Association of lifestyle with valvular heart disease progression and life expectancy among elderly people from different socioeconomic backgrounds | BMC Medicine

Study population

The UK Biobank is a large-scale prospective cohort study with an open-access protocol [20]. Approximately 500,000 individuals aged 40–69 years were recruited from 22 assessment centers across England, Scotland, and Wales between 2006 and 2010. Baseline information on participants’ lifestyle, socioeconomic, and health-related factors were obtained through questionnaires, interviews, and physical measurements, while blood samples were also collected and processed. All participants provided written informed consent. This study was approved by the National Information Governance Board for Health and Social Care and the North West Multi-Center Research Ethics Committee. Our analysis was restricted to participants who were 60 years old and above. We excluded individuals with missing data on lifestyle factors, SES, or other covariates.

Lifestyle factors

A lifestyle score was created based on 5 risk factors (tobacco smoking, obesity, physical inactivity, unhealthy diet, and unhealthy sleep patterns) measured by the UK Biobank questionnaire at baseline. We adopted three healthy lifestyle criteria based on the cardiovascular health promotion goals of the American Heart Association [21]: not smoking for over a year, maintaining a body mass index (BMI) below 30, and engaging in regular physical activity (at least 150 min per week of moderate intensity, 75 min per week of vigorous intensity, or a combination of moderate and vigorous activities totaling at least 150 min per week). A healthy diet was determined according to the 2021 dietary guidance for cardiovascular health from the AHA [22]. Since 6 of the 13 recommended dietary priorities were unavailable in the UK Biobank, a healthy diet pattern was identified based on increased consumption of fruits, vegetables, whole grains, and fish and decreased or no consumption of refined grains, processed meats, and unprocessed red meats. Additionally, a healthy sleep pattern was defined by the presence of at least four of the following characteristics: being an early riser, sleeping 7–8 h per day, seldom or never having insomnia symptoms, not snoring, and not experiencing excessive daytime sleepiness [23]. We assigned one point to each favorable lifestyle factor and categorized overall lifestyle into three levels: poor (zero to two points), moderate (three points), and ideal (four to five points). Table S1 in Additional file 1 describes the lifestyle factors in greater detail.

SES and covariates

The Townsend Deprivation Index (TDI) was sourced from national census data within the UK Biobank, signifying an area-level socioeconomic status (SES) measure based on variables including car ownership, household overcrowding, owner occupation, and employment [24]. The Townsend scores are a continuous variable, where higher values indicate a lower SES. Participants were divided into two groups, with those scoring above the 60th percentile defined as having low SES.

We also selected two individual-level measures of SES available within the UK Biobank: household income and educational qualifications. Average household income information was self-reported, with low income defined as less than £31,000. Participants also disclosed their highest educational qualifications, choosing from a list that included “College or university degree,” “A levels/AS levels or equivalent,” “O levels/GCSEs or equivalent,” “CSEs or equivalent,” “NVQ or HND or HNC or equivalent,” “Other professional qualifications,” “None of the above,” or “Prefer not to answer.” According to the International Standard Classification of Education [25], we categorized educational attainment into two levels: low, which included “O levels/GCSEs/CSEs or equivalent” and “None of the above,” and high, which encompassed “College or university degree,” “A levels/AS levels or equivalent,” “NVQ or HND or HNC or equivalent,” and “Other professional qualifications.

Covariates including age, ethnicity, the medication use for cholesterol and blood pressure, and a history of hypertension, dyslipidemia, diabetes, coronary artery disease, atrial fibrillation, cardiomyopathy, heart failure, chronic kidney disease, and cancer were collected via a self-administered touchscreen questionnaire. Operative procedures for VHD were coded according to the Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4 (OPCS-4), using hospital inpatient records (K25, K26, K311, K312, K341, K351, K352).

Outcomes

The outcome was major VHD, defined as a composite of the two most common types in the UK Biobank: aortic stenosis and mitral regurgitation. We excluded congenital and rheumatic valve diseases. Incident cases of VHD were identified through linkage with Hospital Episode Statistics (HES) and national death registries. For our analysis, we considered only the first VHD event in compound cases, disregarding any subsequent events. The International Classification of Diseases 10th Revision (ICD 10) was used to code VHD cases. Vital status was obtained from Death Registries. The date of each VHD event was pinpointed as the earliest recorded occurrence within the UK Biobank data. A detailed outcome definition can be found in Table S2 in Additional file 1. Participants were followed-up in the database from their assessment date until July 31, 2021, for Scotland, October 31, 2022, for England, or February 28, 2018, for Wales.

