To be honest, I’ve never been a big vancomycin fan. My primary objections to vancomycin are roughly three-fold:
- Vancomycin is given too broadly, in clinical situations where MRSA is extremely unlikely (e.g. urosepsis or community-acquired intra-abdominal sepsis).
- Vancomycin levels are often high, potentially causing nephrotoxicity.
- In the rare patient who does wind up having MRSA, vancomycin levels are often sub-optimal, leading to treatment failure (or at least it seems that way).
However, these aren’t purely problems with the drug itself – they’re largely problems with how we use the drug. We should be able to address them. We’ve already explored issue #1 (misuse of vancomycin) in pneumonia and urosepsis. This post will focus on #2-3: pharmacokinetics.
Why vancomycin levels matter (now more than ever)
The therapeutic window of a drug is the space between the therapeutic dose and the toxic dose. Determining the correct dose is harder for drugs with a narrow therapeutic window: if you’re a bit high you’ll encounter toxicity, whereas if you’re a bit low you’ll loose efficacy.
The therapeutic window of vancomycin has narrowed over the past couple decades. When vancomycin was first used in the 1980s, the target trough level was 5-10 mg/L.1,2 5-10!! How quaint! Unfortunately, over time, bacteria have grown more resistant to vancomycin (a phenomenon known as “MIC creep”). To counter this, the latest guidelines on vancomycin use in 2009 recommended targeting a trough level of 15-20 mg/L for severe infections.3
Another reason to avoid low troughs is that troughs <10 mg/L may promote the emergence of resistant bacteria.4,5 Not only is sub-therapeutic vancomycin ineffective, it may actually render the infection harder to eradicate subsequently.6,7
Bacteria evolved to resist vancomycin, but unfortunately the human kidney hasn’t: kidneys are just as sensitive to nephrotoxicity as they were twenty years ago. The target trough levels we’re shooting for now (15-20 ug/mL) would have been considered toxic levels ten or twenty years ago. In fact, many studies show that trough levels above 15 ug/mL are a risk factor for vancomycin nephrotoxicity.8–11
Thus, the therapeutic window for vancomycin has narrowed. We are forced to walk a fine line between under-dosing vancomycin (therapeutic failure, fostering of antibiotic resistance) and over-dosing vancomycin (nephrotoxicity).
What are we shooting for?
The antibacterial efficacy of vancomycin is both time- and concentration-dependent. The best way to account for both factors is the area under the concentration-time curve during 24 hours (AUC24). Clinical success is predicted by the ratio of the AUC24 to the minimum inhibitor concentration (MIC) of the bacterial strain (AUC24/MIC). The target AUC24/MIC ratio is generally accepted to be >400 to achieve clinical efficacy.
Staph species sensitive to vancomycin have a MIC of 1 ml/L or lower. This implies that when using vancomycin for empiric antibiotic therapy, we should target an AUC24 >400 mg*hours/L.
It’s unclear how high we can push the AUC24 without causing nephrotoxicity. The risk of nephrotoxicity increases gradually with higher exposures to vancomycin (figure above). Overall, it seems reasonable to maintain the AUC24 below 600 mg*hours/L.12–15 Thus, the target AUC24 lies between 400-600 mg*hours/L.
Grudge match: Trough vs. AUC
Calculating the AUC requires a bit of work. A shortcut approach to guarantee clinical efficacy (AUC>400) is to keep the trough level above 15 mg/L. As shown below, mathematically this forces the AUC to be at least 400:
The pharmacokinetic simulation shown below illustrates the relationship between trough and AUC.16 A trough above 15 mg/L guarantees that the AUC will clinically effective (>400). The problem with targeting a trough level between 15-20 mg/L is that occasionally the AUC will be insanely high.
