A Precautionary Tale in Psychiatry

Over the past 50 years, psychiatry has increasingly become psychiatric medicine coincident with the enormous developments in our understanding of and ability to effectively use clinical psychopharmacology to treat patients with psychiatric illnesses. There have been both increased understanding of the molecular mechanisms underlying the effects of psychiatric medications and increased numbers of psychiatric medications. The latter has occurred in tandem with a similar explosion in the availability of medications to treat a host of other medical conditions. In fact, the repertoire of available medications expands virtually every few weeks.

While this phenomenon provides an opportunity to treat more and more patients with increasing efficacy, it also poses challenges for the treating physician. Specifically, the availability of more medications means that the prescriber must be knowledgeable about all of those medications-including any prescribed by another physician.

Most drugs (with the exception of anti-infectives) are given to change patient biology, which is how they treat disease. For this reason, drugs are an important acquired form of biological variance in patients. In fact, drug treatment likely creates more clinically meaningful biological variance than do genetic differences in people. The reason is the widespread and complex practice of multiple medication use (MMU), which is often necessary but sometimes dangerous. The extent and complexity of MMU underscores the rigorous training needed by clinicians to prescribe medications safely and effectively and to avoid adverse outcomes.

For these reasons, this article reviews the scope and complexity of MMU in clinical practice and presents a theoretical framework along with basic principles to aid the practitioner in avoiding unintended and adverse consequences of MMU.

DEFINITION OF A DRUG-DRUG INTERACTION
A drug-drug interaction (DDI) involves alteration by a coprescribed drug of the nature, magnitude, or duration of the effect of a given dose of another drug.

Change in nature of the effect
This part of the above definition refers to the clinical scenario in which 2 drugs taken together produce an effect that is not expected from either drug alone. Such an effect is often sudden and catastrophic. Examples in psychiatry include serotonin syndrome, hypertensive crisis, delirium, and seizures. Examples in general medicine include cardiac arrhythmias and sudden death.

While the clinical consequences of such DDIs may be readily apparent, the fact that they are caused by a DDI may still be missed. Instead, the consequences may be misattributed to a variety of causes, including other medical causes and even suicide attempts.^1,2 While some prescribers may only think of DDIs in terms of such sudden and catastrophic outcomes, DDIs of this type occur much less often than do DDIs that result in a change in the magnitude and duration of the patient's response to the medication treatment, whether desired or undesired.

Change in magnitude and duration
A change in magnitude occurs when the effect observed is greater or less than the effect that would have been predicted based on the dose given of the affected drug. A change in duration occurs when the effect is longer or shorter than the effect that would have been predicted based on the dose given of the affected drug. In other words, the magnitude or duration of the effect observed is one that the prescriber could have anticipated had a different dose of the affected drug been given to the patient.

For this reason, DDIs that produce a change in magnitude or duration are more difficult to detect in practice because they may easily be attributed to the patient being either "sensitive" or "resistant" to the effects of the prescribed drug. While this superficial explanation is true, the treating physician may not realize that the "sensitivity" or "resistance" is not inherent in the patient but is caused by an acquired change in the patient resulting from the presence of a coprescribed drug. These DDIs may also mimic many different clinical scenarios, including the apparent worsening of the disease being treated,^3-5 the appearance of a new disease,⁶ lack of efficacy (ie, the patient is resistant to beneficial drug effect),⁷ poor tolerability (ie, the patient is sensitive to adverse drug effect),^8-10 and withdrawal symptoms or drug-seeking behavior on the part of the patient.^11-14

Clinical relevance of DDIs
There has been some debate on how to determine when a DDI is clinically relevant.¹⁵ A conservative position is that only DDIs that result in a serious adverse event as defined by the FDA should be considered clinically relevant. The FDA's definition of a serious adverse event is any event that produces one or more of the following outcomes:

• Death (or is life-threatening).
• Persistent or significant disability/ incapacity.
• Hospitalization or prolongation of existing inpatient hospitalization.
• Congenital anomaly/birth defect.
• Cancer.
• Clinically significant overdose.
• Other important medical events that occur in an emergency department or ambulatory surgical cen- ter and require medical or surgical intervention to prevent death or hospitalization.

