Rising prevalence of BMI ≥40 kg/m2: A high‐demand epidemic needing better documentation

Summary Whilst previously rare, some surveys indicate substantial increases in the population with body mass index (BMI) ≥40 kg/m2 since the 1980s. Clinicians report emerging care challenges for this population, often with high resource demands. Accurate prevalence data, gathered using reliable methods, are needed to inform health care practice, planning, and research. We searched digitally for English language sources with measured prevalence data on adult BMI ≥40 collected since 2010. The search strategy included sources identified from recent work by NCD‐RisC (2017), grey sources, a literature search to find current sources, and digital snowball searching. Eighteen countries, across five continents, reported BMI ≥40 prevalence data in surveys since 2010: 12% of eligible national surveys examined. Prevalence of BMI ≥40 ranged from 1.3% (Spain) to 7.7% (USA) for all adults, 0.7% (Serbia) to 5.6% (USA) for men, and 1.8% (Poland) to 9.7% (USA) for women. Limited trend data covering recent decades support significant growth of BMI ≥40 population. Methodological limitations include small samples and data collection methods likely to exclude people with very high BMIs. BMI ≥40 data are not routinely reported in international surveys. Lack of data impairs surveillance of population trends, understanding of causation, and societal provision for individuals living with higher weights.

due to lack of awareness and limited evidence base. This problem is compounded by stigmatization over size and weight by the media, within society and among health professionals. 17 Furthermore, conventional behavioural weight management interventions have minimal impact for BMI ≥40. 18 Only a minority of individuals access bariatric surgery, 19 and effective nonsurgical interventions are not yet widely available. 20 Thus, once individuals reach a very high BMI, the potential for sustained weight loss to improve quality of life and reduce secondary medical complications is limited.
The severity and impact of muscular-skeletal complications reduce physical activity and mobility, increasing dependence on others and putting weight loss further out of reach. Consequently, individuals with BMI ≥40 face reduced life expectancy, 21 multimorbidity, 22 disability, 23 and reduced quality of life. 24 In turn, this disease burden produces multifaceted demands on health and social care services, raising direct and indirect costs. 25 Until recently, evidence on direct costs of BMI ≥40 have been limited. 25 Total health care costs in the United Kingdom rise linearly, and double, as BMI increases 26 from 20 to 40. A recent systematic review of international health care costs and BMI found costs for people with BMI ≥40 to be 50% greater than for people with BMI 18.5 to 24.9. 27 Costing studies typically exclude underresearched wider care costs, such as social care 28 and nursing home usage related to functional disability and often long-term provision, 28,29 so current estimates are likely to underestimate the full costs. Forecasts indicate that increased resource usage will continue, including costs required to structurally adapt care facilities to the needs of people with BMI ≥40, alongside providing suitable equipment and training for staff, that is currently missing. 30 The present scoping review explores the extent of international prevalence data on BMI ≥40. It focusses on measured data, given the potential for error and bias with self-reported anthropometry by individuals who are overweight. 31,32 We assessed the extent and quality of epidemiological reporting for the BMI ≥40 category internationally, with a view to improving the documentation of this emerging highdemand population in future national surveys, to enable development of a reliable evidence base to guide effective care.

| METHODS
The primary epidemiological reports being investigated are health surveys, undertaken by governments for population surveillance to inform strategic policy priorities. Health surveys are not usually reported in the academic literature, unless for a secondary analysis focussed on a specific issue or subpopulation, often with considerable time lag. Thus, database search terms for the primary survey would need to be very broad, making identification through academic databases highly resource intensive, with results prone to being incomplete and outdated. For the present scoping review, an alternative search strategy was therefore applied, based on the sources identified by the most recent systematic review of international BMI survey data, with additional searches to update and supplement these sources, outlined below.
Four key approaches were used to identify potential data sources:

Building on previous work
The NCD Risk Factor Collaboration (NCD-RisC) 2017 study on global BMI trends was chosen as a basis for the initial search, on the basis of its size and rigour: It used 2416 sources of measured data, collected up to 2016. 33 These sources were compiled from a systematic medical database search, supplemented by a worldwide network of researchers identifying and accessing national measurement surveys via interested parties, including World Health Organization (WHO). Full details are given in the published paper and its appendix. 33 All 2416 sources in the NCD-RisC appendix were screened as per the inclusion criteria in Table 1 in Appendix A. Results from conference proceedings were deliberately included, as they can provide the first exposure of research analysis, highlighting new or updated data sources, whilst full articles commonly take much longer to publication, if indeed they are ever published in full. Sources were searched both for BMI ≥40 data, but also to identify any new sources/health surveys not already located through parts 1 and 2 of the search process.

