Email this   Print this

Multiple Medications, Multiple Considerations

SHELDON H. PRESKORN, MD

Journal of Psychiatric Practice, January 2001, 48-52

This column will be a brief detour from the series on the Human Genome Project to discuss some recent results that expand on a previous focus of the column -- the problems posed by the patient who is on multiple medications. This column then is another in the clinical pharmacology case conference (CPC) series.^1-4 Following the CPC format, this column begins with a real-life case. The reader will then be asked a series of questions. Finally, the "take home" messages of the case will be discussed.

The case

JQP is a 52-year-old male receiving his care through the Veterans Administration (VA) healthcare system. He is seeing four different prescribers and is on eight medications: acetaminophen, cimetidine, codeine, erythromycin, ibuprofen, metoprolol, paroxetine, and thiothixene. To simplify the discussion, we will focus only on the following four drugs: codeine, erythromycin, metoprolol, and paroxetine.

Clearly, there is a distinct indication for each of these medications. Poetic license was taken in Table 1 to assign a prescriber to each medicine for the sake of the discussion. The reader is asked to consider these questions:

• Do these medications interact?
• If so, how?
• If so, how would the interaction present clinically?
• What might a prescriber do in response to that clinical presentation?

In the spirit of a CPC, take a moment, if you wish, to formulate your responses to these questions.

Background

Before giving responses to these questions, we will provide some background on how this case came to our attention. Our research group has been conducting an examination of the frequency and nature of polypharmacy in a Veterans Administration Integrated Service Network (VISN). The national VA health system is grouped into 22 regional VISNs for administrative and management purposes.

Current prescription data were extracted from the computerized pharmacy database on 4,857 patients selected at random and 2,779 patients selected solely based on the fact that they were taking at least one antidepressant. The group not taking an antidepressant was on average taking four systemically administered prescription drugs, whereas the group taking an antidepressant was on average taking five medications. This result, along with other findings from our study and the factors summarized in Table 2, makes it clear that being on an antidepressant appears to be a risk factor for being on a larger number of medications.

An intentionally conservative approach was taken in collecting these data. First, the count did not include locally acting medications (e.g., topical skin products).

Vitamins, minerals, herbal products, and illicit drugs were also not counted. Locally acting drugs were excluded because it was believed that most would not achieve a high enough systemic concentration to pose the risk of a clinically meaningful drug-drug interaction (DDI). Vitamins, minerals, herbal products, and illicit drugs were not included because the ability to obtain an accurate count was questionable. Instead, the focus was on systemically administered drugs because they would be recorded in the pharmacy dispensing records. Parenthetically, the fact that vitamins, minerals, herbal products, and illicit drugs were excluded does not mean that these products cannot interact in a clinically meaningful way with prescription drugs or with each other but simply that the goal was to understate rather than overstate the results.

The database was queried to determine the most commonly used combinations. The goal was to establish whether any of these combinations contained drugs that were likely to interact either pharmacodynamically or pharmacokinetically or whether there was insufficient information to know whether they would interact. For example, it is estimated that enough is known about the metabolism of only 20% of commonly prescribed drugs to know whether they are susceptible to a specific cytochrome P450 (CYP) enzyme-mediated DDI. This is because many drugs were developed and marketed before the technology was available to determine how they are metabolized. Once drugs are marketed, it is rare for further work to be done on their metabolism. Thus, an older drug may be susceptible to a CYP-enzyme-mediated DDI, but this may not be known and, therefore, cannot be included in a drug interaction alert system.

Much of our knowledge about clinically significant DDIs has been discovered only after drugs have been marketed. That knowledge is generally restricted to DDIs that produce a serious and immediate adverse effect that is clearly distinct from the natural course of the patient's underlying health problems. Take for example, sudden death in a patient taking terfenadine (Seldane) and ketoconazole for a vaginal yeast infection.^8,9 That is an obvious and unexpected adverse outcome in a physically healthy young patient. However, sudden death, while obvious, would not be unexpected in an elderly patient with a history of myocardial infarctions. In such a patient, the consequence of the DDI might easily be misattributed to underlying health problems. In a similar manner, the adverse central nervous system effects of a DDI may also be confused with a worsening of the patient's underlying neurological or psychiatric disease, as discussed in earlier columns.^1,3

Moreover, our knowledge is primarily limited to the effect of drug A on drug B. Existing drug alert systems are based on such a binary approach. The problem is that many patients, like JQP, are taking multiple medications. The questions then are: Do these medications interact and are the interactions limited to one drug affecting another or are they more complex?

