Frequency of Thrombolytic Therapy in Patients With Acute Ischemic Stroke and the Risk of In-Hospital Mortality: The German Stroke Registers Study Group

Data were collected prospectively by the treating hospital physician of an individual stroke patient. Information was documented continuously from admission to discharge on a standardized form. After discharge, these forms were sent to the regional stroke register. There, all forms were checked for plausibility and completeness before data entry. In case of implausible or incomplete data, the respective hospitals were contacted, and the data were corrected. Before starting the current analysis, we tested the logistics of data pooling and the feasibility of data analysis.¹⁷

Variable definitions and the methods of data collection were standardized among the registers. All registers agreed to a mandatory data set before pooling.¹⁴ Standardized questionnaires were used, and each study physician was guided by a manual of operation. Information on the following variables was assessed: thrombolytic treatment (intravenous use of tPA, intra-arterial thrombolysis, none); time from stroke onset to hospitalization (admission within 3 hours of stroke onset, admission after 3 hours of stroke onset, unknown stroke onset); admission from other acute facility; hypertension (reported blood pressure systolic ≥160 mm Hg or diastolic ≥95 mm Hg, or patient’s self-report of treated hypertension); diabetes mellitus (reported pathological elevated fasting blood glucose level, patient’s self-report of diabetes, or use of antidiabetic drugs); hypercholesterolemia (reported pathological elevated plasma total cholesterol level ≥240 mg/dL or use of lipid-lowering medication); previous stroke (neurological deficit >24 hours before the current event); atrial fibrillation (documented by ECG); neurological deficits of current stroke (weakness or paresis, aphasia, dysarthria, and disturbances of consciousness); discharge destination (dead, home, residential/nursing home, rehabilitation unit, other hospital); and pathological subtype of stroke (cerebral infarction, primary intracranial or subarachnoidal hemorrhage, transient ischemic attack, undefined type). Stroke was defined according to the World Health Organization criteria.¹⁸ Stroke subtype was diagnosed from the results of the first CT or MRI scan. The experience of an individual hospital in tPA administration was defined by the number of patients treated with tPA per year. This definition is based on the study of Katzan and colleagues.⁹ In that study, the proportion of patients receiving thrombolytic therapy was substantially increased in hospitals administering tPA to >5 patients per year. Therefore, in our analysis, hospitals administering tPA to ≤5 patients in 2000 were classified as centers with limited experience, and hospitals treating >5 patients were classified as experienced centers, regardless of other hospital characteristics.

Obviously, stroke patients in observational studies are not assigned at random to receive thrombolytic treatment. Therefore, the current data set was analyzed with a propensity score approach. The propensity score defines the probability of each individual patient to be treated based on a given set of covariates.¹⁹ The use of propensity score analyses balances the distribution of covariates between treatment and control groups and therefore minimizes the influence of potential biases.^20,21 Variables related to the decision to treat patients with tPA were included in a logistic regression model. The resulting propensity score for the treatment of ischemic stroke patients with tPA included the following 14 variables: age (continuous), sex, diabetes mellitus, hypertension, hypercholesterolemia, atrial fibrillation, previous stroke, admission from other acute facility, admission within 3 hours of stroke onset, presence of limb paresis or weakness, presence of aphasia, presence of dysarthria, disturbed level of consciousness, and the treating hospital. Interaction terms between sex, time of hospitalization, neurological deficits, and risk factors were added to the model. Calculation of the propensity score was restricted to centers administering tPA in 2000. The multivariate regression model for the propensity of being treated with tPA had a c statistic of 0.94 (95% CI, 0.93 to 0.95), which represents the area under the receiver-operating characteristic curve and indicates very good discrimination between patients who did and who did not receive thrombolytic therapy.

