Zacher 2014.

Methods	Country: Australia Setting: 25 café strips: 18 Melbourne suburbs; 7 Adelaide suburbs Date: mid‐October 2012 ‐ mid‐April 2013. (Pre‐study carried out mid‐October 2011 ‐ mid‐April 2012) Design: Observational study convenience sample pre and during/post standardised packaging (PP) implementation
Participants	No recruitment of participants: passive observation only. Selection of locations: For the pre‐standardised packaging phase, street segments were selected (referred to as ‘café strips’) from a range of socioeconomic areas in Melbourne and Adelaide that were known to have many popular cafés, restaurants and bars. Fieldworkers sampled every venue in their assigned café strip/s which had outdoor seating visible from the footpath. New venues were added to the sample if they had opened between phases. 520 unique venues of which 480 venues allowed smoking and had patrons present at least once in either phase. At least 1 patron was present for 2391 observations pre‐PP and for 2219 observations post‐PP (total n = 4610)
Interventions	Branded = pre‐PP, branded packs and 30% front‐of‐pack warnings and 90% back‐of‐pack warnings. Standardised (plain) = 75% front‐of‐pack warnings and 90% back‐of‐pack warnings. Set of 14 HW: 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). New set of 14 warnings divided into 2 sets to be rotated after 12 months (so 7 new warnings during the post‐ phase of this study)
Outcomes	[Secondary behavioural): At each café strip venue, fieldworkers recorded observations using notepads/smartphone data collection applications, including: (1) number of people smoking or holding/rolling/lighting a cigarette. Package display was recorded slightly differently in each phase. Pre‐standardised packaging , fieldworkers noted: (2) number of packages visible on the table; (3) number of packages orientated (a) face‐down; (b) standing on their side, top or bottom; (c) in a case or tin (not the original packaging); (d) completely concealed by a telephone, wallet or some other object, so that the fieldworker was unable to ascertain its orientation; (e) with an unknown orientation (i.e. too far away/inadvertently obscured) Post‐standardised packaging phase: fieldworkers noted the same things, but separately for fully‐branded packs, PPs and packs of unknown packaging. 5c ‐ 5e classified as unknown packaging Analysis summary: Multi‐level Poisson models were employed to test the effect of phase (pre‐ or post‐standardised packaging) on the prevalence of pack display among patrons. Random intercepts were included for café strip and venue in all models to adjust SEs for correlations among observations within the same venue and among venues within the same café strip. In order to analyse the data as the rate of packs to patrons, the count of patrons was used as an offset term, meaning that for an observation (i.e. 1 for each venue in each wave) to be included in the analysis at least 1 patron had to be present. The rates of smokers to patrons and packs to active smokers were tested similarly, using appropriate offset terms (patrons and active smokers, respectively). The analysis of packs to active smokers excluded observations where no active smokers were recorded, due to the offset term. For each outcome, unadjusted models were first run, followed by models adjusting for city, SES, presence of children, month, day/time, temperature and wind speed. A series of analyses were then conducted (using the adjusted models) to examine interactions between phase and city, SES, presence of children and day/time to determine whether or not the rates of packs to patrons and smokers to patrons changed more between phases in some situations than others. To determine whether any effects were absent in October/November (when plain packs were first emerging onto the market) but present or stronger in December onwards (when all packs sold were required to be plainly packaged), a 2‐category variable was also created for month, comparing October/November to December–April observations, and tested its interaction with phase. Multi‐level Poisson models were then used to examine whether phase had any effect on face‐up pack orientation, concealment or external case use. Compared rates of face‐up orientation and pack concealment among fully‐branded packs pre‐standardised packaging to rates among plain packs post‐standardised packaging, excluding packs in external cases and packs in unknown orientations from the total count of packs pre‐standardised packaging to ensure that rates had comparable denominators in both phases. Only observations for which at least 1 known‐orientation fully‐branded (pre‐ PP) or plain (post‐PP) pack was recorded were included in these analyses due to the offset term. The rate of case use was analysed for all observed packs in both phases, because the authors could not determine whether the original package was fully branded or standardised; accordingly, at least one pack had to be observed for an observation to be included in the analysis. Interactions between phase and covariates were also examined. A sensitivity analysis was then conducted, limiting the sample to venues which were open for business in both the pre‐ and post‐PP phases. The analytical procedure mirrored the steps for the main analysis
Funding source	“This study was funded by Cancer Council Victoria, Cancer Council South Australia, and Quit Victoria.”
Conflicts of interest	“None”
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Selective reporting (reporting bias)	Low risk	Comment: This built on a previous study (Wakefield 2013) and objectives for this study were as expected
Sampling Method	High risk	Quote: “in October 2011, we selected a convenience sample of 25 cafe strips'...'Our results were consistent even when limiting the sample to stores that were observed in both phases, suggesting that the findings were not biased by inclusion of slightly different stores before and after plain packaging” Comment: Convenience sample. Selection of location remained consistent over the pre‐ and post‐ time periods. Findings did not change when new stores were included
Measurement of independent variable	Low risk	Comment: Australia, law enforced
Measurement of dependent variable	Low risk	Quote: “We used Krippendorff’s alpha, which is valid for count data, to calculate inter‐rater reliability for numbers of patrons, active smokers, packs and packs orientated face‐up, and the presence of children. The alpha scores for most outcomes were high (patrons: α = 0.998; smokers: α = 0.897; packs: α = 0.895; presence of children: α = 1.000). For face‐up pack orientation, an acceptable alpha of 0.795 was achieved overall, and limiting the analysis to venues with consistent observations for number of packs observed resulted in an alpha of 0.881.“ ….“The strength of this study is that, unlike survey questions on pack display behaviour, our measures are objective and not subject to recall or social desirability biases. Observational methods similar to those used in this study have been shown to be accurate [23–25], and our measures of inter‐rater reliability were acceptable to high.” Comment: Methods for observations appear sound and reliability high
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: “Our results were consistent even when limiting the sample to stores that were observed in both phases, suggesting that the findings were not biased by inclusion of slightly different stores before and after plain packaging” Comment: When analyses were restricted to those present in both phases of data collection, the results were consistent.
Statistical methods	Low risk	Comment: Appropriate