Five years ago, I spent time as a fellow in a lab at the Safra Center for Ethics at Harvard that was devoted to studying “institutional corruption,” and what I particularly appreciated about the lab was that it provided a clear method to investigate and conceptualize the problem. The framework was this: Identify “economies of influence” that may corrupt the behavior of individuals within the institution, document the corruption, and explore the resulting social injury.
Lisa Cosgrove and I joined together in studying the “institution” of psychiatry through this lens, and we focused on its institutional behavior since 1980, when the American Psychiatric Association published the third edition of its Diagnostic and Statistical Manual. This was when the APA adopted its disease model for categorizing mental disorders, with the profession then taking up the task of selling this new model to the public. This was the particular guild interest that arose in 1980, and has shaped its behavior ever since. We’re all familiar with the second “economy of influence” that has exerted a corrupting influence on psychiatry—pharmaceutical money—but I believe the guild influence is really the bigger problem.
In our book Psychiatry Under the Influence, we then documented the corrupt behavior, which could be found in every corner of psychiatry: the false story told to the public about drugs that fixed chemical imbalances in the brain; the biasing of clinical trials by design; the spinning of results; the hiding of poor long-term results; the expansion of diagnostic categories for commercial purposes; and the publishing of clinical trial guidelines that inevitably promoted the use of psychiatric drugs.
At the end of our investigation, I had a new way of conceptualizing the social injury that was caused by this corruption: our society had organized itself around a false narrative, one that was presented to us as a narrative of science, but was belied by a close examination of the actual evidence.
Now, with the publication of a new study in the American Journal of Psychiatry titled “The Long-term Effects of Antipsychotic Medication on Clinical Course in Schizophrenia,” we have a new opportunity to observe this “institutional mind” of psychiatry at work. The article, authored by former APA president Jeffrey Lieberman and seven other psychiatrists, is meant to serve as an evidence-based review that defends the profession’s current protocols for prescribing antipsychotics, which includes their regular long-term use. By closely examining this review, we can assess, once more, whether this is a profession that can be trusted to honestly evaluate its evidence base and use that evidence to guide its care.
In 2010, I published Anatomy of an Epidemic, and in that book, I wrote about the long-term effects of antipsychotics, and concluded that there is a history of science that leads to this conclusion: on the whole, antipsychotics worsen the long-term outcomes of people diagnosed with schizophrenia and other psychotic disorders. Joanna Moncrieff has similarly written about the hazards of antipsychotics in her book The Bitterest Pills and in published papers. So too has Peter Gøtzsche in his book Deadly Psychiatry and Organized Denial and various journals.
All of this criticism helped to promote further inquiry into this concern, which put psychiatry on the hot seat: did it prescribe these drugs in a way that caused more harm than good? Did their protocols for the drugs—immediate use for all first-episode psychotic patients and then “maintenance” use of the drugs—need to be rethought?
In their newly published study, Lieberman and colleagues perform what they describe as an “evidence-based” review of these questions. They conclude that there is no compelling evidence that supports this concern. In a subsequent press release and a video for a Medscape commentary, Lieberman has touted it as proving that antipsychotics provide a great benefit, psychiatry’s protocols are just fine, and that the critics are “nefarious” individuals intent on doing harm.
Joanna Moncrieff has already published a blog on Mad in America that is critical of the study, and, in particular, of the authors’ dismissal of studies related to the effect of antipsychotics on brain volumes. Miriam Larsen-Barr, who has done research on user attitudes toward antipsychotics, also wrote a blog criticizing the study and press release, focusing on how the authors ignore user accounts about how the drugs affect their lives.
In this MIA report, I simply want to look closely at how Lieberman and his collaborators reviewed the literature and individual studies. We can then see whether they have done so in a way that reveals the mind of a group interested in truly investigating the question of the long-term effects of antipsychotics, with the patients’ well-being uppermost in their thoughts, or whether it reveals the “mind” of a group interested in protecting guild interests.
Then, at the end of this exercise, we can ask this essential question: If we can’t trust the profession to develop “evidence-based” treatments that put the interests of patients first, as opposed to their own guild interests, what should our society do?
The eight authors describe themselves as “an international panel of experts in antipsychotic pharmacology, neuroimaging, and neuropathology.” The task they gave themselves was to “review the preclinical and clinical evidence supporting potential adverse effects of antipsychotics on long-term outcomes.”
From an “institutional corruption” perspective, there is additional information that needs to be added to that description: All eight are psychiatrists, and thus there is a “guild” interest present in this review, given that they are investigating whether one of their treatments is harmful over the long-term. As for the influence of pharmaceutical money, five of the eight report financial ties to drug companies, with at least three serving on speaker bureaus for multiple companies. Lieberman reports no such conflicts of interest, but he has a past filled with such conflicts: in 2009, he disclosed having had financial ties to 15 drug companies in the previous two years.
The authors, in their introduction, cite papers by Moncrieff, Gøtzsche and me as raising concerns about the long-term effects of antipsychotics. In terms of clinical importance, the review should serve as an examination of evidence relevant to how antipsychotics are being prescribed. Are they being used in a manner that best serves patients, or in a manner that increases the likelihood that a person who suffers a psychotic break will have a poor long-term outcome?
Psychiatry’s current protocol calls for psychiatric drugs to be prescribed to all psychotic patients, and for “maintenance therapy” over the long term. Placebo-controlled trials provide the evidence for the first practice. Withdrawal studies, where patients stabilized on medications are randomized to continued treatment or to withdrawal (usually abruptly so), support the second.
