Exploring the Underlying Mechanisms Linking Adiposity and Cardiovascular Disease: A Prospective Cohort Study of 404,332 UK Biobank Participants

: Obesity is causally associated with multiple cardiovascular outcomes but effective population measure to control obesity is limited. This study aims to decipher to which extent excess atherosclerotic cardio-vascular diseases (ASCVD) and heart failure (HF) risk due to obesity can be explained by conventional risk factors. This is a prospective cohort study of 404,332 White UK Biobank participants. Participants with


Introduction
& O besity is associated with elevated mortality risk and account for over one-third of morbidity worldwide, most notably cardiovascular disease (CVD). 1,2By 2025, it is estimated that obesity will affect more than 18% of men and more than 21% of women across the globe.There is emerging evidence that obesity may be a modifiable risk factor for multiple types of CVD, such as heart failure (HF), and atrial fibrillation. 3,4besity is thought to cause CVD through multiple mechanisms and pathways. 5,6Most widely established, obesity is known to increase triglyceride-rich lipoproteins, most notably very-low-density lipoprotein. 7econdly, obesity has also been found to induce chronic, low-grade inflammation, 8,9 which could interact with the circulating lipids to promote atherosclerosis. 9More recently, it has been suggested that hemodynamic and renal function may play a role in obesity predisposing to CVD. 10 Nonetheless, most of the available evidence on how obesity could lead to CVD is based on animal experiments 11,12 or from piecewise clinical studies. 13,14A meta-analysis attempted to systematically disentangle the mechanisms linking obesity to atherosclerotic CVD (ASCVD), including coronary artery disease and stroke, 15 and suggested that blood pressure (BP), cholesterol and blood glucose were all mediators.However, that study did not examine emerging factors, such as systemic inflammation and renal function, nor conducted formal mediation analysis which could have obviated biases. 16Importantly, there is also scarce evidence of the mechanisms underpinning the association between obesity and HF, an increasingly prevalent CVD. 17 Therefore, this study aims to systematically explore the mechanisms linking obesity with ASCVD and HF under a counterfactual framework.Specifically, this study aims to decipher to which extent excess ASCVD and HF risk due to obesity can be explained by traditional risk factors such as lipids, BP, and glycated hemoglobin (HbA1c), as well as emerging markers of liver, 18 kidney function, 19 and systemic inflammation. 20

Study Design
The UK Biobank is a prospective cohort that recruited over 500,000 participants from the general population who were aged 37-73 years and were registered with NHS general practitioners between 2006 and 2010. 21Participants attended 1 of 22 assessments centers across England, Scotland, and Wales, where they completed a self-administered, touchscreen questionnaire and face-to-face interview to collect information on their lifestyle, health and socioeconomic characteristics, and trained research staff measured their height, weight, and BP and obtained blood samples.The UK Biobank obtained ethical approvals from the Northwest Multicenter Research Ethics Committee, the Community Health Index Advisory Group, the Patient Information Advisory Group, and the National Health Service National Research Ethics Service.All participants provided informed consent to participate and be followed up through data linkage.
This study included 404,332 white UK Biobank participants and who were free from CVDs, and any other chronic diseases (including cancer and chronic obstructive pulmonary disease) at baseline and had a body mass index (BMI) 18.5 kg/m 2 (Fig 1).Other ethnic groups were excluded because they may have different BMI cut-offs for obesity. 22articipants with chronic health problems at baseline or conditions and who were underweight were excluded to minimize reverse causation.

Exposures
Obesity was defined as general obesity and central obesity.BMI was calculated as weight/height 2 .Standing height (cm) was measured by a Seca 202 device (SECA).Weight and bioimpedance were measured by the Tanita BC-418MA body composition analyzer (Tanita Corporation of America).Waist and hip circumference were measured in centimeters.Trained nurses did all measurements following a standardized protocol.Wessex nonstretchable sprung tape measurements were used to determine waist and hip circumferences. 23By dividing the waist circumference by the hip circumference, the waist-to-hip ratio was obtained. 23General obesity was defined as BMI 30 kg/m 2 .Central obesity was classified as waist-to-hip ratio (WHR) >0.95 and >0.85 for males and females, respectively. 24

