Too much of a good thing: defining antimicrobial therapeutic targets to minimize toxicity

Abstract

Antimicrobials are a common cause of drug toxicity. Understanding the relationship between systemic antimicrobial exposure and toxicity is necessary to enable providers to take a proactive approach to prevent undesired drug effects. When an exposure threshold has been defined that predicts drug toxicity, therapeutic drug monitoring (TDM) can be performed to assure drug exposure does not exceed the defined threshold. Although some antimicrobials have well-defined dose-dependent toxicities, many other exposure-toxicity relationships have either not been well-defined or in some cases, not been evaluated at all. In this review, we examine the relationship between exposures and toxicities for antibiotic, antifungal and antiviral agents. Furthermore, we classify these relationships into four categories: known association between drug exposure and toxicity such that clinical implementation of a specific exposure threshold associated with toxicity for TDM is supported (category 1), known association between drug exposure and toxicity but the specific exposure threshold associated with toxicity is undefined (category 2), association between drug exposure and toxicity has been suggested but relationship is poorly defined (category 3), and no known association between drug exposure and toxicity (category 4). Further work to define exposure-toxicity thresholds and integrate effective TDM strategies has the potential to minimize many of the observed antimicrobials toxicities.

Introduction:

Antimicrobials are one of the most common drug classes associated with toxicity.(1–4) They are among the most frequent agents to cause drug-induced liver injury and kidney injury and serial monitoring of hepatic, kidney, and hematologic markers is necessary to identify onset and severity of drug-induced toxicity.(5, 6) This current practice of screening for toxicity is reactive, not preventive. When a toxicity is noted, the medical provider is challenged to decide whether a change should be made to dosing or selection of the antimicrobial regimen. However, an understanding as to why the toxicity occurred is typically unavailable. Most toxicities are considered dose-dependent, rather than idiopathic, suggesting that toxicity is related to how much drug is given. In actuality we know toxicity is due to individual exposure that results from a given dose, with variability in exposures across individuals primarily due to differences in the absorption, metabolism, distribution, and clearance processes that determine the drug concentrations over time in that person.., In theory, administration of personalized dosages, which account for patient characteristics that influence drug disposition, can facilitate safer use of drugs and reduce the likelihood of dose-dependent toxicities.

Therapeutic drug monitoring (TDM) is the process of quantifying drug concentrations in patient specimens, typically blood (i.e. serum or plasma), and using that information to guide dosing in an individual patient.(7) Traditionally, TDM has been implemented in a non-specific manner with the goal to ensure that drug concentrations fall within a therapeutic range. The therapeutic range is bound by minimum effective concentrations at the lower ranges and toxicity at higher concentrations. While toxicity targets are strictly driven by the drug exposure parameters (e.g. AUC, C_max, C_min, etc.), efficacy targets for antimicrobials must also account for susceptibility of the infecting organism to the drug. In the case of bacterial infections, the minimum inhibitory concentration (MIC) contributes to the efficacy target (e.g. AUC/MIC, C_max/MIC, etc.), which precludes use of a single concentration-driven target for efficacy.

Because of the complexity of defining a therapeutic window for antimicrobials, many experts advocate for use of more personalized dosing strategies, such as through target concentration intervention.(8) These strategies utilize pharmacokinetic (PK) principles to derive individualized doses that will most likely achieve specific drug targets. Such approaches account for MICs, patient condition, and various other factors that may influence the relationships between dosing, exposures, and outcomes (efficacy and toxicity) to provide individualized dosing that maximizes likelihood of efficacy and minimizes the risk of toxicity. However, even with use of more advanced and personalized dosing strategies, a specific therapeutic target must be defined.

Several stipulations must be met to justify implementation of TDM for any given antimicrobial agent (Table 1). As a result, TDM has routinely been used to guide dosing for only a handful of antimicrobials, such as the glycopeptides and aminoglycosides.(9, 10) Outside of these two classes, TDM has not typically been implemented due to the presumed wide therapeutic range for most antimicrobials, lack of a clearly defined therapeutic target, or infrequent use of a specific agent. More recently, the value of TDM for antimicrobial efficacy has been espoused.(7) Particularly in cases of critically ill patients, in patients with modified pharmacokinetics (e.g. dialysis, burns, ECMO, neonates), or in the treatment of multi-drug resistant organisms that require higher exposures to overcome resistance, attention has been given to the value of TDM to ensure sufficient drug concentrations are achieved to optimize efficacy.(11) This has led to increased adoption of TDM for agents other than aminoglycosides and vancomycin. However, quantification of antimicrobial drug exposure remains infrequently utilized to guide dosing due to potential toxicity concerns.

Table 1.

Requirements for practical use of therapeutic drug monitoring of antimicrobials

Requirement	Rationale	Stipulation(s)
Lack of predictable dose-concentration relationship	Variability in the inter-individual PK of a drug must be significant enough to preclude reliable prediction of drug concentrations based on dosing alone.	The degree of variability between patients that would be considered “enough” depends on the therapeutic index of the drug and the clinical importance of attainment of the therapeutic target.
AND
A validated drug assay exists	Need to have a reliable means to measure drug concentrations that will inform clinical dosing decisions in a time-frame that is relevant to the care of the patient.	The rapidity in which drug concentration measurements need to be available depend on the use of the drug (i.e. more rapid results are needed in a critically ill patient than in an outpatient).
AND
Drug concentration quantification reliably reflects free drug	The free fraction of drug exerts the drug effect(s). Thus, must be able to quantify free drug directly or be able to infer the free fraction of drug from the total concentration.	Variability in protein binding both across and within patients will influence the free fraction of drug. Interpretation of total drug concentration will be more difficult for drugs with a higher degree of protein binding, in which small changes in protein binding will have a larger effect on the free drug concentration.
AND AT LEAST ONE OF THE FOLLOWING
Established effective concentration(s)	Therapeutic target for effectiveness (Cmax:MIC, T>MIC, etc.) must exist for a drug-bug pair in vivo, ideally in the patient population of interest. Use of TDM may identify individuals who require dosing increases to optimize effectiveness.	MIC of organism being treated will have a significant impact in determining the effectiveness target.
Established toxic concentration(s)	Therapeutic target (concentration, AUC, etc.) for toxicity. TDM may identify individuals who require dosing decreases to minimize risk of adverse effects.	MIC of organism being treated does not play a role in determining the toxicity target, except to recognize that the effectiveness target (Cmax:MIC, T>MIC, etc.) may result in overlap with toxicity target (Cmin>x, AUC>y, etc.) when accounting for the MIC

Abbreviations: AUC, area under the curve; C_max, maximal concentration; C_min, minimum concentration; MIC, minimum inhibitory concentration; PK, pharmacokinetics; T>MIC, time above minimum inhibitory concentration; TDM, therapeutic drug monitoring.

