Author + information
- Received February 5, 2019
- Revision received March 18, 2019
- Accepted April 9, 2019
- Published online August 5, 2019.
- Mohamad Alkhouli, MDa,b,∗ (, )
- Fahad Alqahtani, MDa,
- Abdulrahman Tarabishy, MDc,
- Gurpreet Sandhu, MDb and
- Charanjit S. Rihal, MDb
- aDivision of Cardiology, Department of Medicine, West Virginia University, Morgantown, West Virginia
- bDepartment of Cardiovascular Diseases, Mayo Clinic School of Medicine, Rochester, Minnesota
- cDivision of Neuroradiology, Department of Radiology, West Virginia University, Morgantown, West Virginia
- ↵∗Address for correspondence:
Dr. Mohamad Alkhouli, West Virginia University School of Medicine, 1 Medical Center Drive, Morgantown, West Virginia 26505-8059.
Objectives The aim of this study was to assess temporal trends in the incidence of ischemic stroke among patients undergoing percutaneous coronary intervention (PCI), predictors of post-PCI ischemic stroke, and the impact of post-PCI ischemic stroke on in-hospital morbidity, mortality, length of stay, and cost.
Background Data on the incidence and outcomes of ischemic stroke in patients undergoing PCI in the contemporary era are limited.
Methods The National Inpatient Sample was used to identify patients who underwent PCI between January 1, 2003, and December 31, 2016. The incidence of post-PCI ischemic stroke was calculated, and its predictors were assessed. In-hospital outcomes of patients with and those without post-PCI stroke were also compared.
Results The adjusted incidence of post-PCI ischemic stroke increased during the study period from 0.6% to 0.96% following PCI for ST-segment elevation myocardial infarction, from 0.5% to 0.6% following PCI for non–ST-segment elevation myocardial infarction, and from 0.3% to 0.72% following PCI for unstable angina or stable ischemic disease (ptrend <0.001). Carotid disease, cardiogenic shock, atrial fibrillation, and older age were the strongest predictors of post-PCI ischemic stroke. Post-PCI stroke rates were lower at high-volume versus low- to intermediate-volume centers. Thrombolytics, cerebral angiography, and mechanical thrombectomy use increased over time but remained infrequent. After propensity score matching, in-hospital mortality was higher among patients with post-PCI stroke (23.5% vs. 11.0%, 9.5% vs. 2.8%, and 11.5% vs. 2.4% in the ST-segment elevation myocardial infarction, non–ST-segment elevation myocardial infarction, and unstable angina or stable ischemic heart disease cohorts, respectively; p < 0.001). Post-PCI stroke was associated with a >2-fold increase in length of stay, a >3-fold increase in nonhome discharges, and a >60% increase in cost.
Conclusions The incidence of post-PCI ischemic stroke increased significantly over the past decade, partially because of the increasing complexity of patients undergoing PCI over time. Further studies are needed to systematically assess contributors to this worrisome trend and to identify effective strategies for its mitigation.
Acute ischemic stroke is an uncommon but often devastating complication of percutaneous coronary intervention (PCI). The reported incidence of post-PCI ischemic stroke has varied widely among different studies. In single-center registries, the incidence of post-PCI stroke ranged from 0.37% to 1.3% (1–4). In recent randomized controlled trials, the incidence of post-PCI stroke was 0.8% to 1.4% with PCI for acute myocardial infarction and 0.4% to 0.8% with PCI for unstable angina (UA) or stable ischemic heart disease (SIHD) (5–8). However, contemporary nationwide data on post-PCI stroke are limited. The latest national analysis was performed using the CathPCI Registry and documented an overall stroke rate of 0.22%. However, that study included noncontemporary cohorts of patients who were treated with PCI prior to 2007 (9). The past decade has witnessed significant increases in the number of older and higher risk patients undergoing PCI and in the complexity of PCIs being performed (e.g., chronic total occlusion intervention, mechanical circulatory support [MCS]–assisted PCI) (10–13). Whether those changes are associated with temporal changes in the incidence and outcomes of post-PCI stroke remain unknown. We hence used the National Inpatient Sample to assess: 1) temporal trends in the incidence of ischemic stroke among patients undergoing PCI; 2) predictors of post-PCI ischemic stroke; and 3) the impact of post-PCI ischemic stroke on in-hospital morbidity, mortality, length of stay, and cost.