Ischemic stroke patients treated with tPA were matched individually to patients not receiving tPA within the same hospital. For constructing the matched samples, the nearest available matching on the estimated propensity score method as described by Rosenbaum and Rubin²² was used. Treated subjects within a hospital were ordered at random; then, the first treated subject was selected and matched with the control subject with the nearest logit of the propensity score within the same hospital. Both cases and controls were then removed, and the next treated subject was chosen. Altogether, the propensity scores of 90% of the matched control subjects were within the defined caliper (one quarter of an SD of the logit of the propensity score of subjects treated with tPA²²).

The t test was used to test differences in continuous variables, and the χ² test was used for those in proportions. To estimate the odds ratio (OR) and the resulting 95% CI for the matched case-control pairs, conditional logistic regression was performed. The risk of in-hospital mortality was calculated before and after adjustment for the number of patients treated with tPA per hospital. Subgroup analyses were performed by running different regression models for patients admitted within 3 hours and after 3 hours of stroke onset. Eight patients treated with tPA and 745 patients not treated with tPA were excluded from statistical analyses because of missing values. The 77 patients treated with intra-arterial thrombolysis were removed from the study because it was restricted to intravenous application. To estimate potential selection effects in this hospital-based cohort, we compared its age and sex distributions to those of the unselected population-based Erlangen Stroke Project, which includes all hospitalized and all nonhospitalized stroke patients within a defined region of 100 330 inhabitants.²³ All tests were 2 tailed, and statistical significance was determined at α=0.05. Statistical analyses were performed with the SPSS 10.0 software package.

A total of 384 patients were treated with tPA (range per hospital, 1 to 32); 351 of them were admitted within 3 hours of stroke onset (range per hospital, 1 to 27) (Table 1). We found that 2.9% of all ischemic stroke patients and 10.4% of patients admitted within 3 hours of stroke onset received thrombolytic therapy.

The overall cohort had a mean propensity score of 0.04 (median, 0.002; range, 0 to 0.88) for receiving thrombolytic therapy, indicating a low probability of being treated with tPA within the ADSR network. For patients treated with tPA, the mean propensity score was 0.28 (median, 0.24; range, 0.001 to 0.88) compared with 0.03 (median, 0.002; range, 0 to 0.81) for patients not receiving treatment. Patients receiving tPA were significantly younger, more often male, and more often admitted within 3 hours of stroke onset (Table 2). Patients treated with tPA had a significantly lower prevalence of diabetes, hypertension, and previous stroke and a higher prevalence of atrial fibrillation. Differences in the presence of neurological signs were also observed, indicating a higher stroke severity among patients receiving thrombolytic therapy. After matching for propensity score, no significant differences in baseline characteristics, comorbidities, and presence of neurological deficits were found between patients treated with tPA and matched control patients not receiving thrombolytic treatment (Table 2).

In-hospital mortality for patients treated with tPA was significantly higher than for matched control subjects (11.7% versus 7.4%, respectively; P=0.047) and for all patients not treated with tPA (4.5%, P<0.0001) (Table 3). Stratification of the patients according to admission within 3 hours of stroke onset yielded a similar but nonsignificant mortality difference between cases and controls in those with early admission (11% versus 7.6%, respectively; P=0.12). This difference increased considerably among those admitted later than 3 hours after stroke onset (18.8% versus 6.3%, respectively; P=0.13). However, because of the small number of cases and controls, this difference was not statistically significant.

Substantial differences in rates of in-hospital death were observed after stratification by hospital experience with tPA use (Table 3). Hospitals that administered tPA to ≤5 patients in 2000 had a significantly increased rate of in-hospital death for patients treated with tPA compared with matched controls not receiving thrombolytic treatment (24% versus 8.6%, respectively; P=0.024). In contrast, no significant differences in rates of in-hospital death were observed between patients treated in hospitals administering >5 thrombolytic therapies per year and matched controls (9.4% versus 7.2%, respectively; P=0.32). To investigate a potential bias by differences in time under observation, length of stay was compared between hospitals administering 1 to 5 thrombolytic therapies and hospitals administering tPA to ≥5 patients. No statistically significant difference in mean length of stay for tPA patients and matched controls was found among less experienced and experienced centers (13.3 versus 13.2 days, respectively; P=0.91).

