Mexican Americans with Atrial Fibrillation have More Recurrent Strokes than Non-Hispanic Whites

Abstract

Background and Purpose

Atrial fibrillation is a common cause of stroke with a known preventative treatment. We compared post-stroke recurrence and survival in Mexican Americans (MAs) and non-Hispanic whites (NHWs) with atrial fibrillation in a population-based study.

Methods

Using surveillance methods from the Brain Attack Surveillance in Corpus Christi (BASIC) Project, cases of ischemic stroke/TIA with atrial fibrillation were prospectively identified January 2000-June 2008. Recurrent stroke and all-cause mortality were compared by ethnicity with survival analysis methods.

Results

A total of 236 patients were available (88 MA, 148 NHW). MAs were younger than NHWs, with no ethnic differences in severity of the first stroke or proportion discharged on warfarin. MAs had a higher risk of stroke recurrence than NHWs (Kaplan Meier estimates of survival free of stroke recurrence risk at 28-days and 1-year were 0.99 and 0.85 in MAs and 0.98 and 0.96 in NHWs; p=0.01, log-rank test), which persisted despite adjustment for age and gender (hazard ratio 2.46, 95% CI: 1.19, 5.11). Severity of the recurrent stroke was higher in MAs than in NHWs (p=0.02). There was no ethnic difference in survival after stroke in unadjusted analysis or after adjusting for demographics and clinical factors (hazard ratio 1.03, 95% CI: 0.63–1.67).

Conclusions

MAs with atrial fibrillation have a higher stroke recurrence risk and more severe recurrences than NHWs, but no difference in all-cause mortality. Aggressive stroke prevention measures focused on MAs are warranted.

INTRODUCTION

Atrial fibrillation (AF) is the most common sustained heart rhythm abnormality, affecting over 2.2 million adults in the United States.¹ AF is an important risk factor for both incident and recurrent stroke.² Cardioembolic stroke, most of which is due to AF, is the most lethal subtype of ischemic stroke³ and has a higher risk of disability than other stroke subtypes.⁴ Strokes due to AF are largely preventable with warfarin therapy, which results in a 64% relative risk reduction.⁵ However, the use and effectiveness of warfarin has been suggested to be suboptimal in Hispanics and other minority groups.^{6, 7}

Mexican-Americans (MAs) are the largest subgroup of Hispanic Americans, the largest minority population in the United States.⁸ MAs have a higher incidence of stroke than non-Hispanic whites (NHWs),⁹ yet have better post-stroke survival.¹⁰ However, little is known about post-stroke outcomes in MAs with AF. The objective of this study was to investigate ethnic differences in stroke recurrence and post-stroke mortality in a population-based study of individuals with AF and stroke.

METHODS

Case Identification

This study was based on the population-based Brain Attack Surveillance in Corpus Christi (BASIC) study. Detailed BASIC methods have been published previously.^{9, 11, 12} Briefly, active and passive surveillance were used to capture strokes among residents of Nueces County, Texas aged 45 and older. Cases were ascertained actively by searching admission and emergency department logs for a set of validated screening diagnostic terms.¹¹ Passive surveillance involved utilizing International Classification of Disease, Ninth Revision, code searches (codes 430–438, excluding codes 433.x0 and 434.x0 (x = 1–9), 437.0, 437.2, 437.3, 437.4, 437.5, 437.7, 437.8, and 438) for stroke hospital or emergency department discharges and stroke billing codes in neurology offices. Study neurologists validated all cases of stroke using source documentation, blinded to ethnicity and age. Corpus Christi is approximately 150 miles from Houston and San Antonio, and the seven hospitals in the community serve as the regional referral center for the surrounding communities, allowing for complete case capture of acute neurological events. Ischemic stroke was defined as the acute onset of a focal neurological deficit specifically attributable to a cerebrovascular distribution that persisted for ≥ 24 hours and was not attributable to another disease process.⁹ Transient ischemic attack (TIA) carried the same definition as ischemic stroke except that symptoms resolved within 24 hours.⁹

