Food safety interventions in low‐ and middle‐income countries in Asia: A systematic review

Effective and sustainable interventions are necessary for long‐term improvement of food safety. This review provides a summary of food safety interventions evaluated in selected low‐ and middle‐income countries in Asia between 2000 and 2020. A systematic screening of published articles from PubMed and CabDirect databases was carried out with the aid of Rayyan QCRI software. A total of 25 studies were considered in the review. A ‘before and after’ study design was the most frequently used design (64%), while five studies (20%) used a randomized control trial (RCT) design. Interventions applied focused on training to improve knowledge, attitudes, and practices (KAP) towards safe food (60%) or on specific technologies (40%). Nine of the 25 studies were specific on the value chain considered: cattle (1), poultry (1), pigs (4) and fish value chains (3). Except for one study, all interventions reported some level of success; 17 were rated as having a high level of success, defined differently across studies. However, there is a clear evidence gap for the efficacy and cost‐effectiveness of food safety interventions in market settings, both consumer‐ and vendor‐facing. A rigorous and standardized assessment of intervention effectiveness and sustainability is recommended, to not only identify areas of improvement, but also to ensure scaling of interventions with demonstrated evidence of success and sustainability.

middle-income countries (LMICs), with the continent of Africa leading in terms of per capita burden, seconded by South-East Asia. In the South-East Asian sub-regions, SEAR B and SEAR D, the foodborne disease burden is estimated at 690 and 710 DALYs lost per 100,000 population, respectively (Havelaar et al., 2015). The most affected demographic are children below the age of 5 years, who account for disproportionately more illnesses and deaths (Havelaar et al., 2015). The African and South-East Asian WHO regions also have lowest national food safety scores, just 45% and 51%, respectively in 2021 compared with 78% for the European region (SPAR, 2021 https://extra net.who.int/sph/spar).
There is an urgent need to invest in food safety especially in developing countries where most of the health burden accrues, and economic costs are more than USD 100 billion per year (Jaffe et al., 2019). Food safety interventions (FSI) can be categorized by the point within the value chain where they are implemented, moving from farm to fork; pre-harvest interventions may include the use of biocontrol methods in the management of fungal toxins (Ola et al., 2022), or vaccine use in animals (Charlier et al., 2022) or post-harvest activities such as spraying animal carcasses with lactic acid (Lambré et al., 2022) or exposing fresh produce to ultraviolet light (Yemmireddy et al., 2022). Interventions focusing on transport often aim to maintain a cold chain and traceability (Moore et al., 2020). Interventions targeting food processing facilities can mitigate disease risk in processed foods, for example, the pasteurization of dairy products reduces the risk of exposure to consumers (Ruegg, 2003). Apart from upgrading market infrastructure, few interventions have targeted retail (Grace et al., 2018). Education and training interventions can target actors at any level in the value chain (Singh et al., 2016;Takanashi et al., 2013) while awareness creation typically targets consumers (Bass et al., 2022). A recent review of food safety training interventions (Global Alliance for Improved Nutrition (GAIN), 2022) categorized training by the underlying theory, audience, training environment and curriculum. It found the topics, training duration, and learning outcomes were highly variable, and a wide range of media were used (from lectures, to guided practice, to parades, to behavioural nudges).
Attempts to improve food safety especially in LMICs face a number of obstacles including (i) weak public health systems leading to poor disease and pathogen identification; (ii) absence of rigorous epidemiologic investigations to establish source attribution, blurring logical assertions on whether the cause of the illness was unsafe food, drinking water, animals, the environment, people or a combination of factors, and (iii) poor reporting of FBD (Grace, 2017). A plan that aims to address these gaps would provide a good basis for successful implementation of food safety interventions.
National food control systems have been defined as a combination of mandatory regulatory approaches (e.g., inspection) with preventive and educational strategies along the entire food chain that assures the quality, legality, and safety of foods (FAO and WHO, 2019). National food control systems can be considered along four dimensions: inputs (including legal framework and recourses); control (including inspection and surveillance); communication (with domestic and international stakeholders); and evidence (including risk analysis and performance monitoring) (FAO and WHO, 2019). Considering these dimensions, legal frameworks in LMICs are often out-of-date and inappropriate while, by definition, resources are constrained (Grace, 2015). Control relies on food inspection but the limited effectiveness of this is shown by the fact that much or most of the food sampled in LMICs fails to comply with standards (Roesel et al., 2014); moreover, a situational analysis of six African countries found that the chance that a livestock products sold in a mass domestic market would be inspected was never more than one in 2000 and most commonly zero (Jabbar & Grace, 2012). Assessments of food safety in LMICs commonly note the lack of inspectors and inadequate guidance and governance concluding sanitary inspections may contribute little to food safety (Ntawubizi et al., 2020).
A review of available interventions, their efficacy and costeffectiveness, is needed to inform decisions on what can be scaled up and what modifications may be needed to suit use in different contexts. We had previously conducted a systematic literature review (SLR) on food safety interventions in Africa, the region with the highest per capita FBD burden (Grace et al., 2018), and following from this, wished to conduct a SLR on food safety interventions in Asian low-and middle-income countries, the region with the second highest per capita FBD. The review was undertaken as part of a donor-funded project 'EatSafe; Evidence and Action Towards Safe, Nutritious Food' with the findings expected to inform the design of interventions to implement as part of the program, as well as to inform future programming decisions. Ethical approval was not sought as the study only involved review of existing literature. No human subjects were interviewed.

