Submitted to the Senate of Tel Aviv University

Similar documents
THE FRAMINGHAM STUDY Protocol for data set vr_soe_2009_m_0522 CRITERIA FOR EVENTS. 1. Cardiovascular Disease

The Impact of Smoking on Acute Ischemic Stroke

Redgrave JN, Coutts SB, Schulz UG et al. Systematic review of associations between the presence of acute ischemic lesions on

Epidemiologic Methods I & II Epidem 201AB Winter & Spring 2002

Clinical Features and Subtypes of Ischemic Stroke Associated with Peripheral Arterial Disease

<INSERT COUNTRY/SITE NAME> All Stroke Events

Biostatistics and Epidemiology Step 1 Sample Questions Set 1

Stroke 101. Maine Cardiovascular Health Summit. Eileen Hawkins, RN, MSN, CNRN Pen Bay Stroke Program Coordinator November 7, 2013

Supplementary Online Content

ATTENTION-DEFICIT/HYPERACTIVITY DISORDER, PHYSICAL HEALTH, AND LIFESTYLE IN OLDER ADULTS

Antiplatelet Therapy in Primary CVD Prevention and Stable Coronary Artery Disease. Καρακώστας Γεώργιος Διευθυντής Καρδιολογικής Κλινικής, Γ.Ν.

Autonomic nervous system, inflammation and preclinical carotid atherosclerosis in depressed subjects with coronary risk factors

Cardiovascular Disorders Lecture 3 Coronar Artery Diseases

Transient Atrial Fibrillation and Risk of Stroke after Acute Myocardial Infarction

Modelling Reduction of Coronary Heart Disease Risk among people with Diabetes

Supplementary Online Content

Frequency of Cardiac Risk Factors in. Ischemic

Andrew Cohen, MD and Neil S. Skolnik, MD INTRODUCTION

Guidelines on cardiovascular risk assessment and management

AN EVALUATION OF CARDIOVASCULAR RISK IN EARLY BEREAVEMENT

A common clinical dilemma. Ischaemic stroke or TIA with atrial fibrillation MRI scan with blood-sensitive imaging shows cerebral microbleeds

Cardiovascular Diseases and Diabetes

Critical Review Form Therapy

Current Clinical Trials for Stroke Survivors in NJ and Philadelphia Areas

Lecture 8 Cardiovascular Health Lecture 8 1. Introduction 2. Cardiovascular Health 3. Stroke 4. Contributing Factors

DECLARATION OF CONFLICT OF INTEREST

Supplementary Appendix

ICD-10-CM - Session 2. Cardiovascular Conditions, Neoplasms and Diabetes

REPORT FROM THE CANADIAN CHRONIC DISEASE SURVEILLANCE SYSTEM:

KEEPING YOUR PATIENT OUT OF THE HOSPITAL BY PREVENTING A SECOND STROKE

Shawke A. Soueidan, MD. Riverside Neurology & Sleep Specialists

Biases in clinical research. Seungho Ryu, MD, PhD Kanguk Samsung Hospital, Sungkyunkwan University

Transient Ischemic Attacks and Risk of Stroke in an Elderly Poor Population

NORTH MISSISSIPPI MEDICAL CENTER MEDICAL CENTER. Stroke: Are you at risk? A guide to stroke risk factors & resources at ACUTE STROKE UNIT

CEREBRO VASCULAR ACCIDENTS

Clinical Studies 129

Alan Barber. Professor of Clinical Neurology University of Auckland

Stroke secondary prevention. Gill Cluckie Stroke Nurse Consultant St. George s Hospital

Appendix XV: OUTCOME ADJUDICATION GUIDELINES

VACCINE ACTIVE SURVEILLANCE I

3. Screening Subject Identification Screening Overview

Canadian Best Practice Recommendations for Stroke Care. (Updated 2008) Section # 3 Section # 3 Hyperacute Stroke Management

Antihypertensive Trial Design ALLHAT

Dr Julia Hopyan Stroke Neurologist Sunnybrook Health Sciences Centre

Dr Ben Turner. Consultant Neurologist and Honorary Senior Lecturer Barts and The London NHS Trust London Bridge Hospital

Nicolas Bianchi M.D. May 15th, 2012

Raluca Pavaloiu et al. - Clinical, Epidemiological and Etiopathogenic Study of Ischemic Stroke

Exclusion: MRI. Alcoholism. Method of Memory Research Unit, Department of Neurology (University of Helsinki) and. Exclusion: Severe aphasia

Comparability of patient-reported health status: multi-country analysis of EQ-5D responses in patients with type 2 diabetes

Repeat ischaemic heart disease audit of primary care patients ( ): Comparisons by age, sex and ethnic group

PTHP 7101 Research 1 Chapter Assignments

Vague Neurological Conditions

University of Wollongong. Research Online. Australian Health Services Research Institute

a. Ischemic stroke An acute focal infarction of the brain or retina (and does not include anterior ischemic optic neuropathy (AION)).

Table S1. Read and ICD 10 diagnosis codes for polymyalgia rheumatica and giant cell arteritis

Appendix 2C - Stroke Services in Fife

2003 World Health Organization (WHO) / International Society of Hypertension (ISH) Statement on Management of Hypertension.

Pre-Hospital Stroke Care: Bringing It To The Street. by Bob Atkins, NREMT-Paramedic AEMD EMS Director Bedford Regional Medical Center

Downloaded from:

INTERNAL VALIDITY, BIAS AND CONFOUNDING

inter.noise 2000 The 29th International Congress and Exhibition on Noise Control Engineering August 2000, Nice, FRANCE

Know Your Number Aggregate Report Comparison Analysis Between Baseline & Follow-up

BILATERAL BREAST CANCER INCIDENCE AND SURVIVAL

STROKE UPDATE ANTHEA PARRY MAY 2010

Stroke Prevention. For more information about stroke, call University Hospital s Heart Line at 706/ or toll free at 866/

Ischemic Stroke in Critically Ill Patients with Malignancy

Acute stroke. Ischaemic stroke. Characteristics. Temporal classification. Clinical features. Interpretation of Emergency Head CT

Coronary Heart Disease. Raja Nursing Instructor RN, DCHN, Post RN. BSc.N

Understanding Stroke

Baldness and Coronary Heart Disease Rates in Men from the Framingham Study

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE General practice Indicators for the NICE menu

