Printed from:
Do you feel lucky?


Journal of Practical Psychiatry and Behavioral Health, January 1998, 37-40

What does Clint Eastwood have to do with clinical psychopharmacology? As the character, Dirty Harry, in the 1972 movie, he posed the question which is the title of this column. That question has universal application to many situations in life, including what we decide to pay attention to as physicians. If we feel lucky, we can ignore issues where we think there is a wide margin for error without the risk of serious consequences.

The title of one of the movies that launched Clint Eastwood’s movie career - The Good, the Bad, and the Ugly - is also relevant to the prescriber of medications. As any experienced practitioner knows, the consequences of using drugs in combination can range from good to bad to downright ugly. Sometimes, the physician has to count on luck when treating a patient; however, most of us prefer most of the time to rely on knowledge. After all, we know that the rules of chance favor the house, not the player, and that nature generally sides with the hidden flaw.

So what are the possible odds posed by the practice of polypharmacy? The answer is well over 642 billion. In this column, I will explain how these odds were derived and discuss how to avoid the bad and the ugly without having to rely on luck alone.

The topic of this column is central to the debate about the sort of training needed to prescribe medications. In psychiatry, this debate has been fueled by two independent phenomena: the effort to reduce health care costs and the development of psychiatric medications with incredibly wide therapeutic indexes (i.e., the difference between the usual therapeutic and the toxic dose). For example, older drugs such as lithium and tricyclic antidepressants had therapeutic indices of 5 or less, meaning five times the usual therapeutic dose could potentially cause serious toxicity, even in a medically healthy patient. In contrast, the therapeutic index of many of the newer antidepressants is so large that fatalities do not occur even when substantial overdoses of these medications are taken alone. That fact has given rise to the question: What does the prescriber need to know to write a prescription? Some have even proposed that many of the newer antidepressants are so safe they could be made available as over-the counter treatments.

However, these arguments ignore the fact that these medications are generally not taken alone but rather in combination with other medications. This fact can alter the safety concerns that the prescriber must consider. I have discussed this issue in several earlier columns (see my columns in the March and September, 1997 issues) and described cases of drug-drug interactions and the serious consequences that can occur even when generally safe medications are taken in combination and the prescriber did not consider their potential to interact either pharmacodynamically and/or pharmacokinetically to alter the clinical outcome.


Obviously, patients receiving monodrug therapy cannot experience a drug-drug interaction. Conversely, being on more than one medication does not mean that patients will necessarily experience a drug-drug interaction, but they are at potential risk. Moreover, the more medications a patient is taking, the greater likelihood that he or she will experience such an interaction. To evaluate the extent of this problem, we need to know the percentage of patients in clinical practice taking more than one medication at the same time.

To answer this question, my colleagues (Cheryl Carmichael, Anne Harvey, Ph.D., and Dale Horst, Ph.D.) and I surveyed different practice settings to determine the incidence and nature of polypharmacy in different types of clinical practice. To date, we have surveyed four types of clinical practice: primary care; general psychiatric outpatient care; outpatient and inpatient care at a university affiliated Veterans Administration Medical Center; and care of HIV-positive outpatients in a specialty clinic. In each of these settings, we surveyed only patients who were on at least one antidepressant and then ascertained the frequency and nature of polypharmacy in those patients (Table 1).

Table 1. Percentage of patients on antidepressants who have the potential to experience a drug-drug interaction as a function of treatment setting
Clinical Setting Number of Patients Percent prescribed only an antidepressant Percent prescribed at least one other medication Percent prescribed 3 or more other medications
Primary Care 2045 28% 72% 34%
Psychiatry Clinic 224 29% 71% 30%
VA Med Center and Clinics 1076 7% 93% 68%
HIV Clinic  66 2% 98% 77%

In each of the practice settings we surveyed, the vast majority of patients receiving an antidepressant were also taking one or more other medications. However, the nature and extent of polypharmacy varied among the different settings. As expected, the extent of polypharmacy was greater in the more medically ill populations (i.e., the VA and HIV populations) than in the primary and psychiatric care settings.

