Study population
The individual-level data collected from participants enrolled in the UK Biobank (Application Number: 65711) were utilized in this study. The detailed study design and population of the UK Biobank have been previously described [15, 16]. Briefly, the UK Biobank is an ongoing prospective cohort that incorporated data between 2006 and 2010 from 22 assessment centers across the United Kingdom; the participants were aged between 40 and 69 at recruitment. Demographics, healthy lifestyle information, and other potentially health-related information were obtained through touch screen questionnaires, face-to-face interviews, physical examinations, and biological samples.
In the present study, among the 502,356 participants, we excluded individuals who withdrew from study (n = 85), had missing quality-controlled genotyping data (n = 16,231), had missing information on metabolic related factors and BMI data, or were underweight (BMI 2) (n = 66,833). Furthermore, participants with a prior history of cardiovascular events or cancer were excluded. Finally, 479,461 participants with at least one outcome (cardiovascular outcomes or all-cause mortality) were included. To examine whether transitions in metabolic status, BMI status, and BMI-metabolic status (time window for the transition: from baseline in 2006–2010 to the second survey in 2012–2013) altered the aforementioned outcomes, 18,058 participants were enrolled in the subsequent analysis (Fig. 1).
Assessment of metabolic health status and BMI categories and their transitions
According to the National Cholesterol Education Program-Adult Treatment Panel III (NCEP-ATP III) criteria [17], MetS was defined as the presence of 3 or more following abnormal components: (1) waist circumference > 102 cm in men and > 88 cm in women; (2) systolic blood pressure (SBP) ≥ 130 mmHg or diastolic blood pressure (DBP) ≥ 85 mmHg or antihypertensive agents; (3) serum glucose ≥ 6.1 mmol/L or antidiabetic agents; (4) serum triglyceride (TG) ≥ 1.7 mmol/L or antihyperlipemic agents; (5) high-density lipoprotein cholesterol (HDL-C) 18]. BMI categories were classified into three categories based on WHO guideline [19]: normal weight (18.5 ≤ BMI 2), overweight (25 ≤ BMI 2), and obesity (BMI ≥ 30 kg/m2). According to the combination of metabolic status and BMI categories, we classified participants into 6 groups: MHN, MHOW, MHO, metabolically unhealthy normal weight (MUN), metabolically unhealthy overweight (MUOW), and metabolically unhealthy obesity (MUO).
Furthermore, transitions in metabolic status (MH throughout, MH to MU, and MU throughout), BMI status (normal weight throughout, normal weight to overweight, overweight throughout, overweight to obesity, and obesity throughout), and BMI-metabolic status (MHN throughout, MHN to metabolically healthy overweight or obesity [MHOO], MHOO throughout, MHOO to metabolically unhealthy overweight or obesity [MUOO], and MUOO throughout) were characterized from baseline to the second survey.
Definition of genetic predisposition
To evaluate genetic predisposition, PRSs for CVD, coronary disease (CAD), myocardial infarction (MI), stroke, heart failure (HF), and atrial fibrillation (AF) were constructed for each participant. In brief, the PRSs of CVD, CAD, and AF were extracted from ‘Standard PRS (Category 301)’ provided by the UK Biobank PRS Release. Furthermore, 31 single SNPs related to MI, 32 SNPs related to stroke, and 12 SNPs related to HF were employed to determine the PRSs for MI, stroke, and HF, respectively, as reported in published genome-wide association studies (Tables S1–S3) [8,9,10]. To mitigate the impact of SNP deletions, we utilized the following formula to calculate the PRSs for each individual [20]:
$$ {\text{PRS}}_{{\text{j}}} = \mathop \sum \limits_{{\text{j}}} \frac{{{\text{S}}_{{\text{i}}} \times {\text{G}}_{{{\text{ij}}}} }}{{M_{{\text{j}}} }} $$
(1)
In this formula, ‘S’ presents the effect value (beta/odds ratio), ‘G’ symbolizes the allele dose (with each SNP being recoded as 0, 1, or 2, according to the number of risk alleles), and ‘M’ presents the total number of SNPs. The subscript ‘i’ denotes the sequence number of the SNP, whereas the subscript ‘j’ pertains to the sequence number of the individual. Additionally, we categorized individuals into three distinct groups in line with their PRSs: low (quintile 1), intermediate (quintiles 2–4), and high (quintile 5), as detailed previously.
