You are on page 1of 12

Therapeutic Index

By Raed D. Abughazaleh1 and Timothy S. Tracy2


Keywords: Toxicity, efficacy, median lethal dose (LD50 ), median toxic dose (TD50 ), median effective dose
(ED50 )

Abstract: Therapeutic index is an expression of the relative toxicity as compared to efficacy


of a given compound. In animal studies, it can be computed as the ratio of the median
lethal dose (LD50 ) to the median effective dose (ED50 ). In humans, it is frequently expressed
as the ratio of the median toxic dose (TD50 ) to the ED50 . Several issues limit the utility
of the therapeutic index, and to-date there remains no official list of drugs considered
to possess narrow therapeutic indices. Mortality is a poor measure of a drug’s toxicity
and raises ethical concerns in animal studies. Similarly, ED50 can be a poor measure of a
drug’s effectiveness because of lack of consistency in endpoint biomarkers and because
the ED50 may be different for multiple conditions the drug may be used to treat. When
LD50 and ED50 are obtained in animals, any extrapolation to humans is limited. In addition,
multiple factors of interindividual and intraindividual variability exist in humans and can
limit the utility of a therapeutic index estimate. Interindividual variability can be caused by
either pharmacokinetic or pharmacodynamic differences, and may play an important role
in brand-to-generic switching of narrow therapeutic index drugs. Intraindividual variability
can also be related to physiologic factors that affect absorption, distribution, metabolism,
or excretion of a drug. In conclusion, use of a therapeutic index requires an understanding
of its limitations, and may be most useful in the preclinical stages of drug testing where
limited experience of the drug’s administration is available. To date, there remains no
official guidance to research laboratories and pharmaceutical companies on the optimal
use of therapeutic index.

The term “therapeutic index” has been used for several decades, yet the concept remains widely misunder-
stood and its significance is controversial. In this respect, although the general concept of the therapeutic

1 University of Minnesota College of Pharmacy, Minneapolis, Minnesota


2 University of Minnesota College of Pharmacy, Department of Experimental and Clinical Pharmacology, Minneapolis,
Minnesota

c John Wiley & Sons, Inc. and


This article was originally published online in 2007 in Wiley Encyclopedia of Clinical Trials,
republished in Wiley StatsRef: Statistics Reference Online, 2014.

c 2007 by John Wiley & Sons, Inc.


Copyright 1
Therapeutic Index

index is describable by most individuals, issues such as what defines a “narrow therapeutic index” drug,
pharmacokinetic and pharmacodynamic (See Pharmacokinetics and Pharmacodynamics) variability, in-
terindividual and intraindividual variability, and adequacy of biomarkers and surrogate endpoints among
others continue to confound determination and clinical utility of the “therapeutic index.” In this article,
definitions of therapeutic index, “narrow therapeutic index,” and the aforementioned confounding issues
will be discussed, and a perspective on the clinical utility of a therapeutic index will be presented.

