Iam not an expert in managed care.Instead, I have spent a considerableamount of time trying to understanddrug—drug interactions (DDIs) in terms ofthe mechanisms underlying such interactions,the way they present in clinical practice,and how they can affect patientoutcome. During this time, I have come tobelieve that most people involved in healthcaredelivery seriously underestimate theclinical and, hence, economic importance ofDDIs on treatment outcomes.
Most physicians only think about DDIs interms of serious, catastrophic outcomes,such as death, seizures, arrhythmias, anddelirium. Although those outcomes areimportant, they represent only the tip of theiceberg. DDIs can present in almost any wayclinically imaginable. When making presentationsto physicians across the country, Ihave also been struck by the frustration, andsometimes hopelessness, that they feel whenfaced with edicts from managed care to prescribeone drug over another or to switch apatient from one medication in a therapeuticclass to another because it is cheaper. Yetsuch changes can potentially alter the rest ofthe patient's drug regimen, sometimes dramatically,leading to negative patient outcomesand increased healthcare costs.
This paper will review the impact of DDIsand therapeutic substitutions. Specifically, Iwill cover:
The Complexity of Medication Regimensin Practice
To put this issue into perspective, considerthe following 4 questions:
Physicians' Desk Reference
3500 according to the 20041 5.2 Ã— 1017 149 trillion Approximately 20
These numbers are, of course, mathematicalpossibilities. As the reader, you mightdismiss them in several ways. You might ask,"Well, how many people are actually taking5 different medications?" Or, you may say,"These numbers are mathematical possibilities,but physicians must use common drugcombinations." You might also say, "Theenormity of these numbers is deceiving,because drugs can be grouped into meaningfulclasses."
The problem is that none of these counter argumentsis correct. Based on a recentsurvey,2 7% of all Americans older than theage of 18 years were taking 5 prescriptiondrugs the week before the survey. Based onour own studies,3 the average outpatientseen in a Veterans Administration (VA)health setting is taking 5 medications. Ofcourse, VA patients are typically older andare likely to take more medications. Agingpatients tend to use medications to preventor slow the development of long-term problems,as opposed to using drugs only to treatacute problems. Given the anticipatedgrowth in the elderly population in theUnited States in the next 20 years–nearly35 million Americans were older than 65years of age in 2000 and that number isexpected to grow to 83 million by 20504–alarge portion of the US population will soonbe at increased risk for complex multiplemedication use.
Based on our surveys examining polypharmacyin VA outpatients, the mediannumber of drugs a VA patient younger than60 years of age taking an antidepressant versusnot taking an antidepressant, was 5 versus2, respectively.5 For patients 60 yearsand older, the comparable numbers were 6versus 4, respectively.5 Overall, taking anantidepressant was a greater risk factor forcomplex multiple medication use than age.Moreover, 70% of younger VA patients nottaking an antidepressant were taking aunique drug regimen (defined solely on thebasis of the drugs the patient was takingwithout regard to dose or schedule) versus80% of older VA patients not taking antidepressantsand 86% and 97% of younger andolder VA patients taking antidepressants,respectively. Of note, 363 to 394 differentdrugs were used to treat these various patientpopulations. As discussed above, that isapproximately one tenth of the 3500 prescriptionson the US market. Thus, the numberand diversity of medication regimensused to treat patients in a conventional practicecould be substantially higher than thoseused by these VA outpatients, and thus, thediversity of their medication combinationswould also be expected to be higher than wasfound in these VA outpatients. This diversitymeans that few, if any, prescribers in the VAsystem have extensive experience with thetotality of the effects of all of the medicationstheir patients are receiving. The same is likelytrue for other prescribers, especially whenthe patient is seeing more than 1 prescriber,which is becoming increasingly more commonas the US population ages.
The Mechanisms Underlying DDIs
Prescribers generally think of drugs interms of the disease they treat and theirtherapeutic class, but, in fact, drugs can beclassified in at least 4 different ways accordingto (1) structure, (2) pharmacodynamics,(3) pharmacokinetics, and (4) therapeuticuse. Drugs interact as a result of their pharmacokineticand pharmacodynamic propertiesrather than their therapeuticindications. Pharmacokinetics is the studyof the time course of drug and metabolitelevels in different fluids, tissues, and excretaof the body, and of the mathematicalrelationships required to develop models tointerpret such data. However, the merepresence of a drug within the plasma andother corporal systems is not a measure ofits activity. This is where pharmacodynamicscomes into play as the study of the biochemicaland physiological effects of drugsand their mechanisms of action.
