Use of Tissue-Type Plasminogen Activator for Acute Ischemic Stroke: Title and subTitle BreakThe Cleveland Area Experience FREE

In December 1995, the National Institute of Neurological Disorders and Stroke (NINDS) rt-PA Stroke Study Group reported a benefit with intravenous tissue-type plasminogen activator (IV tPA) for patients with acute ischemic stroke of less than 3 hours' duration.¹ The Food and Drug Administration (FDA) subsequently approved IV tPA for acute ischemic stroke in June 1996. The use of IV tPA has caused a dramatic change in the way acute ischemic stroke is approached. However, IV tPA carries a 10-fold increased risk of intracerebral hemorrhage (ICH).¹ To minimize the risks of thrombolytic treatment, guidelines have been devised by the American Heart Association,^{2 - 3} American Academy of Neurology,⁴ and NINDS.⁵ Because of the risks of hemorrhage and the narrow therapeutic window, IV tPA as a treatment for acute ischemic stroke remains controversial.^{6 - 9}

Several series on tPA use outside the NINDS trial setting have reported symptomatic ICH rates similar to the NINDS trial^{10 - 18} (Table 1). However, most of these series involved stroke centers experienced in the administration of IV tPA, and several of the sites participated in the NINDS or Alteplase ThromboLysis for Acute Noninterventional Therapy in Ischemic Stroke (ATLANTIS) trials of IV tPA in early stroke.^{10 - 12 ,15 - 16 ,18} Data were obtained by voluntary reporting in other series.^{13 - 14 ,17}

Information on the use of tPA in the community setting has been hampered by the lack of a systematic method of data collection. Cleveland, Ohio, has been in an unusual position to study the use of IV tPA in the community setting. From 1991-1999, standardized data were collected on every stroke patient admitted to 29 Cleveland area hospitals by the Cleveland Health Quality Choice (CHQC) project. Risk adjustment was done using CHOICE prediction models that use proprietary formulas to predict in-hospital mortality and length of stay (LOS). Using the CHQC data collection system, we assessed the rate of IV tPA use in Cleveland, the incidence of symptomatic ICH, and in-hospital patient outcomes after tPA treatment. We also assessed physician compliance with national treatment guidelines for IV tPA use in acute stroke.

The CHQC system was used to collect data on all patients admitted for stroke to 29 Cleveland area hospitals over a 1-year period from July 1997 through June 1998. This included all hospitals in the metropolitan Cleveland area except for the local Veterans Affairs hospital and comprised all hospitals in Cuyahoga, Lake, Lorain, and Geauga counties. Hospitals had a mean bed size of 324 (range, 46-1153). Twenty-one were private not-for-profit institutions; 7, private for-profit; and 1, a public hospital. Ten of the hospitals had accredited residency programs (8 medical, 2 osteopathic). Stroke patients were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) principal diagnosis codes 431 to 436 and 997.02.

Data were collected in 6-month "waves." Hospital identity was anonymous, although the volume of stroke admissions was available. No specific informed consent statement was signed, and no patient identifying data were produced. Retrospective chart reviews were performed by independent, trained quality assurance and medical records personnel at each hospital guided by an operations manual. Demographic, physiologic, and laboratory variables, as well as all secondary ICD-9-CM diagnosis and procedure codes, were collected for each patient with stroke.²⁰ Several processes were established to ensure data reliability^{20 - 21} including double-keystroke data entry and independent evaluation of reliability of data abstraction at each hospital.

In 1997, a separate data form was developed for patients receiving IV tPA. Data points on this form included presence of symptomatic ICH, specialty of physican involved in tPA administration, time from symptom onset to initiation of tPA, blood pressure control (before and after tPA treatment), use of antithrombotic medications within 24 hours of start of tPA treatment, and National Institutes of Health Stroke Scale (NIHSS) score documentation.

