Ball 2015.

Methods	Design: RCT Timing: Prospective Allocation to group: Random Number of clusters or sites: N/A Number of individuals: 574 Length of intervention: 3 months (+ 6 months additional follow‐up after the end of the intervention)
Participants	General description of participants: Female, adult customers of Coles supermarkets in Melbourne, Australia Age: Adults (age 18 to 60 years was an inclusion criterion, mean age was 44 years in the CG and 43 in the IG) Inclusion criteria: Quote: "Women were required to either hold or be willing to obtain a Coles store loyalty (FlyBuys) card, which was provided to shoppers at no cost, and use this card when they shopped at Coles supermarkets over the 9‐mo study period. Additional eligibility criteria were that women were aged between 18 and 60 y; the main household shopper; able to speak, read, and write English and provide written informed consent to participate; willing to give information about total household income; willing to have their Coles sales data collected and analyzed; and the only woman in their household taking part in the study" Exclusion criteria: N/R Recruitment: Quote: "Coles and Loyalty Pacific Pty Ltd. (FlyBuys) staff (...) composed the sampling pool from which a random sample of 5000 was drawn for a recruitment mailing to participants’ home addresses. At the same time, a media release that targeted local newspapers was undertaken in catchment areas to encourage additional participation" Weight status at baseline: N/R SSB consumption at baseline: Self‐reported intake of SSB at baseline was 13 ml/day in the CG and 13 ‐ 25 ml/day in the 2 IGs. SSB purchases at baseline were 115 ml/day in the IG and 93 ‐ 102 ml/day in the 2 IGs Equity considerations: Quote: "The sample was close to evenly split according to catchment area (44% from low–socioeconomic status and 55% from high–socioeconomic status areas) and educational level (50% tertiary educated). The majority (71%) of women were married, and 53% of women had at least one child living at home. (...) Supplementary analyses were conducted to determine whether intervention effects were moderated by socioeconomic position (by testing 3‐way interaction terms, intervention by time by educational level/household income). Results (data not shown) indicated no strong evidence of moderation by education or income (i.e., intervention effects were consistent across women with low and high education for all 19 outcomes at times 2 and 3 and across low‐ and high‐income women for all time 2 outcomes and all but 3 time 3 outcomes)." 27% of participants in the CG and 25% in the IG were not born in Australia, and 100% were women
Interventions	Intervention: A 20% price reduction on water and low‐calorie beverages implemented through a supermarket loyalty card scheme. (Quote: "[P]articipants in the price‐reduction intervention arm received a 20% price discount on target items, which was applied at the checkout on swiping their FlyBuys card at any Coles store for a 3‐mo period. Participants were sent a list of discounted items (all fruit and vegetables, including fresh, tinned, and frozen, and diet or low‐calorie carbonated beverages or water) at the start of the intervention and midway through the intervention period. This discount was applied over and above any other usual store discounts. Low‐calorie carbonated beverages were included in the discount because carbonated beverages are among the most popular and affordable beverage for Australian consumers with average prices per liter below those of other nonalcoholic beverages") Behavioural co‐intervention: None. (The study had 4 arms: control, price reduction, skills building, and price reduction plus skills building. We report data on the comparison control versus price reduction only) Control: No intervention
Outcomes	Measures of SSB intake: Purchases of SSB assessed continuously throughout the study period with electronic sales data, and self‐reported intake of SSB assessed at baseline, at 3 months (the end of the intervention) and after 6 months additional follow‐up Measures of intake of alternatives to SSB: Purchases of water and diet beverages assessed continuously throughout the study period with electronic sales data, and self‐reported intake of water and diet beverages assessed at baseline, at 3 months (the end of the intervention) and after 6 months additional follow‐up Anthropometric measures: N/R Adverse outcomes: The study does not report how data on adverse outcomes were collected, and if adverse outcomes were observed or not, but notes an increase in the self‐reported intake of SSB in the IG (see above) Other outcomes: Intervention costs, assessed with a comprehensive economic evaluation, reported in a separate publication
Context and implementation	Setting: Supermarkets Sector: Retailing Country: Australia Year(s) when implemented: 2013 ‐ 2014 Mode of implementation: Pilot trial by researchers in co‐operation with a supermarket chain Level of implementation: Setting‐based intervention
Declarations	COI: "None of the authors reported a conflict of interest related to the study" Funding: "The Supermarket Healthy Eating for Life study was supported by a National Health & Medical Research Council Project Grant (594767). KB is supported by a National Health & Medical Research Council Principal Research Fellowship (1042442). SM is supported by an Australian Research Council Future Fellowship (FT100100581). LG is supported by a National Health & Medical Research Council Early Career Fellowship (1035100). (...) Coles supermarkets and the National Heart Foundation of Australia provided in‐kind support for this study" Trial registration: Retrospectivly registered at Current Controlled Trials Registry, Trial ID ISRCTN39432901, www.isrctn.com/ISRCTN39432901 Protocol availability: Published as a separate paper
