Francesco Bollorino: First of all, Russ, please tell me more about you and your work at FDA in order to allow our readers to know you better…..

Russel Katz: My name is Russell Katz. I received my MD degree in 1977 from the Albert Einstein College of Medicine in New York City. I completed my training in Neurology at the Einstein Hospitals in 1982, and began working at the Food and Drug Administration (FDA) in the Division of Neuropharmacological Drug Products in 1983. I became the Team Leader (supervisor) of the Neurology Group in the Division in 1985, and the Deputy Director of the division as well in 1986. I became the Director of the Division in 1999.

Our division is responsible for the regulation of clinical research with investigational drugs to treat neurologic and psychiatric illnesses in the United States. In this capacity, we review all protocols for studies in humans with investigational (unapproved) drugs, as well as review adverse event data from these trials. We frequently meet with drug companies to help design the studies that will be used to establish the safety and effectiveness of these drugs that will ultimately result in the marketing of these drugs for specific indications.

In addition to regulating research with investigational drugs, we are responsible for reviewing all of the safety and effectiveness data submitted by sponsors to support the marketing approval of drugs to treat either neurologic or psychiatric illness. These data are submitted in New Drug Applications (NDAs). If a drug is already approved for one indication, and the application is submitted to support a new indication, I have the authority to make the final decision about whether or not the new indication should be approved. If the application is for a drug that has never been approved for any indication before, we review the data in my division and make a recommendation to our superior, who has the authority to make the decision about whether these applications should or should not be approved.

My division has about 55 employees, divided into numerous groups. We have three groups of clinicians; a group of neurologists (the Neurology group), a group of psychiatrists (the Psychiatry group), and the safety group. Each application is assigned to a reviewer in either the neurology or psychiatry group; this reviewer is responsible for reviewing the protocols and clinical information in that application. The safety group consists of several physicians trained in internal medicine (and some in epidemiology as well) who help us evaluate specific safety issues that may arise, either during the development of the drug, in an NDA, or even after the drug is marketed. In addition to these clinicians, the division has pharmacologists, whose primary role is to review animal toxicity data. These animal data are required (both rodents and non-rodents) to examine the toxicities that the drug is capable of producing (including microscopic changes), so that the studies in humans can be adequately designed to monitor any potential toxicities that may occur in humans. The division also has administrators, as well as pharmacists, called Project Managers, who are responsible for ensuring that the various projects are completed in a timely manner; project managers also serve as the primary contact between the division and the industry sponsors, and are also responsible for writing the many letters that we send to companies, as well as scheduling the many meetings with industry. On any given day, I may have 6-7 meetings, a number of which may be with an industry sponsor.

In addition to the staff described, the division frequently consults with clinical pharmacologists, who are experts in drug metabolism and drug interactions, chemists, who are responsible for ensuring that new drugs are manufactured safely, are pure and potent, statisticians, who are critical to analyzing the results of clinical studies and who provide important input into the design of clinical trials, and epidemiologists and other experts in other medical areas. All of these specialities are critical in the adequate assessment of the properties of new drugs.

A: about your methodology at FDA: which are the criteria and the methodologies, with whom you evaluate the safety and the effectiveness of drugs in general and of new antipsychotics drugs in particular?

R: Regarding the criteria for effectiveness for new drugs, the law that we work under, the Federal Food, Drug, and Cosmetic Act (FD&C Act; the Act), requires that we find that "substantial evidence" of effectiveness exists before a drug may be found to be effective. Basically, this typically means evidence from more than one "adequate and well-controlled" clinical study. Under rare conditions, the Agency may approve a drug on the basis of only one such study, with "confirmatory evidence". We have never approved a drug for a psychiatric indication on the basis of only a single trial; we believe that independent replication is crucial to concluding a drug is effective.

