Skip to main content
NCBI home page
BMJ Open logoLink to BMJ Open
. 2025 Oct 7;15(10):e092962. doi: 10.1136/bmjopen-2024-092962

Associations of body fat and inflammation with non-communicable chronic diseases and mortality: a prospective cohort study of the UK Biobank

Natasha Wiebe 1,✉, Marcello Tonelli 2
PMCID: PMC12506142  PMID: 41062143

Abstract

Abstract

Objective

Certain leading medical organisations are considering alternatives to the Body Mass Index (BMI) as a predictor of the risk for non-communicable chronic disease (NCD) or death. Our objective was to evaluate the associations between various measures of body fat and the risk of incident NCDs or mortality, independent of inflammation.

Design

Population-based prospective cohort study (the UK Biobank cohort).

Setting

The UK.

Participants

Adults (aged between 40 and 69 years) were accrued between March 2006 and October 2010 and followed until December 2022. There were 500 107 participants: the median age was 58 years (IQR 50–63) at baseline, 45.6% were male and 94.7% were white.

Exposures

BMI, waist-to-hip ratio (WHR), body fat percentage measured by bioimpedance analysis (BIA; fatBIA), C-reactive protein (CRP) and various other measures of body fat obtained by dual-energy X-ray absorptiometry (DXA; including visceral adipose tissue (VAT)) and magnetic resonance imaging (MRI).

Outcomes

All-cause death, cardiovascular disease (heart failure, hypertension, myocardial infarction, pulmonary embolism and stroke), cancers (breast, colorectal, endometrial, oesophageal, kidney, ovarian, pancreatic and prostate), diabetes, asthma, gallbladder disease, chronic back pain and osteoarthritis.

Results

The 5th and 95th percentiles for measures of body fat were BMI 20.5 (considered ‘healthy’) and 37.0 kg/m2 (considered ‘unhealthy’), WHR 0.71 and 0.94 and BIA 24.8% and 47.6% in females, and BMI 22.0 (considered ‘healthy’) and 35.4 kg/m2 (considered ‘unhealthy’), WHR 0.83 and 1.05 and BIA 15.5% and 34.7% in males. BMI was strongly correlated to fatBIA (0.85 in females and 0.80 in males) but less so with WHR (0.46 in females and 0.59 in males). All measures of body fat were positively associated with the incidence of NCDs, but only WHR remained positively associated with death after full adjustment (HR 95th percentile vs 5th percentile (95% CI): BMI 0.80 (0.76 to 0.84), WHR 1.21 (1.16 to 1.28) and BIA 0.80 (0.76 to 0.84) in females; BMI 0.89 (0.85 to 0.93), WHR 1.19 (1.14 to 1.24) and BIA 0.89 (0.85 to 0.92) in males). Simpler models that adjusted for age, sex, CRP, WHR and either BMI or fatBIA gave similar results. Associations between body fat and the incidence of NCDs after accounting for the competing risk of death were also similar.

Conclusions

BMI was strongly correlated with fatBIA, but WHR and visceral adipose tissue percentage were less so. All measures of body fat were associated with the incidence of NCDs, but only WHR was independently associated with mortality. These findings support the hypothesis that body fat may be protective against death and that the excess risk associated with higher WHR may be mediated by something other than body fat.

Keywords: Obesity, EPIDEMIOLOGY, PUBLIC HEALTH, Body Mass Index, Cardiovascular Disease, Death

STRENGTHS AND LIMITATIONS OF THIS STUDY.

  • This was a large prospective study (over half a million participants), with a wide array of fat measures and a long follow-up period (median 14 years).

  • There was no data on weight cycling, weight stigma, weight-based discrimination and hyperinsulinaemia/insulin resistance, which are common in higher-weight patients; thus, important residual confounding remains possible.

  • The study participants were not representative of the general UK population; in particular, they had fewer health conditions.

Introduction

Obesity is associated with an increased prevalence of non-communicable chronic disease (NCD),1 particularly type 2 diabetes, cardiovascular and pulmonary diseases and cancer.2 3 However, obesity (as assessed by high Body Mass Index (BMI)) is not associated with excess mortality in many clinical populations.4 In fact, after adjustment for key confounders, such as inflammation and/or fasting insulin, higher BMI was associated with a lower risk of mortality in a representative population of US adults.5 Critics have argued that collider bias may explain this ‘obesity paradox’, but studies in general populations thus far, where selection bias does not occur, do not support this theory.4 6

Both inflammation and hyperinsulinaemia are associated with excess risk of incident NCDs; however, neither has been shown to be caused by obesity.7 There is some evidence from human and animal studies that chronic elevations of C-reactive protein (CRP) may precede or even cause adult-onset obesity and possibly insulin resistance (IR).8,12 Thus, obesity may arise as part of a protective response to disease development.

Most studies examining the association between obesity and adverse clinical outcomes have relied on BMI as a marker of body fat. However, several organisations, such as the American Medical Association,13 recommend consideration of alternatives to BMI, including visceral fat, body adiposity index, body composition, relative fat mass and waist circumference, in part because of BMI’s origins in white supremacy and eugenics.14 Moreover, the BMI encompasses lean as well as fat mass.

While many systematic reviews have noted that mortality is lower for people with higher BMI,415,29 few, if any, studies evaluated whether body fat (however measured) may reduce the risk of incident NCDs after accounting for inflammation. In this prospective cohort study of UK adults, we evaluated the associations of various body fat measures, independent of inflammation, with the development of NCDs and all-cause mortality. Given the potential importance of survivorship bias for analyses of non-fatal outcomes, such as incident NCDs, we repeated all analyses using a competing risks framework.

Methods

We report this prospective cohort study according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.30

Patient and public involvement statement

Neither the patients nor the public were involved at any stage of the research process, nor will they be involved in the dissemination of the results.

Data sources and cohort

We used the UK Biobank database (www.ukbiobank.ac.uk)31, which incorporates data from over 500 000 adults between the ages of 40 and 69 years in England, Scotland and Wales. The UK Biobank includes data on demographics, social variables and behaviours, baseline medical conditions, imaging, prescriptions and supplements, biological samples, surgeries, genomics and clinical outcomes. We used the database to assemble a cohort of adults with at least one measure of body fat. Participants were accrued between 13 March 2006 and 1 October 2010 and were followed until death, loss to follow-up or study end (31 December 2022), whichever occurred first.

