Abstract
The continued failure in approving new drugs for treatment of acute stroke has been recently set back by the failure of the NXY-059 (Stroke-Acute Ischemic NXY Treatment (SAINT) II) trial. The disappointment was heightened by the latter study being viewed as a most promising compound for stroke drug development program based on the preclinical data. Since the SAINT I/II development program included many of the STAIR (Stroke Therapy Academic Industry Round table) guidelines, yet have still failed to achieve the expected efficacy, there is a clear need to continue and analyze the path forward for stroke drug discovery. To this end, this review calls for a consortium approach including academia, government (FDA/NIH), and pharmaceutical industry partnerships to define this path. It is also imperative that more attention is given to the evolving discipline of Translational Medicine. A key issue in this respect is the need to devote more attention to the characteristics of the drug candidate nature—target interaction, and its relationship to pharmacodynamic treatment end points. It is equally important that efforts are spent to prove that phenotypic outcomes are linked to the purported mechanism of action of the compound. Development of technologies that allows a better assessment of these parameters, especially in in vivo models are paramount. Finally, rational patient selection and new outcome scales tailored in an adaptive design model must be evaluated.
Continued failures of acute stroke clinical trials have been the subject of much debate, and there still remains hope for the next ‘magic bullet’ or mechanism, which will prove to be therapeutically active. To formulate better an optimal path for more successful research and development of treatments for acute ischemic/embolic stroke, an important initiative has been developed by teams of stroke scientists and clinicians termed (Stroke Therapy Academic Industry Round table)‘STAIR’ (STAIR, 1999, 2005). The STAIR forum, although not organized by any independent or recognized authority, held several conferences, which served as a venue to debate to deal with many critical issues in preclinical and clinical stroke drug development. The guidelines of the STAIR forum have been influential in stroke research in academia and pharmaceutical companies. However, difficulty in identifying novel stroke therapies has been highlighted recently by the SAINT II (Stroke-Acute Ischemic NXY Treatment II) study (Pearson, 2006), where an ‘antioxidant’ agent, NXY-059, failed in what has been considered an ‘improved and promising’ modern trial well founded on extensive preclinical data (Green and Ashwood, 2005). The negative outcome of the SAINT II study occurred despite the convictions of many in the experimental and clinical stroke community regarding the importance of the mechanism of action in acute ischemic neuroinjury and the impressive dossier of stringent preclinical and clinical pharmacokinetic studies associated with reactive free radicals (RFR) scavengers (Hess, 2006; Lees et al, 2006). Further optimism has been generated from the SAINT I trial, a large phase III study (Lees et al, 2006). In this brief review, we focused on the need for better and more effective and comprehensive strategies in stroke drug research and development, particularly in using Translational Medicine approaches for successful treatment of acute ischemic/embolic stroke. We emphasize the need for better biomarkers before commencing full-scale clinical trials and hope to encourage and intensify the debates in this most difficult area.
Modern drug discovery requires detailed knowledge of the molecular drug target selected for modulation, including its location, expression patterns in normal and disease cells, and the role of the target in the pathophysiology of the disease. Although many successful marketed drugs have been realized in spite of poorly defined mechanism of action, this is not the case for most stroke trials, where the neuroprotective agents have used quite well-characterized mechanism. One possible reason for stroke drug failure might be rooted in insufficient target validation in ischemic brain injury. Thus better target validation should be sought whenever possible. In reference to RFR as ‘the target’, there are difficulties in defining what the discrete reactive molecule(s) is. Of the core reactive substances, N, O, H, Cl, CSN, numerous molecular species are derived leading to vast number of RFR species markedly differing in biological half-life span, redox potential, targets of interaction (proteins, complex carbohydrates, lipids, and nucleic acids), and temporal existence. The lack of knowledge of which of the RFRs play a key role in stroke in reference to their targeted structures and temporal frames is a fundamental gap that enhances the risk in trials.
Recommendation: We suggest that more attention is given to target validation as part of the preclinical package and efforts are made to validate the target role in stroke pathophysiology in humans. A quantitative description of the molecular target location, its temporal expression and functional profile in relation to disease pathophysiology should be studied in detail. Although many marketed drugs have been developed without clear understanding of the role of the target in the pathophysiology of the disease, nonetheless it is now becoming obligatory for successful drug development.
Drug discovery aims to pursue compounds/biologicals, which are inherently potent and highly selective modulators (agonists or antagonists) of a discrete molecular target. Such compounds interact with their molecular targets in a way that does not inactivate the target itself but rather modulate its function. It is critical to establish an exposure-response relationship of the target to define the pharmacodynamic response. These correlations can define the extent of target modulation needed to generate the desired response while sparing undesired biological activities and assessing whether the benefit can be garnered in the range of the affordable exposure. The ability to establish such parameters with RFR is unfortunate mainly due to the elusive nature of the RFRs, their targets, and short half-life of the active species. Furthermore, drugs are molecules that interact with a specific target but do not fundamentally alter its biophysical properties. This is categorically distinct from the type of interaction that occurs with an antioxidant modality. This latter event requires chemical interaction that inactivates the target while chemically altering the drug. Such situations are most difficult to design rationally because the redox coefficients cannot be sufficiently inclusive and programmed for all the molecular species that exists at a critical time points. It is therefore difficult to track discrete target-compound interactions and their pharmacodynamic consequences.