BASIC methodology includes an extended chart abstraction of a randomly selected subset of cases which includes a review of electrocardiogram (ECG) reports. The study population consisted of BASIC patients with ischemic stroke or TIA between January 1, 2000 and June 30, 2008 who underwent extended abstraction and were found to have either a chart history of AF, or AF noted on the admission ECG. TIAs were included up to July 31, 2007, as BASIC stopped surveillance for TIA after that date. Some patients in the current study were included in prior reports of stroke recurrence¹³ or post-stroke mortality¹⁰ from BASIC; however, the majority (160 of 236, 68%) of patients in the current study were not included in these prior manuscripts due to a longer period of case ascertainment and minor differences in inclusion criteria.

Demographics and Stroke Risk Factors

Stroke risk factors, including coronary artery disease, diabetes, hyperlipidemia, hypertension, previous stroke, smoking, insurance status and presence of a primary care physician were obtained from the medical record. National Institutes of Health Stroke Scale (NIHSS) at the time of index event and at the time of stroke recurrence was abstracted from the medical record.¹⁴ Use of warfarin or antiplatelet agents at discharge from the index event and at the time of presentation with recurrent stroke was abstracted from the medical record. Education level was obtained from patient or proxy interview and was treated as a dichotomous variable defined by completion of high school. Ethnicity was determined from the medical record, as we have previously shown 97% agreement between medical record and patient self-report (kappa=0.94).⁹

Identification of Recurrent Stroke and Mortality

Recurrent ischemic stroke or intracerebral hemorrhage was determined from BASIC surveillance through June 30, 2008. All-cause mortality through the same date was determined from the medical record for in-hospital mortality, or through Texas Department of Health (TDH) databases or the Social Security Death index (SSDI). Through 2005, deaths were identified as previously described,¹⁰ with a match on four of five identifiers (first name, last name, date of birth, social security number, and permanent residence) needed to confirm a case as deceased. After 2005, the methodology for identifying out of hospital deaths changed slightly due to delays in receiving the TDH database and the TDH no longer providing social security numbers. The SSDI then served as the primary method for identifying deaths using the same matching criteria. After the SSDI search, patients without known mortality were linked to the TDH death certificate database using first and last names, date of birth and permanent address. All four items had to be identical to confirm a match, though some manual review was required for special circumstances such as transposed numbers.

Statistical analysis

Age was compared by ethnicity with a Student’s t-test, and NIHSS was compared by ethnicity using Wilcoxon rank-sum test for the index stroke and at the first recurrent stroke. Stroke risk factors, demographics, and anti-thrombotic medication (at discharge from index event and at time of first recurrent stroke) were compared with chi-square tests or Fisher’s exact tests. Stroke recurrence was compared by ethnicity with Kaplan Meier estimates and the log-rank test, with cases censored at the time of death or at the end of follow-up (June 30, 2008). Multivariable Cox proportional hazards regression was performed to compare recurrence-free survival by ethnicity. Since we observed a small number of recurrent strokes, our pre-specified analysis plan was to adjust for only demographics (age and gender) based on the recommendation to have 10 events per covariate to ensure model validity.¹⁵ However, due to the possibility of strong confounding by diabetes and education, additional exploratory models were performed adding diabetes and education (together and separately) to the model. Only the first recurrent event was included in the survival analyses, since only 4 patients had more than one recurrent event.

Post-stroke all-cause mortality through the end of follow up was compared by ethnicity with Kaplan Meier estimates and the log-rank test, with cases censored at the end of follow-up. Multivariable Cox proportional hazards regression was performed to compare survival by ethnicity. Covariates for the mortality model were pre-selected based on factors which were significantly associated with mortality or confounded the ethnicity-mortality relationship in a prior analysis of post-stroke mortality in this population.¹⁰ Gender was also pre-selected to be in the model based on prior literature showing an association between gender and post-stroke survival.¹⁶ Final model covariates included age (treated continuously), gender, ethnicity, NIHSS (treated continuously), coronary artery disease, stroke recurrence (modeled as a time dependant variable), diabetes, and education (dichotomous). Since stroke recurrence risk differed by ethnicity and recurrent stroke could be along the causal pathway between ethnicity and mortality, an additional exploratory analysis was performed excluding recurrent stroke. Analysis was performed in SAS 9.1.3 (SAS Institute, Cary, North Carolina).