1 Definition
Therapeutic index is most commonly defined as the ratio of the median lethal dose (LD50 ) to the median
effective dose (ED50 ), as determined in preclinical animal studies (See Preclinical treatment evaluation).
The median LD50 is the dose that results in death in half the population of animals to which it is
administered, whereas the median ED50 is the dose required that elicits the desired pharmacologic effect
in half the studied population [1] . Therapeutic index, therefore, reflects the selectivity of a drug to elicit a
desired effect rather than toxicity. A higher (or greater) therapeutic index suggests a higher tolerance to a
dose increase beyond ED50 , or a wider “margin of safety.”
Drugs with a low therapeutic index are frequently also referred to as “narrow therapeutic index”
drugs [2] . According to the Code of Federal Regulations (CFR) §20.33, the United States Food and Drug
Administration (FDA) defines narrow therapeutic index drugs as those that “exhibit less than a 2-fold
difference in median lethal dose (LD50 ) and median effective dose (ED50 ) values, or have less than a 2-fold
difference in the minimum toxic concentrations and minimum effective concentrations in the blood,
and safe and effective use of the drug products requires careful dosage titration and patient monitoring.”
However, the FDA does not claim narrow therapeutic index to be a formal designation, and does not
mention it in the publication Approved Drug Products with Therapeutic Equivalence Evaluations [The Orange
Book] [3–5] . Others have defined narrow therapeutic index drugs as those “for which relatively small changes
in systemic concentrations can lead to marked changes in pharmacodynamic response” [6] . To date, no
official listing of narrow therapeutic index drugs has been published, and the FDA has taken the position
that each drug is unique, and thus different drugs should not be clustered into discrete groups [3] . Despite
the FDA’s position, at least two state legislative attempts have been made to define narrow therapeutic
index drugs, including legislation proposed by the state of New York in 1997 (NY proposed Assembly Bill
8087.A).
The therapeutic index of a drug can be determined through the use of quantal dose-effect plots. This
methodology plots the Gaussian normal distribution of the percentage of subjects either responding to
or dying from increasing doses. From this curve, cumulative response and death curves are plotted, and
the ED50 and LD50 are obtained (Figure 1).
Throughout all of this, one must recognize that a determined therapeutic index from preclinical animal
studies may not directly extrapolate to the human situation when used to estimate a drug’s prospective
utility and safety in humans. Certainly, it is obvious that determination of an LD50 in humans is not
possible. This has led to alternative methods of estimating the therapeutic index in humans via methods
using the therapeutic dosing ranges of a drug and related measures. For instance, the median toxic dose
TD50 may be substituted in place of the LD50 in the numerator of the therapeutic index determination.
In this case, the median TD50 is the dose that produces a defined toxicity (See Toxicity (Adverse Events))
in 50% of the subjects examined (in this case, it may be human subjects or animals). Apart from the
obvious advantage of not requiring a lethal endpoint, one can also gather data more relevant to the human
situation.
Certain safety factor (CSF) is another variation of therapeutic index in which TD(LD)50 is replaced by
TD(LD)1 , and ED50 is replaced by ED99 . The obvious theoretical advantage is from minimizing toxicity
in the experimental subjects. However, it is sometimes not possible to achieve 99% efficacy in subjects
2 c 2007 by John Wiley & Sons, Inc.
Copyright
Therapeutic Index

Figure 1. Quantal dose-effect plot. Each curve represents cumulative frequency distribution of animals ex-
hibiting efficacy or toxicity. Therapeutic index is calculated from dividing median lethal dose (LD50 ) by median
effective dose (ED50 ).
even at the maximal dose of a drug. Using a 1% toxicity threshold (TD1 ) may be too ambitious as well
because placebo may cause a higher than 1% incidence of mild side effects. Finally, use of LD1 as a toxicity
measure may compromise experimental accuracy as the lethal dose determinations require very large
sample sizes to assess 1% lethality with reasonable statistical certainty. The CSF has yet to be validated in
the literature as a reliable method of estimating therapeutic utility of an agent. Though few publications
have used this method, the work of Zhou et al. [7] describes an interesting application of the CSF as
related to intravenous administration of emulsified isoflurane. In this study, the investigators found that
emulsified isoflurane could be given safely by the intravenous route and that the CSF for emulsified
isoflurane was comparable with that of intravenous propofol.
Another method proposed as a clinical alternative for therapeutic index and closely related to the
estimation using TD50 and ED50 is the use of the therapeutic concentration range (See Therapeutic Dose
Range; Minimum therapeutically effective dose) upper and lower limits as an estimation of the therapeutic
index. Table 1 lists the proposed therapeutic ranges for some commonly monitored drugs [8] .

2 Issues Associated with the Determination of TD50

The use of TD50 as the primary measure of toxicity evaluation in estimating the therapeutic index
is confounded by not so obvious factors that are associated with its determination and merit further
examination. Toxicity from drugs can be far more complex to assess and understand than efficacy. With
efficacy evaluation, one is looking for a specific effect or a set of specific effects to monitor using specific
endpoints, which are typically based on previously known mechanisms of action and extrapolated ED50 ,
and are generally predictable in terms of onset, magnitude, and duration. Toxicities to a certain extent
can be predictable, such as those that are extensions of the drug’s known mechanism(s) of action and that
tend to be consistent across a class of drugs. For example, a primary toxicity of alpha-receptor blockers
c 2007 by John Wiley & Sons, Inc.
Copyright 3
Therapeutic Index

Table 1. Proposed therapeutic ranges for some commonly monitored drugs

Drug Therapeutic Range

Amikacin 20–30 :g/mL (peak), <10 :g/mL


(trough)
Carbamazepine 4–12 :g/mL
Digoxin 1–2 ng/mL
Gentamicin 5–10 :g/mL (peak), <2 :g/mL (trough)
Lidocaine 1–5 :g/mL
Lithium 0.6–1.2 mEq/L
Phenytoin 10–20 :g/mL
Procainamide 4–10 :g/mL
Quinidine 1–4 :g/mL
Theophylline 10–20 :g/mL
Tobramycin 5–10 :g/mL (peak), <2 :g/mL (trough)
Valproic acid 50–100 :g/mL
Vancomycin 20–40 :g/mL (peak), 5-15 :g/mL
(trough)

