1 Boring Old Man

That was way, way too complicated. The short version is, "I found a better example." I’m starting over.

A Double-blind, Randomized, Placebo-Controlled Trial of Fluoxetine in Children and Adolescents With Depression
^{by Graham J. Emslie, MD; A. John Rush, MD; Warren A. Weinberg, MD; Robert A. Kowatch, MD; Carroll W. Hughes, PhD; Tom Carmody, PhD; and Jeanne Rintelmann}
Archives of General Psychiatry. 1997, 54:1031-1037.

^{Background: Depression is a major cause of morbidity and mortality in children and adolescents. To date, randomized, controlled, double-blind trials of antidepressants (largely tricyclic agents) have yet to reveal that any antidepressant is more effective than placebo. This article is of a randomized, double-blind, placebo controlled trial of fluoxetine in children and adolescents with depression.
Method: Ninety-six child and adolescent outpatients (aged 7-17 years) with nonpsychotic major depressive disorder were randomized (stratified for age and sex) to 20 mg of fluoxetine or placebo and seen weekly for 8 consecutive weeks. Randomization was preceded by 3 evaluation visits that included structured diagnostic interviews during 2 weeks, followed 1 week later by a 1-week, single-blind placebo run-in. Primary outcome measurements were the global improvement of the Clinical Global Impressions scale and the Children’s Depression Rating Scale – Revised, a measure of the severity depressive symptoms.
Results: Of the 96 patients, 48 were randomized to fluoxetine treatment and 48 to placebo. Using the intent to treat sample, 27 (56%) of those receiving fluoxetine and 16 (33%) receiving placebo were rated "much" or "very much" improved on the Clinical Global Impressions scale at study exit (chi 2=5.1, df=1, P=.02). Significant differences were also noted in weekly ratings of the Children’s Depression Rating Scale – Revised after 5 weeks of treatment (using last observation carried forward). Equivalent response rates were found for patients aged 12 years and younger (n=48) and those aged 13 years and older (n=48). However, complete symptom remission (Children’s Depression Rating Scale – Revised) occurred in only 31% of the fluoxetine-treated patients and 23% of the placebo patients.
Conclusion: Fluoxetine was superior to placebo in the acute phase treatment of major depressive disorder in child and adolescent outpatients with severe, persistent depression. Complete remission of symptoms was rare.}

The other primary outcome measure, the weekly CDRS-R score, was examined as a continuous variable. Figure 2 shows the traditional method of dealing with patient attrition during the course of the treatment [ie, weekly CDRS-R scores for each group are presented with the LOCF]. The last available observation is filled in for the values for patients who discontinued study participation before 8 weeks… Comparing the weekly CDRS-R scores for each treatment group using t tests, the first week that the groups were significantly different was week 5. At week 5, the mean CDRS-R score for the fluoxetine-treated group [39.8±13.2] was lower than the placebo group [46.8±16.6] [t=-2.28, df=94, P=.03]. To make the most efficient use of the available data without resorting to a completion analysis or LOCF analysis, the rate of change [slope] and baseline CDRS-R score [intercept]) were estimated from linear regressions on each patient and for each group. The estimated baselines were similar [54.2 for the fluoxetine-treated group vs 53.8 for the placebo group]. However, the fluoxetine-treated group slope of -2.75±-2.52 was significantly different from the placebo group slope of -1.27±-2.86 [t=2.68, df=94, P<.001].

These results may be interpreted as follows. The fluoxetine-treated group began with an average CDRS-R score of 54.2 and their scores improved by 2.75 U per week to end with an estimated week 8 score of 32.2. While placebo patients began with a similar average CDRS-R score [53.8], their score improved only 1.27 U per week to the end of the study with an estimated exit score of 43.6. Using the empirical Bayesian analysis of estimating slopes [intercept estimates were unavailable] gives an estimated slope of -2.60±2.36 for the fluoxetine treated group and a significantly smaller slope of -1.32±3.56 for the placebo group [t=2.08, df=94,P=.04].

Study X065 was the first Clinical Trial of an SSRI [Prozac] in Adolescents. It was one of the two studies used by Lilly to get approval for Prozac in the treatment of MDD in Adolescents [the only SSRI approved for kids since it got through the FDA before the black box warning was imposed]. To my knowledge, the only sight we have of the primary raw data from this study is the top graph in Figure X065.2 above. The published graph [LOCF] and the % change lower graph are ways to make things look better on paper. And the verbiage in the article [quoted above] is an indirect way of claiming a difference using regression slopes, in lieu of showing us that simple raw data graph [which clearly shows no difference].

So that’s why I want the raw data. If I had seen that top X065.2 graph, I would’ve simply said, Prozac doesn’t help depressed kids, no matter what the manipulated versions showed. So would you. That’s why they didn’t show it. There’s a second point. Gibbons et al used this data in his meta-analysis that said the SSRIs were efficacious in adolescents. All I see is even more evidence that his meta-analysis is flawed…

^{[what they saw]                                                                                        [what we saw]}