Although many types of study designs could theoretically demonstrate effectiveness, almost all of the studies on which we rely for a finding of "substantial evidence of effectiveness" are placebo controlled studies. The most important factor in being able to interpret a study is, in our view, a finding that the new drug is better than the control group. By better, we generally mean that a statistically significant difference between the drug and control group (in almost all cases, this means that the p-value for the drug-control comparison is 0.05 or less) is shown on a meaningful clinical outcome. For psychiatric illness, these outcomes are usually mean scores on a standard rating scale (for example, the HAM-D in studies of patients with Major Depressive Disorder). We believe that a finding of no difference between a new drug and a standard drug is almost always impossible to interpret, because such an outcome has at least two interpretations; either both drugs worked, or both drugs did not work. Although the standard drug in such studies is almost always a drug that is approved for that indication, we still do not have, usually, assurances that the standard drug actually worked in this particular study. Because we usually cannot know this, it is possible that the standard drug did not work in a particular study; if the standard drug did not work in that particular study, and the new drug was shown to be equal to the standard drug, we might conclude that the new drug was effective when, in fact, it was not. So, we believe that we can only conclude that a drug is effective if the new drug is superior to the control.

We believe that placebo controlled trials are ethical in almost all cases (even when drugs are already approved for the indication being studied), because we do not believe that exposing patients to placebo for relatively short periods of time in most cases (6-8 week for depression or schizophrenia studies, for example) causes any permanent harm, or increases the risks of injury. In these studies, if patients start to deteriorate, they can be removed from the trial and be treated with an effective treatment. Further, the placebo control design is the most efficient; that is, we can get an answer that we can interpret with the fewest number of patients.

Most of the studies that serve as the basis for drug approval in psychiatry are relatively short (3-8, 12 weeks) even though the drugs may be used for years for chronic conditions. For this reason, we also require that sponsors perform longer term studies, to determine if the drugs work for longer periods of time. Because most patients cannot be treated with placebo for very long periods, the most common design used to document longer term effectiveness is the randomized withdrawal design.

In this design, patients are treated with the drug, in an open-label setting, and must meet criteria for a clinical response for at least 6 months. After that point, patients are randomized to continued treatment or placebo, and are followed until they meet certain relapse criteria. The time to relapse is then analyzed in this randomized phase. If the patients in the drug group have a longer time to relapse than the placebo patients, the trial is successful, and the duration of the effect is considered to be the duration of the pre-randomization, open label phase, during which they met responder criteria. We want this open-label phase to be as long as possible (again, at least 6 months), because we want to answer the clinically important question of whether or not the drug is continuing to work for at least 6 months. Recently, we have decided that this long-term study should be part of an NDA submission, whereas in the past (even the recent past) we have only required sponsors to perform these studies after the drug has been approved. There has been controversy about the ethics of taking patients off a drug that has been treating them successfully, and, in some cases (in schizophrenia, for example) these studies are difficult to perform in the United States, because many institutional review boards do not think they are ethical. We disagree, because we believe that physicians currently treat patients for long periods of time with these drugs (even if the only data available are from short-term studies), and it is critical to determine if the drugs actually work in the long-term. Also, in these designs, if the patient is doing poorly in the randomized phase, they are considered to have met relapse criteria, and they are removed from the study and treated with an effective drug, so they are not kept on placebo for an unacceptably long period of time.

Obviously, the specific designs of studies that are used to demonstrate effectiveness vary from indication to indication (for example, in not every indication are mean differences in rating scales the primary outcomes; for drugs to treat panic disorder, we usually look at the number of full panic attacks as a primary outcome, as an example), but the major considerations for most indications are as described above.

Regarding safety, an important consideration is that sufficient numbers of patients be exposed in order to get a reasonable idea of the safety issues with these drugs. According to the International Conference on Harmonisation, a total of at least 1500 subjects typically need to be exposed to drug, with at least 300-600 exposed to a therapeutic dose (or greater) for at least 6 months, and at least 100 for one year. Most psychiatric drugs have more than this minimum exposure, but not necessarily much more. We evaluate the drug’s effects on vital signs and EKGs in relation to the timing of the dosing (that is, we measure these at least at the maximum plasma level, and at multiple other times as well), and routinely measure standard laboratory measures. Of course, we look very closely at all reported adverse events, and in particular at all deaths, serious adverse events, and discontinuations due to adverse events; we are particularly interested in the comparison of the incidence of these events between drug (and different doses of the drug) and placebo.

