Chapter 18. Principles of Antimicrobial Therapy

1. Antibiotic therapy should be based on the infection site, host defenses, antibiotic toxicity, antibiotic pharmacokinetics and pharmacodynamics, and the regional resistance organisms and antibiotic susceptibility patterns.

2. Methicillin-resistant Staphylococcus aureus, which is generally spread by contact, has become a major cause of serious hospital-associated infections and ensuing morbidity despite the development of newer effective antibiotics.

3. The term catheter-associated infection is used for surveillance and is a diagnosis of exclusion based on the presence of a catheter within 48 hours of a bloodstream infection and no other identifiable source.

4. Ventilator-associated pneumonia typically arises as an extension of upper airway bacteria and is related to duration of intubation. Several "bundles" are recommended for prevention, although their efficacy is controversial.

5. Most guidelines recommend avoiding central venous catheters when possible, removal as soon as it is no longer needed, and placement using full sterile procedures with barrier precautions.

6. Intra-abdominal infections are often associated with mixed flora and generally require anaerobic as well as aerobic antibiotic treatment.

7. Recommended surgical prophylaxis for infective endocarditis is evolving. In contrast with past guidelines, it is currently recommended for only a few conditions.

In general, antimicrobial drugs may be used in 3 different modes: therapeutic, prophylactic, and preemptive. In the therapeutic mode, they are prescribed to treat established clinical infection. This requires prompt diagnosis of clinical infection and a clear understanding of the pharmacologic principles governing treatment of such infections. In the prophylactic mode, antimicrobials are prescribed to all members of a given population before an event (eg, surgery) to prevent infection. Successful prophylactic programs require that the antimicrobial therapy be sufficiently nontoxic, inexpensive, and efficacious to justify the intervention. Finally, in the preemptive mode, antimicrobial therapy is administered to a subgroup of individuals based on either laboratory markers or clinical epidemiologic characteristics that place them at significant risk of a serious clinical infection (eg, patients undergoing organ transplants).1 Effective preemptive therapy requires the careful delineation of the factors that justify antimicrobial intervention at a point when clinical disease is not yet manifest.2-4 Clearly, there is some overlap between prophylactic and preemptive modes.

The purpose of this chapter is to present the pharmacologic and clinical principles that underlie all 3 forms of antimicrobial use, distinguishing between what is known, what is suggested by consensus but lacks definitive evidence, and what needs further study. The focus is primarily on antibiotics rather than antifungal and antiviral drugs. Specific information on newer antibiotics such as cyclic lipopeptides, glycylcyclines, ketolides, oxazolidinones, streptogramins, and newer fluoroquinolones can be found in review articles5-8 and newer textbooks such as Mandell et al.9 Aside from the specifics of a particular antibiotic, it is important to recognize the different characteristics among the various antibiotic classes. For example, some antibiotic classes such as tetracyclines are bacteriostatic in contrast to bacteriocidal antibiotics which directly kill bacteria, bacteriostatic antibiotics prevent bacteria from growing but generally require leukocytes to kill them. Another important difference among antibiotic classes is their mechanisms of action, especially the distinction between antimicrobials that directly affect cell-wall synthesis and those that affect protein or nucleic acid synthesis. β-Lactams and vancomycin are examples of the former, whereas aminoglycosides, macrolides, glycylcyclines, and fluoroquinolones are examples of the latter. Other important distinctions include the rate of development of resistance, induction of β-lactamases, presence of postantibiotic effect, pharmacokinetics, and pharmacodynamics.

When antimicrobials are used prophylactically for surgical patients, initial therapy is often empirical, directed against the most likely microbes while taking into account their antimicrobial resistance patterns not only in a region, but also in a particular hospital or even an area within the hospital, such as an intensive care unit (ICU). Most commonly, surgical antimicrobial prophylaxis is directed against the organisms most likely associated with a particular type of surgery. Other important considerations in selecting a prophylactic antimicrobial agent are its toxicity, its spectrum, its antimicrobial properties (ie, bacteriostatic or bacteriocidal), post-antibiotic effect, the likely source of bacteria (eg, colon, skin, biliary tract), and the drugs' pharmacokinetics and pharmacodynamics.

It is of great concern that a significant proportion of antimicrobial use, particularly broad-spectrum agents, both in the inpatient and outpatient setting, are often inappropriate, both because they are unnecessary and because they do not cover the infecting organisms.10-12 Antibiotic use has been clearly associated with the emergence and dissemination of antibiotic-resistant organisms, with this occurring far more commonly in ICUs than in non-ICU inpatient areas and more commonly in inpatient than outpatient settings.

Minimal Inhibitory Concentration and Minimal Bactericidal Concentration

Antimicrobial therapy designed to treat active infection is predicated on the reliable delivery of the drug to the site of infection at the concentration and duration required to contain or eradicate the infection. The effects of antimicrobials may be either to kill the targeted organisms (bacteriocidal) or to inhibit their growth (bacteriostatic). Although it would seem likely that bactericidal therapy would be inherently superior to bacteriostatic therapy, in most instances of clinical infection, either form of therapy is effective. There are, however, a few instances in which there is an absolute requirement for a bactericidal regimen, largely because the infection is inaccessible to the body's defenses or the patient is immunocompromised. Predictably, they are as follows:

• Systemic infection in a severely neutropenic patient
• Staphylococcal and, presumably, other forms of osteomyelitis
• Cardiovascular infection (eg, endocarditis)
• Central nervous system (CNS) infection: meningitis, cerebritis, and abscesses
• Infection in the presence of a foreign body, such as a joint prosthesis

If there are no bactericidal regimens for the invading organism, then high-dose and prolonged bacteriostatic therapy in association with aggressive surgery, if possible, becomes the alternative management strategy.

Modern microbiologic techniques not only permit the isolation and identification of the infecting microbe, but also can determine the susceptibility to different antibiotics, thus guiding the choice of antibiotics for a particular patient. When exposed in vitro to possible therapies, bacteria may be killed or their growth may be inhibited without being killed. The lowest concentration required to prevent growth is termed the minimal inhibitory concentration (MIC; variants of which are the MIC50 and MIC90, which are the concentrations of drug required to inhibit 50% and 90%, respectively, of strains of the same species, in contrast to MIC, which measures inhibition of 1 strain). Similarly, the minimal bactericidal concentration (MBC) is the lowest concentration required to kill the bacterial isolate (with the MBC90 defining the lowest concentration of drug needed to kill 90% of isolates of the same species). The MBC is technically much more difficult and time-consuming to determine than the MIC; it is subject to multiple potential errors and is generally not important for routine clinical care.13

These in vitro results are then interpreted in terms of blood levels that are attainable when appropriate doses of the antimicrobial agent in question are administered. One of the 3 results is usually reported: susceptible, resistant, or intermediate. Susceptibility suggests that blood levels attainable with the therapeutic doses of an antibiotic should be effective in the management of infection due to this organism. Whether such efficacy is actually obtained depends on other factors, particularly activity and penetration in the site of infection. For example, blood levels may not provide adequate concentrations at sites such as the eye, the brain, or the prostate gland, and even with adequate concentrations, antibiotics can be inactivated by host factors. For example, daptomycin is inactivated by pulmonary surfactant and aminoglycosides and polymyxins by low pH. Thus in vitro susceptibility is only the first step in choosing the appropriate antimicrobial.

Susceptible sensitivity means the MIC is at or below the concentration achieved with the usual recommended dose of the antibiotic. Intermediate sensitivity means that MICs of the organism(s) in question are higher than blood and tissue levels that are achievable without potential toxicity. For the most part, the use of antimicrobial agents to which the bacterial isolate exhibits intermediate susceptibility should be avoided. However, particularly in dealing with a broadly resistant infection in the urinary tract, the issue is not the attainable blood levels, but rather the relationship between the MIC of the isolate and the attainable urine concentration. For example, tetracycline can often be used to treat "resistant" Pseudomonas aeruginosa strains isolated from urine because the levels in urine may be 100 times higher than in blood.

Resistant sensitivity generally means the MIC is greater than the concentration achievable in therapeutic or nontoxic doses. However, it can also mean that clinical efficacy has not been reliable or that special resistance mechanisms are likely to develop, such as induction of genes encoding inactivation of the antibiotic.

Pharmacologic Principles in the Successful Use of Antimicrobial Agents

The application of certain pharmacologic principles is important in optimizing the effectiveness of antimicrobial therapy. These principles are particularly important in treating infections, because unlike most anesthetic drugs, a direct measure of efficacy either does not exist or is delayed for long enough that by the time it is determined that the treatment failed, adequate treatment might be much more complex or possibly ineffective.

Pharmacokinetics and Pharmacodynamics

The terms pharmacokinetics and pharmacodynamics14,15 are used to describe the interaction of the drug administered with the individual receiving the therapy. Pharmacokinetics (PK) describes the quantitative relations between the administered doses of a drug and the concentrations of that drug in different body compartments. The most common PK models assume that there are only 2 compartments in the body, central and peripheral, which obviously do not correspond exactly to physical locations. Nonetheless, PK takes into account the effects of drug absorption, distribution, metabolism, and excretion. Pharmacodynamics (PD) describes the quantitative effects, both toxic and beneficial, of the plasma or tissue concentrations of a drug on the individual.16,17 PK/PD models quantify the relationships between PK and PD. Perhaps the most useful measure of antimicrobial levels is the area under the time-concentration curve (AUC) divided by the dosing interval. This permits estimation of the average drug concentrations over the dosing interval. However, such data are usually available only for blood (the central compartment) and therefore may not accurately reflect the value at the tissue or site of interest. For example, aminoglycosides are highly concentrated in the kidney. Nonetheless, determining the PK–PD relationship is the foundation of efforts to define the appropriate therapeutic antimicrobial dose.

The volume of distribution is the ratio of the drug concentration in blood or plasma, assumed to be in equilibrium with all spaces where it distributes, to the amount of drug administered. For many drugs the volume of distribution vastly exceeds total body water. This large apparent volume of distribution may occur because of tissue or protein binding if only free drug is measured. This value is also altered by lipid solubility, the drug's blood–tissue partition coefficient, perfusion, volume status, and pH.

High serum protein binding of a drug can alter the PK behavior of the molecule without significant effects on the PD profile by serving as "buffers" so that there is little change in the concentration of free drug, despite its distribution to extravascular sites. Because albumin is the major binding protein, most binding usually occurs in the plasma, and serum albumin concentration can markedly affect the PK of highly bound drugs. In addition, at sites of active inflammation, there can be leakage of serum proteins, accounting for higher concentrations of drug at these sites.

The nature of the capillary bed at a particular site is also an important factor in the distribution of antimicrobials. At most sites the capillary bed has small pores (fenestrations) that allow for the unencumbered movement of substances with molecular weights of less than 1000 daltons (the great majority of antibiotics are of that size—the so-called small molecules). There are, in addition, specialized sites, such as the retina, the brain, and the prostate gland, where the capillaries are unfenestrated. Because the antimicrobial molecules must pass through these capillaries, diffusion is enhanced by high lipid solubility, small size, and the degree to which they are not ionized. β-Lactams, because they are highly ionized at plasma pH, may exhibit impaired penetration at these sites.18 A final mechanism affecting the distribution of antimicrobial molecules is active transport pumps. These pumps act on organic anions and are present in the choroid plexus, the retina, the proximal tubules of the kidney, and the biliary ducts. β-Lactam drugs are pumped out by these active transport pumps, with probenecid competitively inhibiting this mechanism.

Pharmacokinetics and the Choice of Antimicrobial Therapy

Three PK/PD issues should be considered in the choice of antimicrobial therapy: the ability of the drug to penetrate to the site of presumed infection, the presence of local factors at the site of infection that might modify the efficacy of particular antimicrobial including bacterial resistance and the presence of biofilms and the route of delivery that should be used. Thus if infection of the CNS is suspected, drugs that reach effective concentrations within the CNS should be used in doses adequate to achieve this objective (Table 18-1). However, there are exceptions. For example, although not recommended, meningitis due to high-level penicillin-resistant Streptococcus pneumoniae has been successfully treated with antibiotics that typically do not penetrate the CNS well. This may be a result of increased penetration resulting from inflammation. Moreover, serum levels may not be indicative of cerebrospinal fluid (CSF) levels. For example, cefotaxime serum levels peak and decline rapidly, whereas CSF levels are much lower but remain elevated much longer than serum levels, probably because the cerebral capillary endothelium is not fenestrated and lacks pinocytotic vesicles.19

Local factors that modify the effectiveness of particular antimicrobial agents include the following:

1. Inadequately drained abscesses. For example, aminoglycosides are far less effective in the presence of pus (both aminoglycosides and vancomycin are bound by pus, limiting their antimicrobial activity) and in environments with low pH and low oxygen tension.20

2. Low Po2. Some antibiotics (eg, aminoglycosides) require oxygen-dependent transport to penetrate the outer membranes of susceptible bacteria. This explains why aminoglycosides are generally ineffective against anaerobic bacteria.21

3. The presence of β-lactamases. In mixed microbial infections, organisms such as Bacteroides fragilis produce β-lactamases, which cause local inactivation of β-lactam antibiotics despite their in vitro sensitivity to other bacteria.20 Production of β-lactamases may also be induced by certain antibiotics.

4. Binding to hemoglobin. Penicillins and tetracyclines bind to hemoglobin, thus decreasing the efficacy of these drugs for infected hematomas.22

5. Presence of foreign bodies. In most instances, the presence of a foreign body at the site of infection will prevent the elimination of the microbial invaders, even with the best antimicrobial regimens.23

6. The inoculum effect. High bacterial counts can increase the MIC of some antibiotics.

7. Decreased bioavailability. Oral administration of some antimicrobials (eg, trimethoprim-sulfamethoxazole, fluoroquinolones, fluconazole) can achieve blood levels that are comparable to those obtained with parenteral therapy if gastrointestinal function is adequate. However, absorption of some antimicrobials (eg, levofloxacin) is inhibited by food or substances containing divalent metal ions. A number of other factors can affect bioavailability, including the first-pass effect, in which drugs absorbed in the small intestine are metabolized in the liver; drug-metabolizing enzymes in the gut; and the presence of diseases that interfere with absorption, such as inflammatory bowel disease, parasitic infestation, and diarrhea of any etiology. Nonetheless, even if therapy is initiated intravenously, completion of a course of treatment can be accomplished with a more cost-effective oral regimen once the patient stabilizes.

