Do Antipsychotics Worsen Schizophrenia in the Long-Run?

Randomization, association, and the arrow of causality.

Posted Aug 08, 2014

Puberty, Edvard Munch

A common question from patients with a new diagnosis of schizophrenia who have responded to initial treatment with antipsychotic medications is whether they must continue taking them for the rest of their lives. On this issue, most psychiatrists hold the conservative view that yes, most patients with schizophrenia should take their medications for the rest of their lives in order to prevent relapse. If a patient does opt to discontinue their medications, it is recommended that this be done slowly and cautiously, under close medical supervision.

In recent years however, opinions to the contrary have been floated in the popular press, giving the impression of a growing sense that antipsychotic medications may not be necessary in the long run and that they might even make things worse. Last year, an op-ed piece appeared in The Washington Post with the provocative title, “A psychiatrist thinks some patients are better off without antipsychotic drugs.”1 The author, psychiatrist Sandra Steingard, described how her own conservative view on long-term antipsychotic use shifted dramatically after she read Anatomy of an Epidemic by Robert Whitaker.2 After reading the book and reviewing the studies cited in it, she came to agree with Whitaker’s view that people with schizophrenia who stay on antipsychotic medications do worse than those who discontinue them.

Whitaker, a science writer by trade, has built a name for himself arguing this position while peddling his award-winning book through his own website ( and blog ( and in numerous media interviews (see for example a recent interview from CBC Canada here). To his credit, Whitaker has performed an exhaustive review of long-term follow-up (“longitudinal”) studies of schizophrenia – he knows the literature. And yet, his various arguments have been refuted by respected voices from within academic psychiatry including E. Fuller Torrey, Ronald Pies, Daniel Carlat, and Andrew Nierenberg.3-6

Rather than simply “pile on” to these responses, I want to carefully examine the specific claim – Whitaker’s or otherwise – that antipsychotic medications worsen schizophrenia. For the record, I did read Anatomy of an Epidemic when it came out back in 2010.  But while my initial reaction of surprise and curiosity was similar to that of Dr. Steingard’s, my own subsequent conclusions were rather different after I’d reviewed the data. Here’s why.

To begin with, I’m a psychiatrist who specializes in the treatment of hospitalized patients with schizophrenia. Perhaps to some, that might mean that I’m inherently biased by the psychopharmaceutical industrial complex, but it also means that Monday through Friday for the past 15 years, I’ve spent my days directly observing the effects of antipsychotic treatment in patients under my care. While some psychiatrists find hospital work with severe mental illness daunting, I often say that my job is relatively easy because the single most common reason that patients with schizophrenia end up in the hospital is that they discontinue their medications. I’ve seen countless patients who were doing well on their medications as an outpatient experience a relapse in the wake of medication discontinuation, only to regain stability after simply restarting them in the hospital. After discharge, when patients later stop their medications again, they find themselves back in to the hospital. It’s a cycle all too well known to any clinician working with patients who suffer from schizophrenia. And yes, they suffer – let’s dispense with any romantic notions that those with schizophrenia are just eccentrics that have trouble conforming to society’s oppressive standards of normalcy.

But determining the effect of antipsychotic medication on psychotic symptoms and relapse requires more than personal observation and experience, which is subject to bias. That’s why we have hundreds if not thousands of clinical trials comparing antipsychotic treatment to placebo in the treatment of acute psychotic symptoms. Though there have certainly been “negative” studies along the way – typically those in which placebo response rates were high and similar to improvements with medications – the bulk of these studies dating back to the 1950’s have shown that on average, antipsychotics consistently offer greater improvement of troubling psychotic symptoms such as delusions and hallucinations compared to treatment with placebo. Numerous studies have also demonstrated that when patients who are stabilized on antipsychotic medications stop taking them, they relapse at a greater rate within 1 to 2 years than those who continue taking medications. Even Whitaker doesn’t dispute that antipsychotics are valuable for symptom control in the short-term.

Given this demonstrated antipsychotic effect, it is ethically difficult to justify longer placebo-controlled experiments, just as it was ethically wrong to withhold treatment for syphilis in the infamous Tuskegee Experiment that is now required study in bioethics training. But a handful of longer-term studies have nonetheless been performed through the years and it is these studies that Whitaker uses to frame his position that antipsychotics worsen schizophrenia in the long run.

Martin Harrow and colleagues have published two papers that examined 139 patients with schizophrenia in the Netherlands who were either on or off antipsychotic medications over a 20-year study period.7,8  In both analyses, Harrow found that those not on antipsychotics had a lower severity of psychosis and significantly greater rates of recovery compared to those taking antipsychotics. This surprising finding led Harrow to suggest a “recovery paradox” in which antipsychotics help in the short-term, but may lose efficacy in the long-term. 

However, many are familiar with the phrase “association doesn’t prove causality” and that certainly applies here. If medicated patients fared worse in the long run, could there be another explanation other than a paradoxically toxic long-term effect of medications? 

