Author + information
- Received May 5, 2003
- Revision received August 19, 2003
- Accepted September 24, 2003
- Published online May 5, 2004.
- Arnold J. Greenspon, MD, FACC*,* (, )
- Robert G. Hart, MD†,
- David Dawson, MD‡,
- Anne S. Hellkamp, MS§,
- Marc Silver, MD, FACC∥,
- Greg C. Flaker, MD, FACC¶,
- Eleanor Schron, MS, RN#,
- Lee Goldman, MD, FACC**,
- Kerry L. Lee, PhD§,
- Gervasio A. Lamas, MD, FACC††,
- MOST Study Investigators
- ↵*Reprint requests and correspondence:
Dr. Arnold J. Greenspon, Jefferson Heart Institute, Jefferson Medical College, 925 Chestnut Street, Philadelphia, Pennsylvania 19107, USA.
Objectives This study was an analysis of factors associated with stroke in a population of patients paced for sinus node dysfunction in a large prospective clinical trial (Mode Selection Trial [MOST]).
Background The effects of dual-chamber versus single-chamber ventricular pacing on subsequent stroke in patients with sinus node dysfunction are not known.
Methods A total of 2,010 patients with sinus node dysfunction were randomized to ventricular or dual-chamber pacing and followed for a median of 33.1 months.
Results The median participant age was 74 years. During 5,664 patient-years of follow-up, 90 strokes (11 hemorrhagic) occurred. By life-table analysis, the rate of stroke was 2.2% (95% confidence interval [CI] 1.6 to 2.9) at one year and 5.8% (95% CI 4.5 to 7.1) at four years. The incidence of stroke was not significantly different in dual-chamber (4%) as compared with ventricular-paced patients (4.9%) (hazard ratio [HR] 0.82, 95% CI 0.54 to 1.25, p = 0.36). Multivariable analysis demonstrated that significant predictors of stroke included prior stroke or transient ischemic attack, Caucasian race, hypertension, prior systemic embolism, and New York Heart Association functional class III or IV (p < 0.05); pacing mode remained non-significant after adjustment for these factors (p = 0.37). Clinically reported atrial fibrillation after implantation was a risk factor for stroke in this cohort after adjustment for other predictors of stroke (p = 0.042, HR 1.68 [95% CI 1.02 to 2.76]).
Conclusions Clinical characteristics, but not mode of pacing, were associated with subsequent stroke in patients paced for sinus node dysfunction.
The most common indication for implantation of a permanent pacemaker is sick sinus syndrome (1). It is estimated that the yearly incidence of stroke in this population is approximately 6% to 10% (2–7). Several large randomized clinical trials have established that atrial fibrillation (AF), which commonly occurs in patients with sinus node dysfunction (8,9), is associated with an annual risk of stroke of approximately 5%, which may be significantly lowered to 1.7% with appropriate anticoagulant therapy (10). Previous trials have suggested that physiologic pacing may decrease the incidence of AF (11–14). Although there is a direct relationship between AF and stroke, the effects of physiologic pacing on stroke are less clear (11–16). The Mode Selection Trial (MOST) was a large, prospective, randomized, clinical trial designed to analyze the role of physiologic pacing on the outcome of patients receiving a pacemaker for symptomatic sick sinus syndrome (15). The present study is an analysis of factors associated with stroke in this patient population.
The MOST study was designed as a five-year trial to compare ventricular rate-modulated (VVIR) pacing with dual-chamber rate-modulated pacing (DDDR) in patients whose sinus node dysfunction required permanent pacing for bradycardia (15). The first patient was enrolled on September 25, 1995, and the last patient on October 13, 1999. Follow-up ended January 31, 2001. The 2,010 enrolled patients at 91 clinical sites (Appendix) had a median follow-up of 33.1 months.
