Column - The slippery slide

This column is one of series of case-based discussions illustrating basic principles of clinical psychopharmacology relevant to optimal patient care. After reading the case presentation, readers can take a moment to formulate how they might have managed the patient. I will then present what was done, the outcome, and the important take-home points, at least from my perspective.

Joe was a 35-year-old man with chronic undifferentiated schizophrenia. His illness had begun in his late teens with a psychotic break. Over the next 15 years, Joe had many such breaks that required hospitalizations. His functioning deteriorated due to the persistence of psychotic symptoms between frank breaks and due to the emergence of deficit symptoms. He was treated with a variety of conventional antipsychotics, including both depot antipsychotics alone and in combination with other psychiatric medications (e.g., mood stabilizers) without great success. His psychotic symptoms would improve but not completely resolve and his deficit symptoms were not appreciably affected. Given this history, Joe was judged to be suffering from treatment refractory schizophrenia.* Joe was therefore given a trial of clozapine and did remarkably well on this monodrug therapy at a dose of 250 mg/day! Joe's case was one of those that make you wonder what is it about clozapine that is so special and why other so-called "atypicals" have not been able to match it--but that is an issue for another column. Admittedly, Joe did not become well, but he did improve substantially with a virtually complete resolution of his psychotic symptoms and a definite improvement, although not a complete resolution, of his deficit symptoms (e.g., amotivation, asociality, anhedonia). He felt much better and functioned better than he had in years.

Joe continued to do well for the next year. Then he presented to his outpatient psychiatrist with complaints of anxiety. This anxiety was "free floating," in that it had no specific focus, nor was there any apparent environmental precipitant. Joe's social system had not changed and he had no new financial, living, or health concerns. There was also no evidence of a re-emergence of psychotic symptoms. Joe was compliant with his clozapine and there had been no change in his physical status that might have affected his clozapine clearance.

Joe's outpatient physician decided to add an anti-anxiety medication, buspirone, at a dose of 5 mg three times a day. At the end of 2 weeks, Joe had not improved. The physician therefore tried to optimize the dose by increasing it to 30 mg/day. A week later, on a Tuesday night, Joe went to a local emergency room complaining of persistent anxiety, initial insomnia, and feelings of being "overwhelmed" and hopeless, and was admitted. This admission resulted in a change of physicians because Joe was being followed in a group practice in which some of the physicians specialized in outpatient management while others specialized in inpatient care. Nevertheless, procedures were in place for communication between the inpatient and outpatient physicians and a summary of Joe's outpatient care was sent to the inpatient team.

Given the complaints of initial insomnia and feelings of hopelessness, Joe was diagnosed as having a major depressive episode superimposed on his long-standing schizophrenia. Based on that assessment, nefazodone was added. The dosing was straight from the package insert: 100 mg twice a day with the intention to increase to 150 mg twice a day in a week assuming good tolerability and persistence of the target symptoms of anxiety, initial insomnia, and feelings of hopelessness.

During Joe's physical examination upon admission, he had some expiratory wheezes and stated he had suffered from asthma in the past. As a result, a primary care physician was consulted who examined Joe and started theophylline 300 mg/day.

If you wish, take a moment at this juncture to reflect on this case. What are the salient issues from your perspective? Do you agree with the clinical management? If not, what would you have done differently and why? What do you think the outcome was?

The outcome in Joe’s case is ordinary rather than exceptionally dramatic or convoluted. He had no serious adverse outcome. He did not die nor did he have a seizure. If not for this column, Joe’s case would most likely not have made it into the medical literature. Instead, it would have gone to the repository where old medical records are kept and forgotten. Nevertheless, Joe’s case presents issues that are worth discussing. As regular readers of this column know, we can conceptualize such cases within the framework defined by the equation in Figure 1.

The initial inpatient plan involved more aggressive pharmacotherapy--but Joe did not get better. In fact, he told his regular outpatient physician (who happened to be doing the group’s rounds on the first weekend Joe was in the hospital) that he felt worse in terms of fears about not getting better. He was visibly more anxious and agitated. Following the advice in Kenny Roger’s song, "The Gambler," this physician decided it was time to fold rather than to hold the cards that had been dealt. In other words, he stopped all the medications except clozapine on Saturday. The next day, he increased the dose of clozapine to 300 mg/day. Later that week, Joe had improved sufficiently to be discharged home. For the past several years, he has continued to do well in continued outpatient care.

