What the obesity epidemic does not need: A cancel culture

Abstract

A paper recently published in Lifestyle Medicine highlights the importance of informed and respectful debate as part of the scientific endeavour. Appearing in the first edition of the journal, Jacob et al.¹ examined data from 7403 participants from the 2007 Adult Psychiatric Morbidity Survey (APMS), a nationally representative survey of the English adult population conducted by the National Centre for Social Research and Leicester University.² They found that there was a negative association between Verbal IQ, estimated using the National Adult Reading Test (NART), and obesity. During peer review, the manuscript was praised for its clear and robust statistical analyses. However, due to the nature of the topic, editorial review included assessing the manuscript for inappropriate or discriminatory language or conclusions, as well as ensuring both scientific and analytical merit in line with the journal's scope. Perhaps unsurprisingly, the paper still resulted in significant vocal debate on social media (Twitter), including calls for the paper's retraction. As Twitter does not uniformly allow for reasoned discussions, formal letters to the editor were solicited by Wiley to highlight any significant issues and allow for formal response by the original authors. This process is still an important aspect of moving scientific research forward, and as a journal we follow both International Committee of Medical Journal Editors (ICMJE)³ and the Committee on Publication Ethics (COPE) publishing principles,⁴ which include giving a platform for rebuttal to published articles and encouraging logical and reasoned scientific debate. Both the letter to the editor by Redsell et al.⁵ and the response by Jacob et al.⁶ are included in this issue. While we stand by the publication of the manuscript – as further outlined below – we believe that it is the job of the scientific community at large to iteratively move a field forward based on discussions such as those highlighted here. Therefore, to some extent, each reader can and will make up their own minds based on the evidence, following scientific principles to do so. We hope this commentary assists that process.

Using weight or normative estimators of body composition, such as the body mass index (BMI), as predictors of individual health or health outcomes is both controversial and highly emotive.⁷ This is clear from the tone of both the letter by Redsell et al.⁵ and the response by Jacob et al.⁶ One thing that we feel is absolutely necessary in order to move discussions in this field forward is the separation of our personal feelings about a topic from the scientific arguments, as much as that is possible. In line with that, both the letter writers and article authors were given opportunities to tone down their language, and both articles have been lightly edited (in language but not content) to more constructively moderate the discourse.

A discrete focus on the scientific arguments and process of a manuscript is important because Redsell and co-authors present unsubstantiated accusations such as that the manuscript is ‘ill-advised (at best)’, and include veiled threats that ‘this paper is also likely to cause harm, which may risk reputational damage’. The latter assertion was largely focused on how the findings may be portrayed by the media. We absolutely agree that accurate reporting of scientific findings in the media – as well as ensuring that research is not discriminatory – is an essential part of science publication and communication. However, the broader implication in the letter appears to be that research into populations at particular risk of, or potential downstream (health) consequences of, obesity should be avoided because it may be inaccurately portrayed in the media and used to propagate weight stigma. This argument also includes non-sequiturs in the assumptions that risk factors for obesity usually regarded as non-modifiable are not of interest for potential targeting of modifiable risk factors, and that patients and the media will necessarily fuel negative stereotypes. Patients, including the patient representative who helped write the editorial in our last issue,⁸ are clear that they want to see the evidence for themselves, which the open access model supports. Therefore, it would be a curious approach to sanitise the content due to potential for misinterpretation rather than accept submitted articles based on their scientific quality.

Redsell et al.⁵ also suggest that the original study suffered from a number of methodological flaws and breached two ethical principles (beneficence and justice). Though there is some discussion regarding the utility of the National Adult Reading Test in this setting, no clear methodological flaws – either in the analysis or the interpretation – were raised. We agree that the quality of a data used to examine a specific question can almost always be improved. For instance, it must be recognised that societal inequities are significant drivers of both obesity and 'performance' in normative tests such as those for intelligence quotient (IQ)⁹ with the history, use, and interpretation of the original IQ tests grounded in both sexism and racism.^{10, 11} Jacob et al. performed multiple sensitivity analyses to adjust for individual background and environmental factors that might influence the outcome, but the direction and magnitude of the effect remained consistent. Importantly, though the authors could have been more explicit about other unknown confounders, as well as the fact that socioeconomic parameters such as race, income, and education are often proxies for much greater social inequities and a wide range of lived experiences,¹² they performed appropriate adjustments with the available data and were clear about the limitations. This does not preclude the possibility that different results will be seen in the future if datasets containing a broader range of equitable cognitive tests and accurate body composition measurements are collated and analysed. Advances in epidemiology like other branches of science is iterative and while no clinical research study is perfect, constructive criticism is a key driver of improving study design and conduct.