The design of the study has been described (17). In brief, patients with documented sinus node dysfunction undergoing implantation of a dual-chamber pacing system were eligible for admission into the trial if they were age 21 years or older, were in sinus rhythm at the time of randomization, and had adequate atrial and ventricular sensing and pacing thresholds. Before implantation, all patients had a clinical assessment including history and physical examination, baseline New York Heart Association (NYHA) functional class for heart failure, a mini-mental state examination (18), and a quality-of-life evaluation that included the 36-item Medical Outcomes Study Short-form General Health Survey (19). Comorbidities were assessed with the Charlson Comorbidity Index (20). The site investigators recorded history of prior stroke or transient ischemic attack (TIA).
All patients received a DDDR pacemaker. Pacing mode was then randomly assigned to either VVIR or DDDR pacing by the Data Coordinating Center at the Duke Clinical Research Institute. Randomization was stratified by history of stroke and by clinical site.
Following implantation, all patients were seen at a clinical site four times during the first year and then twice yearly thereafter. A complete history and physical examination, including neurologic evaluation, were performed at each follow-up visit. If a stroke was suspected, the clinical center was required to consult a local neurologist who would then obtain a brain imaging study (either computerized tomography or magnetic resonance imaging). The clinical site neurologist was responsible for documenting the clinical and laboratory findings of the stroke on a separate case report form. Events were classified on the basis of case report forms, hospital records, and brain imaging. On the basis of these findings the stroke was characterized as primary hemorrhagic, ischemic, or unknown. Hemorrhagic strokes were subclassified as intraparenchymal, subarachnoid hemorrhage, or subdural hematoma. A clinical events committee composed of six cardiologists and two neurologists classified the cause of death and adjudicated all suspected strokes unaware of the mode of pacing. Atrial fibrillation was adjudicated by an electrocardiographic core lab.
Warfarin use data from the follow-up period were used to determine whether patients were taking warfarin during any of the following time periods: at hospital discharge, during the first year, during the second year, or after the second year. Limitations of medication use data precluded any finer characterization of warfarin use patterns.
The two pacing mode groups were analyzed on an intention-to-treat basis. Cumulative event rates were calculated by the Kaplan-Meier (21)method. Single-predictor Cox proportional hazards models (22)were used to assess the relationship of baseline variables to stroke in univariable analyses. Multivariable analysis of baseline factors, in which all variables were considered simultaneously for their relationship to stroke, also used a Cox proportional hazards model with time to first stroke as the outcome variable. Cox models for the association of post-randomization AF with stroke included AF as a time-dependent covariate, which allows patients to move from the no-AF risk class to the AF risk class at the first documented occurrence of AF. To assess the role of potential predictors of stroke without the possible confounding effect of warfarin use, univariable and multivariable analyses were also performed in the subgroup of 1,204 patients who did not receive warfarin at any time.
To assess the relationship between AF and subsequent warfarin use and, therefore, the extent to which warfarin use might affect the AF-stroke risk relationship, we used unadjusted logistic regression models to examine the association of AF in the first year post-randomization with warfarin use in the second year among patients with no warfarin use in the first year.
The 2,010 patients who were entered into the trial had a median age of 74 years (interquartile range of 68 to 80 years). Of these, 996 patients were randomized to VVIR pacing and 1,014 were randomized to DDDR pacing. Baseline characteristics and cardiac risk factors, including age, female gender (VVIR = 48% vs. DDDR = 47%), and minority race (VVIR = 14% vs. DDDR = 16%), were similar in both groups (Table 1). The prevalence of risk factors for stroke, such as hypertension (VVIR = 61% vs. DDDR = 63%), hypercholesterolemia (VVIR = 34% vs. DDDR = 37%), cigarette smoking (VVIR = 9% vs. DDDR = 8%), and diabetes mellitus (VVIR = 20% vs. DDDR = 24%), was also similar in the two groups. Although a slight imbalance existed in other clinical variables favoring less sick patients in the VVIR pacing group, we used multivariable analyses to control for these differences. Of those patients who had a history of AF or flutter, 76% had their most recent episode within the three weeks before randomization.