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by the Institutional Review Board of the University of Michigan as well as the individual Texas hospitals.

RESULTS

A total of 5,681 cases of ischemic stroke or TIA were identified between January 1, 2000 and June 30, 2008. Of these, 1,660 were randomly selected for extended chart review including ECG reports, with a total of 236 cases of ischemic stroke/TIA and AF included.

Table 1 shows demographic and baseline characteristics of the study population. Compared with NHWs, MAs were younger, less likely to have completed 12 years of education, more likely to be diabetic, and less likely to have a primary care physician. There were no significant ethnic differences in other stroke risk factors as shown in Table 1. There was no ethnic difference in the proportion of cases discharged on warfarin at the time of their baseline stroke/TIA (MA, 37%; NHW, 40%; p=0.68, 21 cases with missing data), or the proportion discharged on no antiplatelet or anticoagulant medication (MA, 16%; NHW, 19%; p=0.54, 21 cases with missing data).

Table 1.

Demographic characteristics and stroke risk factors (n=236)

	Non Hispanic White(n=148) median (IQR) or n (%)	Mexican American(n=88) median (IQR) or n (%)	p value
Age	82 (75, 88)	78 (68, 83.5)	<0.01
Female Gender	79 (53%)	54 (61%)	0.23
Initial NIHSS	4 (2, 9)	5 (2, 8)	0.84
Coronary Artery Disease	80 (54%)	48 (55%)	0.94
Ischemic stroke as presenting syndrome	115 (78%)	70 (80%)	0.74
Diabetes Mellitus	39 (26%)	44 (50%)	<0.01
Hyperlipidemia	44 (30%)	26 (30%)	0.98
Hypertension	115 (78%)	74 (84%)	0.23
History of Previous Stroke or TIA	51 (34%)	29 (33%)	0.81
Current Smoking	17 (12%)	7 (8%)	0.35
Primary Care Physician	143 (97%)	79 (90%)	0.03
High School Education*	111 (77%)	20 (23%)	<0.01
Health insurance coverage†	144 (99%)	83 (95%)	0.20

NIHSS: National institutes of health stroke scale, IQR: inter-quartile range

^*4 cases with missing data.

^†3 cases with missing data

Thirty-three individuals had at least one recurrent event (19 MA and 14 NHW) over a median follow-up time of 427.5 days (inter-quartile range (IQR) 90.5, 1164). All recurrent events were ischemic strokes with the exception of one intracerebral hemorrhage in a MA. Among those with recurrent stroke, no ethnic difference at the time of the recurrence was found in the proportion taking warfarin (MA 38%, NHW 36%, p=0.92, 3 cases with missing data) or taking no antiplatelet/anticoagulant medication (MA 7%, NHW 19%, p=0.60, 3 cases with missing data). The median NIHSS at the first recurrent event was significantly higher in MAs (median 10, IQR: 3, 18) than in NHWs (median 4, IQR: 3, 6; p=0.02). Three MAs and one NWH had a second recurrent stroke during the study period, and one MA had three recurrent strokes during the study period.

MAs had a higher risk of stroke recurrence than NHWs (p=0.01). Kaplan Meier estimates of survival free of stroke recurrence at 28-days and 1-year were 0.99 and 0.85 in MAs and 0.98 and 0.96 in NHWs (Figure 1). The pre-specified Cox proportional hazards analysis showed that the higher risk of stroke recurrence in MAs persisted despite adjustment for age and gender (hazard ratio (HR) 2.46, 95% confidence interval (CI) 1.19–5.11) as shown in Table 2. The exploratory analysis adjusting for diabetes, high school education, age, and gender slightly attenuated the relationship between ethnicity and recurrence with the 95% confidence interval including one, though MAs remained about twice as likely to have stroke recurrence (HR 2.04 95% CI 0.83–5.03) than NHWs. Adding diabetes and education separately to the model with demographics revealed that the change in the ethnic association with risk of recurrence was more due to education (MA HR 2.05 95% CI 0.84, 5.01) than to diabetes (MA HR 2.44 95% CI 1.16–5.16).

