Implementing the 2020 Vancomycin Guidelines

Dr. Kristi Kuper: Well hello everyone, and welcome to the webinar today entitled Vancomycin Dosing: Implementing the 2020 Guidelines. My name is Dr. Kristi Kuper and I am Director of Clinical Pharmacy for DoseMeRx, which is part of Tabula Rasa HealthCare and I will serve as the moderator for today’s webinar.

So to get us started, I have just a couple of housekeeping announcements. We are making our panel members even more famous by recording this webinar, and in the coming days, you’ll receive the link to the recording via the e-mail address that you used to register. The response to this webinar has been amazing, and we have almost 2,000 participants registered. So as you can imagine, we will need to keep all lines in Listen Only mode.

Now, you have the ability to download a copy of the handout by clicking on the gray document icon, which is located on the right side of the screen in your webinar control panel. We encourage questions, so please feel free to type them into the Q&A box, and our speakers will be answering as many questions as time permits at the end of the webinar.

We’re going to go ahead. And before we move on to the formal presentation, we’d like to do a quick poll with the audience today. We’re interested in knowing where your institution is in their journey with converting from trough to AUC-based dosing for IV Vancomycin. What you’ll see is: This poll question will populate on your screen, and it takes just a few seconds. There are five responses that you can choose from.

First of all: A) I have not started yet, B) We are in the early planning stages, C) We are currently implementing AUC dosing, D) We have already implemented AUC dosing, or E) Other or not applicable. So, if you just take a minute and click on the answer that most likely applies to your institution, that would be great. And this will be really helpful for our speakers and it gives them a sense of who are listening in today.

We’ll give it just one other minute or a few seconds. Okay, looks like we’ve got quite a number of people answering. Let’s go ahead and close the poll. Great, okay, so it looks like we have quite a diversity here of responses. Great. We appreciate you responding and that, again, really gives us a really good sense of who is listening in today. Okay, so let’s go ahead and get started. I’m personally very excited to introduce our exceptional speakers for today’s webinar.

Our first speaker is Dr. Tom Lodise. Tom is an accomplished researcher and clinician and is one of the lead authors on the recently published 2020 Vancomycin Monitoring Guideline. He is a professor of pharmacy at Albany College of Pharmacy and Health Sciences in Albany, New York. In addition to his academic responsibilities, he’s also a practicing IV clinical pharmacist at the Stratton VA Medical Center.

Our second speaker is Dr. Ethan Smith. Ethan is the program coordinator for the Antimicrobial Stewardship Program at Cedars-Sinai Medical Center. Cedars-Sinai is one of the largest non-profit academic medical centers in the US that’s located in Los Angeles, California. In addition to this role, he’s also an assistant professor at the University of Southern California and the University of California San Francisco Colleges of Pharmacy.

With that, I’m going to go ahead and turn it over to Dr. Lodise. Tom?

Dr. Tom Lodise: Thank you for the introduction. Kristi, can you hear me?

Dr. Kristi Kuper: Yes.

Dr. Tom Lodise: Okay, great. I just want to make sure I wasn’t on mute. Welcome, everybody. Thank you joining us today and taking time out of your busy schedule to discuss the updated consensus guidelines. In some ways, we can shift some of our practices from trough-only monitoring to AUC-guided dosing. A few things, we only have a limited time here and I’ve done a few of these types of webinars, and what I’ve found is: The best part is by far the audience participation.

Rather than having a little bit of build-up, I thought the best way to begin this presentation is to jump right into the major changes. These are some of the simplified recommendations. First and foremost, in patients with suspected or definitive serious MRSA infection — so again, the emphasis on serious MRSA infections, and individualized AUC/WIC target ratio, 400 to 600 assuming a vancomycin MIC of one, should be advocated to achieve clinical efficacy while improving patient care.

With the assumption of a broth MIC of 1 microgram per mL, that equates to a targeted AUC range of 400 to 600 per day. In contrast, due to previous guidelines, which recommend checking [INAUDIBLE 00:05:31] after the fourth dose, given the importance of early, appropriate therapy, vancomycin targeted exposure should be achieved early in the course of therapy, preferably within the first 24 to 48 hours.

A few things that come up is, in a situation where you might have an MIC by the gold standard broth microdilution method, when it’s greater than 1, it’s important to recognize that achieving an AUC/MIC target of greater than 400 is low with conventional dosing. Higher doses may be needed, but there is likely some toxicity risk associated with that and it’s a risk versus benefits situation. I encourage you to rely on your clinical judgment.

Also, a thing that came through as far as developing the guidelines, is what happens if you have a broth microdilution MIC of .5 or .25? Our current recommendation right now is not to decrease the vancomycin target AUC range. The reason is we don’t have data to show that AUC de-escalation still results in efficacious outcomes.

Other things: As part of the recommendation, trough-only monitoring, where they target a 15 to 20, is no longer recommended based on efficacy and nephrotoxicity data in patients with serious infections due to MRSA. I got a question that always comes as well, who do I apply these results to? Again, a question that always comes is: Who do I apply these results to? The emphasis is really on patients which serious MRSA infections.

And currently, there’s insufficient evidence to provide recommendations on whether trough-only or AUC-guided vancomycin monitoring should be used among patients with non-invasive MRSA infections or other types of infections due to other pathogens. Those are really the major changes within the guidelines. We have some recommendations for loading doses and maintenance doses in specialized populations, including those with obesity, requiring replacement therapy, as well as pediatrics.

The biggest thing is this move away from trough-only monitoring. Again, the whole reason why the 2009 consensus guidelines recommended troughs of 15 to 20 is really a surrogate of achieving an AUC/MIC ratio of 400. When we moved to this practice, what we had was an opportunity to really evaluate the outcomes associated with both efficacy as well as toxicity. And despite this move across all institutions throughout the world, the clinical benefits of maintaining a trough of 15 to 20 have not been well-described.

There’s one study that’s notable coming out of Detroit that did identify a link between clinical success in vancomycin trough values among patients with MRSA bacteremia infections. And what they saw was this inverted U-curve where failure was less than 15%, dropped to 40% and increased again above 20%. Looking at this, I think this is somewhat along the lines of an unstable exposure response relationship, but I would argue a failure rate of 40% really doesn’t get us to the top of the dose response curve.

