Carbapenem use: five agents and a bedside decision map
Carbapenem use should be reserved for severe infections with a resistant gram-negative risk, then matched to the agent whose spectrum fits the site. Four indications justify starting: suspected ESBL or AmpC Enterobacterales, severe infection with MDR risk, mixed aerobic-anaerobic sepsis, and empirical therapy in immunocompromised patients. Meropenem and imipenem cover Pseudomonas aeruginosa; ertapenem does not. Reassess at 48 to 72 hours and stop on clinical stability.
Why carbapenem use is a high-stakes decision
At 2 a.m. in a respiratory ward, a 78-year-old man with an acute exacerbation of COPD and type 2 respiratory failure is on day 4 of intubation. His temperature is 39.1 degrees Celsius, white cell count 18.6 x 10^9/L, procalcitonin 2.4 ng/mL, and sputum culture returns Klebsiella pneumoniae resistant to imipenem. The resident raises meropenem from 1 g every 8 hours to 2 g every 8 hours, lengthens the infusion to 3 hours, and calls infectious diseases.
That same morning in clinic, a 62-year-old woman with community-acquired pneumonia and no comorbidity has been receiving ertapenem 1 g every 24 hours for 3 days at another hospital. Her CT shows progression. Two years earlier she was repeatedly admitted for bronchiectasis, and Pseudomonas aeruginosa had grown in her sputum.
What separates those two rooms is what this article is about. Carbapenems are the strongest card in the antibacterial deck, and the easiest to play wrong. Some clinicians withhold them when a bloodstream infection caused by an extended-spectrum beta-lactamase (ESBL) producer is being held on a cephalosporin; others reach for them at the first fever, treating the last line as an antipyretic. Everything that follows answers five questions: whether to start, which agent, how to give it, how long, and when to stop.
Ward | Patient and findings | What actually went wrong |
Respiratory ward, 2 a.m. | 78-year-old man, COPD exacerbation, type 2 respiratory failure, intubation day 4, 39.1 degrees Celsius, WBC 18.6 x 10^9/L, PCT 2.4 ng/mL, sputum Klebsiella pneumoniae resistant to imipenem | A resistant isolate in a ventilated patient with a device and prior exposure; the response is a higher dose, a 3-hour infusion, and a specialist consult |
Outpatient clinic, next morning | 62-year-old woman, community-acquired pneumonia, no comorbidity, ertapenem 1 g every 24 hours for 3 days, CT progression; bronchiectasis and prior Pseudomonas aeruginosa in sputum 2 years earlier | The agent has no Pseudomonas aeruginosa activity, so the regimen was functionally no antibacterial therapy at all |
Table 1. Two bedside scenes that frame the five prescribing questions.
Why carbapenems resist beta-lactamases and where they still fail
Start with the penicillin nucleus, a penam. A carbapenem replaces the sulfur at position 4 with carbon, builds a double bond between positions 2 and 3, and carries a trans-hydroxyethyl side chain at position 6. Those three changes are the reason most beta-lactamases cannot hydrolyze the ring. Enzymatic hydrolysis, the most common bacterial resistance tactic, largely fails against this class, including the ESBL and AmpC enzymes that make third-generation cephalosporins unreliable. That explains both why the class works and why it has to be rationed.
Surviving beta-lactamases is not the same as being invincible. Killing by carbapenems is time-dependent: efficacy tracks the fraction of the dosing interval during which the free drug concentration stays above the minimum inhibitory concentration (MIC), written as percent fT>MIC, not the height of the peak. Every dosing move discussed later, a larger dose, a shorter interval, or a longer infusion, is a way of lengthening that time.
Four holes are built into the class and have to be checked before any carbapenem use is ordered. Stenotrophomonas maltophilia is intrinsically resistant through its L1 metallo-enzyme. Enterococcus faecium is intrinsically resistant, although Enterococcus faecalis is usually susceptible. Methicillin-resistant Staphylococcus aureus (MRSA) is not covered. Atypical organisms such as Mycoplasma and Chlamydia are not covered either, and neither are fungi. Any empirical plan that expects one drug to sweep the field leaks at exactly these points.
Organism or group | Why it escapes | What to do instead |
Stenotrophomonas maltophilia | Intrinsic resistance via the L1 metallo-beta-lactamase | Do not use; select by susceptibility, for example trimethoprim-sulfamethoxazole or levofloxacin |
Enterococcus faecium | Intrinsic resistance; Enterococcus faecalis is usually susceptible | Do not use; select ampicillin, vancomycin, or linezolid by susceptibility |
MRSA | No activity | Do not use; add vancomycin, linezolid, or daptomycin |
Atypical pathogens (Mycoplasma, Chlamydia) | Not covered | Do not use; combine a macrolide or a respiratory fluoroquinolone |
Fungi | No activity | Do not use; treat separately on mycological evidence |
Table 2. Organisms that carbapenems cannot catch. When any of these is plausible, the regimen needs a partner drug rather than a bigger dose of the same one.
Resistance belongs in the same calculation. The 2025 full-year report of a national antimicrobial resistance surveillance program covered 80 hospitals and 516,309 clinical isolates. Those are not niche findings.
Organism | Imipenem resistance | Meropenem resistance |
Klebsiella pneumoniae | 20.1 percent | 20.8 percent |
Pseudomonas aeruginosa | 22.6 percent | 18.8 percent |
Acinetobacter baumannii | 72.0 percent | 71.9 percent |
Table 3. Resistance rates in the 2025 full-year surveillance report, based on 516,309 isolates from 80 hospitals.
The message is plain: carbapenem resistance is no longer a rare event. Before writing any empirical prescription, ask whether this patient carries the soil for a resistant organism. Four markers matter: hospitalization or broad-spectrum antibacterial exposure within the previous 90 days, an invasive device such as an endotracheal tube, central venous catheter, or indwelling urinary catheter, prior isolation of a resistant organism, and a ward where resistant organisms are circulating.
