Principles of Effective Antibiotic Therapy — Part 2
Lecture Handout — Slides 63+
1 Pharmacokinetic/Pharmacodynamic (PK/PD) Indices
1.1 Three Patterns of Antibiotic Killing
One of the most important advances in modern antibiotic therapy is the recognition that antibiotics do not all kill bacteria in the same way. The relationship between drug concentrations and antimicrobial efficacy is described by pharmacokinetic/pharmacodynamic (PK/PD) indices, which link measurable drug exposure parameters to microbiological outcomes. Understanding these indices allows clinicians to rationally optimize dosing regimens rather than relying on one-size-fits-all approaches.
Three fundamental PK/PD patterns have been identified:
Concentration-dependent killing (Peak/MIC). For these antibiotics, the height of the peak concentration relative to the MIC is the primary driver of bacterial killing. The higher the peak above the MIC, the more rapid and complete the bactericidal effect. Drugs in this category include aminoglycosides, fluoroquinolones, and polymyxins. Clinical implication: when treating infections with higher MICs, it may be preferable to give a single higher dose rather than multiple smaller doses.
Time-dependent killing (T > MIC). For these antibiotics, efficacy depends on how long the drug concentration remains above the MIC during the dosing interval. Increasing the peak concentration does not substantially improve killing — what matters is duration of exposure. This pattern is characteristic of beta-lactams (penicillins, cephalosporins, carbapenems, monobactams) and macrolides. Animal and clinical studies have established that maintaining concentrations above the MIC for at least 40–50% of the dosing interval is generally associated with efficacy.
Exposure-dependent killing (AUC/MIC). These antibiotics exhibit a mixture of both time- and concentration-dependent activity. The total drug exposure over 24 hours — represented by the area under the concentration–time curve (AUC) divided by the MIC — best predicts outcomes. Vancomycin, tetracyclines, and linezolid fall into this category.
1.2 How PK/PD Indices Are Identified
1.3 Dose Fractionation Studies in Antimicrobial Pharmacodynamics
Dose fractionation experiments are a cornerstone methodology for identifying which PK/PD index best predicts antimicrobial efficacy. The fundamental design is elegant: a fixed total daily dose is administered to infected animals (typically neutropenic murine thigh or lung infection models) but divided into different dosing intervals — for example, the same total mg/kg/day given as a single dose (q24h), split into two doses (q12h), three (q8h), or six (q4h).
Because the total drug exposure (AUC) remains constant across all regimens while the shape of the concentration-time curve changes dramatically, you can tease apart which PK/PD driver matters most:
If the q24h regimen wins (produces the greatest bacterial kill or survival), the drug achieves the highest peak concentrations, and efficacy is driven by Cmax/MIC. This is the classic pattern for aminoglycosides and daptomycin — concentration-dependent killing where you want to hit hard with large, infrequent doses.
If all regimens perform equally, efficacy depends only on total exposure regardless of how it’s distributed over time, pointing to AUC/MIC as the predictive index. Fluoroquinolones, azithromycin, and most antifungals (including the azoles and echinocandins) follow this pattern. Notably, AUC/MIC-driven drugs can still show concentration-dependent killing in time-kill curves; the fractionation study reveals that what ultimately matters in vivo is cumulative exposure.
If the q4h regimen wins, splitting the dose into frequent small administrations keeps concentrations above the MIC for the longest fraction of the dosing interval, identifying %T>MIC as the key driver. This is the signature of beta-lactams and, to a degree, linezolid — time-dependent killing where prolonged exposure above threshold matters more than peak height.
As the figure below illustrates, the three PK/PD curve shapes are visually distinct: the q24h-optimized profile shows tall, sharp peaks well above MIC; the AUC-driven profile shows equivalent areas under all regimens; and the T>MIC-optimized profile shows low, sustained concentrations hovering above the MIC line.
These experiments, pioneered largely by Craig and Andes at Wisconsin, translated directly into clinical dosing strategies — continuous or extended infusions for beta-lactams, once-daily aminoglycoside dosing, and AUC-guided vancomycin monitoring. They remain the standard preclinical framework for establishing PK/PD targets that then inform clinical breakpoints, dose optimization, and Monte Carlo simulation work.