ESM 1. Survey questionnaire sent to French GPs. Correct answers are in bold. Part 2: Clinical cases: (Good answer are in bold) Clinical Case 1:

Department of Anesthesiology. Clinical Base Year Neurosurgery Curriculum. Residency Program Director Department of Neurosurgery

Retrospective Study on the Safety and Efficacy of Clopidogrel in the Treatment of Acute Cerebral Infarction

Risk Factors for Ischemic Stroke: Electrocardiographic Findings

GALECTIN-3 PREDICTS LONG TERM CARDIOVASCULAR DEATH IN HIGH-RISK CORONARY ARTERY DISEASE PATIENTS

SUMMARY 8 CONCLUSIONS

JUSTUS WARREN TASK FORCE MEETING DECEMBER 05, 2012

Blood pressure and total cholesterol level are critical risks especially for hemorrhagic stroke in Akita, Japan.

Cardiovascular Disease


Antithrombotic therapy in patients with transient ischemic attack / stroke (acute phase <48h)

STROKE INTRODUCTION OBJECTIVES. When the student has finished this module, he/she will be able to:

Cerebrovascular Disorders. Blood, Brain, and Energy. Blood Supply to the Brain 2/14/11

Bias. A systematic error (caused by the investigator or the subjects) that causes an incorrect (overor under-) estimate of an association.

Cardiac Pathophysiology

10/8/2018. Lecture 9. Cardiovascular Health. Lecture Heart 2. Cardiovascular Health 3. Stroke 4. Contributing Factor

Supplementary webappendix

Alan Barber. Professor of Clinical Neurology University of Auckland

Intermediate Methods in Epidemiology Exercise No. 4 - Passive smoking and atherosclerosis

TRANSIENT ISCHEMIC ATTACK (TIA)

Lucia Cea Soriano 1, Saga Johansson 2, Bergur Stefansson 2 and Luis A García Rodríguez 1*

Hypertension The normal radial artery blood pressures in adults are: Systolic arterial pressure: 100 to 140 mmhg. Diastolic arterial pressure: 60 to

Branko N Huisa M.D. Assistant Professor of Neurology UNM Stroke Center

Controlling Bias & Confounding

Stroke is the third-leading cause of death and a major

Patient with Daily Headache NTERNATIONAL CLASSIFICATION HEADACHE DISORDERS. R. Allan Purdy, MD, FRCPC,FACP. Professor of Medicine (Neurology)

Statistical Fact Sheet Populations

/ / / / / / Hospital Abstraction: Stroke/TIA. Participant ID: Hospital Code: Multi-Ethnic Study of Atherosclerosis

Canadian Stroke Best Practices Table 3.3A Screening and Assessment Tools for Acute Stroke

Transcription:

TEL AVIV UNIVERSITY SACKLER SCHOOL OF MEDICINE DEPARTMENT OF EPIDEMIOLOGY AND PREVENTIVE MEDICINE TRIGGERING RISK FACTORS FOR ISCHEMIC STROKE A CASE-CROSSOVER STUDY Thesis Submitted for the Degree Doctor of Philosophy by SILVIA KOTON Submitted to the Senate of Tel Aviv University April 27, 2003

II This work was carried out under the supervision of Professor Manfred S. Green

III Acknowledgments To Professor Manfred S. Green for his valuable guidance, encouragement, confidence and for initially stimulating my interest in this area To Prof. Nathan Bornstein and Dr. David Tanne for their remarkable cooperation, for the continual assistance and support throughout this work To the medical, nursing and paramedical teams in the neurology and medical wards at the Tel Aviv Sourasky Medical Center and the Chaim Sheba Medical Center for making me feel a member of the team To Dr. Nira Koren-Morag and Mr. Gil Harari for the statistical counselling Special thanks to my husband, Avihu, for his moral support, encouragement and help, and to my children Elad, Omer and Alon for their understanding and constant encouragement.

IV Contents Page number Figures list Tables list Abstract 1.0 Background 1-8 1.1 General 1 1.2 Conventional risk factors for stroke 1-2 1.3 Prevalence of stroke risk factors among Israeli stroke patients 3-4 1.4 Seasonal and circadian variation in stroke incidence 4 1.5 Stroke triggering factors 4-5 1.6 Comparison with triggers for myocardial infarction 5-7 1.7 Potential mechanism underlying triggering of stroke 7 1.8 The case-crossover study design 8 2.0 Potential contribution of the study 8-9 3.0 Research objectives 9-10 4.0 Research hypotheses 10 5.0 Methods 11-22 5.1 Study design 11 5.2 Target population 11 5.3 Study sample 11-12

V Page number 5.4 Study variables 12-15 5.5 Data collection forms 15-16 5.6 Details of procedures and sources of data 16-17 5.7 Potential biases 17-20 5.8 Statistical analyses 20-22 6.0 The pilot study 22-40 7.0 Results 41-74 7.1 Distribution of identified stroke patients according to inclusion status 41-43 7.2 The study sample 43-50 7.3 Exposure to potential triggering factors using the preceding day as control period 51-67 7.4 Exposure to potential triggering factors using regular exposures as control period 68-70 7.5 Exposure to fever as a potential triggering factor 71 7.6 Meteorological data 71-74 8.0 Discussion 75-92 9.0 Conclusions 92-93 10.0 References 94-103 Appendixes 1. Patient s questionnaire (Hebrew) 2. Medical files form (Hebrew) 3. TOAST classification definitions 4. Modified Rankin Scale (mrs)

VI Figures list Page number Figure 1: Distribution of patients by sex and age groups, pilot study 34 Figure 2: Distribution of patients by place of birth, pilot study 35 Figure 3: Distribution of patients by education, pilot study 35 Figure 4: Distribution of stroke patients according to inclusion status 41 Figure 5: Distribution of age by sex, n=200 45 Figure 6: Distribution of study subjects by sex and level of education, n=20 46 Figure 7: Distribution of study subjects by place of birth, Jews only, n=200 47 Figure 8: Distribution of the Jewish population aged 45 years and over, by place of birth, Israel 2000. 47 Figure 9: Distribution of study subjects by type of event, n=200 48