Again, we all say we did this 15 to 20. Now, we did at our institution, and the important thing to recognize is why we felt good about getting people within that range… There’s really not a lot of data to support that maximized clinical outcomes. The other thing when you consider with dosing, not only does the drug need to be effective, but it also has to be safe. What we did find it getting people into that 15 to 20 range resulted in an increased rate of acute kidney injury relative maintaining troughs less than 15. This is a meta-analysis I was involved in 7 years ago looking across the literature at the time and saw a very strong relationship between troughs greater than 15, an AKI relative maintaining troughs less than 15, and there was an exposure response relationship noted.

In JAC, Journal of Antimicrobial Chemotherapy, there’s a subsequent meta-analysis published in this month’s edition which really found a similar relationship. So, a few things about the relationship between vancomycin exposure and acute kidney injury: One thing that always comes up is: What came first, the chicken or the egg? And so, what we did is this is a study we did many years ago looking at initial vancomycin dosing. Again, this was at a time when the new recommendations for maintaining troughs in the 15 to 20 were released.

What we saw was some had already [INAUDIBLE 00:10:09] dosing practices at our institution. As a positive control, we include all the [INAUDIBLE 00:10:15] four or more grams a day. And really, what we found is, first and foremost, over the first few days of therapy, there’s no difference in the toxicity curve. That’s the one thing we need to realize with vancomycin acute kidney injury: Most allergic reactions occur after four or five days. Now, I would argue that doing the first few days of therapy, any bumps [INAUDIBLE 00:10:40] and this applies to the aminoglycosides as well, is really due to hyperperfusion of kidneys and patients with sepsis.

And what do you see over time is a separation of the curves. The other thing I want you to make note here is, again, we think about vancomycin acute kidney injury. There’s a background rate among patients who are using it. If you go into an ICU, many patients are going to develop acute kidney injuries for reasons other than vancomycin. However, what the literature does tell us is when you do use vancomycin, you have more intensive exposures, you have more AKI relative to maintaining vancomycin and less intensive exposure, depending on the patient population you’re interested, the more vulnerable, the more AKI you’ll see, the less vulnerable, the less AKI you’ll see. That’s why you see a difference in vancomycin AKI rates in inpatients versus those within the ICU.

The other thing that always comes up is, we do this whole thing with vancomycin, and it’s always thought to be a surrogate of achieving an AUC/MIC ratio of 400. And what we’re finding is, it’s probably not a good clinical surrogate. So, a trough is just an exposure at the end of a dosing interval, and it really varies at a function of how often the dose is given: Is it two or three times a day? As well as the patient clearance.

Probably, a more informative exposure variable is the AUC, which is integrated exposure over time. So, a trough ensures a lower bound of the AUC value. But what you see is up to a three-fold to four-fold variation at the top end. So, a trough of 15 to 20 is always going to get you at a 400. It could resolve in subsequent AUC value upwards of 1,000, 1,200, even up to 1,600.

The reason why I make note of this is, as we dig a little deeper into the data, what we find is clinically, you get AUCs above, 600, 700, 800. That’s where you really begin to see an increased risk of acute kidney injury with vancomycin relative to lower AUC values. Furthermore, there’s been a lot of nights worked on in animal systems by Dr. Mark Sheets. And really, what he has found is AUC is the driver of acute kidney injury in vancomycin in his animal’s system.

So, it’s a biological plausibility. We know vancomycin is not sedative stress on proximal renal tubular cells. That speaks to the intensity of exposure. Intensity of exposure is best captured by looking at the AUC versus the trough. While the trough ensures a minimal AUC, there’s considerable variance between the lower and the upper bound.

And the last thing I want to say is, what we have found in practice is this whole idea moving to AUC from a toxicity standpoint makes a difference. There’s a nice study published by Mike [INAUDIBLE 00:13:38]. This was an NIH-sponsored study, prospective multi center. And really, what he has shown is, over a three-year period of time shifting to AUC-guided dosing really reduced AKI rates by 50%.

And this is a study coming out of the [INAUDIBLE 00:13:52]. This is a pre and post-analysis and I really like this. Really, what they did, their institution, moving to AUC monitoring, they were all able to reduce acute kidney injury relying on the vancomycin consensus guideline definition of the 0.5 or 50% increase in [INAUDIBLE 00:14:09] from baseline, from about 10% to 5%.

And really, what you see in the AUC-guided group versus the trough-only monitoring group, it’s all separation of the curves over time, largely occurring four or five days out consistent with most non-immunological reactions. And if you look at the hood in this study and look at the AUC-guided group and the trough-guided group, what you see is just less intensive exposures in the AUC-guided group. Troughs, the median trough was 12. The median AUC was 474, and the interquartile range was 360 to 611.

In contrast, you look at that trough group, the median trough wasn’t even 15, it was 14.2. What they did is they put most patients in that AKI AUC sweet spot above 700 with about 25% of people having AUC above 880. Again, the whole reason that always comes up is, we do see that from a biologic standpoint, AUC to be the driver. It appears to be when you get above 600, 700, 800, [INAUDIBLE 00:15:17] increased events…

And furthermore, we have some prospective validation that shows if you shift to AUC-guided dosing, you result in lower AKI. The thing that always comes up is, « You know, in a lot of my patients, vancomycin-associated AKI is mild. And usually, it resolves within 1 to 2 weeks of discontinuation. So, we had this idea of, even if it does occur, these patients are going to resolve.

The one thing you need to realize, when your serum creatinine increases by 0.5%, there’s already considerable reductions in the GFR. And in the case of vancomycin, there’s already considerable damage [INAUDIBLE 00:15:55] fronts. And what you’ll find is when you have a 0.5 increase, you have a 50% reduction of GFR needs to occur before you see that subsequent bump. So, even though it does respond, there’s a lot of acute kidney injury there.

What we find is, even patients who have mild cases of vancomycin associated to acute kidney injury had increased mortality, increased attributal hospital stay, and increased healthcare resource utilization. We recently did a study, and what we found is even among patients with skin infections, what we found is a 5-day increase length of stay among individuals who were largely similar baseline with the exception of some having AKI and those who have not.