Source: Adapted from Schumacher 1995 [8] .

is orthostatic hypotension, and fibrinolytic agents are notorious for increasing the risk of bleeding, both
toxicities being obvious extensions of the respective drugs’ mechanisms of action.
However, there are many forms of toxicities that are idiosyncratic, meaning that the mechanism re-
sponsible for the toxicity is unknown or unexplainable. For example, the exact mechanism underlying
why the statin hypolipidemic agents (See Statins) can cause muscle toxicity including myopathy and
rhabdomyolysis [9] remains elusive. This toxicity is rather unpredictable, can occur anytime during treat-
ment course, can be associated with virtually all drugs from the class, and does not appear to be dose
related. However, there is consensus that the toxicity is related to the potency of the statin such that the
incidence of muscle toxicities is more likely with the more potent statin agents [9] . It is true that there are
established risk factors for such toxicities; however, these are not prospective predictors of the toxicity but
mere associations. Any one or a combination of at least three mechanisms are currently proposed for this
statin-associated muscle toxicity, but the exact cause(s) remain theoretical and not definitively established
[9]
.
Drug toxicity can sometimes be of an allergic nature and can occur as type I immediate, type II cytotoxic,
or type III immune-complex hypersensitivity reactions [10] . Type I hypersensitivity is immunoglobulin
E (IgE) mediated and can be manifested in its most severe form, systemic anaphylaxis, such as that
associated with penicillins. Type II hypersensitivity reactions are IgG mediated and can be severe as
well. Examples of type II hypersensitivity reactions include quinidine-induced hemolysis or quinine-
induced thrombocytopenia. Lastly, type III hypersensitivity reactions are also IgG mediated and can
result in conditions like serum sickness, such as is noted with the penicillin-type drugs. Each of these
types of hypersensitivity reactions is typically unpredictable when it initially occurs and can be greatly
influenced by an individual’s specific biological characteristics, such as previous exposure to similar
compounds or antibody titers. Determination of therapeutic index is therefore of little value with respect
to hypersensitivity reactions because there may not be a clear association of drug dose and observed
toxicity. Thus, margin-of-safety determinations are extremely difficult when hypersensitivity reactions
occur as an element of the drug’s toxicity profile.
Another area where determination of therapeutic index lacks guidance is in the case of toxicities
following chronic administration of the drug that goes beyond the timeframe of clinical trial evaluation.
In such cases, long-term toxicities are usually identified during postmarketing surveillance studies (See
4 c 2007 by John Wiley & Sons, Inc.
Copyright
Therapeutic Index

Postmarketing surveillance of new drugs and assessment of risk), which is unfortunately after the drug
has become available to the general public. For example, amiodarone is an antiarrhythmic agent that
exhibits a very long half-life (t1/2 = 35 days), resulting in substantial accumulation of the drug for very
long periods of time. In this case, such long-term accumulation can ultimately result in toxicities, like
pulmonary fibrosis, that do not become evident for a substantial period of time following initiation of drug
administration [11] . Such toxicities may be beyond the scope of evaluation in preclinical animal studies
and may not be reflected in determinations of the TD50 , thus rendering determination of the therapeutic
index meaningless in this situation.