Figure 1. Survival free of stroke recurrence.

Kaplan Meier Curve showing survival free of stroke recurrence by ethnicity (p=0.01 log-rank test)

Table 2.

Multivariable models for stroke recurrence and all-cause mortality

	Hazard Ratio	95% confidence interval
Stroke Recurrence (n=236)
Mexican American*	2.46	1.19–5.11
Age	1.01	0.97–1.04
Female†	0.94	0.47–1.91
All Cause Mortality (n=232‡)
Mexican American*	1.03	0.63–1.67
Age	1.06	1.04–1.09
Female†	0.60	0.41–0.87
NIHSS at index stroke§	1.07	1.04–1.09
Coronary artery disease	1.19	0.82–1.72
High School Education	0.99	0.63–1.54
Recurrent stroke	3.66	2.24–5.97
Diabetes	1.03	0.71–1.51

^*Reference: non-Hispanic White

^†Reference: male

^‡Four cases were excluded from the full model for mortality due to missing data on education

^§Treated linearly, NIHSS: National Institutes of Health Stroke Scale

Considering all-cause mortality, a total of 139 individuals died (52 MA, 87 NHW) during the follow up period (median follow up time 523.5 days (IQR: 131, 1247.5). Kaplan Meier estimates of survival at 28 days and 1 year were 0.91 and 0.69 in MAs and 0.82 and 0.65 in NHWs (Figure 2). There was no difference in survival by ethnicity (p=0.99). Cox proportional hazards analysis showed no difference in survival by ethnicity in unadjusted analysis (HR 1.00, 95% CI 0.71–1.42), or when adjusted for age, gender, initial NIHSS, coronary artery disease, education, recurrent stroke, and diabetes (HR 1.03, 95%CI 0.63–1.67). Age, initial NIHSS, recurrent stroke, and gender were significantly associated with survival as shown in Table 2. The exploratory analysis removing stroke recurrence from the multivariable model also showed no ethnic difference in mortality (1.21 95%CI 0.76–1.94).

Figure 2. All-cause mortality.

Kaplan Meier curve showing survival after stroke by ethnicity (p=0.99, log rank test)

DISCUSSION

This population-based study of stroke/TIA patients with AF found that MAs have more than double the risk of recurrent stroke and a greater severity of recurrent stroke than NHWs. There was no ethnic difference in post-stroke mortality. Higher risk of recurrence in MAs could be explained by ethnic differences in warfarin use or monitoring, though we found no ethnic differences in the proportion of cases on anticoagulant or antiplatelet medication at discharge or at the time of stroke recurrence. We did not have more detailed data on outpatient use or monitoring of warfarin which has been shown to be suboptimal in Hispanic Medicare beneficiaries with AF.⁶ Therefore, we cannot exclude the possibility that ethnic differences in warfarin management contributed to the observed ethnic differences in stroke recurrence. Therapeutic use of warfarin has been associated with a lower stroke severity.¹⁷ The greater severity of recurrent stroke in MAs could therefore be consistent with less effective control of anticoagulation in this group. Others have also suggested that warfarin may be less efficacious in minority populations than in NHWs.^{6, 7}

The higher burden of recurrent stroke in MAs could also be explained by a greater frequency of non-cardioembolic recurrent strokes in MAs than in NHWs. Indeed, MAs were more likely than NHWs to have diabetes, which is an important risk factor for small vessel disease¹⁸ and is a risk factor for stroke in AF.¹⁹ However, it is unlikely that the excess small vessel disease in MAs is responsible for the ethnic difference in recurrence, as adjustment for diabetes in exploratory analyses did not alter the higher risk of recurrent stroke in MAs. We did not formally classify the recurrent stroke subtype in the current work since a diagnosis of small artery occlusion requires that cardioembolic sources be absent,²⁰ and therefore no cases in this study would have been classified as small vessel disease as all had AF.