The five agents and three distinct roles
Five carbapenems are available in routine clinical practice: imipenem-cilastatin, meropenem, ertapenem, biapenem, and panipenem-betamipron. They are not five brands of the same drug. They occupy three distinct positions: broad-spectrum agents for severe infection (imipenem, meropenem, biapenem, panipenem), a narrow-spectrum precision agent (ertapenem), and a dedicated central nervous system option (meropenem carries a meningitis indication).
Agent | Usual adult dose | Spectrum and distinguishing features | One-line position |
Imipenem-cilastatin | 0.5 g q6h; 1 g q6h for severe infection; daily maximum 4 g | Covers Pseudomonas aeruginosa and Acinetobacter baumannii; strongest gram-positive activity in the class; needs cilastatin to inhibit renal dehydropeptidase; highest CNS toxicity; high bone tissue concentration | Abdominal, bone and joint, and mixed infections; not for meningitis |
Meropenem | 1 g q8h; 2 g q8h for meningitis; daily maximum 6 g | Covers Pseudomonas aeruginosa and Acinetobacter baumannii; stable against renal dehydropeptidase so no inhibitor is needed; can be infused over 3 hours; best CNS safety in the class | Default for severe infection and for CNS infection |
Ertapenem | 1 g q24h; no dose escalation | No Pseudomonas aeruginosa, Acinetobacter baumannii, or enterococcal coverage; half-life about 4 hours allows once-daily dosing; stable against ESBL and AmpC; must not be diluted in glucose | Mild to moderate ESBL infection, community intra-abdominal infection, colorectal surgical prophylaxis |
Biapenem | 0.3 g q12h; 0.3 g q8h for severe infection; daily maximum 1.2 g | Covers Pseudomonas aeruginosa; relatively fewer CNS adverse reactions; the 0.3 g unit is an order of magnitude below the other agents | Common domestic agent; unit conversion is the most frequent local prescribing error |
Panipenem-betamipron | 0.5 g q12h; 1 g q12h for severe infection; daily maximum 2 g | Spectrum close to imipenem; needs betamipron to reduce renal cortical accumulation; carries a meningitis indication | Limited availability, usually a backup |
Table 4. Doses and positions of the five agents. Doses assume normal adult renal function; local labeling and hospital protocols govern, and renal adjustment appears in a later section.
The three dividing lines that change the choice
Set the five agents side by side and only three differences actually change a clinical decision. Three facts are easier to carry than a whole spectrum matrix.
Dividing line | Clinical meaning |
Activity against non-fermenting gram-negative bacilli | Ertapenem does not cover Pseudomonas aeruginosa or Acinetobacter baumannii; using it against Pseudomonas aeruginosa is equivalent to giving no antibacterial at all |
Penetration into cerebrospinal fluid | Only meropenem and panipenem carry a bacterial meningitis indication; the labels of imipenem and ertapenem carry no CNS indication |
Seizure risk | Imipenem has the highest affinity for GABA-A receptors and the highest risk; in a patient with epilepsy or underlying CNS disease, avoid imipenem first |
Table 5. The three dividing lines. Half-life, protein binding, and intramuscular availability rarely change a decision and belong in second place.
Dividing line one is activity against non-fermenters. Ertapenem is the only narrow-spectrum carbapenem, and the price is the loss of Pseudomonas aeruginosa and Acinetobacter baumannii coverage. That narrowness is the point: it withholds selection pressure from the two hardest non-fermenting organisms while retaining activity against ESBL and AmpC Enterobacterales. It fits a patient with confirmed or strongly suspected ESBL Enterobacterales infection and no Pseudomonas risk: community-acquired pneumonia, community intra-abdominal infection, or outpatient parenteral antimicrobial therapy (OPAT).
Dividing line two is cerebrospinal fluid. Only meropenem and panipenem have a bacterial meningitis indication in the label, and only meropenem reaches reliable cerebrospinal fluid concentrations in the presence of meningeal inflammation, at 2 g every 8 hours. Neither imipenem nor ertapenem has a CNS indication, and imipenem should be avoided in patients with pre-existing CNS disease.
Dividing line three is seizure risk. The C2 side chain of imipenem is the most basic in the class and blocks GABA-A receptors most strongly, giving it the highest seizure risk. Meropenem sits lowest, with serious seizure-like events reported in fewer than 0.1 percent of non-meningitis patients. One detail is easily missed: an active safety monitoring study put overall CNS adverse event rates for imipenem and meropenem close together, 3.98 percent versus 3.51 percent. Meropenem produces more mild events instead, dizziness, sleep disturbance, and delirium. What needs watching is not only a convulsion but a sudden change of consciousness in an older patient once the drug is started.
When carbapenem use is justified: four indications and three exclusions
The threshold fits in one sentence: this is a severe-infection drug for a resistant background, not the default drug for a severe patient. Any one of four indications makes a carbapenem a candidate.
1. Known or strongly suspected ESBL or AmpC Enterobacterales infection of moderate to severe grade. Bloodstream infection, hospital-acquired pneumonia and ventilator-associated pneumonia, complicated intra-abdominal infection, complicated urinary tract infection, and febrile neutropenia.
2. Severe infection with risk of multidrug-resistant (MDR) gram-negative organisms. Hospitalization or broad-spectrum antibacterial exposure within 90 days, invasive devices including endotracheal intubation, central venous catheters, and indwelling urinary catheters, prior isolation of a resistant organism, or a ward with resistant organisms in circulation.
3. Severe mixed aerobic and anaerobic infection. Severe intra-abdominal infection, deep diabetic foot infection, aspiration pneumonia with lung abscess. A carbapenem covers both sides as monotherapy here, so metronidazole does not need to be added.
