Neurology -- Correspondence for Krapf et al., 65 (5) 690-695

Correspondence published:

Reply from the author: Hilmar Krapf, MD (29 December 2005)

Effect of mitoxantrone on MRI in progressive MS: Results of the MIMS trial: Stephen E. Nadeau (29 December 2005)

Reply from the author 29 December 2005

Hilmar Krapf, MD,
Department of Neurology
University of Tübingen, Hoppe-Seyler-Str.3, D- 72076 Tübingen, Germany
Send Correspondence to journal:
Re: Reply from the author

hilmar.krapf{at}dgn.de Hilmar Krapf, MD

Dr. Nadeau suggests that the statistical analysis is not sensitive enough to detect the difference in change of Gadolinium- enhancing (GE) lesions between patients treated with mitoxantrone 12 mg/m² (M 12) compared to placebo. He further argues that a possible dilution effect of patients with no GE lesions at baseline increases the Type II error and that the appropriate group for analysis should have been only patients with GE lesions at baseline.
We do not agree with this opinion. First, patients with no GE lesions at baseline are at risk to have GE lesions at month 12 and therefore can not be excluded from the analysis without introducing a severe bias into the results. Secondly, we already adjusted the analysis for the presence of GE lesions at baseline by using a stratified Wilcoxon-test according to van Elteren as described in the statistical methods of the paper. Thirdly, the use of only 1/3 of the patients in the analysis would decrease the power to detect any difference dramatically.
We agree with the correspondent that there are dramatic effects of M12 on number of GE-lesions and only small effects can be seen in the placebo group. The p-value of p=0.105 is a strong suggestion for a difference between these treatment groups. However, according to the non- randomized nature of this cohort and the small number of patients under investigation, no other valid statistical analysis would show a “significance” without introducing a substantial bias.
Disclosure: The MIMS study was supported by Wyeth-Lederle. R. Gonsette, MD, H.-P. Hartung, MD, and S.P. Morrissey, MD, have received honoraria from Wyeth-Lederle and Immunex Corporations for educational presentations (less than four and participation in a meeting at the Food and Drug Administration, Bethesda, MD). O.Z. was an employee of Lederle/Germany until 2000, and S.P. Morrissey, MD, was employed by Wyeth GmbH Germany from 2003 to 2005. O. Zenker, MD, and T. Zwingers, MA, report no conflicts of interest.

Effect of mitoxantrone on MRI in progressive MS: Results of the MIMS trial 29 December 2005

Stephen E. Nadeau,
University of Florida
GRECC-182 VA Medical Center, 1601 SW Archer Road, Gainesville, FL 32608-1197
Send Correspondence to journal:
Re: Effect of mitoxantrone on MRI in progressive MS: Results of the MIMS trial

snadeau{at}ufl.edu Stephen E. Nadeau

Krapf et al and the Mitoxantrone in Multiple Sclerosis study group (MIMS) [1,2] describe a disappointing impact of mitoxantrone treatment on MRI gadolinium enhancing lesions. Coming from this highly respected group, this article is likely to be influential. The MIMS investigators have also distinguished themselves for their statistical sophistication yet the disparity between the data and the conclusions born of statistical analysis is dramatic. The authors should further clarify the statistical approach which was not adequately detailed.
The authors concluded that at months 12 and 24, neither mitoxantrone 5 mg/m2 (M5) nor mitoxantrone 12 mg/m2 (M12) produced a statistically significant difference in the total number of scans exhibiting gadolinium enhancement (GE), relative to the placebo group. However, in the M12 group, 10 of 34 subjects had GE at baseline and 1/34 had GE at 2-year follow-up (p = 0.0064, Fisher's exact test). This is a dramatic result: M12 eliminated GE in all but one patient and the statistical analysis testifies to the low likelihood that this result could occur by chance. In the placebo group, 8/36 subjects had GE at baseline, 6/30 at 24 months (p = 0.7668, Fisher's exact test).
The authors also conclude that there was no impact of M12 on change in number of GE lesions between baseline and 24 months (p=0.105)(Wilcoxon test). At first glance, the Wilcoxon test appears to be an appropriately conservative statistic, but was it the optimal statistic? In the M12 group at baseline, 24 of the 34 subjects had no GE. The outcome in these subjects was substantially determined by a ceiling effect: they could not improve, they could only get worse. Since only one M12 subject had a GE scan at 24 months, at worst the MRI in only one of these 24 subjects got worse.
It is possible that the M12 subgroup of interest in this analysis should have been the group that had GE at baseline, and that the failure by the authors to find statistically significant results in their analysis could reflect the dilution caused by inclusion of the 24 subjects without lesions at baseline - in essence a Type II error.
References
1. Krapf H, Morrissey SP, Zenker O, Zwingers T, Gonsette R, Hartung H-P, et al. Effect of mitoxantrone on MRI in progressive MS. Results of the MIMS trial. Neurology 2005;65:690-695.
2. Hartung HP, Gonsette R, König N, et al. Mitoxantrone in progressive multiple sclerosis: a placebo-controlled, double-blind, randomised, multicentre trial. Lancet 2002;360:2018-2025.
The author reports no conflicts of interest.