A prior stroke was reported in 11% (n = 108) of the VVIR pacing patients and 11% (n = 116) of the DDDR pacing patients. The median time since the most recent stroke was similar in both groups (VVIR = 2.4 years [interquartile range 0.4 to 7.6] vs. DDDR = 2.3 years [interquartile range 0.6 to 6.0]). A previous TIA was reported in 8% (n = 77) of VVIR pacing patients and 7% (n = 66) of DDDR pacing patients. The median time since the most recent TIA was also similar in both groups (VVIR = 0.9 years [interquartile range 0.0 to 2.9] vs. DDDR = 1.5 years [interquartile range 0.0 to 4.2]). Prior systemic embolism was also similar in the two groups (VVIR = 3%, n = 26 vs. DDDR = 3%, n = 30).
Patients were followed for a median of 33.1 months (interquartile range 21.4 to 46.3 months), which represents 5,664 patient-years of follow-up. Over the course of the trial, there were 90 strokes (Table 2). The one-year, three-year, and four-year event rates in the MOST study were 2.2% (95% confidence interval [CI] 1.6 to 2.9), 4.9% (95% CI 3.8 to 6.0), and 5.8% (95% CI 4.5 to 7.1), respectively. There were 49 strokes (4.9%) in the VVIR pacing group versus 41 in the DDDR pacing group (4%) (unadjusted hazard ratio [HR] for DDDR:VVIR = 0.82 [95% CI 0.54 to 1.25], p = 0.36) (Fig. 1). Two patients had more than one stroke in the VVIR pacing group as compared with six patients in the DDDR pacing group. There was no difference in the type of stroke recorded in each group. Most strokes were due to ischemia (VVIR = 88%, n = 43 vs. DDDR = 78%, n = 32) as opposed to hemorrhage (VVIR = 10%, n = 5 vs. DDDR = 15%, n = 6).
Prior stroke, prior TIA, prior neurologic event (either stroke or TIA), prior systemic embolism, age, hypertension, and NYHA functional class were associated with subsequent stroke (p < 0.05). Twenty other baseline variables were not found to be associated with stroke, including neurologic deficit on baseline physical examination, gender, weight, race, diabetes, prior myocardial infarction, history of congestive heart failure or AF, therapy with aspirin or coumadin at admission, antiarrhythmic drug therapy, and pacing mode. A history of AF was not a predictor of subsequent stroke (p = 0.22).
A multivariable analysis of baseline factors as predictors of subsequent stroke was performed when all variables were considered simultaneously (Table 3). Significant predictors included prior stroke or TIA (p = 0.014, HR 1.80 [95% CI 1.12 to 2.89]), Caucasian race (p = 0.024, HR 2.44 [95% CI 1.12 to 5.32]), hypertension (p = 0.028, HR 1.70 [95% CI 1.06 to 2.72]), prior systemic embolism (p = 0.037, HR 2.43 [95% CI 1.06 to 5.60]), and NYHA chronic heart failure functional class III or IV (p = 0.044, HR 1.67 [95% CI 1.01 to 2.77]). Atrial fibrillation following randomization was a significant risk factor for stroke (p = 0.042, HR 1.68 [95% CI 1.02 to 2.76]) after adjustment for other significant predictors of stroke. Pacing mode was not predictive of subsequent stroke (p = 0.37).
Warfarin and aspirin
At the time of hospital discharge after pacemaker implantation, approximately one-half of the patients in the trial were taking aspirin (VVIR = 46%, n = 458 vs. DDDR = 49%, n = 496) whereas 23% of the VVIR pacing group (n = 233) and 24% of the DDDR pacing group (n = 238) were taking warfarin. Therefore, approximately three-quarters of the patients in the MOST study were taking either aspirin or warfarin immediately after implantation.
Neither warfarin (p = 0.5) nor aspirin use (p = 0.2) at hospital admission was a negative risk factor for stroke. In the MOST study, 773 patients were taking warfarin at some point during the study. Post-enrollment AF in the first year was closely associated with new warfarin therapy in the second year (p = 0.001, OR = 6.6 [95% CI 3.7 to 11.9]).
