1. Summary of PICOS and of AMSTAR scores of included systematic reviews, and how existing reviews informed the PICOS of a new Cochrane Review.

Domain	Finding	How it informed our review question or methods
Setting	12 reviews did not specify the setting 11 reviews stated the community as the setting 3 reviews stated the setting was LMICs 3 reviews specified a school as the setting	We chose the community as the setting, defined as a group of people with diverse characteristics who were linked by social ties, share common perspectives and engage in joint action in geographical locations or settings (MacQueen 2001).
Participants	5 reviews did not specify the types of participants for inclusion 11 reviews included infants and children (up to school‐aged children) 1 review included adults and adolescents 6 reviews included pregnant women or mothers in the immediate postpartum period. 1 of these also targeted other adults who could be linked to women who may have breastfed. Many of these were assessing interventions on breastfeeding or complementary feeding. 1 review included only parents of children aged 2–5 years, as it assessed influence of parenting practices on children's dietary habits 2 reviews included all people living in a community 3 reviews included only poor people who were recipients of some service, e.g. recipients of a government conditional cash‐transfer programme	As existing reviews specifically addressed specific high‐risk groups, we did not focus on these. Instead, we included all individuals across all ages that belonged to the community where relevant interventions had been implemented.
Intervention (including its duration)	14 reviews addressed interventions related to the availability of food, 5 of which also assessed interventions influencing utilisation of food, such as nutrition education 13 reviews assessed interventions addressing food utilisation 7 reviews assessed interventions addressing access to food (2 of which had a low AMSTAR score of 4) 28 reviews did not specify the duration of the intervention, and only 1 included interventions with a minimum duration of 3 months. As a result, the duration and the follow‐up times of the interventions varied considerably within and across reviews	Of the 14 reviews that addressed food availability, 5 also assessed food utilisation (e.g. combination of community gardens and nutrition education). As fewer reviews addressed food access, we included interventions that had addressed this dimension of food security. We included interventions with any duration but extracted outcomes that were measured ≥ 3 months after implementation.
Control	18 reviews did not specify a control group 6 reviews compared the intervention with either no intervention, an alternative intervention or placebo 3 reviews did not have any control group 2 reviews stated that included studies needed to have a control group, but did not specify further	We included studies in which these interventions, individually or in combination, were compared to no intervention or to other eligible intervention.
Outcomes assessed	The specific outcomes assessed across the included reviews varied considerably and often they were not clearly specified at the outset. The most common and important outcomes reported in these reviews were related to dietary intake, anthropometric measurements, and biochemical and clinical indicators, to describe the impact of the intervention on nutritional status. Other outcomes measured included food purchase or expenditure, food production, morbidity and mortality, and breastfeeding initiation rates or duration. Often, reviews measured the same outcome in different ways. For example, anthropometric indicators assessed differed, as did their classifications, across the included reviews. This makes it difficult to compare results across reviews and to reach a conclusion about the effectiveness of a specific intervention.	The most commonly specified outcomes measured food and nutrition security, and nutritional status. We also focused on these outcomes. Examples included: diet diversity scores and hunger measures; and anthropometric, biochemical and dietary intake indicators. We clearly defined, a priori, the specific outcome measures and metrics that we included in our review.
Study designs	11 reviews did not specify which study designs they would include 3 reviews included only RCTs 1 review included only CCTs 1 review included only impact evaluations 13 reviews included a variety of study designs, which included ≥ 2 of: RCTs, BAS, quasi‐RCTs, analytical cohort studies, ITS, CCTs, randomised field trials and CSS However, the definitions of the study design labels used were not always clear and varied across the included reviews.	The study design labels used varied across included reviews and were not always clearly defined. We included both randomised and non‐randomised studies, as we expect that existing RCTs in the area of food security would be scarce. We wanted to include the best available evidence for our review question. We clearly defined the type of study designs included in our review.
Search strategies	Most reviews ran comprehensive searches. They used a comprehensive set of keywords and searched a variety of relevant databases. Only 5 reviews did not indicate search terms either in the text or in an appendix. 2 reviews conducted searches until 2012 11 reviews searched until 2010–2011 9 reviews searched before 2010 7 reviews did not specify the date of the last search	Our review included updated searches across a variety of relevant databases and websites. We drew on common keywords used across these included reviews.
Reporting	The methods sections of most reviews were often not reported clearly. The reporting of results in these reviews, in terms of characteristics of included studies, was also poor.	Poor reporting of the characteristics of included studies makes it difficult to assess the context in which these results were obtained. Thus, it is difficult to generalise the results. We clearly reported on the characteristics of included studies, so that the context in which the interventions were implemented was clearly understood.
AMSTAR scores	9 reviews were of low quality (AMSTAR score: 0–4) 11 reviews were of moderate quality (AMSTAR score: 5–8) 8 reviews were of high quality (AMSTAR score: 9–11) 1 review did not have a score as it did not include any studies	Of the 8 high‐quality reviews, 5 assessed interventions that aimed to improve food availability or utilisation (or both), and 3 assessed interventions addressing food access. The other 2 included reviews that addressed food access were of low quality (AMSTAR = 4). We contributed to the evidence base on interventions addressing food access by producing a high‐quality systematic review that assessed the effectiveness of the interventions on relevant outcomes, such as nutritional status.

BAS: before‐and‐after study; CCT: controlled clinical trial; CSS: cross‐sectional study; ITS: interrupted time series; LMIC: low‐ and middle‐income country; RCT: randomised controlled trial.