The effect of any drug alone when used totreat a patient is determined by 3 variablesas expressed in Equation 1:
This equation states that the effect of thedrug (whether therapeutic, nuisance sideeffects, or toxicity, and whether intended ornot) is a function of what the drug does tothe body (the first variable) multiplied bywhat the body does to the drug (the secondvariable) as modified by the biology of thespecific patient (the third variable).Equation 2 illustrates that drug concentrationis determined by 2 variables: (1) dosingrate and (2) clearance.
The clearance for most drugs (approximately85%) is determined by the functionalactivity of the cytochrome P450 (CYP450)drug metabolizing enzymes.6 These enzymesconvert drugs into polar metabolites, whichcan then be eliminated in the urine.Changing the functional activity of CYPenzymes generally results in a change in drugclearance, which is in essence like a changein the opposite direction in the dosing rate.
Increases in clearance without a compensatorychange in dosing rate can drop thedrug level below the point where it is effective,which can result in loss of efficacy withrelapse or withdrawal symptoms. As aresult, the patient can appear to be resistantto the efficacy of the drug, or may evenappear to be drug seeking, if the affecteddrug has an abuse potential. Conversely,reductions in clearance without a compensatorychange in the dosing rate can raisethe drug level and increase both the frequencyand the severity of the drug's dosedependentadverse effects. As a result, thepatient can appear to be sensitive or intolerantof the adverse effects of the drug, mayappear to have the emergence of a new disease,or may have more serious adversereactions, including death.Several factors contribute to the pharmacokineticsof the drug, including its absorptionfrom the site of administration, itsdistribution into the various compartmentsof the body, its metabolism or biotransformationinto more polar substances, and itselimination, typically in the urine.
Biological variance among patients playsa major role in explaining why somepatients respond differently to a given doseof a drug, whether the reason is alteredpharmacokinetics or altered pharmacodynamics.Such biological differences can shiftthe dose-response curve, making a patienteither more or less sensitive to the effects ofthe drug compared with the average patient,whether the net result is altered therapeuticefficacy or altered frequency or severity ofadverse effects.
Biological variables can be divided into 4major categories: genetics, age, disease, anddifferences in the internal environment. Theinternal environment is a function of biologicalvariations resulting from the effects ofother medications the patient is taking. Toput the impact of this last variable into perspective,consider the fact that the coadministrationof a drug capable of substantiallyinhibiting the CYP450 enzyme responsiblefor the rate-limiting metabolism of a coprescribeddrug can have a greater effect on theclearance of the coprescribed drug thansevere hepatic or renal disease.
In essence, drug treatment is an acquiredbiological variance among patients, whichcan substantially alter the effects of otherdrugs patients are taking (eg, a DDI). Theeffects of such DDIs can present in almostany way clinically imaginable (Table 1).That is the crux of the dilemma and the importanceof DDIs for everyone involved inhealthcare delivery: the patient, the prescriber,the healthcare system, and drugmanufacturers (Table 2).
Differences Among TherapeuticallySimilar Drugs and the Potential for DDIs
Therapeutic indication is the way mostprescribers and most managed care systemsthink of drugs. However, drugs do not interact based on their therapeutic indications.Indeed, drug interactions are a function ofthe pharmacodynamic and pharmacokineticproperties of the coprescribed drugs, asexpressed in Equation 1.
Antidepressants represent a good case inpoint. There are at least 8 pharmacodynamicallydistinct classes of antidepressants.7One of these classes is the serotonin selectivereuptake inhibitors (SSRIs). Thesedrugs were the first class of rationally developedantidepressants. They were in the discoveryphase of drug development in the1970s and entered human efficacy trials inthe late 1970s and early 1980s. They weredeveloped on the basis of what was knownabout the pharmacology of the tricyclic antidepressants(TCAs), such as amitriptyline.8The distinguishing pharmacological characteristicsof TCAs were a function of theirability to bind and block multiple differentneural targets (Figure 1). As a result, TCAscan and did cause many different types ofadverse effects: some were nuisance (eg, histamine-1 receptor blockade), but otherswere serious and potentially life threatening(ie, cardiac arrhythmia due to the inhibitionof fast Na+ channels) (Figure 2).
