Martin Harrow along with his colleagues T.H. Jobe and R. N. Faull has published another paper on the long-term outcome of people who experienced a psychotic episode. Funded by a grant from the Foundation for Excellence in Mental Health Care, this paper adds to our knowledge of an extremely important and valuable study.
In this blog, I am going to first review some material that may be familiar to many of you. I will then talk about this latest paper.
In the late 1980’s Harrow began the Chicago Follow-up Study. A group of 139 individuals were assessed during a hospitalization for psychosis. They were then assessed at 2, 4.5, 7.5, 10, 15 and 20 years. The sample included 70 individuals who met the diagnostic criteria for schizophrenia spectrum disorders and 69 individuals diagnosed with mood disorders. The sample was comprised of consecutively admitted patients to two Chicago hospitals who were considered in the early phase of their disorder; for 41% this was a first admission and for 25% it was the second admission.
Of those 70 diagnosed with schizophrenia spectrum disorder (and I will use the notation SZ as they do in the paper to refer to this group), they evaluated 59 at 20 years. The other 11 were evaluated at all time points up through 15 years.
This is considered a naturalistic study. No randomization occurred. The researchers did not offer treatment. They were interested in what happens after a person has a hospitalization at a relatively young age (mean age at initial hospitalization was 23).
In earlier papers, Harrow has discussed general outcome including ability to work and quality of relationships. He has found that many people go on to do fairly well. Robert Whitaker highlighted this study in his book, Anatomy of an Epidemic because of the fact that the individuals with an early schizophrenia diagnosis who did not remain on neuroleptic medication did much better than those who did stay on the drugs. In fact, they did better than those diagnosed with a mood disorder who stayed on the medication even though the prognosis for individuals diagnosed with a mood disorder is thought to be better than those diagnosed with schizophrenia. This was an important part of Whitaker’s general argument that there was accumulating evidence that long-term exposure to neuroleptic drugs worsens outcome.
Those who are critical of Whitaker’s interpretation have generally said that he confused correlation with causation. They point out that it is possible that the individuals who stopped the drugs had a condition that had a more favorable outcome and that the drugs were stopped BECAUSE they were doing better. They point out that we have always known there was a cohort of individuals who will recover from an initial psychotic episode and it would make sense that we would see some people over 20 years off medications and doing well. This study alone, they say, does not prove that the drugs worsen outcome.
While these critics misinterpret Whitaker, who clearly understands the distinction between causation and correlation, their points about alternative explanations of these findings are valid. Whitaker asserts, however, that while not conclusive these findings are nevertheless worrisome and he wrote about other studies that supported his hypothesis. We now have the Wunderink study, a randomized control trial comparing continuous drug treatment to intermittent symptom-focused treatment, and the results were similar; those who were on drugs intermittently and only when symptoms were present overall did much better than those who remained on the drugs continuously.
So what is new about this paper? Harrow, Jobe and Faull looked at the presence of psychotic symptoms (hallucinations, delusions, disorganized thought) over the 20 years and compared them in three groups: those on drugs at every follow up visit (group 1), those on drugs on some but not all visits (group 2), and those on no drugs at any of the visits starting at the 2 year point (group 3). Of the SZ individuals there were 25 in group 1, 24 in group 2 and 15 in group 3.
I want to think about this paper with both hypotheses in mind: hypothesis A is that the drugs do not impact outcome but that we are merely seeing the variance of outcome among people who once were psychotic and hypothesis B is that the drugs contribute to worse outcome.
Harrow found that at every point, those in group 1 had a much higher level of psychotic symptoms than those in group 3. For instance at 4.5 years, 76% of group 1 had some psychotic symptoms as compared to 27% of those in group 3 and at 20 years, the numbers were 56% as compared to 24%. In addition, when looking at how many individuals in each group had psychotic symptoms at a minimum of 4 of the 6 follow up visits, 76% of individuals in group 1 met this criteria as compared to 7% in group 3.
This still does not prove that the drugs caused this differential outcome. These results are not inconsistent with the hypothesis that those who were more symptomatic are the ones to be prescribed the drugs. However, it does raise a question as to how effective these drugs are. As Harrow writes, “These results suggest that, longitudinally, the antipsychotics are not effective in eliminating or reducing psychosis for the great majority of SZ.”
However, there is another finding that does challenge hypothesis A. This comes from Figure 2 in his paper on page 2. At 2 years, 74% of individuals in group 1 have psychotic symptoms as do 60% of those in group 3. These differences are not statistically significant. But the lines diverge at year 4.5 and continue to diverge over the next 15 years. At 4.5 years, 86% of group 1 have psychotic symptoms as compared to 21 % of group 3. By year 20, the difference is 68% vs. 8%.
