Global genomics research increases animal health metadata​

Global organizations such as the World Organization for Animal Health (WOAH) and Food and Agriculture Organization (FAO) have taken the lead in compiling data related to animal disease metrics. However, as genomic data increases with the decreasing cost of sequencing, metadata related to animal health and disease can be used to supplement GBADs' existing data. There are estimates that predict that there are up to 40 billion gigabytes of genomic data generated every year and that over the next decade genomics research will generate between 2 and 40 exabytes of data (Stephens et al. 2015). The National Centre for Biotechnology Information (NCBI) is an example of a set of databases that holds information pertinent to GBADs. As more researchers are conducting genomic work on livestock and other economically important species, data on breed, location, and individual condition and disease are being recorded. In particular, NCBI's BioSamples database stores the accompanying metadata to genetic sequences uploaded to NCBI. As sequencing becomes more cost efficient, animal metadata that can be used by GBADs will accumulate. There has been a remarkable spike in the number of metadata entries concerning livestock in the past 7 years (Figure 1). These metadata accompany genomic studies carried out by not just universities, but also organizations around the world. NCBI has already been used to create databases for other researchers to use (Hu et al. 2022). The AnimalMetagenome DB (http://animalmetagenome.com) holds metagenomic data for 4 domestic species and an abundance of wild species.

Figure 1. NCBI entries in the BioSample database concerning species covered in GBADs (cattle, pig, goat, sheep, chicken, llama, equids, camel). This figure shows the pattern of number of entries since 2005.

Geographic spread of genomic data​

One of GBADs' main goals is to close global data gaps concerning animal health. In order to standardize basic animal health data across countries, location and breed data are incredibly important. Classifying the global range of animals will assist in understanding the spread of diseases originating from animals. Users can input location data as a basic attribute in the BioSamples database. In particular, entries can contain the location of the submitter, sampling location, and even specific sampling coordinates. These data can reflect where certain species densely populate or which countries and locations have a lack of data. Although DNA sequencing has decreased in cost, it is still a sizable purchase; therefore, these data can also indicate whether there are inherent biases in the presence of genomic sampling across countries. Using rentrez, an R package to access NCBI (Winter 2017), we were able to extract metadata entries for genomic studies involving livestock and other economically important species. Figure 2 illustrates the geographic spread of these data, where it is shown that there are clusters of data in highly studied areas. Conversely, there are clear gaps in data across Africa and parts of Asia (Figure 2).

Figure 2. Static snapshot of an interactive figure displaying sampling locations for livestock and other species relevant to GBADs based on NCBI BioSample metadata.

Now, these are just the data that contained location information for each entry. As mentioned earlier, these data are important baseline data on global species occurrence; however, the BioSamples database also holds data more pertinent to GBADs, such as disease. We explored the mined metadata from NCBI and found that in our study species, 98 percent of entries did not contain viable disease data. By 'viable', we mean information that is disease related, as many entries had random characters or non-related information. The remaining 2 percent of data held diseases and disease agents that overlapped with WOAH's disease agent list (WOAH, personal communication of internal document). For example, bovines had 16 diseases that overlapped with the WOAH list (Figure 3).

Figure 3. Static snapshot of an interactive figure displaying sampling locations for cattle with disease information present in the NCBI entry. The diseases listed overlap with the diseases WOAH is monitoring.

Inconsistent user input leads to gaps in data​

NCBI houses data for a range of organisms from bacteria to whales; however, when mining data for livestock and economically important species, close to 350,000 entries were returned. Gathering data for GBADs species (cattle, chicken, pig, sheep, goat, etc) revealed variation in the amount of data. The figure below illustrates this variation, where cattle, chicken, and pigs returned the most amount of data. The completeness of the data for these returned entries varied; however, they made up the largest proportion of GBADs-related data. Cattle entries, in particular, accounted for 27% of the total number of entries. The combination of two other species, llamas and camels, made up less than 1% of the total returned entries.

In addition to species information, NCBI also allows for users to input breed data. As breed information can be useful for understanding the prevalence and spread of disease, it is important to know which breeds form the majority of the data. Figure 4 illustrates the proportion of specific breeds for each species gathered for GBADs, where each different colour bar indicates a different breed. As shown in the plot, the majority of entries did not contain breed information. This is an important finding, as breed information should be one of the main parameters reported, especially for economically important species. This plot also shows potential gaps in the data housed in NCBI, as certain breeds are not as prevalent in the data. This can provide knowledge about where resources, time, and funding should be allocated to equalize the data across breed and species.