This is a classic dose fractionation result for a cephalosporin. The same dataset (CFU/thigh at 24 hours in a neutropenic mouse thigh infection model) is plotted against all three candidate PK/PD indices simultaneously:
In the Cmax/MIC panel (left), the data points are scattered with no coherent relationship — you can achieve the same peak-to-MIC ratio and get wildly different bacterial burdens, ranging from near-baseline (~7.3 log, the dashed stasis line) to maximal kill or no kill at all.
The AUC/MIC panel (centre) shows a similar scatter. Total exposure alone does not predict outcome.
The %T>MIC panel (right) reveals a tight sigmoidal relationship (R² = 0.94). As the percentage of the dosing interval with concentrations above MIC increases, bacterial burden drops progressively. Below about 25–30% T>MIC, there is essentially no efficacy (CFU remain at or above stasis). Maximal kill plateaus once T>MIC reaches roughly 60–70%.
This is the definitive demonstration that beta-lactams are time-dependent killers. The clinical translation is direct: you don’t need higher peaks, you need longer exposure above MIC — hence the rationale for extended or continuous infusions of beta-lactams, and why dosing interval matters more than dose size for this drug class.
1.4 From Animal Model to Clinical Validation — AUC/MIC Target for Fluoroquinolones
This figure below illustrates a critical step in PK/PD-guided drug development: translating an animal-derived target to human outcomes.
The left panel shows ciprofloxacin in a murine P. aeruginosa pneumonia model. Mortality rate is plotted against the 24-hour AUC/MIC. There is a steep sigmoid relationship: mortality is essentially 100% at low AUC/MIC values, then drops sharply to near zero once AUC/MIC exceeds approximately 125. This identifies the animal-derived PK/PD target.
The right panel validates this same threshold in humans — patients with ventilator-associated pneumonia treated with fluoroquinolones. When clinical and microbiological cure rates are stratified by achieved AUC/MIC, there is a dramatic jump in outcomes at the 125–250 bracket (88% microbiological cure, 81% clinical cure) compared to lower strata. Below an AUC/MIC of 125, cure rates are poor (≤40% microbiological, ≤30% clinical). Interestingly, outcomes plateau or even slightly decline at very high AUC/MIC values, which likely reflects confounding — patients with highly susceptible organisms (low MICs, thus high AUC/MIC ratios) may have had less virulent infections.
Taken together, these panels demonstrate the “bench-to-bedside” PK/PD pipeline: animal dose fractionation identifies AUC/MIC as the driver and defines a target of ~125, and clinical outcome data in critically ill patients confirms that this same threshold discriminates success from failure. This AUC/MIC ≥ 125 target for fluoroquinolones against Gram-negatives has become one of the most widely cited PK/PD benchmarks and underpins breakpoint setting, Monte Carlo dosing simulations, and clinical dosing recommendations.
These studies have consistently shown that specific PK/PD targets predict clinical success. For example, for beta-lactams, a T > MIC of 40–50% is typically associated with microbiological efficacy, while for aminoglycosides a Peak/MIC ratio of 8–10 is optimal (Craig, 1998; Drusano, 2004).
1.5 PK/PD Characteristics of Common Antibiotic Classes
The table below summarises the PK/PD classification of major antibiotic families and the corresponding optimisation strategies.
For concentration-dependent drugs (aminoglycosides): give infrequent high doses (e.g., once-daily dosing). For time-dependent drugs (beta-lactams): give frequent doses or use extended/continuous infusions. For AUC/MIC-dependent drugs (vancomycin): optimise total daily exposure through AUC-guided monitoring.
2 Application: Model-Based Antibiotic Dosing Optimisation
2.1 From Population Models to Individualised Therapy
The practical importance of PK/PD indices is that they enable model-based dosing — a mathematically guided approach to individualising antibiotic therapy for each patient. This is particularly critical in the intensive care unit, where the pharmacokinetics of antibiotics are profoundly altered and the coefficient of variation in drug exposure can reach 80–100%.