Follow-up and outcome ascertainment
Participants without CVD or cancer were followed up from the date of baseline examination until the first occurrence of current study outcomes, loss to follow-up, or the censoring date (October 12, 2023, defined as the end date of disease and mortality data collection), whichever came first.
The primary outcomes included all-cause mortality, nonfatal CVD morbidity, and nonfatal CVD mortality. Nonfatal CVD consists of nonfatal CAD (I20–I25), MI (I21–I23, I24.1, and I25.2), stroke (I60-I64), HF (I11.0, I13.0, I13.2, I50.X), and AF (I48), which were identified according to International Classification of Diseases, Tenth Revision (ICD-10) codes. The records on the incidence of CVD, CAD, MI, stroke, HF, and AF were obtained by linking with the primary care system, hospital inpatient records, and the death registry. Mortality information was determined by matching with the death registries of the National Health Service Information Centre [15].
Assessment of covariates
A series of covariates in the present study were obtained through touch-screen questionnaires or face-to-face interviews, including age, sex, race (white, mixed, Asian or Asian British, black or black British, Chinese, and other), Townsend Deprivation Index (with higher values representing lower socioeconomic status), annual household income (£; 100,000), educational attainment, sleep duration, healthy diet (yes or no), physical activity (low, middle, or high), smoking status (never, previous, or current), and alcohol intake frequency (never, special occasion only, one to three times a month, once or twice a week, three or four times a week, and daily or almost daily). Levels of educational attainment were classified into 6 levels: (1) no qualifications, (2) Certificate of Secondary Education or Ordinary Levels/General Certificate of Secondary Education or equivalent, (3) Advanced Levels/Advanced Subsidiary Levels or equivalent, (4) other professional qualification, (5) National Vocational Qualification or Higher National Certificate or equivalent, and (6) college or university degree [21]. A healthy diet was based on eating at least 5 portions of a variety of fruits and vegetables every day, following the NHS guidelines [22]. The self-reported physical activity level was assessed using the well-validated International Physical Activity Questionnaire-Short Form [23]. We addressed missing covariates by employing a missing indicator category for categorical variables and substituting mean values for continuous variables.
Statistical analysis
The participants’ baseline characteristics, encompassing sociodemographic characteristics, socioeconomic status factors, and metabolic risk factors, are presented as mean ± standard deviation (SD) for continuous variables and as percentages for categorical variables. The chi-square (χ2) test was used for comparing categorical variables, while analysis of variance or Student’s t test was performed for continuous variables. Cox proportional hazard models, with duration of follow-up as the time scale, were utilized to evaluate the associations of exposures (metabolic status, BMI status, BMI-metabolic status, and their transitions) and PRS with cardiovascular outcomes and all-cause mortality across BMI categories. The proportional hazard assumption was examined using Schoenfeld residuals. Two Cox proportional hazard models were fitted. Model 1 was adjusted for age, sex, race, Townsend Deprivation Index, annual household income, educational attainment, 22 assessment centers, and the first 5 principal components of ancestry. Model 2 was further adjusted for family history of diabetes, family history of high blood pressure, and lifestyle factors including sleep duration, healthy diet, physical activity, smoking status, and alcohol intake frequency, based on Model 1. Furthermore, the subgroup analysis of females was additionally adjusted for pregnancy history and menopausal status.
We analysed the effect of metabolic status, BMI status, and BMI-metabolic status on all-cause mortality, CVD morbidity, and CVD mortality stratified by different levels of PRS. To investigate the joint association between exposures (metabolic status, BMI status, BMI-metabolic status, and their transitions) and PRS, we established the following new product terms: six categories for metabolic status and PRS (2 × 3), nine categories for BMI status and PRS (3 × 3), eighteen categories for BMI-metabolic status and PRS (6 × 3), nine categories for transitions in metabolic status and PRS (3 × 3), fifteen categories for transitions in BMI status and PRS (5 × 3), and fifteen categories for transitions in BMI-metabolic status and PRS (5 × 3). Hazard ratios (HRs) and their 95% confidence intervals (CIs) for each outcome across these groups were calculated. Likelihood ratio tests were used to evaluate the significance of the multiplicative interaction term by comparing models with and without this term.
Several subgroup analyses were conducted to examine the stability and possible variations of the primary results, stratified by age (P value less than 0.05 was considered to indicate statistically significant.