rt-PA plus Eptifibatide versus rt-PA Alone in Acute Ischemic Stroke: A Propensity Score Matched Post Hoc Analysis

Abstract

Background and Purpose

The CLEAR-ER trial demonstrated safety of rt-PA plus eptifibatide in acute ischemic stroke (AIS). CLEAR-ER randomized AIS patients (5:1) to 0.6mg/kg rt-PA plus eptifibatide versus standard rt-PA (0.9mg/kg). IMS III randomized AIS patients to rt-PA plus endovascular therapy versus standard rt-PA. ALIAS Part 2 randomized patients to albumin ± rt-PA versus saline ± rt-PA. Our aim was to compare outcomes in CLEAR-ER combination arm patients to propensity score matched rt-PA only subjects in ALIAS Part 2 and IMS III.

Methods

The primary outcome was 90-day severity-adjusted mRS dichotomization based on baseline NIHSS. Secondary outcomes were 90-day mRS dichotomization as “excellent” (mRS 0–1); mRS dichotomization as “favorable” (mRS 0–2); and nonparametric analysis of the ordinal mRS.

Results

Eighty five combination arm CLEAR-ER subjects were matched with 169 ALIAS Part 2 and IMS III trials’ rt-PA only patients (controls). Median age in CLEAR-ER and control subjects was 68years; median NIHSS in the CLEAR-ER subjects was 11 and in control subjects 12. At 90 days, CLEAR-ER subjects had a non-significantly greater proportion of patients with favorable outcomes (45% vs 36%, unadjusted RR 1.24, 95% CI 0.91–1.69, p=0.18). Secondary outcomes were 52% vs 34% excellent outcomes (RR 1.51, 1.13–2.02, p=0.007); 60% vs 53% favorable outcome (RR 1.13, 0.90–1.41, p=0.31); and ordinal Cochran-Mantel-Haenszel p=0.10.

Conclusion

rt-PA plus eptifibatide showed a favorable direction of effect that was consistent across multiple approaches for AIS outcome evaluation. A phase III trial to establish the efficacy of rt-PA plus eptifibatide for improving AIS outcomes is warranted.

Introduction

Twenty years after completion of the NINDS recombinant tissue plasminogen activator (rt-PA) stroke trial,¹ intravenous (IV) rt-PA remains the only proven therapy for acute ischemic stroke (AIS). The recently completed 126-patient phase II Combined Approach to Lysis Utilizing Eptifibatide and rt-PA in Acute Ischemic Stroke - Enhanced Regimen (CLEAR-ER) trial found that the addition of eptifibatide, a platelet glycoprotein (GP) 2b/3a inhibitor that prevents platelet aggregation, to IV rt-PA had a safety profile and direction of effect in favor of the combination therapy over IV rt-PA.² While this direction of effect persisted after statistical adjustment, there were baseline imbalances in the trial in favor of the combination arm with regard to age and baseline NIH Stroke Scale (NIHSS) score. In this paper, we compared combination therapy patients from CLEAR-ER to contemporaneously enrolled IV rt-PA arm patients in the phase III Interventional Management of Stroke (IMS) III³ and the Albumin in Acute Stroke (ALIAS) Part 2⁴ trials. We compared outcome using four approaches variably proposed as optimal for acute stroke clinical trials.^5–9

Methods

This was a post-hoc propensity matched analysis of data from three previously published randomized clinical trials. The CLEAR-ER trial was a multi-center, double-blind, randomized safety study. AIS patients treated with IV rt-PA within three hours of symptom onset were randomized to 0.6mg/kg rt-PA plus eptifibatide (135mcg/kg bolus and a two-hour infusion at 0.75mcg/kg/min) (combination arm, n=101) versus standard rt-PA (0.9mg/kg) (n=25).¹⁰ The IMS III trial was a multi-center multi-national randomized clinical trial of IV rt-PA plus endovascular therapy (n=434) versus IV rt-PA (n=222) in AIS patients treated with standard dose IV rt-PA within three hours of symptom onset.³ The ALIAS Part 2 trial was a multi-center multi-national randomized clinical trial of albumin (n=422) versus saline (n=419).⁴ ALIAS Part 2 patients who were eligible for rt-PA were treated with rt-PA per standard of care.