In order to explore this possibility, we need to briefly discuss study design. In a typical short-term study of any medication compared to placebo, various “controls” are used as a part of the scientific method to minimize other factors that might contribute to an effect. For example, as I discussed in a previous blogpost, outcomes are rated in a “blinded” fashion to avoid the bias of observer expectation.9 Likewise, every effort is made to equally balance factors potentially contributing to treatment response (e.g. baseline symptoms, duration of illness, socioeconomic or marital status, etc.) between study groups. Most often, this balance is achieved by simple “randomization” such that the assignment of each study subject to one of the two treatment groups is based purely on chance. If group assignment is randomized such that there is an equal likelihood of being assigned to medication or placebo, then the other contributing factors tend to even out. This is why, for clinical outcomes, the “randomized, double-blind, placebo-controlled study” is regarded as the “gold-standard” for study design. When I was learning research methods as a budding clinical trials investigator, I was taught that randomization was the most important starting point of a controlled study. Without it, you can’t be confident that observed differences are due to the treatments being compared.

The studies by Harrow were not randomized trials. Whether a patient took or didn’t take antipsychotic medications was simply a personal choice such that in the long run, those that were able to remain off antipsychotic medications for 20 years were a “self-selected group.” This raises the very real possibility that the unmedicated group – who were not “cured” of schizophrenia and continued to experience mild psychotic symptoms – might have had a less malignant form of the illness such that medications were not absolutely necessary. And it’s still possible that if the unmedicated subjects had taken antipsychotic medications, they might have done even better than they did. It’s also possible that if medicated subjects had not been on antipsychotics, they would have done even worse. But these possibilities were not tested in the studies.

Without randomization, it’s difficult to judge the direction of the “arrow of causality” for the association between antipsychotics and poorer long-term outcomes in the Harrow studies. That is, instead of the conclusion that antipsychotics caused worsening in medicated patients, it’s just as likely that certain characteristics of a patient’s individual illness such as the type or severity of their symptoms determined whether or not someone needed to take antipsychotic medications. Ever since the term “schizophrenia” was coined a hundred years ago, it has been regarded as a constellation of different illnesses with distinct causes rather than a single disease with a single cause. Therefore, some patients have forms of psychosis that resolve spontaneously and those with chronic symptoms can have markedly different clinical courses. It makes sense then that those with the more refractory forms of schizophrenia will both tend to be on medications as well as fare less well than those who aren’t. Based on clinical experience, this is an obvious explanation for the data that doesn’t require entertaining a paradox about antipsychotic medications.

When counseling patients about taking antipsychotics in the long run, psychiatrists often invoke a comparison to diabetes, in which medications like insulin are usually required indefinitely. However, some patients with type II diabetes are able to lose weight and restore their body’s insulin function so that they don’t need medications any longer. And yet, no one is claiming that insulin makes diabetes worse, even though increasing doses are usually needed as the pancreatic islet cells that make the body’s own insulin deteriorate during the natural course of the illness. It’s the same situation with high blood pressure – long-term medication is dictated by persistence of the condition rather than some toxic effect of antihypertensive medications.

And so, despite the ethical risk, what we really need to answer the question of whether patients with schizophrenia should stay on antipsychotic medications indefinitely is a randomized, controlled trial of antipsychotics in the long-term treatment of schizophrenia. 

No surprise then that when just such a randomized study was published last year in JAMA Psychiatry, it was met with much anticipation and fanfare. In this study, Wunderink and colleagues followed 257 patients with psychosis (not just schizophrenia) after 6 months of initial treatment with antipsychotic medications.10 Those patients who responded to treatment were randomized to continuation of medication or drug reduction/discontinuation for 18 months11 and later up to 7 years. Dose reduction/discontinuation involved decreasing the dose of medication “if feasible,” but increasing the dose if symptoms re-emerged. At the 18-month mark, dose reduction/discontinuation resulted in significantly greater relapse rates compared to medication maintenance, but at 7 years, the results flipped, with greater rates of functional remission (defined as having a low score on a scale that measures social functioning) as well as both functional remission and symptomatic remission together (having a low score on the social functioning scale and another scale that measures psychotic symptoms) in the dose reduction group. There were no differences in symptomatic remission alone between the two groups.

Equipped with the results of a randomized study, Whitaker has since bolstered his argument regarding the apparent toxicity of antipsychotics.12 No surprise there. But in discussing the Wunderink study, Thomas Insel likewise wrote in an op-ed piece that “antipsychotic medication… appeared to worsen prospects for recovery over the long-term.”13 Quite a damning comment considering that Insel is not only a psychiatrist, but the director of the National Institutes of Mental Health.

But does the Wunderink study really demonstrate anything about antipsychotics worsening schizophrenia? Examining the study more carefully, it should be understood that most of the subjects in the dose reduction/discontinuation arm of the study actually remained on antipsychotic medication, albeit at a reduced dose. Some discontinued and had to be restarted. In addition, for subjects in the medication maintenance group, the dose of medication was also decreased whenever possible such that some of them ended up completely off medications as well. In this way, the study was not so much a comparison of being on or off medications, but a comparison of being on higher versus lower antipsychotic doses. Lower doses won out – a time honored observation that has consistently led to longstanding clinical recommendations to use the lowest possible effective antipsychotic dose in the treatment of schizophrenia.