If you take a step back and looking at the larger literature, the deleterious outcomes associated with acute kidney injury are well described within the kidney literature. And the other thing we’re beginning to say is, it’s just not a short term negative outcome to worry about, but there’s long term consequences. And data suggests that patients with acute kidney injury, regardless of the cost, is often accompanied by remote organ dysfunction or organ cross-talk, which increases a patient’s susceptibility to a number of conditions over time, whether it be cardiovascular events, infections due to immunosuppressions, and changing in their metabolic system.

Again, it’s not just we have this mild event, serum creatinine goes up, it goes down, and the patient has normalization. Rather, there’s this whole supply of events that occur in other organ systems and really sets the patients up for bad outcomes over time. Again, when I think about the reasons for shifting to AUC, a lot of it is driven by minimizing acute kidney injury. And we’re beginning to see that acute kidney injury really increase above an AUC of 600.

So, that only defines the upper bound and the therapeutic range that we came forward with within the guidelines. The next question is: Well, what’s the lower range? If you told me to go, « I can’t go above 600?. » How low can I go? Historically, we’ve always relied on this thought that the pharmacodynamic target for Vancomycin is an AUC/MIC ratio in excess of 400 for patients with serious MRSA infections.

This is largely derived in animal studies done in 1987 by Steve [INAUDIBLE 00:18:15], and we have some clinical validation as well. We’re looking at patients with staph aureus pneumonia that really supported this target. The one thing you recognize in a lot of these clinical evaluations is they basically use the simple formula based on total daily dose and estimated renal function. This is the way they derived the vancomycin clearance to estimate the AUC.

Now, we’re all clinicians on the phone. We all recognize that there’s considerable interpatient variability. And it’s nearly impossible to generate a valid estimate of vancomycin exposure variables in a given individual based on GFR filtration formulas alone and do it in a wide interpatient exposure variability. The one thing I always say is: ICU, the more different physiologic states in a given population, you can have interpatient variability and exposure profiles upwards to 50, 60, 70%.

This is the whole reason why we do TDM among these patient populations. So the question is: Well, what’s the lower bound? This was a study I was involved in. It was the Prospective Observational Evaluation of the Association between Day 2 Vancomycin Exposure Profiles and Failure among Adult, Hospitalized Patients with MRSA Bloodstream Infections or simply PROVIDE for short.

This was a prospective study, multi-center, across the United States among individuals who were confirmed bloodstream infections. Relying on some previous identified AUC/MIC targets using a Bayesian method to estimate the Vancomycin exposure profile with limited blood concentration data, this was a method that [INAUDIBLE 00:19:58] validated as an acceptable way to estimate the AUC, rather than collecting 5 or 6 samples.

We looked at the relationship between Day 2 exposures and failure, and we power to study, included 250 valuable patients have sufficient power and detect a 15% to 20% difference and failure rates between a dichotomous day 2 AUC/MIC exposure variables. Just to jump to the chase, give that a little bit short a time, we have AUC/MIC index at the [INAUDIBLE 00:20:29] delusion as well as the etest. Again, these targets of 650 and greater than 320 were identified in the previous publication.

And interesting, what we found is, looking at the failure rates, they’re actually more pronounced. And those who would exceed these critical dates who AUC/MIC thresholds. Now, the multivaried analysis, even though the point estimate suggested that higher AUC/MIC ratios [INAUDIBLE 00:20:59] that was not the case, as 95% confidence found spanned zero in both risk difference analysis.

What we did find is acute kidney injury was more pronounced among individuals above these Day 2 AUC/MIC thresholds. Now, as we know, when we think about acute kidney injury, MIC is not informative. It’s a drug exposure that’s the biologic driver of acute kidney injury. I’m thinking about vancomycin. So, what we did is we did a desirability of outcome ranking analysis. And really, what a DOOR analysis does is it considers all the potential outcomes possible in a given individual that’s of interest in [INAUDIBLE 00:21:41].

And what we did for our analysis here, we had on the one extreme death, on the other extreme, you had treatment success with no AKI, and any of the intermingling of the two in-between. So, what you want to see in these DOOR analysis, you want to see a lot of purple. A lot of purple means survival with treatment success and no AKI. And really, what we’ve found is looking at quintiles of exposure, and the reason why we looked at quintiles is – we have 250 patients. We want to break into different buckets of patients where we had a meaningful sample within each bucket to make some inferences.

We saw no real differences in success when you factored in clinical success with acute kidney injury. But what you found in the lower two quintiles, we saw maximal rates of treatment success, no death without acute kidney injury. The question is: What are those quintiles? One was 94 to 387, and the second was 392 to 515. I will tell you that lower quintile, even though at a lower bound, it’s 94, most individual were 350 or higher.

People always say, it’s how low can we go. We may be able to go below 400, I cannot confirmly say that because of our current dosing, contemporary dosing, most patients are going to be above 400. That was the whole essence of doing drops of 15 to 20. These data suggest we might want to go lower, but there’s inclusive evidence to suggest that. So for now, based on this analysis and some other, that’s how we came up with that lower bound of 400.

Again, the other thing I always ask is: Why do we drop the MIC? The MIC is less important for several reasons. If you look at large-scale surveillance studies, most vancomycin MICs [INAUDIBLE 00:23:31] is .5 or 1.

In the PROVIDE study, over 95% of the isolates were 1 or less. There’s inherent [INAUDIBLE 00:23:42] to MIC measurement what a range of accuracy of plus or minus 1, 2 log dilutions. It’s acceptable to call it a 1, it’s acceptable to call it .25. MICs are not typically available within the first 72 hours of the index culture, so we do have to do [INAUDIBLE 00:24:01] dosing even if we know it’s MRSA, and there’s a high degree of variability between MIC testing methods used in institutions relative to the broth microdilution MIC method.

So the question becomes is: We do want to shift to AUC-guided dosing. How do we get there? Ethan will talk a little bit more about the Bayesian method, but we really have one or two ways. We could do a Bayesian approach or we could reconfigure our first order of patients that were accustomed to using the vancomycin to determine what the AUC is and then to do optimized dose.

