BJA Education, 23(12): 480e487 (2023)

doi: 10.1016/j.bjae.2023.09.001
Advance Access Publication Date: 19 October 2023

Matrix codes: 1A02,

2C07, 3C00

Empirical decision-making for antimicrobial therapy

in critically ill patients
M. Ippolito1,2 and A. Cortegiani1,2,*
1
University of Palermo, Palermo, Italy and 2Policlinico Paolo Giaccone, Palermo, Italy
*Corresponding author: [email protected]

Keywords: antibiotics; infections; intensive care units; sepsis

Learning objectives Key points

By reading this article you should be able to:
The initiation of empirical antimicrobial therapy

Explain the main determinants of the appropri- is part of a complex decision-making process.
ateness of empirical antimicrobial therapy in
Multidisciplinary teams should be involved early.
critically ill patients.
Consider the most probable infection sites, mi-

Identify risk factors for multidrug-resistant in- croorganisms and local epidemiology.
fections and pharmacokinetic/pharmacody-
Risk factors for multidrug-resistant pathogens
namic issues. must be assessed.

Implement multidisciplinary team discussions
Strategies to optimise dosing are needed in criti-
on challenges in the management of infections in cally ill patients.
critically ill patients.

Apply best practices for empirical antibiotic mortality is higher in ICU-acquired infections or in cases caused
antimicrobial therapy in critically ill patients. by antibiotic-resistant microorganisms (e.g. vancomycin-
resistant Enterococci [VRE], Klebsiella resistant to third-
generation cephalosporins or carbapenems, or carbapenem-
The prevalence of infections among critically ill patients resistant Acinetobacter species) compared with community-
admitted to the intensive care unit (ICU) is high: an average of acquired infections.1
54% of patients have suspected or proved infection, including Critically ill patients with sepsis or septic shock benefit
22% ICU-acquired infections.1 According to a recent interna- most from receiving timely and appropriate antimicrobial
tional observational 24-h point-prevalence study, the risk of therapies. Several studies have shown a reduced length of
stay and mortality associated with early and appropriate an-
timicrobials administration.2,3 The last published version of
the Survival Sepsis Campaign Guidelines made a strong
Andrea Cortegiani MD is a consultant in anesthesiology and
recommendation in favour of the immediate administration
intensive care medicine at the University Hospital Policlinico Paolo
of antimicrobials (ideally within 1 h from recognition) to adult
Giaccone and associate professor in anesthesiology at the University
patients with septic shock or a high likelihood of sepsis.4 For
of Palermo. His main research interests are infection and sepsis in
patients with possible sepsis without shock, the guidelines
critically ill patients, non-invasive respiratory support and research
recommend a rapid assessment of the likelihood of infectious
methodology. He is the current chair of the scientific committee of the
compared with non-infectious causes of acute illness and
Italian Society of Anesthesia, Analgesia and Intensive Care
suggest treatment with antimicrobials within 3 h if concern
(SIAARTI) and an editorial board member of the British Journal of
for infection persists.4
Anaesthesia.
After appropriate clinical samples are taken, appropriately
Mariachiara Ippolito MD is an early career anesthesiologist and chosen antibiotic(s) should be given in a timely manner while
intensivist working in the Department of Anesthesia and Intensive balancing both the need for activity against the causative
Care at the University of Palermo, with research interests in infection pathogens and the avoidance of unnecessary broad-spectrum
and sepsis and mechanical ventilation. coverage, which can potentially lead to antimicrobial

Accepted: 5 September 2023

© 2023 The Author(s). Published by Elsevier Ltd on behalf of British Journal of Anaesthesia. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
For Permissions, please email: [email protected]
 Empirical antimicrobial therapy in ICU

