Cigarette Smoking A Risk Factor for Cerebral Infarction in Betsy B. Love, MD; Jos\l=e'\Biller, MD; Michael \s=b\ To assess the
impact of cigarette stroke in young adults (15 to 45 years old), we compared smoking data from 181 patients with cerebral infarction with that of 307 control subjects matched for age, gender, geographic location, and hospital admission dates. While controlling for these matching variables and hypertension, an analysis based on a conditional logistic regression model indicated that a smoker was 1.6 times more likely to have a cerebral infarction than a nonsmoker (95% confidence interval, 1.07 to 2.42). There was a cumulative dose effect with each additional pack-year causing a greater risk of having a cerebral infarction. In fact, after adjusting for all other risk factors, there was a significant quadratic component to the dose-response relationships, with the result that individuals with a larger number of pack-years were invariably the stroke patients. There was no significant difference in smoking status among the various subtypes of cerebral infarction (atherosclerotic, nonatherosclerotic vasculopathy, cardioembolic, smoking
on
hematologic
related,
undetermined). These data indicate that cigarette smoking is an important risk factor for cerebral infarction in young adults. Risk factor modification through cessation of smoking may reduce the risk of ischemic stroke in young adults. (Arch Neurol. 1990;47:693-698)
Accepted for publication
December 22, 1989. From the Division of Cerebrovascular Diseases, Department of Neurology (Drs Love, Biller, Adams, and Bruno) and Division of Biostatistics, Department of Preventive Medicine (Dr Jones), University of Iowa College of Medicine, Iowa City. Presented in part at the meeting of the World Federation of Neurology, Research Group on
Neuroepidemiology, Cincinnati, Ohio, April 22, 1988.
Reprint requests to the Division of Cerebrovascular Diseases, Department of Neurology, University of Iowa Hospitals and Clinics, Iowa City, IA 52242 (Dr Love).
P.
Young Adults
Jones, PhD; Harold P. Adams, Jr, MD; Askiel Bruno, MD
"\7[7"hile cigarette smoking has been ' associated with coronary artery disease in many studies, only recently has its role in cerebrovascular disease been more firmly established. In the Framingham study,1 prospective data accumulated over more than three de¬ cades demonstrated a strong relation¬ ship between cigarette smoking and all types of stroke. The increased stroke risk among cigarette smokers was dose related and independent of arterial hypertension. Another recent prospec¬ tive cohort study found a significantly increased risk of stroke among young and middle-aged female cigarette smokers.2 Due to the importance of identifying modifiable risk factors for cerebral infarction in young adults, we studied smoking habits in patients aged 15 to 45 years old with cerebral infarction to determine whether ciga¬ rette smoking was a risk factor in this '
age group. MATERIALS AND METHODS
Study Population and Design We reviewed our experience with indi¬ viduals entered in our ongoing prospective registry of cerebral infarction in young adults aged 15 to 45 years old. These pa¬ tients were hospitalized at University of Iowa Hospitals, Iowa City, between July 1977 and December 1987. All patients were evaluated by the investigators. Risk factors for cerebral infarction were reviewed. Causes of cerebral infarction were pursued by cardiac, hématologie, and artériographie studies. The cause of cerebral infarction was classified as atherosclerotic, nonatherosclerotic vasculopathy, cardioembolic, hématologie related, or undetermined based on clinical and paraclinical investi¬ gation. Atherosclerosis was the presumed cause of cerebral infarction if the patient had two or more established risk factors of atherosclerosis in the absence of any other identifiable causes. If the results of the in¬ vestigation were inconclusive or the cause
could not be established, we labeled it as cerebral infarction of undetermined cause. We focused on cigarette smoking as a risk factor, collecting information regarding current and prior smoking status, number of cigarettes smoked daily, total number of years smoked, and pack-years (number of packs smoked daily multiplied by the num¬ ber of years smoked). Patients were cate¬ gorized as smokers, nonsmokers, or former smokers. Smokers were defined as those currently smoking any number of ciga¬ rettes on at least a weekly basis. Nonsmok¬ ers did not use tobacco at the time of exam¬ ination or in any consistent manner in the past. Former smokers had ceased smoking at the time of examination, but had used tobacco in the past. Each case was compared with one or two computer-generated hospitalized control subjects that were matched for age (±5 years and between 15 and 45 years old), gender, hospital admission dates (±2 years), and county of residence. All control subjects were hospitalized in either the neurology or internal medicine services. Control subjects were without known cere¬ brovascular disease, coronary artery dis¬ ease, or other smoking-related illnesses (including any type of malignancy of the
oropharynx, larynx, lung, esophagus, stom¬ ach, bladder, and pancreas; bronchitis, chronic obstructive pulmonary disease, peptic ulcer disease, or peripheral vascular disease). Smoking status of the control subjects was noted by chart review. The control subjects were classified in the same manner as
described for the
cases
into the
categories of smoker, nonsmoker, or former smoker. If cigarette-smoking data were not available for a control subject, another randomly selected, matched control subject was
used.
