MIC, PK/PD Indices, and Antibiotic Breakpoints

No comments

I discussed C. auris in my last post and pointed out that, due to the relative novelty of the organism, there are no “defined” breakpoints when it comes to drug-bug combinations. While I did say that fluconazole, in many instances, was considered “resistant”, the truth is I cannot say that with 100% confidence as there are no definitions when it comes to breakpoints. So I figured this would make an interesting topic to talk about. 

Breakpoints and antibiotic susceptibility are related, with the former term being used to define susceptibility and resistance for a specific drug-bug combination. The definition of breakpoint varies, but the usual definition is as follows: MIC for any antibiotic that distinguishes the wild-type bacteria from those that are resistant (1). Another definition is the clinical breakpoint, which refers to the MIC that separates strains where there is a high likelihood of treatment success from treatment failure. These typically take into account the minimum inhibitory concentration of a certain bug drug combination, the pharmacokinetics/dynamics, animal and to a certain extent, patient data to create categories that are easily applicable to specific scenarios. In other words, multiple factors come into play when generating this magic number.

Three organizations are in charge of establishing breakpoints and categories (either Sensitive, Intermediate, or Resistant) in an attempt to standardize the values; the USDA Center for Drug Evaluation and Research, the Clinical and Laboratory Standards Institute (CLSI) and the European Committee on Antimicrobial Susceptibility Testing (EUCAST). Most labs in the US refer to the CLSI. 

Measuring MIC

I discussed this in a post a while back, but I think it is helpful to go over this prior to proceeding. 

Most of the drug-bug susceptibility testing occurs in vitro, with two methods being used: dilution and diffusion (2). Dilution refers to testing a cultured organism with different, serial dilutions, of an antibiotic in a two-fold dilution scheme (100, 50, 25, 12.5, etc) with a standardized concentration. The organism is defined as susceptible to an antibiotic when the MIC value is at or below the concentration achieved with the usual recommended dose of the antibiotic. Of course, inoculating bacteria with each type of dilution would take a while, so now microdilutions are used, an example of which is below:

Agar diffusion method, on the other hand, uses an agar plate swabbed with a standardized concentration of the test organism and then paper disks containing a defined antibiotic concentration are placed on the bacteria. The diameter of the zone of inhibited growth around the disc is measured:

Keep in mind this measurement is made in vitro, meaning this number has little clinical applicability without context. Which is why I tried to refrain from calling C. auris “resistant” to fluconazole or “susceptible” to anything else since that number was obtained in isolation, outside of clinical data or pharmacokinetic or pharmacodynamic data. In other words, it is a number that does not take into account the actual patient.

Pharmacokinetic and Pharmacodynamics

To review, pharmacokinetics deals with how the body handles the drug (absorption, distribution, eleminatino) while pharmacodynamics deals with how the drug acts on the organism  (in this case, the bactericidal activity and toxic effects, 3). The relevant calculations are not important for this conversation, but the concept is. Pharmacodynamic properties can be explored in vitro, with fixed drug concentrations being exposed to a bacterial species and measuring viable counts (4). Of course, in vitro data doesn’t really help much in a living being with varying volumes of distribution and altered mechanism for elimination (i.e. that ICU who is third-spacing due to renal failure). One way to approach this while using PK and PD data is to use PK/PD indices (3). These allow for the definition of a pharmacodynamic target, which is the minimum PK/PD index that is aimed for when treating patients and is typically based on pre-clinical and clinical drug-bug exposure response relationships. The pharmacodynamic target (PTD) is a value that maximises the changes of success.

PK/PD indices are things you have probably heard or read about but didn’t understand. You can think of these as predictors of efficacy, all related to MIC. Animal studies have demonstrated that different classes of antibiotics have different PK/PD indices (4). These are: 

  1. Time above MIC 
  2. Ratio of AUC/MIC over 24h
  3. Peak concentration/MIC

In other words, bactericidal activity can be thought of as

  • Concentration-dependent killing: more drug = more killing. This is why aminoglycosides are dosed with the trough in mind and once a day. 
  • Minimal concentration-dependent killing: essentially, get the drug to a good enough level and keep it there. This comes into play with beta-lactams, where you want to keep the concentration above the MIC

As an example, see here that, increasing the concentration of ticarcillin does not lead to better efficacy and actually plateaus (3):