| Selection of studies (search process)
The study was designed to capture food safety interventions conducted in Asia. The studies had to have been reported between the year 2000 and 2020 and published in English to be included in the Impacts • This review provides an overview of food safety training and technology-specific interventions that have been evaluated in selected countries in Asia.
• Training interventions to increase knowledge, attitude and practices on food safety were common.
• The review highlights the gaps in designs meant to assess impact or success, cost-effectiveness and sustainability of interventions. review. We limited the literature search to 2000 because we needed to put temporal bounds on the study, and based on our knowledge of the enormous changes in SE Asian food systems over the last two decades, we believed that studies before 2000 would be less relevant. A total of 20 Asian countries were purposively considered in the review: Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, Sri Lanka, Cambodia, Laos, Myanmar, Thailand, Vietnam, Brunei, Indonesia, Malaysia, Papua New Guinea, New Guinea, Philippines, and Timor-Leste. Observational studies, secondary data sources, or field experimental studies on interventions aimed at improving food safety or evaluation of food hazards or health impacts were included.

| Exclusion criteria
Exclusion criteria included: interventions not related to foodborne hazards, those only conducted in laboratories or controlled research farm settings (on-farm trials), those focused on prevalence or risk factor analysis instead of interventions, or those not implemented in the selected Asian countries, or those published before 2000 or after 2020.

| Keywords and publication search
Search strategies were developed using the population, intervention, comparison and outcome (PICO) format using Medical Subject Headings (MeSH). Keywords were defined and combined into appropriate syntaxes (Table 1). Searches were done in PubMed and CabDirect databases, on the 4 November 2020 and outputs exported to Rayyan QCRI software (Ouzzani et al., 2016) from where duplicates were sought and removed.

| Article selection and screening process
Screening of articles was done in Rayyan QCRI software . The search   results including the publication title, author list, journal, year, issue,   page numbers, abstract and Digital Object Identifier (DOI) numbers, when available, were downloaded and uploaded to Rayyan®. After the duplicate removal was done, the articles were subjected to title and abstract screening against predefined inclusion and exclusion criteria. The screening was done independently by two reviewers.
The included articles were then subjected to full-text screening.
Reasons for the full-text exclusion have been provided (Figure 1).
Any discordance in the classification of articles was resolved by a third reviewer.