VII Page number Figure 10: Distribution of stroke cases by TOAST classification (TIA not included), n=184 48 Figure 11: Percent of patients exposed to potential triggers during the two hours before the stroke without similar exposures during the same two hours the preceding day, n=200 52 Figure 12: Percent of stroke patients exposed to at least one potential trigger1, by TOAST classification, n=68 57 Figure 13: Distribution of stroke patients by TOAST classification, n=184 58 Figure 14: Odds Ratio for patients exposed to potential triggers during the two hours before the stroke compared to annual exposures 69

VIII Tables list Page number Table 1: Test-retest agreement coefficients for exposure to potential triggers during the day of the stroke 25 Table 2: Test-retest agreement coefficients for exposure to potential triggers during the day previous to the stroke 26 Table 3:Test-retest agreement coefficients for usual exposure to potential triggers during the last year 27 Table 4: Test-retest agreement coefficients for presence of known risk factors 28 Table 5: Inter-observer agreement coefficients for exposure to potential triggers during the day of the stroke 29-30 Table 6: Inter-observer agreement coefficients for exposure to potential triggers during the day previous to the stroke 31 Table 7: Inter-observer agreement coefficients for usual exposure to potential triggers during the last year 32 Table 8: Inter-observer agreement coefficients for presence of known risk factors 33 Table 9: Number and percent of patients reporting risk factors, pilot study 36

IX Page number Table 10: Stroke type according to TOAST classification, pilot study 37 Table 11: Reported signs and symptoms, pilot study 38 Table 12: Descriptive variables of the sample and excluded stroke patients, n=405 43 Table 13: Study sample characteristics, n=200 45 Table 14: Distribution of sample and Israel population by place of birth and age group (percent), Jews only 47 Table 15: Prevalence of risk factors for stroke according to medical files, absolute number and percentage, N=200 49 Table 16: Duration of risk factors, patients reporting risk factors only 50 Table 17: Medications as reported in the patients medical files, absolute number and percentage, N=200 50 Table 18: Odds Ratio for patients exposed to potential triggers during the two hours before the stroke compared to the same two hours the preceding day 53

X Page number Table 19: Odds Ratio for patients exposed to at least one potential trigger during the two hours before the stroke compared to the same two hours the preceding day, analysis by sex, age group and stroke type 54-55 Table 20: Odds Ratio for patients exposed to at least one potential trigger during the two hours before the stroke compared to the same two hours the preceding day, analysis by selected health characteristics 56 Table 21: Odds Ratio for patients exposed to at least one potential trigger during the day of the stroke compared to the preceding day, by exposure time 60 Table 22: Odds Ratio for patients exposed to sudden posture change as response to a startling event immediately before stroke onset during the day of the stroke compared to the preceding day 61 Table 23: Odds Ratio for patients exposed to negative emotions during the day of the stroke compared to the preceding day, by exposure time 62

XI Page number Table 24: Odds Ratio for patients exposed to anger during the day of the stroke compared to the preceding day, by exposure time 63 Table 25: Odds Ratio for patients exposed to sudden temperature changes during the day of the stroke compared to the preceding day, by exposure time 64 Table 26: Odds Ratio for patients exposed to positive emotions during the day of the stroke compared to the preceding day, by exposure time 65 Table 27: Odds Ratio for patients exposed to heavy eating during the day of the stroke compared to the preceding day, by exposure time 66 Table 28: Odds Ratio for patients exposed to heavy physical exertion during the day of the stroke compared to the preceding day, by exposure time 67 Table 29: Odds Ratio for patients exposed to potential triggers during the two hours before the stroke compared to regular exposures during the last year 68-69 Table 30: Odds Ratio for patients with fever at admission compared to patients with fever one week before 71

XII Page number Table 31: Daily average temperature, barometric pressure and humidity characteristics, data from the meteorological service 72 Table 32: Odds Ratio for patients exposed to at least one potential trigger during the two hours before the stroke compared to the same two hours the preceding day, analysis by meteorological conditions. 73-74

XIII Abstract Background: Cerebrovascular disease is the commonest and most important neurological disease of adult life, causing long-term severe disability, and among the most common causes of death in western countries. Several conditions and life-style factors have been identified as risk factors for stroke, of which hypertension has been recognized as the most important modifiable risk factor for both cerebral infarction and intracerebral hemorrhage. There is emerging interest in the role of psychosocial factors such as anger trait, expressed anger and long-term psychological distress, in the pathogenesis of stroke. Conventional risk factors and the newer psychosocial risk factors only partially explain the individual risk of stroke and the variance between populations, and do not explain why stroke occurs at a particular time. Circadian and seasonal variations in stroke onset and the association between acute infections and stroke incidence, suggest the existence of triggering mechanisms. However, there are very few reported studies on triggering risk factors for stroke. The role of preceding infection as a risk factor for ischemic stroke was investigated in several case-control studies and results from a pilot study on the potential stroke triggering effect of routine activities have been recently reported. To the best of our knowledge, this is the first comprehensive case-crossover study on stroke triggering factors and the first study to investigate the potential stroke triggering effect of sudden

XIV changes in posture, positive and negative emotional stress, and sudden changes in environmental temperature. Aim: To determine whether exposure to sudden changes in body posture in response to a startling event, emotional stress, anger, sudden physical effort, heavy eating, sudden changes in environmental temperature, and high body temperature can trigger the onset of stroke. Methods: The research was designed as a case-crossover study, an epidemiological method that uses each patient as his own control and especially developed to assess the effect of potential triggers on the incidence of acute events. A total of 684 consecutive stroke patients were interviewed. Of these, 200 were eligible for the study. Patients were interviewed 1-4 days after the event using a validated questionnaire especially designed for this purpose. Additional data were collected from patients files. Patients cognitive state was assessed using the Mini Mental State test. Patients scoring 26 or less were not interviewed. Additional exclusion criteria were aphasia and dementia. Exposures during a two-hour hazard period prior to onset of the stroke were compared to two types of control data: exposures in the same two-hour period the day before the ischemic event and the usual level of exposure during the previous year. Clinical and laboratory data were collected and stroke type determined based on the TOAST criteria by a stroke neurologist, blinded to potentially triggering variables. Statistical analyses for