In multivariate analyses, patients treated with tPA in hospitals with limited experience in tPA application had an increased risk of in-hospital death (OR, 3.3; 95% CI, 1.1 to 9.9) (Table 4). No significant increase in risk of in-hospital mortality was found for patients treated with tPA in experienced hospitals (OR, 1.3; 95% CI, 0.8 to 2.4). Restricting analyses to patients admitted within 3 hours of stroke onset demonstrated similar risks, although the OR of inpatient death in hospitals with limited experience in the use of tPA did not reach statistical significance (OR, 2.8; 95% CI, 0.9 to 8.6).

About 3% of all ischemic strokes and 10.4% of patients admitted within 3 hours of stroke onset were treated with tPA within the hospitals of the ADSR network. These proportions were comparable to those in previous multicenter studies reporting on the rates of tPA use outside clinical trials: from 1.6%⁸ to 6.3%⁷ of all ischemic strokes and 10.4% of patients admitted within 3 hours of stroke onset.⁹ The rates of patients with ischemic stroke receiving thrombolysis varied from 0.5% to 18% between hospitals. The highest rate of tPA administration in our study (18%) was similar to that reported from a single-center study in Germany with a special designed referral system for early hospital admission of acute stroke patients.⁵ Previous studies on thrombolytic therapy in routine clinical practice found rates from 8%⁴ to 13%^3,9 of patients receiving tPA beyond the 3-hour time window. In our analysis, 8.5% of patients were treated with tPA, although they were admitted to hospital >3 hours after stroke onset. However, the true rate of thrombolytic therapies administered outside the 3-hour time window in our study might be higher because a median time of 55 minutes from admission to hospital to the start of tPA treatment must be added to the admission time.²⁴

Half of the hospitals within the ADSR network administered tPA. Thrombolytic therapy was administered more often in neurological departments, in hospitals providing stroke unit services, and in centers treating >100 ischemic stroke patients per year. This variance in administration of tPA might be caused by differences in institutional attitudes, patient characteristics, or individual concerns that treating physicians have about safety and efficacy of thrombolytic therapy. A recently published study by Katzan and colleagues²⁵ reported that only 30% of attending neurologists were very convinced about the efficacy of tPA use, whereas 62% were very concerned about the risk of thrombolysis.

