Principles of Therapeutics

Over a century ago, Claude Bernard formalized criteria for gathering valid information in experimental medicine. However, application of these criteria to therapeutics and to the process of making decisions about therapeutics has, until recently, been slow and inconsistent. Although the diagnostic aspects of medicine are approached with sophistication, therapeutic decisions often are made on the basis of impressions and traditions. Over the past three decades, the principles of human experimentation have been defined, and the techniques for evaluation of therapeutic interventions have progressed to the point that it should now be considered absolutely unethical to apply the art, as opposed to the science, of therapeutics to any patient who directly (the adult or child) or indirectly (the fetus) receives drugs for therapeutic purposes. Therapeutics must now be dominated by objective evaluation of an adequate base of factual knowledge. This philosophy has been popularized recently under the terminology of 'evidence-based medicine.'

Conceptual Barriers to Therapeutics as a Science

The most important barrier that inhibited the development of therapeutics as a science seems to have been the belief that multiple variables in diseases and in the effects of drugs are uncontrollable. If this were true, the scientific method would not be applicable to the study of pharmacotherapy. In fact, therapeutics is the aspect of patient care that is most amenable to the acquisition of useful data, since it involves an intervention and provides an opportunity to observe a response. Recently, it has become evident that many of the important aspects of disease cannot be adequately assessed by objective data. For instance, dyspnea may not be predicted by measures of pulmonary function with spirometry, and the pain of angina pectoris often is not well correlated with ST-segment depression on the electrocardiogram. Nonetheless, there is a tendency to ignore or denigrate the subjective, symptomatic 'soft' data in favor of the objective 'hard' endpoints. The challenge for clinical investigators has been to objectify or 'harden' subjective measures so that they are useful for quantifying drug responses. It is now appreciated that clinical phenomena that are important to patientssuch as dyspnea, pain, and ability to functioncan be defined, described, and quantified with some precision. The approach to complex clinical data has been artfully discussed by Feinstein (1983, 1999).

Another barrier to the realization of therapeutics as a science was overreliance on traditional diagnostic labels for disease. This encouraged the physician to think of a disease as static rather than dynamic, to view patients with the same 'label' as a homogeneous rather than a heterogeneous population, and to consider a disease as a single entity even when information about pathogenesis was not available. If diseases are not considered to be dynamic, 'standard' therapies in 'standard' doses will be the order of the day; decisions will be reflexive. Needed instead is an attitude that makes the physician responsible for recognition of and compensation for changes that occur in pathophysiology as the underlying process evolves. For example, the term myocardial infarction refers to localized destruction of myocardial cells caused by interruption of the blood supply; however, decisions about therapy must take into account a variety of autonomic, hemodynamic, and electrophysiological variables that change as a function of the time, size, and location of the infarction. Failure to take all such variables into account while planning a therapeutic maneuver may result in ineffective therapy in some patients while exposing others to avoidable toxicity. A diagnosis or label of a disease or syndrome usually indicates a spectrum of possible causes and outcomes. Therapeutic experiments that fail to control for the known variables that affect prognosis yield uninterpretable data. Often, if not usually, not all of the relevant variables are known. In such cases, the response to a therapeutic intervention can be a clue to sort out the parameters that contribute to the response and may be a way to discover the underlying variables that contribute to the disease.

A third conceptual barrier was the incorrect notion that data derived empirically are useless, because they are not generated by application of the scientific method. Empiricism often is defined as the practice of medicine founded on mere experience, without the aid of science or a knowledge of principles. The connotations of this definition are misleading; empirical observations need not be scientifically unsound. In fact, concepts of therapeutics have been greatly advanced by the clinical observer who makes careful and controlled observations of the outcome of a therapeutic intervention. The results, even when the mechanisms of disease and their interactions with the effects of drugs are not understood, are nevertheless often crucial to appropriate therapeutic decisions. Frequently, the initial suggestion that a drug may be efficacious in one condition arises from careful, empirical observations that are made while the drug is being used for another purpose. Examples of valid empirical observations that have resulted in new uses of drugs include the use of penicillamine to treat arthritis, lidocaine to treat cardiac arrhythmias, propranolol and clonidine to treat hypertension, and sildenafil for male erectile dysfunction. Conversely, empiricism, when not coupled with appropriate observational methods and statistical techniques, often results in findings that are invalid or misleading.

Clinical Trials

Application of the scientific method to experimental therapeutics is exemplified by a well-designed and well-executed clinical trial. Clinical trials form the basis for therapeutic decisions by all physicians, and it is therefore essential that they be able to evaluate the results and conclusions of such trials critically. To maximize the likelihood that useful information will result from the experiment, testable hypotheses of the study must be clearly defined, homogeneous populations of patients must be selected, appropriate control groups must be found, meaningful and sensitive indices of drug effects must be chosen for observation, and the observations must be converted into data and then into valid conclusions. The sine qua non of any clinical trial is its controls. Many different types of controls may be used, and the term controlled clinical trial is not synonymous with randomized, double-blind, placebo-controlled trial. Selection of a proper control group is as critical to the eventual utility of an experiment as the selection of the experimental group. Although the randomized, double-blind controlled trial is the most effective design for avoiding bias and distributing unknown variables between the 'treatment' and 'control' groups, it is not necessarily the optimal design for all studies. It may be impossible to use this design to study disorders that occur rarely, disorders in patients who cannotby regulation, ethics, or bothbe studied (e.g., children, fetuses, or some patients with psychiatric diseases), or disorders with a typically fatal outcome (e.g., rabies), where historical controls can be used.

There are several requirements in the design of clinical trials to test the relative effects of alternative therapies. (1) Specific outcomes of therapy that are clinically relevant and quantifiable must be measured. These may include subjective assessments, which are important in determining whether a therapy improves the patient's well-being. Quality of life can be assessed by the experimental subject and can be tabulated objectively and incorporated into evaluation of a therapy (Guyatt et al., 1993). Wherever possible, well-defined clinical endpoints, i.e., survival or pain relief, should be used, rather than an intermediate endpoint or 'surrogate' marker (Fleming and DeMets, 1996; Bucher et al., 1999). A surrogate marker is a clinical sign or laboratory test that correlates with the clinical outcome of a disease. Blood pressure, blood cholesterol, CD4 lymphocyte count in acquired immunodeficiency syndrome (AIDS), and premature ventricular complexes are examples of surrogate markers that have been used as endpoints in clinical trials. Although surrogate markers often are useful to reduce the length and sample size of a clinical trial, the results of such trials may be misleading, as the Cardiac Arrhythmia Suppression Trial (CAST) demonstrated (Echt et al., 1991). In CAST, the antiarrhythmic drugs encainide, flecainide, and moricizine were effective in suppressing ventricular arrhythmias (the surrogate marker) in patients following a myocardial infarction, but the drugs nonetheless increased mortality. The ultimate test of a drug's efficacy must rest with actual clinical outcomes. (2) The accuracy of diagnosis and the severity of the disease must be comparable in the groups being contrasted; otherwise, false-positive and false-negative errors may occur. This is a particular issue in developing therapies for poorly understood syndromes, such as fibromyalgia and chronic fatigue syndrome. (3) The dosages of the drugs must be chosen and individualized in a manner that allows relative efficacy to be compared at equivalent toxicities or allows relative toxicities to be compared at equivalent efficacies. (4) Placebo effects, which occur in a large percentage of patients, can confound many studiesparticularly those that involve subjective responses; controls must take this into account (Temple, 1997). (5) Compliance with the experimental regimens should be assessed before subjects are assigned to experimental or control groups. The drug-taking behavior of the subjects should be reassessed during the course of the trial. Noncompliance, even if randomly distributed between both groups, may cause falsely low estimates of the true potential benefits or toxicity of a particular treatment. (6) Sample size should be estimated prior to beginning a clinical trial so that the trial has the power to detect a statistically significant effect if, in fact, such an effect exists. Depending upon such factors as the overall prognosis and variability of the disease and the anticipated improvement and variability in outcome or toxicity from the new treatment, very large numbers of subjects may be needed; otherwise, the possibility of a false-negative result is high (i.e., no statistically significant differences between the two treatments will be found, even though differences actually exist). It may be very difficult to determine whether or not a new therapy is equivalent to existing therapy without the use of a placebo. Even with large sample sizes, there can be uncertainty. Unless there is a substantial effect of therapy that has been consistently demonstrated in previous trials, it may be impossible to assure that either the standard therapy or the new therapy has a significant effect, i.e., is better than a putative placebo, even if the two are shown to be statistically equivalent. (7) Ethical considerations may be major determinants of the types of controls that can be used and must be evaluated explicitly (Passamani, 1991). For example, in therapeutic trials that involve life-threatening diseases for which there already is an effective therapy, the use of a placebo is unethical, and new treatments must be compared with 'standard' therapies.