Body fat and markers of inflammation

We obtained measures of baseline body fat via dual-energy X-ray absorptiometry (DXA; total tissue fat percentage measured by DXA (fatDXA) and visceral adipose tissue percentage measured by DXA (VATDXA)), abdominal MRI (subcutaneous adipose tissue percentage measured by MRI (SATMRI), VAT percentage measured by MRI (VATMRI), total abdominal adipose tissue index measured by MRI (indexMRI; defined as (VATMRI volume+SATMRI volume)/height squared)), bioimpedance analysis (BIA; body fat percentage measured by BIA (fatBIA)) and physical measurement (BMI, waist-to-hip ratio (WHR)). BIA measurements were made with the Tanita BC418MA body composition analyser, which has been validated in at least two studies using DXA32 or hydrostatic weighing33 as reference standards. Both the DXA and the BIA are affected by hydration status. We did not investigate waist circumference and other physical measures, as a systematic review by Jayedi et al34 has shown that many of these physical measures, including the WHR, had similar associations with mortality.

Available markers of inflammation were CRP, neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR). Red cell distribution width (RDW), apolipoprotein B and glycated haemoglobin, all of which are also related to inflammation,35,37 were obtained at baseline. The blood sample collection and physical measurements (including BIA) occurred on the same day.

Longitudinal outcomes

In addition to all-cause mortality, we studied obesity-related NCDs identified by a systematic review by Guh et al2 from 2009. Their review found evidence linking obesity with 18 NCDs: five cardiovascular diseases (specifically heart failure, hypertension, myocardial infarction, pulmonary embolism and stroke), eight cancers (specifically breast, colorectal, endometrial, oesophageal, kidney, ovarian, pancreatic and prostate), type 2 diabetes (expanded in the current study to include type 1 diabetes), asthma, gallbladder disease, chronic back pain and osteoarthritis. Death and cancer incidence were defined using available registries. Asthma, myocardial infarction and stroke were defined using validated administrative algorithms developed by the UK Biobank Outcome Adjudication group.38 The remaining seven outcomes, although not validated, were defined similarly: the first incidence of a three-character International Classification of Diseases code obtained from primary care data, hospitalisations, the death registry or self-reported conditions during nurse-led interviews.

Covariates

We obtained the following baseline variables: demographics (age, sex and ethnicity), social variables (rural/urban residence, Townsend deprivation index,39 household income, employment status, education, social supports—friend and family visits, leisure/social activities and the ability to confide in others), behaviours (smoking status, alcohol intake, physical activity (ie, the International Physical Activity Question (IPAQ)),40 fruit and vegetable intake and sleep duration), comorbidities and other health-related measures (walking pace, forced expiratory volume in 1 s, hand grip strength (maximum of left and right hands) and some form of a government-accommodated disability). Comorbidities were Guh et al’s obesity-related conditions2 (cardiovascular disease, cancer, diabetes, asthma, osteoporosis, gallbladder disease and chronic back pain) and self-reported insomnia. The Townsend deprivation index incorporates unemployment, lack of car and home ownership and household overcrowding. Higher scores indicate more deprivation. Gender was not available.

Statistical analyses

We did analyses with Stata MP V.16.0 (www.stata.com) and report baseline descriptive statistics as counts and percentages or medians and interquartile ranges, as appropriate. We tested differences across BMI groups using χ2 and Kruskal-Wallis tests. We set the threshold p value for statistical significance at 0.05. We calculated all pairwise correlations between the various body fat measures and with CRP by sex.

We used Cox regression to determine the associations between baseline measures of body fat and time to clinical outcomes. In initial exploratory regressions with mortality, we determined that variables were best parametrised as follows: (1) quadratic and linear terms for body fat and most laboratory measures and (2) a natural logarithm of CRP (lnCRP). Since quantities of body fat are expected to differ between the sexes, we interacted sex with each body fat measure. We report age-sex-lnCRP-adjusted associations (primary analysis) and fully adjusted associations (using all covariates mentioned above except for IPAQ and walking pace, which may be influenced by a participant’s weight). We determined that the proportional hazards assumption was satisfied by examining plots of the log-negative-log of within-group survivorship probabilities versus log-time after adjustment for covariates. Missingness of each variable is reported in online supplemental table S1. No outcomes or measures of body fat were imputed, but missing values for covariates were imputed in the regressions: the median value for continuous variables and the most frequent value for categorical variables.

We report percentile-based HRs (the 95th percentile vs 5th percentile) for the measures of body fat with corresponding 95% CIs. The 5th percentile of BMI fell within the normal range of BMI for both females and males (ie, 20.5 and 22.0 kg/m2), and the 95th percentile of BMI fell within the moderate obesity range for females and males (ie, 37.0 and 35.5 kg/m2). Additionally, we plotted continuous HRs across the 1st percentile to the 99th percentile range (referent at the 5th percentile) in the natural units of four body fat measures (BMI, WHR, fatBIA and VATDXA) superimposed on histograms showing the population distribution of each measure of body fat.

In sensitivity analyses, (1) we explored the independent effects of various subsets of covariates (ie, age-sex, social determinants, behaviours, laboratory, comorbidities and other health-related measures) on the risk of outcomes, (2) we modelled mortality as a competing risk using a flexible modelling approach outlined by Lambert41 (Weibull hazards regression with three or four knots) and (3) we adjusted for WHR together with BMI or fatBIA to explore the independent associations of WHR from BMI and fatBIA.

Results

Of 502 178 participants, 500 107 had a measure of body fat and thus were included in this study. Median follow-up time was 13.8 years (range 4 days to 16.8 years). There were 44 145 (8.8%) deaths, 76 982 (21.6%) new diagnoses of CVD, 33 088 (6.8%) first diagnoses of cancer, 25 384 (5.4%) new diagnoses of diabetes, 13 182 (3.0%) new diagnoses of asthma, 21 070 (4.4%) new diagnoses of gallbladder disease, 35 500 (7.8%) new diagnoses of chronic back pain and 40 133 (9.0%) new diagnoses of osteoarthritis.

