Aspirin and Ticlopidine for Prevention of Recurrent Stroke in Black Patients: Title and subTitle BreakA Randomized Trial FREE

Blacks are disproportionately affected by stroke, yet they have been underrepresented in clinical trials.^{1 - 8} Recommendations for stroke prevention in this population have been based largely on trials that have included few black participants. This may not be an optimal practice because blacks are among those with a higher prevalence of major cardiovascular risk factors, a different distribution of atherosclerotic occlusive cerebral vascular lesions, vascular biological differences such as low renin hypertension, and a different pattern of use of medical procedures and access to care^{8 - 13} that could influence outcome.

A subgroup analysis of the Ticlopidine Aspirin Stroke Study (TASS)^{14 - 15} suggested a more favorable risk-benefit profile for nonwhites than whites. Specifically, among the 495 black and 108 nonwhite and nonblack study participants, there was a 24.1% relative risk reduction (RRR) for stroke and death at 2 years favoring ticlopidine (500 mg/d) over aspirin (1300 mg/d), and 10% fewer serious adverse events (SAEs).¹⁵ Overall in TASS, there was a 12% RRR for nonfatal stroke or death from any cause (P = .05) favoring ticlopidine at 3 years.

The current study was designed in 1993, with the belief that a targeted recurrent stroke prevention study for blacks was justified given their disproportionate stroke burden, promising data for ticlopidine as a recurrent stroke preventive treatment in nonwhites, and the lack of previous substantial representation of blacks in stroke clinical trials. The primary outcome of the African American Antiplatelet Stroke Prevention Study (AAASPS) was the composite end point of recurrent stroke, myocardial infarction, or vascular death.

A description of the design and methods of AAASPS has been reported previously¹⁶ in accordance with criteria proposed by the Consolidated Standards of Reporting Trials.^{17 - 18} That article¹⁶ included a discussion of barriers to black participation in clinical trials and how they were overcome,^{12 ,19 - 20} the rationale for study drug selection, relationships established with primary care physicians, management of cardiovascular risk factors, and other major aspects of the study. The diagnosis of stroke and stroke subtype was determined after review of source documents and case report forms and by application of criteria from the Trial of ORG 10172 in Acute Stroke Treatment (TOAST)²¹ by local principal investigators for the entry stroke and by the AAASPS adjudication committee for all outcome events. Entry and outcome stroke cases received computed tomography or magnetic resonance imaging of the head.

At the time our study design was developed in the early to mid-1990s, we believed that there was uncertainty about the preferred aspirin dose for recurrent stroke prevention.²² Given this uncertainty, we opted for an aspirin dose of 650 mg/d in accordance with the recommendations by Barnett et al.²²

Eligibility criteria included black (African American) race; 29-85 years of age inclusive; noncardioembolic ischemic stroke with onset at least 7 days but not more than 90 days; cranial computed tomographic scan or magnetic resonance image of the brain consistent with occurrence of the entry cerebral infarct; measurable neurological deficit that correlates at onset with entry cerebral infarct; informed consent; and availablity of patient to be followed up in an outpatient treatment program.¹⁶

Exclusion criteria included transient ischemic attack, subarachnoid hemorrhage, cardiac source embolism, iatrogenic stroke, nonatherosclerotic stroke, postoperative stroke occurring within 30 days of operation, or carotid endarterectomy as primary treatment measure for entry cerebral infarct; mean arterial blood pressure higher than 130 mm Hg on 3 consecutive days; modified Barthel index of less than 10; history of dementia or neurodegenerative disease; severe comorbid condition (eg, cancer) judged to limit survival during 2-year follow-up; enrollment in another clinical trial; allergy or sensitivity to study drugs; woman of childbearing potential; gastrointestinal tract bleeding, bleeding diathesis, or platelet or other hematologic abnormality (judged to be a contraindication for administration of study drugs) currently active or clinically active in the past year; hematuria or positive stool guaiac test related to major bleeding source; and prolonged prothrombin time or partial thromboplastin time, blood urea nitrogen level higher than 40 mg/dL, serum creatinine level higher than 2.0 mg/dL (176.8 µmol/L), thrombocytopenia or neutropenia defined by the lower limit of normal for the platelet count or white blood cell count (unless absolute neutrophil count of at least 1800/mm³), or liver function tests 2 or more times the upper limit of normal. All sites had to receive formal approval from their institutional review board before the study could commence at a local site.

