Published online before print October 19, 2006, doi: 10.1161/01.STR.0000248166.30486.4f
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Stroke. 2006;37:2878.)
© 2006 American Heart Association, Inc.
Letters to the Editor |
Differences in Response to Reperfusion Therapies in Acute Stroke Between Men and Women: Mediated by Sex or by Chance?
Institute for Clinical Research and Health Policy Studies, Tufts-New England Medical Center, Boston, Mass
Foothills Hospital, Calgary, Alberta, Canada
To the Editor:
Drs Sandercock and Lewis remind us of some of the very real hazards of subgroup analysis. Performing multiple "one-variable-at-time" subgroup analyses will inevitably yield spurious results.1,2 This problem of spurious false-positive results, however, is not one that is directly addressed by increased sample size or statistical power, as they imply. Further, they seem to argue that the proper response to the risk of spurious false-positive results is to cease all analyses on variation in the effect of thrombolysis in stroke until the IST-3 Trial is completed. We respectfully disagree.
Although multiple "one-variable-at a-time" subgroup analyses may sometimes yield misleading results, for treatments with both risks and benefits such as recombinant tissue plasminogen activator for stroke, reporting only the summary result is also likely to be misleading, because the average effect might not even reflect the risk-benefit trade-offs seen in typical patients within the trial.1–3 This is a true dilemma, and we believe that both terms of this dilemma deserve respect (because patients are harmed by type II as well as by type I error). Although completely satisfactory solutions to this dilemma may not be possible, one approach that we believe has promise is the use of extant databases to develop multivariate risk-benefit models.1–3 These models, informed in part by subgroup analyses such as ours, can then be used to explore treatment-effect heterogeneity when future randomized clinical trial data become available. Hypothesis-generating analyses on "old data" can guide a priori hypothesis-driven analyses on new data (as Sandercock and Lewis appear to implicitly concede at the very end of their editorial).
Regarding specifically our "inappropriate" analyses on the influence of gender on treatment effect, in the pooled analysis of the intravenous recombinant tissue plasminogen activator trials, because we had no a priori expectation of finding this effect, our primary hypothesis, as noted in the discussion, was that the effect arose by chance. Since uncovering this effect, there have been some confirmatory signals, which suggest the results might not be spurious. Perhaps, the most important of these is the relatively consistent finding that women with ischemic stroke do substantially worse than men in the absence of therapy,4–9 a finding confirmed by a recent (and entirely appropriate) subgroup analysis of the International Stroke Trial to which Dr Sandercock contributed.10 At the same time, no such effect is seen among lytic-treated patients, either in the combined trial database (n=1069, odds ratio for good outcome in men 0.92 [0.72 to 1.18]) or in CASES (n=1135, odds ratio 1.05 [0.82 to 1.24] [M.D.H., unpublished data, 2001]). The presence of a gender effect among the untreated and its absence among the treated implies a treatment-effect interaction of precisely the kind we found.
Additionally, the previous analysis by Kent et al was an adjusted analysis; outcomes in subgroups were presented in unadjusted form just for clarity and transparency. Both unadjusted and adjusted analyses are also reported in the PROACT-2 analysis. The choice for modified Rankin Scale 
2 for the PROACT-2 analysis was based on the fact that this was the primary outcome for the study, and appropriate for the severe stroke severity of enrollees. Selecting the modified Rankin Scale 1, which is reported in our analysis as a secondary analysis, could have been justified on other grounds (eg, consistency with the "inappropriate" analysis of the intravenous trials) and using this outcome would not have uncovered a treatment-effect interaction, as Sandercock and Lewis correctly point out.
Despite appearances to the contrary, we think the methodological disagreements between ourselves and the editorialists are more superficial than they seem. We agree that sorting out which patients benefit from thrombolytics in stroke will require new trials like the IST-3 and ECASS-3. Indeed, multivariate risk-benefit stratification may be most appropriate for interventions informed by a redundancy of prior clinical trials and subgroup analyses.11,12 Thus, we believe that the pending trials present us with more reasons, not less, to torture the old data.
Acknowledgments
Disclosures
None.
References
1. Hayward RA, Kent DM, Vijan S, Hofer TP. Reporting clinical trial results to inform providers, payers, and consumers. Health Affairs. 2005; 24: 1571–1581.
2. Hayward RA, Kent DM, Vijan S, Hofer TP. Multivariable risk prediction can greatly enhance the statistical power of clinical trial subgroup analysis. BMC Medical Research Methodology. 2006; 6: 18.[Medline] [Order article via Infotrieve]
3. Rothwell PM, Mehta Z, Howard SC, Gutnikov SA, Warlow CP. Treating individuals 3: from subgroups to individuals: general principles and the example of carotid endarterectomy. Lancet. 365: 256–265.
4. Lofgren B, Nyberg L, Osterlind PO, Gustafson Y. In-patient rehabilitation after stroke: outcome and factors associated with improvement. Disability & Rehabilitation. 1998; 20: 55–61.[Medline] [Order article via Infotrieve]
5. Weimar C, Ziegler A, Konig IR, Diener HC. Predicting functional outcome and survival after acute ischemic stroke. J Neurol. 2002; 249: 888–895.[CrossRef][Medline] [Order article via Infotrieve]
6. Glader EL, Stegmayr B, Norrving B, Terent A, Hulter-Asberg K, Wester PO, Asplund K; Riks-Stroke Collaboration. Sex differences in management and outcome after stroke: a Swedish national perspective. Stroke. 2003; 34: 1970–1975.
7. Di Carlo A, Lamassa M, Baldereschi M, Pracucci G, Basile AM, Wolfe CD, Giroud M, Rudd A, Ghetti A, Inzitari D; European BIOMED Study of Stroke Care Group. Sex differences in the clinical presentation, resource use, and 3-month outcome of acute stroke in Europe: data from a multicenter multinational hospital-based registry. Stroke. 2003; 34: 1114–1119.
8. Roquer J, Campello AR, Gomis M. Sex differences in first ever acute stroke. Stroke. 2003; 34: 1581–1585.
9. Kapral MK, Fang J, Hill MD, Silver F, Richards J, Jaigobin C, Cheung AM; Investigators of the Registry of the Canadian Stroke Network. Sex differences in stroke care and outcomes: results from the Registry of the Canadian Stroke Network. Stroke. 2005; 36: 809–814.
10. Niewada M, Kobayashi A, Sandercock PA, Kaminski B, Czlonkowska A; International Stroke Trial Collaborative Group. Influence of gender on baseline features and clinical outcomes among 17, 370 patients with confirmed ischaemic stroke in the international stroke trial Neuroepidemiology. 2005; 24: 123–128.[CrossRef][Medline] [Order article via Infotrieve]
11. Kent DM, Hayward RA, Griffith JL, Vijan S, Beshansky JR, Califf RM, Selker HP. An independently derived and validated predictive model for selecting patients with myocardial infarction who are likely to benefit from tissue plasminogen activator compared with streptokinase. Am J Med. 2002; 113: 104–111.[CrossRef][Medline] [Order article via Infotrieve]
12. Rothwell PM, Warlow CP. Prediction of benefit from carotid endarterectomy in individual patients: a risk-modelling study. European Carotid Surgery Trialists’ Collaborative Group. Lancet. 1999; 353: 2105–2110.[CrossRef][Medline] [Order article via Infotrieve]
This article has been cited by other articles:
 |
|
 |
|
  D. M. Kent, A. M. Buchan, and M. D. Hill The gender effect in stroke thrombolysis: Of CASES, controls, and treatment-effect modification Neurology, September 30, 2008; 71(14): 1080 - 1083. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||