3 Issues Associated with Determination of the LD50

Upon careful consideration, it becomes apparent that issues related to the LD50 and its determination are
abundant [12] . For example, mortality can be an imprecise measure of a drug’s toxicity because toxicity is
generally a continuum and not an absolute endpoint. In fact, one may be willing to accept a certain level
of toxicity in exchange for a given benefit depending on the drug in question (e.g., anticancer agents).
In addition, there are species differences that prevent accurate extrapolation of LD50 (and ED50 ) from
animals to humans. This can include such factors as interspecies pharmacodynamic and pharmacokinetic
differences. One must be cognizant of the correlations from animals to humans of these types of studies
when determining the potential information that may be gained.
A second factor that may profoundly affect the determination of LD50 (and TD50 for that matter) is
the effect of drug formulations on the toxicities/effects noted. For example, slow or sustained release
(SR) drugs tend to show a steady and slow rise in serum concentration, resulting in a blunted maximum
concentration (Cmax ) and a longer time to maximum concentration (Tmax ) as compared with immediate
release (IR) preparations (Figure 2). The end result may be that one experiences drug concentrations
within the toxic range after administration of an immediate release formulation, but a sustained release
formulation might not achieve this same drug concentration. Frequently, solution formulations are used
in animal studies for ease of administration and result in immediate drug absorption whereas the final
formulation for human use may be a tablet (and potentially a sustained release tablet or capsule) that will
give a very different pharmacokinetic profile with potentially lower maximum concentrations. Likewise,
use of other routes of administration such as intramuscular or transdermal administration may also give
pharmacokinetic profiles analogous to a sustained release preparation, and thus have potentially a more
blunted peak concentration. However, it should be noted that this effect of sustained release formulations
can cut both ways in that it can also confound the determination of the ED50 because this kind of kinetic
profile can potentially alter the period of time that effective concentrations are reached.
Finally, one must consider conditions where nonlinear pharmacokinetics occur with the drug of inter-
est. Generally, nonlinear pharmacokinetics occur due to changes in one or more of the drug’s absorption,
distribution, metabolism, or elimination characteristics in response to change in dose [13, 14] . In such a
case, the concept of superposition is violated, and thus the increase in drug bioavailability (See Bioavail-
ability and Bioequivalence) or concentrations resulting from an increase in dose is greater (or lesser)
than would be predicted based on the change in size of the dose (Figure 3). For instance, the steady-state
concentrations of fluoxetine are significantly higher and the half-life longer than would be predicted from
single-dose administration [15] . This occurs due to inhibition of fluoxetine metabolism by its metabo-
lite norfluoxetine, which also exhibits a long half-life. Thus, the product of metabolism (norfluoxetine)
inhibits parent drug metabolism, resulting in substantially higher drug concentrations upon multiple
dosing. Without chronic dosing models and measurement of toxicity, this type of phenomenon would be
missed and would not be reflected in the therapeutic index either.
c 2007 by John Wiley & Sons, Inc.
Copyright 5
Therapeutic Index

Figure 2. Effect of dose formulations on toxicity and effectiveness of a drug. Immediate release formulations
(IR) typically result in higher Cmax and lower Tmax values than do sustained release formulation (SR). As a result,
a given dose of a drug may exhibit a milder toxicity profile if given in a sustained release formulation than would
the same dose given in an immediate release formulation.

An additional classic example of nonlinearity is seen with phenytoin, which follows the behavior
of Michaelis-Menten kinetics (See Michaelis–Menten Model; Michaelis-Menten equation) with respect
to saturation of metabolism (see Figure 3). As the drug’s serum concentration rises, the catalytic site
in its metabolizing enzymes saturates with the drug molecule, and maximum catalytic velocity (Vmax )
is quickly reached. As a result, the drug’s metabolism reaches a plateau and causes its accumulation.
The normal therapeutic concentration range of phenytoin is 10–20 :g/mL; however, within the range
of doses that result in concentrations within this therapeutic range, an incremental increase in dose
results in a disproportionately greater increase in concentration, meaning that the difference between
subtherapeutic and toxic concentrations is traversed rapidly with a small change in dose [13, 14, 16] . Another
consequence of this saturable metabolism exhibited by phenytoin is the disproportionality between the
rate of administration and time to reach steady state, where the time to reach steady state increases
progressively with increasing rate of administration, again due to the saturation of metabolism [14] .
An additional factor potentially contributing to phenytoin’s nonlinear kinetics is its limited aqueous
solubility and subsequent absorption, where its peak serum concentrations are actually reduced at higher
oral doses. In one study, a 1600-mg single oral dose achieved a serum concentration peak of 10.7 mg/L,
but dividing it into four 400-mg doses every 3 hours achieved a peak level of 15.3 mg/L [17] .
As we can see, nonlinear kinetics can pose a serious hurdle in interpreting a drug’s therapeutic index.
One can imagine the level of complexity and unpredictability of behaviors of such drugs when considering
other factors such as interindividual variability, which can be as high as 50% for phenytoin [18] . Fortunately,
not many drugs behave like phenytoin, and a list of some of these drugs and the mechanism resulting in
the nonlinearity can be found in Table 2 [13, 15] .
6 c 2007 by John Wiley & Sons, Inc.
Copyright
Therapeutic Index

Figure 3. Nonlinear pharmacokinetics results from changes in one or more of a drug’s absorption, distribution,
metabolism, or elimination characteristics. This manifests in nonlinear dose-concentration curve, and may lead
to underestimation or overestimation of the target concentration in response to incremental increases in dose.
In the special case of metabolic saturation, nonlinearity is caused by reaching the maximum catalytic velocity
(Vmax ) at a given substrate concentration ([S]), as seen in the hyperbolic plot (inset).