resistance. Moreover, overuse of antibiotics can cause harm,

such as drug toxicity, modification of microbiota in terms of
reduced diversity and richness and opportunistic infections
such as Clostridioides difficile.5 Thus, the use of rapid and
appropriate empirical antimicrobial therapy is pivotal to Infection site
improve patients’ outcomes,4 but decision-making is partic-
ularly challenging in the context of critical care. It should be
guided by local epidemiology, the probable source of the
Dose Most probable
infection and patient-specific factors including the severity of optimisation pathogens
illness, known microbiological colonisation status (e.g. car- Best practice for
riage or colonisation of methicillin-resistant Staphylococcus empirical antibiotic
aureus [MRSA]/VRE/extended-spectrum b-lactamases [ESBL]/ treatment in
AmpC b-lactamases [AmpC]/carbapenem resistance or other critically ill patients
multidrug-resistant [MDR] pathogens) and risk factors for in-
fections caused by MDR pathogens.6 However, even after Risk factors
Clinical
for MDR
comprehensive clinical reasoning, 14e78% of patients may severity
receive inappropriate empirical antimicrobial therapy, pathogens
increasing the risk of negative outcomes.7e9 Molecular
Other issues must be considered of paramount importance microbiology
when selecting an antimicrobial therapy, including pharma- results
cokinetic/pharmacodynamic (PK/PD) considerations typical of
critically ill patients (e.g. acute kidney injury [AKI], hypo-
albuminaemia, continuous renal replacement therapies, vas- Fig 1 Mind map on the best practice for empirical antimicrobial therapy.
oplegia and capillary leak syndrome, altered volume body
compartments and third fluid space accumulation). Rapid
microbiology technologies can help by quickly detecting percutaneous drainages) or diagnostics (e.g. CT scans) and
pathogens and clinically relevant major genetic determinants their appropriate timing.11
of antibiotic resistance from common biological samples, thus Moreover, the infection site influences the choice of anti-
widening the potential for early and appropriate antimicrobial microbial even before microbiological results, because of the
therapy. The complexity of antimicrobial therapy in critically tissue penetration characteristics of different drugs. The tis-
ill patients can be compared to a ‘puzzle composed of pieces that sue availability of each drug varies based on both the drug and
must merge together to ensure both clinical cure and prevention of the tissue characteristics. Examples of tissues with significant
resistance spread’.10 This highlights the need to take into ac- barriers to drug penetration are the CNS and the lungs,
count infection site, patients’ clinical conditions, PK/PD and whereas sites easily reached are usually the bloodstream and
characteristics of antimicrobials. Indeed, the long turnaround the urinary tract.
time of standard microbiological cultures and in vivo antimi- In addition to tissue-dependent determinants of antimi-
crobial susceptibility testing patterns, although pivotal for crobial penetration (i.e. the presence of specialised mem-
selecting targeted treatment, hamper their usefulness in branes or the degree of vascularisation), other determinants
making decisions about empirical treatment. depend on the drug (molecular weight, degree of protein
The aim of this review is to provide practical insights and binding, lipid solubility and degree of ionisation).12 Low mo-
best practice principles to the intensivist for the initial pre- lecular weight, lipophilicity and low degree of protein binding
scriptions of empirical antimicrobial therapy in critically ill facilitate tissue and cellular penetration, with easier avail-
patients (Fig. 1). Full descriptions of the principles of antimi- ability of the antimicrobial at the infection site. Conversely,
crobial stewardship and de-escalation are outside the scope of drugs with high molecular weight, a hydrophilic nature and a
this review. Indeed, the initiation of empirical therapy is the high degree of protein binding, are less available at the
start of a complex decision-making process that must include infection site, thus higher doses are needed to reach appro-
a plan for stopping or de-escalation of treatment, which is a priate effect-site concentration. For example, quinolones are
cornerstone of antimicrobial stewardship in the ICU. lipophilic and reach high concentrations in all tissues.12
Some settings and groups, such as immunocompromised Among the hydrophilic drugs, aminoglycosides have low
patients or fungal/opportunistic infections are not covered in penetration into common sites of infections, such as the lungs
this review, and they would need specific reasoning on a wider and low respiratory tracts, because of limited distribution to
range of probable pathogens and infections. extravascular spaces.12,13 b-Lactams are exceptions. They are
hydrophilic agents for which it is possible to achieve high
distribution to interstitial compartments, owing to their wide
Sites of infection and favourable therapeutic window, allowing high doses to be
An effective empirical antimicrobial therapy should be first used. This results in higher tissue concentrations (e.g. respi-
guided by accurate identification of the most probable site of ratory system, abdomen, bone and skin)1 especially in pa-
infection. This is based on clinical presentation and diagnostic tients with altered volume of distribution (e.g. those with
information. The identification of the source allows for sepsis).3 Another important determinant is the presence of
reasoning on the most commonly involved pathogens, inflammation, which may increase tissue distribution of an-
appropriate microbiological sampling, and the eventual need timicrobials because of leaky tissue/blood barriers, the open-
for source control procedures (Table 1).11 Indeed, the sus- ing of intercellular tight junctions and oedema formation. An
pected infection site may trigger multidisciplinary consulta- example is the increased penetration of b-lactams (e.g. cef-
tions, often pivotal for source control procedures (e.g. surgery, triaxone) into the cerebrospinal fluid (CSF) in meningitis,

BJA Education - Volume 23, Number 12, 2023 481

Empirical antimicrobial therapy in ICU

Table 1 Main characteristics and suggested sampling for sites of infection.

*Please consider that the selection of empirical antibiotic therapy should not be based only on the ‘best’ penetration in the infection
site. For example, vancomycin is widely used for several infections in different body sites (i.e. CNS, soft tissue, abdomen, bloodstream)
despite its pharmacokinetic characteristics.

Suggested Examples of source Peculiarities Notes on antibiotics penetration*

microbiological control
sampling

Lungs Bronchoalveolar Drainage of pleural The epithelial lining fluid is b-lactams have good penetration, but
lavage, protected empyema the main site to reach. higher doses can be considered to
specimen brushing, Diffusion of antibiotics is increase it. Consider fluoroquinolones.
or endotracheal hampered by the alveolar Linezolid has good penetration. Colistin
aspirate; blood eblood barrier, as the and aminoglycosides have poor
cultures alveolar wall has no penetration; nebulisation can be
fenestration considered
Urinary tract Urinary cultures; Removal of urinary Good penetration for Consider b-lactams, aminoglycosides,
blood cultures catheter, antimicrobials with fluoroquinolones achieve good
nephrostomy urinary excretion concentrations
Abdomen Percutaneous Drainages, Enterobacterales, Consider b-lactams, aminoglycosides,
drainage or open debridement, or enterococci or Candida spp. fluoroquinolones and tigecycline have
surgical sampling both are commonly involved good penetration
during source
control; blood
cultures
Central CSF sample; blood Removal of invasive Penetration dependent on Consider b-lactams and rifampicin have
nervous cultures devices; surgical integrity of bloodebrain good penetration. Consider
system decompression barrier and inflammation aminoglycosides with intrathecal
of meningitis (e.g. reduced administration.
in case of acidosis) Vancomycin i.v. or intrathecal is an
option. Meropenem high dose is
commonly used in the UK.
Colistin has poor penetration,
intrathecal administration can be
considered
Soft tissues Surgical sampling; Surgical Despite rigid structure of Consider clindamycin,
and bones blood cultures debridement bones and low b-lactams, daptomycin and linezolid
vascularisation of soft have good penetration
tissue, most antibiotics
have good penetration
Bloodstream Blood cultures Removal of venous Easy to reach. Effect site As it is easy to achieve good
catheters concentration is equal to concentrations, choice has to be based on
plasma concentration minimum inhibitory concentration of the
microorganisms, alteration of patient’s
body fluids compartments (i.e. septic
shock)

