Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Fluoxetine for depressed adolescents: harms likely to exceed benefits
- Peter R Mansfield, Anne L Tonkin, Jon N Jureidini. (28 May 2005)
-
Re: Fluoxetine for depressed adolescents: harms likely to exceed benefits
- John S. March (14 August 2005)
-
Re: Fluoxetine for depressed adolescents: harms likely to exceed benefits
- Peter R Mansfield, Jon N Jureidini. (16 August 2005)
-
Re: Re: Fluoxetine for depressed adolescents: harms likely to exceed benefits
- John S March (16 September 2005)
-
Public release of raw data from the TADS trial
- Howard Mann (16 September 2005)
-
Re: Re: Re: Fluoxetine for depressed adolescents: harms likely to exceed benefits
- Michael Kozart (25 May 2007)
|
|
|||
|
Peter R Mansfield, Research Fellow Dept. General Practice, University of Adelaide, SA 5005, Australia, Anne L Tonkin, Jon N Jureidini.
Send response to journal:
|
March et al have not addressed our main concerns about the reporting of the TADS trial of fluoxetine for depressed adolescents.<1,2,3> They have not justified their accusations against us but we concede that our comments on the design of their trial were not clear. Our main point was that unblinded comparisons can mislead. March et al have not acknowledged that point. March et al claim “clinical meaningful” benefits from fluoxetine but their trial found no significant difference from placebo on their most informative primary endpoint, the Children's Depression Rating Scale (CDRS -R; P = 0.10). They have presented a table of absolute differences in the percentage of subjects experiencing benefits (27%) and harms (4%). March et al don’t disclose the confidence intervals nor what benefits and harms they included or excluded. In the original TADS report the point estimates for percentages of responders were 60.6% on fluoxetine alone vs 34.8% on placebo suggesting an extra 25.8% crossed the line to be classified as a responder. However there was no significant difference on the average scalar score so there must have been a group of subjects whose scores worsened or improved less on fluoxetine than the average improvement on placebo (the squeezing the middle effect).<4> March et al have not disclosed how large that group was. Point estimates in the original report suggest absolute increases in harms of 10.3% for psychiatric adverse effects, 6.6% for harm related events and 4.7% for suicide related events. Some subjects suffered more than one event. Many adverse events were of greater magnitude than the average benefit. We conclude that, as with other antidepressants, fluoxetine may have a tiny average benefit. However the magnitude of benefit is unlikely to exceed the magnitude of harms that may be less frequent but more severe. 1. TADS. Fluoxetine, cognitive-behavioral therapy, and their combination for adolescents with depression: treatment for adolescents with depression study (TADS) randomized controlled trial. JAMA 2004;292:807-20. 2. Jureidini J, Tonkin A, Mansfield PR. TADS study raises concerns. BMJ 2004;329:1343-4. 3. March JS; TADS Group. Authors of TADS study reply to letter raising concerns. BMJ. 2005;330:730-1. 4. Lenzer J. Journalists on Prozac. BMJ 2004;329:748 Competing interests: All authors are members of Healthy Skepticism www.healthyskepticism.org |
|||
|
|
|||
|
John S. March, Professor and Chief, Child and Adolescent Psychiatry Duke University Medical Center, Durham, NC 27705
Send response to journal:
|
To the editor: Re: http://bmj.bmjjournals.com/cgi/eletters/330/7493/730-b#108211 For Peter Mansfield and colleagues (1 2), “healthy skepticism” appears to reflect antimedication biases that are ideological in nature and not based on data. Intent-to-treat analyses from the TADS clearly demonstrate that the fluoxetine-containing conditions yield substantially larger effect sizes relative to CBT and to PBO on independent evaluator and teen reported dependent measures (3,4). The difference between scalar and dichotomous outcomes on the CDRS-R reflects scientifically interesting but clinically unimportant method variance (in particular, specifying a random intercept in a random coefficient regression model), which we are planning to address in a separate paper. We respond here to their assertion that we are in some way hiding patients who worsened on fluoxetine by “squeezing the middle,” an inane assertion first advanced by the BMJ’s medical reporter, Jeannie Lenzer (5). As part of a larger effort to evaluate safety outcomes, we examined shifting to worse from