But while the initial treatment group allocation was randomized, the subsequent dose changes in both treatment groups were based on clinical response and occurred at the whim of the treating psychiatrists. Therefore, in the final analysis that indicated that those subjects who successfully decreased or discontinued their antipsychotic medication regardless of initial group assignment had significantly greater rates of symptomatic remission, functional remission, and recovery compared to those that remained on higher doses, these outcomes were not based on random group assignment. In this way, the Wunderink study wasn’t actually a randomized controlled trial after all.

Therefore, rather than supporting a paradoxical hypothesis of long-term antipsychotic toxicity, the results of the studies by both Wunderink and Harrow could be better explained by hypothesizing that differences in illness severity or symptom type determine to what extent long-term antipsychotic medication is necessary.  For example, in addition to antipsychotic status, the Wunderink study found that less severe negative symptoms (e.g. amotivation, asociality, alogia, etc.) and not living alone were associated with recovery. Such factors may help to predict who will need antipsychotic medications in the long run.

If the association between antipsychotic treatment and poorer outcome can be explained by two opposing directions of causality, how does one decide what to believe? In such cases, we tend to fall victim to something called “confirmation bias” – interpreting the data based on our own preconceived notions. Whitaker, who has been called a “darling of the antipsychiatry movement14 and has established a successful career on the premise that antipsychotics worsen schizophrenia, is likely to stick to his guns. Just the same, psychiatrists who have witnessed both short-term and long-term improvement on antipsychotic medications, as well as inevitable relapses when patients stop their medications, are going to stick to ours.

Perhaps we’ll have to await a real randomized controlled trial to answer the question more clearly. But in the meantime, a balance viewpoint seems appropriate.

While antipsychotic medications represent an important cornerstone in the treatment of schizophrenia, we desperately need safer and more effective medications. These medications are palliative rather than curative, such that even a good response often leaves a patient with residual symptoms, functional impairment, and a variety of side effects that range from merely annoying to life-threatening. This helps to explain why, randomization issues aside, those on antipsychotic medications did not fare all that well in the long-term studies discussed above. 

Also, while many clinicians who face these limitations in the effectiveness of antipsychotic medications often tend to raise the dose of medications out of desperation, erring on the side of lower doses and reducing polypharmacy (the use of multiple antipsychotics together) is probably best for most patients.

In addition, it is now well established that psychotherapy and other “psychosocial” interventions such as supported employment should be integrated into the treatment of schizophrenia in order to optimize recovery.  It is no longer sufficient for most patients with schizophrenia to be seen by a psychiatrist a few times a year for 15-minute “med checks.”  

Finally, both clinical experience and long-term studies suggest that some patients can in fact do without long-term antipsychotic medication. In fact, I have stopped antipsychotic medications on numerous occasions without subsequent difficulties particularly when, after careful review, I have questioned a patient’s previous diagnosis of schizophrenia. Such patients with questionable schizophrenia do end up in research studies by the way, so that it’s understandable that non-randomized studies support the effectiveness of medication discontinuation in self-selected patients.

Of course, every patient wants to have the kind of schizophrenia that doesn’t require long-term medications, just as every patient wants to not have schizophrenia at all. Unfortunately, that hope doesn’t predict the benefit from and need for antipsychotic medication. Neither does the unfounded belief that antipsychotics cause a worsening of schizophrenia in the long run.



1.  Steingard S. A psychiatrist thinks some patients are better off without antipsychotic drugs.  The Washington Post, 12/6/2013.

2. Whitaker R. Anatomy of an Epidemic:Magic Bullets, Psychiatric Drugs, and the Astonishing Rise of Mental Illness in America. New York, Crown Publishers, 2010.





7.  Harrow M, Jobe TH, Faull RN.  Do all schizophrenia patients need antipsychotic treatment conttinously throughout their lifetime?  A 20-year longitudinal study. Psychological Medicine  2012; 42:2145-2155.

8.  Harrow M, Jobe TH, Faull RN.  Does treatment of schizophrenia with antipsychotic medications eliminate or reduce psychosis?  A 20-year multi-follow-up study.  Psychological Medicine (online March 2014)


10.  Wunderink L, Nieboer R, Wiersma D, Sytema S, Nienhaus FJ.  Recovery in remitted first-episode psychosis at 7 years of follow-up of an early dose reduction/discontinuation or maintenance treatment strategy.  JAMA Psychaitry 2013; 70:913-920.

11. Wunderink L, Nienhuis FJ, Sytema S et al. Guided discontinuation versus maintenance treatment in remitted first-episode psychosis: Relapse rates and functional outcome. J Clin Psychiatry 2007;68:654-661.