My goal is not to debate the literature in detail, but to illustrate, using the policy’s stated logic, why the evidence is insufficient for a level A recommendation. I ask you to accept one statement: Controversy in interpretation of the data exists, precluding consensus in our specialty. Let’s focus on the Level A recommendation for tPA for stroke within 3 hours of symptom onset. For your reference, here is the definition of an ACEP Level A recommendation:

“Level A recommendations. Generally accepted principles for patient management that reflect a high degree of clinical certainty (i.e., based on strength of evidence Class I or overwhelming evidence from strength of evidence Class II studies that directly address all of the issues).

I applaud the panel for noting that NINDS was the only positive trial, along with a Class II meta-analysis, including NINDS, supporting this recommendation. However, I do have concerns that this amount of evidence was deemed sufficient for a Level A recommendation. The policy also points out that other large, randomized trials have been conducted (e.g. Australian Streptokinase trial, Multicenter Acute Stroke Trial–Italy, Multicenter Acute Stroke Trial–Europe and ECASS I). No benefit was shown, and “Some were halted early because of excessive mortality in the treatment arm.” Unfortunately, these studies were discounted, as their design differed from NINDS in timing of administration, dose, or thrombolytic agent. ECASS II, a Class I study with the same dose, but a longer time frame (6 hours), was also discussed. The document states there was no difference in outcomes, even for the 20% of patients treated within 3 hours. However, the parenchymal intracranial hemorrhage (ICH) rate was 12% in the tPA group compared to 3% of controls. Atlantis (also 6 hours) was also reviewed.

They noted it was discontinued after enrolling 142 patients due to an increased ICH rate in those treated within 5-6 hours. The protocol was modified to 0-5 hours, and after enrolling 31 patients was modified again to 3-5 hours. Again, the panel concluded that no difference in efficacy was shown, but a difference in ICH rate of 7% v. 1% was experienced. In 59 patients treated within 3 hours, 61% of the tPA vs. 26% of the controls met the primary endpoint (NIHSS score of 0-1 at 90 days).

Despite the limitations noted by the panel in many of these studies, they cited the Class II meta-analysis as evidence for support of their Level A recommendation. That meta-analysis included the data from some of these studies (NINDS, ECASS, ATLANTIS, and EPITHET). How can one discount a study’s results to later accept a reanalysis via compiled data from a meta-analysis using those same studies? If this approach is acceptable, why was Dr. Hoffman and Dr. Schriger’s graphic reanalysis of NINDS discounted? The policy reports that the authors concluded that tPA had a minimal change on NIHSS scores at 90 days. The policy appropriately points out that change in NIHSS was not a primary outcome of NINDS. However, the graphic reanalysis provides a value added perspective. The small beneficial difference demonstrated in NINDS disappears when change in NIHSS score is measured. The clinical policy’s analysis of the data may just as reasonably be interpreted to show that the data are inconclusive or too limited to prove benefit or safety.