Chapter 103. Critical Care Pharmacology

• Critical care therapeutics should be individualized to maximize therapeutic effect while minimizing the potential for adverse drug reactions.
• The appropriate loading dose is determined primarily by the volume of distribution of the drug in the patient.
• The maintenance dose is proportional to the clearance and the desired steady-state plasma concentration.
• Elimination half-life is inversely proportional to clearance and directly proportional to volume of distribution.
• Steady-state conditions are obtained after the passage of three to five half-lives.
• Prospective consideration of possible drug-patient, drug-disease, and drug-drug interactions minimizes the potential for undertreatment or adverse drug reactions.
• Therapeutic drug monitoring may follow purely pharmacodynamic parameters or additionally use plasma levels to calculate pharmacokinetic parameters.
• Therapeutic drug monitoring attempts to ensure adequate therapy and to prevent, detect, and appropriately report adverse drug reactions.
• Systemwide changes in management of critically ill patients, including physician order entry systems and dedicated intensivists and pharmacists, can potentially decrease the incidence of adverse drug reactions.

Individualization of critical care therapeutics through the application of pharmacologic principles is intended to reconcile important features of ICU management including polypharmacy, altered drug disposition, and cost considerations, in the design of a rational drug regimen. Critically ill patients routinely receive multiple medications, and the potential for adverse drug reactions (ADRs), particularly drug-drug interactions, increases in proportion to the number of agents received. Furthermore, physiologic changes resulting from critical illness may alter several aspects of drug disposition in a manner that is often difficult to predict based on available information. Finally, rational critical care therapeutics is a major component of providing cost-effective critical care, because of the substantial fraction of the average hospital pharmacy budget consumed by critical care therapeutic agents.

Individualization of pharmacotherapy attempts to avoid ADRs caused by drug overdosage or undertreatment, including ADRs caused by drug-disease, drug-drug, and drug-patient interactions. Correct drug dosing is frequently complicated in critical illness by alterations in bioavailability, volume of distribution, and elimination.1 Errors in choosing a therapeutic regimen are frequently poorly tolerated by these patients; rapid efficacy may be necessary for survival, but physiologic reserve may also be inadequate to withstand the effects of drug intoxication. Concentrating efforts on optimal dosing of drugs that have a low therapeutic index (low ratio of toxic to therapeutic plasma level) is therefore particularly important. Consideration of pharmacokinetic (PK) and pharmacodynamic (PD) principles, and their application to design an approximate patient model for individualized therapeutics, should precede addition of any new drug to an ICU patient's regimen. Furthermore, the dynamic physiology of these patients mandates frequent reassessment of the accuracy of this model, updating drug regimens as required. Therapeutic drug monitoring is used to titrate therapy with drugs that have both a low therapeutic index and readily measurable plasma concentrations that are proportional to drug effects (therapeutic and toxic). Finally, any change in the status of an ICU patient must be considered to be potentially the result of an adverse drug reaction, especially in the absence of a clear-cut explanation.

Pharmacokinetics is the study of drug movement through the body,2–7 encompassing all aspects of drug disposition (“what the body does to the drug”), including (1) entry, (2) distribution, and (3) elimination. The three most important quantifiable PK parameters considered in evaluating the disposition of a drug are therefore (1) bioavailability (fraction of administered dose reaching the systemic circulation), (2) volume of distribution, and (3) clearance (elimination by biotransformation and/or excretion). Pharmacokinetic models divide the body into one or more compartments (see below), so that drug disposition may be described by a series of mathematical equations. Pharmacodynamics refers to the relationship of drug movement to pharmacologic response (“what the drug does to the body”). The relationship between PK data and the PD phenomenon of drug effect is based on the drug-receptor complex theory, which states that there is a correlation between drug concentrations in the plasma and at the receptor site, and that drug effect is proportional to the extent of drug-receptor complex formation.

Bioavailability

The fraction of administered drug reaching the systemic circulation is termed its bioavailability. Intravascular injection generally results in 100% bioavailability. Other parenteral routes (intramuscular, subcutaneous, transdermal, intraperitoneal, and inhalational) and enteral (oral, through enteral feeding tubes, and rectal) of administration often achieve bioavailability which is both less complete and less predictable, although selected agents administered by the sublingual, inhalational, and other routes may undergo extensive and rapid absorption. Formulation properties (e.g., elixir vs. tablet) and the physicochemical properties (e.g., lipid solubility and degree of ionization at gastric pH) of the administered drug can substantially affect systemic bioavailability, particularly following enteral administration. Enterally administered agents, except drugs administered by sublingual or rectal routes, are uniquely subjected to the potential influence of first-pass (presystemic) metabolism of drug by gut flora and by intestinal and liver enzymes (see below), which tends to decrease the systemic bioavailability of the parent compound.4,6,8 Presystemic metabolism may result in the production of metabolites that are inactive (e.g., most propranolol metabolites), or active (4-hydroxypropranolol from propranolol, morphine from codeine, or enalaprilat from enalapril) following enteral substrate administration. Agents requiring metabolism to an active metabolite to elicit pharmacologic response, such as enalapril and codeine, are commonly referred to as prodrugs. Locally administered drugs (e.g., intraperitoneal antibiotics for peritonitis or intraocular β-blocking agents for glaucoma) are intended for local effects, but it is important to remember the potential systemic absorption and systemic effects of these drugs. In addition to the factors that determine extent of absorption, drug formulation can be manipulated to alter the rate of drug absorption (e.g., standard-release vs. extended-release preparations), which will in turn potentially blunt the peak drug concentration obtained after drug administration. Important considerations in choosing the route of drug administration in the critically ill patient will be further discussed below.

Drug Distribution and Elimination

The simplest compartmental PK model describes the body as a single homogenous compartment. An amount or dose of drug (D) is administered intravenously, instantaneously distributed into a space of volume (Vd), and the immediate serum or plasma concentration (C) measured. Given these simplifications, the volume of this theoretical compartment Vd (the apparent volume of distribution) may be calculated: Vd = D/C, since C = D/Vd. By assuming instantaneous distribution and concentration sampling (before any elimination has occurred), this model describes the theoretical volume into which the administered drug must be distributed in order to produce the observed drug concentration (Fig. 103-1A). The volume of distribution of most drugs does not usually correspond to a physiologic volume (such as the extracellular fluid volume [0.2 L/kg] or total body water space [0.5 to 0.7 L/kg]), but is often many times larger than the total body volume; drugs with a large volume of distribution are usually characterized by low plasma protein binding, high lipid solubility, and extensive tissue binding.

Drugs whose disposition may be described by a one-compartment kinetic model exhibit log-linear plasma concentrations as a function of time, thus exhibiting “simple” or “linear” pharmacokinetics (see Fig. 103-1B). Since the absolute rate of drug elimination is a linear function of its plasma concentration, and the fraction of drug eliminated per unit time is constant, a “first-order” kinetic model applies. Following a single intravenous bolus (and assumed instantaneous complete distribution), plasma drug concentration declines owing to a first-order elimination process (biotransformation and/or excretion), so that a semilogarithmic plot of the logarithm of plasma drug concentration (log Cp) versus time yields a linear graph. Back-extrapolation of this graph to time zero permits estimation of the plasma drug concentration immediately following the bolus (Cp0), and thus an estimate of Vd, since Vd = D/C. The slope of this plot is called the elimination rate constant k (or ke).

The kinetics of drug elimination by this first-order process are described by the equation:

where CpT2) and Cp(T1) are the plasma drug concentrations at later and earlier measurement points, respectively; T2 − T1 is the time elapsed between these two time points; and k is the elimination rate constant (in units of time–1).

The elimination half-life (t1/2) of the drug, which is the amount of time required for the plasma drug concentration to decrease by 50%, can be determined from the elimination rate constant by the equation:

Using the above equation, it can be shown that the drug concentration declines by 50% during each half-life, so that little of the bolus dose remains after four to five half-lives (Table 103-1). Alternatively, if a drug is eliminated by a saturable process, elimination may become saturated at high concentration, so that kinetics become zero-order (a constant amount of drug, rather than a constant fraction, is eliminated per unit time). The half-life of elimination of such agents is concentration-dependent, because of capacity-limited (saturable) clearance, and PK assessment requires the use of Michaelis-Menten kinetics. Small dose escalations may result in disproportionately large plasma concentration increments when the maximum metabolic capacity is exceeded. Ethanol, phenytoin, and salicylic acid are examples of drugs with capacity-limited clearance, and many other agents develop saturation of metabolic pathways in overdose (see Chap. 102).

If the same drug given above as a bolus to estimate simple (single-compartment) PK parameters is instead administered as a continuous intravenous infusion, and the rate of infusion exceeds the rate of simultaneous (first-order) elimination, then the plasma drug concentration will increase as accumulation occurs. Eventually, since the amount of drug excreted per unit time increases with increasing plasma drug concentration, the elimination rate becomes equal to the rate of administration; the plasma drug concentration at this time, which from then onward remains constant (unless the rates of administration or elimination change), is termed the steady-state concentration (Cpss) (Fig. 103-2, upper panel). Accumulation of an agent exhibiting linear pharmacokinetics proceeds in a fashion which is the mirror image of its elimination following a single intravenous bolus (Table 103-2). Accumulation of intermittently administered drugs differs from constant intravenous infusion only in the presence of peak and trough plasma drug concentration fluctuations; steady-state still develops over the same number of half-lives, with peak-to-trough fluctuations around a mean Cpss (see Fig. 103-2, lower panel). Therefore one can use the concept of half-life to estimate the time to attainment of a steady-state plasma drug level (and pharmacologic effect) following initiation of therapy, and to offset of drug effect following discontinuation of the agent, regardless of the dosing regimen in use.