XV matched case-control studies adapted to the case-crossover design were performed. The term statistically significant in this report implies a significance level of 0.05. Results: Of the eligible patients, 49.4% were included in the study sample. The remaining 50.6% could not be included for reasons not related to the study variables. There is no statistically significant difference between the sample and eligible excluded patients in the distribution of age, sex, address and hospital. Seventy-six patients (38%) reported exposure to at least one potential triggering factor during the two hours hazard period before onset of stroke symptoms, compared to nine patients exposed during the control period the day before. The odds ratio for patients exposed to at least one potential triggering factor during the two hours hazard period was 8.4 (95% CI 4.5-18.1). The most common reported potential triggers were negative emotions and sudden changes in body posture in response to a startling event. Negative emotions in the day of the stroke were reported by twentynine patients (14.5%), compared to two patients the day before (OR=14, 95%CI 4.4-89.7). Sudden changes in body posture in response to a startling event were reported by twenty-four patients (12%), compared to one patient the day before (OR=24, 95%CI 5.1-428.9). The highest OR was calculated for sudden changes in body posture in response to a startling event. ORs significantly higher than 1 were obtained also for exposures to negative emotions and anger.

XVI The Odds Ratio for patients exposed to at least one potential trigger during the two hours before the stroke compared to their annual exposures was 15.1 (95% CI 11-20.1). The highest OR was found for negative emotions, followed by heavy eating (OR=35.6 and OR=21.8 respectively). Fever (38 0 C and over) at admission was compared to reports of fever during the same day of the week preceding the stroke. Two patients had fever at admission (self reported confirmed by PO measure at admission), one patient reported having fever one week before the stroke. There were no reports of fever in both periods. The estimated odds ratio was 2 (95% CI 0.2-43.0). Information on daily average temperature, barometric pressure and humidity was obtained from the meteorological service for each of the stroke incidence days. In the assessment of the influence of meteorological variables on the potential triggering effect of the various triggers on stroke incidence, no significant relation between temperature, humidity and barometric pressure and triggering effects was found. Conclusions: Potential stroke triggering factors are present in almost 40% of the patients. Sudden changes in body posture in response to a startling event, negative emotions and anger appear to be independent triggering risk factors for stroke. The sample does not include deaths and patients that did not survive the first 24 hours after the stroke, which may represent a more serious type of stroke. Therefore it may not be possible to extrapolate the study findings to that sub-population of patients. We assessed the possible influence

XVII of selection and misclassification biases. There was no evidence to suggest that these biases had any substantial effect on the findings. Confounding bias was mostly controlled by the study design. Changes in medications, smoking habits and diet were assessed by stratification in the statistical analysis using the last year exposures as control period. There was no evidence of any confounding due to changes in medical treatment or changes in diet habits during the one-year control period. Due to the small number of patients reporting changes in smoking habits, their potential confounding effect is likely to be small. The findings contribute to understand why patients experience a stroke suddenly at a certain moment, through the recognition of potentially new immediate risk factors for ischemic stroke that may interact with and complement conventional risk factors. The study results could be an important component of stroke prevention strategies. Preventive measures that could be developed as a result of this study include giving the public adequate information about the various potential triggers for stroke and changing lifestyle, which is particularly important for elderly people. An example of this kind of life-style change may be avoiding sudden changes in body posture in elderly people. Future studies should be directed at investigating the triggering mechanism of stroke, identifying factors which may increase susceptibility to triggers and investigate methods for protecting individuals at high risk for triggered stroke incidence. Keywords: ischemic stroke, transient ischemic attack, triggers, risk factors, epidemiology, case-crossover study, Israel

1 1.0 Background 1.1 General Cerebrovascular disease is the commonest and most important neurologic disease of adult life in western countries, causing long-term severe disability. Cerebral infarctions account for 80% of stroke cases (1-5). In Israel, as in the U.S., stroke is the third most common cause of death. In 1997, a total of 2,905 deaths (8% of total deaths) were caused by stroke, 1,390 of them among men and 1,515 among women, leading to crude mortalty rates of 48.3/100,000 and 51.7/100,000 respectively (6). Agespecific mortality rates increase with age, from about 20/100,000 for ages 45-49 to almost 1,000/100,000 for ages 75 and over for both sexes (7). In 1993, 13,889 hospitalization cases were due to stroke, about half in each sex (8). In several European countries, stroke hospital-related cost has been reported to rank first of all diseases (9). Due to the high long term human and financial costs, efforts should be focused on stroke prevention. 1.2 Conventional risk factors for stroke Several conditions and life-style factors have been identified as risk factors for stroke. According to a recent scientific statement published by the American Heart Association, nonmodifiable risk factors include age, sex ethnicity and family history of stroke while modifiable risk factors include hypertension, smoking, diabetes, asymptomatic carotid stenosis, atrial fibrillation and other cardiac diseases which contribute to the risk of

2 thromboembolic stroke, sickle cell disease, and hyperlipidemia. Additional conditions, among them obesity, physical inactivity, poor diet, hyperhomocysteinemia and alcohol abuse, are reconized as less well documented or potentially modifiable risk factors (9). The most important modifiable risk factor for both cerebral infarction and intracerebral hemorrhage is hypertension (9-11). There is a direct, continuous and independent relationship between stroke incidence and both systolic and diastolic blood pressure (9). Control of hypertension decreases the incidence of fatal and non-fatal strokes (12). Hypertension also interacts with diabetes to increase the risk of stroke (13). In a prospective cohort study recently published, increasing BMI was positively associated with an increase in the risk of stroke in men after adjustment for hypertension and diabetes. This study suggests that preventing obesity may reduce risks of stroke (14). There is emerging interest in the role of psychosocial factors in the pathogenesis of stroke. The relationship between psychosocial factors, such as anger trait, expressed anger and long-term psychological distress, and stroke incidence has recently been reported (15-17) and Israeli researchers have reported that psychosocial factors predict long term stroke mortality among men in Israel (18). Conventional and newer psychosocial risk factors only partially explain the individual risk of stroke and the variance in stroke incidence among different populations.