While that finding is not surprising, it is important for several reasons. First, these populations are more frail and thus more susceptible to the adverse consequence of a drug-drug interaction. Second, there is a greater likelihood that an adverse outcome due to a drug-drug interaction might be misidentified as a consequence of these patients’ underlying frail health than would be the case in a medically stable population. Therefore, these populations, due to their frail health, are more likely to experience a drug-drug interaction due to the increased extent of polypharmacy; the severity of the adverse outcome is likely to be greater; and the correct cause (i.e., an adverse drug-drug interaction) is less likely to be identified. Ironically, the misidentification of the problem may lead to even more polypharmacy, since another medication may well be added to treat a perceived worsening of the underlying medical illnesses.

As we expected, the nature of the polypharmacy also varied as a function of the type of medical setting. HIV-positive patients obviously had a much great likelihood of being on a variety of antiviral and other anti-infective medications. The VA patients were more likely to be taking a variety of medications for chronic illnesses such as hypertension, diabetes, arthritis, and cardiac, renal, hepatic, and gastrointestinal diseases. Psychiatric patients were more likely to be receiving other psychiatric medications for comorbid psychiatric syndromes, as augmentation strategies, or to treat adverse effects. The primary care patients were a mini-version of the VA population.


Even though the various medications a patient is taking may have been prescribed for completely different reasons and for completely different targets (sites of action), they may still interact. For example, most HIV patients with adequate access to care are now receiving one or more protease inhibitors. These medications were rationally developed to inhibit a viral enzyme whose function is essential for the production of infective HIV. By blocking this enzyme, these drugs prevent viral replication and appreciably slow the progression of the disease. Thus, these drugs are targeted at a site of action that is not normally found in the human body, but is instead an expression of the virus’s genetic code.

HIV patients are at increased risk for experiencing a depressive episode and are therefore likely to be treated with an antidepressant. Generally, the physician will favor one of the newer antidepressants because of their wide therapeutic index, as discussed earlier. Say that an HIV patient is prescribed a selective serotonin reuptake inhibitor (SSRI), a reasonable scenario since these are the most widely used of the newer antidepressants. The SSRIs were rationally developed to block the uptake pump for serotonin on serotonin neurons. In this scenario, is any interaction possible, given that the HIV drugs and the antidepressant were rationally developed to target completely different sites of action?

Almost all of the protease inhibitors also inhibit the cytochrome P450 CYP enzyme 3A3/4, which is responsible for approximately 50% of known oxidative drug metabolism. Let us say our HIV patient is on fluoxetine or paroxetine-which is a safe bet since the former is the most widely prescribed SSRI in this country and paroxetine is also quite popular. While these two drugs inhibit the serotonin uptake pump like other members of this class of antidepressants, they also produce substantial inhibition of CYP 2D6 at their usually effective minimum antidepressant dose. This CYP enzyme is responsible for approximately 30% of oxidative drug metabolism. Thus, protease inhibitors and these two SSRIs have a common, if unintended, target which is the CYP enzyme system responsible for the bulk of oxidative drug metabolism. Moreover, the two CYP enzymes they affect are responsible for approximately 80% of known oxidative drug metabolism. Now look back at Table 1 and see how many other medications our HIV patient is likely to be taking in addition to these drugs. Our HIV patient is clearly a prime candidate for CYP-mediated drug-drug interactions. To quote the 60s ballad, Big John, "If the right one (CYP 3A inhibition) doesn’t get you, the left one (CYP 2D6 inhibition) will."


Clinicians must therefore be knowledgeable and mindful of an enormous number of variables when prescribing medications, even those that by themselves are quite safe. To demonstrate the amount of knowledge that must be considered, we did some mathematical calculations of the number of possible combinations of medications that a given patient could be taking given the data we collected at the four practice sites. Altogether, 600 different medications were prescribed to the approximately 3500 patients at these four sites. The most medications any single patient was taking was 21! This was a VA patient, but, as you probably would guess, patients in the HIV clinic were not far behind.