A meta-analysis of clinical trial data demonstrated no significant excess of early deaths after tPA use.¹ Four previous studies reported on in-hospital mortality after tPA use outside clinical trials in multicenter settings.^4,7–9 The proportion of patients who died in hospital varied from 9%⁷ to 15.7%.⁹ Two studies suggested that administering tPA outside clinical trials did not cause an excess in inpatient death^4,7; the other reports showed an increased risk of in-hospital mortality among tPA patients.^8,9 These divergent results might be caused by differences in the respective control groups chosen for the comparison with the mortality of tPA patients. The studies,⁴ which found tPA use to be safe in routine clinical practice, compared their outcomes with the results of the National Institute of Neurological Disorders and Stroke rt-PA Stroke Study Group trial.² The studies, which showed an increased risk of inpatient death after thrombolysis, compared in-hospital mortality after tPA use with patients not receiving tPA recruited from the same study population that gave rise to tPA patients.^8,9 Characteristics of patients included in randomized controlled trials might not be representative of study populations in routine clinical practice. Therefore, a comparison of outcomes between patients treated with tPA in community settings with control subjects from randomized controlled trials is difficult. A further explanation for the divergent results on outcome after tPA treatment could be the fact that, as in our analysis, rates of in-hospital-death varied with the number of thrombolytic therapies administered by the respective hospital per year. The 2 studies that found tPA use to be safe in clinical routine were performed in 13 hospitals experienced in the administration of tPA⁴ and in 20 hospitals experienced in the use of tPA or integrated in a special stroke treatment network.⁷ The observed in-hospital mortality rate of 9.4% for tPA patients treated in hospitals with >5 tPA applications per year in our study was comparable to the mortality reported in these 2 studies (9.5%⁴ and 9%⁷). An increased risk of inpatient death after thrombolysis was found in studies that either included all 29 hospitals within a defined region⁹ or reported on 137 unselected community hospitals that participate in the Healthcare Benchmarking Database.⁸ The multicenter approach suggested that, similar to our analysis, hospitals with different experience levels in tPA use were included in these studies, although this was not explicitly addressed in the study reports. The overall risk of in-hospital mortality for patients treated with tPA compared with patients not receiving tPA was similar in the study of Reed and colleagues⁸ (OR, 1.8; 95% CI, 1.2 to 2.7) and in our analysis (OR, 1.7; 95% CI, 1 to 2.8). The higher proportion of in-hospital deaths in the study of Katzan and colleagues compared with our results (15.7% versus 11.7%, respectively) might be due to the fact that, in that specific study, a higher proportion of hospitals with limited experience in tPA use (≤5 thrombolytic treatments per year) was found than in our study (75% versus 54%, respectively).⁹ It is noteworthy that the rate of in-hospital deaths among matched participants not treated with tPA in the analysis of Katzan and colleagues was similar to our results (7.2% versus 7.4%, respectively).⁹ The degree of the individual hospital’s experience in tPA use can be affected by several factors, eg, expertise in CT scan reading, the way classification of stroke severity is done, the identification of exclusion criteria, implementation of training programs to educate physicians, or creation of specific guidelines for tPA use. Previous studies suggest that between 31%²⁶ and 50%⁹ of patients receiving tPA in routine clinical practice were not treated in accordance with national treatment guidelines. Thus, differences in outcome after tPA use between hospitals with limited and high experience in application of tPA might be caused by different rates of violations against existing protocols on thrombolytic treatment. The only available information in this context in our study was the time difference between stroke onset and hospital admission. Therefore, we were unable to further clarify whether protocol violations contributed to the observed differences of in-hospital deaths between hospitals with different levels of experience in tPA treatment. Future studies should aim to carefully assess potential protocol violations, taking into account the sensitiveness of the issue, to reveal their potential impact.

Our study has several strengths and limitations. In contrast to randomized controlled trials, which balance all observed and unobserved factors between treatment and control groups, propensity score analyses balance only covariates that were assessed and used to construct the score. However, in several instances, propensity score matching provides better control for matched factors than randomized controlled trials.²¹ We cannot exclude that some unobserved differences in patient characteristics between the cases and control subjects contributed to our results. However, we minimized potential confounding by regional differences in treatment processes or in patient characteristics by recruiting control subjects from the same hospitals as cases. Some of the detected variations in early outcome might be caused by differences in clinical characteristics (eg, different proportions of patients with cerebellar or brain stem infarcts) or stroke severity between patients treated with tPA and untreated patients. However, we controlled the statistical analyses for neurological deficits present on admission, especially for a disturbed level of consciousness, which was identified to be a good predictor of stroke severity.²⁷ We were not able to present detailed explanations about the decision to exclude patients admitted within 3 hours of stroke onset from thrombolytic therapy or about reasons to treat patients admitted after 3 hours of stroke onset with tPA. However, our study provides valid and reliable data because data collection within the stroke registers of the ADSR network was standardized, a predefined variable set was used,¹⁴ and the feasibility of the procedures was tested.¹⁷ In our study, no information was provided on mortality in the time period after discharge from acute hospital. Therefore, a potential long-term benefit of thrombolysis on mortality could not be evaluated.

In summary, our study identified considerable variance in the impact of tPA use on in-hospital mortality of patients with ischemic stroke in routine clinical practice. Hospitals performing >5 thrombolyses per year had a lower in-hospital mortality compared with hospitals with fewer tPA treatment applications. Our results provide evidence that thrombolytic treatment in patients with ischemic stroke should be administered preferably in hospitals experienced in tPA use.

