Infection and Risk of Ischemic Stroke
Differences Among Stroke Subtypes
Background and Purpose— Although prior studies have demonstrated that 25% to 35% of stroke patients have had a recent infection, the role of infection as a risk factor remains unclear. Our aim was to characterize the effect of infectious/inflammatory syndromes on stroke risk.
Methods— Case-control and crossover analyses of 233 cases and 363 controls aged 21 to 89 years were performed. Cases were patients hospitalized with a first ischemic stroke at a Los Angeles, California, medical center. Controls were outpatients in the hypertension, diabetes, and general medical clinics. All subjects were administered a neurological examination, an infection/inflammation (I/I) examination, and an interview to elicit recent I/I history at baseline (within several days of stroke onset) and again approximately 2 months later. Three physicians classified subjects by the presence or absence of I/I within 1 month of the index dates, based on findings of the I/I examination, the interview report, and laboratory results.
Results— Infections, either total or specific, were not found more frequently in cases than controls. However, patients with a recent respiratory tract infection suffered more often from large-vessel atherothromboembolic or cardioembolic stroke than did patients without infection (48% vs 24%, P=0.07). The age- and sex-adjusted relative risk estimate for these subtypes was 1.75 (95% CI, 0.86 to 3.55). The risk was notably high for those without stroke risk factors: 4.15 (95% CI, 1.22 to 14.1) for normotensives, 2.71 (95% CI, 1.04 to 7.06) for nondiabetics, and 1.74 (95% CI, 0.74 to 4.07) for nonsmokers. Patients with a recent respiratory infection also had a more severe neurological deficit on admission than those without infection (P=0.05).
Conclusions— Our results suggest that respiratory tract infection may act as a trigger and increase the risk of large-vessel and/or cardioembolic ischemic stroke, especially in those without vascular risk factors.
Increasing evidence has linked infections to atherosclerosis, myocardial infarction, and stroke.1,2 Although initially formulated nearly a century ago,3,4 the hypothesis that infections might be causal agents in atherosclerosis and cardiovascular disease received little attention until the 1970s. Then, control of epidemic respiratory diseases was suggested as 1 reason for the decrease in coronary heart disease morality and sporadic influenza as a possible trigger of incident disease.5,6 Since then, case-control and cohort studies have found infection, especially respiratory and dental infections, associated with myocardial infarction and other coronary heart disease.7–12 Although prior studies have demonstrated that as many as 25% to 35% of patients with ischemic stroke have a history of infection within the preceding month,13–17 the role of infection as a stroke risk factor remains unclear.
We conducted a study of ischemic stroke with the use of both case-control and crossover analyses to characterize the prevalence and timing of infectious/inflammatory syndromes potentially associated with increased risk. We also investigated whether the association of ischemic stroke with recent infection was restricted to stroke subtypes and whether recent infection influenced the severity of the neurological deficit.
Stroke cases were identified by review of admitting records of the emergency department and the neurology unit of the Los Angeles County/University of Southern California Medical Center for diagnoses possibly related to stroke. The study neurologist reviewed the medical charts of potential cases and identified those eligible for the study. Eligible cases included all patients with a diagnosis of a first ischemic stroke between August 24, 1994, and October 29, 1998, who were aged 21 years or older and residents in southern California at least 2 months, except those with exclusionary conditions. Patients were ineligible if stroke onset had occurred >4 days before hospitalization or while in the hospital or if the patient was a participant in a double-blind treatment trial. Also excluded were patients who had an intracerebral hematoma, tumor or other central nervous system mass lesion, recent (<1 month) myocardial infarction, recent (<3 months) deep vein thrombosis, hepatic or renal failure, recent (<5 years) malignancy except basal cell carcinoma of the skin, lupus erythematosus, polyarteritis nodosa or Takayasu’s disease, immunodeficiency disorder, active tuberculosis (>3 months therapy), chronic (>3 months) urinary tract infection, or chronic (>3 months) bronchitis; was HIV-positive or pregnant; had recently (<3 months) used street drugs; or had recently (<1 month) sustained significant trauma.