In short, achieving a trough >15 mg/L ensures efficacy but not safety. This issue has been shown clinically as well:17
Another illustration of why trough levels are a poor surrogate for AUC is that the relationship between trough and AUC is strongly influenced by the dosing interval:
Ultimately the trough level simply fails as a measurement of vancomycin exposure. By itself, the trough is a poor predictor of treatment success.18 When compared with the trough, the AUC better predicts which patients will respond to therapy.19–21
To ensure optimal AUC/MIC, especially in critically ill patients, estimation of the AUC should be mandatory –Montiero et al 20187
Why conventional dosing fails
Conventional dosing consists of giving a weight-based dose (e.g. 15 mg/kg q12) and then checking a trough level after four doses, targeting a trough of 15-20 mg/L for serious infections. There are numerous reasons this strategy fails, especially among the sickest ICU patients:
- Trough levels cannot ensure administration of a safe dose. Any strategy which is targeted solely to trough levels is fundamentally flawed.
- The volume of distribution (Vd) is often abnormal among ICU patients – especially patients with volume overload, burns, pregnancy, morbid obesity, or systemic inflammation.
- Some patients in the ICU experience augmented renal clearance. In response to stress, these patients have elevated renal drug clearance. When treated with standard doses of vancomycin, they will have low vancomycin levels.
- Many patients in the ICU suffer from acute kidney injury, which often isn't accurately reflected by their creatinine levels (it may take several days for creatinine levels to re-equilibrate following a change in renal function). These patients are at risk for accumulating high vancomycin levels.
- For reasons #2-4 above, the estimated vancomycin dose will often be wrong. Waiting until the fourth dose to check a trough may cause patients to receive the wrong dose for 48 hours.
If vancomycin is dosed using conventional strategies among critically ill patients, patients will commonly receive inadequate or toxic doses. A multinational observational study found that target drug levels were achieved in only half of patients!22
Re-evaluation of current vancomycin dosing recommendations in critically ill patients is needed to more rapidly and consistently achieve sufficient vancomycin exposure –Blot et al 201422
Equation-based pharmacokinetic drug modeling: a more precise approach.
Vancomycin dosing can be modeled nicely using a single-compartment model, where vancomycin distributes into a single compartment and is subsequently eliminated via the kidneys. This model appears to work fairly well, except for the first few hours after a dose of vancomycin (while the drug is being distributed to the tissues).
Two points define an exponential decay curve. Thus, using a single-compartment model, two drug levels are required to define vancomycin pharmacokinetics. This is slightly more effort-intensive than simply measuring a trough, but it provides considerably more information:
- The volume of distribution (Vd) is measured (not estimated based on the patient’s weight). This avoids errors encountered in patients who are pregnant, obese, or edematous.
- The constant of elimination (Ke) is measured. This eliminates the vagaries of estimating renal function based on a random creatinine level.
- Measurement of these parameters allows calculation of the patient's AUC.
The nuts and bolts of equation-based pharmacokinetic dosing are explored here. This does require measuring more vancomycin levels. The advantage is that this allows for immediate, personalized targeting of therapeutic drug levels. The math involved can be performed by a variety of pharmacokinetic calculators including a Microsoft Excel spreadsheet:
Alternative method of pharmacokinetic calculations: Bayesian strategy
Bayesian strategies start with a population database that relates pharmacokinetic parameters to several pieces of patient data (e.g., age, gender, weight, creatinine). This database is used to create a computer model which predicts an individual’s pharmacokinetics from several pieces of information.
The main advantage of a Bayesian strategy is reduction in the number of vancomycin levels that need to be checked (e.g. rather than checking two levels, only one level may be required). However, Bayesian strategies have a number of limitations:
- The volume of distribution is often estimated (not measured directly).
- Bayesian models are developed within a certain population, with a specific ethnic makeup and range of body types. Such models may fail in other populations or unusual body types. Alternatively, an equation-based approach makes zero assumptions and therefore works equally well on any human being (or mammal for that matter23).