Postantibiotic Effect

Some antibiotics continue to suppress the growth of certain bacteria after the antibiotic is no longer detectable in the media—a phenomenon termed the postantibiotic effect (PAE). For example, after a 2-hour exposure of susceptible staphylococcal cultures to penicillin, growth is still inhibited for 1.4 to 1.6 hours after multiple washings removed the penicillin from the culture medium.24 Although PAE can be demonstrated for virtually all antimicrobials, the bacteria affected and the duration of the effect are highly variable. β-Lactam antibiotics, with the exception of carbapenems such as imipenem and meropenem, have significant PAEs only for some gram-positive bacteria. In contrast, most antibiotics that inhibit protein or nucleic acid synthesis in gram-negative bacteria generally exhibit a PAE for these organisms. For example, for gram-negative bacteria, aminoglycosides and fluoroquinolones produce PAEs for 1 to 4 hours, whereas β-lactams have essentially no PAEs for these bacteria. Importantly, the PAE duration tends to be reduced for aminoglycosides and fluoroquinolones in acidic media, such as encountered in necrotic tissue. A related but poorly understood phenomenon is that during the PAE phase, bacteria are more susceptible to killing by leukocytes, an effect termed postantibiotic leukocyte enhancement (PALE).25,26 In addition, the PAE itself may be prolonged by leukocytes.

Concentration-Dependent and Time-Dependent Killing Agents

The distinction between time-dependent and concentration-dependent killing coupled with the presence or absence of a PAE has important clinical ramifications, particularly with respect to dosing intervals. For antimicrobial/bacteria combinations that exhibit time-dependent killing, especially when there is no PAE, the duration that the antimicrobial level is above the lethal range (usually about 4 times the MIC) is important, whereas concentrations above the bacteriocidal levels cause little additional bacterial killing. Vancomycin and β-lactams with the exception of carbapenems typically exhibit time-dependent killing. There is general agreement that for β-lactams to be effective, their concentration must be well above the MIC for most of the dosing interval. This is best achieved using continuous infusions for at least part of the dosing interval or prolonging each bolus administration. This should be especially true for gram-negative bacteria, which usually do not exhibit PAE for β-lactams. Although this concept is supported by animal data, it is puzzling that studies in humans have not generally shown superiority of continuous infusions.27-29 Nonetheless, for this class of drugs, it seems logical to adjust dosing frequency to maintain levels in a bacteriocidal range for as long as logistically possible.30

In contrast, for antibiotic/bacteria combinations that exhibit concentration-dependent killing and significant PAEs, the bacteriocidal rate and the duration of the PAE are dose-dependent (eg, for aminoglycosides, fluoroquinolones, daptomycin, colistin, metronidazole, ketolides, and possibly azithromycin). Therefore, it may be more important to achieve very high peak levels and allow the trough to decrease below the MIC because efficacy will still be maintained by the PAE, whereas toxicity may be minimized. For these drugs, the AUC divided by the MIC is more important than the dosing frequency or the time greater than the MIC. For combinations of antibiotics, the situation is more complicated. The PAEs for gram-positive bacteria are increased either additively or synergistically when aminoglycosides are added to cell-wall active antibiotics. However, this does not occur with gram-negative bacteria, except when imipenem or meropenem is used with an aminoglycoside or for certain bacteria such as the enterococci.31

These considerations led to the concept that administering aminoglycosides in larger doses once daily would prove more beneficial and less toxic than the usual standard of 3 times daily.32 In particular, because aminoglycosides exhibit concentration-dependent killing and have a PAE, concentrations that reach a level 8 to 10 times the MIC will rapidly kill all susceptible bacteria and suppress the survival of higher MIC mutants. Because of the PAE, suppression can be sustained during the period that the concentration drops below the MIC. The maintenance of drug-free intervals for a period before the next dose might also help prevent adaptive resistance, which may occur from decreased aminoglycoside uptake by bacteria exposed to lower levels of the drug. Finally, because renal uptake is saturable, the very high initial levels are not accompanied by proportionally increased renal uptake. Consequently, renal toxicity does not increase. Numerous studies show that in most patients and even in febrile neutropenic patients, these tenets seem to hold; efficacy is at least equal to multiple daily dosing, but surprisingly, renal toxicity is not decreased.33,34

A third group consists of antimicrobials that are predominately bacteriostatic and have prolonged PAEs. Because of the prolonged PAE, efficacy is more dependent on an AUC that is greater than the MIC rather than time above the MIC. This group includes macrolides, clindamycin, quinupristin/dalfopristin, tetracyclines, tigecycline, and linezolid.15

Resistance

General Concepts

Bacteria may be intrinsically resistant to antimicrobials, or they may acquire resistance. Intrinsic resistance reflects a natural resistance to an antibiotic. It is demonstrated when bacterial growth is not inhibited or the bacteria are not killed by the antibiotic, even though no new genetic information has been acquired. For example, vancomycin is ineffective against gram-negative bacteria because it cannot penetrate their outer membrane. Acquired resistance reflects a genetic alteration in a microorganism that renders a once effective antimicrobial ineffective. The general mechanisms for bacterial resistance to antibiotics are as follows: (1) decreased permeability to the antimicrobial, thus preventing its entry; (2) increased efflux pumps that pump antimicrobials out of gram-positive bacteria or keep the antimicrobial concentrations in the space between the inner and outer membranes of gram-negative bacteria and/or the cytoplasms low; (3) antimicrobial inactivation; (4) modification of the antimicrobial target, interference of binding to the target, or overexpression of the target; (5) binding of the antimicrobial by false binding sites; (6) development of pathways that bypass the target.

The genes encoding these phenotypes may be chromosomal or plasmid/transposon in origin, inducible, or constitutive. Methicillin-resistance in Staphylococcus, vancomycin-resistance in Enterococcus, and broad-spectrum β-lactam resistance in gram-negative bacteria have important diagnostic and therapeutic implications, particularly in the hospital setting.35-40

In gram-negative bacteria, antibiotics usually must penetrate the hydrophobic lipopolysaccharide outer membrane via specialized channels, termed porins. Alterations in the permeability of porins to specific antibiotics can either prevent sufficient quantities from entering the bacteria or limit the rate of entry to the extent that even a relatively low inactivation rate is sufficient to render them ineffective. The former is a frequent cause of P aeruginosa resistance to aminoglycosides and carbapenems, and the latter allows many gram-negative bacteria to inactivate β-lactams via β-lactamases. Increased active efflux of antimicrobials is a less common mechanism of resistance but can be important for tetracyclines, macrolides, streptogramins, fluoroquinolones, some β-lactams, and possibly meropenem.

Inactivation is the predominant mechanism of resistance for several classes of antibiotics. Aminoglycosides can be inactivated by both gram-positive and gram-negative bacteria, usually by plasmid-mediated enzymes. These enzymes modify specific sites on the aminoglycoside molecule, such that N-acetylation, O-nucleotidylation, or O-phosphorylation takes place. These modified aminoglycoside molecules bind poorly to the ribosome target, rendering them inactive in normally susceptible bacteria. Target modifications in penicillin-binding proteins account for methicillin resistance in staphylococci and penicillin resistance in pneumococci and enterococci. Bypass pathways account for vancomycin resistance in Enterococcus faecium and resistance of many bacteria to folate antagonists, such as trimethoprim-sulfamethoxazole.

Specific Mechanisms

Perhaps the most comprehensively studied and categorized inactivating enzymes are the β-lactamases. There are at least 890 different β-lactamase protein sequences (www.lahey.com/studies), one or more of which can inactivate virtually any β-lactam, monobactam, or carbapenem-based antimicrobial.41 They are carried on chromosomal genes, plasmids, transposons, and integrins, the latter often producing multidrug resistance (MDR). The classification schemes for β-lactamases are based on both phenotype and genotype and continue to evolve. One scheme groups them into classes A, B, C, and D based on amino acid motifs. However, it is also common to classify them into the more subjective Bush-Jacoby functional groups 1, 2, and 3 based largely on which β-lactams they hydrolyze and which are blocked by the β-lactamase inhibitors. Certain β-lactam antibiotics induce β-lactamases. For example, clinical failure associated with the emergence of resistance often occurs when serious Enterobacter infection is treated with a cephalosporin despite initial cephalosporin susceptibility.42 Thus cephalosporins should be used carefully in the setting of infection with the previously mentioned gram-negative pathogens.35

Although β-lactamases are produced by some gram-positive bacteria, notably staphylococci (but not streptococci) and many anaerobes, the efficacy of these β-lactamases is limited because they are diluted by the extracellular space, in contrast to being confined in the periplasmic space of gram-negative bacteria, where they reach high concentrations.38,43 Although plasmid-encoded β-lactamases cause staphylococcal resistance to many penicillins, resistance of some staphylococci and other gram-positive bacteria can also occur by completely different mechanisms. For example, methicillin-resistant Staphylococcus aureus (MRSA) is caused by the chromosomal presence of the mecA gene, which decreases the binding affinity of methicillin to penicillin-binding protein 2a. Similarly, ampicillin-resistant E faecium has an altered penicillin-binding protein 5 that has a greatly reduced affinity for these antibiotics and most β-lactams. In contrast, in the majority of cases, vancomycin resistance in enterococci is related to a multigene plasmid, which is easily spread among enterococci. This plasmid alters a terminal amino acid group in one of the cross-linking amino-acids that prevents vancomycin from binding to the growing cell membrane.44,45

It is astonishing that the first time vancomycin-resistant enterococci (VRE) were observed was in 1986,46 and within 15 years it has come to represent more than 24% of ICU enterococcal isolates in the United States. Unfortunately, there are limited options for the treatment of severe infections related to VRE, which are typically also resistant to ampicillin and aminoglycosides. Several new antibiotics (eg, tigecycline, a glycylcycline; quinupristin/dalfopristin, a streptogramin; and linezolid, an oxazolidinone) that have recently been approved by the US Food and Drug Administration (FDA) have efficacy against VRE and other resistant gram-positive bacteria.7

There is great concern that S aureus will acquire the vancomycin-resistance determinant from enterococci. This was first shown to be possible in the laboratory,47 and scattered clinical cases have now been reported of MRSA strains with both decreased vancomycin susceptibility (known as vancomycin-intermediate S aureus [VISA]) and vancomycin resistance (known as vancomycin-resistant S aureus [VRSA]).48 The mechanism for this diminished susceptibility appears to be due to a thickening of the cell wall leading to decreased antibiotic penetration and sequestration of vancomycin, rather than the transfer of vancomycin resistance genes from Enterococcus leading to an altered vancomycin binding target.49-51

Emerging Bacterial Resistance

In the last decade, there has been a dramatic increase in the development and dissemination of antibiotic-resistant organisms. As shown in Table 18-2, resistance of enterococci, especially E faecium, to vancomycin has shown the greatest increase, whereas MRSA has remained stable up to 2007, which is the last year when such data were available. Many factors have contributed to the marked increase in resistance; however, indiscriminate broad-spectrum antibiotic use is an important contributor.52 Other factors that select for resistance include dosing that is too low or given for too short a duration. Thus it is critically important to use antimicrobials judiciously, as overuse facilitates the propagation of pathogens that are difficult or impossible to treat. One example of antimicrobial misuse is in farming, where the widespread deployment of antimicrobials to animals has resulted not only in resistant flora on farms, but also in the spread of such resistant organisms to humans.

Combination Antimicrobial Therapy

A common practice in the hospital use of antimicrobial agents is "double coverage" in an effort to improve efficacy and prevent resistance from developing. Indeed, in vitro studies often suggest benefits from such an approach, with at least the following justifications.

Prevention of the Development of Antimicrobial Resistance

Combination therapy to prevent development of resistance is most effective when the mechanism for resistance involves chromosomal mutations, with the merits of combination therapy being most easily seen for tuberculosis. The microbial burden in a particular patient and the known rate of mutational resistance to each of the antituberculous drugs are the major determinants of the most effective regimen for a particular patient. Thus, in a nonimmunosuppressed host, the pulmonary microbial burdens are as follows: primary tuberculosis of approximately 104 organisms, noncavitary reactivation 106 to 108 organisms, cavitary disease 108 to 1010 organisms, and so forth. The rates of mutation to resistance for tuberculosis antimicrobials are approximately 1 in 106 for isoniazid and ethambutol, 1 in 108 for rifampin, and 1 in 105 for streptomycin. Therefore, the treatment for primary tuberculosis typically requires a single drug (usually isoniazid), noncavitary reactivation disease requires 2 drugs, and so on. Although other factors enter into the design of the therapeutic regimen, the approach outlined fits with the efforts to avoid resistance. However, chromosomal mutation is an uncommon mechanism of resistance (the great majority of bacterial resistance is due to the acquisition of a plasmid or transposon that mediates resistance). Thus the prevention of development of resistance is not an adequate reason for combination therapy of pyogenic infection.

Known or Suspected Polymicrobial Infection

This can be a valid reason for deploying multiple drugs. For example, a perforation of the distal colon will result in the release of anaerobes (particularly B fragilis), gram-negative bacilli, and bowel streptococci (including enterococci) into the peritoneal cavity. This event had traditionally required a 2- or 3-drug regimen (eg, ampicillin, metronidazole, and gentamicin) to deal with the range of aerobic and anaerobic organisms. However, today there are single antibiotics such as carbapenems, β-lactam/β-lactamase combinations, newer fluoroquinolones, and tigecycline with a spectrum of activity that will accomplish the same task.

Initial Therapy

When the initial assessment of a patient suggests the possible presence of a therapeutic emergency, broad-spectrum therapy is obligatory, at least until the results of cultures are available. The issue that remains controversial is whether this is best accomplished with a single drug or multiple drugs. Current evidence does not generally support the older concept that a combination of a β-lactam with an aminoglycoside provides a higher probability of success than a single drug, even if P aeruginosa is likely to be involved. Nonetheless, commonly therapy is initiated with a broad-spectrum drug or combination of drugs, which may have overlapping spectra with different resistance patterns to ensure the initial coverage is adequate. However, once cultures and sensitivities are available, the regimen should be changed to the drug or drugs with the narrowest spectra that provide sufficient coverage.

Decreased Toxicity

A principle of cancer chemotherapy is that in some cases the toxicity can be decreased by using multiple drugs with different mechanisms of action and toxicities at lower doses. Unfortunately, the efficacy of this approach in the treatment of significant infection has not been demonstrated.

Synergistic Therapy

In vitro and in animal models of serious infection, enhanced killing with certain antibiotic combinations for some species of both gram-positive and gram-negative bacteria can be demonstrated. This has been most evident with combinations of β-lactams and aminoglycosides. Whether this can be demonstrated in clinical situations is not yet determined. Moreover, whether this truly constitutes synergy is debatable, as the methodology (eg, the checkerboard method) usually does not adhere to standard methods for demonstrating drug synergy. The best evidence of the utility of enhanced efficacy of combination therapy comes from the study of patients with enterococcal endocarditis. Treatment of enterococcal endocarditis with penicillin (or ampicillin) alone results in an inferior clinical outcome compared with that achieved with the addition of streptomycin or gentamicin to the penicillin. Enterococcal strains with high-level aminoglycoside resistance (MIC >2000 μg/mL) do not have the potential for bactericidal therapy and have a distinctly poorer outcome. Although combination of a penicillin and gentamicin may also shorten the requisite duration of therapy for some nonenterococcal gram-positive endocarditis, outcomes are not improved.