Amberg: Neurologische Abteilung des Klinikums St Marien; Aschaffenburg: Neurologische Klinik des Klinikums; Bad Aibling: Neurologische Klinik Bad Aibling GmbH; Bad Kissingen: Medizinische Klinik des St Elisabeth-Krankenhauses; Bad Neustadt: Neurologische Klinik GmbH; Bad Reichenhall: Innere Abteilung des Staedtischen Krankenhauses; Bayreuth: Neurologische Klinik des Krankenhauses Hohe Warte; Eggenfelden: Abteilung fuer Innere Medizin des Kreiskrankenhauses; Erlangen: Neurologische Klinik der Universität Erlangen-Nürnberg; Guenzburg: Klinik fuer Neurologie des Bezirkskrankenhauses; Haar: Neurologische Klinik des Bezirkskrankenhauses; Hausham: Geriatrie des Krankenhauses Agatharied; Hoechstadt an der Aisch: Abteilung fuer Innere Medizin des Kreiskrankenhauses St Anna; Ingolstadt: Neurologische Klinik des Klinikums; Lohr am Main: Krankenhaus des Bezirkes Unterfranken; Muenchen: Neurologische Klinik der Ludwig-Maximilians-Universitaet Muenchen, Neurologische Klinik der Technischen Universitaet Muenchen, Abteilung für Neurologie des Staedtischen Krankenhauses Bogenhausen, Abteilung für Neurologie des Staedtischen Krankenhauses Harlaching; Murnau: Abteilung fuer Innere Medizin II des Klinikums Garmisch-Partenkirchen; Nuernberg: Neurologische Klinik des Klinikums, Medizinische Klinik 4 des Klinikums; Neuendettelsau: Innere Abteilung des Diakoniewerkes Neuendettelsau; Rosenheim: Neurologische Klinik des Klinikums; Rotenburg: Innere Abteilung des Krankenhauses Rotenburg o.d.T. GmbH; Starnberg: Medizinische Klinik des Kreiskrankenhauses GmbH; Traunstein: Innere und Neurologische Abteilung des Kreiskrankenhauses Traunstein; Werneck: Innere Abteilung des Krankenhauses Markt Werneck; Wolfratshausen: Innere Abteilung des Kreiskrankenhauses; Wuerzburg: Neurologische Klinik der Universität Wuerzburg. Coordinating center: Geschäftsstelle der Bayer, Arbeitsgemeinschaft für Qualitätssicherung in der stationären Versorgung, BAQ, München.

Hamburg: Neurologische Klinik des Universitätsklinikums Hamburg-Eppendorf, Neurologische Abteilung des Allg Krankenhauses Barmbek (LBK Hbg), Neurologische Abteilung des Allg Krankenhauses St Georg (LBK Hbg), Neurologische Abteilung des Allg Krankenhauses Harburg (LBK Hbg), Neurologische Abteilung des Allg Krankenhauses Wandsbek (LBK Hbg), Neurologische Abteilung des Allg Krankenhauses Altona (LBK Hbg), Neurologische Abteilung des Klinikums Nord (LBK Hbg), Abteilung für Innere Medizin des Evangelischen Krankenhauses Alsterdorf, Medizinische Abteilung des Bethesda Allg Krankenhauses Bergedorf, Neurologische Klinik des Katholischen Marienkrankenhauses, Innere Abteilung des Evangelisches Amalie Sieveking Krankenhauses, Neurologische Abteilung des Allg Krankenhauses Eilbek (LBK Hbg), Abteilung für Neurologie und Psychiatrie des Albertinen Krankenhauses, Abteilung fuer Neurologie des Krankenhauses Itzehoe, Medizinische Abteilung des Wilhemsburger Krankenhauses “Gross Sand,” Innere Abteilung des Allg Krankenhauses Eilbek (LBK Hbg), Neurologische Abteilung des Krankenhauses Buchholz, Buchholz in der Nordheide. Coordinating center: Geschäftsstelle der Arbeitsgemeinschaft Externe Qualitätssicherung (EQS), Hamburg.