The National Inpatient Sample (NIS) (14) is the largest publicly available all-payer administrative claims-based database that contains information about patient discharges from approximately 1,000 nonfederal hospitals in 45 states. Data from the NIS are stratified to represent 20% of U.S. inpatient hospitalizations across different hospital and geographic regions (random sample). National estimates of the entire U.S. hospitalized population are calculated using the standard Agency for Healthcare Research and Quality weighting methodology.
The NIS was queried to identify patients who underwent PCI between January 1, 2003, and December 31, 2016, using International Classification of Diseases-Clinical Modification procedural codes (Online Table 1). We used International Classification of Diseases-9th Revision-Clinical Modification codes for PCI performed prior to October 1, 2015, and International Classification of Diseases-10th Revision-Clinical Modification codes for PCI performed on or after that date. The use of the NIS database for PCI numbers was validated by Epstein et al. (15), who reported a mean difference of 0.2% in quarterly PCI counts between Medicare claims and the NIS. Patients who underwent coronary bypass grafting during the same admission were excluded because it would not possible to determine if the stroke was secondary to PCI or to the bypass surgery. The overall cohort was then divided into 3 groups according to the indication for PCI: ST-segment elevation myocardial infarction (STEMI), non–ST-segment elevation myocardial infarction (NSTEMI), and UA or SIHD (Online Figure 1). Patients with acute ischemic stroke following PCI were then identified using International Classification of Diseases-9th Revision-Clinical Modification and International Classification of Diseases-10th Revision-Clinical Modification codes (Online Table 1). These codes have been found to have excellent sensitivity and specificity in several studies (16–18).
The study’s primary endpoints were: 1) the annual incidence of post-PCI stroke stratified by the indication for PCI (STEMI, NSTEMI, or UA or SIHD); 2) predictors of post-PCI stroke; and 3) in-hospital outcomes (morbidity, mortality, resource use, and cost) of patients with post-PCI stroke compared with those without post-PCI stroke diagnoses.
Temporal trends in the incidence of post-PCI stroke
We calculated the crude annual incidence of post-PCI stroke stratified by the indication for PCI (STEMI, NSTEMI, and UA or SIHD). The statistical significance of the trend in the incidence was assessed using the Cochrane-Armitage test. To determine whether the change in stroke rate over time was related to changes in the characteristics of patients, the hospital, or PCI complexity or techniques, we applied 2 risk adjustment methods. First, a multivariate regression model was constructed using generalized estimation equations with exchangeable working correlation matrix. This model was constructed to account for clustering of outcomes within hospitals and has been validated in several studies (14,19,20). Second, a marginal structure model was constructed to allow further adjustment of time-dependent variables (21). Each observation received a weight inversely proportional to the estimated probability of stoke, which was then fitted to a marginal structural model. Each calendar year was compared with 2003 as the reference year. To obtain the yearly risk-adjusted stroke rate for the study period, we multiplied the adjusted odds ratio for each year (2004 to 2016) by the observed stroke rate for the reference year. Variables included in this model are listed in Online Table 2.
To confirm the validity of the results in different settings, we performed additional sensitivity analyses. In theory, it is possible that some patients had acute ischemic strokes and then underwent PCI during prolonged hospitalizations. The inclusion of these patients might lead to a falsely elevated rate of post-PCI stroke rate. To minimize the chance of this potential confounding issue, we calculated the incidence of post-PCI stroke in patients who underwent PCI within 48 h of admission. Also, we hypothesized that rates of post-PCI stroke might be higher at low-volume PCI institutions. Hence, we grouped the institutions into 3 tertiles on the basis of PCI volume (low, intermediate, and high) and calculated the incidence of post-PCI stroke at intermediate- and high-volume institutions only. This analysis was performed on a subcohort of patients who underwent PCI prior to 2012, as the chance in NIS sampling methodology in 2012 precludes accurate of PCI volume assessment on or after 2012.
Predictors of post-PCI stroke
Predictors of stroke after PCI were assessed in univariate logistic regression analysis. Those with p values of <0.10 were then further assessed in a multivariate logistic regression analysis.
