Electronic letters to:

Research:
Michael D. Hill, Alastair M. Buchan for The Canadian Alteplase for Stroke Effectiveness Study (CASES) Investigators
Thrombolysis for acute ischemic stroke: results of the Canadian Alteplase for Stroke Effectiveness Study
CMAJ 2005; 172: 1307-1312 [Abstract] [Full text] [PDF]
*eLetters: Submit a response to this article

Electronic letters published:

[Read eLetter] Stroke researchers appear to have an endless capacity for self-delusion
Jeffrey Mann   (16 May 2005)

Stroke researchers appear to have an endless capacity for self-delusion 16 May 2005
Top
Jeffrey Mann
None

Send letter to journal:
Re: Stroke researchers appear to have an endless capacity for self-delusion

jmannemg{at}earthlink.net Jeffrey Mann

In their paper, the authors conclude that "the outcomes of stroke patients undergoing thrombolysis in Canada are commensurate with the results of clinical
trials."  Which particular clinical trial's results do they use for comparison purposes?

The authors decided to compare the CASES stroke patients' rate of excellent stroke outcome result (mRS<1) to the results from the NINDS trial (see figure 3 in their paper) and they demonstrate that 31% of their CASES stroke patients had an excellent stroke outcome result compared to 39.9% of the NINDS tPA patients and 27.7% of the NINDS placebo patients. To make their comparative results look even better they use an adjusted mRS<1 value of 36.8% (a previously unused adjustment mRS<1 value), rather than an adjusted mRS<1 value of 31%, to make their comparison. However, why did they arbitrarily choose to compare their adjusted mRS<1 value of 36.8% to the NINDS placebo group's rate of excellent stroke outcome value of 27.7%. Why didn't they compare their rate of excellent stroke outcome value of 36.8% to the comparable value for the placebo patients from other tPA-for-stroke RCTs?  

In my analysis of tPA-for-stroke RCTs [1], I noted that when one specifically looks at stroke patients treated between 91-180 minutes, that the NINDS placebo patients had an "average" rate of excellent stroke outcome value of 25%, while the placebo patients from the ECASS/ECASS II/ATLANTIS trials had an 'average" rate of excellent stroke outcome value of 35.8%. Also, Kent et al. published the pooled results for all the male patients treated between 0-6 hours from the NINDS/ECASS II/ATLANTIS trials [2], and they could not demonstrate that tPA was effective in male patients on "average" (using an unstratified pooled analysis for all the male patients), because the "average" rate of excellent stroke outcome in male placebo patients was 36.7% compared to 38.5% for tPA-treated patients. Why was the "average" rate of excellent stroke outcome value of 36.7% so much higher than the comparable value for the NINDS trial's placebo patients (who were treated between 91-180 minutes) and who had an "average" rate of excellent stroke outcome value of 25% -- even though both placebo groups had a median baseline NIHSS score of 13-14? I discussed this subject at great length in my critical essay [1], and I demonstrated that if one uses a single "average" rate of excellent stroke outcome value, based on a pooled analysis (unstratified analysis), to quantify tPA's "apparent" therapeutic effect, that a significant imbalance in baseline stroke severity can radically alter the OR results -- even if one presumes that tPA is equally effective throughout the stroke severity range. The NINDS placebo group's rate of excellent stroke outcome value of 25% was artefactually low -- because there were far less very mild stroke patients and far more very severe stroke patients in the placebo group compared to the tPA-treated group. Also, the NIHSS 11-15 stroke severity placebo subgroup had an unexpectedly low rate of excellent stroke outcome value of 14%. Kent et al. recently published the stroke severity subgroup results from their pooled male gender analysis in a response letter to Stroke [3] -- I noted there was apparently no significant imbalance in patient numbers in the different stroke severity subgroups, and I also noted that the NIHSS 11-15 placebo stroke severity subgroup (which constituted 32.6% of the total number of placebo stroke patients) had a rate of excellent stroke outcome value of 33.5%. If the latter rate of excellent stroke value is  the "correct" value for untreated stroke patients who have a bNIHSS stroke severity score of 11-15, and if there is no significant imbalance in patients numbers between the stroke severity subgroups, then it would be more appropriate to use an "average" rate of excellent stroke outcome value of 36.7% as the comparator value for placebo group patients. Under those "comparative" circumstances, there is obviously no evidence that the CASES stroke patients derived a significant therapeutic benefit from tPA therapy (36.8% versus 36.7%).

I previously recommended that the rate of excellent stroke results from all the tPA-for-stroke trials should be plotted as a scattergram [3]. This particular stratified statistical technique would eliminate the ability of "baseline stroke severity imbalances" to be a confounder when assessing tPA's therapuetic effect, and it would enable stroke researchers to obtain a more accurate perspective of tPA's true efficacy in acute ischemic stroke. However, it would appear that stroke researchers cannot stop using a pooled (unstratified) analysis to quantify tPA's "aparent" therapeutic effect, while at the same time always using the same "single" inappropriately low placebo group rate of excellent stroke outcome value of 27.7% as the sole comparator value, and this one-sided view of reality could be perceived to be a form of self-delusion. Reality demands that "like" be compared to "like" and the burden of proof lies with stroke researchers -- they have to demonstrate that they have chosen the appropriate comparator placebo group's results when they attempt to quantify the "average" rate of excellent stroke outcome difference between untreated and tPA-treated stroke patients using a pooled analysis.  

 

References:

 

1. Mann J. A Critique of the Reanalysis of the NINDS trial.

Available at http://jeffmann.net/soapbox/IngallCrtitique.htm

2. David M. Kent, Lori Lyn Price, Peter Ringleb, Michael D. Hill, and Harry P. Selker.  Sex-Based Differences in Response to Recombinant Tissue Plasminogen Activator in Acute Ischemic Stroke: A Pooled Analysis of Randomized Clinical Trials. Stroke 2005 36: 62-65.

3. David M. Kent, Lori Lyn Price, Harry P. Selker, Peter Ringleb, and Michael D Hill.  Sex-Based Differences in Response to Recombinant Tissue Plasminogen Activator in Acute Ischemic Stroke - response letter. Stroke 2005 36: 929.

4. Mann J. A personal analysis of the NINDS trial using patient-level data.

Available at http://jeffmann.net/soapbox/NINDSpersonalanalysis.html

Conflict of Interest:

None declared