What principles should guide interactions between population health researchers and the food industry? Systematic scoping review of peer‐reviewed and grey literature

Summary There is no explicit consensus amongst population health researchers regarding what constitutes acceptable or effective interactions with the food industry. This has led to confusion and disagreements over conflicts of interest, which can undermine the integrity of science. To clarify this issue, we aimed to systematically identify the key principles developed by population health researchers to prevent or minimize conflicts of interest when interacting with the food industry. Databases of peer‐reviewed literature were searched. In addition, an advanced Google search, a request to experts seeking related documents, and hand searching of references were undertaken. Thematic analysis of the extracted data was undertaken. We examined 54 eligible documents describing guidelines for population health researchers when interacting with the food industry. Fifty‐six principles were identified and synthesized in five themes. There were high levels of agreement in themes relating to research governance, transparency, and publication but less agreement and guidance on how principles should be applied in relation to funding and risk assessment. There is agreement on some of the general principles for preventing and minimizing conflicts of interests for population health researchers when interacting with the food industry. However, for issues such as assessing the appropriateness of an industry partner, greater clarity and consensus are required.

by funding bodies and research institutions. 4,5 However, the primary purposes of the food industry (to maximize profit) and population health researchers (to further knowledge to inform health improvement) are often poorly aligned, leading to the potential for conflicts of interest. These real and perceived conflicts of interest can undermine the credibility of research and researchers, resulting in an erosion of trust amongst the general public and policy makers and scepticism of published research. [6][7][8][9] Despite these concerns, there is currently no explicit international consensus for population health researchers regarding what constitutes acceptable or effective interactions with the food industry. 10,11 A significant factor informing acceptable interactions is preventing the damaging impacts of actual or perceived conflict of interests.
Whilst guidance regarding acceptable interactions and conflicts of interest involving the food industry has been developed for nongovernment organizations 12 and for policy makers, 13,14 limited work has been undertaken specifically for population health researchers, addressing the unique challenges they face. Consensus-driven principles for nutrition researchers (including those working in population health) have been developed in the United States of America by representatives of US government agencies, the food industry, and professional nutrition associations to guide research partnerships. [15][16][17] However, the involvement of the food industry in the development of researcher engagement principles is considered problematic by many. 8,18 The absence of clear consensus on what are acceptable interactions between population health researchers and the food industry has given rise to disagreements and confusion, which can further undermine credibility and integrity of nutritional and dietary public health science, further eroding trust and exacerbating scepticism.
Confusion remains for a number of reasons. Firstly, there is a lack of clarity around the term "conflict of interest." One commonly cited definition of conflict of interest is "a set of conditions in which professional judgment concerning a primary interest (such as a patient's welfare or the validity of research) tends to be unduly influenced by a secondary interest (such as financial gain)." 19 Financial gain tends to be the focus for conflict of interest guidance. 20 However, financial gain is not the only factor that may lead to conflicts of interest for researchers. The desire for recognition, academic advancement, and success in publication and funding are other powerful influences. 21 Furthermore, conflicts of interest can also arise from researcher's personal relationships and business associations. 11 A conflict of interest can also be perceived to exist even when it does not. Perceived conflicts of interest can be just as important as actual conflicts of interest in undermining trust in research and researchers. 9,22,23 This perception is not always unwarranted, with many examples of the food industry donating funds to, or collaborating with, universities as part of a deliberate strategy to improve their reputation and to influence the evidence base. 7,[24][25][26][27] Secondly, the type of interaction that population health researchers may have with the food industry is varied and may have different consequences. Financial transactions, for example, research grants, are the most publicized form of interaction. 7 These are often supported by government and research institutions, particularly in the face of declining public funds for research. 17,28 However, interactions may take other forms, such as in-kind funding, which includes access to resources, travel costs, or honoraria or direct dialogue, which could include providing advice on industry initiatives. 13 There is limited acknowledgement of the existence of these other forms of interaction or guiding principles about them.
Finally, the "food industry" includes a heterogeneous range of companies and products. Different food products have different consequences for health, and many companies have a mixed portfolio of products. Furthermore, companies themselves may engage in health promoting, health damaging, or health neutral behaviours beyond the products they sell. This heterogeneity makes it difficult to come up with overarching principles for appropriate interactions.
To help clarify what constitutes appropriate interactions between population health researchers and the food industry, we aimed to systematically harvest from the literature, synthesize and analyze the key principles that have been identified to help prevent, or manage actual or perceived conflict of interest.

| METHODS
Our protocol was informed by the Arksey and O'Malley framework for conducting a scoping review. 29 A scoping review is a rapid form of knowledge synthesis designed to identify gaps in the evidence base.
It is an iterative process guided by a requirement to identify all relevant literature regardless of study design. 29 The review protocol was registered with PROSPERO, registration no. CRD42017060539.

| Search strategy
We conducted a number of different searches between March and June 2017 to identify relevant peer-reviewed and grey literature. Firstly, a Box 1 Search strategy, inclusion, and exclusion criteria Search strategy: The search strategy included subject heading and text word terms appropriate to each database. Combinations and synonyms of the following terms were used: "conflict* of interest," OR "public*private partnership*" OR "university-industry relations" AND "diet" OR "nutrition" OR "food" OR "obesity" OR "food industry" AND "manag*" OR "guiding principle*" OR "codes of conduct" OR "framework" OR "standard."

