Because cervical auscultation is still a widely used and simple clinical method for identifying persons with asymptomatic carotid disease, we initiated the Asymptomatic Cervical Bruit Study in 1988 [9] to explore the natural history of asymptomatic patients with varying degrees of carotid stenosis. We also sought to test the effectiveness of aspirin in higher-risk patients with stenosis of at least 50% in reducing the risk for cerebral and cardiac ischemic events and to assess the effect of aspirin on the rate of progression of atherosclerotic carotid disease. Although some researchers [13, 14] have suggested that aspirin be considered as a therapeutic option for these patients, the effectiveness of the drug has never been adequately tested.

We report the results of the randomized controlled trial portion of our study in which we tested the effectiveness of aspirin in patients with a high degree of carotid stenosis. The design and operation of the study have already been reported in detail [9].

Methods

Top Methods Results Discussion Author & Article Info References

Study Organization

The Asymptomatic Cervical Bruit Study included six participating hospitals, a clinical coordinating office, a methodologic and statistical center, and two committees: one for centrally adjudicating events and the other for monitoring safety. The clinical coordinating office, located at the Montreal General Hospital, was responsible for the overall daily conduct of the study, including the handling of all protocol-related matters such as the completeness of the study documentation. The methodologic and statistical center was the Division of Clinical Epidemiology at the Montreal General Hospital; its responsibilities included quality assurance and storage and analysis of data. The central adjudication committee verified patient eligibility and conducted blinded review of all outcome events reported in the study. The monitoring committee, composed of members not involved in the clinical activities of the study, was responsible for monitoring safety and applying the rules for stopping the study drug on the basis of the planned interim analyses. The members of this committee had access to the randomization codes.

Patient Eligibility

Eligible participants were neurologically asymptomatic patients with an audible cervical bruit in whom duplex ultrasonography indicated the presence, in at least one artery, of a carotid lesion that reduced the diameter of the artery by at least 50%. Sources of referred patients included family practice units and internists and general practitioners from the community and hospital-based subspecialty clinics; all sources had been informed of the study and of our interest in assessing patients with asymptomatic cervical bruits.

Although age alone was not an exclusion criterion, patients were excluded if they had a history of symptomatic ischemic cerebrovascular disease, valvular heart disease other than mitral valve prolapse, nonvalvular atrial fibrillation, recent (< 3 months before study entry) myocardial infarction or unstable angina, previous carotid endarterectomy, medically necessary use of aspirin or regular use of nonsteroidal anti-inflammatory drugs, use of anticoagulant agents, life expectancy of less than 5 years, and allergy to or intolerance of aspirin compounds. All patients gave informed consent, and the ethics committee of all participating institutions approved the study.

Treatment Regimens

Aspirin was supplied as 325-mg enteric coated tablets in plastic bottles that contained enough tablets for 6 months (approximately 200 tablets). The placebo tablets were identical in appearance and packaging. Eligible patients were allocated to receive either one aspirin or placebo tablet per day on the basis of a centrally determined blocked randomization arrangement. This arrangement ensured the equal distribution of treatment regimens within each clinical center. The clinical nurse coordinator verified compliance by counting pills at each follow-up visit. Patients were instructed not to take any aspirin-containing compounds and to notify the study personnel of any change in their therapeutic regimen. For safety purposes, the treatment codes were only available centrally to the monitoring committee and locally to the pharmacist-in-chief of each institution.

Baseline and Follow-up Evaluation

The initial clinical assessment was done by a qualified neurologist and included a complete medical history with emphasis on neurologic aspects; a general physical examination; and a complete neurologic and vascular examination. Data were initially collected on the presence of vascular risk factors such as hypertension (defined by history or a blood pressure of at least 160/95 mm Hg); diabetes (defined by history or a fasting blood glucose level of at least 150 mg/dL [more than equals 8.2 mmol/L]); coronary artery disease (defined by history or evidence on initial electrocardiogram of previous myocardial infarction); hyperlipidemia (defined by history or a total cholesterol level of at least 250 mg/dL [6.5 mmol/L]); history of intermittent claudication or peripheral vascular surgery; and smoking status. Baseline laboratory assessment included a complete blood count; blood chemistry studies, including blood sugar and lipid profile; electrocardiogram; and duplex examination of the neck. We have previously shown that duplex ultrasonography is highly accurate compared with angiography [9]. Patients were scheduled for repeat clinical and ultrasonographic assessments at 6-month intervals. At each follow-up visit, the nurse coordinator and attending neurologist recorded information on the occurrence and timing of any neurologic or cardiac events, new medical conditions, adverse events, and compliance with all prescribed medications, including the study regimen.

