Published online before print April 27, 2006, doi: 10.1161/01.STR.0000222922.77194.db
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Stroke. 2006;37:1367.)
© 2006 American Heart Association, Inc.
Letters to the Editor |
Response to Letter by Chen et al
Neurological and Mental Health Division, The George Institute For International Health, Royal Prince Alfred Hospital and The University of Sydney Australia
Institute of Health Sciences and Public Health Research, Academic Unit of Psychiatry and Behavioural Sciences, School of Medicine, University of Leeds, United Kingdom
Response:
Chen and Guo reviewed our recently published systematic review of pharmacological strategies for the management of depression after stroke1 and performed a meta-analysis of the continuous end points in this review. We had refrained from pooling these end points because of the considerable heterogeneity and methodological shortcomings in the studies reviewed, including small sample sizes and multiple methods used to assess depression both within and across studies. Moreover, Chen and Guo added data from a crossover study2 and excluded data from a study where mean and standard deviation data were extracted from a published figure.3
We conducted our systematic review and meta-analysis to Cochrane standards,4 in line with our peer-reviewed Cochrane protocol. Cochrane guidelines do not allow post hoc decisions about whether to include or exclude a trial in a systematic review in order to avoid selection bias. We maintain that the data in our systematic review should not be pooled for the following reasons:
1. Data should not be pooled if the trials under consideration are clinically or statistically heterogeneous.
2. Randomization only deals with bias and confounding if a trial is large enough for the randomization to overcome the play of chance. Because small trials were included in our review, there is a real possibility that groups had unbalanced characteristics at baseline, with the exception of exposure to treatment, so that between-group differences in outcomes may be misleading without appropriate adjustments being made for baseline differences.
3. There are several issues to consider when including crossover trial data in a meta-analysis.5 Although they are too numerous to be covered here, it is important to note that we could not assume there was no carry-over effect of the intervention. Also, outcome information from this trial was not available from paired analyses, and no data were available on within-individual comparison of treatment effects.
We thank Chen and Guo for their interest in our analysis, but maintain that the data they present are not sufficiently robust to influence clinical practice for the reasons outlined.
References
1. Hackett ML, Anderson CS, House AO. Managment of depression after stroke: a systematic review of pharmacological therapies. Stroke. 2005; 36: 1092–1097.
2. Robinson RG, Schultz SK, Castillo C, Kopel T, Kosier JT, Newman RM, Curdue K, Petracca G, Starkstein SE. Nortriptyline versus fluoxetine in the treatment of depression and in short-term recovery after stroke: a placebo-controlled, double-blind study. Am Journal of Psychiatry. 2000; 157: 351–359.
3. Lipsey JR, Robinson RG, Pearlson GD, Rao K, Price TR. Nortriptyline treatment of post-stroke depression: a double-blind study. Lancet. 1984; 1: 297–300.[Medline] [Order article via Infotrieve]
4. Alderson P, Green S, Higgins JPT. Cochrane Reviewers’ Handbook 4.2.2 (updated December 2003). 2004; In: The Cochrane Library, Issue 1.
5. Elbourne DR, Altman DG, Higgins JPT, Curtin F, Worthington HV, Vail A. Meta-analyses involving cross-over trials: methodological issues. International Journal of Epidemiology. 2002; 31: 140–149.
Related Article:
Stroke 2006 37: 1365-1366.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||