Re: Restoring Study 329: efficacy and harms of paroxetine and imipramine in treatment of major depression in adolescence
Restoring Study 329 illustrates three deficiencies of our medical literature.
While the authors concisely outline study design, endpoints, sample size determination, and ultimate failure of the study, their reanalysis fails by ignoring three principles of hypothesis testing. There are no pre-specified outcomes. There is no assessment of statistical power to detect a difference between the populations (Fisher RA 1925 Statistical Methods for Research Workers, Edinburgh: Oliver&Boyd) . Worse, no account is made for multiple comparisons (J Hsu Mltiple comparisons: theory and methods. CRC Press, Chicago, 1996, Bland JM, Altman DG. Multiple significance tests: the Bonferroni method. BMJ 1995; 310: 170). Neglecting multiple comparisons will results in fictitious associations (Austin PC, Mamdani MM, Juurlink DN, Hux JE Testing multiple statistical hypotheses resulted in spurious associations: a study of astrological signs and health. J CLin Epidemiol 59(9): 964-9 (2006)). The first deficiency of our medical literature is authors’ limited appreciation of statistics.
The second deficiency is readily demonstrated. Reviewers have a limited understanding of statistics.
Two good reasons for not permitting open access to clinical trial data are self- evident.
More telling and more damning is the third deficiency: the willingness of editors to headline unsupported claims of causal associations. No correction, no retraction, no apology, no comment. Shame on you.
Rapid Response:
Re: Restoring Study 329: efficacy and harms of paroxetine and imipramine in treatment of major depression in adolescence
Restoring Study 329 illustrates three deficiencies of our medical literature.
While the authors concisely outline study design, endpoints, sample size determination, and ultimate failure of the study, their reanalysis fails by ignoring three principles of hypothesis testing. There are no pre-specified outcomes. There is no assessment of statistical power to detect a difference between the populations (Fisher RA 1925 Statistical Methods for Research Workers, Edinburgh: Oliver&Boyd) . Worse, no account is made for multiple comparisons (J Hsu Mltiple comparisons: theory and methods. CRC Press, Chicago, 1996, Bland JM, Altman DG. Multiple significance tests: the Bonferroni method. BMJ 1995; 310: 170). Neglecting multiple comparisons will results in fictitious associations (Austin PC, Mamdani MM, Juurlink DN, Hux JE Testing multiple statistical hypotheses resulted in spurious associations: a study of astrological signs and health. J CLin Epidemiol 59(9): 964-9 (2006)). The first deficiency of our medical literature is authors’ limited appreciation of statistics.
The second deficiency is readily demonstrated. Reviewers have a limited understanding of statistics.
Two good reasons for not permitting open access to clinical trial data are self- evident.
More telling and more damning is the third deficiency: the willingness of editors to headline unsupported claims of causal associations. No correction, no retraction, no apology, no comment. Shame on you.
K. A. Papp, MD, PhD, FRCPC
Canada
Competing interests: No competing interests