The primary hypothesis of this randomized, double-blind, investigator-initiated clinical trial was that ticlopidine (500 mg/d) was more effective than aspirin (650 mg/d) in preventing the composite outcome of recurrent stroke, myocardial infarction, or vascular death (death due to ischemic or hemorrhagic stroke, myocardial infarction, sudden death, pulmonary embolism, heart failure, visceral or limb infarction, or a vascular procedure or operation) among blacks with noncardioembolic ischemic stroke who were followed up for up to 2 years. The prespecified secondary hypotheses were that the incidence of the outcome end points of recurrent stroke or death; nonfatal or fatal stroke; recurrent stroke, myocardial infarction, or death from all causes; vascular death; death from all causes; or myocardial infarction would be lower in ticlopidine-treated patients compared with aspirin-treated patients followed up for up to 2 years.

The ratio of those receiving ticlopidine and aspirin was 1:1, and the sequence was stratified by site to balance the treatment groups. All study personnel were masked (blinded) from treatment assignment with the exception of 1 study statistician who developed the randomization algorithm. After written informed consent was obtained, local study site personnel called a dedicated automated telephone registration system for AAASPS, which was operated by the Moffitt Cancer Research Institute at the University of South Florida in Tampa, to register a study participant.²³

Ticlopidine was packaged as 250-mg tablets and aspirin as 325-mg coated tablets. The placebo tablets had identical physical properties. Study medications were dispensed from plastic bottles. Medication compliance was determined by pill count at each follow-up visit. A dose of 250 mg of ticlopidine was administered with a placebo aspirin tablet twice a day with meals. A dose of 325 mg of aspirin was administered with a placebo ticlopidine tablet twice a day with meals.

Before entry into the study, at 12 months, 24 months, and at any time a study participant experienced an outcome event or terminated from the trial, the following laboratory studies were obtained: complete blood count and platelet count; blood urea nitrogen; serum creatinine; total cholesterol, low-, high- and very low-density lipoprotein cholesterol, and triglycerides; total bilirubin, alanine aminotransferase, aspartate aminotransferase, lactate dehydrogenase, and alkaline phosphatase; serum glucose and electrolytes; and urinalysis. Complete blood count and platelet count were performed every 2 weeks during the first 3 months of the study or at any unscheduled time the local investigative team deemed that it was indicated.

Study participants were examined in person at baseline; every 2 weeks during the first 3 months; and at 6, 10, 12, 16, 20, and 24 months; and at any unscheduled time the investigative team deemed that it was indicated for reason of patient safety, medication compliance, or the occurrence of outcome events or SAEs. Telephone contact was made during study months for which patients did not have an in-person examination to screen for medication compliance, outcome events, and SAEs. Because blacks are more likely to experience cyclical neutropenia,¹⁵ and the incidence of severe neutropenia with ticlopidine use is uncommon (<1%),¹⁵ we believe that study personnel remained blinded in relation to laboratory data.

Before entry into the trial, acute hospital care and stroke diagnostic evaluations were at the discretion of the local principal investigator or primary care physician. Study patients received verbal instruction and written materials about stroke prevention, and a booklet about the AAASPS program that was developed in collaboration with the AAASPS community advisory board. This booklet also listed medications that were to be avoided or were contraindicated. Study patients were educated about possible SAEs of the study medications and given contact information to use in the event questions or problems arose. For study patients who reached an outcome end point, local principal investigators were given the option to treat these patients with open-label aspirin or to transition patients to community care for stroke prevention according to their primary care physicians.

The AAASPS included several mechanisms to ensure patient safety: a predetermined laboratory "panic" value system whereby the main laboratory notified the local investigative team and the clinical safety monitor of a critical value; an internal AAASPS inhouse safety committee; and an external data safety and monitoring board that was appointed by the National Institutes of Health.

