Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Concerns about the TADS study
- Jon N Jureidini, Anne L Tonkin, Peter Parry, David B Menkes, Peter R Mansfield, Chris J Doecke (25 September 2004)
-
Lenzer's Piece Exemplifies Journalistic Errors
- Nicholas A. DeMartinis (11 October 2004)
-
Dr DeMartinis is Mistaken
- Jeanne Lenzer (11 October 2004)
-
Statistical Analysis in TADS
- Stephen L. Black (11 October 2004)
-
Reply to Stephen Black and Nicholas DeMartinis
- Peter R Mansfield (15 October 2004)
-
CBT versus both placebos and drugs, or a mistake?
- Susan F Anspach (7 December 2004)
|
|
||||||||||||||||||||
|
Jon N Jureidini, Head, Department of Psychological Medicine Women's and Children's Hospital, North Adelaide, 5006 SA, Australia, Anne L Tonkin, Peter Parry, David B Menkes, Peter R Mansfield, Chris J Doecke
Send response to journal:
|
We have additional concerns about the methodology and reporting of the Treatment
for Adolescents With Depression Study (TADS).[1]
MethodologyValid comparison between groups requires that they are subject to the same methodology. Yet TADS consists of two separate randomised studies: a double-blind comparison of fluoxetine (109 subjects) vs placebo (112); and an unblinded comparison between CBT alone (111) and fluoxetine+ CBT (107). The lack of patient-blinding and placebo-control in the latter group is likely to exaggerate the benefit seen in the fluoxetine + CBT group, who receive more face-to-face contact and know (as do their doctors) that they are not receiving placebo. Comparing results across all four groups is therefore misleading. The authors’ claim that a CBT + placebo arm would have been ‘both too expensive and too artificial to have clinical relevance’ is unconvincing. Invalid comparisons are neither value for money nor relevant.EfficacyThe valid finding from TADS is the lack of a statistical advantage of fluoxetine over placebo on the primary endpoint, the Children’s Depression Rating Scale (CDRS-R; p = 0.10) but this was not mentioned in the abstract. Putting together that result with the small or absent advantages to fluoxetine with other endpoints (see table) and in other studies,[2] the benefits of fluoxetine, like all other antidepressants, are of doubtful clinical importance for children.
Adverse EffectsThere are a number of reasons why adverse events and suicidal behaviour may be of greater magnitude than the TADS paper suggests. Despite small numbers, more subjects leaving the study than reporting adverse effects, and the splitting of adverse events into multiple groups, there were significantly more psychiatric adverse events in the fluoxetine group than the placebo group (chi-squared (1 df), p = 0.047). Despite small numbers and the exclusion of known suicidal behaviour, TADS found a trend to more suicidal behaviour (6 attempts in the fluoxetine groups, versus 1 in the no-fluoxetine groups), consistent with other trials of SSRIs. We are less reassured than the authors by the fact that no attempt was fatal. Suicide is a rare event so that a study the size of TADS should be expected to miss a significantly increased risk. In summary, the data do not support the TADS authors’ optimistic conclusions. The benefit-harm balance of SSRI therapy for depression in childhood and adolescence has yet to be shown to be favourable. 1. Treatment for Adolescents With Depression Study Team. Fluoxetine, Cognitive-Behavioral Therapy, and Their Combination for Adolescents With Depression: Treatment for Adolescents With Depression Study (TADS) Randomized Controlled Trial JAMA; 2004; 292:807-820 2. Jureidini J, Doecke C, Mansfield P, Haby M, Menkes D, Tonkin A. Efficacy and safety of antidepressants for children and adolescents BMJ. 2004; 328:879-83Competing interests: None declared |
||||||||||||||||||||
|
|
||||||||||||||||||||
|
Nicholas A. DeMartinis, Assistant Professor University of Connecticut School of Medicine, Farmington, CT 06030
Send response to journal:
|
