Most recently updated on 13th January 2018
Hello! This is the questions, clarifications and corrections section of the Lost Connections website. As with my previous book ‘Chasing The Scream’, I’m really keen to make sure I have been entirely accurate, and I went through a detailed fact-checking and editing process. Once a month I’ll post an update here, and have any errors corrected for future editions. (You can see the corrections page for Chasing The Scream here to get a sense of what I’m talking about.)
I’m also very happy to have an open conversation about areas where some people think I have made an error and I consult the facts and disagree.
I was emailed and asked to respond to a few people’s queries and comments about the book.
RESPONSE TO STUART RITCHIE
Stuart Ritchie at the University of Edinburgh has asked a question and put forward some studies which he argues contradict some of what I concluded from my interviews and research. He asked where the figure I put forward, that 65 to 80 percent of people taking anti-depressants become depressed again within a year. They are footnoted in the book. I initially learned about this claim from Dr Steve Illardi, Associate Professor at Clinical Psychology at the University of Kansas, in his book ‘The Depression Cure’ (see pages 44 to 45), as Dr Ritchie correctly surmises. I then read through the studies Dr Illardi cites, and discussed in depth this field of research with Professor Irving Kirsch, Professor John Ioannidis, Professor Joanna Moncrieff, Dr David Healey, Dr Lucy Johnson, Dr Sam Everington, Dr Vincent Felitti, Professor Michael Chandler, and with people who defend anti-depressants more strongly like Dr Peter Kramer.
Some of the studies Dr Illardi cites, which I also drew on, are: Corey-Lisle, P. K. et al., “Response, Partial Response, and Nonresponse in Primary Care Treatment of Depression,” Archives of Internal Medicine 164 (2004): 1197–1204; Trivedi et al., “Medication Augmentation after the Failure of SSRIs for Depression,” New England Journal of Medicine 354 (2006): 1243–1252.
In the book, I explain in detail why I was persuaded by the people I spoke to, and by my reading of the evidence, that the research from the Star*D trial – a specific investigation into the effects of anti-depressants – struck me as particularly important. Some studies that are worth looking at in relation to this: Diane Warden et al., “The STAR*D Project Results: A Comprehensive Review of Findings,” Current Psychiatry Reports 9, no. 6 (2007): 449–459; A. John Rush et al., “Acute and Longer-Term Outcomes in Depressed Outpatients Requiring One or Several Treatment Steps: A STAR*D Report,” American Journal of Psychiatry 163 (2006) : 1905–1917; Bradley Gaynes et al., “What Did STAR*D Teach Us? Results from a Large-Scale, Practical, Clinical Trial for Patients With Depression,” Psychiatric Services 60, no. 11 (November 2009), http://dx.doi.org/10.1176/ps.2009.60.11.1439; Mark Sinyor et al., “The Sequenced Treatment Alternatives to Relieve Depression (STAR*D) Trial: A Review,” Canadian Journal of Psychiatry 55, no. 3 (March 2010): 126–135, doi: 10.1177/070674371005500303 ; Thomas Insel at al., “The STAR*D Trial: Revealing the Need for Better Treatments” Psychiatric Services 60 (2009): 1466–1467 . Warden et al., “The STAR*D project results: A comprehensive review of findings,” Current Psychiatry Reports 9, no. 6 (Dec. 2007): 449–459. (Not every one of these studies is in agreement, of course.)
In the book I also discuss how the Star-D trial evidence is disputed by Peter Kramer, in a way I found unpersuasive: I explain to the reader that anyone who wants to look it up can find his critique in Ordinarily Well, 192–3.
I also refer people to the research Dr Ed Pigott has done which argues that 90 percent of people who take anti-depressants experience a relapse or stop taking the drugs because the side-effects are so severe. I thought his analysis, while making some important points, went too far, which is one reason why I went for the lower figure given by Dr Illardi.