For TDM to be effective at preventing toxicity, a clear relationship between drug exposure (i.e. concentrations) in blood and toxicity must exist.(12) TDM is only useful and necessary if the results of drug concentration measurement are actionable. From a toxicity perspective, this means that a target should be defined that, when achieved, minimizes the risk of adverse events. Traditionally, C_min targets are used because antibiotics are administered intermittently and troughs are simple to interpret. However, clinicians should distinguish when C_min is the driver of toxicity (e.g. aminoglycosides) versus when it is a surrogate of another drug exposure measure (i.e. AUC), such as with vancomycin. When C_min is used as a surrogate, the correlation between troughs and AUC may vary among subjects, limiting its effectiveness in toxicity prevention. In addition, it should be noted that many clinically available drug assays measure total drug concentrations, although only free (unbound) drug exerts toxic effects. There is often wide variability in protein binding among individuals, which can have a significant impact on the unbound fraction of drug.(13) As free drug assays become available, recognizing the differences in total versus free drug concentrations as related to toxicity will be important.

In this review, we will examine the exposure-toxicity relationships of commonly used anti-infective agents. Where a relationship between a specific drug exposure and toxicity exists, we will summarize the existing data that supports a specific toxicity target. We focus our review on antimicrobial agents that are not traditionally the focus of TDM with the goal to summarize the literature on potential toxicity targets. We also seek to highlight antimicrobial agents for which an exposure-toxicity relationship is apparent, but a target is not yet defined. Drugs that are routinely subject to TDM and for which there is an extensive body of literature surrounding exposures and toxicity (e.g. vancomycin and aminoglycosides) will not be covered.

Classification strategy:

We performed a narrative review of the therapeutic targets for prevention of antimicrobial-associated toxicities. We focused on systemic antimicrobials including antibacterial, antiviral and antifungal agents. This review was not intended to be a systematic review or meta-analysis. Because of the myriad of potential dose-dependent toxicities that have been reported among patients treated with antimicrobials, and the heterogeneity of available studies, we felt that all data could not be reasonably summarized for this work. We instead focused on drug-toxicity relationships that have been most thoroughly evaluated and could be summarized in the context of TDM, and explored drug-toxicity relationships that have not been as clearly defined. Given the breadth and number of available antimicrobial agents, we focused on representative examples from categories defined below (in Table 2). We additionally did not include glycopeptides or aminoglycosides, given the extensive existing toxicity literature and experience with TDM for these drug classes. Our goal was to characterize drug-toxicity relationships that could be used to guide TDM in the future. In an era of increasing antimicrobial resistance, defining exposure thresholds associated with toxicity is important so that aggressive dosing can be used, as needed, without fear that higher doses alone will lead to deleterious outcomes.

Table 2.

Definitions of exposure-toxicity relationships to support use of toxicity targets for TDM

Category	Definition	Drug of class (toxicity)
1	Known association between drug exposure and toxicity such that clinical implementation of a specific exposure threshold associated with toxicity for TDM is supported	Voriconazole (neurotoxicity, hepatotoxicity) Flucytosine (bone marrow suppression) Linezolid (thrombocytopenia) Polymyxins (nephrotoxicity)
2	Known association between drug exposure and toxicity but the specific exposure threshold associated with toxicity is undefined	β-lactam agents (neurotoxicity) Daptomycin (skeletal muscle myopathy)
3	Association between drug exposure and toxicity has been suggested but relationship is poorly defined	Trimethoprim-sulfamethoxazole (hematologic and renal toxicities) Amphotericin (nephrotoxicity) Ganciclovir (cytopenias) Acyclovir (neutropenia) Itraconazole (tremor) Posaconazole (pseudohyperaldosteronism)
4	No known association between drug exposure and toxicity	Fluoroquinolones (tendonopathy) Rifamycins (hepatotoxicity)

We defined key exposure-toxicity relationships based on the available literature. For each, we sought to classify the drug-toxicity associations into one of 4 categories (Table 2). These categories were developed to summarize the strength of the exposure-toxicity data in regard to clinical implementation of a specific toxicity target. Category 1 includes drugs for which there is a clearly defined exposure-toxicity relationship that adequately supports clinical use of toxicity targets for TDM. Category 2 defines dose-dependent toxicities that have been sufficiently described in the literature, but for which a specific TDM target remains undefined. These drug-toxicity relationships require further or continued research. Category 3 includes drug exposure-toxicity associations that have been reported but for which insufficient data exist to support use of TDM. Category 4 comprises drug toxicities that do not appear to be exposure-dependent based on available data.

Category 1 – Drug exposure-toxicity relationship is defined

Aminoglycosides and vancomycin have substantial data supporting the use of TDM to mitigate toxicity among their recipients. We identified four additional drug/drug classes (voriconazole, flucytosine, linezolid, polymyxins) for which there are consistent data supporting specific exposure-toxicity relationships. We believe that data support the clinical use of an upper bound to the therapeutic range for TDM for these agents.

Voriconazole and neurotoxicity

TDM is recommended for several of the broad-spectrum triazole antifungal agents (i.e. itraconazole, voriconazole, posaconazole) to optimize therapeutic efficacy and to avoid potential toxicity.(14, 15) Gastrointestinal symptoms and hepatotoxicity are among the most common toxicities associated with triazoles, and these undesired effects can result in the need to change treatments.(16) Although serum concentration targets exist for itraconazole, voriconazole, and posaconazole, a defined exposure-toxicity relationship to date has only been elucidated for voriconazole (Table 3).

Table 3.

