AUTHOR COMMENTARIES

His most-cited paper, "Effectiveness of antipsychotic drugs in patients with chronic schizophrenia," (Lieberman JA, et al., N. Engl. J. Med. 353[12]: 1209-23, 22 September 2005), is a Highly Cited Paper in the field of Clinical Medicine, with 463 cites to date. In addition, this paper is the most-cited paper on schizophrenia published in the past two years.

At present, Dr. Lieberman is the Lawrence E. Kolb Chairman of Psychiatry at the Columbia University College of Physicians and Surgeons and Director of the New York State Psychiatric Institute. He also holds the Lieber Chair and Directs the Lieber Center for Schizophrenia Research in the Department of Psychiatry at Columbia.

Your 2005 New England Journal of Medicine article continues to be cited at a remarkable rate. What prompted that study? Why test so many antipsychotic drugs so relatively late in the game?

To understand why the study was necessary, you have to understand the history and development of antipsychotic drugs. These drugs were invented in the 1950s and were then known as neuroleptic drugs. The term was derived from the ability of these drugs to cause a kind of psychomotor paralysis, a characteristic slowness or rigidity of the animals or people who took them.

But at the same time these drugs produced a near-miraculous quelling of psychosis, for which there had never been an effective treatment. After the prototype of this kind of drug was developed, a series of similar drugs, essentially analogues, were introduced. They all had the same therapeutic property but varied in terms of the potency and the therapeutic dose at which they had to be administered. Through this process of essentially imitating the prototype with relatively specific or minor refinements in potency and side effects, a whole cohort of antipsychotic drugs were developed, which eventually came to be known as first-generation or typical antipsychotic drugs.

For close to 40 years these drugs were used, and they worked, albeit with limitations of which the medical field and the patients who took them became increasingly aware. This period witnessed no real progress, no real innovation, in terms of improving the effectiveness of treatment and improving upon the pharmacology of these antipsychotic medications.

When did this change and what did it?

In 1989 clozapine was introduced, and this was really a different kettle of fish. It was the first antipsychotic medication that had therapeutic efficacy, but did not produce this characteristic psychomotor paralysis. It was called an atypical neuroleptic or antipsychotic. In addition, clozapine had a greater capacity to alleviate psychotic symptoms. This was demonstrated by testing it in patients who were not responsive to existing medications and finding that these patients improved substantially with clozapine. So clozapine proved to be a huge sensation. It was touted as a breakthrough drug and was even the subject of a cover story of Time magazine in 1990.

But like all good things, there was a catch. In this case, there was a very serious potentially deadly side effect called agranulocytosis—toxicity against white blood cells. This side effect occurs in about one in 150 people who take the drug. If it isn’t detected in time, it renders people susceptible to infection and possible death by sepsis. As a result the drug was approved by the FDA only to be used in people who did not respond to other antipsychotic medications and with the requirement that they have their blood tested on a weekly basis for as long as they took the medication.

Did the pharmaceutical industry come up with other drugs that had the same efficacy of clozapine but not this potentially fatal side effect?

Yes, clozapine motivated the pharmaceutical industry—you could say galvanized it—to develop a clozapine-like drug that did not have agranulocytosis as a side effect. So other drugs were developed, tested, approved, and introduced, beginning in 1994 with a drug called risperadone; then 1996, olanzapine; 1997 was quetiapine; and in 2001, ziprasidone. These drugs were ostensibly clozapine-like, meaning they had better therapeutic efficacy and less or no potential for the neurologic side effects that the first-generation drugs had.

These drugs began to be used preferentially over the older drugs, to the point that by the end of the decade—the end of the century—upwards of 90% of all prescriptions for antipsychotic medications were for the newer medicines. One result was a tremendous increase in the cost of antipsychotic medications. In the U.S., the cost of these medications, pre-clozapine, was on the order of $300 million per year. By 1999, it was about $7 billion.

How much better were these second-generation drugs?

If you looked at the data, at the clinical trials that were done comparing first- and second-generation drugs, it wasn’t obvious that the second-generation drugs were actually better. It was by no means absolutely conclusive. And if they were better, they didn’t appear to be all that much better.

So the National Institute of Mental Health (NIMH), as one of its initiatives to stimulate large practical clinical trials in mental illness, awarded us a contact to do a study evaluating the effectiveness of antipsychotic medications. What they said is, "We would like to know how the first-generation and second-generation drugs compare. And if the second-generation drugs are better, how do these compare to each other and which are the best?"

