Pharmacologic Principles for Combination Therapy

Myron L. Toews and David B. Bylund

Department of Pharmacology, University of Nebraska Medical Center, Omaha, Nebraska

This article discusses the pharmacologic basis for understanding the drug does to the body” and pharmacokinetics as the study
the therapeutic actions of drugs, particularly for their use in combi- of “what the body does to the drug.” Both pharmacokinetics
nations. The focus is on principles underlying combination therapy and pharmacodynamics provide important targets and rationales
in general, including examples from diseases other than chronic for the use of combination therapy.
obstructive pulmonary disease (COPD). Pharmacodynamic aspects
of drug action are covered, with an emphasis on recent advances
PHARMACODYNAMICS: DRUG ACTIONS AT THEIR
in the understanding of drug–receptor interactions and of drug
TARGET SITES
agonism. Pharmacokinetics and drug-induced adaptive changes in
receptors and cell signaling pathways are summarized, emphasizing Basic Aspects of Drug Action
their importance for potential combination therapies aimed at pro-
Drug binding to receptors. The effects of nearly all drugs are
longing drug action. An organizational framework for three differ-
ent approaches to combination therapy is then proposed; the mo-
mediated by interactions with specific receptors, either the classi-
lecular rationales for each approach are described together with cally defined receptors for hormones, neurotransmitters, and
classic examples from other diseases, and then their application to growth factors, or less classical drug receptors that nonetheless
combination therapy in COPD is discussed. Finally, terminology for obey the laws of drug-receptor interactions (2, 3). These drug-
the independent and interactive effects of drug combinations is receptor interactions can generally be described by the equation
discussed, and approaches to the quantitative analysis and visual [RD] ⫽ [RT] ⫻ [D] / (KD ⫹ [D]), where [RD] is the concentration
display of the effects of drug combinations are introduced. The basic of receptor-drug complex, [RT] is the total receptor concentra-
principles reviewed here provide the pharmacologic foundation for tion, [D] is the concentration of free drug (not bound to recep-
subsequent articles in this issue that address the combinations in tor), and KD is the equilibrium dissociation constant for the
current use for COPD, and they point to novel strategies for poten- binding of free drug D to free receptor R to form the RD
tial future approaches to combination therapy in COPD. complex, defined by the reaction R ⫹ D ← → RD. The analogous
form of the equation used in radioligand binding assays may be
Keywords: desensitization; dose–response curves; drug–receptor inter-
more familiar: B ⫽ Bmax ⫻ F / (KD ⫹ F), where B is the concentra-
actions; pharmacodynamics; pharmacokinetics; synergism
tion of bound ligand, Bmax is the total number of binding sites,
A firm foundation in the basic pharmacologic principles that F is the free drug concentration, and KD is again the equilibrium
govern the actions of all drugs is essential to understanding dissociation constant for the binding reaction. Both of these
the effects of combination drug therapy for chronic obstructive equations yield the classical hyperbolic dose–response curve
pulmonary disease (COPD). The specific actions of the drugs when data are plotted versus the free drug concentration and
used in combination therapy must be understood first, but their the classical sigmoid dose–response curve when plotted versus
individual actions are subject to modification by the presence the log of the free drug concentration. In either plot, the KD
of a second drug. Unique new actions may occur only with value is the concentration of drug that gives 50% of the maximal
combination therapy and not with either drug given alone, add- drug–receptor complex. KD is a measure of the affinity of the
ing yet another layer of complexity. The most important basic drug for the receptor (or of the receptor for the drug), with
principles of pharmacology that are essential for understanding smaller KD values representing higher affinities and larger KD
combination therapy are reviewed here, together with a new values representing lower affinities. Both types of plots reveal
organizational framework for the diverse rationales for using the saturability of the reaction that arises from the finite number
drug combinations and a brief overview of terminology and of receptors available for interacting with the drug.
analytical methods essential for understanding the individual Receptor-mediated drug effects. The end points of interest for
and interactive effects of drugs used in combination therapy. treating COPD or other diseases are the cellular or clinical effects
The field of pharmacology is divided into two broad areas: that result from formation of a drug–receptor complex, not the
pharmacokinetics, which deals with safely getting the right drug–receptor interaction itself. In the simplest case, the effect
amount of active drug to the right location for the right amount of a drug binding to its receptor is directly proportional to the
of time; and pharmacodynamics, which deals with understanding concentration of the RD complex, and the equation for the
the effects of the drug at its site of action (1–3). Alternatively, drug’s effect is essentially identical to that for its binding to the re-
pharmacodynamics has been described as the study of “what ceptor. In mathematical terms, E ⫽ Emax ⫻ D / (Ka ⫹ D), where
E is the effect at a given drug concentration D, Emax is the
maximal effect when all of the receptors are occupied by active
drug, and Ka is the concentration of drug that gives half of this
maximal response and defines the drug’s potency. In this simple
(Received in original form April 14, 2005; accepted in final form July 6, 2005) case, the drug’s potency for inducing a response is identical to
This study was supported by The University of Nebraska Medical Center, Omaha, its affinity for binding to the receptor; however, this is not usually
Nebraska. the case.
Correspondence and requests for reprints should be addressed to Myron L. Toews, Agonism and antagonism. Drugs that increase receptor activa-
Ph.D., Professor, Department of Pharmacology, University of Nebraska Medical tion when they form the RD complex are agonists. Drugs that
Center, 985800 Nebraska Medical Center, Omaha, NE 68198–5800. E-mail:
bind but do not alter receptor activation are antagonists. Al-
[email protected]
though antagonists do not alter the activation state of the recep-
Proc Am Thorac Soc Vol 2. pp 282–289, 2005
DOI: 10.1513/pats.200504-037SR tor, their binding nonetheless can result in effects that are clini-
Internet address: www.atsjournals.org cally important, because their occupancy of the receptor’s
Toews and Bylund: Pharmacology of Combination Therapy 283