Based on this knowledge, SSRIs weredesigned to reproduce only 1 of the effects ofthe TCAs: the ability to block the serotoninuptake pump (Figure 3). The developmentof these drugs was accomplished by isolatingthe various mechanisms in test tubes, andthen using in vitro techniques to determinethe structure-activity relationship needed toachieve high affinity for the serotoninuptake pump and low affinity for all of theother targets affected by TCAs. In otherwords, the drug developers "dialed in" thedesired effect and "dialed out" the undesiredeffects of TCAs.
The drug development approach used todiscover the SSRIs represents the essence ofrational drug discovery as we know it today.As a result of this process, the SSRIs sharemany pharmacodynamic properties, including(1) antidepressant and anxiolytic efficacy,(2) nuisance effects limited to thoseresulting from indirect serotonin agonism,and (3) a wide therapeutic index and thussafety in overdose.8
Although rational drug developmentaccounts for these similarities among thevarious SSRIs, such discovery is always limitedto what is known at the time the drug isbeing developed. That fact is critical tounderstanding the differences between theSSRIs. These drugs were developed in the1970s–a decade before the first CYP450enzyme (CYP2D6) was isolated throughtechniques from molecular biology.9
The Role of the CYP450 Enzymes
CYP enzymes are important because theymediate the biotransformation of mostdrugs.9 This biotransformation, in turn, isthe rate-limiting step in the clearance.Hence, clearance is what determines thedosing rate of most drugs as expressed inEquation 2.
Clinical trials, which determine the usualrecommended dosing schedule for drugs, arein essence population pharmacokinetic studies.The goal is to determine the usual dosefor the usual patient in the trial–who hasusual clearance and as a result accumulationof the right concentration of the drug toengage the right mechanism of action to theright degree to produce an optimal response.
A minority of drugs can affect the functionalactivity of CYP enzymes either byinhibiting or inducing them. Inhibition of aspecific CYP enzyme by a drug (ie, the "perpetrator")results in a decrease in the clearanceof drugs dependent on that CYPenzyme for their rate-limiting biotransformation(Equation 2). Thus, inhibition of aCYP enzyme most often has the same effectas increasing the dose of the affected (or"victim") drug. The increased concentrationof the victim drug produces dose-dependenteffects expected for the drug, but more thanwould usually be expected for the dose thepatient is taking because the reduced clearanceresults in greater than expected drugaccumulation. Conversely, stopping the perpetratorresults in increased clearance of thevictim drug and hence a fall in its levels,unless the dose is increased to offset theincrease in clearance.
The increases in levels of the victim drugcan present in a myriad of ways, rangingfrom simple intolerability to the apparentemergence of a new disease to catastrophicoutcomes. The reduction in levels of thevictim drug can present as loss of efficacyor withdrawal symptoms, and can evenlead the clinician to suspect that thepatient is demonstrating drug-seekingbehavior, particularly when the victim drugis a drug of potential abuse, such as an opiateor benzodiazepine.
CYP enzyme induction is the mirror imageof CYP enzyme inhibition, so that starting aninducer is like stopping an inhibitor and viceversa. Such DDIs can lead the clinician totake further action to diagnose and intervene,and thus inadvertently further increase thecost of health services as well as drug costs.7
Although frequently inhibition and inductionare in effect similar to changing thedose of the victim drug and can be compensatedfor by dosing adjustments, that is notalways the case. The inhibition may preventthe conversion of a prodrug like codeine intothe active metabolite, morphine, and hencethe patient does not experience painrelief.10,11 In this case, no dose adjustment ofcodeine can be made to compensate for theinability to convert it into the active metabolite.Conversely, induction may increase theproduction of an unusual metabolite includingreactive metabolites, which can mediateserious toxicities.12 This latter type of CYPenzyme-mediated DDIs is fortunately rare,but can lead to serious adverse patient outcomeswhen they do occur.