However, the withdrawal studies don’t tell of how maintenance therapy affects patients over the long term. They simply reveal that once a person has stablized on the medication, going abruptly off the drug is likely to lead to relapse. The focus on long-term outcomes, at least as presented by critics, provides evidence that psychiatry should adopt a selective-use protocol. If first-episode patients are not immediately put on antipsychotics, there is a significant percentage that will recover, and this “spontaneous recovery” puts them onto a good long-term course. As for patients treated with the medications, the goal would be to minimize long-term use, as there is evidence that antipsychotics, on the whole, worsen long-term outcomes.
This is the clinical relevance of this “concern.” Does psychiatry need to rethink its drug-use protocols in order to give patients the best chance to recover and function well over the long term?
The good news here is that the authors, with this paper, are seeking to address the main issues raised by critics. Furthermore, as they performed their review, they revisited many of the studies we “critics” have written about. To a large degree, they have retraced the historical path I wrote about in Anatomy of an Epidemic, which provides a sense of agreement on the evidence that there is to be found and reviewed.
Deconstructing Their Review: Part One
The first part of their review is devoted to assessing the evidence base for use of antipsychotics to treat psychotic episodes, including their use in first-episode patients. Is there evidence that supports treating all patients in this way, or is there reason to develop protocols, particularly for first-episode patients, that would give them an opportunity to recover without exposure to the drugs?
A. Placebo-controlled trials in psychotic patients
In 2009, Leucht published a meta-analysis of 38 trials of second-generation antipsychotics and reported a response rate of 41% for the drug-treated patients versus 24% for the placebo group. Lieberman and colleagues cite this study as providing solid evidence for regularly prescribing antipsychotics to patients experiencing psychotic episodes. “This magnitude of therapeutic effect compares favorably with many of the most effective treatments in non-psychiatric fields of medicine,” they wrote.
What’s missing from their review
While placebo-controlled studies may show whether a drug is more effective than placebo in curbing a target symptom (over a short period of time), they do not provide evidence of the benefit-harm effect that the drug has on all patients so treated. To make that assessment, reviewers need to calculate the NNT in the trial data, which is the number of patients that need to be treated to get one additional positive response. This identifies both the percentage of patients that benefit from the treatment and the percentage that may be harmed by exposure to a treatment they don’t benefit from (e.g., those who would have recovered anyway or are non-responders to the treatment.)
In the meta-analysis cited by Lieberman, the NNT is 6. Although 41% of drug-treated patients are responders, 24% of the placebo patients are too, producing a net increase in responders of 17%. That means for every six patients you treat with the drug, you will get one additional responder, and this is the group that could be said to benefit from the treatment.
The remaining 83% fall into the harm category. There are the non-responders (59%) who didn’t benefit from the treatment but are now exposed to the adverse effects of antipsychotics, and the placebo responders (24%) who would have gotten better without such treatment but are similarly exposed to the drugs’ adverse effects.
If you have a treatment with minimal side effects, then an NNT of six may justify a one-size-fits-all protocol. The additional chance of responding to the treatment justifies taking a risk of exposure to a drug with minor side effects. But in this case, given that antipsychotics have such severe side effects, the Leucht meta-analysis of placebo-controlled trials provides an evidence-based argument for avoiding antipsychotics as a first-line therapy for all patients, and attempting instead to see if other approaches—psychosocial approaches, sleep aids, and so forth—can produce a similar response rate over the short term (41%), without exposing everyone to the hazards of the drugs. The drugs might then be tried as a second-line therapy for non-responders.
The point, for the purposes of this MIA report, is this: reviewers seeking to promote their drug treatment as effective will look solely at whether it produces a superior response to placebo. This leads to a one-size-fits-all protocol. Reviewers that want to assess the benefit-harm effect of the treatment on all patients will look at NNT numbers. In this instance, the NNT calculations argue for selective use of the drugs, for you would want protocols that allowed for “placebo responders” to emerge without being exposed to the drugs, and did not call for drugs to be prescribed on a continual basis to “non-responders.”
B. Placebo-controlled trials in first-episode patients
The Leucht study cited by Lieberman implies that 24% of psychotic patients will recover, or at least partially recover, from a psychotic episode within a six-week period (the usual length of placebo-controlled trials.) However, the trials in the Leucht meta-analysis were conducted in patients who had been exposed to antipsychotics, with the “placebo group” composed mostly of patients abruptly withdrawn from the drugs. They aren’t a true placebo group, and this raises an essential question: What is the recovery rate for first-episode patients who aren’t treated with antipsychotics, and how does this compare to the recovery rate for first-episode patients treated with the drugs?
In their paper, Lieberman and colleagues provide this answer: “No placebo-controlled trials have been reported in first-episode psychosis patients.”
This is a rather startling confession. Since the introduction of antipsychotics in 1955, there has never been a placebo-controlled study in medication-naïve patients to see if the drugs are effective in this group. Yet this initial drug treatment regularly becomes a defining moment in their lives, because psychiatry’s usual protocol is to maintain patients on antipsychotics indefinitely.
This alone should give psychiatry reason to re-examine its practice for treating first-episode patients. It is not evidence-based, and it would seem that the profession, which likes to present itself as practicing evidence-based medicine, would want to remedy this extraordinary deficiency.
C. Studies that have reported recovery rates in first-episode patients, or a cohort largely composed of first-episode patients, that involved treating them with psychosocial care and without antipsychotics
1) Odegard and McWalter
These are two early studies cited by Lieberman and colleagues that compared recovery rates immediately before and after the introduction of antipsychotics. They wrote that these studies showed “increased rates of discharge and reduced rates of hospitalization associated with initial antipsychotic treatment.”