In the book I explain why I found the evidence from many of the drug research trials unpersuasive. Stuart Ritchie puts forward four studies which he asks me to consider:
“This 2003 meta-analysis of 31 studies found a 1-year relapse rate of 19% with antidepressants (vs. 41% with placebo): https://www.ncbi.nlm.nih.gov/pubmed/12606176
This 2008 meta-analysis of 23 studies found a 1-year relapse rate of ~20% with antidepressants (vs. ~40% with placebo – see Figure 3): https://t.co/V8TawlMS1E
This 2009 meta-analysis of 11 studies found a 1-year relapse rate of 23% with antidepressants (vs. 51% with placebo):
This 2016 meta-analysis of 72 brief studies (~8 months) found a relapse rate of 23.3% with antidepressants (49.4% w/discontinuation); and of 37 longer studies (~2yrs) found a relapse rate of 24.5% (vs. 55.6% w/placebo):
I asked Professor Irving Kirsch of Harvard University, who has done so much to reveal the flaws in many of these trials, for his response to Dr Ritchie, and he wrote:
“The relapse prevention studies referred to [by Ritchie] are discontinuation trials. It’s what happens to people who improve on drug and then are either kept on the drug or switched to placebo. What this ignores is what happens when people improve on placebo and are kept on placebo. [Professor Kirsch sent me graphics which illustrate this.] Note that people started on placebo and continued on it have the same low relapse rate as those started on drug and continued on drug. Also see the other three slides. Unlike other treatments, those on drugs have to stay on drugs to avoid high rates of relapse. Also note that the STAR*D trial produced an exceptionally high rate of relapse (over 90% according to Ed Pigott’s analysis) even without discontinuation.”
I explain in the book: “The debate about whether chemical antidepressants work—for some other reason we don’t fully understand—is still ongoing. There is no scientific consensus. Many distinguished scientists agree with Irving Kirsch; many agree with Peter Kramer.”
Anyone who would like to read a more detailed analysis of the flaws in the drug testing trials – one that I find persuasive, some others may not – should check out Professor Irving Kirsch’s book ‘The Emperor’s New Drugs’, which goes into this scientific debate in much greater detail, and people who want to read a defence – one which has many good points but ultimately, for reasons I explain in the book, I was not persuaded by – should read Peter Kramer’s ‘Ordinarily Well.’
I’d also like to take this opportunity to reiterate my position. My argument is not that chemical anti-depressants are bad, or that people who feel benefit from them (including some of the people I love) should stop taking them. It is that their benefits are real but limited, and we need to expand the menu of help that is available, and change our culture to deal with the reasons people are feeling so bad. All sides of this debate agree – at the very least – that there is a very significant number of depressed and anxious people who are not helped by anti-depressants, and this is just one of many reasons why we need to have a much bigger debate about depression.
WHAT ABOUT PUBLICLY-FUNDED TRIALS OF ANTI-DEPRESSANTS?
Dr Ritchie has pointed out that there are some publicly-funded trials of anti-depressants which also indicate they are more effective than placebos. Many people understandably infer from this that this fact demonstrates their effectiveness, since there is no vested interest for publicly-funded bodies to back these drugs unless the effects are real. I found this point persuasive when I first learned about it in my research several years ago, and I explored it with some of the scientists named above during the research for the book, and I then looked at many of publicly-funded studies, including some of those cited by Dr Ritchie.
As I dug deeper, I was broadly persuaded by the analysis of these trials put forward by Professor Kirsch and others, and I touch on why in the book.
Before I quote the analysis Professor Kirsch offers, I have to step back for a moment to explain something. Scientists measure the depth of someone’s depression using something called the Hamilton scale, which was invented by a scientist named Max Hamilton in 1959. The Hamilton scale rages from 0 (where you’re skipping along merrily) to 51 (where you’re suicidal). To give you a yardstick: you can get a six-point leap in your Hamilton score if you improve your sleeping patterns.
Concerning the trials Dr Ritchie raises, Professor Kirsch explains: “About half of studies, regardless of funding, find support [for chemical anti-depressants] in the sense that they show a statistically significant difference between drug and placebo. The question is, are these differences clinically meaningful?”