Was there an interaction in JQP?

Figure 1 shows the known interactions among the four drugs JQP was taking. Codeine is an inactive prodrug that must be converted by CYP 2D6 to morphine to produce analgesia.^10,11 Metoprolol is a beta-blocker whose clearance is dependent on CYP 2D6 mediated biotransformation.¹² Paroxetine produces substantial inhibition of CYP 2D6 under usual dosing conditions.^13,14 While paroxetine is metabolized by CYP 2D6 at low concentration, it saturates this enzyme under usual dosing conditions, which accounts for its nonlinear pharmacokinetics.^13,14 At higher concentrations, paroxetine is most likely dependent on CYP 3A mediated biotransformation for its elimination. ^15,16 CYP 3A is substantially inhibited by erythromycin under usual dosing conditions.^17-19 The inhibition of CYP 3A by erythromycin should produce an increased accumulation of paroxetine, which in turn would produce more inhibition of CYP 2D6, which in turn would lead to less conversion of codeine to morphine and more accumulation of metoprolol. Obviously, drug interactions can be complex.

How might this interaction present?

Due to inhibition of the conversion of codeine to morphine, the patient should have less than optimal pain control. This could be construed correctly as lack of efficacy on the part of codeine -- but as a result of a DDI rather than because the patient is unresponsive to morphine analgesia. However, this effect might also be misconstrued as the patient seeking more opiates. The lack of pain relief might also be seen as consistent with the patient's clinical depression.

A sufficient accumulation of metoprolol can lead to profound hypotension as a result of reduced cardiac output due to the chronotropic (i.e., decreased heart) and ionotrophic (i.e., decreased stroke volume) effects of blockade of beta adrenergic receptors. However, more modest increases in metoprolol levels might simply present as fatigue, which could be misconstrued as a worsening of clinical depression.

Figure 1 - Cytochrome P450 (CYP) mediated drug-drug interactions between codeine, erythromycin, metoprolol and paroxetine.

What might be done in response to the DDI?

An apparent worsening of the patient's depressive syndrome could lead to a decision to increase the paroxetine dose, further worsening the problem. Such a dose increase would worsen the situation by increasing the levels of paroxetine and further inhibiting CYP 2D6 function and thus further increasing the levels of metoprolol. Eventually, the adverse effects on blood pressure resulting from the elevated metoprolol levels would become apparent but the cause might still not be understood by the prescriber.

Given the issues illustrated in this discussion, it is easy to understand why physicians see the results of such interactions but miss the cause, as discussed in an earlier column.² In this case, the situation could easily be interpreted as the patient being "resistant" to the analgesic effects of codeine, which would then lead to him being switched to another medication, and as being "sensitive" to metoprolol, which would lead to a dose reduction or switch to another antihypertensive medication. The patient could also be seen as being "sensitive" to paroxetine, since he would be expected to develop higher than expected levels and hence effects for the dose prescribed because of the inhibition of paroxetine clearance due to the CYP 3A inhibition caused by erythromycin. Again, the prescriber of paroxetine might either lower the dose of this medication or switch to another medication.

Each of these "resistant" and "sensitive" interpretations would actually be correctbut not because of something inherent in the patient but rather because of the other medications that he was taking. Ironically, the patient's "sensitivity" and "resistance" would change when the course of treatment with erythromycin ended. If the dose of the medication had been reduced without understanding why the patient was sensitive to it, then the levels could drop below a critical value to be effective and the patient could relapse, leading to the need for more clinical attention. This explains why it is so important for such interactions to be recognized in healthcare systems: they may not result in a clinically serious adverse outcome because of prescriber adjustments but they can still cost money because of the increased clinical attention that is required.

Such a scenario might be even more likely in this case because four different prescribers were treating the patient. Nevertheless, correct identification of a clinically meaningful DDI is a problem even when there is only one prescriber, as illustrated in the case of neuroleptic malignant syndrome that was described in an earlier column.² In that case, the prescriber did successfully manage the patient's adverse DDI. However, a lengthy (and expensive) stay in an intensive care unit was needed to do so. Despite successfully managing the patient, the prescriber wrote up the case for publication not realizing it resulted from a complex DDI.

Another example of the expense incurred because of DDIs is the case of an apparent severe worsening of Parkinson's disease which led to a 2-week hospitalization and the involvement of multiple consultants, described in another previous column.³ In that case, the prescriber held firm to the belief that the problem was a worsening of the patient's underlying condition, until the patient fully recovered more than 2 months after stopping the offending agent (i.e., fluoxetine) due to the protracted time needed to clear that drug and hence its resulting adverse effect. That case also suggests that prescribers may prefer not to recognize the role they may have played in the patient's adverse outcome as a result of their prescribing practices.

