more clearly III…

Posted on Monday 12 March 2012

I ended my last post by saying I’d "go rest a while" – but it wasn’t as restful as I’d planned. Ever since the second Gibbons article came out [Benefits From Antidepressants], I’ve had a nagging doubt about the outcome, particularly in children. The reported response rates and remission rates just seem too good to me compared to the studies he used for his analysis. At a time when the efficacy of SSRIs is being heavily criticized, he says:
The results do not support previous findings that antidepressants show little benefit except for severe depression. The antidepressants fluoxetine and venlafaxine are efficacious for major depressive disorder in all age groups, although more so in youths and adults compared with geriatric patients. Baseline severity was not significantly related to degree of treatment advantage over placebo.
As mentioned last time, they didn’t even look at the venlafaxine studies in kids [more clearly II…], so that "fluoxetine and venlafaxine are efficacious for major depressive disorder in all age groups" comment is out of line, but even the fluoxetine results seem overly optimistic to me.

Obviously, I can’t reproduce their calculations since their datasets are not available and their methodology is dense and beyond my knowledge or resources, but what I could do is look at the ways the original studies reported their results. There are many ways people regularly report outcomes: the mean changes in a variable; the % responding based on some criteria; the % remitting based on some criteria; the Number Needed to Treat, a number derived from comparing placebo and treatment responses. NNT means, "How many patients would I have to treat to beat the placebo results?" Obviously, the lower the NNT, the greater the efficacy. Looking at Gibbons’ study, the change in CDRS-R score was not computed but rather expressed as complex regression trajectories requiring full data [as they say, "beyond the scope…"]. But he did report % responding, % remitting, and the NNT values in the more usual ways:

    Response = a 50% reduction in CDRS-R
    Remission = a CDRS-R < 28
    Number Needed to Treat [NNT] = 1÷[PtreatmentPplacebo] where P is proportion meeting criteria

Looking at the studies, LYAQ was immaterial since what I was looking at was efficacy [no significant efficacy found]. The other three [X065, HCJE, and TADS] all had remission rates using the exact same criteria used by Gibbons. Unfortunately, X065 and TADS used different response criteria [CGI improvement of 1 or 2] which is another test and untranslatable to Gibbons’ study. Study HCJE did use the CDRS-R for response criteria like Gibbons, but they set the bar at 30% reduction. Fortunately, by a quirk they also had a table that gave the response rate for 50% reduction which fit Gibbons criteria. So there were some direct comparisons, and I could calculate the NNT values.

The reported outcomes were for the full duration of the study, whereas Gibbons cut off his data at 6 weeks [I’m still unsure quite why], but I reasoned that if his truncating the data effected the outcome, it would decrease efficacy if anything – and I was looking for an inflation. So I thought running down the numbers might be instructive [the not restful part]. So here, for reminders, are the four clinical trials in his analysis, shown graphically as before with the data Gibbons included in bold and the omitted data shaded. LYAQ [24%] would decrease the overall efficacy. TADS, HCJE, and X065 would  increase the efficacy, proportional to to the number of subjects [31%, 31%, and 14% respectively]. The sources for the data follow the graphs:


This was a small study that used the CGI to determine response, but did give the remission values using the CDSR-R like Gibbons:
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-squared=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.


In the published version of the HCEJ study, they reported their outcome with this table:
Their predefined outcome parameters included:
    Response = a 30% reduction in CDRS-R
    Remission = a CDRS-R < 28
They met the Remission Criteria [same criteria as Gibbons]:
Significantly more fluoxetine treated patients [41.3%] than placebo-treated patients [19.8%] met the prospectively defined criteria for remission [p < .01].
They didn’t meet their Response Criteria [30% reduction in CDRS-R] but they were close and included this table to show us that if only they’d chosen another percentage, they would have made the grade:
I don’t know if that’s kosher or not. But that’s not why I’m showing the table. It’s because this table allows us to see their response rates at a 50% reduction in CDRS-R [outlined] which happens to be the criteria that Dr. Gibbons et al used for their outcome parameter, so we can compare the outcomes.


The TADS study used the CGI for response, but did include the remission figures in the same format as Gibbons:
The rate of remitted subjects at the end of treatment, based on the dichotomized CDRS-R, was 102 of 439, or 23%. There was a significant overall treatment [Wald x2= 17.08, df = 3, p = .0007] and site [Wald x2 = 21.5, df =12, p = .04] effect. More specifically, the COMB group with a 37% remission rate was superior to all of the other treatment conditions [FLX: 23%, p = .02; CBT: 16%, p = .0004; PBO: 17%, p = .0009]. However, the FLX, CBT, and PBO rates were not statistically different from one another [p 9 .05]….