Figure 4. Barplot illustrating the proportion of different breeds reported for each species investigated in the NCBI BioSamples database. The pink indicates the number of entries with no breed information. All other colours indicate proportion of different breeds for each species.

Next steps in utilizing these data​

Based on our first look at the available metadata on NCBI, there are a few tasks that could make the process more streamlined. Additionally, the following steps would increase metadata for secondary uses.

1. Standardize parameters. Standardizing parameters would improve the data cleaning process and assist in equalizing data across different institutions. Currently, the free text set up of the BioSample database makes it very difficult to compare across entries, as misspellings and extra characters can inhibit comparisons. As discussed elsewhere (Goncalves and Musen 2019), bolstering the underlying infrastructure of the NCBI BioSamples database will assist in standardizing data and make these data more accessible to researchers for secondary uses outside of genomics.

2. More focus on metadata. There are major gaps in data that range from smaller details such as disease prevalence to important parameters such as breed or location. As NCBI gets accessed more and more for metadata, it will be imperative that users input as much data as possible. Basic data such as species, breed, and location should be mandatory for users to input.

3. Preprint data uploads. Currently, there is a lag between sampling and DNA sequencing and when these data get uploaded to public databases such as NCBI. This lag can vary from a few months to years. Therefore, data that could be used in secondary projects such as this one could be released up to a few years after sampling. Therefore, I recommend uploading genomic data with its accompanying metadata to NCBI as soon as possible or when preprints are submitted.

NCBI and other genomic databases hold data that can be useful for more than just genomic projects. For GBADs, these data can provide valuable information about presence of local breeds in normally unsampled locations as well as occurrences of diseases in sampled individuals. Additionally, it provides the first look into how genomic projects may supplement GBADs' aim to gather animal health data and metrics.

References:

Goncalves R, Musen MA (2019) The variable quality of metadata about biological samples used in biomedical experiments. Scientific Data, 6, 190021.

Hu R, Yao R, Li L, et al. (2022) A database of animal metagenomes. Scientific Data, 9, 312.

Stephens ZD, Lee SY, Faghri, F, et al. (2015) Big Data: Astronomical or Genomical? PLoS Biology, 3, 1002195.

Winter, DJ (2017) rentrez: An R package for the NCBI eUtils API. The R Journal, 9, 520-526.

Figure 1: The Roadmap to Reproducibility

“The whole point of science, the way we know something, is not that I trust Isaac Newton because I think he was a great guy. The whole point is that I can do it myself … Show me the data, show me the process, show me the method, and then if I want to, I can reproduce it.”

Brian Nosek - Washington Post¹

Reproducible science requires well-documented methods, code, and making data available. It means providing transparency in what you are doing through the whole scientific process to foster trust in the process and outcomes and allowing others to leverage past work. There is no place for the scientific skeptic when your research is reproducible.

The data that GBADs is using comes from many different sources and is used in models that then produce more datasets, and that act as inputs to other models. This daisy-chain of data-model-data-model-data is not confined to the work of a single scientist; we have collaborators working all over the globe. It is crucial that the underlying data is available, and all our methods are reproducible so we can build upon each other’s work and allow others to use our estimates confidently.

With the aim of making all our processes reproducible and transparent, GBADs is embarking on the “Roadmap to Reproducibility”. In this blog post we invite you to travel along the road to reproducibility with us. Buckle up your seat belts as we avoid the fiery blazes of the “Reproducibility Crisis” before heading to our final destination, the “Data Utopia”!

The “Reproducibility Crisis”​

In 2016, a survey of 1,576 researchers from Nature revealed that there is a "crisis of reproducibility" in the scientific community². Of the participants "more than 70% of researchers have tried and failed to reproduce another scientists’ experiment and more than half have failed to reproduce their own experiments." Many factors contribute to irreproducible research (Figure 2), where pressure to publish and selective reporting were among the most highly rated. However, unavailable raw data and unavailable methods or code were also rated as high contributors. While pressure to publish requires a culture shift in research, unavailable raw data and methods or code are problems we can solve.

