Schumann 2008.

Methods	Setting: community, Germany Recruitment: from participants in a general population health examination survey
Participants	847 smokers (ex‐smokers in study not included here); 46% female (full sample), average age 44 years (full), average cpd 15 Controls more likely to be in preparation (32% vs 20%) and to have past year quit attempt
Interventions	∙ Assessment only (completed questionnaires on 3 occasions) ∙ Expert system ‐ tailored 3 to 4‐page letter and 8 to 26‐page SoC‐matched booklet at 0 months, 3 months, and 6 months
Outcomes	Abstinence at 24 months; sustained 18 months' follow‐up (other measures of abstinence also reported) Validation: none
Notes	Tailoring ‐ 67% got 3 letters, 21% 2, 13% only 1 72% reported reading some materials
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Each participant was assigned a unique computer‐generated random number between 0 and 1; the data file was sorted by ascending random numbers; participants were then consecutively assigned to the 3 study conditions
Allocation concealment (selection bias)	Low risk	No opportunity to alter allocation or exclude
Blinding (performance bias and detection bias) All outcomes	Low risk	Although no biochemical validation, written contact only; participants in control group do not appear to have known of intervention; all participants engaged in long‐term questionnaires re smoking status, so differential misreport judged unlikely
Incomplete outcome data (attrition bias) All outcomes	Low risk	Somewhat greater loss in intervention (34%) than in control (27%) Meta‐analysis includes those lost as smokers Study authors report that generalized estimation equation gave similar results