binding site can prevent endogenous agonists from increasing activating the receptor, but rather they were inverse agonists,
receptor activation. A third group of drugs bind and activate drugs capable of altering receptors and signaling pathways in
the receptor in a manner similar to agonists but fail to cause the the opposite direction from classical agonists. Current under-
maximal response, even when sufficient drug is present to occupy standing of inverse agonists is that they bind preferentially to
all of the receptors. These drugs are partial agonists, and the the inactive R form of the receptor and stabilize it in the same
extent to which they increase receptor activation is termed their way that agonists preferentially bind and stabilize the R* form.
“efficacy.” The efficacy of partial agonists is defined in relation Furthermore, just as there are partial agonists that cause less
to the maximal response possible or the response to the historical than full activation of receptors even with full receptor occu-
reference drug for the receptor, that drug being called a full pancy, there are also partial inverse agonists that do not fully
agonist to distinguish it from partial agonists. Full agonists have prevent the formation of R* even with full receptor occupancy.
an efficacy of 1; antagonists have an efficacy of 0; and partial The current explanation for these partial effects is that these
agonists have efficacy values greater than 0 but less than 1. Partial drugs can bind and stabilize both the R and R* conformations
agonists with low efficacy frequently behave as antagonists in to various extent. Full agonists primarily bind and stabilize the
clinical use. For example, some ␤-adrenergic antagonists, such R* state, full inverse agonists primarily bind and stabilize the
as pindolol, are partial agonists and are sometimes described R state, and the entire spectrum between these extremes is
as having intrinsic sympathomimetic activity. It is important to occupied by partial agonists, antagonists, and partial inverse
emphasize that potency is a measure of the amount of drug agonists. These effects are quantified as the efficacy of the drug,
required to generate a given response, whereas efficacy is a with values ranging from 1 for a full agonist to ⫺1 for a full
measure of the magnitude of the response that is generated by inverse agonist. In current models, the only true antagonists are
the drug; these parameters are largely independent of each other. compounds with an intrinsic efficacy of exactly zero. Most drugs
Advanced Concepts in Drug Action and Receptor Theory used as antagonists are likely to exhibit at least some partial
agonist or partial inverse agonist activity; however, in clinical
Complexities in drug–receptor interactions and responses. The sim- practice, many of these may not be significantly different from
plifying assumption that a drug binds to a single state of its classically defined antagonists. For COPD therapy, it is impor-
receptor to induce a single response that is directly proportional
tant to consider the possibilities that some of the pathology may
to its binding to the receptor is occasionally valid in the labora-
arise from receptor constitutive activity, that some of the effects
tory. It is especially valid when a specific receptor proximal effect
of existing drugs could be mediated by inverse agonism rather
is the end point being measured, such as G protein activation
than simple antagonism, and that new drugs with inverse agonist
for G protein–coupled receptors or transcriptional activation of
activity might provide novel therapeutic benefit.
a specific gene for steroid receptors. It is now known, however,
Multiple active conformations of receptors. A related and al-
that receptors can have multiple active states that can send
most certainly equally important concept is that many receptors
multiple signals, making receptor activation more complex. Also,
can exhibit multiple active conformations, each of which can
in clinical studies and in medical practice, the endpoints being
measured are generally many steps distal to the initial drug– interact with different effectors to modify the activity of different
receptor binding interaction, through complex signal transduc- downstream signaling targets to cause different effects. Drugs
tion cascades and response mechanisms. In these cases, the final can bind to and stabilize these various active conformations
effect of the drug is still determined by its binding to its receptor, selectively, making it possible to design drugs to target only a
but both the drug’s cellular or clinical potency and its cellular subset of the effects mediated by that receptor. Perhaps the best-
or clinical efficacy exhibit a much more complex relationship to characterized examples of this concept are the selective estrogen
the fraction of receptors occupied by the active drug. Several of response modifiers, such as tamoxifen and raloxifene, which bind
the key factors that contribute to differences between simple to estrogen receptors and act as agonists for some of the effects
receptor occupancy and actual receptor-mediated drug re- of estrogen but as antagonists for other estrogen receptor effects.
sponses are highlighted here, together with the opportunities Different estrogen response elements on DNA are preferentially
they may present for new approaches to COPD combination activated or inactivated by these different estrogen receptor
therapy. conformations, allowing a different set of genes to be expressed
Constitutive activity and inverse agonism. It is now firmly estab- in response to each of these drugs, even though they all use the
lished that receptors can exhibit constitutive activity and activate same receptor (7, 8). It seems likely that glucocorticoid receptors
cellular responses even in the absence of an activating ligand. and response elements exhibit similar phenomena and that selec-
Although these phenomena were first observed for mutated or tive glucocorticoid response modifiers could become as impor-
highly overexpressed receptors in isolated cell systems, there is tant as those for estrogen receptors, with a likely beneficial
evidence to support their relevance in more physiologic systems increase in specificity of clinical effects as a result.
and in some pathologies (4–6). This discovery has led in large Similar multiple active conformational states have been dem-
part to the current concept that receptors can exist in an inactive onstrated for G protein–coupled receptors, with different confor-
conformation (usually represented by R) or an active conforma- mations activating different G proteins and different down-
tion (R*) and that agonist ligands stabilize the R* conformation stream signaling pathways. In the case of G protein–coupled
by binding to it. The extent to which agonists may also bind receptors, this concept is sometimes referred to as “signal traf-
to the inactive R form of the receptor and actively drive the ficking,” with different ligands directing receptor responses down
conversion to the active R* form remains to be established. In different intracellular signaling pathways (9, 10); “conformation-
either case, the effect of agonist binding is to increase the fraction selective agonism” is proposed here as a more appropriate term.
of receptors in the active R* form and to increase receptor- These concepts provide new opportunities for monotherapy with
mediated signaling. novel conformation-selective drugs. An additional exciting possi-
The identification of constitutively active receptors quickly bility is that combination therapies could target the interconver-
led to the realization that many drugs previously considered to sions among these various active states, with one drug serving
be antagonists could actually decrease the constitutive activity as the activator and the other drug controlling the likelihood of
of these receptors. They were not simply antagonists with no the receptor being in the desired conformation for that drug to
effect of their own except to prevent agonists from binding and bind and stabilize.
284 PROCEEDINGS OF THE AMERICAN THORACIC SOCIETY VOL 2 2005