Outcomes
The outcomes measured were incident ASCVD, and incident HF ascertained from hospital admissions and deaths.Date and cause of hospital admissions were obtained through record linkage to Health Episode Statistics (England and Wales) and Scottish Morbidity Records (Scotland).Date and cause of death were obtained from death certificates held by the National Health Service Information Centre (England and Wales) and the National Health Service Central Register (Scotland).Dates and causes of hospital admissions were obtained through record linkage to Health Episode Statistics (England and Wales) and Scottish Morbidity Records (Scotland).Comprehensive details regarding the linkage procedures are at http://content.digital.nhs.uk/services.At the time of analysis, death records were available up to the end of September 2021 for England and Wales and the end of October 2021 for Scotland.Hospital admission data were available up to September 30, 2021 for England, July 31, 2021 for Scotland and February 28, 2018 for Wales.ASCVD was defined, in accordance with the American College of Cardiology definition, 25 as fatal CAD (ICD-10 codes: I10-25), fatal/nonfatal myocardial infarction (MI) (I21), and fatal/nonfatal stroke (I60-64).HF was defined as either a hospitalization or death record with ICD-10 codes: I11.0, I42.0, I42.6-42.7,I42.9, or I50.

Mediators
This study included 14 candidate mediators in 6 categories based on clinical trials and Mendelian randomization studies.The 6 categories included (1) lipids: low density lipoprotein cholesterol (LDL-c), 26 triglycerides (TG), 27 apolipoprotein B (ApoB), 28 and lipoprotein(a) 29 ; (2) BP: systolic BP (SBP) and diastolic BP (DBP) 30 ; (3) metabolic marker: HbA1c 31 ; (4) liver function marker: alanine aminotransferase (ALT), gamma-glutamyl transferase (GGT) 32 ; (5) kidney function: estimated glomerular filtration rate (eGFR), 33 urine albumin-to-creatinine ratio (uACR) 34 ; (6) others: C-reactive protein (CRP), 35 and hematocrit (HCT). 36P was measured by a nurse using an automated machine (or manually if unavailable), and the mean of available measurements was derived.LDL-c, TG, ApoB, Lp(a), HbA1c and CRP were measured in dedicated central laboratories with acceptable external validation correlations between 2014 and 2017. 37,38The enzymatic selective protection method was used to analyze LDL-c and TG; and the immune-turbidimetric method was used to analyze ApoB, Lp(a), and CRP (Randox Laboratories; Crumlin, County Antrim, United Kingdom using a Beckman Coulter AU5800 Platform). 39levels were used by the immunoturbidimetric method.Circulating concentrations of ALT, GGT were determined using the enzymatic rate method (Beckman Coulter AU5800), and urinary albumin was used in reagents and calibrators sourced from Randox Bioscience.eGFR was computed using the 2009 Chronic Kidney Disease Epidemiology Collaboration equation based on single serum cystatin C measurements taken during baseline assessment center visits by using linear regression to estimate the logarithm of measured GFR from standardized creatinine levels, sex, race, and age. 40In our study, it is chosen over the estimation based on serum creatinine and cystatin C because it has better power in predicting CVD. 19,41Within 24 hours of the blood draw, reticulocyte parameters were assessed on a COULTER LH 750 System and the blood contents on a Beckman-automated hematology sanalyses. 42CT was reported from the instrument by mean (red blood cell volume (MCV) £ red blood cell count (RBC)) /10. 42The urine albumin-creatinine ratio (uACR) is a recognized indicator of urine albumin excretion that urine samples were collected at baseline in all UK Biobank participants. 43,44A random urinary spot was used as a measure of electrolyte excretion. 44

Covariates
Ethnicity, dietary intake, television viewing, smoking, and alcohol consumption were self-reported.The validated International Physical Activity Questionnaire was used to assess self-reported physical activity. 45Townsend area deprivation index was obtained from the postcode of residence and is derived using aggregated data on unemployment, car and home ownership, and household overcrowding. 46Baseline prevalent conditions were self-reported in a nurse-led interview.