Figure 2: Factors that contribute to irreproducible research. Figure obtained from Baker, 2016².

Embarking on the Road to Reproducibility​

Stop 1: People and processes​

Code and data availability are important for reproducibility, however there are people behind the code and data working on processes to make it reproducible.

As such, we have established processes and best practices for the use of data in GBADs, which are communicated in the Data Governance Handbook, and on our Documentation Site. These processes involve the following:

1. Documentation of changes to data and data cleaning practices
2. Documentation of metadata standards used to provide information about the data
3. Where and how data and metadata are stored, and how they are disseminated
4. Best practices for documenting code in GitHub repositories

We also rely on people to comply to the processes we have established. Since some of the data that is used by GBADs does not have metadata, we rely on the establishment of a contact point for the data source to ensure that we can get context on how data were collected, how it can be used, by whom, and for which purposes, and what categories in the data represent.

Stop 2: Acquisition and ingestion of data​

To acquire data, we identify data that is relevant to the estimation of models. For example, livestock population by country and species and live weights are inputs to biomass calculations.

The way we acquire data depends on the format that it is available. There are 3 main ways data has been acquired and ingested:

• When data is available via Application Programming Interfaces (APIs), the data is acquired directly from the source, being transformed before made available via the GBADs’ API;
• When data is available via direct download, the data is downloaded and formatted in database tables before being made available via the GBADs’ API, and;
• When data is available in PDF tables, web scraping scripts scrape data from tables and make it available via csv files before being formatted in database tables and made available via the GBADs’ API.

Each of these processes are documented. The lineage of the data is traced in a Graph Database to ensure that we can trace any changes to the data and make our processes transparent and reproducible.

Stop 3: Data quality​

We quality-check each data source that is acquired by GBADs. Sometimes there are internal errors in the aggregation of categorizations where sub-categories do not add up to a “super-category”. For instance, if poultry is split up into backyard and commercial poultry, these categories should add up to “poultry”. In other cases, there may be a sudden spike in the number of animals in a country. In this case, we need to investigate by comparing the value to other data sources.

All quality-checking and respective changes are recorded. Once data has been “cleaned”, the “cleaned” version is provided via the GBADs’ API and dashboards. This way, there is consistency in quality assessments and each collaborator isn’t doing this independently. This ensures consistency in outputs and improves reproducibility of GBADs’ estimates and data collectively.

Stop 4: Code​

Code that is used to acquire and ingest data, clean data, and create models is available via the GBADs’ GitHub repositories. The code is well documented and there is information about how to run the code, which datasets were used, and who was involved in development.

Final Destination: Data Utopia​

In the Data Utopia, data can be harmonized and re-used for subsequent models or purposes. The idea here is that all members of GBADs are using the same data and not duplicating data cleaning, ingestion, or acquisition efforts. By making the data and methodology for cleaning reproducible, the underlying data is consistent and ready-to-use. In our Utopia, data is visualized and available via dashboards and can be accessed through the API. The dashboards also feature a metadata tab where the methodology, code, and provenance information are provided to ensure all members can access the code and raw data that are displayed and available in the dashboards.

Note: We acknowledge the fact that not all data can be made available. While we are currently working with Open Governmental Data, we anticipate the controlled governance of private and sensitive data, which will not be openly available in raw form without permission from the data holder according to data agreements and licenses.

References:

GBADs Informatics has brought on many new experts in 2022 who have a wide range of talents and interests. Together, the “herd” of 7 student developers, 2 graduate researchers, and 4 post-docs and research associates are advancing our goals in GBADs and exploring new avenues to improve GBADs and increase the utility of the program in creative ways. Check out our current team in the About section of the website and read more about our team’s work over the spring semester below!

Nuts and Bolts of the Knowledge Engine

Kassy has been building a graph database to improve the findability of data and analyze the interoperability of the data sources that GBADs collects. She has also been busy supporting students in their work, liaising with other themes to maintain progress, and spreading the word about GBADs and FAIR data as a Datasphere Initiative Fellow – check out some of her recent presentations on our Highlights page or our Presentations and Publications.

Amardeep developed a modular approach to dashboard creation and produced tutorials for anyone to learn how to produce a basic dashboard in the GBADs style. You can access his tutorials in the docs section of this site.