4. Empirical therapy for moderate to severe infection in an immunocompromised patient. Febrile neutropenia, after solid organ or hematopoietic stem cell transplantation, or during long-term corticosteroid or immunosuppressant treatment.
Three situations should stop carbapenem use before it starts. First, fever without evidence of infection: postoperative absorption fever, tumor fever, drug fever, and thrombosis-related fever are not treated by a carbapenem and only generate resistance. Second, mild or localized infection in an outpatient who can take oral therapy; this class sits in a restricted prescribing tier and outpatient use breaches stewardship rules. Third, prophylaxis: apart from ertapenem for elective colorectal surgery, the class has no prophylactic indication.
Step | Decision and destination |
Step 1 | Confirm a severe infection, not a fever by itself |
Step 2 | Is there soil for an MDR gram-negative organism? If not, return to a conventional regimen and skip the carbapenem |
Step 3 | Is coverage of Pseudomonas aeruginosa and Acinetobacter baumannii needed? If yes, meropenem or imipenem. If no, prefer ertapenem to reduce selection pressure |
Step 4 | Give the full dose, extend the infusion, and send cultures before the first dose |
Step 5 | Reassess at 48 to 72 hours. If effective, de-escalate and shorten the course. If not, look for an undrained focus, identify the enzyme type, and change the regimen |
Table 6. A five-step decision pathway. If any one of the three gates fails, carbapenem use is not justified as a safety net.
Choosing the agent by infection site
Once step 3 is settled, the choice usually narrows to meropenem or imipenem when non-fermenter coverage is required, with CNS risk and local availability deciding between them. When it is not required, ertapenem is the more precise option. The table answers it directly by site.
Infection site | First choice | Key reminder and what to avoid |
Community-acquired pneumonia, hospitalized with comorbidity | Ertapenem 1 g q24h plus a macrolide or respiratory fluoroquinolone | Carbapenems are not first-line empirical CAP therapy; with Pseudomonas risk switch to meropenem and never use ertapenem |
Hospital-acquired pneumonia or ventilator-associated pneumonia | Meropenem 1 to 2 g q8h by extended infusion, or imipenem 0.5 to 1 g q6h | Pseudomonas aeruginosa and Acinetobacter baumannii coverage is usually needed, so ertapenem is inappropriate; add MRSA cover according to local epidemiology |
Community intra-abdominal infection, low risk | Ertapenem 1 g q24h as monotherapy | Source control comes first; monotherapy is not inferior to a combination in low-risk patients |
Hospital-acquired intra-abdominal infection, high risk | Meropenem or imipenem at full dose | Treat as severe when APACHE II is high, age is advanced, albumin is low, or debridement and drainage are not achievable |
Bacterial meningitis | Meropenem 2 g q8h, empirically combined with vancomycin | Neither imipenem nor ertapenem; meropenem must be infused over 3 hours |
Complicated urinary tract infection or pyelonephritis, ESBL | Ertapenem 1 g q24h | Prefer an oral agent when susceptibility and tolerance allow; imipenem offers no advantage here |
Bone and joint infection needing gram-negative coverage | Imipenem 0.5 g q6h, high bone concentration, or meropenem | Long course; attend to gram-positive coverage and debridement |
Febrile neutropenia | Meropenem, imipenem, or biapenem as monotherapy | Never ertapenem, which lacks Pseudomonas coverage; add MRSA coverage and fungal assessment if hemodynamics are unstable |
Table 7. Choosing the agent by infection site. Doses are starting points for adults with normal renal function; actual administration follows the next two sections.
Dosing around percent fT>MIC
Efficacy does not track how high the peak concentration climbs. It depends on the fraction of time the concentration stays above the MIC. For gram-negative organisms, bacteriostasis needs 20 to 30 percent fT>MIC and bactericidal effect needs 40 percent; in severe infection, with a high MIC or a resistant organism, the target is often raised to 100 percent fT>MIC or even 4 to 6 times MIC. Written into an order, that means three things: enough drug, enough frequency, and a longer infusion.
Administration | Concentration curve and T>MIC |
Standard 30-minute infusion | High peak and rapid decline, so T>MIC is short; the vertical axis is plasma concentration, the horizontal axis is time, and MIC is the threshold line |
Extended 3-hour infusion | Lower peak but a longer plateau, so T>MIC is clearly extended and target attainment is higher |
Table 8. Why the same dose works better when the infusion is longer. Target 40 percent fT>MIC; design for 100 percent fT>MIC in severe infection or with a high MIC.
Agent | Usual dose | Severe infection or high MIC | Infusion |
Imipenem-cilastatin | 0.5 g q6h | 1 g q6h, daily maximum 4 g | 30 to 60 minutes; not diluted in lactated Ringer solution |
Meropenem | 1 g q8h | 2 g q8h for meningitis and combination therapy against CRO | Extend to a 3-hour pump infusion to raise target attainment |
Ertapenem | 1 g q24h | No escalation, still 1 g q24h | 30 minutes; normal saline only |
Biapenem | 0.3 g q12h | 0.3 g q8h, daily maximum 1.2 g | 30 to 60 minutes |
Panipenem-betamipron | 0.5 g q12h | 1 g q12h, daily maximum 2 g | 30 to 60 minutes |
Table 9. Adult dosing and infusion method. Ertapenem is highly protein bound, so raising the dose does not raise free drug concentration; the habit of escalating for severity does not apply to it.
Three operational details decide whether extended infusion actually happens. Dissolve 1 to 2 g of meropenem in 100 mL of normal saline and run it by pump over 3 hours at a constant rate rather than estimating a gravity drip. After reconstitution, leave it at room temperature for no more than 4 hours, so extended infusion has to be prepared immediately before use. Write "pump over 3 hours" explicitly in the nursing record; otherwise the next shift may quietly return to a 30-minute drip and the gain disappears without anyone noticing.