Controls were identified by review of patients seen in the hypertension, diabetes, and general medicine clinics of the same medical center. Like cases, controls were 21 years of age or older and residents of southern California for at least 2 months. They also had no history of stroke and were excluded for the same conditions listed for the cases. It was desired that the age and racial distributions and frequencies of hypertension, diabetes, and smoking of controls be similar to that of cases, and an effort was made to approach potential controls with the needed characteristics. The Institutional Review Board of the University of Southern California School of Medicine approved this project.
At the initial visit, each subject (cases and controls) underwent several procedures. The study neurologist administered a physical examination and completed the National Institutes of Health (NIH) stroke scale to quantify the severity of neurological deficit.18 A physician trained in performing an infection/inflammation (I/I) examination evaluated the subject and completed an I/I form. Temperature, blood pressure, and pulse were recorded; a urine sample was obtained; and venipuncture was performed. Subjects were interviewed by the project coordinator to obtain basic demographic information (race/ethnicity, marital status, education, height, weight); history of smoking and alcohol consumption; and medical history of hypertension, diabetes, heart valve problem, heart attack, arrhythmia, transient ischemic attack, cancer, tuberculosis, peptic ulcer, head and other trauma, and headaches. An interviewer blinded to the subject’s status and study hypothesis conducted a structured baseline interview with each subject to elicit recent (<1 month) history of I/I. Because cases may have been interviewed several days after stroke onset, controls were randomly assigned a similar lag time, and the period of review was fixed to 1 month before the start of the lag time. If the subject was unable to respond directly, his or her next-of-kin was consulted as a proxy respondent.
Medical records of cases and controls were reviewed to obtain information on blood pressure, diabetes, myocardial infarction, arrhythmia, and transient ischemic attack. For cases, results of chest x-ray, ECG, CT, MRI, MRA, echocardiogram, angiogram, and carotid duplex scans were obtained. The study neurologist also reviewed the medical record to classify the etiologic subtype of the stroke.19
Each subject was asked to return for a follow-up visit ≈2 months after the initial visit. (The follow-up visit was scheduled for 1 month after the initial visit for the first 4 months of the study.) The neurological examination, I/I examination, and venipuncture were repeated. Temperature, blood pressure, and pulse were recorded, and the blinded interviewer conducted a follow-up interview to elicit occurrences of infections and inflammations since the baseline interview.
Three physicians (2 neurologists and an infectious disease physician) reviewed the findings of the I/I examination, the I/I interview report, and laboratory results to classify all subjects by the presence or absence of I/I within 1 month of the index date. The review was done without knowledge of the person’s case/control status or baseline/follow-up visit. All visits were reviewed twice. Situations in which the ratings did not agree were returned for discussion and resolution at a consensus conference.
I/I was classified as “definite” when the condition was observed on physical examination or it was reported on interview that the subject saw a health professional or took medicine for the condition. The condition was classified as “possible” when it was reported on interview but did not meet the criteria for “definite.” The definitions of infections and inflammations are available on request. For upper respiratory tract infection, symptoms included fever or chills plus 1 of the secondary criteria: sore throat, cough or sneeze, colored sputum, rhinorrhea, ear pain, sinus pain, or otorrhea; in the absence of fever or chills, 2 of the secondary criteria were needed.
Stroke cases were compared with controls by standard statistical procedures. Analyses were adjusted for age (<40, 40 to 44, 45 to 49, …, 65 to 69, 70+ years) and sex. Relative risks (RRs) and 95% CIs were calculated by univariate and multivariate logistic regression techniques for stratified case-control studies.20 Reported probability values are 2-sided.