When Professor Kirsch studied the data provided in private to the Food and Drug Administration (FDA) in the US when anti-depressants were first being submitted for release to the public, he found that the average improvement an individual experiences when they take them is 1.8 percent on the Hamilton Scale – which is a little less than a third of the improvement you get if you improve your sleep patterns. There is a real effect, but it is relatively small, on average. (There is a debate about whether it works better with people who have severe depression – I go through this debate in the book.)
Professor Kirsch argues that this improvement is too small to be regarded as clinically significant. Others think he has somewhat under-estimated the effect, and others argue that an improvement of 1.8 percent is worth the cost and side-effects of the drug.
Professor Kirsch adds: “Also, various design features can augment the apparent [small] effect. These include [trials which are]:
- Excluding people who are mildly or moderately depressed, which make up the majority of patients given antidepressant prescriptions
- Excluding patients who get better on placebo during the wash out period
- Replacing patients who fail to show improvement on the drug during the first two weeks (e.g., in the fluoxetine and sertraline trials submitted to the FDA)
- Breaking blind by patients and clinicians [This is a really important concept explained in detail in Professor Kirsch’s book ‘The Emperor’s New Drugs’] So even that small, clinically insignificant difference, may be an artefact.”
So: nobody disputes that publicly funded trials show some improvement over placebo. I explain this clearly in the book. The debate is – how big is the effect? Some people think it’s too small to solve the problem; others believe it is not, at least for some people.
(There is, of course, no dispute that in addition to this effect, the drug companies exaggerated the effects of these drugs in their trials. They had to make a huge pay-out in the US courts for doing this, following a detailed legal case, concerning the exact drug I took for thirteen years – for a little background on this, check out this, this, this, this, this, this and this.)
It was looking at this evidence, and much more, that led me to my conclusion about chemical anti-depressants: that they have some real effect, but for most people, it is modest, and we need lots of other solutions alongside the chemicals. This is why I am not opposed to these drugs, but I think they need to be complemented by many more strategies to solve depression and anxiety.
RESPONDING TO DEAN BURNETT
One person emailed to ask me to ask me to respond to a post by Dr Dean Burnett entitled ‘Is Everything Johann Hari Thinks About Depression Wrong?’ He makes a series of points that I am happy to respond to. (Dr Burnett makes it clear he is responding to an article I wrote, extracted in the Guardian/ Observer, rather than the book.) He makes four points:
Point one: He claims that it is an overstatement to say that, since the grief exception was removed from the DSM, grieving people can be immediately diagnosed with depression. I asked Dr Joanne Caciatorre, the leading expert on the grief exception and a specialist in treating traumatic grief, to read his reply. She’s the person who – along with others – taught me about this phenomenon. She replied:
“There is no bereavement exclusion anymore in the DSM [Diagnostic and Statistical Manual], thus a diagnosis of MDD [Major Depressive Disorder] can be made when a person has the normal symptoms of grief for two weeks. And a person could be dx’d [diagnosed] with MDD one minute after the death of a baby if he or she had symptoms of MDD for the prior two weeks (many pregnant women feel weepy, experience changes in sleep patterns, loss of energy, fatigued, have trouble concentrating, etc). Of course, any wise clinician would not diagnose MDD under such tragic circumstances even if those symptoms predated the birth and death of the baby by two weeks. That isn’t the point: the point is that there is no protection against such carelessness and it contributes to the mistreatment of families suffering grief. And so, grieving parents are diagnosed within days and weeks at their therapist’s or doctor’s discretion.”
She continues: “This happens all the time, and it manifests in overprescribing trends. For example, in one robust sample of 235 grieving parents, nearly 40% were prescribed psychiatric medications and of those 32% of prescriptions were written within 48 hours (44% within a week and 75% within a month) (Lacasse & Cacciatore, 2014).”
So while Dr Burnett says the event I described cannot happen, the leading expert on this question – who has studied it in depth, and published numerous scientific papers on it – says it can. I wish this were false. Alas, it’s not – and, as Dr Caciatorre says, it reveals a lot about how we don’t understand human suffering, in ways very relevant to the debate about depression.