Data from blacks and nonwhites in TASS^{14 - 15} were used to develop sample size estimates for AAASPS. We anticipated that AAASPS patients would generally be in poorer health than those in TASS. A 2-year event rate of 25% was projected for the AAASPS aspirin treatment group. A total of 1410 patients would be required to achieve 80% power to detect a 25% RRR with respect to the 2-year primary event rate based on use of a 2-tailed log-rank test with an overall α level of .05. Allowing for an increase of 15% for attrition due to voluntary withdrawal and loss to follow-up, and an additional 7% to allow for interim analyses using the Lan-DeMets strategy with O'Brien-Fleming boundaries, a total sample size of 1800 was estimated to yield 306 primary outcome events.^{24 - 25} Interim analyses were performed at information times of 15%, 23%, 35%, 67%, and 76%, after which the blinded treatment phase of the study was terminated.

All analyses were conducted using SAS (SAS Institute Inc, Cary, NC) and S-Plus (Mathsoft, Cambridge, MA) statistical software according to the intention-to-treat principle and were based on available patient data through March 31, 2002. Categorical values were compared using χ² tests and the Fisher exact test. Continuous data were compared using Wilcoxon and t tests. Time-to-event curves were calculated using the Kaplan-Meier method and were compared between treatment groups using the log-rank test. Cox proportional hazards modeling was also used to compare treatment groups while adjusting for various covariates such as age, sex, and entry stroke subtype and severity.²⁶ Secondary end points, as well as prespecified subgroups (sex, entry stroke subtype, and stroke severity, and various categories of elapsed time between onset of the index stroke and start of treatment), were analyzed using similar methods. Study patients who dropped out of the study were censored at the time of their last visit.

Fultility analyses were performed and presented to the data safety and monitoring board (DSMB) appointed by the National Institutes of Health (NIH) as part of study interim analysis reports. These analyses were conducted to assess the likelihood of observing a statistically significant result if the trial was to continue to full information (the observance of 306 primary outcome events). The method of futility analysis was that of conditional power, as defined by the conditional probability of rejecting the hypothesis of no treatment difference at the end of the trial given current interim data. This method was chosen before the trial began. The conditional probability was computed by simulating the number and timing of future events based on the current event rates. Simulations were performed using S-Plus. A conditional power close to zero indicated little chance of crossing the statistical boundary were the trial to go to completion.

The main AAASPS organizational components are listed at the end of the article, with a listing of local sites, the investigative team at each local site, and the number of study patients enrolled at each site. Also, we performed routine audits at 27 centers that were our larger enrollment centers, and at 4 centers for cause (3 sites for possible difficulty with data collection and follow-up procedures, and at 1 site for possibly entering a participant for whom several key data points did not match the source documentation). In all audits, all issues were resolved, and no major action was required to be taken by the AAASPS clinical or data management centers.

Recruitment commenced on December 12, 1995, and was completed on October 1, 2001. A total of 1809 patients were enrolled in the study; 902 in the ticlopidine group and 907 in the aspirin group. The number of patients enrolled by study year was 354 in 1996, which included December 12, 1995 through December 31, 1996; 326 in 1997; 315 in 1998; 303 in 1999; 314 in 2000; and 197 from January 1 through October 1, 2001.

Futility analyses were computed as part of the statistical analyses presented to the NIH-appointed DSMB as part of study interim analysis reports. The latest such report was based on available patient data through March 31, 2002, and suggested that there was less than a 1% chance that ticlopidine-treated patients would have a superior outcome compared with aspirin-treated patients if the trial were to continue to completion. After careful review of this report and consultation with study staff, the study scientific advisory committee, and other external experts, the DSMB recommended that the blinded phase of the study end on July 15, 2002. At that juncture, study participants were given the option of remaining in the study taking study-sponsored open-label aspirin or transition into the community for stroke prevention therapy according to their community physician.

Figure 1 summarizes the status of patients in the trial by treatment group. Three hundred seventy participants (41.0%) in the ticlopidine group and 403 participants (44.4%) in the aspirin group completed the 24-month examination. In general, all types of withdrawal were slightly more common in patients receiving ticlopidine, but a statistically significant difference was observed for withdrawal due to SAEs (15.1% for ticlopidine vs 11.5% for aspirin; P = .02). The time-to-occurrence of lost to follow-up and voluntary withdrawal did not differ significantly between treatment groups (P>.25 for each comparison).

