(terms, types, and treatment implications)
An Interview with Jack E. Rosenblatt M.D.,
Editor, Currents Plus (Psychiatry)
CPP: Dr. Rosenblatt, allow us to begin, not with benefit of pharmacotherapy in psychiatry, but rather with risk—that conferred specifically by drug-drug interactions.
Rosenblatt: I believe that you are asking about pharmacokinetic drug-drug interactions, wherein one drug changes the activity of another by changing how much of it there is at its site of action. The activity of the drug interacted with changes because its activity depends upon how much of it is present at a relevant site (a site that is relevant because it is where a drug exerts its therapeutic effects or initiates its adverse effects). Higher concentrations typically give rise to effects of greater magnitude or duration, while lower ones are associated with effects that are smaller or briefer. The concentrations of a drug at a relevant site are referred to collectively as the drug’s “bioavailability” (which is calculated by measuring the area under a curve of drug blood concentrations at intervals after dosing). Since the bioavailability of a drug that gives rise to a specific effect at a precise location is not yet precisely known, we instead measure the concentration of the drug in the blood or other body fluids as a proxy of its bioavailability at the precise location where it is working. We can do this because a drug’s concentration in the blood or other body fluids establishes a constant relationship, or equilibrium, with its concentrations in compartments smaller than that of blood in the systemic circulation. Bioavailability is studied in association with “clinical activity” of drugs, which may be therapeutic or adverse, manifest as intensification or diminution of preexisting clinical activity, and change after adding or discontinuing the other (interacting) drug.
In one kind of pharmacokinetic drug-drug interaction, interacting drugs change the bioavailabilities of the drugs interacted with by inhibiting or inducing (that is, increasing) the activities of the CYP450 cytochrome enzymes that catalyze the first metabolic conversions of drugs, usually in the liver and/or small bowel mucosal cells). When one drug prevents the breakdown and excretion of the other; the other accumulates and acts like a larger dose. Collectively, these first metabolic conversions are referred to as Phase I metabolism, and include a diverse group of commonly occurring reactions (including oxidation, epoxidation, hydroxylation, N- and O-dealkylations, and aromatization), catalyzed by so-called Phase I P450 cytochrome enzymes.
When a Phase I P450 cytochrome enzyme is inhibited, the concentration of its substrate drug rises, and the result (increased bioavailability) is tantamount to exposure to a higher dose: If the drug’s bioavailability had been low before the interaction, its rise might be accompanied by emergence of therapeutic effects. If it rises further and exceeds the upper threshold of the drug’s therapeutic range, adverse effects predominate and eventually prove treatment-limiting.
A pharmacokinetic drug-drug interaction can confer benefit: After ziprasidone (Geodon) had been in use for some years, researchers found that patients were dropping out of treatment and experiencing unfavorable outcomes from the lower of two doses compared; at the higher dose, more patients improved, stayed in treatment, and had better outcomes. Ziprasidone is metabolized by an enzyme called aldolase hydroxylase and to a lesser degree by CYP3A4; had low-dose patients been treated with an inhibitor of either (for a psychiatric or non-psychiatric indication), increased ziprasidone blood concentrations, continuation in treatment, and better symptomatic relief may have occurred.The same issue occurs every day and in every medication-treated patient, because every drug has a lower threshold of therapeutic effect in every patient and there will always be a difference between no and some effect.
When the activity of a Phase I P450 cytochrome enzyme is induced (that is to say, increased, by a coadministered drug), the bioavailability of the substrate drug falls and its clinical effects, both therapeutic and adverse, diminish. When a drug is inactive until converted to active form by hepatic enzymes (a “pro-drug,”) induction of those enzymes (by, for example, carbamazepine or St Johns wort or broccoli or the polycyclic aromatic hydrocarbons in tobacco smoke) increases formation of the “active principle” (the metabolite responsible for clinical effects) and the result resembles that of a higher dose. If the conversion enzyme is blocked, active metabolite concentrations fall and clinical effects diminish.
CPP When Drug 1 inhibits the enzyme, Drug 2 concentrations rise and the effects of Drug 2 increase; when Drug 1 induces (that is, increases) the activity of the enzyme, Drug 2 concentrations fall and the effects of Drug 2 diminish.
Rosenblatt: Right. This is the kind of pharmacokinetic interaction that involves the P450 cytochrome enzymes, and it occurs predominantly (but not exclusively) in the liver. The P450 cytochrome enzymes that operate in human drug metabolism, designated by family, subfamily, and numbers identifying the genes that express them, bear the same names as do the genes themselves (except that the latter are italicized in print, according to current convention). They include CYP1A2, CYP2B6, CYP2C9, CYP2C19, CYP2D6, CYP3A4, and CYP2E1. (A few other genes belonging to the same families and subfamilies are expressed by additional genes designated by other gene numbers.)
CPP: Yet, according to your definition earlier, pharmacokinetic drug-drug interactions have potential to arise at every stage of drug disposition with potential to affect a drug’s bioavailability, do they not? Drugs or substances that affect absorption after dosing; uptake by (and metabolism within) mucosal cells of the small intestine; first-pass hepatic extraction; distribution to the liver and other organs and tissues; hepatic metabolism of drugs from the systemic circulation; hepatic conjugation, renal and biliary excretion, transmembrane transport of drugs—all are potential sites of pharmacokinetic interactivity.
Rosenblatt: They are. This kind of pharmacokinetic drug-drug interaction (wherein blood concentrations and clinical effects of drugs increase or decrease when other drugs change the activities of the enzymes that metabolize them) represents a subgroup of drug-drug interactions designated “pharmacokinetic,” and it is the most important subgroup, because it is the most frequently attributed. One drug causes more or less of the other’s presence and the clinical activity goes up or down, respectively.
Pharmacokinetic drug-drug interactions represent one of two principal types of drug interactions: The other is called a pharmacodynamic drug-drug interaction, wherein one drug increases or reduces the clinical activity of another drug without changing its bioavailability. Additivity of hypotensive effects during coadministration of beta blockers and angiotensin II receptor antagonists is an example; each lowers blood pressure by a different mechanism of action that does not change the bioavailability of either.
Pharmacodynamic interactions are more clearly discernible with respect to somatic effects, such as effects of antimuscarinic agents on some neuroleptic-induced extrapyramidal effects, those of neuroleptics on stimulant-exacerbated tics, and those of dopaminergic agonists, such as amantadine on neuroleptic-induced hyperprolactinemia, but they are more difficult to identify when improvement or worsening is expressed as psychiatric change. They are detected and typically attributed to one or the other of the drugs that “finally started working” or “was working and then stopped working.” Examples of some of the more conspicuous ones include caffeine (which can induce or worsen panic attacks) and benzodiazepines, such as lorazepam or alprazolam, which can mitigate them; another is stimulant-induced worsening of tics and their improvement by coadministered atypical neuroleptics. One drug opposes the activity of the other by dint of pharmacodynamic interactivity, not because benzodiazepines lower caffeine blood levels or atypical neuroleptics lower stimulant concentrations.
Unfortunately, pharmacodynamic drug-drug interactivity becomes hobbled by a context-dependent semantics: Whether two drugs interact pharmacodynamically may depend upon the indications for which the drugs are coadministered. (The converse has not been developed or studied, but could be.) One can argue, for example, that an SSRI and a benzodiazepine do not interact pharmacodynamically when coadministered for a primary depressive disorder, but do interact pharmacodynamically in a patient with panic disorder, since benzodiazepine monotherapy does not substantially mitigate depressive symptoms, and both drugs mitigate panic (one acutely and the other, prophylactically). Bupropion is effective for depression, but typically not effective for generalized anxiety or panic; in its case, coadministration of a benzodiazepine would not interact pharmacodynamically with respect to antidepressant and anxiolytic effects.
CPP: Do some drugs interact both pharmacokinetically and pharmacodynamically?
Rosenblatt: Yes, and more than one pharmacodynamic interaction may occur simultaneously when two drugs have multiple effects (as with the three pharmacodynamic interactions between caffeine, a proconvulsant, stimulant, and anxiogenic compound, and clonazepam, an anticonvulsant, sedative-hypnotic, and anxiolytic). Another example is coadministration of olanzapine and carbamazepine for acute mania: Both have antimanic activity, and that of carbamazepine may be additive to that of olanzapine, eventuating in antimanic activity exceeding that of each as monotherapy.
Carbamazepine also has potential to attenuate antimanic effects in olanzapine-treated persons by inducing the metabolism of olanzapine (a substrate of CYPs 1A2 and 3A4) and thereby reducing its bioavailability. Measuring olanzapine blood concentrations in that instance can inform about relative benefit of continuing coadministration of olanzapine and carbamazepine, of raising olanzapine dosage, or of replacing carbamazepine with an alternate drug with additive antimanic effects (such as lithium, valproic acid, or atypical neuroleptics) that do not induce hepatic metabolism of olanzapine.
Pharmacodynamic interactivity has potential to increase detection of pharmacokinetic interactivity, because drugs causing additive improvement are coadministered more often than drugs that do not. An example pertains to coadministration of lamotrigine and valproic acid (which interact pharmacokinetically and pharmacodynamically when coadministered for seizure disorders, for which both are effective, but only pharmacokinetically when coadministered for prophylaxis of bipolar depression, for which lamotrigine is more effective than valproic acid).
Estimates of risk and benefit of treatment differ when the same coadministered drugs are prescribed for different indications. Because both lamotrigine and valproic acid have anticonvulsant effects, coadministration may allow additivity of lower dosages and bioavailabilities of each to convey anticonvulsant effects. If lamotrigine bioavailability were lower during coadministration of lower than usual lamotrigine doses and valproic acid bioavailability lower than usual for lower than usual doses of valproic acid, anticonvulsant prophylaxis may be accompanied by lower risk of adverse effects (including serious rash, which is dose/concentration dependent). Because the adverse effects of lamotrigine and valproic acid differ, the pharmacodynamic interaction during treatment of seizure disorders could render both drugs more tolerable and/or safer at the lower bioavailabilities it may allow. (Therapeutic drug monitoring would be needed, however, because even at lower lamotrigine doses during coadministration with valproic acid, lamotrigine bioavailability may be high enough to confer risk of rash, and that is because lamotrigine and valproic interact pharmacokinetically, the latter having been reported to double the bioavailability of lamotrigine across its normative therapeutic dosage range.)