A few things with Bayesian: This is a software program, and why it seems somewhat complex… It’s rather consistent what we’ve typically done with vancomycin is aminoglycocides in the past. If your structural model for vancomycin in those cases is a 2-compartment model [INAUDIBLE 00:24:56] parametized. You have your parameter estimates for buy-ins an clearances and their associated exposure. And based on your prior, you’re going to get some empiric dosing. Over time, what you’re going to enter in is dosing and collect the PK data.

And then as part of this, you have your targeted PK exposures. Starting out, what you’re going to do is it’s going to say, based on your height, age, weight, depending on what your PK model is, it’s going to give out some empiric dosing. And as you gain knowledge on that patient through dosing and subsequent levels round-up dosing, you’re going to get individualized PK estimates which is going to revise the PK estimates depending on what your target exposure is. In the case of vancomycin, it’s an AUC of 400 to 600. It will give you revised dosing [INAUDIBLE 0:25:46] to get within that range.

Again, the process is very familiar to what we’re used to doing. So, a few things about this: Again, we’ll provide innovative treatment schemes from loading dosing regimen. This is something we always struggle with in practice, particularly with the trough monitoring approach… Or saying give a loading dose, give another dose here, and after a certain period of time, they’ll give you a maintenance regimen.

Concentrations do not need to be collected at steady state, which is a real advantage. We think about vancomycin, as we all know, the importance of early, appropriate therapy. The battle is won or lost from an efficacy standpoint within the first 24 to 48 hours. You can include covariates as part of your Bayesian prior population PK model. And again, you had the ability to individualize which prior to pick, depending on what patient population you planted this vancomycin.

In terms of sample collecting, two is better than one. There is some data. I suggest one is okay, but that is typically in your more stable patient populations. The other way, which I know a lot of institutions is just doing an equation-based approach. What you do is you get a post-distributional peak and a trough. And what you do is, you calculate the KE, get back, extrapolate the peak, peak minus trough, divide it by the KE, we’ll give you an AUC productive and dosing interval.

A very nice thing about this: This can be programmed into the electronic medical records system, but please be aware it’s preferred that PK data on the same dosing interval, it only provides a snapshot of that AUC sampling intervals. So, if it’s dosed every [INAUDIBLE 00:27:23] three times and you get unreliable estimates and not steady-state conditions. You can use it there but it’s not preferred.

But in someone who’s in the fluctuating regimen, it’s really hard to use this. There’s really no easy way to say someone’s going from QA to Q12 in the past day or two to figure out what’s that AUC for a given 24 hours. Here’s just an example. I know I’m getting a little bit short on time here, but this is an example. You have your peak. You have your AUC to calculate your KE. You backextrapolate to the theoretical peak concentration at the start of infusion, peak minus trough divided by AUC, divided by the KE gives you the AUC for that dosing interval.

So, the question becomes: Which one of these methods is preferred? This is a nice study by [INAUDIBLE 00:28:14] which I had the opportunity to be involved in. And really, what we had is we had a large dataset of patients who receive vancomycin. We have intensive PK, and what we did for every patient is calculate at their true AUC with all the PK. We’ve looked at three PK data sets, and what we’ve found is, using a Bayesian method or an equation-based approach, both provide you reliable estimates of the [INAUDIBLE 00:28:39] standard AUC with a full complement of PK samples.

The one thing I would say is, the reason why two is better than one: This is just our experience in looking at patients with obesity. Just looking after trough data, which is on the right here, you get a reasonable ratio of the gold standard versus the [INAUDIBLE 00:29:01] AUC, but look at the variance there. In contrast, when you have two samples, you really tighten it up. And depending on the approach, here, we just use different priors.

You might under or overestimate slightly. That’s why we say when we want to target an AUC, shoot for 525 because depending on what program, and use, and what model, what prior, there’s probably a little bit of variance. Some [INAUDIBLE 00:29:26] but from a clinical standpoint, they’re all acceptable.

Just in conclusion, despite widespread integration into clinical practice, the clinical benefits, maintaining higher trough values have not been described. No improvements in efficacy with increases in AKI. Based on what we know, again, this is our current recommendations. Again, researchers [INAUDIBLE 00:29:49] process. Best evidence tell us we should move to AUC-guided dosing largely to reduce the occurrence of AKI, looking at the consequences of AKI, even if it’s mild in nature. We have a few different approaches: Bayesian, software, first-order PK calculators, to calculate the AUC with limited PK sampling. And I think stewardship teams really play a critical role in implementation assessment of Vancomycin AUC-guided [INAUDIBLE 00:30:15]. With that, I’ll stop. I’ll pass it over to Ethan and hope he’ll give you some practical experience with what he has to add with the AUC-guided dosing approach.

Dr. Ethan Smith: Thanks, Tom. Can everyone hear me on the line? For the latter half of this presentation, we’ll review some of the strategies for successfully implementing an AUC-guided approach to vancomycin dosing. And I did want to note that based on the initial poll question, it appears most folks are just getting started in this journey. So, hopefully, this will be a helpful tool kit in your journey.

Before we dive into AUC based dosing implementation strategies, let’s set the stage with a quick patient case. Here, we have a 32-year-old, otherwise healthy male who presents to the ED with a three day history of what appears to be a lower extremity soft tissue infection. In the ED, he has sepsis physiology. So they draw some blood cultures, start him on vanc and ceftriaxone, and admit him with medicine.

Shortly after admission, you get a rapid diagnostic page from your micro lab about his blood cultures resulting with MRSA. Your hospital protocol for serious MRSA infections prescribed to tropical of 15 to 20, and clinically, he’s improved. But inappropriately drawn vancomycin trough results at 13.6. At this point, what should be done?

At this point, I’d invited everybody to answer the poll question on the screen. Do we want to continue the 1 gram Q8 regimen that he’s on? Do we want to increase the dose to 1,250? Do we want more information? And as a stewardship pharmacist, I’ll tell you, answer D is not correct. We don’t switch to that, though. It looks like we have some people that are saying, about half of us saying we want more information, and then the remainder of the responses are to either increase the dose or continue the 1 gram Q8 hour regimen.