Bad Zwesten: Neurologische Akutklinik; Darmstadt: Neurologische Klinik des Klinikums; Eschwege: Medizinische Abteilung des Kreiskrankenhauses; Friedberg: Innere Abteilung des Kreiskrankenhauses Friedberg; Fulda: Neurologische Klinik des Klinikums; Frankfurt/Main: Neurologische Klinik der Universitaet Frankfurt/Main, Neurologische Klinik des Krankenhauses Nordwest, Neurologische Klinik des St Katharinen-Krankenhauses, Neurologische Klinik der Staedtischen Kliniken Hoechst; Gelnhausen: Medizinische Klinik des Kreiskrankenhauses; Kassel: Neurologische Klinik des Klinikums, Geriatrische Abteilung des Kurhessisches Diakonissenhauses; Limburg: St Vincenz Krankenhaus; Lindenfels: Innere Abteilung des Luisenkrankenhauses; Marburg: Neurologische Klinik der Universitaet Marburg; Nidda-Bad Salzhausen: Asklepios Neurologische Klinik; Offenbach: Neurologische Klinik des Klinikums, Medizinische Klinik I des Klinikums; Schwalmstadt: Fachklinik fuer Neurologie der Hepataklinik; Seeheim-Jugenheim: Medizinische Abteilung des Kreiskrankenhauses; Weilmuenster: Neurologische Klinik des Klinikums Weilmuenster; Wiesbaden: Medizinische Klinik I des St Josef-Hospitals, Neurologische Klinik der Dr. Horst-Schmidt-Kliniken. Coordinating center: Geschäftsstelle Qualitätssicherung Hessen, Eschborn.

Arnsberg: Medizinische Klinik des Marienhospitals; Bad Pyrmont: Neurologische Klinik des Bathildis Krankenhauses; Bocholt: Medizinische Klinik des St Agnes Hospital; Bochum: Neurologische Klinik Bergmannsheil, Neurologische Klinik des St Josef-Hospitals, Medizinische Klinik des St Elisabeth-Hospitals; Bottrop: Neurologische Klinik des Knappschafts-Krankenhauses; Castrop-Rauxel: Neurologische Klinik des Evangelischen Krankenhauses; Dortmund: Neurologische Abteilung des Knappschaftskrankenhauses, Neurologische Klinik der Städtischen Kliniken, Medizinische Klinik des St Josefs-Hospitals, Medizinische Klinik des Krankenhauses Bethanien, Geriatrische Klinik des Hüttenhospitals; Duelmen: Neurologische Klinik des Franziskus Hospitals; Emden: Neurologische Klinik des Hans-Susemihl-Krankenhauses; Gelsenkirchen: Neurologische Klinik des Evangelischen Krankenhauses; Hagen: Neurologische Klinik des St. Johannes Hospitals; Hamm: Neurologische Klinik des St Marien-Hospitals; Hattingen: Neurologische Klinik des Evangelischen Krankenhauses; Heesen: Medizinische Klinik der St.Barbara Klinik; Herdecke: Neurologische Klinik des Gemeinschaftskrankenhauses; Herne: Geriatrische Klinik des Marienhospitals; Lippe-Lemgo: Neurologische Klinik des Klinikums; Luebbecke: Medizinische Klinik des Krankenhauses; Minden: Neurologische Klinik des Klinikums; Muenster: Neurologische Klinik der Universitaet Muenster, Medizinische Klinik des Franziskus-Hospitals; Osnabrueck: Neurologische Klinik des Klinikums; Paderborn: Neurologische Klinik des St Vincenz Krankenhauses; Recklinghausen: Neurologische Klinik des Knappschafts-Krankenhauses, Geriatrische Klinik des Elisabeth-Krankenhauses; Warendorf: Medizinische Klinik des Joseph-Hospitals; Wuppertal: Neurologische Klinik des Klinikums. Coordinating center: Institut für Epidemiologie und Sozialmedizin, Universität Münster.