Inclusion criteria:
Documents that evaluate, compare, use, or describe a process, framework, or practise for identifying and/or preventing and/or managing conflict of interest in privatepublic partnerships between population health researchers comprehensive search was conducted on the full holdings of PubMed and SCOPUS databases with no restriction on date or country of publication, although only documents in English were reviewed. The search strategy and inclusion and exclusion criteria can be found in Box 1.
A Google advanced search using the same search terms, inclusions, and exclusions as above, was undertaken to identify grey literature.
The addition of "filetype: PDF" was used to further refine the search as most relevant documents were in a PDF format. Where more than 50 results were found, only the first 50 were scanned, as initial searches revealed no new documents after this.
Additionally, over 200 key informants in the field, identified via international public health nutrition associations and other databases, were contacted and asked to nominate documents for inclusion.
Finally, reference lists of included documents were hand searched for additional eligible documents.
Database searches were managed using EndNote X5. One author (K.C.) screened retrieved titles and abstracts for eligibility.
The full texts of potentially eligible documents were then retrieved and K.C. reviewed these to determine whether they met the inclusion

| Data extraction and synthesis
Documents from all sources were analyzed until concept saturation was reached; that is, no new information, themes, or descriptions of principles for addressing conflicts of interest were found. 30 To support this, a data extraction form was developed and reviewed by all authors. The extracted data included publication year, country, whether there was a definition of conflict of interest, the principles or activities for preventing or managing conflict of interest, the target users, whether competing interests were disclosed, and whether the reviewer perceived the authors had a conflict of interest based on their place of employment. Data extraction was undertaken by K.C.
No formal quality assessment was undertaken following standard practise for scoping reviews. 29 Key extracted principles were imported into NVivo v11 for analysis. 31 An inductive process of coding and analysis was used.
This allowed the initial codes and then the subsequent themes and names for the themes to flow from the data. 32 This process was initially undertaken by K.C.; M.W. duplicate coded 15% of included documents. A flow diagram to illustrate how the themes interact was also constructed to aid analysis.
In order to explore the origins and evolution of ideas in this body of literature, we also undertook a network analysis of all citations.  Six (11%) were focused on issues specific to low and/or middle income countries, and the remainder did not specify the audience or context. Forty one percent of documents were clearly in favour of necessary interactions with private industry (n = 22), 24% were against (n = 13), and the remainder took a neutral position. For those documents clearly in favour of private industry interactions, half (n = 11) had either declared funding by the food or beverage industry or had authors who were employed by the food and beverage industry. The most comprehensive guidance for those working in population health was designed for non-governmental organizations 34,35 or government agencies considering private investment in nutrition intervention programmes. 13,36 The most comprehensive guidance specifically designed for researchers was developed by Rowe et al, 15,37 and this was referred to as an accepted standard in other documents (particularly those with a pro-industry engagement stance). 10,16,17,[38][39][40][41] In the majority of cases, stated principles appeared to be based on authors' opinions, rather than empirical research. Our network analysis of citations (see Figure 2) revealed two documents that had a higher number of citations than the others, 34 Table 1).

| Structure and governance of funding
The research funding process frequently featured as a critical component for preventing conflicts of interest. This included two elements: the appropriateness of accepting funds from the food industry and the governance processes for accepting funds. There was a variety of principles related to the appropriateness of accepting funding;  1.5 Researchers should not accept funds from any commercial organization 45,46 For those who accept funding from the food industry 1.6 Researchers should have no commercial interest in the product being researched 42 1.7 Funding from industry should reflect the full cost of the research (eg, using a standard academic costing framework) and not more than this amount 42 1.8 Industry funding should be nondesignated 51 1.9 There should be no involvement of the food industry funder in any aspect of a research project 27 1.10 There should be limited involvement of the funder in any aspect of the project 52 2. Undertake thorough risk assessment 2.8 Avoid companies whose objectives and/or goals are related to the increased production, supply or demand of "unhealthy food" products and/or to the promotion of unhealthy and unsustainable ways of eating and producing food 35  3.2 Plan research so it is designed objectively and is scientifically sound in its approach 37,39,41,44,67 3.3 Establish upfront control and ownership of the data by the researcher/s but provide accessibility to data and analyses to the industry funder 37,40,67,68 3.4 Data analysis should be done by statisticians independent of the researcher/s who designed and conducted the study 8,42 3.5 Undertake random audits of data provided by food companies for research projects 10 3.6 Secure oversight of the research by a nonconflicted third party 10 For those accepting funding from the food industry, there was also a range of views on the role that the food industry should take in a research project. Several of the statements were ambiguous and provided little detail on the practicalities of this arrangement.
For example, there should be "little involvement from the funder" 22 but with no explanation of what constitutes "little involvement."