- Numerous different Bayesian models exist, so success obtained with one Bayesian model doesn’t guarantee that a different Bayesian model will work. Alternatively, equation-based approaches yield consistent results, making this easier to generalize results across different centers.
- Equation-based strategies are available for free. Alternatively, Bayesian models require purchasing a computer program. This may create pro-Bayesian bias in the literature, because papers validating Bayesian models may be written by authors with conflicts of interest.24,25
- An equation-based approach is easily adaptable to other situations (e.g. pediatric patients, aminoglycosides). This is a versatile, timeless tool which is worth learning. In contrast, spending time and money to learn a Bayesian computer program is a dubious investment because any computer program might not exist in 10-20 years.
- An equation-based approach provides a real-time measurement of the glomerular filtration rate. This is probably more accurate than our typical approaches to estimating glomerular filtration rate. Serial measurement of real-time glomerular filtration rate could provide useful information to detect acute kidney injury rapidly.
- One study directly comparing an equation-based strategy with five Bayesian programs found that the equation-based approach was equivalent or superior to the Bayesian programs.25
In fairness, a Bayesian approach appears adequate for more stable patients (e.g. floor patients with a stable creatinine level). Alternatively, an equation-based approach may be superior in the ICU, where patients tend to have the most bizarre pharmacokinetics (e.g. augmented renal clearance, severe systemic inflammation with expanded volume of distribution).
Pharmacokinetic equations: evidence
The principles of pharmacokinetic equations were described in 1976 by Sawchuk and Zaske.26 They haven’t changed over the past forty years. Although no multi-center RCTs has been performed on these equations, decades of experience support their efficacy. Some newer studies using pharmacokinetic equations are as follows.
Hong 2015: Morbid obesity
This is a before/after study which compared traditional vancomycin dosing to two-point pharmacokinetic equations in patients with morbid obesity.27 In both groups, the therapeutic goal was to achieve a specific trough level (10-15 mg/L for mild infections; 15-20 mg/L for severe infections). The accuracy of the first trough level was the same between groups. Dose adjustment using pharmacokinetic equations doubled the likelihood of achieving a target trough level on the second measurement (65.2% vs 31%; p = 0.024).
The volume of distribution and drug clearance measured using pharmacokinetic equations was compared to the total body weight and estimated creatinine clearance (figure above). Total body weight failed spectacularly in predicting the volume of distribution, which makes sense (inflammation may increase the volume of distribution by a factor of 2-3). Two patients were identified with augmented renal clearance. Overall this study suggests that conventional weight-based dosing will often fail these patients, who may be better served with personalized equation-based dosing.
Fitch 2016: Patients in multi-center hospital network
Finch et al performed a before-after study documenting the transition from conventional trough-based vancomycin dosing to dosing based on two-point pharmacokinetics at four hospitals.15 Following introduction of two-point pharmacokinetics to target an AUC>400, patients had lower vancomycin troughs. This is exactly what would be predicted based on the relationship between trough levels and AUC explored above.
Patients in the post-intervention group were sicker (e.g. higher comorbidity and APACHE-II scores). In unadjusted analysis there was no difference in nephrotoxicity between the two groups. However, using a variety of different multivariable models, the use of two-point pharmacokinetics was associated with a reduction in nephrotoxicity.
Truong 2018 & Miller 2018: Pediatric and adult patients at Loma Linda University
This hospital transitioned from conventional vancomycin dosing to equation-based pharmacokinetic modeling. Their pharmacokinetic strategy involves checking two levels after the first vancomycin dose, which were used to optimize subsequent doses. Pharmacokinetic modeling was used to target a vancomycin trough level (not an AUC24).
In both adult and pediatric populations, the implementation of pharmacokinetic modeling hastened the achievement of target vancomycin troughs.28,29 Although choosing to target a trough level is debatable, these studies demonstrate that pharmacokinetic modeling works and it will facilitate reaching whatever goal you choose.
