What are the effects of sociodemographic variables on the association between a weighted lifestyle score and mortality in the UK Biobank cohort?

Conference:

SAPC ASM 2024 - Bristol

Talk Code:

4E.2

Presenter:

Hamish Foster

Twitter:

@hamishfoster

Co-authors:

Dr Hamish M E Foster, Prof Jason M R Gill, Prof Frances S Mair, Dr Carlos A Celis-Morales, Dr Bhautesh D Jani, Dr Barbara I Nicholl, Prof Duncan Lee, Prof Catherine A O’Donnell

Author institutions:

University of Glasgow

Problem

Unhealthy lifestyles are associated with disproportionate mortality among deprived populations where need for lifestyle support is greatest. Lifestyle scores can support individuals make healthy change across a range of factors to prevent adverse health outcomes. However, barriers to use and effectiveness of scores in more deprived populations include: 1) physiological (e.g., weight or blood pressure) score components, requiring time and resources, 2) arbitrary weighting of different lifestyle factors (e.g., smoking given the same weighting as physical inactivity) thereby missing opportunities to convey more accurate personalised risk, and 3) failure to account for additional risk associated with deprivation (e.g., as done for ASSIGN/QRISK but not yet done for lifestyle scores). We aimed to create a simple weighted lifestyle score and examine the effects of sociodemographic variables on the association between score and mortality.

Approach

Prospective analysis of 462,235 UK Biobank participants aged 37-73 years. A weighted lifestyle score was developed using 11 self-reported lifestyle factors (LFs): smoking, alcohol, physical activity, TV time, intake of red meat, processed meat, salt, oily fish, fruit and vegetables, sleep, social participation. Cox models adjusted for demographics and health conditions were used to examine associations between individual LFs and all-cause mortality to determine score weightings. Weightings were combined into a lifestyle risk score to then explore the effects of deprivation, sex, ethnicity, and age on the association between weighted score and all-cause and CVD mortality.

Findings

Over 12.0 years median follow up, 30,687 (6.6%) participants died including 4,632 (1.0%) CVD deaths. Each LF was independently associated with both outcomes and hazard ratios (HR (95%CIs)) ranged from 2.20 (2.03, 2.15) for smoking to 1.02 (1.00, 1.05) for low oily fish intake. Weighted score (maximum 30 points indicating unhealthy) comprised 14 points for smoking, 1 each for unhealthy levels of intake of oily fish, red meat, processed meat, salt, and 2 each for remaining factors. There were dose-response increments for all-cause and CVD mortality HRs with each additional score point. Associations were stronger in more deprived quartiles and among men. With least deprived and lowest score category as reference, all-cause mortality HRs for highest (unhealthiest) score was 2.67 (2.43, 2.92) in the least deprived and 4.71 (4.43, 5.01) in the most deprived. Equivalent figures but with women and lowest score category as reference, were 3.07 (2.88, 3.26) among women and 4.66 (4.44, 4.89) among men.

Consequences

An extended weighted lifestyle score comprised of 11 self-reported factors has strong associations with mortality, particularly among more deprived and male participants. Deprivation and sex could be incorporated into a simple lifestyle risk score that could convey personalised risk and inform policy and future interventions in areas of deprivation.

Submitted by:

Hamish Foster

Funding acknowledgement:

HF is supported by Medical Research Council Clinical Research Training Fellowship (grant number MR/T001585/1). Remaining co-authors received no funding for this work.