Yong 2015.

Methods	Country: Australia Setting: National phone or web surveys (International Tobacco Control Policy Evaluation Project) Date: Wave 1: September 2011 ‐ February 2012; Wave 2: February ‐ May 2013 Design: Pre‐post longitudinal cohort study
Participants	Nationally representative (random digit dialling) probability sample of smokers aged 18+ (smoked at least 100 cigs in lifetime; smoked at least once in past 30 days). Participants were recruited by telephone (random‐digit dialling), but they could choose to complete the survey by phone or by web Wave 1: n = 1104, Wave 2: n = 1093 (Note: 1525 unique individuals (853 with 1 data point and 672 with 2 data points) who provided a total of 2197 person‐wave observations for GEE analyses) Pre‐ Mean age = 46.24 Post‐ Mean age = 48.48 GEE sample Mean = 47.35 Men: Wave 1: 502; Wave 2: 507 Women: Wave 1: 602; Wave 2: 586
Interventions	IV: own brands vs standardised Branded = own brands before standardised packaging implementation Standardised (plain) = Dark brown‐green colour (Pantone 448C), with the brand name in the same typeface (Lucida Sans) and font size and colour (Pantone Cool Gray 2C). 75% pictorial HW on front, 90% back
Outcomes	[Secondary behavioural]: forgoing cigarettes and avoidance behaviours [Secondary non‐behavioural]: 1) quit intentions. At each wave, assessed smokers’ quit intentions using the question: “Are you planning to quit smoking—within the next month, within the next 6 months, sometime in the future beyond 6 months, or are you not planning to quit?”.2) HWL salience. Assessed using 2 questions: “In the last month, how often, if at all, have you noticed the warning labels on cigarette packages?”; and “In the last month, how often, if at all, have you read or looked closely at the warning labels on cigarette packages?”, both rated on a 5‐point response scale from ‘never’ to ‘very often’. Initial exploratory analyses indicated that the policy changes had different effects on the 2 measures, thus they were used as separate measures rather than combined into a scale. 3) HWL cognitive reactions. Assessed using 3 questions: “To what extent, if at all, do the warning labels make you think about the health risks of smoking?”; “To what extent, if at all, do the warning labels on cigarette packs make you more likely to quit smoking?”; “In the past 6 months, have warning labels on cigarette packages led you to think about quitting?”. The first 2 questions had response options: “Not at all, A little, Somewhat, and A lot” and the last one had: “Not at all, Somewhat, and Very much.” Responses to the 3 questions were combined into a scale by averaging them. 4) HWL behavioural reactions. Assessed using 2 questions, 1 assessing forgoing behaviour: “In the last month, have the warning labels stopped you from having a cigarette when you were about to smoke one?” (Never, Once, A few times, Many times); and the other assessing avoidance behaviour “In the last month have you made any effort to avoid looking at or thinking about the warning labels—such as covering them up, keeping them out of sight, using a cigarette case, avoiding certain warnings, or any other means?” (Yes/No) N.B. Attentional orientation (AO) When you look at a cigarette pack, what do you usually notice first—the warning labels, or other aspects of the pack, such as branding?” Analysis summary: Smokers’ reactions and avoidance orientation (AO) to health warnings (HWLs) pre‐implementation and post‐implementation of the standardised packaging and enhanced health warnings law, were computed for descriptive purposes using weighted data. GEE models were employed to examine pre–post changes by testing for significant main effect of survey wave while controlling for sociodemographic and smoking‐related variables. Dichotomous outcome variables such as avoidance and AO were modelled using binomial distribution with logit link function. Outcome variables such as noticing, reading, cognitive reactions, forgoing and quit intentions were treated as quasilinear and modelled as continuous variables using Gaussian distribution with identity link function as initial exploration indicated that these variables when dichotomised were less sensitive in detecting an effect due to loss of information. Parameters were estimated using unstructured correlation structure with robust variance estimation procedure. GEE modelling of pre–post changes was limited to smokers only (both recontacted and newly‐recruited smokers) at both survey waves, as ex‐smokers are less likely to be exposed to the pack HWLs. To examine whether the pre–post changes differed by AO patterns, difference scores were employed as outcomes and linear regression analyses conducted (since the difference scores were generally normally distributed) to test for group differences in outcomes by regressing the difference scores onto a dummy variable used to represent the 4 different patterns of change across waves in AO towards the HWLs (i.e. brand‐brand; brand‐warning; warning‐brand and warning‐warning). For ease of interpretation, a relevant subgroup was chosen as the reference group for comparison purposes. This set of analyses included only smokers who provided data on both survey waves. To assess effects of attrition, baseline differences were examined in covariates between those retained and lost and found those lost to the study were more likely to be highly educated, complete a phone survey and be recruited into the study in the year before the baseline wave. These variables were controlled for in all regression analyses. Finally, additional GEE analyses were conducted to examine associations of upstream HWL reactions and AO with warning‐stimulated cognitive reactions (midstream outcome) and quit intentions (downstream outcome), to determine whether the strength of the associations differed between pre‐policy and post‐policy implementation by testing for any significant interactions between survey year and reactions on the outcome of interest
Funding source	“The ITC Four Country Survey is supported by multiple grants including R01 CA100362, P50 CA111236 (Roswell Park Transdisciplinary Tobacco Use Research Centre), P01 CA138389 (Medical University of South Carolina), P30 CA138313 (Hollings Cancer Center Support Grant, Medical University of South Carolina) and an ITC pilot study grant (Medical University of South Carolina), all funded by the National Cancer Institute of the USA, Robert Wood Johnson Foundation (045734), Canadian Institutes of Health Research (57897, 79551), National Health and Medical Research Council of Australia (265903, 450110, APP1005922), Cancer Research UK (C312/A3726), Canadian Tobacco Control Research Initiative (014578) and Centre for Behavioural Research and Program Evaluation, National Cancer Institute of Canada/Canadian Cancer Society.”
Conflicts of interest	“KMC has served in the past and continues to serve as a paid expert witness for plaintiffs in litigation against the tobacco industry. GTF and JFT have each served as a paid expert witness or consulting expert for governments in countries whose policies are being challenged by parties under trade agreements. DH has served as an expert witness on behalf of national governments in legal challenges to packaging regulations, as well as an advisor to regulatory agencies for tobacco packaging policies. RB was a member of an expert advisory committee that advised the Australian government on the research done to support the introduction of the plain packaging legislation.”
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Selective reporting (reporting bias)	Low risk	Comment: objectives as expected and reported
Sampling Method	Low risk	Comment: random‐digit dialling, could be completed by phone or web
Measurement of independent variable	Low risk	Comment: The date of the implementation of standardised packaging was known and well enforced
Measurement of dependent variable	Low risk	Comment: used commonly‐used measures
Control for confounding	High risk	Comment: Enhanced pictorial warnings were implemented at the same time as standardised packaging so it is difficult to separate the effects. Hence confounding rated high even though other factors had been controlled for.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "To assess effects of attrition, we examined baseline differences in covariates between those retained (n=788) and those lost (n=316) and found those lost to the study were more likely to be highly educated (p=0.04), complete a phone survey(p<0.001) and be recruited into the study in the year before the baseline wave (p=0.006). These variables were controlled for in all regression analyses." Comment: Controlled for differences between those followed up and those not in analyses
Statistical methods	Low risk	Comment: Appropriate