Online supplemental table S1 summarises demographics and clinical characteristics by BMI categories: <18.5, 18.5–24.9, 25.0–34.9 and ≥35.0 kg/m2. The participants in the highest BMI category were the most deprived according to the Townsend deprivation index, had the lowest income, did the most unpaid voluntary work, were most likely to be non-white and to have some kind of disability accommodation. They were most likely to have quit smoking and to drink only on special occasions, had the most friends or family visits, but felt the least likely to be able to confide. They had the slowest walking pace and the lowest IPAQ scores (although neither variable accounted for a participant’s weight). They had the most insomnia and the most of every obesity-related comorbidity. They had the highest CRP, RDW and glycated haemoglobin, and they had the highest measures of body fat on all measures except for VATMRI.

The participants in the lowest BMI category were the most likely to be unemployed, a homemaker or on disability income. They had the fastest walking pace and were most likely not to drink alcohol, but were the most likely to smoke. They had the fewest visits from friends and family, and they were the least likely to participate in social activities. They had the highest NLRs and PLRs.

There were a number of strong correlations between body fat measures (table 1).42 FatBIA and indexMRI were both strongly correlated with BMI (FatBIA : 0.85 in females and 0.80 in males; and indexMRI: 0.82 and 0.79, respectively). Further, fatBIA was strongly correlated with fatDXA and indexMRI (fatDXA: 0.78 in females and 0.74 in males and indexMRI: 0.78 and 0.75, respectively). FatDXA was very strongly correlated with the indexMRI (0.91 in females and 0.92 in males). Lastly, VATMRI was strongly correlated with VATDXA (0.84 in females and 0.78 in males). SATMRI and VATMRI were dropped from further analyses due to lower power (sample sizes<5% of the total).

Table 1. Correlations of body fat and CRP by sex.

BMI WHR FatBIA FatDXA VATDXA IndexMRI SATMRI VATMRI lnCRP
BMI – 0.59 0.80 0.66 0.59 0.79 0.35 0.32 0.33
WHR 0.46 – 0.62 0.55 0.60 0.60 0.17 0.46 0.37
FatBIA 0.85 0.46 – 0.74 0.62 0.75 0.31 0.34 0.38
FatDXA 0.70 0.35 0.78 – 0.80 0.92 0.35 0.46 0.37
VATDXA 0.58 0.57 0.59 0.71 – 0.85 0.03 0.78 0.35
IndexMRI 0.82 0.47 0.78 0.91 0.79 – 0.37 0.53 0.37
SATMRI 0.48 0.39 0.49 0.57 0.41 0.66 – −0.28 0.09
VATMRI 0.32 0.58 0.37 0.46 0.84 0.54 0.26 – 0.23
lnCRP 0.50 0.34 0.52 0.45 0.41 0.48 0.30 0.34 –
Open in a new tab

Data from females are shown in the lower triangle, and data from males are shown in the upper triangle. Correlations are bolded if ≥0.80.

BMI, Body Mass Index; CRP, C-reactive protein; FatBIA, body fat percentage measured by bioimpedance analysis; FatDXA, total tissue fat percentage measured by dual-energy X-ray absorptiometry; indexMRI, total abdominal adipose tissue index measured by MRI; lnCRP, natural logarithm of C-reactive protein; SATMRI, subcutaneous adipose tissue percentage measured by MRI; VATDXA, visceral adipose tissue percentage measured by dual-energy X-ray absorptiometry; VATMRI, visceral adipose tissue percentage measured by MRI; WHR, waist-to-hip ratio.

In females, the correlations between body fat measures and lnCRP tended to be around 0.41–0.50, except for WHR which was 0.34. In males, these correlations tended to be around 0.33–0.38, whereas for WHR, it was 0.37.

In the age-sex-lnCRP-adjusted and fully adjusted regressions, lnCRP was significantly associated with every outcome (online supplemental table S2).

Online supplemental table S3 shows all the age-sex-lnCRP-adjusted associations between body fat and outcomes. Figures1 2 are an abbreviated version of online supplemental table S3 (together with the results of the competing risk analyses). Despite adjustment for lnCRP, almost all measures of body fat were positively associated with every obesity-related NCD (where power allowed) in both sexes (online supplemental figures S1–S7). Diabetes (HRs from 4.43 to 16.32) and gallbladder disease (HRs from 2.75 to 7.16) were the most strongly associated with a measure of body fat.

Figure 1. Associations of body fat measures with clinical outcomes with and without the competing risk of death in females. Results from six models are presented above (3 measures of body fat×accounting or not accounting for death). Each marker is an HR or SHR comparing the 95th percentile versus the 5th percentile of a body fat measure. The coloured horizontal lines represent the 95% confidence limits. The models are adjusted for age, sex, lnCRP and one of three measures of body fat (linear and quadratic terms). Sex interacts with each measure of body fat. The faded markers do not adjust for the competing risk of death. The non-faded markers account for the competing risk of death. BMI, Body Mass Index; fatBIA, body fat percentage measured by bioimpedance analysis; lnCRP, natural logarithm of C-reactive protein; SHR, subdistribution HR and WHR, waist-to-hip ratio.

Figure 1

Open in a new tab

Figure 2. Associations of body fat measures with clinical outcomes with and without the competing risk of death in males. Results from six models are presented above (3 measures of body fat×accounting or not accounting for death). Each marker is an HR or SHR comparing the 95th percentile versus the 5th percentile of a body fat measure. The coloured horizontal lines represent the 95% confidence limits. The models are adjusted for age, sex, lnCRP and one of three measures of body fat (linear and quadratic terms). Sex interacts with each measure of body fat. The faded markers do not adjust for the competing risk of death. The non-faded markers account for the competing risk of death. BMI, Body Mass Index; fatBIA, body fat percentage measured by bioimpedance analysis; lnCRP, natural logarithm of C-reactive protein; SHR, subdistribution HR and WHR, waist-to-hip ratio.