Outcomes of post-PCI stroke
We compared the incidence of in-hospital death, major morbidities, cost, discharge destination, and length of stay between patients with and those without post-PCI stroke. Descriptive statistics are presented as frequencies with percentages for categorical variables. Mean, SD, median, and 25th and 75th percentiles are reported for continuous measures. To estimate the cost of hospitalization, the NIS data were merged with cost-to-charge ratios available from the Healthcare Cost and Utilization Project. We estimated the cost of each inpatient stay by multiplying the total hospital charge by cost-to-charge ratios. To reduce the effect of selection bias, a propensity score–matching model was developed using logistic regression to derive 2 matched groups for comparative outcomes analysis. Patients who underwent PCI with or without subsequent stroke were entered into a nearest neighbor 1:1 variable ratio, parallel, balanced propensity-matching model using a caliper of 0.01 without replacement to ensure perfect matching. Variables included in the propensity score–matching models are presented in Online Table 3. Matched categorical variables are presented as frequencies with percentages and were compared using the McNemar test. Matched continuous variables are presented as mean ± SD and were compared using a paired-samples Student’s t-test. A type I error rate of <0.05 was considered statistically significant. Weighted data were used for all statistical analyses. All statistical analyses were performed using SPSS version 24 (IBM, Armonk, New York).
A total of 8,753,574 patients underwent PCI during the study period. The indication for PCI was STEMI in 9.8%, NSTEMI in 23.5%, and UA or SIHD in 66.7%. The incidence of post-PCI ischemic stroke was 0.56% overall, but was higher after PCI for STEMI (0.97%) and PCI for NSTEMI (0.81%) than after PCI for UA or SIHD (0.41%) (p < 0.001). Patients who had acute ischemic strokes following PCI were older, were more likely to be female, had a higher prevalence of key comorbidities and cardiogenic shock, and were more likely to have been treated with MCS devices compared with those who did not have post-PCI strokes (Table 1). The incidence of post-PCI ischemic stroke increased significantly between 2003 and 2016 among patients undergoing PCI for STEMI (from 0.6% to 1.3%), NSTEMI (from 0.5% to 1.0%), and UA or SIHD (from 0.3% to 0.9%) (p < 0.001 for all) (Central Illustration). These upward trends persisted after risk adjustment using 2 different risk adjustment models (Online Table 4, Central Illustration) and in 2 sensitivity analyses: 1 including only patients who underwent PCI within 48 h of admission (Online Table 5) and 1 excluding PCIs performed at low-volume centers (Online Tables 6 and 7).
In a logistic regression analysis, carotid artery disease, cardiogenic shock, atrial fibrillation, and older age were the strongest predictors of post-PCI acute ischemic stroke (Table 2). Notably, the prevalence of these variables in the overall cohort of patients undergoing PCI increased significantly over time (Online Figure 2). Other predictors included female sex, Hispanic race, hypertension, diabetes, chronic kidney disease, vascular disease, coagulopathy, use of MCS, anemia, PCI for STEMI or NSTEMI, and a more recent PCI year (Table 3). Among patients with post-PCI stroke, rates of thrombolytic agents, cerebral angiography, and mechanical thrombectomy were low at 4.56%, 4.73%, and 0.65%, respectively (Table 4). In a secondary logistic regression analysis including annual PCI volume as a variable, patients who underwent PCI at institutions in the first and second volume tertiles had higher odds of post-PCI stroke compared with those who underwent PCI at institutions in the third volume tertile (odds ratios: 1.11 [95% confidence interval: 1.03 to 1.19] and 1.19 [95% confidence interval: 1.15 to 1.23], respectively; p < 0.001) (Online Table 8).
Outcomes of post-PCI stroke in the unmatched cohorts
Compared with patients who did not experience strokes, those with post-PCI stroke had higher in-hospital mortality (23.5% vs. 5.4%, 9.5% vs. 1.7%, and 11.6% vs. 1.0% in the STEMI, NSTEMI, and UA or SIHD cohorts, respectively; p < 0.001 for all). These higher mortality rates among patients with post-PCI stroke persisted overtime (Figure 1). They also had higher incidences of vascular complications, acute kidney injury, new dialysis requirements, gastrostomy, blood transfusion, and mechanical ventilation (Table 5). The rate of nonhome discharges, length of stay, and hospital cost were substantially higher in the post-PCI stroke group. The morbidity and mortality of post-PCI stroke at intermediate- to high-volume PCI institutions were similar to those in the overall cohort (Online Table 9).
