Interventions in Small Island Developing States to improve diet, with a focus on the consumption of local, nutritious foods: a systematic review

Introduction Food security in Small Island Developing States (SIDS) is an international policy priority. SIDS have high rates of nutrition-related non-communicable diseases, including obesity and type 2 diabetes, micronutrient deficiencies and, in many, persistent childhood stunting. This is associated with an increasing reliance on imported processed food of poor nutritional quality. Calls have been made for strengthening local food systems, resilient to climate change, to increase the consumption of nutritious locally produced food. We aimed to systematically review interventions intended to improve diet in SIDS, and specifically explore whether these interventions applied a local food approach. Methods The search strategy was applied to 11 databases, including in health, social science and agriculture. Screening of titles, abstracts and data extraction was undertaken in duplicate. Risk of bias was assessed using Cochrane tools. Narrative synthesis of the results was undertaken. The study protocol was registered (PROSPERO registration number: 2020CRD42020201274). Results From 26 062 records, 154 full texts were reviewed and 24 were eligible. Included studies were from the Caribbean, Pacific, Mauritius and Singapore. Five were a randomised study design, one an interrupted time series analysis, eight controlled and ten uncontrolled pre-test and post-test. Nine studies included some aspect of a local food approach. Most interventions (n=15) included nutrition education, with evidence of effectiveness largely limited to those that also included practical skills training, such as vegetable gardening or food preparation. Three studies were considered low risk of bias, with the majority (n=13) of moderate risk. Conclusion There is a lack of robust evidence on interventions to improve diet in SIDS. The evidence suggests that multifaceted approaches are likely to be the most effective, and local food approaches may promote effectiveness, through mechanisms of cultural and contextual relevance. Further development and evaluation of interventions is urgently required to increase the comparability of these studies, to help guide policy on improving nutrition in SIDS.


Supplementary
Nonrandomised controlled before-after study

Marine protected areas
Dietary intakeenergy, protein, fat/ 24 hour recalls and food consumption survey.
Dietary intake -members of villages with effective MPAs had higher energy and protein intakes (particularly marinederived protein) than those that did not have MPAs or had ineffective ones. Poorest dietary protein/fat ratio was found in the village with the worst MPA governance (attributed to high dependency on cash that subsistence economy for livelihood security and to availability of cash rather than effective marine governance and management. Dietary intake -statistically significant higher mean calcium frequency scores post-intervention (p <0.001).
Suggest that using a theory based (health belief model) nutrition education program and focusing on consumption of calcium rich foods (foodbased approach) over calcium supplementation contributed to the success of the program and BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Dietary intake -no consistent evidence for improved dietary intake in the intervention group. Some evidence that aspects of diet worsened compared to the control group; SSB intake (OR (adj) 1.69 (p=0.005)), snacking after school (2.21 (0.001)), eating takeaway or fried foods after school (2.76(0.001)), breakfast consumption (0.63 (0.02)). However, reported purchasing of snack food from shop or take away reduced (OR 0.55 (p 0.001)). Dietary intakesignificant increase in local fruits consumption (1.2 to 2.9 (p<0.001) mean days/week), local vegetables consumption (2.8 to 4.6 (p<0.001), and imported vegetables (0.7 to 2.0 (p<0.001). Local fish and seafood consumption increased from 2.5 to 4.4 (p<0.001) mean days/week with no significant change in imported sources. There was a significant increase in local and imported meat sources (p 0.004; 0.03) and increase in both local and imported sources of starch/flour products, however the local source was not a significant increase (p 0.17; 0.06). There was also a significant increase in Focus on more familiar and less sensitive traditional crops already planted in the village. Emphasise the importance of education and supervision as component of this type of intervention.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)  Purchase -foods high in sugar.

Online hypothetical grocery store
Purchase -a text-only warning label generated a statistically significant reduction in labelled products purchased. None of the secondary outcomes (total sugar purchased (g), sugar purchased per dollar spent (g per $), total spending ($) and total expenditure on high-in-sugar products ($)) were statistically different across groups.) S2 Nonrandomised controlled pre/post-test study Nutrition education (also PA component).