The results of clinical trials of new therapeutic agents or of old agents for new indications may have severe limitations in terms of what can be expected of drugs when they are used in an office practice (Feinstein, 1994). To reduce variability, patients for experimental trials often are selected to eliminate coexisting diseases and concomitant therapy. Such trials usually assess the effect of only one or two drugs, not the many that might be given to or taken by the same patient under the care of a physician. Clinical trials usually are performed with relatively small numbers of patients for periods of time that may be shorter than are necessary in practice, and compliance may be better controlled than it can be in practice. These factors lead to several inescapable conclusions:

As has been implied above, therapy as a science does not apply simply to the evaluation and testing of new, investigational drugs in animals and human beings. It applies with equal importance to the treatment of each patient as an individual. Therapists of every type have long recognized and acknowledged that individual patients show wide variability in response to the same drug or treatment method. Progress has been made in identifying the sources of variability. Important factors are presented in Figure 31; the basic principles that underlie these sources of variability have been presented in Chapters 1: Pharmacokinetics: The Dynamics of Drug Absorption, Distribution, and Elimination and 2: Pharmacodynamics: Mechanisms of Drug Action and the Relationship Between Drug Concentration and Effect. The following discussion relates to the strategies that have been developed to deal with variability in the clinical setting. (See also Appendix II.)

Pharmacokinetic Considerations

Interpatient and intrapatient variation in disposition of a drug must be taken into account in choosing a drug regimen. For a given drug, there may be wide variation in its pharmacokinetic properties among individuals. For some drugs, this variability may account for one-half or more of the total variation in eventual response. The relative importance of the many factors that contribute to these differences depends in part on the drug itself and on its usual route of elimination. Drugs that are excreted primarily unchanged by the kidney tend to have smaller differences in disposition among patients with similar renal function than do drugs that are inactivated by metabolism. Of drugs that are extensively metabolized, those with high metabolic clearance and large presystemic (first-pass) elimination have marked differences in bioavailability, whereas those with slower biotransformation tend to have the largest variation in elimination rates among individuals. Studies in identical and nonidentical twins have revealed that genotype is a very important determinant of differences in the rates of metabolism (Penno and Vesell, 1983). For many drugs, physiological and pathological variations in organ function are major determinants of their rate of disposition. For example, the clearance of digoxin and gentamicin is related to the rate of glomerular filtration, whereas that of lidocaine and propranolol is dependent primarily on the rate of hepatic blood flow. The effect of diseases that involve the kidneys or liver is to impair elimination and to increase the variability in the disposition of drugs. In such settings, measurements of concentrations of drugs in biological fluids can be used to assist in the individualization of drug therapy. Since old age and renal or hepatic diseases also may affect the responsiveness of target tissues (e.g., the brain), the physician should be alert to the possibility of a shift in the range of therapeutic concentrations.

A test should not be performed simply because an assay is available. More assays of drugs are available than are generally useful. Determinations of concentrations of drug in blood, serum, or plasma are particularly useful when well-defined criteria are fulfilled: (1) There must be a demonstrated relationship between the concentration of the drug in plasma and the eventual therapeutic effect that is desired and/or the toxic effect that must be avoided. (2) There should be substantial interpatient variability in disposition of the drug (and small intrapatient variation). Otherwise, concentrations of drug in plasma could be predicted adequately from dose alone. (3) It should be difficult to monitor intended or unintended effects of the drug. Whenever clinical effects or minor toxicity are measured easily (e.g., the effect of a drug on blood pressure or blood coagulation), such assessments should be preferred in the decision to make any necessary adjustment of dosage of the drug. However, the effects of some drugs in certain settings are not easily monitored. For example, the effect of Li⁺ on manic-depressive illness may be delayed and difficult to quantify. For some drugs, the initial manifestation of toxicity may be serious (e.g., digitalis-induced arrhythmias or theophylline-induced seizures). The same concepts apply to a number of agents used for cancer chemotherapy. Other drugs (e.g., antiarrhythmic agents) produce toxic effects that mimic symptoms or signs of the disease being treated. Many drugs are used for prophylaxis of an intermittent, potentially dangerous event; examples include anticonvulsants and antiarrhythmic agents. In each of these situations, titration of drug dosage may be aided by measurements of concentrations of the drug in blood (4). The concentration of drug required to produce therapeutic effects should be close to the value that causes substantial toxicity (see below). If this circumstance does not apply, patients could simply be given the largest dose known to be necessary to treat a disorder, as is commonly done with penicillin. However, if there is an overlap in the concentrationresponse relationship for desirable and undesirable effects of the drug, as is true for theophylline, determinations of concentration of drug in plasma may allow the dose to be optimized. All four of the above-described criteria should be met if the measurement of drug concentrations is to be of significant value in the adjustment of dosage. Knowledge of concentrations of drugs in plasma or urine also is particularly useful for the detection of therapeutic failures that are due to lack of patient compliance with a medical regimen or for identification of patients with unexpected extremes in the rate of drug disposition.

Assay of drugs to assist the physician in achieving a desired concentration of drug in blood or plasma (i.e., 'targeting' the dose) is another example of the use of an intermediate or surrogate endpoint of therapy in place of the ultimate clinical goal. Surrogate markers also can be applied in other ways; one is to provide an indication for a change in the choice of drug therapy. Measurements of concentrations of drugs in plasma and/or measurements of one or more pharmacological effects of the drug can provide an indication of probable lack of efficacy. Other issues of importance with regard to the measurement and interpretation of drug concentrations are discussed in Chapter 1: Pharmacokinetics: The Dynamics of Drug Absorption, Distribution, and Elimination and Appendix II.

Pharmacodynamic Considerations

Considerable interindividual variation in the response to drugs remains after the concentration of the drug in plasma has been adjusted to a target value; for some drugs, this pharmacodynamic variability accounts for much of the total variation in responsiveness among patients. As discussed in Chapter 2: Pharmacodynamics: Mechanisms of Drug Action and the Relationship Between Drug Concentration and Effect, the relationship between the concentration of a drug and the magnitude of the observed response may be complex, even when responses are measured in simplified systems in vitro, although typical sigmoidal concentrationeffect curves usually are seen (see Chapter 2: Pharmacodynamics: Mechanisms of Drug Action and the Relationship Between Drug Concentration and Effect). When drugs are administered to patients, however, there is no single characteristic relationship between the drug concentration in plasma and the measured effect; the concentrationeffect curve may be concave upward, concave downward, linear, sigmoid, or an shape. Moreover, the concentrationeffect relationship may be distorted if the response being measured is a composite of several effects, such as the change in blood pressure produced by a combination of cardiac, vascular, and reflex effects. However, such a composite concentrationeffect curve often can be resolved into simpler curves for each of its components. These simplified concentrationeffect relationships, regardless of their exact shape, can be viewed as having four characteristic variables: potency, slope, maximal efficacy, and individual variation. These are illustrated in Figure 32 for the common sigmoidal log doseeffect curve.

