For this discussion, we define polypharmacy as the use of two or more medications to treat the same condition, use of two or more drugs of the same chemical class, or use of two or more drugs with the same or similar pharmacologic actions to treat different conditions, such as one benzodiazepine for generalized anxiety and another for insomnia. Thus the use of lithium and valproic acid together to treat bipolar disorder would be considered polypharmacy, whereas the use of sertraline and metformin to treat a patient who has depression and type II diabetes would not. We also would not count as polypharmacy the addition of a second medication to treat side effects of the first, such as the use of amiloride to treat lithium-induced polyuria.

Recent surveys have repeatedly documented the high frequency of polypharmacy in the treatment of seriously ill psychiatric patients (1,2). Polypharmacy is prevalent in other medical specialties as well, particularly those dealing with chronic diseases and the elderly population. Although there are many instances of "rational polypharmacy," most would agree that any use of multiple medications may increase the risk of adverse effects, drug interactions, patient noncompliance with treatment, and medication errors. In some cases the problem may be compounded by the fact that the patient is also taking over-the-counter or alternative medicines of which the physician is unaware. Polypharmacy has been associated with elevated mortality rates (3).

Obviously, in many instances management with multiple medications is necessary and constitutes rational polypharmacy. The classic example is the use of benztropine to manage side effects from haloperidol. Another example is the use of lithium to augment a partial response to an antidepressant and thus increase its efficacy. Certainly polypharmacy may be unavoidable or necessary for patients suffering from several chronic conditions.

However, irrational polypharmacy occurs too frequently. Examples include the use of several benzodiazepines or several antipsychotics at the same time. Given the possibility of problematic drug interactions in any polypharmacy, whether rational or irrational, it can never be approached with too cavalier an attitude. As we learn more about the cytochrome P-450 system and its substrates, inducers, and inhibitors, the best ways to avoid problems are to refrain from polypharmacy when one can and to plan carefully when one cannot.

Several different causes of irrational polypharmacy deserve mention. They can occur singly or in combination. The first is a combination of fear and laziness. A patient is unstable or doing poorly, so the clinician adds another medication, and perhaps it helps a little. Since the patient is still not doing well, the clinician is afraid to withdraw any of the drugs, and so the patient's list of medications gets longer and longer with no pruning—or improvement. Polypharmacy of this type can also occur when the patient is doing well and the clinician elects not to rock the boat. Using a benzodiazepine adjunctively to treat sleeplessness or anxiety in the early stage of treatment for panic or depression may be justified, but once the antidepressant has taken effect, continuing the benzodiazepine may do no more than create a drug dependence.

A second cause of irrational polypharmacy is what may be called sloppy diagnosis. For example, a patient with paranoia whose auditory hallucinations increase at night may describe his fear as nervousness or insomnia and receive a prescription for benzodiazepines rather than having his dosage of antipsychotic adjusted. Similarly, a patient who exhibits psychotic agitation may receive valproate or lithium rather than an adjustment of the dosage of antipsychotic because the agitation is presumed to be due to a schizoaffective mania—a condition that may be overdiagnosed clinically. Evident in these scenarios is the need to gather detailed information about the patient's symptoms and the symptoms' relationship to the primary illness as a whole rather than simply to treat each individual symptom.

A botched or stuck cross-titration is a third cause of irrational polypharmacy. While attempting to change the patient's medication, the clinician lowers the dosage of drug A while adding and raising that of drug B. In this cross-titration there may be a period during which the patient is doing worse because neither drug A nor drug B is at a high enough level. Some clinicians will then go back to the starting dosage of drug A but continue treatment with drug B. Alternatively, during the cross-titration, the patient may be doing better, and the clinician may assume that the improvement is due to the combination of drugs; the clinician stops there and leaves the patient on both drugs to avoid rocking the boat. Generally with cross-titration it is safer to get the new drug to a therapeutic level for a sufficient period before withdrawing the old one. However, many medication changes may not need cross-titration. For example, in a switch from perphenazine to olanzapine or another atypical antipsychotic, careful cross-titration may be unnecessary if the olanzapine is started at a sufficient dosage.

A variation of the stalled titration can occur when a clinician attempts to taper or discontinue sedatives. The patient may describe anxiety, which can represent either symptom reemergence and rebound or withdrawal effects from too aggressive a taper. The clinician might then conclude that the use of benzodiazepines is justified and restore the previous regimen.

A fourth cause of irrational polypharmacy is blind adherence to specifications listed in the Physicians' Desk Reference. The Food and Drug Administration sets the maximum dosage of a medication in collaboration with the drug company's somewhat arbitrary recommendations, usually with a wide margin of safety. If sertraline has a maximum dosage of 200 mg a day, and the patient's response suggests that more would help, some clinicians have added 20 mg of paroxetine instead, to avoid overdosing their patients, even though they know that ingestion of a whole bottle of sertraline would not be overly dangerous.