Certainly, few would argue that a DDI that causes a serious adverse event is not clinically relevant. However, this threshold, while having high specificity, lacks sensitivity (ie, too many false negatives).

A more liberal approach is that a clinically relevant DDI is any DDI that results in an untoward or unfavorable outcome of any severity, including an outcome that is less desirable than could be reasonably expected. Some might argue that this approach, while highly sensitive, lacks specificity (ie, too many false positives).

A third approach is that a clinically relevant DDI is one that results in a change in treatment, whether by the physician, the patient, or a third party such as the FDA.^2,3 Such a change could include:

• The discontinuation of one or more of the drugs involved.
• The addition of another drug to the treatment regimen.
• Additional office visit(s).
• Additional diagnostic tests.
• The removal of the drug from the market.

Parenthetically, several drugs were removed from the market over the past decade not because they were dangerous when used alone but because they were dangerous only when used in combination with other specific drugs (eg, terfenadine and ketoconazole).

A conceptual model
Two equations that provide a framework for understanding the effect of any medication when used alone or in combination with other drugs are presented in Figure 1. Equation 1 identifies the 3 variables that determine the patient's response (ie, the effect) to any medication. They are (1) the affinity for and intrinsic activity of the drug at its site(s) of action, (2) the concentration of the drug achieved at its site(s) of action, and (3) the specific biology of the patient that can make him or her an outlier (ie, either sensitive or resistant) on the usual dose-response curve for a drug. Equation 2 makes the point that the concentration achieved at the site of action is a function of the dose patients take relative to their ability to clear the drug from their body.

In the first equation, the third variable (ie, the patient's biology) is a modifier of the first and/or second variables in the same equation. The equation further lists the 4 major sources of interindividual variability among patients: genetics, age, disease, and environment (internal).

Relative to the third variable in Equation 1, prescribers regularly take age (particularly when dealing with very young and very old patients) and disease (eg, liver or renal failure) into account when prescribing a medication. For example, the clinical adage about starting low and going slow when initiating therapy with a new medication in an elderly patient uses age as a surrogate for the physiological changes associated with advanced age that can make the patient more sensitive to the effects of a drug.

The internal environment under the third variable in Equation 1 is made of what the person consumes and includes the drugs they are taking. Figure 2 illustrates the ways in which drugs can interact either pharmacodynamically or pharmacokinetically.

Recently, there has been considerable interest and enthusiasm about the potential for pharmacogenomics to improve the efficacy, safety, and tolerability of medications by specifically selecting a medication based on the genetic uniqueness of the patient,or "individualized treatment." That enthusiasm is based on the anticipation that improved understanding of genetics will allow prescribers to better take into account genetically determined interindividual variability among patients, which can alter the outcome of drug therapy (ie, the genetic component under the third variable in Equation 1).

While there is little doubt that genetics, age, and disease are important sources of biological variance and can alter the outcome of drug therapy, there is an underappreciation of the importance of the biological variance produced by drug treatment. That source of interindividual biological variance is what underlies the occurrence of clinically relevant DDIs.

BIOLOGICAL VARIANCE PRODUCED BY MULTIPLE MEDICATION USE
There are approximately 3500 different chemical entities as prescription drugs on the US market.¹⁶ Given that number, 520 quadrillion different combinations of up to 5 drugs could theoretically be prescribed to a patient. Why use 5 drugs in this example? According to a recent report by the US Department of Health and Human Services, 7% of Americans aged 18 years or older had taken 5 or more prescription drugs in the week before the survey.¹⁷ The same report found that 20% of all Americans aged 65 years or older had taken 5 or more prescription drugs in the week before the survey.

Since the turn of the millennium, an average of 17 new drugs has entered the US market every year. That translates into a new drug becoming available for prescription every 3 weeks. That, in turn, means that every 3 weeks a treating physician can encounter a patient receiving a medication that had previously not been available. Depending on the pharmacodynamics and pharmacokinetics of the new drug, it may also mean a source of new biological variance among patients.

