Neurology -- Correspondence for Elterman et al., 57 (8) 1416-1421

Correspondence published:

Randomized trial of vigabatrin in patients with infantile spasms: John P Osborne, Andrew L. Lux, Stuart W. Edwards, Eleanor Hancock, Anthony L. Johnson, Colin R. Kennedy, Finbar J K O'Callaghan, Richard W. Newton and Christopher M. Verity (4 June 2002)

Randomized trial of vigabatrin in patients with infantile spasms 4 June 2002

John P Osborne
Royal United Hospital Bath United Kingdom,
Andrew L. Lux, Stuart W. Edwards, Eleanor Hancock, Anthony L. Johnson, Colin R. Kennedy, Finbar J K O'Callaghan, Richard W. Newton and Christopher M. Verity
Send Correspondence to journal:
Re: Randomized trial of vigabatrin in patients with infantile spasms

mpsjpo{at}bath.ac.uk John P Osborne, et al.

The study of Elterman et al. is a significant advance in our understanding of the treatment of infantile spasms, particularly in the clear definition of response and relapse in terms of both the clinical and neurophysiological response, in its size and in the use of actuarial curves (survival curves) for responders [1]. There are, however, some weaknesses in the published data, and in their interpretation and analysis.
The results were reported on almost 80% (142/179) of children randomized to receive treatment rather than an intention-to-treat analysis of the 179 children allocated a treatment - as recommended in the CONSORT statement [2]. The data show that 16 results are missing from those exposed to high dose and nine from the low-dose group; two cases were not exposed, and for 10 cases the report form is not available. These numbers (37 of those randomized compared to only 32 responders) are large enough to distort the findings, if the responses in these infants differed from the responses in those reported.
The reported primary response rate in children on "high-dose" vigabatrin (normal dose in the UK) was low at 24/67 (36%) and for low-dose vigabatrin it was 8/75 (11%). Of those with tuberous sclerosis (TS), 13/25 (52%) responded. Only 19/117 (16%) with other underlying aetiologies responded. This confirms again that vigabatrin is most efficacious when treating tuberous sclerosis. The proportions on low and high dose are not given by aetiology however and the paper should have stated these and the odds ratios (the Mantel-Haenszel test statistic) of a response for the two different groups: those with and those without tuberous sclerosis [3].
Turning to the actuarial curves, the x-axis should be calibrated as "time from randomly allocating treatment" rather than "days of vigabatrin therapy" since the duration of therapy depends not only on efficacy but also on tolerability. Figure 5 suggests, for example, that all children with TS were responders by 71 days, where in fact two had been withdrawn from treatment and one was a non-responder. Children who stopped treatment within the 3-month period because of lack of treatment effect or presence of adverse effect should, rather than being right-censored, be classified as non-responders.
Figure 4 seems to show a persistently poorer outcome in children who were initially treated with low-dose, and the difference persists despite the study protocol allowing a change to higher doses of vigabatrin and the addition of other treatments. This could be due either to a long-term adverse effect of an ineffective initial dose of vigabatrin on the response rate, or to right censoring, or to selection bias or to chance. The placebo controlled study of Appleton et al showed no significant difference in primary outcome but the group initially treated with placebo for 5 days appeared to have a significantly better outcome at 6 months than the group initially treated with vigabatrin [4]. This is the opposite effect to that seen by Elterman et al. We think these differences are most likely due to chance, but the discrepant findings argue potently for further studies.
This study ought not to claim that it confirms both the safety and efficacy of vigabatrin. The adverse effect that is potentially most serious and irreversible is vigabatrin-attributed visual field loss. There is no validated method for excluding such losses in children of this age. The best estimate of risk for such losses must be at least one-third - the same as that risk in adults or older children. [5]
Most importantly, the reported increase in response rate over the 3- month follow up period could be due to treatments other than vigabatrin since these were allowed after 21 days. Regrettably, no information on other treatments is given.
References:
[1] Elterman RD, Shields WD, Mansfield KA, Nakagawa J; and the US Infantile Spasms Vigabatrin Study Group. Randomized trial of vigabatrin in patients with infantile spasms. Neurology 2001;57:1416-1421.
[2] Moher D, Schulz KF, Altman DG. The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials. JAMA 2001;134:657-662.
[3] Lang TA, Secic M. How to Report Statistics in Medicine: Annotated Guidelines for Authors, Editors, and Reviewers. Philadelphia, PA: American College of Physicians; 1997:249.
[4] Appleton RE, Peters ACB, Mumford JP, Shaw DE. Randomised, placebo -controlled trial of vigabatrin as first-line treatment of infantile spasms. Epilepsia 1999;40:1627-1633.
[5] Wild JM, Martinez C, Reinshagen G, Harding GFA. Characteristics of a unique visual field defect attributed to vigabatrin. Epilepsia 1999;40:1784-1794.