Perhaps Joe would have done equally well or even better if he had been left on all four medications. In clinical practice, we generally cannot rigorously answer such a question because we deal with a single patient at a time. We do not have the luxury of having multiple Joes that we can randomly assign in a double-blind fashion to either clozapine alone at a higher dose or clozapine plus buspirone, nefazodone, and theophylline to determine which approach is better. Uncertainty is thus an inescapable part of clinical practice--particularly when the outcome is anything less than a serious and unexpected adverse outcome.

In this regard, Joe’s case is quite different from previous ones I have discussed in this column.^2-5 In those earlier cases, we had the following reasons to conclude that the treatment those patients received was the necessary and sufficient cause of their outcome:

1. The outcome in those cases was inconsistent with the expected natural course of the illness. While a sudden fatal arrhythmia might be expected in an elderly patient or a patient with a previous history of cardiac disease, it is not expected in a healthy child ("What Happened to Tommy?", November 1998).⁵ The same is true for the development of neuroleptic malignant syndrome (NMS) in a 48-year-old woman ("I Don’t See ‘Em," September 1997).³

2. All other possible causes of the adverse outcome could be ruled out in those cases. For example, the autopsy in the case of the child’s sudden death revealed no other pathological cause of death.⁵

3. The outcome was consistent with the pharmacology of the drugs involved: the mania in the 17-year-old girl ("Do You Believe in Magic?", March 1997),² the NMS in the 48-year-old woman,³ the delirium in the 67-year-old woman ("A Message from Titanic," July 1998),⁴ or the sudden death in the child.⁵

4. In several of those earlier cases we had laboratory results that were consistent with the adverse outcome being the result of the treatment (e.g., the substantially elevated tricyclic plasma level and the electrocardiogram findings in the case of sudden death).⁵

5. The patients in several of these earlier cases unwittingly served as their own control.^2-4 They did not have the problem before the treatment. They developed it after the treatment was started in a time course consistent with the pharmacology of the medications involved. The problem resolved when the medications were stopped in a time course that was also consistent with the pharmacology of the medications involved.

6. Theoretically we could have rechallenged the patient with the suspected culprit to see if the adverse event would re-occur. Obviously, we rarely take this last step for ethical reasons.

In cases like Joe’s, we have none of these six pieces of evidence. First and foremost, we had no serious and unexpected outcome. After all, that is the goal of good clinical care and the reason for the admonishment to "first do no harm." It is for this reason that Joe’s case is ordinary. So, with this caveat of acknowledged uncertainty, what principles can we derive from this case?

Diagnosis subserves a central function in medicine. Ideally, it should tell us something about the pathophysiology and/or pathoetiology underlying the patient’s problems and in this way guide us in terms of what intervention is needed. For that reason, it is the first factor under the biological variance term in the equation in Figure 1. For example, a patient with a viral pneumonia needs a different treatment than one with a bacterial pneumonia. We believe that psychiatric diagnosis is also meaningful in some way in terms of the basic processes underlying the patient’s signs and symptoms, even if we do not know precisely how at the present time. If we did not have this belief, then diagnosis would not be the pivotal inclusion criterion for most, if not all, clinical trials of new psychiatric medications.

Joe’s case also illustrates the relatively recent practice of giving multiple psychiatric syndromic diagnoses to the same patient. Joe went from having one to several diagnoses in a matter of a few weeks. For many years, chronic undifferentiated schizophrenia was Joe’s only diagnosis. Following his hospitalization, his diagnoses included: a depressive episode and a nonspecific anxiety disorder or at least anxiety complaints along with schizophrenia and asthma. Consistent with the change in the number of diagnoses, Joe’s treatment regimen went from one to multiple medications (i.e., from clozapine alone to clozapine plus buspirone, nefazodone, and theophylline). The reason is simple: Physicians use diagnosis to guide treatment decisions (see Figure 1). In other words, the concept of "comorbidity" greases the polypharmacy slide following the mantra, "A drug for every diagnosis." This approach is not far from the practice of treating every symptom with a drug.

One of Joe’s new diagnoses on admission was a depressive episode. Was that diagnosis helpful? It certainly led to the addition of nefazodone, but did Joe actually have a depressive episode? In favor of this diagnosis is the fact that Joe had recently developed anxiety, initial insomnia, and feelings of hopelessness. He also complained of impaired concentration/attention and had mild psychomotor agitation. His deficit symptoms overlapped with depressive symptoms, in that he had a reduced sex drive and a baseline impairment in interest in his surroundings and in activities; however, these symptoms had not worsened coincident with his emerging problems with anxiety and sleep. On the other hand, Joe readily attributed his feelings of hopelessness to his concern that his anxiety was not better. He denied feelings of sadness or guilt. He had no change in his appetite.