The intended goals of public health epidemiology, and what can and cannot be concluded from a particular dataset or analysis, is a crucial aspect of evidence-based medicine. Redsell et al.⁵ do rightly state that the original manuscript only ‘explores the association between IQ and obesity, not causation’, while also accusing the authors of implying causation. The former is absolutely true, as is the case with any epidemiological study, but Jacob et al⁶ clearly do not attempt to assume causality, presenting previous work that could drive causality (if present) in either direction. Multiple factors may explain the results, from unknown confounding by socioeconomic and environmental factors to an inflammatory phenotype associated with visceral adipose tissue that may impact the brain,^{9, 13} and while these questions remain it is imperative that objective research continue. The way the results are written by Jacob et al. is largely driven by the requirements of multivariate statistical analyses, where a response variable (in this case obesity as a binary outcome) must be chosen, along with associated predictor variables. The outcomes of these models include an odds ratio for obesity, but do not imply that the predictors (e.g., Verbal IQ) are directly causative.

Redsell et al.⁵ do also highlight papers that support the potential for better metrics and assessments for use when studying this question in the future, though one might argue that their representation of cited literature differs from that stated in the papers themselves. For instance, citing McGurn et al. as evidence that the performance of the NART was only modest when that paper describes a strong enough correlation to validate the use of the test.¹⁴ As with any standardised test, we agree with Redsell et al.⁵ that the variance in NART scores (as well as BMI) due to other factors such as material and social deprivation is both possible and likely, but this is at least partly addressed by the statistical adjustments and does not invalidate the results as presented. We also agree that self-reported BMI does not perfectly reflect clinical measurements,¹⁵ as also acknowledged by Jacob et al.⁶ in the original manuscript, although self-report is generally regarded as a valid approach. If anything, self-report is likely to lead to an under-estimate of BMI so is likely to have underestimated the prevalence of obesity in this dataset and as such diluted any associations.

Correct attribution and citation is important, so the letter includes full references inserted at the proofing stage, which clarifies which publications (especially newspapers) may be more likely to misinterpret the science and create headlines, albeit that the references are not recent or relating to policy in the country of publication. We do operate in an international environment, so global awareness is important too, both in not bowing to excessive perceived pressure from the approach of certain UK newspapers nor being influenced by censoring in more totalitarian regimes. There is a balance to strike here: recent events in the Washington Capitol and social media's involvement and subsequent repudiation of longstanding approaches show this is a developing area.

We do strongly believe that every person has the right to feel safe, comfortable, and happy in their body, and multiple studies have indeed shown that damaging weight stigma is incredibly prevalent in the popular media, contributing significantly to society's negative viewpoints around body weight.¹⁶ Work by the “health at every size” (HAES) movement is both important and growing, often highlighting the fact that societal weight stigma and diet culture can significantly negatively impact the health and quality of life of certain targeted individuals,^17-20 as well as the fact that BMI or body composition may not be deterministic with respect to long-term health outcomes.^{7, 21, 22} These are important issues must be addressed head-on, for instance by tackling dangerous aspects of diet culture and media misrepresentation of weight and health, as well as clearly separating out epidemiological work designed to improve risk stratification and policy at the societal level from the nuanced, holistic, and personalised approaches that should be taken to improve an individual's health trajectory with their consent and concordance, which will most often include interventions that are weight neutral. These societal issues should not, however, prevent obesity research at the population and mechanistic levels being performed to identify those at risk, or potential interventions, particularly within systems and institutions. Identifying at-risk populations to help institute systemic change, rather than place health solely as the responsibility of the individual^{23, 24} is an important goal of research using population datasets such as that presented by Jacob et al.,¹ and these studies are never intended to be used to make statements about individuals. However, there is compelling evidence that obesity is a strong driver of poor health outcomes, more powerful for many than physical inactivity.²⁵ Depending on the criteria used, 7–50% of those with obesity are metabolically healthy,²⁶ but even metabolically healthy obesity appears to be associated with an intermediate risk profile for outcomes such as cardiovascular disease.²⁷ Recent systematic reviews have also suggested that (i) weight loss interventions may improve health-related quality of life, with the effect potentially correlated with magnitude of weight loss,²⁸ and (ii) that behavioural weight management interventions result in improvements in a number of mental health outcomes, including body image concerns and self-efficacy,²⁹ without negatively impacting overall mental health. None of this negates an individual's experience of weight and weight stigma, but instead highlights the importance of objective assessment of the evidence at a higher level coupled with individualised patient-centric approaches. Therefore, those who want to help those impacted may be better served by encouraging the use of valid tools for both population and individual risk assessment, as this can inform empowering choices by the individuals, when shared in appropriate ways, which will vary by country, setting and individual.³⁰

In summary, with respect to the original manuscript and subsequent communications, we respectfully maintain that the original manuscript meets our criteria for publication. Retraction is not justified on the basis of the arguments presented. However, we welcome rational scientific debate and hope that by publishing both the letter and reply we can encourage a constructive collaborative approach to destigmatising obesity rather than shaming those who do not avoid addressing difficult and complex issues.

FB is Editor-in-Chief for the Wiley open access journal Lifestyle Medicine and Director of Science and Research for the British Society of Lifestyle Medicine. SK is a bariatric physician, Regional Director for South Wales & on Nutrition Interface Group for the British Society of Lifestyle Medicine. TW is an Associate Editor for the Wiley open access journal Lifestyle Medicine and is a Trustee of British Society of Lifestyle Medicine.