High Prevalence of Stroke Symptoms Among Persons Without a Diagnosis of Stroke or Transient Ischemic Attack in a General Population:  The REasons for Geographic And Racial Differences in Stroke (REGARDS) Study FREE

A substantial portion of the population has clinically undiagnosed stroke. Silent infarctions have been documented by magnetic resonance imaging (MRI) in 11% of community-dwelling adults aged 55 to 64 years in the Atherosclerosis Risk in Communities (ARIC) cohort¹; this incidence increased from 22% at the age of 65 to 69 years to 43% for individuals older than 85 years in the Cardiovascular Health Study.² In both studies, MRI-detected silent infarctions were in individuals who had not been told by a physician that they had had a stroke or transient ischemic attack (TIA), but these same individuals could have experienced stroke symptoms that failed to reach a threshold required for clinical diagnosis. It is possible that many subclinical strokes are nonetheless symptomatic.

Evaluating persons with symptoms of stroke or TIA is an important part of stroke prevention owing to the high risk of stroke.^3- 5 Such evaluation requires that the public and health care professionals recognize and initiate prompt action when major stroke symptoms are reported. A 1999 survey by the National Stroke Association found that 2.3% of a random sample of 10 112 US residents had been told by a physician that they had had a TIA, but only 64% had seen a physician within 24 hours of the event.⁶ An additional 3.2% reported TIA symptoms but had not sought medical attention.⁶ A report on 61 019 adults surveyed across 17 states and the US Virgin Islands found that only 17.2% could correctly classify all stroke symptoms and would call 911 if they thought someone was having a stroke.⁷

Few studies have looked at the association of stroke symptoms with stroke risk factors in a general population. The ARIC study reported a higher prevalence of risk factors, including diabetes mellitus, current smoking, and hypertension, among those with stroke symptoms.⁸ The REasons for Geographic And Racial Differences in Stroke (REGARDS) study provides an opportunity to assess self-reported stroke symptoms and the association of these symptoms with well-established stroke risk factors in a national population sample that included a large African American population.

Beginning in January 2003, a national cohort of community-dwelling individuals older than 45 years were recruited, with approximately equal representation of white and African American individuals and men and women.⁹ Twenty percent of the sample was randomly selected from the “buckle” of the stroke belt (coastal plain region of North Carolina, South Carolina, and Georgia), 30% from the stroke belt states (remainder of North Carolina, South Carolina, and Georgia plus Alabama, Mississippi, Tennessee, Arkansas, and Louisiana), and the remaining 50% from the other 42 contiguous states. Individuals were identified from commercially available lists of residents and recruited using an initial mailing followed by telephone contact. Using a computer-assisted telephone interview, trained interviewers obtained demographic information, medical history, and indexes of quality of life. Consent was obtained verbally and later in writing. A brief physical examination, including blood pressure measurements, blood samples, and an electrocardiogram (ECG), was conducted in person 3 to 4 weeks after the telephone interview. The institutional review boards of the University of Alabama at Birmingham, University of Vermont and State Agricultural College (Burlington), Wake Forest University School of Medicine (Winston-Salem, NC), Alabama Neurological Institute (Brookwood Medical Center), and University of Arkansas for Medical Sciences approved the study methods. Additional methodological details are provided elsewhere.⁹

Participants were asked if a health care professional had ever told them that they had had a stroke or TIA and whether they had ever experienced the sudden onset of each of 6 stroke symptoms using the Questionnaire for Verifying Stroke-Free Status (QVSFS) (Table 1).^10- 12 The QVSFS is a validated questionnaire proposed as a quick screening instrument for identification of stroke-free individuals in a general population. This instrument was shown to be sensitive (sensitivity, 0.97), although specificity is only modest (specificity, 0.60).¹¹ Prevalent stroke symptoms were defined as answering yes to 1 or more symptoms vs no to all symptoms.

The Framingham Stroke Risk Score (FSRS), which estimates the 10-year probability of stroke given the participant's demographic and risk factor profile, was used as a summary index of the stroke risk factor burden for each participant. This risk score includes age, sex, systolic blood pressure, use of antihypertensive medications, current smoking status, history of heart disease, diabetes, left ventricular hypertrophy, and atrial fibrillation.^13- 14 Systolic blood pressure was defined as the average of 2 measurements obtained by a trained technician using a standard protocol and regularly tested aneroid sphygmomanometer, measured after the participant was seated for 5 minutes. History of heart disease was defined as any self-reported myocardial infarction or “heart attack,” coronary artery bypass surgery, coronary angioplasty or stenting, or evidence of myocardial infarction from ECG. Diabetes was defined as a fasting glucose level greater than 126 mg/dL (7.0 mmol/L), nonfasting glucose level greater than 200 mg/dL (11.1 mmol/L), or self-reported medication use for glucose control. Left ventricular hypertrophy was defined by centrally adjudicated ECG. Atrial fibrillation was assessed by self-report or the study ECG. Age, race, sex, use of antihypertensive therapy, educational level, annual family income, health insurance, and perceived health status were defined on the basis of interview data.