I’m sure there can be several arguments to be made here based on varying hospital protocols and dosing strategies, but from my perspective, I would want to know more information before making a final decision. The trough may not be at our prescribed goal, but we also saw Tom present some data on the poor correlation between troughs in AUCs, and the patient is clinically improved so let’s see where we stand.

In this instance, we see that on the current regimen, the patient’s AUC was already therapeutic at about 582. Had we pushed the dose to get a trough of 15 to 20, in this case 17, the corresponding AUC would have exceeded 700. And again, based on some data that we saw Tom present, we see that toxicity starts to increase when we push our troughs past 15 or our AUCs above 700 to 800.

A dose increase here would be unlikely to increase effectiveness of that drug but would put the patient at an increased risk for toxicity, and that’s potentially
by several-fold. I don’t want to spend too much time on this slide here but I do think that this is an important communication tool when implementing an AUC- based dosing strategy. This tool was originally developed for business and sales, but it’s effective here as well. When you’re trying to sell someone a product, you have to provide a succinct reason for why someone should replace their current product with yours. Here, we have to break a highly engrained reflex that serious MRSA infection equals the vancomycin trough of 15 to 20.

If you don’t provide the right insight path for why trough-based dosing approaches are flawed, implementing a new AUC-based dosing strategy is going to be difficult. So if we fill in our AUC-based dosing pyramid, we start here at the bottom. We saw that troughs were a poor surrogate for the AUC. So, the insight to be had here is that troughs are therefore a suboptimal dosing target. That we can easily and reliably calculate AUCs in clinical practice, particularly with Bayesian dosing strategies, this should be the dosing metric that we strive for because the current trough-based dosing strategies are outdated.

Now, this is just one example and one piece of the puzzle. But remember, this is probably one of the biggest paradigm shifts in drug dosing and monitoring to come about in the last decade or so. Accordingly, we want to make sure that our communication strategies are effective in expressing the need and rationale for making the switch. The next several slides are arranged as a stepwise series of questions we asked ourselves as we navigated the waters of switching from trough-based to Bayesian-based vancomycin dosing at Cedars.

Accompanying some of the questions, I’ve included key resources or pieces of data to support all of you in answering these questions at your respective institutions. Not all of these questions may apply or there may be additional questions to ask, but I believe these questions should provide a solid starting foundation.

Step one is of course to identify the scope of the problem. Are you currently using trough-based or AUC-based processes? Some may be going right from trough to AUC strategies, whereas others may be moving from first-order AUC calculations using Excel or some other tool over to Bayesian. Also, who is currently performing the dosing? This is who you’re going to be targeting for more in-depth education.

I think the biggest single barrier, I would anticipate folks will encounter, again, based on our experience at Cedars-Sinai, is breaking that 15 to 20 trough habit. A lot of the assumptions and strategies that we use with first-order trough-based monitoring may not apply to Bayesian-dosed vancomycin dosing, so how you do this will vary but ultimately will depend on the resources available to you and your clinician’s underlying knowledge, experience, and leadership.

Once we’ve defined the scope of our problem, we then move to determining how we’re going to calculate the AUC. Do we want to use first-order two-level PK like Tom talked about, equation-based dosing using an Excel calculator, or do we want to use a Bayesian-based platform? If we go the equation-based route using Excel, who has the expertise to program and manage this?

If we want to follow the new guidelines to the letter, our preference would then be to use Bayesian-based methods. In addition to some of the advantages of Bayesian-based methods that Tom had highlighted previously, you also have the advantage of more advanced modeling potential, to have EHR integration for improved workflows, and you really get some more flexibility around your level timing.

You have to consider which program will be most cost-effective based on what models you’re interested in. Some Bayesian-based platforms will offer different models or how many users you may have. For example, if you’re rolling the software out to all of your clinicians, or are you just rolling this out to select group. For example, if your institution has a pharmacy-dosing protocol, maybe you need a smaller group of users such as just your pharmacists. And really, this is one of the key tables that we presented to our leadership in making a decision to Bayesian or not to Bayesian, if you will.

So if we further compare the methods of first-order pharmacokinetic AUC calculations, we see some additional advantages to Bayesian-based methods. Whereas our first order methods are fairly static, we have to extrapolate over various dosing intervals, or if we’re changing dosing intervals, that becomes problematic. Bayesian methods are adaptive and predictive. Rather than requiring two levels to be drawn preferably at steady-state like with the first-order equations, Bayesian-based methods can be drawn prior to steady-state within the first 24 to 48 hours.

There are some considerations around using 1 or 2 levels, which we’ll discuss shortly. And with Bayesian methods, what matters most is the accurate recording of the level draw times and the dose administration times. As long as those deltas between those two variables can be accurately calculated, that level is considered valid from a Bayesian dosing standpoint. So, if a peak is drawn late or a trough is drawn early, that’s okay. Our Bayesian methods can adapt, whereas first-order equations require those precisely-timed peaks and troughs.

This, with Bayesian dosing, really allows you to have some more flexibility in your level of timing and schedule with other labs such as morning labs, rather than requiring a separate stick. Now that we’ve made a decision to proceed with Bayesian-based dosing methods, how do we justify the cost of the software? The major justification for the cost of this Bayesian platform is going to come from the cost avoidance associated with excess vancomycin toxicity when dosed for a trough of 15 to 20.

[INAUDIBLE 00:40:16] colleagues from Detroit Medical Center published their experience in switching from trough-based to AUC-based dosing. Post-implementation, they observed a 50% relative reduction in the rate of nephrotoxicity and patients receiving vancomycin. This is very much a back of the napkin calculation, but if we correlate this reduction in toxicity, with the excess cost of in-hospital acute kidney injury, we see substantial cost avoidance.

In this hypothetical medical center, which provides 1,500 courses of vancomycin per year, we see approximately 45 cases of nephrotoxicity avoided, resulting in excess of $600,000 in cost avoidance after factoring in the cost of Bayesian software. Even if this calculation were a twofold, over-estimate, substantial cost avoidance still exists.

