The Association Between Hay Fever and Stroke in a Cohort of Middle Aged and Elderly Adults

Study Population

We undertook an analysis of the Atherosclerosis Risk in Communities Study (ARIC) public use data set, a prospective cohort performed in Washington County, Maryland; Forsyth County, North Carolina; Jackson, Mississippi; and the suburbs of Minneapolis, Minnesota.⁸ The original cohort consisted of 15,792 participants who were enrolled in the ARIC study between 1987 and 1989. At the time of their first visit, the participants were between the ages of 45 and 64. Monitoring has consisted of periodic history and physical examination, annual telephone calls, and hospital discharge and death certificate surveillance.

For the purposes of this study, we performed an analysis using information from the participants' fourth visit (1996 to 1998; visit 4) as our baseline because this is the only visit in which participants were questioned about a history of hay fever. Participants were followed through 31 December 2001. We excluded patients with a history of stroke or coronary heart disease because of their high risk of stroke.² Because asthma and hay fever seem to have a similar pathophysiology, we initially performed an analysis that excluded asthma participants. We then performed an analysis that included participants with asthma to determine whether asthma may be an effect modifier or a confounding variable.

Definition of Hay Fever

For purposes of our study, hay fever was defined as a report by a participant of ever having been told by a physician that he or she had hay fever. We relied on participants' self-reporting of hay fever because formal allergy skin testing was not available in the ARIC dataset. We used the term hay fever rather than allergic rhinitis because it was the term used by the ARIC interviewers. Because of the seasonal nature of hay fever, we included participants who ever reported hay fever rather than only participants who were suffering with hay fever at the time of visit 4.

Determination of Stroke Events

The methods used in the ARIC study for evaluating stroke have been extensively outlined by Rosamond et al.⁹ Three methods were used to determine the incidence of stroke. First, yearly interviews were conducted with participants by telephone. Second, the records of local hospitals were reviewed for admissions of study participants during the previous year. Third, death certificates were reviewed from the vital statistics offices in the states involved in the study. In the analysis of hospital records, strokes were identified based on ICD-9 discharge codes 430 through 438, which encompass the different stroke subtypes. In addition, strokes were identified if there was a mention of stroke in a radiologic report. This information was abstracted by a trained nurse who reviewed the hospital discharge records. Stroke cases were initially classified by a computer algorithm and then were confirmed by an expert reviewer based on criteria taken from the National Survey of Stroke.¹⁰ Discrepancies between the computer and the expert reviewer were settled by another expert reviewer. Strokes were then further stratified as “possible,” “definite,” or “probable” based on radiologic imaging and autopsy, if available.¹¹ We included stroke events classified as either definite or probable. Strokes were also defined as ischemic or hemorrhagic but, because of the small number of hemorrhagic strokes (17 out of 125), we used both ischemic and hemorrhagic strokes for our stroke events without differentiating between the 2 types. We also performed an analysis of ischemic strokes alone.

Control Variables

Several variables have been identified as risk factors for stroke and thus may be potential confounders of any relationship between hay fever and stroke. These include sex, race, smoking, body mass index, diabetes, hypertension, alcohol consumption, asthma, and hyperlipidemia.^2,6,12–17

Analysis

We used the statistical package SAS (Statistical Analysis System, Cary, NC) for all analyses. We used χ² tests to examine, for each of the control variables, the bivariate relationships between participants with and without a history of hay fever. We computed an unadjusted Cox regression to determine the relationship between hay fever and risk of future stroke. Two multivariate Cox regressions were performed: one with adjustment for age and smoking because those 2 variables were significant in bivariate analysis; and one including all pre-specified control variables. In addition to our primary analysis we conducted a further analysis to determine whether participants with a history of hay fever and antihistamine usage, a possible proxy for severity of hay fever, were at greater risk of stroke than participants with a history of hay fever who did not use antihistamines. We determined if the participants used antihistamines based on the medical history obtained during visit 4.