Nitin has been working away at the permissions and access restrictions infrastructure needed to ensure GBADs data is secure, and that we can apply access restrictions for different users seeking to access the data portal of the knowledge engine. This infrastructure will be critical for maintaining private data security.

Rehan has become our expert for livestock data scraped from the Central Statistics Agency of Ethiopia’s pdf reports. He has, for the first time, made those data available in a database and is working to develop reports to compare trends on regional and national levels, and identify data discrepancies. He has also been working to compare those data to global data. Reports coming soon, check out how to access details via our API!

Polishing our Products

Matthew has put a lot of work into developing and launching this site for documentation of GBADs resources, tutorials, documentation, and presentations. The site is constantly evolving and more material being added, so check back frequently!

Kurtis began working with us to improve the design of our dashboards and visualizations and to develop a cohesive visual standard to use across GBADs. His style guides have been shared across other themes and with external partners like First Analytics.

(GIF courtesy of Kurtis Sobkowich)

Cross-Theme Collaboration

Grace worked with the Human Health theme to develop an approach to link livestock health changes to human dietary health impacts. She also wrote a blog on the challenges in quantifying the health impacts of animal-sourced food consumption.

William rapidly adapted a python module for expert elicitation (Anduryl) to function in R. He shared this with the team working on the Ethiopian case study, who deployed it in their recent workshops. Find his tutorial on the software here.

Le quickly caught up to speed on progress on the graph database and ontology and has been collaborating with the Ontology and Attribution theme to develop ways to improve the interoperability of data from different major data sources. They are also working to map ontologies to databases.

Kassy and Deb worked with Yin from the Populations and Production Systems theme to structure their biomass calculations and code for use with different data sources and for incorporation in the knowledge engine.

Exploring New Frontiers

Cassandre has been mining NCBI for genomic data on livestock species that might be used to help classify production systems. You can view her seminar on her work here. She is also exploring other databases that could be mined for similar info on breed characteristics such as size and health traits.

Neila is working on the future of animal welfare in GBADs. She has been collaborating with colleagues from Liverpool to develop a proposal to incorporate this topic in future iterations of GBADs, building networks with animal welfare experts around the world, and exploring the existing animal welfare data ecosystem. Read her intro to animal welfare legislation in her recent blog.

Last but certainly not least, Deb and Theresa continued to contribute to GBADs planning and steered the herd to success! They are critical for problem-solving, connecting with partners, and developing our roadmap for the future of GBADs Informatics.

Summer Preview

Keep an eye out for an update on our summer activities, which includes the roll out of numerous dashboards, thanks to the approach Amardeep developed:

• Sky is producing updated dashboards for population and multiple biomass statistics
• Matthew and William are working on a dashboard to display calculations of the Total Economic Value of global livestock done by the PPS theme
• Ethiopia data story dashboard from Rehan, Sky, and Grace

Also, welcome to our newest herd member Faraz, who is working on data “healing” of OIE population data. Welcome also to Emily, who has joined the herd for summer fun!

Catch Up with the Herd

If you want to see GBADs Informatics in action, there are two upcoming opportunities. Check out the GBADs session at SciDataCon on June 22, where Deb, Kassy, and Le will be presenting alongside Ontology and Attribution colleague Stephen Kwok. The Informatics Theme will participate in a special session dedicated to GBADs and will also be presenting 3 talks and 4 posters at ISVEE from August 7-12. We will be covering a range of topics, including meta-ontology frameworks, animal welfare, food systems models, interdisciplinary work in the context of OneHealth, implications of data discrepancies, graph databases, and a survey of major data aggregation initiatives. We hope to see you there!

With an increasing human population, the number of animals raised for food has considerably increased in the last few decades. The world produced 800 million tons of milk, 340 million tons of meat in 2018 (Figure 1) and 86.67 million metric tons of eggs ¹ in 2019.