Association between antifungal therapy and serum/plasma concentrations and toxicities

Agent	First Author (Year)	PK parameter	Results
Voriconazole	Tan (2006)(25)	Cavg	Median plasma voriconazole concentration for patients experiencing a visual adverse event (VAE) was 3.52 μg/mL, and for those not experiencing a VAE, it was 2.72 μg/mL. Significant relationship between plasma voriconazole concentration and odds of a VAE (P =011). Model predicted a 4.7% increase in the odds of a VAE for every 1 μg/mL increase in plasma voriconazole concentration.
	Pasqual (2008)(21)	Cmin	Five serious neurological adverse events in patients with voriconazole levels >5.5 μg/mL (31%), compared with none among patients with levels ≤5.5 μg/mL (0%; P = 002). OR for severe toxicity after a 2-fold increase of voriconazole levels in blood was 284 (95% CI, 0.96–84,407; P = .05).
	Jin (2016)(122)	Cmin	Meta-analysis demonstrating the incidence of hepatotoxicity was significantly increased with trough concentrations >3.0, >4.0, >5.5 and >6.0 μg/mL. The incidence of neurotoxicity was significantly increased with trough concentrations >4.0 and >5.5 μg/mL.
	Hamada (2012)(22)	Cmin	Meta-analysis increase in incidence for values greater than 4.0 μg/mL was significant (OR 2.23, 95% CI 1.12–4.46, P = 0.02) when examining neurological adverse effects.
	Ueda (2009) (26)	Cmin	Eight of 11 patients with ≥ 6 μg/mL experienced hepatotoxicity.
Flucytosine (5-FC)	Kauffman (1977)(30)	Cmax	4 patients with serum concentration ≥ 125 μg/mL of 5-FC preceding and during initial period of bone marrow suppression as compared to 14 patients without bone marrow suppression and levels < 125 μg/mL.
	Stamm (1987)(29)	Cmax	5-FC toxicity occurred in 23 of 37 patients with serum 5-FC concentrations ≥ 100 μg/mL (X2 = 8.08, p = 0.005). Hepatic injury occurred in 6 of 7 patients with 5-FC concentrations ≥ 100 μg/mL (X2 = 8.17, p = 0.004) and 12 of 20 patients with blood dyscrasias (X2 = 5.39, p = 0.02).

Voriconazole is recognized as having significant intra- and inter- patient pharmacokinetic variability seemingly unrelated to dose.(17) Of the triazoles, the relationship between higher drug serum concentrations and associated toxicity has been most clearly defined with voriconazole. The defined serum concentrations to avoid toxicity can be challenging to achieve given the complex and still incompletely understood pharmacokinetics, particularly in children.(18) The current recommended trough serum concentrations range between 1.0–6.0 μg/mL.(15) Neurotoxicity including visual and auditory hallucinations have been associated with higher trough concentrations, though the serum concentration cut off varies based on study ranging from 4.0 to >6.0 μg/mL.(19–21) Hamada et al performed a meta-analysis to exam the cutoff values for neurological adverse effects and voriconazole concentrations. The analysis included 17 studies and the increase in incidence for values greater than 4.0 μg/mL was significant (OR 2.23, 95% CI 1.12–4.46, P = 0.02) when examining neurological adverse effects.(22) In a study that included 61 patients, neurological adverse events occurred in six patients with 4 of 6 patients having a voriconazole trough concentration between 4.5–5.1 μg/mL, and one patient with a trough > 5.5 μg/mL suggesting that lowering the currently recommended targeted threshold should be considered.(23)

Voriconazole and hepatotoxicity

Mixed evidence exists between the relationship of hepatotoxicity and higher voriconazole levels and no accurate cut off value has been defined to predict liver dysfunction.(24, 25) In a study evaluating the efficacy and safety of voriconazole in the treatment of invasive aspergillosis, 6 of 22 patients with plasma concentrations of voriconazole >6 μg/mL developed liver dysfunction or liver failure.(17) A logistic regression analysis performed by Pascual et al. failed to demonstrate a significant association between voriconazole levels and hepatotoxicity; the odds ratio (OR) for hepatotoxicity of a 2-fold increase in voriconazole levels was 1.4 (95% CI, 0.7–3).(21) Tan et al retrospectively evaluated safety and PK data from 10 phase 2 or 3 voriconazole trials and found no statistically significant relationship between plasma voriconazole concentrations and alanine aminotransferase (ALT) abnormalities, whereas statistically significant, but weak, associations were identified between plasma voriconazole concentrations and aspartate aminotransferase (AST), alkaline phosphatase (ALP), and bilirubin abnormalities.(25) Ueda et al performed TDM on 34 immunocompromised patients and observed hepatotoxicity in 8 of 11 cases when voriconazole trough concentrations were ≥ 6 μg/mL.(26) Several additional studies have determined a relationship between voriconazole exposure and toxicity supporting the recommendation of TDM to achieve a therapeutic target.(19) Future studies may reveal the currently accepted threshold of exposure may need to be slightly modified to further decrease risk of adverse effects while assuring adequate exposure for efficacy.

Flucytosine (5-FC) and bone marrow suppression

Flucytosine (5-FC) is concomitantly administered with other antifungals for the treatment of invasive fungal disease.(27, 28) The use of 5-FC is reserved for difficult to treat infections due to its safety profile. A decrease in renal function is commonly observed with amphotericin B and when administered with 5-FC, this renal dysfunction results in the accumulation of 5-FC and subsequent toxicity, therefore TDM is recommended. Because 5-FC is used in combination with other drugs that also have known toxicities, assumptions are made that specific 5-FC toxicities including bone marrow, hepatic and gastrointestinal toxicities.(29) A 2-hour post dose concentration exceeding 100 μg/mL has been observed more frequently in patients experiencing toxicity. Specifically, peak 5-FC serum concentrations of ≥ 125 μg/mL have been associated with reversible leukopenia.(30) Serum 5-FC levels of ≥ 100 μg/mL for two or more weeks have been observed in patients who develop hepatic injury or blood dyscrasias, though a single elevated level was not associated with toxicity recognizing suggesting that sustained elevated serum concentrations contributes to toxicity risk (Table 3).(29, 31)