Between January 25, 2003, and November 30, 2005, 191 028 telephone numbers called to recruit participants had a final disposition. Defined according to standards recommended by Morton et al,¹⁵ the response rate (percentage agreeing to be interviewed among known eligible candidates contacted plus an adjustment for estimate of likely eligible participants among unknown eligible participants) was 40.7% (31 556/77 526). At the time of this analysis, 20 677 participants completed both the telephone survey and in-person component; in-person examinations were pending for 2314, and 8565 participants could not or did not complete the in-home examination. Participants who self-reported either stroke or TIA were excluded (2215), resulting in an analysis cohort of 18 462 participants.

Logistic regression was used to establish the association of prevalent stroke symptoms with age, sex, race, region (stroke belt, stroke buckle, or other), and the FSRS. Along with these factors, the potential confounders of educational level, annual family income, health insurance, and perceived health status were evaluated using multivariate analysis. The FSRS includes items that reflect both age and traditional stroke risk factors. Because we anticipated that the prevalence of stroke symptoms would increase with age, it would not have been clear whether an association between stroke symptoms and the FSRS were attributable to an association with age or to the other risk factors that are included in the FSRS. In multivariate analysis, the association with the FSRS was assessed after adjustment for age, allowing these associations to be attributable to the risk factors independent of age (since association with age was removed by covariate adjustment).

Table 2 gives the baseline characteristics, stroke risk factors, and FSRSs. A total of 7567 (41.0%) of the participants were African American, 9463 (51.3%) were women, and 6534 (35.4%) resided in the stroke belt. The overall mean (SD) age was 65.8 (9.0) years, with a range of 45 to 96 years.

Table 3 provides a description of the prevalence of each of the stroke symptoms overall and by race and sex strata, age strata, and region. One or more stroke symptoms were reported by 3292 (17.8%) of 18 462 participants,with the most common symptom being “a sudden numbness or a dead feeling on one side of the body” (8.5%) and the least common being “suddenly lost ability to understand people” (2.7%). There were statistically significant differences in the frequency of the symptoms by race and sex group for all of the symptoms except “sudden painless loss of vision,” with higher rates generally observed for African Americans. The proportion of participants with the symptoms did not markedly increase with age; however, the proportion with symptoms was significantly higher among those in the stroke belt for 2 of the 6 symptoms. Table 4 provides the distribution of the total number of symptoms, with 11.0% having 1 symptom, 5.0% having 2, 1.0% having 3, and less than 1% having 4 or more symptoms.

Results of the univariate and multivariate logistic regression models predicting 1 or more symptoms are provided in Table 5. In the univariate model, there was a higher likelihood of the presence of symptoms among participants with higher FSRSs, a relationship that was maintained in the multivariate model. Additionally, the odds of 1 or more symptoms was 1.6 times greater among African American than white participants, a statistically significant difference that was only partially mediated by the adjustment for other factors in the multivariate model. In both the univariate and multivariate models, the presence of symptoms was not significantly different between the stroke belt and the rest of the United States. The odds of 1 or more symptoms was 1.2 times greater among women than men in the univariate model; however, multivariate adjustment mediated this difference. The likelihood of symptoms was not associated with age in the univariate model but was lower among older participants in the multivariate model (notably the FSRS, which also includes age in the formulation). In univariate analysis, the prevalence of 1 or more stroke symptoms was more common among participants in lower socioeconomic status strata, as indexed by an income of less than $25 000, educational level of less than a high school diploma, or the absence of health insurance. In multivariate analysis, this association persisted for both income and educational level but not for health insurance. In both analyses, those with fair or poor perceived general health were more likely to have 1 or more stroke symptoms than those with excellent, very good, or good perceived general health.

