Circadian variation in the onset of ischemic stroke. Histogram represents the number of total events occurring in each hour of the day. Superimposed is the overall best fitting curve calculated by rhythm analysis, resulting from 2 significant harmonics with 24- and 12-hour periods (the variables of the curve are given in Table 1 under "Total Population"). The horizontal line represents the MESOR (midline estimating statistic of rhythm), which represents the rhythm-adjusted mean over the period analyzed.
Casetta I, Granieri E, Fallica E, la Cecilia O, Paolino E, Manfredini R. Patient Demographic and Clinical Features and Circadian Variation in Onset of Ischemic Stroke. Arch Neurol. 2002;59(1):48-53. doi:10.1001/archneur.59.1.48
Studies have reported circadian variation in the onset of ischemic stroke, which may carry important pathophysiological implications. However, there is no detailed information about circadian variations among the subtypes of stroke.
To determine whether subgroups of patients with ischemic stroke with specific clinical characteristics would exhibit different circadian patterns, to more systematically examine the role of possible triggering or precipitating factors.
Design and Setting
Analysis of the effects of demographic, medical, and pathophysiological factors on the circadian pattern of an unselected series of patients with ischemic stroke consecutively admitted to our hospital.
The study included 1656 patients. As in other studies, the peak of stroke onset occurred in the morning, with a second peak in the evening. Circadian variation in ischemic stroke onset was shown to be independent of clinical variables considered.
Our study confirms the circadian rhythm of stroke reported in previous studies. There is a chronological pattern of ischemic stroke in the morning, which appears to be independent of the presence of risk factors and of clinical stroke subtypes. The role of circadian variability of blood pressure (present in patients with and without hypertension) and a concurrent morning hypercoagulability are suggested as possible determinants of this pattern. Preventive pharmacological interventions aimed at specifically targeting the morning rise in risk factors could be advantageous in reducing the overall risk of ischemic stroke.
DATA INDICATE that onset of several major cerebrovascular diseases are not randomly distributed over time.1- 10 The existence of a particular chronobiological pattern in the onset of acute cerebrovascular diseases, characterized by circadian, circaseptan, and circannual rhythms (1 day, 1 week, and 1 year, respectively), has been detected.10 A significantly higher occurrence in the morning has been reported1- 8 and confirmed by a recent meta-analysis.9 Although a well-defined pattern of ischemic stroke onset has been proved, few studies8,11 have addressed the possibility that different chronobiological patterns may be detected in different clinical subgroups of patients with ischemic stroke. The aim of the present study was to investigate whether such circadian variation in ischemic stroke onset could delineate subgroups of patients with different demographic or pathologic variables among a large population of patients with ischemic stroke.
Ferrara is a town in northeastern Italy, with a mean population of about 170 000. Its only hospital is St Anna Hospital, which is the sole teaching center for the school of medicine of the local university. St Anna Hospital also serves the entire province of Ferrara as a center where most patients with acute stroke are evaluated. House calls are performed by family physicians during the day and by emergency department physicians at night and on holidays, at no charge. In the emergency department, key physicians, including neurologists and neurosurgeons, are available 24 hours a day throughout the year. Between January 1, 1994, and December 31, 1997, a consecutive series of 1656 patients with ischemic stroke were recorded. The study area and methods of case collection have been described previously.12
The diagnosis of stroke was made by a neurologist and was defined, according to the World Health Organization criteria, as rapidly developing clinical symptoms or signs of focal or global loss of cerebral function, with symptoms lasting more than 24 hours or leading to death, with no apparent cause other than a vascular origin.13 In all patients, laboratory investigations included computed tomographic scan or magnetic resonance imaging, blood tests, 12-lead electrocardiogram, chest radiography, carotid duplex imaging, transcranial Doppler, cerebral angiography, echocardiography (transthoracic or transesophageal), and assessment of prothrombotic syndromes. Additional tests were conducted in selected patients.