Then your study was prompted by the NIMH?

Yes. They issued an RFA, a request for applications, to submit a proposal for a study that would provide evidence and hopefully answer this question of the comparative effectiveness of the various antipsychotic medications.

How many drugs did you test and on how many subjects?

We initially framed the question of wanting to test first-generation drugs against second-generation drugs, and then, secondly, comparing second-generation drugs to each other. So the design involved all the existing second-generation antipsychotic drugs approved by the FDA. Then it compared them to one first-generation drug, which we used as a proxy for all of them, since none were considered to be any better or worse than any of the others. We selected as our proxy a drug that was of intermediate potency and less widely used than some of the others.

Seems like an odd choice. What was your thinking?

The prevailing clinical practice was to use the high-potency first-generation drugs—these were the ones that produced the greatest amounts of neurologic side effects. We reasoned that if we’re going to retest these drugs, we should use one that is not obviously going to produce noxious side effects. We wanted to use a drug that would at least have a reasonable chance of being tolerated, but still worked by the same mechanism as all the other first-generation drugs. We selected a drug called perphenazine.

How many subjects did you have in the study?

Well, we carried out the study at sites in 57 locations in the U.S. and enrolled 1,500 patients. The study called for every patient to be in for a year and half. What’s important to understand is that the patients included in the study were representative of people with chronic schizophrenia.

We allowed for people to be in the study who had other psychiatric or medical co-morbidities and might require other adjunctive medications, in addition to their experimentally assigned anti-psychotic medication. This meant we could generalize the study sample and results to patients in the real world.

Is that what you meant when you said the NIMH funded you to run a "practical" randomized controlled trial?

Yes, exactly. This is somewhat of a different type of study design that is kind of intermediary between what are called classical randomized, controlled trials and observational or naturalistic studies.

Did you find what you expected to find?

When we began this study we fully believed and hypothesized that the second-generation drugs would be far superior and would be shown to be both more effective and safer than perphenazine. What we found, much to our surprise, was that the drug that did the best overall was olanzapine, but it was only by a small amount and there was absolutely no difference between the first-generation drug perphenazine and the other second-generation medications.

There was a high switch rate from all the medications, indicating that over that year-and-a-half period, the majority of the people voted with their feet to try and find a medication that worked better than the one to which they were originally allocated. That said that, yes, these treatments work, but not ideally, and that neither the patients nor their physicians are wholly satisfied with them.

Was olanzapine better at controlling side effects or symptoms or both?

It was better at controlling symptoms but it produced substantial side effects—principally weight gain. And when we looked at perphenazine, it did not produce any significant degree of neurologic side effects, so it was as effective as the other medications and as comparatively safe.

How did the pharmaceutical companies respond to your study?

Not happily.

Did they try to dispute the results?

Yes. Each company has had their own take on the study and its results.

Has anyone else replicated your results?

As it turned out, at the same time we were doing our study, colleagues in England, unbeknownst to us, were doing a very similar study. This was the CUtLASS study, and it had been commissioned by the National Health Service in the U.K., also to compare first-and second-generation drugs. CUtLASS had a slightly different design, but it was asking the same questions—it was also a practical clinical trial, and it basically came up with the identical results.

Even down to the slight superiority of olanzapine?

Yes, it did show some superiority for olanzapine within the patients who received second-generation drugs, but not enough people were treated with olanzapine to give the statistical power to show a significant difference from the others. But on the question of first generation vs. second generation, CUtLASS found no difference between the two classes.

How do you make sense of your results, considering the belief that these second-generation drugs were "breakthroughs" in treatment?

I think there were two main reasons for this disconnect between the exuberant claims and the disappointing performance of these second-generation drugs in the two trials. The first reason is what I call the efficacy-effectiveness gap. This is something well known to researchers as the explanation or as the basis for why treatments often don’t work as well in clinical practice as they appear to do in the clinical trials on which the FDA approvals are based.

Is this because the clinical trials are set up to show the maximum possible benefits?

Right. Phase two and phase three trials, the ones submitted to the FDA for review, are short term; they’re conducted at least partially in hospital settings; they involve a selected sub-sample of patients for whom the drug will eventually be used; and they exclude patients who have different types of co-morbid conditions and so are more complicated to treat.

