Gina Kolata’s article about the efficacy of tPA in acute ischemic stroke in The New York Times was dismissive and misleading. And those are some of the milder words emergency physicians used to describe it as an online controversy erupted.
Ms. Kolata’s article misrepresented the data examining thrombolytics in acute CVA, and it distorted the debate surrounding this literature. Worst of all, she depicted anyone who questioned the evidence supporting tPA as uninformed charismatic bloggers rather than the expert clinicians and methodologists they are. Ask Jerome Hoffman, MD, emergency medicine’s preeminent opponent of tPA for stroke, who said Ms. Kolata characterized him as “a snake oil salesman.” (James R. Roberts, MD, interviewed Dr. Hoffman on the EMN Live podcast: http://bit.ly/EMNLive.)
Most importantly, Ms. Kolata’s article failed because it did not identify the core argument at the center of the debate on tPA for acute ischemic stroke. (“For Many Strokes, There’s an Effective Treatment. Why Aren’t Some Doctors Offering It?” The New York Times. March 26, 2018; https://nyti.ms/2pZXnkE.) But I’m getting ahead of myself.
Stroke care drastically and permanently changed in November 1995. The New England Journal of Medicine published NINDS-2, and ushered in the interventional era of acute ischemic stroke. (N Engl J Med 1995;333:1581.) No longer were we powerless in managing these patients. We could finally offer them more than an aspirin to chew, a corner in which to sit, and a consultation with a neurologist the next morning. But NINDS-2 was not the first trial examining thrombolytic therapy for acute ischemic stroke. In fact, three trials preceded it, all of which were negative (NINDS-1, MAST-I, ECASS-1) and two found an increase in mortality for patients given thrombolytics. (N Engl J Med 1995;333:1581; Lancet 1995;346:1509; JAMA 1995;274:1017.)
These were magically forgotten with the impressively positive NINDS-2, which demonstrated a 13 percent absolute increase in patients given tPA who were alive and independent (mRS of 0 or 1) at 90 days. (N Engl J Med 1995;333:1581.) Supporters justified the six percent absolute increase in symptomatic intracranial hemorrhage by arguing that it did not increase 90-day mortality (21% vs 17%).
But what about those three negative trials? What made NINDS-2 different from those? Was it the agent? Supporters claimed that tPA was the superior thrombolytic, and said we should ignore trials studying other agents. Was it time? NINDS examined patients who received tPA within 180 minutes of symptom onset (half within 90 minutes). Two of the earlier trials examined patients who received thrombolytic therapy over a much broader treatment window.
Was it the patient population? The NINDS authors used strict selection criteria to determine which patients were acceptable candidates. And, of course, a fourth reason was proposed by a less enthusiastic contingent: random chance. This more skeptical group posited that an intervention with little or no efficacy would eventually demonstrate positive results simply by chance alone if it were studied enough times. They reminded the more eager supporters of tPA therapy that taking these results at face value without further validation, even if the NINDS-2 findings might be true, was not only bad science but even worse medicine.
Ignoring the Data
The FDA, despite these warnings, fast-tracked tPA’s approval for acute ischemic stroke if given in under three hours, and all other trials attempting to validate this benefit were abandoned.
Six consecutive trials were published following NINDS, all examining time windows greater than three hours. All six were negative, and four demonstrated harm. (N Engl J Med 1996;335:145; JAMA 1996;276:961; Lancet 1998;352:1245; JAMA 1999;282:2019; Stroke 2000;31:811; Lancet Neurol 2009;8:141.) It was not until the 2008 publication of ECASS-3 that another trial examining thrombolytics for ischemic stroke demonstrated benefit. (N Engl J Med 2008;359:1317.) These benefits, though not as impressive as those of NINDS-2, were convincing enough to make us forget the multitude of negative trials examining similar time windows. Unlike NINDS-2, we were unable to claim ECASS-3 was different from these other negative trials examining similar patients during similar time windows. Instead, they were ignored.
Meta-analyses have been performed since then, guidelines have been written, and financially incentivized performance measures have been created. Each of these add some authoritative legitimacy, but none can circumvent the fact that the science is incomplete. No matter how small a p-value you tag onto a meaningless effect size, how many industry-sponsored names you add to a level 1A recommendation, or how big a financial incentive you attach to administering tPA, it does not replace a high-quality trial validating the original NINDS trial.
Why in the intervening 23 years since the NINDS trial do we exist in a state of evidentiary uncertainty? The fundamental underpinning of scientific inquiry is the ability to reproduce results. Why have we ignored one of the most basic tenets of the scientific method and not validated the NINDS findings? The answer is simple: A second positive trial offers nothing from a monetary standpoint. But a negative trial disproving the first would invalidate its findings, and harm those who stand to gain financially from the wide, unrestricted use of tPA.
Elliot Grossbard, MD, a Genentech scientist, stated it best in an internal communication about further trials comparing streptokinase with tPA for myocardial infarction: “We don’t know how another trial would turn out. And if we don’t come out ahead, we would have a tremendously self-inflicted wound. … [Another study] may be a good thing for America, but it wasn’t going to be a good thing for us.” (Marsa L. Prescription for Profits: How the Pharmaceutical Industry Bankrolled the Unholy Alliance Between Science and Business. New York: Scribner, 1997.)
Such methodologically negligent practices are not isolated to the use of tPA in acute ischemic stroke, but are rampant throughout modern medical science. They are the natural consequence of a society that values profit over scientific truth. The larger issue of science in medicine is a far more difficult conversation to have, so instead we discuss p-values, number need to treat, and other methodological quandaries.
The question should not be, Is tPA effective for treating acute CVA, but why, after 23 years, do we still not know the answer?