| Undertake a risk assessment
For those organizations or researchers considering interacting with the food industry, a key theme was the need to undertake a thorough risk assessment before proceeding. This included not only assessing the suitability of a food company, as well as the type of interaction involved.
In order to assess the suitability of a food company for an interac-

| Ensuring high levels of transparency
The most common, and sometimes only, strategy noted to prevent and manage potential conflict of interests was full disclosure. 24,65,70 However, the definition of full disclosure varied. Several documents stated that "conflicts of interest should be declared" without defining a "conflict of interest." 7,24,56,70,74 Others provided clarification by specifying that conflicts of interest that were financial, personal, or professional should be disclosed 8,68,72 or that the interests of the spouse or partner, minor children, employer, or business partners of all involved in the research partnership should also be disclosed. 13,64 The other element of this theme was ensuring that full disclosure occurred with all elements of research dissemination whether in the form of a project report, journal article, conference presentation, press release, or a public talk. 22 A flow diagram summarizing the researcher decision-making process was synthesized from the literature, using a process of thematic analysis and theorizing, to aid further analysis (see Figure 3). It demonstrated that there is general agreement on the decision-making process, but the specifics of decision making for certain steps were unclear.

| Principal findings
Our systematic scoping review identified 56 principles from 54 documents. These covered five key areas: governance of funding, risk assessment, maintaining standards of governance, ensuring transparency, and improving publication standards. The review has highlighted the wide range of principles proposed to prevent or manage conflicts of interest between population health researchers and the food industry. A high level of agreement between authors exists on some of these principles, whilst others remain more contentious, particularly those related to whether it is appropriate to interact with the food industry and what kinds of interactions are considered problematic.
There was limited guidance on how the principles should be applied in practise.

| Relationship to existing knowledge
There was more consensus and detailed guidance related to three themes: "ensuring high levels of transparency," "maintaining high standards of research governance," and "improving publication standards." These encompass established and foundational principles of research and involve unequivocal decision making. So it is perhaps unsurprising that they had high levels of support. However, within the theme "ensuring high levels of transparency," the concept of "full disclosure" meant different things in different documents. Although there was generally support for individual financial disclosure, acknowledgement of the need for more extensive disclosure was more limited. Agreeing on a standard for full disclosure is important as it provides those reading or listening to findings with a basis for drawing their own conclusions regarding potential for bias and confidence in the findings. 77,78 The flow diagram (Figure 3) highlights the importance of risk assessment processes in making decisions about interacting with the food industry. However, few documents specified criteria or a structured process to guide this. The lack of specific criteria highlights the moral as well as scientific values that should be considered when assessing these issues. 79 Without clear guidance in this area, researchers may have trouble conducting risk assessments. Furthermore, the ownership arrangements of many food companies make term. Without acknowledging these risks, it is hard to identify and prevent or manage them. 55,57 Amongst those authors who were clearly against interactions with industry, there was limited acknowledgement of its potential benefits. 42 Furthermore, there was a lack of acknowledgement that bias can occur from a number of perspectives, and whilst not as extensively documented, this has been found to occur in researchers' opposition to industry interests resulting in overstatement of results. 80,81 The biases created by this emotive and value-based environment-both conscious and unconscious-make it difficult for all parties to critically assess the research opportunities offered by interaction with the food industry and related research designs and methods. Finding a way to objectively assess such research opportunities and methods is critical to advancing the integrity and credibility of population health research on diet and nutrition.

| Strengths and weakness of the study
Whilst much has been written on the topic of conflicts of interest between researchers and the food industry, 7,26,27,77 this is the first study to systematically review and synthesize the range of principles that have been proposed to manage relationships between population health researchers and the food industry so as to prevent or minimize conflicts of interest. The comprehensiveness of the search and the use of two independent reviewers for 15% of the screening, exclusions, data extraction, and coding increase the reliability and validity of the results.
It is important to note that the material analyzed in the review was largely narrative in nature and in the form of discussion papers or guidance. It was not clear in most cases whether the principles cited or related guidance were empirically derived. However, the purpose of this review was to identify exhaustively the range of principles that are currently considered in preventing and managing conflict of interest and then to critically analyze and synthesize these.

| CONCLUSIONS AND IMPLICATIONS
In the light of constrained public sector funding for research, particularly in low-and middle-income countries and the proposition that some population health research questions on diet and nutrition would benefit from industry involvement, it is likely that researcher interaction with the food industry will continue and may benefit population health. For those population health researchers considering interaction with the food industry, our findings highlight that there is a level of consensus on the principles relating to standards of research governance, transparency, and publication. However, in relation to assessing the appropriateness of potential industry interaction and the type of interaction, greater clarity is required to ensure that trust in dietary public health research remains high.
One way to achieve greater clarity on these issues is through formal consensus building. In addition, rather than abstract, high-level principles, clear action-guiding recommendations for assessing oppor-

CONFLICTS OF INTEREST
The authors have no financial relationships with any organizations that might have an interest in the submitted work. To accord with the principles espoused in this research, we have prepared a detailed declaration of interests for all co-authors, which can be found in a supplementary file.