Among the 1,204 patients in the MOST study who did not receive warfarin during the course of the study, there were 50 strokes. In these patients univariable analysis showed the same variables with an association with stroke risk as for all patients: age, NYHA functional class, prior AF, hypertension, and prior stroke or TIA. The direction and magnitude of the associations were similar to those in the entire MOST study sample. In the multivariable model, the only independent predictors of stroke were prior stroke or TIA and age.
The major finding of this analysis of a large prospective randomized clinical trial of patients with sick sinus syndrome was that baseline clinical characteristics, but not mode of pacing, were associated with subsequent stroke. The clinical predictors of stroke included prior neurologic event (either stroke or TIA), prior systemic embolism, age, hypertension, and NYHA functional class before pacemaker implantation.
Incidence of stroke in sinus node dysfunction
In the MOST study, the incidence of stroke in the first year after implant was 2.2% (95% CI 1.6 to 2.9). Early observational studies of patients with sinus node dysfunction estimated the yearly incidence of stroke at 6% to 10% (2–5). In a nonrandomized study of 507 patients paced for sinus node dysfunction between 1980 and 1989, the actuarial incidence of stroke was 3% at one year and 5% at five years (16). In the first small, randomized trial of physiologic pacing in sinus node dysfunction, the thromboembolic event was 11.5% in patients followed for a mean of 40 months (12).
The role of antithrombotic therapy in the prevention of stroke in patients with AF has been assessed in a number of large multicenter trials (10). The annual risk of stroke in the control patients ranged from 1% in those younger than 65 years to 8.1% in those older than 75 years with more than one risk factor, such as hypertension, prior stroke, or diabetes. By comparison, the annualized rate of stroke in those patients treated with antithrombotic therapy was between 1% and 1.7% (10). On the basis of these studies, antithrombotic therapy is now recommended as primary prevention for ischemic stroke, except in those patients who are under age 65 years and have no risk factors (23,24).
Although patients paced for sinus node dysfunction represent a different population of patients than those with primary AF, antithrombotic therapy may nonetheless influence their risk of stroke. Therefore, one explanation for the low incidence of stroke in the MOST study is the higher use of aspirin and warfarin than in earlier studies. In the MOST study, a total of 72% of patients were receiving antithrombotic or antiplatelet therapy at the time of hospital discharge: 48% of patients received aspirin and 24% received warfarin. The frequency of use of these agents in early observational studies of sinus node dysfunction is not known. In the earliest trial of physiologic versus ventricular-based pacing in sick sinus syndrome, Andersen and colleagues reported that 95 of 225 patients (42%) received aspirin while only seven of 225 (3%) received warfarin, perhaps explaining a relatively high stroke rate (12). By contrast, in the Canadian Trial Of Physiologic Pacing (CTOPP), which began after the value of antithrombotic therapy for stroke prevention was documented, 34% of patients were taking aspirin and 11% anticoagulant drugs at the time of randomization (10,11). Long-term use of these agents was not reported.
Impact of AF on subsequent stroke
The association between AF and stroke is well established in the clinical literature (10). The analyses of AF and stroke risk in most studies, however, are confounded by the concurrent use of antithrombotic therapy. Therefore, a priori, one might expect an attenuation of the known risk of stroke in patients that develop AF. Indeed, 46% of the MOST study patients had a history of prior AF, yet history did not predict subsequent stroke. However, an adjusted analysis of AF following randomization as a predictor of stroke was significant: patients with AF post-randomization had a 70% higher relative risk of stroke than patients without AF, even after adjustment for important clinical predictors. This finding may suggest that the routine clinical use of antithrombotic therapy in the MOST study population was insufficient to completely blunt the risk of AF.
