The heating comes on, the clocks have gone back, and hospital wards start filling with a pattern that clinicians have recognised for decades: more heart attacks in January than in June. What has been harder to pin down is why. Two studies published a day apart at the end of October attack that question from opposite directions — one from the calendar, one from the thermometer — and the gap between their answers is more instructive than either result alone.
A cohort of 453,597 people, sorted by season
The first is a UK Biobank analysis in Open Heart on 28 October. Researchers at the University of Glasgow followed 453,597 adults recruited from 22 assessment centres between 2006 and 2010 — 56% women, mean age 56 — and modelled seven outcomes against the calendar using a cosinor method, which fits a repeating 12-month wave rather than comparing isolated months.
Across follow-up there were 33,857 deaths, of which 10,807 (31.9%) were cardiovascular. All-cause mortality peaked in February. The rate in winter was 62.2 per 10,000 person-years against 51.5 in summer. Cardiovascular deaths ran at 20 per 10,000 person-years in January versus 16 in August; non-cardiovascular deaths also peaked in winter, at 42 in January against 35 in September.
The hospital data were more specific. Of 12,273 first heart-failure admissions, 9,520 for myocardial infarction, 7,137 for stroke and 14,492 for peripheral vascular disease, the sharpest seasonal signal belonged to heart attacks: 17 per 10,000 person-years in January against 14 in September. Stroke admissions showed the same direction, 13 against 11.
None of that is a surprise on its own. It is the kind of finding that gets published because the cohort is large enough to test whether the pattern holds inside subgroups — and here it largely did, across sex, age, ethnicity and deprivation. The patterns survived adjustment for risk factors and clustering by assessment centre.
Where the pattern breaks
Two results complicate the clean story. Heart-failure and peripheral-vascular admissions peaked in spring, not winter. And the seasonal swing was larger in participants under 65 than over it, with the older group’s deaths and hospitalisations peaking later in the year — the opposite of what most cold-weather research has reported.
The authors offer a temperate-climate explanation: in the UK the difference between a January and an August average is modest, so a real effect in older adults may be too small for the model to locate reliably. Their own limitations section is unusually frank about something more basic. They had no detailed weather, temperature or air-pollution data. The conclusion that colder temperatures rather than warmer ones drive the excess is an inference from the seasonal shape, not a measured exposure–response curve.
A second dataset that does have a thermometer
The second study, in JAMA Network Open on 29 October, supplies what the UK cohort lacks. Researchers linked daily mean ambient temperature, assigned by residential postcode, to Medicare claims for 1,517,756 heart-failure hospitalisations among 996,312 patients aged 65 and over between 2007 and 2022 (mean age 78, 60.3% female). The design was a time-stratified case-crossover: each patient’s admission day was compared with nearby non-event days for the same person, a structure that controls for everything stable about that patient — their housing, their baseline health, their postcode.
Against a reference of 75 °F, colder days were associated with more heart-failure admissions, with a 14-day lag. At each region’s fifth-percentile temperature the odds ratios were 1.62 in the West North Central region (95% CI 1.25–2.11), 1.40 in the West (1.28–1.53), 1.34 in the Southwest (1.11–1.62) and 1.34 in the Northwest (1.08–1.65). Days above 75 °F showed no association with heart-failure admissions, nationally or regionally.
That last null is the paper’s real argument. The familiar U-shaped curve of temperature and cardiovascular risk — harm at both extremes — did not hold for heart-failure hospitalisations in this population. Cold mattered; heat did not move admissions. The authors are careful to say this should not be extended to heart-failure deaths, where heat signals have been stronger, and they raise the possibility of “harvesting”: the most fragile patients may die before they can be admitted, flattening the hospitalisation curve.
Both can be right
Read side by side, the two studies look contradictory — a spring peak in one, a cold-driven increase in the other. They are measuring different things. A season is a three-month box; a daily temperature is a number. An admission recorded in April may follow a cold snap in March, and the UK authors themselves suggest the spring heart-failure peak “may be related to the warming that occurs after winter” — a lag rather than a new insult. The UK cohort counted only first admissions, in a volunteer sample with a mean age of 56; the Medicare analysis counted any heart-failure admission in everyone over 65. One lacked exposure data; the other lacked the full seasonal sweep of a cohort study.