An empirical trial of an antidepressant is probably a common practice when faced with such a constellation of symptoms. If the patient improves, then we are likely to interpret that fact as supporting the diagnosis. In this case, Joe never had a fair trial of nefazodone but we know that his depressive-like symptoms resolved when he was treated with a higher dose of clozapine alone. Given that outcome, should we conclude that clozapine is an antidepressant?

Most psychiatric diagnoses represent syndromes. They are thus only one step up the ladder from the level of a symptomatic diagnosis (e.g., headache). We do not need to be ashamed of this fact--it is simply the level of our knowledge. However, we do need to realize that different patients with the same syndromic diagnoses may have quite different illnesses in terms of pathophysiology and/or pathoetiology and may thus respond quite differently to the same treatment.

In essence, diagnosis is to the care of patients as a hypothesis is to the conduct of research. The role of a hypothesis is to serve as an educated guess about the cause and effect relationship underlying some phenomenon. Diagnosis plays the same role in the practice of medicine. Like hypotheses, diagnoses are only as good as they are helpful. Researchers design studies to test their hypotheses. If the hypothesis does not predict the outcome correctly, it is either rejected or modified so that it is useful. In medicine, a therapeutic trial is the equivalent of the research study. If the outcome does not match the predicted outcome, then the diagnosis should be reconsidered. Admittedly, a fair test of a diagnosis may mean more than one therapeutic trial given the vagaries of psychiatric treatment (i.e., no single medication from any therapeutic class works for everyone with the same syndromic diagnosis). The point is that the treatment of every patient can be considered an experiment in which the physician is testing his or her diagnostic impression.

Another important principle that we can derive from this case is the importance of deciding that a drug has failed. In Joe’s case, the drug in question was buspirone. Regardless of whether you think buspirone was a good choice for Joe or not, it obviously did not produce the desired effect. If it had, Joe would not have ended up in the emergency room. Based on the equation in Figure 1, buspirone may not have worked because the dose was inappropriate to achieve a concentration that engaged the desired site of action (presumably the 5-HT1A receptor) to the right degree to treat Joe’s anxiety. If that was the assessment of the inpatient team, then the buspirone dose could have been adjusted in an attempt to optimize the response. Instead, another drug was added to the mix. With that move, Joe was on three psychiatric medications rather than two.

Another equally plausible explanation for buspirone’s therapeutic failure is that it simply did not act on the right mechanism of action to help Joe. If that was the opinion of the inpatient team, then it would seem prudent to stop it. A trial of a medication is just that--a trial. It works or it doesn’t. If it works, we continue it. Unfortunately, we sometimes also continue medications when they do not work, as in Joe’s case and in other cases I have presented in these columns.²

When we declare therapeutic failure, it is important to recall that it was the medication that failed rather than the patient or the physician. There is no need for emotions in arriving at this decision. While we do not want our treatments to fail, they are bound to do so on occasion. Accepting that fact can help us avoid the unnecessary use of multiple medications and unwittingly slipping down the slide of polypharmacy.

When patients do not do well, our first thought is frequently that they are not getting well despite the treatment. However, we should also consider the possibility that they may not be getting well because of our treatment. As I discussed in the case of the 17-year-old girl, unusual accumulation of the 1-PP metabolite of buspirone may cause anxiety and even manic-like symptoms ("Do You Believe in Magic," March 1997).² The production of this metabolite is mediated by the cytochrome P450 (CYP) enzyme 3A; however, its clearance is mediated by CYP 2D6. The 17-year-old girl became a phenocopy of CYP 2D6 deficiency as a result of co-treatment with paroxetine 20 mg/day.² Joe had approximately a 10% chance of being genetically deficient in CYP 2D6 because of his ethnic origin.

Such genetic factors are listed under biological variance in Figure 1 because they can shift the dose-response curve for a medication to a clinically significant degree. If Joe was genetically deficient in CYP 2D6, then treatment with buspirone may have played a causal role through its 1-PP metabolite in the worsening of his anxiety, thus contributing to his admission into the hospital. The point here is that we must be prepared to consider that the patient may not be improving because of our treatment rather than in spite of it.

