Unfortunately, we have done it again. Another potentially promising method for acute stroke management has failed in a well-conducted clinical trial. In the IMAGES study,1 2589 patients were randomized to magnesium sulfate (MgSO4) or placebo, with administration started within 12 hours after symptom onset. The investigators found that although the drug is safe in acute stroke patients, it did not reduce the chances of death or disability. After looking at the data for a glimmer of hope, only subcortical stroke patients appeared to benefit, and because this was only a secondary analysis, it should only be regarded as a hypothesis generator.
Why did this happen? In preparation of this large trial, there were some preclinical data showing signs of efficacy for up to 12 hours after vascular occlusion.2–4 Then there was a dose optimization pharmacokinetic study that showed that MgSO4 could be given in a regimen to patients, which provided levels of drug that produced neuroprotection in the rodent stroke models.5 The study was appropriately powered, and it was simplified to reduce the burden on the local investigators.
There were early indications that the trial would fail. For example, only 1 study was cited4 to show that efficacy might extend to 12 hours after vascular occlusion. However, the animal model studies can be criticized for other reasons. I admit that in previous years, I considered it sufficient to simply show that a neuroprotective agent was efficacious when administered 5 minutes after occlusion induction using our stroke models. Despite our previous hopes, I now think that concept is mistaken. The animal models showed us things we did not want to know, such as that the time course of neuroprotection, in the agents we were testing, is relatively short. Putative neuroprotectants do not work if treatment is delayed past clinically reasonable time windows, such as 2 to 3 hours after stroke onset. Further justification for the use of MgSO4 in the IMAGES trial was based on only rat models, and there is no proof that a rodent model has predictive value for acute neuroprotective treatment in humans. Because the best estimate of the time window for treatment of ischemic stroke is 6 hours6 in the absence of extensive collateral circulation, it is unlikely that any neuroprotective, alone, can show such remarkable efficacy. Indeed, in the IMAGES trial, only ≈3% of the patients were treated within 3 hours of stroke onset and only approximately one-third received treatment within 6 hours. The only 2 clinical trials that produced statistically significant results in favor of intravenous therapy7,8 required treatment to start within 3-hour windows. It is relatively easy to recruit patients for longer windows, but hyperacute (<3-hour windows) therapy studies are much more expensive to conduct. To do so, an infrastructure must be in place to evaluate patients and initiate treatments quickly. Personnel must be available to stop whatever else they are doing and go immediately to the emergency department to manage the patients as the emergencies they truly are. The acute stroke patient will probably require urgent brain and vascular imaging by appropriately trained technicians, blood testing, electrocardiogram (EKG), and physicians and nurses with expertise in acute stroke management.
Why was MgSO4 studied? The IMAGES trial was sponsored by the Medical Research Council of the UK. Because the drug is widely available and inexpensive and because the design was a large simple trial, the study was relatively low-cost. Expensive proprietary drug trials, sponsored by pharmaceutical companies, have also been unsuccessful, so the means of funding cannot be the sole reason for failure. However, the lack of adequate reimbursement to the investigators for the real costs of therapy means that only a very long time window can be adequately investigated. The opportunity costs of rushing to the emergency department—and leaving a busy clinic—are considered by practitioners, subconsciously, if not overtly. Acute stroke therapy will be very expensive, even if very potent drugs are found. Regardless of the mechanism of drug action, the sooner that therapy is started, the better the chance to show benefit. “Time is brain.”
When a phase III clinical trial fails, the drug is usually not tested again. This occurs despite the fact that such an initial study may point the way to a better trial design. However, in this instance, magnesium therapy for acute stroke may not be dead. Probably the clearest signal from the IMAGES trial is that the time window duration was far too long. Also, very little tPA was used, despite the fact that it is the only currently approved treatment for acute stroke. It is logical to think that removing a vascular obstruction is likely to improve access of a neuroprotective to threatened tissue and produce some sort of positive interaction, because thrombolytics and neuroprotectants work by different mechanisms.
The FAST-MAG investigators have now completed a phase II trial,9 showing that it is feasible to administer MgSO4 safely before a patient arrives at the hospital. Because all indications are that hyperacute initiation of MgSO4 administration is safe, it is reasonable to try again, using a prehospital delivery strategy.
Even if MgSO4 does not prove to be effective, the idea that field administration can be accomplished will be a useful demonstration of the feasibility of that patient management method. For drugs that are truly safe, this is the next step toward development of a method of combination treatments, which many of us think is the best way to treat acute stroke. An ideal agent should be shown to be safe in all forms of stroke, including primary hemorrhage.
So, where does IMAGES leave us? One conclusion might be the obvious nihilistic interpretation: acute stroke therapy is just too difficult; thrombolysis is too hazardous; and combination trials with thrombolytics plus neuroprotectants are too complicated. By such reasoning, the implication is that spending more of our precious clinical research resources is a waste of time and money.
However, there is a much more hopeful message from the IMAGES results. There is no easy solution, and no matter how simple, a trial design that ignores the lessons of previous trials is doomed to repeat past failures. We have no other choice than to pursue the difficult trial design features that have become clearer over the past several years: (1) putative neuroprotectants must be combined with a thrombolytic; (2) therapy must begin within 4 to 5 hours, and preferably in much less than that; (3) treatments must be given using treatment regimens that are comparable with preclinical testing; (4) neuroprotectant drugs should be studied in patients with primary intracerebral hemorrhages to make possible field administration before transport and brain imaging. Of course, nothing can be guaranteed. Some neuroprotectants will fail despite doing everything correctly. But IMAGES demonstrates, I hope for the last time, that the old approach, no matter how convenient or inexpensive, will continue to fail. Also, it will be necessary to develop health systems that can educate the public and paramedics so that more patients seek care urgently. Our job is to assure that once a patient does access the system, we are ready to mobilize everything we need. Stroke therapy will be difficult, but there is no way to fool Mother Nature.
Section Editor: Marc Fisher MD
- Received May 4, 2004.
- Accepted May 4, 2004.
Izumi Y, Roussel S, Pinard E, Selyaz, J. Reduction of infarct volume by magnesium after middle cerebral artery occlusion in rats. J Cereb Blood Flow Metab. 2000; 11: 1025–1030.
Muir KW, Lees KR. Dose optimization of intravenous magnesium sulfate after acute stroke. Stroke. 1998; 29: 918–923.
Zivin JA. Factors determining the therapeutic window for stroke. Neurology. 1998; 50: 599–603.
Saver JL, Kidwell C, Eckstein M, Starkman S; the FAST-MAG Pilot Trial Investigators. Prehospital neuroprotective therapy for acute stroke: results of the Field Administration of Stroke Therapy—Magnesium (FAST-MAG) pilot trial. Stroke. 2004; 35: e106–e108.