CPP: What about coadministration of lamotrigine and valproic acid for bipolar disorder?
Rosenblatt: In bipolar disorder, valproic acid and lamotrigine interact pharmacokinetically in the same way, but no longer interact pharmacodynamically, because, unlike valproic acid, lamotrigine is ineffective for mania. Its evidence-based effect in bipolar disorder is confined to prophylaxis against depressive recurrence in remitted patients. Effectiveness for acute depression has been reported, but not established for either drug. Nor has that of the atypical neuroleptics, but clinical consensus withal favors lamotrigine for acute depression in persons already receiving atypical neuroleptics for mania.)
Valproic acid has antimanic properties, both acutely and prophylactically. A few early reports hinted at antidepressant effects of valproic acid for bipolar depression, but acute antidepressant effects and prophylaxis of depressive recurrence by valproic acid have not been established. Valproic acid dosage and bioavailability would not be expected to change during coadministration with lamotrigine for prophylaxis of depressive recurrence (because lamotrigine does not affect the bioavailability of valproic acid), but the pharmacokinetic interactivity of lamotrigine and valproic acid may give rise to a higher lamotrigine bioavailability associated with increased risk of rash and other lamotrigine adverse effects. The rationale for coadministering valproic acid with lamotrigine for bipolar disorder would be broader-spectrum therapeutic effectiveness, affording well-documented prophylaxis against different poles of illness, while introducing less (or, as some believe, no) risk associated with lamotrigine of the mood-destabilization that accompanies antidepressant treatment.
Overall, the clinical importance of pharmacodynamic interactivity derives from its potential to allow reduced dosages of coadministered drugs that have similar therapeutic and different adverse effects. The therapeutic effects are additive, while the adverse effects may be no greater than during monotherapy with low (sometimes subtherapeutic) dosages of each. Strangely, this is rarely cited in critiques of polytherapy, and may reflect a “negative affect bias” and “representativeness (cognitve) bias” devolving from more numerous and more conspicuous adverse effects of coadministering two or more drugs with each mutually reinforcing the other. Such artifacts have received greater attention since they were characterized and quantitated by Amos Tversky and Daniel Kahneman a generation ago in economics and psychology, and more recently by the latter in an evidence-abundant, carefully-reasoned theory of different types of cognitive processing in humans (Thinking Fast and Slow, Daniel Kahneman)–a healthy rendering for medical science, especially psychiatry, wherein evidence consists of interpretations of observations, Until relatively recently, observations have been a preponderant focus of searching for artifacts and biases that inhere in experimental methodology and have potential to affect how they are interpreted. Those interpretations are what we recognize as evidence.
CPP: Might broad-spectrum effects of polytherapy (often needed for favorable global outcome in patients with comorbidity) suggest that drugs for development might be more effective if they had broad profiles of activity? A broad spectrum of neurochemical effects has been posited to account for the superiority of clozapine for psychotic disorders, for example, and of clomipramine for obsessive-compulsive disorder.
Rosenblatt: Long-known broad spectrum therapeutic effects of drugs that interact pharmacodynamically may have inspired development of individual drugs with multiple neurochemical effects, likewise intended to afford a broad therapeutic spectrum. Clozapine, as you point out, whose often replicated superior antipsychotic effect (and apparently causal association with reduced suicide rate) is interpreted to devolve from a broad clinical spectrum representing activity at multiple neurotransmitter receptors, transport sites, ion channels, and second-messenger transductions, was the first to be plumbed at depth and characterized as a “dirty” drug. Consistently, its “cleaner” demethylated metabolite, desmethylclozapine, with its narrower spectrum of neurochemical effects, has been found no more effective than placebo for psychotic symptoms. Atypical neuroleptics as a class were posited to have broader spectrum effects in schizophrenia than those of the first-generation antipsychotics (with respect to their putative efficacy for secondary or “deficit” symptoms of residual schizophrenia), and this was attributed to their affinities of binding to different classes of 5-HT2 receptors.
A few years after clozapine was approved in the U.S., the same descriptor (“broad spectrum”) was applied to clomipramine, whose superior therapeutic effect in obsessive-compulsive disorder has been attributed to inhibition of multiple neurotransmitter reuptake sites. Neither its dechlorinated congener imipramine nor the latter’s desmethyl metabolite desipramine are as effective for obsessive-compulsive disorder. Had clomipramine proved less prolific of adverse effects, it might have been regarded as a superior antidepressant (as clozapine is regarded a superior antipsychotic) by dint of anti-obsessional effects in depressed persons with subsyndromal OCD.
That reprises the broad-spectrum efficacy of clozapine apropos of drug development: Does it exert specific clinical effects on a dimension of psychosis or on comorbidity thereof that may give rise to overall superiority for psychotic disorders? If so, it would not be by dint of mitigation of secondary symptoms, because clozapine has been reported superior overall to other atypical neuroleptics that affect secondary symptoms no less.
CPP: Will you give more detail about how the P450 cytochrome enzymes are involved in pharmacokinetic drug-drug interactions?
Rosenblatt: Their metabolic activity represents the first step in clearing a preponderance of drugs in our current pharmacopoeia—those that depend upon Phase I metabolism, which, as I mentioned, comprises a small number of commonly occurring metabolic conversions. When the activities of the enzymes that catalyze those reactions are inhibited or induced, and thereby change drug bioavailability and clinical activity in the direction expected, that is, by definition, pharmacokinetic drug-drug interactivity–a special case of the correlation of drug bioavailability and drug effect.
The P450 cytochrome enzymes (so called because they were detected in spectrophotometric assays by their absorption of light of wavelength 450 billionths of a meter) appear in the molecular historical record about 400 million years ago, when global environmental conditions no longer precluded survival on land of aquatic organisms with otherwise successful terrestrial adaptations. The P450 cytochrome enzymes are thought to have allowed those organisms to metabolize and excrete toxic alkaloids of vegetation, providing a safely consumable energy source (a reproductive advantage) for them and thereby for organisms higher in the foodchain.
The evolutionary value of iron in cytochrome enzymes is conferred by its ability to share electrons with oxygen; oxygen bound to iron of cytochrome enzymes becomes bound to products of the reactions they catalyze, rendering them more soluble in water and thereby more efficiently cleared. By the time humans evolved (about two million years ago), P450 cytochrome enzymes had evolved to facilitate clearances both of endogenous compounds with relatively low water solubility (such as steroid hormones, cholesterol, and metabolic byproducts of lipid metabolism), and environmental chemicals (from the airborne combustion products of Stone Age hearthfires, to alkaloids extracted from plants for medicinal use, to modern synthetic toxins and drugs).
Of the 57 cytochrome P450 enzymes identified in humans, a subgroup (comprising CYPs 1A2, 2B6, 2C9, 2C19, 2D6, 2E1, 3A4, and, when expressed, CYPs 2A6 and 3A5) contributes to hepatic drug metabolism. Concentrated in the liver, small bowel, biliary system, and kidney (and present in other organ systems or tissues, where they participate in steroid and lipid metabolism), drug-metabolizing cytochrome enzymes may exist in functional metabolic units (called “metabolosomes” by one of the groups studying them (R. Fujiwara and T. Itoh, Pharmacological Research Perspectives, 2(5): e00053, October 2014. Ed.), wherein their activities are coordinated with those of hepatic conjugating enzymes (such as uridinyl glucuronosyltransferases, glutathione-S-transferase and others that increase water solubility of drug metabolites for renal excretion). Molecular adaptations that increase water solubility have been an important attribute of successful responses to selection pressure throughout reproductive history; it has favored survival of genes that operate in the clearances of both endogenous reaction products and exogenous exposures.
Activities of P450 cytochrome enzymes are coordinated also with activities of cell membrane transport proteins, such as para-glycoprotein, so-called multi-drug resistance protein, and organic ion transport proteins that affect influx and efflux of drugs to and from the interiors of cells, in one direction to extracellular fluid and eventually to excretory conduits (lumina) in liver, small bowel, and kidney, and in the other, to the interiors of cells, enabling longer exposures of drugs to metabolic enzymes, intracellular pathogens to antimicrobials, and malignant cells to chemotherapeutic agents. The metabolic function of P450 cytochrome enzymes comports with their dense concentrations near excretory conduits in liver, small bowel, and kidney, and it may as well with those of “progeny-protective” placental and mammary tissues and endothelial cells of brain capillaries that constitute a layer of the blood-brain barrier.
CPP: What is known about inhibition and induction of cytochrome enzymes?
Rosenblatt: Four modes of cytochrome enzyme inhibition have so far been described: In competitive inhibition, enzymatic conversion of a substrate to reaction products is reduced because another drug blocks access of the substrate to the enzyme’s active site. It does so reversibly, leaving the enzyme intact after it dissociates from the active site. The potency of competitive inhibition is related to the extent to which the inhibitor out-competes the substrate for binding to the enzyme’s active site. This is related both to its concentration near the site and to its affinity of binding the site (the strength with which it binds the active site of the enzyme, affecting displacement of or by other drugs and its duration of binding before it dissociates from the active site). High-affinity competitive inhibitors (those that bind the enzyme’s active site strongly) can cause clinically significant interactivity at lower concentrations (dosages), which is why 20 milligrams of one SSRI, with a higher affinity for the active site of CYP2D6, may raise the concentrations of coadministered desipramine more than does 20 milligrams of another SSRI with low affinity for the enzyme’s active site.
Non-competitive (allosteric) enzyme inhibition occurs when one of two coadministered drugs binds the enzyme at a location other than its active site or heme group, and remotely affects its three-dimensional conformation, reducing or abolishing the binding affinity of the substrate drug. This kind of inhibition may or may not be reversible, and may or may not leave the enzyme intact. Appearance of and severity of drug-drug interactions of this type are variable, and depend upon how much inhibitor is present and for how long it is. A non-competitive inhibitor may bind to and remain bound to an enzyme whether or not substrate drugs are bound to its active site. Contrariwise, an uncompetitive or anti-competitive inhibitor prevents metabolic conversion of a substrate only after the substrate has bound the enzyme’s active site. As with non-competitive inhibition, uncompetitive inhibition may be reversible and may leave the enzyme intact.