Preintervention
Dietary intakedaily energy and fibre, fruit and vegetable.
Dietary intake -no significant changes between baseline and post intervention kcal intake, fruit, vegetable or fibre consumption. S8 Clusterrandomised parallel-group trial Health-focused support meetings to promote fruit and veg consumption (also promoted PA and cervical screening).

Preintervention
Dietary intakedaily and 7 day fruit and vegetables.
Food frequency questionnairebased on previous 7 days. local vegetables consumption (2.8 to 4.6 (p<0.001), and imported vegetables (0.7 to 2.0 (p<0.001). Local fish and seafood consumption increased from 2.5 to 4.4 (p<0.001) mean days/week with no significant change in imported sources. There was a significant increase in local and imported meat sources (p 0.004; 0.03) and increase in both local and imported sources of starch/flour products, however the local source was not a significant increase (p 0.17; 0.06). There was also a significant increase in imported sugary drinks and foods (p 0.011; 0.04).

S15
Nonrandomised uncontrolled pre/post test study sat fat, trans fat, carbohydrate, protein). % consuming < EAR was significantly lower in the participant group for six out of 17 vitamins and minerals, when adjusted for energy intake; Vitamin A, B1, Folate, Iron, Magnesium, Zinc, and mean % AI for calcium was significantly greater in participants when adjusted. Authors considered study didn't address high intake of fat and sat fat or sodium which is concern.
Multi-level intervention S20 Nonrandomised controlled before-after study
Nutrition knowledge, attitude and behaviour -salt related.

STEPS survey
Dietary intake -no significant difference in mean salt intake (P=0.588) as measured by 24-h urinary excretion after raking the samples and adjusting for potential confounders.
Nutrition knowledge attitude and behaviour -significant increase in the proportion of participants who understood that high salt consumption could cause serious health problems (from 81 to 90%, P=0.049); decrease in always or often add salt to food when eating (from 50 to 33%, P=0.002) and always or often ate processed foods (from 60 to 49%, P=0.020); increase in using one or more methods to control salt intake (from 73 to 93%, P<0.001), particularly through the use of spices rather than salt in cooking (from 48 to 76%, P<0.001); no change in the proportion who knew the recommended daily salt intake was less than 5 g (22 vs. 20%, P=0.638) or who thought that they consumed too much salt and that lowering salt was important in their diet (P>0.183 for each).; reductions in the proportion of the population who BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Key: green = low risk, yellow = moderate/some concern risk, red = high/serious risk, blue = critical risk. *indicates the studies which applied a local food approach.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