Outcome Events

The intention-to-treat main analysis was based on the first event in the composite end point, which consisted of transient ischemic attack, stroke, myocardial infarction, unstable angina, or death. We also did secondary analyses using the following combinations of outcomes: 1) transient ischemic attack, stroke, myocardial infarction, unstable angina, and death from vascular causes; 2) stroke, myocardial infarction, and death from vascular causes; 3) transient ischemic attack and stroke; 4) stroke and death from vascular causes; and 5) myocardial infarction, unstable angina, and death from vascular causes. We used specific criteria to define the outcomes of interest [9, 15, 16]. The type of stroke was documented by computed tomographic scan or magnetic resonance imaging of the brain, and stroke severity was assessed by the Barthel Index [17]. All patients who survived their initial outcome event were followed until the end of the study.

The central adjudication committee reviewed all outcome events using all available information (hospital records, death certificates, and necropsy reports). The committee was blinded to treatment and to the degree of carotid stenosis. If committee members disagreed, they submitted the relevant data to independent external reviewers.

Statistical Planning

On the basis of the literature review, we initially assumed that the annual rate of the composite outcomes in the placebo group would be 17% [18]. To ensure an 80% power to detect a 40% risk reduction with aspirin (one-tailed test; equals 0.05), 242 patients had to be recruited into the study over 2.5 years and followed for an additional 3 years [19]. We chose a one-tailed test because we expected a priori that aspirin would reduce the risk for clinical events.

Before we compared the two treatment groups, however, we realized that the overall event rate was lower than anticipated. We therefore had to increase our sample size to 350 patients by extending the recruitment period. We planned three interim analyses and based the rules for stopping the study on the group sequential testing method [20].

Statistical Analyses

We used descriptive statistics to compare the baseline distribution of the demographic variables and cardiovascular risk factors of the two groups. We used the Fisher exact test and chi-square tests to assess the differences in the distribution of the reasons for early discontinuation of therapy with the study medication and the differences in the proportion of patients whose stenosis progressed. The distribution of stroke severity, measured by the Barthel Index, was compared using the Fisher exact test.

The intention-to-treat main analysis was based on a composite of four nonfatal end points (transient ischemic attack, stroke, myocardial infarction, and unstable angina) and death. We analyzed the time to the first end point using methods for survival data to handle right censoring and differences in the duration of follow-up. We used Kaplan-Meier nonparametric estimates to compare the probability of event-free survival in the two groups. We estimated annual event rates as the ratio of the number of events to the total number of patient-years of follow-up; this rate corresponded to the maximum likelihood hazard rate estimated in the exponential model.

Hypothesis testing was based on the Cox proportional-hazards model, with a significance level of 0.05 [21]. Given the low incidence of events, the hazard ratio (aspirin-placebo) is a good approximation of the risk ratio. Univariate Cox analyses provided P values equivalent to the log-rank testing and crude estimates of the hazard ratio with corresponding 95% CIs. Multivariate Cox analyses allowed us to account for possible imbalances in the distribution of known cardiovascular risk factors in the two arms when testing the effect of aspirin and provided the adjusted hazard ratio estimates. In all multivariable Cox models, the estimated effect of aspirin has been adjusted for age and all binary risk factors that indicate the presence or absence of the specific conditions or characteristics listed in Table 1. We chose risk factors on a priori grounds; these factors did not depend on the significance of particular risk factors in our analyses. For a patient with no history of a given condition and for whom the measurement of a corresponding continuous risk factor was missing, the absence of the respective binary risk factor was assumed in Table 1 and for the purpose of multivariable analyses. To examine the possibility that the effectiveness of aspirin varies according to the sex of the patient, we added sex-treatment interaction to all proportional hazards models and tested the significance of the interaction. We also did separate analyses using data for men only and for women only.