Figure 2

Open in a new tab

The histograms in online supplemental figures S1–S7 and figure 3 show that females tended to have smaller WHRs and smaller quantities of VATDXA but larger quantities of fatBIA than males. The distribution of BMI was quite similar between the sexes, although the distribution for females was wider. In contrast to the reporting of 95th percentile versus 5th percentile HRs (where comparisons using percentiles offset any between-sex differences in absolute body fat), these figures using the natural units show that females were at higher risk for NCDs than males at lower WHRs and VATDXA but were at lower risk with higher fatBIA. The risk was similar between females and males for comparable levels of BMI.

Figure 3. Associations of body fat measures with all-cause mortality. Results from four models are presented above (4 measures of body fat). Plots show HR plotted against BMI (top-left), WHR (top-right), fatBIA (bottom-left) and VATDXA (bottom-right). Curves in pink represent females and curves in blue represent males. Solid lines represent HR adjusted for age, sex and lnCRP, and dashed lines represent fully adjusted HR. The range of body fat represents the 3rd percentile to the 97th percentile within sex. Histograms underlying the HR plots show the distribution of the various measures of body fat by sex (pink for females, blue for males and purple for overlapping distributions). BMI, Body Mass Index; fatBIA, body fat percentage measured by bioimpedance analysis; lnCRP, natural logarithm of C-reactive protein; VATDXA, visceral adipose tissue percentage measured by dual-energy X-ray absorptiometry and WHR, waist-to-hip ratio.

Figure 3

Open in a new tab

The age-sex-lnCRP-adjusted findings for mortality were not consistent across measures of body fat, nor between sexes (figure 3, online supplemental table S3). The DXA and MRI measures of body fat were not significantly associated with mortality in either sex, perhaps because of inadequate statistical power due to their much lower sample size (~10% of the total N). These measures were dropped from further analyses. In females, WHR was positively associated (HR 1.69 (95% CI 1.61 to 1.77)), BMI was not significantly associated (HR 0.98 (95% CI 0.93 to 1.03)) and fatBIA was negatively associated with mortality (HR 0.93 (95% CI 0.89 to 0.98)). In males, all three—WHR, BMI and fatBIA—were positively associated with mortality (HR 1.91 (95% CI 1.83 to 1.99); HR 1.11 (95% CI 1.07 to 1.16); HR 1.17 (95% CI 1.12 to 1.22)).

After full adjustment (figures4 5, online supplemental table S4), all measures of body fat remained positively associated with all obesity-related NCDs. In contrast, BMI and fatBIA were negatively associated with mortality in both sexes (HR 0.80 (95% CI 0.76 to 0.84) and HR 0.80 (95% CI 0.76 to 0.84) in females; HR 0.89 (95% CI 0.85 to 0.93) and HR 0.89 (95% CI 0.85 to 0.92) in males). However, WHR remained positively associated with mortality in both sexes (HR 1.21 (95% CI 1.16 to 1.28) in females; HR 1.19 (95% CI 1.14 to 1.24) in males). In sensitivity analyses with adjustment for various subsets of covariates (figures4 5, online supplemental table S4), the laboratory measures and comorbidities/health-related measures appeared to have the largest attenuating effects on the associations with mortality.

Figure 4. Associations of body fat measures with clinical outcomes adjusted for various sets of covariates in females. Results from 18 models are presented above (3 measures of body fat × 6 sets of covariates). Each marker is an HR comparing the 95th percentile versus the 5th percentile of a body fat measure. The coloured horizontal lines represent the 95% confidence limits. Each model is adjusted for one of the three measures of body fat: BMI, WHR or fatBIA. The model represented by the black asterisks adjusted for all covariates is listed in table 1 except for IPAQ and walking pace, which may be determined, at least in part, by a participant’s body fat. The model represented by the blue squares was adjusted for age and sex, but not lnCRP. The model represented by the yellow circles was adjusted for age, sex and social variables (Townsend deprivation index, rural residence, household income, employment, postsecondary education, non-white ethnicity, friends and family visits, leisure/social activities and ability to confide). The model represented by the orange triangles was adjusted for age, sex and behaviours (alcohol consumption, fresh fruit consumption, sleep duration, smoking status and raw vegetable consumption). The model with the maroon diamonds adjusted for age, sex and laboratory measures (lnCRP, apolipoprotein B, glycated haemoglobin, NLR, PLR and RDW—all quadratic terms except CRP). The model represented by the upside-down green triangle adjusted for age, sex and baseline comorbidities (cardiovascular disease, cancer, diabetes, asthma, osteoarthritis, gallbladder disease and chronic back pain) and other related health measures (forced expiratory volume in 1 s, hand grip strength, government-accommodated disability and insomnia). BMI, Body Mass Index; CRP, C-reactive protein; fatBIA, body fat percentage measured by bioimpedance analysis; IPAQ, International Physical Activity Questionnaire; lnCRP, natural logarithm of C-reactive protein; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; RDW, red cell distribution width and WHR, waist-to-hip ratio.

Figure 4

Open in a new tab

Figure 5. Associations of body fat measures with clinical outcomes adjusted for various sets of covariates in males. Results from 18 models are presented above (3 measures of body fat × 6 sets of covariates). Each marker is an HR comparing the 95th percentile versus the 5th percentile of a body fat measure. The coloured horizontal lines represent the 95% confidence limits. Each model is adjusted for one of the three measures of body fat: BMI, WHR or fatBIA. The model represented by the black asterisks adjusted for all covariates is listed in table 1 except for IPAQ and walking pace, which may be determined, at least in part, by a participant’s body fat. The model represented by the blue squares was adjusted for age and sex, but not lnCRP. The model represented by the yellow circles was adjusted for age, sex and social variables (Townsend deprivation index, rural residence, household income, employment, postsecondary education, non-white ethnicity, friends and family visits, leisure/social activities and ability to confide). The model represented by the orange triangles adjusted for age, sex and behaviours (alcohol consumption, fresh fruit consumption, sleep duration, smoking status and raw vegetable consumption). The model with the maroon diamonds was adjusted for age, sex and laboratory measures (lnCRP, apolipoprotein B, glycated haemoglobin, NLR, PLR and RDW—all quadratic terms except CRP). The model represented by the upside-down green triangle adjusted for age, sex and baseline comorbidities (cardiovascular disease, cancer, diabetes, asthma, osteoarthritis, gallbladder disease and chronic back pain) and other related health measures (forced expiratory volume in 1 s, hand grip strength, government-accommodated disability and insomnia). BMI, Body Mass Index; CRP, C-reactive protein; fatBIA, body fat percentage measured by bioimpedance analysis; IPAQ, International Physical Activity Questionnaire; lnCRP, natural logarithm of C-reactive protein; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; RDW, red cell distribution width and WHR, waist-to-hip ratio.