Parting shot: use vancomycin more selectively & more precisely
Pharmacokinetic dosing might seem like a lot of work, but that’s not necessarily a terrible thing:
- This might cause us to re-consider whether or not the patient truly needs MRSA coverage (rather than just reflexively blanketing every critically ill patient with vancomycin).
- This might cause us to re-consider whether linezolid might be better for some patients (e.g. patients with pneumonia or patients with extremely dynamic renal function). Because linezolid is mostly metabolized by the liver, it doesn't require renal adjustment.
Given the narrowing therapeutic window for vancomycin, it’s probably appropriate to use it more selectively. However, when we do choose to use it, we should do so with thoughtfulness and precision.
- Increasing resistance to vancomycin is making it harder and harder to achieve vancomycin levels which are both safe and effective.
- Efficacy and toxicity of vancomycin appear most closely related to the area under the concentration-time curve over 24 hours (AUC24). The optimal balance of efficacy versus toxicity seems to occur within an AUC24 range between ~400-600 mg*hr/L.
- Traditionally, vancomycin doses have been adjusted to target a specific trough level. However, it is increasingly clear that the trough level is an inadequate measurement of vancomycin exposure. Furthermore, the practice of waiting until the fourth dose to measure the trough level may expose the patient to days of suboptimal therapy.
- Up-front two-point pharmacokinetic modeling allows calculation of drug clearance and volume of distribution. This facilitates immediate optimization of drug doses, maximizing efficacy while simultaneously avoiding nephrotoxicity.
- Equation-based pharmacokinetic modeling isn't rocket science. This can be done with free online calculators or an Excel spreadsheet (explored further here).
Related
- Nuts & bolts of vancomycin PK monitoring (PulmCrit)
References
- Pulmcrit wee: The cutoff razor - April 15, 2024
- PulmCrit Blogitorial – Use of ECGs for management of (sub)massive PE - March 24, 2024
- PulmCrit Wee: Propofol induced eyelid opening apraxia – the struggle is real - March 20, 2024
Do you have experience with pharmacist management of vancomycin? I would be curious your general impression.
We often have more time to specifically dedicate to vancomycin phamcokinetic calculations than the physicians do. I have only been at facilities that had vancomycin dosing done primarily by pharmacists.
Is it still a rather common practice for physician management of vancomycin dosing utilizing “conventional” methods?
Yikes, I feel like I’m about to get myself in trouble with this.
I think pharmacists generally do a better job with this. The ideal approach might be to have a pharmacy service which systematically manages this..
That being said, I still think that *everyone* needs to understand this, for a few reasons:
1) Understanding the pharmacokinetics can help physicians understand what is going on with the patient (e.g. renal failure etc).
2) Not every hospital everywhere has a pharmacy service 24/7/365, so physicians should be able to do this on their own.
Josh, Thanks for the vote of confidence for pharmacists. As a clinical pharmacist that heavily advocates pharmacist directed pharmacokinetic services, I agree with your thoughts. Everyone does need to understand at least the basics of how this is done and what these numbers mean. At our teaching hospital we do everything we can to help our residents understand how and why we are making our recommendations for vancomycin. Not all of them will go to places that have extensive pharmacy services, so they need to know how to do this as well. I will say that were I think special… Read more »
Nice approach. I’d add some form of pharmacokinetic modeling for the loading dose too – Regimes that use fixed doses tend to lead to initial over-or underdosing.
I wonder if clearance of vancomycin has ever been looked at as a proxy for GFR? It’s led me to a diagnosis of complications of uremia once so far (slightly high creatinine in a patient with no legs).
What do you use to model pk in dialysis patients?
Agree that the loading dose needs to be chosen carefully. This is a bit tricky because recommendations on loading dose are all over the map.