We do not intend to suggest that clinicians should feel free to increase doses without limit. At some point, other options should be considered. An example recently related to us was the use of 80 mg of olanzapine twice a day. Although that dosage is probably not overly dangerous, why not try clozapine? Combining medications with psychotherapy or other psychosocial approaches may also be a better approach than polypharmacy (4). Clinicians also should avoid succumbing to pressure from others, such as family members or insurance companies, to prescribe medications or dosages that are not in the patient's best interest.

A fifth cause of irrational polypharmacy is either an inadequate knowledge of receptor pharmacology or a lack of attention to it. Two examples we have observed will illustrate this. In the first, a patient suffering from schizophrenia was doing poorly on risperidone, so haloperidol was added. Since risperidone is a strong antagonist for the dopamine D₂ receptor, adding haloperidol increases the risk of side effects. In the second example, an incomplete response to sertraline, a selective serotonin reuptake inhibitor, led to augmentation with venlafaxine, which both blocks norepinephrine uptake and duplicates the actions of sertraline on serotonin. A switch would have been more appropriate than augmentation in this case.

The need to rush may also lead to irrational polypharmacy. Short hospital stays may create pressure to try to hasten a therapeutic response, although there is no evidence that such an approach is effective.

Yet another cause of irrational polypharmacy is poor exemplars. For example, an industry-sponsored seminar may use speakers who are less than expert and who promote the use of the sponsoring company's medication in combination with other medications for applications that have not been well studied. The recommendations of these "experts" can encourage poorly planned polypharmacy.

A final cause of irrational polypharmacy is what may be called magical thinking. This approach usually stems from word-of-mouth reports about differences among medications of the same class—the benzodiazepines or the antipsychotics, for instance. The clinician might believe that one is better for sedation, control, or some other effect and that another of the same class is better for some other property, so the patient receives prescriptions for two very similar drugs.

Rational polypharmacy may be divided into two types: the validated and the empirical. We recommend that validated strategies, when they exist, be tried before other strategies if single-agent therapies in adequate dosages for an adequate trial duration have failed. For example, lithium augmentation of antidepressants has been validated in double-blind trials, and therefore it has greater weight than augmentations suggested by case reports or theory alone.

Clinicians should examine all of their patients and the medications they are taking. Patients who are taking multiple medications, especially medications of the same class, deserve particularly close scrutiny. Additionally, with the growing popularity of alternative therapies, this review should include asking all patients about over-the-counter medications or herbal remedies they might be taking.

Practicing rational polypharmacy often means conducting mini-experiments, even though they are uncontrolled and unblinded. Such trials may be more informative if the clinician uses objective ratings to evaluate the patient's response to medications. Familiarity with some of the widely used rating scales, such as the Clinical Global Impression Scale, the Hamilton Depression Scale, and the Brief Psychiatric Rating Scale, to name but a few, can help provide more comprehensive and objective ratings of patients' progress. We offer two brief examples here, both of patients taking clozapine.

Patient 1 is taking 600 mg of clozapine a day and has a plasma level of 700 ng/mL but is tolerating the drug. However, his score on the Positive and Negative Syndrome Scale (PANSS) is 55, which is high. You decide to increase D₂ blocking by adding risperidone or haloperidol—haloperidol would be much less expensive—even though previous trials of these medications alone showed little effect. The patient's clozapine level stays stable, but his PANSS score drops to 34. You can justify this polypharmacy even if there is no good research to support it.

Patient 2 is also taking 600 mg of clozapine a day, and you add valproate not just to protect against seizures—which may be justified—but also to improve the patient's response. It works, but you cannot be certain about whether the improvement is due to the valproate's effect on the brain or its effect on the liver, which leads to increased clozapine levels and possible toxicity. It should also be noted that there are no controlled data to suggest that the addition of valproate is effective in augmenting treatment response among patients with schizophrenia.

As psychopharmacology becomes more sophisticated, with new medications targeting either single receptors or performing specific combinations of actions, the possibilities for both rational and irrational polypharmacy increase exponentially. Empirical research is unlikely to investigate more than a small fraction of these possible combinations. In the absence of research-based guidance, clinicians will still need to approach the problem of poorly responsive patients with knowledge and rationality.

We propose the following principles for practicing polypharmacy.

• Therapeutic trials of sufficient duration and with optimal dosages of a single agent should be tried first.

• Polypharmacy may sometimes be useful or necessary.

• Polypharmacy does not hasten a therapeutic response.

• Polypharmacy may increase the risk of morbidity and mortality.

• Polypharmacy is frequently ineffective and expensive.

• Benzodiazepines are usually unnecessary as adjuncts to long-term therapy.

• Objective target symptoms or structured ratings should be used to evaluate the results of treatment.

• Ineffective or minimally effective medications should be discontinued.

• Researched strategies for polypharmacy should be tried first.

• Both pharmacodynamic and pharmacokinetic interactions should be considered.

The authors are affiliated with the department of psychiatry and the behavioral sciences at the Keck School of Medicine of the University of Southern California, 2020 Zonal Avenue, Los Angeles, California 90033 (e-mail, king7548@pol.net). Dr. Simpson is editor of this column.