Notes	This study had 4 arms (control, price reduction, skills building, price reduction plus skills building). We report data on the comparison control versus price reduction only
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "A total of 642 women were randomly assigned to one of 4 conditions by using a computer‐generated block‐randomization sequence produced and implemented by an independent statistician that involved blocks of 4 and 8 in varying combinations stratified by supermarket catchment area"
Allocation concealment (selection bias)	Low risk	Quote: "Allocation concealment was enabled via the secure storage of the randomization sequence separately from the participant database, which was accessible only by the data manager and statistician. Eligible participants were added to the database by the research fellows and assistants who were blinded to the allocation sequence. Only after the baseline survey had been completed and returned with signed consent did the data manager allocate participants to study arms"
Similarity of baseline outcome measurements (selection bias)	High risk	Baseline beverage outcome measurements differed substantially ‐ purchase and consumption data differed by up to 100% between the CG and the 2 IGs of interest. While this was taken into account in the analysis, the large magnitude of the differences suggest that there might have been unobserved baseline differences for which adjustment was not possible. Quote: "Values of sugar‐sweetened beverage purchasing were highly variable at baseline with highest values in the control group; subsequent increases in intervention groups could have reflected a regression to the mean"
Similarity of other baseline characteristics (selection and performance bias)	Unclear risk	CG and IG were compared for socioeconomic status, country of birth, education level, marital status, household annual income, and number of children at home. Substantial differences are notable across all characteristics. The statistical models used did control for most demographic variables, but not for household size: Quote: "[P]urchasing was assessed at the household level, and these servings may have been spread across multiple individuals. It was difficult to adjust for household size in these analyses because some households had a single loyalty card, whereas other households had multiple cards that may or may not have been linked"
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Attrition was approximately 10%, and the analysis was per‐protocol. Baseline beverage purchases and consumption were systematically different among those who were lost to follow‐up compared to those who stayed in the study: Quote: "At baseline, participants who were excluded from compared with those included in the analysis reported lower intakes of (…) tap water (4.4 ± 3.0 compared with 5.5 ± 3.1 servings/d, respectively) and purchased fewer (…) diet beverages (159.6 ± 340.2 compared with 513.6 ± 1324.7 mL/wk, respectively) and more nondiet beverages (1302.2 ± 2496.1 compared with 691.7 ± 1416.0 mL/wk, respectively)"
Blinding (performance and detection bias) Subjective outcomes	High risk	Participants were not blinded to the intervention, and consumption data is self‐reported
Blinding (performance and detection bias) Objective outcomes	Unclear risk	Participants were not blinded to the intervention, and purchasing data could be manipulated by the participants (purchasing data were collected through supermarket loyalty cards, which had to be swiped at the checkout at each visit to a supermarket, and which might have been used selectively. Participants reported in the process evaluation surveys to have made purchases for other persons than themselves. Furthermore, it is possible that participants made stocking purchases during the intervention period to benefit additionally from the price reductions)
Contamination (performance bias)	Low risk	It seems unlikely that the CG received the price reduction intervention
Selective reporting (reporting bias)	Unclear risk	The trial was registered, but only retrospectively. The protocol was published prior to its implementation, and mentions total grocery shopping purchases (collected through FlyBuys loyalty cards) as an outcome, which is not reported in the primary report. In particular, reporting on certain potential adverse outcomes, e.g. compensatory behaviour, might be incomplete due to confidentiality agreements. Quote: "Our agreement with industry partners did not permit us to analyze substitution effects (i.e., whether and how the discount affected the purchasing of nontargeted products)"
Other bias	High risk	The methods used for outcome assessment seem to have been unreliable, as shown by the huge discrepancies between self‐reported beverage consumption and beverage purchases ‐ e.g. in the whole sample at baseline, participants seem to have purchased only 53% of the bottled water they consumed at Coles supermarkets, while they seem to have consumed only 13% of the SSB they bought there. The study authors raise doubts about the reliability of the self‐reported data, and in the process evaluation it is mentioned that participants reported that even with the discount, beverages were more expensive at Coles supermarkets than in other shops, which is why they bought their beverages primarily from other sources. Study authors also mention the possibility that FlyBuy cards were not used for all purchases. Neither the purchasing data from the FlyBuys cards nor the self‐reported consumption data therefore seem to be sufficiently reliable