Renal adjustment with the Cockcroft-Gault equation
Carbapenems are cleared mainly by the kidney; failing to adjust the dose is the single most common dosing error. Older patients are hit hardest: their serum creatinine looks normal because muscle mass is low, while true creatinine clearance may be only about 40 mL/min. Calculate with the Cockcroft-Gault equation; do not take the eGFR printed on the laboratory report and match it directly against a dose band.
Agent | Reduction by creatinine clearance (CrCl) | Dialysis and CRRT |
Imipenem-cilastatin | CrCl 30 to 59: 0.5 g q8h or 0.25 g q6h; CrCl 15 to 29: 0.5 g q12h or 0.25 g q8h; CrCl below 15: 0.25 g q12h | Hemodialysis 0.25 g q12h plus a dose after dialysis; CRRT 0.5 g q6h to q8h |
Meropenem | CrCl 26 to 50: 1 g q12h; CrCl 10 to 25: 0.5 g q12h; CrCl below 10: 0.5 g q24h | Hemodialysis 0.5 g q24h dosed after dialysis; CRRT 1 g q8h to q12h with extended infusion |
Ertapenem | CrCl 30 or above: 1 g q24h; CrCl below 30: 0.5 g q24h | Hemodialysis 0.5 g q24h within 6 hours after dialysis; CRRT 1 g q24h |
Biapenem | CrCl 30 to 59: 0.3 g q12h; CrCl below 30: 0.3 g q24h | Hemodialysis 0.3 g q24h after dialysis; CRRT 0.3 g q12h |
Table 10. Dose adjustment in renal impairment. Reduction bands for biapenem and panipenem differ slightly between hospital protocols, so local labeling and clinical pharmacist input govern.
Three traps show up again and again. A normal serum creatinine in an older patient is not normal renal function; the Cockcroft-Gault calculation frequently moves the dose down one band. Renal function changes during the illness, declining early in sepsis and recovering later, so recalculate every 48 to 72 hours rather than writing one order that lasts until discharge. A patient on continuous renal replacement therapy (CRRT) is not exempt from adjustment: actual clearance depends on the delivered dose and the filter, and underdosing is seen more often than overdosing.
Nine prescription errors seen most often
These are the problems that come up most often on rounds and in prescription review. The first four concern whether carbapenem use is justified at all; the last five concern how to use it. Nine concrete correction rules follow from them, and each is a place where carbapenem use can be tightened without losing efficacy.
# | What the chart shows | Why it fails and what to do |
1 | Starting meropenem for a temperature of 38.5 degrees Celsius with no focus, no inflammatory support, and negative blood cultures | Postoperative absorption fever on day 1 to 2, tumor fever, drug fever, and low-grade fever from pulmonary embolism are not infections; the drug is ineffective and damages flora |
2 | Escalating as soon as a culture is positive | Pseudomonas aeruginosa in sputum, coagulase-negative staphylococcus on a catheter tip, and enterococcus in wound drainage are usually colonization or contamination; judge specimen quality, white cells on smear, and clinical signs |
3 | Treating ertapenem as a universal carbapenem | No Pseudomonas aeruginosa, no Acinetobacter baumannii, no enterococci; in a patient with bronchiectasis, repeated admissions, long-term steroids, or structural lung disease it equals no antibacterial therapy |
4 | Expecting coverage of MRSA, Stenotrophomonas maltophilia, or Enterococcus faecium | All three are intrinsically resistant or insensitive; add a separate drug rather than another gram of meropenem |
5 | Meropenem every 12 hours | The half-life is about 1 hour, so q8h is required with normal renal function; q12h sharply shrinks percent fT>MIC and belongs only to moderate or severe renal impairment |
6 | No dose escalation and no extended infusion in severe infection | 1 g q8h over 30 minutes frequently misses target for organisms with MIC 2 to 4 mg/L; use 2 g q8h pumped over 3 hours for severe infection, high MIC, or combination therapy against CRO |
7 | Ertapenem diluted in 5 percent glucose | Stability falls markedly in glucose-containing solutions; use normal saline only and do not mix with other drugs in the same line |
8 | Coadministration with valproate | Carbapenems lower valproate concentrations substantially, and neither concentration monitoring nor dose escalation of valproate fixes it; switch the antibacterial instead |
9 | Extending the course indefinitely | "Two more days to consolidate" is the main form of overuse; most infections can stop within 7 days and intra-abdominal infection 4 days after source control |
Table 11. The nine recurring errors. Items 1 to 4 are indication errors; items 5 to 9 are administration errors.
Agent | Reduction in valproate concentration |
Meropenem | 80 to 90 percent |
Ertapenem | about 60 percent |
Imipenem | 30 to 40 percent |
Table 12. Magnitude of the valproate interaction by agent. A patient taking valproate for epilepsy should have the antibacterial changed, not the anticonvulsant dose increased.
Agent-specific traps worth memorizing
Imipenem-cilastatin has the highest seizure risk. A history of epilepsy, cerebrovascular disease, brain tumor, intracranial infection, or unadjusted renal impairment pushes that risk higher. With underlying CNS disease or a need for CNS coverage, move directly to meropenem. Do not combine with valproate and do not dilute in lactated Ringer solution, where stability is poor. High bone tissue concentration and the strongest gram-positive activity in the class are its strengths, so abdominal and bone and joint infection remain reasonable uses.
Meropenem is the steadiest agent, with two details that are often missed. The first is frequency: every 12 hours is for renal impairment, while a patient with normal kidneys needs every 8 hours. The second is the meningitis dose: an adult with purulent meningitis needs 2 g every 8 hours with an extended infusion, because 1 g every 8 hours may not deliver enough into cerebrospinal fluid. It also produces the largest drop in valproate of any agent in the class, close to discontinuing the anticonvulsant.