Potency

The location of the concentrationeffect curve along the concentration axis is an expression of the potency of a drug. Although often related to the dose of a drug required to produce an effect, potency is more properly related to the concentration of the drug in plasma to approximate more closely the situation in isolated systems in vitro and to avoid the complicating factors of pharmacokinetic variables. Although potency obviously affects drug dosage, potency per se is relatively unimportant in the clinical use of drugs as long as the required dose can be given conveniently and there is no toxicity related to the chemical structure of the drug rather than to its mechanism. There is no justification for the view that more potent drugs are superior therapeutic agents. However, if the drug is to be administered by transdermal absorption, a highly potent drug is required, since the capacity of the skin to absorb drugs is limited.

Maximal Efficacy

The maximal effect that can be produced by a drug is its maximal, or clinical, efficacy (which is related to, but not precisely the same as, the term efficacy as discussed in Chapter 2: Pharmacodynamics: Mechanisms of Drug Action and the Relationship Between Drug Concentration and Effect). Maximal efficacy is determined principally by the properties of the drug and its receptoreffector system and is reflected in the plateau of the concentrationeffect curve. In clinical use, however, a drug's dosage may be limited by undesired effects, and the true maximal efficacy of the drug may not be achievable. The maximal efficacy of a drug is clearly a major characteristicof much greater clinical importance than its potency. Furthermore, the two properties are not related and should not be confused. For instance, although some thiazide diuretics have similar or greater potency than the loop diuretic furosemide, the maximal efficacy of furosemide is considerably greater.

Slope

The slope of the concentrationeffect curve reflects the mechanism of action of a drug, including the shape of the curve that describes drug binding to its receptor (see Chapter 2: Pharmacodynamics: Mechanisms of Drug Action and the Relationship Between Drug Concentration and Effect). The steepness of the curve dictates the range of doses that are useful for achieving a clinical effect. Aside from this fact, the slope of the concentrationeffect curve has more theoretical than practical usefulness.

Biological Variability

Different individuals vary in the magnitude of their response to the same concentration of a single drug or to similar drugs when the appropriate correction has been made for differences in potency, maximal efficacy, and slope. In fact, a single individual may not always respond in the same way to the same concentration of drug. A concentrationeffect curve applies only to a single individual at one time or to an average individual. The intersecting brackets in Figure 32 indicate that an effect of varying intensity will occur in different individuals at a specified concentration of a drug or that a range of concentrations is required to produce an effect of specified intensity in all of the patients.

Attempts have been made to define and measure individual 'sensitivity' to drugs in the clinical setting, and progress has been made in understanding some of the determinants of sensitivity to drugs that act at specific receptors. For example, responsiveness to -adrenergic receptor agonists may change because of disease (e.g., thyrotoxicosis or heart failure) or because of prior administration of either -adrenergic agonists or antagonists that can cause changes in the concentration of the -adrenergic receptor and/or coupling of the receptor to its effector systems (Iaccarino et al., 1999; see also Chapter 10: Catecholamines, Sympathomimetic Drugs, and Adrenergic Receptor Antagonists). Receptors are not static components of the cell; they are in a dynamic state that is influenced by both endogenous and exogenous factors.

ConcentrationPercent or Quantal ConcentrationEffect Curve

The concentration of a drug that produces a specified effect in a single patient is termed the individual effective concentration. This is a quantal response, since the defined effect is either present or absent. Individual effective concentrations usually are lognormally distributed, which means that a normal variation curve is the result of plotting the logarithms of the concentration against the frequency of patients achieving the defined effect (Figure 33A). A cumulative frequency distribution of individuals achieving the defined effect as a function of drug concentration is the concentrationpercent curve or the quantal concentrationeffect curve. This curve resembles the sigmoid shape of the graded concentrationeffect curve discussed above (Figure 32), but the slope of the concentrationpercent curve is an expression of the pharmacodynamic variability in the population rather than an expression of the concentration range from a threshold to a maximal effect in the individual patient.

The dose of a drug required to produce a specified effect in 50% of the population is the median effective dose, abbreviated as the ED₅₀ (Figure 33B). In preclinical studies of drugs, the median lethal dose, as determined in experimental animals, is abbreviated as LD₅₀. The ratio of the LD₅₀ to the ED₅₀ is an indication of the therapeutic index, which is a statement of how selective the drug is in producing its desired versus its adverse effects. In clinical studies, the dose, or preferably the concentration, of a drug required to produce toxic effects can be compared to the concentration required for the therapeutic effects in the population to evaluate the clinical therapeutic index. However, since pharmacodynamic variation in the population may be marked, the concentration or dose of drug required to produce a therapeutic effect in most of the population will usually overlap the concentration required to produce toxicity in some of the population, even though the drug's therapeutic index in an individual patient may be large. Also, the concentrationpercent curves for efficacy and toxicity need not be parallel, adding yet another complexity to the determination of the therapeutic index in patients. Finally, no drug produces a single effect, and, depending on the effect being measured, the therapeutic index for a drug will vary. For example, much less codeine is required for cough suppression than for control of pain in 50% of the population, and thus the margin of safety, selectivity, or therapeutic index of codeine is much greater as an antitussive than as an analgesic.

Other Factors That Affect Therapeutic Outcome

The variation in pharmacokinetic and pharmacodynamic parameters that accounts for much of the need to individualize therapy has been discussed. Other factors, listed in Figure 31, also should be considered as potential determinants of success or failure of therapy. The following presentation serves as an introduction to these subjects, some of which also are discussed in Chapter 1: Pharmacokinetics: The Dynamics of Drug Absorption, Distribution, and Elimination and Appendix II.

Age

Most drugs are developed and tested in young to middle-aged adults. At each extreme of the age spectrum, individuals differ both in the way they handle drugs (pharmacokinetics) and in their response to drugs (pharmacodynamics). These differences may require substantial alterations in the dose or dose regimen to produce the desired effect in the young or in the very old.

Children

Most medications have not been developed or specifically evaluated in children, and formulations often are inadequate for proper administration. Thus, development of new drugs for children and rational use of old compounds require an integrated approach to pharmacokinetic, pharmacodynamic, and formulation issues. There is no reliable, broadly applicable principle or formula for converting doses of drugs used in adults to doses that are safe and effective in children. When the drug manufacturer does not provide adequate information about pediatric dosage, there can be substantial risk in deriving a dose for children and infants from an adult dose by, for example, simply reducing the dose based upon body weight or surface area. In general, pathways of drug clearance (hepatic and renal) are limited in the newborn, particularly the premature infant. The unique physiology of the newborn has led to past therapeutic disasters, such as gray-baby syndrome (inadequate glucuronidation of chloramphenicol with drug accumulation) and sulfonamide-induced kernicterus (displacement of bilirubin from plasma proteins in the face of increased bilirubin production from fetal erythrocyte turnover, decreased bilirubin conjugation, acidosis, and decreased blood-brain barrier). Careful pharmacokinetic studies in the newborn coupled with clinical therapeutic drug monitoring have markedly improved our knowledge of neonatal developmental pharmacology and resulted in safe therapeutics.

Pathways of drug clearance develop variably over the first year of life and may be influenced by induction of drug-metabolizing enzymes (e.g., phenobarbital exposure). Precise developmental patterns have not been mapped out for most isoforms of cytochrome P450. For CYP1A2, studies using caffeine as a model substrate have revealed the pattern shown in Figure 34 (Lambert et al., 1986). Such a pattern has been noted for many compounds (e.g., theophylline, anticonvulsants) where a very limited metabolic clearance in the newborn matures during the first year of life (albeit with considerable intersubject and metabolic pathway variability) and ultimately achieves weight-adjusted clearance values that exceed those of adults. At puberty, clearance begins to decline, earlier in girls than in boys, to adult levels. The mechanisms regulating such developmental changes are uncertain, and other pathways of drug clearance likely mature with different patterns (deWildt et al., 1999). The critical point is that, at times of physiological change (the premature, the neonate, puberty), major changes in pharmacokinetics are likely to occur, variability is likely to be greatest (both within the same patient over time and among patients), and dosing adjustment, often aided by therapeutic drug monitoring for drugs with narrow therapeutic indices, becomes critical to safe, effective therapeutics. The 7-day-old neonate may be very different pharmacokinetically from the same patient as a newborn, and doses that were appropriate for a 10-year-old on a weight-adjusted basis might well result in overdose for the same patient at age 14.