The above pharmacodynamic markers suggest that if you manipulate the frequency of dosing and the load, and that it varies depending on the type of antibiotic you use. One of the pharmacodynamic measures you will often see used is the area under the curve for concentration vs time aka the AUC. While I know some people are getting algebra flashbacks here, this is a neat way to get a “feel” of how much drug was given over a 24hr period, given the variability when it comes to distribution and elimination. This can be used with other PD parameters such as the MIC and is presented as a ratio, AUC/MIC. Thus, the higher the ratio (the higher the AUC, or more drugs given over a 24hr period OR the lower the MIC) the more efficacious the drug will be. Of course, this is related to the peak above the MIC, with higher peaks being related to more efficacious killing. You can see how the MIC plays a role here; the smaller the number, the higher the ratio and thus, better killing. It is not surprising that peak/MIC and AUC/MIC are related. It should be noted, however, that only the “free” drug in serum plays a role in the bactericidal activity, which varies from drug to drug. 

In another example, sing gatifloxacin and Salmonella enterica, the efficacy of killing as correlated when using AUC/MIC and Concentration above MIC, but not time above MIC (5):

Similarly, the  rate of success in oropharyngeal candidiasis was associated with the AUC (again, the concentration of a drug over a period of time, usually 24hrs) and the MIC (6):

As can be seen from this table, the clinical PK-PD target  changes depending on the type of antibiotic:

PK/PD indices are related to both the concentration-time of exposure and MIC of the organism to that specific antibiotic, allowing you to obtain a PDT value that achieves clinical success in 90% of cases. The target is obtained usually from in-vitro and animal studies. So the breakpoint should take how the body handles the drug, the effect of the drug in the body, and how the antibiotic best achieves its effect in the body, and all of that should translate into a value which is the MIC. You should notice, however, that those indices are obtained from static situations. In reality, the exposure in different patients vary widely, and depends on the volume of distribution and rate of clearance. Those changes change the PDT and thus, the breakpoint

Defining Breakpoints

EUCAST defines susceptible organisms as those who, on a level of antimicrobial activity, are associated with a high likelihood with therapeutic success. There are a few things to notice here, first is the phrasing of “likelihood of therapeutic success.” We will revisit this in a bit. The other is the level of “activity” suggesting that the antibiotic in and of itself is not solely responsible for whether or not an organism is susceptible or resistant. The concentration-time profile plays a role in the success of a drug-bug combination and this, along with the MIC, create a relationship that is associated with success (6). Given this, how does one define breakpoints? The MIC, on its own, is the antibiotic effect in vitro. Despite that, breakpoints are defined with the MIC in mind and serve as a reference point to which other parameters such as PD end points or clinical outcomes can be compared to (4). 

The measurement of breakpoints starts with obtaining a set of MIC values from a large number of strains of a single species, and plotting them on a histogram (4). In this example, over 70,000 strains of Staphylococcus aureus have their MIC against vancomycin plotted. Using a cut-off MIC of 2, you can see that well over 90% fall under this range:

This, of course, requires the observation of several thousand isolates. Ideally, these strains would be “wild-type” meaning they do not harbor any type of resistance mechanisms. From here, a cut-off value is obtained to account for roughly 90% of the isolates. Statistical analysis can be used if the range of MIC values are known to better estimate the optimal wild-type cutoff or this can be done by observation alone. One exception to this rule is the identification of resistance genes. For instance, if a Staph aureus strain harbors the mecA gene, then you would see an “R” next to penicillins, regardless of the MIC. The same applies to vanA/B for E.faecalis. 

Once the wild-type MIC value is determined (remember, this is just an in vitro number that has no clinical relevance yet), the next step is to obtain PK/PD values to predict maximum efficiency for each drug bug combination (4). This is called target attainment analysis and tends to shoot for 90% success rate. The calculations are a bit complicated, but this sample should give you an idea. Healthy volunteers were given 500mg of levofloxacin and PK/PD parameters were obtained and compared to an organism with different MICs. The MIC at which the target attainment dropped off significantly was 1:

Critical illness changes the volume of distribution and elimination mechanisms, so healthy volunteers help up to a point. The target attainment can be improved upon with clinical studies.  Here, using 750mg of IV levofloxacin and the distribution of MIC for E. cloacae and P. aeruginosa, you can see how an MIC of 1 may work for E. cloacae but not so great for pseudomonas, as it misses a not so small percentage of Pseudomonas isolates (4):