Ertapenem is narrow for a reason, but its boundaries are hard. Three of them: no Pseudomonas aeruginosa or Acinetobacter baumannii coverage, no CNS indication, and no glucose dilution. The advantages are just as concrete: a half-life of about 4 hours allows once-daily dosing and suits community settings and OPAT; high protein binding means severe infection is not a reason to escalate; stability against ESBL and AmpC makes it the most precise option for mild to moderate ESBL infection.
Biapenem uses a dose unit an order of magnitude smaller. The usual dose is 0.3 g every 12 hours with a daily maximum of 1.2 g. When converting from meropenem, writing "1 g biapenem" is a threefold overdose. It has relatively fewer CNS adverse reactions, but the local label carries no meningitis indication, so meropenem remains the choice for CNS infection.
Panipenem-betamipron is a backup with a required partner. It has to be combined 1:1 with betamipron, which reduces accumulation of panipenem in the renal cortex. Cross-allergy is possible in patients allergic to penicillin or another carbapenem, so take an allergy history first. Availability is limited and use is usually confined to specific formularies.
When the isolate is carbapenem resistant: identify the enzyme first
Once susceptibility reports resistance to imipenem or meropenem, the next question in carbapenem use is not how much more drug to add. It is which enzyme the organism carries. Carbapenemases fall into three classes and the drug logic is entirely different for each: class A serine enzymes, dominated by KPC in the local setting; class B metallo-beta-lactamases, namely NDM, IMP, and VIM; and class D enzymes, OXA-48 and OXA-23, the latter dominating carbapenem-resistant Acinetobacter baumannii (CRAB). First-half 2025 interim data put KPC detection at about 82.9 percent in carbapenem-resistant Klebsiella pneumoniae, metallo-beta-lactamase detection at 78.4 percent in carbapenem-resistant Escherichia coli, and OXA-23 at 99 percent in CRAB.
Enzyme class | Agents to consider | Key point |
Class A serine enzyme, mainly KPC | Ceftazidime-avibactam or meropenem-vaborbactam | Dominant mechanism in carbapenem-resistant Klebsiella pneumoniae (CRKP); vaborbactam does not inhibit metallo-enzymes; confirm the enzyme type before treating |
Class B metallo-beta-lactamase: NDM, IMP, VIM | Ceftazidime-avibactam is inactive; use aztreonam-avibactam or cefiderocol, often with polymyxin or tigecycline | Metallo-enzymes are common in Escherichia coli; carbapenem monotherapy is essentially ineffective |
Class D, mainly OXA-23 | Sulbactam-based combination with polymyxin or tigecycline | OXA-23 detection is extremely high in CRAB; monotherapy failure rates are high; combine with susceptibility and local epidemiology |
Table 13. Enzyme type decides the regimen. Availability of newer beta-lactam and inhibitor combinations varies widely between hospitals, so the formulary, susceptibility result, and infectious diseases consult govern the final choice.
Enzyme type | Common organisms in the reported setting | What works and what does not |
KPC, class A | Mainly Klebsiella pneumoniae | Ceftazidime-avibactam and meropenem-vaborbactam work; a carbapenem alone should not be used |
NDM, IMP, VIM, class B metallo-beta-lactamase | Escherichia coli, some Klebsiella, Pseudomonas aeruginosa | Ceftazidime-avibactam does not work; use aztreonam-avibactam, cefiderocol, or a polymyxin-containing regimen |
OXA-48, class D | Klebsiella species | Ceftazidime-avibactam works; meropenem-vaborbactam does not |
OXA-23, class D | Acinetobacter baumannii | Sulbactam-based combination; monotherapy failure rate is high |
Table 14. Enzyme type and drug correspondence. When a report says only "resistant" without an enzyme type, ask the microbiology laboratory to add carbapenemase testing.
One situation remains where a carbapenem may still contribute. When the meropenem MIC is 8 mg/L or lower, high-dose extended infusion can be one component of a combination regimen. Above 8 mg/L, escalation alone cannot work and a drug with a different mechanism is required.
De-escalation and duration: three questions before stopping
At 48 to 72 hours after empirical therapy starts, a formal review is not optional. It asks exactly three questions: is the patient clinically better, are microbiology results back, and is the focus controlled. De-escalation or stopping only enters the discussion once all three have answers.
Time point | Stage | Key action |
Day 0 | Empirical therapy | Draw cultures and give the first dose within 1 hour |
48 to 72 hours | First assessment | Temperature, white cell count, CRP, PCT, cultures |
Day 3 to 5 | De-escalation | Narrower spectrum, oral switch, stop combinations |
Day 5 to 7 | Stopping assessment | Most infections can stop within 7 days |
Beyond 7 days | Documented reason | Undrained abscess, bone and joint infection, CNS infection |
Table 15. The treatment timeline. Three signals permit stopping: afebrile for 48 hours, stable hemodynamics, and a focus that is controlled or drained. Do not extend a course merely because CRP is still slightly elevated; inflammatory markers lag behind clinical improvement.
Infection | Recommended duration | Note |
Hospital-acquired or ventilator-associated pneumonia | 7 days | Stop once clinically stable; longer courses do not improve outcomes |
Intra-abdominal infection | 4 days after source control | When drainage is incomplete, duration is set by source control rather than by the drug |
Complicated urinary tract infection or pyelonephritis | 7 to 14 days | Switch to oral therapy as early as susceptibility allows |
Gram-negative bloodstream infection | 7 to 14 days | Repeat blood cultures until negative and look for a focus such as a catheter or abscess |
Bacterial meningitis | 7 to 21 days | 7 days for Neisseria meningitidis, 14 for Streptococcus pneumoniae, 21 for Listeria |
Febrile neutropenia | 48 hours after defervescence and neutrophil recovery | With a positive blood culture or a defined site, use the site-specific duration |
Table 16. Duration by infection type. The clock starts at source control or clinical stability, not on the day carbapenem use began.