Using a case-crossover design,21,22 in which each case acts as his/her own control, we assessed whether the date of stroke and the date of follow-up were preceded by infection. Using discordant pairs only (cases who had an infection either before the stroke or before follow-up), we calculated the RR.
Of 347 eligible stroke patients, 35 were not asked to participate for various reasons (eg, death, language barrier, unable to give informed consent, early discharge, oversight). Of those invited to participate, 79 refused, leaving 233 (75%) for analysis. Of the 588 potential control subjects invited to participate, 226 refused, leaving 362 (62%) for analysis.
Table 1 gives the baseline characteristics of cases and controls. Subjects ranged in age from 21 to 89 years. The mean age of cases was similar to that of controls (57 vs 55). More cases than controls were male (56% vs 48%). The racial/ethnic background also differed; smaller proportions of controls than cases were non-Hispanic white or Asian, and a greater proportion was black. The distributions of the 3 major stroke risk factors, hypertension, diabetes, and smoking, were similar for cases and controls. The eligible cases not enrolled were similar in age and sex to those enrolled; however, a greater proportion of nonparticipants was black.
Follow-up visits were completed by 186 cases and 323 controls. An additional 9 cases and 4 controls consented to the follow-up interview only. Reasons for no follow-up of 38 cases and 35 controls were refusal (9 cases and 29 controls), moved out of area or into a nursing home (7 cases and 1 control), lost to follow-up (11 cases and 5 controls), and death (11 cases). The mean±SE number of days between the initial and follow-up visit was 82±3.0 (median, 62).
A baseline I/I physical examination was performed on 172 cases and 360 controls. The medical records of an additional 55 cases were reviewed to abstract the same information. Follow-up I/I examinations were performed on 185 cases and 321 controls. Consensus conference ratings were completed for 592 baseline visits and 522 follow-up visits.
The classification of definite I/I syndromes within 1 week of the baseline visit by case/control status is given in Table 2. Foot, dental, and respiratory tract infections were the most common infections in both cases and controls. Infections, either total or specific, were not found more frequently in cases than controls. In fact, the frequency of infection was greater in controls for almost all types of infection. Infection rates seen at follow-up were similar to those at baseline. Excluding diabetics from the analysis did not materially change the proportion of cases or controls with any infection. The percentage was unchanged for cases (43%) and was reduced from 56% to 52% in controls. For respiratory tract infection, the proportion decreased more in controls (13% to 9%) than in cases (11% to 10%).
Table 3 shows the demographic and clinical features of cases with and without a definite respiratory tract infection within 1 week of stroke. The profile of sex, age, ethnicity, and vascular risk factors was similar in both groups. However, patients with infection had a more severe neurological deficit on admission as assessed by the NIH stroke scale than those without infection (P=0.05). Similar results were observed when infection was defined as that occurring within 1 month of stroke, because only 2 additional patients were classified as having had respiratory tract infections. When “possible” infections were included, the number of cases with respiratory tract infection doubled (25 to 53), but the distributions of demographic and clinical features were similar, except that those with “possible” infection tended to be younger.
Patients with recent respiratory infection suffered more often from large-vessel atherothromboembolism- or cardioembolism-derived cerebral ischemia than did patients without infection (P=0.07). Almost half of the cases with respiratory infection had a large-vessel or cardioembolic stroke versus less than one fourth in the uninfected group. Small-vessel disease was diagnosed as the cause of cerebral ischemia primarily in patients without recent infection. The age- and sex-adjusted RR estimates of the individual stroke subtypes for 1-week respiratory tract infection are given in Table 4. For large-vessel and cardioembolic strokes combined, the RR is 1.75 (95% CI, 0.86 to 3.55). The small number of large-vessel and cardioembolic strokes limited examination for a possible effect modification by other variables. Nonetheless, there was some indication that the effect of respiratory infection on stroke risk was greater in those without stroke risk factors. The RRs (95% CIs) for large-vessel or cardioembolic disease were 0.85 (95% CI, 0.31 to 2.36) in those with hypertension versus 4.15 (95% CI, 1.22 to 14.1) in those without, 1.13 (95% CI, 0.37 to 3.51) in those with diabetes versus 2.71 (95% CI, 1.04 to 7.06) in those without, and 0.89 (95% CI, 0.21 to 3.83) in smokers versus 1.74 (95% CI, 0.74 to 4.07) in nonsmokers. The effect of infection was 2.14 (95% CI, 0.88 to 5.24) in those aged 57 or less versus 1.38 (95% CI, 0.42 to 4.58) in those older.
