zero does not establish that an event cannot occur outside the studied population or period. Request the missing appendix before summarizing the result.

[1]

05

Keep a useful reading note, not a personal verdict

A note for the invented original table can say: ‘100 assessed per group, same weekly collection in weeks 1–4; A: six people, 12 episodes; B: eight people, eight episodes. Seriousness and attribution need their own fields.’ For a real report, add its citation and unanswered questions. This note does not rank treatments or tell you to start, stop or alter care.

Questions about your own experience belong with a qualified professional. NHS information links UK local cessation services; confirm arrangements with the relevant service. This page collects no symptoms, medicine details or research records. Share necessary personal facts directly under your professional service's privacy procedures.

[3]

06

What to keep in mind

• Name the unit before interpreting a number.
• Check collection, denominator and observation window.
• A missing field, an observed zero and a causal assessment are not interchangeable.