From the Editor

Is the placebo effect getting stronger with time?

image001Placebo pills – greater importance?

Since the first writings of English physician John Haygarth at the turn of the 19th century, the placebo effect is something well documented and well discussed – but not well understood. How can people respond to sugar pills and the like?

Even more oddly, the placebo effect seems to be changing with time, at least in terms of antidepressant medications. In a classic paper published about a decade and a half ago, Walsh et al. found that the placebo effect was getting more pronounced over the years. Dr. Timothy Walsh joked in a Washington Post interview that “[t]hey’re making placebos better and better.” Besides being an interesting finding, there are larger issues – start with the implications to drug development. After all, if the placebo effect is rising, it becomes more challenging to develop a drug that bests it.

Have things changed since the publication of the Walsh et al. paper?

In this week’s Reading, we consider the new Furukawa et al. paper. This study, which reviewed 250 plus randomized controlled trials that involved more than 26,000 patients and included unpublished data, found that the placebo effect isn’t increasing. Also in the Reading: an editorial commenting on the Furukawa et al. paper.

Please note – there will be no Reading next week.


Placebo and Antidepressants

“Placebo response rates in antidepressant trials: a systematic review of published and unpublished double-blind randomised controlled studies”

Toshi A. Furukawa, Andrea Cipriani, Lauren Z. Atkinson, Stefan Leucht, Yusuke Ogawa, Nozomi Takeshima, Yu Hayasaka, Anna Chaimani, Georgia Salanti

The Lancet Psychiatry, October 2016 Online First


It has been widely accepted that placebo response rates in antidepressant trials have been increasing over the past 30 years. The first systematic study on this topic revealed that among the 75 placebo-controlled antidepressant trials up to 2000, there was a positive correlation between the proportion of responders on placebo and the year of publication. A very similar association was found in paediatric antidepressant trials. Other studies examined pre–post raw change scores or their effect sizes in the placebo groups and confirmed a significant association between the thus-measured placebo response and the year of study publication…

We did a comprehensive and systematic search for published and unpublished double-blind placebo controlled trials of first-generation and second-generation antidepressants (the search was part of an update of a network meta-analysis about antidepressants in major depression). We focused on the placebo response rates and, using the largest dataset so far, we aimed at addressing the following questions: whether the believed increase in placebo response rates had persisted up to 2015, and if not, whether we could replicate the previous findings, and what patient and study characteristics could influence the studied associations.

image003Toshi A. Furukawa

So opens a paper by Furukawa et al., just published (online first) by The Lancet Psychiatry.

Here’s what they did:

· The authors looked for randomized controlled trials comparing antidepressants to placebos. They included many antidepressants, including “all the second-generation antidepressants licensed in Europe, USA, Australia, and Japan, and, of the older drugs, we selected the two tricyclics included in the WHO list of essential medicines (amitriptyline and clomipramine), plus trazodone and nefazodone because they had very distinct effect and tolerability profiles.”

· They searched through various databases (like Medline) and also major drug approval agencies, clinical trial registries and pharmaceutical company websites up to 8 January 2016.

· Patients needed to be 18 or older and to have unipolar depression.

· Two investigators reviewed the abstracts; several of the investigators did the data abstraction, with study quality determined by a rating scale (Cochrane risk of bias tool).

· Statistical analyses were done, including calculation of the proportion of responders by half decade, drawing on results of Hamilton Rating Scale at week 8 (and other validated scales).

Here’s what they found:

· The literature search identified 303 placebo-controlled trials – but adequate data was available for only 252 studies (that included 26,323 patients).

· “Between 1978 and 2015, placebo response rates ranged widely from 0% to 70% but the weighted mean proportion of responders appeared to converge towards 35% to 40%, especially after the early 1990s…” (See the figure below for the mega-regression of log-portion of response to placebo.)

· “The structural break test suggested that the break date was the year 1991 (p=0.04)…” In other words, since 1991, the placebo rate has remained constant.

· The increase in placebo prior to 2000 could be replicated – but was better explained by methodological factors.


They note in the discussion:

In our study, we found that the average placebo response rate in antidepressant trials has remained constant in the range of 35–40% since 1991, year at which we found a structural break. We were able to replicate the reported increase in placebo response rates before 2000, but this association was no longer significant when controlling for methodological factors (such as shorter duration of trials and preponderance of single-centre studies) were frequent in the very early trials and became less often used since 1990s.

One interesting aspect to the study: its findings contradict those of the Walsh et al. paper. The authors speak to this in the discussion section, noting several strengths in their work. For example, they draw on 252 studies (vs. Walsh’s 75). They also note that the previously low response rate of the 1980s was tied to studies with short trial periods and single-site – in other words, the “rising” placebo rate was just methodological artifact.

In a Comment piece that runs with this paper, psychologist Paul Enck of the University Hospital Tubingen, puts the findings in a larger context.

The Comment piece can be found here:

Placebo effect size in randomised controlled trials (RCTs) exhibits a remarkable variability across different medical subspecialties, across different diseases within one subspecialty, and across different RCTs within one disease. Among the many attempts to identify driving factors for the placebo response on the one hand, and its variability on the other, few have received as much attention as has the study by Walsh and colleagues in 2002, claiming that drug and placebo response rates in RCTs in depression were rising. According to their data, the average placebo response across 75 studies rose significantly from an average of 20% in 1980 to 35% in 2000, while the drug response rose from 40% to 50–55% for both tricyclic antidepressants and selective serotonin reuptake inhibitors (SSRI) during the same period.

If so, Fava´s prediction was that sooner or later drug trials in depression would no longer be feasible…

image005Paul Enck

So opens Enck’s Comment. Noting the comprehensiveness of the Furukawa et al. paper, he feels that the whole discussion may now be over.

That said, larger questions remain:

The remaining unanswered questions are: what caused the increase of the placebo response observed before 1991? What initiated the ‘structural break’ around the turn of the century? And why were these time trends and their change not observed in other medical subspecialties?

He doesn’t have answers but notes the need for further study.

To see what initiated the rise and break in placebo responses in depression may require a more subtle and historical approach that takes into account changes in the way drugs have been tested in the past that cannot be known from published reports (eg, individualised patient data, changes in patient recruitment methods, personal preferences of the principal investigator, limitations by local institutional review boards, and other factors.

A few thoughts:

1. The Furukawa et al. paper is a very impressive – it pulls together hundreds of papers and draws out a clear result.

2. 1991 is a key year. Why? There is no answer any of the questions asked by Enck.

3. Of course, it’s important to put this in a larger perspective: for a phenomenon that we don’t really understand, we now know that it’s not changing with time.

4. Mulling the placebo effect and this major study, I’m reminded of Shakespeare’s observation that: “There are more things in heaven and earth, Horatio, / Than are dreamt of in your philosophy.”

Reading of the Week. Every week I pick articles and papers from the world of Psychiatry.