In other words, they are stating that these studies support the immediate use of antipsychotics in first-episode patients, and that this use leads to superior long-term outcomes. It isn’t placebo-controlled evidence, but evidence of a different sort that supports their current practice.
What’s missing from their review
They did not provide any data from these two studies, which, if presented, would argue for selective use of antipsychotics. Here are the data:
In both studies, more than 60% of first-episode patients treated without antipsychotics recovered and never relapsed in follow-up periods of three to five years. They had an episode of psychosis, rather than becoming chronically ill (and thus in need of continual drug treatment.)
It is also notable that the recovery rate—and the longer-term stay-well rate—didn’t really change with the introduction of antipsychotics. At this time, there was not yet the practice of maintaining discharged patients on antipsychotics, and thus the good stay-well rate seen in these studies is for patients who are not being maintained on medications long term.
In the early 1960s, May randomized 228 first-admission patients to psychotherapy, milieu treatment, antipsychotic, ECT, and drug plus psychotherapy. There was no placebo control, but the milieu group provided something close to an “untreated” group. “Patients assigned to antipsychotic (alone or in combination) or to ECT spent less time in the hospital over the subsequent three years and exhibited superior social functioning,” Lieberman and colleagues wrote.
This study is being presented as showing that initial use of the drug led to a better three-year outcome for patients so treated.
What’s missing from their review
They fail to report this important result: 59% of the patients randomized to milieu therapy were deemed “successes,” and over the three years, May reported, “the initial of successes from milieu functioned over the follow-up at least as well, if not better, than the successes from the other treatment.”
Indeed, the milieu successes had the best scores of any of the five groups on social relationships and overall adjustment, which led May to write: “Even though some treatments get worse results on the average, their successes might be of higher, or more enduring quality.”
This is a finding relevant to whether a significant percentage of first-episode patients can be successfully treated without medications, and how such “successes” fare over the longer term. Once again, it shows that about 60% fell into this group, and that this group of “successes” appeared to be of “more enduring quality.”
In the early 1960s, the NIMH conducted a study of 229 acutely ill inpatients with schizophrenia who were randomized to one of three antipsychotics or to placebo. Fifty percent were first-episode patients. While the drug-treated patients had higher recovery rates over the short-term, Lieberman and colleagues noted that “individuals who received placebo during the initial 6-week trial were less likely to relapse after hospital discharge compared within individuals who received phenothiazine.”
Having reported on the better one-year outcomes for the placebo patients, Lieberman and colleagues then dismissed that favorable result in this way: “Because dropout rates due to non-response during the initial randomized trial differed substantially between treatment groups (2% in the phenothiazine group compared with 29% in the placebo group), it is highly likely that the sample for the follow-up study of discharged patients contained a disproportionate number of poor-prognosis patients in the phenothiazine group, since these subjects would not have achieved discharge if treated with placebo.”
What’s missing from their review
First, they failed to report that 67% of the placebo group were discharged, and thus were in this “best outcome” group at the end of one year. This is a finding that appears to duplicate the results in the May study.
Second, in their effort to dismiss the better results for the placebo patients, Lieberman and colleagues assume that it was “poor prognosis” placebo patients who dropped out from the study. However, as modern longitudinal studies have shown (such as Harrow), the drop-outs may be better characterized as “non-compliant” patients, and this is the group that has the better long-term outcomes.
Third, when Schooler sought to explain the better 1-year outcomes for those treated with placebo, she found that the placebo patients were “more likely to have fathers who were mentally ill,” which was a risk factor that “increased the likelihood of rehospitalization.” This difference in the patient groups, Schooler wrote, “would make a higher rehospitalization rate of placebo patients more, rather than less, likely.”
In this 1970s study “of acutely psychotic patients, most of whom were medication naive,” 127 were randomized to either placebo or chlorpromazine, and followed for three years after discharge. Lieberman and colleagues wrote that Rappaport “reported poorer outcomes over a 3-year follow-up” in the chlorpromazine group, but then dismissed the better outcomes for the placebo patients.
The design, they wrote, is “confounded by unequal attrition rates during the initial treatment phase—45% of placebo-treated patients dropped out compared with 26% of patients treated with chlorpromazine. After correction for the unequal loss of poor-prognosis patients in the placebo group, differences in the outcome were no longer significant between groups.”
What’s missing/wrong with their review
First, at the end of three years, there were 41 in the placebo group who were still in the study versus 39 in the drug group. So there was no unequal “attrition” by study end (see graphic below).
Second, Rappaport wrote that it was “during the follow-up period” that there was a larger attrition of subjects from the placebo group, rather than in the initial treatment phase, as claimed by Lieberman. If this is so, then there must have been a larger attrition of subjects from the drug group in the initial treatment phase (for the numbers in each group to be equal at study end.)
Third, of the 80 patients still in the study at the end, 14 in the placebo group were seen as “bad prognosis” patients at the start of the study, and 14 in the drug group were so designated at the start. Thus, there was no “unequal loss of poor prognosis patients” that confounded the three-year results.
Fourth, the differences in outcomes between the two groups were dramatic. Two thirds of those randomized to placebo were off medication at the end of the follow-up period, and it was this group that, by far, had the best three-year outcomes. At the same time, 57% of those randomized to drug remained on an antipsychotic for the three years, and it was this medication-compliant group—following the prescribed protocols of today for antipsychotic use—that had the highest rehospitalization rate, roughly nine times higher than for the never-exposed group.