The overall rate of response, defined as a CGI-I score of 1 or 2 in the intention-to-treat sample was 229 of 439, or 52% [TADS, 2004]. As previously reported, response rates by treatment arm were as follows: 71.0% [COMB], 60.6% [FLX], 43.2% [CBT], and 34.8% [PBO].


As mentioned above, this study was kind of a ringer, but fluoxetine had no response or remission different from placebo, so it’s contribution to overall efficacy would be to diminish it. The Response, Remission, and NNT are non-applicable [eg ∞].


Now, here are the results reported by Dr. Gibbons in his efficacy meta-analysis:
Youth Studies of Fluoxetine
In youth studies of fluoxetine, the estimated average rates of change over 6 weeks were −15.96 CDRS-R units for placebo and −20.58 CDRS-R units for fluoxetine [MMLE=−4.62; SE=1.26; P<.001], indicating 29.0% greater improvement for fluoxetine. The estimated response rates were 29.8% for fluoxetine vs 5.7% for placebo [OR=6.66; 95% CI, 3.07- 14.48; P<.001; NNT=4.16]. Remission rates were 46.6% vs 16.5% for fluoxetine and placebo, respectively [OR=4.23; 95% CI, 2.64-6.77; P<.001;NNT=3.33]. The finding of higher remission rates than response rates questions the validity of the CDRS-R remission threshold score of 28. No effect of baseline severity on treatment efficacy was found [P=.90]. For patients with low severity, the rates of change in symptoms were −17.60 CDRS-R units for fluoxetine vs −12.56 CDRS-R units for placebo. For patients with high severity, the rates of change were −28.86 CDRS-R units for fluoxetine vs −24.40 CDRS-R units for placebo. The estimated differences were 5.04 CDRS-R units [95% CI, 2.56 to 7.52] for low severity and 4.45 CDRS-R units [95% CI, −0.58 to 9.49] for high severity. Estimated response rates for treated vs placebo receiving patients were 23.0% vs 3.2% [difference of 19.8%], respectively, for low severity and 40.2% vs 17.2% [difference of 23.0%], respectively, for high severity. Estimated remission rates for treated vs placebo-receiving patients were 54.1% vs 19.4% [difference of 34.7%], respectively, for low severity and 28.9% vs 7.5% [difference of 21.4%], respectively, for high severity.

The outcome criteria for Gibbons et al are:

    Response = a 50% reduction in CDRS-R
    Remission = a CDRS-R < 28

Using the values from above, we can construct the following table comparing the reported results between the pertinent input studies [X065, HCEJ, and TADS] and Gibbons’ outcome for the meta-analysis of all four studies. Recall that the lower the Number Needed to Treat, the more efficacious the treatment [NNT = 1 ÷ diff]. And it all comes down to the red numbers in the table below – the outcome variables [NNT] from the original studies that fit the same criteria used by Gibbons:

    X065   HCEJ   TADS   GIBBONS
PARAMETER Pbo Flx diff Pbo Flx diff Pbo Flx diff Pbo Flx diff

Response CGI 33% 56% 23% 34.8% 60.6% 25.8%
NNT 4.35     3.88    
Response CDRS-R 16.8% 33.9% 17.1% 5.7% 29.8% 24.1%
NNT   5.85     4.15  
Remission Rate
22.9% 31.3% 8.4% 19.8% 41.3% 21.5% 17% 23% 6% 16.5% 46.6% 30.1%
NNT 11.9   4.65   16.67   3.33  

As I said, there was no way for me to adjust the number to make up for the fact that Gibbons didn’t use all the data, but I still think if anything it would decrease efficacy. For the numbers I could compute, Gibbon’s outcome was better that any of the studies he used for his meta-analysis. And that’s before the decrease from the negative study LYAQ and whatever efficacy he lost by cutting off the end of the others. I find that  remarkable  impossible. How can the meta-analysis be better than any of the pieces used to derive it? better for remission in all cases? better in the one case where we can make direct response rate comparisons? better absent the effect of a large negative study with 24% of the overall subject pool?