The most important mode of cytochrome enzyme inhibition is “mechanism-based inhibition,” which occurs when the inhibitor binds to the enzyme’s active site and is itself converted to reaction products that bind the enzyme irreversibly and terminate its activity. Enzyme activity is restored only after new enzyme is generated. As with other modes of cytochrome enzyme inhibition, clinical severity depends upon extent of accumulation of unconverted substrate. An example is the interaction of HMG CoA reductase inhibitors (“statin drugs”) with a subgroup of calcium channel blockers that are converted by CYP3A4 into reaction products that bind to and inhibit CYP3A4 irreversibly. Enzyme activity is restored after activation of the CYP3A4 gene and cellular expression of more CYP3A4. (Statin drugs that are substrates of CYP3A4 and have high first-pass hepatic extractions are associated with greatest accumulations of statin drugs during mechanism-based CYP3A4 inhibition, and the most likely to cause disintegration of muscle cells (rhabdomyolysis), complicated by renal failure.)
The mechanism of cytochrome enzyme induction has become better understood over the last decade, and the findings suggest that it may represent activation of processes operating in physiologic enzyme regulation: drugs that induce cytochrome enzymes are bound by a large nuclear protein (such as the pregnane X receptor), which binds to another (such as the retinoid R receptor), to form a dimer (complex of two bound molecules) that can enter cell nuclei, where the nuclear protein-ligand (inducer) binds to the regulatory region of a gene that codes for the enzyme family and subfamily induced by the ligand, which initiates transcription of the enzyme gene and expression of more enzyme. Inducers have been associated in the liver with cytochrome P450 CYP’s 3A4 and 1A2, and, less commonly, with CYP’s 2C19, 2C9, CYPs 2E1 and 2B6.
CPP: Not CYP2D6—the enzyme that adverted psychiatry to the clinical significance of the P450 cytochrome enzymes? In the early Nineties, its inhibition by coadministered paroxetine and fluoxetine was found to increase desipramine bioavailability by up to ten-fold.
Rosenblatt: Findings about the inducibility of CYP2D6 (important, relative to its large number of substrates, including, as you point out, many indicated for psychiatric disorders) have been mixed: some authors have found no evidence of hepatic CYP2D6 inducibility; others have (by dexamethasone and rifampicin); yet others have reported extra-hepatic CYP2D6 induction (particularly in brain). One has reported hepatic induction of CYP2D6 by nicotine and by alcohol in African green monkeys. Another has reported increased CYP2D6 activity in hepatocytes of humans taking the herbal medicines valerian and Ginkgo biloba. That CYP2D6 activity increases during pregnancy has been reported by yet another group. A consistent body of evidence about inducibility of CYP2D6 in human liver (akin to that of CYP3A4) has not yet emerged.
CPP: What does it suggest about CYP3A4 induction?
Rosenblatt: CYP3A4 is regarded as the most inducible of the hepatic cytochrome enzymes, because it is responsible for the Phase I metabolism of more substrates than are any of the other P450 cytochrome enzymes. More is understood about its underlying mechanistic processes than about those of other cytochrome enzymes, but investigation has been more prolific of questions than of answers (applause sign on): for example, induction is thought to be initiated by activation of the nuclear receptor protein pregnane X by substrates of CYP3A4, specifically; yet, activation of the pregnane X receptor initiates synthesis of any CYP3A4, and some compounds that activate pregnane X do not increase the activity of CYP3A4. Moreover, some compounds that increase CYP3A4 activity do not activate pregnane X. Diverse “coactivators” have been posited to affect the likelihood that the nuclear receptor complex will activate or suppress transcription of the genes for cytochrome P450 enzymes, which may explain why some drugs have been reported to both inhibit and induce CYP3A4. Among other hypotheses inspiring research is one that involves a “super moderator” (hepatocyte nuclear factor 4-alpha (HNF-4-alpha)) whose initial involvement is liver development, whereafter it is posited to act as a “transactivator,” controlling expression of genes for nuclear protein receptors (also known as xenosensors) that initiate transcription of diverse cytochrome enzymes, conjugation enzymes, and membrane transport proteins when signaled about the presence within cells of an exogenous chemical (such as a prescription drug) or endogenous substrate. That finding (and hypothesis) may help explain how activities of hepatic cytochrome enzymes, conjugating enzymes, and membrane transporters are coordinated.
CPP: When you mentioned pro-drugs earlier, you said that inhibitors and inducers of P450 cytochrome enzymes had effects on drug activity opposite to those of drugs that require no conversion to active moieties. Will you elaborate on that?
Rosenblatt: A pro-drug is an inactive compound that undergoes Phase I metabolism to form one or more active reaction products responsible for the compound’s clinical effects. Inhibition of the enzyme that catalyzes conversion of an inactive drug to an active metabolite (moiety) reduces therapeutic drug activity and side effects (of the therapeutic moiety), while induction of the enzyme increases therapeutic and adverse effects of the active moiety. The analgesic tramadol is an example: during inhibition of CYP2D6, which catalyzes O-demethylation of tramadol to its active moiety, tramadol accumulates and has been reported to induce hyperserotonergic toxicity (serotonin syndrome) and seizures. Because inhibition of CYP2D6 reduces formation of tramadol’s active moiety, O-desmethyltramadol, it is accompanied by reduction of analgesic potency. Someone who takes tramadol for severe head pain, for example, may find the same dose ineffective after she begins treatment with fluoxetine for depression. A potent inhibitor of CYP2D6, fluoxetine inhibits formation of tramadol’s active moiety and may increase the tramadol dosage needed to relieve her head pain. Beecause accumulation of tramadol is reported to increase risk of seizure and of hyperserotonergic toxicity, she might be better off with an SSRI that has very weak inhibitory effects on CYP2D6, such as citalopram or escitalopram.
Codeine is another pro-drug whose analgesic effects derive from an active metabolite (morphine), whose formation depends upon CYP2D6. As with tramadol, inhibition of CYP2D6 eventuates in reduced clinical effect (morphine-induced analgesia) and accumulation of codeine. Two other prodrugs whose activity may change by dint of inhibition or induction of cytochrome P450 enzyme activity are the antihypertensive losartan (a pro-drug requiring CYP2C9 for conversion to its active moiety) and clopidogrel, a substrate of CYPs 2C19 and 3A4, whose antiplatelet effects devolve from an active metabolite formed by CYP3A4. Inhibition of both may cause clopidogrel to accumulate and cause adverse effects, while affording little if any protection against clot formation.
CPP: Several of the first and some of the second-generation antidepressants and neuroleptics are converted to active metabolites, are they not? Some have been developed and marketed for the same indications as their parent drugs.
Rosenblatt: Several drugs have active metabolites (examples include imipramine, amitriptyline, fluoxetine, and potentially other antidepressants), but they are not considered pro-drugs because, like their metabolites, they too are active. As you mention, some active metabolites or active enantiomers of active parent drugs are developed and marketed as new drugs (9-hydroxyrisperidone, desvenlafaxine, escitalopram, for examples); many more have been tested and not developed (because they lacked effectiveness, were less effective than their parent drugs, or induced adverse effects). Activity is generally preserved in a tertiary amine compound that has been demethylated or didesmethylated. That would apply to aripiprazole, but I know of no data about comparative effectiveness or tolerability of its desmethyl- or didesmethyl-metabolites. I suspect that pharmaceutical firms have been studying them. Aripiprazole is among the most tolerable of the atypical neuroleptics and the only one that does not induce hyperprolactinemia.
CPP: Are some cytochrome P450 enzymes more important than others with respect to human drug metabolism?
Rosenblatt: Metabolizing the highest proportion of FDA-approved drugs (among the hepatic cytochrome P450-dependent enzymes) are those belonging to Family 3, Subfamily A. Fifty to sixty percent of currently approved drugs are reported to be substrates of CYP3A4.
Arguably the most important of the hepatic P450 cytochrome enzymes, CYP3A4 has distinctive features favoring resilience with respect to drug clearance and response to chemical challenge. As with those imparting resilience to the immune system in responding to foreign protein, the features redounding to the advantage of CYP3A4 in limiting exposure to accumulation of endogenous substances and exogenous toxins derive from its capacity for diversity of response and efficient integration with complementary processes that increase efficiency of excretion.
CPP: What might convey advantages in that respect to CYP3A4?
Rosenblatt: Its large active site. It is larger than that of other P450 cytochrome enzymes and able to simultaneously accommodate more than one of a large number of Phase I metabolic reactions, yielding greater diversity of metabolic products that then become substrates of Phase II metabolism by enzymes rendering them more water soluble for excretion in urine (including glutathione-s-transferase and uridinyl glucosuryltransferases for hepatic glucuronidation), before they become substrates of Phase III processes (those of membrane transport proteins that mediate their efflux into the lumen of the small bowel).
Suggesting efficiency from coordination, inhibitors and inducers of CYP3A4 during Phase I metabolism inhibit and induce, respectively, enzymes and proteins that operate in Phases II and III. Some substrates are cleared by Phase II or Phase III processes, and do not require metabolic conversion by Phase I enzymes. When Phase I enzymes (the P450 cytochrome enzymes) are required for clearance, the activity of those belonging to family-subfamily enzymes other than CYP3A may coordinate with clearance processes “downstream.”
Further increasing diversity and efficiency of CYP3A4 during drug clearance is the distinctively large regulatory region of the CYP3A4 gene, with its binding areas for nuclear proteins (including the pregnane X and retinoid R receptors) that enter the cell nucleus conveying “information” in the form of endogenous and environmental compounds (ligands) bound to them, which may convey information about which gene to express. Activity of CYP3A4 increases or decreases as a function of how much is expressed by the CYP3A4 gene under regulation by environmental signals.
CPP: FDA characterizes substrates as “sensitive” and inhibitors and inducers as “strong,” “moderate,” and “weak.” What do those characterizations mean, and how relevant are they clinically?
Rosenblatt: FDA defines substrate “sensitivity” quantitatively, as the extent of an increase in substrate concentration after inhibition of the cytochrome P450 enzyme responsible for its first metabolic conversion (Phase I metabolism). FDA’s sensitive substrate is one whose bioavailability increases five-fold or more during coadministration of an inhibitor of its P450 cytochrome enzyme.
