Evidence that Tenecteplase Is Noninferior to Alteplase for Acute Ischemic Stroke: Meta-Analysis of 5 Randomized Trials

TNK (tenecteplase), a genetically modified variant of ALT (alteplase), is a newer generation fibrinolytic agent with ease of administration advantages over ALT itself. TNK has a higher specificity for fibrin, a longer half-life, and reduced binding to PAI-1 (plasminogen activator inhibitor)-1 which leads to greater resistance to inactivation by PAI-1.¹ Its pharmacokinetic profile allows TNK to be administered as a single bolus. TNK is an established treatment for acute myocardial infarction, where in head-to-head trials against ALT it has shown equal therapeutic efficacy and fewer major bleeding complications compared with ALT.^2,3

TNK is a promising agent for acute ischemic stroke (AIS). With its greater fibrin specificity, TNK theoretically has the potential to be superior in efficacy and safety compared with ALT for AIS. However, even if only equivalent to, rather than better than ALT, TNK would still be a useful agent for AIS. Since TNK requires only a one-time bolus for administration, compared with the 60 minutes continuous infusion required for ALT, TNK could be administered more efficiently in large vessel occlusion patients, permitting faster start of subsequent endovascular mechanical thrombectomy, especially in drip-and-ship patients who could have lytic administered at an outside hospital and then be immediately transferred in a standard advanced life support ambulance staffed by paramedics, rather than having to await the availability of a critical care transport ambulance staffed by nurses knowledgeable in continuous infusion pump management. TNK has been investigated for AIS in several randomized trials, both against supportive care and against ALT as an active comparator.

The most recent AIS guideline of the American Heart/Stroke Association advanced a new recommendation that TNK could be considered as an alternative to ALT in select patients with AIS with minor neurological impairment and no major intracranial occlusion.⁴ However, this recommendation was based on informal consideration, rather than formal meta-analysis, of completed randomized control trials. Accordingly, we undertook a formal, noninferiority meta-analysis.

The authors declare that all supporting data are available within the article (and its in the online-only Data Supplement) and the cited published randomized trials.

This formal meta-analysis was conducted in accordance with the PRISMA guidelines (Preferred Reporting Items for Systemic Reviews and Meta-Analyses).⁵ We searched PubMed (Jan 2005 to August 2018) using the search strategy tenecteplase AND alteplase AND acute ischemic stroke. Studies were included if they met the following criteria: (1) randomized clinical trial; (2) patients enrolled with acute cerebral ischemia, with brain imaging performed before enrollment to exclude hemorrhage; (3) allocation to TNK versus active comparator ALT; and (4) treatment initiated acutely, within 6 hours after last known well time.

The primary efficacy end point analyzed was disability-free outcome (modified Rankin Scale [mRS] score, 0–1) at 3 months post-stroke. Additional efficacy outcomes were functional independence (mRS, 0–2) at 3 months and reduced level of disability overall 7 mRS levels (shift analysis) at 3 months. Safety outcomes were symptomatic intracranial hemorrhage (sICH) and mortality. Symptomatic hemorrhage events in individual trials were identified using the sICH definition employed in that trial. Two raters (Dr Burgos and Dr Saver) independently abstracted end point data, and any discrepancies were resolved by consensus review.

In the lead statistical analysis, the noninferiority margin was set at 6.5%, as per a recent major noninferiority design trial of different fibrinolytic regimens for AIS.⁶ In addition, 2 more stringent noninferiority margins were explored: 5%, based on an older survey of stroke experts to establish the minimally clinically important difference for stroke therapies⁷; and 1.3%, based on a more recent stroke expert survey that was designed to mitigate anchoring and centrality bias.⁸ Noninferiority margins were set at the same values for the mRS 0 to 2 efficacy end point. For shift analysis, the noninferiority margins applied to the meta-analytic common odds ratio (cOR) for improved outcome overall 7 mRS levels was established using the outcome distribution in the ALT group, applying 6.5%, 5%, and 1.3% improvements at the mRS 2 to 1 transition, and then applying exactly proportionate improvements at all other mRS transitions (yielding thresholds of cOR 0.87, cOR 0.89, and cOR 0.97). For mortality and sICH, the noninferiority margins were set at 1%, based on known provider and patient intolerance of outcome differences in both these major safety end points, and for sICH, the low rate base rate with ALT treatment.