Statistical analysis

Participants of the UK Biobank were followed up from baseline until the date of diagnosis, death, withdrawal, or completion of follow-up, whichever came first. To investigate the role of lifestyle on the progression of VHD, we employed a multistate model with a Gompertz distribution [26]. This model included three states: “NO-VHD,” “VHD,” and “death.” Three transitions were observed in participants: (1) from no VHD to incidence of VHD, (2) from no VHD directly to death, and (3) from VHD diagnosis to death. We only considered the first occurrence of transitioning into a new state and did not allow for reversal.

When analyzing transitions 1 and 2, participants who had prevalent VHD at the baseline were excluded. For transition 3, we included those with prevalent VHD at the baseline as well as those who developed VHD during follow-up. The multistate survival analysis was controlled for covariates including age, ethnicity, histories of hypertension, dyslipidemia, and diabetes, as well as medication use for cholesterol and blood pressure. Age at diagnosis of VHD was further adjusted in the analysis of transition 3. We estimated hazard ratios (HRs) to measure the association of lifestyle alone, the combination of lifestyle and SES, and lifestyle within SES-stratified subpopulations with VHD incidence and mortality. In addition, we calculated the dose–response relationship between lifestyle category and health outcomes, in which lifestyle category was treated as a continuous variable. To explore multiplicative interactions, a product term combining the socioeconomic and lifestyle scores was additionally included in our model.

We assessed the life expectancy of individuals with or without VHD categorized by lifestyle categories across SES levels. First, the Gompertz multistate model was employed, utilizing age from 60 years as the time scale, to determine the age-specific and SES-specific transition rates for each transition. Next, the baseline prevalence of lifestyle levels was calculated by every 10-year age bracket and by SES among participants without VHD at baseline and those with VHD during follow-up. Then, we used SES-specific hazard ratios by lifestyle, and the proportions of lifestyle categories, to weight transition rates as described in previously published methods [27]. Finally, using the weighted age-specific and SES-specific transition rates, we constructed life tables for people with and without VHD [28, 29]. We assumed that individuals currently without VHD could develop the condition in the future. Our life tables concluded at 100 years, with methodological details provided in the supplemental method in Additional file 1. Years of life gained due to healthier lifestyles was calculated by subtracting the life expectancies of participants with ideal and moderate lifestyles from those with a poor lifestyle of the same age and SES. The 95% confidence interval for the life expectancy was estimated by conducting 2000 Monte Carlo simulations. To account for gender-specific differences in life expectancy, all analyses were conducted separately for men and women.

We confirmed the robustness of our analyses by performing sensitivity analyses in which (1) the continuous lifestyle score was considered rather than categorical variable; (2) the five lifestyle factors were analyzed separately, instead of being aggregated into a single score; (3) additional adjustments were made for potential confounders in the lifestyle-VHD relationship, such as coronary artery disease, atrial fibrillation, cardiomyopathy, heart failure, chronic kidney disease, and cancer; (4) the Cardiovascular Health (CVH) metric, introduced by the American Heart Association (AHA), was used instead of lifestyle; (5) aortic stenosis and mitral regurgitation were considered separately, instead of being aggregated into a general VHD category; (6) low socioeconomic status was indicated by a TDI percentile above 40, rather than 60th percentile; (7) individual-level measures of SES, such as household income and educational attainment, were used; (8) participants who developed VHD or died within the first 2 years of follow-up were excluded; (9) for the analysis of VHD patients (transition 3), we made additional adjustments to account for any operations related to VHD; (10) participants without VHD were assumed to be immune to the disease, and their expected lifetimes were recalculated.; (11) multiple imputation was conducted to impute all missing independent variables and test the influence of missing data.

All statistical analyses were performed using the R software 4.0. All P values were two-sided and considered statistically significant when less than 0.05.