In addition, the measures they use to assess effectiveness are usually ratings of symptoms, whereas in clinical practice, you’re measuring how well the patients are functioning, how they feel subjectively, as well as what the severity of their symptoms are.

The second reason for the disparity between the results and our expectations is simply that the claims for the superiority of second-generation antipsychotics were greatly exaggerated, and this may have been encouraged by an overly expectant community of clinicians and patients who were eager to believe in the power of new medications. They were hoping for something new and better, they were told here it is, and they bought into it.

At the same time, these newer drugs were very aggressively marketed by the pharmaceutical industry. Although marketing is supposed to be based on evidence, that evidence was largely coming from the clinical trials. It was not representative of how the drugs would necessarily perform in the real world, in the broader population of patients.

One of my colleagues in England was interviewed on this subject and he said that it wasn’t that clinicians were misled or deceived. Rather they were sort of "beguiled" into thinking that these new drugs were better. It’s like you go into a supermarket or an automobile showroom and somebody shows you a generic version of a product and you compare it to one that has very fancy and attractive packaging. There was a sort of slick presentation to these drugs that gave the perception they were not only newer but better.

Your NEJM paper has obviously been very highly cited. Do you think the results of your study have actually influenced clinical practice?

I think one big effect the paper has had is to change the debate in the field. So rather than thinking it’s a foregone conclusion that these drugs are better and that it is substandard care or possibly even malpractice to be using the older drugs, this is no longer the case.

At the same time, if you look at prescription rates and marketing data, there hasn’t been a rush back to the older medications. The question is, what are the incentives and what are the motivational forces?

The main motivational force, apart from knowledge, is financial. But when doctors decide what medications to use, they don’t think about how much it costs; they think about how effective it might be and what are the side effects. Now we have an entire generation of doctors who have been trained only on the new medications. They’ve never even used the first-generation drugs.

Do you think this will change?

It could. One reason is that the government, insurance companies, the pharmaceutical-benefits companies, the Veterans Administration, and the state Medicaid offices are all deciding how to use data from our study and from CUtLASS, and whether the formularies and the reimbursement policies should encourage consideration, if not use, of first-generation drugs. If these drugs work as well, more or less, and cost a fraction of what the newer medications cost—the ones that are still under patent—then from a public health standpoint we should be trying to be more efficient and economical. So changes in policy may affect practice patterns in response to these results.

However, there are also significant political considerations. Patients, the American people, and the doctors don’t like to be told what to do and they don’t like to be deprived of choice. People want to have a right to have what they want, and ostensibly the very best treatment, even if it costs a heck of a lot more. And this is true, of course, particularly when they’re not paying for it.

What people forget is that ultimately we are paying for it, either from our tax dollars or higher insurance premiums. I’m just a scientist and physician, but as a result of this study, I and my colleagues have been drawn into a policy debate and it’s ongoing.

If a patient comes into your office tomorrow and you diagnose him with schizophrenia, what drug do you give him?

For first-episode patients, that’s still an open question. For chronic patients, I would review what treatments had been already used, what’s worked, and what side effects have appeared. At the end of the day, if there are no such special considerations emerging from the treatment history, then there is no reason not to prescribe perphenazine or another intermediate potency first-generation drug, because I know it’s as good as any other drug and particularly if cost is an issue. For treatment-resistant patients who haven’t gotten better on anything else, then clozapine is the drug to be used.

For first-episode patients, we don’t know. Those patients are relatively rare. They weren’t a part of our study or the British study. But the beginning of the illness is the point when the patient is most responsive to treatment. So it’s possible that a slight difference in effectiveness may have exaggerated benefits in this group. There is a study looking at first-episode patients that’s just been completed in Europe. We’re waiting for the results, which will be the next important piece of evidence or chapter in this story.

How have you personally followed up on this research?

We’re doing a couple of studies, trying to determine how to manage the weight-gain side effect of these second-generation drugs. Is it better to switch to a drug that doesn’t cause weight gain or to manage the weight gain?

We’re also doing a study comparing long-acting injectable first- and second-generation antipsychotic drugs. Here the same cost differential applies. We’re also spending a lot of time and effort looking for new and novel drugs. It’s clear even if the new drugs are a little bit better, they’re still far from being panaceas. We desperately need mechanistically novel, new medications. So we’re now very involved in a program of experimental therapeutics.

Jeffrey A. Lieberman, M.D.
Columbia University
New York, NY, USA