Linezolid and thrombocytopenia

Linezolid is an oxazolidinone antibiotic most often utilized in the treatment of infections causes by resistant gram-positive bacteria, such as methicillin-resistant Staphylococcus aureus and vancomycin-resistant Enterococcus. It has excellent bioavailability and good tissue penetration, but its use is associated with notable toxicities, including bone marrow suppression, lactic acidosis, and peripheral neuropathy.(32) For linezolid, C_min has been shown to correlate well with AUC.(33, 34) As a result, trough sampling has been recommended as a single time-point measurement for both efficacy- and toxicity-based TDM, particularly for prolonged courses (>2 weeks) in which the risk of toxicity is increased.(32)

Of linezolid’s toxicities, exposure has been most consistently linked to development of thrombocytopenia. And, the good correlation between AUC and C_min has allowed for a C_min-toxicity threshold to be defined. Despite heterogeneity in the populations studied, a C_min >6–10 mg/L has consistently been associated with an increased risk of thrombocytopenia.(35–40) In a study of Taiwanese adults, 13 of 36 (36%) with linezolid concentrations available developed thrombocytopenia during treatment.(38) The median C_min and AUC₂₄ were both significantly higher in the patients with thrombocytopenia (C_min: 13.0 vs 7.2 mg/L, p = 0.027; AUC₂₄: 451 vs 291 mg·h/L, p = 0.025) and C_min was a significant predictor of thrombocytopenia on multivariable Cox regression.(38) In a similar study among Chinese adults, linezolid C_min was significantly higher in patients with thrombocytopenia (n = 31, median = 8.81 mg/L) compared to those without (n = 39, median = 2.88 mg/L, p <.0001).(40) And, C_min was also a significant predictor on multivariable logistic regression.(40) In a study of 84 Chinese adults treated with linezolid, 21% developed thrombocytopenia.(39) The estimated probability of thrombocytopenia was 50% for those with C_min of 7.85 mg/L and 95% when C_min was 10.55 mg/L.(39) Meanwhile, in a study of 45 Italian adults, the probability of thrombocytopenia was 50% when C_min was 6.53 mg/L or AUC₂₄ was 280.74 mg·h/L, and 95% when C_min was 9.96 mg/L or AUC₂₄ was 343.02 mg·h/L, based on logistic regression analyses.(35)

Polymyxins and nephrotoxicity

Polymyxin antibiotics, colistin (polymyxin E) and polymyxin b, have been in clinical use since the 1950s. They have regained favor recently due to increasing prevalence of multi-drug resistant gram-negative bacteria, including carbapenem-resistant Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacteriaceae. These agents exert their activity by interacting with lipopolysaccharide on the surface of gram-negative bacteria, disrupting outer membrane integrity and causing cell death.(41) While these agents tend to be reserved for the sickest of patients, a 2018 systematic review identified the prevalence of nephrotoxicity to be greater than 25% in adults treated with either of the intravenous polymyxin agents.(42) As with many drugs, dose and duration of therapy have been associated with increased toxicity for both colistin and polymyxin b.(43) And, studies have found that the polymyxins carry higher nephrotoxicity risk than other agents used for similar indications,(44, 45) most notably β-lactams, which largely limits their use to treatment of infections caused by bacteria resistant to other available agents.

Several studies have evaluated the association between serum concentrations and nephrotoxicity. Typically, colistin concentrations fluctuate little during a single dosing interval when the drug is given three times daily,(46) so both C_min and average C_SS have been used for TDM. In a study of 102 adult patients treated with intravenous colistin methanesulfonate sodium (CMS), 25.5% developed acute kidney injury (AKI) on day 7.(47) C_min at steady-state (after 3 days of treatment) was the only variable associated with day 7 AKI on multivariable logistic regression (odds ratio 4.7; 95% confidence interval [CI] 2.38–9.29), even after adjusting for cumulative dose, age, and comorbidity score. It was also associated with AKI at end of therapy (odds ratio 2.1; 95% CI 1.33–3.42) when adjusting for other significant covariates.(47) When C_min was divided into quartiles, patients with steady-state C_min >2.2 mg/L had significantly higher rates of AKI on day 7 (65.4%) and at end of therapy (84.6%).(47) In a study of 153 critically ill patients not requiring renal replacement therapy at initiation of IV colistin, the rate of AKI was significantly greater when average steady-state concentrations (C_SS,ave) were >1.88 mg/L among patients with baseline creatinine clearance was <80 mL/min/1.73 m² and was significantly greater among patients with baseline creatinine clearance was >80 mL/min/1.73 m² when average C_SS,ave > 2.25 mg/L.(48) Meanwhile, Lakota and colleagues performed a meta-analysis of 17 studies to evaluate the association between steady-state polymyxin b exposure (AUC_0–24) and nephrotoxicity.(49) They found a significant linear association between polymyxin b exposure and nephrotoxicity and defined an AUC_0–24 of 100 mg·h/L as the upper bound of the therapeutic range for this agent, which corresponds to an average C_SS of 4.2 mg/L.(49)

Based on this evidence, international consensus guidelines now recommend TDM and use of adaptive feedback control be used wherever possible for both colistin and polymyxin b.(50) For both drugs, a 24-hour AUC_SS target of 50–100 mg·h/L, corresponding to a C_SS,ave of 2–4 mg/L, is recommended.(50) It is important to note, however, what form of drug is being measured for TDM. Because CMS is a pro-drug, it is converted to colistin both in vivo following administration and in vitro followings samples collection. If samples are not handled and stored appropriately, quantification of colistin concentrations can be affected.(51) Clinicians also should be aware of when laboratories report total colistin concentrations, which reflect the sum of all CMS derivatives plus colistin, versus colistin concentrations alone. Additionally, advanced chromatographic methods allow for separate quantification of colistin A (polymyxin E1) and colistin B (polymyxin E2) within samples.(51) Similarly, polymyxin b, which is a mixture of polymyxin B1 and B2, can be quantified as total drug (polymyxin b) or polymyxin B1 and B2 concentrations.(51)

Category 2 - Drug exposure-toxicity relationship is apparent but exposure threshold is undefined

Several drugs have gained attention due the observation that toxicities are more common among recipients of higher dosages. For others, drug exposure-toxicity relationships have been purported for drugs cleared by the kidney when toxicities occur more commonly among patients with renal failure. Despite such observed associations, further investigation is required to solidify the exposure(s) at which toxicities are most likely. Below are examples of drugs for which an exposure threshold associated with toxicity likely exists but the specific threshold is not certain based on current evidence.

