Recent bacterial and viral infection is a risk factor for cerebrovascular ischemia

In 1897, Sigmund Freud reported that acute infantile hemiparesis was associated with an ongoing or recent infectious disease in about one third of cases. Postmortem studies revealed a mainly vascular origin of neurologic symptoms.¹ Some years earlier, Marie had already described cases of children with recent or acute infection and sudden hemiparesis that may have been of vascular origin. This association was similar to that which he had found to exist between infection and multiple sclerosis.² More recent studies indicated that acute infection is among the most important risk factors for ischemic stroke in children.^3,4 Syrjänen et al.⁵ showed that acute infection was an important and independent risk factor for ischemic stroke in adult patients under the age of 50.

In a first case-control study including older subjects, we found that the role of recent infection as a stroke risk factor was not restricted to young and middle-aged patients. Infection in the week before ischemia, especially bacterial and respiratory tract infection, were most important in this respect.^6,7 The recognition of infectious disease as a stroke risk factor could be important for stroke prevention and may help to understand epidemiologic features of stroke that are insufficiently explained by established risk factors (e.g., the seasonal variation of stroke incidence).⁸ In our previous study, we investigated control subjects from the general population. Because different methodological approaches are necessary to establish infection as a risk factor, we designed a second case-control study with hospital control subjects. A further aim of this study was to investigate pathogenetic mechanisms possibly linking infection and stroke. Previous studies provided evidence that acute infection may lead to a hypercoagulable state.^9,10 However, the pathogenesis of infection-associated stroke is still insufficiently understood.

Methods. We performed a case-control study investigating patients with acute cerebrovascular ischemia and hospitalized patients with nonvascular and noninflammatory neurologic diseases as a control group. Control subjects were individually matched to patients by sex, age, and season of admission. Between August 1995 and January 1996, we examined 166 consecutive patients with acute cerebrovascular ischemia who were admitted to the Neurology Department of the University of Heidelberg. The patients, 56 women and 110 men, were 61.2 ± 13.8 (mean ± SD) years of age(range, 22 to 85 years). Among the patients, 130 had ischemic stroke and 36 had transient ischemia with symptoms resolving within 24 hours and without infarction on neuroimaging. All patients were examined by cranial CT to exclude cerebral hemorrhage. A second CT or cranial MRI was performed in 93 patients. Results from the following technical investigations were available in patients: extracranial (n = 155) or transcranial (n = 150) Doppler sonography, digital subtraction, CT or MR angiography (n = 66), ECG and/or Holter monitoring (n = 152), and (transthoracic and/or transesophageal) echocardiography (n = 99).

Between August 1995 and March 1996, we investigated 166 control subjects hospitalized for neurologic diseases. Exclusion criteria for the control group were neurovascular diseases, infectious and autoimmunologic disorders of the nervous system, cerebral or spinal metastases from malignant tumors, paraneoplastic disorders, and neurologic diseases due to alcohol or drug abuse. The first potential control subject admitted after a respective patient of the same sex and comparable age (±10 years) was asked for participation. A larger age difference within a case-control pair was accepted when no control subject could be identified within 2 weeks after patient admission. Because of a lower than usual admission rate of potential control subjects, the average interval between the examination of a patient and his or her respective control subject was 35 days. All subjects asked agreed to participate. The control subjects, 56 women and 110 men, were 57.9± 15.3 (range, 19 to 87) years of age. None of the control subjects had postponed an earlier date of admission due to symptoms of acute infection. The control subjects suffered from primary tumors of the brain and spinal cord (n = 47), neurodegenerative diseases and movement disorders (n = 40), disorders of the cranial nerves and the peripheral nervous system (n = 24), diseases of the vertebral column and the cervicocranial junction (n = 22), epileptic seizures (n = 15), or other disorders (n = 18).