We did separate subgroup analyses for patients with stenosis of less than 80% and those with stenosis of 80% or more using Bonferroni correction to account for multiple comparisons. The significance of the differences between the effect of aspirin in the two categories was assessed by testing the interaction between treatment and the degree of stenosis in the Cox model. We used the efficacy approach to repeat all analyses; with this approach, patients were censored when therapy with the study medication was permanently discontinued.

To investigate the extent to which the therapeutic benefit of aspirin decreases with an increasing duration of follow-up, we used a generalization of the Cox model in which the constant hazard ratio is replaced by a flexible function of time [22]. We modeled a time-varying hazard ratio using a computer program written in the C language (unpublished manuscript [a copy of this paper and details of the computer program are available from the authors]). The regression spline method was used to estimate the pattern of changes in the hazard ratio over time and to assess the significance of these changes using the likelihood ratio test [23].

Results

Top Methods Results Discussion Author & Article Info References

Patient Screening and Entry

Of the 1995 patients screened, 1621 (81.4%) did not meet the inclusion criteria. The main reasons for exclusion were a history of neurologic symptoms (27.3%), medically necessary use of aspirin or other antiplatelet medications (26.7%), and carotid stenosis of less than 50% (26.2%). Thus, 374 patients were enrolled in the randomized controlled trial over 6 years (1988 to 1994) from six hospitals. Two of these patients, both in the aspirin group, were later found to be ineligible by the eligibility and adjudication committee.

Patient Characteristics at Entry

Of the remaining 372 patients, 188 were randomly assigned to receive aspirin and 184 to receive placebo. Fifty-three percent of patients were female (mean age, 66.7 years). Both groups had similar demographic characteristics, vascular risk factors, and severity of carotid stenosis (Table 1).

Adherence to Protocol

The mean duration of follow-up was 2.4 years (median, 2.3 years), and the mean duration of therapy with study medication was 1.9 years for the placebo group and 2.0 years for the aspirin group. Only two patients (both in the placebo group) were lost to follow-up, after 1.5 and 2.7 years in the study, respectively. The treatment code was broken in 11 instances, always at the request of the treating neurologist. Seven of these 11 patients were in the placebo group, and 4 were in the aspirin group. The treatment code was broken because of the onset of ischemic neurologic symptoms in 9 patients and because of gastric bleeding in 2 patients. Three hundred eighteen patients (85.4%) fully complied with the study protocol, 162 in the placebo group and 156 in the aspirin group. Compliance was established by pill counts at regular follow-up visits. The 54 patients who temporarily discontinued therapy with study medication were compliant for an average of 83.4% of the duration of follow-up. The proportion of patients with early permanent discontinuation of study therapy was 55.9% in the placebo group and 52.1% in the aspirin group (Table 2). Excluding the occurrence of the end points, the other most frequent reasons were the need for other treatments and new medical problems.

Adverse Effects

Side effects consisted primarily of gastrointestinal disturbances. Fourteen patients reported gastrointestinal disturbances of varying degree; 6 patients had to be hospitalized. Two hospitalized patients (1 each in the placebo and aspirin groups) had nonfatal gastric hemorrhage that required blood transfusions. No cases of intracerebral hemorrhage were recorded during the study.

Interim Analyses

The first interim analysis was based on data collected until 31 December 1992. By that time, 283 patients had been recruited and 62 clinical end points had been observed, 30 in the placebo group and 32 in the aspirin group. The results suggested that aspirin did not decrease the risk for adverse outcomes (crude hazard ratio [aspirin-placebo], 1.04 [P = 0.88]). Projections of the final results, which were based on computer simulations, indicated that aspirin would not yield a significant risk reduction. However, we decided that to provide a conclusive negative result, the 95% CI for the hazard ratio should exclude the relative risk reduction of 40% or more. Computer simulations indicated that 2 additional years of recruitment would be necessary to ensure a probability of more than 80% that the final analysis would yield a conclusive result. Another interim analysis was planned for the end of April 1994. The trial was stopped on the basis of the results of this analysis.

Final Analyses

The distribution of the clinical end points and the corresponding estimated annual event rates for the two groups are shown in Table 3. The annual hazard rates for all clinical end points combined were 12.3% in the placebo group and 11.0% in the aspirin group.