Daptomycin and myopathy

Daptomycin is a parenteral cyclic lipopolypeptide antibiotic with activity solely against gram-positive bacteria. It is predominantly used for treatment of infections caused by vancomycin-resistant enterococci and methicillin-resistant Staphylococcus aureus, as well as other resistant gram-positive bacteria. Daptomycin demonstrates concentration-dependent activity with AUC/MIC as the parameter best associated with clinical efficacy.(52) There is substantial inter-individual variability in plasma concentrations at approved dosages resulting in significant heterogeneity in the actual dosages prescribed in clinical practice.(52–54) Myopathy, marked by elevations in serum creatine phosphokinase (CPK) levels, is the most notable toxicity of daptomycin. a large study of 3042 patients treated with daptomycin, the incidence of myopathy (defined as rise in CPK >2.5x upper limit of normal) was 4.2% and 25 (0.8%) had rhabdomyolsis.(55) Given the large variability in exposures achieved at approved doses, many experts espouse use of high-dose daptomycin (>6 mg/kg) for treatment of serious infections, but toxicity concerns have limited routine use of such higher doses.(56, 57)

In pre-clinical studies in dogs, skeletal muscle toxicity was more closely related to dosing interval (i.e. fractionated doses versus once-daily dosing) than to C_max or AUC.(58) This has led to the use of once-daily administration of daptomycin in adults. Bhavnani et al. evaluated the relationship between daptomycin exposure and myopathy among 108 adult patients with S. aureus bacteremia.(59) Although only 6 patients met criteria for myopathy, both C_min and AUC were associated with CPK elevation on simple logistic regression analyses. When assessed via classification and regression tree (CART) analysis, a cut-point for C_min of 24.3 mg/L was most significantly associated with CPK elevation (OR 33.0; 95% CI 4.60 – 237).(59) Similarly, in another study of 38 patients treated with daptomycin, of whom 5 (13%) had elevations in CPK >2.5 times the upper limit of normal, the median C_min was significantly higher among patients with muscle toxicity (23.9 vs 9.2 mg/L, p < .05).(60) However, several other studies have failed to confirm this cut point or demonstrate a clear dose-toxicity relationship. In a case-control study involving 128 cases with myopathy matched to controls without myopathy, there were no differences in dosages administered to cases or controls.(55) Similarly, in a study of 63 adult patients treated with daptomycin who underwent routine clinical TDM, there was no association between CPK elevations and dose or C_min.(56) Meanwhile, in a retrospective study of 86 patients treated with daptomycin at dosages of 2.7 to 13.8 mg/kg at a single institution, 25% of C_min values were above 25 mg/L.(53) However, there was no association between C_min and CPK elevations.(53) Ultimately, while an exposure-toxicity relationship may exist for daptomycin, there are insufficient data to support clinical use of a specific C_min as the upper bound for the therapeutic window.

β-lactam antibiotics and neurotoxicity

β-lactam agents are the most commonly prescribed class of antibiotics among hospitalized patients.(61) They have long been viewed as safe agents with a wide therapeutic index, for the most part. As a result of increasing data supporting the association between β-lactam concentrations and treatment success among adults with gram-negative infections,(62, 63) TDM is becoming more common for these agents, particularly in critically ill patients.(64) And, with the rise in drug resistant gram-negative bacteria, along with identification of numerous populations with large variability in pharmacokinetics (e.g. critically ill, obese, burn patients, cystic fibrosis), more aggressive dosing strategies have been considered to ensure attainment of effective drug concentrations.

Unfortunately, several β-lactam agents have become increasingly recognized as potential causes of neurotoxicity.(65) Reports of β-lactam-associated neurotoxicity have implicated cefepime,(66–69) carbapenems,(65, 68) piperacillin,(65, 68) flucoxacillin,(65) and other agents in this class (70, 71) as potential causes. While the mechanism underlying this adverse drug effect is not firmly established, a myriad of presenting signs and symptoms have been reported including encephalopathy, cognitive impairment, seizures, and coma.(70) Most often, these adverse events have occurred in critically ill adults and in those with renal impairment,(66, 68, 70, 72) suggesting that delayed drug clearance may lead to higher drug exposures and subsequent toxicity.

Several studies have examined the link between neurotoxicity and β-lactam serum concentrations, most often in patients undergoing routine TDM. In studies of patients receiving β-lactams as intermittent infusions, significantly higher C_min concentrations have been reported among subjects experiencing neurotoxicity.(65, 68, 69) Meanwhile, higher steady-state concentrations (C_ss) have been reported among patients with neurotoxicity during receipt of continuous infusions.(67) Overall, these studies demonstrate a clear exposure-toxicity relationship for β-lactams and neurotoxicity. However, the neurotoxic potential of these agents may differ by drug and class (i.e. penicillins, cephalosporins, carbapenems), although no comparative safety studies have been conducted.

The specific cut-point that best distinguishes those with and without toxicity has varied across studies and drugs, precluding clinical use of a specific cut-point for TDM (Table 4). Additional controlled studies will be needed to determine the true target for toxicity. While C_min has been the PK parameter most often associated with neurotoxicity, this is largely because C_min has been obtained as part of routine TDM for efficacy. It is not clear that C_min is the driver of toxicity as, to our knowledge, no published studies have evaluated whether AUC or C_max better define the risk of toxicity. Ultimately, further research is needed to define the therapeutic window for these β-lactam agents, particularly in critically ill patients with renal failure.

Table 4.

