| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Stroke. 2007;38:1967.)
© 2007 American Heart Association, Inc.
Comments, Opinions, and Reviews |
From Department of Neurology, University of Miami Miller School of Medicine, Miami, Fla.
Correspondence to Myron D. Ginsberg, MD, Department of Neurology (D4-5), University of Miami Miller School of Medicine, PO Box 016960, Miami, FL 33101. E-mail mginsberg{at}med.miami.edu
| Abstract |
|---|
|
|
|---|
Key Words: albumin clinical trial ischemia neuroprotection stroke
| Introduction |
|---|
|
|
|---|
The SAINT II Trial, which followed the SAINT I Trial of NXY-059, was presumably intended to attempt to replicate the "positive" results of SAINT I to satisfy FDA requirements for the eventual licensing of this agent for the indication of acute ischemic stroke therapy. The results of SAINT I, published in February 2006, in the worlds most prestigious journal of clinical medicine,2 had raised the expectations of the broader stroke community hoping for the eventual emergence of an efficacious neuroprotectant, but also attracted considerable criticism. In SAINT I, ischemic stroke subjects were randomized to treatment with either NXY-059 or placebo within 6 hours of stroke onset; the primary outcome was disability assessed at 90 days on the modified Rankin scale (mRS). The SAINT I authors reported that NXY-059, compared with placebo, significantly improved the overall distribution of mRS scores (P=0.038; OR, 1.20; 95% CI, 1.01 to 1.42). By contrast, NXY-059 treatment failed to improve National Institutes of Health Stroke Scale (NIHSS) scores relative to placebo.2 Interpretability of the possible clinical importance of the SAINT I mRS results, shown in the Table, is highly problematic, and suspicion was justifiably aroused that the magnitude of the treatment effect, even if statistically significant, might be too low to be clinically meaningful. For example, if one were to define favorable outcome more stringently as an mRS score of 0 or 1 at 90 days, the data of the Table show that 31.0% and 33.4% of the placebo and NXY-059 groups, respectively, attained this outcome. The percentage difference between groups is 2.4%, and the predicted number needed-to-treat to achieve 1 positive outcome (1/0.024) would be 41.7, a treatment effect of only modest clinical importance.
|
In a recent article in this journal, Koziol and Feng3 have re-analyzed the outcome measures of the SAINT I Trial with laser-like incisiveness and remarkable clarity, and have laid bare its shortcomings. The next three paragraphs summarize their salient critical points. The thoughtful article by Saver is also germane.4
The SAINT I analysis of mRS outcome used the Cochran-Mantel-Haenszel procedure, which relied on previous knowledge of stratification factors that were not published, and an unstratified analysis was not presented. The failure to present the results of an unstratified analysis represents a major shortcoming of SAINT I. When Koziol and Feng applied the straightforward nonparametric Mann-Whitney statistic to the mRS data, they failed to achieve statistical significance favoring NXY-059 over placebo in either 1-sided (P=0.0768) or 2-sided (P=0.153) tests.3 It is best to report the results of both adjusted and unadjusted analyses, irrespective of the prespecified primary mode of statistical analysis. When a result is positive only on a prespecified adjusted analysis but not on unadjusted analysis, this often is a signal of a borderline result that may not hold up in a confirmatory study.
Koziol and Feng3 asked the crucial question of relevance: "How does an observed mean difference of 0.13 U on the mRS scale between the 2 arms translate to the benefit of individual patients?" They provided an alternative calculation of an observed effect size of 0.17 points in SAINT I and concluded that the SAINT I trialists "... appear to have conflated statistical with clinical significance."3 By way of contrast, in the NINDS tissue plasminogen activator (tPA) trial (part 2),5 the percentage of subjects achieving the stringent 3-month mRS outcome of 0 to 1 was 39% for the tPA group versus 26% for the placebo group—an absolute effect size difference of 13%, resulting in a
0.5 mRS point group difference. This difference is
3-fold greater than the group mRS difference of SAINT I.
Koziol and Feng3 called attention to the highly unusual procedure in SAINT I of regarding the mRS and the NIHSS as coprimary outcome measures but not adjusting statistically for the use of 2 measures, eg, by declaring a priori an
=0.025,2 a level that would have abrogated the significance of the SAINT I results.