Receptor expression levels and receptor reserve. Because of the Adaptive Changes in Receptors and Signaling Pathways as
signal amplification that can occur at many steps in the signal Targets for Combination Therapy
transduction pathways activated by receptors, and because fac- Drug receptors and their associated signaling pathways are not
tors other than the concentration of drug-receptor complex may static entities, but rather are subject to adaptive changes in many
be limiting for the final response, the maximal cellular or clinical of their properties. With constant or repeated exposure to ago-
response is often obtained with concentrations of drug far below nist drugs, receptors can undergo a desensitization in their ability
those required to achieve maximal receptor occupancy. This to bind their ligands or to initiate signals in response to ligand
results in phenomena referred to as “spare receptors” (so termed binding. Multiple mechanisms are often involved, and the time
because the cell has more receptors than needed for a maximal courses for onset and reversal of each component of the overall
response) or “receptor reserve” (because the cell can lose some changes in responsiveness can differ. Covalent modification of
of its receptors and still maintain full responsiveness). Receptor the receptors or other signal pathway molecules is often involved
reserve shifts the dose–response curve for drug effects to the and typically occurs rapidly. With somewhat longer treatment,
left, moving it away from the curve for drug binding to the many receptors undergo internalization into endocytotic vesicles
receptor. The concentration of drug that gives half-maximal re- or redistribution to other cellular compartments, such as move-
sponse in this context is usually referred to as the EC50 (effective ment into or out of caveolae or lipid rafts, where their accessibil-
concentration for a 50% response), because it is a complex factor ity to ligand or their ability to propagate signals may be different.
affected by many variables rather than a true constant. The ratio With even longer exposure, there is often a decrease in the
of the EC50 to the KD for binding is often used as a quantitative total number of receptors because of enhanced degradation,
indicator of the extent of receptor reserve. The key outcome of decreased synthesis, or both. This decrease in receptor number
these relationships is that the cellular or clinical effectiveness of is referred to as downregulation, although it also can contribute
a drug changes in a complex way as the number of receptors to functional desensitization. Much recent work has focused on
changes (e.g., because of disease states or drug-induced up-regu- the multiple mechanisms for adaptive regulation of G protein–
lation or down-regulation of receptors or their coupling to down- coupled receptors, with the ␤2-adrenergic receptor that is the
stream signaling pathways). In general, an increase in the number target of several COPD drugs being the best characterized (11).
Many other types of receptors are subject to similar adaptive
of receptors increases the maximal response in the absence of
changes with chronic activation, although the specific molecular
a receptor reserve but shifts the dose-response curve to the left
details are different. Conversely, drugs that block or decrease
(higher potency, lower EC50 value) in the presence of a receptor
receptor activity are likely to induce receptor upregulation or
reserve (i.e., the original receptor number was already sufficient
increased signal capacity, allowing homeostatic maintenance of
to generate a maximal response). Conversely, a decrease in the the proper level of sensitivity and responsiveness.
number of receptors leads to a rightward shift in the dose-response These adaptive changes provide novel possibilities for combi-
curve (decreased potency, higher EC50 value) if there are spare nation therapies. For example, a second drug could be used to
receptors; however, as the number of receptors becomes the decrease desensitization of the response to the first drug and
limiting factor for the downstream effect, the maximal response prolong the first drug’s actions. Alternatively, a second drug
also begins to decrease. This concept provides possibilities for might be used intentionally to induce a tissue-selective desensiti-
combination therapy, with a second drug used to increase or zation or downregulation of the first drug’s receptors, enabling
decrease the expression of the receptors that are the target for the first drug to have more selective actions. Much effort has
the first drug. These concepts are also relevant to the adaptive been placed on advancing the understanding of the mechanisms
upregulation and down-regulation of receptor expression that of receptor desensitization, and this work may someday lead to
can occur in response to chronic drug exposure, discussed further combination therapies with one drug to attain the desired re-
later. sponse and a second drug to help maintain that response.

PHARMACOKINETICS: DRUG DELIVERY AND RATIONALES FOR COMBINATION THERAPY

DURATION OF ACTION Clinical Rationales for Combination Therapy
Pharmacokinetic Principles for Combination Therapy From a clinical therapeutics perspective, there are perhaps only
two broad rationales for using drug combinations. The first and
Pharmacokinetics is generally divided into four components: (1 )
most obvious is to obtain a greater therapeutic effect with the
absorption, (2 ) distribution, (3 ) metabolism, and (4 ) elimination
combination than can be achieved with either drug alone. The
(1). Before a drug can reach its target cells and receptors to
second is to obtain the same therapeutic effect as could be
mediate its effects, it must be effectively absorbed and then
obtained with only one of the two drugs, but with fewer deleteri-
distributed to the desired site of action. The second drug for ous side effects or dose-limiting toxicities. Presumably, an ideal
combination therapy could be a drug that improves the rate of combination therapy would accomplish both of these goals (12).
absorption, extent of absorption, or the distribution of a first In contrast to these two simple clinical rationales, there is a
drug already known to be effective in causing the desired end much larger range of pharmacologically based reasons for using
point. A second drug might be used to target the effects of the a combination of two drugs rather than a single drug. These
first drug to the desired site of action, without the second drug two-drug combinations target diverse mechanisms related to all
having any actions of its own at that site. Metabolism and elimi- aspects of drug action, as discussed in more detail in the following
nation of a drug can decrease either the amount or duration of sections.
action of active drug, and a second drug for combination therapy
could be a drug that decreases either the metabolism or elimina- Mechanistic Rationales for Drug Combinations:
tion of the active first drug. In these cases, the second drug has An Organizational Framework
no direct effect of its own on the desired end point but only acts Overview. For the discussion below, the pharmacologic rationales
by increasing the effective concentration of the active first drug for combination therapy are divided into three general ap-
at its therapeutic target site. proaches, referred to as classes. It should be stressed, however,
Toews and Bylund: Pharmacology of Combination Therapy 285