Two experienced abstractors audited the charts of all patients receiving IV tPA to ensure the accuracy of the collected data, with special attention to the coding of symptomatic ICH. Because of the low number of patients receiving IV tPA at any individual hospital, there was a potential for abstractors reviewing charts to overlook IV tPA use. Transfer of stroke patients from an outside emergency department was another potential source of abstraction error. To minimize underdetection of tPA use, an independent audit was performed of all identified patients with stroke hospitalized within 3 hours of symptom onset as well as all emergency department transfers within 12 hours (n = 800). To cross-check the accuracy of tPA use identification, we also determined the frequency of IV tPA use at 8 hospitals by means of hospital-specific tracking systems already in place. Finally, we obtained corroborating information from local neurologists.

Comparisons of baseline variables and outcomes were performed among all patients with ischemic stroke (ICD-9-CM codes 434 and 436) receiving and not receiving tPA. A subgroup of ischemic stroke patients admitted within 3 hours of symptom onset but who did not receive IV tPA was identified. A bootstrap resampling method²² was used to match these patients to the tPA-treated patients by frequencies for age and history of stroke. Patients with a diagnosis of ICD-9-CM code 433 (occlusion/stenosis of precerebral artery) and 435 (transient cerebral ischemia) were excluded from these analyses because of known diagnostic inaccuracies associated with these codes.^{23 - 24}

Outcome measures included ICH rate, in-hospital mortality, LOS, and discharge destination. To adjust for severity of illness, modifications of the CHOICE predicted mortality and predicted LOS models were used. In the CHOICE models, the in-hospital stroke mortality rate and LOS are predicted using proprietary formulas developed specifically for patients with stroke. The models were based on demographic and clinical data from the first 48 hours of hospitalization including comorbid conditions, admission vital signs, physical findings, and results of laboratory and other diagnostic tests.^{20 ,25 - 26}

Initial models were developed in patients discharged in 1991 based on clinical variables that were independently (P<.01) related to in-hospital death or LOS in logistic or linear regression analyses. Initial models were validated in an independent cohort of patients (discharges, July-December 1992). Performance of the mortality models was assessed by receiver operating characteristic (ROC) curve analysis²⁷ to examine discrimination and the Hosmer-Lemeshow test²⁸ to examine calibration. The ROC curve area, which is numerically equivalent to the c statistic, represents the proportion of times that patients who died had a higher predicted risk of death than patients who survived for all possible pairwise comparisons of patients who died and who were discharged alive. An ROC curve area of 0.5 represents a model with no discrimination, while an area of 1.0 represents a model with perfect discrimination. Performance of the LOS models was assessed by the explained variance (R²) and by analysis of residuals. As data from subsequent time periods were collected, models were reestimated using the larger data set to provide more stable estimates of regression coefficients for each variable.

Intracerebral hemorrhages and subsequent "do not resuscitate" (DNR) orders are potential direct results of tPA treatment. Because the purpose of the prediction models was to adjust for severity of illness based on baseline variables rather than those due to tPA treatment, the variables ICH and DNR were excluded from the models.

Predicted mortality and LOS were compared among 3 subgroups: patients treated with tPA, matched patients admitted within 3 hours of stroke onset and not treated with tPA, and all ischemic stroke patients. Statistical significance was determined using χ² or Fisher exact test for categorical variables and the t test for quantitative variables. The Wilcoxon rank sum test was used for quantitative data series with nonnormal distributions, except for LOS, which was normalized by a log transformation.

A total of 4986 patients with the primary diagnosis of stroke were admitted to 29 Cleveland area hospitals from July 1997 through June 1998. There were 4345 patients with ischemic strokes (ICD-9-CM codes 433, 434, and 436), of which 3948 were diagnosed under ICD-9-CM code 434 or 436. Of these patients, 17% were admitted within 3 hours of stroke onset. Intravenous tPA was administered to 70 patients (1.8% of ischemic strokes) and was given at 16 of the 29 hospitals to a median of 3 patients (range 1-17) (Table 2). The rate of IV tPA use among hospitals varied from 0% to 10.2% of stroke admissions and was not associated with total admission volume of patients with stroke.

Baseline characteristics and comorbid conditions were similar in patients treated with tPA and the matched patients seen within 3 hours but not treated with tPA except for initial blood pressure and valvular heart disease (Table 3). The general ischemic stroke population was older and had higher rates of prior symptomatic cerebrovascular disease and diabetes mellitus. Recorded neurological signs were similar among all 3 groups except for Babinski sign, which was documented more frequently in patients receiving IV tPA.