Just as MICs for a strain can vary, so can the PK/PD parameters for each individual patient. Ideally, one would obtain thousands upon thousands of data points from various different patients in different scenarios, but this is not feasible. Remember, the number you are shooting for in terms of breakpoint should allow you to have a high chance of success throughout different scenarios. How would one go about obtaining such a vast number of data points? Use a simulation named after a casino in Monaco. The Monte Carlo simulation allows you to use existing PK/PD parameters to simulate a much larger simulation assuming a sort of distribution (4, 6). This generates more “samples” and picks a target that allows you to successfully treat the vast majority of patients. Once you get these values, you can display them in 2 ways. One is as a table, where the range of MIC is plotted against the PK/PD parameter that depends on the type of antibiotic in the y-axis, resulting in the probability of target attainment i.e the probability of getting a cure for a specific dose. Again, these varies on MIC and PK/PD parameters:

In the above example, a MIC of 4 allows you to get 100% cure through many different PDTs for 1g of ceftazidime given TID. Another way is to plot the results:

Here, an MIC of 2 allows you a 50% target attainment, but that of 1 allows you nearly 100% for linezolid. Again, you can see that the chances of cure are related to a lower MIC. 

Another consideration when determining breakpoints is the probability of resistance. The relationship between exposure and resistance actually follows an inverted “U” shape (7). In one example where the effect of quinolone on pseudomonas was investigated, the size of the resistant subpopulation of bacteria was amplified at several AUC/MIC ranges up until a certain point, after which there is a decline in resistant colonies:

Changes in initial inoculum size also play a role, with higher levofloxacin concentrations required at 107 vs 106 CFU of pseudomonas:

This may play a role when it comes to high-inoculum infections such as abscesses, infective endocarditis, or thrombophlebitis and may explain the relatively high failure rates in these types of infections. Interestingly, the longer therapy goes, the more difficult it is to suppress amplification of a resistant subpopulation, with shorter regimens having a good bactericidal effect. In one study plotting garenoxacin against S. aureus, a regimen with a AUC/MIC ratio of 100 was compared with one of 280. Both regimens had efficacious killing for the first 5 days but the resistant population increased in the low AUC/MIC regimen. This may also reflect the high relapse rate in difficult to treat infections such as infective endocarditis. 

Overall, the best approach to establishing MIC breakpoints involves comparing PK/PD cutoffs, clinical cut-offs, and wild-type cut-offs, based on a table as below (4):

Of course, this is largely subjective. A review notes that breakpoints should be established as such:

  1. Assess the MIC distribution for a collection of isolates, ideally from global sources
  2. Apply PK/PD cutoff to this collection
  3. After this, establish clinical cut-offs and consider them as a validation tool for PK/PD cutoffs. 

As mentioned previously, you only get one cutoff per drug-bug combination and it does not take into account any infection site. For instance, the MIC may not play much of a role for certain antibiotics in urinary tract infections if the antibiotic is excreted in the urine as the concentration there would be much higher. At the other end, CSF penetration may affect the concentration of drug in cases of meningitis. For most of the breakpoints, these are established with “bacteremia” in mind rather than a specific type of infection. 

Utility of Clinical Data

For the most part, many of the breakpoints are established prior to the broad use of antibiotics and with wild type bacteria that are typically not resistant to a certain antibiotic. While a lot of animal and basic science data goes into determining breakpoints, many times these do not pan out once in clinical use. In a pair of retrospective cohort studies published in the same paper (8), 30 day mortality was compared when an MIC of 32-64 was used in pseudomonas for pip-tazo versus a lower MIC of <16. Primary endpoint was higher in the high MIC group when pip-tazo was used compared to the low MIC, where that cohort had similar outcomes when pip-tazo and an alternative was used:

Further, logistic regression analysis found that treatment with pip-tazo in those whose MIC was 32-64 had a significant higher likelihood of death:

Similarly, another study evaluated clinical outcomes in patients who got cefepime for gram negative bacteremia (9). Among 197 patients, 28 day mortality was higher in those whose cefepime MIC was 8 (the threshold for susceptibility per CLSI at the time):