that this is not a rigorous classification scheme, because many are many different approaches within this class of combinations,
drug combinations have multiple rationales and mechanisms of as illustrated with typical examples next.
interaction and do not fall cleanly into one single class or another. An example of using two drugs to obtain a greater response
This organizational framework may be useful as a guide for than with either drug alone, where the two drugs have similar
understanding current combinations, however, and as a stimulus targets and actions, is in Parkinson’s disease. Levodopa is a
for encouraging new approaches using novel combination strate- prodrug precursor that is converted to dopamine, and it is now
gies. used together with bromocriptine, a direct-acting dopamine re-
Class 1 combinations: rationales and examples. Class 1 combi- ceptor agonist targeting the same receptor. The antibiotic cotri-
nations include two (or more) drugs that each target different moxazole uses two drugs acting at different points in a single
aspects of the disease. This is perhaps the most obvious rationale pathway to achieve greater inhibition than either drug alone. It
for using drug combinations, because many diseases are known is a combination of sulfamethoxazole, which blocks folic acid
to be multicomponent or multifactorial. It is in this class of synthesis by inhibiting dihydropteroate synthetase, and trimeth-
combinations where the actions of Drug A and Drug B are oprim, which acts at a later step in nucleotide synthesis to inhibit
most likely to be independent, at least in their specific actions; dihydrofolate reductase.
however, they might have interactive effects on the overall im- Cancer chemotherapy provides multiple examples of the use
provement in patient status. Hypertension is a multicomponent of drug combinations to achieve the desired therapeutic end
disease treated with class 1 combinations, including drugs acting point while avoiding dose-limiting toxicities of the individual
on completely different tissues or organs. Treatment of hyperten- agents, each of which is theoretically sufficiently cytotoxic to kill
sion is likely to include some combination of diuretics acting on cancer cells on its own. One such regimen combines cisplatin,
the kidney to decrease blood volume; vasodilators acting directly dosage of which is limited by nephrotoxicity; etoposide or vin-
on vascular smooth muscle cells; ␣2-adrenergic agonists acting blastine, each of which are dose-limited by bone marrow suppres-
centrally to reduce sympathetic activation; angiotensin-con- sion; and bleomycin, which is dose-limited by pulmonary toxicity.
verting enzyme inhibitors acting on the endothelium to decrease In this case, the individual drugs also have different mechanisms
angiotensin levels; and ␤-adrenergic antagonists acting on the of cancer cell cytotoxicity, attacking the disease by multiple
heart to decrease myocardial contractility and on the kidney to pathways while also avoiding damage to nonmalignant cells.
inhibit renin release. A variant of class 2 combination therapy is to use a second
Class 1 combinations in COPD therapy. Class 1 combinations drug to decrease a single, specific, unwanted effect of the first
are a common approach to COPD therapy, because COPD is drug. In prostate cancer, gonadotropin-releasing hormone ago-
widely accepted as a multicomponent disease. A case can be nists, such as leuprolide, are used chronically to effect down-
made that more such combinations for COPD therapy are regulation of gonadotropin-releasing hormone receptors, inhib-
needed. The most obvious class 1 combination in COPD therapy iting stimulation of testosterone synthesis. The androgen receptor
is the use of ␤2-agonists plus corticosteroids. The predominant antagonist flutamide is used in combination early in this therapy
therapeutic effect of ␤2-agonists is bronchodilation by activation to block the actions of the increased testosterone that occurs
of ␤2-receptors on the smooth muscle cells. In contrast, the pre- initially with leuprolide, in the period before the gonadotropin-
dominant therapeutic effect of corticosteroids is to decrease releasing hormone receptors become downregulated. Another
inflammation, primarily by inhibiting the actions of inflammatory example is the use of a potassium-sparing diuretic such as spiro-
cells and inflammatory mediators. The use of muscarinic antago- nolactone, together with thiazides, to decrease the excess loss
nists together with corticosteroids is a similar example of a class of potassium that can occur with thiazides alone.
1 combination, with muscarinic antagonists acting to decrease Another class 2 approach is to use a second drug to decrease
bronchoconstriction. Other aspects of these combinations are the workload for the first drug, making it more effective. In type
that each includes one agent primarily targeting acute aspects 2 diabetes, the ␣-glucosidase inhibitor acarbose lowers glucose
of the disease (␤2-agonists or muscarinic antagonists) and one absorption from the gastrointestinal tract, decreasing the level
agent primarily targeting longer-term elements of disease pro- of plasma glucose that other drugs, such as the sulfonylurea
gression (corticosteroids), and that each contains one agent for glipizide or injected insulin, must metabolize or store. Similarly
symptomatic relief (the bronchodilators) and one that is perhaps in hypertension, angiotensin-converting enzyme inhibitors, such
truly disease-modifying (antiinflammatory corticosteroids). Ad- as captopril, decrease the concentration of angiotensin II, so
ditional mechanisms and target cell types are also involved in that angiotensin receptor antagonists, such as losartan, have less
the effects of all three of these classes of drugs, as highlighted angiotensin II to counteract.
elsewhere in this issue. There is a need for additional drugs to Class 2 combinations in COPD. The use of bronchodilators
target other aspects of COPD, most likely to be used in further in combination is an example of class 2 combinations for treating
combinations with the previously mentioned drugs. Aspects of COPD. ␤2-Agonists, muscarinic antagonists, and theophylline
COPD that are not currently targeted effectively are mucus all target the bronchoconstriction component of COPD, act on
production or secretion and cough, and the airway structural airway smooth muscle cells, and elevate cyclic adenosine mono-
remodeling that occurs with disease progression and remains phosphate as a key component of their mechanisms. ␤-Agonists
essentially irreversible. and muscarinic antagonists both target the G protein–mediated
Class 2 combinations: rationales and examples. Class 2 includes input to the enzyme adenylyl cyclase, but ␤-agonists act to in-
many combinations of drugs that target a single disease compo- crease stimulatory input by Gs, whereas muscarinic antagonists
nent and often a single cell type or even a single response path- act to decrease the inhibitory input by Gi. These two classes of
way in that cell type, but with different sites of action. This drugs use different receptors and G proteins but share the same
specific targeting allows greater effects or reduced drug doses pathway from adenylyl cyclase to bronchodilation.
and reduced toxicities. In some of these cases, a single drug may Combining theophylline with ␤-agonists can be viewed as
be safe but insufficiently effective on its own. In other cases, using theophylline to decrease the workload for ␤-agonists, be-
drugs that are sufficiently powerful on their own may be avail- cause theophylline prevents the breakdown of the cyclic adeno-
able, but side effects and toxicity may prevent their use at fully sine monophosphate that is formed in response to ␤-agonist
effective doses. In either case, combining two effective drugs stimulation. From an alternate perspective, the use of other
achieves the desired end point while avoiding toxicity. There bronchodilators with theophylline serves as an approach to low-
286 PROCEEDINGS OF THE AMERICAN THORACIC SOCIETY VOL 2 2005