Part of the problem with existing studies is the lack of stratification of patients into "therapeutic emergency" versus "a need for parenteral therapy" versus "outpatient management." As a result, even with the lack of definitive evidence, our approach to staphylococcal endocarditis is as follows: stable patients get single-drug therapy with nafcillin; those who are more ill (hemodynamic instability, evidence of emboli, etc) receive potentially synergistic therapy with a minimum of 2 drugs. Although many experienced clinicians use potentially synergistic therapy for other serious gram-negative infections, particularly those occurring in compromised hosts, definitive evidence is lacking and most current data favor monotherapy.

Effective antimicrobial therapy requires the integration of a sizable body of information regarding the patient, the invading microbe(s), and the antimicrobial agents themselves into an effective antimicrobial prescription. The factors that must be considered include the following.

Identity of the Organism

The identity of the infecting organism(s) must be known, or at the very least, it must be possible to make a high-probability assessment of the most likely culprit(s). The initial choice of antimicrobial agents is usually based on probability assessments of the most likely pathogens causing a particular clinical syndrome arising in a particular environment, with subsequent adjustment of the regimen once specific microbiologic information becomes available. Important factors to be weighed in determining the likely infecting organism include the probable source (eg, vascular access device, lung, or urinary tract) and whether the infection is likely to be community-acquired or hospital-associated. Other issues of importance include recent antibiotic use, the presence of such growth factors as iron and elevated blood sugar, and the existence of devitalized tissue.

Urgency of Treatment

The key question is whether one is dealing with a therapeutic emergency or a diagnostic dilemma. In the former case, the broadest possible regimen should be utilized initially, with later modifications based on precise microbiologic information. This is often termed front loading of the antimicrobial regimen. In the latter case, narrower spectrum initial therapy is possible, with fine-tuning of the regimen once specific information is available.

Epidemiology

Epidemiologic considerations must also be factored into antimicrobial decision making. When dealing with primary illnesses acquired in the community, the nature and timing of possible exposures must be considered. Important exposures include the geographically restricted systemic mycoses (Blastomyces dermatitidis, Coccidioides immitis, and Histoplasma capsulatum), Mycobacterium tuberculosis, influenza, group A streptococci, Legionella, and meningococcus. Even more important is the special epidemiology of the hospital-associated environment, be it a nursing home, a patient with a recent admission to a hospital or recent antibiotic treatment, a hospital, or a specialized area of the hospital such as an ICU. Specifically, the microbiologic flora within a hospital or even a particular ICU within the hospital and their respective resistance patterns are essential information. The rising incidence of infection with difficult to treat organisms such as MRSA, VRE, and multiresistant gram-negative bacilli has a profound influence on the initial choice of antibiotics. For example, vancomycin is usually preferred over nafcillin as initial therapy for suspected staphylococcal infection, and advanced spectrum β-lactam agents such as imipenem rather than cefazolin or gentamicin for gram- negative infections that could be caused by highly resistant Klebsiella species.53,54 It is likely that these problems with antibiotic resistance will increase, not just in terms of these organisms, but also with rising incidences of infections with penicillin-resistant pneumococci55,56 and ampicillin-, vancomycin-, and gentamicin-resistant enterococci.56

These considerations highlight the importance of the reservoir of a given organism and the selective pressure that reservoir is under (eg, Salmonella and Campylobacter in domestic animals vs S pneumoniae in the community at large and multidrug-resistant gram-negative rods in the nosocomial environment).57,58 By understanding the reservoir, the prevalence of antimicrobial resistance in the reservoir, and the risk of exposure of a patient to a given organism, one can determine appropriate initial antimicrobial therapy.

The implication of these principles is 2-fold. First, every hospital and patient area within the hospital must engage in ongoing surveillance of the most common organisms and antimicrobial susceptibility patterns causing infection in their particular areas. Second, decisions about antimicrobial use should take into consideration such information, particularly for patients who have been subjected to the selection pressures of previous courses of antibiotics or environments that may have rendered them more vulnerable to colonization and/or invasion with resistant flora.

Host Factors

A number of host factors must be considered in formulating an antimicrobial regimen. These can be divided into 3 groups: (1) those that increase a patient's risk for a specific type of infection; (2) those that increase a patient's risk of complications from an infection (eg, the presence of prostheses such as artificial joints, synthetic vascular grafts, or cardiac valves); and (3) those that increase the risks of treating an infection with specific drugs because of patient comorbidities, such as renal insufficiency.

Specific Drug Risks

Perhaps the most important consideration is the history of previous adverse reactions to the drugs in question. Great precision is needed in defining the nature of a history of an adverse reaction. For example, nausea, vomiting, and diarrhea, particularly after oral administration, do not preclude the use of an antimicrobial, especially intravenously. In contrast, a history of anaphylaxis, Stevens-Johnson syndrome, toxic epidermal necrolysis, or allergic interstitial nephritis does preclude further use. Commonly, the exact nature of the allergy is unclear. In this circumstance, either the class of drug implicated should be avoided or else skin testing with a subsequent desensitization regimen should be performed.

The presence of renal dysfunction in the patient being treated can also have a profound effect on antimicrobial therapy because many antimicrobials are cleared by the kidney and therefore require adjustment of the dosing regimen. Unless such adjustments are made, toxic side effects may occur, such as injury to the 8th cranial nerve and further renal injury with aminoglycosides,59 seizures with penicillins and carbapenems,60 and bleeding due to platelet dysfunction (added to that already caused by a uremic state) induced by ticarcillin, mezlocillin, and piperacillin.61 Although many antimicrobials are cleared or metabolized by the liver, hepatic dysfunction usually requires less dosage adjustment than renal dysfunction. In general, little dosage adjustment is required until the bilirubin exceeds approximately 5 mg/dL (Table 18-3). However, with hepatic dysfunction, chloramphenicol, clindamycin, the tetracyclines, and the antituberculous drugs isoniazid and rifampin should be avoided, if possible (in the case of the antituberculous drugs, if they are required for emergency therapy, then special effort should be made to monitor blood levels in the face of serious hepatic dysfunction).62 When both renal and hepatic dysfunctions are present, then therapy with essentially all β-lactam antimicrobial agents should be carried out with great care.63

Impaired Host Defenses

Special considerations need to be extended to those patients who have specific impairments in host defenses. Impairment may be anatomical, such as cutaneous ulcerations or mucosal abnormalities, or it may be secondary to functional host defense defects, such as neutropenia, asplenia, malignancy, immunosuppressive therapy, diabetes mellitus, or human immunodeficiency virus (HIV) infection. Patients with prosthetic heart valves, prosthetic joints, vascular grafts, or other prostheses can suffer dire consequences from metastatic spread of infection if initial therapy is inadequate. Therefore, such patients should be considered for front-loading of the regimen, ideally with a bactericidal drug, even if their initial clinical state would not normally require it.

Pregnancy

A critically ill patient who is pregnant, and for whom termination of the pregnancy is not an option, represents a particular challenge. First, the pharmacokinetics of many antibiotics are altered in pregnancy due to a larger volume of distribution and increased glomerular filtration rate.64 Second, the database on the safety of different antibiotics during pregnancy is woefully incomplete. The safety classifications of various antimicrobials during pregnancy are listed in Table 18-4. Importantly, this list is constantly updated. Therefore, before using any antimicrobial, current data should be checked. Aminoglycosides and isoniazid should only be prescribed if absolutely necessary because the former may be associated with 8th cranial nerve dysfunction in the infant and the latter with an increased incidence of psychomotor retardation, myoclonus, and seizures.

General

When choosing the appropriate antibiotic therapy, it is critical to decide whether the infectious process is nosocomially acquired or present at admission (ie, community-acquired). In 2005, the Centers for Disease Control and Prevention (CDC) published rigorous definitions of nosocomial infections in various body sites.65 The CDC defines health care–associated infections (HAI) as those that develop during hospitalization but are neither present nor incubating upon the patient's admission to the hospital. This generally includes infections that occur more than 48 to 72 hours after admission and within 10 days after hospital discharge. Under the aegis of the CDC, the National Healthcare Safety Network (NHSN), has subsumed the surveillance functions of the National Nosocomial Infections Surveillance system (NNIS) in collecting and analyzing HAI surveillance data. Figure 18-1 shows the site distribution of 28,502 antimicrobial-resistant HAIs collected by the NHSN in 463 hospitals from January 2006 until October 2007.66 These data show that approximately 80% of antimicrobial-resistant HAI pathogens were associated with 3 sources: 35.3% with intravascular catheters, 30.1% with bladder catheterization, and 15.9% with ventilators. This accounts for the enormous regulatory interest in minimizing or eliminating infections in these 3 sites. The primary pathogens associated with these various sites are shown in Table 18-5. Note, however, that these are from surveillance data and therefore may overestimate actual infection rates because the former include associations that may lack definitive evidence of the causative device or organism.

It is important to recognize that gram-positive pathogens, particularly Staphylococcus epidermidis, S aureus, and enterococci are the leading bloodstream pathogens, whereas S aureus and gram-negative bacteria, particularly P aeruginosa, are the predominant pathogens associated with nosocomial pneumonia. In addition to S aureus and P aeruginosa, Escherichia coli and Candida albicans are the primary pathogens found in urinary tract infections, with the great majority of these being associated with urinary tract instrumentation.67-73 These data are consistent across continents.73 It is of interest that Candida species now account for the fourth most common cause of nosocomial bloodstream infection. Ten years ago, it was 24th, a testimonial to the increasing importance of this organism, particularly in ICU patients and possibly to the overuse of antibiotics.

As indicated in Fig. 18-1, the incidence of nosocomial infections is highly associated with the use of devices such as ventilators, vascular access catheters, and urinary catheters. This is likely a result of the breakdown of normal host defenses and clearance mechanisms. In addition, these devices may support the growth of various microorganisms in different ways. For example, the development of biofilms on urinary and central venous catheters enhances the adherence of certain microorganisms such as E coli and may impair the penetration of antimicrobials or alter the resistance patterns, whereas the condensation of water in ventilator circuits is a conducive environment for Pseudomonas and Acinetobacter.71 The device-associated infection rates are surprisingly constant for different devices and types of ICUs. For example, from the most recent NHSN data, the maximum and minimum rates per 1000 device days for catheter-associated urinary tract infections was 25.7 in medical–surgical ICUs and 2.4 in burn ICUs; for central line–associated bloodstream infections, the maximum was 20.6 in medical–surgical ICUs, and the minimum was 2.7 in neurosurgical ICUs; and for ventilator-associated pneumonia, the maximum was 25.1 in medical–surgical ICUs and the minimum was 3.5 in cardiac medical ICUs.67 Surprisingly, these rates are consistently near the lowest in burn ICUs.

Multidrug-Resistant Bacteria

Methicillin-Resistant Staphylococcus aureus

S aureus is the leading cause of hospital-associated infections. The mortality rate for invasive S aureus infections approaches 20%, and an increasing percentage of all such infections are resistant to methicillin (MRSA). S aureus developed resistance to penicillin in the 1940s due largely to its inducible production of a penicillinase that hydrolyzes penicillin and most other β-lactams. Methicillin and oxacillin, which are resistant to penicillinase, were developed in the late 1960s. S aureus resistance to these newer drugs emerged shortly after they were developed. Today, at least 80% of S aureus isolates are resistant to methicillin and most β-lactams, including penicillin and cephalosporins. β-Lactams work by inactivating penicillin-binding proteins (PBPs), which are enzymes that are essential to the formation of the cell wall. Resistance to methicillin (and oxacillin) is due to PBP2a, which is encoded by the mecA gene. PBP2a has low affinity for β-lactams and is not present in methicillin-sensitive S aureus (MSSA). Aside from multiple mechanisms that impair the efficacy of antimicrobial therapy, S aureus can produce a wide variety of substances that promote adherence to cells and artificial materials (eg, catheters), as well as a panoply of exotoxins, all of which enhance its virulence. Because of the virulence of S aureus coupled with the resistance of MRSA to most antibiotics, its containment and treatment of infections caused by it are major problems for hospitals and other chronic care facilities.9

Currently approximately 60% of S aureus infections in hospitalized patients are MRSA, although this varies somewhat regionally.74 Initially, the hospital-acquired and community-acquired MRSA differed markedly. MRSA are often classified into species by pulsed-field gel electrophoresis of DNA fragments into USA100 to USA700.75 Initially, hospital-acquired MRSA (HA-MRSA) comprised USA100 and USA 200, which are resistant to all but a few antibiotics (eg, vancomycin). In contrast, the community-acquired variant (CA-MRSA) largely included USA 300 and USA 400 genotypes, which are sensitive to many antibiotics (eg, clindamycin), excepting most β-lactams.76 However, this distinction has blurred as an increasing number of HA-MRSA are now caused by USA300 CA-MRSA, which evolved independently of HA-MRSA. CA-MRSA usually carries genes that encode Panton-Valentine leukocidin (PVL), a pore-forming exotoxin that is associated with necrotizing pneumonia as well as skin, soft tissue, and leukocyte necrosis. Although it is not clear which strains are most virulent, PVL is thought to be important in the virulence of USA300.