Figure 5

Open in a new tab

When accounting for the competing risk of death (figures1 2, online supplemental table S5), most of the associations between body fat and the risk of incident NCDs were attenuated slightly in females but were somewhat amplified in males, especially for diabetes. In contrast, accounting for the competing risk of death largely abrogated the associations between measures of body fat and incident cancer for both sexes.

After adjustment for WHR and BMI or fatBIA, the associations between body fat and mortality remained different from those between body fat and the risk of incident NCDs (figures6 7, online supplemental table S6). WHR maintained positive associations with mortality, but in these analyses BMI (females: HR 0.75 (95% CI 0.71 to 0.79); males: HR 0.67 (95% CI 0.64 to 0.71)) and fatBIA (females: HR 0.73 (95% CI 0.69 to 0.77); males: HR 0.78 (95% CI 0.74 to 0.82)) were negatively associated with mortality in both females and males.

Figure 6. Associations of body fat measures with clinical outcomes adjusted for two measures of body fat in females. Results from two models are presented above. Each marker is an HR comparing the 95th percentile versus the 5th percentile of a body fat measure. The coloured horizontal lines represent the 95% confidence limits. The models are adjusted for age, sex, lnCRP and two measures of body fat (linear and quadratic terms). The first model was adjusted for BMI and WHR. The second model was adjusted for fatBIA and WHR. Sex interacts with measures of body fat. BMI, Body Mass Index; fatBIA, body fat percentage measured by bioimpedance analysis; lnCRP, natural logarithm of C-reactive protein and WHR, waist-to-hip ratio.

Figure 6

Open in a new tab

Figure 7. Associations of body fat measures with clinical outcomes adjusted for two measures of body fat in males. Results from two models are presented above. Each marker is an HR comparing the 95th percentile versus the 5th percentile of a body fat measure. The coloured horizontal lines represent the 95% confidence limits. The models are adjusted for age, sex, lnCRP and two measures of body fat (linear and quadratic terms). The first model adjusts for BMI and WHR. The second model adjusts for fatBIA and WHR. Sex interacts with the measures of body fat. BMI, Body Mass Index; fatBIA, body fat percentage measured by bioimpedance analysis; lnCRP, natural logarithm of C-reactive protein and WHR, waist-to-hip ratio.

Figure 7

Open in a new tab

Discussion

We found that BMI was strongly correlated with body fat as assessed by BIA, but WHR and VATDXA were less so. All of these measures were positively associated with the incidence of NCDs despite the adjustment for CRP and other available characteristics. In contrast, only WHR remained positively and significantly associated with all-cause mortality after adjustment for potential confounders. These findings support the hypothesis that body fat may be protective against death and that the excess risk associated with higher WHR may be mediated by something other than body fat.

With or without adjustment for CRP, age, sex, social determinants, behaviours, other inflammation-related laboratory measures, baseline conditions and further health-related measures, higher body fat was associated with a significantly increased risk of incident NCDs. The findings from competing risk analyses suggest that the latter is not due to survival bias, except perhaps for cancer, where the associations between higher body fat and higher risk were no longer apparent. Similarly, our findings do not support a clinically meaningful effect of collider bias (where higher-weight patients may be screened and diagnosed more frequently than lower-weight patients), especially for males, where consideration of competing risks tended to amplify rather than attenuate the subdistribution HRs for incident NCDs.

In contrast, the fully adjusted associations between measures of body fat and all-cause mortality were negatively associated when BMI or BIA was used to assess body fat but were positively associated for WHR. Unlike WHR, the BIA-measured body fat as a predictor acted similarly to BMI throughout the analyses. Moreover, in age-sex-lnCRP-adjusted models, WHR remained associated with higher mortality after adjustment for BMI or fatBIA, whereas BMI and fatBIA were negatively associated with mortality after adjustment for WHR. There was no evidence that visceral fat, as measured by DXA, had any particular advantages for predicting risk of mortality, although these analyses were underpowered (figure 3).

Several studies43 44 have noted that WHR is more strongly associated with hyperinsulinaemia and IR than BMI. Our previous work with the National Health and Nutrition Examination Surveys (NHANES US study cohort5 found that adjustment for fasting insulin was sufficient to uncover an association between higher BMI and lower mortality in a general population of US adults, particularly in males. Furthermore, four large studies (totalling ~10 000 participants) in adults without diabetes8 9 11 12 show that rises in markers of inflammation precede increases in insulin level. Randomised trials45 46 and a meta-analysis7 (with 60 studies) demonstrate that changes in insulin levels precede changes in BMI. This sequence suggests that greater adiposity may buffer the hyperglycaemic stress response during acute illness (ie, increased endogenous glucose production),47 thereby reducing mortality and potentially other consequences of NCDs. Additionally, larger fat cells can sequester more excess glucose. Adipose tissue is metabolically active and further contributes to immune function48 through apoptotic cell clearance, extracellular matrix remodelling, angiogenesis and serves as an immunological effector site in pathogen defence.49 Moreover, adiposity provides greater metabolic reserves,50 not only of fat but also from increased muscle and bone mass, which may slow the progression of cachexia. While obesity is associated with increased risk of falls, non-sarcopenic obesity may mitigate the risk of subsequent fractures.51

Our study also has important limitations that should be considered when interpreting results. First, as with any observational study, residual confounding remains possible. We did not have data on weight cycling,52,56 weight stigma,57,59 weight-based discrimination (eg, less access to therapeutic surgeries and diagnostic imaging) and hyperinsulinaemia/IR—all of which are more common among higher-weight patients. Second, the study population was not representative of the general UK population; the participants were more likely to be white, female, wealthier and to have fewer health conditions.60 Therefore, our findings may be less generalisable to non-white people, males, those who are less wealthy and those with more health conditions. Third, while no study has established that body fat causes poor outcomes, some of the covariates in the fully adjusted analyses might mediate the associations between body fat and outcomes. Although we cannot rule out this possibility, mediation would be expected to bias towards the null (not past the null) and is unlikely to explain the finding that the association between mortality and higher BMI and fatBIA inverts (shifts from positive to negative) following statistical adjustment, regardless of whether models include age, sex, lnCRP and WHR, or the full set of covariates. Fourth, 7 of the 11 NCDs were defined using non-validated administrative algorithms and, therefore, would be subject to misclassification. Lastly, we used single imputation for missing values (mode for categorical variables and median for continuous variables), which would also be expected to bias towards the null in both the fully and age-sex-lnCRP-adjusted analyses.