If the sensitivity criterion is intended to mark a lower limit of clinical significance for interactivity of substrate drugs and inhibitors of their Phase I metabolic enzymes, it may need refinement when applied to substrate drugs with narrow therapeutic margins, such as tricyclic antidepressants. Drugs with narrow therapeutic margins manifest therapeutic effects at bioavailabilities that increase on a steeply sloped segment of a curvilinear relationship between bioavailability and clinical effects to a margin above which adverse effects predominate.
Clinically significant pharmacokinetic drug-drug interactions may result in variable increments of bioavailability in vivo, and the lower limits of increments of tricyclic bioavailability may be lower than the five-fold or greater increment stipulated by FDA’s criterion. The increments of desipramine bioavailability during coadministration with the potent CYP2D6 inhibitor paroxetine, for example, ranged from 3.5- to 10-fold in an early study. The tricyclic antidepressants are significantly more likely to induce medically severe adverse effects at (imipramine-equivalent) bioavailabilities of 300 nanograms per milliliter or higher. The lower limit of enhanced risk is therefore three-fold (or less) higher than the therapeutic ranges identified for imipramine and amitriptyline (100 to 300 ng/ml) and most of the therapeutic range identified for nortriptyline (50 to 150 ng/ml), wherein also occur not only modal adverse effects, such as fatigue, anticholinergic effects, and postural hypotension, but also adverse effects that are not uncommon and more serious, such as mood-switching to hypomania or mania in persons with bipolar disorder, mood instability and affective recurrence in persons with remitted bipolar disorder, cardiac conduction abnormalities, and reduced seizure thresholds in vulnerable patients (such as persons with epilepsy who have been noncompliant with antiepileptic medications, and those with impending delirium tremens or abrupt benzodiazepine discontinuation).
CPP: Is FDA’s sensitivity dimension not sensitive enough?
Rosenblatt: Unlikely to be, in my opinion. Drawing attention to some substrates by designating them “sensitive” when their bioavailabilities increase by five-fold or more leaves a gap within which some drug substrates may confer higher risk of severe adverse effects at the lower part of the range, with respect to drugs with narrow therapeutic ranges. Tricyclic antidepressants illustrate this. The sole clinical adversity emerging at tricyclic bioavailabilities that conform to the sensitive-substrate criterion of a five-fold or greater increase in bioavailability would be lethality of overdose (amitriptyline having been reported lethal after an overdose approximately ten-times higher than daily dosage).
Many drugs are substrates of one or more cytochrome enzymes in addition to their primary cytochrome enzyme, the additional enzymes serving as potential “escape pathways” of metabolism when the activity of primary cytochrome enzymes is low (because of inheritance of alleles associated with no or low enzyme activity or coadministration of drugs that inhibit the primary enzyme). Not yet known is the extent to which the clinical value of escape pathways might be negated by redundancy of enzyme inhibition (inhibitors of one cytochrome enzyme inhibiting an additional one, two, or three, as exemplified by fluvoxamine, which inhibits CYPs 1A2, 2C9, 2C19, and 3A4 (each moderately to strongly)).
Findings of the last few years show more variability in the potencies of enzyme inhibition than had been suspected (including different potencies of the same inhibitors of the same enzymes for different substrates, and additionally, variability of metabolism of different enantiomers of the same substrate. Moreover, different degrees of inhibition by the same inhibitor have been reported for cytochrome enzyme associated with different alleles of the same cytochrome enzyme gene. Those findings suggest that a more nuanced categorization may be needed for substrates of P450 cytochrome enzyme activity.
A “sensitivity” concept based on a five-fold increment of substrate accumulation after inhibition of one well-documented model inhibitor is in fact a “low sensitivity” concept, since many drugs whose bioavailabilities increase by smaller increments have been implicated in clinically severe pharmacokinetic drug-drug interactions. I favor a classificatory method that applies the range of increased bioavailabilities reported in association with exposure to inhibitors of their phase I enzymes, as in, “Desipramine has a sensitivity of 3.5 to 10 when coadministered with paroxetine.” Because bioavailability is related to dosage, its actuality would dawn quickly and intuitively (as quickly as it takes to think of the difference clinically between a dosage of desipramine of 100 milligrams per day and one of 350 to 1000 milligrams per day).
CPP: What about FDA’s stratification of inhibitors as weak, moderate, and strong?
Rosenblatt: I do not demur regarding FDA’s stratification of cytochrome enzyme inhibitors as strong, moderate, and weak (denoting increments of substrate bioavailability by at least five-fold, two- to five-fold, and 1.25- to 2.0-fold, respectively), so long as they refer to effects of enzyme inhibitors on enzymes converting specific substrates or enantiomers (as in, Itraconazole is a strong inhibitor of CYP3A4-metabolism of aripiprazole, of which paroxetine is an inhibitor with potency proportional to the extent to which aripiprazole is a substrate of CYP2D6).
CPP: What contributes to “normal” variability in the activities of P450 cytochrome enzymes? If large enough, one might wonder whether the same things contribute to risk of pharmacokinetic drug-drug interactions.
Rosenblatt: Some do and some could. One source of variability comprises the single nucleotide polymorphisms (SNP’s) of the genes that code for cytochrome enzymes—versions of the genes that differ by one nucleotide. At least one has been identified with each of the hepatic, drug-metabolizing P450 cytochrome enzymes (more than two dozen are associated with CYP3A4 alone), but they are not among the most important contributors to variability of enzyme activity.
A source that is among the most important contributors is the multitude of genetic alleles of CYPs 2C9, 2C19, 2D6, and 3A4. Their phenotypes are defined as differences in enzyme activity. For example, apropos of CYP2D6, they denote activity in “poor metabolizers,” “normal metabolizers” (the phenotype of the “wild-type” allele), “extensive” or “rapid” metabolizers, and, even “ultra-rapid metabolizers” (a CYP2D6 phenotype thought to result from genetic duplication—two copies of the wild-type gene). Genotypes so far identified with a “null” (no activity) phenotype have not been identified for all cytochrome enzymes. One allele (CYP2D6*17) has been reported to code for CYP2D6 enzymes whose activities are typically low, but variable among the substrates that interact with them. The activity of CYP2D6 in persons with that allele, for example, is reported to metabolize dextromethorphan, risperidone, and haloperidol (all substrates of CYP2D6) with significantly different activities, giving rise to variability within variability.
The preponderance of variant alleles for cytochrome enzyme genes is associated with reduced activity (and higher bioavailabilities of their substrates), with the “dose” of inactive allele related to the magnitude of reduction in enzyme activity. For example (and as noted with other cytochrome enzyme alleles), homozygous CYP2C9 genotypes are associated with less activity than are heterozygous genotypes of the variant alleles CYP2C9*2 and CYP2C9*3, and the heterozygous genotypes are associated with less CYP2C9 activity than is the homozygous wild-type genotype (CYP2C9*1). Relative to activity of persons who are homozygous for the wild-type allele, Japanese patients with homozygous genotypes of variant alleles have been reported to show a 90-percent reduction in activity, compared with a 60-percent reduction in patients with the heterozygous genotypes (H. Takahachi and H. Echizen, Clinical Pharmacokinetics 40:8:587-603, 2001; Ed.). Moreover, rates of CYP2C9 alleles associated with low activity have been reported to vary with respect to racial extraction, with eight to 20 percent of a Caucasian sample found to have CYP2C9 *2 or *3, respectively, compared with zero percent and two to five percent of an Asian sample, respectively (H. Takahachi and H. Echizen, ibid.; Ed.). (The most recently reported anomalous allele of CYP2C9 (CYP2C9 *12), has been associated with 60 percent less activity than that of the enzyme associated with the wild type allele (CYP2C9 *1) (T.J. O’brien and colleagues, Clin Chim Acta 23:424:73-75, 2013); Ed.) The differences among frequencies of low-activity alleles in different racial-ethnic groups has potential to vary widely. The most recently reported anomalous allele of CYP2C9, CYP2C9*12, has been associated with 40 percent of the activity of the wild type allele, CYP2C9 (*1) (T.J. O’brien and colleagues, Clinica Chimica Acta 23:424:73-75, 2013). Previously documented differences that have been reported to span a range from zero to 45 percent of some racial extractions.
As noted with other cytochrome enzyme alleles, homozygous genotypes are associated with less activity than are heterozygous genotypes CYP2C9*2 and CYP2C9*3, variant alleles of CYP2C9*1. Japanese patients with homozygous genotypes of the former have been reported to show 90 percent reduced activity, compared with 60 percent reduced activity in heterozygous patients. (T.J. O’Brien and colleagues, Clinica Chimica Acta 23:424:73-75, 2013).
Low activity associated with some alleles of cytochrome enzyme genes results not from a low-activity enzyme, but rather from low expression of an allele with normal activity (such as CYP3A4*22). Not all P450 cytochrome enzyme families have “null” forms without activity and not all are associated with ultra-rapid metabolizer phenotypes.sssss
Another “layer” of variability with potential clinical significance inheres in the enzymes of Phase II metabolism (those that increase water solubility of Phase I metabolites to facilitate excretion in urine or bile, such as uridinyl glucuronosyltransferases, sulfatransferases, or glutathione-S-transferase), and in those that promote clearance by allowing intracellular drug or metabolite to traverse cell membranes (“efflux”). The activities of the elements of Phases II and III of metabolism contribute, as do those of Phase I, to the variance of drug clearance. As your question implies, it is logical to wonder whether contributors to variance of drug clearance at baseline may become factors of risk during pharmacokinetic drug-drug interactivity, and the logic is corroborated in observations that inhibitors and inducers of CYP3A4 inhibit and induce, respectively, (the cell membrane drug transporter) para-glycoprotein. Similar correlations may obtain between inhibitors and inducers of other P450 cytochrome enzymes and other membrane transporters (such as multi-drug-associated proteins 2 and 3 and organic ion transporting polypeptide 2). Also contributing to variance of clearance (and, in extremis, severity of drug-drug interactivity) may be inhibition of P450 cytochrome enzymes by Phase II glucuronides (as suggested in a recently published clinical case report of inhibition of CYP2C8 by the glucuronide of clopidogrel).