The results of the intention-to-treat Cox proportional-hazards analyses of the effect of aspirin on the risk for composite outcomes are as follows: In the univariate analysis, the hazard ratio was 0.904 (CI, 0.615 to 1.330; P = 0.61); in the multivariate analysis, the hazard ratio was 0.988 (CI, 0.667 to 1.464; P = 0.95). The results for ischemic events only were as follows: In the univariate analysis, the hazard ratio was 1.000 (CI, 0.672 to 1.493; P = 0.99); in the multivariate analysis, the hazard ratio was 1.079 (CI, 0.718 to 1.621; P = 0.71). All comparisons indicate that the two groups did not significantly differ, a conclusion reflected by the high P values and hazard ratios close to 1.0. This conclusion remained valid when we controlled for known vascular risk factors. The nonsignificant results were replicated in the efficacy analyses (all P values greater than 0.5).

The 95% CI for the crude hazard ratio (intention-to-treat analysis) ranged from a 38% relative risk reduction with aspirin (hazard ratio, 0.62) to a risk increase of 34%. The results thus allow us to rule out, with 95% confidence, the possibility that aspirin may reduce the risk for the composite outcomes by 40% or more. After adjustment for known vascular risk factors in the multivariate Cox model, the lower boundary of the CI is higher than 0.65, which excludes the relative risk reduction of 35% or more.

According to subgroup analyses, aspirin had no significant effects in patients with stenosis of either less or more than 80%; in addition, the treatment-by-stenosis interaction was not significant. We found no evidence of the effect of aspirin on the risks for different combinations of end points in any of the secondary analyses. In addition, no difference in the effectiveness of aspirin between men and women has been found because the sex and treatment interaction has not been significant in all proportional hazards models. Moreover, in all separate sex-specific analyses for either men or women, the effect of aspirin was not significant. The severity of strokes as measured by the Barthel Index did not significantly differ between the aspirin and placebo groups (P = 0.64).

Figure 1 shows the Kaplan-Meier estimates of the probability of event-free survival in the two trial arms, for the composite outcomes with deaths from nonvascular causes included and excluded, respectively. Whereas the estimates for the two groups are similar in the later phase of follow-up, the probability of survival for the patients receiving aspirin during the initial phase appears to be higher than that for the patients receiving placebo. This suggests that the effect of aspirin may change over time. The likelihood ratio test, based on the regression spline generalization of the Cox model, indicates that changes in the crude and adjusted hazard ratio (aspirin-placebo) over time for the intention-to-treat analysis of the composite outcomes are statistically significant (P = 0.035 and 0.025, respectively). The solid curve in Figure 2 shows the estimated pattern of changes in the adjusted hazard ratio as a function of the duration of follow-up. In the first year of follow-up, the hazard ratio is less than 1.0, which suggests that in the initial phase of the treatment, patients receiving aspirin may be at a lower risk than patients receiving placebo. The pointwise 95% CIs for the hazard ratio Figure 2 are less than 1.0 for part of the first year, which indicates that the risk reduction with aspirin may be significant over that limited time interval. However, the estimated hazard ratio increases rapidly in the second year and for the rest of the follow-up period remains above 1.0. This finding suggests that aspirin may be associated with a slightly increased risk in the later phase of follow-up.

View larger version (16K): [in this window] [in a new window]   Figure 1. Top. Probability of event-free survival for the composite end point (all vascular events and death): Kaplan-Meier estimates (intention-to-treat analysis). Bottom. Probability of event-free survival for vascular events (nonvascular deaths excluded): Kaplan-Meier estimates (intention-to-treat analysis). Values below each graph indicate the number of patients at risk in the placebo and aspirin groups.

View larger version (45K): [in this window] [in a new window]   Figure 2. Time dependence of the effect of aspirin on the risks for the composite end point. The solid curve represents the estimated hazard ratio (aspirin-placebo), adjusted for the common risk factors, as a function of the duration of follow-up. The dotted curves indicate 95% CIs for the hazard ratio. The dashed horizontal line represents a hazard ratio of 1.0, which corresponds to an equal risk in the two groups. Areas below this line represent risk reduction due to aspirin; areas above the line represent risk increases.

Thirty-seven patients in the placebo group (20.1%) and 44 patients in the aspirin group (23.4%) progressed to a higher stenosis category. The difference in the proportions between the two treatment groups was not significant (P = 0.44).