There are several reasons that MRSA has attracted so much attention. First, regardless of the species, MRSA infections are a major cause of morbidity and mortality. Second, until recently, vancomycin was the only antimicrobial that could be used to treat HA-MRSA. Third, MRSA is highly contagious, spread primarily by contact. Fourth, both patients and non-patients frequently carry MRSA in their nares without any signs or symptoms. Consequently, they may inadvertently spread the organism, causing infections among susceptible contacts. Lastly, MRSA causes many types of infection, ranging from furuncles and carbuncles to metastatic infections, necrotizing pneumonia, osteomyelitis, and endocarditis. It is also responsible for a substantial percentage of catheter-associated bloodstream infections. There are many strategies intended to prevent HAIs in hospitalized patients, many of which center around handwashing and contact precautions, and the incidence of MRSA infections seems to be declining.77-79 Although it is tempting to attribute this to mandatory reporting in some states as well as an emphasis on a variety of measures intended to help prevent MRSA infections, the decrease began before widespread implementation of these measures and may reflect a natural biologic trend.80 The role of nasal screening and decontamination, usually with topical mupirocin, is controversial and probably not generally cost effective, except possibly in patients with recurrent infections.81

Although CA-MRSA and HA-MRSA both are resistant to β-lactams because of PBP2a, CA-MRSA is usually sensitive to clindamycin, trimethoprim-sulfamethoxazole, and doxycycline. Current guidelines recommend empirical treatment with one of these drugs for outpatient treatment of purulent cellulitis, abscesses with constitutional symptoms, or with comorbidities such as diabetes or immunosuppression.81 Otherwise, drainage alone may suffice. Notably, these antimicrobials are not effective against HA-MRSA. Pending sensitivities, it should be assumed that all staphylococcal infections in hospitalized patients are multidrug resistant. Therefore, initial treatment is limited to vancomycin, linezolid, daptomycin, or tigecycline, with certain caveats as reviewed by Neuner et al.82 Daptomycin should not be used for MRSA pneumonia because it is inactivated by pulmonary surfactant. Linezolid may be at least as effective as vancomycin for treating MRSA pneumonia and may have an advantage because, at least in vitro, it inhibits the production of several toxins, including PVL. Of concern, however, is its potential for inducing toxicity in combination with drugs that inhibit serotonin uptake because it is weak inhibitor of monoamine oxidase. Unfortunately, strains of S aureus that have partial (VISA) or complete (VRSA) resistance to vancomycin are emerging. Fortunately, aside from the previously mentioned antibiotics, several new antibiotics, including the cephalosporins ceftaroline and ceftobiprole, that have activity against MRSA are currently undergoing development.

Multidrug-Resistant Gram-Negative Bacteria

The plethora of bacterial resistance mechanisms and the mobility of some of the responsible genes have posed an ongoing problem in the management of both gram-negative and gram-positive infections. However, for gram-negative bacteria, this problem has increased with the emergence of strains that are resistant to 3 or 4 different antibiotic classes. Such multidrug resistance (MDR) has been a problem for some strains of Klebsiella pneumoniae. More recently, the incidence of infections caused by Acinetobacter baumannii, which is intrinsically MDR, has increased. In the past, these organisms were usually sensitive to carbapenems, but lately they have acquired serine and metallo-β-lactamases (carbapenemases), which hydrolyze carbapenems as well as all other β-lactams.83 As a result, some strains of A baumannii, P aeruginosa, and other species of Enterobacteriaceae (eg, E coli) have exhibited resistance to 4 different antibiotic classes.84 The genes responsible for production of these carbapenemases usually reside on plasmids that have apparently been transmitted from K pneumoniae to other Enterobacteriaceae, including E coli. These genetic islands often carry with them resistance mechanisms for multiple classes of non–β-lactam antibiotics. Between 2006 and 2008, approximately 60% of A baumannii infections throughout the United States were MDR, in contrast to only 10% of P aeruginosa and 15% of K pneumoniae.84

MDR in Acinetobacter species have become particularly problematic. They are usually restricted to hospitals, especially ICUs, and community-acquired infections are uncommon. However, because they are found in soil and water, they may be present in trauma and burn patients on admission. Moreover, for reasons that are not understood, they are also being found with increasing frequency in military personnel injured in the Middle East. Acinetobacter species can colonize skin, respiratory tract, and gastrointestinal tract, and because they are transmitted by contact, meticulous handwashing and contact precautions are essential in preventing hospital outbreaks. Furthermore, they can survive desiccation for weeks. Consequently, eradication of outbreaks may require closing the offending area (eg, an ICU) for complete decontamination. A baumannii infections occur mostly in ICU patients with ventilator-associated pneumonia and catheter-associated bloodstream infections, whereas P aeruginosa and K pneumoniae are most prevalent in catheter-associated urinary tract infections.

These MDR bacteria may be resistant to all antibiotics except colistin (polymyxin E). Unfortunately, colistin is nephrotoxic. Because it was approved before the FDA's promulgation of more rigorous standards, the appropriate doses and dosing schedules are not well established. Therefore, minimization of cross contamination is vital to contain these bacteria.

Catheter-Related Infections

Intravascular access devices are common causes of bacteremia and fungemia. A primary bloodstream infection (BSI) is defined as at least 1 positive bacterial or fungal culture drawn from a peripheral site without any known source. Although a BSI is usually central to defining infections attributable to a catheter, the terminology has become somewhat confusing. A central line–associated bloodstream infection (CABSI) is defined as a BSI when an arterial or central venous catheter (CVC) was in place within the past 48 hours.85 Importantly, as noted earlier, this definition is used for surveillance rather than clinical care and has several limitations.86 First, using the total catheter days as the denominator for CLABSI as per CDC recommendations implies that for a given patient, the probability of getting a CLABSI is constant over the CVC dwell time. Yet, clearly, dwell time does matter, which implies that the probability per day is not constant. Moreover, if this were not true, there would be no infection-control reason to advocate minimizing CVC dwell times. Therefore, in assessing CLABSI rates, a correction should be applied for dwell times rather than normalizing the rate per 100 catheter days. Second, because there are no accepted criteria for when to obtain blood cultures, the incidence of BSI within 48 hours of the presence of CVC may underestimate CLABSI or even central line–related bloodstream infections (CRBSI) unless blood cultures are obtained every time a catheter is removed. Finally, statistical comparisons among rates cannot be made unless the total number of catheter days is known or confidence limits are included in the data. The latter practice was used by the CDC,87 but not in reporting by individual states.

In contrast, a suspected CRBSI is defined as a BSI in a patient with a CVC or arterial catheter in place, whereas a definite CRBSI is defined as a BSI with the same organism grown from the catheter tip. If the catheter is still in place, a CRBSI is confirmed if blood sampled through the catheter grows more than 15 colony-forming units (CFU) of the same organism on blood media. Alternatively, if a peripheral blood culture cannot be obtained, a CRBSI may be confirmed if blood is sampled through at least 2 lumens of a multilumen catheter and 1 lumen yields a colony count at least 3 times greater than the other lumen(s).88 However, this definition is occasionally supplanted with a laboratory-confirmed bloodstream infection (LCBSI), which is defined either as a CRBSI or a patient with a CVC and with fever (>38°C) or chills or hypotension not related to an infection at another site and with a common skin contaminant (eg, Corynebacterium spp., Bacillus spp., Propionibacterium spp., coagulase-negative staphylococci including S epidermidis, viridans group streptococci, Aerococcus spp., or Micrococcus spp.) grown from 2 or more blood cultures drawn on separate occasions.89 The most widely used diagnostic technique to culture a catheter tip is a semiquantitative method, which involves rolling the catheter tip or the introducer tip used with a pulmonary artery catheter across an agar plate and then counting the number of CFUs after overnight incubation. Although a positive culture drawn through a catheter is relatively nonspecific and has a higher rate of false-positive results than peripheral blood cultures, a negative culture generally rules out a catheter-related infection because of the low false-negative rate.90 It is difficult to find current data on the risk of bacteremia associated with various vascular devices, probably because of the large variation in type of catheter (number of lumens, catheter material, antibiotic coated or not), techniques (skin preparation), care of the site, and barrier precautions between regions of the country and hospitals within a region.

To put this risk in perspective, it is worth noting that 50,000 to 100,000 patients in US hospitals develop nosocomial bloodstream infections each year, with approximately 70% of these infections related to arterial or central venous catheters of various types.67 Such infections are associated with a mortality rate of 25% to 35% and a 2- to 3-fold increase in attributable mortality. In 2009, The Infectious Diseases Society of America (IDSA) published guidelines revised from 2001 for the management of intravascular catheter-related infections.88 The guidelines generally divide CRBSI into early (<14 days after placement) or late, and percutaneous or tunneled catheters. The guidelines state that the initial choice of antibiotics depends on the severity of the patient's clinical disease, the risk factors for infection, and the likely organisms associated with the specific intravascular device. Understanding the pathogenesis and virulence of the typical organisms involved permits this problem to be approached rationally.

For nontunneled short-term catheters, the organism usually comes from skin flora that colonizes the external surface of the catheter. The most common of these is S epidermidis, followed by S aureus, Candida species, and enteric gram-negative bacteria. In contrast, long-term and tunneled catheters are typically colonized intraluminally from the catheter hub. The most common microorganism is still S epidermidis, but for these catheters, it is followed by enteric gram-negatives and then S aureus, with P aeruginosa becoming more important. Gram-negative bacteria are more common with femoral catheters, probably because of perineal flora, and are therefore discouraged. They are also more common when there are gram-negative infections of the respiratory tract or surgical wounds or drains because they increase the incidence of gram-negative skin colonization, thereby providing a reservoir of organisms for such infections. This pathogenic mechanism explains the increasing infection rate associated with central line position: femoral > internal jugular > subclavian. Presumably, the incidence of infections is higher for internal jugular catheters than for subclavian catheters because the former is more likely to be exposed to orotracheal secretions. The use of antibiotic-impregnated catheters has been shown to prevent microbial contamination of the catheter and subsequent bacteremia in some but not all studies. In one study, catheter-associated bacteremia decreased from 3.4% to 0.3% with the use of a minocycline- and rifampin-impregnated catheter.91 However, these results should be interpreted cautiously because these coatings may lead to false-negative catheter cultures.

Initial therapy of suspected CRBSI usually includes vancomycin because of the high prevalence of methicillin-resistant S aureus and S epidermidis in the hospital environment. However, the latter microorganism poses a particular problem because it is both the most common CRBSI but also the most common contaminant. For this reason, the 2009 IDSA guidelines recommend confirmation via additional blood cultures obtained via the catheter and peripherally before initiation of therapy or catheter removal for S epidermatitis-suspected CRBSI. Additional coverage directed at gram-negative organisms should be considered based on local susceptibility data, especially if gram-negative infection is present at other bodily sites or if the patient manifests cardiovascular instability. In the latter case and for immune-compromised patients, therapy should also be directed at multidrug-resistant organisms such as P aeruginosa. Empiric therapy directed at vancomycin-resistant enterococci is rarely indicated, as this organism is an unusual cause of acute hemodynamic instability. Antifungal therapy is also not initiated unless the patient has been receiving total parental nutrition, prolonged use of broad-spectrum antibiotics, the catheter is in a femoral artery or vein, the patient is immune-compromised, or there is microbiologic evidence of fungal infection (see later discussion). Fluconazole is acceptable therapy if the patient has not received it within the past 3 months and the risk of Candida krusei or Candida glabrata is low. Otherwise, an echinocandin is recommended. Once the offending bloodstream pathogen is identified and its resistance profile known, focused antimicrobial therapy should be used. There are no definitive data for duration of therapy, but the catheter should be removed and therapy prolonged (4-8 weeks) for complicated CRBSI such as suppurative thrombophlebitis, endocarditis, osteomyelitis, or metastatic seeding. For uncomplicated CRBSI, generally the catheter should be removed and, if needed, replaced de novo rather than over a guidewire if possible. Therapy is recommended for 7 to 14 days.

There are several important issues to consider in the treatment of the common CRBSI pathogens. As mentioned, S epidermidis is often a contaminant. However, when it does cause a BSI, it typically behaves as a relatively avirulent pathogen. Consequently, it usually can be treated with removal of the catheter and a short course of antibiotics. Nonetheless, this organism has a propensity to adhere to prosthetic devices such as cardiac valves and artificial joints. Therefore, when S epidermidis bacteremia occurs in such a patient, one must carefully evaluate the prosthetic device for evidence of secondary infection. In light of these considerations, patients with indwelling prostheses should be considered at higher risk for consequences of central catheter placement.

S aureus, on the other hand, is an extremely virulent organism, which in the setting of bacteremia, often disseminates, causing osteomyelitis, endocarditis, and other severe, tissue-destructive infections. Thus when a bacteremia with this organism is confirmed, a careful evaluation for metastatic infection should occur, and prolonged therapy (2 to 4 weeks) is often recommended. Enterococcal BSI behaves in a similar manner as S epidermidis and typically responds to removal of the catheter. However, endocarditis may occur, particularly in the setting of prolonged bacteremia. The optimal therapy for this organism is ampicillin plus an aminoglycoside, although single-drug therapy with some of the newer antibiotics such as daptomycin and tigecycline show therapeutic promise. When C albicans is cultured from the blood, the patient should be evaluated for metastatic infectious foci (eg, hepatic, ocular, and skin), the catheter should be removed, and antifungal therapy initiated, typically with fluconazole, unless C glabrata or C krusei is prevalent in the environment or isolated, in which case an echinocandin should be used.92 If metastatic foci of candidal infection are found, then the optimal management of these complications will determine the duration of therapy.

Prophylaxis

Surgical

Need for Prophylaxis

The need for antibiotic prophylaxis for surgery depends on the risk of a surgical site infection (SSI), generally defined as purulence within the wound. The occurrences of SSIs are related to the wound classification, patient-related factors such as immunocompetence, the bacterial milieu, hospital infection rate for various procedures, and factors relating to the wound itself.93 As shown in Table 18-6, wounds are usually classified as clean (class I), clean-contaminated (class II), contaminated (class III), and dirty/infected (class IV).94 The increasing risk of SSI from class I to class IV is related to the wound's bacterial burden, although dirty/infected wounds are often already infected. However, careful microscopic examination shows that even clean wounds are contaminated with skin flora.95

Staphylococcal species are the most common wound pathogens in most SSIs. Antibiotic prophylaxis is debatable for some clean procedures such as an inguinal hernia repair or mastectomy. However, for other clean procedures (especially for neurosurgical, cardiothoracic, and vascular surgery; hip or knee arthroplasty; and any procedure in which bone is excised or a prosthesis is inserted), gram-positive coverage, in the past with vancomycin but currently usually with cefazolin, has been recommended. However, because of the increasing incidence of MRSA in both hospitals and in the community, the trend seems to be shifting back toward vancomycin. Trials comparing vancomycin with cefazolin to prevent surgical site infections in cardiac surgical patients have had variable results, but in one study the incidence of SSI in coronary artery bypass surgery was decreased when vancomycin was used for prophylaxis instead of cefuroxime,96 and The Society of Thoracic Surgeons recommends vancomycin prophylaxis for selected patients.97

Prophylactic antibiotics should be administered for all clean contaminated and contaminated wounds, as well as for hysterectomies and most invasive urologic procedures. Sterilization of the urinary tract is recommended before any urologic procedure if possible. Even with successful treatment of a urinary tract infection, deep-seated infection of the prostate gland can be reactivated by manipulation and/or surgery. Prophylaxis is advised for high- and moderate-risk patients undergoing procedures involving infected tissues and should include anti-staphylococcal antibiotics for the prevention of cellulitis and osteomyelitis. Similar coverage is advised for patients receiving prosthetic cardiac valves. Patients with urinary tract infections should receive antibiotics such as fluoroquinolones, third-generation cephalosporins, or an aminoglycoside that are active against gram-negative bacilli. Convincing evidence for prophylactic antimicrobial benefit is also found for patients undergoing endoscopic manipulation of an infected biliary tree or urinary tract. For these conditions, antibiotics such as ampicillin-sulbactam and piperacillin-tazobactam are reasonable choices, although the increasing frequency of E coli resistance to the former drug has led to suggestions to avoid it.