Also important because of the magnitude of differences of activity associated with them are phenoconversions from high to low activity resulting from drug- or foodstuff-induced inhibition of P450 cytochrome enzyme activity. The potent CYP2D6 inhibitor paroxetine, for example, may increase concentrations of the CYP2D6 substrate desipramine 10-fold (although the mean increment, about five-fold, is still high). Not as large but nonetheless considerable are the increments of substrate bioavailability associated with decrements of activity of CYPs 1A2, 2C9, 2C19, 3A4, and 2E1 during coadministration of potent inhibitors. Some of the azole antifungal agents) inhibit the activity of several, prompting some to label them “universal inhibitors.” I am not aware of an inhibitor of similar magnitude in association with CYP2B6 (of which bupropion is a sensitive substrate), although a few low-potency inhibitors of CYP2B6 have been identified (the SSRIs, for example).
The sources of greatest variability of cytochrome P450 enzyme activity are genetic and pharmacologic. In practice, potent inhibitors of enzymes that metabolize coadministered substrates can be avoided and genotyping may not be necessary, if initial dosages of substrate are low and dosage is adjusted by small increments at intervals no shorter than those required for attainment of steady-state blood concentrations of the preceding dosage.
CPP: Do other important “normal” sources of variability exist?
Rosenblatt: They do, but are subject to differences in the propensities of parent compound and metabolite to induce therapeutic or adverse effects. Poor metabolizers of CYP2D6, for example, may be at less risk of cardiac adverse effects of encainide (prolongation of QTc, a putative indicator of risk for torsade de pointes cardiac arrhythmia) than are extensive or ultra-rapid metabolizers, because the metabolite is significantly more likely to induce them. The same logic applies to desirability of beta adrenergic blockade in propafenone-treated patients: like encainide, propafenone is a CYP2D6 substrate and is a more potent beta-blocker than is its Phase I metabolite. The advantage therefore may accrue to the poor metabolizers. It does not with pro-drugs that require CYP2D6-catalyzed conversion to active metabolites: tramadol and codeine are less potent for that reason in persons with low CYP2D6 activity, and tamoxifen requires CYP2D6 (and other cytochrome enzymes) for conversion to its active metabolite endoxifen for prophylaxis of breast cancer recurrence.
Genetic variability is important in that respect, and there are different kinds. One source of variability comprises the single nucleotide polymorphisms (SNP’s) of the genes that code for cytochrome enzymes—versions of the genes that differ by one nucleotide. Each of them exists for each enzyme (there are more than two dozen associated with CYP3A4), but they are not the most important contributors to variability of enzyme activity, generally, and also are not with respect to the variability of activity among P450 cytochrome enzymes.
CPP: You used the term “phenoconversion” a few minutes ago. What does it mean and what significance does it have?
Rosenblatt: It denotes a change in phenotype. Phenotype in the context of P450 cytochrome enzymes refers to activity of a cytochrome enzyme, which may be high or low in association with the same genotype. We were just talking about normal and low activity of CYP2C9, associated with wild-type and variant alleles, respectively. Apropos of the CYP2D6 enzyme and gene, the wild-type genotype is associated with normal activity (in “extensive metabolizers”); “intermediate metabolizers have a genotype that includes one or more variant alleles, while “poor metabolizers” have homozygous genotypes of variant alleles. Ultra-rapid metabolizers have two copies of the normal, wild-type gene.
The same genotype may be associated with different phenotypes: someone with the extensive metabolizer genotype, for example, may have the CYP2D6 enzyme activity (phenotype) of a poor metabolizer when treated with an inhibitor of CYP2D6, such as paroxetine or fluoxetine; or when afflicted with an inflammatory illness associated with high blood concentrations of cytokines (which inhibit CYP2D6). The genotype of CYP2D6 ultra-rapid metabolizers consists of a duplication of the wild-type allele, giving rise to a phenotype with a summed activity twice that of the enzyme expressed by one wild-type allele. CYP2D6 activity also may appear high during coadministration of dexamethasone or rifampin, according to some reports.
Low activity associated with an allele of a cytochrome enzyme gene may derive not from a low-activity enzyme, but from reduced expression of a normal-activity enzyme (as reported with respect to CYP3A4*22). Phenotype may be more informative than genotype for predicting likelihood of pharmacokinetic drug-drug interactivity and therapeutic response.
CPP: What are some of the best documented and most clinically significant pharmacokinetic drug-drug interactions?
Rosenblatt: Allow me first to point out that some have devolved from induction as well as from inhibition of P450 cytochrome enzymes. As I mentioned earlier, induction results from increased expression of a cytochrome enzyme gene, and more enzyme (hence, greater enzyme activity). Should a pharmacokinetic drug-drug interaction devolve from induction by one drug of the cytochrome enzyme that metabolizes a coadministered drug, reduction or loss of the latter’s effects may occur, because the induced enzyme has lowered its bioavailability.
Loss of therapeutic response to antiretroviral agents has been reported during coadministration of carbamazepine (a CYP3A4 inducer), as has rejection of renal grafts associated with failed immunosuppression, attributed in one instance to reduced cyclosporine bioavailability and in another, to reduced bioavailability of tacrolimus during coadministration of St Johns wort (whose constituents include a potent inducer of CYP3A4). Other inducers of CYP3A4 include phenytoin, avasimibe, oxcarbazepine, ritonavir, bosentan, efavirenz, etravirine, modafinil, nafcillin, amprenavir, aprepitant, armodafinil, echinacea, pioglitazone, prednisone, and rufinamide. Artemisinin is an inducer of CYP2C19; nevirapine induces CYP2B6; and inducers of CYP1A2 include montelukast, polycyclic aromatic hydrocarbon combustion products of tobacco smoking, smoked meats, charbroiled meats, cruciferous vegetables (cauliflower, Brussels sprouts), moricizine, omeprazole, and phenobarbital. Some induce more than one P450 cytochrome enzyme, such as phenytoin, rifampicin, and carbamazepine; modafinil (or R-modafinil) induces CYP3A4 and has been reported to induce CYP1A2. The mechanism of cytochrome enzyme induction is thought to be a “special case” of normative enzyme expression activated to a non-normative degree. How they do so is unknown.
We have been talking about genetic sources of variability within the cytochrome P450 enzymes, but environmental sources are no less influential. Environmental influence may be more variable and more elusive than are genetic influences. Unlike lifespan stability of the genome, environments change as people age. As persons age and become afflicted with more pharmacotherapy-sensitive conditions, they perforce become exposed to more medicines, of which many are substrates, inhibitors, and/or inducers of cytochrome enzymes. As usual, genetic and environmental variations interact, each contributing to variability of outcomes.
Another “layer” of variability inheres in the enzymes of Phase II metabolism (those that increase water solubility of Phase I metabolites to facilitate excretion in urine or bile, such as uridinyl glucuronosyltransferases, sulfatransferases, or glutathione-S-transferase), and in those that promote clearance by allowing intracellular drug or metabolite to be transported through cell membranes to the outsides of cells (“efflux”).
The activities of the elements of Phases II and III of metabolism contribute, as do those of Phase I, to the variance of drug clearance. As your question implies, it is logical to wonder whether contributors to variance of drug clearance at baseline may be related to (or become) factors of risk during pharmacokinetic drug-drug interactivity, and the logic is corroborated in observations that inhibitors and inducers of CYP3A4 inhibit and induce, respectively (the cell membrane drug transporter) para-glycoprotein. Similar correlations may obtain between inhibitors and inducers of other P450 cytochrome enzymes and other membrane transporters (such as multi-drug-associated proteins 2 and 3 and organic ion transporting polypeptide 2). Also contributing to variance of clearance (and, in extremis, severity of drug-drug interactivity) may be inhibition of P450 cytochrome enzymes by Phase II glucuronides (as suggested in a recently published clinical case report of inhibition of CYP2C8 by the glucuronide of clopidogrel).
CPP: What makes a pharmacokinetic drug-drug interaction clinically significant?
Rosenblatt: Clinical significance of pharmacokinetic drug-drug interactions is determined by whether change in the bioavailability of a drug interacted with results in treatment-limiting adverse effects or loss of therapeutic effect. Whether change in bioavailability is associated with either depends upon the breadth of the therapeutic ranges of the drugs interacted with, and how close drug-drug interactions bring their bioavailabilities to upper or lower thresholds of the therapeutic ranges. One factor of risk, therefore, is the bioavailability of the drug interacted with before the interaction, and how close it is to the lower or upper thresholds of the drug’s therapeutic range.
Clinical significance also is affected by whether there may be “escape” pathways of metabolism when the normative metabolic pathway is inhibited or saturated, and also, whether patients have illnesses affecting organ systems involved in clearance processes.
Another consideration is the pre-interaction activity of the enzyme of which the drug interacted with is a substrate. Someone who has an enzyme of low or no activity because they carry a variant allele of the gene for the enzyme may show less change in bioavailability when the enzyme is inhibited than someone with a normally active enzyme. If an escape pathway is operating before onset of the interaction and contributes substantially to maintaining bioavailability at dosage, inhibition of the normative, Phase I enzyme by an inhibitor that does not also inhibit the enzyme of the escape pathway would be less likely to result in clinically significant pharmacokinetic interaction. With no escape pathway, however, bioavailability of a drug that is a substrate of a low-activity enzyme expressed by a variant allele, or an enzyme of low activity expressed by a normal-activity wild-type allele during coadministration of an inhibitor of the enzyme, addition of another inhibitor of the enzyme and of its escape enzyme would be more likely to raise bioavailability above the upper threshold of the drug’s therapeutic range and result in clinically significant interactivity. This is where assay of the activity of the enzyme at baseline, along with the baseline bioavailability, may predict likelihood of interactivity with different drugs under consideration for coadministration (and favor selection for coadministration by their propensities to inhibit or induce escape pathways in persons with low-activity alleles of enzymes of phase I metabolism).
Among the most important moderators of clinical significance of pharmacokinetic interactivity is whether an orally administered drug is a substrate of CYP3A4 with a large first-pass hepatic extraction. If so, coadministration of either an inhibitor or inducer of CYP3A4 will disproportionately affect systemic bioavailability by changing the activity of CYP3A4 in small bowel mucosal cells and in hepatocytes. Evidence suggests that inhibition and induction of CYP3A4 also inhibits or induces, respectively, the conjugating enzyme and membrane transport protein (para-glycoprotein) that effluxes CYP3A4 substrate drugs from both hepatocytes and small bowel mucosal cells that absorb them from the bowel lumen.