Here are the outcomes that Lieberman and colleagues did not present:
Finally, in an effort to investigate the higher attrition rate for the placebo patients in the follow-up phase, Rappaport did this analysis: he threw out the four worst scores in the chlorpromazine/off group, and with this smaller sample size, there were now no significant differences between the never-medicated group and the patients randomized to drug who then stopped taking the medication during the follow-up phase. Rappaport also threw out the five worst scores in the chlorpromazine/on group, but even with this manipulation, the significant differences in outcomes between the placebo/off and chlorpromazine/on groups didn’t disappear.
In sum, Lieberman and colleagues, in their review of this study, didn’t report the actual outcome data from this study, and dismissed the much better outcomes for the off-med groups with an inaccurate description of Rappaport’s investigation into the effect of attrition rates. Lieberman and colleagues, in their bottom-line summary of the study, stated that there were no significant differences at three years between those randomized to placebo and those randomized to drug, when, in fact, the study told of placebo/off patients doing quite well, and medication compliant patients doing quite poorly.
Rappaport made this clear in the study abstract: “Our findings suggest that antipsychotic medication is not the treatment of choice, at least for certain patients, if one is interested in long-term clinical improvement. Many unmedicated-while-in-hospital patients showed greater long-term improvement, less pathology at follow-up, fewer rehospitalizations, and better overall functioning in the community than patients who were given chlorpromazine while in the hospital.”
Lieberman and colleagues cited this Chinese study in a section of their paper devoted to this question: “What is the clinical evidence that initial treatment affects long-term outcome?” In this section, they also cited the Odegard, McWalter, May, Schooler, and Rappaport studies, and thus they are positioning the Ran study as one that provides findings on patients treated initially with and without medications, who are then followed for some time.
Ran’s study, they said, was a “14-year follow-up comparing medicated with never-medicated individuals,” and they described the results in this way: “Outcomes were poorer in the never medicated group, including rates of partial and complete remission, homelessness, and mortality. Approximately 10% of individuals who had never been treated achieved remission, and approximately 8% were partially remitted compared with 30% and 37%, respectively, in individuals who regularly took medication.”
Given the context, the readers are left to understand that this study provided evidence that initial treatment with medication, and subsequent “regular” medication use over the next fourteen years, produced superior outcomes.
What’s missing/wrong with their review
This study did not compare the long-term outcomes of non-medicated psychotic patients, beginning with the identification of such patients early in the course of their illness, to a similar group of medicated patients. Lieberman and colleagues have completely misrepresented this study: its design, the makeup of the patient cohorts, their medication use, and, apparently just for good measure, their outcomes.
In the study, Hong Kong investigators surveyed a rural community of 100,000 in China, and identified 510 people who met the criteria for a diagnosis of schizophrenia. This was in 1994, and at this baseline moment, there were 156 who had never been treated and 354 who had received antipsychotic medication at least once, and thus were deemed the “treated” group.
The “untreated” group were on average 48 years old and had been ill for 14 years. Anyone in this rural community who, in the years before 1994, had suffered a psychotic episode and recovered without treatment would not have shown up in this study. In layman’s terms, the investigators had identified a group of “chronically crazy” people for their “untreated” group they would now follow for 14 years.
The “treated” group were not, in fact, a “regularly” medicated group, not even at baseline. They were simply a cohort of patients who had received antipsychotic medication at least once before the start of the study. There is no information given on how many in this group were taking a medication in 1994, when the study began.
At this baseline moment, the chronic, untreated group was much more severely ill than the “treated” cohort, the investigators noted. The untreated cohort was “significantly older, less likely to be married, more likely to have no family caregiver and to live alone, had a lower education level, and fewer family members.” The untreated group also came from families with a significantly lower economic status, and they were more likely to have been abused by their families. In addition, the never-treated group had a “longer duration of illness; higher mean scores on the PANSS positive subscale; and had higher PANSS negative subscale and general mental scores.” Eighty-three percent had “marked symptoms/or were deteriorated,” compared to 54% of those in the treated group.” Only 17% percent were in partial or complete remission at baseline, compared to 47% of those in the treated group.
Over the next 14 years, there were many in the “untreated” cohort who continued to go without treatment, and the researchers reported the remission rate for this “never-medicated” cohort at study end. As for the “treated” group, the investigators did not report on their medication use during the 14 years. They do note that in the last year of the study, 2008, only 11% of this “treated” group took a second-generation antipsychotic even once. There is no information given about whether any in this group were taking a first-generation antipsychotic at that time.
In their review, Lieberman and colleagues did not mention this. They present it as a long-term study that compares medicated to unmedicated individuals, when in fact it is a study that compares a select group of older chronic patients to a less ill group that, at some point in their lives, were “treated.” Patients in this second cohort who weren’t taking an antipsychotic at baseline and did not take any medication over the next 14 years would still have been counted in the outcomes for the “treated” group. Yet, Lieberman and colleagues describe them as “regularly medicated” during the 14 years.
In addition, Lieberman and his co-authors also got the remission numbers wrong. At the end of 14 years, the remission/partial remission rate for the chronically ill, unmedicated group was 30% (not 18%), and it was 57% for the treated group (not 67%.) They deflated the remission rate for the group they described as “unmedicated,” and inflated the rate for the group they wrongly described as “medicated.”
And this is a study they cited as providing evidence that initial treatment with medication produced better long-term outcomes.
In the past 15 years, Jaakko Seikkula has regularly reported on the five-year outcomes of first-episode psychotic patients treated with Open Dialogue Therapy in Western Lapland, a protocol that avoids immediate use of antipsychotics. Two thirds of their patients have not been exposed to antipsychotics at the end of five years, and are asymptomatic and functioning well. However, Lieberman and colleagues do not cite this study, even though it provides evidence, consistent with earlier studies, that 60% of first episode patients, when provided with psychosocial care, can recover without the use of antipsychotics.