Timing of Prophylaxis

It is generally accepted that intravenous administration of prophylactic antibiotics should be initiated no sooner than 1 hour before incision. Theoretically, this ensures adequate tissue levels before surgery begins. The major impetus for this practice comes from a study that retrospectively analyzed the timing of prophylactic antibiotic administration for patients for whom a wound infection was reported.98 These investigators concluded that patients who developed SSIs were more likely to have received prophylaxis between 24 and 2 hours before surgery or after skin incision. Often unnoticed in this study was that the incidence of wound infection was statistically unchanged if the antibiotics were given within 2 hours before incision or within the first 4 hours after incision. When these data were subjected to a multiple logistic regression, the only variables related to wound infection were underlying disease, nursing service, type of surgery, duration of surgery, and time after the start of surgery when the first dose of prophylactic antibiotic was administered. Notably, of the 41 SSIs, 58% were resistant to the antimicrobial drug used. Thus although this study is widely quoted as showing that prophylaxis must be given within 2 hours of the incision, these data do not fully support this conclusion. Moreover, this study was retrospective, so at best it could only detect associations rather than cause and effect. It also did not address when the antibiotic infusions were completed.

Although it seems logical to have adequate tissue levels of prophylactic antimicrobials before skin incision, supporting data are only inferential, and it is likely that the importance depends on the type and size of the inoculum. Nonetheless, administration of prophylactic antibiotics within 1 hour of surgery (2 hours for vancomycin) has become a standard of care that is publicly reported as part of the Surgical Care Improvement Project (SCIP). Moreover, the Centers for Medicare and Medicaid Services (CMS) may reduce hospital reimbursement if it is not reported. CMS also will not pay additional costs associated with mediastinitis after cardiac surgery or SSIs after bariatric surgery or certain orthopedic procedures.

In toto, SCIP has 6 surgical infection prevention measures, but only 3 are core measures: (1) timely administration of prophylactic antibiotics as described previously, (2) administration of antibiotics recommended for a given procedure, and (3) discontinuation of prophylactic antibiotics within 24 hours after surgery (48 hours for cardiac surgery). The latter exception occurred because of a document drafted by The Society of Thoracic Surgeons that made the following points: (1) although prolonged use of antibiotics can lead to the emergence of resistant infections, there are no data that this can occur with administration for under 48 hours; (2) antibiotic prophylaxis for 48 hours is "clinically effective in minimizing infectious complications in cardiac surgery" and is as effective as prophylaxis administered for more than 48 hours; and (3) antibiotic prophylaxis should not be used for indwelling catheters of any type or chest tubes.99 There is general agreement on this last point for all types of surgery, with the possible exception of transplantation.

Despite the CMS mandate, to date no study has convincingly shown that adherence to the 3 SCIP core measures has reduced the surgical infection rate. A study of vancomycin prophylaxis in 2048 cardiac surgical patients was interpreted as showing that vancomycin initiated from 16 to 60 minutes before incision and administered over 1 hour was associated with the lowest rate of surgical site infections.100 However, the relative risks were not significantly increased for any time from 16 to more than 180 minutes before surgery. Results were similar for odds ratios calculated using a logistic regression that adjusted for various patient variables. Thus this study actually found that SSI rate was not affected if the vancomycin infusion was completed essentially any time before surgery or as long as 30 minutes after surgery began. A single prospective study101 and 2 purportedly prospective studies that were actually based on data mined from several databases102,103 failed to show any relation between postoperative infections and timing of antibiotic prophylaxis. Most recently, the largest retrospective data–based cohort study also failed to show a relation between SSI and any of the 3 SCIP core measures or all 3 combined, although a reduction in SSI rate was associated with meeting any 2 or more of the 6 infection prevention measures.104 This engendered an editorial view that "The current metrics does not appear to discriminate between effective and noneffective care at the patient or hospital level" and that "investing resources in SCIP reporting is no longer cost-effective."105 Finally, an animal study widely cited as documenting the need to have the antimicrobial given before the incision lacked statistical testing and used very high inocula and varying doses of antimicrobials.106

Other Measures to Prevent Surgical Site Infections

Aside from maintaining strict sterile procedures, hair removal on the morning of surgery that does not create microabrasion (eg, using clippers rather than razors) is logical and effective, although there are few data to support hair removal at all. Showering with chlorhexidine has not been shown to be effective, possibly because of the interval between showering and surgery. Successful treatment of distant infections and drain removal as soon as possible is also logical although not carefully studied. Surgical technique such as careful approximation of tissue planes and avoidance of hematomas may be the single most important factor. One important factor that is now part of CMS quality measures is maintenance of normothermia both intraoperatively and postoperatively. Normothermia presumably reduces SSI rates because of increased skin blood flow. A seminal study of 200 patients undergoing colorectal surgery randomized to an intraoperative temperature of 34.5° or 36.5° yielded SSI rates of 19% and 6%, respectively.107 Postoperative oxygen supplementation has met with mixed results. Controlling blood glucose to less than 200 mg/dL in diabetic cardiac surgical patients reduced the incidence of deep wound infections.108 However, it is unclear whether these results can be extrapolated to noncardiac surgery, although the concept seems logical.

To compare SSI rates among hospitals or even physicians, adjustment must be made for patient risk. One of the earliest predictors of postoperative SSI was the Study on the Efficacy of Nosocomial Infection Control (SENIC) index.109 This index gave 1 point to each of the following: operative time greater than 2 hours; abdominal procedures; contaminated or dirty procedures; and at least 3 discharge diagnoses, the latter probably serving as a surrogate for patient health. On the basis of data accrued from 1970 and 1975 to 1976, patients with 4 points had approximately a 30% SSI rate, whereas it was 1% for patients with 0 points. This index was subsequently modified into the National Nosocomial Infection Surveillance (NNIS) index, which gave 1 point for each of the following: an American Society of Anesthesiologists score of 3, 4, or 5; a contaminated or dirty procedure; and length of procedure greater than 75% of that expected for each operation. One point was subtracted for laparoscopic surgery.110 Interestingly, using this NNIS index, the SSI rates were similar for all surgical classifications except clean operations. Although many factors relating to both the patient and the environment have been identified, they have not been incorporated into risk predictors.111 In fact, since the NNIS index, there has not been any further development in multivariable risk predictors. Therefore, although publicly available, comparisons of hospital SSI rates are problematic because of inadequate risk stratification.

ICU

Comparable successes with prophylactic antimicrobial regimens in ICU patients have been more difficult to prove, as have attempts to prevent infection with selective gut decontamination regimens. The latter involves using either nonabsorbable antimicrobial agents or fluoroquinolones to eliminate the aerobic gram-negative flora while leaving the anaerobic flora intact, which provides some protection against colonization with a variety of potential pathogens, which is termed colonization resistance. Similarly, aerosolized antibiotics, particularly polymyxin or aminoglycosides, have not been shown to prevent pneumonia. Topical antibiotic ointments also have not been shown to decrease the incidence of intravenous access–related bloodstream infection.

Preemptive Therapy

Recently the concept of preemptive therapy has come to the fore. Preemptive therapy was initially defined in transplant patients, where the initiation of ganciclovir therapy in bone marrow transplant patients with evidence of cytomegalovirus (CMV) replication either in the blood or in the respiratory secretions, at a time when they were asymptomatic, prevented the development of otherwise life-threatening CMV pneumonia. In organ transplant patients, the initiation of preemptive ganciclovir during intensive antirejection therapy markedly decreases the incidence of systemic CMV infection normally associated with such therapy.2,3

The efficacy of preemptive therapy for fungal infection in, for example, abdominal fecal contamination has not been well established. Data for fungal prophylaxis suggest that fluconazole may be effective prophylaxis for certain high-risk patients, such as those undergoing bone marrow transplants or reoperation for gastric or upper small bowel perforations, anatomic sites where large numbers of Candida species are normally found. Nevertheless, there is no evidence that this reduces mortality.112,113 However, such studies are difficult to interpret because the diagnosis of invasive fungal infection is often challenging, and distinguishing between candidal colonization and invasion is problematic. Several studies have shown that when a patient is colonized with a fungus at 3 or more sites, there is a 30% to 60% incidence of invasive disease, with a high associated mortality. There is some agreement that preemptive therapy should be initiated only after recent abdominal surgery with recurrent gastrointestinal perforations or anastomotic leaks and is not necessary for the initial surgery, even if there is fecal soilage, so long as it has not been present for a relatively long period.114 However, this issue is further complicated by the emergence of fluconazole-resistant C albicans and the change in common fungal species to more resistant species such as C krusei and C glabrata, which are not sensitive to fluconazole but can be treated with echinocandins.

Bacterial Endocarditis

The guidelines for prophylaxis for endocarditis have changed markedly since first published in 1955 and continue to evolve, including the most recent update published in 2008,115 which supplanted those published in the preceding year.116 Although these guidelines are generally agreed on, they are not supported by any Class I evidence (benefit >>> risk) that definitively show that antibiotic prophylaxis prevents bacterial endocarditis during procedures that can produce a bacteremia. Nonetheless, given the consequences of endocarditis and the minimal risk associated with prophylaxis, the use of antibiotics to prevent cardiac infection is sensible for high-risk procedures. The most recent guidelines (2008) modify recommended prophylaxis for dental procedures. They are based on 4 rationales: (1) endocarditis is more likely to result from random bacteremias associated with daily activities than dental, gastrointestinal, or genitourinary procedures; (2) prophylaxis would prevent few if any cases of endocarditis; (3) the risk of prophylaxis exceeds the potential benefit; and (4) maintenance of oral health is more important than prophylaxis.

The major changes in these recommendations are that prophylaxis for dental procedures is reasonable only for cardiac conditions associated with the highest risk of an adverse outcome from endocarditis. This includes patients with prosthetic valves, patients with a history of infective endocarditis, patients with unrepaired complex cyanotic congenital heart disease or repaired with prosthetic material within the past 6 months, and cardiac transplant patients with regurgitation from a structurally abnormal valve. Prophylaxis is recommended for these patients only for procedures that involve manipulation of gingival tissue or the periapical region of teeth or perforation of the oral mucosa. Prophylaxis is not recommended for genitourinary or esophageal procedures or colonoscopy. Interestingly, in contrast to older guidelines, these new guidelines no longer recommend prophylaxis for patients with mitral valve prolapse, even those with audible murmurs or thickened valve leaflets. However, this recommendation is not universally accepted and is an enlightening controversy for several reasons.117 First, prolapse can occur in normal mitral valves under conditions that reduce the end-diastolic volume of the left ventricle, such as hypovolemia or enhanced contractility, especially in young adults. Second, prolapse without regurgitation is not thought to increase the risk of bacterial endocarditis because the regurgitant jet seems to cause the abnormalities that increase bacterial adherence to the valve. Thus it was thought that only patients with mitral valve prolapse who also have mitral regurgitation should receive antibiotic prophylaxis, although this is no longer recommended.

Procedures such as central catheter placement through skin that is otherwise normal and has been cleansed with povidone-iodine or chlorhexidine are not associated with significant bacteremias and therefore do not warrant prophylaxis. Regardless of the source, viridans streptococci (α-hemolytic streptococci) are the most common cause of endocarditis, followed by S aureus and enterococci. If prophylaxis is used, a single oral dose of 2 g of amoxicillin given 1 hour before the procedure or intravenous ampicillin or penicillin is recommended. Clindamycin or azithromycin is alternative for penicillin-allergic patients. Endocarditis prophylaxis is recommended for some clean procedures, such as abdominal and lower extremity vascular procedures, craniotomies, orthopedic procedures with hardware insertion, and any procedure that includes implantation of permanent prosthetic material. In contrast, the need for prophylactic antibiotics for orthopedic procedures such as laminectomies and spinal fusions is controversial.

The rate of endocarditis-causing bacteremia with genitourinary tract surgery or instrumentation is high in patients with urinary tract infections and prostatitis and in prostatic surgery. E faecalis is the most common bacterial species, but Klebsiella species are also common. The recommended prophylaxis for these high-risk patients undergoing such procedures is ampicillin. Vancomycin may be substituted for ampicillin in penicillin-allergic patients. Gentamicin may be added in high-risk patients. Attempted sterilization of the urinary tract before any procedure is also thought to be beneficial. Prophylaxis is not recommended for uncomplicated vaginal delivery, cervical biopsy, or manipulation of an intrauterine device in the absence of infection.

Pneumonia

Patients with pneumonia had been divided into 2 general categories: those with community-acquired pneumonia (CAP) and those with hospital-acquired (ie, nosocomial) pneumonia (HAP). However, much of health care is now delivered in nonhospital settings such as nursing homes, where bacterial infections have a risk of being caused by MDR organisms similar to that of hospitalized patients, as is the severity of such infections. Recognition of this led to classifying pneumonia in patients exposed to such environments as health care–associated pneumonia (HCAP). Accordingly, in 2005 pneumonias were categorized into the following 3 groups:118 (1) hospital-acquired pneumonia (HAP), which occurs at least 48 hours after admission; (2) ventilator-associated pneumonia (VAP), which occurs at least 48 hours after endotracheal intubation; and (3) HCAP, which occurs in patients who were hospitalized in an acute care hospital within the past 90 days or who within the past 30 days resided in a nursing home or a long-term care facility or received intravenous antibiotic therapy, wound care, hemodialysis, or chemotherapy. These patients either present with pneumonia or develop it within 48 hours of admission.

Whether HCAP should be considered a single entity or should be subdivided into subcategories by, for example, host, environment, hospital exposure, and immunosuppression is controversial.119,120

To facilitate the study of pneumonia in hospitalized patients, the NHSN published detailed diagnostic criteria.65 Despite this, caution must be exercised in interpreting studies on pneumonia because defining criteria may vary from those used by NHSN.121 Nonetheless, there are clear differences in the microbiology of CAP and HCAP, whereas that of HCAP, HAP, and VAP is similar.

Community-Acquired Pneumonia

It is useful to consider CAP in 3 different categories. First are the typical pneumonias of conventional bacterial origin, which are characterized by the abrupt onset (within <24 hours) of fever, chills, systemic toxicity, cough, purulent sputum production, and dyspnea, often after a preceding viral illness. Second are the atypical pneumonias, characterized by a subacute onset of fever, nonproductive cough, and malaise, with a gradual progression over a several days. Legionnaire disease, which combines features of both, has an abrupt onset of fever, rigors, nonproductive cough, systemic toxicity, and increasing dyspnea, often after a several day prodrome of gastrointestinal upset, headache, malaise, and encephalopathy. Although there is considerable overlap among these presentations, initial therapy can be guided by such categorization and by considering the different etiologies within each category.