CPP: What have been the best-documented clinical manifestations of pharmacokinetic drug-drug interactions resulting from inhibition or induction of cytochrome P450 enzymes?
Rosenblatt: Among the earliest reported examples of interactivity attributed to cytochrome enzyme inhibition in the modern era is the cardiac conduction abnormality torsade de pointes (French for “twisting of the points,” a descriptor of its electrocardiographic appearance). Best documented in that regard is not the relationship between drug exposure and torsade de pointes itself, but rather between the drug exposure and QTc prolongation, which is considered a powerful predictor of the risk of torsade de pointes (but withal has not been prospectively observed to be the endpoint of a “progression of” QTc prolongation). What has been shown conclusively is a direct relationship between the bioavailabilities of some drugs (such as the CYP3A4 substrates cisapride, astemizole, and terfenadine) and the durations of corrected QT intervals in persons receiving them during simultaneous exposure to CYP3A4 inhibitors
Among the other well-documented and clinically serious pharmacokinetic drug-drug interactions is, as mentioned, that of desipramine (a substrate of CYP2D6) and moderately to highly potent inhibitors of CYP2D6, such as paroxetine, fluoxetine, bupropion, sertraline at dosages of 100 milligrams per day or higher, duloxetine at dosages of 120 milligrams per day or higher, and CYP2D6 inhibitors with non-psychiatric indications (such as cinacalcet , terbinafine, or quinidine). The tricyclic antidepressants other than desipramine also are substrates of CYP2D6, and (like desipramine) also substrates of other P450 cytochrome enzymes (examples including doxepin, a substrate of both CYP2D6 and CYP2C19 and amitriptyline, metabolized by CYP1A2 as well as by CYP2D6). The pharmacokinetic interactivity between tricyclic antidepressants and inhibitors of CYP2D6 derives clinical importance from the narrow therapeutic margin of tricyclic antidepressants (whose lethal dosages may be as little as ten-fold higher than those of modal therapeutic daily dosages). The prescription rate of tricyclic antidepressants has diminished but is not yet stable, since prescription rates for prophylaxis of migraine headache, acute treatment of neuropathic pain, pruritic rashes, and gastritis have increased since FDA approved the first selective serotonin reuptake inhibitor, fluoxetine (Prozac), in late 1987, which, with other newer-generation antidepressants, have supplanted their use for treatment of depressive disorders.
Another well-documented and potentially serious pharmacokinetic interaction is that between warfarin and inhibitors of CYP2C9. Anticoagulant effects of warfarin derive from its S-isomer (which is four- to five-times more potent as an anticoagulant than warfarin’s D-isomer) and is preponderantly metabolized (7-hydroxylated) by CYP2C9). That the “phenoconversion” of CYP2C9 from normal to low activity (by, for example, coadministration of warfarin with a CYP2C9 inhibitor) is clinically significant is demonstrated by the well-documented lower warfarin dosage needs of patients with low-activity, variant alleles of CYP2C9 gene. (As with cytochrome enzyme genes generally, variant alleles of CYP2C9 are associated with low activity.)
Compared with that of the normal-activity, “wild-type” allele (CYP2C9*1), the activities of every variant so far studied (alleles 2,3,4,5,6,7,8,9,11, 12 , and 13) have been low, allowing adequate anticoagulation at low warfarin dosage. In a study reported nearly 15 years ago (H. Takahashi and H. Echizon, Clinical Pharmacokinetics 40(8):587-603, 2001), eighty-six percent of a sample of warfarin-treated patients in the U.K. who achieved adequate anticoagulation at low warfarin dosages (1.5 milligrams per day or lower) were found to have the variant alleles *2 or *3 (either as homozygotes or heterozygotes with the wild-type allele, *1). In a randomly selected control group of responders to warfarin at any dosage, thirty-eight percent had variant alleles. Other authors have reported greater risk of gastrointestinal bleeding in patients with low-activity, variant alleles of CYP2C9 during treatment with nonsteroidal anti-inflammatory agents (which, like S-warfarin, are substrates of CYP2C9). Incommensurate bioavailabilities, therapeutic responses, and adverse effects of other substrates of CYP2C9 at unexpectedly low dosages have been published or otherwise communicated in patients with variant alleles during exposure to substrates of CYP2C9. Similar findings have been reported with respect to substrates of other cytochrome enzymes whose genes exist as variant alleles (2C19, 2D6, and 3A4).
Other interactions documented clinically significant are those between some of the HMG CoA reductase inhibitors, such as lovastatin or simvastatin (substrates of CYP3A4 with large first-pass hepatic extractions) and some of the calcium channel blockers (those that potently inhibit CYP3A4, such as diltiazem and amlodipine); those of cyclosporine (CYP3A4 substrate) during treatment with potent CYP3A4 inhibitors, such as itraconazole; as I mentioned, those of drugs associated with QTc prolongation (the CYP3A4 substrates cisapride, astemizole, and terfenadine) during consumption of double-strength grapefruit juice (a CYP3A4 inhibitor (by dint of its constituent furanocoumarin bergamottin) in small bowel mucosal cells after single oral dosing, and an inhibitor of hepatic CYP3A4 during repeated grapefruit consumption); those associated with increased frequency of clinically significant gastrointestinal bleeding associated with nonsteroidal anti-inflammatory agents (sensitive substrates of CYP2C9) during coadministration with inhibitors of the latter.
Precedents of drug-drug interactions attributable to induction of CYP1A2 and of CYP3A4 can be considered a heuristic for estimating clinical significance of drug-drug interactivity from effect sizes of newer, less well-studied substrates. Reductions of bioavailability and therapeutic response have been reported in olanzapine-treated patients during periods of cigarette smoking and of smoking cessation. Reduced CYP3A4 substrate bioavailability (by hyperforin, a constituent of St. Johns wort) has been documented to be clinically significant in association with graft rejection attributed to induction of metabolism of immunosuppressants (and CYP3A4 substrates) cyclosporine and tacrolimus. So, too, has been antiretroviral failure in HIV disease during coadministration of the potent CYP3A4 inducer carbamazepine.
CPP: Do you know of any clinical application wherein routine therapeutic monitoring of drug bioavailability might raise the standard of care in psychiatry?
Rosenblatt: It can be useful when resorted to selectively rather than routinely. Routine measurement is useful for determining attainment of steady-state blood concentrations, which I have found helpful in improving tolerability of treatment and increasing medication compliance by reducing the likelihood of “overshooting” lowest effective dosages early in treatment.
Selective measurement of drug blood concentrations may provide information that yields “real world” risk reduction. In my experience, it has done so during assessment of abrupt changes in clinical state, when, for example, a depressed or euthymic patient with bipolar disorder manifests mood-switching to hypomania, which may have resulted from an increase in antidepressant bioavailability or a reduction of mood stabilizer blood concentrations; it has when patients begin to experience progression of initially subtle clinical effects of uncertain etiology (such as mild activation or restlessness, consistent with neuroleptic-induced akathisia or with mood-switching from depression to dysphoric hypomania or mania, (the former consistent with increase and the latter with reduction of neuroleptic bioavailability)); it has also informed optimal dosing for some patients, such as rapid metabolizers whose drug blood concentrations remain low despite modally effective dosages, or slow metabolizers deemed “placebo responders,” who may evince clinically significant drug effects at implausibly low dosages.
Monitoring drug bioavailability and/or cytochrome enzyme activity also may predict potentially important pharmacokinetic drug-drug interactions resulting in diminished or lost efficacy of treatment (as when an olanzapine-treated patient may become more symptomatic after resuming cigarette smoking, wherein combustion products of tobacco induce CYP1A2 activity and reduce olanzapine bioavailability, or when oxcarbazepine may be inducing CYP3A4 activity and thereby reducing alprazolam bioavailability accompanied by recurrence of panic attacks).
Allow me to elaborate on my assertion that routine measurement of drug blood concentrations has potential to yield net benefit clinically in determining when steady-state blood concentrations are attained. Doing so can help prevent raising dosage prematurely and “overshooting” optimally effective and tolerable dosages. Blood drug concentrations also can be used to calculate elimination half-lives, from which we can calculate time required to achieve steady-state blood concentrations. Without knowing when a drug’s blood concentrations reach steady-state, we might, for example, start someone on a medication that appears effective within 48 hours, but does not achieve steady-state blood concentrations until the fifth day; because of the improvement observed after 48 hours, we discharge the patient on day three, and then, two days later, learn that the patient has developed treatment-limiting adverse effects. The patient stops the medication, relapses, and is back in the hospital. That could happen with psychiatric drugs of any class, given the variability of durations required for attainment of steady-state blood concentrations.
CPP: Define steady-state blood drug concentrations, if you would.
Rosenblatt: A steady-state drug blood concentration is the blood concentration at which the amount of drug entering the systemic circulation equals the amount cleared from it. It typically occurs four to five elimination half-lives after the first dose of a regimen of regular, repeated dosing. (The elimination half-life is the time between peak plasma drug concentration after dosing (called Tmax) and the time at which it diminishes by one-half.)
Of the pharmacokinetic parameters, time to reach steady-state blood concentration is among the most clinically important with respect to adjusting dosage in non-emergent situations, because it preserves the relationship between dosage and bioavailability and informs the timing of dosage adjustments for optimal tolerability and continued compliance.
The time required to reach steady-state blood concentration after starting a drug or changing its dosage defines an interval during which dosage should remain constant (to reduce the likelihood that eventual steady-state drug blood concentrations may be higher or lower than needed for therapeutic effect and thereby confer higher risk than necessary of adverse effects or treatment-resistance). In practical terms, knowing the time at which steady-state drug blood concentrations develop may reduce likelihood of adverse effects, and increase the likelihood that responsive patients will be maintained on lowest effective dosages.
Drug action throughout the day reflects bioavailability within a range between minimum (“trough”) blood concentrations before dosing and maximal (peak) blood concentrations (CMaxs) occurring at times after dosing (Tmaxs). Those blood concentrations depend upon dose, and differ among drugs; they may vary with routes of administration, times of last feeding during oral dosing, effects on absorption and/or clearance of coadministered drugs, and diseases of liver, kidney, or heart.