Association between β-lactam serum/plasma concentrations and neurotoxicity

Agent	First Author (Year)	PK parameter	Results
Cefepime	Beumier (2015)(68)	Cmin	No significant correlation between Cmin for ceftazidime/cefepime and neurotoxicity (p = .37)
	Huwyler (2017)(69)	Cmin, Css	Patients with neurotoxicity had higher Cmin (intermittent infusions; 52.2 vs. 21.3 mg/L, p < .001) and higher Css (continuous infusions; 48.8 vs. 33.8 mg/L, p > .05)
	Lamoth (2010)(123)	Cmin	Median Cmin significantly higher patients with neurotoxicity (28.0 vs. 7.2 mg/L, p < .0001); logistic regression found association between high Cmin and neurological adverse events (odds ratio: 1.42 per mg/L; 95% confidence interval: 1.00 to 2.03; p = 0.05
	Vercheval (2020)(67)	Css	Patients with neurotoxicity had higher Css (71.8 vs 49.6 mg/L, p = .036); ROC analysis defined optimal cut-point of 63.2 mg/L to identify neurotoxicity (AUC = .742)
Flucloxacillin	Imani (2017)(65)	Cmin	Mean Cmin significantly higher in patients with neurotoxicity (p = .01); threshold concentration for which there is 50% risk of developing a neurotoxicity event: Cmin >125.1 mg/L
Meropenem	Beumier (2015)(68)	Cmin	Significant correlation between higher Cmin and toxicity (p = .01)
	Imani (2017)(65)	Cmin	Mean Cmin significantly higher in patients with neurotoxicity (p = .04); threshold concentration for which there is 50% risk of developing a neurotoxicity event: meropenem, Cmin >64.2 mg/L
Piperacillin	Beumier (2015)(68)	Cmin	Significant correlation between higher Cmin and toxicity (p = .05) in patients treated with piperacillin/tazobactam; threshold at which toxicity occurred not reported
	Imani (2017)(65)	Cmin	Mean Cmin significantly higher in patients with neurotoxicity (values shown in figure but not reported; p < .01); threshold concentration for which there is 50% risk of developing a neurotoxicity event: piperacillin, Cmin >361.4 mg/L
	Quinton (2017)(124)	Css	Mean Css significantly higher in patients experiencing neurotoxicity: 159.9 vs 91.3 mg/L; p = 0.0016); piperacillin concentration >157.2 mg/L was an independent risk factor of neurotoxicity on multivariable analysis (odds ratio: 14.86, p = .03)

Category 3 – Drug exposure-toxicity relationship suggested but poorly defined

For many agents, the specific exposure parameter (AUC, C_min, etc.) that is most closely associated with toxicity has yet to be defined. This could be because quantification of exposure is difficult, the toxicity is rare (making it difficult to study), or simply because few studies have attempted to define an exposure-toxicity relationship. Category 3 includes drug-toxicity relationships that are suggested, often because of reports of increased toxicity with use of higher dosages or accumulative exposure, but for which the specific exposure parameter driving toxicity is not defined. Without a specific target, TDM cannot currently be used to mitigate toxicity.

Trimethoprim/sulfamethoxazole and hematologic and renal toxicities

Trimethoprim/sulfamethoxazole (TMP/SMX) is an antimicrobial used for both treatment and prevention of bacterial, fungal, and parasitic infections. It is a combination of two sulfonamide agents that act synergistically to inhibit folate metabolism. Because the drug contains two components, toxicities are numerous. Hypersensitivity reactions, electrolyte disturbances, acute kidney injury, gastrointestinal disturbance, and hepatic, hematologic, and CNS toxicities have been reported and are all more common among recipients of higher dosages of this drug.(73–75) In a study of 52 adults with HIV receiving TMP/SMX as treatment of Pneumocystis jirovecii infection, 40% experienced a drug-related adverse event.(73) Daily doses >16 mg/kg (TMP component) was independently associated with toxicity on Cox proportional hazards analysis.(73) In a large study comparing ambulatory patients treated with high-dose TMP/SMX to patients treated with standard-dose TMP/SMX, electrolyte disturbances (p = .002), hypersensitivity reactions (p = .008), and acute kidney injury (p = .04) were all more common in high-dose recipients.(74) And, in a large study involving more than 6000 TMP/SMX recipients with data in an electronic medical record database, more recipients of high-dose TMP/SMX (>5 mg/kg/day of TMP) sustained episodes of hyperkalemia (3% vs 1%, p <.001) and acute kidney injury (2% vs 0.7%, p <.001) than standard-dose recipients.(75)

The links between serum TMP and SMX concentrations and toxicities are less clear. In a study of 279 patients who had peak SMX levels obtained as part of TDM during TMP/SMX treatment (range of dosages: ~2.5–23 mg/kg/day), there was a weak correlation between daily dose and peak SMX level (R²: 0.34) and no relationship between peak SMX concentration and hematologic, electrolyte, or renal abnormalities.(76) Meanwhile, in a study of 37 patients with HIV treated with TMP/SMX for Pneumocystis jirovecii pneumonia, higher steady-state concentrations of TMP (>8 mg/L) but not SMX were associated with development of neutropenia and thrombocytopenia (timing of sampling and statistics not reported); liver function tests and rash did not vary based on TMP or SMX levels.(77) In a clinical trial of TMP/SMX for treatment of Pneumocystis jirovecii pneumonia among patients with AIDS, 122 had C_ave determined at steady-state.(78) Neutropenia, anemia, and azotemia were more common among subjects with TMP C_ave >8 mg/L compared to <5 mg/L (all p<.05).(78) Liver function tests and rash were not related to TMP C_ave and SMX C_ave was not associated with any adverse effects.(78) In a study by Joos et al. of 40 patients with Pneumocystis jirovecii pneumonia, 17 (43%) discontinued TMP/SMX during a planned 3-week course due to severe adverse effects.(79) There were no differences in serum TMP or SMX levels among patients with severe toxicities and those that completed the treatment course, however reactive SMX and TMP metabolites were not measured but could be contributing to toxicities.(79) Finally, in a study of 104 patients receiving low-dose, thrice weekly TMP/SMX for prophylaxis, 9 (8.7%) discontinued therapy due to adverse effects.(80) Concentrations of TMP and SMX were universally low precluding determination of exposure-toxicity or -efficacy associations.(80)

Toxicities of TMP/SMX are myriad and common when given at the higher dosages necessary for treatment of Pneumocystis jirovecii pneumonia. However, the bulk of data on TMP/SMX toxicities stem from studies of adults with HIV, which may not translate to other patient populations. Dosing of TMP/SMX is based on the TMP dosage. The mechanism of the various toxicities associated with each drug component differ, which means that exposure-toxicity relationship may also vary. Additionally, development of these toxicities even when TMP/SMX is given at low doses suggests that toxicities may have a component that is dependent on duration of exposure. Ultimately, additional studies are needed to define the specific toxicity target for TMP.

