Item-based analysis of the effects of duloxetine in depression: a patient-level post hoc study

Abstract

Oft-cited trial-level meta-analyses casting doubt on the usefulness of antidepressants have been based on re-analyses of to what extent the active drug has outperformed placebo in reducing the sum score of the Hamilton Depression Rating Scale (HDRS-17-sum) in clinical trials. Recent studies, however, suggest patient-level analyses of individual HDRS items to be more informative when assessing the efficacy of an antidepressant. To shed further light on both symptom-reducing and symptom-aggravating effects of a serotonin and noradrenaline reuptake inhibitor, duloxetine, when used for major depression in adults, we hence applied this approach to re-analyse data from 13 placebo-controlled trials. In addition, using patient-level data from 28 placebo-controlled trials of selective serotonin reuptake inhibitors (SSRIs), the response profile of duloxetine was compared to that of these drugs. Duloxetine induced a robust reduction in depressed mood that was not dependent on baseline severity and not caused by side-effects breaking the blind. A beneficial effect on depressed mood was at hand already after one week; when outcome was assessed using HDRS-17-sum as effect parameter, this early response was however masked by a concomitant deterioration with respect to adverse event-related items. No support for a suicide-provoking effect of duloxetine was obtained. The response profile of duloxetine was strikingly similar to that of the SSRIs. We conclude that the use of HDRS-17-sum as effect parameter underestimates the true efficacy and masks an early effect of duloxetine on core symptoms of depression. No support for major differences between duloxetine and SSRIs in clinical profile were obtained.

Fig. 1

a–c Week-by-week mean change for HDRS-17-sum (a), the depressed mood item (b) and the psychic anxiety item (c) for duloxetine- and placebo-treated subjects, respectively. Lines represent estimated means from a linear mixed model. Effect sizes at week 1, 2, 4, 6 and 8, respectively: a 0.05, 0.20, 0.27, 0.37, 0.37; b 0.20, 0.31, 0.37, 0.46, 0.44 c; 0.23, 0.29, 0.34, 0.42, 0.43. n.s. non-significant (p = 0.16); *** = p < 0.001

Fig. 2

a, b Week-by-week mean change for the suicidality item for duloxetine and placebo-treated subjects, respectively, in subjects aged 25 years and above (a) subjects aged 18–24 years (b). Lines represent estimated means from a linear mixed model. Baseline scores a 0.62 points, standard deviation (SD) 0.76; b 0.53 points, SD 0.70. Effect sizes at week 1, 2, 4, 6 and 8, respectively: a 0.15, 0.23, 0.26, 0.23, 0.24; b 0.22, 0.22, 0.13, 0.25, 0.20. n.s. non-significant (p = 0.10, 0.11, 0.42, 0.16, 0.21); ***p < 0.001

Fig. 3

a, b Estimated endpoint means and effect sizes for the depressed mood item in placebo-treated subjects and in duloxetine-treated subjects with or without adverse events a only counting initial adverse events (reported during week 1–2 and b counting any adverse event throughout the trial. Placebo: n = 998; duloxetine without early adverse events: n = 713; duloxetine with early adverse events: n = 1 029; duloxetine without any adverse event throughout the trial: n = 449; duloxetine with any adverse event throughout the trial: n = 1 293. Effect sizes: a duloxetine with adverse events vs placebo: 0.46; duloxetine without adverse events vs placebo: 0.34, duloxetine with adverse events vs duloxetine without adverse events: 0.12 b duloxetine with adverse events vs placebo: 0.44: duloxetine without adverse events vs placebo: 0.36; duloxetine with adverse events vs duloxetine without adverse events: 0.12. *p = 0.02 (a), p = 0.04 (b); ***p < 0.001

Fig. 4

Relative effect sizes versus placebo for duloxetine and four SSRIs (grouped) using the sum of all other HDRS-17 items as covariate. Placebo (duloxetine studies): n = 1 345; duloxetine: n = 2 230; placebo (SSRI studies): n = 2 581; SSRIs: n = 5 681 (citalopram n = 744; fluoxetine n = 754; paroxetine n = 2 981; sertraline n = 1 202). HDRS-17 items and p-values for the contrast duloxetine vs SSRIs: 1 = depressed mood (p = 0.63), 2 = feelings of guilt (p = 0.35), 3 = suicidality (p = 0.89), 4 = early insomnia (p = 0.59), 5 = middle insomnia (p = 0.59), 6 = late insomnia (p = 0.24), 7 = work & activities (p = 0.82), 8 = psychomotor retardation (p = 0.49), 9 = psychomotor agitation (p = 0.26), 10 = psychic anxiety (p = 0.65), 11 = somatic anxiety (p = 0.93), 12 = gastrointestinal symptoms (p = 0.23), 13 = general somatic symptoms (p = 0.39), 14 = sexual dysfunction (p = 0.54), 15 = hypochondriasis (p = 0.45), 16 = weight change (p = 0.80) and 17 = insight (p = 0.78)