The prevalence of stroke symptoms in this national sample of a general stroke- and TIA-free population was high, with almost 18% reporting at least 1 symptom. Stroke symptoms were more likely among African American compared with white individuals, those with lower income and less education, those with fair to poor perceived health status, and those with higher FSRSs. The last finding suggests that at least some of these symptoms may represent stroke events that did not reach the threshold required for clinical diagnosis. This theory is supported by other studies that reported a high prevalence of cerebral infarctions according to MRI in those without recognized stroke.^1,16- 17 These undiagnosed or unrecognized events could have a substantial impact on cognitive functioning or personality^18- 20 and could also be powerful harbingers of subsequent major strokes.^21- 22

As we hypothesized a priori, the magnitude of the racial difference in prevalence of stroke symptoms was mediated by multivariate adjustment; however, significantly higher odds of stroke symptoms among African Americans persisted. Unmeasured aspects of socioeconomic status could be contributing to the lack of a complete mediation of the racial differences. Data on socioeconomic status are limited in the REGARDS study, where some measures recommended by Braveman et al,²³ including past socioeconomic experiences and occupation, are not available. The recent report by Avendano et al²⁴ of an elderly cohort in New Haven, Conn, showed that in individuals younger than 75 years there was an increased stroke risk for those with lower income and educational levels, but in those 75 years or older this relationship reversed. In the REGARDS cohort 83% were younger than 75 years, and similar to the findings of Avendano et al, we observed a higher prevalence of stroke symptoms among those with low income and low educational level. Lower perceived general health is a strong predictor of adverse health outcomes^25- 27 and was the most potent correlate of stroke symptoms in REGARDS, where participants with fair to poor perceived health were more likely to report stroke symptoms compared with those with excellent to good perceived health.

Few studies have evaluated symptoms of stroke in an epidemiologic setting. The only other population-based, longitudinal cohort studies to assess stroke or TIA symptoms are the ARIC study²⁸ and the Renfrew/Paisley Study,²⁹ which evaluated a cohort of residents aged 45 to 64 years from the towns of Renfrew and Paisley in Scotland. The ARIC questionnaire asked whether the participant had ever had a sudden episode that involved any of the following 6 neurologic symptoms: speech dysfunction, loss of vision, double vision, weakness or paralysis, numbness or tingling, and dizziness or loss of balance. Of 12 205 participants, 47% reported the sudden onset of at least 1 stroke symptom during their lifetime.³⁰ In the Renfrew/Paisley Study, 15 113 participants were asked if they had ever experienced, without warning, any of the following 4 symptoms, assumed to be indicative of TIA: suddenly lost power of an arm, suddenly lost power of a leg, suddenly been unable to speak properly, or suddenly lost consciousness.²⁹Table 6 contrasts the prevalence of stroke and TIA symptoms in these 3 studies. Although the prevalence of several symptoms was comparable across studies, differences in the wording of some questions make comparisons problematic. Each study asked about speech problems, with 3.8% of REGARDS participants reporting a sudden loss of ability to express yourself verbally or in writing and 2.6% of ARIC participants reporting a speech dysfunction compared with only 1.8% of the Renfrew/Paisley Study participants reporting the inability to speak properly. The prevalence of weakness was 5.8% in REGARDS, weakness or paralysis prevalence was 2.3% in ARIC, and the prevalence of lost power of an arm or leg was 2.3% and 2.4%, respectively, in the Renfrew/Paisley Study; however, REGARDS specified unilateral symptoms and did not ask about paralysis, whereas ARIC did not restrict symptoms to unilateral events and included paralysis. Both REGARDS and ARIC asked about numbness, reporting an 8.5% prevalence in REGARDS and about twice that in ARIC; however, REGARDS specified unilateral symptoms, whereas ARIC did not. In addition, REGARDS asked about “numbness or a dead feeling,” whereas ARIC asked about “numbness or tingling.” Tingling is a symptom that connotes a peripheral neuropathic process rather than focal cerebral ischemia. Both ARIC and REGARDS asked about visual symptoms; however, the format of these questions makes the results difficult to contrast. Additionally, the presence of 1 or more symptoms was higher in ARIC, largely from the question asking about dizziness (not asked in REGARDS or the Renfrew/Paisley Study); however, it is uncertain whether this finding represents a true difference between the populations or differences in the questions.

The ARIC investigators have further classified symptoms into TIA or stroke or not, based on a validated TIA or stroke algorithm, and found a prevalence of 12.9%.³⁰ Although the same algorithm cannot be applied to REGARDS because the stroke symptom questions were not the same, it can be projected that some of the REGARDS participants who reported at least 1 stroke symptom either had a mild stroke or TIA that was not diagnosed or they do not recall the physician diagnosis. The strong association with the FSRS supports the theory that these are likely to be stroke events in at least a subset of the population with symptoms.