Outcomes of post-PCI stroke in the matched cohorts
After propensity score matching, in-hospital mortality remained substantially higher among patients who had ischemic strokes after PCI compared with those who did not (23.5% vs. 11.0%, 9.5% vs. 2.8%, and 11.5% vs. 2.4% in the STEMI, NSTEMI, and UA or SIHD cohorts, respectively; p < 0.001 for all). Similarly, the incidence of key complications remained higher in patients with post-PCI strokes (Table 5). Length of stay was >2-fold longer in patients with post-PCI strokes, who also had a 3-fold higher incidence of nonhome discharge. Post-PCI stroke was associated with a >60% increase in the cost of hospitalization in all 3 cohorts (Table 5).
The main findings of the present investigation are as follows: 1) the incidence of post-PCI ischemic stroke is low but has increased significantly over the past decade, partially because of the increasing complexity of the patients treated and the intricacy of the PCI techniques themselves; 2) carotid artery disease, cardiogenic shock, atrial fibrillation, and older age were the strongest predictors of post-PCI ischemic stroke, and low institutional annual PCI volume was also independently associated with increased risk for post-PCI stroke; 3) cerebral angiography and mechanical thrombectomy are used in a small but increasing minority of patients who have post-PCI strokes; and 4) post-PCI stroke is associated with substantial morbidity, mortality, resource use, and cost.
The incidence of post-PCI stroke has been assessed in several single-center studies, national registries, and meta-analyses of PCI trials (1–7,9,22). The reported incidence of post-PCI stroke varied across this study from 0.22% to 1.6% but was consistently higher after PCI for STEMI and NSTEMI than after PCI for UA or SIHD. However, the majority of these studies included selected groups of patients and were performed more than a decade ago. Hence, their applicability to current PCI practice is questionable given the temporal changes in the risk profile of patients treated (older patients, more comorbidities, cardiogenic shock, and so on), and the increasing complexity of the PCI procedures themselves (radial access, chronic total occlusion, MCS devices, and so on) (11–13,23,24). Our analysis used a large national all-payer database inclusive of patients across all age groups and found a significant increase in the incidence of post-PCI stroke over the past 14 years. However, after risk adjustment, the magnitude of this upward trend was less pronounced, suggesting that a large number of the excess stroke events in recent years can be explained by the increasing complexity of patients treated with PCI. Nonetheless, further studies are needed to confirm and assess the underlying causes of these worrisome trends.
Identifying predictors of post-PCI stroke is essential to develop effective prevention strategies. Prior studies have found older age, female sex, vascular disease, renal insufficiency, history of stroke or transient ischemic attack, heart failure, use of MCS devices, and vein graft interventions to be independent predictors of post-PCI stroke (1,9,25). Our multivariate logistic regression analysis confirmed these predictors but also identified novel predictors of post-PCI stroke, including non-White race, cardiogenic shock, carotid artery disease, atrial fibrillation, use of atherectomy devices, and institutional procedural volume. Certain technical and procedural factors (access site, PCI target, intra- and post-procedural anticoagulation and/or antiplatelet therapy, and so on) are not captured in this database, and hence their predictive value for post-PCI stroke is uncertain. However, prior studies have found that patient comorbidities and clinical presentation are more powerful predictors than procedural characteristics (1,9,26). Although several of the predictors of post-PCI stroke in our study are not modifiable, their recognition is essential for comprehensive risk stratification of patients referred for PCI.
Our study documents a substantial negative impact of post-PCI stroke on in-hospital morbidity, mortality, resource use, and cost. In addition, it showed that in-hospital mortality in patients who experienced ischemic strokes after PCI did not improve over the past 14 years. These findings are in line with those of previous studies and highlight the unmet need for effective prevention and treatment strategies of post-PCI stroke. Patients who experience post-PCI strokes during hospitalization are, in theory, in an ideal setting for prompt recognition and management of the stroke. This includes the use of thrombolytic agents and/or early referral for cerebral angiography with or without mechanical thrombectomy if anatomically suitable.
Mechanical thrombectomy is associated with improved functional outcomes compared with usual care alone in patients with acute ischemic stroke due to large artery occlusion (27). However, this study showed low use rates of thrombolytic agents, cerebral angiography, and mechanical thrombectomy (4.56%, 4.73%, and 0.65%, respectively), although these rates slowly increased over time. Reasons for these low use rates are complex and may include the lack of access to acute stroke interventions at many hospitals and the infancy of the interventional stroke field in general. Nonetheless, given the substantial persistent morbidity and mortality associated with post-PCI ischemic stroke, further investigations are needed to assess the utility of modern stroke interventional techniques in the management of patients with post-PCI ischemic stroke.
