Summary of findings 2. Labelling compared to no intervention.

A Labelling compared to no intervention in children, youth and adults: impact on primary outcomes
Outcomes (follow‐up) No. of clusters or participants No. of studies Certainty of Evidence	Impact
A.1 Traffic‐light labelling
SSB sales (12 months) 42 points‐of‐sale in 2 hospitals in 2 cities 2 ITS studies ⊕⊕⊕⊝ MODERATEa	Boelsen‐Robinson 2017 (units of red‐labelled beverages sold):−56% (95% CI −67 to −45) Hartigan 2017 (share of red‐labelled beverages among all beverages sold): −25 percentage points (P < 0.001)
Revenueb (12 months) 42 points‐of‐sale in 2 hospitals in 2 cities 2 ITS studies ⊕⊕⊝⊝ LOW	Boelsen‐Robinson 2017 (total vending machine revenue):−21% (95% CI −29 to −12) Hartigan 2017 (monthly revenue from beverages): Increase (from USD 34,624 at baseline to USD 35,390 during the intervention, no statistical analyses shown)
A.2 Nutritional rating score shelf‐labels in supermarkets
SSB sales (7 to 11 months) 442 stores from 4 chains in 2 countries 1 ITS and 1 CBA study ⊕⊕⊝⊝ LOW	Cawley 2015 (units of SSB sold):−27.3% (no P value or CI reported) Hobin 2017 (share of beverages with zero stars (mainly SSB), coefficient estimate): −0.026, P < 0.001
Revenueb (7 to 11 months) 442 stores from 4 chains in 2 countries 1 ITS and 1 CBA study ⊕⊕⊝⊝ LOW	Cawley 2015 (total unit sales): −4.9% (95% CI −9.7 to 0.07) Hobin 2017 (total revenue, coefficient estimate):+0.042, P < 0.01
Compensatory consumptionb (7 to 11 months) 442 stores from 4 chains in 2 countries 1 ITS and 1 CBA study ⊕⊕⊝⊝ LOW	Cawley 2015 (number of zero‐star rated items sold per week in the average food and beverage category): −3183 units/week (95% CI −5454 to −913). Hobin 2017 (average star rating of all products sold, coefficient estimate): +0.01 (P < 0.001).
A.3 Menu‐board calorie labelling in chain restaurants and cafés
Beverage calories per transaction (4 to 12 months) 353 stores from 4 chains in 6 cities 1 controlled ITS, 2 CBA studies ⊕⊝⊝⊝ VERY LOWc,d	Bollinger 2011 (beverage calories per transaction): −0.3% (P < 0.01) Elbel 2013 (beverage calories per transaction):No effects (data not shown) Finkelstein 2011 (beverage calories per transaction): +1.7 kcal (95% CI −1.5 to 4.9)
Revenueb (11 months) 316 stores from 1 chain in 2 cities 1 controlled ITS ⊕⊝⊝⊝ VERY LOWc,e	Bollinger 2011 (total store revenue, regression coefficient):+0.005, P > 0.05
Fast‐food restaurant visits* (4 months) 23 stores from 2 chains in 2 cities (1 CBA study) ⊕⊝⊝⊝ VERY LOWc,e	Elbel 2013 (number of fast‐food restaurant visits): +0.9 visits/week (P = 0.07)
Compensatory consumption* (4 to 12 months) 353 stores from 4 chains in 6 cities and counties 1 controlled ITS and 2 CBA studies ⊕⊝⊝⊝ VERY LOWc	Bollinger 2011 (calories from foods and beverages per transaction):−6.0% (95%CI −6.2 to −5.8) Elbel 2013 (calories from foods and beverages per transaction):−3.8 kcal/transaction (95% CI −125 to 119) Finkelstein 2011 (calories from foods and beverages per transaction):+18.5 kcal/transaction (95% CI −11 to 48)
A.4 Emoticon labelling in school cafeterias
Sugar‐sweetened milk (4 months) 186 students in 2 schools 1 ITS study ⊕⊕⊝⊝ LOW	Siegel 2016a (share of students selecting chocolate milk): −16 percentage points (−27 to −4)
Total milk selection* (4 months) 186 students in 2 schools 1 ITS study ⊕⊝⊝⊝ VERY LOWf	Siegel 2016a (share of students selecting any milk):+2 percentage points (no statistical analyses shown)
CBA: Controlled‐before‐after study; CI: Confidence interval; ITS: interrupted‐time‐series study; NRCT: non‐randomised controlled trial; RCT: randomised controlled trial; SSB: sugar‐sweetened beverages.

^aUpgraded for magnitude of effect: We judged the effects on beverage sales to be large, and unlikely to have arisen by chance or through bias.
 ^bOutcomes included as potential adverse outcomes.
 ^cDowngraded for risk of bias: We judged Elbel 2013 and Finkelstein 2011 to be at unclear risk of bias in several domains. In both studies, calories per transaction were substantially lower in the intervention group than in the control group at baseline, and in Finkelstein 2011 the study authors hypothesise that differences in baseline outcome measurements may explain the lack of observed effects. In Finkelstein 2011 the control restaurants were located in counties adjacent to the county where the intervention was implemented. Restaurants in the intervention and control groups may have been frequented by the same customers, leading to contamination, which would have biased results towards null. Elbel 2013 reports only the non‐significance of effects observed for the outcome of interest to this review (calories from beverages per transaction), and may have been underpowered to detect effects for this outcome, which was not the primary outcome of the study.
 ^dDowngraded for indirectness: All three studies report only indirect measures of SSB intake, namely beverage calories per transaction. Moreover, the only study at low risk of bias, (Bollinger 2011), was implemented in a specific setting – Starbucks cafés in New York City – and the generalisability of its results to other settings may be limited.
 ^eDowngraded for imprecision: The 95% CI is large and includes zero.
 ^fDowngraded for imprecision: The study reports that no statistically significant effects were observed without providing an exact effect estimate, P value or 95% CI.