Figure 1: World meat production by livestock type from 1961 to 2018. (Source:Our World in Data)

Food animals are raised in different production systems where proper housing, management, nutrition, disease prevention and treatment, handling, and slaughter are the responsibility of humans. There is a large variety of production systems² but some of the most familiar are the intensive system confining animals in large numbers and relatively small spaces and the extensive system providing animals with grazing opportunities. In almost all production systems, animals can confront a number of conditions that might impair their overall wellbeing. The study of an animal’s physical, physiological and mental state is referred to as animal welfare ³. According to the World Organization for Animal Health, animal welfare is “the physical and mental state of an animal in relation to the conditions in which it lives and dies”⁴. Based on the European Union⁵ and the UK Animal welfare Act 2006⁶, reducing animal suffering by enabling preventive action to be taken before suffering occurs is the main point highlighted when defining animal welfare. Although steps have been taken to address the topic, livestock animals still go through welfare issues including chronic fear and pain, injuries and diseases, and movement deprivation to name a few. It took us a long time to come to a clear definition of animal welfare and ways to improve it at the farm level. In this blog, I will explore the main events that impacted the concept of animal welfare and how humans are studying this topic in farm animals. I will also present an overview on how we attempt to frame the concept of animal welfare in the Global Burden of Animal Diseases (GBADs) program analytical structure.

Key time points in the history of Animal welfare

For centuries, domesticated animals have played a major role in the life of humans. Through traditional farming techniques, these animals have always assisted humans in activities related to agriculture, transport, and trade. With the beginning of the industrialization, farming systems’ main goal was to increase productivity in order to feed a poor, but growing human population. This resulted in developing housing technologies to confine more animals and enhancing productivity by providing high quality feed. However, confining animals in crowded environments started to result in poor body conditions with the appearance of injuries, lameness, and diseases. It is with the publishing of the book “Animal Machines”⁷ by Ruth Harrison in 1964 that steps in a sequence of events to promote the concept of animal welfare as a public concern started to be taken (Figure 2).

Figure 2: Main events that contributed to setting Animal welfare as a public concern in Europe and the world.

Another important event was the forming of the Brambell Committee by the UK government, whose purpose was to report on welfare conditions in British livestock farming. In 1965, the Committee issued its Report to Enquire into the Welfare of Animals Kept under Intensive Livestock Husbandry Systems ⁸. This report paved the way to issue a list of concerns that the animal should be free of to ensure a decent welfare condition. This list was referred to as the five freedoms⁹ which was published after the creation of the Farm Animal Welfare Council in 1979. These freedoms are:

1. Freedom from Hunger and Thirst: by ready access to fresh water and a balanced and nourishing diet.
2. Freedom from Discomfort: by providing an appropriate environment including shelter and a resting area.
3. Freedom from Pain, Injury or Disease: by prevention or rapid diagnosis and treatment.
4. Freedom from Fear and Distress: by ensuring housing and management conditions which avoid mental suffering.
5. Freedom to Express Normal Behavior: by providing sufficient space for self-care (grooming, dust-bathing, stretching) and nesting, opportunities for social behavior with animals of their own kind.

Producing more meat at a lower cost comes now with a social pressure about the quality-of-life animals are experiencing. This paved the way to national, regional and international animal welfare regulations, private codes and practices¹⁰. The Five Freedoms were widely used as the basis in developing animal welfare guidelines, assessment protocols and audits. After introducing a Protocol on Animal Welfare in the Amsterdam Treaty¹¹ in 1997, the European Union (EU) was the first to publish animal welfare directives to ensure all country members adopt minimum guidelines when keeping animals for food production (Directive 98/581)⁵. Species specific EU directives were published later along with regulations specific to animal transport (Directive 1/20059)¹² and slaughter (Directive 1099/2009)¹³. Regarding international organizations, the OIE first published its Animal Welfare Standards⁴ for terrestrial and aquatic animals in 2004. National and international groups like the Royal Society for the Prevention of Cruelty to Animals (RSPCA) adopted their own guidelines and protocols to provide an auditing and final product labelling service.

Most recent initiatives for animal welfare are taking place in Europe. The ‘End the Cage Age’¹⁴ is a tool proposed by the European Commission to enhance citizens’ direct participation in policy making. It was launched in 2018 with the main demand to ban the use of cages in animal farming, providing animals with the opportunity to perform normal behaviors. The European Commission will propose legislation to prohibit the use of cages for laying hens, pullets, broiler and layer breeders, quail, ducks, geese and rabbits. This in addition to prohibiting farrowing crates and sow stalls and individual calf pens where not already prohibited. Another European initiative is the Farm to Fork Strategy¹⁵ which aims to accelerate Europe’s transition to a sustainable food system. In addition to promoting a neutral or positive environmental impact, the Commission announced that existing animal welfare legislation would be fully revised by 2023. The new legislation aims to ensure a higher level of protection, be broader in scope, easier to enforce and aligned with the latest scientific evidence.