Pharmacodynamic differences between children and adults have led to unexpected outcomes of therapy and adverse effects. For example, while antihistamines and barbiturates generally sedate adults, these drugs cause many children to become 'hyperactive.' Of great concern are the effects of medications, particularly when used chronically, on physical and cognitive development. Chronic therapy with phenobarbital can have a significant effect on learning and behavior in children. Tetracyclines deposit in developing teeth, with resultant permanent staining. While children are at risk for all the side effects of chronic corticosteroid therapy seen in adults, such drugs will also stunt linear growth. Children, however, are not always at increased risk for adverse drug effects. For example, while young children appear to be at higher risk for hepatotoxicity from valproic acid than are adults, they are at much lower risk for hepatotoxicity from isoniazid and possibly acetaminophen overdose.

In 1997 Congress passed the Food and Drug Administration Modernization Act (FDAMA). One of this act's aims is to enhance the amount of available information on drug use in children. FDAMA and the Final Rule that followed give the FDA the ability to request information on marketed drugs and reward pharmaceutical companies with 6 months of additional marketing exclusivity for studies that adequately address the request. For drugs in development, a plan to obtain data in children must be negotiated with the FDA prior to the drug's approval for marketing. The data required for children depend on the disease being studied. If the disease is similar in adults and children and there is no known reason to suspect the drug to behave differently in children, then the bulk of the adult efficacy data can be extrapolated to children if similar drug exposure in children and adults can be assured. Thus, pharmacokinetic data in children are required along with adequate exposure for safety assessment, but the FDA efficacy standard for adequate and well-controlled trials prior to approval of the drug for children may not be required. This approach by the FDA is novel and has already resulted in the accumulation of much new data in children. The rest of the world has its eyes on this initiative, and discussions are under way in Europe regarding the requirements for approval of drugs for children (Conroy et al., 2000).

Pediatric formulations of old and new drugs remain a problem for practical therapeutics. While toxicity of the vehicles used to administer drugs (e.g., diethylene glycol toxicity from elixir of sulfanilamide) led to the Pure Food and Drug Act of 1938, the 'gasping syndrome' associated with excess administration of drugs preserved with benzyl alcohol was described in the newborn as recently as the 1980s. For intravenous medications, formulations are often too concentrated for proper measurement of the tiny doses required for newborns. Oral formulations frequently present major problems with palatability and possible adverse reactions to flavoring and coloring agents. Particularly for pediatric suspensions, syrups, and chewable tablets, different preparations of the same drugs, while being equivalent from the point of view of bioavailability, may differ in acceptability to a specific patient.

The Elderly

As adults age, gradual changes in drug kinetics and effects result in an increase in the interindividual variability of doses required for a given effect. The pharmacokinetic changes result from changes in body composition and the function of drug-eliminating organs. The reduction in lean body mass, serum albumin, and total body water and the increase in percentage of body fat result in changes in the distribution of drugs depending on their lipid solubility and protein binding. The clearance of many drugs is reduced in the elderly. Renal function declines at a variable rate to about 50% of that in the young adult. Hepatic blood flow and the function of some of the drug-metabolizing enzymes also is reduced in the elderly, but the variability of this change is great. In general, the activities of cytochrome P450 enzymes are reduced, but conjugation mechanisms are relatively well maintained. Frequently, the elimination half-life of drugs is increased as a consequence of a larger apparent volume of distribution (of lipid-soluble drugs) and/or a reduction of the renal or metabolic clearance.

Changes in pharmacodynamics also are important factors in treating the elderly. Drugs that depress the central nervous system produce increased effects at any given plasma concentration. Physiological changes and loss of homeostatic resilience can result in increased sensitivity to unwanted effects of drugs, such as hypotension from psychotropic medications and hemorrhage from anticoagulants, even if dosage is appropriately adjusted to account for the age-related pharmacokinetic changes.

The proportion of our population in the elderly and very old age groups is increasing. These individuals have more illnesses than younger people and consume a disproportionate share of prescription and over-the-counter drugs. These factors, combined with the changes in pharmacokinetics and pharmacodynamics that occur with aging, make the elderly age group a population in whom drug use is likely to be marred by serious adverse drug effects and drug interactions. It is a population that should receive drugs only when absolutely necessary for well-defined indications and at the lowest effective doses. Prospectively defined endpoints, appropriate use of therapeutic drug monitoring, and frequent reviews of the patient's drug historywith discontinuation of those drugs that did not achieve the endpoint desired or are no longer requiredwould greatly improve the health of the elderly population. On the other hand, appropriate therapy should not be withheld because of these concerns. Outcomes data with a number of drug interventions have proven that the elderly can benefit at least as much as, and often more than, the young in the treatment of chronic diseases such as hypertension and hypercholesterolemia (LaRosa et al., 1999). Furthermore the natural history of chronic diseases of the elderly, such as osteoporosis and prostate hyperplasia, can be halted or reversed by appropriate drug therapy.

Gender

Although there may be some pharmacokinetic or pharmacodynamic differences between the sexes, early drug development until recently has been performed exclusively in males because of FDA guidelines that prohibited the participation of women of childbearing potential. In the 1990s, the FDA readdressed the importance of including women in early clinical trials, and the previous guidelines were revised to allow the participation of women in all phases of drug development. It is expected that at the time of drug approval, the database will be sufficiently complete to allow a rational assessment of the pharmacokinetic, pharmacodynamic, and safety issues in each sex (Sherman et al., 1995; Harris et al., 1995).

DrugDrug Interactions

The use of several drugs often is essential to obtain a desired therapeutic objective or to treat coexisting diseases. Examples abound, and the choice of drugs to be employed concurrently can be based on sound pharmacological principles. In the treatment of hypertension, a single drug is effective in only a modest percentage of patients. In the treatment of heart failure, the concurrent use of a diuretic with a vasodilator and/or a cardiac glycoside often is essential to achieve an adequate cardiac output and to keep the patient free from edema. Multiple-drug therapy is the norm in cancer chemotherapy and for the treatment of certain infectious diseases. The goals in these cases usually are to improve therapeutic effectiveness and to delay the emergence of malignant cells or of microorganisms that are resistant to the effects of available drugs. When physicians use several drugs concurrently, they face the problem of knowing whether a specific combination in a given patient has the potential to result in an interaction, and if so, how to take advantage of the interaction if it leads to improvement in therapy or how to avoid the consequences of an interaction if they are adverse.

A potential drug interaction refers to the possibility that one drug may alter the intensity of pharmacological effects of another drug given concurrently. The net result may be enhanced or diminished effects of one or both of the drugs or the appearance of a new effect that is not seen with either drug alone.

The frequency of significant beneficial or adverse drug interactions is unknown. Surveys that include data obtained in vitro, in animals, and in case reports tend to predict a frequency of interactions that is higher than actually occurs. While such reports have contributed to skepticism about the overall importance of drug interactions, there are potential interactions of definite clinical importance, and the physician must be alert to the possibility of their occurrence. Estimates of the incidence of clinical drugdrug interactions range from 3% to 5% in patients taking a few drugs to 20% in patients who are receiving 10 to 20 drugs. Because most hospitalized patients receive at least six drugs, the scope of the problem clearly is significant. The recent successful treatment of AIDS with multiple drugs, including several that have potent effects to alter the activity of drug-metabolizing enzymes, has heightened the public awareness of drug interactions. Recognition of beneficial effects and recognition and prevention of adverse drug interactions require a thorough knowledge of the intended and possible effects of drugs that are prescribed, an inclination to attribute unusual events to drugs rather than to disease, and adequate observation of the patient. Automated monitoring of prescription orders in the hospital or outpatient pharmacy may decrease the physician's need to memorize potential interactions. Nevertheless, knowledge of likely mechanisms of drug interactions is the only way the clinician can be prepared to analyze new findings systematically. It is incumbent upon the physician to be familiar with the basic principles of drugdrug interactions in planning a therapeutic regimen. Such reactions are discussed for individual drugs throughout this textbook.