although their penetration in the CSF in the absence of penetration into abdominal and pulmonary sites, or tigecy-
meningeal inflammation is generally poor.14 The knowledge cline, the blood concentrations of which are usually low).18
of these drug properties and pathophysiological changes is
needed when considering the prescription of antimicrobials
for effective drug treatment. These will govern whether the
Risk factors
drug can easily reach the infection site at the usual dose or if a The European Centre for Disease Prevention and Control
higher dose is needed, or if indeed the drug has very poor (ECDC) and the Centers for Disease Control and Prevention
penetration in the effect site, independent of dosage requiring (CDC) define multidrug resistance as ‘acquired non-
alternate options. susceptibility to at least one agent in three or more antimi-
Alternative routes should also be considered if tissue crobial categories’. In the same document, extensively drug-
penetration is poor. These include aerosolised antimicrobial resistant (XDR) and pan drug-resistant (PDR) bacteria have
agents that penetrate the epithelial lining fluid poorly (e.g. also been defined, respectively, as ‘non-susceptibility to at
adjuvant inhaled colistin); or the intrathecal route for CNS least one agent in all but two or fewer antimicrobial cate-
infections in patients with an existing external ventricular gories’ and ‘non-susceptibility to all agents in all antimicrobial
drain, which may be used to give the antimicrobial drugs.15,16 categories’.6 Antibiotic resistance in the ICU has a high prev-
Complicated intra-abdominal infection (cIAI) is a complex alence because of patient-related risk factors and the pres-
site of infection, often a polymicrobial infection caused by ence of high antibiotic selection pressure.19 Local
Gram-negatives, anaerobes, enterococci, or fungi.17 In some epidemiology has been widely recognised as one of the most
cases (e.g. first-line treatment failure), the peculiarities of important factors for the development of infections from MDR
second-line suboptimal agents should also be known (i.e. pathogens. Moreover, as recently confirmed by the EPIC-III
aminoglycosides or colistin, both characterised by low study, infections caused by MDR pathogens in the ICU have

482 BJA Education - Volume 23, Number 12, 2023

 Empirical antimicrobial therapy in ICU

been independently associated with a higher risk of mortality empirical anti-MRSA, anti-VRE, anti-pseudomonal antibiotics
compared with infections caused by other microorganisms but none of the organisms targeted by these antibiotics were
(i.e. Enterococcus odds ratio [OR]¼2.41, Klebsiella resistant to b- present) were associated with higher mortality.27
lactam antibiotics OR¼1.29, carbapenem-resistant Acineto- These factors were considered by the Surviving Sepsis
bacter species OR¼1.40).1 A prevalence of 25% or more MDR, or Campaign (SSC) 2021 which recommended: (i) using empirical
an ongoing outbreak should alert the physician for infections antimicrobials with MRSA coverage in adults with sepsis or
possibly caused by MDR pathogens, regardless of other septic shock at high risk of MRSA but against their use in case
variables.20 of low risk; (ii) using two antimicrobials with Gram-negative
The adoption of local empirical antibiotic therapy protocols coverage for empirical treatment in adults with sepsis or
may help limit the variability of empirical treatment of in- septic shock and high risk for MDR organisms.4 However, they
fections. Nevertheless, no protocolised approach can disregard advised against this combination therapy in patients at low
the individual physician’s knowledge of risk factors for MDR risk or when susceptibilities are known. Furthermore, a plan
pathogen infections and multidisciplinary teamwork with local for stopping/de-escalating empirical antimicrobial therapy
antimicrobial stewardship programmes.21 Indeed, patient- should be discussed early with the multidisciplinary team to
related factors (including clinical severity) must be assessed limit any potential harm.
case by case by the treating physicians (e.g. intensivists, in-
fectious diseases specialists and microbiologists).22e24
The most significant risk factors for developing MDR
Microbiology
pathogen infections are the presence of a prior colonisation The spectrum of microorganisms involved in community-
status or recent infection caused by MDR pathogens, broad- acquired infections depends on local epidemiology and type of
spectrum antibiotic use during the preceding 90 days, inva- infection (e.g. community-acquired pneumonia, meningitis). In
sive procedures (e.g. abdominal surgery), prolonged mechan- case of hospital- or ICU-acquired infections, in patients with risk
ical ventilation (48 h) or ICU stay, recent hospitalisation for factors for MDR, six bacterial species frequently showing
at least 5 days in the past 90 days and immunosuppres- multidrug resistance and going by the acronym ESKAPE
sion.22,23 There is evidence that MDR pathogens infections are (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae,
not directly associated with poor prognosis, but inappropriate Acinetobacter baumannii, Pseudomonas aeruginosa and Enterobacter
antibiotic therapy may increase the risk for mortality.25 To species), specifically coined. In 2017 the ESKAPE pathogens were
highlight the complexity of treating such cases, the definition assigned a ‘priority status’. Early discussion with the microbi-
of ‘difficult-to-treat’ resistance was proposed, as an in vitro ology team will help with planning diagnostic tests and identi-
intermediate susceptibility or resistance to all b-lactam cate- fying potential pathogens based on local information and
gories, including carbapenems and fluoroquinolones.18,26 patient-specific factors to aid with decisions on antimicrobial
Several scores have been validated to estimate the risk of drug treatment. Traditional microbiological methods such as
MDR infections, mostly aimed at narrowing the spectrum of cultures usually require 48e72 h for the identification of path-
antibiotic treatment.23 Altogether, it is reasonable that ogens, and an even longer time to identify antibiotic suscepti-
community-acquired infections, in the absence of risk factors bility. As a consequence, they do not provide timely information
for MDR, may be treated with first-line therapeutic options for critically ill septic patients. However, despite these limita-
without significant risks for clinical failure. Community-onset tions, they remain the most widely available tests used for
sepsis is not frequently caused by resistant pathogens, except identifying the causative organisms.
where risk factors for community-acquired pneumonia In patients with sepsis or septic shock, rapid microbiolog-
caused by MRSA or Pseudomonas aeruginosa, such as history of ical techniques with turnaround times of around 2e6 h, may
MRSA colonisation, i.v. treatment, bronchiectasis are present provide identification of causative pathogens rapidly. These
(Table 2).27 Thus, the need for MDR coverage should reflect the use the genetic material of pathogens to identify the organ-
clinical severity of the infection (e.g. septic shock), immune isms and their potential antimicrobial susceptibility. A simple
status of the patient and other individual risk factors (Table 2). and inexpensive rapid test is the traditional ‘Gram stain’ of
In a cohort study of 15,000 patients with sepsis and a fresh specimens from sterile areas such as CSF, or synovial
positive culture, both inadequate (i.e. at least one pathogen fluid. They can also be very useful in identifying pathogens
not susceptible to therapy given) and unnecessarily broad- from deep respiratory samples and help to provide more tar-
spectrum empirical antimicrobials (i.e. patients receiving geted antimicrobial therapy.28