All these events continue to shape the way we address animal welfare as citizens and as consumers. In addition to national and international guidelines to ensure farm animals are protected, animal welfare organizations and movements work on increasing consumer awareness about animal welfare. It is in recognizing where we fail as a society in providing the best environment and management to food animals that we will start to realize the impact of poor animal welfare on the global society and economy.

Animal welfare in the scope of GBADs

In addition to the fact that animal disease represents an impairment of animal welfare to a degree, the burden of animal welfare will eventually be included in GBADs’ economic estimations. This will be tackled through an analysis of the welfare factors that contribute to the appearance of disease on one hand and an estimation of the welfare impairment after disease emergence on the other hand. While the methodology is in progress, specific points of interest will be studied:

• The impact of animal welfare policy on farm economics on a national and regional level.
• The impact of the production system (including housing and management) on animal welfare and its economic consequences.
• Production systems, behavioral outcomes and its relationship with disease.

These goals, although seemingly ambitious, will be achieved over a long period of time. Generating data is not currently in the scope of GBADs. However, the use of open data sources, private data and/or expert elicitation data will be the basis of our analyses.

References:

(Graphic from Advancing Nutrition)

Animal-Sourced Foods in the Global Diet

Livestock have long played important cultural, social, and economic roles while also functioning as critical food sources. However, modern technological advances have led to an explosion in livestock production and consumption of animal-sourced foods (ASFs). Increased access to ASFs has been a boon to human dietary health in many ways, but modern livestock production systems are often harmful to the climate and raise animal welfare and sustainability concerns¹. In many high-income countries (HICs), people often eat more ASFs than are recommended, while ASFs are still scarce in low- and middle-income countries (LMICs)¹. Global reference diets advocate for very limited ASF intake² , but this is not healthful for many people.

So, in a world where ASFs are too available for some, and too scarce for others, and livestock still play an important cultural role, how do we balance scarce planetary resources to produce equitable outcomes related to livestock production and human health? This is a question that the Global Burden of Animal Diseases (GBADs) program is working to address by developing a systematic process to determine the burden of animal disease on human health and wellbeing. One piece of this puzzle involves understanding, in better detail, how ASF consumption impacts health in different contexts.

(Graphic from Our World in Data)

Role of ASFs in Health

ASFs are sources of key macronutrients (like protein and fat) and micronutrients (like calcium, Vitamin B12, Vitamin A, iron, and zinc), some of which are difficult to find or less bioavailable in plant sources³. Compared to supplements, whole foods also have bioactive factors and compounds that can enhance nutrient availability.

ASFs are important across the lifespan, but especially during childhood, pregnancy and lactation, and old age. Common issues correlated with low ASF consumption among these populations are anemia, stunting and wasting, micronutrient deficiencies and functional decline (in the elderly). Micronutrient and protein deficiencies in particular can lead to a cycle of impaired gut function and reduced nutrient absorption, as well as reduced immunological functioning and increased susceptibility to infectious disease⁴. These conditions are not restricted to those with undernutrition, either – many of the 772 million people affected by obesity suffer from similar micronutrient deficiencies and related health issues⁵.

(Graphic from Global Nutrition Report 2018)

While micro- and macronutrients have been linked to specific health outcomes, it is unclear in what quantity, how frequently, and how long they should be consumed to achieve lasting benefits in different populations, living under different conditions. Dietary guidelines, recommended daily average nutrient intakes, models of nutrient sufficiency, and other tools exist for different populations, but are sometimes based on reference populations that may not represent the group in question. Even in instances where it is clear how much of a nutrient is required for a specific population, it is unclear how to deploy specific foods to meet those needs. A person’s health status also influences their ability to absorb or use nutrients. Their gut microbiome, intestinal health, and the actual nutrient content of foods grown and stored in different manners all contribute to variance between the projected impacts of ASF consumption and the (inadequately) observed reality.