The conceptually simple linear PK profile of a drug that fits a single-compartment model is actually inadequate to fully describe the disposition of most drugs. Multicompartmental models more accurately describe the rate and extent of drug shifts between these spaces, requiring multiple slope calculations to include the effects of intercompartmental distribution and elimination from the body (Fig. 103-3).3 Such models divide the body into a central (the blood volume and highly perfused tissues, Vdc) and one or more peripheral or tissue (fat, muscle, or skin) compartments. The final volume of distribution (Vdtot) reflects the sum of central and tissue compartments. Vdc is calculated by measuring the instantaneous plasma concentration (Cp0) after an intravenous drug dose is administered (Fig. 103-4A), then Vdtot is calculated by back-extrapolation from the elimination phase of the plasma concentration versus time curve to the y-axis (concentration) at zero time (see Fig. 103-4B). The two-compartment model also involves consideration of another aspect of drug distribution: the rate of distribution of drug from the central to the final volume of distribution is quantified by the distribution half-life (αt1/2), although some of the decline in plasma drug concentration during this period often is due to simultaneous drug elimination during the initial distribution phase. The elimination half-life (βt1/2) corresponds to the half-life considered using a single-compartment approach.

Despite these refinements, the principles of linear pharmacokinetics hold for most agents in practice. Thus from a semilog plot [log concentration (y) versus time (x)], the slope is ke (the elimination rate constant) following full distribution (Vdtot), and the rate of drug clearance (CL) is based on the balance of distribution and excretion rates:

It is often incorrectly assumed that CL may be calculated from this equation. In fact, both CL and Vdtot are quantified to determine ke. In other words, CL and Vdtot determine ke: CL is a physiologic value, a measurement of the volume of plasma from which a substance is completely cleared per unit time (in units of volume per unit time, usually milliliters per minute); Vdtot is a theoretical volume determined by a postdistribution plasma concentration measurement; thus ke is a derived variable, and the rate of elimination may thus be considered to reflect the balance of drug distribution to tissues and drug elimination from the body,9 so that it is useful to consider this relationship as follows:

and, since

then

In other words, the elimination half-life of drug is prolonged by a large volume of distribution or low clearance value, and shortened by limited distribution or a vigorous clearance process.6 The measured clearance of a substance (CLTOT) usually represents the algebraic sum of several clearance processes; in many cases, this may be represented as follows:

For most drugs, it is appropriate to consider the magnitude of each of these component processes and the effects of the altered physiology of critical illness on their relative importance in achieving clearance of parent drug and metabolites. Clearance of any agent by an eliminating organ is proportional to, but cannot exceed, blood flow to the eliminating organ: CL = Q × (CA − CV), where Q is organ blood flow, and CA and CV are the drug concentrations in the arterial and venous blood supplies to the eliminating organ, respectively. These and other PK parameters relevant to the individualization of critical care drug therapeutics will be discussed below.

Pharmacodynamic Modeling

It is a basic tenet of the above approach that pharmacologic response is proportional to plasma drug concentration. Four additional determinants of pharmacologic response merit emphasis in this regard: (1) the “effect site” concept, (2) the importance of the free (unbound) plasma drug concentration, (3) interindividual variability in drug response, and (4) dose-dependent alterations in receptor activation and pharmacologic effect.

Effect Site

The effect site concept was introduced to more completely describe the relationship of drug disposition to pharmacologic response. This term refers to the tissue compartment containing the receptors that must be bound in order to elicit drug effect.10 If the effect site is located in a rapidly perfused compartment (Vdc), then intravenous drug administration will elicit immediate pharmacologic response. The antiarrhythmic effect of lidocaine is an example of such behavior, and thus initial serum lidocaine levels correlate directly with drug effect. In fact, the well-established practice of empirically administering additional lidocaine boluses following an initial load is based on the necessity of maintaining adequate serum levels, which tend to decline as the drug distributes into other tissues or is eliminated.3,11 The distribution half-life (αt1/2) of lidocaine is only 8 minutes (see Fig. 103-3), so the plasma level following an initial bolus may decline to subtherapeutic levels before a simultaneously administered continuous infusion results in drug accumulation sufficient to attain the desired steady-state plasma level (Cpss). By contrast, plasma digoxin levels do not correlate with drug effect until the postdistribution phase (2 to 4 hours following an IV bolus), because distribution to tissues is required for binding to the target tissue receptors. The usual total digoxin loading dose of 1 mg is generally administered in smaller fractions (typically 0.5 mg first, with 0.25 mg following at 6 and 12 hours later), because the full effects of each loading dose fraction cannot be assessed until tissue distribution has occurred, at which time the loading procedure can continue, if it is necessary and safe to do so.

Free or Unbound Plasma Drug Concentration

Only free (unbound) drug is available for receptor binding, so that alterations in plasma protein binding may profoundly affect the interpretability of measured total (free plus bound) plasma drug concentrations. For example, plasma protein binding of phenytoin is diminished by either hypoalbuminemia or uremia, the former owing to lesser absolute availability of albumin binding sites, and the latter thought to be a result of competitive occupancy of these sites by “uremic substances.” As a result of this phenomenon, total plasma phenytoin concentrations are often misleading in such patients. Phenytoin is normally 90% plasma protein-bound, so that the usual total plasma concentration range is 10 to 20 mg/L, and the free level is 1 to 2 mg/L. If plasma albumin is decreased from 4 g/dL to 2 g/dL, then an apparently therapeutic total plasma phenytoin level of 16 mg/L may be associated with a toxic free phenytoin level of over 7 mg/L; this is because the unbound or free fraction increases from the customary 10% toward 50%. It is therefore wise to monitor plasma free phenytoin levels in patients with hypoalbuminemia or renal insufficiency.

Interindividual Response Variability (Sensitivity)

Another point to note is the variability between individual responses to a given plasma drug level; sensitivity to drug action may differ between individuals or groups. For example, elderly subjects are more sensitive to the sedative effects of benzodiazepines (and many other sedatives) than younger subjects;12 American subjects of Chinese descent are twice as sensitive as American white men to the β-blocking effects of propranolol;13 black subjects are less sensitive to the vasodilatory effect of isoproterenol than are white Americans.14 Recent evidence implicates pharmacogenomics as the cause of variability of drug effect between races, and even individuals within races.15

Dose-Dependent Effects

The predominant drug-receptor complex mediating a drug's effect may change according to drug concentration, resulting in variable pharmacologic responses at different plasma levels. For example, dopamine predominantly activates dopaminergic receptors (and causes mesenteric and renal vasodilation) only at infusion rates of up to 2 μg/kg per minute. β1-Adrenergic (inotropic) and α-adrenergic (vasoconstrictor) activation occur at doses as low as 2 to 4 μg/kg per minute and 5 to 10 μg/kg per minute, respectively; as a result, the hemodynamic effect of dopamine infusion changes with increasing dose.15 Despite such concerns, we expect the attainment of drug levels in the usual therapeutic range to elicit the desired effect in most patients.

Most drugs used in the ICU are dosed according to broadly applicable population-based PK and PD parameters and titrated to pharmacologic response only if this is readily quantifiable, rather than being subjected to rigorous PK modeling with therapeutic drug level monitoring. The latter approach is applied to drugs that have a low therapeutic index and plasma drug concentrations that correlate with drug effect. The intermediate-complexity approach of monitoring physiologic parameters (PD monitoring) of drug effect is more readily applied in the ICU (and operating room) than in other patient care settings because of the routine use of monitoring devices. Sedation, neuromuscular paralysis, seizure suppression, diuretic agent action, and the effects of cardiovascular drugs (antiarrhythmic, chronotropic, inotropic, vasodilator, and vasoconstrictor actions) and bronchodilators are commonly assessed in this fashion, in conjunction with formal PK monitoring using drug levels for appropriate agents.

What follows is a framework for drug administration based upon physiologic modeling of the patient and application of known drug characteristics to this patient model, in order to optimize ICU therapeutics and minimize the potential for iatrogenic adverse events. Furthermore, this approach includes routine surveillance to prevent adverse drug reactions, and to assess the potential for such a phenomenon to be the underlying cause of a change in patient status.

The PK and PD principles outlined above may be applied routinely to attempt optimal design of drug regimens for ICU patients. As shown in Fig. 103-5, focused consideration of PK and PD parameters to answer a series of questions can provide a framework for therapeutic individualization. Individualization of drug therapy requires consideration of the effects of multiple factors responsible for variability in drug response (aside from disease severity); these include: (1) drug-patient interactions (including body habitus, age, gender, and race); (2) drug-disease interactions; and (3) drug-drug interactions. This approach only yields an approximation of ideal therapeutics, because information addressing drug disposition in this population is often incomplete, and its interpretation is further complicated by the dynamic physiology of critical illness. This method simply attempts to maximize the potential for rational therapy based upon available information. As a result, any drug regimen in critically ill patients requires early and frequent reassessment. The customary prudent approach to drug dosing in settings in which altered disposition is anticipated but poorly quantifiable, embodied in the phrase “start low and go slow,” is generally inappropriate in this setting, in which immediate therapeutic effect may be vital. Nevertheless, a rational approach to critical care therapeutics should reduce the incidence of iatrogenic pharmacotherapeutic complications of critical illness.

The key principles underlying this approach, which relates PK parameters to choices of drug regimen, may be summarized as follows:

1. The appropriate loading dose is determined by estimation of the volume of distribution of the agent in the patient; the regimen for administering this dose is chosen based upon the desired effect site and the distribution half-life for the agent.

2. The maintenance dose is chosen to equal the elimination rate of the agent at steady state (the product of CL × Cpss).

3. Steady-state plasma drug level and stable drug effect are not obtained until three to five half-lives (see Table 103-2 and Fig. 103-2) have passed.

