1 Boring Old Man

Years ago I joked, "What they might as well do is make a table with all the DSM diagnoses down the side and all the psychiatric meds across the top and then just fill in the blanks." I was making fun of the endless clinical trials that were choking our journals. One of my more astute partners laughed as she said, "You’re sooooo far behind. They’ve already done that. Now they’re doing them two drugs at a time. Much bigger table!" But of all the combo drug strategies that came along, the one I least understood was adding atypical antipsychotics to antidepressants. That made some sense with severe cases of psychotic depression, but those are not common and most people being given the combination treatment didn’t seem to have those symptoms. What was the rationale for antipsychotics? Beats me. The drugs are anxiolytic, but there are plenty of much safer things to do to treat anxiety. So I’ve generally been suspicious that the atypical augmentation was a marketing ploy. Here, we are presented with a new meta-analysis of the studies to date that met solid study criteria. These are all industry funded studies with one being a joint industry/NIMH study:

Adjunctive Atypical Antipsychotic Treatment for Major Depressive Disorder: A Meta-Analysis of Depression, Quality of Life, and Safety Outcomes
PLoS Medicine
by Glen I. Spielmans, Margit I. Berman, Eftihia Linardatos, Nicholas Z. Rosenlicht, Angela Perry, and Alexander C. Tsai.
March 13, 2013
[full text on-line]

^{Background: Atypical antipsychotic medications are widely prescribed for the adjunctive treatment of depression, yet their total risk–benefit profile is not well understood. We thus conducted a systematic review of the efficacy and safety profiles of atypical antipsychotic medications used for the adjunctive treatment of depression.}

^{Methods and Findings: We included randomized trials comparing adjunctive antipsychotic medication to placebo for treatment-resistant depression in adults. Our literature search [conducted in December 2011 and updated on December 14, 2012] identified 14 short-term trials of aripiprazole, olanzapine/fluoxetine combination [OFC], quetiapine, and risperidone. When possible, we supplemented published literature with data from manufacturers’ clinical trial registries and US Food and Drug Administration New Drug Applications. Study duration ranged from 4 to 12 wk. All four drugs had statistically significant effects on remission, as follows: aripiprazole [OR], 2.01; 95% CI, 1.48–2.73], OFC [OR, 1.42; 95% CI, 1.01–2.0], quetiapine [OR, 1.79; 95% CI, 1.33–2.42], and risperidone [OR, 2.37; 95% CI, 1.31–4.30].}

^{The number needed to treat [NNT] was 19 for OFC and nine for each other drug. All drugs with the exception of OFC also had statistically significant effects on response rates, as follows: aripiprazole [OR, 2.07; 95% CI, 1.58–2.72; NNT, 7], OFC [OR, 1.30, 95% CI, 0.87–1.93], quetiapine [OR, 1.53, 95% CI, 1.17–2.0; NNT, 10], and risperidone [OR, 1.83, 95% CI, 1.16–2.88; NNT, 8]. All four drugs showed statistically significant effects on clinician-rated depression severity measures [Hedges’ g ranged from 0.26 to 0.48; mean difference of 2.69 points on the Montgomery–Asberg Depression Rating Scale across drugs]. On measures of functioning and quality of life, these medications produced either no benefit or a very small benefit, except for risperidone, which had a small-to-moderate effect on quality of life [g = 0.49].}

^{Treatment was linked to several adverse events, including akathisia [aripiprazole], sedation [quetiapine, OFC, and aripiprazole], abnormal metabolic laboratory results [quetiapine and OFC], and weight gain [all four drugs, especially OFC]. Shortcomings in study design and data reporting, as well as use of post hoc analyses, may have inflated the apparent benefits of treatment and reduced the apparent incidence of adverse events.}

^{Conclusions: Atypical antipsychotic medications for the adjunctive treatment of depression are efficacious in reducing observer-rated depressive symptoms, but clinicians should interpret these findings cautiously in light of [1] the small-to-moderate-sized benefits, [2] the lack of benefit with regards to quality of life or functional impairment, and [3] the abundant evidence of potential treatment-related harm.}

The whole point of a randomized placebo controlled clinical trial is to see if a medication has the desired clinical effect compared to placebo on some predefined outcome, with three usual measures of success: statistics, strength of effect, and overall improvement. The outcomes for the first two focus on some objective parameter [ie rating scale] and measure the percent achieving response or remission, or they measure the magnitude of difference in the actual outcome scores. The strength of the drug’s effect on the non-parametric [yes-no] parameters of response or remission are quantified by the Odds Ratio [OR] or the Number Needed to Treat [NNT]. The strength of the parametric [continuous scale] magnitude of difference is measured by the effect size. If you already know this stuff, this paragraph is boring and simplified. If you don’t know it, it may still be Greek, but I maybe it will help below.