baseline across 12 weeks of treatments for the intent-to-treat sample using the CGI-Improvement scores as rated by an independent evaluator blind to treatment assignment. Consistent with the robust overall effect of time and, independently, of treatment3, far more subjects improved than worsened; those assigned to COMB again faired best. Specifically, zero of 107 (0%) of patients assigned to COMB, 2 of 109 (1.8%) in FLX, 0 of 111 in CBT (0%) and 2 of 112 (1.8%) in PBO shifted from baseline to a CGI-I score of 6 (much worse) or 7 (very much worse). Broadening the range to include a CGI-I score of 5 (worse) resulted in worsening in 2.8% of subjects in COMB, 3.7% in FLX, 9.9% in CBT and 7.1% in PBO. Thus, as before, the data directly contravene the assertions made by Mansfield and colleagues. It is a fundamental principle of evidence-based medicine (EBM) that the utility or value placed on a treatment (as contrasted to its significance or effect size) is open to clinical interpretation (6). If Mansfield and colleagues don’t like the idea of taking psychotropic medications, they are free not to recommend them, but they should be careful not to confuse their clinical preferences with science. John S. March, MD, MPH for the TADS Team Duke University Clinical Research Institute References 1. Mansfield PR, Tonkin A, Jureidini J. Fluoxetine for depressed adolescents: harms likely to exceed benefits. BMJ.com 2005. 2. Jureidini J, Tonkin A, Mansfield PR. TADS study raises concerns. BMJ 2004;329(7478):1343-d-1344. 3. TADS. Fluoxetine, cognitive-behavioral therapy, and their combination for adolescents with depression: Treatment for Adolescents With Depression Study (TADS) randomized controlled trial. Jama 2004;292(7):807-20. 4. March JS. Authors of TADS study reply to letter raising concerns. Bmj 2005;330(7493):730-1. 5. Lenzer J. Journalists on Prozac. BMJ 2004;329(7468):748-. 6. March JS, Chrisman A, Breland-Noble A, Clouse K, D'Alli R, Egger H, et al. Using and teaching evidence-based medicine: the Duke University child and adolescent psychiatry model. Child Adolesc Psychiatr Clin N Am 2005;14(2):273-96, viii-ix. Competing interests: Consultant to Lilly, Pfizer, Wyeth, GSK, Jazz; Drug for NIMH-funded studies from Lilly, Pfizer; Speaker for none |
|||
|
|
|||
|
Peter R Mansfield, Research Fellow Dept. General Practice, University of Adelaide, SA 5005, Australia, Jon N Jureidini.
Send response to journal:
|
Healthy Skepticism strives against both pro and anti-drug bias. We warn against techniques for misleading people including failure to disclosure crucial information and use of fallacious arguments including ad hominem and straw man attacks. March et al have still not disclosed all the evidence required to evaluate their assertions. We request that they re-read our letters and respond appropriately to what we really wrote. Our main concern remains March's overemphasis of the TADS comparison of placebo vs fluoxetine with CBT (which was not blinded) and underemphasis of the blinded comparison of placebo and fluoxetine which did not find a statistically significant difference on the scalar primary endpoint. Sometimes binary endpoint data for continuous variables give the impression that the treatment makes a worthwhile difference whereas closer examination of the more informative scalar data gives a different impression. Focusing on one endpoint or the other can lead to different decisions about whether to use a drug or not. That is not a clinically unimportant difference. Lenzer's "squeezing the middle" hypothesis makes sense of the discrepancy between the statistically significant difference on the TADS trial binary endpoint but no such difference on the scalar endpoint. Contrary to his claim to have refuted that hypothesis March has not provided the data required. We ask March et al to let us see how large the "squeezing the middle effect" is by providing us with a table of before vs after CDRS-R scores for all subjects in the fluoxetine and placebo groups of their trial. This would enable us to compare the full range of improvement, no change and deterioration in the fluoxetine and placebo groups. We also request the adverse event data for those groups. Until all the required data is disclosed black or white conclusions about the TADS trial are not justified. Meanwhile we urge caution because so far no antidepressant has been clearly shown to do more good than harm for children and adolescents. Competing interests: Competing interests: All authors are members of Healthy Skepticism www.healthyskepticism.org |