4. Many adverse drug reactions are predictable and preventable by individualization of therapeutics and by consideration of known or potential drug interactions.

Patient characteristics are used to estimate physiologic and pathophysiologic variables affecting drug disposition. Several patient characteristics should be routinely considered, including age, gender, drug allergies, body habitus, volume status, plasma protein concentrations, and parameters of organ function (gastrointestinal tract, circulatory, renal, and liver function).

Age, gender, body habitus, and volume status are variables used to estimate expected parameters of drug disposition, based upon population PK/PD data. For most dosing calculations, the lesser of lean (or ideal) and actual body weight is used. Lean body weight (LBW) is calculated from the patient's height: For males, LBW (kg) = 50 + (2.3 × each inch above 5 feet). For females, LBW (kg) = 45.5 + (2.3 × each inch above feet). Some drugs are dosed according to actual weight or to patient body surface area, the latter obtainable from published nomograms using height and weight.

Using these data, we approach the design of individualized drug therapy for ICU patients as follows:

Route of Drug Administration

Oral bioavailability of the agent may be inadequate to achieve systemic effect owing to luminal conditions (alkaline pH or tube feeds) or first-pass metabolism (by luminal bacteria, intestinal enzymes, or hepatic enzymes). Intravenous drug administration is often preferred in the ICU, even for administration of highly bioavailable drugs, for several reasons. Rapid onset and offset of effect may be desirable to rapidly initiate therapy, while retaining the capacity for titration. Gastrointestinal dysfunction (ileus or bowel wall edema) may also result in unpredictable drug delivery by the enteral route, although there is no evidence (in subjects with decompensated congestive heart failure, renal failure, or cirrhosis) that this theoretical concern is actually relevant.17 Formulation properties (e.g., extended-release preparations) do not affect the extent of absorption, but rather the rate of absorption and potentially the peak drug concentration after each dose.

Loading Dose

Many therapies initiated in the ICU are intended to have a rapid onset of effect. Whether administered by continuous intravenous infusion or by intermittent intravenous bolus or oral dosing, plasma drug concentration and therapeutic effect will not reach steady-state levels until three to five half-lives have passed; such a delay may be unacceptable, particularly for drugs which have a prolonged half-life (which may be a normal feature of a drug's disposition [e.g., digoxin], or caused by organ dysfunction and impaired elimination [e.g., aminoglycosides with renal failure]). Conversely, drugs that have a very short half-life (e.g., nitrovasodilators, esmolol, atracurium, and propofol) may achieve a rapid therapeutic effect when administered by infusion but without a loading bolus, since three to five half-lives pass in a matter of minutes. Administration of a loading dose, calculated as follows, rapidly achieves the desired target level (but does not alter the time required to attain steady-state conditions):

where Cp is the desired plasma concentration.

The appropriate loading dose is primarily determined by the distribution characteristics of the drug (Vd) and the body habitus and volume status of the patient. Usual volume of distribution values are readily available for most drugs; however, the effects of alterations in body habitus, nutrition, and volume status are much more difficult to quantify. Estimates of drug distribution characteristics are further confounded in critically ill patients by dynamic hypercatabolic losses of fat and lean body mass, accompanied by massive third-space volume retention or intravascular volume depletion, often associated with development of hypoalbuminemia and relative hyperglobulinemia. Following a period of marked positive fluid balance in a hospitalized patient, the weight increment may be used to estimate the total body water increase. Otherwise, estimates of altered body water content are usually empiric, although therapeutics in some specific disease states have been studied in sufficient detail to provide useful information. For example, a decreased aminoglycoside loading dose is required in volume-depleted acute renal failure. Furthermore, “uremic substances” that accumulate in renal failure appear to displace some drugs from tissue-binding sites, thus reducing their Vd regardless of volume status. As a result, the appropriate digoxin loading dose is decreased by 50% in the presence of renal failure; Vd is also decreased for methotrexate and insulin in renal failure patients. Conversely, edematous states such as sepsis,18,19 cirrhosis, nephrosis, and congestive heart failure often increase Vd. Determination of the ideal or lean body weight is difficult in the presence of obesity, cachexia, or combinations of the above factors, such as in a typical hypercatabolic patient with sepsis, acute renal failure, and postresuscitation volume overload. Loading dose regimens may also be altered based upon the relationship between the desired onset of effect to the distribution pattern at the effect site, as discussed above in contrasting the standard loading dose regimens for lidocaine and digoxin.

Volume of distribution for a given drug is reported in units of volume per weight in a normal patient. In the critically ill patient, the alterations in weight and body fluid compartments will alter the distribution of the drug. A reasonable approach to approximate these alterations in the calculation of the loading dose of hydrophilic drugs is to adjust the normal volume of distribution in proportion to the estimation of the patient's water compartment. It is reasonable to assume that water comprises approximately 10% of adipose tissue weight (estimated as actual body weight minus lean body weight). Additionally, the net weight gain secondary to fluid resuscitation will contribute to the volume of distribution for water-soluble drugs. Mathematically, we can express this by the following:

Maintenance Dose

Administration Regimen

In the critical care setting, the administration regimen usually consists of a choice between intermittent intravenous bolus therapy and continuous intravenous infusion. Factors considered in this decision include: drug characteristics (half-life and therapeutic index), patient characteristics, the desired pharmacologic effect, and cost/staffing considerations. When drugs with a low therapeutic index are dosed intermittently, the fluctuations (peak-to-trough) of plasma concentration may require formal monitoring of plasma concentrations and PK parameter estimates to ensure adequate therapy without toxicity. Administration by infusion eliminates the peak-to-trough plasma concentration fluctuations associated with intermittent parenteral boluses, which may be accompanied not only by failure to achieve continuous therapeutic effect, but also by activation of rebound counterregulatory effects during trough periods (negating prior and subsequent drug action). Continuous intravenous infusion may thus improve therapeutic efficacy of some agents. Loop diuretic agent infusions have been reported to augment sodium excretion compared to equivalent intermittent dosing, probably because of a combination of effects: increased cumulative renal tubular diuretic agent exposure (the product of time and concentration) and avoidance of periods of physiologic rebound salt conservation.20,21 Since diuretic effect onset is delayed when using only continuous intravenous infusion (until drug accumulates; see Fig. 103-2), the ideal regimen to maximize natriuresis may include an initial bolus dose to achieve the required luminal threshold drug concentration and induce immediate natriuresis. Likewise, gastric acidity is also far better controlled by continuous intravenous infusion of H2 blockers than by equivalent intermittent dosing.22

Continuous infusion of short-acting agents may also be desirable to allow titration of effect; nitrovasodilators, esmolol, propofol, and atracurium may be used for optimal control of vasodilation, β-blockade, sedation, or neuromuscular paralysis, respectively. It is widely assumed that continuous infusion of agents that have a short elimination half-life guarantees rapid reversal of drug effect following cessation of the infusion, but various factors may retard offset of effect, as is the case for reversal of sedation using agents administered by continuous infusion.23,24 Potential explanations for such alterations in drug disposition or response during continuous infusion compared to intermittent bolus therapy include compartmentalized tissue distribution, accumulation of active metabolites, or saturation of clearance mechanisms. Intermittent bolus administration titrated to specific sedation parameters is less likely than continuous infusion to result in undetected drug or metabolite accumulation if excretory mechanisms deteriorate or become saturated, unless a routine assessment of time to awakening is performed daily in patients receiving continuous infusion. Clinically, it has been shown that daily interruption of sedation of mechanically ventilated patients results in decreased duration of ventilation, likely due to the minimization of drug accumulation.25 Conversely, tolerance to the effects of several drugs (a pharmacodynamic phenomenon) occurs if a drug-free interval cannot be included in the administration regimen, requiring escalating dosages of agents such as nitroglycerin, dobutamine, and opiate analgesics to maintain a therapeutic effect. Finally, the convenience for nursing staff of administering agents by continuous infusion rather than intermittent bolus translates into decreased staffing expenditures.

Clearance

Clearance includes all processes that eliminate the drug from the body—both excretion and biotransformation. Because the total body clearance of a drug involves the actions of multiple organ systems, the estimation of the predominant rate of elimination and route of elimination is often complicated, and warrants further discussion.

What Is the Predicted Elimination Rate of the Drug?

The predicted elimination of the drug usually corresponds to the drug administration regimen that elicits an optimal therapeutic response in most subjects. Agents with a low therapeutic index may be subjected to therapeutic drug monitoring, aiming for a maintenance dose equaling the product of CL × Cpss. The desired Cpss is selected based on the therapeutic response required (e.g., target plasma lidocaine level for suppression of ventricular arrhythmias) and the clearance rate is estimated based on published data (usually obtained from healthy patients). At steady state, the rate of administration (Ra) equals the rate of excretion (Re). Ra is dose (mg) divided by interval (minutes), and Re is CL (mL/min) × Cpss (mg/mL).

What Are the Predominant Routes of Elimination of the Parent Drug and Its Metabolites (Particularly Those That Are Pharmacologically Active or Toxic)?

Renal insufficiency (Table 103-3), hepatic disease (Table 103-4), or circulatory dysfunction (Table 103-5) may affect clearance of parent drug or metabolites. There are several well-known examples of drugs with metabolites that are pharmacologically active or even toxic. Accumulation of active or toxic metabolites in the presence of renal failure is probably the most common clinical scenario in which this feature of drug disposition is important (Table 103-6). Nonrenal (usually hepatic) elimination of parent drug or metabolites has been increasingly documented to be quantitatively important in subjects with renal impairment (as discussed below). Likewise, renal drug or metabolite elimination assumes an increased role in subjects with liver disease.