The counterintuitive part: selection errors outnumber dosing errors
When a patient is not improving, the instinct is to escalate. Across these decision points escalation is often the wrong move; the more useful question is whether the right agent was chosen at all.
Giving ertapenem to a ventilated patient with structural lung disease and prior Pseudomonas aeruginosa is a selection error, not an insufficient dose. Adding grams or extra days to that prescription cannot produce activity that the molecule does not have. The ventilated patient in the opening vignette needed a different drug, and the response came after the switch, not after a dose increase.
A second reversal sits in the CNS data. Meropenem is rightly described as the safer agent for the brain, yet an active monitoring study found overall CNS adverse event rates of 3.98 percent for imipenem and 3.51 percent for meropenem, essentially close. What differs is severity and phenotype. Watching only for convulsions misses the mild delirium and sleep disturbance in older patients, which make up much of the meropenem signal.
A third reversal in carbapenem use concerns ertapenem and severity. For every other agent in the class, severe infection argues for a higher dose. Ertapenem is highly protein bound, so a higher total dose does not raise free drug concentration, and the escalation habit becomes a plain error. A fourth sits in CRRT: clinicians often assume no adjustment is needed, while underdosing is the more common observed problem because clearance depends on delivered treatment dose and filter.
So a failing regimen should be audited in a fixed order. Confirm the organism is covered by the chosen molecule, then confirm the dose and interval against renal function, then confirm the infusion duration, and only then consider resistance mechanisms and a new drug. Moving straight to the last step wastes a day and adds selection pressure.
Special-use tier rules that govern the prescription
Carbapenems sit in the restricted special-use tier of antimicrobial management everywhere. Two documents still in force shape the prescription: a national regulation on clinical antimicrobial use effective since 2012 and a 2017 ministerial notice on antimicrobial resistance containment. Four constraints decide whether the order stands up to review.
Constraint | What it requires |
No outpatient use | Restricted-tier antibacterials must not be used in an outpatient clinic, with no exception; OPAT has to draw from another tier |
Consultation before the prescription | Indications must be strictly applied and a designated specialist in infectious diseases, respiratory medicine, critical care, clinical microbiology, or pharmacy must agree before an authorized physician writes the order |
Emergency override limited to 1 day | In a life-threatening emergency an override is allowed, but the quantity cannot exceed 1 day of supply, the indication must be documented in detail, and paperwork must be completed within 24 hours |
Dedicated file management and sampling rate | Carbapenems and tigecycline are tracked in dedicated files, and the pre-use microbiological sampling rate must be not less than 80 percent |
Table 17. The four stewardship constraints. In prescription review, the sampling requirement is checked most concretely: without a blood culture or a qualified sputum specimen drawn before the first dose, even a well-written progress note fails.
For the clinician on a night shift, the correct sequence is to draw cultures, give the first dose immediately, document the indication and the reason for the override in the progress note, and complete the consultation the next morning. Delaying the first dose for hours while waiting for a consult raises mortality in septic shock, and forgetting the paperwork afterward creates a separate problem. Both halves of that sequence have to happen.
Three bedside cases
Case 1: not every carbapenem needs to be meropenem. A 72-year-old man developed fever of 39.2 degrees Celsius with chills on day 3 after endoscopic retrograde cholangiopancreatography for a common bile duct stone. White cell count was 19.8 x 10^9/L, procalcitonin 8.6 ng/mL, and blood culture grew Escherichia coli that was ESBL positive and susceptible to both ertapenem and meropenem. Ultrasound showed mild intrahepatic duct dilation without abscess. He received ertapenem 1 g every 24 hours together with an assessment for relief of biliary obstruction. Temperature normalized at 48 hours, blood cultures were negative on day 6, and the drug was stopped on day 7. With no Pseudomonas risk and no CNS infection, ertapenem and meropenem are equivalent in efficacy, but ertapenem applies less selection pressure, needs one dose a day, and costs less. Biliary drainage decided the outcome, not an upgrade to meropenem.
Case 2: the wrong agent cannot be rescued by a bigger dose. A 78-year-old man with COPD was on day 4 of intubation with fever of 39.1 degrees Celsius, purulent sputum, and a new right lower lobe infiltrate. He had already received ertapenem 1 g every 24 hours for 3 days without defervescence. Sputum culture grew Pseudomonas aeruginosa susceptible to meropenem at an MIC of 1 mg/L. He was switched immediately to meropenem 1 g every 8 hours pumped over 3 hours, with airway drainage, positioning, and oral care. Temperature fell to 37.8 degrees Celsius on day 3, he was extubated on day 7, and the total course was 7 days. Ertapenem has no Pseudomonas activity, so this was a selection error rather than underdosing; raising the dose or extending the course would never have worked. Structural lung disease, repeated admissions, and long-term steroids mean the empirical regimen must cover Pseudomonas aeruginosa by default.
Case 3: two hard errors stacked in one patient. A 65-year-old woman with an 8-year history of epilepsy managed with valproate was admitted to intensive care with severe intra-abdominal infection and given imipenem-cilastatin 1 g every 6 hours at another hospital. On day 2 she developed myoclonus followed by a generalized tonic-clonic seizure, and her valproate concentration had fallen from 62 to 11 mcg/mL. Imipenem was stopped, meropenem 1 g every 8 hours with extended infusion was started, neurology switched seizure control to levetiracetam, and the abdominal abscess was drained percutaneously. She was afebrile on day 5 and antibiotics stopped on day 9. The two failures compounded: the agent with the highest seizure risk in the class was chosen, and it ran into a certain drug interaction. For a patient taking valproate, treat a carbapenem as contraindicated; changing the antibacterial beats increasing the valproate dose.
