Due to the narrow therapeutic window in the NINDS rt-PA Stroke Trial and the small percentage of patients presenting within the first 3 h of symptom onset, efforts were made to evaluate rt-PA beyond 3 h of stroke onset. The Alteplase ThromboLysis for Acute Noninterventional Therapy in Ischemic Stroke (ATLANTIS) study was a double blind, randomized, placebo-controlled trial evaluating the safety and efficacy of 0.9 mg/kg IV rt-PA in patients with acute ischemic stroke [21]. Inclusion criteria were similar to the NINDS rt-PA Stroke Trial except for the time window studied, an age restriction (patients ≥80 years of age excluded), and exclusion for subtle signs of early infarction on CT involving >33 % of the middle cerebral artery (MCA) territory. The ATLANTIS study was initially designed to evaluate patients treated within 0–6 h (Part A). In December of 1993, after enrollment of 142 patients, the time window was changed (Part B) to 0–5 h due to concerns of the Data Monitoring and Safety Board (DMSB) about the 5–6 h group. Time from onset to treatment in Part B was further modified to 3–5 h following FDA approval of rt-PA for patients with acute ischemic stroke treated within 3 h.

Of 142 patients enrolled in Part A, 22 were entered within 3 h of stroke onset and 46 between 5 and 6 h of stroke onset. While a higher percentage of patients in the rt-PA treatment group improved by ≥4 points on the 24-h NIHSS (21 % placebo vs. 40 % rt-PA, P = 0.02), more placebo patients improved by ≥4 points at 30 days (75 % placebo vs. 60 % rt-PA, P = 0.05), and rt-PA patients had higher rates of sICH (0 % placebo vs. 11 % t-PA, P < 0.01) and mortality at 90 days (7 % placebo vs. 23 % t-PA, P < 0.01) [22].

In Part B of the ATLANTIS study, 613 patients were enrolled, including 39 patients entered within 3 h of stroke onset and 24 patients entered after 5 h. The trial was ended prematurely in July 1998 based upon a Data and Safety Monitoring Board (DSMB) analysis indicating that “IV rt-PA treatment was unlikely to prove beneficial” [21]. Placebo and rt-PA treated patients had median NIHSS scores of 11 and median onset-to-treatment times of 270 min and 276 min, respectively. There were no differences in the primary endpoint (the percentage of patients with a NIHSS <  1 at 90 days) or any of the secondary endpoints between placebo and rt-PA treated patients, with the exception of >11 point improvement on the NIHSS at 30 and 90 days (“major neurologic recovery”), which favored the rt-PA group (P = 0.02 and P = 0.03, respectively). Patients treated with rt-PA had a higher rate of sICH (1.3 % placebo vs. 6.7 % rt-PA, P < 0.001) and a trend toward higher 90-day mortality (6.9 % placebo vs. 10.9 % rt-PA, P = 0.08). Analyses using only the “target population” of 547 patients—enrolled within 3–5 h of stroke onset without protocol violations—were similar to intention-to-treat analyses.

In 2008, the ECASS III study, a double blinded, 1:1 randomized, placebo controlled trial of 0.9 mg/kg IV rt-PA administered between 3 and 4.5 h from symptom onset was published [23]. The study design was nearly identical to the ECASS II study (discussed later) with a couple of modifications. Patients were not eligible if they were over the age of 80 years, were on any anticoagulant regardless of coagulation test results, had a baseline stroke severity above NIHSS 25, and if they were diabetic with a prior history of stroke. Initially the study intended to enroll patients within 3–4 h from symptom onset but after enrolling 228 patients the study protocol was amended to extend the window to 4.5 h. The primary efficacy endpoint was defined as a favorable mRS at 90 days (0–1). Secondary efficacy endpoints at 90 days included a global outcome measure, the Barthel Index ≥95, and mRS 0–2. Safety endpoints included 90-day mortality, sICH defined by any intracranial hemorrhage associated with a clinical deterioration (NIHSS > 4), and any intracranial hemorrhage.

The study enrolled 821 patients from July 2003 through November 2007. Baseline demographics between the two treatment groups were statistically similar; with median baseline NIHSS of 9 and 10 for rt-PA and placebo groups respectively. In the intention to treat (ITT) population, the rt-PA arm demonstrated superior 90 day outcomes compared to placebo (mRS 0–1, 52.4 % vs. 45.2 % respectively, OR 1.34 [1.02–1.76]). The secondary endpoint of the global outcome for the rt-PA arm was also superior compared to placebo (OR 1.28 [1.00–1.65]). The secondary endpoint of mRS 0–2 was not significantly different in the ITT population (OR 1.30 [0.95–1.78]) but was statistically superior in the per-protocol population (PPP) (OR 1.41 [1.01–1.96]). Neither the ITT or PP populations demonstrated statistically significant difference between treatment arms (BI score PP OR 1.33 [0.99–1.80] in favor of rtPA). As in prior studies, greater rates of intracranial hemorrhage were noted in the rt-PA arm yet mortality rates were similar. Symptomatic hemorrhage per the ECASS III definition was noted in 2.4 % vs. 0.2 % of the rt-PA vs. placebo arms (OR 9.85 [1.26–77.32]). Using the NINDS definition of sICH, rtPA continued to demonstrate higher rates (7.9 % vs. 3.5 %, OR 2.38 [1.25–4.52]). As noted in prior meta-analyses, ECASS III reiterated the importance of earlier treatment with decreasing chances of favorable outcome with longer onset to treatment times and the authors emphasized treating as quickly as possible.