Is a Dose Reduction or Escalation Required, Owing to Impaired (Renal, Hepatic, or Circulatory Dysfunction) or Augmented (Induction of Metabolism or Extracorporeal Drug Removal) Clearance of the Drug or Its Metabolites?

As outlined below, glomerular filtration rate (GFR) may be estimated routinely to a reasonable approximation, and the effects of renal dysfunction on drug clearance may be estimated with some degree of precision (see Table 103-3). The effects of varying levels and etiologies of hepatic and circulatory dysfunction (see Tables 103-4 and 103-5) on drug disposition are far more difficult to predict. Estimation of renal and hepatic clearance functions and factors that alter drug metabolism and excretion will be further discussed below. Biliary, pulmonary, cutaneous, and extracorporeal elimination may be important for clearance of some specific agents and will not be discussed in detail here. For further information regarding drug clearance by extracorporeal devices, see Chap. 102; as a general rule, an increase in drug clearance by 30% or more is regarded as significant. Dialyzability by hemodialysis (HD) or peritoneal dialysis is suggested by water solubility, low molecular weight (<500 Da; up to 5000 Da with high-flux membranes), low protein binding (<90%), small volume of distribution (Vd <1 L/kg), and a low intrinsic clearance (<500 mL/70 kg per minute).3 HD clearance is additionally affected by the porosity and surface area of the membrane used, and the blood pump and dialysate flow rates. Drug clearance by hemofiltration (slow continuous ultrafiltration, continuous arteriovenous hemofiltration, or continuous venovenous hemofiltration) may be achieved by either transmembrane sieving (convection) or membrane drug adsorption (e.g., it requires 20 mg of aminoglycosides such as gentamicin or tobramycin, which are polycationic to saturate each new AN69 hemofilter, which is anionic); the addition of diffusive clearance by use of countercurrent dialysate flow (continuous arteriovenous hemodiafiltration or continuous venovenous hemodiafiltration) augments small-solute clearance (since the capacity of these substances to cross the membrane is limited by the concentration gradient, not particle size). Data regarding altered drug disposition (changes in Vd or clearance) induced by extracorporeal membrane oxygenation (ECMO) or plasmapheresis are available only for agents which have been specifically studied (e.g., aminoglycosides, opiates, and phenytoin), and in the case of ECMO, studies have been performed almost exclusively in pediatric patient populations.

Renal Excretion

Renal clearance of drugs or metabolites is usually achieved by glomerular filtration, and diminished clearance owing to renal insufficiency is therefore proportional, caused by a decline of GFR. Renal blood flow (RBF) averages approximately 1300 mL/min, which is about 20% to 25% of cardiac output. Renal plasma flow (RPF, 650 mL/min) is about 50% of RBF, and 20% of RPF undergoes glomerular filtration, so that GFR averages 130 mL/min. The remaining 80% of RPF circulates in peritubular capillaries, where constituents may be secreted into or reabsorbed from renal tubules. Only 1.8 L of urine is produced from the 190 L per 24 hours of daily glomerular filtrate (130 mL/min × 1440 minutes), because water and most solutes are predominantly reabsorbed (e.g., the fractional excretion of sodium in a stable subject is 1%, thus 99% of filtered sodium is reabsorbed).

Glomerular filtration of plasma constituents is primarily limited by size (≤50,000 Da), water solubility, plasma protein binding (only free drug is filtered), and volume of distribution (extensively tissue-bound substances are less likely to be renally excreted).4,6 Some drugs also undergo significant renal tubular secretion or reabsorption (passive or active). Passive reabsorption of weak acids or bases from the renal tubular lumen may be influenced by urinary alkalization or acidification, respectively. Active tubular secretion by proximal tubular cationic or anionic pumps may be subject to competitive inhibition, resulting in decreased renal clearance and prolonged plasma half-life of the lower affinity substance. For example, probenecid inhibits tubular penicillin secretion, prolonging penicillin elimination half-life. Similarly, trimethoprim and cimetidine inhibit renal tubular creatinine secretion, reducing creatinine clearance and elevating serum creatinine, without affecting actual GFR (see below).

The plasma clearance of creatinine (creatinine clearance; CrCl) provides a close approximation of GFR, because creatinine is produced from muscle at a constant rate and freely filtered (it has a molecular weight of 113 Da, is water soluble, and is not protein bound). The amount of creatinine filtered (the product of the GFR and the plasma creatinine, PCr) is (theoretically) equal to the amount of creatinine excreted during the same period (the product of the urine creatinine concentration UCr, and the urine flow rate V); thus, GFR = (UCr × V)/PCr. Measurement of CrCl requires a timed urine collection to quantify the urinary excretion rate and a midpoint plasma creatinine sample (PCr). CrCl normally slightly overestimates GFR because some plasma creatinine is also secreted by renal tubules, augmenting the measured clearance value beyond that due to filtration. This effect is magnified with the development of progressive glomerular disease, as hypersecretion of creatinine by remnant tubules accounts for an increasing fraction of declining serial CrCl values.26 In critically ill patients, collection of a urine sample during a period of stable renal function with a steady plasma creatinine value may not be possible; short (0.5- to 4-hour) collections have been used in an attempt to overcome this problem. Inulin clearance more closely approximates GFR, but is impractical for clinical use, since determination requires inulin infusion, and the assay is difficult to perform. Other clearance techniques are equally impractical for routine clinical use, and more complex devices measuring real-time changes in renal function are not widely available for clinical applications at this time.27

The steady-state serum creatinine concentration is commonly applied to the Cockcroft-Gault formula to estimate GFR, without urinary collection, because of the relationship between patient age and weight (muscle mass), serum creatinine value, and CrCl.28 For males:

For females multiply the above value by 0.85.

Confounding factors include: requirement of a steady-state serum creatinine value (SCrss), estimation of ideal body weight, the empiricism of the age and gender estimates used in development of the formula,29–31 and the effects of some medications on tubular creatinine secretion.32 It is particularly difficult to account accurately for the effects on the estimated GFR calculation of diminished muscle mass (and thus creatinine generation) in the elderly, or in patients with cirrhosis, spinal cord injury, cachexia, or other causes of muscle wasting. Nevertheless, a rough estimate of the current GFR is obtained and should be used to guide dosing calculations.33–35 In fact, one study found that this formula, using a “corrected” serum creatinine concentration of 1 mg/dL in cachectic patients, provided a more accurate GFR estimate in critically ill patients than either short (half-hour) or long (24-hour) creatinine clearance measurements.35 The Modification of Diet in Renal Disease (MDRD) Equation offers another method to approximate GFR; the equation is found to be more accurate than the Cockcroft-Gault equation and takes into account more clinical factors, such as serum albumin and blood urea nitrogen, that likely reflect the patient's clinical status more accurately, but is much more complicated to use.36 Websites are available (such as www.nephron.com) to facilitate the use of the MDRD Equation. Another interesting and accurate approach uses measured aminoglycoside clearance, which occurs entirely by glomerular filtration, as a surrogate measurement of GFR.37

Following GFR estimation, dosage is usually adjusted based upon published criteria for the agent in question. These maintenance dose adjustments for renal insufficiency are made as follows:38–41

(i.e., percentage increase in interval or percentage decrease in dose). The dosing interval may be increased (interval extension), the size of individual doses decreased (dose reduction), or a combination of both approaches may sometimes be necessary.

For the interval extension method, the usual size dose is given, but the intervals between individual doses are lengthened. This method is useful for drugs characterized by long plasma half-lives in the presence of renal impairment and a wide therapeutic range. This approach is convenient but results in large plasma concentration fluctuations between peak and trough levels; in drugs with a low therapeutic index, toxic or subtherapeutic levels may result. For example, aminoglycosides are ideally suited to such a dosing strategy: The peak plasma concentration correlates with therapeutic efficacy, but trough levels must be monitored and kept low to minimize toxicity.

Alternatively, in drugs with a low therapeutic index, and those for which a constant level is preferred, the size of individual doses is reduced. This achieves a more constant plasma concentration, with less peak-trough fluctuation; however, increased toxicity because of higher average trough levels may result. For example, antiseizure drugs must be dose-adjusted in this fashion.

In the presence of renal insufficiency, hepatic metabolism accounts for the bulk of nonrenal elimination of drugs that are normally renally excreted. Hepatic biotransformation of some drugs is altered in patients with renal insufficiency; the cause of this phenomenon is unknown. Most available data come from studies of patients with chronic renal insufficiency, including those with end-stage renal disease.39 Unfortunately, the effect of acute renal failure on hepatic drug biotransformation is little studied, and specific studies in this population have yielded some unexpected results. Normally, about 30% of vancomycin is cleared by nonrenal routes (approximately 40 mL/min). Nonrenal vancomycin clearance is decreased to 6 mL/min in patients with chronic renal failure on maintenance dialysis. The elimination pattern in acute renal failure (ARF) is more complex. Nonrenal elimination is approximately halved early in the course of ARF (16 to 17 mL/min), and decreases to values closer to those of subjects with chronic renal insufficiency (CRI) (10 mL/min) after 1 week of ARF.43 This phenomenon requires an appropriate dosage adjustment, which is facilitated in this case by the routine availability of plasma level monitoring. Conversely, nonrenal clearance of imipenem in acute renal failure is nearly double that found in chronic renal insufficiency (95 mL/min vs. 51 mL/min), approaching that found in patients with normal renal function (130 mL/min; comprising over half of normal total clearance of approximately 230 mL/min). As a result, the total clearance of imipenem in anuric ARF patients is 108 mL/min, compared with 64 mL/min in CRI, and 230 mL/min in normal subjects. Therefore, the daily dose requirement is appropriately reduced from 2000 to 4000 mg/d hours in normal subjects to 1847 mg/d in ARF patients—a near doubling of the dose that would be administered based on CRI data (1111 mg/d).44 In this case, the risk of seizure because of impaired clearance and imipenem accumulation is balanced against the risk of undertreatment, but accurate dosing would not be achieved based solely on clearance data from CRI patients, and plasma drug levels are not routinely available for clinical use in most institutions. Acute renal failure data should be used whenever possible; absent this information, CRI data must be used, but with an appropriate cautionary approach.