We have recently concluded that the atypical antipsychotic drugs cause an increase in cerebrovascular adverse events and death in elderly patients with dementia and psychotic symptoms. We do not really know the mechanism of these events, but, for example, a recent review of 17 controlled studies with 4 atypical antipsychotics showed that there were increased deaths on drug compared to placebo in 15 of these studies. We do not know the mechanism of these deaths. As you know, we had concluded recently also that these drugs cause loss of glucose control in some patients; we also do not understand the mechanism of these events.

This last conclusion was based not on data from controlled clinical trials, but from post-marketing data. In this country, physicians and other health care workers are not required to report adverse events for drugs that are approved that they think are drug related (if companies become aware of these events, they are required to report them to the FDA, at least on an annual basis in most cases). However, sometimes we do get enough reports of adverse events to be able to conclude whether or not the drug seems to be causing the event. Typically, in this case, we count the number of reports of a given adverse event reported with a drug and then we try to estimate the number of patients who have been treated with the drug (this estimate is based on estimates of the number of prescriptions written and some assumptions about how this translates to months or years of use). These data are then used to calculate a "reporting rate", which is the number of cases reported to us that could reasonably be related to the drug divided by the number of patients (or patient-years) of exposure. This is a reporting rate, not a true incidence, because we know that not all events of a particular type are reported to us (we believe that the number of cases reported to us is far smaller, maybe 10% or less, of the actual number of cases that occurred) and we are not really sure of the number of patients who are actually treated. However, we then try to figure out what the background rate of the specific event is in a relevant population (ideally, the population with the same disease who have not received the treatment). We then compare the reporting rate to the background rate, making certain assumptions about how much the reporting rate actually differs from the true incidence. If the reporting rate is about equal to the background rate, or certainly if it is greater, all other things being equal, we probably will conclude that the drug caused the event. It is very common to base various safety conclusions on these sorts of analyses, especially when the event in question is very rare, because we would never have enough exposure to the drug prior to approval to be able to detect rare events. It is only after approval that enough patients are exposed to a treatment to detect rare events, and it is in these settings that these analyses are most helpful. Of course, when the event we are concerned about has a high background rate, it may be very difficult to determine that a drug caused this event, because typically the reporting rate will be far below the background rate; this was the case with the atypical drugs and loss of glucose control, given that diabetes is so common in the background population. I’d like to make an important point in our assessment of safety (and effectiveness too). We rely on the findings from the empirical clinical data when we are considering questions of safety and effectiveness. That is, we are far less concerned about whether we understand the mechanism of the finding. For example, as I mentioned, our analyses of clinical trials revealed that the antipsychotic drugs cause an increase in mortality in elderly patients with dementia. We really do not have any good understanding about why this should be so, but the data confirm that it is true, so this is the basis for our conclusion. This is typically the way we reason (certainly for effectiveness also, we really do not know with any certainty why these drugs work-we have theories, of course-but we base our decisions about effectiveness on the findings from well controlled studies).

A: about antipsychotics drugs: which are the confirmations of safety and effectiveness you usually approve and which are the elements of doubt you are at the moment investigating, searching proffs in positive and negative?

R: As far as antipsychotic drugs, I think I’ve mentioned the major issues for both safety and effectiveness questions. I would only point out that fully investigating some of these questions (mostly safety, but sometimes effectiveness also) can take a very long time, for several reasons. One reason, of course, is that there are many such potential safety signals (arising from trials, post-marketing reports, literature reports, etc.), and we don’t have enough staff to address all of them instantly. Also, many times we ask the sponsors to perform additional analyses, and this also takes time. Further, when a safety question arises (for example, the glucose question), we are very interested to evaluate all of the drugs in a class, or with a particular indication. In these cases, we have to not only look at our own data (say, post-marketing reports), but also we have to ask all of the involved companies to analyze their own data; this can take a very long time. Another example, not an antipsychotic, but a good example all the same, was the question of whether Serzone (nefazodone, the anti-depressant) caused liver failure. Again, the question was raised by post-marketing reports (the events were rare), but the sponsor questioned whether the background rate we were comparing their reporting rate for liver failure was appropriate. We used a rate based on the rate of liver failure in the general population. The sponsor believed that the background rate of liver failure in the relevant population, that is the population of patients with depression, was greater than that in the general population. If that was true (and there was some reason to believe that it was), comparing the reporting rate of liver failure for nefazadone to the general population background rate would have been inappropriate. Therefore, a large study was done to further examine what the background rate of liver failure was in the depressed population; this took several years (incidentally, it turned out that the background rate of liver failure in the depressed population was about exactly the same as the rate in the general population). But we thought it was important to examine this question before we concluded that nefazadone caused acute liver failure. It is very important to us to try to get the best answer possible, because changing the label with new safety information (which is usually the outcome of these analyses) can have important consequences. For example, if we put a strong warning on the label on a specific product, we may drive doctors to use other products in that class, but if those other products cause the same adverse event, and we don’t know about it, that could be very problematic.