Among the 192 cases for whom we had consensus data on I/I status at both baseline and follow-up, 32 were discordant for “definite” respiratory tract infection within 1 week: 19 had infection before stroke but not before follow-up, and 13 cases had infection before follow-up but not before stroke. The RR of stroke among those with infection was 1.46 (95% CI, 0.72 to 3.15). The small number of discordant pairs limited detailed subgroup analysis; the RR estimates in all subgroups were >1, but the CIs were wide and included 1 (Table 5). Analysis by stroke subtype showed no increased risk of ischemic stroke due to small-vessel disease in persons with a preceding respiratory tract infection (RR=1.00). Of the 3 cases with cardioembolic stroke that were discordant for infection at baseline and follow-up, all had respiratory infection at baseline (RR=∞). Risk of large-vessel disease was increased (RR=2.00) but not significantly.
Our results suggest that respiratory tract infection during the week or month preceding stroke is not a major risk factor for ischemic stroke. Increased risk was limited to the large-vessel and cardioembolic strokes and was greatest in those without vascular risk factors.
Previous studies of infection and stroke have found that most infections among stroke cases were located in the respiratory tract.13–15,23 A case-control study in Germany24 found that recent infection was associated with cardioembolic stroke. In the subgroup with febrile infection (19 of 197 strokes), almost half the patients (9 of 19) suffered from cardioembolic stroke (P=0.04). Small-vessel disease was diagnosed as the cause of cerebral ischemia only in patients without recent infection. Another study by the same researchers17 also found recent infection significantly increased the risk of cerebrovascular ischemia from cardioembolism (RR=3.25; 95% CI, 1.06 to 10) and tended to elevate the risk for arterioarterial embolism (RR=7.0; 95% CI, 0.86 to 57).
The diagnosis of stroke subtype used in our study is subject to error. Many patients have more than 1 pathogenic process, and identifying the different mechanisms is complex. Because the possible misclassification of stroke subtype is unlikely to be related to infection, this error should not bias our results. We are, however, limited by the small number of large-vessel and cardioembolic strokes, which results in wide CIs for the RR estimates. Our population of stroke patients was drawn from a large, inner-city hospital where risk factors for small-vessel disease are highly prevalent and often poorly treated. It is therefore not surprising that small-vessel disease was the predominant stroke subtype.
Because many conditions that predispose to infection also carry risk for stroke (eg, diabetes), selecting control subjects requires care when studying their possible association. Our controls were from the same medical facility as the cases and thus, tended to be of the same low socioeconomic status. We also attempted to obtain controls of similar sex, age, and risk factor distributions as the cases. Although they were outpatients, our controls reported high infection rates, which varied little between baseline and follow-up. Infections are more frequent in populations that have a lower socioeconomic status.25 In addition, those patients accessing the medical center facilities are more likely to be sick than the general population. The rate of respiratory tract infection in our controls was 13% compared with 3% to 4% in community controls in Finland and Germany.13,15 However, the rate of infection in cases was comparable to the 10% to 30% seen in other studies.13–15,17,23
Because of problems inherent in control selection, we also did a case-crossover analysis comparing cases at time of stroke to themselves at follow-up. This study design can be used to study the effect of a brief and transient exposure period on the risk of an acute outcome, such as stroke and has the advantage that patient characteristics (which are relatively stable during the 2-month follow-up period), including vascular risk factors, are controlled for in the analysis. The results for the case-crossover analysis confirmed the major finding of the case-control study: respiratory tract infections increase the risk of large-vessel and cardioembolic stroke but not small-vessel disease. Thus, uncontrolled confounding in the case-control analysis is unlikely to explain the observed results.