In conclusion, BMI was strongly correlated with body fat as assessed by BIA, whereas WHR and VATDXA were less so. All measures of body fat were associated with the incidence of NCDs, but only WHR was independently associated with mortality.

Supplementary material

online supplemental file 1
bmjopen-15-10-s001.pdf (1.8MB, pdf)
DOI: 10.1136/bmjopen-2024-092962

Acknowledgements

Anita Lloyd, MSc, PStat, University of Alberta, performed the technical review; Sophanny Tiv, BSc, University of Alberta, provided graphics support. This research has been conducted using the UK Biobank Resource under Application Number 144943.

This research has been conducted using the UK Biobank Resource under Application Number 144943. This work uses data provided by patients and collected by the NHS as part of their care and support. The study data were provided by the UK Biobank, NHS England and the National Safe Haven. The interpretation and conclusions contained herein are those of the researchers and do not represent the views of the UK Biobank, NHS England or the National Safe Haven.

Footnotes

Funding: This study was supported by MT via the University of Calgary (David Freeze Chair in Health Services Research). The sponsors had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript; nor in the decision to submit the manuscript for publication.

Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2024-092962).

Provenance and peer review: Not commissioned; externally peer reviewed.

Patient consent for publication: Not applicable.

Ethics approval: This study involves human participants and was approved by the University of Alberta Research Ethics Board (Pro00134819) and the University of Calgary Research Ethics Board (pSite-23–0044). All participants provided written informed consent (National Research Ethics Service, reference 11/NW/0382) before taking part.

Data availability free text: These data are publicly available from https://www.ukbiobank.ac.uk/.

Patient and public involvement: Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.

Data availability statement

Data are available in a public, open access repository.