A: Considering the high social impact of seriuos mental illness how can we evaluate from the economical point of view the use of new antipsychotics in front of the use of older first generation ones?

R: Fortunately, or unfortunately, we do not consider economic questions when we decide whether or not a drug should be approved. The FDA has no authority to consider drug cost; other federal agencies do consider cost for various reasons (for example, to determine which drugs will be reimbursed for by Medicare), but the FDA does not consider economics at all. Further, the law under which we work does not require that new drugs be more effective (or safer) than already approved drugs; the law only requires "substantial evidence of effectiveness", and says nothing about comparative efficacy or safety. Of course, we do consider, particularly with regard to safety, how a new drug compares to other drugs, but usually this question is not tested directly (that is, we usually don’t have comparative data from a single study in which both treatments are tested). For example, if a new drug to treat insomnia (not considered a life threatening condition) has a major safety question (say, aplastic anemia), we may very well conclude that such a serious adverse event (even if relatively rare), is not acceptable for a drug to treat this condition, given that there are many approved treatments for this condition that do not cause this serious adverse event. However, for drugs to treat very serious conditions (for example, depression, schizophrenia, etc.), we usually tolerate serious adverse events, even if other drugs approved to treat this condition do not cause this serious adverse event. For example, nefazadone caused acute liver failure (rarely), but no other antidepressant did. Nonetheless, we decided to keep it on the market, because we believe that depression is a very serious disease, and the physician and patients can decide if they want to take the risk. Underlying this decision is the belief that patients may respond to one drug but not another, and for serious diseases, it is better to have many choices, as long as the labeling adequately warns the doctor about the serious risks. Incidentally, we rarely, if ever, have any actual data that some patients may respond to one drug and not another; we just take this on faith.

Regarding comparative effectiveness, that is a difficult question. That is, companies sometimes want to say that their drug is more effective than another drug. Studies to definitively show this are complicated (for example, we think that such studies should adequately study multiple doses of each drug), and are rarely performed. I do not believe that we have ever granted such a superiority claim, except perhaps for clozapine, which demonstrated in an adequate study, that patients who failed another antipsychotic actually did better on clozapine than a different antipsychotic. Even here, I don’t believe we say that clozapine is better than any other specific drug. Again, the law does not in any way require such a showing of superiority.

A: Some antipsychotic drugs (either "old" and "new") produce an increase of appetite and in conseguence of caloric contribution. Is this the only one reason why we can see a decrease of the tolerance for glucose in patients under treatment with such type of drugs or have you found other metabolic processes involded?

R: Regarding the mechanism of loss of glucose control with the antipsychotics, I can only say two things: 1) we do not know the mechanism, although there are theories, and 2) we found this adverse event with drugs that did not seem to cause weight gain, as well as with drugs that do cause weight gain, and we did not think we could say that those drugs that cause weight gain cause more of a problem with glucose control than those that did not cause weight gain. Therefore, we really do not understand the mechanism of this adverse event.

A: Are there data of evidence which support or deny the fact that the coming of new antipsychotic drugs has modified the prognosis of serious mental hillnesses? To make some examples to be clear: decrease of time of hospitalization, increase of number of patients under treatment, decrease of breaks in treatments….

R: I really cannot answer the question of whether or not the new antipsychotic drugs have modified the prognosis of mental illness. We do not have the sort of data that you are asking about.