Table 2 Main risk factors for MDR.

MDR Gram-negative Local epidemiology; invasive procedures (e.g. abdominal surgery)

Presence of a prior gastrointestinal colonisation status or recent infection caused by MDR pathogens
Broad-spectrum antibiotics use during the preceding 90 days
Prolonged mechanical ventilation (48 h) or ICU stay
Recent hospitalisation for at least 5 days in the past 90 days
Immunosuppression
MRSA4 Local epidemiology, history of MRSA infection or colonisation
Recent i.v. antibiotics
Recurrent skin infections or chronic wounds
Invasive devices
Haemodialysis
Recent hospital admission and severity of illness

BJA Education - Volume 23, Number 12, 2023 483

Empirical antimicrobial therapy in ICU

Table 3 Summary of resistance mechanisms caused by b-lactamases according to the Ambler classification or type of resistant
bacteria and insight for clinical reasoning. HAP, hospital-acquired pneumonia; VAP, ventilator-associated pneumonia. *Treatment
options are here summarised to guide clinical reasoning. Physicians should refer to protocols based on local epidemiology and to the
most updated guidelines to select the evidence-based recommended antimicrobial therapy for their patients.

Classes Resistance Examples of treatment options*

Class A, extended-spectrum b- Resistant to third generation or higher Carbapenems

lactamases (ESBLs) cephalosporins (e.g. ceftriaxone)

Class A, serine carbapenemases Resistant to b-lactams including Meropenemevaborbactam; ceftazidime

(KPC) carbapenems eavibactam; fosfomycin; colistin

Class B, metallo-b-lactamases Resistant to all b-lactams except for Ceftazidimeeavibactam þ aztreonam;

(NDM, VIM, IMP1) aztreonam. Note that they are usually also meropenemevaborbactam þ aztreonam;
producing ESBL (resistance to aztreonam) cefiderocol; colistin þ fosfomycin

Class C, AmpC b-lactamase Resistant to cephalosporins Carbapenems, ceftolozaneetazobactam

Class D, oxacillinases (OXA-23, Frequently combined with ESBL and AmpC. Ceftazidimeeavibactam; cefiderocol;
OXA-48, OXA-48-like etc.) Resistance against cephalosporins and fosfomycin; colistin
carbapenems and vaborbactam or
relebactam

Methicillin-resistant Resistant to b-lactams. Some species may be Vancomycin; linezolid; daptomycin;

Staphylococcus aureus also vancomycin-resistant (VRSA) ceftaroline; telavancin; ceftobiprole
Pseudomonas aeruginosa Rifampin, tetracycline, chloramphenicol, Piperacillinetazobactam; levofloxacin;
trimethoprimesulfamethoxazole; cefepime; ceftazidime.
frequently resistant to fluoroquinolones In difficult-to-treat cases: ceftolozane
and many b-lactams, including etazobactam, ceftazidimeeavibactam or
carbapenems imipenemecilastatinerelebactam (when
not susceptible to carbapenems and
traditional b-lactams); cefiderocol in case of
metallo-b-lactamases producing strain
Acinetobacter baumannii Glycopeptides, lincosamides, macrolides, Combination therapy with at least two
streptogramins; frequently resistant to all b- active agents for severe infections;
lactam including carbapenems polymyxin (e.g. colistin) þ ampicillin/
sulbactam; cefiderocol (in combination if
possible); inhaled colistin may be
considered for HAP/VAP caused by CRAB