The total rate of ICH was 22% in patients receiving tPA (16/70, 95% confidence interval [CI], 13.7%-34.4%). Symptomatic ICH occurred in 15.7% of these patients (11/70, 95% CI, 8.1%-26.4%), and was fatal in 6 patients. Ten of the symptomatic hemorrhages occurred within 24 hours of tPA administration. Patients with symptomatic ICH were older and had higher blood glucose levels than patients without ICH (Table 4). There was no statistically significant association between symptomatic ICH and hospital stroke admission volume.

Six additional cases of IV tPA use were identified in patients admitted under primary ICD-9-CM codes other than stroke. Three of these patients were admitted for cardiac disease and developed stroke while in the hospital. One additional patient was admitted with nausea and vertigo and was diagnosed as having benign positional vertigo. He developed other focal neurological symptoms on day 3 of hospitalization and was given IV tPA. No additional information is available on the remaining 2 patients. One of these 6 patients had a fatal ICH. When these patients are included in the number of patients receiving tPA, the total symptomatic hemorrhage rate remains essentially unchanged at 15.8% (12/76).

The in-hospital mortality rate for the IV tPA-treated patients (15.7%) was significantly higher than that for the matched patients seen within 3 hours but not treated with tPA (7.2%) (P = .01) and the general population of patients with stroke not receiving tPA (5.1%) (P<.001) (Table 5). Six of the 11 deaths among the patients receiving tPA occurred in patients with symptomatic ICH. The mortality rate of the patients receiving tPA was also significantly higher than the CHOICE predicted mortality of 7.9% (P<.006). Mortality rates for patients seen within 3 hours but not treated with tPA and the general population of patients with stroke not receiving tPA were similar to their predicted values.

Patients receiving IV tPA also had significantly longer hospital stays than patients seen within 3 hours but not receiving tPA and all patients with stroke who did not receive tPA. However, the predicted LOS of the patients receiving tPA was also longer, and there was no significant difference between actual and predicted LOS for either the patients receiving tPA or patients not receiving tPA. R² for the LOS prediction model was 0.245, explaining approximately 25% of the statistical variability. Factors in the LOS model that were significantly different between the tPA patients and other subgroups were placement of a percutaneous gastrostomy (PEG) tube during hospitalization and an abnormal serum bicarbonate level within the first 2 days of admission. Patients treated with tPA were discharged home significantly less often than those seen within 3 hours but not treated with tPA.

Deviations from national treatment guidelines^{2 - 4} were identified in 50% of patients treated with IV tPA (Table 6). There was no statistically significant association between the presence or type of protocol deviations and symptomatic ICH (Table 4). The use of antiplatelet agents or anticoagulants within 24 hours of tPA administration was the most frequent protocol deviation and occurred in 37.1% of patients. Intravenous heparin was the most common agent started within that period (n = 16), followed by aspirin (n = 7), combination of IV heparin plus aspirin (n = 2), or ticlopidine (n = 1). Patients treated within 5 minutes past the 3-hour time window (n = 4)were not counted as having deviated from protocol. Treatment beyond this time occurred in 12.9% cases and ranged from 3 hours 10 minutes to 6 hours 13 minutes from symptom onset.

Neurologists were directly involved in 95.5% of IV tPA treatment decisions and were usually present at the time of IV tPA administration (Table 6). Neurologists made the final interpretation of the head computed tomography scan in 43 (86%) of the 50 cases in which this information was recorded. The NIHSS score was documented in 40% of cases, with a median score of 12 (range, 4-21). Recommended blood pressure monitoring parameters were followed in 47.8% of patients. Deviations from blood pressure guidelines included both frequency of monitoring and exceeding maximum blood pressure limits in 86.1% of variant cases. Admission orders typically stated "vital signs per routine." An additional 2 patients (5.5%) had only incorrect maximum blood pressure limits, and 3 patients (8.3%) did not have blood pressure checked with recommended frequency.