- Duration-of-untreated psychosis studies
To make their argument that initial treatment with antipsychotics leads to better long-term outcomes, Lieberman and colleagues cited a meta-analysis of 33 studies that found “modest correlations between a longer duration of untreated psychosis and residual positive and negative symptoms and impairments in social functioning.” They also noted one “quasi-experimental study” that found a reduction in “the gap between onset of psychotic symptoms and treatment with antipsychotics was associated with improved long-term outcome,” with this study lasting two years.
What’s missing from this review
These duration-of-untreated psychosis studies do not compare initial treatment of a psychotic patient with an antipsychotic to a no-medication regime. All of the patients in these studies are treated with antipsychotics; the difference is that one group has been ill a shorter time than the second group before getting into treatment.
As Sandy Steingard wrote in an MIA blog several years ago, there isn’t a clear picture, from these studies, of whether shortening the time of psychosis before treatment with an antipsychotic produces a meaningful difference over longer periods of time. The results have been conflicting. But researchers did find that a longer delay in patients getting psychosocial care led to an increased likelihood that patients would be suffering from negative symptoms at six years.
Early treatment with psychosocial care may produce a benefit, and perhaps that early treatment would be even greater if combined with the absence of neuroleptic treatment.
Deconstructing their Review, Part Two
In the second part of their study, Lieberman and colleagues focus on the “long-term effects of antipsychotic medication on clinical course in schizophrenia.” They reviewed longitudinal studies, evidence on whether antipsychotics shrink the brain, and concerns that antipsychotics induce a dopamine supersensitivity that make patients more biologically vulnerable to psychosis over the long term.
A. Longitudinal studies
This is the best prospective, longitudinal study that has been conducted in the modern era. Harrow, with funding from the NIMH, followed a large group of psychotic patients, including 64 diagnosed with schizophrenia, for more than 20 years, and periodically assessed their outcomes, with patients grouped according to their use of antipsychotics. Here is what Lieberman and colleagues wrote—in entirety—about this study, which they lumped together, in the same sentence, with their report on a Finnish longitudinal study by Moilanen.
The two studies, they wrote, “found superior outcomes in individuals who were previously treated but not taking medication at the time of assessment compared with individuals taking medication. However, nonmedicated patients had more favorable premorbid characteristics in the study by Harrow and colleagues and had less severe illness in the study by Moilanen and colleagues, suggesting that in naturalistic uncontrolled studies, medication status may in part be the consequence of whether patients are doing well or poorly rather than the cause.”
In short, they acknowledge that the outcomes were “superior” in the Harrow study and also in Moilanen’s, but dismiss these outcomes by stating that the unmedicated patients were less severely ill at baseline.
What is missing/wrong with their review of Harrow’s study
There is no other longitudinal study in the literature that rises to the level of Harrow’s in terms of providing rigorous, comprehensive data on the different long-term course of medicated and unmedicated psychotic patients. He enrolled 200 psychotic patients into the study, from two Chicago-area hospitals (one public, and one private), and they were a young group, with a mean age of 22.9 years. Nearly half were first-episode patients, and another 21 percent had only one previous hospitalization. All were treated in the hospital conventionally with medications, and after they were discharged, Harrow conducted regular follow-ups that charted their medication use, their symptoms, and their functioning. He divided his patients into different subgroups: good-prognosis and bad-prognosis schizophrenia patients, and patients with milder disorders. At the end of 15 years, he still had 77% of the 200 patients in the study, which is an extraordinary feat. In drug studies, the drop-out rates in even short-term trials will be much higher than that.
Yet, Lieberman and colleagues, in a paper said to investigate the long-term effects of antipsychotics, don’t present any details that would tell of this study’s rigor, and they don’t present any of the outcome data. Moreover, their explanation for the unmedicated patients —that it was because they had a better initial prognosis—is belied by Harrow’s data, and the reason that this can be shown is precisely because Harrow divided the patients into groups with different prognoses.
There was a dramatic difference in outcomes, across all domains, for the unmedicated patients, which showed up beginning with the 4.5-year follow-up and remained during all of the subsequent assessments. For those with a schizophrenia diagnosis, the long-term recovery rate for the off-med group was 40% versus 5% for the medicated group. The unmedicated patients were much more likely to be working, they scored better on cognitive tests, and they had less anxiety at every follow-up assessment. Perhaps most notable of all, they were much less likely to have psychotic symptoms at each follow-up, starting with the 4.5-year assessment. Here are the results that Lieberman and colleagues failed to present and discuss:
Having failed to present this data, Lieberman and colleagues then dismiss the “superior outcomes” for the unmedicated patients by stating that this group had more “favorable premorbid characteristics.” There is an interesting history behind this explanation for the difference in outcomes, which psychiatry has clung to even after it was revealed to be false.
Harrow, in his 2007 report, did attribute the better outcomes in the unmedicated group to a difference in premorbid characteristics, as it was the good-prognosis patients who were more likely to go off medication. However, that explanation was belied by this fact: the good prognosis schizophrenia patients who got off did better than the good prognosis schizophrenia patients who stayed on, and that was true for all other subgroups (bad-prognosis patients who got off did better than the bad-prognosis patients who stayed on, and that was the same for those with milder psychotic disorders.) Most compelling of all, schizophrenia patients who stopped taking antipsychotics did better than those with milder disorders who kept on taking the medications.
In his more recent publications, Harrow has acknowledged that this is so. In every subgroup, the on-med groups had notably worse long-term outcomes. In a 2013 article, he summed up his thoughts in this way:
“How unique is that the apparent efficacy of antipsychotics could diminish over time or become ineffective or harmful? There are many examples for other medications of similar long-term effects, with this often occurring as the body readjusts, biologically, to the medications.”
