The research team, led by Jiayue Zhang and Zhongxin Xu of the Department of Neurology at the China-Japan Union Hospital of Jilin University, set out to answer a question that has long hovered over environmental epidemiology: does smoking modify the way temperature affects stroke risk? Smokers already carry a heavier burden of vascular disease, and laboratory evidence suggests that both smoking and cold exposure stress the cardiovascular system through overlapping mechanisms, including increased blood viscosity, vasoconstriction, and inflammatory activation. If those pathways interact, smokers might be expected to show an amplified response to cold snaps. Testing that hypothesis in a real-world population required a dataset with precise timing of symptom onset, reliable smoking histories, and a climate cold enough to produce meaningful temperature variation — conditions that Changchun, a city with harsh winters in Jilin Province, provides in abundance.

The methodological core of the study is the distributed lag non-linear model, or DLNM, a statistical tool that has become a workhorse in climate-health research. Unlike simpler regression approaches that ask whether today’s temperature predicts today’s outcomes, a DLNM can simultaneously model how the effect of a temperature exposure varies with its intensity and how that effect is distributed across time — the lag dimension. The researchers combined this framework with quasi-Poisson generalised linear models, appropriate for count data such as daily stroke admissions, and adjusted for relative humidity, long-term trends, seasonality, day of the week, and public holidays. This adjustment strategy is critical, because temperature is strongly correlated with season, and failing to disentangle the two can produce spurious associations.

The team examined six distinct temperature indicators rather than relying on a single measure: daily average, apparent, minimum, and maximum temperatures; the diurnal temperature range, which captures how much the temperature swings within a single day; and the temperature change between neighbouring days, which captures rapid shifts in weather. Each indicator tells a different meteorological story. Average and minimum temperatures reflect sustained cold, while the diurnal range and day-to-day change probe the physiological cost of thermal variability. The apparent temperature, which incorporates humidity, attempts to capture what the body actually experiences. This multi-indicator approach allowed the researchers to distinguish between the effects of absolute cold and the effects of temperature volatility — two dimensions of winter weather that are often conflated.

The results paint a temporally nuanced picture. Lower average, apparent, minimum, and maximum temperatures were all associated with increased daily counts of first-ever ischaemic stroke, but only after a delay of approximately two to three weeks. This lagged pattern is longer than many earlier studies have reported and suggests that the physiological cascade linking cold exposure to arterial thrombosis may unfold over a matter of weeks rather than hours. By contrast, a lower diurnal temperature range — meaning less variation between daytime highs and nighttime lows — was associated with increased case counts mainly on the same day, at lag zero. The temperature change between neighbouring days showed weaker, direction-dependent associations concentrated within the first two weeks: days of warming were followed by more strokes, while cooling episodes were followed by fewer, at least in the short term.

What might explain a two-to-three-week delay between a cold snap and a surge in strokes? The authors point to plausible biological pathways. Cold exposure is known to raise blood pressure, increase plasma fibrinogen and cholesterol concentrations, promote platelet aggregation, and trigger inflammatory responses — changes that could gradually destabilise atherosclerotic plaques or promote thrombus formation over days to weeks. Respiratory infections, which peak in cold weather and are themselves recognised stroke triggers, may also contribute to the delayed pattern. Alternatively, the lag could partly reflect behavioural and healthcare-seeking dynamics, as cumulative winter stress on vulnerable individuals builds toward a clinical event. The study’s design cannot definitively separate these mechanisms, but the consistent lag structure across multiple temperature indicators strengthens the case that the delay is real rather than a statistical artefact.

The headline negative result concerns smoking. Formal interaction tests found no evidence that smoking status modified the temperature–stroke association: the unadjusted P values for the interaction terms ranged from 0.314 to 0.623, and after Holm correction for multiple comparisons all adjusted P values were 1.00. In other words, smokers and non-smokers in this cohort responded to temperature in statistically indistinguishable ways. The cohort itself was nearly evenly split — 1,205 smokers and 1,220 non-smokers — giving the analysis reasonable power. For public health messaging, this suggests that cold-weather stroke warnings should not be targeted exclusively at smokers; the entire population appears to carry similar temperature-related risk, even though smokers face elevated baseline risk from other mechanisms.

The study also ventured into the laboratory. Among 1,688 patients with complete bloodwork, the researchers explored whether laboratory indicators — including inflammatory markers, lipid fractions, and haematological measures — differed across categories of absolute temperature exposure. After applying false discovery rate correction to guard against spurious findings, some indicators did differ across temperature categories in the overall cohort and among non-smokers, but not among smokers. The authors frame these analyses as exploratory, and the absence of differences in the smoking group is consistent with the null interaction finding, though the biological interpretation remains tentative. Notably, the study’s exploratory nature and its correction strategy reflect a growing emphasis in epidemiology on distinguishing confirmatory findings from hypothesis-generating ones.

The study has limitations worth noting. It was conducted in a single city over roughly fifteen months, and the authors excluded records from 2022 to reduce the confounding influence of COVID-19-related changes in healthcare utilisation — a decision that shortened the observation window and may limit generalisability to other climates and healthcare systems. Case counts were defined by symptom-onset date, which is more precise than admission date but still depends on accurate patient recall. As with all time-series studies, the design identifies population-level associations rather than individual-level causal effects. Nevertheless, the consistency of the lagged cold effect across four separate temperature indicators, and the careful adjustment for seasonality and humidity, lend credibility to the central conclusion.

The practical implications are concrete. The authors argue that their findings support sustained cold-season preparedness — not just alerts on the coldest days, but health system readiness that persists for weeks after cold exposure, given the delayed risk peak. They also recommend post-exposure health monitoring for vulnerable individuals, alongside the familiar pillars of smoking cessation and vascular risk factor management. As climate change reshapes temperature patterns worldwide, including more frequent and intense cold extremes in some regions, understanding the precise timing of weather-related stroke risk becomes increasingly valuable for hospitals, emergency services, and public health agencies. This study adds an important temporal dimension to that effort: the danger of a cold snap, it seems, does not pass when the weather warms — it lingers, quietly, for weeks.

Subject of Research: Lagged associations between ambient temperature and first-ever ischaemic stroke onset, and effect modification by smoking status, using distributed lag non-linear models

Article Title: Lagged associations between ambient temperature and first-ever ischaemic stroke onset by smoking status: a distributed lag non-linear model study

Article References: Zhang, J., Li, Z., Wang, C., Chen, F., Cheng, J., Shi, X., & Xu, Z. (2026). Lagged associations between ambient temperature and first-ever ischaemic stroke onset by smoking status: a distributed lag non-linear model study. BMC Public Health. https://doi.org/10.1186/s12889-026-29964-9

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29964-9

Keywords: ischaemic stroke, ambient temperature, distributed lag non-linear model, smoking status, time-series study, cold exposure, diurnal temperature range, public health, Changchun, climate and health, stroke risk, epidemiology