So, now that we’ve decided on a Bayesian-dosing platform, we have a solid framework for justifying the costs, what do we need to think about next? Well, do we have a clinical pharmacy expert who can coordinate these efforts? And is there a physician champion we can partner with? Do we need any committee approvals? Probably at a minimum, your antimicrobial stewardship or pharmacy and therapeutics committees. But are there any IT-related committees which need to think about here?

I know our IT folks, one of their major requests for us was using a single sign-on process where our hospital credentials could be used to log into our Bayesian dosing platform. That was something that we had to review with EIS. Any contracts will also likely require review from our legal teams. And once we have an idea of who needs to approve these metrics or these recommendations, we need to think about how we want to package that information.

The cost avoidance data presented earlier will likely be a strong selling point. But in structuring these communications, think back to our insight pyramid. Once the initial planning phases have successfully been completed and we have our key stakeholder buy-in, we need to think about revising our dosing policies. Do we need to have any exclusion for AUC-based dosing?

I have some key groups that we might want to exclude listed on this slide, though this may vary depending on the AUC methods that are being employed. In your dosing protocols, providing some guidance in organism and MICs will be helpful, as Tom discussed. I think this will likely be a frequently asked question based on our experience.

Remember, when referring to AUC, we’re really talking about that AUC/MIC ratio. But seeing most staph aureus isolates have an MIC of one or less, that denominator doesn’t really come into play for the most part. If the NIC ends up being less than one, as Tom discussed, don’t adjust your AUC goal. There’s just not any evidence for doing that yet. However, if it’s above one, you may want to provide some guidance on considering alternative therapy, particularly in that patient that’s failing to respond to vancomycin.

But the big question is, what about non-staph aureus isolates? Most of our evidence for AUC to MIC dosing for vanco comes from staph aureus bacteremia. That said, there are some preliminary data suggesting these AUC targets for efficacy are similar in enterococci, plus staph aureus is one of the most virulent organisms we encounter.

So if our AUC goal of 400 to 600 is appropriate for these severe staph aureus bacteremias, it would likely be appropriate for other less virulent gram-positive organisms as well. But again, there’s limited data, so this is just something that you have to make a decision at your institution risk versus benefit. But remember, troughs of 15 to 20 have always been a surrogate for exceeding that AUC minimum of 400 anyway, so take that with a grain of salt.

Additional dosing policy updates will likely include guidance on choosing the right model within your Bayesian platform, particularly if you purchase multiple. For example, maybe you purchase standard model, a complex or critically ill model, and an obese model. What patient populations do you apply these to? You want to have some standard processes in place so that the dosing is consistent across clinicians. As you heard, Tom discussed the updated guidelines recommend 2-level PK using Bayesian methods. But can you broadly implement this for all patients at your institution?

I think for many, including where I’ve practiced, two levels for all patients certainly can be challenging. What if patients go home on vancomycin? How do we continue their therapy? It’s another question we asked. Hopefully, for some of those that are on this webinar, there’s going to be access to Bayesian dosing on the outpatient side. But if not, I generally recommend that if the patient has a therapeutic AUC at a given trough value, that same trough value will be targeted on the outpatient side, though it does get tricky if patients renal functions or dosing intervals change. So just having some guidance there is helpful and is a point of education for your clinicians that are dosing.

We raised the question on the previous slide earlier in the presentation: Is there any data for supporting single-level sampling with Bayesian dosing methods? Turner and colleagues asked the same question and compared Bayesian methods to a reference AUC using richly sampled PK data. Overall, they found that the trough-only sampling was quite accurate compared to the reference AUCs. So, in practice, if we’re aiming for the middle of our AUC range, so approximately around 500, then the slight variation here within that single level of sampling would probably have minimal effect on appropriate target attainment.

And a single Bayesian or single-level Bayesian AUC calculation is more accurate representation of overall vancomycin exposure compared to a single-trough value in the 15 to 20 range. That said, the two-level sampling did improve the accuracy of the Bayesian calculations in this study. So as Tom mentioned, we would still want to perform two-level sampling: in-patients with either severe infections or critical illness, or those with potentially dynamic renal function or other factors that may further alter vancomycin PK.

We’re getting into the home stretch. It’s time to pick a practical go live date and assign various roles and responsibilities for the implementation process here, develop some training procedures, and I would recommend using a case-based competency assessment to ensure that those using the software can demonstrate the skills necessary to use the software effectively prior to dosing an actual patient.

If we have a per-pharmacy dosing protocol at our institutions, we’ll also want to be sure and arrange some abbreviated provider-oriented education on the change in our dosing procedures for vancomycin for those that are most likely to order the drug. And I would also encourage using some real-time education in the form of a comment that accompanies vancomycin results in your EHR. Just as a reminder that levels may be different than expected, i.e. not in that 15 to 20 range because it’s either a peak, or a random level, or some other value that’s being used to calculate an AUC.

If you’re coming from an institution, either large or small, there’s going to be a lot of moving parts in this transition. Your Bayesian dosing champion or expert may not be able to address all questions or all clinical areas of your department at once. This is really where a super user process might come into play. I think those of you on the call that are in leadership roles probably know who your high-performing staff are, so definitely seek them out and use them to your advantage. Choose providers from different areas or shifts, so that there’s more likely to be at least a few super users working at any one time.

Those super users should receive more intensive training to answer more nuanced questions and we should really encouraged those super users to be the frontline experts. Then that Bayesian dosing champion or coordinator can focus on more complicated questions or issues as they arise. And also, consider scheduling huddles with these super users maybe on a bi-weekly, or daily, or however often basis. And those implementation coordinators would join that meeting to ensure that all of the issues are addressed in real time.

I think importantly, we don’t want to forget these super users when it comes time to yearly performance reviews, as well. We realized there were a lot of moving parts here. How do we keep it all together? This is really where our implementation checklist, or our roadmap for our transition, comes into play. We want to try and break down larger tasks into smaller goals and include clearly-defined roles and responsibilities for all those involved. We want to ensure that our deadlines are appropriate, and realistic, and are set so that they have ample time to consider all of the steps that need to occur so that you can meet your go live date at the designated time.