As noted above, Lieberman and colleagues lump this study in with Harrow’s and dismiss it in the same sentence, stating that the unmedicated patients were less severely ill at baseline.
What is missing/wrong with this review
In this study, Finnish researchers identified 70 patients born in 1966 who were diagnosed with schizophrenic psychoses. They assessed the status of the patients in 2000, when they were 34 years old (with a mean duration of illness of 10.4 years.) At that time, the 24 patients off medication were doing much better than the 46 patients on antipsychotics: they were more likely to be working, more likely to be in remission, and had better clinical outcomes. In short, it wasn’t that the non-medicated group in this study were “less severely” ill when initially diagnosed, but rather that a decade after initial diagnosis, those off medication were doing much better than those on the medication.
The researchers then followed the 70 patients for another nine years, and those who used antipsychotics less than 50% of the time were more likely to function well, be in remission, and have a good clinical outcome.
As was the case with their presentation of the Harrow study, Lieberman and colleagues didn’t report the data, and they explained the “superior outcomes” away with a claim—that the unmedicated group was less severely ill—that is not found in the study itself.
There is one other longitudinal study that provides long-term outcome data based on the patients’ use of antipsychotics. This one is a Danish study of 496 first-episode schizophrenia-spectrum patients diagnosed from 1998 to 2000. At the end of 10 years, there were 303 patients still in the study, 121 of whom were deemed “non-compliant” and off medication. There were no differences at baseline between the non-compliant and on-medication groups at 10 years.
Here are their 10-year outcomes:
This study was not cited by Lieberman and colleagues.
In their review of longitudinal studies, Lieberman and colleagues spent one sentence on the Harrow and Moilanen studies, didn’t cite the Danish study, and then wrote that “other naturalistic studies have found improved outcomes in individuals with schizophrenia compared with those who did not.” They gave no details of these other naturalistic studies, but cited two reports by Tiihonen and his Finnish colleagues.
What is wrong with this review
The Tiihonen articles are not outcome studies at all, and they are not studies of specific known patients. They are both reports by a Finnish group that mined a national registry of drug prescriptions to report relapse and mortality rates for psychotic patients, organized by their current antipsychotic use.
In the first study, they identified 2230 adults hospitalized for a first episode of schizophrenia from 1995 to 2001, and then charted their medication use for every 30-day period following discharge (based on a national database for community prescriptions of antipsychotics). Any relapse or death during a 30-day period was chalked up to outcomes for the particular antipsychotic they were on, or chalked up to “not antipsychotic drug” if they didn’t fill a prescription during that month. Given this methodology, anyone who came off a medication and relapsed would show up in the outcomes for “no antipsychotic drug,” and if someone committed suicide after discontinuing a medication, this too was chalked up to the outcomes for “no antipsychotics.” They found that there was a higher risk of relapse during months of “no medication,” which, as much as anything, may be due to the high risk that comes with being placed on an antipsychotic and subsequently trying to come off it.
In the second study by this same group, they reported on mortality rates of patients with schizophrenia based on length of exposure to antipsychotics over a 11-year period. The researchers reported that “long-term treatment with antipsychotic drugs is associated with lower mortality compared with no antipsychotic use.” They did so by mining the same database of community prescriptions of antipsychotics.
This study, while not an outcome study, has been cited as evidence that long-term use of antipsychotics reduces mortality in schizophrenia. However, as critics of this study have found, deaths of hospitalized patients were not counted, which led to an exclusion of “64% of deaths on current antipsychotics.” In addition, the group with the lowest mortality rate, by far, were those who used antipsychotics less than six months over the 11-year period, as opposed to those who used antipsychotics regularly.
As such, the Tiihonen studies cannot be described as “naturalistic studies” that found “improved outcomes” in medicated patients. They weren’t studies of known patients; they didn’t chart outcomes over longer periods of time (severity of symptoms, cognition, and functional measurements); and there is reason to argue that mortality rates were lowest in those patients who used antipsychotics for only a short time.
B. Brain shrinkage
Lieberman and colleagues review a number of studies regarding the effects of antipsychotics on brain volumes, and whether such changes are associated with clinical worsening, cognitive decline and functional impairment. The conclusion they came to can be summarized in this way:
- It’s not clear whether the brain shrinkage seen in schizophrenia patients over time is due to the drug or the disease.
- It’s not clear whether this loss of gray matter in schizophrenia patients is associated with clinical worsening, cognitive impairment, or functional impairments.
- The matter needs further study.
What’s missing from their review
An effort to assess the long-term effects of antipsychotics requires, most of all, seeing if evidence of different types fits together. The question of their long-term effects arises from the longitudinal studies: why do the unmedicated patients in Harrow, Molainen, and Wils have better long-term outcomes, and on every domain that is measured? The conventional belief has been that antipsychotics are an essential treatment, and yet the longitudinal data belies that belief. This leads to a search for research that might provide a biological explanation for the outcomes seen in the longitudinal data.
The brain-shrinkage data needs to be viewed in that light. What does research show? There is evidence from multiple researchers, in the U.S. and abroad, that antipsychotics shrink the brain, and that this shrinkage is dose related. The same neurotoxic effect can be seen in studies in rats and monkeys, and in those animal studies, the shrinkage cannot be attributed to the disease. And as Joanna Moncrieff noted in her blog in response to Lieberman’s article, there “are no studies that show progressive brain changes in people diagnosed with schizophrenia or psychosis in the absence of antipsychotic treatment.”