4 Use of Biomarkers in the Determination of ED50

Death of the animal is the only biomarker for assessing LD50 . In the case of TD50 , the biomarker is simply
the toxicity observed and is most generally an extension of the drug’s pharmacologic effects, except in
the case of hypersensitivity reactions (see the previous discussion). However, determining biomarkers
for pharmacologic effect (i.e., ED50 ) can be a more difficult task. For example, if one was studying a
new thrombolytic molecule in mice, a decision must be made on whether the efficacy endpoint should
be time to complete lysis of the clot, time to partial lysis, degree of thrombus lysis, or even a surrogate
endpoint marker such as degree of vascular occlusion. Because other investigators studying the effects
of drugs of the same class might choose a different biomarker than previously used, the appropriateness
c 2007 by John Wiley & Sons, Inc.
Copyright 7
Therapeutic Index

Table 2. Mechanisms and examples of important drugs that exhibit nonlinear kinetics

Poor aqueous solubility


Phenytoin
Griseofulvin
Saturation of carrier-mediated absorption
Amino-$-lactam antibiotics
Levodopa
Saturation of first-pass metabolism
Verapamil
Fluorouracil
Propranolol
Hydralazine
Saturable plasma protein binding
Valproic acid
Disopyramide
Saturable metabolism
Ethanol
Phenytoin
Theophylline
Salicylate
Autoinhibition
Fluoxetine
Autoinduction
Carbamazepine
Phenytoin
Rifampicin

Source: Adapted from Ludden 1991 [13] and Otton et al. 1993 [15] .

of comparing therapeutic indices of any two drugs may be questionable. It is possible to use composite
or multiple biomarkers in the determination of ED50 , but this situation may be complex and difficult to
interpret.
An additional component of ED50 that merits consideration is whether one should consider either
dose or drug concentration as the correlate to effect. It is difficult, and perhaps often impossible, to
derive an exact measure of drug exposure at the target molecular site of drug action, but more accurate
assumptions will naturally produce more representative and potentially useful therapeutic indices. It may
be argued that the applied dose is a sufficient measure of exposure, where ED50 would translate into
the drug dose that achieves the desired effect in half the subjects studied. Accounting for the dose as a
surrogate endpoint of exposure obviates the need for drug concentration analyses and pharmacokinetic
parameter estimations (such as Cmax , Tmax or area under the curve). However, when only dose is used,
one is making the assumption that the drug concentrations achieved are identical for all individuals at
a given dose. This is obviously an oversimplification as we will discuss with respect to pharmacokinetic
variability [6] .
To further complicate matters, the effect of some drugs such as antibiotics (i.e., antibacterial action) is
concentration dependent (e.g., aminoglycosides and fluoroquinolones), whereas for other antibacterials
the effect is time dependent (e.g., penicillins and cephalosporins). Additionally, for some drugs such
as the serotonin reuptake inhibitors (SSRIs), the primary antidepressant effects occur through synaptic
remodeling, which can take weeks to months [1] . These differences illustrate that in some cases, the
temporal nature of the drug’s pharmacodynamics rather than pharmacokinetics dictate the time frame
of action.
8 c 2007 by John Wiley & Sons, Inc.
Copyright
Therapeutic Index

Table 3. Examples of narrow therapeutic index drugs affected by polymorphic genes

Drug Polymorphic genes

Cyclosporine CYP3A5, MDR1


Phenytoin CYP2 C9
Carbamazepine CYP3A4
Warfarin CYP2 C9
Thioguanine Thiopurine methyltransferase (TPMT)
Irinotecan UGT1A1
Tamoxifen CYP2D6

