It has been argued recently that the drug-placebo difference is diminishing in clinical trials of depression. Recently a paper suggested that the placebo response in trials of antidepressant agents increased abruptly after 1991 and remained stable afterwards at the level of 35–40%.1 A possible explanation for this change could be the introduction of Diagnostic and Statistical Manual Third Edition (DSM)-III and DSM-III-Revised (R) and the change in paradigm concerning depression, which came with them since the early 1980s.2 3 With the introduction of operationalised diagnostic criteria, the concept of depression both changed and broadened. Neither the Hamilton Rating Scale for Depression (HAM-D) nor the Montgomry-Asberg Depression Rating Scale (MADRS) correspond to this new concept since they were both developed or last updated before 1980 and overall the properties of these scales are problematic.4
We used the Kirsch database which is publically available through previous publications to investigate whether there was an abrupt change in the placebo and also in the medication response rate.5 6
The plotting of standardised mean change from baseline, separately for placebo and for drug arm as well as for the standardised mean difference between the two arms (Cohen's D) with the distant weighted least squares regression method is shown in figure 1.
Figure 1.
Distant-weighted least squares regression lines of Cohen's D concerning the active drug arm, the placebo arm and their difference. Note that this graph is only to give a visual impression of the data since such an analysis violates the within studies randomisation.
Of course such a separation of placebo and drug arms violates the within studies randomisation, but this is not uncommon in many such analyses and it is inevitable if one wishes to see the isolated effect on either intervention arm. However, the plotting of the difference between arms respects this within studies randomisation.
The plots in figure 1 confirms the presence of an abrupt change in the years 1988–1990 as previously reported for the placebo arm,1 but a similar increase was also present for the drug arm. There was a slight decrease in the difference between drug and placebo arms until 1988–1990 but the difference started increasing afterwards. The two intervention arms manifested parallel lines with an increase for both after 1989, but after 1995 the drug response rate stabilised while the placebo response is decreasing.
The above observations are based on a limited number of studies, but the overall database is considered to be one of the most reliable databases available. They confirm a change around the years 1988–1990 just after the introduction of DSM-III-R and suggest that the effect was similar for both arms. They do not support a diminishing placebo-active drug difference, while, on the contrary there is some support for an increasing difference.
It is possible that while the trials before 1990 were based in more ‘endogenous’ and ‘severely’ depressive patients, after 1990 they used the DSM criteria and recruited a more broadly defined population with milder forms of depression, a fact which was made worse with the domination of multicentred studies which included personnel with low research experience and training and also low interest in research per se. Such a difference in the quality of patients is highly unlikely to be caught by either the 17-item HAM-D or the MADRS.2
These milder forms of depression could be inherently more responsive to placebo, but it is unknown whether they are also more responsive to antidepressants or the increase is essentially due to an increased response to add-on benzodiazepines, which are allowed in almost all trials for both arms up to a certain dosage, as an open-label add-on according to the opinion of the therapist.
Thus a possible explanation of the increase in the placebo and the active drug response is the change in the quality of study samples because of the change in the definition of depression and the resulting high response to benzodiazepines.
Footnotes
Competing interests: KNF has received grants and served as consultant, advisor or CME speaker for the following entities: AstraZeneca, Bristol-Myers Squibb, Eli Lilly, Ferrer, Gedeon Richter, Janssen, Lundbeck, Otsuka, Pfizer, the Pfizer Foundation, Sanofi-Aventis, Servier, Shire and others.
Provenance and peer review: Not commissioned; internally peer reviewed.
References
- 1.Furukawa TA, Cipriani A, Atkinson LZ, et al. Placebo response rates in antidepressant trials: a systematic review of published and unpublished double-blind randomised controlled studies. Lancet Psychiatry 2016;3:1059–66. 10.1016/S2215-0366(16)30307-8 [DOI] [PubMed] [Google Scholar]
- 2.Fountoulakis KN. DSM-III-R change in definition might have affected placebo response to antidepressants. Lancet Psychiatry 2017;4:21–2. 10.1016/S2215-0366(16)30372-8 [DOI] [PubMed] [Google Scholar]
- 3.Cipriani A, Furukawa TA, Atkinson LZ, et al. DSM-III-R change in definition might have affected placebo response to antidepressants—authors’ reply. Lancet Psychiatry 2017;4:22–3. 10.1016/S2215-0366(16)30415-1 [DOI] [PubMed] [Google Scholar]
- 4.Fountoulakis KN, Samara MT, Siamouli M. Burning issues in the meta-analysis of pharmaceutical trials for depression. J Psychopharmacol (Oxford) 2014;28:106–17. 10.1177/0269881113504014 [DOI] [PubMed] [Google Scholar]
- 5.Huedo-Medina TB, Johnson BT, Kirsch I. Kirsch, et al.'s (2008) calculations are correct: reconsidering Fountoulakis & Moller's re-analysis of the Kirsch data. Int J Neuropsychopharmacol 2012;15:1193–8. 10.1017/S1461145711001878 [DOI] [PubMed] [Google Scholar]
- 6.Fountoulakis KN, Veroniki AA, Siamouli M, et al. No role for initial severity on the efficacy of antidepressants: results of a multi-meta-analysis. Ann Gen Psychiatry 2013;12:26. 10.1186/1744-859X-12-26 [DOI] [PMC free article] [PubMed] [Google Scholar]