ering the concentration of theophylline needed, thereby avoiding mechanisms for pharmacokinetic enhancement of its actions.
the significant side effects that occur with higher doses of theoph- This view of salmeterol may provide a useful model for potential
ylline. Although their shared ability to modulate cyclic adenosine new agents composed of a class 3 combination of active princi-
monophosphate is the most widely accepted view of the actions ples. Whether formulated as a single drug or used as separate
of these agents, it is also possible that some of the benefits of constituents, combinations for selective targeting or for over-
using them in combination arise from unknown or underappreci- coming adaptive losses in drug sensitivity are approaches that
ated additional effects that may be different for each of the should find greater use as these processes become better under-
individual drugs. stood.
Class 3 combinations: rationales and examples. Class 3 combi-
nations include one drug that is useful on its own but is insuffi- PHARMACOLOGIC ANALYSIS AND VISUALIZATION
ciently effective or too toxic, and a second drug that does not OF PHARMACOLOGIC EFFECTS OF
share the same activity as the first drug and may have no useful DRUG COMBINATIONS
effect on its own, but that can enhance the effectiveness of
the first drug by either pharmacokinetic or pharmacodynamic Basic Terminology for Analyzing Combination Effects
mechanisms. It should be noted that these enhancers may modu- Using combinations of drugs necessitates considering the quanti-
late not only the magnitude of the response to the active drug tative nature of the combination effects and the terminology for
but also the duration of action of the active drug, with either describing differences between the actual combination effects
type of action being of potential therapeutic benefit. and what is expected based on the known effects of the individual
The antibiotic Augmentin combines an active drug with an drugs (18–20). In the simplest quantitative analysis, the effects
inhibitor of its degradation; the ␤-lactamase inhibitor clavulanate of the combination can equal the sum of the expected effects
is included to prevent the target bacteria from inactivating the of the two drugs alone, referred to as “additive”; less than the
active drug amoxicillin. Another approach is to use a second expected sum, “subadditive”; or greater than the expected sum,
drug to alter the tissue or cellular distribution of the first active “superadditive.” With regard to mechanisms of action, the effects
drug, exemplified by Sinemet (carbidopa and levodopa), a com- of the two drugs can be either independent, if neither drug alters
bination used in Parkinson’s disease. The therapeutic goal is to the action of the other, or interactive, if one drug in some way
increase dopamine levels in the brain by having levodopa be alters the action of the other drug. Some sources use the term
converted to dopamine by dopa-decarboxylase in the brain. “independence” to describe what is typically called additivity
Dopa-decarboxylase in the periphery can metabolize levodopa (20). The authors prefer to use the terms “additivity” for the
to dopamine, however, decreasing the amount of levodopa deliv- quantitative nature of the combination and “independence” for
ered to the brain (dopamine does not cross the blood-brain the mechanistic (noninteractive) nature of the combination. The
barrier) and increasing unwanted side effects of dopamine in concept of interaction is inherent to the nature of the class 3
tissues outside the blood-brain barrier. Carbidopa is a dopa- combinations described previously, where one drug by definition
decarboxylase inhibitor that does not cross the blood-brain bar- alters the actions of the other. In contrast, the targeting of differ-
rier, selectively inhibiting dopa-decarboxylase in the periphery ent aspects of disease with the class 1 combinations does not
and indirectly targeting levodopa action to the brain. necessarily imply independence in their actions, because either
Class 3 combinations in COPD. There are no clear examples drug could alter the effects of the other on their different end
of intentional class 3 combinations in current COPD therapy,
points, in addition to its primary effect. Similarly, class 2 combi-
although such effects may contribute to some extent to COPD
nations can be either independent or interactive.
therapy. Corticosteroids do not directly cause bronchodilation,
but they can upregulate expression of ␤2-receptors, potentially Mechanisms for Subadditive and Superadditive Effects
enhancing the effect of the ␤2-agonist (13). There is mixed evi- of Drug Combinations
dence on the question of whether corticosteroids may prevent
or counter agonist-induced desensitization and downregulation For drugs with independent actions, subadditivity is most likely
of ␤2-receptors, and these effects may vary with cell and tissue to occur when tissue responsiveness rather than drug effective-
type or with specific receptor polymorphisms (14–16). Altering ness is the limiting factor. If the sum of the responses to the two
␤-receptor expression and desensitization is clearly not the pri- drugs individually is greater than the maximal response possible
mary rationale for this combination, but these effects may con- by the system, then subadditivity occurs, without any true inter-
tribute to some extent to the effectiveness of the combination. action between the drugs (Figure 1, top panel). Subadditivity for
There may be merit in considering the actions of some drugs interactive drugs occurs when one drug interferes with the action
used as monotherapy as being class 3 combinations with endoge- of the other to decrease its effect, for example by enhancing its
nous mediators. Theophylline as monotherapy can elevate cyclic degradation or by accelerating the downregulation of its recep-
adenosine monophosphate, presumably acting as an enhancer of tors (Figure 1, middle panel). Classical antagonism is a special
the actions of endogenous agents that stimulate cyclic adenosine case of subadditivity caused by interference, where the second
monophosphate formation, the constitutive activity of Gs-cou- interacting drug has no action of its own except to block the
pled receptors, or the basal activity of adenylyl cyclases. Simi- action of either the first drug or the endogenous agonist.
larly, muscarinic antagonists given alone can prevent the endoge- Superadditivity can occur only for interactive combinations.
nous activation of muscarinic receptors that results from the Although many studies use the term “synergistic” for all effects
presence of endogenous agonist or perhaps from constitutive of combinations that are greater than expected from simple
activity of muscarinic receptors. additivity, the authors prefer to use “enhancement” and “syner-
It is intriguing to think of salmeterol as a class 3 combination gism” as separate terms for superadditive effects, with these
of two active ingredients within a single compound (17). Its terms providing additional mechanistic information. Enhance-
saligenin headgroup is the active drug in ␤2-receptor activation, ment is used when the second drug has no effect on its own and
whereas its arylalkyl-oxyalkyl tail is a built-in enhancer that only increases the effectiveness of the first drug (Figure 1, bottom
targets the active drug to its specific site of action, where the panel, left). In contrast, “synergism” is reserved for the case in
drug then binds tightly to prolong the headgroup’s duration of which each of the drugs has clearly demonstrable effects on its
action at this site. Salmeterol provides an active drug plus two own but where the effects of the combination are clearly greater
Toews and Bylund: Pharmacology of Combination Therapy 287