We took a detailed history from all subjects using a standardized questionnaire that focused on signs and symptoms of infection during the last 4 weeks and past history of recurrent and chronic infection. If information could not be obtained directly from the patient, a next-of-kin was interviewed (patients, n = 18; control subjects, n = 0). We diagnosed a recent infection as described previously.⁶ Briefly, diagnosis of infection required the combination of two or more related symptoms typical of infection or at least one typical symptom together with measurement of increased body temperature (>37.5 °C) or a corresponding microbiological finding indicating acute infection or a related pathologic finding compatible with infection on technical examination (e.g., radiograph). To be accepted, symptoms of infection had to occur before ischemia or admission. Infection was not diagnosed if information was lacking or insufficient (patients, n = 10; control subjects, n = 0) or if diagnosis was doubtful or another etiology (e.g., allergy or vasculitis) was possible. When at least one sign or symptom typical of infection was present, standardized microbiological tests were performed. In case of symptoms of respiratory tract infection, this included detection of specific antibodies against adeno virus; respiratory syncytial virus (RSV) (IgG and IgM; ELISA); influenza A and B virus (IgG and IgA); immunofluorescence assay (IFA); and mycoplasma pneumoniae (IgM; IFA); complement fixation (CF) antibodies against adeno, influenza A and B, parainfluenza, RSV, echo and coxsackie virus; and an antistreptolysin test (hemolysis inhibition test). If related symptoms were present, additional cultures of sputum or a tonsillar swab was performed. A chest x-ray was done if pneumonia was suspected. Symptoms of gastrointestinal infection prompted stool cultures and tests for antibodies against Salmonella enteridis (tube agglutination test), Campylobacter jejuni (CF test), and Yersinia enterocolitica (microagglutination test). If symptoms of urinary tract infection were present, a urine sediment and urine cultures were investigated. In case of skin infections, cultures of local swabs were performed. If sepsis or endocarditis was suspected, we investigated blood cultures and performed echocardiography. Serologic tests were repeated after 10 or more days. Changes of antibody titers of two or more steps or the presence of IgM (IgA) type antibodies were considered as diagnostic. Suppliers and cutoff values of serologic tests were identical to those given previously.⁶ The CF tests, not previously performed, were provided by Roche Serologie(Munich, Germany). The diagnosis of a bacterial origin of infection was based on the proof of causative bacteria or on the presence of purulent secretion or typical roentgenologic evidence (e.g., chest radiograph in pneumonia). The diagnosis of a viral origin required the serologic proof of causative viruses. In absence of such proof, a viral origin of respiratory infection was suspected when no signs or symptoms of a bacterial infection were present. Subclinical infection was not investigated.

The severity of the neurologic deficit in patients was quantified using the National Institutes of Health (NIH) scale.¹¹ Etiologic subtypes of cerebrovascular ischemia were defined as reported previously.⁷ Here, the criteria are briefly summarized. Embolism from large-artery atherosclerosis was defined as presence of a stenosis > 50% diameter reduction of a brain-supplying artery corresponding to clinical symptoms and with location and morphology typical of atherosclerosis on Doppler ultrasound or angiography. Cardioembolism was defined as presence of high- or medium-risk sources of cardiac emboli according to the classification of ORG 10172 in Acute Stroke Treatment investigators.¹² Thromboembolism of undetermined origin was defined as cerebral ischemia in supply territories of large intracranial arteries with no or two or more identifiable sources of embolism. Microangiopathy was the presence of one of five lacunar syndromes (pure motor stroke, pure sensory stroke, sensorimotor stroke, ataxic hemiparesis, dysarthria-clumsy hand syndrome) and infarction < 1 cm diameter. Cerebral ischemia of hemodynamic origin was defined as extraterritorial infarcts on CTs or MRIs and stenoses > 80% of occlusions of corresponding supply arteries. Cervical artery dissection was defined as typical findings on angiography or cervical MRI such as mural hematoma or string sign. Cerebral vasculitis was defined as abnormal immunologic serum parameters together with typical neuroradiologic findings (MRI, angiography). Cerebrovascular ischemia of unknown origin was defined as patients with incomplete workup or unknown etiology despite extensive investigations. Complete workup comprises at least one follow-up CT or MRI when first imaging did not reveal an acute ischemic lesion, extra- and transcranial Doppler ultrasound, electrocardiogram and Holter monitoring, transthoracic echocardiogram, and routine blood analysis.