Mathematically, the first new drug marketed in 2007 can be used in approximately 150 trillion different 5-drug combinations (ie, the new drug plus up to 4 other drugs already on the market when it enters the market). Needless to say, the safety and tolerability of all of those potential combinations were not tested before the new drug was approved. That is the reason why the careful and well-trained prescriber is the major safeguard against clinically relevant DDIs. That is also a major reason why some of the adverse effects of new medications come to light only after they have been marketed. The reason is that those adverse effects only occur because of the presence of another drug (ie, DDIs such as the fatal arrhythmias that occurred when terfenadine was used in combination with ketoconazole).

PREVALENCE OF MMU
Multiple pharmacoepidemiological studies have documented that MMU is widespread and growing in the United States and Europe.^18-26 As previously mentioned, a random survey of Americans by the US Department of Health and Human Services¹⁷ found that 7% of Americans 18 years or older took 5 or more prescription drugs in the week before the survey. Not surprisingly, those numbers increase when the individuals surveyed are patients.

For example, a recent pharmacoepidemiological study revealed that the average outpatient in the Veterans Administration Health System (VAHS) took 4 or 5 drugs depending on his age and whether he was taking antidepressants.^19,20 Even more impressive was the observation that up to 96% of these patients (depending on age and whether they were taking an antidepressant) were on unique regimens (ie, no other patient in the sample of a 1000 or more patients was taking the same specific combination of medications). As already discussed, drugs are a source of clinically important biological variance among patients. This VAHS study thus showed that an enormous number of these patients are biologically unique from every other patient when viewed solely in terms of what medications they are taking.

There are at least 5 reasons why patients are taking more than 1 medication (Table 1). Of all of these reasons, perhaps the most compelling is that the patient has more than 1 illness. Even in this situation, the treating physician must consider whether the various medications being taken-even though aimed at different conditions-might interact to adversely affect the patient.

Further increasing the complexity of the situation, the patient who has different medical conditions is likely to have more than 1 physician from different specialities treating him or her. That poses a potential communication problem among those different prescribers because their individual prescribing actions may have a combined effect in the patient that is different from what they might have otherwise anticipated.

This issue was underscored by the aforementioned VAHS study in which the percentage of patients taking 8 or more medications doubled for every prescriber that a patient was seeing up to 4 prescribers, at which point 80% of such patients were taking 8 or more medications.¹⁹

PSYCHIATRY AND MMU
While MMU is common in all patient populations, there is evidence that psychiatric illness increases the risk of MMU. In the VAHS survey, antidepressant use was a greater determinant of MMU than older age.^18,19 While the reasons for this observation have not been definitively elucidated, the following are possible explanations:

• Patients with psychiatric illness frequently have other comorbid medical conditions as well as multiple psychiatric syndromes.
• Patients with psychiatric illness frequently present with somatic complains that may lead to symptomatic treatment. This may be particularly true for patients presenting in primary care before an underlying psychiatric illness has been recognized.
• Patients with psychiatric illness have been found to be high utilizers of health care services, and consequently, may see multiple prescribers. As discussed above, there is a dose- response curve between the number of prescribers a patient sees and his likelihood of taking 8 or more medications.

Another factor is the growing use of polypsychopharmacology. For example, from the 1970s to the 1990s, there was a 14-fold increase in the likelihood that a patient being seen by the biological treatment branch of the intramural program of the National Institute of Mental Health would be receiving 3 or more psychiatric medications.²⁷

Of note, the adverse effects of psychiatric medications alone or in combination can present as changes in brain function; hence, psychiatric and neurological signs and symptoms may be mistaken for changes in the underlying nature of the patient's neuropsychiatric illness. That applies today more than in the past because advances in psychiatric drug development have resulted in medications that are more highly and selectively targeted to brain mechanisms than were earlier psychiatric medications such as tricyclic antidepressants and low-potency antipsychotics. Those earlier drugs often had substantial peripheral effects that occurred before adverse effects on the CNS did. As a result of newer drugs being more specific in their effects on the brain, the careful clinician must be alert to the possibility that the worsening of the patient's mental status and behavior may be caused by the medications they are taking rather than simply attributing it to a worsening of their underyling illness.

Table 2 and Table 3 present a list of pharmacological and clinical principles that may be helpful to consider when treating a patient who is taking more than 1 medication.²⁸