Statistical Analysis
The data
were
aggregated into matched
sets, each set consisting of
a case
and 1 to
individually matched control subjects. Analysis was performed using conditional logistic regression,3 which modeled the odds of having a cerebral infarction as a function of risk factors. The pool of potential risk 2
Downloaded From: http://archneur.jamanetwork.com/ by a University of California - San Diego User on 03/13/2016
factors included arterial hypertension, dia¬ betes mellitus, cigarette-smoking status, number of cigarettes smoked per day, packyears, and pack-years squared. Interactions among these variables and interactions of these variables with the matching variables of gender and age were also considered. Pack-years and squared pack-years were highly correlated, indicating the term squared pack-years contained linear as well as quadratic information about smoking history. To alleviate this problem, packyears was centered by subtracting the average number of pack-years. The square of this variable was used to assess whether stroke incidence was quadratic in centered
pack-years.
In the first phase of the analysis, each of the potential risk factors was individually investigated for its association with stroke incidence. Second, an extensive search was made to find the multivariable regression model that best described the data. "Best" was defined as the model with maximum log-likelihood3 and with each variable being either significant at the .1 level or involved in a significant interaction. Third, regres¬ sion diagnostics for conditional logistic regression" were used to determine if the final results remained unchanged when any potential outlier was removed. It should be noted that in conditional logistic regres¬ sion, a case is only compared with his/her own matched control subject(s). This com-
parison is made
on the measured charac¬ teristics or variables. Since the case and matched control subject(s) are identical on the matching variables of age and sex, one cannot assess the possibility of these being risk factors, and so these terms are not in¬ cluded as main effects in the regression model. However, they can appear in inter¬ action terms with nonmatching variables, such as age X hypertension, since these fac¬ tors may differ between case and matched control subject(s).
RESULTS
There were 181 patients with cere¬ bral infarction and 307 matched con¬ trol subj ects. Among the cases, 96 were men (53% ) and 85 were women (47% ), with a mean age of 34.5 years. The control group consisted of 159 men (52%) and 148 women (48%), with a mean age of 34.4 years. Most cerebral infarctions were sec¬ ondary to atherosclerosis, nonathero¬ sclerotic vasculopathies, or embolism of cardiac origin (Fig 1). Despite ex¬ tensive investigation, 12% had cere¬ bral infarction of undetermined cause (Fig 1). Table 1 lists the smoking sta¬ tus of the cases according to presumed cause of cerebral infarction. Although there was a seemingly greater percent¬ age of smokers in the undetermined category, this difference was not sta¬
tistically significant. A comparison between the smoking status of the cases and control subjects
Fig 1.—Cause
is illustrated in Table 2. The data in those figures do not reflect matched
of cerebral infarction in 181 young adults.
analysis.
Table 1.
Stroke
Subtype
and
Smoking Status
—
Percent
Stroke Subtype Atherosclerotic
Nonatherosclerotic Cardioembolic
No.
Smoker
49
53
39
54
Nonsmoker
Former Smoker
vasculopathy 46
Hématologie Undetermined
Table 2. Smoking Status of Cases and Control Subjects
The set of potential risk factors including hypertension, diabetes, smoking status, number of cigarettes smoked per day, and centered packyears were individually assessed for their associations with case/control status. The separate odds ratios, values, and 95% confidence intervals are given in Table 3. Smokers were 1.43 times more likely to have a cerebral infarction than nonsmokers, a margin-
Table 3.—Univariate Estimates of Relative Risk for Cerebral Infarction*
—
No. of Percent
All cases All control
subjects Men Cases Control
subjects Women Cases Control
subjects
Smoker
Nonsmoker
smoker
50
47
3
43
51
6
Ex-
54
41
5
48
47
5
45
54
1
40
55
5
95% Confidence
Complete
Variable_Matched Setst_Odds Ratio_ _Interval Hypertension_180_2.900_