The meta-analytic software employed was RevMan 5.3. For the 1 trial that had multiple comparisons (2 TNK dose tiers versus same ALT control group), unit-of-analysis error was avoided by splitting the shared control group into 2 half-sized groups.⁹ Both random and fixed effects analyses were performed. The random-effects analyses were considered the lead approach, as they make fewer assumptions. The fixed effects analyses, which are less subject to small trial overweighting, were considered sensitivity analyses.⁹

The systematic literature search yielded 81 records for detailed screening, among which 5 were independent randomized trials meeting study entry criteria, (Figure I in the online-only Data Supplement). These trials enrolled a total of 1585 patients, (828 TNK, 757 ALT). Trial characteristics are shown in the Table. Across all trials, mean age was 70.8, 58.8% of patients were male, mean National Institutes of Health Stroke Scale (NIHSS) at baseline was 7, and mean time from last known well to treatment start was 148 minutes. Risk of bias was intermediate to low for all studies (Figure II in the online-only Data Supplement). All patients with ALT received standard 0.9 mg/kg ALT dosing, 10% as a bolus, followed by the remaining 90% over 60 minutes. TNK dosing was one-time bolus only, at doses of 0.1 mg/kg in 6.8% of patients, 0.25 mg/kg in 24.6%, and 0.4 mg/kg in 68.6%.

For the primary end point freedom from disability (mRS, 0–1) at 3 m, data were available from all 5 trials, on 1585 patients (Figure 1). Crude cumulative rates of disability-free outcome were TNK 57.9% versus ALT 55.4%. Informal, random-effects meta-analysis, the risk difference was 4% (95% CI, −1% to 8%). The lower 95% CI bound of −1% fell within the lead noninferiority margin of −6.5%, as well as within the more stringent noninferiority margins of −5% and −1.3%. There was no evidence of modification of treatment effect by TNK dose level, I²=0%, interaction P=0.38, although testing for heterogeneity of the lowest, 0.1 mg/kg dose tier was underpowered, as the low dose tier accounted for only 4.3% of the data when weighted by SE. Considering the remaining dose tiers, both the 0.25 mg/kg dose alone and the 0.4 mg/kg dose alone met the lead noninferiority criterion. There was no evidence of modification of treatment effect by enrollment in a trial with no or uncommon use of concomitant endovascular thrombectomy or a trial with planned thrombectomy for all patients (Figure III in the online-only Data Supplement).

Considering the secondary efficacy analyses, for the additional efficacy end point of functional independence (mRS, 0–2) at 3 m, data were available on 1473 patients from 4 trials (Figure 2). Crude cumulative rates of independence were TNK 71.9% versus ALT 70.5%, risk difference 8% (95% CI, −4% to 20%). The lower 95% CI bound of −4 fell within the lead noninferiority margin of −6.5%, as well as within the more stringent noninferiority margin of −5%, but crossed the most stringent noninferiority margin of −1.3%. No effect modification by concomitant nonuse or use of endovascular thrombectomy was noted (Figure IV in the online-only Data Supplement). For the end point of level of disability at 3 m across all 7 levels of the mRS, data were available for 1397 patients from 3 trials. Crude summary distributions combining all trials are shown in Figure 3A and showed nominally more highly desirable outcomes (mRS, 0 and mRS, 0–1) with TNK versus ALT with relatively similar outcome rates for other mRS thresholds. Formal meta-analysis of the cOR for less disabled outcome found overall cOR, 1.21 (95% CI, 0.93–1.57), with the lower bound meeting criteria for noninferiority on the lead and the intermediate stringent thresholds, but crossing the most stringent noninferiority margin of 0.97 (Figure 3B).

