Light at night, and a heart that remodels

Design

Population-based cohort with wrist-worn light sensing, later cardiac MRI, and Cox models for incident events

Population

11,071 UK Biobank participants with accelerometry and subsequent cardiac MRI; 73,286 for outcome analyses

Primary outcome

Cardiac MRI structure and function phenotypes, and incident cardiovascular disease and mortality

Effect

High vs no night-time light above 3 lux: indexed LV mass +2.4% (95% CI 1.6-3.1), wall thickness +1.5%, myocardial contraction fraction -1.9%, global strain impaired 2.7-2.9%. Over 8-10 years: heart failure +29% (12-49), atrial fibrillation +15% (4-27), myocardial infarction +24% (7-44), stroke +33% (11-59), cardiovascular death +26% (5-52). 24-49% statistically mediated by shorter sleep

Eleven thousand and seventy-one UK Biobank participants wore a wrist device with a light sensor for a week, and had cardiac magnetic resonance imaging three years later. A larger group of 73,286 contributed to outcome analyses. The exposure was night-time light above 3 lux — roughly a streetlight through thin curtains, or a television left on.

High exposure was associated with a pattern of concentric left ventricular hypertrophy and subclinical functional decline: 2.4% greater indexed LV mass (95% CI 1.6-3.1), 1.5% greater wall thickness, 1.9% lower myocardial contraction fraction, and impaired strain in all three directions of about 2.7 to 2.9%. The right ventricle and left atrium showed the same direction of change. Relationships were linear across the dose range, and 24 to 49% of the association was statistically mediated by shorter sleep duration.

Over 8 to 10 years of follow-up, high exposure went with 29% higher heart failure risk, 15% higher atrial fibrillation, 24% higher myocardial infarction, 33% higher stroke and 26% higher cardiovascular mortality.

The percentages on the imaging measures are small — a 2.4% difference in indexed mass is not something you would see on a report. What makes the study worth reading is the consistency across chambers and the dose-response, which is the pattern you would expect if this were real.

• Ask about the sleeping environment when sleep is raised: streetlight, television, phone, shared room
• Frame the advice as darkness plus duration, since much of the association ran through sleep length
• Do not overstate it: these are small structural differences in an observational cohort
• Relevant in Indian urban housing, where light control is often not within the household's power
• No intervention has been tested; this is exposure data only

Why it matters

It gives a physical, addressable target for sleep advice rather than the usual exhortation to sleep more.

Don't overread it

An observational cohort with a single week of light measurement cannot show that reducing light at night changes cardiac structure or events.

The statistics, in plain English

Percentage differences of 1.5 to 2.9% in imaging measures are far below what any single scan could distinguish; they are population averages detectable only because the cohort is large. The outcome hazard ratios are more substantial, but light exposure was measured once for seven days and treated as a lasting exposure. Mediation analysis attributing 24 to 49% of the effect to sleep duration is a statistical decomposition of observational data, not a demonstration that shortening sleep is the mechanism.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.