State of the Evidence on ASFs and Health Outcomes

The evidence base for the impact of ASF consumption on health, particularly among key risk groups and life stages, is disappointingly sparse^3,1,6,7. The health impacts of dietary changes are notoriously difficult to capture. Dietary changes largely only cause long-term impacts after, well, a long time, and it is difficult to conduct rigorous randomized controlled trials (RCTs) of diets for more than a few weeks. Long-term observational cohort studies are often plagued by high costs, poor participant retention rates, and logistical issues. Robust cross-sectional epidemiological studies provide evidence for relationships between consumption and risk of health outcomes, but these often are limited to HICs and cannot be used to make causal inferences.

Recent reviews have summarized the minimal and often inconclusive empirical research on ASF consumption impacts. Reviews concerning elderly populations largely focus on protein intake in HICs. They suggest that protein from ASFs may reduce risk of functional decline and may be preferable to plant-based protein in maintaining muscle mass^8,9. Another review of the impact of livestock-derived foods on the nutritional health of pregnant women could not even find any studies to assess³. The same review found mixed results for the impact of milk supplementation of varying amounts and lengths on linear growth in children, despite the known relationship between milk consumption and growth-promoting biological factors. RCTs related to ASF consumption and health have mostly been conducted among children in LMICs, yet a systematic review of such studies found inconsistent results and overall very low study quality¹⁰.

The Lulun project illustrates a rigorous and high-quality RCT and highlights difficulties in assessing ASF consumption impacts. The study involved egg supplementation for six months among 6–9-month-old children in Ecuador, resulting in increased weight and height gain and substantially reduced stunting risk¹¹. However, a repeat of the study in Malawi saw no such effect, potentially due to higher baseline consumption of ASFs or greater exposure to gastrointestinal disease risk factors¹². Even the positive effects observed in Ecuador may have minimal long-term impact – a follow-up study two years later noted similar levels of growth faltering across intervention and control groups¹³. Interestingly, egg consumption in either group after the study ended was correlated with reduced growth faltering at the later follow-up point, suggesting that sustained egg consumption conferred benefits.

A related field of study considers the impact of nutrition-sensitive agriculture, interventions which may be more reflective of potentially long-term sustainable approaches to improving ASF access. These interventions are typically aimed at improving smallholder agriculture through training, behavior change, and/or access to agricultural resources. Several such projects have demonstrated improvements in ASF consumption among participants, though the pathways through which agricultural projects impact nutrition are more complicated than in ASF supplementation studies¹⁴. Most studies of nutrition-sensitive agricultural programs are not adequately designed to assess nutrition outcomes, though, and so far have only demonstrated weak effects on health indicators such as stunting⁶.

(Simplified impact pathway for nutrition-sensitive agricultural interventions. Graphic from FAO)

Some Ways Forward

There are many ways to improve data collection surrounding the impact of ASFs on health, a few of which are highlighted here. Longer follow-up periods are recommended for RCTs, as the longevity of health benefits gained from ASF consumption interventions are unclear. Using different study endpoints may also improve our understanding of how ASF consumption changes our bodies during and after an intervention. Weight or circulating micronutrient levels may bounce back short term with an intervention – but do metabolic and immunological indicators change in the same time frame?

In the fields of nutrition-sensitive agriculture and livestock production, interdisciplinary partnerships between researchers and program implementers can help overcome cost barriers and ensure that appropriate nutrition outcomes, such as dietary diversity, are embedded in programs from inception^15,16.

Finally, our understanding of the impact of ASF consumption can grow by understanding how processing impacts the nutrient content of food, understanding how culture and behavioral norms influence consumption, and by developing more accurate recommended intake levels for different populations.

GBADs’ Role

While not directly involved in strengthening data collection on the impact of ASFs on health, GBADs has a role to play in calculating how to achieve sustainable, equitable production and consumption of ASFs to improve human health. GBADs will

• Demonstrate inefficiencies in livestock production and across value chains

• Provide in-depth data on production systems to identify the most efficient systems for a given context

• Provide high-quality estimates of ASF production by commodity and by location, to determine where ASF access could be improved via improved livestock health

• Quantify how poor livestock health contributes to poor human health

• Contribute to strengthening the connection between livestock production and nutrition sectors

GBADs outputs will be useful for experts from nutrition, environmental science, and related disciplines to help generate an approach to ASF production and consumption that balances the health of humans, animals, and the planet.