For the safety end point of symptomatic ICH, data were available for 1585 patients from all 5 trials. Crude summary sICH rates were TNK 3% versus ALT 3%, risk difference 0% (95% CI, –1% to 2%; Figure V in the online-only Data Supplement). For death, in 1585 patients from all 5 trials, crude mortality rates at 3 months were TNK 7.6% versus ALT 8.1%, risk difference 0% (95% CI, –3% to 2%; Figure VI in the online-only Data Supplement). Though point estimates were favorable, TNK CIs crossed the narrow noninferiority margins that were set for both sICH and death. There was no evidence of modification of treatment safety effects by TNK dose level (Figures V and VI in the online-only Data Supplement).

For all comparisons, sensitivity analyses using a fixed effects model yielded similar results to the analyses using random-effects models.

In this formal meta-analysis of head-to-head trials in AIS, TNK demonstrated noninferiority to ALT. For the primary efficacy end point, freedom from disability (mRS, 0–1) at 3 months, TNK in combined dose analysis exceeded each of the assessed noninferiority margins, and each individual tested dose exceeded the lead noninferiority margin. Similar findings were present on the secondary efficacy end points of functional independence (mRS, 0–2) and lower degree of disability (shift analysis). Fewer trials contributed data to these analyses, reducing power, but TNK still surpassed 2 of the 3 assessed noninferiority margins, including the lead threshold. For the safety end points of mortality at 3 months and sICH, point estimates were also favorable for TNK, though CIs crossed the tight noninferiority margins set for these outcomes.

Compared with the initial trials establishing ALT as a proven stroke therapy, and compared with patients treated in routine practice, the patients collectively enrolled in the analyzed TNK versus ALT trials were similar in age, sex, and timing of therapy start, but had milder presenting stroke deficit severity (NIHSS 7 in the current study versus NIHSS 12 in the 2 pivotal National Institute of Neurological Disorders and Stroke r-tPA trials and versus NIHSS 10 in US practice).^10,11 This study’s findings of noninferiority are, therefore, more secure for patients with milder (eg, NIHSS, 1–14) deficits than patients with more severe (eg, NIHSS ≥15) deficits. However, the one trial that focused solely on patients with large vessel occlusions and severe deficits did provide strong signals of noninferiority.¹²

The findings of this study are consistent with and substantially add to prior meta-analyses of TNK versus ALT. Most prior systematic reviews included fewer trials, and all were undertaken within a superiority rather than noninferiority framework.^13–16 The current study is the first to demonstrate that, when the accumulated evidence is collated, TNK has now met important criteria for noninferiority.

With regard to different doses of TNK, for the lead efficacy outcome of freedom from disability (mRS, 0–1), there was no evidence of heterogeneity of treatment effect across the 3 studied TNK concentrations. However, power to detect differences was constrained for the low, 0.1 mg/kg dose, evaluated in the fewest patients, compared with the 0.25 mg/kg and 0.4 mg/kg dose tiers. In addition, there was some evidence of heterogeneity of treatment effect by TNK dose for the secondary efficacy outcome of functional independence (mRS, 0–2), suggesting that additional trial testing of different dosages is warranted.