Discussion

Top Methods Results Discussion Author & Article Info References

We found that aspirin provided no sustained protective effect in asymptomatic persons with advanced carotid disease. The inclusion of transient ischemic attacks as an end point was justified because the occurrence of these attacks increases the risk for subsequent stroke and necessitates the use of an antiplatelet agent [24, 25] or surgery [26]. On the basis of our results, we can with a high degree of confidence exclude a relative risk reduction of 35% or more for the end points of interest with the regular use of aspirin. The exclusion of this level of risk is clinically sound considering the similar degree of effectiveness of aspirin in the primary prevention of ischemic disease in patients with nonvalvular atrial fibrillation [27] and other vascular conditions [24, 25, 28] and the fact that these patients are also asymptomatic and have a clearly lower absolute risk than recently symptomatic patients with a similar degree of carotid disease [26]. Our main results are similar to those of the Veterans study [14]. However, unlike those investigators, we did not find that aspirin reduced the occurrence of transient ischemic attacks. Differences in study designs may explain this discrepancy; the Veterans study was not designed to test the effect of aspirin.

Both treatment groups had similar demographic characteristics and risk factor profiles. The adjudication process ensured the validity of the outcome data. Few patients were lost to follow-up, a noteworthy achievement given that these patients were asymptomatic and thus were less likely to be motivated for long-term follow-up in a clinical trial. The compliance with the study regimen was also good, and the long-term use of compounds with an antiplatelet effect was avoided. We agreed on the aspirin dose of 325 mg/d because it is in the midrange of what has been suggested for the prevention of ischemic events of all types [24, 25, 27, 28] and because it is less likely to cause intolerance and complications in asymptomatic older persons. Indeed, few adverse events occurred during the study. Most of these events were gastrointestinal, and hospitalization was required in only a few cases. No hemorrhagic strokes were recorded during the trial, an outcome at variance with some recent studies [29, 30] that found an increased incidence of intracranial hemorrhage or disabling strokes in asymptomatic persons receiving aspirin. This finding could be partly explained by the smaller size of our patient sample and the relatively shorter follow-up period. Furthermore, it is unlikely that hemorrhagic strokes were missed in our study because a computed tomographic scan of the brain was mandatory in all patients presenting with a neurologic event.

The efficacy of aspirin has now been well established in patients with symptomatic ischemic vascular disease [24, 25]. However, evidence of the overall protective effects of this antiplatelet agent in low-risk persons [29, 30] is not as convincing, especially the evidence on stroke and death from vascular causes. This has made it difficult to make clear recommendations about aspirin prophylaxis [31].

According to available published information [32], patients with asymptomatic carotid disease probably represent a group with an intermediate risk for vascular events; their risk for cardiac events is probably at least equal to the risk for stroke itself [5, 33]. Although recent randomized studies have addressed the efficacy of carotid endarterectomy in these patients [11], only a few reports of uncontrolled studies [14, 18, 34-36] have mentioned the putative effects of aspirin. None was specifically designed to test the effectiveness of aspirin. In addition, the Antiplatelet Trialists' Collaboration [24] did not specifically address the potential indication of aspirin in asymptomatic carotid disease because of a lack of data from randomized clinical trials. However, these investigators did report a lack of benefit in low-risk persons, a finding consistent with our results. The degree to which an over-representation of women in our group could also be partially responsible for the lack of efficacy is unclear; however, a separate analysis comparing the effect of aspirin in men and women did not suggest this possibility. For some specific end points (such as death from vascular causes), the estimated risks were higher in the aspirin group than in the placebo group. Moreover, for end points in which the difference in risk favors aspirin, the difference was minor and not statistically significant. In the latter situation, to further assess the clinical relevance of a possible therapeutic effect of aspirin, we calculated the expected number of patients who would have to be treated so that one end point would be prevented [37]. Assuming that the difference in hazard rates observed in our sample corresponds to the true difference in respective event rates in the population, we calculated that 77 patients would have to be treated for 1 year to prevent one event of the vascular or nonvascular type. However, more than 200 patients would have to be treated to prevent one stroke or death. These numbers are substantially higher than those previously reported with regard to the use of aspirin in patients with transient ischemic attacks or other vascular conditions [37].