A clinically important question is: Given the drugs used in clinical practice and the number of medications being prescribed for some patients, how many different combinations is it possible to prescribe? Summing the possible combinations of 600 medications taken 2 at a time (179,700), plus combinations taken 3 at a time (35,820,200), plus combinations taken 4 at a time (5.346 x 109), plus combinations taken 5 at a time (6.37 x 1011), adds up to 642 billion possible combinations for 2 to 5 medications from 600 choices (Table 2). If we account for all possible combinations from 2 to 21 (the largest number of drugs being taken by a single patient) that could occur with 600 different medications, this works out to 3.127 x 1038 possible combinations of medications!

Table 2. Possible combinations of 600 drugs using a maximum of 5 drugs at a time.
Exact Number Used in Combination Calculation Number of Possible Combinations
Two Drugs (600)    =      600!      =                (600) (599)             =
  (2)            2! 598!                            2 
Three Drugs (600)   =      600!      =             (600) (599) (598)          =
   (3)           3! 597!                             6 
Four Drugs (600)   =      600!      =          (600) (599) (598) (597)    =
   (4)           4! 596!                             24 
Five Drugs (600)   =      600!      =      (600) (599) (598) (597) (596)  =
   (5)           5! 595!                            120 
Up to 5 drugs =                        642,645,014,870

Obviously, there are far too many combinations for anyone to memorize. So how do we manage this information? Some have in essence suggested that we stick our collective heads in the sand and repeat to ourselves: "We and our patients have done perfectly OK without thinking about these matters. Everything will be fine." I do not advocate that approach, even for ostriches. Nevertheless, if that is the approach you have chosen, I wish you good luck.

On the other hand, it would also not be sensible to suggest that clinicians memorize billions of possible combinations. Even if it were possible to do so, it would be inadequate-particularly since more drugs are being developed every day. Talk about continuing medical education!


  1. Be aware that interactions can occur and consider them as possible reasons why the patient has not responded as you expected. In other words, consider the possibility that the patient might not have gotten better, and may even have gotten worse, because rather than in spite of the medication you selected.
  2. Be knowledgeable about the major interactions and, when you add a medication, consider the potential for interactions. Obtain information from sources such as the package insert, pharmacology texts, your hospital pharmacy or drug information service, and computer information services including Medline and drug-drug interaction software packages.
  3. When possible, use drugs that do not affect unnecessary targets, since effects on extraneous targets increase the risk of unnecessary drug-drug interactions in addition to other safety or tolerability problems. Such unnecessary targets may be receptors, as is the case with the older tricyclic antidepressants, or cytochrome P450 enzymes, as is the case with some of the newer antidepressants. In fact, one goal of rational drug development is to discover drugs that avoid such targets, thus reducing the risk of drug-drug interactions. This is obviously an important issue given the widespread incidence of polypharmacy in patients receiving antidepressants and other psychiatric medications.
  4. When you must use a drug with multiple targets and are uncertain about the potential for interactions, follow the adage from geriatric medicine to "start low and go slow." In such situations, go back to the starting point and be especially alert to the possibility that clinical worsening may occur because of, rather than in spite of, the medications that the patient is taking.
  5. Remember that stopping medications is also "doing something."
  6. Who should prescribe? Obviously, it should be a person who has the training necessary to follow the guidelines given above safely, effectively, and efficiently. Otherwise, you and your patient will need to count on luck.


  1. Janicak PG, Davis JM, Preskorn SH, Ayd FJ Jr, Principles and Practice of Psychopharmacotherapy. 2nd ed. Baltimore, Md: Lippincott, Williams & Wilkins; 1997
  2. Ciraulo DA, Shader RI, Greenblatt DJ, Creelman WL, Drug interactions in psychiatry. Baltimore: Williams & Wilkins; 1995
  3. Preskorn SH, Clinical Pharmacology of Selective Serotonin Reuptake Inhibitors. Caddo, Okla: Professional Communications, Inc; 1996
  4. Gilman AG, ed, Goodman and Gilman’s pharmacological basis of therapeutics, 9th ed. New York: McGraw-Hill; 1996
  5. Melmon KL, Morrelli HF, Hoffman BB, Nierenberg DW, Clinical pharmacology: Basic principles in therapeutics, 3rd ed. New York: McGraw-Hill; 1992