Point two: He argues that Professor Irving Kirsch is a marginal figure in the field and should not be taken seriously. In fact, he is a leading Professor at Harvard University, and I explored in depth his relationship to the rest of the field in my research, as I lay out in Lost Connections. Professor John Ioannidis – who the British Medical Journal calls “the scourge of sloppy science”, and the Atlantic Monthly says “may be one of the most influential scientists alive” – is one of the people I discussed Professor Kirsch’s findings with. He explained to me that research into chemical anti-depressants is absolutely dominated by researchers who take money from the drug companies who profit from these drugs, and that they frequently even write up the finished scientific reports. In this context, he explained, warnings from Professor Kirsch and others that chemical anti-depressants have more limited effects for most people than is widely believed, and serious side-effects, will not get much traction, for obvious reasons.
Speaking of this kind of pharmaceutical research, “I think that this is a field that is seriously sick,” Professor Ioannidis told me. “The field is just sick and bought and corrupted, and I can’t describe it otherwise.” This is the only context in which Professor Kirsch, who doesn’t take money from the drug companies, is marginal – not for intellectual reasons, but for commercial ones, and I think that’s a reason to respect him more, not less. (He is certainly not the only person to refuse money from the drug companies, and it’s worth stressing that – as a I say in the book – not everyone who refuses money from the drug companies is as sceptical as Professor Kirsch about the benefits of the drugs.)
Point three: Dr Burnett argues I present some of my wider arguments – that there are deeper social and psychological causes of depression and anxiety which must be dealt with – as if they are new, when in fact they are mainstream and well-known within the scientific community. But I explicitly state in the article that they are well-known and mainstream in the field, writing: “This is radical, but it is not, I discovered, some maverick position. In its official statement for World Health Day in 2017, the United Nations reviewed the best evidence and concluded that “the dominant biomedical narrative of depression” is based on “biased and selective use of research outcomes” that “must be abandoned.”” I talk more in ‘Lost Connections’ about how this broader argument is the position of the World Health Organisation and much psychiatric training.
He says many of these insights have been known since the 1970s. That’s correct – which is why in the article, I explain that Professor Michael Marmot made his discovery in the 1970s, and the book is full of the stories of discoveries made in the 1970s.
The problem is not that scientists don’t know this stuff about the deeper causes of depression and anxiety – I state again and again that they do. It’s that (1) it has not been explained to the public, (2) that these insights haven’t been used to reshape the services we offer to depressed and anxious people, and (3) that it hasn’t been used to reshape public policy. Indeed, there has been a retreat from these perspectives – not philosophically, but in practice. As Professor Laurence Kirmayer, the head of Social Psychiatry at McGill University, told me (speaking mainly of the US and Canadian systems, which Dr Burnett is not part of): “Psychiatry has undergone a real constriction from this bio-psycho-social approach. While some people still pay lip service to it, mainstream psychiatry has become very biological.” He added “It’s very problematic.” We have ended up with “a grossly oversimplifed picture” of depression that he said “doesn’t look at social factors . . . But at a deeper level for me, it doesn’t look at basic human processes.”
This is why, while these insights are not new to specialists in the field, they are new to us, the public – which is why I wrote a book to explain it to a wider public than those who read psychiatric journals where so much crucial work is being done, and to try to figure out its huge implications for how we all live.
Point four: He says I claim that only one kind of treatment is available on the British National Health Service for depression, when in fact they are many. Far from neglecting this topic, I have a whole chapter on the best of this NHS diversification in my book. (It’s chapter seventeen for anyone who wants to check it out). Of course, it’s not possible to get every point I wanted into every article. He’s right to pick up on the fact I say “there’s only one thing on the menu for depressed and anxious people” – for space reasons, I cut the words “too often, there’s only one thing on the menu for depressed and anxious people.” I should have kept those words in to be clearer.
A RESPONSE TO DR CARMINE PARIANTE
Dr Carmine Pariante wrote an article in the Independent entitled ‘As a psychiatrist, I know Johann Hari is wrong to cast doubt on anti-depressants’. Dr Pariante and I concur on a lot, and I admire his work. He wrote to me: “As a psychiatrist, I value your emphasis on the importance of social connections, and I hope that you will recognize in my response the many points on which I completely agree with you.”
