Emotional blunting by mechanism of second-generation antidepressants: comparing bupropion and serotonin reuptake inhibitors

Missing data will be carried forward using LOCF analyzed on an ITT basis as in the original trials, for all three data-sets.Question 1: Is emotional blunting experienced more commonly with bupropion and/or venlafaxine compared to placebo?With pooled data from WXL101497 and AK130939, we will use t-tests (or Wilcoxon tests if data does not appear normally distributed after plotting) to compare MADRS Item 8 change scores across: bupropion vs. placebo, and venlafaxine vs. placebo.We will also use proportion z tests to compare the % with worsening Item 8 scores, with two comparisons: bupropion vs. placebo, and venlafaxine vs. placebo.Question 2: Do bupropion and serotonergic antidepressants differ in their propensity to cause emotional blunting?Here we will also use the same tests to make two comparisons: bupropion vs. venlafaxine (pooled data from WXL101497 and AK130939) and also bupropion vs. escitalopram (data from LOC114589).Power analysis for Question 1 and 2:For Q1, with a two-tailed test, assuming alpha .05 and 90% power, we could detect small effect sizes of d = .28 or higher with n = 265 per group (.23 if n = 385 per group) for tests of change scores.For Q1 and Q2, assuming alpha .05 and 90% power, and n = 265 per group, we could detect proportion differences of 9-12% depending on the baseline proportion in the sample.A proportion difference of 10%, or a d of .23-.28, would be on the lower limit of “clinical significance” as per convention in antidepressant research, so this is considered adequate.Question 3. Is emotional blunting associated with depression improvement, suicidality, or sexual dysfunction? AND Do these associations also exist in placebo groups?This analysis will only use pooled data from WXL101497 and AK130939. The analysis here is somewhat exploratory, and there are no pre-specified hypotheses per se. Rather, the goal is simply to see if any associations exist, and to examine the shape of the distributions etc. We will create scatter plots with MADRS Item 8 Change Scores on the X-axis and the following variables on the Y-axis: MADRS total score (excluding item 8) % reduction, MADRS item 2 (reported sadness) change scores, MADRS item 10 (suicidality) change scores, CGI change scores, and CSFQ (sexual functioning) change scores. These will be done separately by treatment condition. If there are obvious linear or non-linear monotonic relationships, we can calculate r or Tau-b, respectively, and test for statistical significance. We cannot select tests of association in advance because we do not have any past research to tell us what type/shape of association/distribution we should expect to see.The rationale for conducting the analysis for question 3 is:1. An association between blunting and sexual dysfunction is predicted by past research and would help validate our single item measure. If this association is also present, and of a comparable size, in the placebo group, then this would tell us that it is not due to medication per se, and that would be a useful finding in and of itself.2. Finding positive correlations between emotional blunting and improvement (i.e., as blunting gets worse, depression/mood scores decrease) is expected by the theory that medication-induced blunting contributes to treatment effects observed in clinical trials. Alternatively, finding a negative correlation (i.e., as blunting gets worse, depression scores increase/worsen) would support the opposite notion, that blunting is an adverse effect related to treatment non-response, and therefore, does not account for treatment effects reported in clinical trials.However, both of these interpretations would only make sense if the associations are only present in the two medication groups. If they are also present, and of a similar size, in the placebo group, then it does not make sense to discuss these associations as being "drug-induced".Limitations1) The use of a non-validated single-item measure of emotional blunting. 2) The analysis for Question 3 is exploratory, although this is felt to be appropriate at this time because research in this area is scarce.3) We cannot make inferences about long-term treatment, although some existing studies have failed to find a correlation between blunting and longer treatment duration (1, 6).Strengths1) Potential to establish convergent validity of a single-item measure of emotional blunting that is already used in numerous clinical trials and permits further exploration with archived data.2) Ability to make causal inferences given experimental design and placebo control3) Ability to explicitly test the claim that emotional blunting accounts for the treatment effect reported in antidepressant clinical trials