Our data indicate that a recent respiratory tract infection is associated with more severe stroke. Similarly, in a case-control study in Germany, the neurological deficit on admission (assessed by the Scandinavian Stroke Scale) was more severe in patients with infection than those without.24 However, in another study by the same authors,17 those with recent infection did not have a more severe neurological deficit as measured by the NIH stroke scale than those without. Macko et al16 reported that 13 stroke cases with preceding (<1 week) I/I had a mean score on the Toronto stroke scale of 51±68 versus 32±31 in 24 stroke cases without a preceding infection. Neither Ameriso et al14 nor Bova et al23 found differences in the severity of neurological deficits between patients with and without prior infection.
Infections may induce thrombosis and brain infarction by several mechanisms. Some mechanisms link infections with large-artery atherogenesis. Other mechanisms suggest a prothrombotic state. Infection and inflammation cause many systemic effects, including changes in lipid metabolism, platelet aggregation, lysis, blood coagulation, alterations in endothelial function, spasms in vascular smooth muscle, atheroma instability, and subsequent plaque rupture. Compared with stroke patients without infection, those with infection were found to have increased fibrin D-dimer concentrations and anticardiolipin antibody titers.14 Likewise, infection-associated stroke patients had lower levels of activated protein C and ratios of active tissue plasminogen activator to plasminogen activator inhibitor-1 and elevations in C4bp antigen levels than noninfection stroke cases or controls.26 Only the levels of C-reactive protein showed significant differences between stroke patients with and without infection in the study by Grau et al.24 The difference in fibrinogen levels between the groups was not statistically significant (P=0.08). The markers of coagulation and fibrinolysis (antithrombin-3, plasminogen, plasminogen activator inhibitor-1, thrombin-antithrombin complexes, prothrombin fragment F1+2, and fibrin D-dimer) and the marker of endothelial damage (thrombomodulin) were not different between groups. This indicates a link between ischemic stroke, infection, and a procoagulant state. However, another study found no difference between stroke patients with and without preceding infection with respect to factor VII and factor VIII activity, fibrin monomer, von Willebrand factor, fibrin D-dimer, C4b-binding protein, protein S, anticardiolipin antibodies, interleukin-1 receptor antagonist, soluble tumor necrosis factor-α receptor, interleukin-6, interleukin-8, and neopterin.17 The pathological link between infection and stroke is still insufficiently understood.
None of the studies performed to date can prove a causal role of infection in ischemic stroke. However, our study suggests that infection may act as a trigger and may temporarily increase the risk of large-vessel and/or cardioembolic ischemic stroke, especially in those without vascular risk factors. Our results in a racially and ethnically diverse sample can be generalized to most urban, multiethnic populations in the United States.
This study was supported in part by NIH grants P01-NS31945 and M01 RR-43 from the GCRC Branch of the National Center for Research Resources.
- Received June 19, 2002.
- Revision received August 21, 2002.
- Accepted September 9, 2002.
Nieto FJ. Infections and atherosclerosis: new clues from an old hypothesis? Am J Epidemiol. 1998; 148: 937–947.
Mattila KJ, Valtonen VV, Nieminen MS, Asikainen S. Role of infection as a risk factor for atherosclerosis, myocardial infraction, and stroke. Clin Infect Dis. 1998; 26: 719–734.
Osler W. Diseases of the arteries. In: Osler W, MacCrae T, eds. Modern Medicine: Its Theory and Practice in Original Contributions by Americans and Foreign Authors.Vol 4. Philadelphia: Lea & Febiger; 1908: 426–427.
