Vancomycin therapeutic drug monitoring and population pharmacokinetic models in special patient subpopulations

Abstract Vancomycin is a fundamental antibiotic in the management of severe Gram‐positive infections. Inappropriate vancomycin dosing is associated with therapeutic failure, bacterial resistance and toxicity. Therapeutic drug monitoring (TDM) is acknowledged as an important part of the vancomycin therapy management, at least in specific patient subpopulations, but implementation in clinical practice has been difficult because there are no consensus and agglutinator documents. The aims of the present work are to present an overview of the current knowledge on vancomycin TDM and population pharmacokinetic (PPK) models relevant to specific patient subpopulations. Based on three published international guidelines (American, Japanese and Chinese) on vancomycin TDM and a bibliographic review on available PPK models for vancomycin in distinct subpopulations, an analysis of evidence was carried out and the current knowledge on this topic was summarized. The results of this work can be useful to redirect research efforts to address the detected knowledge gaps. Currently, TDM of vancomycin presents a moderate level of evidence and practical recommendations with great robustness in neonates, pediatric and patients with renal impairment. However, it is important to investigate in other subpopulations known to present altered vancomycin pharmacokinetics (eg neurosurgical, oncological and cystic fibrosis patients), where evidence is still unsufficient.

worldwide. 1 These facts turn urgent the need to develop strategies to improve both vancomycin prescribing and monitoring.
The aims of the present work were as follows: to define the state of art of vancomycin TDM, based on the three most recent guidelines from USA, Japan, and China, to review TDM in specials populations and resume population pharmacokinetic (PPK) models of vancomycin.

| ME TH ODOLOGY
The methodology followed was similar to that used by Ye and coworkers. 1 Three guidelines were selected as the basis of this work: the American from 2009 (AME), 2 the Japanese from 2013 (JPN) 3 and the Chinese from 2016 (CHN) 4 guidelines on vancomycin TDM.
These guidelines were analyzed, compared, and recomendations were evaluated. Level of evidence (LoE) and grade of recommendation (GoR) used in this work are shown in Table 1. The LoE and GoR used are in agreement with the GRADE system, 5 also used in the CHN guideline. 4 For comparison purposes, the graduation scale used in AME 2 and JPN 3 guidelines were transformed by two researchers, independently, and then discussed for consensus.
The "Population Pharmacokinetic Model review" addressed the items "Dose adjustment method" and "Special populations" since these were the items less detailed in the referred guidelines. This review was carried out using Ovid MEDLINE and EMBASE electronic databases.
The search equations used were: "Vancomycin" AND "Pharmacokinetic Model" for articles written in English, in humans until July 2017. A total of 63 records were found and two investigators independently screened the identified titles and abstracts to select articles. Of these, 31 records were excluded, either because they were not carried out in humans, do not directly concerned vancomycin or were general reviews of antibiotic use. The 32 full texted articles were tested for eligibility and nine were discarded since the models were not defined or were developed without using nonlinear mixed effects modelling.
Finally, 23 articles were included in the qualitative synthesis ( Figure 1), from now on referred as the "PPK Model review".

| Indication and relevance
In the guidelines investigated, only two guidelines refers situations where TDM is recommended. All clinical conditions where TDM is indicated present low quality of evidence. Therefore, these are considered as strong recommendations.
The literature review gathered evidence for the recommendation of TDM in additional special population groups [hematologic, neurosurgery, extracorporal membrane oxygenation (ECMO), neonates and pediatric cystic fibrosis (CF)]. Table 2 summarizes data relative to the analysis of the clinical relevance of TDM for specific patient conditions. Authors could not find additional data on the clinical relevance of TDM, other than that described by the CHN guideline. 4 However, a meta-analysis, including one randomized controlled trial (RCT) and five cohort studies, showed that TDM significantly increases the rate of clinical efficacy and decreases the rate of nephrotoxicity in patients treated with vancomycin. 6 More recently, another RCT, comparing groups of patients treated with vancomycin (simple infections or infections with MRSA), with and without TDM also confirmed this conclusion. 7 In fact, patients in the TDM group were discharged from the hospital more rapidly, reached clinical stability faster, had shorter courses of vancomycin and the time to initial target trough concentration was shorter. Moreover, in the MRSA infection subset, patients in the TDM group were also discharged from the hospital more rapidly, reached clinical stability faster, had shorter courses of vancomycin and attained initial target troughs in <5 days vs ≥5 days. 7 Data on TDM implementation on hospitals is also scarce: in the last 10 years, only one paper concerning TDM implementation was identified. However, this paper describes that, in France, implementation of vancomycin TDM is quite high: vancomycin TDM was available in 97% (477/490) of hospitals. 8 Unfortunately, a study carried out in Scandinavian countries showed that, in most cases, vancomycin TDM does not comply with recent recommendations/ guidelines. 9 Medical center implementation of the AME guideline, 2 with associated training, resulted in a significant short-term improvement in vancomycin dosing and TDM. The appropriateness of the prescribed dose increased from 51% of patients during the pre-period to 78% during the postperiod (P < 0.0001). Similarly, overall appropriateness of sampling of vancomycin troughs at steady state improved from 36% to 55% (P < 0.03). Specifically, the appropriate timing of troughs (within 30 minute of the next dose) increased from 37% (64/ 173) during the pre-period to 78% (149/191) during the postperiod (P < 0.0001). 10 Another study claims that identification of improvement opportunities in TDM methology and implementation over a 1-year period allowed a 37.5% reduction in inappropriately held vancomycin doses, although about 10% of doses remained as held inappropriately. 11 Moreover, this study highlights the difficulties in identifying barriers to change and modify healthcare worker behaviour. 11