If dosage is raised before attainment of steady-state blood concentration of the initial dosage, you may see optimal therapeutic effects as drug blood concentrations rise to the lower threshold of a therapeutic range of drug concentration (and before they reach steady-state). As they continue to rise to steady-state or higher concentrations, therapeutic effects may become compromised by adverse effects. The same applies to dosage reductions following an earlier dosage reduction, wherein steady-state blood concentrations for the new, lower dose have not yet been attained. If dosage is lowered to mitigate a dosage-related adverse effect, the latter may improve or remit at a lower drug blood concentration that continues to fall to a steady-state drug concentration of the newer, lower dosage. That newer, lower dosage, however, may give rise to a steady-state blood concentration below the drug’s threshold blood concentration for therapeutic effect. If that occurs, therapeutic effects may be lost along with adverse effects.
If you know the mean interval required for a drug to produce steady-state blood concentration after dosage change, and if you know your patient’s consistent deviation from that, along with the relationship between change in dosage and change of drug blood concentration for the drug (and in your patient), increasing personal experience with that drug will allow greater precision in dosing for maximal benefit and minimal risk over intervals as brief as pharmacokinetic and clinical parameters permit.
No new technology is needed to apply an old principle (prediction of time to reach steady-state blood concentrations), whose practical value is to enhance the timing of dosage adjustments and inform the timing of assessing their effects with the intention of increasing the likelihood of achieving post-discharge bioavailabilities that are both effective and tolerable.
CPP: If a patient were to show partial response with good tolerability of treatment, and you wished to increase dosage, upon what basis would you determine when and by how much?
Rosenblatt: Conventional therapeutic range is helpful in that regard, and my decisions about dosing derive from it. When therapeutic effects of a drug emerge and predominate in a patient at one bioavailability (expressed as the area under the curve of serum concentration over time) and intolerable or unsafe adverse effects do so at another, higher one, the intervening bioavailabilities constitute a drug’s therapeutic range for a patient at a given time, age, and state of health.
Therapeutic range of drug dosage is a proxy for therapeutic range of drug bioavailability. For most drugs at steady-state, drug dosage is directly correlated with drug bioavailability below the upper thresholds of their therapeutic ranges. Also for most drugs, bioavailability within their therapeutic ranges is directly correlated with clinical effects. In fact, one can define a therapeutic range as a span wherein dosage and/or bioavailability correlate positively with therapeutic and/or common, non-treatment limiting adverse clinical effects. The upper threshold of the range is where the slope for adverse effects becomes more steeply positive and that for therapeutic effects becomes zero or negative. Therapeutic ranges of dosage in patient populations are similarly shaped, but may be lower or higher, depending principally upon pharmacologic and genetic factors (the former, exposures to coadministered inhibitors of the enzymes that catalyze the first steps of drug metabolism, and the latter, frequencies of variant alleles of genes, which typically express enzymes lower in activity than those expressed by their wild-type, normal alleles).
Lower enzyme activities are associated with higher bioavailabilities at comparable dosages. Therefore, therapeutic ranges of drug bioavailability may be indistinguishable between populations that have different therapeutic ranges of dosage. The highest-activity phenotype of CYP2D6 (“ultra-rapid metabolizers”) is unique, in that it devolves from duplication of the normative, wild-type allele. Highest-activity phenotypes of other cytochrome genes are attributed to wild-type alleles (unduplicated) and/or exposures to enzyme inducers (in medicines, foodstuffs, and herbal formulations). Those patients’ ranges of therapeutic dosage have higher upper and lower thresholds, when dosed for lowest effective dosages.
Typically, the upper limit of a therapeutic range is defined (in units of bioavailability or dosage) as the lowest blood concentration or dosage associated with a preponderance of treatment-limiting adverse effects. Accordingly, drugs that afford benefit at relatively low serum concentrations or dosages and continue to afford benefit and remain tolerable and safe at high serum concentrations or dosages are, by definition, drugs with wide therapeutic ranges. When a drug is effective but tolerated over a small range, it is deemed to have a narrow therapeutic range. Drug-drug interactions are more likely to be treatment-limiting in patients receiving drugs with narrow therapeutic ranges, characterized by steeply-sloped dose-response relationships, wherein small changes in dosage or bioavailability are associated with disproportionately large changes in drug adverse effects and/or clinical response. Drug-drug interactions are better tolerated (and pose less risk) when drugs that are interacted with have wide therapeutic ranges.
In practice, persons who are taking a drug with a wide therapeutic range at relatively high dosage may experience dose-related effects considered adverse in package labeling, but, in the event, may be therapeutic. An example that everyone knows is antipsychotic-induced sedation in a patient who is severely agitated or aggressive, and/or has severe sleep disorder. In those patients, sedation may be a therapeutic side effect during early treatment.
CPP: From what you were saying earlier, does it follow that high pre-interaction drug blood concentrations are risk factors for clinically significant interaction due to inhibition of a drug’s Phase I enzyme, and that low ones are risk factors with respect to drug interactions due to induction of the cytochrome enzyme of which the drug is a substrate?
Rosenblatt: Smaller-percentage increases of drug bioavailability associated with pharmacokinetic drug-drug interactions become more important at higher dosages or blood concentrations of drugs interacted with, when their documented toxicities are associated with supratherapeutic blood concentrations. Inhibition of desipramine metabolism by sertraline (between 25 and 75 milligrams per day) increases the bioavailability of desipramine by thirty to fifty percent (which may be clinically significant if initial desipramine blood concentrations are 300 ng/ml or higher).
No minimum percentage increase of bioavailability in the context of pharmacokinetic drug-drug interactivity has been established safe at final tricyclic antidepressant blood concentrations exceeding 300 ng/ml. (Patients with steady-state desipramine, imipramine, amitriptyline, and nortriptyline concentrations of 300 ng/ml or higher account for published or reported cases of clinically significant toxicity that include information about blood drug concentrations.)
Similarly, small percentage reductions of drug bioavailability (via enzyme induction, for example) are more likely to prove clinically significant when initial bioavailability is low and clinical adversity of drug-drug interactivity has been documented to manifest preponderantly as loss of therapeutic activity.
Every drug therapeutic range has a lower therapeutic threshold associated with emergence (or disappearance) of therapeutic effect, below which enzyme inducers may reduce bioavailability and conduce to relapse, particularly in patients whose drug blood concentrations are low before an inducer is coadministered. If the interacting drug (Drug B) induced the activity of the cytochrome enzyme that metabolized a drug (Drug A) by a mean of 30 percent, and it caused blood concentrations of Drug A (with a therapeutic threshold of, say, 70 ng/ml) to fall in a patient whose blood concentrations were 85 ng/ml before the inducer was coadministered, coadministering something other than Drug B should be considered.
Generally, concern is greater with respect to drugs that raise bioavailability by inhibiting cytochrome enzyme activity in patients with already high bioavailabilities (the patient receiving a high lamotrigine dosage, to which valproic acid has just been added, for example). In those whose initial bioavailabilities of a drug are low, the more exigent risk is induction of its phase I cytochrome P450 enzyme, which, by lowering bioavailability, could reduce or abolish therapeutic effects (the long-term cigarette smoker, whose delusional disorder has improved during treatment with olanzapine in a smoke-free inpatient unit, whose discharge home is imminent, for example).
Lisdexamfetamine (Vyvanse) is an inactive lysine-amphetamine dimer that red blood cell peptidases split to release dextroamphetamine, already available and indicated for attention-deficit/hyperactivity disorder. Its onset differs from that of immediate-release dextroamphetamine by a latency necessitated by the duration required for its conversion to free lysine and free d-amphetamine. Yet, its total duration of stimulant action need not exceed that of immediate-release d-amphetamine at some hypothetical dosage of the latter that achieved a (safe and tolerable) bioavailability exceeding a threshold for clinical activity in advance of Tmax (time to reach peak blood concentration) and continuing to exceed it for a half-life or longer after Tmax.
I suspect that unwanted stimulation and perhaps hypertension would keep that hypothetical dosage hypothetical. I mention it, however, as a reminder that duration of action is a function of bioavailability exceeding a threshold for therapeutic activity, and that an increased dosage of an immediate-release formulation of a drug may, by virtue of its higher bioavailability, eventuate in a longer duration of action than that of a lower dosage, and if safe and tolerable at peak blood concentration, may be credibly competitive with a more costly extended-release formulation, with respect to duration of action. I don’t know what that dosage may be in the case of d-amphetamine or what dosage of lisdexamfetamine might have a similar or shorter duration of action, but I do know that I have seen no studies of durations of action after a range of immediate-release stimulant doses, and no studies of how they compare with those of a range of dosages of extended-release formulations of the same stimulant. The principle would apply no less to immediate-release and extended-release formulations of other drugs.
CPP: The rate of prescription psychotropic drug exposure in the U.S. was reported in a proprietary study by Medco Health Solutions (now Express Scripts) to have increased overall from 2001 to 2010. The study’s sample comprised 2.5 million adults, adolescents, and children, among whom the overall rate of exposure to psychotropic medications (antidepressants, antipsychotics, anxiolytics, and ADHD) was twenty percent in 2010. Will increased risk of drug-drug interactions be the logical consequence of increased drug exposure?
Rosenblatt: Allow me to make a point about evidence-based rates of pharmacokinetic drug-drug interactions, because the answer to your question about extent of patients treated pertains to them: The most frequently reported drug-drug interactions are not necessarily those that occur most frequently. The principal artifacts that bias the estimates are extent of use of individual drugs in the population under study and ascertainment errors, which “self-select” for detection by the clarity and conspicuousness of the “endpoints” of toxicity attributed to them (among them historically: seizures, graft rejection, hepatic failure, unstable level of consciousness, and sudden death of cardiovascular origin). The first artifact is clear-cut: If more Drug or Drug Class A is being prescribed in a population than Drug or Drug Class B, and each is associated with the same true rate of an adverse effect, more cases of the adverse effect will occur in association with Drug A. The second artifact is only as clear-cut as is the attribution of causality of an adverse event to a drug or to the illness (including its diatheses or complications) for which it is being prescribed
I am familiar with the Medco study. One of its findings was that prescription rates of psychotropic drugs increased over the duration of the study (2001 through 2010) to 20 percent overall, with an increase to 25 percent in women. That represented a change of a little more than 20 percent in adults over ten years. Use of atypical antipsychotics increased more in men, but was higher in women, as were rates of use of other classes of psychotropics. The largest increase occurred among children, who withal showed a reduction in treatment with stimulants and an increased use of atypical neuroleptics. Use of stimulants increased markedly among young to middle-aged women (20- to 44-years-old), as did use of anxiolytics, such as alprazolam or other benzodiazepines. Anxiolytic use declined in women sixty-four and older, who at that age sustained their earlier, high rates of antidepressant usage.