References

  • 1.Nyberg ST, Batty GD, Pentti J, et al. Obesity and loss of disease-free years owing to major non-communicable diseases: a multicohort study. Lancet Public Health. 2018;3:e490–7. doi: 10.1016/S2468-2667(18)30139-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Guh DP, Zhang W, Bansback N, et al. The incidence of co-morbidities related to obesity and overweight: a systematic review and meta-analysis. BMC Public Health. 2009;9:88. doi: 10.1186/1471-2458-9-88. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Wiebe N, Stenvinkel P, Tonelli M. Associations of Chronic Inflammation, Insulin Resistance, and Severe Obesity With Mortality, Myocardial Infarction, Cancer, and Chronic Pulmonary Disease. JAMA Netw Open . 2019;2:e1910456. doi: 10.1001/jamanetworkopen.2019.10456. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Wiebe N, Lloyd A, Crumley ET, et al. Associations between body mass index and all-cause mortality: A systematic review and meta-analysis. Obes Rev. 2023;24:e13588. doi: 10.1111/obr.13588. [DOI] [PubMed] [Google Scholar]
  • 5.Wiebe N, Muntner P, Tonelli M. Associations of body mass index, fasting insulin, and inflammation with mortality: a prospective cohort study. Int J Obes (Lond) 2022;46:2107–13. doi: 10.1038/s41366-022-01211-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Flegal KM, Troiano RP. Changes in the distribution of body mass index of adults and children in the US population. Int J Obes Relat Metab Disord. 2000;24:807–18. doi: 10.1038/sj.ijo.0801232. [DOI] [PubMed] [Google Scholar]
  • 7.Wiebe N, Ye F, Crumley ET, et al. Temporal Associations Among Body Mass Index, Fasting Insulin, and Systemic Inflammation: A Systematic Review and Meta-analysis. JAMA Netw Open . 2021;4:e211263. doi: 10.1001/jamanetworkopen.2021.1263. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.de las Heras Gala T, Herder C, Rutters F, et al. Association of changes in inflammation with variation in glycaemia, insulin resistance and secretion based on the kora study. Diabetes Metabolism Res . 2018;34 doi: 10.1002/dmrr.3063. [DOI] [PubMed] [Google Scholar]
  • 9.Herder C, Færch K, Carstensen-Kirberg M, et al. Biomarkers of subclinical inflammation and increases in glycaemia, insulin resistance and beta-cell function in non-diabetic individuals: the Whitehall II study. Eur J Endocrinol. 2016;175:367–77. doi: 10.1530/EJE-16-0528. [DOI] [PubMed] [Google Scholar]
  • 10.Li Q, Wang Q, Xu W, et al. C-Reactive Protein Causes Adult-Onset Obesity Through Chronic Inflammatory Mechanism. Front Cell Dev Biol. 2020;8:18. doi: 10.3389/fcell.2020.00018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Park K, Steffes M, Lee D-H, et al. Association of inflammation with worsening HOMA-insulin resistance. Diabetologia. 2009;52:2337–44. doi: 10.1007/s00125-009-1486-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Yan Y, Li S, Liu Y, et al. Temporal relationship between inflammation and insulin resistance and their joint effect on hyperglycemia: the Bogalusa Heart Study. Cardiovasc Diabetol. 2019;18 doi: 10.1186/s12933-019-0913-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.American Medication Assocation Support removal of bmi as a standard measure in medicine and recognizing culturally-diverse and varied presentations of eating disorders and indications for metabolic and bariatric surgery: report 09 of the council on science and publish health (a-23) [1-Sep-2023]. https://www.ama-assn.org/system/files/a23-csaph07.pdf Available. Accessed.
  • 14.Erdman Farrell A. Fat Shame: Stigma and the Fat Body in American Culture. New York: New York University Press; 2011. [Google Scholar]
  • 15.Xie H, Wei L, Zhang H, et al. Association of systemic inflammation with the obesity paradox in cancer: results from multi-cohort studies. Inflamm Res. 2024;73:243–52. doi: 10.1007/s00011-023-01832-x. [DOI] [PubMed] [Google Scholar]
  • 16.Bundhun PK, Li N, Chen MH. Does an Obesity Paradox Really Exist After Cardiovascular Intervention?: A Systematic Review and Meta-Analysis of Randomized Controlled Trials and Observational Studies. Medicine (Baltimore) 2015;94:e1910. doi: 10.1097/MD.0000000000001910. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Grymonprez M, Capiau A, De Backer TL, et al. The impact of underweight and obesity on outcomes in anticoagulated patients with atrial fibrillation: A systematic review and meta-analysis on the obesity paradox. Clin Cardiol. 2021;44:599–608. doi: 10.1002/clc.23593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Kwon Y, Kim HJ, Park S, et al. Body Mass Index-Related Mortality in Patients with Type 2 Diabetes and Heterogeneity in Obesity Paradox Studies: A Dose-Response Meta-Analysis. PLoS One. 2017;12:e0168247. doi: 10.1371/journal.pone.0168247. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Li S, Wang Z, Huang J, et al. Systematic review of prognostic roles of body mass index for patients undergoing lung cancer surgery: does the ‘obesity paradox’ really exist? Eur J Cardiothorac Surg. 2017;51:ezw386. doi: 10.1093/ejcts/ezw386. [DOI] [PubMed] [Google Scholar]
  • 20.Li Y, Li C, Wu G, et al. The obesity paradox in patients with colorectal cancer: a systematic review and meta-analysis. Nutr Rev. 2022;80:1755–68. doi: 10.1093/nutrit/nuac005. [DOI] [PubMed] [Google Scholar]
  • 21.Lin G-M, Li Y-H, Yin W-H, et al. The Obesity-Mortality Paradox in Patients With Heart Failure in Taiwan and a Collaborative Meta-Analysis for East Asian Patients. Am J Cardiol. 2016;118:1011–8. doi: 10.1016/j.amjcard.2016.06.056. [DOI] [PubMed] [Google Scholar]
  • 22.Lv W, Li S, Liao Y, et al. The “obesity paradox” does exist in patients undergoing transcatheter aortic valve implantation for aortic stenosis: a systematic review and meta-analysis. Interact Cardiovasc Thorac Surg. 2017;25:633–42. doi: 10.1093/icvts/ivx191. [DOI] [PubMed] [Google Scholar]
  • 23.Marcks N, Aimo A, Januzzi JL, Jr, et al. Re-appraisal of the obesity paradox in heart failure: a meta-analysis of individual data. Clin Res Cardiol. 2021;110:1280–91. doi: 10.1007/s00392-021-01822-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Mei X, Hu S, Mi L, et al. Body mass index and all-cause mortality in patients with percutaneous coronary intervention: A dose-response meta-analysis of obesity paradox. Obes Rev. 2021;22:e13107. doi: 10.1111/obr.13107. [DOI] [PubMed] [Google Scholar]
  • 25.Nie W, Zhang Y, Jee SH, et al. Obesity survival paradox in pneumonia: a meta-analysis. BMC Med. 2014;12:61. doi: 10.1186/1741-7015-12-61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Niedziela J, Hudzik B, Niedziela N, et al. The obesity paradox in acute coronary syndrome: a meta-analysis. Eur J Epidemiol. 2014;29:801–12. doi: 10.1007/s10654-014-9961-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Padwal R, McAlister FA, McMurray JJV, et al. The obesity paradox in heart failure patients with preserved versus reduced ejection fraction: a meta-analysis of individual patient data. Int J Obes. 2014;38:1110–4. doi: 10.1038/ijo.2013.203. [DOI] [PubMed] [Google Scholar]