The approach works as follows:
Population pharmacokinetic models are developed by studying hundreds or thousands of patients who received the drug. These models identify the clinical covariates (e.g., renal function, weight, albumin level, dialysis status) that significantly affect drug clearance and volume of distribution.
An a priori prediction is generated for a new patient based on their clinical characteristics. This provides a probability estimate that a given dosing regimen will achieve the PK/PD target (e.g., sufficient T > MIC for a beta-lactam).
Therapeutic drug monitoring (TDM) — measuring an actual drug level in the patient’s blood — is used to validate or refine the model prediction. If the measured concentration differs from the prediction, the model is adjusted (Bayesian updating) to produce a dosing regimen specific to that individual patient.
As the patient’s clinical status changes (e.g., declining renal function, fluid shifts), the model can be continuously updated to maintain optimal exposure.
2.2 Worked Example: Meropenem Dosing in the ICU
Consider a 35-year-old patient (70 kg, 170 cm) admitted after a motor vehicle accident who is becoming hypotensive and requires empirical broad-spectrum antibiotic therapy. Meropenem 1 g infused over 30 minutes every 8 hours (the standard licensed dose) is initiated.
Using a population PK model for meropenem in critically ill patients, we can predict the concentration–time profile. For a susceptible pathogen with an MIC of 2 mg/L (e.g., E. coli), the model estimates a T > MIC of approximately 93% — comfortably above the 40–50% target needed for efficacy.
But what if the MIC is 16 mg/L? Now the time above MIC falls to approximately 18–20%, well below the target. Simply doubling the dose to 2 g achieves only marginally better T > MIC, while producing dangerously high peak concentrations (approaching the seizure threshold of approximately 100 mg/L). The additional drug is essentially “wasted” — eliminated renally without contributing to bacterial killing.
The solution exploits the PK/PD principle: since meropenem is time-dependent, we need to increase duration of exposure, not peak concentration. By infusing 1 g over 7.5 hours (instead of 30 minutes) every 8 hours, the T > MIC improves dramatically — approaching 100% even against the higher MIC target. Meropenem must be changed every 8 hours because of limited stability in solution, so a 7.5-hour infusion (with 30 minutes for the changeover) maximises infusion time within this constraint.
This model-based approach also allows clinicians to ask “what if” questions: What if renal function declines? What if we need to cover a less susceptible pathogen? The interactive TDMx platform (https://www.tdmx.eu/Launch-TDMx/) allows exploration of these scenarios for meropenem and other antibiotic classes such as aminoglycosides (Abdul-Aziz et al., 2015; Kothekar et al., 2020).
2.3 Hepatic Clearance and Drug Interaction Screening
Not all antibiotics are cleared renally. Several important antimicrobials (e.g., metronidazole, rifampin, certain macrolides, antifungals) undergo significant hepatic metabolism. For these drugs, hepatic dysfunction and drug–drug interactions become the primary concerns for dose adjustment.
Clinicians should routinely screen for drug interactions — particularly when using rifampin (a potent inducer of cytochrome P450 enzymes), azole antifungals (CYP inhibitors), or macrolides — using resources such as UpToDate or similar drug interaction databases.
3 Site-Specific Limitations of Antibiotic Activity
3.1 Daptomycin Inactivation in the Lung
Daptomycin is one of the most important antibiotics available for treating methicillin-resistant Staphylococcus aureus (MRSA). It is a rapidly bactericidal, concentration-dependent lipopeptide that was originally developed by Eli Lilly in the 1980s but abandoned after phase II trials revealed rhabdomyolysis (muscle damage leading to renal failure) when given three times daily. A biotech company (Cubist Pharmaceuticals) later applied PK/PD principles to recognise that daptomycin was concentration-dependent — a single daily dose both improved killing and reduced the risk of rhabdomyolysis (which was caused by persistent trough concentrations).