Alterations in Metabolism
The use of a substantial CYP enzymeinhibitor or inducer can be analogous toconverting most patients taking the druginto a phenocopy of genetically-based slowmetabolism. For example, fluoxetine inhibitsseveral CYP enzymes to varying degrees atits usually effective antidepressant dose of20 mg/day (Figure 4).13 One of the CYPenzymes that is inhibited to a substantialdegree is CYP2D6, which is believed tomediate the rate-limiting step in the clearanceof approximately 30% of marketeddrugs.6 This enzyme is polymorphic, meaningthat there are a small percentage of individualswho are genetically deficient in thisenzyme. These individuals, because of theirnonexistent CYP2D6 activity, achieve unusuallyhigh concentration when taking theusual drug dose, resulting in increasedadverse effects, including sudden death.14The rate of this genetic deficiency varies indifferent populations and is highest inCaucasians of northern European extraction,with a prevalence of about 7%.
Fluoxetine at doses of 20 and 40 mg/day,respectively, converts 66% and 95% of normalCYP2D6 metabolizers into phenocopiesof genetic deficiency of this enzyme.15,16 Inother words, treatment with fluoxetine convertsthe majority of normal metabolizersvia CYP2D6 into a phenocopy of an infrequentto rare genotype of CYP2D6 deficiency(Figure 4). That means that thecoprescription of 20 mg/day of fluoxetinewill result in the usual dose of drugs dependenton normal CYP2D6 function for theirclearance will be producing concentrationsthat are 4 times greater than would normallybe expected. Parenthetically, CYP2D6 isonly one of several CYP enzymes inhibitedby fluoxetine at usual antidepressant doses(Table 3).16
The long half-life of fluoxetine and itsactive metabolite, norfluoxetine, make suchCYP enzyme inhibition even more problematicbecause of the protracted gradualbuildup of the levels and effects of fluoxetine/norfluoxetine and the equally protractedtime needed to clear these levels and effects.The half-life of fluoxetine and norfluoxetineis 2 to 4 and 7 to 15 days, respectively, inyoung, healthy individuals, but in healthy,elderly individuals between the ages of 65and 75 years, the half-life of norfluoxetine is21 days.17 That means it takes up to 4months to reach steady state and 4 monthsto completely clear the drug in olderpatients. During this protracted period, levelsof fluoxetine and norfluoxetine will continueto gradually accumulate leading in parallel toa reduction in the functional capacity of aspecific CYP enzyme inhibited by fluoxetine,as illustrated in Figure 5.18 The flip side isthat the clearance of fluoxetine and norfluoxetinetakes a parallel period of time as itsaccumulation, and the same is true for itsconcentration-dependent effects, includingthe inhibition of specific CYP enzymes. Thisgradual accumulation and dissipation of itseffects makes it even more difficult to detectand understand the reason behind DDIsmediated by fluoxetine's effect on CYPenzyme function because the adverse eventscan develop insidiously long after the start ofthe offending agent, fluoxetine, and can persistfor a long period after fluoxetine has beendiscontinued.
Three SSRIs (fluoxetine, fluvoxamine,and paroxetine) under their usual dosingconditions inhibit 1 or more CYP enzymes(more than they inhibit the neuronal uptakepump for serotonin). The latter effect is theapparent mechanism mediating their antidepressantefficacy (Figure 6).7,8,19 Hence, it ispharmacologically impossible for theseSSRIs to treat depression without affectingthe function of specific CYP enzymesresponsible for the clearance of specificcoprescribed drugs. Such inhibition can, inturn, change the accumulation of thesecoprescribed drugs and their concentration-dependenteffects on the patient.
However, CYP enzyme inhibition is not auniversal quality of SSRIs, as evidenced bycitalopram and sertraline. In the case ofthese 2 SSRIs, there is wide separationbetween their ability to inhibit the serotoninuptake pump and its inhibitory effect on anyof the major known CYP enzymes.7,17,20 Inother words, these SSRIs remain selectiveeven when CYP enzymes are considered inthe equation. This is not true for fluoxetine,fluvoxamine, or paroxetine.
The Masked Ways that DDIs CanPresent in Practice
With the exception of anti-infectives, alldrugs are given to patients to change theirbiology. Once that fact is recognized andacknowledged, it is obvious that DDIs canpresent in any way clinically imaginablebecause they change the biology of thepatient.