Overall, there is a decline in both groups but the group on continuous treatment remains much more symptomatic.
There is a suggestion that even though the groups look similar with regard to the presence of psychotic symptoms at year 2, in fact group 1 was more symptomatic (figure 4, page 5). 52% of group 1 had moderate or severe functional impairment at year 2 as compared to 30% in group 3. But these lines also diverge. By year 20, 47% percent in Group 1 are still struggling as compared to 8% in group 3.
This study tells us that there was a group of individuals who at 2 years were symptomatic, 30% of them significantly impaired by their symptoms, who chose to remain off-drug. Overall, this group improved dramatically over the next 15 years.
So let’s go back to the two hypotheses. With hypothesis A, I would have predicted that the rate of psychosis in the group who was not on drug to be lower from early on. After all, the hypothesis is that they were destined to do well no matter what and the drugs were stopped because they were doing well. I would have expected them to diverge from the other group by two years. However, they did not diverge until 4.5 years.
This is how science works. There are theories with supporting hypothesis. Each single study addresses a narrowly defined question. Over time, accumulating studies support or refute these hypotheses and over time, there may be a consensus that the preponderance of the data support a theory. Even then, theories may be refuted. Science is messy.
The impact of this study is sometimes minimized because it is not a randomized controlled study (RCT). It has become axiomatic that RCTs are the gold standard in medicine. David Healy, in his book Pharmageddon, has addressed the misunderstanding involved in this notion. RCTs are helpful in answering a certain kind of question, specifically; what is the impact of altering one variable in a system. When done well, they are most helpful in discerning small differences among treatments. We do not need an RCT to find a large difference. We would not have waited for an RCT to use penicillin because its short-term impact on infection was so apparent. At the same time, RCT’s did not inform us about the problems with long-term use of antibiotics in our society. It took a different kind of approach – an epidemiologic one – to understand those kinds of problems.
This study was a naturalistic one and I would argue that this kind of study is an important contribution to this tremendously complex field. One can have a hypothesis with a naturalistic study and one can see to what extent the study support that hypothesis. I challenge my psychiatric colleagues and others to do a thought experiment. If you had been told of this study in 1990, what might you have expected based on the narrative of the time? How would you have drawn those graphs? If you would have drawn them differently from what the subsequent data revealed, what questions does that raise for you? This is what we need to be discussing right now. To dismiss this work because it is naturalistic is missing the point. This is rich data. It may never be collected again. It informs us and at the very least raises some serious questions about current paradigms of care.
But it does not stand alone. It is part of a body of data. It joins Andreasen’s long-term imaging data. It joins recent studies that question even the short term efficacy of these drugs. It joins Wunderink’s randomized trial. It joins other studies, some done as long at 30 years ago, which suggested the same thing – long term use of neuroleptics hinder recovery.
At the same time, this study raises conundrums – an aspect of the paradox that Whitaker addresses in his notable chapter in Anatomy, “A Paradox Revealed.” Among the 70 people in the SZ group, only 12 were free of psychotic symptoms at each assessment. Seven were in the group who were never on drug, so of the 15 who never took drugs eight individuals were not on antipsychotics and had periods of psychosis. Overall most of them appeared to have ended up psychosis free at 20 years. This suggests to me that at some point, while experiencing psychotic symptoms, these individuals chose to remain off drugs. We have a system of care, however, which strongly urges drug treatment at any sign of psychosis. It is almost axiomatic that if one requires hospital level of care, that drugs will be prescribed. People are labeled with anosognosia if they choose to try to recover without drugs and dire warnings of worse outcome are made by many psychiatrists. I am not sure there is a will to offer respite – and some people need a high level of security in order to be safe – to these individuals when there are drugs available that will diminish the intensity of the psychosis in the short run.
Right now, there is legislation moving through congress – the add-on to the Medicare bill that was passed by congress on Friday and which will go to the senate on Monday – that will take us in a different direction. It will make it more likely that people will be forced to remain on drugs long-term and even less likely or even impossible that alternatives to drug treatment will even be available.
I think there is room for a reasonable approach that might not satisfy everyone but might balance the various interests: respect for personal autonomy, safety, economy, improving long term outcome. I would try to hold off on using drugs for everyone for whom a safe and supportive environment can be found. For others, I might offer a neuroleptic but start low and increase slowly. This was the fundamental approach of the neuroleptic threshold study done over 20 years ago. Then when the person is stable, I would have an open discussion about slowly tapering. Even if a person declines this offer, I have learned that this is a discussion that can be ongoing for many years. I am learning through my own practice that people who may have had trouble coming off drugs earlier can still be successful with tapers even 10 or 15 years later. But this is an approach that does not place prevention of relapse as paramount, but in balance with other considerations including personal choice and the many consequences associated with long-term exposure to these drugs.