Amphotericin B and nephrotoxicity

Amphotericin B is a polyene antifungal commonly associated with toxicities. The deoxycholate formulation is well-tolerated in the neonatal period, but highly associated with nephrotoxicity outside of the neonatal population occurring in ~30% of patients.(81) Additionally, electrolyte disturbances and infusion related reactions frequently observed.(16) Lipid formulations were developed to improve the safety profile and include lipid complex, liposomal and colloid dispersion formulations.(82) Data on exposure as related to efficacy or toxicity are sparse for the polyenes. Despite the high rates of toxicities associated with the polyenes, TDM is not performed in the clinical setting and instead a reactionary approach to toxicity screening including close monitoring of renal function and electrolytes is recommended. A small study with mean serum concentration/time curves of amphotericin deoxycholate in 13 infants and children demonstrated elevated concentrations in all patients receiving 1mg/kg on day 7–18 of therapy. Eleven of these patients experienced hypokalemia despite potassium supplementation, and 6 of 9 patients with available data had elevated serum creatinine concentrations. The authors’ concluded that TDM of deoxycholate amphotericin should be integrated into clinical care given the elevated serum concentrations and high percentage of toxicities.(83) Similarly, a PK study in immunocompromised children receiving liposomal amphotericin B found statistically significant relationship between mean AUC_{0 –24} and probability of nephrotoxicity (OR, 2.37; 95% confidence interval, 1.84 to 3.22; P = 0.004).(84) Given the wide range of dosing used, nonlinear PK, and high rates of toxicity, further evaluation of amphotericin TDM should be considered.

Ganciclovir and valganciclovir and pancytopenia

Ganciclovir and its oral prodrug valganciclovir, are used for the prophylaxis and treatment of cytomegalovirus (CMV) in congenital neonatal infections and immunocompromised host populations. Toxicities associated with ganciclovir including pancytopenia, nephrotoxicity and neurotoxicity. Although interindividual variability in exposure is observed, clear associations between exposure and toxicity is lacking. One of the challenges with determining the relationship between ganciclovir exposure and toxicity is the rates of pancytopenia observed in the patients most likely to receive anti-CMV therapy are high due to underlying disease states including malignancy. In 82 adults receiving ganciclovir for the treatment of CMV infection or disease, no association was observed between toxicity and serum peak or trough ganciclovir concentrations.(85) The occurrence of anemia has been associate with AUC values higher than 50 μg · h/ml, as the incidence of anemia was 51.9% versus 26.6% in patients with AUC ranges <50 μg · h/ml (P = 0.010).(86) However, during a course of IV ganciclovir in 31 immunocompromised patients, no correlation was observed between peak and trough concentrations and neutropenia.(87) To date, no clearly defined exposure threshold associated with toxicity has been identified.

Acyclovir and neutropenia

Close monitoring for toxicity including neurotoxicity, nephrotoxicity and neutropenia is recommended when high dose acyclovir is prescribed.(88, 89) Based on an initial clinical study evaluating high dose IV acyclovir of 60 mg/kg/day in neonates, 21% of infants developed an ANC of <1000/mm³. Serum concentrations were obtained from 13 patients, but an evaluation of the relationship between exposure and neutropenia was not evaluated.(90) In an evaluation of infants receiving prolonged suppressive oral therapy of acyclovir for neonatal herpes, a trend in the development of neutropenia was observed in the treatment vs placebo group, though not statistically significant (P = 0.09) and serum concentrations of acyclovir were not measured.(88) Neutropenia developed in nearly half of 26 neonates receiving prolonged suppressive acyclovir and plasma concentrations of acyclovir were obtained during the course of therapy without a clear association between exposure and neutropenia and neutrophil counts recovered spontaneously in the majority of patients who continued acyclovir therapy.(91) Given the frequent observation of neutropenia in this population close monitoring of neutrophil count is recommended. Although an association between exposure and toxicity has not been defined, the resolution of neutropenia upon decreasing acyclovir dosing or discontinuing drug all together suggests a relationship between exposure and toxicity.(92)

Itraconazole and tremor

Itraconazole is primarily used to treat histoplasmosis. Guidelines for the treatment of histoplasmosis state that the concentration associated with an increased risk for toxicity has not been defined, yet concentrations >10 μg/mL should prompt dosage reduction.(93) Tremor has rarely been observed in patient with elevated itraconazole serum concentrations.(94) Although a retrospective analysis provided evidence of a relationship between higher itraconazole concentrations and adverse events suggesting a serum concentration >17.1 μg/mL predictive for the subsequent development of toxicity,(95) similar findings between itraconazole plasma levels and toxicity have not been replicated.(96) This in part could be due to the method used for measurement and the detection of the active metabolite in addition to parent drug.(97)

Posaconazole and pseudohyperaldosteronism

TDM is recommended with posaconazole to guide treatment efficacy, however no concentration dependent adverse events or toxicity have been well-described.(98) In patients being treated with posaconazole for 24 weeks, few adverse events were observed and no correlation between serum drug concentration and toxicity was noted.(99) Review of two randomized, controlled clinical studies involving more than 500 patient did not reveal a significant relationship between average plasma concentration and adverse events. Although the incidence of adverse events tended to be lower in the in patients with lower C_avg (range 21.5–355ng/ml) as compared to the other patients (range 355–3,650ng/ml), the difference was not statistically significant.(100)

More recently, posaconazole has been associated with posaconazole-induced pseudohyperaldosteronism (PIPH) resulting in secondary hypertension, hypokalemia and metabolic alkalosis. The development of PIPH has been associated with higher median serum posaconazole levels, with levels ≥4.0 μg/mL measured in patients with this adverse event.(101) Thus the role for TDM in posaconazole toxicity evaluation will continue to evolve.