It’s the closest I could come to finding a way to check Gibbons’ work, given the available information. It’s hardly elegant or precise, but it’s enough for me to be genuinely skeptical of this study’s accuracy, and suggests that the Archives of General Psychiatry would be well advised to have someone neutral go over the data and the calculations used in this study before committing it to our medical literature in perpetuity. I think there’s something very wrong with these numbers…
    March 12, 2012 | 10:38 PM

    Gibbons study is all about sending out a marketing message & grabbing press headlines….the numbers don’t need to add up…it’s the same old statistical slight of hand that has been used now for decades getting these drugs approved and into the mega sales market place…you could prove that this study or analysis was completely bogus; and it might get a quick paragraph in the back of some Wednesday morning addition of the Timbuktu newspaper..sadly, just how it works…

    on a brighter note…have you seen this article in the Washington Post..

    March 12, 2012 | 10:44 PM

    Yuk! Good for Adriane…

    As for Gibbons – I take your point, but hope springs eternal, because the opposite it to just watch it happen like people did for twenty five years, and that sure didn’t help…

    March 13, 2012 | 9:03 AM


    From: Journal of Child and Adolescent Psychopharmacology
    Sent: Wednesday, March 07, 2012
    Subject: More Effective Treatments Urgently Needed for Adolescent Depression


    A state-of-the-art issue reporting on the latest research findings on antidepressant medications combined with appropriate therapeutic strategies has been published by
    Journal of Child and Adolescent Psychopharmacology
    Special Issue on Psychopharmacology of Adolescent Depression

    Featuring some of our favourite KOLs: Wagner, Keller, Emslie, Brent, Birmaher…

    This (unusually) free issue, guest edited by G.J. Emslie, MD, lead author in Gibbons’ Lilly Fluoxetine trials, in close collaboration with Gibbons “safe” and “effective” deceipts (aided and abetted by Arch Gen Psych) appears directly intended to “neutralise” the Black Box warning, so as to persuade clinicians and parents/patients that antidepressants are safe and effective for adolescents, increasing prescription and consumption, hence rendering the warning “sales neutral”, ahead of Pristiq and Cymbalta’s FDA indication push. This campaign is likely also intending to persuade those in the FDA making this decision to approve the indication.

    This (unusually) free issue really pushes the case to child psychiatrists. In this issue Guest Editor: G.J. Emslie, MD states “Barriers to effective treatments include public perceptions of effectiveness and safety of treatments” – and they’re working to overcome these unhelpful-to-marketing-perceptions.

    Dr. Emslie receives research grant support from Biobehavioral Diagnostics Inc., Eli Lilly, Forest Laboratories, GlaxoSmithKline, and Somerset; has been a consultant for Biobehavioral Diagnostics Inc., Eli Lilly, Forest Laboratories, INC Research Inc., Lundbeck, Pfizer, Seaside Therapeutics, Shire Pharmaceuticals, and Wyeth; and has been on the Speakers Bureau for Forest Laboratories.

    See video of Emslie, at the same table with Gibbons, laying out the campaign here:
    A Multidisciplinary Approach to Treating Major Depressive Disorder in the Adolescent Patient (Slides With Video)- Faculty and Disclosures Graham J Emslie, MD Disclosure: has received … Vanya J Hamrin, RN, MSN, … husband’s employer is Moore Medical. Robert D Gibbons, an expert witness for Pfizer, Wyeth Pharmaceuticals, and the US Dept of Justice. This appalling pharmaceutical advertisement was “supported by an independent educational grant from Forest Laboratories.” (btw should Gibbons be noting this in his COI?)

    The issue is full of articles pushing the case, including: “An Open-Label Safety and Pharmacokinetics Study of Duloxetine in Pediatric Patients with Major Depression” in which John S March, key NIMH advisor once again lies about COI “Dr. March has not engaged in promotional work, e.g., speakers bureau or training, for over 15 years.” cf prior 2006/7 disclosures of such work.

    In the press release: “There are no radically new treatments on the horizon for the treatment of depression, and so we have to do better with the treatments we have available,” says Graham J. Emslie, MD, Guest Editor of the issue and Director of Child Psychiatry at University of Texas Southwestern Medical Center, Dallas. “Few youths with depression receive adequate treatment.”

    This whole deceptive campaign, which WILL harm our children, deserves the strongest response of informing and warning about this disinformation, on all platforms, journals and media, that responsible clinicians can possibly muster.

    Reuters, Los Angeles Times etc don’t routinely scan Arch Gen Psych contents alerts, I’m sure. There ARE press releases responsible for the publicity. Who is sending them out? Can journalists share this information?

    Question – Aside from letters, blogs etc on this issue, is there opportunity for public submissions to the FDA regarding this decision?

Sorry, the comment form is closed at this time.