At the time the blinded phase of the study was halted, excluding patients with outcome events, 47.1% of the ticlopidine group and 46.5% of the aspirin group had not completed the 2-year follow-up, and 149 participants remained in the ticlopidine group and 176 in the aspirin group (P = .11). Overall, the study database contained an average of 1.54 years of follow-up data per study patient. Median follow-up was 710 days for ticlopidine-treated patients and 716 days for aspirin-treated patients. On average, study participants completed 88% of study visits.

Baseline characteristics of patients in the 2 treatment groups are balanced and are summarized in Table 1. Blacks have a higher risk of lacunar infarction than whites.⁸ Furthermore, our eligibility criteria excluded those with atrial fibrillation, cardiac sources of embolism that would require warfarin therapy, or large artery carotid occlusive disease treated by carotid endarterectomy, which would serve to increase the likelihood of enrolling lacunar infarction patients into the study. Drug compliance was assessed by pill count. Overall, median compliance was 91%; 90% in the ticlopidine group and 92% in the aspirin group.

A total of 245 primary outcomes (the composite of recurrent stroke, myocardial infarction, or vascular death) occurred during the trial (Table 2). There were 133 primary outcomes among the ticlopidine-treated patients and 112 among the aspirin-treated patients. As shown by Kaplan-Meier curves in Figure 2, there was no statistically significant difference between treatment groups in the occurrence of the primary outcome end point (P = .12 by log-rank test; the O'Brien-Fleming boundaries were not crossed). Cox proportional hazards modeling that included key covariates, such as treatment group assignment, age, and baseline risk factors, showed a similar treatment difference (P = .11).

There were slightly more fatal (4 vs 2) or nonfatal (102 vs 84) recurrent strokes among ticlopidine-treated patients, but this was not statistically significant (Table 2). However, the Kaplan-Meier curves indicate time to fatal or nonfatal stroke (Figure 2) approached a statistically significant difference (P = .08 by log-rank test) in favor of the aspirin-treated patients.

Secondary outcomes including any recurrent stroke, death (all cause), vascular death, recurrent stroke or death (all cause), and the composite end point of recurrent stroke, myocardial infarction, or death (all cause) were not statistically significantly different between treatment groups (Table 2). Overall, we observed a 2-year primary event rate of 19.7% among ticlopidine-treated patients and 16.3% among those treated with aspirin.

We compared recurrent stroke subtype and severity between treatment groups and did not find a statistically significant difference (Table 3). Also, we extended the follow-up data for the primary outcome cluster by adding information from April to December 2002. There were 5 additional events: 3 strokes and 1 myocardial infarction in the ticlopidine group and 1 stroke in the aspirin group. The primary outcome findings, however, were not substantially changed (P = .08 by log-rank test).

In addition to the intention-to-treat analysis, we performed an on-treatment analysis for the primary event outcome cluster. For each patient, we computed the percentage of time on blinded study medication prior to the occurrence of an outcome event or completion of the study (whichever came first). We then performed log-rank tests on subgroups of patients based on this percentage (ie, those having a percentage on blinded study medication (≥60% or ≥80%). We observed P = .10 for the 60% cutoff and P = .11 for the 80% cutoff compared with P = .12 overall. Thus, we did not find a statistically significant difference based on intention-to-treat or on-treatment analyses.

Data on patients reporting SAEs appear in Table 4. Overall, 532 patients reported SAEs: 270 (29.9%) among ticlopidine-treated patients and 262 (28.9%) among aspirin-treated patients. Diarrhea (0.3% vs 0.2%), neutropenia (3.4% vs 2.2%), and thrombocytopenia (0.3% vs 0.2%) were slightly more frequent among ticlopidine-treated patients, but this did not reach statistical significance. One case of thrombocytopenia, which occurred early after the enrollment phase of the study was initiated, was diagnosed as a possible case of thrombotic thrombocytopenic purpura. The patient recovered after treatment with plasmapheresis. Gastrointestinal tract hemorrhage was slightly more common in aspirin-treated patients (0.9% vs 0.4%), but this did not reach statistical significance. The time-to-occurrence of a SAE did not differ significantly between the treatment groups (P = .39), but the time-to-occurrence of a SAE did (P = .003) (not shown in Table 4).