Molecular diagnostic techniques are able to identify genetic genetic patterns. Many studies have shown good performances
materials from pathogens along with the most common genetic of such methods30,31; and shown low rates of mismatch and
determinants of resistance (e.g. KPC, OXA 48-like, VIM, IMP, detection failure, whilst having a short turnaround time in pa-
NDM, CTX-M, AmpC, mecA/C and MREJ). The techniques tients with suspected bloodstream infection admitted to the
include polymerase chain reaction (PCR), mass spectrometry ICU. A watchful approach has increasingly been adopted in
(e.g. matrix-assisted laser desorption ionisation time-of-flight centres with the availability of such rapid microbiology tech-
mass spectrometry [MALDI-TOF]), magnetic resonance (i.e. nologies, thus deferring antimicrobial prescriptions until posi-
T2). Such techniques may be adopted as additional diagnostic tive results allow clinical decision-making.32 Furthermore, rapid
tools, especially when supported by multidisciplinary consul- microbiological methods can reduce the duration of inappro-
tations, but also as screening techniques for detecting coloni- priate antibiotic treatment by up to 45% (37.9e52.1%) and pro-
sation status. For this purpose, rectal (i.e. carbapenem-resistant vide some cost-saving.33 More recent development has been T2
Enterobacterales) or nasal swab specimens (e.g. MRSA) may be magnetic resonance techniques (e.g. T2Bacteria and T2Resis-
processed using PCR methods, able to amplify genetic material tance assays), able to detect amplified DNA of the six ESKAPE
and give results in a few hours. They can be performed at ICU pathogens and different resistance determinants from whole
admission and periodically during the ICU stay for microbio- blood specimens. Thanks to its rapidity and ease of sampling, it
logical surveillance (e.g. at day 3, day 7 then weekly) thus has become an established method in the diagnostic work-up of
identifying patients needing preventive isolation. Information bacteraemia and septic shock in many centres.34 In summary,
on antimicrobial susceptibility may be obtained using rapid rapid microbiological techniques have profoundly changed the
phenotypical diagnostic methods (i.e. light-scattering or time- diagnostic approach to infections in critical care and provided a
lapse microscopy). Another method is MALDI-TOF, which al- paradigm shift from empirical antimicrobial therapy to quasi-
lows for fast and accurate identification of pathogens (i.e. 10e30 targeted therapy. However, the interpretation of molecular re-
min) from subcultures, with an antibiogram available at ~24 h.29 sults should be done in the context of the whole clinical picture,
PCR panel methods detect predetermined genetic material of as the identification of genetic material of pathogens from
pathogens grouped according to infection site (e.g. meningitis samples may represent colonisation rather than infection, and
panels, respiratory panels, bloodstream infection panels etc.). result in overtreatment. This is most commonly seen in respi-
They can also provide information on the presence of resistance ratory samples. Thus, in the case of notification of a resistance

484 BJA Education - Volume 23, Number 12, 2023

 Empirical antimicrobial therapy in ICU

molecular pattern, clinical experience and multidisciplinary pathogens, it should be known that these species may be
approach are pivotal for interpreting the results (i.e. colonisa- resistant to all b-lactams, including new b-lactam/b-lactame
tion or infection). inhibitor (BLBLI) combinations such as ceftazidimeeavibactam,
Overall, molecular techniques have allowed more rapid meropenemevaborbactam or imipenemerelebactam but they
microbiological identification. However, the expectations of may be susceptible to cefiderocol. Carbapenem-resistant Enter-
such techniques to limit antibiotic overuse have not materi- obacterales may be susceptible to ceftazidimeeavibactam and
alised, partly because they cannot distinguish between infec- meropenemevaborbactam whereas oxacillinase-producing
tion and colonisation. The results of such tests must therefore may be susceptible to ceftazidimeeavibactam.
be considered as part of a broader decision-making process. Moreover, the characteristics of specific pathogens should
also be considered. Acinetobacter baumannii is a typical
example of a pathogen intrinsically resistant to several groups
Mechanisms of resistance of antimicrobials, including glycopeptides, lincosamides,
Antimicrobial resistance develops from core genetic mutation macrolides and streptogramins. Thus, CRAB are often sus-
or acquisition of mobile genetic elements (e.g. plasmids). ceptible only to colistin or cefiderocol.36
There are several mechanisms which may be combined: Pseudomonas aeruginosa is often intrinsically resistant to
rifampin, tetracycline, chloramphenicol, trimethoprim-
(i) The presence of b-lactamases
sulfamethoxazole and many b-lactams. Thus, difficult-to-
(ii) Modification of permeability of the bacterial wall
treat Pseudomonas aeruginosa may be susceptible to
(iii) The presence of drug efflux pumps in the bacterial wall