The optimal approach to performing meta-analyses to evaluate noninferiority, rather than superiority, has not been well-characterized in prior investigations and consensus statements.⁵ The approach we have taken is a straightforward one: first aggregating data across trials using standard study-level meta-analytic techniques and then applying standard noninferiority testing, rather than superiority testing. This general approach has been taken in another study, albeit in the frame of sequential trial analysis rather than consolidated trial analysis.¹⁷

In noninferiority analyses, selection of the noninferiority margin to be applied can be challenging. Consensus recommendations are that the noninferiority margin be the smallest value that would be a clinically important effect.¹⁸ However, there often is a diversity of opinion in a field regarding what value demarcates the minimally clinically important difference of an outcome. The usual approach in noninferiority analyses of employing only a single noninferiority margin fails to take into account this spectrum of views. Accordingly, for this analysis, we assessed 3 different noninferiority margins of varying stringency, based on use in a prior noninferiority trial or broad surveys of expert physicians. One threshold that already employed in a peer-reviewed clinical trial,⁶ was designated the lead, but the others were analyzed as well to explore the robustness of the findings from the perspective of different stakeholder groups.^7,8 That the lead efficacy outcome surpassed all 3 assessed noninferiority margins indicates a strong evidential basis for noninferiority claims.

The results of this study have direct implications for clinical practice and treatment guidelines. TNK has been widely recognized as offering highly desirable feasibility advantages over ALT in the era of endovascular thrombectomy. Since it can be administered as a one-time bolus dose, rather than as an hour-long infusion, TNK can permit nonthrombectomy hospitals to accelerate their door in—door out times for patients needing intravenous thrombolysis followed by endovascular thrombectomy, by obviating the need for less widely available, nurse-staffed critical care ambulances for interfacility transport. While ALT requires a strategy of drip and ship, with the drip segment requiring 1 hour, TNK enables a strategy of give and go, with the give segment requiring 1 minute. There is a potential drawback to TNK’s all-at-once administration in that, with ALT, the longer infusion may be stopped after only a partial dose if the patient exhibits signs of potential hemorrhage; in addition, serum half-life of ALT is shorter than TNK, allowing hemostasis to return more quickly. However, these possible advantages of ALT in uncommon patients with early bleeding complications have not been shown to importantly alter clinical outcome, and seem unlikely to outweigh the substantial functional improvements achieved with faster transfer to thrombectomy centers. Accordingly, demonstration of noninferiority alone, even in the absence of superiority, would make TNK a useful agent. The prospect of that advantage led the most recent American Heart/American Stroke Association AIS guideline to recognize TNK as an alternative to ALT in patients with minor neurological impairment and no major intracranial occlusion, even in advance of any formal noninferiority analysis providing actual evidential support this recommendation.⁴ The current noninferiority meta-analysis now provides a more robust foundation for recommendations regarding use of TNK as an alternative to ALT.

This study has limitations. First, loss to follow-up is a greater threat to analytic validity in noninferiority than superiority analyses. Some of the included trials did not have complete outcome ascertainment at 3 months. However, rates of loss to follow-up were quite low, mitigating this concern. Second, power to explore differences in efficacy and safety among different TNK dose tiers was limited, especially for the lowest 0.1 mg/kg dose. Third, patients with severe presenting deficits were under-represented in the analyzed trials. Further head-to-head studies in patients with severe deficits are desirable. Fourth, the analyzed trials primarily enrolled patients within 4.5 hours of onset. While there is no clear pathophysiologic reason to expect that the relative performance of TNK and ALT would be different in patients selected for treatment on the basis of imaging evidence of still salvageable tissue >4.5 hours from last known well time, studies in late time windows would be of value.

In conclusion, accumulated clinical trial data indicates intravenous TNK is noninferior to ALT in the treatment of AIS. These findings provide formal support for recent guideline recommendation to consider TNK an alternative to ALT in patients with early cerebral ischemia. The available data are not definitive, as the greatest weight of evidence is from a trial that enrolled patients with mild deficits likely to have good outcomes, diminishing study informativeness about noninferiority, and because the most stringent noninferiority margins were not met for secondary efficacy and safety outcomes. Larger, definitive, noninferiority trials are desirable. While awaiting them, use of TNK in lieu of ALT is reasonable, especially in settings in which endovascular thrombectomy would be expedited.

We thank Jeffrey Gornbein, PhD.