To study pathogenetic pathways possibly linking infection and stroke, we compared hematologic and biochemical parameters in patients with and without recent infection. From these analyses, we excluded all subjects with trauma, surgery, or previous acute vascular diseases within the last month, with noninfectious inflammatory disorders or malignancies. Furthermore, we excluded those subjects in whom recent infection was possible but above criteria were not sufficiently met or who developed infection within the first 2 days after ischemia. For analyses of routine laboratory parameters, we only considered blood samples taken within 48 hours after ischemia. To investigate special inflammatory or hematologic parameters, we selected all patients with ischemic stroke (not transient ischemia) and infection within the preceding week in whom blood samples were available within 24 hours after ischemia. We selected the same number of stroke patients without infection with the aim to create two groups well matched for age, sex, prevalence of vascular risk factors and etiology of stroke, and severity of the neurologic deficit and the interval between ischemia and venipuncture. In these selected subjects, we analyzed the following parameters from citrated plasma (0.9% sodium citrate) samples: factor VII and factor VIII activity (modified activated partial thromboplastin time with factor VII or factor VIII deficient plasma; Dade International, Miami), fibrin monomer (enzyme immunoassay [EIA], Boehringer-Mannheim, Germany), fibrin D-dimer (EIA; Agen Biomedical Ltd., Acacia Ridge, Australia), C4b-binding protein(C4b-BP; microlatex photometric immunoassay; Diagnostica Stago Asnières-sur-Reine, France), protein S (free and bound protein S; EIA; Boehringer-Mannheim), and von Willebrand factor antigen(electroimmunodiffusion; Immuno, Vienna, Austria). From serum we assessed anticardiolipin antibodies (IgG, IgM, IgA; EIA; ELISA, Freiburg, Germany), interleukin-1 receptor antagonist (IL-1RA) and soluble tumor necrosis factor-α receptor I (sTNF-RI) (EIAs; R & D Systems, Minneapolis, MN), IL-6, IL-8, and neopterin (EIA; DPC, Los Angeles, CA). All samples were frozen at -80 °C until measurements were done. Analyses were done blinded for the status of the patients.

In the statistical analysis of risk factors, we used conditional logistic analysis for odds ratio (OR) estimation.¹³ Variables for the final logistic model were selected in order of their p values in univariate analysis. For age adjustment we used the procedure described by Neuhäser und Becher.¹⁴ A possible relation between infection and age on the risk was investigated by including an interaction term into the logistic model. ORs and 95% CI are given in all analyses. The chi-square test and Fisher's exact test were applied to compare sample proportions as appropriate. For the comparison of biochemical and hematologic parameters, we applied the nonparametric Mann-Whitney U-test. We used the statistical software package SAS (SAS, Cary, NC) for the analyses.

Results. Infection within the 1 preceding week was more common in patients (22.3%) than control subjects (8.4%; OR 3.1; 95% CI, 1.57 to 6.1). The prevalence of infection during the 2 to 4 weeks before ischemia or admission was not different between groups. Therefore, we focus on infection within the 1 preceding week. Infection strongly increased the risk of stroke for subjects under the age of 50; in older subjects, there was a trend to an elevated risk. The risk was significantly increased in both men and women(table 1). The risk of ischemic stroke (OR 3.5; 95% CI, 1.60 to 7.7) and the risk of transient cerebral ischemia (OR 3.0; 95% CI, 0.61 to 14.7) were similar.