Certainly, many of the symptoms assessed in this study do not reflect stroke. For example, in REGARDS, during follow-up, the same QVSFS is administered semiannually, with the retrieval of medical records for any participant who answers positively to any of the 6 symptom questions and reports going to a physician or being hospitalized for the symptom. During the tracking process, many nonstroke causes for these symptoms were identified. For “sudden painless weakness” or “sudden numbness or a dead feeling,” some participants have potential alternative explanations for the symptoms, such as surgery or a condition related to the hip or spine. For “sudden painless loss of vision,” some participants reported ocular conditions or procedures that could explain the symptoms, including cataracts, glaucoma, eye surgery, and eye examinations. This experience is consistent with prior validation studies of the QVSFS that showed it to have high sensitivity but modest specificity with regard to detecting stroke symptoms.¹²

It is not known if these REGARDS participants consulted a health care professional for evaluation of symptoms only to have the symptoms judged to be of insufficient magnitude for a clinical diagnosis. Many people do not seek medical attention for stroke symptoms, or if they do, they do not seek it immediately. In addition to studies already mentioned,^6- 7 this finding was documented by the Rochester Epidemiologic Project, reporting that after TIA, multiple recent events had gone unreported by patients.³¹ This lack of attention to medical care could be partially attributed to a low level of knowledge of the symptoms of stroke. A 2003 national random sample of 1024 women 25 years and older found that correct identification of the warning signs of stroke was low across all racial/ethnic and age groups; knowledge of warning signs ranged from 2% to 8% for the symptom of disorientation to 11% to 23% for sudden dimness or loss of vision to 29% to 39% for sudden unilateral weakness or numbness of the face or limb.³²

Strengths of this work include the use of a national general population sample with a large African American population; however, it is always possible that those choosing to participate in any study could be nonrepresentative of the general population. An unknown number of eligible individuals may have been missed because of the inability to determine eligibility in some of the households contacted. A potential bias also exists in that approximately 27% of the participants who initially consented by telephone could not or did not continue to the in-person examination. Although this would affect the estimate of the prevalence of stroke symptoms, it plays a minor role in the association of symptoms with stroke risk factors. The use of the QVSFS, a validated instrument for the identification of stroke-free individuals, is a strength of the study. Although we can only speculate that some proportion of those reporting symptoms have actually experienced a stroke, this link will be strengthened in the future as neurologist-adjudicated incident stroke events become available through the follow-up of the REGARDS cohort. In ARIC, participants with these self-reported stroke symptoms had a 2.8-fold greater risk of a subsequent clinically diagnosed stroke.³³ Similarly, in the Renfrew/Paisley Study, men with 2 or more stroke symptoms had a 2.3-fold increased risk for clinically diagnosed stroke, whereas women had a 2.4-fold increased risk.²⁹ We have also assumed that the FSRS is a valid index of stroke risk in a biracial national cohort. Although the Framingham Coronary Heart Disease Risk Score has been validated in multiethnic groups,³⁴ no similar validation has been performed for the FSRS.

When evaluating a person who screens positively for stroke symptoms but is not otherwise known to have had either stroke or TIA, it may be appropriate to evaluate the person with head imaging for evidence of prior cerebral infarction or an alternative origin for the symptoms. If stroke mimics are excluded, physicians may wish to control vascular risk factors in accordance with secondary stroke prevention guidelines and even consider screening for symptomatic cervical carotid stenosis that might warrant revascularization.³⁵ However, as yet no evidence exists that screening for stroke symptoms is associated with improved stroke or other outcomes on a population basis.

In this article, we have demonstrated a strong association between the presence of stroke symptoms and the stroke risk factor burden as indexed by the FSRS. Coupled with previous reports showing a substantially increased risk of a subsequent stroke in those with stroke symptoms and a substantial prevalence of silent stroke according to MRI, our findings suggest that these commonly reported symptoms may be mild strokes that failed to reach the threshold for clinical diagnosis. Targeted education on the warning signs of stroke and risk factor reduction efforts for individuals who report stroke symptoms may be helpful in improving early recognition and in the prevention of stroke.

Correspondence: Virginia J. Howard, MSPH, Department of Epidemiology, University of Alabama at Birmingham School of Public Health, 1665 University Blvd, Birmingham, AL 35294-0022 (vjhoward@uab.edu).

Accepted for Publication: June 7, 2006.

Author Contributions: Ms Howard and Dr McClure had access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Howard. Acquisition of data: Howard, McClure, and Pulley. Analysis and interpretation of data: Howard, McClure, Meschia, and Friday. Drafting of the manuscript: Howard, McClure, and Meschia. Critical revision of the manuscript for important intellectual content: Howard, McClure, Meschia, Pulley, Orr, and Friday. Statistical analysis: McClure. Obtained funding: Howard. Administrative, technical, and material support: Howard, Meschia, and Pulley. Study supervision: Howard.

Financial Disclosure: None reported.

Funding/Support: This research project was supported by cooperative agreement U01 NS041588 from the National Institute of Neurological Disorders and Stroke, National Institutes of Health, Department of Health and Human Services.

Role of the Sponsor: Representatives of the funding agency were involved in the review and approval of the manuscript but not directly involved in the collection, management, analysis, or interpretation of the data.