We differ a little on some questions (not all) surrounding chemical anti-depressants. I want to go through his two key points separately:
Point one: Dr Pariante argues: “Antidepressants reverse some of the physical effects of depression: increasing production of new brain cells, and reducing inflammation.”
Response: I asked him to send me the study on which he was basing this conclusion, and I looked at it carefully and then asked Professor Kirsch for his analysis. (Again, I had discussed this question during my research with several scientists, but I wanted to return to it here with one of the most distinguished figures in the field to make sure I had understood it.) The study Dr Pariante sent over is here.
Professor Kirsch responded: “Pariante bases this conclusion on an uncontrolled study. With no control group, there is no way of knowing whether the effect was due to the antidepressant, rather than it being a consequence of improvement in depression brought about by the passage of time and/or the placebo effect. For example, using PET, Mayberg and colleagues found that successful treatment with an SSRI was associated with “metabolic increases involving the prefrontal, anterior cingulate, premotor, parietal, posterior insula, and posterior cingulate and metabolic decreases involving the subgenual cingulate, parahippocampus, and thalamus.” The hitch is that pretty much the same changes were observed among participants who had been successfully treated with placebo.”
Point two: Dr Pariante also argued: “Studies using the most rigorous approaches (supported by publicly funded bodies, such as the Medical Research Council, not just pharmaceutical companies as Hari suggests) show that antidepressants lift the pervasive sadness at the core of clinical depression, albeit with varying effects on other symptoms such as sleep or appetite problems.” (The study he sent me for this is here)
I respond to the argument about publicly-funded trials earlier in this post – see above.
Concerning the argument that chemical anti-depressants lift the pervasive sadness at the heart of depression, Professor Kirsch responded: “Pariante bases this on a recent study by Hieronymus et al (2017) in which superiority of SSRI over placebo were found more consistently on a single item of the Hamilton depression scale than on the total score of the scale. There is a reason for constructing scales made of many items, and the reason is that multi-item scales are more reliable that single items in that scale. For example, in the supplementary material of the Hieronymus study, it was revealed that side effects were associated higher single-item depression scores on one of the two SSRIs studied (citalopram) but lower scores the other (paroxetine). Further, the single item on which these findings were based does not measure “the pervasive sadness at the core of clinical depression.” Here is what that the item actually says:
1 DEPRESSED MOOD (sadness, hopeless, helpless, worthless)
0 |__| Absent.
1 |__| These feeling states indicated only on questioning.
2 |__| These feeling states spontaneously reported verbally.
3 |__| Communicates feeling states non-verbally, i.e. through facial expression, posture, voice and tendency to weep.
4 |__| Patient reports virtually only these feeling states in his/her spontaneous verbal and non-verbal communication.
The mean score on this item was between 1 and 2, regardless of whether patients were given drug or placebo, meaning that both groups reported feeling sad, hopeless, helpless, and worthless after treatment, but those who had been given the SSRI were more likely to state them without being asked and those given placebo more likely to wait to be asked. Now I ask you, is this a clinically meaningful difference in outcome?”
I found these arguments by Professor Kirsch persuasive but this is an on-going scientific argument in which they are both far more expert than me, and I’m looking forward to continuing to follow it.
I’m grateful to Dr Cariante and to Professor Kirsch – who agree on a lot – for engaging in this way.
RESPONSE TO ZOE STAVRI
Zoe Stavri argues that I am wrong to draw on the work of Professor Irving Kirsch, for three reasons. I was aware during my research of these criticisms (not specifically from Stavri, but from others), and I was also aware of Professor Kirsch’s response to them, and I was broadly persuaded by Professor Kirsch’s arguments. I asked him to reiterate them specifically for this page.