Figure 2. Fishnet diagram of the response surface for a synergistic drug

combination. The right front edge of the diagram shows the response
to Drug A alone, with an EC50 of 1 and a maximal response of 100.
The left front edge shows the corresponding response for Drug B with
the same EC50 and maximal response. The magnitude of the response
at each pair of drug combinations is shown on the vertical axis. Drug
A and Drug B also interact with each other to produce synergism, each
increasing the magnitude of the response to the other. The response
to Drug A in the presence of a maximal concentration of Drug B is
shown on the left rear wall, and that to Drug B with maximal Drug A
on the right rear wall, with an overall maximal response of 400 at
maximal concentrations of both drugs.

than 1.5 times the sum of the individual effects is often taken
as the criterion for synergism. Note that enhancement can occur
Figure 1. Mechanistic principles for additivity, subadditivity, and super-
because of an increase in potency, an increase in efficacy, or
additivity of combination drug effects. The actions of the individual both, and full dose-response curves in the absence and presence
drugs are indicated beside their respective arrows, and the net effect of the second drug are necessary to delineate fully these effects.
of the combination is indicated at the bottom of each pair. Top panel An example of synergism from the authors’ work, albeit with
illustrates mechanisms for drugs with independent actions, with additiv- endogenous agents rather than with drugs, is the stimulation of
ity occurring if tissue responsiveness allows both Drug A and Drug B human airway smooth muscle cell DNA synthesis by the lipid
to exert their full effects (left), but with subadditivity occurring if tissue mediator lysophosphatidic acid acting by a G protein–coupled
responsiveness is the limiting factor for the total magnitude of response receptor, and epidermal growth factor acting by its receptor
possible (right). Middle panel illustrates additivity (left) versus subaddi- tyrosine kinase pathway (21). Lysophosphatidic acid caused an
tivity (right) for an interactive combination in which Drug B interferes 8-fold stimulation; epidermal growth factor caused a 17-fold
with the actions of Drug A (a threefold reduction indicated by the red stimulation; and lysophosphatidic acid plus epidermal growth
inhibition line), preventing it from exerting its full response, but Drug factor caused a 98-fold stimulation, nearly four times the ex-
B nonetheless has a beneficial effect on its own. The case in which Drug pected value from the sum of the agents separately. In contrast,
B has no effect on its own and only interferes with Drug A action is in similar studies of this same response with endothelin and
simple antagonism. Bottom panel illustrates two possibilities for super- epidermal growth factor in another laboratory (22), there was no
additivity for interactive combinations: enhancement for the left pair stimulation by endothelin alone, but the presence of endothelin
of drugs, in which Drug B increases the effect of Drug A by twofold increased the stimulation by epidermal growth factor from 23-
(green arrow) but has no effect on its own, and synergism for the right
fold to 86-fold, an example of enhancement. The authors of this
pair of drugs, where Drug B exerts a useful effect on its own in addition
latter study referred to their effect as “potentiation”; others have
to its twofold enhancement of the effects of Drug A.
used the term “sensitization” for similar phenomena. Because
these terms have connotations of increased potency or sensitiv-
ity, as opposed to increased efficacy or maximal response, and
than additive (Figure 1, bottom panel, right). Synergism arises because both terms have specific uses for situations in which
from the summation of the individual drug effects plus additional one agent is given as pretreatment rather than in combination,
mechanisms that lead to one or both drugs amplifying the effect the authors prefer the more general term “enhancement” except
of the other. In the organizational framework described pre- in cases where there is clearly synergism. Several extensive trea-
viously, synergism occurs for drugs that are a class 3 combination tises during the past 15 years have addressed the many different
in addition to either a class 1 or 2 combination. A value greater models that have been developed to understand the concepts
288 PROCEEDINGS OF THE AMERICAN THORACIC SOCIETY VOL 2 2005

of independence, additivity, synergism, and antagonism (18–20).