3 1.3 Prevalence of stroke risk factors among Israeli stroke patients A study based on a prospective hospital-based registry in Tel-Aviv found that past medical history of hypertension was the major risk factor, occurring in 52.2% of patients and 39.5% of controls. Other risk factors were identified: ischemic heart disease (29.7% of cases and 13.7% of controls), diabetes mellitus (25.2% of cases and 14.9% of controls), smoking (17.0% of cases and 26.5% of controls), and hyperlipidemia (8.2% of cases and 27.5% of controls). Atrial fibrillation was reported by 14.3% of stroke patients. The prevalence of atrial fibrillation among controls was not assessed (5). A nation-wide case-series study of first stroke in 17-49 years old patients (1992-1993) found that the prevalence of potential risk factors was as follows: smoking - 53.6%, hypertension - 43.4%, hyperlipidemia - 22% and diabetes mellitus - 21%. Among controls from the CINDI study, smoking and hypertension rates were lower (19). A 21-year follow-up prospective study of middle-age and elderly men from a healthy working population in Israel (starting in 1965) showed an independent negative association between HDL-C and ischemic stroke mortality (20). Further analysis from the same cohort concluded that there are ethnic variations in stroke mortality in Israel, beyond those expected by differences in conventional risk factors (21). Among patients with documented coronary heart disease, an increased risk for ischemic stroke/tia incidence was associated with high triglycerides and low HDL

4 cholesterol. The increased risk associated with high triglycerides was found across subgroups of age, sex, patient characteristics and cholesterol fractions (22). 1.4 Seasonal and circadian variation in stroke incidence Stroke shows a circadian variation in time of onset. In a meta-analysis of 31 studies, an average increase of 49% in all types of stroke was found between 6 AM and noon compared with the number expected if no circadian variation was present. Between midnight and 6 AM, there were 29% fewer strokes than expected (23). Increased stroke mortality during winter has been reported by several studies (24-26). The mean ratio of winter to summer mortality from stroke in Israel during 1976-85 is about 1.5 for both sexes in almost all age groups, with very narrow confidence intervals (26). Circadian and seasonal variations in stroke incidence may be related, in part, to the effect of triggering risk factors. 1.5 Stroke triggering factors Triggering effect of personal and environmental factors may explain the circadian and seasonal variations in stroke incidence. However, there are very few reported studies on triggering risk factors for stroke. The role of preceding infection as a risk factor for ischemic stroke was investigated in several case-control studies (27-30). Those studies concluded that infectious syndromes might be associated with an increased risk of stroke.

5 The increased risk seems to be associated primarily to bacterial infections (29), and confined within a temporal window of less than one week (30). The triggering effect of infections was studied in Israel by Bornstein et al (31,32) in a prospective study including 182 consecutive patients hospitalized during August-October 1994. Controls were patients who had undergone a stroke and were followed up at the stroke out-patient clinic. Infections were present in 24.2% of cases in comparison to 9.7% of controls (p<0.0002). Among controls, infections were evenly spread over time, whereas among cases they concentrated during the week prior to the stroke (p<0.0001). The suggested possible mechanism includes cytokines, which are released during many bacterial infections, and other acute phase reactants like fibrinogen, which is known to be a risk factor for stroke (32). 1.6 Comparison with triggering factors for myocardial infarction Various triggering factors have been shown to increase the risk of myocardial infarction (MI) incidence. Myocardial infarction and sudden cardiac death show a marked circadian variation with increased risk during the morning, after awakening and arising (33). Extreme emotional stress superimposed on the stress of awakening was shown to enhance triggering of MI (34). Physical activity, emotional stress, and assumption of upright posture increase serum cathecholamines. A study on precipitating factors of MI

6 found that 40% of 186 patients had at least one of these factors immediately before the onset of symptoms, the most frequent one being a sudden change in position (35). Cathecholamine concentrations and platelet aggregability increase with assumption of the upright posture. This may explain the increased MI incidence between 6 AM and noon (36,37). Platelet activation and secretion occur also in association with emotional stress (38). Physical exertion was found to trigger acute myocardial infarctions. In The Determinants of Myocardial Infarction Onset Study (1989-1992) (39) heavy physical exertion was associated with a transient risk of myocardial infarction during the subsequent hour 5.9 times higher than the risk during periods of lighter or no exertion. Increasing levels of habitual physical activity were associated with progressively lower relative risks. A total of 4.4% of the patients reported heavy physical exertion in the hour before the onset of infarction symptoms (39). In a population-based case-control study in Germany (1989-1991), the relative risk for heavy physical exertion was 2.1. The relative risk was lower for people that exercised regularly (40). Mittleman et al (39) showed that the induction time for myocardial infarction onset following exposure to heavy physical exertion is less than one hour. A proposed mechanism for the triggering of myocardial infarction is the disruption of a vulnerable atherosclerotic plaque in response to hemodynamic stress (39). Sudden changes in temperature, emotional

7 stress, anger and sudden change of position may cause similar hemodynamic stresses caused mainly by changing platelets aggregability. In an Israeli study on exposure to work-related triggers and incidence of an acute coronary syndrome, 24% of the patients reported being exposed to at least one potential trigger during the hour preceding the coronary event (41). 1.7 Potential mechanisms underlying triggering of stroke Potential mechanisms for triggering stroke can be postulated. They may include in-part sympathetic nervous system hyperactivity; hemodynamic effects; deleterious endothelial effects - e.g brief episodes of mental stress, similar to those encountered in everyday life, may cause transient endothelial dysfunction (42), and induction of coagulation abnormalities by acute stress. Underlying mechanisms were suggested for triggering of acute myocardial ischemia, including endothelial dysfunction and injury induced by acute stress and hyperresponsivity of the sympathetic nervous system, manifested by exaggerated heart rate and blood pressure, in response to psychological stimuli (43), but the triggering mechanisms in the pathogenesis of stroke may differ from those underlying triggering of acute myocardial ischemia. Taking in account a possible delay in the effect of triggers on stroke compared to acute MI, a two-hour period of exposure to potential triggers for stroke is considered as hazardous in the present study.