Stroke onset time was defined as the earliest time the patient or a witness noted definite neurological symptoms or signs. It was obtained from patients, their relatives, or bystanders.
Precise determination of the time of symptom onset was possible in 1395 patients. In an additional 187 subjects—although stroke onset time could not be exactly determined because the stroke occurred while they were asleep, or they were unconscious, disoriented, or aphasic and a witness was not available to give reliable information—stroke onset could be assigned to 1 of 4 periods: 12:01 to 6 AM, 6:01 AM to noon, 12:01 to 6 PM, or 6:01 PM to midnight. For the remaining 74 patients, time of onset was unknown.
For each patient, we recorded demographic data; family history of vascular diseases, hypertension, and diabetes; medical history, with particular reference to hypertension, coronary artery diseases, atrial fibrillation, valvular and other heart diseases, previous transient ischemic attacks or strokes, asymptomatic carotid stenosis, bruit, diabetes, hyperlipidemia, peripheral arteriopathies, cigarette smoking, and alcohol consumption; and symptoms and signs at stroke onset and their evolution.
All patients underwent a physical and neurological examination. A quantitative evaluation using the Canadian Neurological Scale14 and a disability status determination according to the Rankin scale15 was assessed on admission. Cerebral infarction in patients was classified as 1 of 4 clinically identifiable subtypes16: total anterior circulation infarcts, partial anterior circulation infarcts, posterior circulation infarcts, and lacunar infarcts. Moreover, the cause of ischemic stroke was classified according to the criteria of the Trial of Org 10172 in Acute Stroke Treatment17 as large artery disease, small artery disease, cardioembolism, other less common determined causes, and undetermined causes, which included patients with multiple potential causes.
Diagnoses were based on clinical features and on results from the imaging and laboratory tests, following the methods of the Trial of Org 10172 in Acute Stroke Treatment investigators.17
For the patients whose strokes were precisely timed, the hour of each event was tabulated, rounding the time consistently to cover 24 hours, and the frequency of the events was computed for each hour of the day. With commercially available software,18 the analysis of circadian rhythm was performed using the cosinor method and a partial Fourier series with up to 4 harmonics (periods of 24, 12, 8, and 6 hours), in which a least squares minimization is used and a cosine function is fitted to the data via a regression method. Among all the possible combinations of the periods chosen, the program permits the selection of the harmonic or the combination of harmonics that best explains the variance of data. The percentage of rhythm (percentage of overall variability of data about the arithmetic mean attributable to the fitted rhythmic function) and the probability value resulting from the F statistic (used to test the hypothesis of zero amplitude) are reported in the results as representative factors of goodness of fit of the approximating curve function and statistical significance of rhythm, respectively. The best fitting curve indicates the period with the greatest percentage of rhythm. Along with the peak time of each single harmonic, the program also calculates peak and trough times (time of occurrence of the absolute maximum and minimum, respectively) of the overall best fitting curve.
Moreover, the χ2 goodness-of-fit test to the null hypothesis of equal distribution of strokes was applied to the 1582 patients whose onset could be reasonably included in one of the four 6-hour periods, using available software (Epi-Info, version 6; Centers for Disease Control and Prevention, Atlanta, Ga; version 6.04b; World Health Organization, Geneva, Switzerland). Differences were considered significant at P<.05.
Precise hours of ischemic stroke were recorded for 1395 patients (mean age ± SD, 74.6 ± 12 years). Forty-four percent of strokes occurred in the morning between 6:01 AM and noon, and the hypothesis for uniform distribution of the onset time was rejected on the basis of the χ2 test for all subtypes (χ23 = 311.77, P<.001). It has been suggested that a morning excess of strokes could be explained by the exclusion of patients with events occurring during the night, when time of onset could not be defined. Assuming that the stroke could have occurred at any time during sleep, we assigned these patients to a 6-hour interval between the time the patient was last known to be asymptomatic and the time at which patients or their relatives first became aware of the event. The null hypothesis of uniformity was still rejected when the 1582 patients whose onset could be categorized into the 6-hour periods were analyzed (χ23 = 255, P<.001). In addition, a worst-case scenario was considered in which the 74 patients whose strokes were untimed were arbitrarily assigned to the periods having the fewest observed strokes. There was still a significant circadian variation in the risk of stroke (1656 patients total, χ23 = 207, P<.001).