What about a DDI alert system?

To our knowledge, no existing DDI system would inform the physician about the totality of the complex interactions going on in the patient JQP nor about their likely clinical presentation nor about what the clinician could do to minimize the risk of a clinically serious adverse outcome. There are DDI systems that might alert the physician to one or more of the interactions but not to all of them and not to their interdependence. One reason for this is that few formal, prospective studies have ever been done to examine the interaction of more than two drugs at one time.

How does the body "think" about DDIs?

This case illustrates the fact that the body does not care why a drug was prescribed (i.e., its indication) nor who prescribed it. The body cares only about the pharmacodynamics and the pharmacokinetics of the drugs. In other words, DDIs can be understood using our basic equation:

(Equation 1) 
Effect = drug affinity for X drug  X biological
         site of action      level   variance 

The first variable is the pharmacodynamics of the drug. The second is its pharmacokinetics. The third is biological variables that can shift the dose-response curve, making the patient either more sensitive or resistant to the effect of the drug. DDIs occur when one drug acts as the third variable to shift the dose-response curve of another drug by interacting either pharmacodynamically or pharmacokinetically with it. In the case presented in this column, each drug directly or indirectly interacted with every other drug prescribed for the patient.

The body also does not care whether the patient has one or more prescribers nor what their medical specialties may be nor whether they are prescribing their specific medications for quite different indications (Table 1). This case thus illustrates that specialists cannot prescribe medication as if they were in a vacuum. Instead, they must consider all other medications the patient is taking, including over-the-counter drugs, herbal preparations, and even dietary intake (e.g., of alcohol and grapefruit juice). Thus, the infectious disease expert should be as knowledgeable about the potential effect of erythromycin on paroxetine as is the psychiatrist.

Unfortunately, this may not be the case, which is one of the reasons the FDA has encouraged the removal of some drugs such as terfenadine from the market. In discussing the decision to remve terfenadine from the market, Peter Honig, director of the FDA office of postmarketing drug risk assessment, was quoted as saying: "Seldane was a huge drug, and it was a fine drug, if used appropriately. But the system (the medical care system) proved incapable of preventing the drug interactions. We've learned our lesson."²⁰

Thus, one way to deal with this problem has been to pull drugs from the market. A second is to not approve them or to screen them out from drug development.²¹ A third is to take them off formularies. A fourth is to develop improved information systems to alert physicians to potential DDIs and then provide advice about how to minimize them (e.g., by a drug substitution or dose adjustment).

There are severe limitations in our knowledge in using the latter approach. Our knowledge of the mechanisms by which drugs interact is still rudimentary in many instances. As mentioned above, there are significant gaps in our knowledge about the precise enzymes that are important in the biotransformation of many drugs as a necessary step in their eventual elimination from the body. That information was available for the four drugs involved in the case of JQP and that fact enabled us to understand the complex way in which those drugs interact, as illustrated in Figure 1. However, that information is lacking for many other drugs.

Even when the information does exist, many DDI programs are not mechanistically driven but rather are simply based on empirical data derived from sources ranging from case reports to formal, prospective trials. Such systems will alert the physician when a specific drug is used in combination with another specific drug but only if an interaction has been reported with those specific drugs. For example, a DDI system might alert a physician when fluoxetine is being used in combination with metoprolol because there has been a case report published on this interaction.²² However, the same system might not alert the physician when paroxetine is used in combination with metoprolol because there have been no case reports or formal studies. Yet, paroxetine would be predicted to have an effect on metoprolol clearance comparable to that of fluoxetine on the basis of the comparable inhibition of CYP 2D6 produced by these two antidepressants at the lowest, usually effective antidepressant dose.^13,14

Given the limitations in our current knowledge and systems, physicians need to be knowledgeable about and cognizant of DDIs. They should consider the possibility of DDIs when selecting medications for a given patient. While they will likely want to use a DDI alert system or may believe that they can rely on the pharmacy to catch any possible DDIs, this is not necessarily the case for the reasons outlined above. This is the reason that physicians will want to be cautious whenever they are prescribing drugs that are known to cause substantial inhibition of CYP enzymes to a patient who is taking other medications, just as they would be cautious when prescribing drugs to patients taking monoamine oxidase inhibitors. They must also consider all the medications the patient is taking, not just the medications they are prescribing. Whenever a patient is taking more than one medication and doing less than optimally, prescribers should consider the possibility that the patient may not be doing well because of the medications they are taking rather than in spite of them. In other words, DDIs should be part of the differential diagnosis in any patient who is taking more than one medication and is doing less than optimally. This issue is important both in terms of patient outcome and the cost of healthcare delivery. If these issues are not kept in mind, DDIs ironically can lead to more polypharmacy and the use of more expensive forms of health care, as illustrated in previous CPC columns.^1-4