Interactions may be either pharmacokinetic (alteration of the absorption, distribution, or elimination of one drug by another) or pharmacodynamic (e.g., interactions between agonists and antagonists at drug receptors). The most important adverse drugdrug interactions occur with drugs that have serious toxicity and a low therapeutic index, such that relatively small changes in drug level can have significant adverse consequences. Additionally, drugdrug interactions can be clinically important if the disease being controlled with the drug is serious or potentially fatal if undertreated.

Pharmacokinetic DrugDrug Interactions

Drugs may interact at any point during their absorption, distribution, metabolism, or excretion; the result may be an increase or decrease in the concentration of drug at the site of action. As individuals vary in their rates of disposition of any given drug, the magnitude of an interaction that alters pharmacokinetic parameters is not always predictable, but it can be very significant.

The delivery of drug into the circulation may be altered by physicochemical interactions that occur prior to absorption. For example, drugs may interact in an intravenous solution to produce an insoluble precipitate that may or may not be obvious. In the gut, drugs may chelate with metal ions or adsorb to medicinal resins. Thus, Ca²⁺ and other metallic cations contained in antacids are chelated by tetracycline, and the complex is not absorbed. Cholestyramine adsorbs and inhibits the absorption of thyroxine, cardiac glycosides, warfarin, corticosteroids, and probably other drugs. The rate and sometimes the extent of absorption can be affected by drugs that alter gastric motility, but this is usually of little clinical consequence. Interactions within the gut may be indirect and complex. Antibiotics that alter the gastrointestinal flora can reduce the rate of bacterial synthesis of vitamin K such that the effect of oral anticoagulants, which compete with vitamin K, will be enhanced. If a drug is metabolized by the gastrointestinal microorganisms, antibiotic therapy may result in an increase in the absorption of the drug, as has been demonstrated for some patients receiving digoxin (Lindenbaum et al., 1981).

Recently, it has become evident that a number of drugs are substrates for various promiscuous transport systems that are present in many cells. P-glycoprotein (PGP) is the best studied of these systems, but many other systems are being discovered, such as the family of organic anion transporter systems. PGP is present in intestinal cells, renal tubular cells, biliary canalicular cells, and cells making up the blood-brain barrier. In the gut, PGP pumps drug into the lumen and thereby limits absorption. In the blood-brain barrier, PGP eliminates drug from the central nervous system (CNS), thus altering drug distribution. In the liver and kidney, PGP transports drug into the biliary canalicula and tubular lumen, thereby enhancing drug elimination. Inhibition of PGP therefore can alter the absorption, distribution, and elimination of drugs and is a topic of much current investigation. Cyclosporin A, quinidine, verapamil, itraconazole, and clarithromycin are examples of drugs that can inhibit PGP, whereas rifampin apparently can induce PGP. It is curious that inhibitors and inducers of CYP3A4 often appear to have similar effects on PGP, although this is not always true (Kim et al., 1999). Much as there has been an explosion of information in the past decade about the CYP drug-metabolizing enzymes, the next decade promises a rich yield of information on PGP and similar transport systems.

Many drugs are extensively bound to plasma albumin (acidic drugs) or -acid glycoprotein (basic drugs). In general, only unbound drug is free to exert an effect or to be distributed to the tissues. Thus, displacement of one drug from its binding site by another might be expected to result in a change in drug effects. Although such binding/displacement interactions occur, they are rarely of clinical significance. This is because the displaced drug distributes rapidly into the tissues: the larger the apparent volume of distribution of the drug, the less is the rise in the concentration of free drug in the plasma. Furthermore, following the displacement, more free drug is available for metabolism and excretion. Thus, the body's clearance processes eventually reduce the free drug concentration to that which existed prior to the drug displacement interaction. As a result, the effect of such an interaction is usually small, transient, and frequently unrecognized. However, the relationship of free drug to the total (bound plus free) drug is changed, and the interpretation of plasma drug assays that measure total drug concentration must be altered.

A few drugs are actively transported to their site of action. For instance, the antihypertensive drugs guanethidine and guanadrel inhibit sympathetic nervous system function after being transported into adrenergic neurons by the norepinephrine-uptake mechanism. Inhibition of this neuronal uptake system by tricyclic antidepressants and some sympathomimetic amines will inhibit the sympathetic blockade and reduce the antihypertensive effects of guanethidine and guanadrel. More drugs may be transported away from their site of action by PGP or other transporters. For example, cancer chemotherapy may be limited by transport of anticancer drugs out of tumor cells by PGP. Attempts have been made to block PGP in order to enhance chemotherapy, thus making use of a drugdrug interaction to enhance clinical efficacy (Krishan et al., 1997).

Interactions involving drug metabolism can increase or decrease the amount of drug available for action by inhibition or induction of metabolism, respectively (see also Chapter 1: Pharmacokinetics: The Dynamics of Drug Absorption, Distribution, and Elimination). Interactions may occur among administered drugs or between drugs and dietary substances [e.g., grapefruit juice (a CYP3A4 inhibitor)], herbal remedies [e.g., St. John's wort (a CYP3A inducer); see Fugh-Berman, 2000], or other chemicals [e.g., alcohol; other organic solvents (CYP2E1 inducers); cigarette smoke; polychlorinated biphenyls (CYP1A2 inducers)]. The effects of enzyme induction or inhibition are most obvious when drugs are given orally, because all of the absorbed compound must pass through the liver prior to reaching the systemic circulation. Additionally, the intestinal mucosa contains substantial amounts of CYP3A4, which can metabolize some drugs before they reach the portal circulation. Therefore, even for drugs that have a systemic clearance that is mainly dependent on hepatic blood flow (e.g., propranolol), the amount of drug that escapes metabolism on the first pass will be influenced by enzyme induction or inhibition. Examples of drugs that are affected by enzyme inducers are oral anticoagulants, quinidine, corticosteroids, low-dose estrogen contraceptives, theophylline, mexiletine, methadone, HIV protease inhibitors, and some -adrenergic blocking agents. Knowledge of the specific pathways of metabolism of a drug and of the molecular mechanisms of enzyme induction can help in planning studies of possible drug interactions, and preclinical drug development commonly includes studies to determine pathways of drug metabolism (Yuan et al., 1999). Thus, if a compound is found to be metabolized by CYP3A4 in in vitro studies, the potential for clinically significant interactions can be focused on studies with commonly used drugs that can either inhibit (e.g., ketoconazole) or induce (e.g., rifampin) this enzyme. Probes for the evaluation of potential drugdrug interactions by the different CYP isoforms in human beings are being developed (e.g., midazolam or erythromycin for CYP3A and dextromethorphan for CYP2D6). The example of arrhythmias triggered by a combination of terfenadine (which has been withdrawn from the market) and ketoconazole highlights the need for such studies in early drug development. In this interaction, ketoconazole inhibits the metabolism of terfenadine (by CYP3A4) to its active metabolite, resulting in high concentrations of unmetabolized terfenadine, which is toxic (Peck, 1993).