The results of cosinor analysis, including only patients with precisely timed strokes, are summarized in Table 1, in which the significant harmonics (24 and 12 hours) and the overall best fitting curve are reported. The sinusoidal test results showed a statistically significant circadian pattern, with a major peak at 8:28 AM. Spectral analysis also detected a significant 12-hour cycle at 8:13 AM and 8:13 PM.
Figure 1 demonstrates the overall best fitting curve resulting from 2 significant components of 24- and 12-hour periods, with a maximum occurrence at 8:28 AM, a second minor peak at 8:13 PM, and a minimum occurrence at 11:28 PM.
The data were similar for both sexes for patients aged 45 to 70 years. Younger patients (<45 years) and older patients (>85 years) did not show any significant circadian rhythms.
Hypertension, diabetes, hyperlipemia, smoking habits, previous vascular events, and treatment with antiplatelet agents or anticoagulant drugs did not modify the circadian pattern of ischemic stroke onset. The circadian variation in onset was independent of clinical characteristics of ischemic strokes (Table 2).
The morning increase and the second minor peak in the evening were detected in patients with atherothrombotic strokes, cardioembolic strokes, lacunar strokes, and strokes of other or unknown mechanisms. A similar pattern was detected in the clinically identifiable subtypes of ischemic stroke,16 except for the partial anterior circulation infarcts subtype (Table 2).
After random reallocation of untimed patients to the time frames having the fewest strokes, the analysis by clinical and demographic subgroups using the χ23 goodness-of-fit test applied to the 1656 patients yielded similar results (all, P<.001, except for therapy with anticoagulant agents): age 45 to 70 years, χ23 = 153; older than 70 years, χ23 = 142.5; normotension, χ23 = 116.5; hypertension, χ23 = 165; diabetes, χ23 = 88.3; dyslipidemia, χ23 = 68.5; no smoking, χ23 = 170.34; smoking, χ23 = 147.6; previous stroke, χ23 = 74.5; first-ever stroke, χ23 = 169; therapy with antiplatelets drugs, χ23 = 74; therapy with anticoagulant agents, χ23 = 8.4, P< .005; atherothrombotic stroke, χ23 = 45.71; cardioembolic stroke, χ23 = 27.79; small vessel disease, χ23 = 80; other determined cause, χ23 = 61.5; undetermined cause, χ23 = 69.6; total anterior circulation infarct, χ23 = 26; posterior circulation infarct, χ23 = 27.8; and lacunar infarct, χ23 = 63.5. Patients with partial anterior circulation infarcts subtype showed a significant increase in strokes from 6:01 AM to noon (χ23 = 9.1, P< .05). The null hypothesis of uniform distribution across the 4 periods was not rejected for patients younger than 45 years or older than 85 years.
Finally, a further evaluation considered patients with stroke onset while asleep, grouped by cause of stroke.17 Lacunar strokes are more likely to occur during sleep (25% of all lacunar strokes) than other types of strokes (χ24 goodness of fit = 36; P<.001).
This study of a large unselected population of patients with stroke confirmed the findings of other studies1- 11 that symptom onset is more frequent in the morning during the first few hours of diurnal activity, with a second peak in the evening and a minimum occurrence during the night. These data, added to those of a recently published meta-analysis,9 strengthen the assumption that a circadian timing of stroke does exist (χ23 for goodness of fit across all reports, including the present one = 1405.75, P<.001).