8 1.8 The case-crossover study design The case-crossover design was introduced in 1991 as a new epidemiological technique, especially developed for studying the effect of transient exposures on the risk of onset of acute events (44). Control information for each patient is based on his/her past exposure experience. The case-crossover design is the scientific way to ask Were you doing anything unusual just before the event? and, in fact, the only possible way to achieve the required data to answer this question. Case crossover studies were conducted to analyze the association between cellular telephone calls and the risk of motor vehicle collision (45) and to assess the influence of exposures to triggering factors on the incidence of acute coronary syndromes (39-41,44,46,47). Concerns about external generalizability may be formulated since case-crossover studies, similarly to case-control studies based on patients interviews, cannot include patients not surviving long enough or too impaired to be interviewed. The casecrossover design is the most adequate for the study of the potential influence of exposures to triggering factors on the incidence of ischemic stroke. The possibility of selection bias does not eliminate the utility of the case-crossover design but possible biases should be addressed. 2.0 The potential contribution of the study Stroke patients reporting onset of symptoms immediately after experiencing an overwhelming emotion are not rare. During the preliminary phase of this

9 study, e.g., we interviewed a relatively healthy man, that stroked immediately after being told that his son has died unexpectedly. This illustrative case emphasizes the importance of studying the triggering influence of specific factors by formal rigorous and novel methodology, designed specifically to assess the effect of short exposures on the incidence of acute events. This study contributes to the understanding of the association between exposure to sudden physical effort, infections, sudden changes in temperature, emotional stress, anger and sudden assumption of the upright posture during defined hazard periods and incidence of stroke. Analysis of the influence of triggers on stroke risk according to stroke type is important in any attempt to understand the mechanisms operative specifically in triggering stroke. This study may contribute to better understanding of the mechanism operating in occurrence of acute stroke through the recognition of potentially new immediate risk factors for ischemic stroke that interact with and complement conventional risk factors. Its findings could be an important consideration in stroke preventive strategies. 3.0 Research objectives 3.1 General goal of the research To determine the influence of exposure to potential triggering factors on the risk of cerebral ischemic events incidence.

10 3.2 Specific objectives A. To determine whether the risk of cerebral ischemic events is increased by exposure to sudden changes in body posture in response to a startling event, emotional stress, anger, sudden physical effort, heavy eating or sudden changes in environmental temperature during a two-hour hazard period immediately prior to the event onset. B. To determine whether the risk of cerebral ischemic events is increased by exposure to high body temperature (a sign of acute infections) during the 24 hours prior to the event onset. 4.0 Research hypotheses A. The risk of ischemic events is increased by exposure to sudden changes in body posture in response to a startling event, emotional stress, anger, sudden physical effort, heavy eating or sudden changes in environmental temperature during a two-hour hazard period immediately prior to the event onset. B. The risk of ischemic events is increased by exposure to high body temperature (a sign of acute infections) during the 24 hours prior to the event onset.

11 5.0 Methods 5. 1 Study design The research was designed as a case-crossover study. Exposure of each patient to sudden physical effort, high body temperature, sudden changes in environmental temperature, emotional stress, anger and sudden change in posture during defined hazard periods were compared with previous exposures. Objective data on exposure to environmental temperature, barometric pressure and relative humidity were also collected. 5. 2 Target population Ischemic events patients hospitalized in the central area of Israel. 5. 3 Study sample A total of 200 patients with cerebral infarctions (stroke and TIA) admitted to two hospitals were interviewed. The sample size was determined using the n-query software based on the following table. Exposure in controls Difference in exposure Type 1 error Power Sample size 5% 10% 5% 80% 105 10% 10% 5% 80% 148 10% 10% 5% 90% 193 5% 5% 5% 90% 388 10% 5% 5% 80% 459

12 According to the table, 193 cases will be needed to detect a difference in exposure of at least 10% between controls and cases with a 5% probability of type 1 error and a 90% study power. 5. 4 Study variables 5.4.1 Cases Ischemic stroke and TIA cases clinically diagnosed by experienced neurologists. The diagnostic classification of stroke used here is similar to that used in the WHO Stroke Registers and in the MONICA Study. Definition of stroke _from the MONICA Manual, 1990 (48): Stroke is defined as rapidly developed clinical signs of focal disturbance of cerebral function lasting more than 24 hours (unless interrupted by surgery or death), with no apparent cause other than a vascular origin. It includes patients presenting clinical signs and symptoms suggestive of subarachnoid haemorrhage, intracerebral haemorrhage or cerebral ischemic necrosis. It does not include transient cerebral ischemia or stroke events in cases of blood diseases (e.g. leukemia, polycythemia vera), brain tumor or brain metastases. Secondary stroke caused by trauma is also excluded. Definition of TIA: A TIA is defined as an acute loss of focal cerebral or ocular function with symptoms lasting <24 hours and that, after adequate investigation, was

13 presumed to be due to embolic or thrombotic vascular disease (49). Patients with isolated vertigo, diplopia, bilateral blindness or drop attacks were excluded since such symptoms may result from either diffuse cerebral ischemia or from non-vascular pathologies (50). Classification of TIAs (50,51): 1. Carotid distribution TIA: Unilateral motor or sensory symptoms or Dysphasia or Transient monocular blindness 2. Vertebrobasilar distribution TIA: Vertigo or Diplopia or Hemianopsia or Bilateral motor or sensory symptoms Dysarthria was not used to classify the vascular distribution of TIA. Study inclusion criteria: Ischemic stroke and TIA cases diagnosed by a neurologist according to the previously specified definitions. Aphasic patients and patients scoring 26 or less on the Mini Mental State (MMS) test were included only in cases in which interview of a family member or caregiver that lives with the patient was possible. The study included recurrent stroke patients with no previous significant disability (mrs<=2), reporting 1-2 previous ischemic events, the last one 6 months or more before the present event.