For study patients who had to permanently discontinue blind phase medication due to a SAE (not shown in Table 4), thrombocytopenia (0.2% vs 0.0%), neutropenia (1.8% vs 1.2%), and rash (1.7% vs 0.6%) were more common among ticlopidine-treated patients, but only the occurrence of rash differed significantly between the treatment groups (P = .02). Gastrointestinal tract hemorrhage requiring premature discontinuation of study medication (not shown in Table 4) was more common among aspirin-treated patients but did not reach statistical significance (0.6% vs 0.2%; P = .45).

Ticlopidine, a unique inhibitor of platelet aggregation,^{27 - 28} was approved for clinical use in the early 1990s by the US Food and Drug Administration to reduce the risk of fatal or nonfatal thrombotic stroke in patients with stroke precursors and in patients who have had a completed thrombotic stroke. Two large clinical trials, TASS¹⁴ and the Canadian American Ticlopidine Study (CATS),²⁹ showed efficacy for ticlopidine in stroke prevention. The CATS intention-to-treat analysis showed a RRR of 23.3% (P = .02) for the composite outcome of stroke, myocardial infarction, or vascular death favoring ticlopidine over placebo. Reversible severe neutropenia and reversible severe skin rash occurred in ticlopidine-treated patients in about 1% of cases and diarrhea in 2% of cases. In TASS, the RRR favoring ticlopidine over aspirin at 3 years for the primary end point of nonfatal stroke or death from any cause was 12% (95% confidence interval [CI], −2% to 26%) with a marginally significant difference in primary events (P = .048). In addition, there was an RRR of 21% (95% CI, 4%-38%; P = .02) for fatal or nonfatal stroke favoring ticlopidine at 3 years, and ticlopidine was more effective than aspirin in both men and women. Diarrhea (20%), skin rash (14%), and severe but reversible neutropenia (<1%) occurred in patients receiving ticlopidine.

Efficacy data from our trial of black patients with noncardioembolic ischemic stroke did not show a reduction of the composite outcome of recurrent stroke, myocardial infarction, or vascular death among ticlopidine-treated patients (P = .12 by log-rank test). The outcome of fatal or nonfatal stroke approached a statistically significant difference (P = .08 by log-rank test) favoring the aspirin treatment group.

The blinded phase of our study was halted by the data and safety monitoring board appointed by the National Institutes of Health after the recruitment and follow-up phases of the study had been ongoing for about 6.5 years because futility analyses indicated a less than 1% chance of ticlopidine being significantly better than aspirin therapy in the prevention of our primary outcome if the trial were to continue to completion. These analyses also indicated a 40% to 50% likelihood of aspirin being significantly better than ticlopidine in reducing the risk of recurrent fatal or nonfatal stroke if the trial were to continue to completion. The decision of the data and safety monitoring board to stop the study was based on the potential futility of ticlopidine use for the primary study outcome end point and the small likelihood, but potential for SAEs among ticlopidine-treated patients.

Based on data from a subanalysis of nonwhite patients in TASS,¹⁵ we believed that there was a substantial likelihood that ticlopidine would be more effective than aspirin in reducing vascular events among a high-risk stroke population. We hypothesized that ticlopidine, a more global platelet inhibitor, would be more effective in black ischemic stroke patients with a substantial stroke and cardiovascular disease risk profile and that aspirin might prove to be an inferior platelet inhibitor in these high-risk patients. In retrospect, we are somewhat surprised by our results because TASS showed a modest reduction in the primary outcome of stroke or all-cause death favoring ticlopidine over aspirin.¹⁴ Furthermore, clopidogrel, a compound in the same drug class as ticlopidine, has also been studied. The Clopidogrel versus Aspirin in Patients at Risk of Ischemic Events (CAPRIE) study showed a modest benefit with 75 mg/d of clopidogrel compared with 325 mg/d of aspirin for the reduction of recurrent stroke, myocardial infarction, or vascular death among patients who entered with either recent ischemic stroke, recent myocardial infarction, or symptomatic peripheral arterial disease.³⁰ In a CAPRIE subanalysis of 6431 patients with stroke, there was only a nonstatistically significant reduction of ischemic stroke, myocardial infarction, or vascular death favoring clopidogrel (RRR, 7.3%; 95% CI, −5.7% to 18.7%; P = .26). However, the CAPRIE subanalysis of stroke patients³⁰ and the TASS subanalysis of nonwhites¹⁵ were not adequately powered statistically to be conclusive.