Point one, from Stavri. She says: ‘Kirsch and colleagues didn’t find that antidepressants don’t work on the majority of people: they found that effectiveness of antidepressants are more effective for severe depression and less effective for mild or moderate depression. That’s a nuanced difference’
Response, from Professor Kirsch (the square brackets are from me): “We found no drug-placebo effect at all for mild or moderate depression and less than clinically significant differences for most people with very severe depression. Those who are depressed enough to show a clinically (as opposed to statistically) significant difference represent approximately 10% of patients presenting in practice with diagnoses of major depressive disorder. However, it is important to look at the criterion used for clinical significance. NICE [The National Institute for Clinical Excellence, the independent body which assesses the efficacy of drugs] set a 3 point difference on the HAM-D [the Hamilton scale, which measures depression] or an effect size of 0.50. Leucht and colleagues examined what changes in the HAM-D correspond to global clinical ratings of improvement. They found that a 3 point difference is equivalent to a clinician rating of “no change”. To get a rating of “minimally improved” you need a change of 7 points on the HAM-D. so far, no one has been able to find a drug-placebo difference anywhere close to that in any group of patients, no matter what the severity.”
Point two, from Stavri: “Other researchers analysed the same dataset as Kirsch and drew different conclusions. Using different statistical modelling, Fountoulakis and colleagues found antidepressants were better than placebo, at all levels of depression severity. Turner and Rosenthal’s interpretation of the data is different to Kirsch’s, suggesting that certain measures can be more important than disappearance of depression, such as quality of life, which has been overlooked in Kirsch’s study, and to be “circumspect but not dismissive” in considering the benefits of antidepressants.’”
Response, from Professor Kirsch: “They made statistical errors in there calculations, which we pointed out in our rejoinder. In any case, some meta-analyses show an association of drug-placebo differences with baseline severity; others do not. Others use unconventional cut-offs for classifying severity, and some use statistical techniques (e.g., grouping Ss into severe v not severe) that all statisticians say are unreliable, akin to throwing away at least 1/3 of the data, and therefore not having enough power to reliably show real differences. More important, no one finds an overall drug-placebo difference that is clinically significant even according to the NICE criteria. So if the effect is the same across levels of severity, then no one is getting a clinically meaningful effect from the drug compared to what they would get on placebo. We used the same measure used by the FDA to make approval decisions.”
Point three, from Dr Stavri: “Kirsch only looked at certain types of antidepressants – four types of SSRIs; your comments re relating specifically to the serotonin hypothesis and don’t relate to antidepressants in general.”
Response three, from Professor Kirsch: “There are few if any differences between the effects of one type of AD [anti-depressant] vs another. Other researchers have used different ADs in their meta-analyses. All come up with the same basic findings as did we. No one gets an effect size above 0.35 (the NICE criterion is 0.50). This includes two meta-analyses done in-house at the FDA on all the clinical trials for which they had the data. The largest found the same 1.8 point difference that we reported in our meta-analyses. Note also, that one gets the same response to Tianeptine (an SSRE) as to SSRIs.”
A WIDER POINT
In his response, Dr Burnett seems to think – I could be wrong – that my book is an attack on general practitioners and psychiatrists who are working hard every day and doing their best. Far from being opposed to psychiatry, I am hugely sympathetic to doctors and psychiatrists working day to day with some of the most distressed people in our culture, who overwhelmingly work really hard to do the best for their patients. They are doing their best with the tools they are given. My own sister was, until very recently, a psychiatric nurse, and I am enormously proud of her and the work she did, which saved lives.
It is the wider society – all the rest of us – that need to do a better job of giving them better tools, of acting on the insights from the best research into mental health, and (most importantly of all) of changing the way we all live so that fewer people are left in a state of profound distress.
We don’t leave the job of dealing with car crashes simply to people who have already been mangled in car accidents and to doctors in emergency rooms. We know that the culture has to change to prevent it – through airbags and seatbelts and speed limits and bans on drunk-driving. We need a comparable cultural shift to deal with the nine deep causes of depression and anxiety I was able to find evidence for in ‘Lost Connections.’
If you spot anything in ‘Lost Corrections’ that you think is not correct, the best way to contact me is to drop me a line at email@example.com – on the 12th of each month I’ll post an update on what I’ve received and respond to queries. (Note: this post replaced a slightly shorter one responding only to Dr Burnett but I haven’t changed any of the points made there, I have just integrated them into this longer response.)