Higher consumption of ultra-processed foods is associated with increased risk of obstructive sleep apnoea: A doubly robust machine learning approach — The Association Specialists
  1. Schedule
  2. Oral Presentations - Structural, commercial and social determinants of health
  3. Higher consumption of ultra-processed foods is associated with increased risk of obstructive sleep apnoea: A doubly robust machine learning approach

Higher consumption of ultra-processed foods is associated with increased risk of obstructive sleep apnoea: A doubly robust machine learning approach (24165)

Tefera Mekonnen 1 , Zegeye Abebe 1 , Muktar Ahmed 1 , Anteneh Mengist Dessie 1 , Teketo Tegegne 2 , Robert Adams 1 , Yohannes Adama Melaku 1
  1. Flinders University, Adelaide, Australia
  2. Deakin University, Melbourne, Australia

Background: Ultra-processed foods (UPFs) are linked to metabolic and inflammatory disorders, but their association with obstructive sleep apnoea (OSA) remains unclear. We examined the potential causal effect of UPF intake thresholds on incident OSA in a large UK cohort.

Methods: We conducted a prospective cohort study of 115,165 UK Biobank participants. UPF intake was quantified as the percentage of total grams/day from web-based dietary assessments collected between February 2011 and April 2012. Three thresholds were examined: ≥25%, ≥50%, and ≥75% versus lower intake. Incident OSA was identified from linked hospital and primary care records through 27 October 2022. Targeted maximum likelihood estimation (TMLE) was used to estimate risk ratios (RRs), adjusting for demographic, socio-economic, lifestyle, dietary, and clinical confounders. Secondary models, subgroup analyses, sensitivity analyses, negative controls, and E-values were also assessed.

Results: Higher UPF intake was associated with progressively increased OSA risk. In TMLE models, RRs were 1·29 (95% CI 1·16–1·43) for ≥25%, 1·55 (1·39–1·73) for ≥50%, and 2·11 (1·83–2·43) for ≥75% UPF intake. Corresponding numbers needed to harm were 284, 139, and 64. Secondary modelling approaches produced similar estimates, and continuous UPF modelling supported a monotonic dose–response pattern. By body mass index status, RRs for UPF intake ≥25%, ≥50%, and ≥75% were 0·90 (0·66–1·24), 1·43 (0·98–2·09), and 2·72 (1·62–4·56) among participants without obesity, and 1·23 (1·06–1·43), 1·35 (1·17–1·55), and 1·57 (1·22–2·01) among participants with obesity, respectively. Adjustment for body mass index and diet quality attenuated, but did not eliminate, the associations. UPF subgroup analyses suggested that beverages contributed most strongly to the overall association. Negative control analyses supported specificity, and E-values indicated greater robustness at higher thresholds.

Conclusion: Higher UPF intake was dose-dependently associated with increased OSA risk. Reducing UPF consumption may help lower OSA burden, although mechanistic and interventional studies are needed.