Respiratory tract infection was the predominant type of infection in both groups and increased the risk for cerebrovascular ischemia more than infection in general (OR 4.3; 95% CI, 1.88 to 9.8) (table 2). Microbial agents that probably caused infectious symptoms were detected in 21 of 37 patients and in 6 of 14 control subjects with recent infection. Several different viruses and bacteria were found in patients(table 3). Infection of bacterial (OR 3.3; 95% CI, 1.34 to 8.3) and infection of viral (OR 3.6; 95% CI, 1.33 to 9.7) origin were both associated with an increased risk for cerebrovascular ischemia (seetable 2). Only eight patients and one control subject with infection had measured increased body temperature (>37.5 °C). Three patients and one control subject had consulted a physician because of the infection. In most patients, symptoms of infection had started to wane when cerebral ischemia occurred. However, in 31 of 37 patients and in 11 of 14 control subjects with infection during the preceding week, at least one remaining sign or symptom of the infection was present on admission. The prevalence of infection showed a seasonal variation in both patients (August to September, 12.5% [6/48]; October to November, 27.4% [17/62]; December to January, 25.0% [14/56]) and control subjects (August to September, 2.9%[1/35]; October to November, 7.9%[3/38]; December to January, 10.3% [6/58]; February to March, 11.4%[4/35]).

In univariate analysis, atrial fibrillation, previous stroke or TIA, coronary heart disease, arterial hypertension, peripheral arterial disease, diabetes mellitus, and a positive family history of stroke were significant risk factors for cerebrovascular ischemia (table 4). In age-adjusted multiple logistic regression analysis including these factors and current smoking, recent infection remained a significant risk factor (OR 2.9; 95% CI, 1.31 to 6.4; p = 0.009). Atrial fibrillation (OR 23.3; 95% CI, 5.05 to 107; p < 0.001), previous stroke or TIA (OR 4.8; 95% CI, 2.24 to 10.4;p < 0.001), and arterial hypertension (OR 2.5; 95% CI, 1.39 to 4.5; p = 0.002) were other significant risk factors. Combined coronary heart disease and peripheral arterial disease (OR 1.9; 95% CI, 0.94 to 4.0; p = 0.072), current smoking (OR 1.7; 95% CI, 0.90 to 3.3; p = 0.10), diabetes mellitus (OR 1.6; 95% CI, 0.71 to 3.4; p = 0.28), and a positive family history of stroke (OR 1.5; 95% CI, 0.77 to 2.7; p= 0.25) were not independently associated with cerebrovascular ischemia. Results summarized in table 1 suggest that age may significantly modify the role of infection as a stroke risk factor. Investigating the interaction between age and infection in our logistic regression model, we found a significant inverse relation between age and infection as a risk factor (p = 0.018).

To gain insight into the pathogenesis of infection-associated stroke, we compared patients with and without infection. Patients with infection tended to be younger (57.4 ± 15.5 years versus 62.3 ± 13.2 years, p = 0.098) and were more often former or current smokers (59.5% versus 40.5%, p = 0.041) than patients without infection. There was no difference between both groups with respect to other risk factors(e.g., hypertension, diabetes mellitus, or previous stroke or TIA) or to etiologic subtypes of cerebrovascular ischemia (e.g., cardioembolism, 35.1% versus 34.9%; patients with versus patients without infection; arterioarterial embolism, 21.6% versus 25.6%). However, the combined diagnoses"thromboembolism of unknown origin" and "unknown etiology despite complete workup" tended to be more common among patients with than among those without recent infection (32.4% versus 20.9%; p = 0.15). Recent infection significantly increased the risk for cerebrovascular ischemia from cardioembolism (OR 3.25; 95% CI, 1.06 to 10.0) and tended to elevate the risk for arterio-arterial embolism (OR 7.0; 95% CI, 0.86 to 57). Among patients receiving a standardized neurologic examination (NIH stroke scale) within 48 hours after ischemia, those with recent infection did not have a more severe neurologic deficit (n = 26; median of NIH Stroke Scale, 5; 25 to 75% quartile, 2 to 19) than those without infection (n = 83; 6; 2 to 14).