|||
|
|
|||
|
John S March, Professor Duke University Medical Center
Send response to journal:
|
To the Editor: We thank Mansfield and colleagues for their continued interest in the Treatment for Adolescents with Depression Study (TADS). Regarding the squeezing the middle hypothesis, we note (1) the TADS was powered on a dichotomized CGI-I score so that it is in no sense a secondary endpoint, (2) the CGI-I and CDRS-R are highly correlated (r = .8, p < .001), and (3), given the overwhelming trend toward improvement across time and by treatment, it is statistically impossible for a substantial trend toward worsening for fluoxetine (or other treatment condition) to be buried in the CDRS-R results (see Table 3 (TADS, 2004)). To be certain, we examined the percent of subjects by treatment group who had a 1-point (minimum possible) or greater deterioration on the CDRS-R at week 12 compared to baseline. Only 21 of 439 (4.7%) patients worsened: 1.9% in combined treatment, 6.4% in fluoxetine, 5.4% in CBT, and 5.4% in placebo. Thus, we conclude (as expected) that the CDRS-R and CGI-I results yield the same overall result, namely that combined treatment is superior and that there is no middle to squeeze hidden in the data. The TADS Team takes seriously questions Mansfield and others raise about the 12-week outcomes. With over 30 papers either “in press” or in preparation and more to come, we expect to address these questions in the process of addressing gaps in the literature about how best to optimize treatment for depressed teens. As we turn our attention to the 36-week outcomes, it is reassuring that secondary analyses of the 12-week data already presented at a variety of academic meetings and slated for publication shortly replicate the primary results favoring combined treatment with regard to safety, function, quality of life, and clinical remission, among others. Regarding the request for raw data, we plan to release a fully de- identified TADS data set into the public domain no later than three years following completion by the last patient of the final study stage (IV). Sincerely, John S. March, MD, MPH For the TADS Team Competing interests: Consultant to Pfizer, Lilly, Wyeth, Jazz; research drug for NIH study from Pfizer, Lilly. |
|||
|
|
|||
|
Howard Mann, Faculty Member University of Utah School of Medicine
Send response to journal:
|
In his latest response in a debate concerning the clinical relevance of the published results of the Treatment for Adolescents with Depression Study(TADS)trial [1], Prof. Marsh responds to requests for access to the trial's raw data by committing to "release a fully de- identified TADS data set into the public domain no later than three years following completion by the last patient of the final study stage (IV)." What justifies this delay ? Publication of the aggregate results of the trial's clinically relevant primary and secondary outcomes may be expected to influence clinical practice. Contentions about these results should be informed by analyses of the raw data as deemed necessary by any party participating in the debate. Public release of an anonymized raw data set concurrently with the publication of a trial's outcomes should be standard practice. The TADS report was published in 2004. What arguments can the TADS investigators advance for not doing so ? Howard Mann, M.D. [1]Fluoxetine, cognitive-behavioral therapy, and their combination for adolescents with depression: treatment for adolescents with depression study (TADS) randomized controlled trial. JAMA 2004;292: 807-20. Competing interests: None declared |
|||
|
|
|||
|
Michael Kozart, Sonoma County Department of Mental Health 3333 Chanate Rd, Santa Rosa, CA 95404 9
Send response to journal:
|
Dear Editor, Having followed the series of exchanges between March et. al, the authors who published reports from the TADS study, and authors critical of those published reports, I remain unsatisfied with the lack of explanation provided by March et al over a basic question posed originally by Jureidini and Tonkin BMJ 2004; 329: 1343-1344. Just why did the original TADS protocol fail to include a specific placebo control for the efficacy of fluoxetine plus cognitive behavior therapy? In other words, results reporting the superior efficacy of fluoxetine plus CBT were not matched against placebo plus CBT. Can the team of March et al provide a simple explanation for this lack of placebo control inclusion in their previous study? Competing interests: None declared |
|||