Five sentences for families
Talking about carbapenem use is hard because of comprehension, not cost. Five answers cover the questions families ask most often, without exaggerating or hiding anything.
- "Why such a strong antibiotic?" Some bacteria make an enzyme that cuts ordinary antibiotics apart. The structure of this class was changed so that enzyme cannot cut it, which is why it still works against resistant organisms. The culture results or the clinical picture suggest an ordinary drug may not be enough.
- "Will there be no drug left later?" That depends on whether carbapenem use is correct, not on whether it happens at all. Three things protect it: take a specimen before starting to identify the organism, stop after an adequate course, and switch to a narrower agent as soon as possible.
- "I heard it can cause seizures?" One agent, imipenem, does carry that risk, so a patient with a seizure history is given a safer member of the same class instead. That consideration is already built into the choice.
- "Can it be switched to tablets?" These agents are intravenous only. Once temperature is stable, eating is possible, and the infection is controlled, treatment continues with an oral drug and the infusions can stop.
- "How many days?" Most infections need about 7 days, but stopping depends on three markers: normal temperature, stable blood pressure, and a controlled focus. The team reviews every 2 to 3 days and will not continue indefinitely.
A one-page practical checklist
- Whether to start: moderate to severe ESBL or AmpC Enterobacterales infection, severe infection with resistant-organism soil, severe mixed aerobic and anaerobic infection, or empirical therapy in an immunocompromised patient. Fever alone is not an indication.
- Which agent: need Pseudomonas aeruginosa and Acinetobacter baumannii coverage, choose meropenem or imipenem; no such need, choose ertapenem; CNS infection, only meropenem at 2 g every 8 hours.
- How to give: meropenem 1 g every 8 hours, 2 g every 8 hours pumped over 3 hours in severe infection; imipenem 0.5 g every 6 hours; ertapenem 1 g every 24 hours with no escalation and normal saline only; biapenem 0.3 g every 12 hours.
- How to reduce: calculate CrCl with Cockcroft-Gault and repeat every 48 to 72 hours; a normal serum creatinine in an older patient usually means one band lower; CRRT is not an exemption.
- What it cannot cover: MRSA, Stenotrophomonas maltophilia, Enterococcus faecium, atypicals, and fungi. Ertapenem adds Pseudomonas aeruginosa and Acinetobacter baumannii to that list.
- Three red lines: no valproate coadministration, no glucose dilution for ertapenem, no outpatient prescribing of a restricted-tier agent.
- When to stop: 7 days for hospital-acquired or ventilator-associated pneumonia, 4 days after source control for intra-abdominal infection, 7 to 14 days for gram-negative bloodstream infection, 7 to 21 days for meningitis. Judge by clinical status and the focus, not by the decimal point on CRP.
Limitations
Doses here are usual adult starting points for carbapenem use under normal renal function, and the source material states that local labeling, hospital protocols, susceptibility results, patient weight, infection site, and local resistance epidemiology should shape every individual prescription. Renal reduction bands for biapenem and panipenem vary between institutions, so the values in the table are a working reference rather than a validated local protocol.
The resistance rates come from one national surveillance program covering 80 hospitals and 516,309 isolates in a single full year, with enzyme distributions drawn from first-half interim data. A single program and a single period cannot describe the flora of a particular ward, and the 72.0 percent imipenem resistance in Acinetobacter baumannii in that report should not be assumed to hold locally. Availability of newer beta-lactam and inhibitor combinations also differs sharply between hospitals, so enzyme-guided options are conditional on the formulary.
This is clinical education material written for prescribers. It is not advice for a specific patient and does not replace the product label, institutional antimicrobial policy, or an infectious diseases and clinical pharmacy consultation. The three cases carry no individual patient identifiers.
Bottom line
Carbapenem use is a five-question discipline: confirm the indication, pick the molecule whose spectrum fits the site and the organism, dose against percent fT>MIC and creatinine clearance, review at 48 to 72 hours, and stop on clinical stability and source control. The costliest mistake in carbapenem use is rarely a dose that is a little low. It is choosing ertapenem when Pseudomonas aeruginosa is plausible, escalating a drug that cannot work, or continuing an effective course for two extra days. Check the organism first, then the dose, then the infusion, and only then the enzyme mechanism.
FAQ
When should carbapenem use be started in a hospitalized patient?
Start only when one of four indications is present: known or strongly suspected ESBL or AmpC Enterobacterales infection of moderate to severe grade, severe infection with MDR gram-negative risk, severe mixed aerobic and anaerobic infection, or empirical therapy in an immunocompromised patient. Fever without evidence of infection, mild infection in an outpatient who can take oral therapy, and prophylaxis are not indications apart from ertapenem for elective colorectal surgery. Carbapenem use in a restricted tier also requires prior consultation and a microbiological sample before the first dose.
Which carbapenem covers Pseudomonas aeruginosa and which one does not?
Meropenem, imipenem-cilastatin, biapenem, and panipenem-betamipron cover Pseudomonas aeruginosa; ertapenem does not, and it also lacks Acinetobacter baumannii and enterococcal coverage. A patient with bronchiectasis, repeated admissions, long-term steroids, or structural lung disease should be assumed to carry Pseudomonas risk, so ertapenem is equivalent to giving no antibacterial at all. Switch to meropenem or imipenem instead of escalating the ertapenem dose, because no dose of ertapenem creates activity the molecule does not have.
Why is ertapenem not used for meningitis?