| Quality assessment
Quality assessment of individual papers was based on the Cochrane assessment of bias criteria (Higgins et al., 2022). The included articles were classified as good, medium or poor quality. Articles that had an unbiased selection of subjects, had methods that were judged by the reviewers to be scientifically sound, with appropriate data analysis and complete results that articles were classified as good quality articles. Medium quality articles acknowledged and accounted for selection bias of subjects, limitations in data analysis, had methods that the reviewers understood, and had results that were valid. Table 2 presents a summary of the data extracted from the relevant articles. The data was summarized as proportions and presented in tables and figures. TA B L E 1 Databases searched, and search syntax used in the review.

Database Syntax No. of hits
PubMed ((foodborne OR "food safety" OR "safety label" OR "food hygiene" OR "food-borne" OR "food borne" OR "kitchen hygiene" OR "street food" OR abattoir OR slaughter* OR "willing* to pay" NOT "organic" NOT "indigenous" NOT "GMO" NOT "exotic") AND ("cluster random" OR Random OR "clinical trial" OR intervention OR trial OR experiment OR impact OR evaluation OR effect OR control* OR manag* OR improve OR achiev*) AND ("South Asia" OR "South East Asia" OR Afghanistan OR Bangladesh OR Bhutan OR India OR Maldives OR Nepal OR Pakistan OR Sri Lanka OR Cambodia OR Laos OR Myanmar OR Thailand OR Vietnam OR "Viet Nam" OR Brunei OR Indonesia OR Malaysia OR "Papa New Guinea" OR "New Guinea" OR Philippines OR "Timor-Leste") AND ("2000/01/01"[PDat]: "2020/08/01"[PDat])) 3150 CabDirect (title: (foodborne OR "food safety" OR "safety label" OR "food hygiene" OR "food-borne" OR "food borne" OR "kitchen hygiene" OR "street food" OR abattoir OR slaughter* OR "willing* to pay" NOT "organic" NOT "indigenous" NOT "GMO" NOT "exotic") OR ab: (foodborne OR "food safety" OR "safety label" OR "food hygiene" OR "food-borne" OR "food borne" OR "kitchen hygiene" OR "street food" OR abattoir OR slaughter* OR "willing* to pay" NOT "organic" NOT "indigenous" NOT "GMO" NOT "exotic")) AND

| Outcome of the systematic screening
The PRISMA flow diagram showing the study selection process is presented as Figure 1. A total of 4237 articles were found after database search: 3049 from PubMed and 1188 from CabDirect. Five more studies were found through reference screening (these were added later in the review process) and underwent the same screening process. Two hundred and eighty-eight (288) articles were duplicates and were thus removed. The total number of articles subjected to title and abstract screening was 3954. Fifty-five (55) of the 3954 met the inclusion criteria and were considered in the next stage of review. Twenty-five (25) articles were selected from the full paper review. The remaining articles (30) were excluded for the following reasons: experimental laboratory studies (11), did not specify the study country (n = 9), the interventions were not related to improving food safety (6), or the full-text articles were inaccessible (4). Six of the articles were scored as having moderate quality and 19 as good quality. No article was judged to be of poor quality. Data were extracted from these 25 articles and subjected to qualitative analysis (see Tables 3 and 4 for a summary of the 25 articles). Sufficient data were not available to conduct a meta-analysis.
The 'before and after' was the most frequently (64%, 16/25) used study design. In this case, the outcome of interest was measured before and after implementation of the intervention but with no randomization or control groups. Five studies reported the use of randomized control trials (RCT). In this case, two groups were considered, one group received the intervention while the other group did not and acted as a control. Two studies introduced a new approach to food safety intervention and measured adoption, while for the remaining two studies, the design as described by the authors does not fit into a RCT or a 'before and after' design.
Training to improve knowledge, attitude, and practice (KAP) towards safe food (60%), and technology-specific interventions (40%) were the main categories of interventions reported. Each of these interventions is described in the following sections. Most of the interventions (72%, 18/25) were applied at the household or food vendor level and focused on consumer safety. Sixteen (64%) studies involved informal food vendors; 7 (28%) were applied in the formal sector settings. Two (8%) were non-specific.
Nine of the 25 studies were specific on the value chain considered: cattle (1), poultry (1), pigs (4) and fish value chains (3 Low level of hazard -All samples from slaughterhouses that used the Halal method lower total place count than the maximum recommended level (4), market/retail level (1), farm/production level (3) or at the level of the processing stage (1).