14 Exclusion criteria: Stroke events in cases of blood diseases (e.g. leukemia, polycythemia vera), brain tumor or brain metastases were not be included. Secondary post-traumatic stroke and haemorrhagic stroke was excluded. Cases with developing stroke symptoms were excluded because the hazard exposure period cannot be determined in those cases. Patients that were asleep during the hazard period and waked up with symptoms were excluded. Definition of event onset: The time (hour and minutes) of appearance of the first sign. 5.4.2 Self-matched controls The same cases, evaluated for exposure the day before the ischemic event (in accordance to the case-crossover design). 5.4.3 Exposure variables Patients were asked to report exposure to heavy physical exertion, high body temperature, emotional stress, anger, sudden environmental cold or heat, and sudden changes in posture during the two-hour hazard period before the ischemic event, during the same control period the day before the ischemic event and usual exposures during the previous year. Physical exertion was categorized according to the scale published by Mittleman et al, 1993 (39). Emotional stress and anger were measured using the PANAS scale (52) and the onset anger scale (47) respectively. An Hebrew

15 previously validated version of the PANAS (53) was used in the present study. The reported split-half reliability coefficient (Spearman-Brown test) used to assess the internal consistency of the Hebrew scales is α=0.76 for the positive affects scale and α=0.94 for the negative affects scale. Testretest correlation coefficients for the various study periods were r=0.78-0.88 for the positive scale and r=0.62-0.82 for the negative scale. The Hebrew version of the PANAS was found to have predictive validity for acute myocardial infarction (41). Data on environmental temperature, barometric pressure and humidity were collected from the Israel Meteorological Service. 5.4.4 Potential confounders The case-crossover design allows control for most potential confounders. Since there is almost complete matching (case serves as control to him/herself), confounding by constant subject characteristics is eliminated. Personal changes in smoking behavior, diet and medications during the study period were controlled for in the statistical analyses. 5. 5 Data collection forms Two data forms were especially formulated: the patient s questionnaire and the medical data sheet. The last one was filled for every patient according to the patient s hospitalization file and included data on the

16 patient s health status, health risk factors, medical history, medications, laboratory tests results, and diagnostic procedures. The patient s questionnaire included demographic data, self-reported risk factors, medical history, medication, stroke signs and symptoms, time of onset, data on exposure to potential triggering factors during 26 hours before the event and frequency of exporure to those factors during the last year. The questionnaire included Hebrew versions of recognized scales: Mittleman s Scale for Physical Exertion (39), the Anger Onset Scale (47) and the Positive Affect and Negative Affect Scales PANAS (52). The scales were validated in the target population. First, they were translated into Hebrew and translated back to English by two different English teachers. The PANAS items were compared to those included in the previously reported Hebrew version (41) and no significant difference was found between the versions. The formulated Hebrew versions of the scales were then included in the patient s questionnaire and validated in the pilot study as reported in section 6- The Pilot Study. 5. 6 Details of procedures and sources of data Patients were interviewed 1-4 days after the event. Additional data were collected from patients files. An effort was done to interview a family member or a caregiver that lives with the patient in order to complete missing data as needed. Patients cognitive state was assessed using the Mini-Mental-State (MMS) test. The study questionnaire included questions

17 about demographic variables, risk factors, health status, stroke type and category, exposures during the hazard period and baseline exposures. In order to try and understand the mechanisms operative specifically in triggering stroke, it is important to study whether the potential triggering effect differs according to stroke type. Therefore, results of ancillary testing were collected and stroke type determined based on the TOAST criteria (54), by a stroke neurologist, blinded to triggering effects. Exposures during a two-hour hazard period were compared to two types of control data: exposure level in the same two-hour period the day before the ischemic event and the usual level of exposure during the previous year. This study design explains why data about exposures during the 26 hours before the stroke were collected. To prevent bias due to reported overestimation of exposure immediately prior to the stroke, the whole 26- hour period was treated as a long hazard period in the interview. Exposure to changes in environmental temperature, barometric pressure and humidity were assessed based on data collected from the Israel Meteorological Service. 5. 7 Potential biases The goal of the present study is to determine the influence of exposure to selected potential triggers on the risk of ischemic stroke and TIA incidence. The case-crossover design, which allows for assessment of exposures to familiar but startling activities and emotions, is the most

18 adequate to obtain the data required to address the study objectives. The case-crossover design has advantages over other designs. Allowing each patient to serve as his own control eliminates control-selection bias, one of the main causes of bias in case-control studies. The case-crossover design is also immune to any potential confounding by constant subject characteristics. On the other hand, some concerns related to biases and results interpretation might be formulated. The potential biases in the present study will be discussed. Selection bias About 20% of stroke patients remain aphasic during the week after the event (Bornstein, personal communication). These patients were not included in the sample unless it was possible to interview a family member or a caregiver. Since there is no reason to suspect that the triggering mechanism or the exposures differ, excluding patients for reasons not related to the study variables should not have any effect on the results. Possible differences between the sample and elegible excluded patients in the distribution of age, sex, address and hospital were assessed. Misclassification bias A. Information bias: Since the main task of a case-crossover study is to compare exposures in the hazard period to those in the control period, an effort was made to accurately measure reported exposures. Especial

19 efforts were done during the questionnaire formulation to assure unbiased collection of data. The high validity and reliability of the questionnaire minimizes potential bias in information. In the assessment of exposure to sudden changes in temperature, objective information on temperature, barometric pressure and humidity was collected in addition to the selfreported data. This information may be considered an indicator of the validity of subjective self-reported data. Differential misclassification bias may lead to overestimation of the risk associated with exposure to triggers. The proposed study design, that treats the 26-hour period as a long hazard period, minimizes the possibility of overestimation of exposures during the hazard period and underestimation of them during control periods. Patients were not aware of the hazard period being studied. Since most of the patients are over the working age (65 years), there is no reason to suspect over-reporting of exposures in order to achieve any kind of secondary gain from their employers. There is still a possibility of random errors in the collected information. Random errors in measurement of exposures may cause non-differential misclassification, which would tend to bias the odds ratio to unity. B. Recall bias: Data based on the patients self-report are prone to recall bias. Differential recall bias may lead to misclassification and to an overestimation of the risk. The possibility that a life-threatening event may

20 lead to biased report was taken into account in the formulation of the study form. It is likely that unusual events such as exceptionally heavy physical exertion, fever, anger and sudden changes in environmental temperature are recalled easily. Asking about exposures during short periods immediately before the stroke onset helps minimizing recall biases. Confounding factors Although no potential confounding by constant subject characteristics is possible in this self-matched design, factors changing over time may still act as potential confounding factors. Potential confounding by changes in smoking habits, diet and medications during the last year will be assessed and controlled for by stratification in the statistical analysis using the last year exposures as control period. 5. 8 Statistical analyses The study sample was characterized according to the distribution of age, sex, origin, chronic diseases (by categories), medications intake (by categories), principal signs and symptoms, prevalence of known risk factors, exposure variables, and severity of the ischemic event. Statistical analyses for matched case-control studies adapted to the casecrossover design were performed: hazard and control intervals pairs were defined for each subject, instead of case and control subjects pairs in matched case-control studies. Odds-ratios and 95% CI were calculated

