I am not sure if this forum has any interest in application of models to patient care or not. If this is only regarding Pumas software usage related, please ignore my question. I see a relevance as models will play a major role in clinical pharmacotherapeutics. If not, I would value the communities views.
We have been using Ctrough (or now AUC) to initiate and adjust vanc dosing. The AUC typically is calculated as AUC(0-24hours) - historically we used only steady-state exposures to adjust doses. The new guidance on vanc target AUC calls for AUC(0-24) - it is not clear if this is after the first dose OR is it at steady-state.
Dear Bob,
This is an open forum for the science of drug development where one of the focuses is on quantitative aspects. We certainly appreciate your interest and encourage you to discuss.
With regards to vancomycin dosing guidelines, the AUC(0-24) in the new guidelines is about steady-state IMHO.
You’ll find a wide range of specialties and perspectives here, but for my part, the answer is an emphatic, “yes!”
I’m glad you mentioned vancomycin because I think it’s perfect for illustrating why model-informed precision dosing is so badly needed. Most adults are started somewhere between 15-20mg/kg q8-12h +/- LD despite well-documented interpatient variability in vancomycin exposure from multiple sources. We account for that variability by using TDM to calculate “individualized” doses, but we’re still just chasing the therapeutic target through trial and error. The information we’re gathering has no predictive power because it’s only relevant to the dosing interval from which samples were collected. The end result is usually suboptimal therapy, with <50% of patients achieving therapeutic targets based on their initial dose. That means ~50% of patients are either supratherapeutic at an increased risk of VIN, or subtherapeutic at risk for therapeutic failure or resistance (at the extreme).
In many cases, a well-developed Bayesian population PK/PD model can accurately predict vancomycin exposure before the first dose is administered and that accuracy improves as patient-specific information becomes available. That means faster, more reliable target attainment with a decreased risk of toxicity or therapeutic failure compared to the “traditional” approach.
Of course, that doesn’t mean there aren’t significant challenges to the widespread adoption of model-informed methods. The effort needed to shift current clinical practice away from the one-size-fits-all approach can’t be overstated (and probably requires its own thread…), but I think we will reap enormous benefits if we put in the work.
This is a really good question. I have always interpreted the recommended “daily AUC target” (AUC24) of 400-600 mg*hr/L as being a steady-state (SS; AUC24,SS) value, but there is definitely some ambiguity in how the recommendation is worded. I think the choice of wording comes from the studies used to define the range and more importantly, their methods for calculating AUC24 (there are 2). The first is a simple “equation” based approach that relies on first-order kinetics and algebraic manipulation. With this approach, you can calculate a reasonable estimate of AUC24 by using two appropriately timed levels collected at steady-state. In other words, this method of calculating AUC24 only works if the patient is at SS, which makes it AUC24,SS by convention. In contrast, the second approach relies on Bayesian estimation, which does not share the same sampling requirements. You can use a single random sample (usually a trough) collected after the first dose to calculate AUC24 (current) and predict the value of AUC24,SS (~3 doses from now) if you continue at the same dose. Since the goal is to maintain an AUC24 between 400-600 throughout the treatment period, I’d argue that the SS designation is implied.
Hello! The guidelines refer to steady-state - From my experience we would order 2 steady-state (or near- steady state) levels within the same dosing interval at 1-2 hours after the end of infusion and then at trough.
I hope that helps and am excited to see AUC-guided dosing happen!
@haden@adunn34 Thank you very much. Wouldn’t you then agree the new guidelines are already outdated? If we have to wait till SS to make dose adjustments.
Also, what is so special about SS to be used as the therapeutic goal? Wouldn’t we agree the first 48-72 are clinically most relevant. Then how does this affect the therapeutic target and dose adjustments. Any work done so we can use first dose levels to dose adjust?
Bob
The equations that are often used in conjunction with TDM to calculate “patient-specific” PK parameters are only valid if we assume that vancomycin is at SS. That’s why most hospitals collect vancomycin samples between the 3rd and 4th dose, they’re accounting for SS without actually saying it. For q8-12hr dosing, that’s somewhere between 24-48 hours of therapy relative to the 12-30 hours of regular dosing needed for vancomycin to reach SS (3-5 elimination half-lives * average elimination half-life of 4-6 hours in adults = 12-30 hours to SS). It’s also worth noting that time to SS is unaffected by the size, number, or frequency of dosing. All that matters is that the drug is given at regular intervals.
Based upon its half-life relative to how often it’s given (2-3x/day), vancomycin accumulates in the plasma. This means that if you adjust your dose based upon a non-SS value, you risk giving the patient more drug than is needed to reach the target of 400-600 mg*hr/L. For example, let’s say that we have a patient, for whom the elimination half-life of vancomycin is 6 hours and we want to start them on 1250mg q12h. Thanks to accumulation [1 / (1 - exp(-kel*tau))], the AUC12,SS is going to be ~35% higher than the AUC12 calculated from the first dose. There’s also the added complication that you can’t superimpose non-SS AUCs so there’s no simple way to get AUC24 from AUC12 in this case. Bottom line, things get very dicey when you try to adjust dosing based upon a non-SS target.