  • 28.Tan X-F, Shi J-X, Chen and M-H. Prolonged and intensive medication use are associated with the obesity paradox after percutaneous coronary intervention: a systematic review and meta-analysis of 12 studies. BMC Cardiovasc Disord. 2016;16:125. doi: 10.1186/s12872-016-0310-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Zhi G, Xin W, Ying W, et al. Obesity Paradox" in Acute Respiratory Distress Syndrome. Asystematic Review and Meta-Analysis PLoS One. 2016;11:e0163677. doi: 10.1371/journal.pone.0163677. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453–7. doi: 10.1016/S0140-6736(07)61602-X. [DOI] [PubMed] [Google Scholar]
  • 31.UK Biobank UK biobank: health research data for the world. [12-Sep-2025]. https://www.ukbiobank.ac.uk Available. Accessed.
  • 32.Pietrobelli A, Rubiano F, St-Onge M-P, et al. New bioimpedance analysis system: improved phenotyping with whole-body analysis. Eur J Clin Nutr. 2004;58:1479–84. doi: 10.1038/sj.ejcn.1601993. [DOI] [PubMed] [Google Scholar]
  • 33.Kelly J, Metcalf J. Validity and Reliability of Body Composition Analysis Using the Tanita BC418-MA. J Exerc Physiol. 2012;15:75–83. [Google Scholar]
  • 34.Jayedi A, Soltani S, Zargar MS, et al. Central fatness and risk of all cause mortality: systematic review and dose-response meta-analysis of 72 prospective cohort studies. BMJ. 2020;370:m3324. doi: 10.1136/bmj.m3324. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Lippi G, Targher G, Montagnana M, et al. Relation between red blood cell distribution width and inflammatory biomarkers in a large cohort of unselected outpatients. Arch Pathol Lab Med. 2009;133:628–32. doi: 10.5858/133.4.628. [DOI] [PubMed] [Google Scholar]
  • 36.Liu S, Hempe JM, McCarter RJ, et al. Association between Inflammation and Biological Variation in Hemoglobin A1c in U.S. Nondiabetic Adults. The Journal of Clinical Endocrinology & Metabolism . 2015;100:2364–71. doi: 10.1210/jc.2014-4454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Faraj M, Messier L, Bastard JP, et al. Apolipoprotein B: a predictor of inflammatory status in postmenopausal overweight and obese women. Diabetologia. 2006;49:1637–46. doi: 10.1007/s00125-006-0259-7. [DOI] [PubMed] [Google Scholar]
  • 38.UK Biobank Outcome Adjudication Group UK biobank: algorithmically-defined outcomes. 2022. [20-Aug-2025]. https://biobank.ndph.ox.ac.uk/ukb/refer.cgi?id=460 Available. Accessed.
  • 39.Townsend P. Deprivation. J Soc Pol. 1987;16:125–46. doi: 10.1017/S0047279400020341. [DOI] [Google Scholar]
  • 40.Guidelines for data processing and analysis of the International Physical Activity Questionniare (IPAQ) - short and long forms. 2005. https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/ipaq_analysis.pdf Available.
  • 41.Lambert PC. The estimation and modelling of cause-specific cumulative incidence functions using time-dependent weights. Stata J. 2017;17:181–207. [PMC free article] [PubMed] [Google Scholar]
  • 42.Correlation and regression 1997. [20-Aug-2025]. https://www.bmj.com/about-bmj/resources-readers/publications/statistics-square-one/11-correlation-and-regression Available. Accessed.
  • 43.Benites-Zapata VA, Toro-Huamanchumo CJ, Urrunaga-Pastor D, et al. High waist-to-hip ratio levels are associated with insulin resistance markers in normal-weight women. Diabetes Metab Syndr. 2019;13:636–42. doi: 10.1016/j.dsx.2018.11.043. [DOI] [PubMed] [Google Scholar]
  • 44.Liu M-M, Liu Q-J, Wen J, et al. Waist-to-hip ratio is the most relevant obesity index at each phase of insulin secretion among obese patients. J Diabetes Complications. 2018;32:670–6. doi: 10.1016/j.jdiacomp.2018.04.006. [DOI] [PubMed] [Google Scholar]
  • 45.Holman RR, Thorne KI, Farmer AJ, et al. Addition of biphasic, prandial, or basal insulin to oral therapy in type 2 diabetes. N Engl J Med. 2007;357:1716–30. doi: 10.1056/NEJMoa075392. [DOI] [PubMed] [Google Scholar]
  • 46.Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33) The Lancet. 1998;352:837–53. doi: 10.1016/S0140-6736(98)07019-6. [DOI] [PubMed] [Google Scholar]
  • 47.Bradley P. Hypothesis: Enhanced glucose availability and insulin resistance enhances an activated immune system and accounts for the obesity paradox. Clin Obes. 2022;12:e12521. doi: 10.1111/cob.12521. [DOI] [PubMed] [Google Scholar]
  • 48.Schipper HS, Prakken B, Kalkhoven E, et al. Adipose tissue-resident immune cells: key players in immunometabolism. Trends Endocrinol Metab. 2012;23:407–15. doi: 10.1016/j.tem.2012.05.011. [DOI] [PubMed] [Google Scholar]
  • 49.Barthelemy J, Bogard G, Wolowczuk I. Beyond energy balance regulation: The underestimated role of adipose tissues in host defense against pathogens. Front Immunol. 2023;14:1083191. doi: 10.3389/fimmu.2023.1083191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Kishore BK. Reverse epidemiology of obesity paradox: Fact or fiction? Physiol Rep. 2024;12:e70107. doi: 10.14814/phy2.70107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Gandham A, Mesinovic J, Jansons P, et al. Falls, fractures, and areal bone mineral density in older adults with sarcopenic obesity: A systematic review and meta-analysis. Obes Rev. 2021;22:e13187. doi: 10.1111/obr.13187. [DOI] [PubMed] [Google Scholar]
  • 52.O’Hara L, Taylor J. What’s Wrong with the “War on Obesity”? A Narrative Review of the Weight-Centered Health Paradigm and Development of the 3C Framework to Build Critical Competency for a Paradigm Shift. SAGE Open; 2018. [Google Scholar]
  • 53.Almuwaqqat Z, Hui Q, Liu C, et al. Long-Term Body Mass Index Variability and Adverse Cardiovascular Outcomes. JAMA Netw Open. 2024;7:e243062. doi: 10.1001/jamanetworkopen.2024.3062. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Swartz AZ, Wood K, Farber-Eger E, et al. Weight Trajectory Impacts Risk for 10 Distinct Cardiometabolic Diseases. J Clin Endocr Metab. 2025 doi: 10.1210/clinem/dgaf348. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Bangalore S, Fayyad R, Laskey R, et al. Body-Weight Fluctuations and Outcomes in Coronary Disease. N Engl J Med. 2017;376:1332–40. doi: 10.1056/NEJMoa1606148. [DOI] [PubMed] [Google Scholar]
  • 56.Hamm P, Shekelle RB, Stamler J. Large fluctuations in body weight during young adulthood and twenty-five-year risk of coronary death in men. Am J Epidemiol. 1989;129:312–8. doi: 10.1093/oxfordjournals.aje.a115135. [DOI] [PubMed] [Google Scholar]
  • 57.Fat" Is Not Necessarily a Bad Word. Hastings Cent Rep. 2020;50:11–6. doi: 10.1002/hast.1180. [DOI] [PubMed] [Google Scholar]
  • 58.Thille P, Friedman M, Setchell J. Weight-related stigma and health policy. CMAJ. 2017;189:E223–4. doi: 10.1503/cmaj.160975. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Hunger JM, Smith JP, Tomiyama AJ. An Evidence‐Based Rationale for Adopting Weight‐Inclusive Health Policy. Social Issues Policy Review. 2020;14:73–107. doi: 10.1111/sipr.12062. [DOI] [Google Scholar]
  • 60.Fry A, Littlejohns TJ, Sudlow C, et al. Comparison of Sociodemographic and Health-Related Characteristics of UK Biobank Participants With Those of the General Population. Am J Epidemiol. 2017;186:1026–34. doi: 10.1093/aje/kwx246. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    online supplemental file 1
    bmjopen-15-10-s001.pdf (1.8MB, pdf)
    DOI: 10.1136/bmjopen-2024-092962

    Data Availability Statement

    Data are available in a public, open access repository.

    Articles from BMJ Open are provided here courtesy of BMJ Publishing Group

    RESOURCES