Indeed, in none of the preclinical studies with NXY-059 (as is the case for all previous antioxidants trials) was such data reported. The pursuit of the ‘antioxidant’ program relied on biomarkers, which only indirectly suggested that the mechanism of action might have been realized.
Recommendation: We recommend that more consideration be given to define the scope and time frames of compound interaction with the target. Knowledge on this interaction in the locale of the injury and its pharmacodynamic consequences might aid in understanding how the drug-target interaction could lead to improved tissue salvage and thus a better functional outcome.
Animal models of human disease are needed and are cornerstones in the science of drug discovery. It is clear that many models are only approximation of human disease and as such are limited. In stroke, the thromboembolic event is central to the condition; yet, preclinical stroke models used for studies with NXY-059 have not studied this key component. Studies on nonhuman primates (NHP) looked at compound exposure at clinical relevant levels, the clinically relevant ‘therapeutic window’ (Marshal et al, 2001; Green and Ashwood, 2005), and also studied chronic functional recovery in NHP (Green and Ashwood, 2005, STAIR, 1999; 2005); however, thromboembolic conditions have not been pursued.
First, in vivo studies with NXY-059, like the majority of previous studies with antioxidant compound that ultimately underwent clinical evaluation, ignored important components of the human ischemic stroke (embolus, vide supra). This key scenario despite its obvious relevance to the human disease was not studied in any of the NXY-059 focal cerebral stroke models. The argument that ‘method of vessel occlusions is not relevant to the brain response to ischemia’ ignores the importance of clot-derived substances (e.g., thrombin) being flushed into the ischemic region by residual flow, possibly confounding the ischemic insult. A better simulation of the human condition focusing on the embolic cause of stroke need to be rigorously implemented in stroke experimental ‘modeling’ including NHP models.
Second, of the key in vivo data generated with NXY-059, the two NHP chronic studies are of special interest. Whereas the first study initiated treatment minutes after the onset of ischemia, the second NHP study initiated treatment 4 h after ischemia and also measured plasma compound levels at the end of drug administration. The second study, while clearly more ‘clinically relevant’ we believe still falls short of Translational needs. Although we do not have a detailed account of the recruitment time in SAINT II, it is likely that some proportion of patients were recruited toward the end of the ‘recruitment window’ of 4 to 6h, and represents a critical time frame not studied in NHP. Furthermore, although the drug exposure in NHPs achieved monitored levels in the clinical program, drug exposure at the end of treatment may not allow for extrapolation in both plasma and brain at the vulnerable period. Also, the lack of appreciable compound concentration in the brain (<1% of exposure) and speculative site of action (microvascular interfaces) erodes confidence in the proposed mechanism of action.
Third, pharmacodynamic biomarkers, such as salvage by the compound of proteins, lipids, and other structures, damaged by the RFR were not studied. In no in vivo study, it has been shown that the compound acting via RFR inhibition abolished RFRs-mediated damage, in the ischemic zone (or other sites), at any of the acute phases of the stroke paradigm studied.
Fourth, the contemporary experimental practice of ‘6 h window for drug efficacy’ in lieu of the current standard for clinical trials has no proof of predictive value because compounds that have demonstrated extended ‘window’ over NXY-059 in experimental studies, followed by long term and robust functional recovery, have not faired better in phase II efficacy trials (Marshal et al, 2001; Bogousslavsky et al, 2002).
Fifth, the demonstration that the compound has salvaged viable, but vulnerable, brain tissue, commonly termed ‘penumbra’ was neither confirmed in experimental models with NXY-059 nor considered in patient selection for the clinical trials. The hope of neuroprotection is the rescuing of neurons or other brain cells within the area compromised by ischemic environment, yet sufficiently viable for eventual recovery. Supportive data along these lines need to be provided.
We believe that clinical trial design for acute stroke treatments can be improved in at least two respects: (1) improved patient selection to maximize selection of patients capable of showing a treatment response and (2) improved clinical end-point selection to improve the specificity and informative of clinical outcome measures.
The use of diffusion/perfusion magnetic resonance ‘mismatch’ imaging to identify ‘salvageable’ brain tissue to define better eligible patients, who could benefit from neuroprotective drugs has been subject of much debate recently. However, this approach has never been validated and also faces difficult technological hurdles in both preclinical and clinical studies. Also, the use of ‘modality-specific’ clinical outcome measures, in addition to the commonly used composite measures (Rankin, Barthel, and NIHSS), might provide more selective information on specific neurological functions (e.g., robotic-driven motor function), which are more objective and reproducible could provide more consistent results. However, this approach has not been validated as yet and like D/PW-MR ‘mismatch’ for salvageable tissue assessment, may also turn to be wrong.
Our commentary is aimed at fostering and encouraging tripartite debate, where academia, government, and pharmaceutical industry scientists can join in developing a comprehensive ‘road map’ for stroke drug discovery and development. Specific issues discussed in the review can serve as possible issues for constructive debates while raising the real issues of validation.