Hepatic Biotransformation

Most drugs are lipophilic, so that they tend to be reabsorbed across renal tubular or intestinal membranes following glomerular filtration or biliary excretion, respectively. These drugs must therefore undergo biotransformation to more polar/hydrophilic compounds to allow urinary or biliary excretion. Drug biotransformation reactions are classified as phase I (functionalization reactions, exposing or introducing a functional group), or phase II (biosynthetic) reactions. Phase I reactions are predominantly oxidations, reductions, and hydrolysis reactions, and are usually catalyzed by enzymes of the cytochrome P450 (CYP450) system, located in the endoplasmic reticulum (Tables 103-7, 103-8, and 103-9). Phase II reactions are conjugations, catalyzed by a variety of (mainly cytosolic) enzymes, which covalently link the parent drug or metabolite to a variety of compounds. Biliary excretion of a conjugated compound may result in enterohepatic recirculation of the parent compound if intestinal flora cleave the conjugate bond. Both phase I and II reactions generally result in a loss of pharmacologic activity, although more active metabolites are less commonly produced. Hepatic drug metabolism is functionally characterized by two patterns of clearance, which may be flow-limited or capacity-limited. The hepatic extraction ratio (ER) is derived as follows: ER = (CA – Cv)/CA, where CA and CV are the drug concentration in hepatic arterial (or portal venous) and hepatic venous blood, respectively. The hepatic ER is low for capacity-limited drugs (which have saturable biotransformation pathways) and high for flow-limited drugs (the liver essentially metabolizes any drug delivered, a process limited only by drug-delivering blood flow).

The bulk of drug biotransformation is performed by hepatic enzymes, and the remainder by renal tubular, intestinal, cutaneous, and pulmonary enzymes. The CYP450 system, comprising at least 12 families (in humans) of heme-containing endoplasmic reticulum enzymes, is the major catalyst of hepatic drug biotransformation reactions (see Table 103-9).5,45–48 Three families (CYP1, CYP2, and CYP3) are responsible for essentially all drug biotransformations, through phase I (primarily oxidative) reactions. Cytochrome P450 nomenclature is based on amino acid sequence homology. Families (designated by numbers, such as CYP3) contain at least 40% homology; subfamilies (designated by capital letters, such as CYP3A) are over 55% homologous; individual P450 enzymes are again designated by Arabic numerals (for example, CYP3A4). The CYP3A subfamily accounts for over 50% of phase I drug metabolism, predominantly by the CYP3A4 subtype. Most other drug biotransformations are performed (in decreasing order of frequency) by CYP2D6, CYP2C, CYP2E1, and CYP1A2 (Fig. 103-6).5 More than one enzyme may metabolize a particular drug, but metabolism of many drugs is dependent upon a single enzyme. Since the CYP450 nomenclature is based on structural criteria, rather than biotransformation function, isozymes of a particular class often lack any common substrates.

Phase II (conjugation) reactions form a covalent linkage between a drug's functional group and one of a number of compounds: glucuronic acid, sulfate, glutathione, acetate, or amino acids. These conjugates are highly polar, usually inactive, and undergo urinary or fecal excretion.46,49

Drug biotransformation may be enhanced or impaired by multiple factors, including age, gender, enzyme induction or inhibition (see Table 103-7), pharmacogenetics, and the effects of hepatic dysfunction or other disease states (including those which decrease hepatic perfusion). Conditions that impair drug biotransformation may result in type A adverse drug reactions, caused by accumulation of toxic concentrations of parent drug or metabolites (see section on adverse drug reactions, below).

Effects of Age

CYP1A2 activity is increased in children compared to adults, causing the well-known increased theophylline dosage requirements in this population.50,51 Similarly, CYP3A4 activity appears to decline in the elderly compared to younger adults,52–54 although age-related declines in hepatic size, hepatic blood flow, or drug binding and distribution may underlie this phenomenon, because in vitro enzyme activity is unchanged with age.55

Effects of Gender

Differences in pharmacokinetic and pharmacodynamic properties of drugs between men and women are more commonly being recognized. These differences can sometimes have major clinical impact, or be as subtle as gender-related differences in metabolism of specific stereoisomers of a particular drug formulation. Differences in rate of absorption and extent of first-pass metabolism have been reported for some drugs, including ondansetron56 and zolmitriptan,57 but these generally lack major clinical significance. Gender differences in volume of distribution, after adjustment for weight effects, have also been identified for a number of drugs. Theophylline exhibits a smaller volume of distribution in females compared to males,58 as do the fluoroquinolones.59 Possible explanations include differences in body composition between men and women, physiologic changes associated with menstrual cycles, and differences in plasma protein binding secondary to hormonal characteristics. Gender differences in drug elimination have been studied in hepatic and renal processes. Clinically significant differences were predominantly linked to gender-specific expression of metabolic enzymes (e.g., CYP3A4 and CYP1A2); gender differences in renal handling of drugs tend to be clinically silent. Pharmacodynamic variability in humans is more difficult to measure, because the measured parameters can often be subjective, such as pain or depression. Nonetheless, relevant gender differences have been identified. Women treated with thrombolytics after a myocardial infarction are more likely to experience intracranial hemorrhage. Other examples include drugs involved in glucose management and arrhythmia treatment.

Pharmacogenetics

Many adverse drug reactions which were formerly termed idiosyncratic are now recognized to be caused by genetic polymorphisms of expression (through autosomal recessive traits) of genes involved in drug disposition and effects, including the hepatic drug metabolizing enzymes (phase I or phase II) in particular individuals and racial groups, which have been most well-studied16,61–64 (see Table 103-9; Tables 103-10 and 103-11). The identified polymorphisms of these metabolic enzymes differ in their rate of metabolism: “extensive metabolizers” are the common wild-type phenotype, “poor metabolizers” exhibit a decreased rate of substrate metabolism, and the rare “ultra-rapid metabolizers” display an increased rate of substrate metabolism. The incidence of expression of the different phenotypes varies by race and ethnic group. The effect of racial origin on drug metabolism is receiving increasing attention in the drug development process,13,65–67 although it must be remembered that pharmacogenetic factors controlling drug disposition and response vary not only according to race and gender, but also between individuals of the same race and gender.

CYP2D6 Polymorphism

The most common P450 polymorphism in Caucasians is of CYP2D6 expression. CYP2D6 is responsible for metabolism of approximately 30 to 40 commonly used drugs, about half of which are psychoactive agents (antidepressants, including all tricyclic antidepressants [TCAs] and selective serotonin reuptake inhibitors [SSRIs], and antipsychotics, including haloperidol68); other substrates include β-blockers (all except atenolol and sotalol, which are renally excreted, not metabolized), other antiarrhythmic drugs, and codeine. Approximately 7% to 10% of the Caucasian population, 2% to 5% of Africans, but only rare Asian subjects lack expression of this enzyme and are poor metabolizers of CYP2D6 substrates.66 Such patients are prone to develop profound bradycardia during standard β-blocker therapy or severe drowsiness when receiving psychoactive drug therapy; in each case the adverse event is caused by accumulation of the parent drug. Conversely, these individuals derive little analgesic effect from codeine, which must be metabolized by CYP2D6 to its more potent metabolite morphine to achieve therapeutic efficacy. Quinidine, fluoxetine, and amiodarone are potent inhibitors of CYP2D6 activity, and convert genetically extensive metabolizers to phenotypically poor metabolizers of CYP2D6 substrates.

CYP2C19 Polymorphism

The most prevalent P450 polymorphism in Asians is of CYP2C19 expression; 15% to 27% of Asians but only 3% to 5% of Caucasians lack expression of this isozyme and are poor metabolizers of substrates such as S-mephenytoin, phenytoin itself, diazepam, and omeprazole.67 Poor metabolizers of CYP2C19 have a better response to omeprazole in eradication of Helicobacter pylori than do extensive metabolizers.

However, prolonged (>1 year) exposure to high serum omeprazole levels is more likely to cause cobalamin deficiency and result in neuropsychiatric and hematologic side effects.69

CYP2C9 Polymorphism

The incidence of CYP2C9 polymorphism is up to 20% in black patients, but only 1% in Caucasians; important substrates include S-warfarin, phenytoin, diclofenac (and other hepatically metabolized nonsteroidal anti-inflammatory drugs [NSAIDs]), glipizide, and losartan. Fluconazole at usual therapeutic doses is a powerful CYP2C9 inhibitor. Clinically useful laboratory studies are being developed to determine CYP2C9 phenotype and guide therapy of drugs with narrow therapeutic indices, such as warfarin and phenytoin.

CYP2E1 Polymorphism

CYP2E1 metabolizes acetaminophen and many volatile anesthetics and is induced by isoniazid and ethanol. Interestingly, 2.4% of Caucasians and 10% of subjects with alcoholic liver disease have a CYP2E1 rapid metabolizer phenotype and metabolize CYP2E1 substrates to a greater extent than most of the general population.

CYP3A Polymorphism

The CYP3A family of enzymes displays a wide range of expression of activity for each enzyme in normal populations, contributing to the difficulty in identifying clinically significant polymorphisms. Additionally, substrates of the CYP3A family enzymes often are metabolized in a multigenic pathway. Many common substrates are metabolized by both CYP3A4 and CYP3A5; the relatively common CYP3A5 poor-metabolizer phenotype may then be clinically obscured by this dual pathway. The role of the CYP3A4 enzyme in drug elimination is further complicated by the presence of polymorphisms that alter the rate of metabolism of some substrates but not others.70 The clinical implications of these polymorphisms remain unclear.