These works provide interesting insights and detailed descrip-
tions of the mathematical analyses and theoretical models for
such effects that are beyond the scope of this article.
Graphic Representation of Combination Drug Effects
The approach to graphic visualization and analysis of the effects
of drug combinations depends on the class, as defined previously.
For the different end points targeted by the agents in class 1
combinations, in which the effects on each end point are largely
independent, separate dose-response curves for each response
are adequate, because the response to each drug in these combi-
Figure 3. Isobolographic analysis of combination drug additivity, sub-
nations is essentially flat for the different endpoint induced by
additivity, and superadditivity. Two different isoboles are shown for
the other drug. For class 3 combinations, in which one of the
the 50% response level for Drug A, whose EC50 is 20, and for Drug B,
drugs has no effect on its own, a family of traditional dose-
whose EC50 is 100. The two points on the axes represent these 50%
response curves in the absence and presence of increasing con- response levels for Drug A alone (at a concentration of 20) and Drug
centrations of the second drug should be adequate to analyze B alone (at a concentration of 100). The isobole lines represent all pairs
and illustrate the interactions. Changes in potency, efficacy, or of concentrations of Drug A and Drug B giving the same magnitude
both may be observed in these graphs. of response, the line of additivity. The dashed line is the linear isobole
For combinations in which both drugs affect the same end predicted from the assumptions used in most treatises, which is valid
point on their own but may in addition have interactive effects for cases in which the two drugs share the same receptor and response.
ranging from antagonism to synergism, the most useful approach The solid line shows the more general isobole for the case in which no
is a three-dimensional group of bar or line graphs of the dose– assumptions about the receptors for the two drugs are required. The
response curves for each agent in the absence and presence of square data points on this line represent calculated pairs of concentra-
increasing concentrations of the other drug. These curves can tions whose individual responses add up to 50% of the maximal re-
then be connected graphically to form a response surface dia- sponse; as examples, other concentration pairs giving the same response
gram or fishnet plot, allowing convenient visualization of the include 5 of Drug A plus 43 of Drug B, 10 of Drug A plus 20 of Drug
measured or predicted effect of every pair of concentrations. In B, and 17 of Drug A plus 4.2 of Drug B. Points C–G represent responses
these graphs, the concentrations of each drug are plotted on the seen with various hypothetical drug combinations. Combination C falls
horizontal axes and the response is plotted on the vertical axis. on the dashed line and is additive in the first model, but in the second
The dose–response curves to each of the drugs alone are on the model it is above the line and subadditive (requiring higher than pre-
front edges of the diagram, and the dose–response curves to dicted doses to attain the indicated effect). Similarly, Combination D is
additive in the first model, but in the second model it is superadditive
each drug in the presence of maximal concentrations of the other
(with the indicated effect level occurring at lower concentrations than
are on the back walls of the diagram. An example of such a plot
predicted). Combination E is not additive in either model but is super-
is shown in Figure 2, for a case in which each drug enhances the
additive in the first model and subadditive in the second model. Combi-
efficacy but not the potency of the other. Standard spreadsheet nation F is superadditive using either model, whereas combination G is
programs and more specialized scientific graphing and analysis subadditive using either model.
programs are capable of generating and analyzing these plots.
Two approaches are commonly used for visualizing and ana-
lyzing the occurrence of simple additivity versus synergism or
interference. The first is an isobolograph, a plot of one or more Response surface or fishnet graphs are also useful for de-
of the isoboles, or equal-response lines (Figure 3). The dose of tecting synergism or interference. The surface of the predicted
Drug A required to achieve a given level of response (e.g., 50%) effects with simple additivity for all combinations of drug concen-
is plotted on one axis, and the dose of Drug B required to achieve trations is plotted (i.e., not only for a single effect level). Here,
the same response is plotted on the other axis. All of the other combinations that exhibit synergism fall above the additivity
concentration pairs of Drug A with Drug B that are expected surface (lower concentrations required to achieve a given re-
to give the same response level, based on simply adding their sponse) and combinations that exhibit interference lie below the
individual effects, are then plotted to generate the isobole line. surface. An interesting plot that can be generated from these
The actual measured response to each tested combination of data, at least in theory, is a plot of the difference between the
Drug A given together with Drug B is then plotted on the same actual response surface for the combinations and the surface
graph. Points that fall on the line are additive, whereas points predicted by simple additivity. This plot generates a new surface
that fall below the line of additivity represent combinations that revealing the pattern of concentrations exhibiting synergism or
are synergistic (requiring lower concentrations of the combina- interference and the extent of this interaction for each pair of
tion than expected to attain the given effect). Points that fall concentrations; readers are referred to more extensive treatises
above the line are subadditive (requiring higher concentrations for examples and additional details (20).
of the combination than expected to attain the given effect).
Although most discussions of this subject present these iso-
ADDITIONAL PHARMACOLOGIC ASPECTS OF
boles as straight lines (18–20), that is the expected outcome only
COMBINATION THERAPY
in selected instances in which the two drugs act on the same
receptor or other target in a competitive manner. For the more This discussion of combination pharmacology has focused on
general case, where the two drugs act on different receptors or rationales and mechanisms for the beneficial effects of drug
different pathways to the same end point, the predicted additivity combinations. It is critical to recognize, however, that combining
isoboles are concave. The isobole lines for both assumptions two or more drugs is always associated with the danger of greater
are illustrated in Figure 3, highlighting the different conclusions side effects. These can include additivity of the known side effects
depending on the assumption (or knowledge) regarding the tar- of each of the drugs or completely unexpected side effects caused
gets of the two drugs. by interactions between them.
Toews and Bylund: Pharmacology of Combination Therapy 289