Multivariate analysis found that 28 day mortality was higher in those whose MIC was >8 and were given cefepime (aOR 9.1, 95% CI 2.2 to 37.5). Comparing those who got another therapy, mortality remained higher in those who got cefepime (aOR 2.0, 95% CI 0.5 to 7.9) but this was not statistically significant. These examples illustrate that, a lot of times, established breakpoints may not translate well into clinical outcomes and as a result they tend to be re-evaluated once further clinical data returns. CLSI frequently updates its breakpoints and these can be accessed for free at: https://clsi.org/standards/products/free-resources/access-our-free-resources/

Disk Diffusion Breakpoints

How do you apply this to disk diffusion? You use a scattergram (4). Here, you plot the MIC against zone diameters and define an error bound where strains known to be resistant on MIC testing would show up as Sensitive on zone diameter in <1% of the cases:

In the above example, a 13mm diameter guarantees an MIC below 4, where very major errors are minimal. Minor errors are usually categorized as intermediate. Ideally, the cut-off would attempt to minimize errors where you get a “resistant” organism showing up as “sensitive” in disc diffusion. 

References:

  1. Mouton JW, Ambrose PG, Canton R, Drusano GL, Harbarth S, MacGowan A, Theuretzbacher U, Turnidge J. Conserving antibiotics for the future: new ways to use old and new drugs from a pharmacokinetic and pharmacodynamic perspective. Drug Resist Updat. 2011 Apr;14(2):107-17. doi: 10.1016/j.drup.2011.02.005. Epub 2011 Mar 26. PMID: 21440486.
  2. [edited by] John E. Bennett, Raphael Dolin, Martin J. Blaser. Mandell, Douglas, and Bennett’s Principles and Practice of Infectious Diseases. Philadelphia, PA :Elsevier/Saunders, 2015.
  3. Craig WA. Pharmacokinetic/pharmacodynamic parameters: rationale for antibacterial dosing of mice and men. Clin Infect Dis. 1998 Jan;26(1):1-10; quiz 11-2. doi: 10.1086/516284. PMID: 9455502.
  4. Turnidge J, Paterson DL. Setting and revising antibacterial susceptibility breakpoints. Clin Microbiol Rev. 2007 Jul;20(3):391-408, table of contents. doi: 10.1128/CMR.00047-06. PMID: 17630331; PMCID: PMC1932754.
  5. Ambrose PG, Bhavnani SM, Rubino CM, Louie A, Gumbo T, Forrest A, Drusano GL. Pharmacokinetics-pharmacodynamics of antimicrobial therapy: it’s not just for mice anymore. Clin Infect Dis. 2007 Jan 1;44(1):79-86. doi: 10.1086/510079. Epub 2006 Nov 27. Erratum in: Clin Infect Dis. 2007 Feb 15;44(4):624. PMID: 17143821.
  6. Mouton JW, Brown DF, Apfalter P, Cantón R, Giske CG, Ivanova M, MacGowan AP, Rodloff A, Soussy CJ, Steinbakk M, Kahlmeter G. The role of pharmacokinetics/pharmacodynamics in setting clinical MIC breakpoints: the EUCAST approach. Clin Microbiol Infect. 2012 Mar;18(3):E37-45. doi: 10.1111/j.1469-0691.2011.03752.x. Epub 2012 Jan 20. PMID: 22264314.
  7. Mouton JW, Ambrose PG, Canton R, Drusano GL, Harbarth S, MacGowan A, Theuretzbacher U, Turnidge J. Conserving antibiotics for the future: new ways to use old and new drugs from a pharmacokinetic and pharmacodynamic perspective. Drug Resist Updat. 2011 Apr;14(2):107-17. doi: 10.1016/j.drup.2011.02.005. Epub 2011 Mar 26. PMID: 21440486.
  8. Tam VH, Gamez EA, Weston JS, Gerard LN, Larocco MT, Caeiro JP, Gentry LO, Garey KW. Outcomes of bacteremia due to Pseudomonas aeruginosa with reduced susceptibility to piperacillin-tazobactam: implications on the appropriateness of the resistance breakpoint. Clin Infect Dis. 2008 Mar 15;46(6):862-7. doi: 10.1086/528712. PMID: 18279040.
  9. Bhat SV, Peleg AY, Lodise TP Jr, Shutt KA, Capitano B, Potoski BA, Paterson DL. Failure of current cefepime breakpoints to predict clinical outcomes of bacteremia caused by gram-negative organisms. Antimicrob Agents Chemother. 2007 Dec;51(12):4390-5. doi: 10.1128/AAC.01487-06. Epub 2007 Oct 15. PMID: 17938179; PMCID: PMC2168001.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s