There are also studies that have found that this shrinkage correlates with increased positive symptoms, increased negative symptoms, and declines in cognition and functioning. This is data that completes the picture: Longitudinal studies tell of worse outcomes for medicated patients, and MRI studies tell of drugs that shrink brain volumes, and that this shrinkage is associated with worse outcomes. The two data sets combine to tell a coherent story.
In her blog, Moncrieff provides a detailed critique of the review by Lieberman and colleagues of the brain-shrinkage literature. But their biggest failure in presenting this research could be said to have begun in the earlier part of their paper, when they failed to present the outcomes from the three longitudinal studies, and pretended that Tiihonen’s two published articles represented naturalistic studies that found better long-term outcomes for medicated patients. They never presented the longitudinal data that fits together with the brain-shrinkage research to present such compelling evidence that long-term use of these medications worsens aggregate outcomes.
Yet, even though they failed to present the longitudinal data, their review of the brain-research still led them into an intellectual quagmire. How could it be that the reduction in brain volumes seen in psychotic patients was related entirely to the disease and not at all to the drugs, when the shrinkage has also showed up in animals treated with the drugs? They came up with this explanation: “The relevance” of such findings “is uncertain, due to species related differences and because animals lack the pathophysiology of schizophrenia.”
With that explanation, they were setting up the notion that drugs that shrink the normal brain don’t affect the “psychotic” brain in the same way, and this led them to speculate about whether the reduction in brain volumes observed in schizophrenia patients, even if due to the drug, is a bad thing. Yes, in normal people the loss in gray matter correlates with a drop in IQ, but perhaps it has the opposite effect in people with psychotic diagnoses. “It is possible that antipsychotics may have deleterious effects on normal brain but protective effects in the presence of schizophrenia-related neuropathology.”
C. Dopamine supersensitivity
In the late 1970s, Lieberman and colleagues write, the concern arose in psychiatric research circles that antipsychotics induced a dopamine supersensitivity—a change in the density of dopamine receptors—that could make patients more vulnerable to relapse than they would be in the normal course of their illness. This could explain why relapse rates are so high in withdrawal studies, and why so many patients treated with antipsychotics become chronically ill.
Lieberman and colleagues do acknowledge that antipsychotics induce a “rapid increase in D2 receptor density.” But, they said, two withdrawal studies provide reason to doubt that this is a problem. In one, the speed of tapering (abrupt vs. gradual) did not produce different relapse rates for the withdrawn patients, and in a second report, a meta-analysis of studies that had charted relapse following withdrawal, the investigators found that withdrawn patients continued to relapse at a fairly steady rate throughout the first year, and while so did the medicated patients, the difference in the relapse rates between the two groups was sustained throughout the twelve months.
With such withdrawal literature as their guide, they concluded that “clinical studies have not provided compelling evidence that treatment with antipsychotics worsens the course of illness, increases risk of relapse, or causes cognitive deficits.”
What’s missing with their review
As was the case with the brain-shrinkage issue, any investigation of drug-induced dopamine supersensitivity needs to start with outcome data that raised the concern in the first place. In the 1970s, reports of higher long-term relapse rates for medicated patients, as compared to patients not treated with antipsychotics, which were found in the studies by Schooler, Bockoven, Rappaport, Mosher, and Carpenter, presented the field with a conundrum. The withdrawal studies they had conducted showed that patients withdrawn from the drugs relapsed at a much higher rate than the drug-maintained patients. This supported a conclusion that the drugs reduced the risk of relapse. But the outcome data told of higher relapse rates for the medicated patients. How to explain this conundrum?
This led Carpenter to ask a profound question: What if the patients had never been put on drugs to begin with? Perhaps the drugs induced a biological change that made patients more biologically vulnerable to psychosis. This could lead to an increased risk of relapse following drug withdrawal, which would show up in the withdrawal studies. Yet, if patients stayed on the medications, drug-induced dopamine supersensitivity could lead to an increased risk that they would become chronically ill. This would explain the poor results for the medicated patients in the longitudinal studies. The withdrawal studies may not have been measuring the return of a disease, but the risk of going off a drug after the patient’s brain has been changed by it.
Chouinard and Jones fleshed out a biological explanation for what appeared to be going on. The drugs blocked dopamine receptors, and in response to the blockade, the density of dopamine receptors increased, and this made patients more biologically vulnerable to psychosis. They conducted studies in their patients that confirmed their theory, and they concluded that this drug-induced change could lead to more “severe symptoms” than the patients had ever experienced before.
That was the research that was published in the early 1980s, and the modern longitudinal studies, which found higher remission rates in unmedicated patients over the long term, provide new evidence to support the conclusion that drug-induced dopamine supersensitivity increases the risk that a person will become chronically ill (see Harrow, Moilanen, and Wils). To specifically investigate this question, Harrow reported on psychotic symptoms over a 20-year period for those who were medication compliant throughout the study and those who never took any medication from year two onward. Here are the results:
That is the story of dopamine supersensitivity. But Lieberman and colleagues did not present this historical context, and they don’t mention the modern longitudinal data, or Harrow’s investigation of it. They didn’t present the research that makes a compelling case that drug-induced dopamine supersensitivity may worsen long-term outcomes.
All is Well in the Land of Psychiatry
What can be seen here, in this deconstruction of the review by Lieberman and colleagues, is that they presented information, time and time again, in a way that protects guild interests and their current protocols for prescribing antipsychotics.
- They never provide data from the studies showing that 60% or so of first-episode patients may recover without the use of antipsychotics.
- They always dismiss the better outcomes for unmedicated patients in cited studies, arguing that it is an artifact of an unequal comparison for some reason or another (Schooler, Rappaport, Harrow, and Moilanen).