ceftolozaneetazobactam or, when producing metallo-b-
(iv) The presence of modified protein binding targets
lactamases, to colistin or cefiderocol as last options.
These mechanisms can be constitutive or inducible, often Of note, newly synthesised antibiotics, such as the most
under the selective pressure of antibiotics (e.g. AmpC b-lac- recent BLBLI combinations (ceftazidimeeavibactam,
tamases gene derepression). Among ESKAPE pathogens, ceftolozaneetazobactam, meropenemevaborbactam, imipe-
MRSA, ESBL-producing or third-generation cephalosporin- nemerelebactam) or cefiderocol, should be used with caution
resistant Escherichia coli, VRE, carbapenem-resistant Acineto- and under strict guidance, balancing the need for early and
bacter baumannii (CRAB) and Pseudomonas aeruginosa are effective therapies with the need to protect their efficacy from
frequently seen. Different pathogens may rely more on one of development of resistance that is of growing concern.37
these mechanisms whereas others may present several
combined mechanisms. Notably, enzymatic degradation
through carbapenemase production is often the main mech-
Dosage adjustments
anism of resistance to b-lactams in Enterobacterales, whereas The major determinants of antimicrobial efficacy are the
Acinetobacter baumannii and Pseudomonas aeruginosa may use a concentration and time of exposure to the antibiotic. These
variety of mechanisms, including efflux pumps, porin can be graphically demonstrated with plots of concentration
expression, antibiotic target mutations and drug-inactivating against time, with the area under the curve showing how long
enzymes. Resistance of MRSA results from the expression of the antibiotic concentration remains above the target level
the gene mecA, producing a penicillin-binding protein (PBP2a) after a dose.
with low affinity for b-lactams. Staphylococcus aureus may also Whilst b-lactam efficacy mostly depends on the duration of
acquire the gene VanA from VRE, especially in the cases of co- exposure to concentrations above the minimum inhibitory con-
infection, thus developing a rare resistance to vancomycin. centration (MIC) (i.e. it is time dependent), other drugs such as
The knowledge of potential mechanisms of resistance may aminoglycosides, have concentration-dependent killing where
help avoid therapeutic errors. Molecular microbiology allows their efficacy depends on how high their serum concentration
rapid identification of pathogens and tailoring of antibiotic is above the MIC.
treatment with a potential for lower risk of treatment failure, The pathophysiological changes in critically ill patients (e.g.
especially when there is detection of genes related to anti- vasoplegia, capillary leak, AKI) and treatments such as renal
microbial resistance. It is, for example, the case of the detec- replacement therapy can affect the PK/PD of antimicrobial
tion of genes related to the production of carbapenemases, the drugs. These need to be considered when providing the optimal
availability of a molecular antibiogram, or a previously known dosing of antimicrobial therapy. When using hydrophilic drugs
colonisation status. such as b-lactams, a loading dose (1.5e2 times the standard
When prescribing antimicrobials on the basis of molecular dose) may need to be considered in septic patients to compen-
microbiology results, physicians should consider susceptibil- sate for the higher volume of distribution.17,21,38 Indeed, criti-
ity patterns (Table 3), considering that the results received will cally ill patients may be treated with concentrations remaining
not be informative on all the available drugs (e.g. identifica- below the targets for effectiveness, thus contributing to negative
tion of carbapenem resistance would not inform about sus- outcomes. A prospective, multicentre point-prevalence study
ceptibility to fosfomycin or colistin). In this light, once a on eight b-lactams has recently shown an association between
phenotypic antibiogram becomes available, a de-escalation maintaining free drug concentration above MIC for at least 50%
from empirical therapy to appropriate targeted therapy or for the entire dosing interval (OR 1.02 and 1.56, respectively;
should be timely performed. Overall, guidelines should be P<0.03) and favourable clinical outcomes.39
consulted for full descriptions of the available evidence and Critically ill patients frequently have altered renal function,
for evidence-based recommendations on most appropriate ranging from augmented renal clearance to functional
therapies and doses.35,36 impairment with the need for renal replacement therapy.40
For example, guidelines strongly recommend the use of a Dose reduction on the basis of creatinine clearance should
carbapenem in case of severe infections caused by third- not be applied routinely in the setting of an ICU, as full anti-
generation cephalosporin-resistant Enterobacterales.36 In case microbial dose should be provided to patients with septic shock
of detection of metallo-b-lactamase (Ambler Class B) producing for probable clinical benefit at least for the first 24e48 h in case