For this analysis, we matched two controls among IMS III and ALIAS rt-PA only subjects for each CLEAR-ER combination arm subject using a propensity score matching approach.^{11, 12} Age, gender, race, baseline mRS, baseline NIHSS score, and time from stroke onset to rt-PA initiation were included in the multivariable logistic model used to generate a propensity score for each subject. The 1:2 matching mechanism was based on a greedy algorithm, with the best match determined by the weighted sum of the absolute difference in propensity score and age between potentially matching individuals, allowing a maximum difference of 0.025 in the propensity score and 6 years for age, with the weight for the propensity score set to be double that for age.¹³

Both CLEAR-ER and IMS III allowed enrollment of patients with baseline mRS >1 while ALIAS Part 2 only allowed patients with baseline mRS of 0 or 1. All datasets were restricted to subjects with baseline mRS of 0 or 1. Of the 101 subjects in the combination arm of CLEAR-ER, 16 were excluded for baseline mRS >1, leaving 85 subjects available for propensity matching. Of the 222 IV rt-PA IMS III subjects, 9 subjects were excluded (1 with baseline NIHSS missing, 4 with baseline mRS >1, 4 with missing 90-day mRS), leaving 213 subjects available for propensity matching. Of the 419 saline patients in ALIAS Part 2, 361 received IV rt-PA and 16 were excluded for missing 90-day mRS, leaving 345 available for propensity matching.

The primary outcome was defined as 90-day severity-adjusted mRS dichotomization based on baseline NIHSS (favorable outcome if mRS=0 with NIHSS ≤7; mRS=0 or 1 with NIHSS 8–14; mRS=0–2 with NIHSS>14).⁶ Secondary outcomes included 90-day mRS dichotomization as “excellent” (mRS 0–1); mRS dichotomization as “favorable” (mRS 0–2); an analysis of the ordinal mRS; and, NIHSS of 0 or 1 at 24 hours.

Relative risks (RR) were determined for the dichotomized efficacy outcome variables. Adjusted models included age, baseline NIHSS, and time to IV rt-PA using Zou’s modified Poisson approach.¹⁴ The non-parametric analysis of the ordinal mRS was performed using the Van Eltren form of the Cochran-Mantel-Haenszel (CMH) test using the full range of the mRS as a 6-category polychotomous outcome (collapsing mRS scores of 5 with 6).¹⁵ Differences in proportions of favorable outcomes were also calculated with 95% confidence intervals (CI).

Results

Eighty five combination arm CLEAR-ER subjects were matched with 169 ALIAS Part 2 and IMS III rt-PA only patients (62 IMS III and 107 ALIAS). Patient characteristics and propensity matching factors are presented in Table 1. Note that 18 out of 107 rt-PA patients in ALIAS Part 2 also had endovascular therapy. At 90 days, CLEAR-ER subjects had a greater proportion of patients with favorable outcomes (45% vs 36%, unadjusted RR 1.24, 95% CI 0.91–1.69, p=0.18). Secondary outcomes were 52% vs 34% excellent outcomes (RR 1.51, 1.13–2.02, p=0.007); 60% vs 53% favorable outcome (RR 1.13, 0.90–1.41, p=0.31); and, a shift (CMH p value =0.10). At 24 hours, 20% of CLEAR-ER subjects had NIHSS 0 or 1 versus 14% of controls; difference in proportions 6% (−4%, 16%, p=0.19). Table 2 shows the adjusted and unadjusted RR and outcomes at 90 days. Table 3 shows the differences in proportions of 90-day outcomes with 95% CI. Symptomatic intracranial hemorrhage rates within 36 hours were 0 (0%, 95% CI 0%–4%) in CLEAR-ER subjects and 5 (3%, 95% CI 1%–7%) in rt-PA subjects. The mRS distributions for the CMH test are shown in the Figure.

Table 1.

Patient Characteristics and Propensity Matching Factors

	CLEAR-ER Combination Arm (N=85)	Control rt-PA Only (N=169)	p- value
Age, median (range)	68 (33–86)	68 (31–84)	0.93
Male	42 (49%)	83 (49%)	0.96
Black	11 (13%)	22 (13%)	0.99
Baseline NIHSS, median (range)	11 (6–31)	12 (6–30)	0.99
Baseline mRS			0.80
0	74 (87%)	149 (88%)
1	11 (13%)	20 (12%)
Time from stroke onset to IV rt-PA, median (range)	113 (54–187)	119 (45–240)	0.76
Medical History
Atrial Fibrillation	27 (32%)	38 (22%)	0.11
Hypertension	71 (84%)	123 (73%)	0.06
Diabetes	26 (31%)	40 (24%)	0.24
Congestive heart failure	6 (7%)	17 (10%)	0.43
Myocardial infarction	12 (14%)	19 (11%)	0.51
Hyperlipidemia	37 (44%)	89 (53%)	0.17
Prior stroke	10 (12%)	29 (17%)	0.26

Table 2.