Between the first 6 months and second 6 months of the study, there was a nonsignificant increase in the rate of IV tPA use (1.4% to 2.1%, P = .12) and a decline in the symptomatic ICH rate (17.9% to14.3%, P = .13). There was also a nonsignficant decline in the occurrence of protocol deviations (Table 5). Documentation of NIHSS scores and adherence to recommended blood pressure monitoring parameters occurred significantly more frequently in the second 6 months of the study.

The rate of IV tPA use for ischemic strokes was only 1.8% in the Cleveland metropolitan area 2 years after FDA approval and 10.4% for ischemic stroke patients arriving within 3 hours of symptom onset. There is little other information on citywide rates of IV tPA use except for in Cologne, Germany, where a centralized admission system is in place for all patients with stroke, and the IV tPA rate is 5.1%.¹⁵ Most reports on IV tPA usage rates are from individual hospitals^{1 ,10 ,15 ,18} and range from 1.1% of ischemic stroke admissions in 2 Houston community hospitals to 5.7% in a Houston-area university hospital that participated in the NINDS trial.

Rates of IV tPA use among the 29 individual Cleveland metropolitan area hospitals ranged from 0% to 10.2% of all ischemic stroke admissions. The majority of patients receiving IV tPA (89%) were admitted to hospitals admitting 150 or more stroke patients per year, although the highest rate was in a hospital that admitted only 39 stroke patients. Two of the 10 institutions with the highest stroke patient volumes did not treat a single patient with IV tPA. This variation may reflect differences in patient populations, institutional attitudes and capabilities, and emergency medical services transfer patterns.

The 3-hour treatment window for IV tPA is a major factor in the low usage rate. Only 17% of ischemic stroke patients arrived at Cleveland hospitals within 3 hours of symptom onset and of these, only 10.4% received IV tPA. The reasons for exclusion in patients arriving within 3 hours cannot be assessed from our data. In the NINDS trial, only 3.6% of patients screened were treated. Only 4% of patients arriving within 125 to 170 minutes after stroke onset received tPA compared with 48% of those arriving within 120 minutes.²⁹ The main reasons for exclusion in the NINDS trial were rapid improvement, ICH on baseline computed tomography scan, minor symptoms, and age.

This is the first published study to suggest that the community experience with IV tPA for acute stroke may be different from the NINDS trial. The symptomatic ICH rate after IV tPA in the Cleveland metropolitan area was 15.7%. This is higher than other post-NINDS reports (Table 1) and 2.5 times greater than in the NINDS trial. In the Standard Treatment with Activase to Reverse Stroke (STARS) study, which includes data collection from 56 centers, symptomatic ICH occurred in 3% of the first 296 patients enrolled.¹² Participation was voluntary and most STARS centers participated in the ATLANTIS IV thrombolysis trial. Hanson et al¹⁷ examined tPA use in community hospitals in the area surrounding Minneapolis, Minn. Data were obtained by voluntary reporting. The symptomatic ICH rate in the first 97 patients treated with IV tPA was 9%.

The higher ICH rate in Cleveland partly reflects the method of case ascertainment that used a standardized and systematic method of data collection by the CHQC project. It is noteworthy that the second highest reported symptomatic ICH rate (12%) was in a preliminary study performed by retrospective review of tPA cases in 7 Indianapolis, Ind, hospitals.¹⁹ While that study involved hospitals in only a single region, the 29 hospitals examined here represented a wide spectrum of teaching and nonteaching facilities. It is likely representative of mid- to large-sized cities in the United States. Because hospital identity was anonymous in this study, we were unable to examine the relationship between hospital characteristics and tPA-related hemorrhagic complications.

Caution must be used in comparing the rates of symptomatic hemorrhage after IV tPA in the Cleveland metropolitan area to that reported in the NINDS trial because the patient populations may be different. We have limited data on initial stroke severity, an important determinant of symptomatic ICH risk.^{30 - 31} In the 40% of patients receiving tPA for whom an NIHSS score was available, the median score was 12, which is similar to the median score of 14 in the NINDS trial. Baseline characteristics such as age, comorbidities, and blood pressure in the Cleveland series were also similar to the NINDS trial with the exception of valvulvar heart disease (19.4% vs 8.3%, respectively). We have no information on patient use of antiplatelet or anticoagulant therapies.