This circadian pattern is similar to that of acute myocardial infarction and myocardial ischemia, sudden cardiac death, and other vascular events.19- 27 Some underlying pathophysiological mechanisms may be common.10,21,28
However, because different times of day may reflect different pathophysiological mechanisms of stroke, subanalyses are important for establishing patterns. In further exploring the circadian variation in various patient subsets, we confirmed that each subgroup of patients with ischemic stroke, stratified according to risk factors, clinical variables, and putative cause of stroke, was identified with morning and evening peaks of stroke onset. The only significant common risk factor for these events was hypertension. However, in the present study, patients with and without hypertension had the same chronobiological pattern of stroke onset. This suggests that blood pressure, with its circadian variability, and not strictly hypertension, plays an important role in the circadian pattern of stroke onset.
The morning increase in stroke onset was attenuated only in patients younger than 45 and older than 85. One can speculate whether the difference is because of age or other differences between patients. Although we cannot exclude the possibility that some differences could not be detected because of small sample size, we can hypothesize that there may be different pathophysiological mechanisms in strokes between younger and older persons.
To our knowledge, this is the first study analyzing circadian pattern of stroke according to sex, stroke type, cause of stroke, age, presence of risk factors, and clinical characteristics. The circadian rhythm of stroke seems to be independent of other considered factors, except possibly younger and older age. This is in agreement with previous studies29,30 on acute myocardial infarction. Although a marked difference in diurnal patterns of myocardial infarction was initially reported in subgroups of smokers, β-blocker users, and patients with non–Q-wave infarction, diabetes, previous congestive heart failure, and previous myocardial infarction,29 further investigation found only minor differences in symptom onset, and multivariate analysis showed that only age older than 70 years and a history of previous myocardial infarction modified the circadian rhythm of symptom onset.30 However, antithrombotic drugs, such as aspirin,31 may modify the temporal pattern of myocardial infarction by attenuating the morning peak. Conversely, in our study, prior use of anticoagulant and antiplatelet agents did not affect morning occurrence of ischemic stroke.
In conclusion, our data confirm the existence of a chronological risk of stroke, although the circumstances surrounding the onset of stroke are not fully understood. In all subgroups of our patients, a statistically significant bimodal circadian variation was present and demonstrated that the circadian rhythm in cerebrovascular diseases is independent of stroke subtypes, patient demographics and clinical features, and presence or absence of risk factors.
A broad implication of our findings may be stroke prevention. Our results confirm that early morning is associated with a higher risk of the onset of stroke symptoms, irrespective of type of stroke. The circadian variability of blood pressure, resembling the temporal biphasic pattern of stroke, together with a concurrent morning prothrombotic condition,32- 34 may create a final negative synergistic effect.
A chronotherapeutic approach has been suggested for cardiovascular diseases.35 One may speculate that antihypertensive agents that target morning rise in blood pressure might be advantageous in controlling this risk factor for stroke. Long-term investigations are addressing this question.36
Accepted for publication August 28, 2001.
Author Contributions:Study concept and design (Drs Casetta, Granieri, Fallica, and Manfredini); acquisition of data (Drs Casetta, Granieri, Fallica, and la Cecilia, and Mr Paolino); analysis and interpretation of data (Drs Casetta, Granieri, Fallica, la Cecilia, and Manfredini); drafting of the manuscript (Drs Casetta, Fallica, la Cecilia, and Manfredini); critical revision of the manuscript for important intellectual content (Drs Casetta, Granieri, Fallica, and Manfredini and Mr Paolino); statistical expertise (Drs Casetta, la Cecilia, and Manfredini); obtained funding (Drs Casetta and Granieri); administrative, technical, and material support (Drs Casetta, Granieri, Fallica, and la Cecilia and Mr Paolino); study supervision (Drs Casetta, Granieri, and Manfredini).
This work was supported by a grant from the Italian Ministry of the University and Scientific and Technological Research (MURST 60%), Rome (Dr Granieri).
Corresponding author and reprints: Ilaria Casetta, MD, Section of Clinical Neurology, Dipartimento di Discipline Medico-Chirurgiche della Comunicazione e del Comportamento, University di Ferrara, Corso della Giovecca 203, I-44100 Ferrara, Italy (e-mail: email@example.com).