During the same time period as the ECASS III study, the European Safe Implementation of Treatments in Stroke (SITS) study was performed across 35 countries and 700 clinical centers [24]. This observational study utilized data from the International Stroke Thrombolysis Registry (ISTR), a voluntary internet-based registry of patients treated with IV rt-PA according to standardized criteria from the SITS-Monitoring Study (SITS-MOST). The SITS-ISTR study compared patients treated within 3–4.5 h from symptom onset to those patients in the registry treated 0–3 h from symptom onset. A total of 664 patients were treated within 3–4.5 h from symptom onset and compared to 11,865 patients in the 0–3 h cohort. The patients in the later treatment arm were slightly younger (65 years vs. 68 years, P < 0.0001), less severe (NIHSS 11 vs. 12, P < 0.0001) and treated later after symptom onset (195 min vs. 140 min, P < 0.0001) compared to the 0–3 h cohort; otherwise the treatment groups were largely similar. No significant difference in the proportion of patients achieving a 90 day mRS 0–1 was noted between the later window and 0–3 h (41 % vs. 40 % respectively). Similarly, rates of sICH (NINDS definition) and mortality did not statistically differ between cohorts (8.0 % vs. 7.3 % and 12.7 % vs. 12.2 % for 3–4.5 h and 0–3 h respectively).

A similar observational study was performed on data from the CASES Registry [25]. Comparing two very similar groups of stroke patients treated with IV rt-PA, those treated in the 3–4.5 h window (129) had similar clinical outcomes to patients treated within 3 h from symptom onset (982) (90 day mRS 0–1 39.4 % vs. 36.5 %, adjusted RR 0.98 [0.8–1.2]). Statistically significant trends were noted in the later treatment epoch where mortality and sICH were greater compared to 0–3 h window (adjusted RR 1.53 [1.15–2.0]; RR 2.4 [1.09–4.2] respectively). Of note, in the CASES 3–4.5 cohort, 36 % of patients would have been excluded by ECASS III criteria (age >85 years or baseline NIHSS > 25), suggesting the ECASS III criteria, empirically chosen for safety, may exclude patients who would receive overall net benefit from IV r-PA in this extended window.

Based on data from ECASS III and SITS-MOST, the European Medicine Evaluations Agency approved the use of IV rt-PA for appropriate patients within 0–3 and 3–4.5 h from symptom onset using SITS-MOST and ECASS III criteria respectively. The American Heart Association/American Stroke Association (AHA/ASA) published a Science Advisory which, after reviewing all available published data, recommended “rtPA should be administered to eligible patients who can be treated in the time period of 3–4.5 h after stroke (Class I Recommendation, Level of Evidence B)” per ECASS III criteria [26]. Despite the FDA declining a request to extend the current US label for rt-PA for ischemic stroke to include 3–4.5 h, the most recent AHA/ASA Guidelines continued to support this recommendation [27].

Two large, randomized trials evaluated the safety and efficacy of IV rt-PA in stroke patients treated within 0–6 h of symptom onset, the European Cooperative Acute Stroke Study (ECASS) and ECASS II [28, 29]. ECASS was published 2 months before the NINDS trial results and 7 months prior to the FDA approval of rt-PA given within 3 h. ECASS was a multi-center, double-blind, randomized trial of 1.1 mg/kg of intravenous rt-PA vs. placebo in patients with acute ischemic stroke treated within 6 h of symptom onset [28]. Patients were eligible if they were ≥18 years old and had a clinical diagnosis of moderate to severe hemispheric stroke. Patients with coma, hemiplegia plus fixed eye deviation, global aphasia, vertebrobasilar stroke, or those with hypodensity in ≥33 % of a MCA on CT were excluded. Primary endpoints were differences of 15 points in the median Barthel Index (BI) and one grade in the median modified Rankin Scale (mRS) at 90 days. Secondary endpoints included 30-day mortality.

In the intent-to-treat population of 620 patients, there was no difference in primary endpoints. Hemorrhagic infarction was more frequent in the placebo-treated group (30.3 % placebo vs. 23 % rt-PA, P < 0.001), but parenchymal hematoma was more frequent in the rt-PA treated group (6.5 % placebo vs. 19.8 % rt-PA, P < 0.001). Death associated with hemorrhage occurred in 19 of the rt-PA treated patients and in 7 of the placebo patients. There was an overall trend toward increased 30-day mortality in the rt-PA patients (12.7 % placebo vs. 17.9 % rt-PA, P = 0.08).

Neurological improvement occurred in a subset of rt-PA treated patients, now known as the “target population”. While planning the trial, the ECASS investigators anticipated that up to 20 % of patients would have major protocol violations, but would be inadvertently randomized and, indeed, 109 patients were deemed major protocol violations after post-hoc review. The protocol violations included 66 with abnormalities on CT scans (mainly major early infarct signs), 12 patients who received prohibited therapy (e.g. heparin in less than 24 h), 20 patients whose follow-up deviated from the 90 ± 14 day time window, and 11 others (e.g. randomized but not treated). The remaining 511 patients were prospectively defined as the target-population; in this population the rt-PA treated patients showed better median 90-day mRS scores vs. placebo (2 vs. 3 respectively, P = 0.035). However, the 90-day BI scores did not differ (P = 0.16), nor did mortality. Overall, the total parenchymal hematoma rate in the target population was 19.4 % for rt-PA and 6.8 % for placebo (Table 5.3). In a post-hoc analysis of the ECASS data, application of the NINDS rt-PA Stroke Trial outcome measures to the intent-to-treat population (minus five patients who were randomized but did not receive treatment) demonstrated benefits for rt-PA patients in achieving a mRS of 0–1 (P = 0.044) and a NIHSS of 0–1 at 90 days (P = 0.001) [30].