We cannot, however, totally exclude an early transient beneficial effect of aspirin. This is interesting in light of recent reports suggesting that the effect of aspirin as an antithrombotic agent may be time-dependent. These reports include the overview by the Antiplatelet Trialists' Collaboration [24], which stated that the effectiveness of aspirin may diminish over the duration of follow-up, and other reports suggesting that the antiaggregant effect of aspirin decreases over time [38]. This phenomenon merits further corroboration in future trials.

Regardless of the ultimate conclusions about the appropriateness of carotid endarterectomy in this clinical setting, we found that aspirin had no important protective effect. This finding thus limits the role of aspirin as a viable long-term alternative in neurologically asymptomatic patients with carotid disease. This conclusion, however, may not apply to patients with concomitant active cardiac disease that requires treatment with aspirin because such patients were excluded from our trial. We also cannot exclude a possible beneficial effect from a higher dose of aspirin or from other antiplatelet agents [15, 39] that have different mechanisms of action.

Appendix

The following are committee members of the Asymptomatic Cervical Bruit Study Group. Clinical Coordinating Office: The Montreal General Hospital, Montreal, Quebec, Canada: France Bourque, Robert Cote, Jacques Leclerc, Ariane Mackey, Duncan McIlraith, and Lisa Wadup; Methodologic and Statistical Center: The Montreal General Hospital: Michal Abrahamowicz, Renaldo Battista, France Bourque, Robert Cote, Ariane Mackey, and Duncan McIlraith; Central Adjudication and Eligibility Committee: Renaldo Battista, Leo Berger, France Bourque, Robert Cote, Maurice Godin, Yves Langlois, Louise-Helene Lebrun, Jacques Leclerc, Luc Marchand, Robert Lewis, Ariane Mackey, and Duncan McIlraith; External Adjudication Committee: Joseph Bogousslavsky (Lausanne, Switzerland), Alastair Buchan (Ottawa, Ontario, Canada), and Ashfaq Shuaib (Saskatoon, Saskatchewan, Canada). Monitoring Committee: Michal Abrahamowicz, John Esdaile, and Steven Grover (Drs. Esdaile and Grover are not involved in the study in any other capacity).

The following centers and personnel participated in the study group: The Montreal General Hospital, Montreal, Quebec: Michal Abrahamowicz, PhD, Renaldo N. Battista, MD, France Bourque, RN, Patrice Bret, MD, Robert Cote, MD, Susan Joseph, RN, Jacques Leclerc, MD, Ariane Mackey, MD, Duncan McIlraith, MD, Lisa Wadup, RN, Roxanne du Berger, Xiao-Ping Hu, and Shan-Shan Wang; Hotel Dieu Hospital, Montreal, Quebec: Debra Black, MD, Vincent Lajoie, RN, Monique Lamarre, RN, Yves Langlois, MD, Brigitte Lecours, RN, Luc Marchand, MD, Aldolfo Perez de Leon, MD, Ghyslaine Roederer, MD, and Susie Roy, RN; St. Luc Hospital, Montreal, Quebec: Lucienne Bourque, RN, Marie-Paule Desrochers, RN, Nicole Daneault, MD, Louise-Helene Lebrun, MD, and Suzanne Fontaine, MD; The Jewish General Hospital, Montreal, Quebec: Joseph Carlton, MD, Normand Miller, MD, and Lisa Wadup, RN; Enfant Jesus Hospital, Quebec City, Quebec: Alice Lajeunesse, RN, Denis Simard, MD, and Claude Roberge, MD; Charles Lemoyne Hospital, Greenfield Park, Quebec: Leo Berger, MD, and Danielle Simard, RN.

Drs. Battista and Abrahamowicz: Montreal General Hospital, Division of Clinical Epidemiology, 1650 Cedar Avenue, Montreal H3G 1A4, Quebec, Canada.

Dr. Langlois: Jewish General Hospital, Division of Vascular Surgery, 3755 Cote St. Catherine Road, Montreal H3T 1E2, Quebec, Canada.

Dr. Mackey: Enfant Jesus Hospital, Division of Neurology, 1401 18th Avenue, Quebec City G1J 1Z4, Quebec, Canada.