Yes, the elimination constant (k) is generally linearly related to the GFR, so pharmacokinetics should be able to diagnose acute kidney injury early. Many vancomycin pharmacokinetic calculators will display an estimated GFR (e.g. https://emcrit.org/squirt/vanco/)
Not sure about dialysis patients. Pharmacokinetics may be a bit less important in the context of chronic dialysis because if you overshoot a bit there is no real risk of AKI.
Hi,
Any comment on vancomycin as a continuous infusion (eg 2g/24 hours)?
It is used reasonably frequently in some centres, and provides a safe trough, adequate AUC, and 100% t > MIC.
It is also easy to monitor – less frequent blood tests; blood tests not time-specific or affected by previous intermittent infusion rate etc.
It also transitions nicely to ward and outpatient care in the (rarer) circumstances that 14 days Vanc is required.
Challenges (eg getting initial dose right) are similar to intermittent dosing.
Thoughts?
This is standard practice in the unit where I currently work. Of course you have to give a loading dose first, or that MRSA-bacteremia might go untreated for days. It works pretty nicely in most patients and is relatively easy to adjust, but the one real caveat is that levels do remain time-specific. A steady-state level wouldn’t be, but this is only achieved when infusion rates have been unchanged for about five half-lives. As we usually check levels daily, they tend to reflect transition states, except in patients with augmented clearance. You do develop a “feel” for the direction you’re… Read more »
Agree, vancomycin continuous infusion sounds like an outstanding option. If you’re at a shop with continuous vancomycin infusions then that’s probably a superior strategy honestly compared to intermittent infusion. There is some PK work that goes into the loading dose & maintenance rate but then it’s very easy to monitor and adjust (you can just check one level; your AUC is simply 24 multiplied by the level). Continuous vancomycin infusions doesn’t seem to be done in the USA, so I didn’t pay a lot of attention to it. Unclear whether the hospital culture here is ready to move in that… Read more »
Stumbled on this site. It is cool. Thanks for your efforts. Its a difficult topic to summarize in a short monograph and you’ve done an admirable job of it. I’m skeptical about changing to TDM using AUC:MIC calculation. One point of clarification; where you show the graph by Chavada et al. (Fig 1 in their report) the AUC values associated with trough of 15 -20 there are 2 values above 6oo . Is this what you mean by “toxicity” ? It would be more correct to say that they have increased risk of toxicity no? But Chavada found on univariate… Read more »
Sorry, one last point. You & readers will notice quite a discrepancy between the correlation of AUC to trough shown in Figures above in the Pai publication and the Chavada study (r^2 of 0.41 vs spearman R of 0.923). Firstly, by visual inspection Chavada data look quite linear so its curious that they would use spearman (a rank order correlation for non-parametric data). I emailed the corresponding author & he didn’t get back to me with an explanation. So I really nerded out and using a plot digitizing webapp I recreated the dataset & found pearson correlation of 0.94 (or… Read more »
Continued from above:
Abulfathi AA et al. JCP 2018, r = 0.80
Suchartlikitwong et al JCP 2019, r^2 = 0.94
Clark et al DMID 2019, r^2 = 0.731
Bel Kamel et al TDM 2017 r^2=0.51
Jin et al Infect & Chemo 2014 r = 0.96
Hahn et al , TDM 2015 r^2 = 0.50
& there may be more
Josh, my department have been reading all of the articles we can on this subject. Some of them contain the equations that many calculators are using. One that seems popular, but which doesn’t seem to be included in the original paper, is the one from Goti et al. Their paper only really shows tabulated results. Any idea where people are getting the actual equations used?
See table 2 in their paper. That will give you the parameter values of CL, Vc, Vp, and Q along with the covariate relationships. Goti et al is a two compartment model so you will need to use two compartment infusion kinetic equations. Pharmacy school/most pharmacokinetic textbooks only describe one-compartment models and their equations. Here is a link to a resource that describes two-compartment kinetic model equations: http://r-forge.r-project.org/scm/viewvc.php/*checkout*/PFIM3.2.2/PFIM_PKPD_library.pdf?root=pkpd