Phase II Enzyme Polymorphisms

A genetic polymorphism for acetylation of procainamide, hydralazine, isoniazid, and sulfa drugs (see Table 103-10) underlies the “slow acetylator” phenotype seen in over 50% of American blacks and whites but only 10% of Asian individuals. Many other phase II reactions are subject to pharmacogenetic variation56 (see Table 103-10), resulting in polymorphic expression of the metabolic capacity for specific agents. Examples include glucuronidation (Gilbert syndrome), and the activities of glutathione-S-transferase (acetaminophen metabolism), thiopurine methyltransferase (azathioprine metabolism), glucose-6-phosphate-dehydrogenase (quinine-induced hemolysis), pseudocholinesterase (prolonged paralysis following succinylcholine), and dihydropyrimidine dehydrogenase (5-fluorouracil toxicity).

Enzyme Induction and Inhibition

Common substrates of the P450 enzyme responsible for metabolism of a particular drug may be inhibitors or inducers of its metabolism and therefore create the potential for drug interactions (see Tables 103-7 and 103-9). Substances that are not substrates of a P450 isozyme may also be inhibitors or inducers of its activity (e.g., fluconazole and CYP2C9, quinidine and CYP2D6). The potential for such an interaction is increasingly predictable, and consequent overdosage or underdosage preventable, owing to identification of the specific P450 isozyme responsible for drug metabolism in recent product package inserts and other standard sources (see section on general drug information reference texts, below). For example, since both cyclosporine and tacrolimus (formerly FK506) are metabolized by CYP3A4, biotransformation of both agents is predictably affected by known inhibitors and inducers of this enzyme (see Table 103-9). Ketoconazole is a particularly potent inhibitor of CYP3A4 and has been used to deliberately lower cyclosporine dosage requirements as a cost-saving measure.71,72 By contrast, phenobarbital induces cyclosporine metabolism to such an extent that concomitant use of these agents is not advisable; likewise, therapeutic cyclosporine levels are difficult to maintain during rifampin therapy. Inducers and inhibitors of phase II enzymes have been less extensively characterized, but some clinical applications of this information have emerged; for example, phenobarbital is used to induce glucuronyl transferase activity in icteric neonates, and both phenobarbital and valproate have been used to modulate chemotherapeutic agent glucuronidation and toxicity.73

Effects of Disease States

Critical illness routinely alters all the physiologic processes involved in drug disposition. It would be surprising if drug disposition and/or response were not significantly altered in such patients. Unfortunately, most data regarding drug disposition in critically ill patients must be extrapolated from other populations; clearly, such information must be interpreted with extreme caution.

Effects of Circulatory Dysfunction

The effects of congestive heart failure (CHF) on drug disposition are the best-studied examples of alterations of drug absorption, distribution, and elimination associated with circulatory insufficiency (see Table 103-5). Secondary decrements in hepatic and renal blood flow result in impaired clearance of some drugs excreted by these routes. Clearance of hepatically eliminated drugs is more likely to be impaired by CHF if their metabolism is flow-limited (i.e., characterized by a high extraction ratio, such as lidocaine). It is doubtful that most such information, even from populations with decompensated CHF, is applicable to the setting of cardiogenic shock. Even fewer data are available regarding disposition of most agents in patients with hypovolemic or septic shock, or even sepsis syndrome without frank hypoperfusion. Just as is the case in CHF, septic hypoperfusion may disproportionately involve the renal or splanchnic circulations, resulting in impairment of renal or hepatic perfusion to a degree that would not be expected based upon global hemodynamic data. Furthermore, experimental evidence suggests that endotoxemia impairs hepatic drug metabolizing-enzyme function74 and that nitric oxide probably mediates much of this inhibition.75

Effects of Liver Disease

The effects of hepatic disease on PK and PD parameters of drug disposition are much more difficult to predict than the consequences of renal disease, based on data available for most agents.76,77 Standard indices of hepatic function, such as the clinical features and biochemical values comprising the Child or Child-Turcotte scores, do not correlate even qualitatively with the capacity of the diseased liver to biotransform xenobiotics.78 Activity of particular cytochrome P450 isozymes or conjugating enzymes may be decreased, preserved, or even increased in the presence of different liver diseases at various stages of severity.17,77,79 Most data suggest that cirrhosis variably affects the hepatic content and activity of particular cytochrome 450 isozymes; contents of CYP3A are not demonstrably changed, CYP1A2 and CYP2E1 activities are decreased, and CYP2C activity may even be increased (as is tolbutamide clearance in cirrhosis).77 In addition, the presence of gastrointestinal hypomotility, hypoalbuminemia, increased or decreased plasma glycoprotein levels, ascites/edema, and altered hepatic blood flow may all alter drug absorption, distribution, elimination, and effects unpredictably (see Table 103-4). Gastric hypomotility does not alter bioavailability; rather it delays absorption and reduces the peak plasma level. Altered levels of the major drug-binding plasma proteins in cirrhosis or other forms of liver dysfunction have complex secondary effects. The frequent presence of hypoalbuminemia decreases binding of acidic drugs, such as phenytoin. Production of α1-acid glycoprotein, which binds many basic drugs (e.g., lidocaine and quinidine), is impaired by severe cirrhosis but increased by inflammatory states (it is an acute-phase reactant). The importance of decreased plasma protein binding (caused by decreased binding protein levels or by displacement by a competing substance) is determined by the hepatic extraction ratio of the drug, and the presence or absence of concomitant impairment of hepatic biotransformation capacity (caused by disease or by competition with an interacting agent). Changes in protein binding do not alter clearance of high-extraction (flow-limited) drugs, but if hepatic metabolism of a low-extraction (capacity-limited) drug is impaired and protein binding of the drug is also decreased, then the plasma free drug level will increase.80 Renal elimination of drugs or metabolites (inactive, active, or toxic) is also commonly impaired in patients with cirrhosis by the associated decrement in glomerular filtration rate, which is often unappreciated because of the poor correlation between plasma creatinine values and GFR in cirrhotic subjects.16 Finally, liver disease patients appear to have increased sensitivity to many drugs or their metabolites; some of this phenomenon is attributable to synergistic sedation (a PD phenomenon), although many such instances are probably caused by unrecognized accumulation of active or toxic metabolites or abnormally increased CNS distribution (e.g., increased blood-brain barrier permeability to cimetidine in cirrhosis).81

In practice, the presence of hepatic dysfunction should prompt a thorough drug regimen re-evaluation to examine all disposition parameters.17,76,78,82 Useful data may be available regarding the effects of cirrhosis on clearance of the agent, permitting an appropriate dose reduction. More commonly, the only information available details the predominant route of drug elimination. In either case, careful dose titration upward from a dose lower than normal is obviously the prudent approach; however, such a cautious approach may not be consistent with the rapid therapeutic effect desired in critically ill patients. Dose reduction is most likely to be necessary when a drug which usually undergoes extensive first-pass metabolism is enterally administered in the presence of liver disease severe enough to impair hepatic clearance of the agent; in such a situation, both increased oral bioavailability and decreased clearance tend to increase plasma drug levels:

Careful therapeutic monitoring of agents that have a low therapeutic index (see Table 103-8), using plasma levels if these are clinically available, represents the optimal approach in cases in which impaired clearance is suspected. Otherwise, monitoring of physiologic parameters during graded upward dose titration may be the best available option. It is important to remember that, because of impaired elimination and the resultant prolongation of half-life, achievement of steady-state plasma drug levels and corresponding maximal effect will be correspondingly delayed. Failure to wait for achievement of steady-state conditions prior to dose escalation may result in drug accumulation to toxic levels. Important drug interactions are also more likely to occur in the presence of impaired drug elimination “reserve,” so that increased vigilance regarding potential protein-binding or hepatic biotransformation interactions is warranted in the presence of liver disease (see below).

Half-Life

When clearance is diminished, elimination half-life increases proportionately, as does the time required to achieve steady-state plasma levels (three to five half-lives). The appropriate time for assessment of full pharmacologic effect and measurement of levels may thus be markedly delayed. For example, the elimination half-life of digoxin is normally 20 hours, but increases to 1 week with end-stage renal disease (ESRD); an ESRD patient discharged from the hospital with a plasma level within the therapeutic range 1 week after starting maintenance therapy with 0.25 mg per day of digoxin will probably subsequently develop digitalis toxicity.

Consideration of Potential Drug-Drug Interactions

Drug-drug interactions may occur owing to alterations in drug disposition (PK interactions), or drug effect (PD interactions).4,83,84 PK interactions occur through effects on four drug disposition mechanisms: (1) gastrointestinal absorption, (2) drug distribution, (3) hepatic metabolism, and (4) renal excretion. These categories will be further discussed below. PD interactions occur by three basic mechanisms: (1) pharmacologic interactions, usually based on binding of an antagonist to the target receptor mediating drug effect (e.g., use of a nonselective β-blocker such as propranolol in a patient requiring inhaled nebulized β-agonist therapy for bronchospasm); (2) physiologic interactions, such as the additive effects of multiple sedatives or vasodilators, or the opposing effects of warfarin and vitamin K on coagulation factor synthesis; and (3) drug-induced changes in the intracellular milieu resulting in altered effects of other agents (the best such example is the precipitation of digitalis toxicity by loop diuretic–induced hypokalemia and hypomagnesemia). It is worth emphasizing the fact that individual sensitivity to drug effect (e.g., sedative effects in the elderly) may predispose particular patients to suffer the consequences of PD drug interactions more readily than others. Prevention of PD drug interactions is best accomplished through a basic understanding of the mechanism of action, common side effects, and population-specific issues (if any) pertinent to any drug prescribed.

