Weather patterns linked to daily afebrile seizures in children, study finds

A sweeping four-year analysis of pediatric seizure admissions in Türkiye has found that weather — long blamed by parents and even some clinicians for triggering childhood seizures — shows no measurable independent effect on the daily number of afebrile seizures in children. The study, published in BMC Pediatrics, systematically tested temperature, atmospheric pressure, humidity, wind speed, wind direction and their daily ranges against daily seizure counts and found no statistically significant association for any of them, challenging a persistent belief that has circulated in families of children with epilepsy for generations.

The research team, led by Eren Sari of Canakkale Onsekiz Mart University together with Canan Ustun of Ankara Etlik City Hospital and Bulent Unay of the University of Health Sciences, Ankara, assembled one of the more granular datasets brought to bear on this question. Between April 2017 and April 2021, they identified 494 eligible afebrile seizure presentations among 429 children at a tertiary university hospital in Ankara. Crucially, they excluded seizures accompanied by fever, removing the major confounder that has muddied earlier work, since febrile seizures — common in young children and strongly linked to infections and rapid temperature changes — have often been lumped together with afebrile events in previous analyses.

The meteorological exposure data came from a weather station located roughly two kilometers from the hospital, close enough that daily readings of mean temperature, atmospheric pressure, relative humidity, wind speed, diurnal temperature range, diurnal pressure range and wind direction could be matched reliably to each calendar day. Over the study window, seizures occurred on 396 of 1,491 calendar days, giving the researchers an unevenly distributed count series — exactly the kind of data structure in which naive statistical approaches can manufacture spurious weather effects where none exist.

To avoid that trap, the team applied generalized additive models within a Poisson regression framework, the standard machinery for analyzing daily count data in environmental epidemiology. Poisson regression is appropriate because the outcome — the number of seizure presentations per day — is a non-negative integer with many zeros and occasional larger values. Generalized additive models, in turn, allow the relationship between each weather variable and seizure counts to take any smooth, nonlinear shape rather than forcing a straight-line assumption, while a smooth term for calendar time soaks up long-term trends, seasonality and slow drifts in hospital utilization that could otherwise masquerade as weather signals.

The analysis also incorporated distributed lag structures spanning zero to three days, a technical choice with real biological rationale. If a weather front, pressure drop or temperature swing truly provoked seizures, the effect might not appear on the same day; barometric changes preceding storms, for instance, have been proposed to act with delays of one to three days. By modeling the cumulative influence of each variable across a four-day lag window, the study allowed delayed and carry-over effects to express themselves — and still found nothing. After adjustment for temporal trends, weekday effects and all meteorological variables simultaneously, every weather parameter returned p-values above the conventional 0.05 threshold.

One signal did emerge, but it was not from the sky. The smooth term for calendar time was highly significant, with a p-value below 0.001, indicating genuine underlying temporal variation in daily seizure counts that the evaluated meteorological variables failed to explain. In other words, seizure presentations were not distributed randomly across the four years, but whatever drove their ebb and flow was not temperature, pressure, humidity or wind as measured at a single urban station. The authors suggest that unmeasured environmental exposures, individual clinical factors such as medication adherence and sleep, or circannual biological rhythms may underlie this residual temporal pattern.

The findings land in a field where the evidence has long been contradictory. Some earlier studies, largely in adult epilepsy populations, reported modest associations between seizure frequency and low atmospheric pressure, thunderstorms, or seasonal temperature extremes; others found nothing once proper statistical controls were applied. In pediatric settings, the literature has been thinner and more contaminated by febrile seizures, which do cluster with weather in an indirect sense because respiratory infections peak in winter. By focusing strictly on afebrile events and applying lag-aware, trend-adjusted time-series methods, the new study addresses both weaknesses at once and comes down firmly on the side of no effect.

The methodological lesson may be the study’s most durable contribution. Time-series data on disease counts are riddled with autocorrelation, seasonality and slow secular trends, and simple month-by-month or season-by-season comparisons can easily generate false positives — a child admitted on a stormy day counts as an anecdote, and anecdotes accumulate into conviction. Proper models such as the Poisson GAMs used here explicitly separate within-day weather variation from between-season structure, and the authors argue that this rigor should be the minimum standard for any future claim that weather triggers seizures. Their null result, they note, is not evidence that meteorology is irrelevant to human health in general, but rather that in this pediatric cohort, at this temporal resolution, no independent weather signal survived scrutiny.

The authors are careful about scope. The study is retrospective and single-center, drawing on one hospital in Ankara with weather data from a single nearby station. Afebrile seizure etiologies in children are heterogeneous — ranging from structural epilepsy and genetic syndromes to metabolic disturbances — and individual susceptibility to environmental stressors could differ across these subgroups in ways a pooled daily count cannot detect. Medication changes, sleep deprivation, screen exposure and stress, all suspected seizure modulators, were not captured. The team calls for prospective, multicenter studies that incorporate additional environmental exposures and individual-level clinical data, which would allow stratified analyses that daily aggregate counts preclude.

For families of children with epilepsy, the practical message is one of reassurance rather than alarm. The study offers no support for keeping children indoors when the barometer falls or for attributing breakthrough seizures to the day’s weather. At the same time, the significant calendar-time signal is a genuine scientific puzzle: something real is modulating when children with seizures arrive at the hospital, and identifying it — whether it lies in infection patterns, behavioral rhythms, treatment dynamics or unmeasured atmospheric variables — is the clear next target for research. What the new work establishes is that the answer will not be found by simply reading a thermometer or a barometer.

The research received no specific funding, was approved by the University of Health Sciences Gülhane Scientific Research Ethics Committee, and is published open access under a Creative Commons Attribution 4.0 license. It was published on 5 September 2026 after peer review, with the accepted manuscript shared early to accelerate access to the findings.