Ertapenem has no bacterial meningitis indication in its label, and imipenem has none either; only meropenem and panipenem carry that indication. Only meropenem reaches reliable cerebrospinal fluid concentrations during meningeal inflammation, and the required dose is 2 g every 8 hours infused over 3 hours, empirically combined with vancomycin. Imipenem should also be avoided in patients with underlying CNS disease because it has the highest affinity for GABA-A receptors and the highest seizure risk in the class.
How long should a meropenem infusion run in severe infection?
For severe infection, a high MIC, or combination therapy against carbapenem-resistant organisms, meropenem 2 g every 8 hours should be pumped over 3 hours rather than the standard 30 minutes. Dissolve 1 to 2 g in 100 mL of normal saline and run it at a constant rate by pump; an extended infusion lowers the peak but lengthens the plateau and raises percent fT>MIC target attainment. Prepare it immediately before use because room-temperature storage after reconstitution must not exceed 4 hours, and write the 3-hour duration explicitly in the nursing record so the next shift does not revert to a 30-minute drip.
How should carbapenem doses change in renal impairment?
Calculate creatinine clearance with the Cockcroft-Gault equation rather than copying eGFR from the laboratory report, and recalculate every 48 to 72 hours because renal function shifts during sepsis and recovery. For meropenem, use 1 g every 12 hours at CrCl 26 to 50 mL/min, 0.5 g every 12 hours at 10 to 25 mL/min, and 0.5 g every 24 hours below 10 mL/min; for ertapenem, reduce from 1 g to 0.5 g every 24 hours below CrCl 30. An older patient with a normal-looking serum creatinine may have a true CrCl near 40 mL/min, and a patient on CRRT still needs adjustment because underdosing is more common than overdosing.
Why can carbapenems not be combined with valproate?
Carbapenems lower valproate blood concentrations substantially: meropenem by 80 to 90 percent, ertapenem by about 60 percent, and imipenem by 30 to 40 percent. Neither monitoring the concentration nor raising the valproate dose overcomes the interaction, so the correct move is to change the antibacterial rather than escalate the anticonvulsant. In one bedside case a patient with an 8-year epilepsy history developed myoclonus and a generalized tonic-clonic seizure on day 2 of imipenem while valproate fell from 62 to 11 mcg/mL; the response was to stop imipenem, switch the anticonvulsant to levetiracetam, and drain the abdominal focus.
What organisms do carbapenems fail to cover?
Four gaps are intrinsic to carbapenem use: Stenotrophomonas maltophilia produces an L1 metallo-beta-lactamase, Enterococcus faecium is intrinsically resistant although Enterococcus faecalis is usually susceptible, MRSA has no susceptibility, and atypical organisms such as Mycoplasma and Chlamydia are not covered. Fungi are not covered either. When any of these is plausible in an empirical regimen, add a separate drug chosen by susceptibility, such as trimethoprim-sulfamethoxazole or levofloxacin for Stenotrophomonas, vancomycin or linezolid for MRSA, and a macrolide or respiratory fluoroquinolone for atypicals.
How many days should treatment last after source control?
For intra-abdominal infection, 4 days after adequate source control is enough, and when drainage is incomplete the duration is determined by source control rather than by the drug. Hospital-acquired or ventilator-associated pneumonia is usually 7 days, complicated urinary tract infection and gram-negative bloodstream infection are 7 to 14 days, and bacterial meningitis ranges from 7 to 21 days depending on the organism. Stop when the patient has been afebrile for 48 hours, hemodynamics are stable, and the focus is controlled or drained; a mildly elevated CRP alone is not a reason to extend the course.
Which carbapenemase type determines the choice of a newer beta-lactam combination?
Send the enzyme type first: class A serine enzymes dominated by KPC respond to ceftazidime-avibactam or meropenem-vaborbactam, class B metallo-beta-lactamases such as NDM, IMP, and VIM require aztreonam-avibactam or cefiderocol and ceftazidime-avibactam is inactive, and OXA-48 responds to ceftazidime-avibactam but not meropenem-vaborbactam. OXA-23, dominant in CRAB, calls for a sulbactam-based combination. First-half 2025 interim data put KPC at about 82.9 percent in CRKP, metallo-beta-lactamases at 78.4 percent in carbapenem-resistant Escherichia coli, and OXA-23 at 99 percent in CRAB.
Is meropenem every 12 hours ever correct?
With normal renal function it is not: the meropenem half-life is about 1 hour and the standard interval is every 8 hours, while every 12 hours sharply reduces percent fT>MIC and is the most common frequency error seen on rounds. Every 12 hours becomes correct only in moderate to severe renal impairment, for example 1 g every 12 hours at CrCl 26 to 50 mL/min or 0.5 g every 12 hours at CrCl 10 to 25 mL/min. In severe infection with a high MIC the direction is the opposite, moving to 2 g every 8 hours with a 3-hour infusion.
What should happen at the 48 to 72 hour review?
Ask three questions: is the patient clinically better, are microbiology results back, and is the focus controlled. Check temperature, white cell count, CRP, PCT, and cultures, then de-escalate on day 3 to 5 by narrowing the spectrum, switching to oral therapy, and stopping combination drugs. If the patient has not improved, look for an undrained focus, request carbapenemase testing, and change the regimen rather than simply adding more of the same drug. Most infections can stop by day 5 to 7, and any course beyond 7 days needs a written reason such as an undrained abscess or bone, joint, or CNS infection.
Can carbapenem use be started in an outpatient clinic?
No. This class is listed in the restricted special-use tier of antimicrobial management, and outpatient use is prohibited without exception, so outpatient parenteral antimicrobial therapy has to use an agent from another tier. Inpatient use also requires prior consultation with designated specialists, a prescription from a physician holding the corresponding authority, and dedicated file management. Emergency override for a life-threatening patient is limited to 1 day of supply with the indication documented and paperwork completed within 24 hours, and the required pre-use microbiological sampling rate is not less than 80 percent.