The bacterial species or virus causing CAP is never identified in approximately 40% to 60% of patients. Table 18-7 shows the frequencies of the most common bacterial pathogens subdivided into typical and atypical pneumonias. S pneumoniae is the most common cause, accounting for approximately 40% of bacteremic pneumonias. Other relatively common typical bacteria are Haemophilus influenzae, S aureus (with an increasing frequency of methicillin-resistant strains), K pneumoniae, and Moraxella catarrhalis. The atypical bacteria are Chlamydia pneumoniae, Mycoplasma pneumoniae, and Legionella pneumophila. The most common viruses are influenza A and B, depending on the season, respiratory syncytial virus, and respiratory syncytial virus and human metapneumovirus.122 The choice of appropriate antimicrobial therapy is helped by Gram stain of sputum, which cannot detect atypical bacteria of viruses. For detection of atypical bacteria, urinary Legionella antigen assay detects more than 80% of cases of pneumonia due to L pneumophila type I, which accounts for approximately 70% of all cases, but not other Legionella species or types; a polymerase chain reaction test detects mycoplasma; and immunoglobulin M titers detect Chlamydia. Respiratory viruses can be detected by the appropriate antigen or nucleic acid amplification tests. However, for most patients, antimicrobial therapy needs to be initiated in the absence of clear-cut microbiologic information.

In patients with respiratory failure for whom a typical pneumonia is suspected and who do not have a predisposing immunologic defect or a history of a gross aspirational episode, initial therapy should be directed against S pneumoniae, H influenzae, and possibly S aureus, particularly if they also have influenza. Recommended empirical therapy for such inpatients is a β-lactam plus a macrolide, usually azithromycin, or a fluoroquinolone, such as levofloxacin, alone.123 For patients with CAP who require an ICU, the recommended empiric therapy is a β-lactam plus a fluoroquinolone or a β-lactam plus a macrolide, although the supporting data are very limited. If S aureus is suspected and the incidence of MRSA in the community is high, adding vancomycin or linezolid may be appropriate until the organism and its sensitivities are confirmed. If an atypical pneumonia is suspected, a fluoroquinolone such as levofloxacin or erythromycin or one of the newer macrolides, azithromycin or clarithromycin, with or without trimethoprim-sulfamethoxazole, would constitute reasonable initial therapy. For patients with a Legionella-like presentation, high-dose erythromycin has been the traditional therapy of choice. However, azithromycin or levofloxacin may be preferable, combined with of one of the regimens used for the treatment of typical bacterial pneumonia (eg, ampicillin-sulbactam or ceftriaxone), if resistant pathogens are a concern. It is also important to consider viral pathogens because they are a common cause of CAP; rapid diagnostic tests are available, and the possibility of antiviral therapy exists. Although the initiation of empiric therapy is often obligatory, invasive diagnostic techniques such as bronchoalveolar lavage or lung biopsy should be considered for any patient with respiratory failure in whom the etiologic diagnosis is not quickly apparent or fails to respond to therapy. This is especially important for individuals with underlying conditions that predispose them to a broader range of pathogens, such as alcoholics and the immunocompromised.

Nosocomial Pneumonia

Nosocomial pneumonia includes both HAP, defined as occurring more than 48 hours after admission, and HCAP. It can be divided further into early onset, which occurs within the first 4 days of hospitalization, and late onset, occurring 5 or more days after hospitalization. The term nosocomial pneumonia will therefore refer to all HCAP and late-onset HAP because early onset non-HCAP nosocomial pneumonia is similar to CAP. It is less likely to result from MDR bacteria and has a better prognosis than late-onset pneumonia. The common organisms are essentially the same as those found in CAP. MRSA that is often sensitive to clindamycin is also increasingly common in non-HCAP or CAP pneumonia.

The etiology of nosocomial pneumonia, especially when acquired within the ICU, is vastly different from the etiology of non-HCAP CAP. The pathogenesis is typically an extension of a tracheobronchitis. Relatively antibiotic-resistant, aerobic, gram-negative bacilli had been the most frequent invading pathogens, but recently this has been surpassed by gram-positive bacteria. This is thought to be due to the increased rate of oropharyngeal and gastric colonization with these organisms. This colonization serves as a reservoir for the introduction of this flora into the lower respiratory tract, usually due to aspiration and impaired clearance. This is especially true among intubated patients, those with previous lung injury, significant atelectasis, or immunocompromised patients.

Simple measures to help prevent VAP have been promulgated by the Institute for Healthcare Improvement (IHI) (www.IHI.org) as a ventilator bundle that includes elevation of the head of the bed to 30 to 45 degrees, daily trials of spontaneous breathing without sedation, daily "sedation vacation," prophylaxis for both peptic ulcer disease and deep venous thrombosis, and daily oral care with chlorhexidine. Notably, only 2 IHI recommendations, elevation of the head of the bed and oral care, directly relate to VAP. Although these recommendations seem logical, there are no convincing data that they have resulted in a decrease in the incidence of VAP.124,125 Other recommendations for preventing VAP include using enteral rather than parenteral nutrition, orotracheal intubation, orogastric tubes, and subglottic secretion drainage and, obviously, minimizing the duration of intubation. Some studies have also demonstrated a reduction in VAP using silver-coated endotracheal tubes. However, this remains controversial, largely because of study design and incremental expense.126

Figure 18-2 shows the bacterial etiologies of HAP and VAP. Although gram-negative bacillary pneumonia had been the major problem, gram-positive bacteria, especially MRSA, have overtaken gram negatives. Regardless, the bacteria causing infection and their antibiotic sensitivity patterns vary widely in different hospitals and even in different areas within a hospital. Thus precise antibiotic recommendations for initiating therapy must be based on ongoing surveillance of the resident bacterial flora and antibiotic sensitivities in the particular hospital area.

Initial therapy should be modified based on sputum examinations including Gram stains, which provide important information about the relative importance of a particular organism found on culture. Initial therapy should be based on a single advanced-spectrum antibiotic with broad gram-negative protection, possibly with an aminoglycoside for patients with more severe illness, although this practice is less common than when these guidelines were published.127 As noted previously, clear-cut evidence in humans that 2-drug, potentially synergistic therapy is more effective than a single drug is controversial, but such therapy is used by many clinicians in the face of rapidly progressive Pseudomonas or Klebsiella infection.128 Because of the increasing prevalence of MRSA, vancomycin or linezolid is often appropriate pending more definitive bacterial identification.

In immunocompromised patients, pneumonia is the most common life-threatening infection. Although a detailed discussion of the approach to pulmonary infection in these patients is beyond the scope of this chapter, certain antimicrobial strategies are worth noting. As outlined in Table 18-8, particular host defense defects are associated with specific infections, and initial therapy should reflect these associations. Even more than other ICU patients, these immunocompromised individuals are especially susceptible to nosocomial infection, both with the resident gram-negative flora and with Aspergillus species. Because of the importance of such infections in these patients, precise diagnosis is essential, using invasive techniques if necessary.

Bacteremia

Bacteremias occurring in hospitalized patients can be considered as arising from 2 separate pathogenic routes. One is a consequence of definable infection at an anatomic site such as the lung, biliary and urinary tracts, pancreas, and gut, as discussed later. The incidence of bacteremia with pneumonia varies according to which organism is causing the infection. Bacteremia occurs in 30% to 50% of patients with pneumococcal pneumonia, whereas this occurs in fewer than 10% with gram-negative or aspiration pneumonia. In pneumococcal infection, the bacteremia is typically caused by hematogenous seeding from the pulmonary infection, whereas in gram-negative VAP the bacteremia is likely related to heavy oropharyngeal colonization leading to skin contamination with consequent seeding of a central venous catheter. Thus, in the setting of VAP, all vascular access devices should be carefully evaluated, especially for duration of implantation, exit site erythema and drainage, and by blood cultures. The initial antimicrobial therapy of bacteremia secondary to invasive tissue infection is identical to that which would be prescribed in the absence of bacteremia.

Urinary Tract Infection

Urosepsis is relatively uncommon unless complicating factors are present, such as obstruction to urine flow, diabetes, advanced age, spinal cord injury, or bladder catheterization. The likely bacteria are also different from those in uncomplicated urinary tract infections (UTI) or asymptomatic UTIs. In these 2 instances, the likely organisms are E coli or other Enterobacteriaceae, which can be treated with a fluoroquinolone or trimethoprim-sulfamethoxazole. In patients with complicating factors, gram-positive and antibiotic-resistant gram-negative bacteria, including P aeruginosa, are more common.129 Consequently, therapy should be initiated with broader-spectrum antibiotics, usually fluoroquinolones or advanced-spectrum β-lactam agents such as ceftazidime, ampicillin-sulbactam, ticarcillin-clavulanate, aztreonam, or imipenem. As with all infections, antimicrobial therapy should be adjusted so that the spectrum is as narrow as possible.

A more common problem in hospitalized patients, especially those with urinary catheters, is nosocomial bacteriuria. The literature on UTI in patients with urinary catheters is confusing because definitions vary in different studies. For patients with indwelling urinary catheters, the Infectious Disease Society of America defines catheter-associated asymptomatic bacteriuria (CA-ASB) as greater than 105 CFU/mL of 1 or more bacteria collected from the catheter in patients without symptoms compatible with a UTI. A catheter-associated UTI (CA-UTI) is a CA-ASB with symptoms and signs (eg, a new-onset fever, altered mental status, lethargy, flank pain, and acute hematuria) that occur without another identified cause.130 The etiology of such infections is far different from that observed in community-acquired infections. Whereas E coli accounts for more than 85% of community-acquired urinary tract infections, it is responsible for only one-third of nosocomial UTIs. Enterococci, P aeruginosa, relatively antibiotic-resistant Enterobacteriaceae such as Klebsiella, Proteus, Enterobacter, S epidermidis, and Serratia, and Candida species now account for the majority of these infections.129,130 Antibiotic therapy can delay the appearance of bacteriuria, but the price to be paid if catheterization is maintained is that when infection does occur, it will be relatively resistant.

The incidence of bacteria in patients with urinary catheters is estimated to be 5% to 10% per day. Thus duration of catheterization is the most important risk factor for CA-ASB and CA-UTI. Moreover, approximately 15% of nosocomial bacteremias are attributable to bacteriuria. Thus catheters should be placed for well-defined reasons and removed as soon as possible. Moreover, Medicare will no longer reimburse a hospital for additional length of stay related to a CA-UTI that was not present on admission (http://www.cms.gov/HospitalAcqCond/06_Hospital-Acquired_Conditions.asp). Aside from minimizing the duration of catheterization, for short-term use (<30 days), studies suggest that the use of silver alloy or antibiotic-coated catheters may reduce the incidence of bacteriuria, although this is not generally recommended.131

Treatment of asymptomatic positive cultures when the catheter is still present is generally not indicated because this usually represents colonization rather than invasive infection, and long-term benefits of such therapy are unlikely. However, when to treat critically ill patients may be problematic because they may not be able to relate symptoms. Usually a quantitative colony count of more than 105 CFU is used as a criterion for treatment, with the choice of antibiotic guided by the culture and the urinary excretion of the antimicrobial.132 However, colony counts can be misleading because of the formation of biofilms and encrustations on the catheter surfaces, particularly from Proteus species, P aeruginosa, K pneumoniae, and Providencia species.130 Treatment is clearly indicated if symptoms develop and/or instrumentation of the urinary tract is to be carried out.

Clostridium Difficile

Clostridium difficile is the leading cause of gastrointestinal infection in the nosocomial environment. This pathogen is an important cause of fever and leukocytosis, which may precede the diarrheal phase. The spectrum of C difficile infection (CDI) may vary from mild and resolving without treatment to toxic megacolon or perforation and associated life-threatening septic shock. The latter requires emergency surgery, which is associated with a high mortality. Surgery should also be considered for any patient who does not respond to medical therapy within 24 to 48 hours. The most common etiology is prolonged antibiotic therapy, not necessarily recent, especially with clindamycin, fluoroquinolones, second- or third-generation cephalosporins, and broad-spectrum penicillins, with the exception of those that are β-lactamase-stable. Patients receiving immunosuppressive therapy appear to be at particularly high risk for severe CDI.

The pathogenesis of this disease is typically mediated by enterotoxin A or cytotoxin B, and bacteremia is extremely rare. The diagnosis is usually confirmed by an enzyme immunoassay for the aforementioned toxins. Recently, a hyper-virulent C difficile strain has emerged, which, although still sensitive to metronidazole and vancomycin, produces much more toxin and is associated with an increased mortality rate. The spores from C difficile are extremely hearty and impervious to antimicrobial therapy, which explains reports of relapse rates varying from 8% to 50% within 2 weeks to several months in successfully treated patients.133 Although oral metronidazole had been the therapy of choice even for relapsing cases, recent studies showed a higher rate of recovery from diarrheal symptoms and a lower rate of recurrence with oral vancomycin, which is now considered the drug of choice for severe infection.9,134 This may be related to metronidazole's complete absorption in the small intestine, reaching the colon via enterohepatic circulation and leakage from the bowel wall, which may also account for the undetectable colonic levels after resolution of diarrhea.

For patients being treated with other antibiotics for an ongoing pyogenic process, the treatment of C difficile is particularly vexing. In this setting, where continued broad-spectrum antibiotic use is required, we recommend continuing the CDI therapy in parallel with the other antimicrobial agents and extending the course of therapy for C difficile after completion of the other antibiotics typically for 5 to 10 days. The extended C difficile course of therapy is required because of the role of antibiotic therapy in provoking CDI by altering the normal bowel flora, which takes days to reconstitute. Patients with a history of CDI are more likely to have a recurrence with subsequent antibiotic therapy, probably because of the presence of latent spores. Patients who are severely ill from CDI may benefit from several adjuvant approaches, including minimization of other antimicrobial therapy, toxin-binding resins such as cholestyramine, fecal enemas, and, in severe cases, surgical resection of the colon (eg, in the setting of toxic megacolon). For a patient who undergoes a colectomy for C difficile toxic megacolon, the rectal stump, if it is left behind, may be a source of residual disease.