• Statistical: Statistical significance is all you need to know if the outcome to not significance. In that case, anything you see in the numbers is just number noise to be ignored. In a clinical trial, if a difference is statistically significant, all you know is to look further. The differences observed may be meaningful or trivial, but the p value [probability] doesn’t discriminate the two. Something can be very significant but clinically irrelevant.
• Strength of Effect: In these studies, the investigator defines the change in the outcome measure that will define response and remission in advance. There are two ways to express the results – Odds Ratio and Number Needed to Treat, both calculated from the same things – the percentages of response or remission compared between placebo and drug. So if 5% respond to placebo and 25% respond to the drug, the Odds Ratio and Number Needed to Treat are:
  
  Obviously, the higher the OR, the better the response, and the lower the NNT the better the response. The NNT is the easiest to say in plain language: You need to treat 5 patients to get one subject who does better on the drug than placebo. The Effect Size is a bit more complicated and requires more information:
  
  The Effect Size varies with the differences in the means and the measurement of variability [overall variance in the sample]. Obviously, the most accurate computation comes from having the raw  data in hand, but there are any number of other ways and formulas to calculate it indirectly, depending on what information you can get. Here’s what the authors said about their computations:

  ^{For continuous outcomes, effect sizes were computed from means and standard deviations when possible. When these were not provided, effect sizes were computed based on means and p-values, or p-values only.}

  … meaning that they did the best they could with what they had to work with.

  In this meta-analysis, the focus was on four atypical antipsychotics and their efficacy as adjuncts in depressed patients who had not responded to antidepressants [treatment-resistant depression]. Here’s a greatly truncated version of the forest plots of the Odds Ratios for Response and Remission with the four drugs Aripiprazole [Abilify], Olanzapine/Fluoxetine Combo [Zyprexa + Prozac], Quetiapine [Seroquel], and Risperidone [Risperdal] just to show the variability within and among the studies:

  
  ^{[Odds Ratios with 95% Confidence Intervals]}

  Now here are the OR, NNT, and Effect Size composite results [shown without confidence limits]:

  

  So what do the numbers mean? There’s the rub. What’s a good Odds Ratio? a strong NNT? a robust Effect Size? Click on the links to discover that even Wikipedia won’t directly answer the question. But there are a couple of other things to mention before returning to interpretation.
• Overall Improvement: The Strength of Effect measures are based on rating scales with specific items related to depression scored by external raters [HAM-D, MADRS, etc]. There are other self rating scales. Here’s what the authors said:

  ^{Continuous measures of quality of life included the Quality of Life Enjoyment and Satisfaction Questionnaire [Q-LES-Q] and the Short Form 36 Health Survey [SF-36]. The only continuous measure of functional impairment employed in these trials was the Sheehan Disability Scale [SDS]. As measures of quality of life and functional impairment varied across studies, we pooled such measures together to create an omnibus effect size for each drug, and across all drugs.}

  The results were:
  
  but they hastened to add:

  ^{With regards to quality of life and functioning, adjunctive quetiapine, aripiprazole, and OFC produced effect sizes that were either not statistically significant or small and clinically negligible in magnitude. Adjunctive risperidone was more efficacious than adjunctive placebo on quality of life/functioning, with a small-to-moderate effect size. The pooled effect across quality of life/functioning measures varied significantly across treatments [QB = 6.88, p = 0.003], with risperidone [g = 0.49] yielding a higher effect than the other three drugs combined [g = 0.11], which did not differ significantly from each other [QB = 4.02, p = 0.13]. However, the effect of aripiprazole on quality of life/functioning was small [g = 0.22] and statistically significant [p = 0.001], whereas the effects of OFC [g = 0.04, p = 0.74], and quetiapine [g = 0.05, p = 0.53] were both not statistically significant and of quite small magnitude. The effect of aripiprazole on quality of life/functioning should be interpreted with caution, as the effect for the drug on the SDS was very small and no longer statistically significant when patients who violated study protocol were excluded from analysis [g = 0.12, p = 0.08]. Similarly, the effect of risperidone on quality of life/functioning should be interpreted tentatively since it is largely driven by post hoc analyses.}

  These authors concluded that from the subjects point of view, their lives weren’t particularly improved by atypical antipschotic augmentation [which, by the way, is the point of treatment in the first place].

• if we had the raw data as these studies were published, they could’ve been checked  in a timely fashion.
• and if we had the raw data as the studies were published, this meta-analysis could’ve been done without all the detective work it took to excavate the facts.
• and if we’d had the raw data when it should’ve been available, a huge cohort of people wouldn’t have taken unnecessary and expensive medications that made some of them sick and didn’t help very many.
• and if we’d had the raw data up front, the drug companies and the KOLs wouldn’t have made so much ill-gotten profit and there would be fewer Mercedes in the parking lots.
• and if we had the raw data from the start, the medical literature could’ve stayed scientific like it was  supposed to be.
• and…