The ability of one drug to inhibit the renal excretion of another is dependent on an interaction at active transport sites. Many of the reported interactions occur at the anion transport site, where, for example, probenecid inhibits the excretion of penicillin to cause the desirable effects of elevated plasma concentrations of the antibiotic and a longer half-life. Similarly, the renal elimination of methotrexate is inhibited by probenecid, salicylates, and phenylbutazone, but in this case methotrexate toxicity may result from the interaction. Interactions at the transport site for basic drugs include the inhibition of excretion of procainamide by cimetidine and amiodarone. An interaction at renal tubular PGP causes inhibition of the excretion of digoxin by quinidine, verapamil, and amiodarone. Finally, the excretion of Li⁺ can be affected by drugs that alter the ability of the proximal renal tubule to reabsorb Na⁺. Thus, clearance of Li⁺ is reduced and concentrations of Li⁺ in plasma are increased by diuretics that cause volume depletion and by nonsteroidal antiinflammatory drugs that enhance proximal tubular reabsorption of Na⁺.

Pharmacodynamic DrugDrug Interactions

There are numerous examples of drugs that interact at a common receptor site or that have additive or inhibitory effects due to actions at different sites in an organ. Such interactions are described throughout this textbook. Frequently overlooked is the multiplicity of effects of many drugs. Thus, phenothiazines are effective -adrenergic antagonists; many antihistamines and tricyclic antidepressants are potent antagonists at muscarinic receptors. These 'minor' actions of drugs may be the cause of drug interactions.

Other interactions of an apparently pharmacodynamic nature are poorly understood or are mediated indirectly. Halogenated hydrocarbons, including many general anesthetics, sensitize the myocardium to the arrhythmogenic actions of catecholamines. This effect may result from an action on the pathway that leads from adrenergic receptor to effector, but the details are unclear. The striking interaction between meperidine and monoamine oxidase inhibitors to produce seizures and hyperpyrexia may be related to excessive amounts of an excitatory neurotransmitter, but the mechanism has not been elucidated.

One drug may alter the normal internal milieu, thereby augmenting or diminishing the effect of another agent. A well-known example of such an interaction is the enhancement of the toxic effects of digoxin as a result of diuretic-induced hypokalemia.

Summary: DrugDrug Interactions

Drugdrug interactions are only one of the many factors discussed in this chapter that can alter the patient's response to therapy. The major task of the physician is to determine if an interaction has occurred and the magnitude of its effect. When unexpected effects are seen, a drug interaction should be suspected. Careful drug histories are important, because patients may take over-the-counter drugs or herbal products, take drugs prescribed by another physician, or take drugs prescribed for another patient. Care must be exercised when major changes are made in a drug regimen, and drugs that are not necessary should be discontinued. When an interaction is discovered, the interacting drugs often may be used effectively with adjustment of dosage or other therapeutic modifications.

Fixed-Dose Combinations

The concomitant use of two or more drugs adds to the complexity of individualization of drug therapy. The dose of each drug should be adjusted to achieve optimal benefit. Thus, patient compliance is essential yet more difficult to achieve. To obviate the latter problem, many fixed-dose drug combinations are marketed. The use of such combinations is advantageous only if the ratio of the fixed doses corresponds to the needs of the individual patient.

In the United States, a fixed-dose combination of drugs is considered a 'new drug' and as such must be approved by the FDA before it can be marketed, even though the individual drugs are available for concurrent use. For such drugs to be approved, they must meet certain conditions. The two drugs must act to achieve a better therapeutic response than either drug alone, so that additional efficacy is achieved or the same effect is achieved with less toxicity (e.g., many antihypertensive drug combinations); or one drug must act to reduce the incidence of adverse effects caused by the other (e.g., a diuretic that promotes the urinary excretion of K⁺ combined with a K⁺-sparing diuretic).

Placebo Effects

The net effect of drug therapy is the sum of the pharmacological effects of the drug and the nonspecific placebo effects associated with the therapeutic effort. Although identified specifically with administration of an inert substance in the guise of medication, placebo effects are associated with the taking of any drug, active or inert.

Placebo effects result presumably from the physicianpatient relationship, the significance of the therapeutic effort to the patient, or the mental set imparted by the therapeutic setting and by the physician. They vary significantly in different individuals and in any one patient at different times. Placebo effects commonly are manifested as alterations of mood, other subjective effects, and objective effects that are under autonomic or voluntary control. They may be favorable or unfavorable relative to the therapeutic objectives. Exploited to advantage, placebo effects can significantly supplement pharmacological effects and can represent the difference between success and failure of therapy.

A placebo (in this context, better termed dummy medication) is an indispensable element of many controlled clinical trials. In contrast, a placebo has only a limited role in the routine practice of medicine. A supportive physicianpatient relationship generally is preferable to the use of a placebo for promoting therapeutic benefits. Relief or lack of relief of symptoms upon administration of a placebo is not a reliable basis for determining whether the symptoms have a 'psychogenic' or 'somatic' origin.

Tolerance

Tolerance may be acquired to the effects of many drugs, especially the opioids, various CNS depressants, and organic nitrates. When this occurs, cross-tolerance may develop to the effects of pharmacologically related drugs, particularly those acting at the same receptor site, and drug dosage must be increased to maintain a given therapeutic effect. Since tolerance does not usually develop equally to all effects of a drug, the therapeutic index may decrease. However, there also are examples of the development of tolerance to the undesired effects of a drug and a resultant increase in its therapeutic index (e.g., tolerance to sedation produced by phenobarbital when used as an anticonvulsant).

The mechanisms involved in the development of tolerance are only partially understood. Tolerance may occur as the result of induced synthesis of the hepatic microsomal enzymes involved in drug biotransformation. Another example of pharmacokinetic tolerance is the development of resistance of cancer cells to drug-induced cytotoxicity due to the induction of PGP, which transports drug out of the cell, thereby reducing the intracellular concentration of the chemotherapeutic agent. The most important factor in the development of tolerance to the opioids, barbiturates, ethanol, and organic nitrates is a type of cellular adaptation referred to as pharmacodynamic tolerance; multiple mechanisms are involved, including changes in the number, affinity, or function of drug receptors. Tachyphylaxis, such as that to histamine-releasing agents and to the sympathomimetic amines that act indirectly by releasing norepinephrine, has been attributed to depletion of available mediator, but other mechanisms also may contribute. The subject of tolerance is discussed in more detail in Chapter 24: Drug Addiction and Drug Abuse.

Genetic Factors

Genetic factors are the major determinants of the normal variability of drug effects and are responsible for a number of striking quantitative and qualitative differences in pharmacological activity. Basic principles of human genetics apply to genetic loci coding for proteins involved in handling of drugs, e.g., drug metabolizing enzymes, carrier proteins, and receptors. Thus (1) allelic variation is common; (2) there are often several different alleles producing variant proteins at a given locus; (3) some allelic variants are 'silent,' with no functional consequences, while others may markedly alter the handling of foreign compounds; (4) gene frequencies for different alleles are likely to vary among different human populations, suggesting the need for vigilance in extrapolation of kinetic and safety data from one population to another; (5) some allelic variants are classified as 'polymorphisms,' variant alleles with a frequency of at least 1%, while other, less common variants are classified as 'rare inborn errors of metabolism.' The consequences of pharmacogenetic variation include: (1) altered clearance of drugs, resulting in a 'functional overdose' in those individuals unable to metabolize the compound; (2) failure to convert a prodrug to an active drug; (3) altered pharmacodynamics (e.g., hemolytic anemia secondary to glucose-6-phosphate dehydrogenase deficiency); and (4) idiosyncratic drug reactions, such as aplastic anemia or hepatotoxicity.

The superfamily of cytochrome P450 enzymes has been extensively investigated for pharmacogenetic variants (Ingelman-Sundberg et al., 1999). For example, an abnormality in CYP2D6 (present in 3% to 10% of various populations) results in deficient metabolism of many compounds. For some of these drugsfor example, the tricyclic antidepressantstoxicity of 'standard' doses may result from accumulation when used in CYP2D6-deficient patients, while for other drugs, either because of a wide therapeutic index (e.g., dextromethorphan) or because multiple pathways are involved in clearance (e.g., propranolol), no dosage adjustment is required.