In addition to mapping out what needs to be done, the implementation checklist is also a means of holding participants accountable to ensure that there’s no inadvertent or extended delays in this process. And for us, the process of transitioning over to Bayesian dosing platform really unfolded over several months from initial concept to final implementation. We had to think about, as we get closer to our implementation date, reminding our various key stakeholders that the transition was going to occur on a certain date and represent against some of those groups that are more likely to prescribe vancomycin just so that they were aware that the transition was occurring.

Alright, so post-implementation. We spent weeks to months getting everything set up and our Bayesian dosing platform is live. We’re ready to dose patients with our Bayesian dosing software. What do we need to do? I think, again, even with all of the best training available, there’s still going to be questions that come up in clinical practice that haven’t been accounted for, or maybe there’s a slight variation of some case that was presented previously. So really, these daily vancomycin rounds that we did for the first several weeks was a great opportunity for those that were on the front lines doing the dosing to interact with our designated Bayesian coordinators, and really ask the experts, so to speak, in real time. It was a really a nice discussion for us.

One pharmacist may have had a question, but it was really a group discussion in our various satellites, and I think it is quite productive. We want to encourage our staff to submit challenging or interesting cases for review, because ultimately, you’re going to revise your vancomycin dosing protocol, but realistically, as it’s rolled out into clinical practice, you may realize that you need some changes to that protocol based on various cases that do occur. Really, for us, this also provided a unique opportunity for some recorded how-to videos which we hosted on our intranet site for staff to watch at their convenience for commonly-encountered dosing situations.

We also noticed that there were a handful of questions that we received several times, so we incorporated these into our living FAQ document for our staff to reference. On both accounts, we received positive feedback. So, the antimicrobial stewardship pharmacist, I can’t help but include a plug for metrics. You want to measure your success and likely your key stakeholders are going to want to see some data. Some potential metrics are listed on the slide, but what you choose will, ultimately, depending on what baseline data exists at your institution or what data can be pulled from electronic or automated reports.

Since you’re going to be measuring this data anyway, don’t forget it as one of your antimicrobial stewardship metrics for your next accreditation visit. And with that, I have two final comments. One, there’s no one-size-fits-all approach here. You have to do what’s best for your institution. Two, the clinical why here should be fairly easily justified based on what Tom and I have presented. The how-to is arguably the biggest challenge. So, implementing a Bayesian platform, rather, AUC-based dosing strategy is primarily an exercise in change management in order for the transition to be successful and an appropriate change management strategy is key.

So now, I’d like to turn it back over to Kristi to discuss some key resources in helping you manage this change.

Dr. Kristi Kuper: Great. Thank you, Ethan, and thank you, Tom, for two amazing presentations. So before we move onto questions, I just wanted to remind you that in the coming days, we’ll be sending a link to the recording for today’s webinar via the email address that you used to register. And in addition, you’ll receive a link to our vancomycin AUC conversion toolkit, which includes a wide range of helpful resources and best practices. And not only what you heard today, but also from hundreds of institutions that utilize DoseMeRx to help support their vancomycin AUC dosing programs.

Also, you can always reach out to us directly through the chat function on our website, which will connect you to someone immediately 24 hours a day, seven days a week. You can help answer any questions or get you connected with more information about DoseMeRx. So, we’re going to go ahead and move on into the questions. And we, not surprising when you have a large webinar, we have about 30 or so questions and we have six minutes. What we’ll do is we’ll just try and get through maybe as many as we can. And after the webinar, we’ll combine them and send out a Frequently Asked Questions document. Some of those folks who have submitted questions today can still get their questions answered.

I’m going to go ahead and start with the first question, which I think is probably best for Tom. Tom, do you have any thoughts on… When you’re using two levels, when you’re monitoring two levels, do you have a recommendation for populations [INAUDIBLE 00:54:17] hold off on getting those levels? Because a lot of cases, [INAUDIBLE 00:54:23] but we don’t know if it’s going to be continued. So, just would love to hear your perspective as well as Ethan’s on how to handle these situations.

Dr. Tom Lodise: That’s a very good question. I would also say if the audience can, I can stay on for a little bit longer after this. I’m not sure what Ethan’s other commitments are, to make sure we get some of the, all the unique questions answered or as many as we can. And this comes up all the time. Ethan, I think, did a good job of setting up the scenarios as to what to do in that empiric situation.

The one thing with the vancomycin guidelines I started out is, all the literature endorsing the AUC of 400 to 600 was really derived in patients with serious MRSA infections. Having said that, the toxicity curve of going above 600 probably applies across all patient populations. Your efficacy curve, while shift is a function of the patient population [INAUDIBLE 00:55:31] but the toxicity curve is rather constant.

It’s like anything else. If it’s someone where you’re just using it just in case and there’s a low suspicion of invasive MRSA infection, you can probably get away with not getting two levels within the first 24 to 48 hours, if there’s strong suspicion of MRSA, probably advisable. So in a situation where… All I would say is I’m actually very interested in what Ethan thinks in a situation where you’re doing it just in case, if it is recover, you are having an [INAUDIBLE 00:56:12], whatever program you pick. It’s probably best to start out with just targeting scanning the lower range.

Again, I cannot tell you how low we can go. There’s this idea that we have to get these very high trough values or AUC values is really not written in the literature. I can tell you that as long as you’re above 400, it appears to be okay for MRSA. Whether that’s true for VRE and other things, I don’t know. It’s always clinical judgment, but please recognize the reason why we do TDM is no PK model can explain more than 50% of the variants upfront. If we did, we wouldn’t have to do TDM.

Ethan, what are your thoughts on that? I just think that there’s clear situations that we should do it. There’s a lot of gray, especially a lot of empiric use. Then again, I think it depends on the institution but how are you guys approaching that at your place?

Dr. Ethan Smith: I couldn’t agree with you more, Tom. I think it’s going to be dependent on the clinical scenario that’s in front of you. But I will say using these Bayesian programs is really a paradigm shift with level monitoring. Early levels with appropriate outcomes with Bayesian dosing don’t necessarily need to be repeated at steady state. That’s the whole advantage of Bayesian dosing, of course, assuming the patient’s clinical status remains stable.