Lenzer’s criticism of journalist's coverage of the TADS trial identifies failures in reporting certain details of the study findings, failure to address methodological issues, and failure to identify potential biases of scientists involved in the TADS study. Her quite valid points are undermined, however, by her repetition of the mistakes of those she criticizes. More concerning is the pattern of these omissions, which suggests the possibility that more than lazy reporting is involved. Lenzer noted that the Prozac monotherapy group failed to perform better than placebo on one of the two primary outcome measures, and through a quote from Peter Mansfield downplayed the significance of the positive findings with the other primary outcome measure. Unmentioned was the planned week 12 CDRS efficacy analysis in which Prozac monotherapy differentiated from placebo. She also didn’t mention the study finding that therapy alone was ineffective, an outcome that was absent from virtually every media report about the study. In fact, the course of improvement of the therapy treatment group was virtually identical to that of the placebo-treated subjects who met for much briefer visits at less than half the frequency of therapy visits. The only unreported facts Lenzer identified questioned medication efficacy. With regard to questioning methodology, Lenzer notes that most reports on the TADS trial did not mention that the Prozac and Prozac plus therapy groups were unblinded, implying that this contributed to the efficacy findings for the combined group. She didn’t mention that the study used blinded raters to moderate the effects of unblinded treatment. Another methodological issue ignored in the media and the majority of recent scientific reviews is the possibility that the failure of antidepressants to differentiate from placebo in industry trials represents a Type II error: that study design issues impaired the ability to determine a true treatment effect of antidepressants. This explanation is consistent with the abundant clinical evidence of efficacy but has been ignored by the media and most of the recent reviews of child antidepressant trials. Again, the only unreported study methodology issues Lenser identifies are those that question medication efficacy for childhood major depression. Lenser again repeats the mistakes of other journalists by failing to identify potential bias in a source in her accompanying article in the U.S FDA’s black box warning on antidepressants and suicide. She quotes several statements by Dr. Peter Breggan on the dangerousness and ineffectiveness of antidepressants, but neglects to mention his stance against all use of medications for psychiatric conditions, including schizophrenia and bipolar disorder. Knowledge of this bias is essential for the reader to assess the veracity of his statements, but readers unfamiliar with Breggan’s writings are left in the dark. The problems with journalistic coverage of the childhood antidepressant treatment debate have indeed been rampant, but mostly characterized by an overemphasis on negative information about medication treatment rather than positive. Lenser seems to imply that most of the problems she identifies in the TADS study coverage are due to lazy reporting, but the pattern of her omissions and her clear knowledge of the issues suggests that bias may play a role in her own failures. The needs of clinically depressed children will only be served by a balanced and objective critical assessment of all treatments for both safety and efficacy, which has unfortunately been lacking in the media and much of the scientific debate. Competing interests: Clinical trial grants: Lilly, Wyeth, Cephalon, Bristol Meyers Squibb Lecture honoraria: Wyeth, Lilly, Pfizer, Astra Zenica, Glaxo, Bristol Meyers Squibb Stock: none |
||||||||||||||||||||
|
|
||||||||||||||||||||
|
Jeanne Lenzer, medical investigative journalist Kingston, NY 12401 USA
Send response to journal:
|
My article, Journalists on Prozac, appeared in the Media Review section of the BMJ – not the news or clinical reviews section. As a media review it is not my job to provide a blow-by-blow recapitulation of the study in question or of the basic clinical issue addressed. My job was to examine why news headlines and text was so enthusiastically positive while the clinical trial itself was far from robust in its methodology and even less robust in terms of its putative positive outcomes. My job, as a critic and as a journalist, is to point out precisely and factually what the divergences were and how some of those divergences came about. I do this not just because I, like many others, am concerned about the distortion of scientific evidence, but because as a journalist I want to educate myself so that I don’t make some of the same errors I believe were made in the reporting that followed the TADS study. Dr DeMartinis says that in my article I identified “failures in reporting certain details of the study findings, failure to address methodological issues, and failure to identify potential