5 Interindividual Variability

There are multiple sources of interindividual, or between-subject, variability that can influence the appli-
cation of a therapeutic index to human drug therapy. Interindividual variability can be divided into two
general categories: pharmacokinetic (PK) and pharmacodynamic (PD). Because both can lead to differ-
ences in efficacy or toxicity of a drug between different individuals, the estimation of a drug’s therapeutic
index will have a certain variability associated with it, and this variability may be substantial. Furthermore,
application of a therapeutic index in making dosing decisions becomes more complex because individuals
respond differently to a given drug; thus, the dose that is efficacious to one person may be toxic to another.
Pharmacokinetic variability can affect all aspects of drug exposure, including bioavailability, the con-
centration time profile, and area under the curve (AUC) of a drug. Pharmacodynamic variability manifests
as interindividual differences in drug effect as measured either directly or with surrogate endpoints. An
increasingly recognized source of PK and PD variability is genetic polymorphisms, which are typically
single nucleotide polymorphisms (SNPs) that exist in our genes and result in differential protein expres-
sion. The PK variability can be influenced by genetics in multiple ways, such as variability in certain
efflux or influx transporter proteins that play a role in transporting a drug molecule within the body or
as variability in drug metabolizing enzymes. For example, a mutation in the multiple drug resistance
gene (MDR-1) encoding the P-glycoprotein (P-gp) transporter has been shown to affect cyclosporine
bioavailability in renal transplant patients [19, 20] and to be predictive of cyclosporine target concentration
in liver transplant patients (See Transplantation) [21] . With respect to pharmacodynamic genetic variation,
mutations in $-adrenergic receptors have been associated with differential response to $-blocker therapy
[22, 23]
as well as other receptors.
Disease and environmental factors may also result in PK and PD variability, such as the degree of
renal or hepatic function present in an individual, concomitant drug therapy (drug interactions), dietary
factors, and so on. As expected, variability may be more clinically relevant in drugs considered to exhibit
a narrow therapeutic index, such as cyclosporine, as compared with drugs that have a wider therapeutic
index such as benzodiazepines. Table 3 lists some polymorphic genes with potential effects on drugs
considered to possess a narrow therapeutic index.
The subject of interindividual variability with respect to narrow therapeutic index drugs has recently
become a contentious one [3] . At issue is whether FDA bioequivalence criteria are sufficient for minimizing
any significant differences in bioavailability and concentration–time profile in narrow therapeutic index
drugs when switching patients from brand-name to generic-name drugs [3, 5] . Several proposals have
recently been put forth by state boards of pharmacy and state legislatures to tighten the regulations
on “generic switching” with respect to narrow therapeutic index drugs [3] . From a pharmacokinetic
standpoint, the FDA currently requires that the 90% confidence interval of a test drug’s rate and extent
of absorption (AUC and Cmax , respectively) be within 80% to 125% of the reference drug in order
for them to be considered bioequivalent, regardless of the drug class or group [4] . In response, the FDA
c 2007 by John Wiley & Sons, Inc.
Copyright 9
Therapeutic Index

reaffirmed its current position and referred to its commissioned taskforce from 1986 that was appointed to
investigate very similar issues at the time. That taskforce subsequently conducted two studies with a drug
considered to possess a narrow therapeutic index, carbamazepine, which demonstrated no differences in
bioequivalence, efficacy, or safety between generic and innovator products [3, 5] .

6 Intraindividual Variability

Another issue of importance in therapeutic index determinations is within-subject variability, or intraindi-


vidual variability. When applying a therapeutic index to any population, one must keep in mind that each
individual in the population has several intrinsic and extrinsic elements of variability that may lead him
or her to experience differences in efficacy and/or toxicity on different occasions of drug administration.
Certainly, when one first thinks of intraindividual variability, the issues of physiologic factors as
potential sources come to mind. For example, differences in gastric pH between dose administrations,
fasted versus fed state, food complexation, differences in gastric emptying rate across days, and other
seemingly subtle but important changes can affect the absorption, distribution, metabolism, or excretion
of a drug and play a role in intraindividual variability [24] .
Another element of intraindividual variability to consider is that a given drug can often be used to
treat several conditions that may require different doses and target concentrations. For example, the
recommended dose range for aspirin when used to prevent recurrent myocardial infarction is 75–325 mg
per day, whereas that for treating rheumatoid arthritis is 3 grams per day to be increased until effect is
achieved [25] . Although platelets are anucleated and exquisitely sensitive to inhibition by aspirin and only
require a low dose to achieve saturation of cyclo-oxygenase-1 (COX-1), monocytes are nucleated, exhibit a
high turnover rate of cyclooxygenases, and thus require higher doses of aspirin to suppress inflammation
and exhibit a dose-related anti-inflammatory effect [1, 26] . Incidentally, gastrointestinal bleeding is more
likely at higher doses of aspirin due to the dose-related suppression of cyclooxygenases in nucleated
enterocytes; therefore, one would expect two different toxicity profiles for aspirin treatment of rheumatoid
arthritis and prevention of myocardial infarction.
The example of aspirin is classic in terms of the different mechanisms of actions of efficacy and toxicity
obtained with different doses. However, aspirin will most likely be assigned a single “therapeutic index,”
and one is left with the task of determining how to apply it to a population given the amount of variability
in efficacy and toxicity that could result from its use for different indications. The question then becomes
whether different indications for a given drug must have different therapeutic indexes based on the dose
range and the specified toxicity endpoint. It is apparent that this can increase the complexity of application
of a therapeutic index, particularly as not all indications and dose ranges are studied in preclinical or early
clinical testing.