What both support is narrower than a headline: the cold half of the year carries more cardiac events, and the association is visible at the daily scale wherever a thermometer is actually part of the dataset.
What cold does to a circulation
A 2026 narrative review in Frontiers in Public Health gathers the physiology. Acute cold activates the sympathetic nervous system, constricts peripheral vessels, raises arterial pressure and myocardial oxygen demand, impairs endothelial function and pushes blood toward a more clot-prone state — haemoconcentration and prothrombotic shifts. That is a coherent mechanism for why a cold morning might tip a compromised heart.
The review is explicit that mechanism is not proof. Its authors write that these findings provide “biological and clinical plausibility” but “do not by themselves establish an effect on population-level cardiovascular mortality.” It is a narrative synthesis, not a systematic review with pooled estimates, and its 26 studies used different temperature metrics, cold definitions, lag structures and outcomes.
Its burden figures are worth knowing, with that caveat attached. A US county-level analysis covering 2000–2020 attributed 42,735 cardiovascular deaths a year — 6.3% of the annual total — to temperatures below each county’s minimum-mortality temperature. In 15 Chinese megacities over 2007–2013 the estimate was 15.8% (95% CI 13.1–17.9%). Globally, one modelling study put 1,801,513 annual cardiovascular deaths in 2000–2019 in the non-optimal temperature range, of which 1,666,814 were on the cold side. The review warns that these come from different populations, periods and definitions and should not be read as directly comparable — and that it could find no consistent gradient by latitude, climate or income. Warm-climate and lower-resource settings remain thinly studied.
What the guidance actually says
The practical content is unglamorous. CDC puts heart disease at one in five deaths in the United States (22%) and notes that about 47% of adults have at least one of three key risk factors: high blood pressure, high cholesterol or smoking. Its symptom list for a heart attack includes chest pain or discomfort, upper back or neck pain, heartburn, nausea or vomiting, extreme fatigue, dizziness and shortness of breath — a set that overlaps with plenty of harmless complaints, which is precisely why NHLBI’s instruction is blunt: if you think you or someone else may be having a heart attack, call 9-1-1 right away, because the longer the heart goes without enough oxygen, the more damage is done to the heart muscle.
On the cold side, CDC’s winter guidance is aimed at hypothermia rather than cardiology, but it names the same population: older adults without adequate food, clothing or heating are at high risk, and hypothermia can occur even above 40 °F if someone becomes chilled from rain, sweat or submersion. The agency’s outdoor advice is layers, a hat, mittens and a water-resistant outer layer.
The review’s framing of “adaptive capacity” — clothing, activity, heating, housing, warning systems, healthcare access — is the closest thing to a public-health lever in this literature, and it comes with its own admission that the relative contribution of those factors “cannot be determined from the available evidence.”
What we did not find
No trial showing that any winter-specific intervention — warming centres, cold-weather alerts, heating assistance — reduces cardiac events. No study identifying a temperature threshold below which risk steps up. No exposure data in the UK cohort, which means its “cold” conclusion rests on the shape of the calendar. Both datasets are observational, and neither can rule out that people who feel unwell in winter are also the people whose other risks are already clustered.
What is left is a reasonable, modest position: winter is when cardiac events concentrate, cold is the most plausible driver, and the numbers are large enough that the season deserves the same attention as the risk factors on a clinic checklist. Anyone with diagnosed heart disease, high blood pressure or a history of heart failure who wants to know what the cold months mean for their own medication, activity or heating should ask a clinician rather than a cohort study. Our medical disclaimer explains the limits of what an article like this can tell you.
Two related pieces from the archive: our report on why a blood-pressure drop on standing matters most when it doesn’t recover, and the walking-intensity trial in frail older adults — both about the same circulation, measured under different conditions.