PK interactions are the subject of many exhaustive compendia, but a few general principles are required to interpret such information. In order for a reported PK drug interaction to be of clinical significance, three criteria should be fulfilled.5,38 First, one of the drugs involved must have a low therapeutic index, so that an alteration in its disposition results in either toxic or subtherapeutic effect (there are important exceptions to this rule). Second, the drug disposition parameter alteration should be on the order of 30% or more, because changes of lesser magnitude are unlikely to be of clinical importance. Finally, only reports in humans should be considered definite, since animal studies of drug disposition do not always accurately reflect human processes. Conversely, the absence of reported data from animal studies does not preclude drug interactions in humans, particularly with respect to newer agents undergoing wide use for the first time.

Absorption

Most such interactions affect the rate rather than the extent of absorption, so that the magnitude and timing of peak plasma drug level is altered, but the fraction absorbed (and thus bioavailability) is unchanged. Interaction mechanisms include physicochemical complexing (e.g., tetracycline and milk products); gastric pH changes (e.g., failure to absorb ketoconazole when gastric pH is made alkaline by coadministration of antisecretory agents); alterations in gastrointestinal motility; effects on gastrointestinal mucosa or flora (e.g., increased digoxin levels after antibiotic therapy; see Chap. 102); changes in mesenteric blood flow; and finally changes in first-pass metabolism (e.g., cyclosporine with ketoconazole71,72,85).

Distribution

Altered drug binding because of displacement by another agent is of far lesser importance as a source of drug interaction than is generally appreciated. In fact, drug displacement from binding sites simply makes more free drug available for excretion or biotransformation.84 The net effect on plasma levels of active drug is thus usually negligible, unless clearance of the displaced drug is also impaired because of organ dysfunction or the inhibition of drug metabolism or excretion by the displacing agent itself.

Hepatic Metabolism

Inducers and inhibitors of hepatic biotransformation are important causes of drug interactions. This is well illustrated by considering common interactions with the enzyme predominantly responsible for cyclosporine metabolism, underlining the importance of considering drug interactions when adding new agents to a patient's drug regimen. Cyclosporine, a potent immunosuppressive drug used in the prevention of allograft rejection, is metabolized by the CYP3A4 enzyme in both the liver and small intestine. Modulation of the pharmacokinetics of cyclosporine has been reported with a large number of drugs. For example, ketoconazole is known to be a potent inhibitor of CYP3A4, and therefore both increases oral bioavailability and reduces the rate of elimination of the drug, increasing blood cyclosporine levels. This predictable interaction can be used favorably to minimize the cost of therapy by decreasing the necessary dose of cyclosporine by co-administration with the cheaper drug, ketoconazole. More commonly, drug interactions result in unfavorable outcomes. Inadvertent co-administration of cyclosporine with a CYP3A4 inducer, such as troglitazone (now removed from the market) or rifampin, can result in rapid drug elimination (a state similar to the previously described “ultra-rapid metabolizer” phenotype) and subtherapeutic cyclosporine levels, with subsequent potential rejection of the transplanted organ.87

There are numerous clinically important examples of such interactions involving the other P450 isozymes also. Theophylline is metabolized by CYP1A2; theophylline toxicity may be precipitated by addition of the quinolone antibiotic ciprofloxacin (an inhibitor of CYP1A2) or by smoking cessation (because of withdrawal of the CYP1A2-inducing effect of cigarette smoking), unless theophylline dosage is reduced appropriately. Only one dose of quinidine (a CYP3A4 substrate) is required to convert a normal, extensive metabolizer of CYP2D6 substrates (including numerous β-blockers, antidepressants, and antipsychotic agents) into a poor metabolizer of these agents. Inhibition of drug biotransformation activity by a drug which is not a substrate of the impaired enzyme is not uncommon: quinidine is a powerful CYP2D6 inhibitor and CYP3A4 substrate; fluconazole is also a CYP3A4 substrate but impairs phenytoin and warfarin metabolism by CYP2C enzymes, in addition to competitive inhibition of metabolism of other CYP3A substrates. Such instances reinforce the need to examine both the known and potential interactions of each additional agent with those in the patient's current regimen.

Finally, agents that diminish hepatic blood flow (vasopressors and cimetidine) may impair clearance of high-extraction drugs such as lidocaine and propranolol.

Renal Excretion

Drug interactions can affect glomerular filtration, tubular secretion, or tubular reabsorption (active or passive). It is obvious that agents that impair or augment glomerular filtration tend to cause a corresponding change in excretion of drugs eliminated by this route, such as aminoglycosides. Active tubular drug secretion is performed primarily by two groups of proximal tubular pumps: an anion pump (which secretes organic acids) and a cation pump (which secretes organic bases). Competitive inhibition of the anion pump by probenecid impairs elimination of penicillin, thus prolonging the half-life of this organic acid. Methotrexate is another anion secreted by this pump, and subject to the same interaction. Cimetidine and trimethoprim inhibit cationic pump secretion of procainamide and some other cationic drugs. Digoxin undergoes distal tubular secretion by a third pump system, which is inhibited by multiple drugs, including quinidine, amiodarone, spironolactone, and several calcium channel blockers; this is the mechanism partly responsible for the elevation of serum digoxin levels routinely encountered following combination of these agents with digitalis therapy. Finally, as discussed in Chap. 102, alterations in urinary pH alter passive distal tubular reabsorption of weak acids and bases, including drugs such as salicylic acid and phenobar bital (a basic drug).

Therapeutic Drug Monitoring

Most drugs are dosed according to population-based PK/PD data. In most cases, critical care therapy is initiated in this fashion, but may be monitored and titrated using readily available pharmacodynamic parameters (sedation, analgesia, paralysis, hemodynamic data, cardiac rhythm, and electroencephalography). Selected agents, usually those with a low therapeutic index, are additionally monitored and dose-adjusted using plasma drug levels and calculations of individual drug disposition parameters. Whether monitored using pharmacodynamic indexes alone, or more precisely with the addition of PK data, ongoing attempts to optimize therapeutics should minimize the occurrence of inadequate therapy or adverse drug reactions (ADR).

Appropriate direct therapeutic drug monitoring involves consideration of many variables. The form of drug measured is important, especially in drugs with high plasma-protein binding. Because only the free fraction of the drug is clinically active, the useful plasma test is one that accurately reflects the free fraction. For example, phenytoin is highly albumin-bound. In the setting of critical illness, hypoalbuminemia, malnutrition, or cirrhosis, total phenytoin levels may be low, while free phenytoin levels are likely to be in the therapeutic range. The timing of plasma drug testing is also important. In drugs such as aminoglycosides, peak and trough levels need to be monitored, since peak levels reflect therapeutic efficacy and trough levels are monitored to ensure adequate drug clearance and to avoid drug accumulation.

Drug regimen adjustment may be precipitated by multiple intercurrent factors, including changes in volume status; alterations (worsening or improvement) in gastrointestinal tract or circulatory, renal, or liver function; the application of extracorporeal therapies which impact drug disposition; and potential drug interactions or other ADRs associated with addition of new agents. Finally, any change in patient status should be considered a potential ADR, and this possible diagnosis assessed using available resources. If an ADR is strongly suspected, this should be reported to the appropriate institutional and extramural authorities, and management altered accordingly.

Adverse Drug Reactions

Drug prescribing intended to achieve beneficial therapeutic effects is necessarily accompanied by the possibility of eliciting an unintended adverse drug reaction (ADR). ADRs are common in the ICU, where multiple drugs are administered to unstable patients who commonly are unable to give a complete medical history and manifest altered drug disposition parameters (including metabolism and excretion). ADRs present diagnostic and therapeutic problems that complicate ICU admission, increasing morbidity, mortality, and length of stay.89

To minimize ADRs, all drugs prescribed by a critical care physician should be reviewed (prior to administration) by the critical care nurse and pharmacist for potential errors including patient identity, known patient allergies, correct dosing, and potential drug-drug interactions. Most preventable ADRs involve (1) prescribing a drug despite a documented allergy to the medication ordered or to a cross-reactive agent; (2) the use of anticoagulants or thrombolytic agents; (3) failure to appropriately monitor and adjust low-therapeutic-index drug administration using plasma drug concentration analysis; and (4) failure to adjust the regimen for administration of renally eliminated drugs in the presence of renal dysfunction.81 Accordingly, we advocate a daily review of all medications given to each ICU patient, focusing on the following issues: (1) drugpatient interactions, including known drug allergies; (2) drugdisease interactions, including presence and severity of organ dysfunction, and appropriate dosing adjustments; (3) potential drug-drug interactions; (4) the possibility of an ADR causing any new corporeal dysfunction; and (5) discontinuation of all unnecessary drugs, in an ongoing attempt to simplify and rationalize therapy. The importance of minimizing the number of drugs prescribed is demonstrated by the fact that there is a 40% probability of developing an ADR when more than 15 drugs are given to a hospitalized patient, compared to a 5% ADR probability when receiving fewer than 6 medications.90

Adverse drug reactions may be classified as type A (an exaggerated pharmacologic effect related to high drug concentration), or type B (an idiosyncratic reaction).61 Increasingly, reactions formerly classified as idiosyncratic (type B) have been shown to be type A reactions based on polymorphic expression of metabolizing enzyme activity and impaired drug metabolism and target effect. Individuals deficient in activity of the necessary metabolizing enzyme develop ADRs caused by accumulation of previously undocumented toxic metabolites; for example, sulfonamide hypersensitivity has been linked to an increase in prevalence of the “slow acetylator” phenotype, and defective detoxification of hydroxylamine metabolites.91,92 Other idiosyncratic phenomena, such as halothane hepatitis and carbamazepine reactions, may also be caused by defective metabolism.92

The diagnosis of an ADR requires a high index of suspicion that signs or symptoms may be drug-related. Distinguishing between an ADR and manifestations of other diseases can be challenging, particularly in critically ill patients with a myriad of problems.