Patients with infection had a higher leukocyte count (n = 28; 10.4± 2.8/nL) on first evaluation than patients without infection (n = 73; 9.0 ± 2.7/nL; p = 0.017). Platelet count, hematocrit, fibrinogen, C-reactive protein, urea, cholesterol, and triglycerides were not significantly different between both groups (data not shown). Blood samples from the first 24 hours after stroke were available in 21 patients with infection and accordingly, we selected 21 stroke patients without infection. Both groups were well balanced with respect to age, sex, vascular risk factors, etiology and severity of stroke, and time interval between ischemia and venipuncture. There was a trend to higher IL-1RA in patients with (604± 297 ng/mL) than in those without infection (485 ± 242 ng/mL;p = 0.14); however, none of the parameters assessed showed a significant difference between groups (data not shown).

Discussion. Infection in the preceding week was significantly more common in patients with cerebrovascular ischemia than in patients with other neurologic diseases. As expected, the prevalence of infection in our hospital control group was higher than among population control subjects in recent studies.^5,6 We excluded patients from the control group with diseases than can be triggered by infection (e.g., MS). However, acute infection can cause deterioration of many diseases in a nonspecific way. Additionally, the presence of a second disorder (e.g., an infection) makes admission to hospital more likely.¹⁵ Furthermore, the quality of information required for the diagnosis of infection was better in the control group than in patients. For all these reasons, we may have underestimated the relevance of infection as a risk factor. A scheduled admission to hospital may be postponed when another disease intervenes. However, none of the control subjects had postponed an earlier date of admission. All control subjects who were asked agreed to participate. Thus, our results were not biased by a lower participation rate in subjects with recent or present infection, a problem that studies with population control subjects may have. A variety of infectious diseases underly seasonal variations. Therefore, we aimed to investigate a control subject shortly after the respective patient. The time interval within the case-control pairs was somewhat longer than desirable. Among all bimonthly periods, the prevalence of infection in the control group was highest in February and March when recruitment of patients had been stopped and only control subjects were examined. Therefore, the delayed examination of control subjects most likely led to an underestimation of the role of infection as a risk factor.

Infection in patients mostly affected the respiratory tract and was not very severe. Most patients with infection still had residual signs or symptoms on admission. Thus, in most patients, the diagnosis of infection was not only based on history but also supported by clinical evidence. Our finding of a close temporal relationship between infection and cerebral ischemia is in accordance with previous studies.^6,16,17 In more than half of our patients with infection, we identified microbial agents that were likely to be causative. This further underlines that the diagnosis of infection was well based. In previous studies, only bacterial infection was associated with an increased risk for stroke.^5,6 Case reports and smaller studies had described an association between childhood stroke and infection due to different viruses.^4,18-20 This is the first case-control study showing that viral infection also contributes to an increased risk. The extension of serologic tests including CF tests probably contributed to the higher proportion of detected viruses as compared with our previous study.⁶ We identified a variety of presumably causative microbes. This suggests that the inflammatory host response during infection rather than microbial agents themselves contribute to the pathogenesis of infection-associated cerebral ischemia. Pierre Marie and Sigmund Freud had already argued against the idea of a "bacille spéciale," a specific micobial agent causing acute childhood hemiplegia.^1,2

For more than a century, the association between stroke and infection was recognized only in children and adolescents in whom established vascular risk factors are uncommon. The studies indicating that acute infection is a stroke risk factor also in adults leave the question unresolved whether the risk conferred by infection may be higher in younger than in older subjects. In this study we found a significant inverse relationship between age and risk by infection. The reason of such age dependence of infection as a risk factor remains unknown. This and previous studies show that approximately one third of the cases of ischemic stroke in children and younger adults are associated with recent infection.^1,4,5,21 Acute infection is therefore among the most important risk factors for stroke in young persons. The prevalence of infection in patients older than 50 was almost identical in this (17.7%) and our previous study (18.5%). The result of the present study in which infection was not a significant risk factor in the older age group was mainly due to a higher rate of infection among older hospital control subjects. Although less important than among younger subjects, acute infection may still be a relevant stroke risk factor in the elderly.