Given the marginal findings and questionable methods of SAINT I, it is natural to question the wisdom of the rationale for proceeding to invest the tens of millions of dollars and thousands of medical personnel hours needed to conduct SAINT II. One factor contributing to this decision may have been the "unnatural exuberance" of the stroke community itself. In an editorial piece in the Emerging Therapies section of the October 2006 issue of Stroke (a section coedited by Dr Kennedy Lees, the lead author of the SAINT I report2), Dr David Hess waxed enthusiastic about NXY-059,6 albeit before the Koziol and Feng analysis3 had appeared in print. Among his comments were the following: "Is NXY-059 a better drug than previous failed agents or was the trial just better designed? Probably both"; "The preclinical data ... were robust and impressive"; "The trial design was also improved over past neuroprotective trials ... the primary outcome measure, a shift in the modified Rankin Scale, was appropriate ... This illustrates the importance of choosing a primary outcome that is realistic and reflects the mechanism of action of the drug"; "If SAINT II replicates the findings of SAINT I, then we have a major advance in the treatment of stroke—a very safe drug that improves most patients a little bit and that reduces ... tPA-related ICH."6 In a companion editorial, Dr Marc Fisher echoed enthusiasm for the mRS outcome measure as used in SAINT I ("... an important innovation ... more appropriate than looking for a cure.").7 In the light of the points raised earlier, these comments must be regarded as insufficiently critical. Nonetheless, it is worth examining these and other points in greater detail.
Dr Hess points out that NXY-059 was backed by a sizeable number of positive preclinical studies in animal models of stroke: that the time window to treatment in SAINT I was 6 hours (with forced stratification to maintain an average of 4 hours at each site); that NXY-059 was very safe; and that target plasma concentrations in SAINT I resembled those known to be neuroprotective in preclinical models of permanent ischemia.6 In this authors opinion, those points are cogent and well-taken. However, as reflected in this discussion, these strengths were offset by glaring weaknesses in clinical trial design: interpretively challenging choice of coprimary end point (shift in mRS); suboptimal statistical analysis; failure to adjust statistically for the second coprimary end point; and exaggeration of the clinical import of the small putative treatment effect itself.
One cannot reflect on the clinical failure of NXY-059 without considering the "STAIR criteria." As its name implies, the Stroke Therapy Academic Industry Roundtable (STAIR) refers to a series of expert consensus conferences on stroke held over the past 7 years. At the first STAIR conference in 1999, a committee-generated set of recommendations emerged as to how preclinically promising neuroprotective agents should be investigated before coming to clinical trials.8 These recommendations included the use of animal models of both permanent and temporary focal ischemia; testing in more than one laboratory; demonstrating efficacy in more than one animal species, including nonhuman primates; testing in both male and female animals; measurement of both behavioral and histological outcomes; showing efficacy with drug administration at 1 hour after ischemia or beyond; and others. In point of fact, the preclinical development of NXY-059 adhered closely to the STAIR criteria.6,8
What, then, can one infer from the marginal results of SAINT I and the failure of SAINT II? A recent report may shed some light on this conundrum.9 This systematic literature survey identified 1026 distinct neuroprotectant drugs or interventions evaluated for acute ischemia, of which 912 were tested in animal models only, 97 in both animals and humans, and 17 only in humans. The authors attempted to assign to each agent a score reflecting its "STAIR quality," ie, the number of STAIR criteria satisfied in preclinical publications (maximum, 10). They found that as "STAIR quality" of an agent increased, its average level of neuroprotection (ie, percentage infarct volume reduction) tended to approach the mean value for the entire group of agents surveyed (25%).9 Clearly, close adherence to the STAIR criteria for preclinical testing does not necessarily favor the emergence of highly efficacious drug candidates (although it may increase the degree of confidence as to the certainty of the preclinical findings obtained), and neither does adherence to STAIR criteria plus experimental efficacy predict clinical success, as NXY-059 well exemplifies.
Nonetheless, many of the principles embodied in the STAIR criteria are meritorious and deserve re-emphasis here. The pre-eminent guiding principle should be scientific rigor. Animal models should be clinically relevant and studied with rigorous attention to physiological monitoring and regulation (particularly, brain and body temperature, blood pressure, arterial blood gases, and glycemic control).10 Therapeutic window studies should demonstrate efficacy at several hours after stroke onset as gauged by long-term as well as short-term functional and histological measures. Pharmacokinetics and pharmacodynamics should be studied preclinically to obtain data that will guide drug dosing in human clinical trials. The use of models of both "permanent" and "transient" focal ischemia has been emphasized, but this distinction may be overly simplistic. Some degree of reperfusion commonly occurs in "permanent" models because of clot lysis or perfusion via collateral channels; on the other hand, "transient" ischemia may be complicated by ongoing postischemic hypoperfusion. In view of our pilot phase clinical data suggesting a possible synergism between tPA-induced thrombolysis and albumin neuroprotection, a more meaningful recommendation might be to study thrombolysed versus nonthrombolysed models of clot-induced ischemia.