biases of scientists involved in the TADS study.” He then calls my concerns “quite valid” but says I undermined those valid points by failing to point out some positive outcomes of the TADS trial. If Dr DeMartinis does indeed find those concerns valid, and he truly believes my job was to restate both sides of the issue as if it was a news report or clinical abstract, I wonder, did he apply those same criteria to himself and did he complain to the Washington Post, the New York Times, National Public Radio and other media that they failed to report the or the less than positive outcome on one of the scales used to measure the results? In regards to my use of Dr Peter Breggin, Dr DeMartinis charges that Dr Breggin takes a “stance against all use of medications for psychiatric conditions.” I don’t know where Dr DeMartinis gets his information but Dr Breggin reassures me that he has no ethical, religious or other proscription against the use of psychotropic drugs. Instead, his concerns are based on his clinical evaluation of the risk/benefit ratio of the drugs he has evaluated. Dr Breggin issued early warnings about tardive dyskinsias caused by neuroleptics at a time when others sniffed at the idea – yet his insights are now widely accepted. Dr Breggin also started warning ten years ago that antidepressants might be related to increased risks of suicide and aggression – something that has just recently been validated by an advisory panel to the US Food and Drug administration regarding antidepressants and children. Given the prescience of his warnings it seems to me that bringing a source like Dr Breggin to the attention of readers is a public service. Jeanne Lenzer Competing interests: author of article |
||||||||||||||||||||
|
|
||||||||||||||||||||
|
Stephen L. Black, Department of Psychology Bishop's University, Lennoxville, Quebec J1M 1Z7
Send response to journal:
|
Peter Mansfield is quoted by Jeanne Lenzer [1] that “six of the seven suicide attempts in TADS were made by adolescents treated with fluoxetine. Only one child not on fluoxetine attempted suicide. This wasn’t statistically significant but it may be clinically significant.” There are two problems with this statement. First, if he is referring to data analysis reported in the TADS study [2] itself, the authors did not say that this result was not statistically significant. Instead, they declined to carry out a statistical test on the grounds that the cases were “too small a number…for statistical comparison” (p. 815). Second, the TADS authors’ reluctance to test this result is presumably based on the traditional rule-of-thumb that the chi-square test not be used when the expected frequency in any cell is less than five. However, this rule was shown to be unsound by Camilli and Hopkins [3], who instead concluded that the “chi-square test was found to be very robust with small expected cell frequencies.” Accordingly, I carried out the chi- square test which the TADS authors declined to do. The data were the 6 suicides out of 216 patients on fluoxetine compared with 1/223 without. The outcome was chi-sq = 3.794, p = 0.051. This result exactly meets the conventional criterion (as normally rounded to two significant figures) of p = 0.05 for declaring a result statistically significant. I would be astonished if anyone quibbled over the additional 0.001 on such an important matter. Consequently, the data of the TADS study do show a statistically significant increase in attempted suicides by adolescent patients taking fluoxetine. This conclusion has particular force as it was the outcome of a study using random assignment of patients to groups. Thus it cannot be dismissed as the result of a pre-existing elevated risk of suicide in those receiving the drug. The conclusion of an increase in actual attempted suicides should be considered in opposition to the authors’ own conclusion, based on clinical evaluation and self-report, that the combination of fluoxetine and CBT “offered the most favourable tradeoff between benefit and risk”. 1. Lenzer, J. (2004). Journalists on Prozac. British Medical Journal, 329: 748. 2. Treatment for Adolescents with Depression Study (TADS) Team (2004). Fluoxetine, cognitive-behavioral therapy, and their combination for adolescents with depression. JAMA, 292: 807-820. 3. Camilli, G., & Hopkins, K. (1978). Applicability of chi-square to 2 x 2 contingency tables with small expected cell frequencies. Psychological Bulletin, 85, 163-167. Competing interests: None declared |