Effectiveness of local food approach to outcome-related components
All nine studies in Table 2 applied a local approach to outcome-related components. This included promoting locally produced food and traditional dietary behaviours such as traditional cooking techniques or foods of cultural significance (n=8), (S3, S5, S6, S13, S14, S20, S21, S24) or applying locally-relevant tools to outcomes measures. (S9) Five of these interventions included a practical food production component, such as teaching skills for planting and harvesting own produce. (S6, S13, S14, S20, S21) Of the eight studies that specifically promoted locally produced food, four showed significant improvements in dietary intake and one study that did not measured dietary intake, showed significant improvement in nutrition knowledge. Two of the effective interventions that demonstrated improved dietary intake were garden-based nutrition interventions. One targeted pregnant women and children in rural Dominican Republic to improve consumption of locally produced food rich in Vitamin A. (S13) The intervention provided garden-based nutrition education, vegetable gardens and laying hens and assessed intake using a FFQ, to significantly increase consumption of any vitamin A rich food, and garden specific vitamin A rich foods in intervention than control communities. The second study aimed to encourage the return to traditional diets to prevent disease in Pohnpei (Federated States of Micronesia), through growing local food using container gardening, and classes on cooking and using charcoal ovens. (S14) This mixed-methods study took a unique approach in focusing on food source as part of their context-specific assessment of dietary intakedistinguishing between consumption of local and imported foods, and found significant increases in the consumption of local and imported fruit and vegetables, local and imported meat, local fish and seafood, imported starch/flour, and imported surgery drinks and foods (Supplementary Table 1).
A third effective intervention that used a local approach to improve dietary intake, was the implementation of Marine Protected Areas (MPAs) in the Solomon Islands. (S24) The study aimed to assess the impact of protecting fish stocks on food security for local communities. Authors observed variation in effectiveness of MPAs in protecting fish stocks, but residents in villages with effective MPAs had higher energy and protein intakes (particularly marinederived) than those that had ineffective or no MPA. Poorest dietary protein/fat ratio was found in the village with the worst MPA governance, and this was attributed to availability of and high dependency on cash over subsistence for livelihood security, rather than effective marine governance and management.
The final local intervention that was effective in improving dietary intake, was a quasiexperimental study in Mauritius that applied an educational approach to increase consumption of locally produced sources of calcium such as milk, fish and dark green leafy vegetables, amongst adults aged 40+ years. (S3) Three other studies included a practical local food production component, similar to the first two garden-based studies described, but these were not shown to effectively improve diet. One of these studies was conducted in Guam, targeted school-aged children and involved the implementation of culturally adapted nutrition lessons at a summer camp. (S6) Local food production was promoted through skill acquisition and taught participants how to grow and cook their own produce using modern and traditional practices whilst emphasising the impacts of imported versus local food. The study used an adapted WillTry tool to assess dietary intake and 'willingness to try' various items, but showed no significant change in dietary intake, or willingness to try, post-intervention. S6) The other two were separate arms of the regional Pacific Obesity Prevention in Communities Project in Fiji (S21) and Tonga, (S20) which involved quasi-experimental studies of community and school-based interventions that included some emphasis on local food production and consumption through vegetable gardens or agricultural training, as a way of building capacity for schools and communities to create solutions to food insecurity. (S20) Despite these studies providing no evidence for significant effects on dietary intake, the Tongan study did have a significant effect on reducing purchasing of some unhealthy food items, including snack foods from a shop or takeaway. (S20) One of the nine studies did not measure dietary intake but evaluated change in nutrition knowledge as the primary outcome measure. This study implemented culturally-appropriate strategies to educate adults in American Samoa about nutrition, and demonstrated an increase in knowledge about items such as high-fat and high-fibre foods post intervention. (S5) Only one study, which was a randomised controlled trial conducted in Singapore, considered the unique nutritional composition of local food in their outcome measure, by analysing dietary records using an online nutrient analysis software which was derived from locally available foods. (S9) The intervention targeted overweight and obese pregnant women with a food coaching app that provided guidance on healthy choices and resulted in no significant changes to macro or micronutrient intakes. (S9) Although the intervention itself did not apply a local approach or focus on the consumption of locally produced foods, it is worth noting that this study was the only study to specifically acknowledge nutrient variation according to food source, and apply that to their analysis; however the findings reported in the manuscript are not specifically discussed in the context of this application.