BJA Education - Volume 23, Number 12, 2023 485

Empirical antimicrobial therapy in ICU

of transient AKI.41 In addition, increased dosage may be needed Declaration of interests

in those undergoing renal replacement therapy. The loading
The authors declare that they have no conflicts of interest.
dose should also be delivered independently of renal function.
There are drugs that clinicians should consider with more
caution in renal dysfunction and AKI (e.g. aminoglycosides), MCQs
selecting less nephrotoxic agents if similarly active.
Some specific settings or populations deserve tailored The associated MCQs (to support CME/CPD activity) will be
attention. Obese critically ill patients need dose adjustment accessible at www.bjaed.org/cme/home by subscribers to BJA
based on calculated lean body mass.38 Complex interactions Education.
between antimicrobial dosage and PK/PD changes are also
present in patients undergoing extracorporeal membrane
References
oxygenation (ECMO) support, with the potential for subopti-
mal dosing. Close discussion with the microbiology and 1. Vincent JL, Sakr Y, Singer M et al. Prevalence and out-
pharmacy team is recommended in these cases. comes of infection among patients in intensive care units
There has been increasing interest in the mode of admin- in 2017. JAMA 2020; 323: 1478e87
istration of antimicrobials, with controversial evidence on 2. Zhang D, Micek ST, Kollef MH. Time to appropriate anti-
extended (i.e. 3-h) or continuous infusions. The rationale is biotic therapy is an independent determinant of post-
based on the potential to maintain more constant plasma infection ICU and hospital lengths of stay in patients with
concentrations compared with rapid intermittent infusions, sepsis. Crit Care Med 2015; 43: 2133e40
avoiding delays in the achievement of effective concentra- 3. Johnston ANB, Park J, Doi SA et al. Effect of immediate
tions.12 In critically ill patients with severe infections, a administration of antibiotics in patients with sepsis in
strategy including loading dose plus extended or continuous tertiary care: a systematic review and meta-analysis. Clin
infusion of b-lactams may be associated with more favourable Ther 2017; 39: 190e202.e6
PK/PD.12 Such an approach has been supported by meta- 4. Evans L, Rhodes A, Alhazzani W et al. Surviving sepsis
analyses, guidelines, guidance documents and clinical prac- campaign: international guidelines for management of
tice.36,42 A recent randomised controlled trial has compared sepsis and septic shock 2021. Intensive Care Med 2021; 47:
continuous with intermittent dosing of meropenem. It found 1181e247
no significative effect on the composite outcome of mortality 5. Szychowiak P, Villageois-Tran K, Patrier J et al. The role of
and the emergence of pandrug-resistant or extensively drug- the microbiota in the management of intensive care pa-
resistant bacteria at day 28 in critically ill patients with tients. Ann Intensive Care 2022; 12: 3
sepsis.43 However, the trial may have been underpowered to 6. Magiorakos AP, Srinivasan A, Carey RB et al. Multidrug-
detect small but clinically important differences in outcomes resistant, extensively drug-resistant and pandrug-
and there was no evidence of harm in the prolonged infusion resistant bacteria: an international expert proposal for
group, as consistently reported in previous studies.42 Indeed, interim standard definitions for acquired resistance. Clin
another large multicentre RCT is ongoing (e.g. NCT03213990) Microbiol Infect 2012; 18: 268e81
and definitive evidence is still lacking. On such a basis, 7. Marquet K, Liesenborgs A, Bergs J et al. Incidence and
intermittent dosing remains an acceptable strategy. outcome of inappropriate in-hospital empirical antibi-
Tailored optimisation of dosing using therapeutic drug otics for severe infection: a systematic review and meta-
monitoring (TDM) is available for a variety of antimicrobials, analysis. Crit Care 2015; 19: 63
antifungals and antivirals, and has been proposed to aid the 8. Savage RD, Fowler RA, Rishu AH et al. The effect of
correct dosing in critically ill adults.44 However, an RCT inadequate initial empirical antimicrobial treatment on
comparing TDM-guided dosing of piperacillinetazobactam mortality in critically ill patients with bloodstream in-
with fixed dosing found no significant differences in 28-day fections: a multi-centre retrospective cohort study. PLoS
mortality or microbiological outcome, despite more consis- One 2016; 11, e0154944
tent attainment of target concentration in patients with TDM 9. Tabah A, Buetti N, Staiquly Q et al. Epidemiology and
(OR 4.5, 95% confidence interval 2.9e6.9, P<0.001).45 Main- outcomes of hospital-acquired bloodstream infections in
taining stable target concentrations may also be useful to intensive care unit patients: the EUROBACT-2 interna-
avoid toxic concentrations and adverse drug effects. tional cohort study. Intensive Care Med 2023; 49: 178e90
The clinical availability of TDM, with the aid of experts to 10. Pea F, Viale P. The antimicrobial therapy puzzle: could
interpret the results, may improve its cost-effectiveness; pharmacokinetic-pharmacodynamic relationships be
nevertheless, its worldwide adoption is still suboptimal. helpful in addressing the issue of appropriate pneumonia
treatment in critically ill patients? Clin Infect Dis 2006; 42:
1764e71
Conclusions 11. De Waele JJ, Girardis M, Martin-Loeches I. Source control
Knowledge of the main factors that determine appropriate in the management of sepsis and septic shock. Intensive
empirical antimicrobial therapy is vital for the early manage- Care Med 2022; 48: 1799e802
ment of patients admitted to critical care. This should consider 12. Jager NGL, van Hest RM, Lipman J et al. Antibiotic exposure at
the most probable infection site and hence, pathogens the site of infection: principles and assessment of tissue
involved, risk factors for MDR, local epidemiology, microbio- penetration. Expert Rev Clin Pharmacol 2019; 12: 623e34
logical results, clinical severity and dosage optimisation stra- 13. Finazzi S, Luci G, Olivieri C et al. Tissue penetration of
tegies. Furthermore, issues such as antibiotic overuse, the need antimicrobials in intensive care unit patients: a system-
for an antimicrobial stewardship programme and the preven- atic reviewdpart I. Antibiotics (Basel) 2022; 11: 1164
tion of resistance emergence cannot be ignored and may be 14. Nau R, So€ rgel F, Eiffert H. Penetration of drugs through the
better considered in a multidisciplinary context. blood-cerebrospinal fluid/blood-brain barrier for treatment