However, daptomycin has a critical limitation: it is inactivated by pulmonary surfactant. When daptomycin enters the lungs, it binds to surfactant and loses antimicrobial activity. This was discovered during a clinical trial for community-acquired pneumonia, in which daptomycin failed to demonstrate efficacy.
Never use daptomycin to treat pneumonia — regardless of the pathogen’s susceptibility. It will not achieve adequate antimicrobial activity in the pulmonary compartment. Although the drug could theoretically treat hematogenously-spread infection to the lung; in alveolar spaces the drug will not be effective.
3.2 Antibiotic Activity in Abscesses
Abscesses and large haematomas present another challenging environment for antibiotics. Several factors conspire to reduce drug activity:
- Aminoglycosides are bound and inactivated by purulent material. Low oxygen tension in abscesses impairs the active uptake of aminoglycosides into bacteria.
- Penicillins and tetracyclines are bound by haemoglobin and have reduced activity in the presence of haematoma formation.
- Low pH within abscesses further reduces the activity of many antibiotic classes.
This is precisely why the management of many infections requires source control — surgical drainage of abscesses, evacuation of haematomas, or removal of infected prosthetic material. Antibiotics alone cannot reliably eradicate bacteria in these protected environments.
4 Principle 7: De-escalate Antibiotic Therapy Based on Microbiology Results and Clinical Biomarker Responses
4.1 The Rationale for De-escalation
Once microbiological data become available — typically 48–72 hours after cultures are obtained — clinicians should narrow the spectrum of antibiotic therapy to target only the identified pathogen(s). This de-escalation strategy reduces collateral damage to the patient’s microbiome, lowers the risk of Clostridioides difficile colitis, and minimises selection pressure for antibiotic resistance (Spellberg, 2025).
Key opportunities for de-escalation include:
- Stopping empirical vancomycin when blood cultures grow only Gram-negative bacilli (no MRSA isolated).
- Stopping broad-spectrum anti-resistant pathogen therapy when cultures reveal susceptible organisms.
- Narrowing from combination therapy to monotherapy when a single effective agent is identified.
Remember: every additional day of unnecessary antibiotic therapy carries an approximately 3% increased risk of adverse drug events (Tamma et al., 2017).
4.2 Biomarkers for Guiding De-escalation
4.2.1 Fever and Leukocytosis
The most classic clinical markers of infection response are temperature and white blood cell count. For most infections, some improvement in both should be evident within 48–72 hours of initiating effective therapy. Persistent fever and leukocytosis beyond this window should prompt reconsideration of the diagnosis, adequacy of source control, or appropriateness of antibiotic selection.
4.2.2 C-Reactive Protein (CRP)
CRP is a non-specific marker of inflammation that rises in response to bacterial infection (particularly Gram-negative sepsis) and declines with successful treatment. However, its clinical utility for guiding antibiotic decisions is debated because it responds to many non-infectious inflammatory stimuli.
4.2.3 Procalcitonin (PCT)
Procalcitonin is a biomarker that rises more specifically in response to bacterial infections, especially Gram-negative sepsis. Multiple studies have demonstrated that procalcitonin-guided algorithms can safely reduce antibiotic duration — when procalcitonin declines below a threshold value, antibiotics can be discontinued with confidence (Schuetz et al., 2012).
In practice, many clinicians use procalcitonin asymmetrically: they ignore negative results (which should prompt antibiotic discontinuation) while using positive results to justify continued therapy. This undermines the very purpose of the test. Procalcitonin requires its own stewardship programme and clear institutional guidelines for appropriate interpretation.
4.3 Gram Stain–Guided Therapy
The initial Gram stain remains a critical tool for early antibiotic selection: it immediately classifies the pathogen as Gram-positive or Gram-negative, allowing prompt narrowing of therapy even before full culture results are available.
A recent randomised clinical trial published in JAMA Network Open (GRACE-VAP trial) demonstrated that Gram stain–guided antibiotic selection for ventilator-associated pneumonia resulted in:
- A 30% reduction in use of anti-pseudomonal agents
- A 40% reduction in MRSA-directed agents
- Higher rates of appropriate antibiotic escalation (7% vs. 1%)