Although most prescribers think of DDIspresenting as a sudden catastrophic outcome,such as sudden death, arrhythmias,or seizures, such DDIs fortunately appear tobe rare. Instead, DDIs most often present inmore subtle and masked ways. As a result,physicians and other prescribers can doextensive workups and add medications totreat what is actually a DDI. The bottom lineis that DDIs can adversely affect patient outcomesin addition to the bottom line of managedcare spreadsheets. (To read aboutreal-life cases of the varied ways DDIs canpresent and how they have been misdiagnosedand treated, visit my Web site atwww.preskorn.com and see the section on"Case Studies" under columns.)
The Importance of CYP Metabolismin Antidepressants
Although most drugs do not inhibit orinduce CYP enzymes, there are a numberthat do. These include quinolone andmacrolide antibiotics and azole antifungals,as well as some SSRIs. However, CYPenzyme-mediated DDIs are particularlyimportant with regard to SSRIs for the followingreasons.
First, antidepressants are used in a largepercentage of the US population. Second,patients taking antidepressants frequentlytake other medications. In primary care andgeneral outpatient psychiatry, our surveysfound that two thirds of patients taking anantidepressant were taking at least 1 otherprescription drug; one third were taking 3more drugs in addition to their antidepressant(Table 4). The more drugs a patient istaking, the greater their likelihood of experiencinga DDI and even multiple DDIs. Third,antidepressants are generally taken formonths to years to prevent relapse after anacute remission or to prevent futureepisodes in a patient at heightened risk forrecurrent major depression. In contrast,antibiotics are usually taken for a couple ofweeks. Thus, the period of time at risk for aDDI is much greater in the patient takingan antidepressant versus one taking an anti-infective. During this protracted period oftime, other drugs may be added or stopped,further complicating the proper identificationof the cause of adverse outcomesresulting from DDIs and delaying the institutionof appropriate and effective countermeasures.
As noted earlier, our VA survey foundthat taking an antidepressant was a greaterrisk factor for more extensive polypharmacythan being older, and that patients takingantidepressants were more likely to be usinga unique drug regimen compared withpatients not taking an antidepressant.3,5
Given that many of the VA patients wereusing unique regimens (ie, 71%-97%), anysingle prescriber in that system would likelyhave little clinical experience with the totaleffects of all of the drugs their patients aretaking. That limited experience furtherhandicaps the prescriber in knowing andavoiding untoward DDIs and in identifyingthem when they do occur.
Mandated Drug Switches and DDIs
Controlling the rising costs of medicationshas been a concern for many healthcaresystems for a long time and morerecently for state governments. When a drugwithin a therapeutic class achieves genericstatus, then there is an obvious temptationto mandate a switch to that generic productas a cost-saving move. The move towarddrug substitution is also encouraged bypharmacy benefit plans that direct prescribersand pharmacists to dispense a specificmedication within a class. Suchsubstitutions may be the result of formularyrestrictions, negotiated drug discounts, ordrug rebate programs. The economic advantagesare indisputable but independent ofthe assumption that all of the medicationswithin a class are identical in all meaningfulrespects. The problem is that each drugwithin an apparently unifying class can varyin many meaningful ways, such that genericor therapeutic substitutions are not asimple matter.
A decade or so ago, I encountered thisissue when a state Medicaid agency mandatedtreatment with a TCA before Medicaidpatients could be treated with the newerSSRIs. TCAs at that time were generic andthe cost was markedly less than that of thenewer, branded products. The economicsappeared simple and straightforward. Themandate was based on the premise that anantidepressant is an antidepressant. Clearly,therapeutic indication is certainly one of theways that drugs can be classified. The problemwith this policy was that it ignored thefact that TCAs were substantially differentfrom the newer antidepressants in manyways. Perhaps the most compelling and mostundeniable difference is that TCAs can kill apatient if taken in even a relatively modestoverdose (ie, a 2-week supply). That is certainlya consideration when treating depressedpatients who are at increased risk fortaking an acute overdose. The risk of lethaloverdose is unequivocally not the case withvirtually any of the newer antidepressants.The Medicaid mandate by this state lastedonly a few weeks and was quickly overturnedbecause of the marked reaction by prescribers,their patients, and loved ones.