| What Now? A Resurgence of Nihilism? |
|---|
|
|
|---|
Skepticism regarding neuroprotection has been formulated in such expressions as, "Why do neuroprotective drugs work in animals but not humans?".14 Such statements are inherently misleading in that they tend to ignore the fact that (as discussed) only a few well-designed and well-executed clinical neuroprotectant trials have been carried to completion, and some clinical trials were thought to have proceeded despite suboptimally conducted laboratory studies.15 Nonetheless, an abundance of published preclinical studies provide scientifically irrefutable proof-of-principle that ischemic neuroprotection, in fact, is feasible with a variety of agents.9 Indeed, the multitude of drug classes capable of achieving experimental ischemic neuroprotection is consistent with pathophysiological evidence that interacting cascades involving multiple biochemical, molecular, and vascular mechanisms contribute to ischemic injury.16–18 The body of experimental evidence supporting the feasibility of ischemic neuroprotection is, in fact, so substantial that it becomes incumbent on both basic, translational, and clinical investigators to examine closely the reasons that specific clinical trials may have failed.
| Why May NXY-059 Have Failed to Show Consistent Efficacy in Clinical Trials? |
|---|
|
|
|---|
-phenyl-N-tert-butyl nitrone], a generic nitrone spin-trap agent extensively characterized in preclinical studies and shown to be protective in animal models of focal and global brain ischemia and traumatic brain injury.19–22
NXY-059, a proprietary compound, was selected for commercial clinical development whereas PBN, a generic agent, was not, despite the fact that properties of NXY-059 might be viewed as suboptimal in several crucial respects. First, NXY-059 is a polar molecule, highly water-soluble, and nonlipophilic (less so than PBN), and as such it crosses the blood–brain barrier only sparingly23,24 and has negligible cellular penetration.25 Thus, it is likely that any anti-ischemic activity of NXY-059 would be confined to the intravascular compartment. By contrast, another novel class of nitrone spin-trapping agents, the azulenyl nitrones, are highly lipophilic; the second-generation azulenyl nitrone, stilbazulenyl nitrone (STAZN), attains forebrain levels of
2.5% of plasma levels within 2 to 3 hours of its intravenous administration.26 Relatively little critical thought appears to have been devoted to whether the poor blood–brain barrier penetrability of NXY-059 might prove to be a fatal flaw in clinical neuroprotection trials, despite the fact that previously studied poorly penetrable antioxidant agents have also been shown to fail under similar trial conditions (eg, tirilazad27,28).
The extent of an antioxidant compounds putative therapeutic efficacy can be gauged from in vitro antioxidant assays and from measurements of the agents oxidation potential. In 1H-NMR–based assays for assessing potency of chain-breaking antioxidants, PBN and NXY-059 were
300-fold less potent in inhibiting free radical-mediated aerobic peroxidation than STAZN, which rivaled the antioxidant potency of vitamin E itself.29 Cyclic voltametry studies have shown that PBN, its derivative S-PBN, and, by inference, NXY-059, have oxidation potentials far higher than those of important biological chain-breaking antioxidants such as vitamin E, glutathione, and beta-carotene; by contrast, azulenyl nitrones oxidize completely at potentials near to or lower than those of endogenous antioxidants.30,31 Correspondingly, the second-generation azulenyl nitrone compound, STAZN, reduced cortical infarct volume by 64% to 97%, and total infarct volume by 42% to 72%, in rats with temporary focal ischemia and 2-hour posttreatment; in more than one-half of STAZN-treated rats, virtually no cortical infarct was evident.32 These results exceed the level of protection reported with NXY-059 in similar models.33
An additional characteristic of NXY-059 is its extremely low potency, necessitating doses 300- to 600-fold higher than were required for the azulenyl nitrone STAZN to achieve a therapeutic effect in preclinical studies.23,26,30,32,33 Low potency is not necessarily a shortcoming, but it means that to achieve a therapeutic effect, very high doses must be administered. It is only because of the favorable safety profile of NXY-059 that these high doses were feasible.