7 Variability in Narrow Therapeutic Index Drugs

By definition, narrow therapeutic index drugs show little intraindividual variability—typically less than
30% analysis of variance coefficient of variation (ANOVA-CV). Otherwise, it becomes difficult to approve
them for human use because patients will suffer continuous cycles of toxic and subtherapeutic drug
concentrations, and the drug would fail in the advanced phases of clinical trials [6] . Table 4 lists some
commonly considered narrow therapeutic index drugs and their estimated intersubject and intrasubject
variability [18] . Note that for all the listed drugs, the intrasubject variability remains below 30%, and less
than the intersubject variability. This means that, for these drugs, practitioners will likely find it hard at
10 c 2007 by John Wiley & Sons, Inc.
Copyright
Therapeutic Index

Table 4. Intra- and intersubject variability for some commonly-considered narrow therapeutic index drugs

Coefficient of variation %

Drug Intersubject Intrasubject

Carbamazepine 38
Conjugated estrogens 42 14–15
Digoxin 52
Levothyroxine 20 <20
Phenytoin 51 10–15
Theophylline (SR) 31 11–14
Warfarin 53 6–11

Source: Adapted from Benet 1999 [18] .

first to find the appropriate dosage regimen for a given patient, but once that regimen is arrived at there
should be little need to readjust it in the future.

8 Summary

The ultimate question is how to apply a therapeutic index clinically. As we have seen, the concept of
therapeutic index is riddled with inconsistencies, and in some cases its use may be tenuous in clinical
practice. The FDA does not endorse the use of therapeutic index, and there is no mention of it in their
publications, Approved Drug Products with Therapeutic Equivalence Evaluations [The Orange Book] [4] . There
is also no apparent FDA guidance on use of therapeutic index for clinical trials and new drug applications
(NDA). However, the therapeutic index may prove useful in the preliminary phases of clinical trials
when it is important to rule out any major toxicity associated with a given drug, and before proceeding
any further with expensive and exhaustive clinical trials. Nevertheless, the ethical issues of experimental
animal testing must be considered .
In the end, therapeutic index provides a semiquantitative estimate of the range of doses (or concen-
trations) that may be administered to a patient (or study subject) to achieve a desired response with an
acceptable level of adverse effects. In addition, a therapeutic index may be useful in choosing among
drugs with similar therapeutic indications.

9 Related Articles
Preclinical treatment evaluation
Toxicity (Adverse Events)
Development toxicity study
Median effective dose
Drug Development
Bioavailability and Bioequivalence
Pharmacokinetics and Pharmacodynamics
herapeutic Equivalence
Therapeutic Dose Rang
Minimum therapeutically effective dose
c 2007 by John Wiley & Sons, Inc.
Copyright 11
Therapeutic Index