Effectiveness of non-local food approach (n=15)
Twelve of these studies measured the impact of intervention on dietary intake. Two of these 12 studies were significantly effective, four showed some significant improvement on some, but not other, measures of dietary intake, and six showed no evidence to be effective.
The two studies that were effective, applied an education approach to teach adults about nutrition via lessons and educational materials. (S4, S15) Both were quasi-experimental studies conducted in the AIMS region, used FFQs to assess dietary intake outcomes and BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) demonstrated significant improvements in diet. One targeted housewives in Mauritius and significantly increased the number of servings of fruit and vegetables consumed compared to pre-intervention, (S4) and the other targeted patients and carers attending general outpatients clinic in Singapore with nutrition information, cooking demonstrations and provision of healthy food samples. (S15) The evaluation focused on wholegrain consumption and found a significant increase in frequency of consumption and purchasing of wholegrains. Both of these interventions also measured nutrition knowledge and showed significant improvements in knowledge scores post intervention.
The four studies that demonstrated mixed effectiveness included educational intervention, food provision and regulations on advertising and marketing. One study targeted hospital staff in Singapore to assess the impact of a plate displaying portion guidance compared to a normal plate and found increased vegetable and reduced carbohydrate proportion over six months, but no significant change in proportion of protein. (S2) One study evaluated a free school meals intervention in Puerto Rico. (S19) Using 24-hour recall method researchers found a significant improvement in nutrient intake for some (six of the 17) vitamins and minerals, but no significant improvements in macronutrient or sodium intake. A Trinidadbased study that used nutrition education to target fruit and vegetable intake in school-aged children, showed no significant association between intervention and fruit or vegetable intake in multivariable regression, but provided evidence that the intervention was associated with lower intakes of fried food, soda and food high fat salt and sugar, and improvements in nutrition knowledge. (S8) Similar to that study, the final study to show a mixed effect demonstrated significant reductions in the consumption of two out of three unhealthy items measured (chips, sweets, burgers), but no improvements in fruit and vegetable intake. This intervention, targeting children, was conducted in Singapore and implemented a national policy-level restriction on advertising and marketing of energy dense, nutrient poor food and drink to children. (S18) This study also measured the impact of intervention on purchase of specific food items, and provided evidence of a significant reduction in purchasing of selected processed/convenience foods, but no difference in fruit or vegetable purchasing.
Two of the six interventions that were ineffective at changing dietary intake, applied a support/guidance approach to empower participants to improve dietary intake (12, 15,) , and one implemented school-based nutrition education. One targeted women aged 40-60 years in Trinidad and Tobago and implemented health-focused support meetings, specifically to promote fruit and vegetable intake, but demonstrated a reduction in fruit and vegetable consumption following the intervention. (S11) One targeted university students, in Puerto Rico, with support sessions that focused on stress reduction and mindfulness to improve diet and physical activity, but showed no significant reduction in consumption of the two variables assess, SSB or bread, between control and intervention groups. (S10) Another intervention that was ineffective, was the implementation of school-based nutrition and physical activity lessons in Puerto Rico that showed no significant effect on fruit, vegetable, fibre or energy intake, post intervention. (S7) Finally, two community level interventions, one in Fiji and the other in Samoa, applied a range of interventions as part of salt reduction programmes. (S22, S23) The Fiji study involved multilevel intervention from individual to national policy level and targeted stakeholders across the food system from producers (advocating for reformulation) through to consumer education. (S22) In Samoa, a community level media campaign advocated for reducing salt intake. (S23) Both used non-randomised, uncontrolled pre/post-test designs and measured salt intake through 24-hour urinary salt excretion, and reported no significant reduction in mean salt intake. However, the Samoan study also measured nutrition (specifically salt- The other four of the 15 studies that did not focus on locally produced food, did not measure dietary intake, but evaluated, and showed mixed intervention effects on, other outcomes; purchasing, (S1, S17) sales (S16) or nutrition knowledge (16) . Two studies measured purchases from an online hypothetical grocery store and were based in Singapore; (S1,S17) one assessed the impact of a hypothetical tax of high calorie items on purchase and expenditure, (S17) and the other assessed the impact of various types of food labelling (information) on product choice. (S1) Both found statistically significant changes in some of the purchasing behaviours measured. The explicit tax was associated with a decrease in unhealthy food purchases, which was not shown in the implicit or fake tax arm of the study, (S17) and a text-only warning label was associated with a significant reduction in purchase of labelled products; however, there was no difference in total sugar purchased or total expenditure on high-sugar products. (S1) One study evaluated the impact of a national 10% tax on sugar-sweetened beverages (SSBs) in Barbados and showed a significant decrease in grocery store sales of SSBs, which was primarily driven by a reduction in carbonated SSB sales, and an increase in sales of non-SSB including significant increase in sales of bottled water. (S16) The final study, conducted in Trinidad and Tobago, measured the impact of a dietitian-led nutrition counselling intervention on nutrition knowledge, and targeted adults with Type 2 Diabetes. (S12) There was no significant difference in total knowledge score when comparing the intervention and control (no counselling) groups, but did demonstrate a significant improvement in attitude and practice scores amongst those who received the intervention compared to the control.