During drug development, compounds may be screened in vitro with human tissue preparations or recombinantly expressed human cytochrome P450 enzymes to ascertain if pharmacogenetic polymorphisms are likely to be involved in metabolism of the drug. Single-dose studies in subjects genotyped for various polymorphisms may help clarify whether the potential for altered drug handling is clinically relevant. For pharmacogenetics to become clinically useful, molecular diagnostic tests for pharmacogenetic variants, done in routine clinical laboratories, must become available, so that a physician can individualize choice of medication or dose regimen based on each specific patient's drug metabolism profile. If a relatively rare but severe adverse reaction to a drug (e.g., a 1 in 5000 risk of hepatotoxicity) is strongly linked to a given pharmacogenetic polymorphism, such pharmacogenetic 'prescreens' could markedly decrease the risk for individual patients and the population as a whole.

Approach to Individualization

After it has been determined that pharmacotherapy is necessary to modify the symptoms or outcome of a disease, the therapist is faced with two types of decisions: the first is qualitative (the initial choice of a specific drug) and the second is quantitative (the initial dosage regimen). Optimal treatment will result only when the physician is aware of the sources of variation in response to drugs and when the dosage regimen is designed on the basis of the best available data about the diagnosis, severity, and stage of the disease, presence of concurrent diseases or drug treatment, and predefined goals of acceptable efficacy and limits of acceptable toxicity. If objectively assessable expectations of drug therapy are not set before therapy is initiated, therapy is likely to be ineffective and continued longer than necessary unless an obvious adverse effect occurs.

In most clinical settings, the decision about the choice of drug is influenced substantially by the confidence the physician has in the accuracy of the diagnosis and estimates of the extent and severity of disease. Based on the best available information, the physician must decide on an initial drug from a group of reasonable alternatives. The extent of this evaluation is itself dependent on many factors, including a cost-benefit analysis of diagnostic tests, and this must be based on the availability and specificity of alternative therapies and the likelihood of a reduction in future utilization of expensive health care. The initial dosage regimen is determined by estimation, if possible, of the pharmacokinetic properties of the drug in the individual patient. The estimate must be based on an appreciation of the variables that are most likely to affect the disposition of the particular drug. These variables have been discussed above (see Figure 31 and Appendix II). Subsequent adjustments may be aided in some instances by measurement of drug concentrations but must ultimately be based on whether the regimen is efficacious, either without adverse effects or at an acceptable level of toxicity.

It has been stated above that every therapeutic plan is and should be treated as an experiment. As such, most of the considerations that were specified in the discussion of clinical trials must be applied to individual patients. Of utmost importance is the definition of specific goals of treatment and the means to assess whether or not these goals are being achieved. Whenever possible, the objective endpoint should be related as closely as possible to the clinical goals of therapy (e.g., shrinkage of a tumor or eradication of an infection). Many clinical goals are, however, difficult to assess (e.g., the prevention of cardiovascular complications associated with hypertension and diabetes). In such cases, it is necessary to use surrogate markers, such as a reduction in blood pressure or the concentration of glucose or cholesterol in plasma. These intermediate endpoints are based on demonstrated (in clinical trials) or assumed correlation of the surrogate marker with the ultimate clinical benefit. In many casessuch as improvement of exercise tolerance in patients with congestive heart failure, the elimination of asymptomatic ventricular arrhythmias, or the change in CD4 lymphocyte count in AIDSthe link between the surrogate marker and the ultimate goal is controversial (Fleming and DeMets, 1996).

The value or utility of each regimen must be assessed at intervals during the course of therapy. The utility of a regimen can be defined as the benefit it produces plus the dangers of not treating the disease minus the sum of the adverse effects of therapy. Another common expression of the usefulness of a regimen is its ratio of risks to benefits (representing a balance between the efficacious and toxic effects of the drug). A definitive evaluation of the utility of a drug is not easy; nevertheless, some sense of the value of a regimen must be established in the minds of the physician and the patient. Knowledge of the usefulness of a given regimen may be a critical determinant of protracted compliance by the patient to a long-term regimen or logical discontinuation by the physician of a marginally efficacious and risky therapy. It must be remembered that the physician, the patient, and the patient's family may have disparate opinions of the utility of a therapeutic regimen. In one study of antihypertensive therapy in which all patients were judged to be improved by the physician, only 48% of the patients considered themselves improved and 8% felt worse. Relatives thought that only 1% of the patients were improved and that 99% had evidence of adverse effects of therapy (Jachuck et al., 1982).

Drug Regulation

The history of drug regulation in the United States reflects the growing involvement of governments in most countries to ensure some degree of efficacy and safety in marketed medicinal agents. The first legislation, the Federal Food and Drug Act of 1906, was concerned with the interstate transport of adulterated or misbranded foods and drugs. There were no obligations to establish drug efficacy and safety. The federal act was amended in 1938, following the deaths of 105 children that resulted from the marketing of a solution of sulfanilamide in diethylene glycol, an excellent but highly toxic solvent. The amended act, the enforcement of which was entrusted to the FDA, was concerned primarily with the truthful labeling and safety of drugs. Toxicity studies were required, as well as approval of a new drug application (NDA), before a drug could be promoted and distributed. However, no proof of efficacy was required, and extravagant claims for therapeutic indications were commonly made (Wax, 1995).

In this relatively relaxed atmosphere, research in basic and clinical pharmacology burgeoned in both industrial and academic laboratories. The result was a flow of new drugs, called 'wonder drugs' by the lay press, for the treatment of both infectious and organic disease. Because efficacy was not rigorously defined, a number of therapeutic claims could not be supported by data. The risk-to-benefit ratio was seldom mentioned, but it emerged in dramatic fashion early in the 1960s. At that time, thalidomide, a hypnotic with no obvious advantage over other drugs in its class, was introduced in Europe. After a short period, it became apparent that the incidence of a relatively rare birth defect, phocomelia, was increasing. It soon reached epidemic proportions, and retrospective epidemiological research firmly established the causative agent to be thalidomide taken early in the course of pregnancy. The reaction to the dramatic demonstration of the teratogenicity of a needless drug was worldwide. In the United States, it resulted in the Harris-Kefauver Amendments to the Food, Drug, and Cosmetic Act in 1962.

The Harris-Kefauver Amendments are sound legislation. They require sufficient pharmacological and toxicological research in animals before a drug can be tested in human beings. The data from such studies must be submitted to the FDA in the form of an application for an investigational new drug (IND) before clinical studies can begin. Three phases of clinical testing (see below) have evolved to provide the data that are used to support an NDA. For drugs introduced after 1962, proof of efficacy is required, as is documentation of relative safety in terms of the risk-to-benefit ratio for the disease entity to be treated. The 1962 amendments also required manufacturers to provide data to support the claims of efficacy for all drugs marketed between 1938 and 1962.

To demonstrate efficacy, 'adequate and well-controlled investigations' must be performed. This generally has been interpreted to mean two replicate clinical trials that are usually, but not always, randomized, double blind, and placebo controlled. Safety is demonstrated by having a sufficiently large database of patients/subjects who have received the drug at the time of filing an NDA with the FDA for approval. As a result of these requirements, the number of patients on the drug, the number of studies, the development cost, and the time required for the clinical studies to complete the NDA have increased. The regulatory review time also increased as a result of the mass and complexity of the data, so that by 1990 the average review time was approaching three years. This increased the inherent tension that exists between the FDA, which is motivated to protect the public health, and the drug developers, who are motivated to market effective and profitable drugs. Competing pressures also exist in the community, where medical practitioners and patient activist groups have criticized the FDA for delaying approval, while some 'watchdog' groups criticize the FDA for allowing drugs on the market that occasionally cause unexpected problems after they are marketed. The FDA has the difficult task of balancing the requirement that drugs be safe and effective and yet allowing useful medications to be made available in a timely manner.