90-Day Outcomes in Combination and rt-PA Only Groups

	Relative Risk (95% CI)	Adjusted Relative Risk (95% CI)*
Outcome
90-day mRS sliding dichotomy	1.24 (0.91, 1.69)	1.24 (0.91, 1.68)
90-day mRS 0–1	1.51 (1.13, 2.02)	1.48 (1.13, 1.94)
90-day mRS 0–2	1.13 (0.90, 1.41)	1.11 (0.91, 1.36)

^*Adjusted models include age, baseline NIHSS, and time to IV rt-PA using Zou’s modified Poisson approach¹⁴

Table 3.

90-Day Differences in Outcome Proportions

	CLEAR-ER (N=85)	Control (N=169)	Difference in Proportions	95% Confidence Intervals	p-value
Outcome
90-day mRS sliding dichotomy	38 (45%)	61 (36%)	9%	−4%, 21%	0.18
90-day mRS 0–1	44 (52%)	58 (34%)	17%	5%, 30%	0.007
90-day mRS 0–2	51 (60%)	90 (53%)	7%	−6%, 20%	0.31

Figure.

mRS Distributions

Based on observed data for the sliding dichotomy outcome measure, the study sample size of 85 in the treatment group and 169 in the control group would achieve 80% power to detect a difference between group proportions of patients with favorable outcome of 0.18, assuming the proportion in the control group is 0.36, using a two-sized Z test with pooled variance at a 0.05 significance level. Furthermore, if we assume the observed group difference in proportion of patients with favorable outcome of 0.09 (45% treatment and 36% control) is the minimum value worth detecting in future studies, a sample size of 466 in each group would achieve 80% power to detect this difference using a two-sized Z test with pooled variance at a 0.05 significance level.

Discussion

In this post-hoc analysis, we found a direction of effect in favor of the combination of rt-PA plus eptifibatide over rt-PA alone in AIS. These findings support a phase III trial to establish the efficacy of rt-PA plus eptifibatide for improving AIS outcomes.

Although sliding dichotomous or ordinal approaches have been preferentially recommended for evaluating mRS outcomes in stroke trials,^{5–9, 16, 17} we found that of the four approaches analyzed, dichotomization at excellent outcomes (mRS 0–1) demonstrated the only statistically significant finding in favor of rt-PA plus eptifibatide when compared with rt-PA. A possible explanation is that for interventions within the early time window studied, dichotomizing at mRS 0–1 may represent a reasonable transition point between “favorable” and “non-favorable” outcomes.⁵ Both positive trials of rt-PA in AIS showed benefit with mRS dichotomized at 0–1.^{1, 18} However, given that the ECASS II trial showed no treatment effect with dichotomization at mRS 0–1 but was positive with mRS 0–2,¹⁹ caution must be taken in interpreting our results and full consideration given to all approaches in selecting an analysis plan for the primary outcome in a planned phase III trial.

Compared with dichotomized favorable outcomes (mRS 0–2), our selected primary outcome, the severity-adjusted dichotomy approach (also referred to as sliding dichotomy or responder analysis) showed a slightly stronger trend in favor of the rt-PA plus eptifibatide group over rt-PA only. The ordinal analysis showed a similar trend as the sliding dichotomy in favor of the rt-PA plus eptifibatide group compared with rt-PA. As such, despite the seemingly stronger signal of efficacy shown by dichotomization at mRS 0 or 1 in this analysis, we favor the sliding dichotomy or ordinal approach for a future trial since there’s no reliable way to predict the distribution of patients who would be enrolled in such a trial. Thus, if patients with more severe strokes are enrolled, dichotomizing at mRS 1 may fail to show a treatment effect as was observed in ECASS II.¹⁹

In addition to the inherent limitations of an unplanned post-hoc analysis, limitations of this report include the small number of available rt-PA plus eptifibatide patients for analysis. Vascular imaging was not required for CLEAR-ER and the impact of combining eptifibatide with rt-PA on the insufficient arterial recanalization rates observed with rt-PA alone^20–22 remains unknown. However, the favorable direction of effect observed with rt-PA plus eptifibatide remained, with comparison groups similar in age and baseline NIHSS (Table 1) and a direction of effect in favor of early improvement with the combination group based on 24-hour NIHSS of 0 or 1.

We also performed analyses using the entire dataset from both ALIAS Part 2and IMS III trials (i.e., including patients who received albumin and/or endovascular therapy) and the results were similar with point estimates more favorable than the results presented in this paper (data not shown). Exclusion of the 18 ALIAS Part 2 patients (leaving 151 rt-PA controls) also did not change our point estimates meaningfully (data not shown).

We conclude that rt-PA plus eptifibatide showed a direction of effect that was consistent across multiple approaches for outcome evaluation in AIS. A phase III trial to establish the efficacy of rt-PA plus eptifibatide for improving AIS outcomes is warranted.