First, ascertainment of stroke diagnosis was the main limitation of this study. Computed tomographic imaging findings and disability score (e.g., Rankin score) in patients who had strokes were not available. The diagnosis of stroke was established solely on the basis of validated International Classification of Diseases-Clinical Modification codes, and these codes are used primarily for billing purposes and can hence be limited by undercoding, overcoding, or erroneous coding. However, the codes used in this study have been shown to have excellent sensitivity and specificity in prior studies (16,17). In addition, besides the CathPCI Registry, there is currently no clinical (nonadministrative) database that is suitable to address the issue of post-PCI stroke on a national level. Even the CathPCI Registry uses self-reporting methods and contains data from large but selective groups of voluntary participating hospitals.
Second, angiographic findings, characteristics of the PCI culprit artery, access site, and perioperative medications are not available in NIS. Hence, the impact of lesion complexity (e.g., chronic total occlusion) and the PCI techniques (e.g., atherectomy) on clinical outcomes could not be assessed.
Third, mechanical thrombectomy has only recently become popular in the management of acute ischemic strokes because of large vessel occlusion following the publication of several trials proving its efficacy (27). The low use rate in our study is likely related to the time period of our analysis (2003 to 2016). Whether the wider adoption of mechanical thrombectomy is associated with its increasing use in patients with post-PCI strokes remains to be studied.
Finally, the NIS allows detailed assessment of hard clinical endpoints during the same hospitalization. However, outcomes of patients beyond hospital discharge are not available.
Acute ischemic stroke after PCI is increasing over time, partially because of the increasing burden of comorbidities among patients undergoing PCI and the increasing complexity of PCI procedures themselves. Further studies are needed to identify effective prevention and management strategy of this PCI complication given its substantial associated morbidity and mortality.
WHAT IS KNOWN? Stroke is a rare but a devastating complication of PCI.
WHAT IS NEW? The incidence of post-PCI stroke is increasing, and its negative impact on short-term outcomes and cost remains substantial.
WHAT IS NEXT? Further research is needed to develop risk scores and risk mitigation strategies for post-PCI stroke.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- mechanical circulatory support
- National Inpatient Sample
- non–ST-segment elevation myocardial infarction
- percutaneous coronary intervention
- stable ischemic heart disease
- ST-segment elevation myocardial infarction
- unstable angina
- Received February 5, 2019.
- Revision received March 18, 2019.
- Accepted April 9, 2019.
- 2019 American College of Cardiology Foundation
- Fuchs S.,
- Stabile E.,
- Kinnaird T.D.,
- et al.
- Korn-Lubetzki I.,
- Farkash R.,
- Pachino R.M.,
- Almagor Y.,
- Tzivoni D.,
- Meerkin D.
- Guptill J.T.,
- Mehta R.H.,
- Armstrong P.W.,
- et al.
- Vranckx P.,
- Frigoli E.,
- Rothenbuhler M.,
- et al.
- Gershlick A.H.,
- Khan J.N.,
- Kelly D.J.,
- et al.
- Elgendy I.Y.,
- Mahmoud A.N.,
- Kumbhani D.J.,
- Bhatt D.L.,
- Bavry A.A.
- Ferrante G.,
- Rao S.V.,
- Juni P.,
- et al.
- Goel K.,
- Gupta T.,
- Gulati R.,
- et al.
- Wayangankar S.A.,
- Bangalore S.,
- McCoy L.A.,
- et al.
- Kolte D.,
- Khera S.,
- Aronow W.S.,
- et al.
- Kokotailo R.A.,
- Hill M.D.
- Ramirez L.,
- Kim-Tenser M.A.,
- Sanossian N.,
- et al.
- Thigpen J.L.,
- Dillon C.,
- Forster K.B.,
- et al.
- Khera S.,
- Kolte D.,
- Gupta T.,
- et al.
- Kolte D.,
- Sardar P.,
- Khera S.,
- et al.
- Valle J.A.,
- Kaltenbach L.A.,
- Bradley S.M.,
- et al.
- Iverson A.,
- Stanberry L.I.,
- Tajti P.,
- et al.
- Song C.,
- Sukul D.,
- Seth M.,
- et al.
- Elgendy I.Y.,
- Kumbhani D.J.,
- Mahmoud A.,
- Bhatt D.L.,
- Bavry A.A.