Isavuconazole and gastrointestinal toxicities

Isavuconazole is the newest generation triazole antifungal agent used to treat invasive fungal disease. A hallmark feature of isavuconazole is the positive safety profile with few associated toxicities, lack of drug-drug interactions and predictable pharmacokinetic profiles in adults.(82, 102) TDM is not currently recommended for isavuconazole given its pharmacokinetic and safety profile.(103) However, serial monitoring of isavuconazole blood concentrations in 19 adults with median exposure of 90 days of therapy showed that levels do increase linearly over time.(104) Higher serum gamma-glutamyl transferase levels and gastrointestinal symptoms were associated with higher isavuconazole levels in multivariate analysis and based on the fixed – time ROC curve analysis, a cutoff of 5.13 μg/mL was identified as a threshold for toxicity.(104) The authors propose 5 μg/mL as the upper normal range to minimize toxicity, especially in patients receiving prolonged courses of isavuconazole or experiencing gastrointestinal symptoms, but further investigation is needed to corroborate these findings.(104)

Category 4 - No relationship between exposure and toxicity

Although all drugs have adverse effects, many appear to have no direct or meaningful relationship to systemic drug exposure. For some antimicrobials, such as cidofovir and foscarnet, the most notable toxicities are infusion related.(105, 106) Other drugs, such as fluconazole and echinocandins, have predictable pharmacokinetics and data do not support that higher exposures within the clinical range result in added toxicity.(107) For others still, such as with fluoroquinolones, toxicities are serious but rare, and systemic exposures are rarely known in individuals experiencing toxicity. For each of these drugs, TDM would not provide benefit in terms of mitigating toxicity. Ultimately, there are too many potential drug toxicities that could fall into this group to summarize fully here. Instead, we present some key examples below.

Antifungals for which no defined relationship between exposure and toxicity exists

Fluconazole is a frequently prescribed azole to treat candidiasis and is generally well-tolerated. PK data demonstrate a predictable linear dose/concentration relationship.(108) Data on the relationship between fluconazole serum concentrations and toxicity are sparse. The most reported adverse reactions include headache, dizziness, rash, and gastrointestinal symptoms. Reversible, asymptomatic elevation of hepatic aminotransferase has been observed in up to 12% of children, yet the relationship between serum concentration and liver enzyme elevation was not explored and has not been further elucidated.(109) In evaluation of high dose fluconazole, more adverse effects deemed possibly related to drug were observed in those receiving ≥ 1600 mg/day.(110) However at the standard dose of 400–800mg, the drug is well-tolerated making the integration of TDM to assess fluconazole toxicity unnecessary when routine dosing guidance is followed.

Given the predictable pharmacokinetics and safe toxicity profile,(12) TDM for the echinocandin antifungal drug class has not been investigated. Caspofungin was well-tolerated in four phase 1 studies with only a single subject exhibiting a possible laboratory determined drug related even with mild increase of transaminases.(111) Avoiding underexposure of echinocandins, especially in critically ill patients, is a possible future role for TDM in optimizing exposure, but likely has little role in minimizing toxicity.(112)

Antivirals for which no defined relationship between exposure and toxicity exists

Cidofovir and foscarnet are both associated with nephrotoxicity which can often be dose-limiting. Cidofovir renal toxicity appears to be related to high concentration of drug within the proximal convoluted tubule cells of the renal cortex while foscarnet is mostly associated with acute tubular necrosis.(106, 113) Mitigation strategies including hydration with both agents around the time of infusion and probenecid co-administration with cidofovir are used to minimize toxicity. The clinical applicability of TDM is limited due the direct infusion related toxicities seen with these antiviral agents.

Antibiotics for which no defined relationship between exposure and toxicity exists

Fluoroquinolones have several black box warnings (tendonitis/tendon rupture, peripheral neuropathy, CNS effects, and exacerbation of myasthenia gravis) and notable safety concerns (aortic dissection, QTC prolongation, Clostridium difficile infection). (114–116) And, published reviews have summarized the risk factors for many of these toxicities.(114–116) For tendon rupture and tendonitis, specifically, there is no clear exposure-toxicity relationship in humans. Although preclinical data suggest that tendonitis occurs more frequently at higher dosages in animals,(117) the systemic exposures achieved in animals with toxicity far exceed those that would be attained in humans. Thus, it is likely that the other risk factors that have been described (older age, steroid use, transplantation, and others) substantially influence who develops tendon-related toxicities during fluoroquinolone use. Based on available data, monitoring drug levels will not inform dosing design to prevent such toxicities.

Rifamycins are important antimicrobials used in the treatment of bacterial and mycobacterial infections. They are rarely used as monotherapy for any infection, making it difficult to decipher their toxicities from those of other co-administered agents. The rifamycins are hepatically metabolized and are involved in several important drug-drug interactions, which may also lead to added risk of toxicities. However, clinical TDM studies have largely found that most systemic exposures of these agents are below the therapeutic range and supra-therapeutic concentrations rarely are achieved.(118–120) As a result, TDM does not appear to have a role in regards to limiting toxicity, although it is advised for efficacy when used to treat tuberculosis.(121)

Conclusion

Advances in PK modeling approaches and the availability of target-oriented clinical software have made personalized antibiotic dosing feasible. The one-size-fits-all dosing strategy, which may be practical and reasonable for many individuals, should no longer be accepted for our sickest and most vulnerable patients. And, with ever-increasing rates of drug-resistant infections, use of personalized antibiotic dosing will become even more crucial to ensure that therapeutic efficacy and safety are achieved. To fully optimize antimicrobial therapies, however, targets for both efficacy and toxicity must be established.

For many antimicrobials, current data are lacking to support the integration of TDM into routine clinical care to minimize toxic effects. Yet, exposure-related toxicities occur with anti-infectives. Further work is needed to better define concentration thresholds associated with frequently observed toxicities. A better-defined exposure-toxicity relationship resulting in effective TDM strategies could help eliminate many of the observed antimicrobial toxicities. To date, much of the available data examining the relationship between exposure and toxicity come from retrospective studies or from case series of patients experiencing toxicity. The studies needed to better define potential exposure thresholds associated with toxicities will require dedicated resources and financial support to elicit useful relationships from Phase 3 studies. However, the long-term benefits of taking a proactive approach to reduce toxicity, instead of the existing strategies to reactively monitor end organ dysfunction followed by dose adjustments or drug cessation, could result in reduced health care savings.