IST-3 broke new ground for misleading statistics in stroke trials with its secondary ordinal analysis, demonstrating “benefit” in the presence of an overall negative, open-label, randomized trial of over 3,000 patients. Now, who here can decipher the Cochran-Mantel-Haenszel test? Team Genentech/Boehringer Ingelheim is hoping you can’t.
This is a retrospective, observational cohort from the Virtual International Stroke Trials Registry looking for a way around the pesky “contraindications” to tPA treatment that currently prevent large groups of patients from receiving treatment. These authors pulled non-treated “controls” from the cohort along with tPA-treated patients who had at least one contraindication or warning to tPA use and compared mRS outcomes at 90 days. The control group was significantly older and sicker – though their strokes were milder (NIHSS 12 [SD 9] vs. NIHSS 14 [SD 8]) – but the authors adjusted only for age and NIHSS. Their conclusion?
“Many of the warnings and contraindications of alteplase may not be justified and licences, guidelines, and protocols should be adjusted to accommodate recent data from registries and real-world experience.”
As the authors correctly note earlier in the paper, observational data combined from heterogenous trials spread over many years is likely rife with differences in care and selection bias. This alone renders their results invalid as anything but hypothesis-generating, rather than practice-changing.
The other issue is their primary outcome and statistical tool of choice:
“The primary outcome measure was the mRS at 90 days, analyzed across the whole distribution of scores with the use of the Cochran–Mantel–Haenszel test,”
Here’s an example of an unadjusted mRS distribution “favoring” alteplase:
Just at first glance, looks pretty much the same – perhaps even favors placebo (top bar). How the heck is this significantly positive towards tPA? Well, the CMH takes into all ordinal levels – even the perturbations between disability/living death/death at the bottom of the scale – as opposed to just the clinically relevant mRS 0-1 or mRS 0-2 that were primary outcomes in other stroke trials. So, the statistical significance in this test has nothing to do with the clinical efficacy of the treatment in question. Then, the adjusted OR of 1.29 (95% CI 1.00 – 1.66) is somehow based on a mélange of “dichotomized outcomes at 90 days (mRS 0–1, mRS 0–2, NIHSS 0–1, and mortality)”.
I’m afraid this simply looks like the authors dragged the VISTA data through every permutation of statistical tools possible until they found a test and a combined endpoint for logistic regression that came out in favor of tPA. And, then, sold it as practice-altering.
Disheartening.
Here’s the disclosures, for your reading pleasure:
BF has received modest honoraria for participation in clinical trials (Sanofi-Aventis). AVA has served as PI of CLOTBUST trial partially funded by Genentech, and currently serves as PI of CLOTBUSTER funded by Cerevast Therapeutics and consultant to Genentech. EB is an employee of Boehringer Ingelheim. JCG, CW, PL, and NKM have no relevant conflicts of interest. AM has served as a consultant for Boehringer Ingelheim, received speaker’s honoraria from Boehringer Ingelheim, and congress expenses from Lundbeck. NW has received expenses from Boehringer Ingelheim for his role as member of the steering committee in relation to the ECASS III trial with alteplase and served as a consultant to Thrombogenics as chairman of the DSMB. SITS International (chaired by NW) received a grant from Boehringer Ingelheim and from Ferrer for the SITS-MOST/SITS- ISTR. His institution has also received grant support toward administrative expenses for coordination of the ECASS III trial. NW has also received lecture fees from Boehringer Ingelheim and from Ferrer. AS and KRL have received research grants, modest honoraria for participation in clinical trials, and have a consultancy or advisory board relationship with manufacturers from drugs (including Boehringer Ingelheim). KRL administered the grant from Genentech for support of this study.
“Thrombolysis in Stroke Despite Contraindications or Warnings?”
stroke.ahajournals.org/…/STROKEAHA.112.674622.abstract