Access to care or socioeconomic factors also may have contributed to the higher risk of recurrence in MAs. However, the Corpus Christi population is non-immigrant, with the majority of MA stroke patients in this community born in the United States.²¹ There was no ethnic difference in health insurance coverage in this community, though MAs were less likely than NHWs to have a primary care physician (97% vs. 90%, p=0.03). While this difference in presence of a primary care physician was statistically significant, the absolute difference may have been too small to be of clinical importance. MAs were significantly less likely than NHWs to have completed high school, and adjusting for high school education in the multivariable model attenuated the relationship between ethnicity and risk of recurrence, though this exploratory analysis should be interpreted with caution due to the small sample. Lack of a high school education has been associated with stroke in both MAs and NHWs in this population.²¹ Lower educational levels might directly contribute to difficulty in following directions for warfarin monitoring and the associated dietary restrictions, or could be a marker for lower socioeconomic status.

Despite the higher risk of stroke recurrence, there was no ethnic difference in post-stroke mortality. This is somewhat surprising, particularly given that recurrent stroke itself was a potent predictor of mortality, and that MAs had more recurrent strokes that were more severe in nature than NHWs. The exploratory analysis removing recurrent stroke from the multivariable mortality model resulted in an increase in the hazard ratio for risk of death in MAs (from 1.03 to 1.21) but it remained non-significant. We have previously shown a higher risk of stroke recurrence and a lower risk of death in MAs compared with NHWs when examining all ischemic strokes in this community.^{10, 13} Better than expected mortality in Hispanics has been seen in other diseases and has been termed the “Hispanic Paradox.” ²² We did not demonstrate lower mortality in MAs than NHWs in this study. However, the finding of similar mortality by ethnicity despite a higher recurrence risk in MAs may still be reflective of a broader protective effect of Hispanic ethnicity against mortality. Ethnic differences in social support networks or do-not-resuscitate orders²³ may also play a role, and this requires further study.

Use of warfarin at the time of discharge or at first recurrent event was 40% or less in both ethnic groups. We cannot comment on the appropriateness of this treatment pattern since we were unable to systematically assess for contraindications to warfarin and did not track outpatient medications or time in range of therapeutic anticoagulation. Therefore, it is possible that patients not prescribed warfarin at discharge had contraindications to treatment or had anticoagulation initiated at a subsequent outpatient visit. However, contraindications to warfarin are unlikely to be the sole reason for the low proportion of patients treated with warfarin. A study of Medicare patients hospitalized with AF which was able to assess for contraindications still found that only two-thirds of ideal candidates were prescribed warfarin.⁶ Therefore, the finding of a relatively low proportion of AF patients treated with warfarin is not unique to this community.

This work has limitations. This was a retrospective review of prospectively collected data. Some potentially important factors such as warfarin contraindications were not systematically assessed. We were not able use a stroke risk stratification scheme for AF such as the CHADS₂¹⁹ score due to lack of data on recent congestive heart failure exacerbation. However, all patients in this study were at high risk of recurrent stroke and were potential candidates for warfarin given their presentation with stroke or TIA. The small sample and the low number of recurrent events limited our ability to adjust for confounders in the recurrence analysis. However, the strength of the association between ethnicity and stroke recurrence despite the small sample suggests that this is an important association which deserves further study.

In conclusion, MAs with AF and stroke/TIA have a higher risk of stroke recurrence, greater severity of recurrent stroke, yet no difference in post-stroke survival when compared with NHWs. Aggressive stroke prevention measures are warranted in this population, and further study is needed to investigate reasons for the higher risk of recurrence and severity in MAs.