486 BJA Education - Volume 23, Number 12, 2023

 Empirical antimicrobial therapy in ICU

of central nervous system infections. Clin Microbiol Rev 2010; analysis of diagnostic test accuracy studies. BMC Infect Dis
23: 858e83 2022; 22: 794
15. Rouby JJ, Zhu Y, Torres A et al. Aerosolized polymyxins for 31. Cortegiani A, Russotto V, Graziano G et al. Use of cepheid
ventilator-associated pneumonia caused by extensive drug Xpert Carba-R® for rapid detection of carbapenemase-
resistant Gram-negative bacteria: class, dose and manner producing bacteria in abdominal septic patients
should remain the trifecta. Ann Intensive Care 2022; 12: 97 admitted to intensive care unit. PLoS One 2016; 11,
16. Ippolito M, Giarratano A, Cortegiani A. Healthcare-asso- e0160643
ciated central nervous system infections. Curr Opin 32. Muzzi M, Viaggi B, Fabbri S et al. The impact of fast
Anaesthesiol 2022; 35: 549e54 microbiology in intensive care units in the era of anti-
17. Niederman MS, Baron RM, Bouadma L et al. Initial anti- biotic resistance: an observational retrospective study.
microbial management of sepsis. Crit Care 2021; 25: 307 Curr Microbiol 2022; 79: 79
18. Kadri SS, Adjemian J, Lai YL et al. Difficult-to-treat resis- 33. Darie AM, Khanna N, Jahn K et al. Fast multiplex bacterial
tance in gram-negative bacteremia at 173 US hospitals: PCR of bronchoalveolar lavage for antibiotic stewardship
retrospective cohort analysis of prevalence, predictors, in hospitalised patients with pneumonia at risk of Gram-
and outcome of resistance to all first-line agents. Clin negative bacterial infection (Flagship II): a multicentre,
Infect Dis 2018; 67: 1803e14 randomised controlled trial. Lancet Respir Med 2022; 10:
19. Ruiz J, Gordon M, Villarreal E et al. Influence of antibiotic 877e87
pressure on multi-drug resistant Klebsiella pneumoniae 34. Giacobbe DR, Crea F, Morici P et al. T2Bacteria and
colonisation in critically ill patients. Antimicrob Resist T2Resistance assays in critically ill patients with sepsis or
Infect Control 2019; 8: 38 septic shock: a descriptive experience. Antibiotics (Basel)
20. Torres A, Niederman MS, Chastre J et al. International 2022; 11: 1823
ERS/ESICM/ESCMID/ALAT guidelines for the management 35. Paul M, Carrara E, Retamar P et al. European society of
of hospital-acquired pneumonia and ventilator- clinical microbiology and infectious diseases (ESCMID)
associated pneumonia. Eur Respir J 2017; 50, 1700582 guidelines for the treatment of infections caused by
21. Cortegiani A, Antonelli M, Falcone M et al. Rationale and multidrug-resistant gram-negative bacilli (endorsed by
clinical application of antimicrobial stewardship princi- European society of intensive care medicine). Clin Micro-
ples in the intensive care unit: a multidisciplinary state- biol Infect 2022; 28: 521e47
ment. J Anesth Analg Crit Care 2013; 3: 11 36. Tamma PD, Aitken SL, Bonomo RA et al. Infectious dis-
22. Dominedo  C, Ceccato A, Niederman M et al. Predictive eases society of America 2023 guidance on the treatment
performance of risk factors for multidrug-resistant path- of antimicrobial resistant gram-negative infections. Clin
ogens in nosocomial pneumonia. Ann Am Thorac Soc 2021; Infect Dis 2023. https://doi.org/10.1093/cid/ciad428
18: 807e14 37. Wang Y, Wang J, Wang R, Cai Y. Resistance to
23. Giannella M, Trecarichi EM, De Rosa FG et al. Risk factors for ceftazidimeeavibactam and underlying mechanisms.
carbapenem-resistant Klebsiella pneumoniae bloodstream J Glob Antimicrob Resist 2020; 22: 18e27
infection among rectal carriers: a prospective observational 38. Timsit JF, Bassetti M, Cremer O et al. Rationalizing anti-
multicentre study. Clin Microbiol Infect 2014; 20: 1357e62 microbial therapy in the ICU: a narrative review. Intensive
24. Aliberti S, Di Pasquale M, Zanaboni AM et al. Stratifying Care Med 2019; 45: 172e89
risk factors for multidrug-resistant pathogens in hospi- 39. Roberts JA, Paul SK, Akova M et al. DALI: defining anti-
talized patients coming from the community with pneu- biotic levels in intensive care unit patients: are current b-
monia. Clin Infect Dis 2012; 54: 470e8 lactam antibiotic doses sufficient for critically ill patients?
25. Montero MM, Lo  pez Montesinos I, Knobel H et al. Risk Clin Infect Dis 2014; 58: 1072e83
factors for mortality among patients with Pseudomonas 40. Bilbao-Meseguer I, Rodrı́guez-Gasco  n A, Barrasa H et al.
aeruginosa bloodstream infections: what is the influence Augmented renal clearance in critically ill patients: a
of XDR phenotype on outcomes? J Clin Med 2020; 9: 514 systematic review. Clin Pharmacokinet 2018; 57: 1107e21
26. Echols RM, Tillotson GS. Difficult to treat: do we need a 41. Gatti M, Pea F. Pharmacokinetic/pharmacodynamic target
new definition. Clin Infect Dis 2019; 69: 1641e2 attainment in critically ill renal patients on antimicrobial
27. Rhee C, Kadri SS, Dekker JP et al. Prevalence of antibiotic- usage: focus on novel beta-lactams and beta lactams/
resistant pathogens in culture-proven sepsis and out- beta-lactamase inhibitors. Expert Rev Clin Pharmacol 2021;
comes associated with inadequate and broad-spectrum 14: 583e99
empirical antibiotic use. JAMA Netw Open 2020; 3, e202899 42. Thabet P, Joshi A, MacDonald E et al. Clinical and phar-
28. Yoshimura J, Yamakawa K, Ohta Y et al. Effect of Gram macokinetic/dynamic outcomes of prolonged infusions of
stain-guided initial antibiotic therapy on clinical response beta-lactam antimicrobials: an overview of systematic
in patients with ventilator-associated pneumonia: the reviews. PLoS One 2021; 16, e0244966
GRACE-VAP randomized clinical trial. JAMA Netw Open 43. Monti G, Bradic N, Marzaroli M et al. Continuous vs
2022; 5, e226136 intermittent meropenem administration in critically ill
29. Verroken A, Defourny L, le Polain de Waroux O et al. patients with sepsis. JAMA 2023; 330: 141e51
Clinical impact of MALDI-TOF MS identification and rapid 44. Abdul-Aziz MH, Alffenaar JC, Bassetti M et al. Antimicrobial
susceptibility testing on adequate antimicrobial treat- therapeutic drug monitoring in critically ill adult patients: a
ment in sepsis with positive blood cultures. PLoS One 2016; Position Paper. Intensive Care Med 2020; 46: 1127e53
11, e0156299 45. Hagel S, Bach F, Brenner T et al. Effect of therapeutic drug
30. Peri AM, Ling W, Furuya-Kanamori L et al. Performance of monitoring-based dose optimisation of piperacillin/tazo-
BioFire Blood Culture Identification 2 Panel (BCID2) for the bactam on sepsis-related organ dysfunction in patients
detection of bloodstream pathogens and their associated with sepsis: a randomized controlled trial. Intensive Care
resistance markers: a systematic review and meta- Med 2022; 48: 311e21

BJA Education - Volume 23, Number 12, 2023 487