| Albumin Neuroprotection Therapy for Ischemic Stroke: A Personal Example of a Journey in Progress |
|---|
|
|
|---|
Preclinical Phase
The use of high-dose human albumin for ischemic neuroprotection originated in serendipitous observations and was pursued in a sequence of preclinical studies beginning in 1997 that established: (1) consistent, high-grade behavioral and histological protection in reversible focal ischemia34,35; (2) lesser but significant protection in permanent focal ischemia36,37; (3) a definable dose-response function and favorable therapeutic window (4 to 5 hours)38; (4) absence of adverse events39; and (5) evidence of multiple participatory mechanisms involving both the intravascular and parenchymal compartments.35–37,40,41 We believe that the therapeutic efficacy of albumin in ischemia, in fact, is directly attributable to its multiplicity of actions. It is of note that while some of the STAIR criteria8 were fulfilled in these studies (namely, testing in both temporary and permanent models; postischemic administration at several doses; measurement of both behavior and histology; exploration of the therapeutic window), many other STAIR criteria were not respected: eg, studies were conducted in only 1 laboratory (our own), in only 1 species (rat, not primates) and 1 gender (male), and only in juvenile rats (not old or diseased).
Pilot Clinical Phase
Based on our preclinical results, the NIH funded a pilot-phase dose-escalation clinical trial of albumin therapy in 82 subjects with acute ischemic stroke. The results of this study, the ALIAS Pilot Trial, were recently published in full in this journal.42,43 Our results showed a high degree of safety (13% incidence of readily manageable mild-to-moderate pulmonary edema). Efficacy analysis performed by comparing lower-dose to higher-dose treatment groups, and by comparing the higher-dose group to historical data from the NINDS tPA trial,5 showed strong suggestive evidence of efficacy using a stringent favorable outcome measure, ie, mRS and/or NIHSS of either 0 or 1 at 3 months. The probability of favorable outcome in the higher-dose albumin subjects was 95% greater than in the comparable NINDS historical cohort. The tPA-treated subjects who received higher-dose albumin were 3-times more likely to achieve a good outcome than subjects with lower-dose albumin,43 suggesting a synergistic interaction of albumin and tPA. Grotta,44 reflecting on translating lessons from laboratory stroke models to the clinic, has stressed the importance of combining neuroprotection and reperfusion strategies.
Several contrasts with SAINT I and II deserve mention. A stringent outcome measure applied to a small population still generated strong suggestions of efficacy, and the expected concordance between mRS and NIHSS outcome measures was in fact observed.43
Randomized Multicenter Clinical Trial Phase
Based on the pilot trial results, the NIH has funded us to conduct a large randomized, multi-center, double-blinded, placebo-controlled efficacy trial, the ALIAS (Albumin in Acute Stroke) Phase III Trial.45 The trial is being conducted at
80 North American clinical sites. Salient design features include: (1) initiation of treatment within 5 hours of stroke onset; (2) albumin dose of 2 g/kg (a preclinically highly effective dose); (3) a stringent primary outcome measure (mRS and/or NIHSS of 0 to 1 at 3 months); and (4) use of 2 independent cohorts (N=900 each), one cohort composed of subjects receiving standard-of-care tPA, and the other not, with separate 1:1 randomization to albumin or placebo for each cohort.
Admittedly, there are many valid approaches to clinical trial design, each with advantages and disadvantages. The design of the ALIAS Phase III Trial, which we believe to be sound, reflects our primary interest in seeking evidence of an albumin treatment effect that results in a normal or nearly normal functional outcome in subjects with ischemic stroke.
| Conclusion |
|---|
|
|
|---|
| Acknowledgments |
|---|
This study was supported by NIH grants NS 40406 and NS 46295.
Disclosures
None.
Received November 30, 2006; revision received December 12, 2006; accepted January 10, 2007.
| References |
|---|
|
|
|---|
This article has been cited by other articles:
![]() |
M. Fisher, D. C. Hess, and K. Lees Issues Pertaining to the Critiques of the SAINT-I Trial Stroke, November 1, 2007; 38(11): e126 - e127. [Full Text] [PDF] |
||||
![]() |
M. D. Ginsberg Response to Letter by Fisher et al Stroke, November 1, 2007; 38(11): e128 - e128. [Full Text] [PDF] |
||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Stroke Home | Subscriptions | Archives | Feedback | Authors | Help | AHA Journals Home | Search Copyright © 2007 American Heart Association, Inc. All rights reserved. Unauthorized use prohibited. |