We are uncertain why ticlopidine was not superior to aspirin in reducing major vascular events in our study patients. We used a lower aspirin dose than in TASS,^{14 - 15} but we can only speculate whether this or some other biological difference in response to therapy might explain our findings. Overall, there is a relative paucity of scientific information about biological differences in drug response by race or ethnic group.³¹ We are not aware of such published information for aspirin and ticlopidine in blacks.

The drop-out rate in our study was generally higher than most recurrent stroke prevention trials. In the CAPRIE study, for example, the number of study participants withdrawn or lost to follow-up was 22 (0.2%) of 9553 among those receiving clopidogrel and 20 (0.2%) of 9546 among those receiving aspirin.³⁰ In our study, the corresponding figures for lost to follow-up or voluntary withdrawal were 15.2% in the ticlopidine treatment group and 13.3% for those receiving aspirin. Most of these were due to voluntary withdrawal, which we defined as the patient deciding to discontinue participation unrelated to an AE but rather to personal preference. We believe that our higher drop-out rate can be attributed to study participants of generally lower socioeconomic status, who have traditionally had less access to medical care, and less involvement in clinical trials.^{8 - 9 ,12 ,20} In an AAASPS exploratory study, we found that voluntary withdrawal occurred for such reasons as distrust of the medical establishment, fear of continuation in a blinded trial, and pressure from family members, primary caregivers, or primary care physicians.²⁰

We believe that our drop-out rate did not limit our ability to detect a treatment effect because we anticipated this level of attrition and accounted for it in our sample size calculation. In addition, we conducted both intention-to-treat and on-treatment analyses and did not find a statistically significant difference between treatment groups for the primary event outcome cluster. Therefore, it is unlikely that our findings are due to a switch to open-label aspirin therapy during the course of the trial thereby diluting the ticlopidine treatment group. We note that the number of primary outcomes was less than the projected number of 306, which was due to premature stoppage of the study. However, had the study continued to completion, we estimated the occurrence of at least 310 outcome events. Thus, the nonborderline nature of the study results, a futility analysis that showed a less than 1% chance of ticlopidine performing better than aspirin, and loss of information that did not occur substantially more in one treatment group than the other, makes it unlikely that loss of information significantly affected our results.

Our recurrent primary event rates were lower than expected. For example, the expected 2-year event rate was projected to be 25% in the aspirin group but was observed to be 16.3%. Our rate, however, for stroke and death is similar to that of the recently completed Warfarin Aspirin Recurrent Stroke Study (WARSS).³² WARSS had a preponderance of lacunar infarctions at entry (56%), similar to AAASPS (67.5%), and the rate of recurrent stroke and death in WARSS was 16.0% at 2 years for the aspirin group. The corresponding rate for our aspirin group, derived by taking the 1-year rate from our aspirin study participants because most had completed 1 year of follow-up, and multiplying this estimate by a factor of 2, was 16.1%; a figure quite similar to that of WARSS. Our study was not designed to determine if the lower-than-expected rates for stroke were due to baseline characteristics in the sample, better control of risk factors, or more widespread use of concomitant therapies. We speculate, however, that this lower-than-expected rate could be related to better control of risk factors as we noted increasing use of cholesterol-lowering agents during the study period.

The SAEs with ticlopidine have been well described.^{14 ,29 ,33 - 36} Our SAE data show that slightly more occurred among ticlopidine-treated patients (29.9%) overall than aspirin-treated patients (28.9%). In addition, there were slightly more clinically reported SAEs of diarrhea in the ticlopidine-treated patients (0.3% vs 0.2%), but more clinically reported SAEs of major gastrointestinal tract bleeding among aspirin-treated patients (0.9% vs 0.4%).