Bainton D, Jones GR, Hole D. Influenza and ischaemic heart disease: a possible trigger for acute myocardial infarction? Int J Epidemiol. 1978; 7: 231–239.
Mattila KJ, Nieminen MS, Valtonen VV, Rasi VP, Kesäniemi YA, Syrjälä SL, Jungell PS, Isoluoma M, Hietaniemi K, Jokinen MJ, Huttunen JK. Association between dental health and acute myocardial infarction. Br Med J. 1989; 298: 779–781.
DeStefano F, Anda RF, Kahn HS, Williamson DF, Russell CM. Dental disease and risk of coronary heart disease mortality. Br Med J. 1993; 306: 688–691.
Gupta S, Leatham EW, Carrington D, Mendall MA, Kaski JC, Camm AJ. Elevated Chlamydia pneumoniae antibodies, cardiovascular events, and azithromycin in male survivors of myocardial infarction. Circulation. 1997; 96: 404–407.
Ameriso SF, Wong VLY, Quismorio FP, Fisher M. Immunohematologic characteristics of infection-associated cerebral infarction. Stroke. 1991; 22: 1004–1009.
Grau AJ, Buggle F, Heindl S, Steichen-Wiehn C, Banerjee T, Maiwald M, Rohlfs M, Suhr H, Fiehn W, Becher H, Hacke W. Recent infection as a risk factor for cerebrovascular ischemia. Stroke. 1995; 26: 373–379.
Macko RF, Ameriso SF, Barndt R, Clough W, Weiner JM, Fisher M. Precipitants of brain infarction: roles of preceding infection/inflammation and recent psychological stress. Stroke. 1996; 27: 1999–2004.
Grau AJ, Buggle F, Becher H, Zimmermann E, Spiel M, Fent T, Maiwald M, Werle E, Zorn M, Ing D, Hengel H, Hacke W. Recent bacterial and viral infection is a risk factor for cerebrovascular ischemia: clinical and biochemical studies. Neurology. 1998; 50: 196–203.
Brott T, Adams HP Jr, Olinger CP, Marler JR, Barsan WG, Biller J, Spilker J, Holleran R, Eberle R, Hertzberg V, Rorick M, Moomaw CH, Walker M. Measurements of acute cerebral infarction: a clinical examination scale. Stroke. 1989; 20: 864–870.
Adams HP Jr, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, Marsh EE III, and the TOAST investigators. Classification of subtype of acute ischemic stroke: definitions for use in a multicenter clinical trial. Stroke. 1993; 24: 35–41.
Breslow NE, Day NE. Statistical Methods in Cancer Research,Vol 1: The Analysis of Case-Control Studies. Lyon, France: International Agency for Research on Cancer; 1980.
Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol. 1991; 133: 144–153.
Mittleman MA, Maclure M, Robins JM. Control sampling strategies for case-crossover studies: an assessment of relative efficiency. Am J Epidemiol. 1995; 142: 91–98.
Bova IY, Bornstein NM, Korczyn AD. Acute infection as a risk factor for ischemic stroke. Stroke. 1996; 27: 2204–2206.
Grau AJ, Buggle F, Steichen-Weihn C, Heindl S, Banerjee T, Seitz R, Winter R, Forsting M, Werle E, Bode C, Nawroth P-P, Becher H, Hacke W. Clinical and biochemical analysis in infection-associated stroke. Stroke. 1995; 26: 1520–1526.
Malaty HM, Graham DY. Importance of childhood socioeconomic status on the current prevalence of Helicobacter pylori infection. Gut. 1994; 35: 742–745.
Macko RF, Ameriso SF, Gruber A, Griffin JH, Fernandez JA, Barndt R, Quismorio FP Jr, Weiner JM, Fisher M. Impairments of the protein C system and fibrinolysis in infection-associated stroke. Stroke. 1996; 27: 2005–2011.