||||||||||||||||||||
|
|
||||||||||||||||||||
|
Peter R Mansfield, Director Healthy Skepticism Inc, 34 Methodist St, Willunga SA 5172 Australia
Send response to journal:
|
I thank Stephen Black and Nicholas DeMartinis for giving me the opportunity to expand on my contribution to Jeanne Lenzer’s excellent review of positive spin on the Treatment for Adolescents with Depression Study (TADS) trial. <1,2> I agree with Stephen Black that the TADS report should have included a statistical analysis of suicide attempts (placebo 0, fluoxetine alone 2, CBT alone 1, CBT with fluoxetine 4). My comment to Jeanne Lenzer that the difference wasn't statistically significant was based on my analysis using the Fisher’s exact test (6 in 216 vs 1 in 223, p=0.067). I choose that test because it is designed to cope with low numbers. However, I concede that it would have been more helpful to describe p=0.067 as "borderline significance" because many people think of significance testing in black vs white terms. After I was interviewed by Jeanne Lenzer the US FDA released an analysis of 26 trials of antidepressant drugs for children and teenagers that shows that overall they do increase suicidal thoughts and behaviour.<3> The TADS trial shows that fluoxetine is very unlikely to be an exception. The lack of statistical significance is likely to be a type II error (false negative) due to the trial being powered only to study efficacy, not adverse effects. Nicholas DeMartinis claims that recent reviews have ignored the possibility that the failure of trials to find evidence of efficacy arises from a Type II error. He does not specify which reviews he is criticising. Contrary to ignoring that possiblity our review <4> showed that the negative results are indeed a Type II error by finding with a meta- analysis that antidepressants do have a beneficial effect. However that effect is very small and likely to be smaller than the harmful effect. It is time to stop shooting the messengers and accept the message that the impression of worthwhile efficacy gained from clinical experience has been an illusion.<5> 1. Lenzer J. Journalists on Prozac. BMJ 2004; 329: 748 2. Treatment for Adolescents With Depression Study Team. Fluoxetine, Cognitive-Behavioral Therapy, and Their Combination for Adolescents With Depression: Treatment for Adolescents With Depression Study (TADS) Randomized Controlled Trial JAMA; 2004; 292:807-820 3. Hammad T. Results of the Analysis of Suicidality in Pediatric Trials of Newer Antidepressants. Presentation to Food and Drug Administration Pspychopharmacologic Drugs Advisory Committee and the Pediatric Advisory Committee September 13, 2004 http://www.fda.gov/ohrms/dockets/ac/04/slides/2004-4065S1_08_FDA- Hammad_files/frame.htm 4. Jureidini J, Doecke C, Mansfield P, Haby M, Menkes D, Tonkin A. Efficacy and safety of antidepressants for children and adolescents BMJ. 2004; 328:879-83 5. Mansfield PR, Jureidini JN, Tonkin AL. Antidepressants do less good than harm for children and adolescents. bmj.com, 8 Oct 2004 http://bmj.bmjjournals.com/cgi/eletters/329/7470/809#77368 Competing interests: None declared |
||||||||||||||||||||
|
|
||||||||||||||||||||
|
Susan F Anspach, writer and teacher home
Send response to journal:
|
The article, "Journalists on Prozac," was well-written and most welcome.
I find, however, one sentence that makes the whole study useless: "71% of teenagers treated with a combination of fluoxetine and cognitive behavior therapy improved compared with 35% with placebo alone"? Where's the study? What does that "alone" mean? This is so broad as to be completely useless. It's apples and oranges, to be completely hackneyed.
Testing with fluoxetine and therapy versus placebo and CBT would be a true test. Testing with fluoxetine and no CBT and placebo with no CBT would be a valid test.
One would be fair to conclude, even had this been a legitimate study, that cognitive behavior therapy doubled the rate of improvement, and the fluoxetine had absolutely nothing to do with it.
One might go further to say that CBT was the cause of the six suicides and placebos allowed for one suicide.
I, of course, know better than this, but that one word, "alone" is baffling. Did the Placebo children REALLY have NO talk therapy? If so, there's no study.
On a separate note:
The title "Journalists on Prozac" makes me think that we need full disclosure from the reporters as well as the subjects on which they're reporting.
Competing interests: None declared |
||||||||||||||||||||