Intra-Abdominal Infections

There are many sources for intra-abdominal infections, and there are substantial differences in presentation and therapy, but treatments share 2 common elements: source control and antimicrobial administration. Source control is defined as any procedure or series of procedures that eliminate infectious foci and correct any anatomic defects contributing to the foci (eg, colonic perforation).114 In addition, in concert with pneumonias, intra-abdominal infections are frequently divided into community-acquired and hospital-associated. As noted previously, hospital-associated infections are thought to be microbiologically similar to true hospital-acquired (ie, nosocomial) intra-abdominal infections. Tigecycline, a relatively new glycylcycline antibiotic that is FDA-approved for complicated intra-abdominal infections, has a spectrum that covers gram-positive, gram-negative, and anaerobic bacteria. Because it covers MRD bacteria including MRSA, S epidermidis, and enterococci (including those that are vancomycin resistant) and is not inactivated by Amp C β-lactamases, it appears to be an excellent single drug for complicated intra-abdominal infections. However, it does not cover P aeruginosa.135 Unfortunately, because of a paucity of prospective studies, guidelines and recommendations are usually based on expert opinion. This section only considers 4 relatively common acute sources of intra-abdominal infections: cholangitis and acalculous cholecystitis, diverticulitis, appendicitis, and pancreatitis.

Cholangitis and Acalculous Cholecystitis

Cholangitis and cholecystitis, though technically an inflammation of the common bile duct and gallbladder, respectively, usually refer to their infection. Cholangitis may result from bile stasis and increased ductal pressure from an obstruction, from bacteria ascending from the small intestine, and possibly via the portal system or lymphatics. Biliary obstruction may also lead to cholecystitis, which usually begins as a sterile inflammatory process causing gallbladder distention and mural ischemia and ultimately leading to infection. Cholecystitis also occurs without obstruction in conjunction with total parenteral nutrition and serious illnesses, including trauma and burns. This entity, termed acalculous cholecystitis, is probably precipitated by factors that reduce gallbladder microcirculation, leading to ischemia. Regardless of etiology, if left untreated, the gallbladder may become gangrenous and ultimately perforate. Because symptoms and signs of both cholangitis and cholecystitis may be masked in the perioperative period, a high index of suspicion is required. Source control via percutaneous cholecystostomy for a critically ill patient may circumvent the need for a cholecystectomy. Nonoperative procedures are also available to decompress the biliary ducts. Once stabilized, procedures that are more definitive may be required.

Bacteria isolated from patients with infected cholecystitis and cholangitis usually reflect normal gut bacterial flora, most commonly E coli, followed by Klebsiella and Enterobacter species. Anaerobes, particularly Clostridium and Bacteroides, are found in approximately 5% to 10% of patients, as is E faecalis, which is usually found in association with other bacteria.136 Although there are no evidence-based guidelines for antibiotic treatment of either entity, by consensus, therapy should include those drugs that are active against these organisms, taking into account local resistance patterns.114 If either entity meets criteria for being health care–associated, as recommended for all such infections, antimicrobial therapy should be broadened to account for higher resistance rates and less common bacteria (eg, P aeruginosa), guided by local epidemiology and sensitivities.

Diverticulitis

Diverticulitis is usually caused by a sigmoid perforation resulting from continued mucus secretion in a diverticulum that has become obstructed at its neck. Management depends on severity, usually defined by the Hinchey classification137: stage 1, localized perforation with pericolic phlegmon; stage 2, perforation with abscess; stage 3, purulent peritonitis; and stage 4, free perforation with fecal peritonitis. Stages 1 and 2 are associated with less than a 5% mortality rate, whereas mortality rates for stages 3 and 4 are approximately 13% and 43%, respectively.138 Stage 1 can be treated initially with antibiotics alone, whereas stages 3 and 4 usually require early operative intervention. Stage 2 is often managed with antibiotics and percutaneous drainage. The bacteriology, as expected, generally reflects that of the colon and is usually polymicrobial. As many as 5 different organisms may be recovered. Anaerobic organisms, especially B fragilis because it outnumbers other bacteria in the colon by roughly 100:1, are common, but aerobic gram-negative bacteria, especially E coli, are also usually present. Thus antibiotic therapy should be directed against both anaerobic and gram-negative bacteria. However, there is only 1 randomized study of different antibiotic regimens, cefoxitin alone or gentamicin plus clindamycin, with the former having a higher clinical success rate. Although enterococci are present in approximately 10% of cultures, the need to cover these organisms is controversial but recommended for health care–associated diverticulitis.114

Appendicitis

The pathophysiology and management of acute appendicitis in many ways resembles that of diverticulitis. Appendiceal obstruction is likely the precipitating event, and the course can range from mild, resolving without any treatment, to gangrene and perforation with abscess formation or peritonitis. Culture data are also similar to those found in diverticulitis, and as many as 14 different bacteria have been recovered from cultures. In addition, as with diverticulitis, although controversial, it is likely that some patients can be managed with antibiotics alone, whereas those with gangrene or perforation require prompt surgery. There are very few studies defining these populations, but there are advocates of a trial of antibiotics without surgery for patients without perforation or gangrene. Generally, if the patient is improving, surgery may be unnecessary, although recurrence rates without surgery are approximately 14% within a year.139 For patients without gangrene, perforation, or abscesses, some advocate that antibiotics be given preoperatively and for a maximum of 24 hours postoperatively, whereas treatment for those with such complications should be continued for approximately 7 days or until signs and symptoms of infection have resolved.114

Acute Pancreatitis

Antibiotic management in acute pancreatitis is challenging because the severity of illness often indicates a therapeutic emergency, yet the traditional markers of infection may not be present and there may be limited supporting clinical data. Patients with severe pancreatitis without infection may present with septic pathophysiology including hypotension, tachycardia, hypoxemia, tachypnea, metabolic acidosis, leukocytosis with a left shift, thrombocytopenia, elevated lactate, and coagulopathy. These findings are all consistent with the pathogenesis of this disease, where an inciting event such as alcohol, gallstones, or trauma leads to pancreatic injury and inflammation, which in turn leads to autodigestion, liquefaction, and necrosis with associated inflammatory cytokine release. If the necrotic pancreatic tissue becomes infected by, for example, biliary reflux, colonic bacterial translocation, or hematogenous seeding, there is an associated increased morbidity and mortality.140 It is important to note that infection is rarely the inciting event but rather a sequel of pancreatic necrosis and often occurs weeks into the hospital course. Abdominal imaging, with a contrast CT scan, has enabled stratification of those patients at risk for developing superimposed infection by an increasing degree of pancreatic necrosis. Unfortunately, imaging, like the physical examination and laboratory evaluation, cannot reliably distinguish an infection from sterile inflammation. Therefore, the time to initiate antibiotic therapy in the presence of documented pancreatic necrosis remains largely based on the patient's clinical state, with deterioration often used as the trigger. Logically, a fine-needle aspirate of necrotic areas should be diagnostic of infection. However, there is a relatively high false-negative rate, and the trend seems to be moving away from its use.141 Although an infected necrotic pancreas has been thought to require source control with surgical debridement or percutaneous drainage in addition to antibiotic therapy and supportive care, there is a trend toward delaying intervention so that the necrotic area can consolidate. However, no studies conclusively support such delays.

It would seem logical to administer antibiotics preemptively, to prevent infection of necrotic pancreatic tissue. However, multiple randomized studies and meta-analyses have failed to find support for this concept, and there are risks of selecting for resistant organisms if the necrotic tissue becomes infected.142 For these reasons, this practice is no longer recommended. Nonetheless, in the critically ill patient for whom it is too risky to delay antimicrobial therapy, targeting the typical infecting organisms, which includes aerobic enteric gram-negative rods and gram-positive cocci, is appropriate. A variety of antibiotics, such as fluoroquinolones, imipenem, ceftazidime, cefepime, metronidazole, clindamycin, chloramphenicol, doxycycline, and fluconazole, have been shown to achieve pancreatic levels above the MIC for the commonly encountered bacteria, but either imipenem or a fluoroquinolone plus metronidazole are commonly recommended. However, because resistant organisms such as S aureus species (which usually are not sensitive to these antibiotics) and Candida species (including C glabrata, which is generally not susceptible to fluconazole) are emerging as significant pathogens in severe acute pancreatitis, it is likely that these recommendations will be modified. Interestingly, multiple studies have indicated that enteral, as opposed to parenteral nutrition, initiated as soon as tolerated may decrease the infection rate and complications in severe acute pancreatitis with no difference between gastric and jejunal routes.143

HIV Infection and AIDS

A remarkable body of information regarding the treatment of HIV infection has emerged since 1981, when acquired immune deficiency syndrome (AIDS) was first recognized. The occurrence of oral thrush, Pneumocystis jiroveci pneumonia, Toxoplasma encephalitis, and other opportunistic infections in apparently healthy gay males was quickly recognized as something unusual; that is, the net state of immunosuppression should not have been great enough to allow such infections to occur. Very quickly, the characteristics of this epidemic emerged: profound and progressive immune compromise and efficient transmission from infected individuals by intimate contact, blood transfusion and organ transplantation, intravenous drug abuse, and perinatal route. In 1984 the identification of the cause of these events, infection with a unique retrovirus now known as the human immunodeficiency virus (HIV), was reported. HIV is now recognized as the cause of a worldwide pandemic of infection, particularly those due to opportunistic organisms, as well as Kaposi sarcoma and other malignancies. More than 30 million individuals are believed to have been infected by HIV since 1981, with devastating consequences.144–146 Table 18-9 lists the CDC definitions of the stages of HIV infection, including AIDS. The long list of AIDS-defining conditions is similar to those found in severely immunocompromised patients.147

Three different phases of disease have been recognized once HIV has been acquired148,149:

Primary HIV Infection

A mononucleosis-like illness is observed in ∼50% of individuals 2 to 6 weeks after infection. Primary infection is associated with a marked increase in plasma viremia, which can exceed 1,000,000 copies per milliliter; a significant decrease in the CD4 T-lymphocyte count; and a large increase in the blood CD8 T-lymphocyte count.

Chronic Asymptomatic Stage

An extended phase of clinical latency occurs, persisting for 10 to 12 years in the majority of individuals. An estimated 20% of individuals, the so-called rapid progressors, have an accelerated course, having full-blown AIDS in less than 5 years; conversely, approximately10%, the so-called slow progressors or non-progressors, remain free of AIDS for 7 to 12 or more years. At the end of this period, the level of viremia rises rapidly and there is a significant decrease in the CD4 T-lymphocyte count. AIDS-defining opportunistic infections begin to appear.

AIDS

In the absence of effective therapy, there is a progressive decrease in CD4-positive lymphocytes and an increase in viral load. These events are correlated with recurrent opportunistic infection, the occurrence of certain malignancies, and death in 2 to 3 years.

The specifics of anti-HIV therapy are constantly evolving, although certain principles remain constant: HIV replication remains at a very high level throughout the stages of illness. This high rate of replication is coupled with a remarkable amount of errors in the function of the reverse transcriptase (the daily production of ∼108 to 1010 virions and a mutation rate of 3 × 10−3). The frequency of these events virtually guarantees the presence and the rapid development of mutants that are responsible for drug resistance. Resistant clones of HIV may be present even before the initiation of any therapy. Such findings mandate that multiple drugs will be needed to treat this infection effectively.

The general principle that applies to HIV therapy is "hit early and hit hard," with multiple drugs being started simultaneously.149-151 Although the precise point when therapy should be instituted is still being studied, at present it is recommended that highly active anti-retroviral therapy (HAART) or ART, which consists of multidrug regimens, be initiated in asymptomatic patients who have circulating CD4 T-lymphocyte counts fewer than 500 cells/mL and/or an HIV RNA load of 5000 to 10,000 copies/mL. Initiation of treatment is also recommended regardless of CD4 count if the viral load exceeds 100,000 copies/mL or if the CD4 count declines by more than 100 per year. It is also recommended for all patients older than 60 years and all pregnant women and when there is coinfection with hepatitis B or C. Further details may be found in Thompson et al.152

Treatment of HIV is very complex, both in terms of drug therapy and because HIV can impair virtually any organ system. Moreover, the antiviral regimens are constantly evolving and may require changes during the course of the disease. There are now a large number of drugs available for HAART regimens, which generally fall into 1 of 5 classes based on mechanism of action (Table 18-10). Aside from the development of resistance, 2 key issues must be considered for all of these drugs: side effects (eg, pancreatitis, hepatotoxicity, and a lipodystrophy syndrome) and drug–drug interactions, including changes in drug metabolism by the hepatic cytochrome P450 enzymes (Table 18-10). This is made even more complex because to help with compliance, combinations of antiviral drugs supplied in a single pill are often appropriate. For these reasons, specialists in HIV should always be involved.

