Koffman 1998.

Methods	Countries: USA and Canada Setting: aerospace manufacturing worksites in Pomona, Rancho Cucamonga and Ontario Design: 3‐arm quasi‐experimental cluster design with no randomization
Participants	Participation rate not known (185 participants followed up at 6 months). Average age 41, average years smoking 22.7, 41% female. Significant baseline difference in age, years at the company, job description, working with chemicals, years smoking, addiction level.
Interventions	The 3 interventions were assigned to one worksite each. 1. Multicomponent package (M): Self‐help ALA package + group cessation sessions in teams of 5 to 7 + monthly telephone counselling for 12 months and maintenance sessions for weight, fitness and stress management. 2. Incentives (MI): as 1, + incentives: US$15 for abstinence each month during the 5‐month programme; US$5 for ‘fading' (smoking no more than 80 cigarettes) in 1st month. Participants organized into teams, and any US$15 forfeited by an individual was added to US$2500 “super grand prize”. 3 top teams (posted on ‘smoking barometer') at end of programme shared super grand prize (by then US$3960), with the top team winning 50% and 2 teams tied for second each winning 25%. 3. Traditional: Self‐help ACS ‘Fresh Start' manual + 5 x 90‐minute group support sessions + videos.
Outcomes	PPA at 6 months, 12 months. Validation: CO (groups 1 and 2 only)
Notes	Control participants paid out US$20 deposit, refundable on programme completion; incentive participants paid out US$50 non‐refundable initiation fee. Suggests that the multicomponent element may be the key factor for efficacy, though confounded by more thorough evaluation at baseline for multicomponent and incentive than for the traditional group. Relevant comparison for this review is MI versus M. Study was funded by General Dynamics, Air Defense Division (participant companies), and materials donated by ALA and ACS. Originally in C&I.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Non‐randomized. Allocation was “quasi‐experimental”, as 1 worksite would not accept randomization.
Allocation concealment (selection bias)	High risk	See above.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Biochemical validation was not consistently used across groups.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not stated.
Other bias	Unclear risk	Allocation was by worksite, but analysis by individual participant. Incentives participants paid a non‐refundable US$50 initiation fee; this might have skewed or deterred participation. 6‐month assessment took place around the final phase of the reward program, so could more accurately be seen as end‐of‐program, rather than true follow‐up. The traditional group received a different initial assessment from the other intervention groups; however further details of the nature of this difference were not given.