| Training and information-sharing interventions
There were 16 interventions on training and information sharing. A summary of the approaches used in delivery of food safety trainings is given in Table 3.

| Technology-specific interventions
A summary of the nine technology-specific interventions is given in Table 4. The interventions used 'before and after' design (6/9) and mostly measured hazard presence (7/9). The studies demonstrated some level of success except one study that failed to demonstrate the success of the intervention. Siagian et al. (2015) analysed the effects of irradiated ready-to-eat foods on the health of immunocompromised residents of the National Narcotic Board in Bogor, Indonesia. Safe traditional ready-to-eat foods were irradiated at an irradiation facility and the process maintained under cryogenic conditions to protect essential dietary nutrients from free radical attack.
The food was then given to immunocompromised people in the intervention group over a period of 21 days. No effect was found on body mass index, but increased albumin levels were reported in the group that consumed irradiated foods. an increase in knowledge, attitude and practice related to food hygiene and handling practices following training Malhotra et al., 2008;Riaz et al., 2016;Samaan et al., 2012). A metaanalysis of studies on training interventions to increase knowledge and attitude of food vendors on food handling practices produced a significantly higher summary effect than those without the training (Soon et al., 2012). World Health Organization. (1999) and Sivaramalingam et al. (2015) recommend that food safety interven- Availability of government services like provision of potable water, relevant infrastructure, and transport services impacts food safety levels (Gabriël et al., 2018).
Nearly twice as many training interventions (16)  Most of the reported food safety interventions were general and did not target a specific value chain and pathogen. However, a few targeted specific hazards, including: Taenia solium, E. coli, zoonotic fish trematodes, faecal coliforms, and faecal Streptococcus.
Unsurprisingly, training was more general (15 out of 16 interventions), while technologies were more likely to target specific hazards (six out of nine interventions).
All interventions focused on specific pathogens targeted zoonotic foodborne pathogens. These may require different types of interventions at different stages of the value chain. The control effort can either be directed to controlling the pathogens in live animals or humans or both (Gabriël et al., 2018). More specific interventions are needed to control specific pathogens for various value chains.
Effective control of fish-borne zoonotic trematodes in a community requires repeat treatments, for 1 year or more, to reduce risks of reinfections . Single chemotherapy of reservoir hosts or humans, or both does not change the reproduction ratio whereas continuous chemotherapy of reservoir hosts or humans led to a decreased reproduction ratio of respectively 1.69 and 1.30, respectively (Boerlage et al., 2013). A practical and sustainable approach to controlling fish-borne zoonotic trematodes in aquaculture can be achieved through identification of critical control points in the fish production system, and protection of fish ponds from con- Sophisticated information technology (e.g., mobile phones, internet) were historically not as widespread, so these media channels were not covered in the studies.
The review noted the absence of comprehensive project design in food safety interventions, the initiatives did not highlight high, medium and low-level outcomes anticipated upon implementation nor clearly defined 'impact' or 'success'. There is a clear evidence gap for food safety interventions in market settings, for both consumers and vendors. Food safety interventions need to leverage existing social networks to improve the dissemination and sustainability of such interventions. There is need for rigorous assessments of intervention effectiveness and sustainability as well as the identification of foreseeable impediments to inform the design of food safety interventions.

ACK N O WLE D G E M ENTS
This study was made possible through support provided by

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.