In the past couple of years, some SSRIs(most notably fluoxetine and paroxetine)have become available as generics. Thesame argument previously used across differentantidepressant classes has now surfacedwithin this drug class (ie, an SSRI is anSSRI). There is some merit to this argument.SSRIs were designed to be as alike as possibleusing 1970s technology. The problem isthat they were designed in the 1970s. Thatwas an era when we had limited understandingand ways of avoiding effects on CYPenzymes. In addition to their marked differenceson CYP enzyme inhibitions, SSRIsalso differ substantially in terms of theirpharmacokinetics, particularly with regardto fluoxetine, which is quite long-lived andhence capable of having sustained effectslong after the drug has been discontinued.21For this reason, fluoxetine can cause late-emergingand long-lasting DDIs, particularlyin elderly patients who take even longer toclear this drug than younger individuals.
A few years ago, a pharmacist for a largemanaged care company asked me how difficultit would be to switch patients fromother SSRIs to fluoxetine. This conversationoccurred when fluoxetine was still underpatent, but the managed care company hadapparently negotiated a quite favorable pricefor fluoxetine and was interested in gettingtheir patients switched to this specific SSRI.I explained that the switch would be relativelysimple in a patient who was not takingother medications. It could be helpful tocross-taper quite short-lived SSRIs (eg,paroxetine) to minimize the risk of SSRI discontinuationsyndrome.22 There was also theproblem that some patients who had benefitedfrom treatment with another SSRImight not respond to or tolerate fluoxetine,but that was a cost-benefit calculation. Ipointed out that the serious problem had todo with the patients who were taking morethan 1 other medication in addition to theirSSRI. In essence, switching such a patientfrom an SSRI with minimal effects on CYPenzymes (eg, sertraline) to fluoxetine witheffects on multiple CYP enzymes was essentiallylike changing the dose of all of thoseconcomitant medications, whose clearancewould be profoundly affected by fluoxetine.The problem was further complicated by thefact that the change would not be obviousinitially, but instead would develop slowlyover several weeks to months as fluoxetineand norfluoxetine accumulated in thepatient. Over time, CYP enzymes wouldbecome more and more inhibited, whichwould gradually lead to more and moreaccumulation of concomitant medicationsdependent on the functional integrity of theinhibited CYP enzymes.
Such mandated switches set the stage forDDIs with all of their attendant problems forall of the participants in the healthcare system.Given the complex regimens that manypatients are taking and with the short timegenerally available for medication checks,mandated switches pose a serious challengefor prescribers and their patients. Based onmy conversations with prescribers acrossthe country, this trend is a source of seriousconcern for many clinicians, especiallywhen they understand the myriad of waysthat DDIs can present.
Mandated switches have a substantial riskof being "penny wise and pound foolish" forthe managed care company. Although perhapsmore subtle than the acute overdoserisk difference between the TCAs and thenewer antidepressants of the early 1990s,the difference in the risk of CYP enzyme-mediatedDDIs with fluoxetine, fluvoxamine,and paroxetine versus citalopram, escitalopram,and sertraline can have significanteffects on patient outcomes and significantcost implications for the healthcare system.The difference is that the cost can be hiddenbecause of the many ways that DDIs canpresent clinically.
Numerous case reports and formal pharmacokineticinteraction studies documentthe risk of adverse outcomes resulting fromDDIs, but controlled clinical studies todetermine the cost of such DDI-inducedadverse outcomes are lacking for obviousethical reasons. To put this matter in perspective,let us say that we wanted to proposea study to an institutional review board(IRB) in which we want to put 2 groups of100 patients each taking 200 versus 1000mg/day of amitriptyline, respectively, toprove that there was an increased risk ofdelirium, seizures, arrhythmia, and suddendeath in the latter group compared with theformer. That study is analogous to prescribing2 groups of patients taking 200 mg/day ofamitriptyline either placebo or 20 mg/day offluoxetine, because fluoxetine on averagecauses an 80% reduction in the clearance ofTCAs, such as amitriptyline.17,18 Those studies,for good ethical reasons, are not likely toreceive IRB approval. Thus, the data are bynecessity based on case reports, formal pharmacokineticsstudies in which the magnitudeof the change in the plasma level of thecoprescribed drug is a surrogate for the riskof serious toxicity and naturalistic surveydata such as the VA studies cited above.