Statistical Analyses
The characteristic of participants stratified by the general obesity categories (nonobese and obese) were summarized using means (standard deviations [SDs]) and frequencies (percentages) for quantitative and categorical variables, respectively.All biomarkers were standardized to sexspecific SD in regression analysis to compare the b coefficients.Multiple linear regressions were performed to evaluate the relationship between general obesity and biomarkers.Cox proportional hazards models were used to analyze the associations between general obesity and CVD.Deaths from other causes were censored at the date of death to eliminate the effect of competing risk.All models were adjusted for age, sex, deprivation, physical activity, TV viewing, dietary intake, alcohol consumption, and smoking, as well as medications for cholesterol, BP and insulin were adjusted in the corresponding factors.The g-formula-based mediation analyses 47 were fitted to identify significant mediators, and their associations with ASCVD and HF separately.As a sensitivity analysis, central obesity was used as exposure instead of general obesity.All analyses were conducted using R version 4.2.1 with the packages survival and CMAverse. 48

Results
Of the more than 500,000 UK Biobank participants, 404,332 with complete data available on the exposures, mediators, and covariates were included in this study (Supplementary Fig 1).The overall prevalence of general obesity was 23.5%.Compared with participants without general obesity, those with general obesity were older, more likely to be deprived, watched more television, performed less physical activity, and were more likely to be male and nonvegetarians (Table 1).
Table 3 shows the associations between obesity and CVDs outcomes by adjustment models.Adjusted for sociodemographic and lifestyle factors, obesity was associated with an increased risk of incident ASCVD (HR 1.30, 95% CI, 1.26-1.35)and incident HF (HR 2.04, 95% CI, 1.96-2.13).The associations were attenuated when candidate mediators were additionally adjusted.The largest attenuation for ASCVD was observed when DBP, SBP, eGFR, triglycerides, and HbA1c were included in the model.For HF attenuation was greatest when eGFR, SBP, and DBP were included in the model (Table 3).
The sensitivity analyses using central obesity are shown in the supplementary materials.The distribution of participants by central obesity was generally similar to that identified using general obesity (Supplementary Table 1).The associations between central obesity and candidate mediators also followed similar patterns as for general obesity (Supplementary Table 2).Adjusted for sociodemographic and lifestyle factors, central obesity was associated with a similarly elevated risk of incident ASCVD (HR 1.33, 95% CI, 1.29-1.37)but a lower magnitude of elevated risk of incident HF (HR 1.67, 95% CI, 1.57-1.71).After adjusting for candidate mediators, there was less attenuation than was observed for general obesity (Supplementary Table 3).The mediation analysis results were largely similar but with a lower proportion of the associations explained by the mediators (Supplementary Table 4).

Discussion
Principal Findings This study investigated the underlying mechanisms between 2 types of obesity-general obesity and central obesity, and 2 main types of CVD outcomes-ASCVD and HF.The results showed that while eGFR, BP, TG, and HbA1c were all significant mediators for ASCVD and HF, they explained a much higher proportion of excess ASCVD risk.To our knowledge, this is the first study which systematically, in a single cohort, showed that the excess ASCVD risk due to obesity could be mitigated through intervening intermediate risk factors, but the excess HF risk could not be.

Strengths and Limitations
This study has several key strengths compared with existing studies.Firstly, we were able to systematically examine an extensive range of mediators, including emerging risk factors, such as liver and kidney function markers, in relation to the 2 major types of CVDs.Secondly, the use of counterfactual-based analysis and the carefully selected covariates provided robustness in estimating the relative importance of the mediators.Using a single cohort, rather than meta-analysis, 44 also ensured the consistency of measurements and eliminated any between-study confounding.The exclusion of participants with other preexisting chronic diseases reduced the chances of reverse causation and confounding by other diseases.However, this study has several limitations.As with any observational studies, reverse causation, and residual confounding could not be ruled out completely, despite robust exclusion and a comprehensive set of covariates.Moreover, the mediation analyses did not model the complex causal relationship between candidate mediators; thus, the proportion mediated should not be summed.While we aim to include a comprehensive list of potential mechanisms, it is not possible to including some emerging risk factors such as troponin, broad metabolomic biomarkers, and cardiac imaging.Last but not least, the UK Biobank is not representative of the UK general population even though exposure-outcome associations have been shown to be consistent with those from representative cohorts. 49mparison With Other Studies Our results are generally consistent with mainly of existing studies.For instance, a meta-analysis study of 9 prospective cohort studies (n = 58,322) showed that BP was the largest mediator, accounting for 37% of the associations between obesity and CHD, with blood glucose and cholesterol also accounting for 17% and 6%, respectively. 50These numbers are very close to our findings.Another collaborative analysis of 58 prospective studies investigated the separate and combined associations between BMI and WHR and the risk of incident CVD, and found that lipids, BP, and history of diabetes were the main mediators. 51Also, a population-based study demonstrated that adjustment for CRP substantially attenuated the association between BMI and CVD substantially. 52nterestingly, a Mendelian Randomization mediation study has estimated that genetically predicted diabetes (41%) was a stronger mediator than SBP (27%), even though LDL-c remained a weak mediator (3%). 53However, this study only investigated risk factors for which there was a good genetic risk score.Similar to our findings, the ARIC study also found that obesity has a stronger association with HF than other CVDs, but the associations was largely unexplained by traditional risk factors. 54Notably, this study included additional emerging risk factors, for example eGFR, which explained the obesity-HF associations, highlighting the potential importance of kidney function.