- They report no data from modern longitudinal studies that tell of much better long-term outcomes for the unmedicated patients.
- In their discussions of drug-induced brain shrinkage and dopamine supersensitivity, they fail to discuss information from the larger body of scientific literature essential to assessing whether these drug effects could explain the poor long-term outcomes seen in the longitudinal studies.
Having reviewed the literature in that guild-protective manner, Lieberman and colleagues then drew these conclusions:
- There is “little evidence” that initial use of antipsychotics or maintenance treatment with the drugs have a “negative long-term effect.”
- There are just a “small number” of patients that may “recover from a first episode of psychosis without pharmacologic treatment or may discontinue medication and remain stable for extended periods of time.”
- Randomized clinical trials (Leucht) and drug-withdrawal studies “strongly support the efficacy of antipsychotics for the acute treatment of psychosis and prevention of relapse.”
They were an “international group of experts,” and they had come to a comforting conclusion for the guild: The drug-use protocols the profession has been using for decades are just fine.
The Press Release (and public denunciation of the critics)
In our book Psychiatry Under the Influence, Lisa Cosgrove and I wrote about the duty of a medical profession to honestly assess and report on the evidence for its treatments in medical journals, and to communicate such information in an accurate manner to the public. The article published by Lieberman and colleagues in the American Journal of Psychiatry served as their communication to the medical profession; a May 5 press release by Columbia University Medical served as their communication to the public.
The press release deserves to be deconstructed too, for it ratchets up the telling of a narrative that supports guild interests. The authors are described as an international group of experts, who conducted this review to respond to critics who were raising concerns about antipsychotics that had the potential to “mislead some patients (and their families),” and thus cause them “to refuse or discontinue antipsychotic treatment.” After conducting a “comprehensive examination” of all possible evidence, they made a definitive finding: “For patients with schizophrenia and related psychotic disorders, antipsychotic medications do not have negative long-term effects on patients’ outcomes or the brain. In addition, the benefits of these medications are much greater than their potential side effects.”
If anyone dared to doubt this finding, Lieberman said in the press release, they “should talk with people whose symptoms have been relieved by treatment and literally given back their lives.”
The drugs were once again cast as essential and even as miracle agents (life-savers), and this became the narrative promoted to the public, and was repeated in stories published by UPI and on such websites as Science Daily and Medical News Today. Meanwhile, Lieberman also touted the importance of the study in a Medscape video, describing it as a “major development in psychiatry.” Dressed in a white coat, he said the study “came to a very firm conclusion as to the unquestionably positive benefit of these drugs,” and then he took aim at those who had dared to question their merits in the first place.
“The critics who gave rise to this notion that antipsychotic treatment adversely affects long-term outcomes were sowing seeds of untruth and, in their pursuit of some ideological goal or need for self-serving acclamations, were ignoring entirely the harm that they were causing many people who unwittingly would accept these as credible statements and follow their guidance. As I film this video, the contents of the new article are embargoed, so I cannot reveal the contents just yet. But let me just say that this is an extremely well-composed paper that provides a scholarly and rigorous review of relevant lines of evidence. It comes to a very clear and definitive conclusion that we all should take note of the findings, apply them in our clinical practices, and use them where need be in the education of patients and against the reputation of individuals who are really trying to create mischief for their own nefarious purposes.”
Such were the communications of Lieberman and the Columbia University Medical Center, and a few days ago I received an email from a friend in Iceland, who has been an activist in the user community for some time, that neatly summed up their effects on societal beliefs.
“I was talking to some doctors the other day,” she wrote. “They were very excited about a newly published article. Many doctors from many countries had contributed to this research and the bottom line was that meds are good and they are surely helping.”
The Harm Done
In his video, Lieberman talks about critics “sowing seeds of untruth” and how such “untruths” can cause harm, and I have to agree that sowing seeds of untruth can cause harm. We can see it so clearly in this case of institutional corruption.
The studies that tell of 60% of unmedicated first-episode patients recovering, and of better long-term outcomes for unmedicated patients, speak of an opportunity for psychiatry to grasp: they could change their protocols and give a chance to people who suffer a psychotic episode to recover and get on with lives unburdened by the many adverse effects of antipsychotics. There is an “evidence base” that tells of new possibilities for people so diagnosed.
But Lieberman and colleagues did not present that possibility in this review. Rather they hid it from view. That is an action that does harm to millions of “patients” and their families, and thus to all society. We will continue to live in a society organizing its care—and its laws regarding psychotic patients—around a false narrative, one told to serve guild interests, rather than the best interests of patients.
The Challenge for Our Society
In a study of institutional corruption, the ultimate goal is to present ideas for solving the corruption. Lisa Cosgrove and I admittedly struggled with this section of our book. While problems in psychiatry have become well known to our society, societal focus has been on curbing pharma’s influence over psychiatry. But how can the influence of its guild interests be curbed?
I really don’t know. The problem is that the power lies with the guild and its academic psychiatrists, who pen articles such as this one. They have the standing in society as experts; their papers are published in “medical journals”; and they have access to the press. Mad in America is meant to serve as a forum for critiquing that conventional narrative, but I am pretty sure that psychiatrists in Iceland will not soon be talking about the “untruths” sowed by Lieberman’s article in the American Journal of Psychiatry.
But I do have one wish. I wish that all psychiatric residents would familiarize themselves with this controversy, and read the research articles that have been cited, and then ask themselves: Is this published report, the ensuing press release, and Lieberman’s video the work of a medical profession they are proud to join? Or are they the work of a medical profession that needs to be thoroughly remade, with this remaking to be their gift to the mental health of people everywhere? That could be quite a legacy for a new generation of psychiatrists.