The goal of HAART therapy is to lower and maintain the HIV viral load to undetectable levels. Such an approach has been quite effective, but drug resistance remains an important impediment in treatment. Clinically important resistance is particularly likely when the level of mutations and the ability of these mutant strains to replicate is relatively high.151,153-155

Invasive Fungal Infection

Advances in therapy for invasive fungal diseases have improved survival of immunocompromised patients (eg, HIV, transplant, and cancer patients and patients with autoimmune diseases). The most common cause of invasive fungal infection has long been Candida species, and this is still true, but there has been an increase in the range of fungal species causing serious infection, as well as an increase in antimicrobial resistance. In addition, new sites of infection are being seen. For example, the ICU use of invasive vascular access devices is becoming an important source for candidemia, whereas this was uncommon in the past. Today, Candida species are the sixth most common nosocomial isolate and fourth most common cause of nosocomial bloodstream infections. As this has occurred, the range of candidal species has changed from azole susceptible C albicans to more resistant non-albicans infections.156–158

The risk of invasive fungal infection is largely determined by the interaction of the following 4 factors156:

1. The environmental exposure to which a patient is subjected is an important factor in the occurrence of many fungal infections. The density of organisms that are aerosolized and then inhaled is the critical first step in the initiation of many fungal infections. These exposures can occur in the community or within the hospital. In the hospital, the patient is vulnerable to organisms that are aerosolized and inhaled on the ward where the patient resides (ie, domiciliary exposure) or to exposure to aerosols while being transported through the hospital for studies or procedures. In both instances, ongoing construction is the most common activity resulting in aerosolization of infectious organisms, particularly Aspergillus species that have resided in the building interstices. In addition, person-to-person spread via the hands of medical personnel is a relatively common event, with the spread of antimicrobial-resistant Candida species being a particular problem.156-160

2. Also important is the patient's net state of immunosuppression, which is a complex function determined by deficits in innate host defenses that may occur in conjunction with many conditions such as underlying diseases and their therapy; infection with immunomodulating viruses including HIV, cytomegalovirus, and the hepatitis viruses; and the presence of protein-calorie malnutrition.156

3. The presence of foreign bodies such as orthopedic prostheses, devitalized tissues, undrained fluid collections, and invasive vascular and urinary catheters contributes significantly to the pathogenesis of invasive Candida infection. Whether one is dealing with candidemia associated with vascular access catheters, peritonitis in association with peritoneal dialysis catheters, or orthopedic prosthesis infection, the chances of successful therapy are greatly enhanced by the removal of the foreign body in association with effective antifungal therapy.156,161,162

4. Darwinian factors can play an important role as well. Thus prolonged therapy with broad-spectrum antibacterial drugs will create an ecologic niche easily occupied by Candida species. The presence of excess growth factors such as glucose and iron can significantly increase the occurrence of mucocutaneous candidiasis, invasive mucormycosis, and other types of invasive fungal infections. Unless the ecologic niche is eliminated, recurrent fungal infections may occur. For example, therapy with the newer azoles can be associated with the development of mucormycosis.156,158,160,161

The fungal species capable of causing invasive infection can be divided into 3 general categories:

1. The geographically restricted systemic mycoses, blastomycosis, coccidioidomycosis, and histoplasmosis are important in North America. In addition, paracoccidioidomycosis in Latin America and penicilliosis in Southeast Asia exhibit similar clinical and epidemiologic patterns. These are dimorphic fungi that grow as molds in soil and as yeast-like forms in tissue. Invasive infection with one of these is greatly amplified by the presence of immunocompromise. Treatment of these infections at present has 2 parts: induction therapy with amphotericin to gain control of the disease and then prolonged oral therapy with an azole to consolidate the antifungal effects. At present, itraconazole is the therapy of choice for this purpose, with the exception of fluconazole for treating coccidioidomycosis.156,161-165

2. The opportunistic fungi are ubiquitous in the environment, where they are nonpathogenic, particularly for normal hosts, but they can cause invasive infection when the inhaled inoculum harbors a high microbial burden and when the host is immunocompromised. These organisms include Aspergillus species, Cryptococcus neoformans, and Sporothrix schenckii. Voriconazole is currently the treatment of choice for Aspergillus infection and fluconazole for cryptococcal infection, often after induction therapy with amphotericin plus flucytosine and saturated potassium iodide or itraconazole for sporotrichosis.156,161,162,166,167

3. The newly emerging fungi now account for approximately 10% of invasive fungal infections, with the major species involved, including Mucorales, Fusarium, and Trichosporon. These species tend to be more resistant to fluconazole, echinocandins, and even amphotericin. Drugs such as voriconazole and posaconazole should be considered as the first choices for therapy. Posaconazole is the first really effective agent for mucormycosis, particularly when combined with surgery.156,161,162

Several types of exposure are important in the development of invasive fungal infections, with the effects of both being amplified if immunocompromise is present. Candidal infections frequently result from contaminated vascular access devices. Less commonly, infections may occur when the mucocutaneous surfaces are compromised. In this case, not only is candidal infection a concern, but invasion by Aspergillus, Mucorales, and other fungal species can also occur.157,158,162,168 The respiratory tract can also be an important portal for fungal infections. Inhalation of the organisms can result in invasive fungal infection of the nasal sinuses and the lungs, with Aspergillus species being the most common invader. The first host response to these organisms is the migration of polymorphonuclear leukocytes, which can kill the inhaled inoculum. If the inhaled organisms escape this first defense, bloodstream invasion with the potential for metastatic spread may occur, as may tissue invasion both at the primary site of infection and metastatic sites. Cell-mediated immunity including alveolar macrophages is then mobilized. With histoplasmosis, persistent infection of macrophages is established, making lipid-associated amphotericin B particularly effective by targeting the macrophages with the lipid moiety. Thus an increased risk of invasive aspergillosis and other such infections (eg, fusariosis) is to be expected in patients with neutropenia and/or impaired cell-mediated immunity.159,160,162,168-170

Drugs that are available for the treatment of invasive fungal infection are as follows:

1. Amphotericin B is a broad-spectrum polyene that acts by binding to fungal cell membranes, specifically ergosterol (for which it has a >500-fold increase in affinity when compared with binding to cholesterol in mammalian cell membranes). Binding to ergosterol results in increased membrane permeability and cytolysis, which are the probable mechanisms of fungal injury and death. Amphotericin B remains the broadest-spectrum antifungal known, producing fungicidal effects against the majority of pathologic fungi, including those that are resistant to other antifungal compounds. The dose-limiting toxicity is to the kidneys, which is particularly important when such nephrotoxic drugs such as gentamicin or cyclosporine are administered to a patient receiving amphotericin. In addition, the administration of amphotericin B usually produces a "cytokine storm," which can include not only fever and chills, but also hypotension. These storms are usually not a problem after the first several days of therapy. Lipid-associated amphotericin appears to decrease but not eliminate both the febrile reactions and the renal toxicity. However, as with all the amphotericin products, the optimal dosing regimen is not known, but the recommended doses are 3 to 5 mg/kg/d for a lipid-associated drug and 0.1 to 1.5 mg/kg/d for the standard amphotericin. The optimal duration of therapy also is unclear. Our practice is to treat until all overt disease is gone and then add a buffer for safety. The duration of this buffer is a clinical decision usually determined by the nature of the original infection and the speed with which the patient responded.

2. Flucytosine is synergistic with amphotericin for the treatment of cryptococcosis. This regimen protects against a single step mutation to flucytosine resistance. Dose-related hepatic and bone marrow toxicity can occur with the use of flucytosine. A common approach is to administer amphotericin and flucytosine for 7 to 10 days to gain control of the process and then complete the course of therapy with oral fluconazole.171,172

3. The azoles act by blocking the cytochrome P450-associated enzyme lanosterol synthetase, which results in the inhibition of ergosterol synthesis. The effect is fungistatic. There are 5 azoles that have been approved for the treatment of systemic fungal invasion. However, ketoconazole and miconazole are essentially of historical interest only and are rarely prescribed today.

   Itraconazole has an appealing spectrum of activity, including Aspergillus and other fungal species. The problem has been poor and unreliable drug delivery. With substitution of an oral suspension, more reliable bioavailability has been achieved, and the utility of the drug should be reassessed. Up to now, the major use of this drug has been in oral "wrap-up" therapy after amphotericin had gained control of the infection.172

   The advent of fluconazole, in contrast, was a major advance in antifungal therapy. Its only weakness is a rather narrow spectrum of activity limited mostly to Candida species and C neoformans. Most C krusei are resistant to fluconazole, as are approximately 10% to 20% of Candida tropicalis.173 Pharmacokinetically, fluconazole penetrates into the urinary tract, the eye, and the brain, as well as the spleen, liver, and other sites. The bioavailability when given by mouth is complete, and thus the dose given by mouth is the same as the parenteral dose. Side effects are relatively uncommon or minor and include a measles-like rash, hepatocellular dysfunction, and minimal gastrointestinal complaints. Fluconazole does interact with cytochrome P450 enzymes, as do all the azoles, and thus can increase the blood levels of such important therapies as cyclosporine and tacrolimus.172

   Voriconazole, a newer azole, can be administered either orally or parenterally.174 Its candidal spectrum is broader than that of fluconazole, with only about 5% of C krusei resistant.173 Voriconazole is fungicidal for Aspergillus and is the most effective of the current anti-Aspergillus drugs, including amphotericin, as well as having efficacy for other resistant molds. Side effects are similar to those of other azoles (rash, hepatocellular dysfunction, nausea, etc.). Voriconazole does have a unique side effect: the occurrence of visual effects, including bright colors, lights, and so forth, akin to that seen with digitalis toxicity. Such symptoms appear to be most common with high doses and appear to be due to retinal dysfunction. All such symptoms disappear when the drug is stopped, and extensive study has failed to reveal any persistent visual or structural consequences. Voriconazole appears to be the drug of choice for invasive aspergillosis. Treatment with voriconazole, however, carries a small risk of selecting for Mucorales infection (mucormycosis).172,175

   Posaconazole is the newest azole and is available only in oral formulation.176,180 Although its antifungal spectrum includes most candidal species and many molds including aspergillosis, clinical data are still limited. It is, however, approved for prophylaxis of candidal infections and aspergillus in immunocompromised patients.

4. The echinocandins are large lipopeptide molecules that damage fungal cell walls by inhibiting the synthesis of 1,3-β-d-glucan, a fungal cell wall component. In vitro, these compounds are fungicidal for Candida and fungistatic for Aspergillus species. These molecules appear to have activity against candidal strains that are amphotericin, fluconazole, and itraconazole resistant. On the other hand, species such as C neoformans and the Mucorales are inherently resistant to echinocandin therapy, presumably because they do not possess significant β-glucan in their cell walls.177

Three echinocandins—caspofungin, micafungin, and anidulafungin—have been approved by the FDA. All 3 must be administered intravenously and have similar antifungal spectra and pharmacodynamics. They all exhibit a post-antifungal effect, which is analogous to the post-antibiotic effect, with durations that vary among different fungi. Interestingly, there is in vitro evidence that they also exhibit the so-called eagle effect, which is an increase in fungal growth at concentrations well above their MIC, but it is unclear if this occurs in vivo.178 The echinocandins seem to be equivalent to fluconazole for the treatment of invasive candidiasis and theoretically should be superior for treatment of Candida normally resistant to fluconazole to be useful in the treatment of both drug-resistant candidiasis and invasive aspergillosis. The results thus far suggest comparable efficacy to that achieved with amphotericin and the licensed azoles. The possibility of achieving better results with combination therapy that includes an echinocandin is particularly intriguing. The cell wall damage that is caused by echinocandins is reminiscent of synergistic therapy of enterococcal infection with ampicillin and gentamicin, with the echinocandin playing the role of a cell wall active agent that potentiates the penetration of additional drugs. Indeed, the echinocandins have been called "the antifungal penicillins."177,179

The Future of Antifungal Therapy

Since the advent of new drugs for the treatment of fungal disease, great progress has been made in terms of efficacy and adverse events. What we need now is a new generation of diagnostic tests that will inform us regarding the appropriate drugs to use, how long to treat, whether multiple drugs should be deployed, and a determination to define microbial load objectively, which will allow us to treat preemptively rather than empirically. The present data on the use of (1-3)-β-d-glucan testing in defining the presence or absence of invasive fungal infection suggest that we are getting closer to that possibility.

One of the greatest frustrations when caring for a potentially septic patient is having to wait 1 to 2 days for culture results to find out if the patient is infected and then to wait another day or 2 to discover the identity and later the susceptibility profile of the infecting organism. With the molecular biologic revolution, new tests have emerged, typically polymerase chain reaction–based technology, enabling the rapid identification of specific pathogens directly from the infected body site. Nonetheless, this technology is prone to many of the same interpretation challenges. For example, does a positive result represent colonization or disease? To help both diagnostically and prognostically, the development of new assays are focusing on the levels (or presence) of mediators in the inflammatory cascade or on circulating bacterial moieties, such as endotoxin or nuclear factor kappa B, to improve our ability to risk stratify patients or determine the infecting pathogen.181-184

A broad array of pathogens can now be identified by molecular techniques, including viruses (HIV, CMV), fungi (Candida, Aspergillus), and bacteria (VRE, E coli, M tuberculosis, Bacteroides).185-190 As learned for HIV-positive patients, monitoring the HIV viral load has become an important parameter in gauging the success of therapy. The development of technology may allow us to provide pathogen-directed therapy earlier in a patient's illness, enabling more focused narrow-spectrum antimicrobial use. This would diminish the selective pressure, which leads to the emergence and dissemination of resistance, as well as yielding novel markers to gauge the duration and intensity of therapy.

Antimicrobial therapy is complicated by the increasing incidence of bacterial, fungal, and viral resistance. As a result, a broader range of antimicrobial agents has come into use, with the basic approach to antimicrobial therapy in these patients being front-loading broad-spectrum antimicrobials. This therapy is then modified to relatively narrow spectrum-specific therapy. The regimens chosen and the doses prescribed are based on the principle that effective therapy is dependent on the delivery of a level of antimicrobial agent to the site of infection that significantly exceeds the MIC of the invading organism. Bactericidal therapy is essential when dealing with cardiovascular infection, CNS infection, prosthesis-associated infection, osteomyelitis, and infection in the neutropenic patient. Combination therapy may be important when dealing with life-threatening enterococcal, P aeruginosa, and perhaps other gram-negative infections, but the definition of synergy varies and the clinical efficacy is not firmly established. Finally, it is hoped that a better definition of high-risk patients will permit the more effective use of preemptive antimicrobial regimens. It is clear that many questions remain unanswered and that many therapeutic decisions must be based on a thorough understanding of the properties of antimicrobials and the pathogenesis of infections, rather than efficacy demonstrated in clinical studies.

• Bonten MJ, Chastre J, Craig WA, et al. Guidelines for the management of adults with hospital-acquired, ventilator-associated, and healthcare-associated pneumonia. Am J Resp Crit Care Med. 2005;171:388-416.
• Craig WA. Pharmacokinetic/pharmacodynamic parameters: rationale for antibacterial dosing of mice and men. Clin Infect Dis. 1998;26:1-10.
• Hooton TM, Bradley SF, Cardenas DD, et al. Diagnosis, prevention, and treatment of catheter-associated urinary tract infection in adults: 2009 International Clinical Practice Guidelines from the Infectious Diseases Society of America. Clin Infect Dis. 2010;50:625-663.
• Liu C, Bayer A, Cosgrove SE, et al. Clinical practice guidelines by the Infectious Diseases Society of America for the treatment of methicillin-resistant Staphylococcus aureus infections in adults and children. Clin Infect Dis. 2011;52:285-292.
• Mermel LA, Allon M, Bouza E, et al. Clinical practice guidelines for the diagnosis and management of intravascular catheter–related infection: 2009 Update by the Infectious Diseases Society of America. Clin Infect Dis. 2009;49:1-45.
• Nishimura RA, Carabello BA, Faxon DP, et al. ACC/AHA 2008 guideline update on valvular heart disease: focused update on infective endocarditis: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines: endorsed by the Society of Cardiovascular Anesthesiologists, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons. Circulation. 2008;118:887-896.
• Rubin RH. Clinical Approach to Infection in the Compromised Host. 4th ed. New York, NY: Kluwer/Academic Press; 2003.
• Rubin RH. Surgical wound infection: epidemiology, pathogenesis, diagnosis and management. BMC Infect Dis. 2006;6:171.
• Solomkin JS, Mazuski JE, Bradley JS, et al. Diagnosis and management of complicated intra-abdominal infection in adults and children: guidelines by the Surgical Infection Society and the Infectious Diseases Society of America. Clin Infect Dis. 2010;50:133-164.
• Teplick R. Caveats for comparing catheter-associated bloodstream infection rates. Crit Care Med. 2011;39:392-394.