What Clinicians Can Do to Managethe Risk of DDIs
One important step is to recognize andacknowledge the risk of DDIs and to use it asa basis for selecting medications. Such risksshould guide drug selections by individualclinicians, and also the managed care organizations,pharmacy benefit companies, andgovernment agencies that govern drug selections.Such considerations should be factoredinto the analysis of the relative valueof the medication. Manufacturers are clearlyconsidering this risk when they select drugsto take into phase I. For example, companiesnow screen their new candidate agents fortheir ability to inhibit human CYP enzymesduring preclinical testing.22 Early identificationof this information aids in the predictionof potential DDIs, which may ultimatelydetermine whether a compound is pursuedin the drug development process.23
Drug manufacturers are attempting tominimize their risk of developing a drug thatwill be difficult to get approved or will be atrisk for being pulled from the market afterbeing approved. Five of the last 10 drugspulled from the market were withdrawnbecause of their potential for either beingthe perpetrator or victim of a CYP enzyme-mediatedDDI.24
Like drug developers, prescribers canselect for their personal formulary drugsindividual drugs within a therapeutic classpartly on the basis of avoiding CYP enzyme-mediatedDDIs.25 An example would bechoosing escitalopram or sertraline as thepreferred SSRI, instead of fluoxetine orparoxetine, on the basis that the former 2SSRIs have a much lower risk of causingclinically significant CYP enzyme-mediatedDDIs than the latter 2 SSRIs. However, amanaged care or pharmacy benefit companycan thwart that approach by mandatinga switch within the SSRI class (ie, mandatedswitch from escitalopram or sertraline tofluoxetine) solely on the basis of acquisitioncosts without taking the risk of CYPenzyme-mediated DDIs into account. Yet,such companies should consider this variablewhen they evaluate medications, particularlyif the company (eg, the VAhealthcare system) carries the financial riskfor the patient's total care rather than simplythe pharmacy costs. This issue is moreproblematic and poses a conflict of interestissue when the formulary decision maker(eg, a pharmacy benefits company) is onlyat risk for the cost of the medication andnot the cost of the increased healthcare thatcan result from a CYP enzyme-mediatedDDI.
Another important way to minimize therisk of DDIs is for the prescriber to keep medicationregimens as simple as possible.However, that is becoming more difficult forseveral reasons. First, the US population isaging and older patients have an increasedrisk of having multiple medical problems.Second, more medications are now availablethan ever before for treating or preventinglong-term illnesses. Third, medicines arenow being used to prevent or delay the development of disease (ie, statins) rather than tosimply treat an acute illness (ie, antibiotics).
Another approach is to use software programsto screen patients for DDIs and makechanges in their medication regimen accordingly.However, there are several limitationsto this approach. First, these programs areonly as good as the data entered, meaningthat the person entering the data must knowall of the medications that the patient is taking.That can be difficult especially if thepatient is seeing more than 1 prescriberand/or using more than 1 pharmacy. Ideally,the person entering the data should alsoknow all of the over-the-counter medicationsand herbal products that the patient istaking. Second, there is extremely limitedformal data to construct computer programsthat work in the real world of patients takingmultiple medications. Even though the averageVA patient takes 5 medications, mostDDI software programs are based on dataabout the effect of 1 drug on another drugrather than the effect of multiple drugsinteracting with each other. Third, theseprograms are generally constructed to serveas alerts rather than information systemsand are frequently plagued by so many false-positivealerts and false negatives that prescribersand pharmacists frequently ignorethem. Fourth, many prescribers find themtoo time-consuming to use in the limitedamount of time scheduled for most "medicationchecks." Fifth, these programs may notbe employed when the prescription is writtenbut only after the fact, causing changesto be made late and increasing both the riskto the patient and the costs to the healthcaresystem.
DDIs have important implications formanaged care. They can adversely impactpatient outcomes and can also have seriouscost implications. The problem posed byunintended and adverse DDIs will continueto grow in importance as the complexity anduniqueness of medication regimens escalatein an aging population using increasinglysophisticated and varied pharmacologicalagents. In this paper, antidepressants andSSRIs in particular were used to illustratethese important points.
Physicians' Desk Reference
1. . Montvale, NJ: ThomsonHealthcare; 2004.