21 for category variables such as physical activity, change in environmental temperature, positive emotions, negative emotions, anger, heavy eating and changes in posture using conditional logistic regression. Further analysis was performed by predetermined subgroups including: stroke type; treatment with β blockers, antiplatelets and diuretics at the time of the stroke; age; sex; history of hypertension. Strata heterogeneity within variables was assessed using the Chi-square test. Two-tailed paired t-test was used to test for the risk factor of environmental temperature, barometric pressure and humidity. In the analysis comparing the exposure in the day of the event with exposure during the previous year, ORs are defined as the ratio of unexposed hours during the last year for subjects potentially exposed to triggers in the day of the stroke, to exposed hours during the last year for subjects unexposed to potential triggers in the day of the stroke. Confidence intervals were calculated based on the variance of the OR natural logarithm, according to the following formulae: 95%CI: e x ; e y x = ln OR - 1.96*[var (ln OR)] y = ln OR + 1.96*[var (ln OR)] var (ln OR)= (a 1 *a 2 ) a 1 * a 2

22 when: (a 1 *a 2 ) is the sum of the products of exposed hours by unexposed hours during the last year for the complete sample. a 1 is the sum of exposed hours during the last year for subjects unexposed to potential triggers in the day of the stroke. a 2 is the sum of unexposed hours during the last year for subjects exposed to potential triggers in the day of the stroke. Differences in exposure proportions between patients reporting changes in smoking habits, diet and medications during the year preceding the stroke and patients that did not report such changes were assessed using the Chi-square test. All tests, besides the strata heterogeneity test, were conducted using the SAS version 6.12 and SPSS version 8 softwares. Chi-square tests for heterogeneity of strata within variables were calculated using the PEPI computer programs for epidemiologic analysis, version 2. The term statistically significant in this report implies a significance level of 0.05. 6.0 The pilot study The reliability and validity of the questionnaire were tested in a pilot study on 30 patients chosen from the target population. Every patient was interviewed twice to enable assessment of test-retest and inter-observer reliability. All the patients were interviewed once by the investigator. Two to four days after the first interview, an additional one was conducted by the same investigator in 15 of the 30 patients (test-retest). A registered

23 nurse working in the neurology department, trained in using the questionnaire, interviewed the remaining 15 patients included in the interobserver test. Data were entered into the data entry system programmed using the SAS software. Reliability was measured using the Spearman correlation test for ordinal and continuous variables and Cohen s kappa test for category variables. The pilot sample was characterized according to the distribution of age, sex, origin, prevalence of known risk factors, and exposure variables. Proportions of patients exposed to the various potential triggers were calculated. 6.1 Pilot study results 6.1.1 Validity of the questionnaire A. Face validity: The majority of the questions was clear and easily understood by the patients. Four of the 18 items in the PANAS scale used to assess positive and negative emotions were occasionally not well understood by some of the patients. In these cases, the item was explained to the patient using a synonym. Both interviewers used the same synonyms when needed. No refusal to answer was reported. The internal consistency of the PANAS scale was measured using the Cronbach s alpha test. All the items in the scale were found to contribute to the internal consistency including the items that were reported by the interviewers as occasionally not well understood.

24 According to these findings, all items were kept in the scale and no changes were performed in the questionnaire. B. Content validity: The data attained through the questionnaire are comprehensive and allow the investigator to extensively assess the variables of interest. 6.1.2 Reliability of the questionnaire A. Test-retest reliability: All patients reported the same day and hour of symptoms onset. The correlation coefficient for the reported time of awakening in the day of the stroke, the previous day and the usual time were r=0.99, r=1.0 and r=0.98 respectively. Agreement coefficients for exposure to potential triggers during the day of the stroke, the day before the stroke, usual exposures and presence of known risk factors are shown in tables 1-4.

25 Table 1: Test-retest agreement coefficients for exposure to potential triggers during the day of the stroke Variable type Agreement coefficient The variable (95% CI) Intensified physical Dichotomous K=1.0 activity Changes in external Dichotomous K=1.0 temperature Sudden changes in Dichotomous K=0.63 ( -0.01;1.28) body posture Anger Categorical K=1.0 Positive PANAS Continuous r=0.49 (-0.03-0.80) (mean) Negative PANAS Continuous r=0.74 (0.36-0.91) (mean) Fever Dichotomous K=1.0 Sore throat Dichotomous Agreement in 14 of 15 cases Other signs of Dichotomous K=1.0 infection Enlarged meal Dichotomous K=1.0 Other special event Dichotomous K=1.0 There is good to very good test-retest agreement in reported exposure to potential triggers during the day of the stroke.

26 Table 2: Test-retest agreement coefficients for exposure to potential triggers during the day previous to the stroke The variable Variable type Agreement coefficient (95% CI) Intensified physical Dichotomous K=1.0 activity Changes in external Categorical Agreement in 14 of 15 cases temperature Sudden changes in Dichotomous K=1.0 body posture Anger Categorical K=1.0 Positive PANAS Continuous r=0.31 (-0.24-0.71) (mean) Negative PANAS Continuous r=0.90 (0.71-0.97) (mean) Fever during the Dichotomous K=1.0 previous week Sore throat Dichotomous K=0.63 (-0.01;1.28) Other signs of Dichotomous K=1.0 infection Enlarged meal Dichotomous K=1.0 Other special event Dichotomous K=1.0 The test-retest agreement in reported exposure to potential triggers during the previous day is very good except for the positive PANAS.

27 Table 3: Test-retest agreement coefficients for usual exposure to potential triggers during the last year The variable Variable type Agreement coefficient (95% CI) Physical activity Categorical K=0.69 (0.37;1.02) Sudden changes in Dichotomous K=0.84 (0.55;1.14) body posture Anger Categorical K=0.91 (0.74;1.08) Positive PANAS Continuous r=0.97 (0.92-0.99) (mean) Negative PANAS (mean) Continuous r=0.90 (0.72-0.97) Recurrent infections Categorical K=1.0 Enlarged meals Categorical K=0.92 (0.75;1.09) The test-retest agreement in reported usual exposures to potential triggers during the last year is very good.