A topic of considerable interest to pharmacologists that is pled receptors: a current perspective. Mol Pharmacol 2003;64:1271–
not discussed here is the question of using drug combinations 1276.
6. Seifert R, Wenzel-Seifert K. Constitutive activity of G protein-coupled
in a fixed-dose single preparation versus using the drugs sepa- receptors: cause of disease and common property of wild-type recep-
rately at their individually most effective doses. Also beyond tors. Naunyn Schmiedebergs Arch Pharmacol 2002;366:381–416.
the scope of this discussion are the potential advantages or dis- 7. McDonnell DP, Connor CE, Wijayaratne A, Chang C-Y, Norris JD.
advantages of combining the activities of the components of a Definition of the molecular and cellular mechanisms underlying the
combination preparation into a single molecule with multiple tissue-selective agonist/antagonist activities of selective estrogen re-
beneficial effects. ceptor modulators. Recent Prog Horm Res 2002;57:295–316.
8. Krishnan V, Heath H, Bryant HU. Mechanism of action of estrogens and
Note that combination therapy need not be limited to combi- selective estrogen receptor modulators. Vitam Horm 2001;60:123–147.
nations of drugs. Single drugs can be used with various interven- 9. Kenakin T. Drug efficacy at G protein-coupled receptors. Annu Rev
tions for additional benefit. This is particularly relevant to COPD, Pharmacol Toxicol 2002;42:349–379.
in which smoking cessation, pulmonary rehabilitation, oxygen 10. Clarke WP, Bond RA. The elusive nature of intrinsic efficacy. Trends
therapy, and lung-volume–reduction surgery may all be benefi- Pharmacol Sci 1998;19:270–276.
11. Gainetdinov RR, Premont RT, Bohn LM, Lefkowitz RJ, Caron MG.
cial in addition to pharmacologic approaches (23). Desensitization of G protein-coupled receptors and neuronal func-
tions. Annu Rev Neurosci 2004;27:107–144.
CONCLUSIONS 12. Reid JL. Pharmacokinetic and pharmacodynamic aspects of the choice
of components of combination therapy. J Hum Hypertens 1995;9:S19–
There are many different rationales for the use of combination S23.
therapy, and the pharmacologic principles for evaluating and 13. Collins S, Bolanowski MA, Caron MG, Lefkowitz RJ. Genetic regulation
understanding their actions are in hand. There are already multi- of ␤-adrenergic receptors. Annu Rev Physiol 1989;51:203–215.
ple examples of the use of combination therapy for COPD, with 14. Moore PE, Calder MM, Silverman ES, Panettieri RA Jr, Shore SA.
Effect of dexamethasone on ␤2-adrenergic desensitization in airway
evidence for beneficial effects. It is clear, however, that better smooth muscle: role of the ARG19 polymorphism. Chest 2003;123:
drugs are needed to treat COPD, perhaps including new combi- 368S–369S.
nations. By applying the pharmacologic principles and consider- 15. Tan KS, McFarlane LC, Lipworth BJ. Concomitant administration of
ing additional possibilities based on the diverse mechanistic ra- low-dose prednisolone protects against in vivo ␤2-adrenoceptor sub-
tionales summarized here, it may very well be possible to develop sensitivity induced by regular formoterol. Chest 1998;113:34–41.
16. Hjemdahl P, Zetterlund A, Larsson K. ␤2-agonist treatment reduces
novel combinations that are more efficacious or have fewer side ␤2-sensitivity in alveolar macrophages despite corticosteroid treat-
effects than existing agents and combinations. ment. Am J Respir Crit Care Med 1996;153:576–581.
17. Coleman RA, Johnson M, Nials AT, Vardey CJ. Exosites: their current
Conflict of Interest Statement : M.L.T. is the recipient of a research grant from
status, and their relevance to the duration of action of long-acting ␤2-
GlaxoSmithKline to study salmeterol and fluticasone, alone and in combination,
on regulation of airway cell epidermal growth factor receptors, in comparison adrenoceptor agonists. Trends Pharmacol Sci 1996;17:324–330.
with other related agents, and he received honoraria from HealthMatters Commu- 18. Tallarida RJ. Drug synergism: its detection and applications. J Pharmacol
nications for serving on the steering committee for this Symposium. D.B.B. does Exp Ther 2001;298:865–872.
not have a financial relationship with a commercial entity that has an interest in 19. Berenbaum MC. What is synergy? Pharmacol Rev 1989;41:93–141.
the subject of this manuscript. 20. Greco WR, Bravo G, Parsons JC. The search for synergy: a critical review
from a response surface perspective. Pharmacol Rev 1995;47:331–385.
21. Cerutis DR, Nogami M, Anderson JL, Churchill JD, Romberger DJ,
References
Rennard SI, Toews ML. Lysophosphatidic acid and EGF stimulate
1. Rescigno A. Foundations of pharmacokinetics. New York: Kluwer Aca- mitogenesis in human airway smooth muscle cells. Am J Physiol 1997;
demic/Plenum Publishers; 2003. 273:L10–L15.
2. Limbird LE. Cell surface receptors: a short course on theory and methods. 22. Panettieri RA, Goldie RG, Rigby PJ, Eszterhas AJ, Hay DWP. Endo-
Norwell, MA: Kluwer Academic Publishers; 1996. thelin-1-induced potentiation of human airway smooth muscle prolifer-
3. Foreman JC, Johansen T. Textbook of receptor pharmacology. Boca ation: an ETA receptor-mediated phenomenon. Br J Pharmacol 1996;
Raton, FL: CRC Press; 2002. 118:191–197.
4. Kenakin T. Efficacy as a vector: the relative prevalence and paucity of 23. Barnes P, Drazen J, Rennard S, Thomson N, editors. Asthma and COPD:
inverse agonism. Mol Pharmacol 2004;65:2–11. basic mechanisms and clinical management. London: Academic Press;
5. Milligan G. Constitutive activity and inverse agonists of G protein-cou- 2002.