In univariate analysis, our study identified several diseases that are known to be stroke risk factors such as atrial fibrillation, hypertension, diabetes mellitus, and a history of stroke or other vascular diseases. The hospital control group in the present study and a population control group from our area in a recent study⁶ had similar prevalences of hypertension (35.5% versus 35.0%), diabetes mellitus (12.0% versus 11.7%), and previous cerebrovascular ischemia (9.0% versus 6.6%). Age and sex distribution in both control groups was similar. This indicates that our hospital control group is representative for the population of our area with respect to above factors. Atrial fibrillation is among the most important stroke risk factors²²; however, our study probably overestimates the relevance of this factor and the CI is extremely wide. Smoking was not found to be an important risk factor in our study. A high prevalence of current smoking in our hospital control group (28.5% versus 18.3% in a population control group⁶) probably contributed to this result. Because smoking is associated with an increased susceptibility to infection, it was included in our multiple logistic regression analysis together with those factors that were significant in univariate analysis. In age-adjusted multivariate analysis, infection remained significantly associated with cerebrovascular ischemia, indicating that acute infection is an independent risk factor.

Most patients with recent infection also had established stroke risk factors. This suggests that in most cases, infection temporarily increased a preexisting risk for stroke. Recent infection increased the risk for cardioembolism and tended to elevate the risk for arterioarterial embolism. Studies on childhood stroke indicated that infection may trigger cerebral vasculitis,^23,24 and in previous studies, we found that cervical artery dissection may be frequently associated with recent infection.^7,25 Therefore, infection appears to increase the risk for several etiologic stroke subtypes.

Various inflammatory pathways may activate coagulation and induce a procoagulant state during and after infection, and this may be one important pathogenetic pathway in infection-associated stroke.^26-28 In subtypes of patients, we investigated several coagulant and inflammatory variables. Sex, age, risk factors, and etiologic subtypes of stroke, the severity of cerebral tissue damage, and time after ischemia may influence such parameters. Therefore, we matched patients with and patients without infection for above variables. We investigated the fibrin generation and degradation products, fibrin monomer and D-dimer, and the activity of coagulation factors of the intrinsic (factor VIII) and extrinsic pathway(factor VII) of coagulation. Inhibition of the anticoagulant protein C/protein S system may link infection and thrombosis.²⁷ The protein S binding protein C4b-BP is an acute phase reactant. An elevation of C4b-BP during infection may result in a lower concentration of free and active protein S that serves as a cofactor of activated protein C. A recent study indicated that levels of activated protein C may be reduced in infection-associated stroke.¹⁰ Anticardiolipin antibodies that inhibit the protein C system are often increased during infection,²⁹ but their role in infection-associated stroke is controversial.^9,30 IL-1 and tumor necrosis factor-α activate the endothelium and stimulate its procoagulant activity.^26,31 We measured the antagonists of both cytokines, IL-1RA and sTNF-RI, which are more stable than the agonists and are usually increased when the agonists are activated.^32,33 Neopterin levels reflect the activation of monocytes that can develop considerable procoagulant capacity.³⁴ Leukocytes play an important role in the pathophysiology of stroke,³⁵ and IL-8 may be an important activator of leukocyte function after ischemia.³⁶ Leukocyte count was higher in patients with than in those without recent infection; however, none of the parameters assessed in patient subgroups showed significant differences between patients with and without infection. One of the major problems investigating the pathogenesis of infection-associated stroke is that cerebral ischemia is itself associated with activated coagulant and inflammatory pathways.^35,37-39 Thus, stimulation of coagulation and inflammation during preceding infection may be masked by later ischemia-associated processes. Furthermore, infectious diseases are heterogeneous and there may exist multiple pathways linking infection and stroke that may be recognized in studies of single cases⁴⁰ but not when larger groups with different mechanisms are analyzed.

This study showed that recent bacterial and viral infection and in particular respiratory tract infection is an important risk factor for ischemic stroke. Infection as a risk factor may contribute to our understanding of the seasonal variation of stroke incidence that shows a peak during winter months when infection is common.⁸ Furthermore, epidemiologic studies found an association between epidemics of respiratory tract infection and the death rate from cardiovascular diseases.^41,42 Better control of infectious disease may have contributed to the decline of stroke during this century. Preventive strategies may consider acute infection as one of the treatable predictors of ischemic stroke.