References

1. Preskorn SH, Do you believe in magic? J Pract Psychiatry Behav Health 1997;3:99-103
2. Preskorn SH, I don’t see ’em. J Prac Psych Behav Hlth. 1997; 3:302-307
3. Preskorn SH, A message from Titanic. J Prac Psych Behav Hlth. 1998;4:236-242
4. Preskorn SH, The slippery slide. J Pract Psychiatry Behav Health 1999;5:50-5
5. Katon W, Schulberg H, Epidemiology of depression in primary care. Gen Hosp Psychiatry 1992;14:237-47
6. Dworkin SF, VonKorff M, LeResche L, Multiple pains and psychiatric disturbance: an epidemiologic investigation. Arch Gen Psychiatry 1990;47:239-44
7. Preskorn SH, Outpatient management of depression: A guide for the primary-care practitioner, 2nd Edition. Caddo, OK: Professional Communications; 1999
8. Carlson A, Morris L, Coprescription of terfenadine and erythromycin or ketaconazole: an assessment of potential harm: Pharmacists can be key in changing prescribing behavior to avoid drug-drug interactions. Science & Practice 1996;NS36:263-9
9. Albengres E, Le LH, Tillement JP, Systemic antifungal agents. Drug interactions of clinical significance. Drug Saf 1998;18: 83-97
10. Chen ZR, Irvine RJ, Bochner F, et al, Morphine formation from codeine in rat brain: a possible mechanism of codeine analgesia. Life Sci 1990;46: 1067-74
11. Sindrup S, Brøsen, Bjerring P, et al, Codeine increases pain thresholds to copper vapor laser stimuli in extensive but not poor metabolizers of sparteine. Clinical Pharmacology and Therapeutics 1991;49: 686-93
12. Belpaire FM, Wignant P, Temmerman A, et al, The oxidative metabolism of metoprolol in human liver microsomes: Inhibition by the selective serotonin reuptake inhibitors. Eur J Clin Pharmacol 1998;54: 261-4
13. Preskorn SH, Clinically relevant pharmacology of selective serotonin reuptake inhibitors. An overview with emphasis on pharmacokinetics and effects on oxidative drug metabolism. Clin Pharmacokinet 1997; 32:1-21
14. Naranjo CA, Sproule BA, Knoke DM, Metabolic interactions of central nervous system medications and selective serotonin reuptake inhibitors. Int Clin Psychopharmacol 1999;14(suppl 2):S35-47
15. Harvey AT, Preskorn SH, Cytochrome P450 enzymes: interpretation of their interactions with selective serotonin reuptake inhibitors. Part I. J Clin Psychopharmacol. 1996;16:273-285
16. Harvey AT, Preskorn SH, Cytochrome P450 enzymes: interpretation of their interactions with selective serotonin reuptake inhibitors. Part II. J Clin Psychopharmacol. 1996;16:345-355
17. Honig PK, Woosley RL, Zamani K, Conner DP, Cantilena LR Jr, Changes in the pharmacokinetics and electrocardiographic pharmacodynamics of terfenadine with concomitant administration of erythromycin. Clin Pharmacol Ther 1992;52:231-8
18. Hersh EV, Adverse drug interactions in dental practice: Interactions involving antibiotics. Part II. J Am Dent Assoc 1999;130:236-51
19. Yasui N, Otani K, Kaneko S, et al, A kinetic and dynamic study of oral alprazolam with and without erythromycin in humans: In vivo evidence for the involvement of CYP3A4 in alprazolam metabolism. Clin Pharmacol Ther 1996;59:514-9
20. Davis R, Appleby J, Patient safety being left behind. USA Today . Oct 11, 2000
21. Preskorn SH, The human genome project and modern drug development in psychiatry. Journal of Psychiatric Practice 2000;6:272-276
22. Walley T, Pirmohamed M, Proudlove C, et al, Interaction of metoprolol and fluoxetine. Lancet 1993;341:967-8. PSYCHOPHARMACOLOGY Journal of Psychiatric Practice 52 January 2001