Potential Mechanisms
While observational studies demonstrate association rather than causation, there is existing evidence that causation is biologically plausible. 55rr Probl Cardiol, August 2023 Generally, obesity indicates an excess amount of adipose tissue.Adipose tissue secretes inflammatory cytokines, such as CRP, and interleukin 6 (IL-6), which interacts with increased BP and LDL-c to cause endothelial dysfunction and, subsequently atherosclerosis. 56Obesity can also cause elevated concentrations of LDL-c, which penetrate the subendothelial region to develop plaques. 57One established mechanism is that obesity predispose for type 2 diabetes which could increase CVD risk substantially. 58Obesity may also cause the renin-angiotensin-aldosterone pathway to become more active, increasing the synthesis of the aldosterone hormone. 10However, it should be noted that the relationship between obesity and eGFR is complex and might be time dependent.Others, however, have proposed additional mechanisms for increased aldosterone synthesis in obese people, including greater direct aldosterone synthesis in adipose tissue and direct adrenal gland activation by the cytokine leptin, which is produced by adipocytes. 10Aldosterone and leptin both activate pathways that result in sodium retention, plasma volume expansion, a rise in systemic inflammation, a rise in renal and cardiac fibrosis, a rise in arterial stiffness, and a reduction in the ability of the left ventricle to relax. 10These pathophysiological deviations finally lead to hospitalization for symptomatic HF. 10,59 It should be noted that CRP might not be a causal factor per se 37 but could be a marker for other chronic upstream inflammatory markers, such as IL-6, a potential causal factor. 38

Implications
The findings of this study suggest that controlling multiple intermediate outcomes of obesity could be effective in limiting the subsequent ASCVD burden in the population.The strength of eGFR as a mediator suggest that the current prevention efforts might be enhanced by monitoring and treating the kidney function of obese people. 60Trials, for example, Dapagliflozin in Patients with Chronic Kidney Disease (DAPA-CKD) 61 which indicated that sodium-glucose co-transporter-2 (SGLT2) inhibitors happen to be the best eGFR drug, 62 to test this hypothesis are warranted.Further studies to examine whether suboptimal but clinical insignificant eGFR (eg, >60 mL/min) should indicate clinical action.
However, the study also illustrated that the underlying mechanisms for ASCVD and HF are very different and none of these mediators investigated could sustainably reduce the population burden of HF.For example, the glucagon-like peptide 1 receptor agonists (GLP1RA) could be a candidate to reduce ASCVD risk due to HbA1c being a much stronger mediator for ASCVD than for HF.Without strong mediators, prevention efforts on HF would need to rely on obesity prevention, particularly since the association of obesity with HF is stronger than that with ASCVD.

Conclusion
Interventions that help individuals living with obesity to maintain healthy levels of lipids, BP, HbA1c, and kidney function could potentially alleviate some of the ASCVD burden.However, HF burden could not be meaningfully reduced without weight management.

FIG 1 .
FIG 1. Proportion of mediation estimated independently for each mediator.Overlaps and sequential mediations were not accounted.

TABLE 2 .
Associations between general obesity and the biomarkers included in the analyses by group category All biomarkers were standardised to sex-specific SD so that the beta coefficients are comparable.All results are statistically significant (P < 0.0001).All analyses adjusted for age, sex, ethnicity, deprivation, physical activity, sedentary behaviour, dietary intake, alcohol consumption, and smoking.*Medications for cholesterol, blood pressure and insulin were adjusted in the corresponding factors.

TABLE 3 .
Associations between general obesity and incident CVD outcome by adjustment models

TABLE 4 .
Mediation analysis for the association between general obesity and CVDAll analyses adjusted for age, sex, ethnicity, deprivation, physical activity, sedentary behaviour, dietary intake, alcohol consumption, and smoking.*Medications for cholesterol, blood pressure, and insulin were adjusted in the corresponding factors.