Not quite, and that’s where the real beauty of the Bayesian method shines through. Unlike the traditional approach (Sawchuk-Zaske, first-order PK equations) that requires you to collect samples at SS to obtain accurate estimates of AUC24,SS, Bayesian forecasting can do it using a single random sample collected at any point after the first dose. That way (as you pointed out) we can adjust our doses much earlier in that critical 24-72hr period following the start of therapy. Early adjustment means faster target attainment, faster time to cure, less days of vancomycin therapy, and a decreased risk of VIN.
It is not practical to expect clinicians to be trained in performing Bayesian modeling in a hospital setting.
“Not quite, and that’s why…” My point is why then have an illusionary or misleading SS target, let us have a AUC first dose target. The SS target suggests there is something special about SS. Clinically it does not make sense. What you are alluding to is your ability to predict AUCSS from first dose, why create two hoops.
I don’t think it’s illusionary, I think it’s practical. If you’re trying to validate a therapeutic target by linking it to a clinical outcome you’d want to assess that target at a point where it’s consistent (i.e., SS), not while it’s changing. Otherwise, you’d confound your results and wouldn’t have a clear signal of how it performed. We know that AUC/MIC is the primary PK/PD driver of vancomycin efficacy and over the course of a handful of studies, primarily in patients with bacteremia, we’ve associated a daily AUC (AUC24) range of 400-600 mg*hr/L with clinical cure and a decreased risk of AKI (assuming MIC = 1). However, clinical cure takes days/weeks to assess and in most cases vancomycin achieves SS in 1-2 days, thus these assessments are made while vancomycin is at SS and the AUC24,SS designation is implied. If the clinical outcome is linked to a PK/PD target that was measured under SS conditions, we can’t perform that same assessment metric obtained after just one dose.
SS is special for 2 reasons. 1) Systemic exposure during ss (AUCtau) is the same for each dosing interval provided the dose is unchanged. That makes monitoring under SS conditions relatively simple. 2) Drug accumulates in the plasma while dosing to SS and that change in exposure with repeated dosing, be it Ctrough, or AUC, has to be accounted for. Vancomycin AUC can increase as much as 200% at SS compared to first-dose depending on individual half-life and how often it’s given. Bayesian forecasting allows us to account for that change without having to actually wait for SS. It’s not a second hoop, it’s a shortcut.
@haden I get what you are saying. I respect your opinion very much. I am not trying to argue. In fact, your thoughts and your work are very inspiring.
What I meant by ‘being illusionary’ is that we teach not-so-good things in our schools. Primarily dating to WWII concepts, when we did not have computers and well developed PK modeling. Today , we do have those. One cannot argue that for a drug with t1/2 of 24 hours, used in ER we need to wait for SS. Nor would we expect for a t1/2 of 4 weeks that we need to wait for reaching SS to elicit its effect on tumor shrinkage, which typically can be seen in 4-8 weeks. We invented SS concept to keep our math overhead low, when the tools were not so well-developed. Now we dont need to. I think we ought to strive to come up with therapeutic targets that independent of time. For eg, warfarin’s target INR is 2-3 (depending on who you ask). That range is independent of time, meaning its not ok to have INR << 2 for even first 4 days of dosing. This should be the same for PK targets. After all, AUCss = function(AUC1st, kel, tau) (this is a unidirectional relationship).
I will take a look at the reference you provided, looks you are taking this individualized pharmacotherapy by its horns. Good for the patients!
@bobbrown Thank you for your kind words. I didn’t see your comments as argumentative and please forgive me if I came across that way in my replies.
Your thoughts on SS are timely since they were echoed by a colleague of mine just this morning in a completely separate context. I understand that we invented SS as a mathematical convenience, but it’s difficult for me to envision a practice scenario in which we don’t have some therapeutic targets that are time-dependent. Even with warfarin. While INR itself isn’t time-dependent, we’re still waiting on accumulation for max effect and a decline in factor II over 5-7d.
Regardless, you’ve given me something to ponder which I greatly appreciate. Welcome to the board, and I look forward to future discussions.
@haden@bobbrown this has been really thought-provoking to read, thank you both for this discussion. I think you bring up some good points @bobbrown - clinically, it doesn’t make sense to strive for a method that requires waiting until steady state (especially when you’re dealing with critically ill patients). I understand your perspective. @haden I am very eager to see Bayesian forecasting implemented I can see your explanation putting a lot of those with reservations at ease.