So then, rates of treatment with psychotropic drugs increased from 2001 through 2010 by a little over two percent per year, with treatment consisting of newer-generation drugs that were no less effective than their predecessors, and were safer and better tolerated. Who would not have expected prescription rates to increase at the modest rate observed, given greater benefit-to-risk?
CPP: What about rates of prescribing psychotropic drugs to children, about whom there exists less evidence of safety?
Rosenblatt: That the percentage of children receiving psychiatric pharmacotherapy now is higher than it was a generation ago is unexceptionable. The reasons for the increase are the same as those that increase prescription rates in adults, and pertain to percentage of the population afflicted with what it treats, extent to which the latter is diagnosed and treated, access to doctors who will prescribe it, robustness of benefit, and acceptable risk.
CPP: To what extent might prescription rates devolve from higher rates of diagnosis of disorders, such as ADHD, for which response rates are high and risks low?
Rosenblatt: The Medco study found that about five percent of children were receiving psychotropic medications in 2010, preponderantly for ADHD and for depression. The response rate in ADHD without comorbidity is high right now (70 to 75 percent with stimulants). I say, “right now,” because, historically in psychiatry, response rates to highly effective treatments fall over time, as occurred with lithium for bipolar disorder. Initial response rates were higher shortly after it was introduced than they are today, which is attributed to increasing heterogeneity of patients receiving lithium.
CPP: You expect response rates of current-generation psychotropics to fall, then, as treatment populations become more heterogeneous.
Rosenblatt: I suspect that the rates with which patients are diagnosed with a disorder for which new treatments are reported highly responsive in preapproval clinical trials or post-approval surveillance studies may in consequence increase, and that that may reflect cognitive bias of diagnosticians and reporting bias of patients and families. Emergence of a highly effective treatment also may induce previously unevaluated persons to present for treatment. It is tempting to speculate that diagnostic bias may have been operating after introduction of lithium, wherein some persons with a diagnosis of unipolar depression may have been rediagnosed as having bipolar depression, and that other revisions favoring diagnosis of bipolar disorder or bipolar spectrum disorder may have been made in diagnoses of borderline personality disorder, intermittent explosive disorder, impulse control disorder, and other disorders wherein abrupt changes of mood or behavior may have been interpretable as changes (or “swings”) of mood state.
When a homogeneous population of persons with a reliably diagnosed disorder becomes more diverse, less uniformly diagnosed, and afflicted with one or more comorbid disorders (that is, less like a research sample and more like an outpatient mental health clinic population), its responsiveness to treatment becomes more variable. I believe that that may be occurring now with stimulant response in children diagnosed with ADHD. The Medco study reported a decline in the rate of stimulant prescriptions for children, accompanied by an increased rate of prescriptions for atypical neuroleptics. Some recent observational evidence suggests that atypical neuroleptics may be no less effective than are stimulants for some children whose diagnoses include ADHD. If systematic studies corroborate that, we will doubtless see more children with ADHD who are treated with atypical neuroleptics or conjointly treated with stimulants and atypical neuroleptics. I know of no clinically significant pharmacokinetic drug-drug interactivity between stimulants and atypical neuroleptics or between the latter and alpha-2 agonists.
Diagnostic heterogeneity impacts treatment responsiveness generally, and when it does, prescription rates (of drugs approved for specific indications) change accordingly. In my opinion (and based solely on personal experience), stimulant prescriptions in children began to decrease when aggressiveness and impulsivity began to be considered manifestations of hyperactivity and inattentiveness that were expected to improve during treatment with stimulants. In most cases, they did not, and in some instances, they worsened during treatment with stimulants. The alternative, which proved highly effective, was an atypical neuroleptic—risperidone initially, and now increasingly aripiprazole (because of its relatively lesser propensity to induce weight gain and metabolic adversity, generally, and its low likelihood of inducing hyperprolactinemia (associated with increased risk of gynecomastia in males, menstrual disturbances in females, and bone demineralization; some evidence also broaches increased risk of breast cancer)).
As the number of persons who respond insufficiently to a drug increase, so do those receiving non-standard or off-label treatments. That group would include those receiving polytherapy (as augmentation or combination treatment), in addition to any drugs they may be receiving for diverse medical disorders. Polytherapy begins at the end of an unexceptionable “syllogism:” If symptoms and/or behaviors meet DSM-V (and beyond) criteria for a disorder, they may be treated; if they may be treated, they may be treated with pharmacotherapy; if they may be treated with pharmacotherapy, they may be treated with pharmacotherapy on- or off-label; if they may be treated on- or off-label, they may be treated off-label with more than one medication. Treatment with more than one medication introduces risk of drug-drug interactivity.
CPP: If comorbidity is defined as more than one diagnosed disorder, can one argue without irony that DSM-V, with its larger number of diagnosable disorders than DSM-IV (each treatable on- or off-label with pharmacotherapy), may increase the likelihood of comorbidity, and thereby of exposure to polytherapy?
Rosenblatt: None of the entities in the prior DSMs or current DSM (particularly those deemed mood disorders) have been independently validated, and most lack professional consensus. Bias in its application is rife and its impacts on healthcare costs have not been systematically scrutinized. The process of formulating a diagnosis of record, for example, is collaborative in some settings (as when made in treatment team meetings) and may evoke unsolicited but influential information about whether or to what extent insurers preferentially reimburse services for different diagnostic entities. Family history may become a source of bias devolving from overidentification of family members with psychiatric diagnoses or personal experiences of pharmacotherapy. Representation bias may conduce to an expectation that academic under-performance is “ADHD until proven otherwise.” Another cognitive bias affecting diagnosis in children is validation by treatment response, a confirmation bias, wherein symptomatic improvement becomes tantamount to accuracy of syndromic diagnosis. (Stimulants improve attention and reduce distractibility in children and adults, whether or not they meet diagnostic criteria for ADHD.)
That current-generation psychopharmaceuticals are more likely to be prescribed than were their first-generation counterparts comports with their superior tolerabilities and wider therapeutic ranges: The more recently introduced drugs afford as much benefit at lesser acute (and in some instances, such as tardive motor syndromes) chronic risk. Their higher ratios of benefit to risk also conduce to greater likelihood that comorbid disorders or symptoms will be treated adjunctively in a larger subgroup of patients who are tolerating partially effective treatment of their primary disorders, but whose global outcomes remain insufficient because of persistent comorbid symptoms.
CPP: Does science lag behind practice with respect to drug-drug interactivity?
Rosenblatt: It lags behind practice in all civilized societies where clinical researchers attempt to answer questions pertaining to toxicity or adversity. Studies about drug-drug interactions are about what can go wrong during treatment, and demonstrating causality conclusively requires rechallenge, which means that you must give someone something that may be harmful under conditions that ensure safety. It is a difficult and risky business. It may require endpoints that are subtle and difficult to discern, and that may bias results in a way that increases the likelihood of false positive findings (as do studies of teratogenicity associated with in utero drug exposures).
To find thresholds of toxicity in humans requires retrospective study of samples that may be non-naturalistic or non-random, animal studies, or in vitro studies. Clinical studies of toxic doses of drugs intended for humans may be as subject to restraint as are contemporary studies of teratogenicity. And in those studies, human findings responsible for categories of safety during pregnancy are typically about what is not found during naturalistic conditions. The most plentiful evidence for human toxicity caused by drug-drug interactivity comes from case reports and analyses of poison center data. The latter do not stratify patients for numbers (or any other feature) of coadministered drugs.
Before high through-put screening (of drug candidates) became routine during commercial drug development, discovery that a drug potently inhibited or induced the metabolism of other drugs, or that its Phase I metabolism was confined to one cytochrome enzyme occurred after it was approved. When its inhibitory effects on cytochrome enzymes became known, Tagamet (cimetidine), formerly a nine-figure-grossing blockbuster, was disused and supplanted by ranitidine (Zantac), at that time the sole alternate to cimetidine. Ranitidine quickly became known as the safer histamine-1 antagonist, and over the short span preceding the introduction of famotidine (less likely than ranitidine to interact pharmacokinetically with cytochrome enzymes), itself became a blockbuster. Mibefradil and perhexilene, with their many pharmacokinetic drug-drug interactions, were withdrawn. The “lesson” was not lost during subsequent drug development.
Even so, the advantages of high throughput screening principally redound to safety and adverse effects. This is done by substitutions on ring structures of chemical nuclei that distinguish major classes of psychotropic drugs by virtue of affinities to neurotransmitter and transporter sites that correlate with a drug’s propensity to induce different types of therapeutic and adverse effects (histaminergic receptor affinities and sedation or weight gain, for example; muscarinic receptor affinities for anticholinergic adverse effects; affinities for binding to different adrenergic receptors and postural lightheadedness). Substituting ring consituents of a candidate drug molecule can be done more predictably to “dial out” adverse effects and raise low thresholds of toxicity that plagued their predecessors. We are left with safer, better-tolerated drugs thereby; but so far, correlations of structure and therapeutic activity have proved less manipulable because determinations of the latter are more complex, multidimensional, and interactive. In practical terms, endpoints for adverse effects and safety are more perspicuous and more reliably measured than are endpoints that proxy improvement and worsening of psychiatric disorders. We are just recently becoming aware that what has traditionally been diagnosed as bipolar depression represents a mixed state of depressive and hypomanic or manic symptoms, which have been characterized for a long time (as has antidepressant-emergent worsening of hypomanic or manic symptoms). When diagnosis corresponds more closely with etiopathology of disorder, we will have better ideas of what to “dial in” and how to do it. Right now, the “maps” for both are relatively crude and overlap like two shadows.