Half of the patients receiving tPA in Cleveland had deviations from national treatment guidelines. We could not determine the reasons for the protocol deviations. Given the high rate of early use of anticoagulants and antiplatelet agents, perhaps some institutions were following other experimental thrombolysis protocols, or patients may have required antiplatelet or anticoagulant agents within 24 hours of tPA administration for conditions such as myocardial infarction or pulmonary embolism.

There was no statistically significant association between protocol deviations and symptomatic ICH in our study, although an association has been reported by others.^{14 ,16 ,19} In a Canadian series of 68 patients receiving IV tPA at 1 institution, patients with protocol deviations had a 27% frequency of symptomatic ICH vs 5% for patients without deviations.¹⁶ A US multicenter series demonstrated a 30% incidence of national guideline deviations among 189 patients treated with IV tPA. The symptomatic ICH rate was 10.7% in patients with protocol deviations vs 3.8% in those without.¹⁴ In our series, there was a trend toward a higher symptomatic ICH rate in patients with no NIHSS score documentation and in those whose blood pressure monitoring parameters varied from national guidelines. These may be surrogate markers for familiarity with thrombolysis and acute stroke management. Patients with symptomatic hemorrhage were also older and had higher recorded glucose levels the first 2 days of hospital admission. Both of these factors have been shown in prior studies to increase the risk of symptomatic ICH.^{13 ,32 - 33}

In the NINDS trial, IV tPA treatment resulted in no change in mortality and a decrease in LOS.^{1 ,34} In our study, stroke patients receiving IV tPA had a significantly higher in-hospital mortality rate than stroke patients who did not receive IV tPA. The CHOICE model adjusts for initial severity of illness. The predicted in-hospital mortality with ICH and DNR removed from the model was similar for patients treated with tPA and those not treated with tPA. However, the actual in-hospital mortality was significantly higher than predicted for the patients receiving tPA, likely due to the occurrence of ICH: 55% of the deaths in patients receiving tPA occurred in patients with symptomatic ICH.

Length of stay was significantly longer in patients receiving tPA than in patients not receiving tPA. However, the predicted LOS in the tPA patients was also longer, and there was no significant difference between the predicted and actual LOS in any of the 3 patient groups. The prolonged predicted LOS in the tPA patients cannot be ascribed to ICH because this variable was taken out of the prediction model. Two factors that were different among groups receiving and not receiving tPA were PEG tube placement and abnormal serum bicarbonate level within the first 2 hospital days. The increased rate of PEG tube placement in tPA patients likely reflects increased stroke severity, although morbidity as a result of tPA-induced ICH is another possibility. The LOS results imply that patients with milder strokes were less likely to receive IV tPA. As previously noted, the presence of mild symptoms was a common reason for exclusion from the NINDS trial.²⁹ Nevertheless, our results do not suggest an early benefit from the use of IV tPA. The CHQC project does not track long-term patient outcomes.

This study has some limitations. Primary ICD-9-CM codes were used to identify the stroke patients in this study. Benesch et al²³ showed ICD-9-CM codes 434 and 436 to be 85% to 90% accurate in identifying patients with stroke. We found 6 additional patients treated with IV tPA who had a primary diagnosis code other than stroke, but we were unable to systematically identify such patients. In addition, we had only limited information on initial stroke severity and did not assess neurological outcomes other than LOS, in-hospital mortality, and discharge disposition.

Paralleling the experience with coronary thrombolysis, the potential exists that IV tPA for acute stroke will often be given by nonneurologists, such as emergency medicine physicians. Although neurologists were involved in almost all of the cases of IV tPA administered in Cleveland, there was a high frequency of deviations from national guidelines. This suggests a need for better professional education for all physicians caring for patients with acute stroke. The low rate of IV tPA use also emphasizes the need for improved public and professional education about stroke. There was evidence of a tPA learning curve in Cleveland. The national blood pressure monitoring guidelines were followed more frequently and NIHSS scores were documented significantly more often in the second 6 months of the study. There was also a downward trend in the occurrence of guideline deviations and an increase in the rate of IV tPA use over time. While these trends are encouraging, especially if replicated nationally, they should not decrease surveillance of IV tPA use in the community setting.