Great discussion everyone! @bobbrown I agree with you that given how far we’ve come, dose adjustments should be based on concentrations after the first dose and not on SS. Fortunately we can do this using Bayesian modeling but clinicians do not have to know anything about these quantitative methods as they can be easily abstracted away from them with the advancements in softwares.
As @haden mentioned, using pioneering CDS such as Lyv, which are very user-friendly, just by entering some patient characteristics (age, sex, height, weight, SCr, etc.) and updating a patient’s concentration values as they become available (even the ones after the first or second dose!) 24-hour AUCs can be forecasted into the future for a specific dosage amount. The software will also recommend what dose must be administered to achieve an AUC of 400-600 mg.h/L. The accuracy of the forecast increases as we input more concentrations.
The Therapeutic guidelines also acknowledge the advantage of Bayesian approach: “An advantage of the Bayesian approach is that vancomycin concentrations can be collected within the first 24 to 48 hours rather than at steady-state conditions (after the third or fourth dose), and this information can be used to inform subsequent dosing.”
Regarding the target itself being SS AUC, I think we should focus on achieving the AUC target as soon as possible while administering safely given how vancomycin can cause infusion related reactions, nephrotoxicity, histamine release, etc.
TL,DR: With advancements in methodologies and softwares, there is no need to wait until steady-state to perform dose adjustments for TDM of vancomycin. With user-friendly clinical decision support systems, dose adjustment approximations can be made within 1-2 days of starting therapy.
Great discussion everyone. I see that it could be acceptable to wait till the third dose (i.e after 1 day) which is the time of reaching ss for vancomycin to decide the appropriateness of the dose if the infection is not severe. Usually vancomycin is reserved for anti-biotic resistant microbial infections which could be critical and need prompt treatment as early as possible. As others suggested, Bayesian methodology to forecast vancomycin concentrations at steady state could be considered. Just want to highlight that in TDM, It is important to measure the drug concentration at ss because at this phase we can assume that the drug concentration in plasma is at equilibrium with concentration at receptors or site of action (in this case the bacteria) and this is important in TDM as we can use concentration as surrogate for drug effect.
Another approach to have idea about AUC0-24 is to use approach as described by "Rodvold KA, Blum RA, Fischer JH, et al. Vancomycin pharmacokinetics in patients with various degrees of renal function. Antimicrob Agents Chemother 1988
@ahmed.salem Intriguing comments. May I know you think it is acceptable to wait for 3-4 days to dose adjust while the patient is suffering with infection ? Remember, these patients are hospitalized. Is it even ethical?
If there are other clinical pharmacists, they can weigh-in as well.
You indicate that reaching PK SS implies reaching equilibrium at the site of action. Can you please provide any scientific basis for this phenomenon? I am unaware of this concept. In ID, for ABSSSI one of the clinical endpoints is resolution of fever in about 48 hours - how does this then depend on the time to PK SS? The newer cancer treatments have a very long t1/2s, resulting in reaching PK SS after months - however typically you should see effects on tumor size by 6-8 weeks - how do we explain the requirement to reach PK SS to define the time to reach equilibrium at the receptor site? Could you please explain. I might be missing something key here. Thank you.
@bobbrown you’re right in that reaching steady state doesn’t always align with how we monitor for efficacy. I think we need to keep these concepts separate in most cases outside of infectious disease. In my eye, ID is a special circumstance because we know that antibiotics are most effective at concentrations at 3-5x the MIC. Once steady state is reached, we then know that we are continuously & consistently going to be reaching that benchmark. Does that mean we should HAVE to wait to SS to dose adjust? Certainly not - I agree with you on this.
If we want a way to assess pre-SS exposure, rudimentary methods won’t cut it. Pre-SS exposure does not accurately reflect if AUC/MIC will remain optimal in the long-term. It’s too highly variable. But with Bayesian software in hand, we can alleviate this issue and implement a monitoring parameter that is both timely and accurate. As a side note, I think this conversation is a great example of why we need active engagement from health care professionals - what may make sense theoretically will not always always work in practice.
The half-life of vancomycin is 4-6 hours so the drug is expected to reach the steady state approximately within one day. As I mentioned, it depends on the clinical context whether the infection is life-threatening or not to evaluate the utility of waiting one day till adjusting the dose. Some empirical formulas based on patients’ covariates were proposed to approximate the individual’s exposure ( I posted one formula) could be used also for initial management. If you used Bayesian methodology, I think will still need to match foretasted exposures at steady state since the current guidelines is based on measuring AUC24 at steady state.
My understanding about why at steady state specifically is that this would be the exposure the level where the drug ultimately ends and will stay at following chronic administrations so matched to efficacy and TDM. However, I agree with you this is not universal, for example, Amiodarone has long half-life (60-140 days) so, it will be risky to wait till steady state to adjust the dose for patients with hepatic impairment.
I can see your explanation putting a lot of those with reservations at ease.