Beginning in the late 1980s with pressure from AIDS activists, the FDA undertook a number of initiatives that have had profound effects in streamlining the process of regulatory approval. These initiatives have all but eliminated the concern about the 'drug lag,' where drugs were available in other countries significantly sooner than in the United States (Kessler et al., 1996). First, the FDA initiated new 'treatment' IND regulations that allow patients with life-threatening diseases for which there is no satisfactory alternative treatment to receive drugs for therapy prior to general marketing if there is limited evidence of drug efficacy without unreasonable toxicity (Figure 35). Second, the agency has established expedited reviews for drugs used to treat life-threatening diseases. Congress has enacted the Prescription Drug User Fee Act (PDUFA), whereby the FDA collects a fee from drug manufacturers that is to be used to help fund the personnel required to speed the review process (Shulman and Kaitin, 1996). Finally, the FDA is becoming involved more actively in the drug-development process in order to facilitate the approval of drugs. A priority review system has been established for drugs in new therapeutic classes and drugs for the treatment of life-threatening or debilitating diseases. By working with the pharmaceutical industry throughout the period of clinical drug development, the FDA attempts to reduce the time from submission of an IND application to the approval of an NDA. This streamlining process is accomplished by the interactive design of well-planned, focused clinical studies using validated surrogate markers or clinical endpoints other than survival or irreversible morbidity. Sufficient data then will be available earlier in the development process to allow a risk-benefit analysis and a possible decision for approval. In some cases, this system may reduce or obviate the need for phase 3 testing prior to approval. Coupled with this expedited development process is the requirement, when appropriate, for restricted distribution to certain specialists or facilities and for postmarketing studies to answer remaining issues of risks, benefits, and optimal uses of the drug. If postmarketing studies are inadequate or demonstrate lack of safety or clinical benefit, approval for the new drug may be withdrawn. In 1997, these changes were codified in the FDA Modernization Act (FDAMA) (Suydam and Elder, 1999). As a result of these initiatives, the review time at the FDA has been dramatically reduced to a period of less than one year, with an ultimate goal of ten months. Details of this act, which includes a variety of other initiatives, such as those discussed above for pediatric drug development, are available on the Internet at www.fda.gov/opacom/7modact.html. These new initiatives by the FDA are based on the assumption that society is willing to accept unknown risks from drugs used to treat life-threatening or debilitating diseases. Some worry that such shortcuts in the drug approval process will result in the release of drugs without sufficient information to determine their utility and proper use. However, as long as the patient's safety can be reasonably ensured, the new plans to accelerate the drug-development process should prove beneficial to patients with such illnesses.

A seemingly contradictory directive to the FDA also is contained in the Food, Drug, and Cosmetic Actthat is, the FDA cannot interfere with the practice of medicine. Thus, once the efficacy of a new agent has been proven in the context of acceptable toxicity, the drug can be marketed. The physician then is allowed to determine its most appropriate use. However, physicians must realize that new drugs are inherently more risky because of the relatively small amount of data about their effects. Yet there is no practical way to increase knowledge about a drug before it is marketed. A systematic method for postmarketing surveillance is an indispensable requirement for early optimization of drug use.

Before a drug can be marketed, a package insert for use by physicians must be prepared. This is a cooperative effort between the FDA and the pharmaceutical company. The insert usually contains basic pharmacological information as well as essential clinical information in regard to approved indications, contraindications, precautions, warnings, adverse reactions, usual dosage, and available preparations. Promotional materials cannot deviate from information contained in the insert.

One area in which the FDA does not have clear authority is in the regulation of 'dietary supplements,' including vitamins, minerals, proteins, and herbal preparations. Until 1994, the FDA regulated such supplements as either food additives or drugs, depending on the substance and the indications that were claimed. However, in 1994, Congress passed the Dietary Supplement Health and Education Act (DSHEA), which weakened the authority of the FDA. The Act defined dietary supplement as a product intended to supplement the diet that contains '(A) a vitamin; (B) a mineral; (C) an herb or other botanical; (D) an amino acid; (E) a dietary substance for use by man to supplement the diet by increasing the total daily intake; or (F) a concentrate, metabolite, constituent, extract or combination of an ingredient described in clause (A), (B), (C), (D), or (E).' Such products must be labeled as 'dietary supplement.' The FDA does not have the authority to require approval prior to marketing of such supplements unless the supplements make specific claims relating to the diagnosis, treatment, prevention, or cure of a disease. However, the common conditions associated with natural statessuch as pregnancy, menopause, aging, and adolescencewill not be treated as diseases by the FDA. Treatment of hot flashes, symptoms of the menstrual cycle, morning sickness associated with pregnancy, mild memory problems associated with aging, hair loss, and noncystic acne are examples of claims that can be made without prior FDA approval. Also, health maintenance and other 'nondisease' claims such as 'helps you relax' or 'maintains a healthy circulation' are allowed without approval. Many supplements with such claims are labeled as follows: 'This statement has not been evaluated by the FDA. This product is not intended to diagnose, treat, cure, or prevent any disease.' The FDA cannot remove such products from the market unless they can prove that there is a 'significant or unreasonable risk of illness or injury' when the product is used as directed or under normal conditions of use. It is the manufacturer's responsibility to ensure that its products are safe.

As a result of the DSHEA legislation, a large number of unregulated products that have not been demonstrated to be safe or effective are widely available. There have been several occasions where such products have been associated with serious adverse effects or have been shown to interact with prescription drugs (see Fugh-Berman, 2000). Under these circumstances, the FDA can act, but the burden is on the FDA to prove the supplements are unsafe. In many ways this situation is analogous to the lack of regulation of drugs that existed prior to the 1938 disaster involving elixir of sulfanilamide, described above. Physicians and patients alike should be aware of the lack of regulation of dietary supplements. Adverse reactions or suspected interactions with such substances should be reported to the FDA using the same mechanisms as for adverse drug reactions (see Adverse Drug Reactions and Drug Toxicity).

Drug Development

Except for concern about governmental interference with the practice of medicine, the average physician has not considered it important to understand the process of drug development. Yet, an appreciation of this process is necessary to estimate the risk-to-benefit ratio of a drug and to realize the limitations of the data that support the efficacy and safety of a marketed product.

By the time an IND application has been initiated and a drug reaches the stage of testing in human beings, its pharmacokinetic, pharmacodynamic, and toxic properties have been evaluated in vitro and in several species of animals in accordance with regulations and guidelines published by the FDA. Although the value of many requirements for preclinical testing is self-evident, such as those that screen for direct toxicity to organs and characterize dose-related effects, the value of others is controversial, particularly because of the well-known interspecies variation in the effects of drugs. Interestingly, although many of the preclinical tests have not been convincingly shown to predict effects that are eventually observed in human beings, the risk of cautious testing of a new drug is surprisingly low.

Trials of drugs in human beings in the United States are generally conducted in three phases, which must be completed before an NDA can be submitted to the FDA for review; these are outlined in Figure 35. Although assessment of risk is a major objective of such testing, this is far more difficult than is the determination of whether or not a drug is efficacious for a selected clinical condition. Usually about two to three thousand carefully selected patients receive a new drug during phase 3 clinical trials. At most, only a few hundred are treated for more than 3 to 6 months, regardless of the likely duration of therapy that will be required in practice. Thus, the most profound and overt risks that occur almost immediately after the drug is given can be detected in a phase 3 study if these occur more often than once per 100 administrations. Risks that are medically important but delayed or less frequent than 1 in 1000 administrations may not be revealed prior to marketing. It is thus obvious that a number of unanticipated adverse and beneficial effects of drugs are detectable only after the drug is used broadly. The same can be more convincingly stated about most of the effects of drugs on children or the fetus, where premarketing experimental studies are restricted. It is for these reasons that many countries, including the United States, have established systematic methods for the surveillance of the effects of drugs after they have been approved for distribution (Brewer and Colditz, 1999).