2. Kaufman DW, Kelly JP, Rosenberg L, Anderson TE,Mitchell AA. Recent patterns of medication use in theambulatory adult population of the United States: theSlone survey. . 2002;287:337-344.
3. Preskorn S, Silkey B, Shah R, Neff M, Jones T, ChoiJ. Complexity of outpatient medication use in relation toage and number of prescribers. . In press.
4. US Census Bureau. 2000.
5. Silkey B, Preskorn SH, Golbeck A, Shah R, Neff M,Jones T. Complexity of medication use among youngerand older outpatients: the role of antidepressants. .In press.
6. Levy RH, Thummel KE, Trager WF, eds. . 1st ed. Philadelphia, Pa: Lippincott,Williams & Wilkins; 2000.
Outpatient Management of Depression:A Guide for the Practitioner
7. Preskorn SH. . 2nd ed. Caddo, Okla:Professional Communications, Inc; 1999.
Clinical Pharmacology of SSRI's
8. Preskorn SH. . 1sted. Caddo, Okla: Professional Communications, Inc;1996.
Trends Pharmacol Sci
9. Gonzalez FJ. Human cytochromes P450: problemsand prospects. . 1992;13:346-352.
10. Chen ZR, Irvine RJ, Bochner F, Somogyi AA.Morphine formation from codeine in rat brain: a possiblemechanism of codeine analgesia. . 1990;46:1067-1074.
11. Sindrup SH, Brosen K, Bjerring P, et al. Codeineincreases pain thresholds to copper vapor laser stimuli inextensive but not poor metabolizers of sparteine. . 1990;48:686-693.
J Pharmacol Exp Ther
12. Maggs JL, Williams D, Pirmohamed M, Park BK.The metabolic formation of reactive intermediates fromclozapine, a drug associated with agranulocytosis inman. . 1995;275:1463-1475.
Metabolic Drug Interactions
13. Shad M, Preskorn SH. Antidepressants. In: Levy R,Thummel K, Trager W, Hansten P, Eichelbaum M, eds.. Philadelphia, Pa:Lippincott Williams and Wilkins; 2004:563-577.
14. Preskorn SH, Baker B. Fatality associated with combinedfluoxetine-amitriptyline therapy. . 1997;277:1682.
15. Preskorn SH. Reproducibility of the in vivo effect ofthe selective serotonin reuptake inhibitors on the in vivofunction of cytochrome P450 2D6: an update (Part I). . 2003;9:150-158.
16. Preskorn SH. Reproducibility of the in vivo effect ofthe selective serotonin reuptake inhibitors on the in vivofunction of cytochrome P450 2D6: an update (Part II). . 2003;9:228-236.
17. Harvey AT, Preskorn SH. Cytochrome P450enzymes: interpretation of their interactions with selectiveserotonin reuptake inhibitors. Part II. . 1996;16:345-355.
18. Preskorn SH, Alderman J, Chung M, Harrison W,Messig M, Harris S. Pharmacokinetics of desipraminecoadministered with sertraline or fluoxetine. . 1994;14:90-98.
J Pract Psych Behav Health
19. Preskorn SH. Drug development in psychiatry andthe human genome project: the explosion in knowledgeand potential targets. . 2001;7:336-340.
J Psychiatr Pract
20. Preskorn SH. The human genome project and moderndrug development in psychiatry. . 2000;6:272-276.
J Clin Psychopharmacol
21. Harvey AT, Preskorn SH. Fluoxetine pharmacokineticsand effect on CYP2C19 in young and elderly volunteers.. 2001;2:161-166.
22. Rosenbaum JF, Fava M, Hoog SL, Ascroft RC, KrebsWB. Selective serotonin reuptake inhibitor discontinuationsyndrome: a randomized clinical trial. . 1998;44:77-87.
Guidance for Industry:In Vivo Drug Metabolism/Drug Interaction Studies–Study Design, Data Analysis, and Recommendations forDosing and Labeling
23. Food and Drug Administration Center for DrugEvaluation and Research (CDER).. Rockville, Md: National Institutesof Health; 1999.
J Psychiatr Pract
24. Preskorn SH. Drug approvals and withdrawals overthe last 60 years. . 2002;8:41-50.
25. Preskorn SH, Flockhart D. 2004 guide to psychiatricdrug interactions. . 2004;11:39-60.