References
[1] L. L. Brunton, ed. Goodman & Gilman’s The Pharmacological Basis of Therapeutics, 11th ed. Boston: McGraw-Hill, 2005,
pp. 126–129.
[2] G. Levy, What are narrow therapeutic index drugs? Clin Pharmacol Ther. 1998; 63: 501–505.
[3] FDA position on product selection for ‘narrow therapeutic index’ drugs. Am J Health Syst Pharm. 1997; 54: 1630–1632.
[4] Center for Drug Evaluation and Research, U.S. Food and Drug Administration. Approved Drug Products with Therapeutic
Equivalence Evaluations [Orange Book], 27th ed. Updated: October 10, 2007. Available at: http://www.fda.gov/cder/
orange/default.htm
[5] FDA comments on activities in states concerning narrow-therapeutic-index drugs. Am J Health Syst Pharm. 1998; 55:
686–687.
[6] L. Z. Benet and J. E. Goyan, Bioequivalence and narrow therapeutic index drugs. Pharmacotherapy. 1995; 15: 433–440.
[7] J. X. Zhou, N. F. Luo, X. M. Liang, and J. Liu, The efficacy and safety of intravenous emulsified isoflurane in rats. Anesth
Analg. 2006; 102: 129–134.
[8] G. E. Schumacher, ed. Therapeutic Drug Monitoring. Boston: Appleton & Lange, 1995, pp. 8–9.
[9] K. A. Antons, C. D. Williams, S. K. Baker, and P. S. Phillips, Clinical perspectives of statin-induced rhabdomyolysis. Am
J Med. 2006; 119: 400–409.
[10] W. Levinson, Medical Microbiology & Immunology: Examination & Board Review. 8th ed. Boston: McGraw-Hill, 2004, pp.
445–452.
[11] H. T. Stelfox, S. B. Ahmed, J. Fiskio, and D. W. Bates, Monitoring amiodarone’s toxicities: recommendations, evidence,
and clinical practice. Clin Pharmacol Ther. 2004; 75: 110–122.
[12] H. P. Rang, M. M. Dale, and J. M. Ritter, Pharmacology, 4th ed. Edinburgh: Churchill Livingstone, 2000, pp. 57–60.
[13] T. M. Ludden, Nonlinear pharmacokinetics: clinical Implications. Clin Pharmacokinet. 1991; 20: 429–446.
[14] R. Malcolm and T. N. Tozer, Clinical Pharmacokinetics: Concepts and Applications, 3rd ed. Philadelphia: Lippincott Williams
& Wilkins, 1995, pp. 406–411.
[15] S. V. Otton, D. Wu, R. T. Joffe, S. W. Cheung, and E. M. Sellers, Inhibition by fluoxetine of cytochrome P450 2D6
activity. Clin Pharmacol Ther. 1993; 53: 401–409.
[16] T. M. Ludden, S. R. Allerheiligen, T. R. Browne, and J. R. Koup, Sensitivity analysis of the effect of bioavailability or
dosage form content on mean steady state phenytoin concentration. Ther Drug Monit. 1991; 13: 120–125.
[17] D. Jung, J. R. Powell, P. Walson, and D. Perrier, Effect of dose on phenytoin absorption. Clin Pharmacol Ther. 1980; 28:
479–485.
[18] L. Z. Benet, Relevance of pharmacokinetics in narrow therapeutic index drugs. Transplant Proc. 1999; 31: 1642–1644;
discussion 1675–1684.
[19] R. H. Ho and R. B. Kim, Transporters and drug therapy: implications for drug disposition and disease. Clin Pharmacol
Ther. 2005; 78: 260–277.
[20] C. R. Yates, W. Zhang, P. Song, S. Li, A. O. Gaber, et al., The effect of CYP3A5 and MDR1 polymorphic expression on
cyclosporine oral disposition in renal transplant patients. J Clin Pharmacol. 2003; 43: 555–564.
[21] L. Bonhomme, A. Devocelle, F. Saliba, S. Chatled, J. Maccario, et al., MDR-1C3435T polymorphism influences
cyclosporine a dose requirement in liver-transplant recipients. Transplantation. 2004; 78: 21–25.
[22] D. E. Lanfear, P. G. Jones, S. Marsh, S. Cresci, H. L. McLeod, and J. A. Spertus, Beta2-adrenergic receptor genotype and
survival among patients receiving beta-blocker therapy after an acute coronary syndrome. JAMA. 2005; 294: 1526–1533.
[23] J. A. Johnson, I. Zineh, B. J. Puckett, S. P. McGorray, H. N. Yarandi, and D. F. Pauly, Beta 1-adrenergic receptor
polymorphisms and antihypertensive response to metoprolol. Clin Pharmacol Ther. 2003; 74: 44–52.
[24] V. P. Shah, A. Yacobi, W. H. Barr, L. Z. Benet,. D. Breimer, et al., Evaluation of orally administered highly variable drugs
and drug formulations. Pharm Res. 1996; 13: 1590–1594.
[25] PDR. Physicians’ Desk Reference, 57th Edition. 2003.
[26] C. Patrono, Aspirin: new cardiovascular uses for an old drug. Am J Med. 2001; 110: 62S–65S.

Further Reading

M. Rowland, L. B. Sheiner, J. L. Steimer, and A. G. Sandoz, Variability in Drug Therapy: Description, Estimation, and Control, A
Sandoz Workshop. New York: Raven Press, 1985.

12 c 2007 by John Wiley & Sons, Inc.


Copyright

You might also like