As we have discussed in earlier columns, medications are added to produce a desirable outcome--however, the opposite may on occasion occur. In terms of the equation in Figure 1, nefazodone has several sites of action including:

Nefazodone-induced inhibition of CYP 3A raises the possibility of CYP enzyme mediated drug-drug interactions. This type of interaction is an important consideration when prescribing clozapine because of its dose-dependent (and hence concentration-dependent) risk of causing seizures.^7,8 If clozapine was principally dependent on CYP 3A, then the addition of nefazodone would be tantamount to increasing the clozapine dose, which could in turn increase the patient’s risk of having a seizure. Fortunately, the first step in the metabolism of clozapine is demethylation which is mediated by CYP 1A2 rather than CYP 3A.^7,8 Nevertheless, CYP 3A has been implicated in the metabolism of clozapine; however, specifics, including the biotransformation it mediates and the potential clinical consequences of changing the activity of this pathway, remain to be elucidated.

While the effect of nefazodone on clozapine is uncertain, in this case, nefazodone, through its inhibition of CYP 3A, would have slowed the conversion of buspirone to 1-PP. That could have led to an improvement in the patient’s anxiety at least to the extent that it had been aggravated by the buildup of this metabolite as discussed above.

The use of theophylline in this case is an example of why psychiatrists should not abrogate their responsibility for the management of the patient’s nonpsychiatric conditions to another physician. Someone needs to be looking at the big picture to avoid problems caused by tunnel vision (i.e., seeing only what is of interest or concern to us relative to our specific specialty to the exclusion of other equally or even more important issues). In Joe’s case, the use of theophylline is an example of tunnel vision.

As shown in the equation in Figure 1, other diseases can be an important contributor to interindividual biological variance and can shift the patient’s response to a medication. The most common examples are disease-induced impairment in hepatic, renal, or left ventricular cardiac function that can alter the pharmacokinetics of drugs to a clinically significant degree. In addition to these illnesses, diseases of the brain can also shift the dose-response curve to centrally active medications. An example of this situation from an earlier column is the 67-year-old woman who developed a profound worsening of her Parkinsonian symptoms following the addition of haloperidol to her treatment regimen.⁴

In a similar way, Joe was a prime candidate for a theophylline-brain disease interaction for a number of reasons. Joe had schizophrenia which had caused him to experience psychotic symptoms in the past. Theophylline can cause psychotic symptoms.⁹ As a result of his schizophrenia, Joe might conceivably have been more susceptible to the psychomimetic effect of theophylline than the average person. Moreover, Joe had been admitted to the hospital because of anxiety and insomnia, which are adverse effects of theophylline.⁹ Thus, theophylline and buspirone may have played a role in the increase in Joe’s anxiety level after his hospital admission.

In addition to the above problems, theophylline is an example of a drug with a long list of clinically significant pharmacodynamically and pharmacokinetically mediated drug-drug interactions. While such lists do not include clozapine, there are several reasons for caution when using these two drugs together. First, clozapine and theophylline both lower seizure thresholds.^7-9 Second, they are both metabolized by CYP 1A2, which raises the possibility of competition for the enzyme.^7-9 Unfortunately, that issue has never been formally tested.

Certainly, it is fortunate that the inpatient team put Joe on nefazodone rather than fluvoxamine. While fluvoxamine and nefazodone both inhibit CYP 3A, of the two only fluvoxamine inhibits CYP 1A2.^6,7 For that reason, co-administration of fluvoxamine has been shown to cause substantial increases in the plasma levels of both clozapine and theophylline.^7,8 Although buspirone is considered benign in terms of drug-drug interactions, its effects have not in fact been studied, making it a wild card in this case.

If the factors described above had not been present, then the use of theophylline might have been appropriate, even considering the modest nature of Joe’s respiratory complaints. However, the presence of these issues appreciably shifted the cost/benefit analysis against using theophylline in Joe. In fact, he did well without theophylline.

There may be other points you think I should have made in this discussion and you may disagree with some of the points I have made here. In either case, I would welcome hearing from you. Pending the receipt of such comments, the most important take-home points from Joe’s case from my perspective are the following:

This column is dedicated to the memory of Ronald L. Martin, M.D., who was professor and chairman of the Department of Psychiatry at the University of Kansas School of Medicine-Wichita and a mentor, colleague, and friend to me and countless others. Ron was a fine physician-educator who trained over 3,000 medical students and 250 psychiatric residents, setting high standards for competence, ethics and professionalism.