We defined serious neutropenia as a laboratory-determined absolute neutrophil count of less than 1000/mm³ and serious thrombocytopenia as a laboratory-determined platelet count of less than 100 000/mm³. In either case, once the absolute neutrophil count or platelet count reached the critical prespecified level, the blinded phase study medication was discontinued. We observed a higher percentage of laboratory-determined serious neutropenia (3.4% vs 2.2%) among ticlopidine-treated patients and about the same percentage of laboratory-determined serious thrombocytopenia (0.3% vs 0.2%) among the treatment groups. All cases of neutropenia were reversible, but one case of thrombocytopenia resulted in possible thrombotic thrombocytopenic purpura in a patient who recovered after treatment with plasmapheresis.

Our data support the prior TASS subanalysis findings¹⁵ suggesting that some key AEs might be less frequent in ticlopidine-treated nonwhites. We observed lower percentages of diarrhea (8.4% in ticlopidine-treated patients vs 9.0% in TASS subanalysis vs 20.4% in TASS overall¹⁴ ) and rash (5.9% vs 9.0% in TASS subanalysis vs 11.9% in TASS overall).

Based on data from the IMS Health's National Prescription Audit Plus and National Disease and Therapeutic Index of Dispensed Prescriptions for nonaspirin antiplatelet agents used specifically or "customized" for stroke or transient ischemic attack patients in 2002, generic and branded ticlopidine was third in market share with an estimated 3.2% of dispensed oral antiplatelet prescriptions; dipyridamole was second at 17.9%; and clopidogrel was first at 78.9% (personal communication from David Milbauer, Marketing Research, Boehringer Ingelheim Pharmaceuticals Inc. Data extracted April 2003). Aspirin use is not captured in the IMS Health Index of Dispensed Prescriptions because it is an over-the-counter medication. With an estimated 4 million stroke survivors and between 600 000 and 750 000 strokes occurring in the United States annually, of which about 85% are ischemic, up to 108 800 persons in the United States could be taking ticlopidine and up to 20 400 persons could receive ticlopidine as an initial stroke prevention therapy each year. Our findings do not support the use of ticlopidine as a first-line agent for blacks for recurrent stroke prevention, provide substantial data to challenge the subanalysis of TASS that suggested benefits of ticlopidine in recurrent stroke prevention in nonwhite patients,¹⁵ and provide new data to challenge current guidelines on secondary stroke prevention.

The implications and relevance of our data as they affect other antiplatelet agents for recurrent stroke prevention in blacks remain speculative as we did not directly compare our study agents with clopidogrel or the combination aspirin plus extended-release dipyridamole. Furthermore, the clinical trials that tested these other agents had relatively few or no black enrollees.^{30 ,37} For clinicians who remain unconvinced about the efficacy or safety of these other agents, aspirin could be the preferred agent for recurrent stroke prevention in blacks. However, administration of these other agents in black patients will vary based on patient and physician preference.

Prior systematic reviews and guideline statements suggest that aspirin is of benefit in a wide range of patients with suspected acute ischemic stroke to reduce the risk of early stroke recurrence³⁸ and as an initial choice of therapy for recurrent stroke prevention.^{39 - 41} Before our study was completed, thienopyridine derivatives (ticlopidine and clopidogrel) appeared to be modestly more effective than aspirin in the prevention of serious vascular events in high-risk patients.⁴² Aspirin is much less expensive than other major antiplatelet agents, is readily available, easy to use, and relatively safe. Head-to-head comparison with other agents indicates that it may be difficult to outperform aspirin as a stroke prevention therapy in some noncardioembolic ischemic stroke patients.^{31 - 32 ,43 - 44}

Our data call into question the superiority of the thienopyridine ticlopidine in black noncardioembolic ischemic stroke patients, and suggest that ticlopidine is unlikely to be superior to aspirin for prevention of recurrent stroke and major vascular events in these patients. Furthermore, ticlopidine may have a less favorable and potentially SAE profile. Therefore, aspirin is a reasonable first choice prevention agent in aspirin-tolerant black patients with noncardioembolic ischemic stroke.