Identification of a temporal relationship between drug initiation and onset of signs or symptoms is useful, but not always possible. For example, supraventricular tachycardia after initiation of theophylline, wheezing after β-blocker administration, or seizures during meperidine therapy may each be caused by, aggravated by, or unrelated to drug effects. Furthermore, patients may develop a type A adverse drug reaction in the ICU when elimination of an agent, which they have received chronically prior to hospitalization, is subject to a drug-disease interaction (e.g., theophylline toxicity precipitated by development of congestive heart failure), or drug-drug interaction (e.g., theophylline toxicity caused by concurrent ciprofloxacin administration). In the absence of another convincing etiology, discontinuation or dose reduction of the suspected drug is indicated. Subsequent improvement in signs or symptoms is presumptive evidence of an ADR; of note, ADRs precipitated by agents with long half-lives may require a longer period for resolution. The degree of certainty in ADR diagnosis increases if rechallenge with the suspected drug leads to reappearance of the presumed ADR; however, readministration of drug for this purpose is not recommended unless the information is crucial for subsequent patient management and alternative drugs are not available. If an idiosyncratic or allergic type reaction is suspected, drug readministration may be hazardous. Prior to considering rechallenge, appropriate consultation is suggested, to evaluate viable alternative therapies, the potential for challenge-induced anaphylaxis, possible desensitization procedures (if any), and formulation of a comprehensive treatment protocol to manage any adverse sequelae of re-exposure to the agent.

Identifying the culprit drug in patients receiving multiple medications can be challenging. The most likely offender should be either dose-reduced or discontinued (as appropriate based on the suspicion of a dose-dependent versus an idiosyncratic/allergic phenomenon), while less likely candidate drugs are continued. If the suspected reaction does not improve, remaining drugs may be discontinued sequentially, beginning with the most likely candidate. In a patient suffering a severe reaction, all medications should be stopped if possible. A common example in the ICU is the patient who develops thrombocytopenia while receiving prophylactic therapy with an H2 blocker and subcutaneous heparin. In this situation, thrombocytopenia frequently represents a manifestation of disease, particularly if hemodynamic instability is present, and not an ADR. Nevertheless, in the presence of severe or progressive thrombocytopenia, an ADR should be considered and medication adjustments made. Heparin use correlates most firmly with drug-induced thrombocytopenia, and this agent should be discontinued first.93 H2 blockers (and indeed most other drugs used in the ICU) are not responsible for the majority of cases of thrombocytopenia and may be continued if clinically important.93 In patients with critical thrombocytopenia, however, risk-benefit analysis supports discontinuation of both drugs pending further evaluation, which may include assay for antiplatelet antibodies (drug-specific tests are possible).

Numerous published tables list drugs associated with ADRs.90,94 These lists are merely suggestive that a sign or symptom may be drug-related. They include both well-documented and poorly-documented drug reactions and are clearly not all inclusive. The absence of a drug from a particular category does not exclude the possibility of drug reaction, since there is always a first identified or reported reaction, especially during the initial period of postmarketing surveillance following approval and release of a new drug for widespread use. In the absence of a simple, dedicated, and exhaustive source of information about ADRs specific to critical care, most clinicians rely on their institution's hospital drug information service, and various library and Internet resources.95 It is important to become familiar with these resources and to utilize government and pharmaceutical industry information services to aid in evaluating potential ADRs. Appropriate information resources include the following:

General Drug Information References

1. Physicians' Desk Reference. Medical Economics Company, Inc. (annual editions).

2. American Hospital Formulary Service Drug Information. Bethesda, MD, American Society of Health-Systems Pharmacists, Inc. (annual editions).

3. Drug Facts and Comparisons. Philadelphia, Lippincott. Phone: 1-800-232-0554 for product information (updated monthly).

4. Hansten PD, Horn JR: Drug Interactions Analysis and Management. Applied Therapeutics, Inc. (updated quarterly).

5. Tatro DS (ed): Drug Interaction Facts. St Louis, Lippincott (updated quarterly).

6. http://medicine.iupui.edu/flockhart/ (updated periodically).

Drug Interaction Software Subscription Services

Most institutions utilize one or more computer software programs available by subscription, such as ePocrates Rx Pro, to help identify drug interactions. These programs are continuously in development, and to date no system has been demonstrated to be superior to an experienced clinician in identifying drug interactions. Barriers to the development of drug-drug interaction (DDI) detection software for widespread use include difficulty of portability and integration into existing hospital computer systems, development and maintenance of the knowledge base, and establishment of formal methodology of evaluation of these systems. Nonetheless, existing available software are a good initial step toward decreasing ADRs due to drug-drug interactions.

Dedicated ICU Intensivists and Pharmacists

Many studies have investigated the benefits of a dedicated team of professionals in the ICU, with the team led by a physician with specialized training in critical care medicine.96,97 This advantage is theorized to be a result of specialized training in particular issues that arise in critically ill patients, a broader perspective of issues that pertain to critically ill patients, and improved continuity of care. In fact, the Leapfrog Initiative—a coalition of some of the nation's largest employers, such as General Electric and General Motors—has identified this as one of the three changes that they believe would most improve safety.98 This same rationalization can be used to justify the role of a dedicated ICU pharmacist. The traditional role of a pharmacist to accept and clarify orders and dispense medications is antiquated. Having a pharmacist who participates in clinical rounds as a full member of the patient care team in the ICU can significantly decrease ADRs and decrease costs of hospitalization as compared to pharmacist review alone.99 The benefits may be explained by improved communication between the health care professionals, optimization of therapy, and improved monitoring and management of adverse drug events. For example, for the treatment of an infection in a critically ill patient, the ICU pharmacist may contribute specialized knowledge in selecting the narrowest-spectrum antibiotic, in selecting an appropriate drug to minimize antibiotic resistance based on patterns of organism resistance in the hospital, in minimizing potential adverse drug reactions, and in selecting the most cost-effective agent. Barriers to implementing the dedicated ICU pharmacist model include the initial investment of money and time to create a new staff position, but these initial costs are likely to be easily overcome by savings in drug costs and prevention of adverse drug events.

Computerized Order Screening Systems

Computerized screening of medication orders at the time of prescription data entry is performed in many acute care practice settings. Such programs allow for real-time checks for potential drug interactions with the patient, disease states, or other drugs. Physician order entry (POE) systems also obviate potential medical errors due to illegible handwriting or transcription errors. Common problems with these programs include a lack of primary literature referencing, routine detection of insignificant interactions, and a significant lag time between introduction of a new drug and interaction data updates. Physicians with minimal technological expertise may also be hesitant to use such technology. Nonetheless, POE systems are another change that the Leapfrog Initiative has identified that may contribute to significant reduction in ADRs.

Reporting Unusual or New Drug Interactions

Medwatch is a voluntary program sponsored by the Food and Drug Administration for reporting adverse events and problems with drugs and other products regulated by this agency. Although it is not a direct information source, contributed data collectively provide continuous new drug information. Medwatch is a major component of the FDA's postmarketing drug product surveillance and has identified ADRs that were not apparent during preapproval clinical trials (e.g., cardiotoxic effects of astemizole). Reports are encouraged even if the practitioner is not certain the product caused the event and whether or not all details are available. The program utilizes a consolidated reporting form (FDA form 3500), which may be submitted by mail or facsimile. Reports can also be sent by computer modem or by phone 24 hours a day 7 days a week. For more information or to report quality problems call 1-800-FDA-1088.

Drug Manufacturer Information

The package insert which accompanies all products contains FDA-approved information regarding the known properties and appropriate use of the drug. Contact information (including toll-free telephone numbers) for pharmaceutical companies can also generally be found in this insert or other package labeling or through the hospital drug information center. Pharmaceutical manufacturing companies usually have drug information support services that can provide up-to-date information pertinent to their products. Manufacturers are required to perform postmarketing surveillance of their products and to collect information about drug interactions and other ADRs.

Computerized Resources in the ICU

Application of computer technology to medical care can improve efficacy and decrease errors at every step of drug administration. Computers can serve as a repository for references. Availability of textbooks, journals, review services and formularies may be valuable at the time of drug prescription. While availability of all references by print may be prohibitive, online references are easily accessible, such as Medline (National Library of Medicine, United States), MDConsult (LLC Ltd Liability Co., DE), Physician Drug References,100 and UptoDate (UptoDate Inc., DE), to name a few.

Most institutions in the United States use computers to store patient medical records. A step further in the integration of computer technology into medical practice is patient information capture. Real-time capture of patient data, such as vital signs, pulse oximetry, and laboratory results, aids in the dissemination of information and more informed decisions for drug prescription and drug effect monitoring. Therapeutic drug monitoring can be more effectively regulated with integration of patient data and drug administration.

Recent advances in handheld devices make all of the aforementioned systems even more convenient and portable. Software for handheld devices, such as Epocrates Rx (Epocrates, Inc., CA) and Medcalc for example, allow for quick and convenient references for patient data interpretation. While computer technology continues to advance, the major barriers to incorporation of these systems into the medical practice include the initial costs of the computer devices and software, the time to install the appropriate software, the education and willingness of health care professionals to use the technology available, and the real possibility of technical malfunction. Future trends in computer technology in the ICU involve expert systems that can simulate human judgment to aid in diagnostic and therapeutic decision making, and data mining that can analyze large amounts of data to recognize relationships that have not been otherwise discovered.