Analysis of the use of warfarin in the MOST study
There were a total of 90 strokes in the MOST study. Therapeutic anticoagulation is known to markedly reduce the risk of stroke in patients with AF (10). Therefore, the pattern of warfarin use in our study population is a potentially confounding variable when one tries to analyze risk factors for stroke. Because the level of anticoagulation was not specifically recorded during the MOST study, it is difficult to determine whether patients were receiving warfarin and, if they were, their level of anticoagulation at the time of stroke. Nonetheless, the close relationship between warfarin use and AF in our study cohort may help explain the low observed rate of stroke in this population. Results in the subgroup of patients who did not receive warfarin at any time are consistent with the analysis of the entire group, suggesting that the factors that affect stroke risk are similar in warfarin and non-warfarin patients.
Role of physiologic pacing in preventing subsequent stroke
In the MOST study, physiologic pacing was not associated with significantly fewer strokes. The MOST study was designed to have a power of >90% to detect a 25% reduction in the combined end point of death or stroke (17). Our power to detect stroke alone was much lower. Given the total number of events (n = 90), our power to detect as significant the difference observed for stroke alone (hazard ratio = 0.82) was only 15%. In an earlier observational study, Sgarbossa et al. (16)demonstrated a strong association between ventricular-based pacing and stroke (hazard ratio = 2.61). In the later Andersen et al. (13)randomized trial, atrial pacing was associated with a significantly lower stroke rate compared with ventricular pacing when patients were followed for a mean of 5.5 years. However, the 22.6% rate of stroke among ventricular-paced patients was approximately eight times higher than other randomized trials of patients with sick sinus syndrome. In the CTOPP study (11), where the annual rate of stroke was low (ventricular pacing = 1.1%, physiologic pacing = 1%) and only 42% of patients were paced for sinus node dysfunction, no association was found. The mean follow-up time in the MOST study was 33 months, so it is possible that longer follow-up or a larger sample size may have demonstrated a benefit of dual-chamber pacing on subsequent stroke. Alternatively, the high use of warfarin and aspirin in the MOST study may also have reduced any of the beneficial effects of physiologic pacing on subsequent stroke and contributed to the low incidence of stroke. A meta-analysis of these and other large randomized pacemaker trials could help explore the issues.
Baseline clinical characteristics, but not mode of pacing, were associated with subsequent stroke in patients paced for sinus node dysfunction. Dual-chamber pacing does not appear to decrease the incidence of stroke significantly in elderly patients paced for sinus node dysfunction. The high use of antithrombotic therapy in the MOST study, particularly in patients with AF, may have contributed to the low stroke rate.
☆ This study was supported by grants U01 HL 49804 (Dr. Lamas), U01 HL 53973 (Dr. Lee), and U01 HL 55981 (Dr. Goldman) from the National Heart, Lung, and Blood Institute of the National Institutes of Health, Bethesda, Maryland. Medtronic Inc., Guidant Corp., and St. Jude Medical donated additional support for study meetings and ancillary studies.
- atrial fibrillation
- confidence interval
- Canadian Trial Of Physiologic Pacing
- dual-chamber rate-modulated
- Mode Selection Trial
- New York Heart Association
- transient ischemic attack
- ventricular rate-modulated
- Received May 5, 2003.
- Revision received August 19, 2003.
- Accepted September 24, 2003.
- American College of Cardiology Foundation
- Lamas G.A.,
- Pashos C.L.,
- Normand S.L.T.,
- McNeil B.
- Fisher M.,
- Kase C.S.,
- Stelle B.,
- Mills R.M. Jr.
- Rubenstein J.J.,
- Schulman C.L.,
- Yurchak P.M.,
- DeSanctis R.
- Andersen H.R.,
- Nielsen J.C.,
- Thomsen P.E.B.,
- et al.
- Sgarbossa E.B.,
- Pinski S.L.,
- Maloney J.D.,
- et al.
- Kaplan S.H.
- Cox D.R.
- Goldstein L.B.,
- Adams R.,
- Becker K.,
- et al.
- Hart R.G.,
- Bailey R.D.