But that’s not necessarily the case If you’re doing trough-based dosing strategies, where maybe if you check a level early to make sure you’re heading in the right direction, you’re still going to repeat that level once you get the steady state. So, I think it just depends.

Dr. Tom Lodise: Yeah. The other thing I would say with the whole two level is, and I think Ethan did a good job, it’s really… If the decision is you’re warned about an invasive MRSA infection, it’s usually just with one dosing interval that you need two. And once you get a patient stabilized, then you get enough information from that first dosing interval that you can transition to trough-only monitoring. Recognize that it’s not are we dosing interval [INAUDIBLE 00:58:22] two levels? It’s just really early on we really want to get informative PK, where the second level has more value, particularly in patients where there’s considerable under patient variability in anticipated exposure profiles.

Dr. Ethan Smith: Totally agree.

Dr. Tom Lodise: We have a couple of questions that are specific maybe to AUC monitoring carrying different indications. We had a question around how comfortable you feel about using an AUC of 400 to 600 for patients with meningitis, and then also osteomyelitis. If you could just share your clinical experience on those two topics, that’d be great.

Dr. Tom Lodise: I’ll start out. I would be curious what Ethan is doing in his institution. From an evidence standpoint, clearly, we have less data there. For meningitis, the infections always fall at the sight of viable tissues and vancomycin’s protein finding it. If there’s good vasculature, you’re going to deliver a lot of drug there. Perhaps that’s a situation. You might want to target the higher end of the dose of the AUC range. Going above that as a risk-for-benefits situation. But if you do want to go above it during the initial phases, I would always make sure once you get a patient stabilized, then pull back because it’s unclear to me: Do you need therapeutic concentrations of drugs for the entire treatment course, or is it just really critical to get that initial burden down?

Same thing with effect of meningitis. When you have inflamed meninges, drug gets in there pretty good. And actually, vancomycin does get into the CSF more so than other drugs. So, same type of thing. If you want to target [INAUDIBLE 01:00:04] from the first day or two, that’s clinical judgment. And the reason why I made such a big deal about vancomycin AKI being a delayed effect, it’s usually after 4 to 5 days of therapy. The higher intensity, the prolonged exposure… It’s really when you see AKI.

If you go higher earlier and back off, there’s still an AKI risk. But every clinical situation really emerge judgment. Guidelines are just recommendations. They provide you with some guard rails, and not every patient fits into the guidelines. Ethan, what do you think about my response? Do you agree or am I just being too academic?

Dr. Ethan Smith: No, I think that’s pretty fair, too. We also have to think about the microbiology of CNS infections, especially community-acquired ones. We’re still using vancomycin. Those MICs that we think about for those organisms, such as strep pneumo, are going to be different than they are for staph aureus. And if we really just go back to the recommendations and the IDSA meningitis guidelines, that trough of 15 to 20 is just referenced back to the 2009 vanco guidelines, which was really expert opinion as a surrogate for an AUC above 400.

We’re just in a circle here where it all comes back to AUC anyway.

Dr. Kristi Kuper: I have one quick question for Tom, and then one for Ethan, and then we’ll conclude today’s webinar. So Tom, can you share your thoughts about using Bayesian dosing to tackle AUC and patients with acute kidney injury or fluctuating renal function?

Dr. Tom Lodise: Yes. The one thing is, when I’m choosing a prior for your Bayesian estimator, I always like the one that has [INAUDIBLE 01:01:48] or some measure of GFR. So, a primer ties [INAUDIBLE 01:01:53]. That gives you the ability to look at time-varying covariance. And with that in someone who is septic, and you take into account where their estimated GFR is related to their vancomycin clearance, it’ll provide a more precise estimate. But clearly, incorporation [INAUDIBLE 01:02:10] GFR to vanco clearance is really as good as serum creatinine which would be [INAUDIBLE 00:02:19] good biomarker of underlying GFR. That’s such a dramatic reduction in GFR before serum creatinine increases.

Actually, vancomycin, in many ways, levels [INAUDIBLE 01:02:30] more indicative of renal function versus any of these formulas. I would say with the next iteration of looking at vancomycin, I think this is where we look at relationship between the kinetic GFR calculator, and vancomycin clearance, and [INAUDIBLE 01:02:48] dynamically-unstable individuals. But I would say the simplest thing to do is probably do a little bit more intensive monitoring [INAUDIBLE 01:02:57] continuous infusion at the ICU because that’s proportional or lineal pharmacokinetics… It’s easier to do dose adjustments there.

But they’re problematic and it’s just an example of a patient population where you have not only interpatient variability to start but also a lot of interpatient variability as the renal function changes.

Dr. Kristi Kuper: Great, thanks Tom. Ethan, this is where it closes out. I think this is a good practical question for you: What is the most common question that you’ve got for your staff when you’re going transitioning from trough-based dosing, which they’ve done forever, to AUC-based dosing? Can you expand on how you’ve been able to address that question?

Dr. Ethan Smith: Sure. We’ve had a few commonly-asked questions, but I think the root of all of those questions really do just stem back to some of the assumptions that we make with first-order trough-based monitoring. It really can’t be applied to Bayesian-based dosing methods. One of our staff might be expecting one result, but the Bayesian platform is saying another. And that’s not necessarily because either a first-order calculator or a Bayesian-based calculator are producing inappropriate answers, they’re just deriving that answer based on different methods.

We have to, at the end of the day, ultimately realize that these tools are helping us make clinical decisions. But at the end of the day, we have to use our clinical judgment in determining what we think is the right thing to do for the patient. I think there’s any number of questions that can be asked, but the root of those really do stem back to that main issue. For us, really developing that FAQ document and those video series that we produced in terms of commonly-encountered dosing scenarios and how they differ in a Bayesian-based process versus how they may have existed in the prior trough-based first-order kinetic realm.

Dr. Kristi Kuper: Great. I think that’s a good way to end today’s webinar. We want to thank all of the participants for joining today, and a special thank you to both Ethan and Tom for a fantastic and informative presentation. This concludes today’s webinar. Thanks, and have a great day.