Digital snowball searching
Whilst NCD-RisC identified sources, individual sources needed located and accessed, often directly from a website of government agencies or international bodies. This was undertaken digitally, seeking previously unknown sources or articles.
All sources were screened using inclusion criteria in Table 1. Prevalence rates are presented graphically in Figure 2A-C. The United States has the highest prevalence, 49 with rates across all adults, men, and women ranging between 5.6% and 9.7%. Other countries with higher rates (5.0%-7.7%) for all adults and women are New Zealand, 45 Kuwait, 42 and Barbados, 47 although rates in men are markedly lower. Australia, 44 Canada, 50 Scotland, 36 England, 35 and Saudi Arabia all have rates in the region of 2.5% to 5.5% for all adults and/or women, again with men notably lower. Germany, Serbia, 40 Spain, Portugal, Poland, Brazil, and Malaysia 52 display the lowest rates between 0.7% and 2.8%. Brunei Darussalam and Seychelles both exhibit disparate prevalence patterns from those above for men and women. to be relatively small, with 13 countries being <8000, and five above this. With prevalence under 10% of these numbers, wide confidence intervals result for BMI ≥40, as seen in Table 2.
Less than half of the countries included have regular surveillance, ranging from annual to three to four yearly (Table 2: England, Scotland, Canada, New Zealand, Australia, and the United States), significantly limiting analysis of trends over time. Reporting from other countries appears to be more ad hoc and unpredictable, without planning for regular surveillance.

| Data quality
To enable assessment of data quality, the column headers of Table 2 highlight basic quality parameters appropriate to national health surveillance, 58     Comparison with analyses of self-report surveys from the United States and Canada, which allow for much larger samples, all show disproportionately larger growth in higher BMI categories, with increases of up to 10-fold in the BMI ≥40 category. [64][65][66] This is despite suggestions that the underestimation of weight for self-report data is likely to be greater with higher BMI. 31 Thus, despite wide confidence intervals in individual survey years and some variations in methodologies, measured trend data evidence from several countries, together with larger-scale self-report datasets, all support a long-term rise in BMI

≥40.
Two key sources of data for NCD-RisC and WOF were STEPS reports and Demographic Health Survey (DHS) country reports.
The STEPwise approach, featuring three different level of "steps" of key risk factor assessments for NCDs, was developed by WHO, to aid countries increase their surveillance capacity. 67 One hundred and thirteen country reports or data sheets were publicly available on the WHO STEPS site at the time of searching. 68  International health surveillance provides reliable health information that is comparable over time and between populations. It has allowed documentation of a nutrition transition that is rooted in the impact of large-scale social changes such as reduction in physical activity, increased urbanization, and greater consumption of processed food. 70 Whilst regional variation remains, the major global concern has moved away from underweight as the primary nutrition issue, towards excess weight as the leading cause of morbidity and mortality through secondary noncommunicable diseases. 23 This has already occurred to such a degree that BMI ≥35 in women now surpasses underweight in 165 countries for women and 113 for men. 3  The BMI ≥40 data were later reported by the analysis of Weiderpass et al of the original STEPS dataset, illustrating that the data had been collected but gone unreported. 42 Table 2  given that the data are already collected. For countries with small numbers to report, the need for caution in interpretation would be dealt with in the same way for low numbers in any category, for example, underweight.

| Causation
The lack of data on BMI ≥40 trajectories by region, nation, age, sex, and class makes it difficult to explore causation. Improved international data, ideally from longitudinal studies, would promote comparison between countries, taking into account their differing social and economic contexts. 73 Together with the emergence of large-scale genome studies looking at the inherited susceptibility to BMI ≥40, 74 reasons for the escalation of high body weight may be more accurately sought. There are some indications of associations with lower socio-economic status in some populations, 75

| Consequences of rise in prevalence
Whilst numbers may appear small in terms of proportion of the whole population, given that these are at national scale, they translate into significant absolute numbers with a large real-life impact on care provision.

| Health risk and comorbidities
It is well recognized that BMI can be a poor proxy for body fat, and in many studies, waist circumference or evidence-based predictive equations for total body fat are better than BMI for assessing health risk, particularly cardiac and metabolic health outcomes. 81 29 As such care of these multiple obesityrelated consequences amplifies health care costs considerably. 89 Given the rise in prevalence and size of these costs, in addition to older costing studies being limited by BMI thresholds that are now too low, this is likely to become of increasing concern globally.

| Planning
As an emerging population, people with BMI ≥40, especially those with BMI ≥50 and ≥60, have needs that are currently often unaddressed by service providers. 4

| Strengths and limitations
This review has concentrated on robustly measured data to establish the international prevalence rates for BMI ≥40. Whilst the rationale for exclusion of self-report data is sound, in that it commonly underestimates BMI, the exclusion also acts as a limitation, for example, by excluding EHIS data, which covers many European countries. The sample sizes possible with measured data are reduced by the need for resources to make measurements, and there is potential bias against including very heavy individuals whose mobility is impaired. In some cases, the upper limit of scales Some studies report the lower threshold of BMI ≥35 particularly those examining Asian populations where different BMI cut-offs relating to overweight and obesity are often applied. 98 Limiting the review to the English language prevented examination of some original data sources, which could only be located in their native language, for example, Spanish for Chile and Mexico. It was not possible to locate English versions of these, and resources did not allow for translation. The OECD reports these original sources in English in its database, but only at BMI thresholds of 25.0 to 29.9 and BMI ≥30, with prevalence for Chile and Mexico the highest in the world, above even those of the United States. 99 Thus, they are likely to have significant BMI ≥40 prevalence. Additionally, whilst the search processes were broad, encompassing a variety of sources and used systematic methods, they were not exhaustive, as might be expected from a formal systematic literature review or meta-analysis.
We believe that they represent a reliable summary of the current evidence base on BMI ≥40, as it is available to decision makers. Some individual sources have not been included, notably those from non-English publications, but they are unlikely to alter the very consistent conclusions.

ACKNOWLEDGEMENTS
Thanks to Welcome Wami and Andy Peters for providing statistical support.

CONFLICT OF INTEREST
The authors declare no competing financial interests in relation to this work. Michael Lean is funded by the University of Glasgow.