To the Editor:
The final conclusion of the recently published article1 on lubeluzole treatment of acute ischemic stroke, that “treatment with lubeluzole within 6 hours of the onset of ischemic stroke resulted in improved clinical outcome at three months with no safety concerns,” may not result from the data presented and the way they were analyzed. The primary aim of this trial was to test the null hypothesis that treatment with lubeluzole results in outcome similar in terms of death at 3 months to placebo treatment. As the primary end points “all deaths” showed no statistically significant difference between groups, the null hypothesis cannot be rejected. The fact that the number of deaths were (not statistically significant) lower in the lubeluzole group may well be due to chance because of small numbers or to differences in baseline characteristics; fewer males and fewer patients with various types of cardiac disease in the lubeluzole-treated group may have favored better outcome in the treatment group. From the analysis description, adjustments for these differences cannot be inferred. Secondly, fewer patients in the lubeluzole group died from hemorrhage, but no reason other than a possible effect of lubeluzole explaining this difference was discussed. Was there perhaps a higher number of patients with hemorrhage in the placebo group in the first place, thus favoring better outcome in the lubeluzole group? The statistically insignificant lower death rate in the lubeluzole group was due in part to the apparently lower number of deaths from congestive heart failure. However, can this be explained by a specific mode of action of lubeluzole, by chance, or by the fact that there were fewer patients with congestive heart failure in the lubeluzole group at baseline? A most prominent effect of lubeluzole might have been expected in deaths directly related to stroke. However, it turns out that lubeluzole did approximately 10% (not statistically significant) worse: odds ratio (OR), 1.11; 95% confidence interval (CI), 0.48 to 2.57 (31/368 lubeluzole versus 27/353 placebo). Considering deaths that were primarily vascular related, lubeluzole again fared worse: OR, 1.19; 95% CI, 0.74 to 1.92 (57/368 lubeluzole versus 47/353 placebo), a finding mainly due to a statistically significant increase in cardiovascular deaths in the lubeluzole-treated group compared with the placebo group (21 versus 7; OR, 2.99; 95% CI, 1.22 to 7.34). This finding is all the more surprising considering the lower number of patients with cardiovascular disease at baseline in the lubeluzole group, and it raises a serious question about the cardiovascular safety of lubeluzole. In this respect, I do not think it is appropriate to combine patients with cardiac pump failure with those who have ischemic heart disease.
Also, the analyses performed on the secondary end points do not allow a definite conclusion about the clinical value of lubeluzole. Various secondary end point analyses were done without adjustment for multiple testing in the same patient sample. Furthermore, a bias by a difference in outcome between lubeluzole and placebo patients who were eventually not included in the analysis cannot be excluded. If we look at a frequently used scale to measure functional outcome after stroke, the Rankin Scale, there was no statistically significant reduction in the patient category “dead, or moderately or severely handicapped” at 3 months: 227/357 lubeluzole versus 237/337 placebo; OR, 0.74; 95% CI, 0.53 to 1.03 (all confidence intervals with Yates’ correction). Because the point estimate indicates a clinically relevant treatment effect, the upper limit of the 95% CI does not exclude that lubeluzole might worsen instead of improve stroke patients’ outcome. Therefore, neither primary nor (relevant) secondary end point analysis allows a definite conclusion in favor of lubeluzole. Some stroke patients (those with ischemic heart disease?) seem to be at an increased risk for myocardial infarction or sudden death by lubeluzole. Until a more reliable estimate of a possible favorable treatment effect or adverse effects in stroke in general and patient subgroups becomes available, the conclusion that lubeluzole improves clinical outcome in stroke patients remains premature.
- Copyright © 1998 by American Heart Association
Grotta J, for the US and Canadian Lubeluzole Ischemic Stroke Study Group. Lubeluzole treatment of acute ischemic stroke. Stroke. 1997;12:2338–2346.
We agree with Dr Lodder that we found no statistically significant effect on overall mortality, as is clearly stated several times in our article. Therefore, his speculation regarding differential effects of lubeluzole versus placebo on mortality due to various causes, such as hemorrhage, congestive heart failure, and stroke, is not justified. None of these differences reached statistical significance (for example, 31/368 lubeluzole patients versus 27/353 placebo patients having stroke-related deaths; CI, 0.48 to 2.57). Had we embarked on such a “fishing expedition” in our analysis, we would have been appropriately criticized.
The secondary analyses of the effect of lubeluzole versus placebo on good outcomes as measured by the Barthel Index, Rankin Scale, and National Institutes of Health Stroke Scale scores were prespecified. Furthermore, dichotomization of these scales into “good” versus “poor” outcome is appropriate, given the nonnormal distribution of scores on all of these scales in stroke patients at 3 months. There is precedent for such analysis in previously successful clinical therapy trials in stroke.R1 Finally, all patients were included in the analysis. The positive effect of lubeluzole was detected by a logistic regression analysis across all three possible outcomes: good, poor, or dead.
We stand by our conclusion that in this study, treatment with lubeluzole was associated with improved clinical outcome with no safety concerns. These findings need to be corroborated.