| Dose adjustement methods
Methods for dose adjustment discussed in the guidelines are summarized in Table 3.
Although the guidelines for TDM of vancomycin do not recommend the use of nomograms, a novel vancomycin dosing nomogram has been recently developed and validated at two Canadian teaching hospitals by Thalakada and co-workers. 12 This nomogram was T a b l e 1 . Level of evidence and grade of recommendation (using GRADE approach) 5

Level of evidence (LoE) Grade of recommendation (GoR)
A (High quality) 1 (Strong recommendation) 2 (Weak recommendation) B (Moderate quality) C (Low quality) D (Very low quality) considered a useful tool that clinicians can use in selecting appropriate initial vancomycin regimens based on age and serum creatinine, to achieve high-target levels of 15-20 mg/L. The authors, however, stressed out that this tool should not replace clinical judgment for patients with unstable and/or reduced renal function. 12 Moreover, creatinine clearance-based nomograms for individualizing vancomycin doses should be used with caution in patients who require substantially prolonged drug exposure such as those with infective endocarditis. 13 The linear regression and Bayesian methods estimate, in general, more accurate dosage regimens. However, these methods require additional resources, such as information technology and healthcare   methods. 14,15 The "PPK Model review", in the adult population, found only six studies and, after analysis of the complete publicatons, three were excluded because they were not population models or were not developed using the nonlinear model of mixed effects methods.

| Special populations
The need for clearer guidelines regarding vancomycin dosing and TDM for patient subpopulations has been recently reported. 19

| Critically ill patients
The selected guidelines do not define special recommendations for this subpopulation. However, the literature review found 13 studies which are discussed below.
Significant challenges in vancomycin use in critically ill patients have been recently identified. 19 There is wide variability in reported practices for antibiotic dosing and monitoring. Therefore, research to develop evidence-based guidelines to standardize practices in critically ill patients is urgently needed. 20 These patients may present very large volume distribution (Vd) as well as supranormal drug clearance. 21,22 Augmented renal clearance has frequently been observed in critically ill patients which was strongly associated with vancomycin pharmacokinetics. As a consequence, two-thirds of these patients present subtherapeutic vancomycin concentrations. 23 Long duration of fasting and massive diarrhea have been associated with elevations in serum vancomycin concentrations, which suggest that TDM might be necessary during enteral vancomycin administration in critically ill patients. 24 Less than 40% of these patients attained therapeutic trough serum concentrations during the first 3 days of therapy. 19   to CI may, therefore, improve clinical outcomes in vancomycin-treated critically ill patients. 22,29 Although optimal administration based on PPK analysis and/or a Bayesian method has improved prediction accuracy, serum concentrations of vancomycin in patients with sepsis often deviate significantly from predicted values. Systemic inflammatory response syndrome (SIRS) duration was identified as influencing vancomycin concentration. Modifying dosing according to SIRS duration will improve prediction accuracy of vancomycin concentration based on TDM. 30 Table 5 describes the two studies proposing PPK models for critically ill patients found in the "PPK Model review".

| Pediatric patients
Only the JPN guideline 3 presented evidence on the advantages of TDM on pediatric patients ( Table 6).
Current recommended vancomycin dosing regimens in pediatric patients (40-60 mg/kg/day), result frequently in subtherapeutic concentrations. 33 Febrile neutropenia, a significant risk factor for augmented renal clearance in this subpopulation, indirectly influenced vancomycin clearance (Clvan) due to increased glomerular filtration rate (GFR). Increasing the initial dose is, therefore, required for achieving optimal therapeutic concentrations in pediatric patients with febrile neutropenia. 34 The probability of achieving an AUC/MIC >400 using only one trough serum concentration and one minimum inhibitory concentration (MIC) in patients receiving 15 mg/kg every 6 hours is variable according to the method used to calculate AUC.
In children, an AUC/MIC of 400 correlates with a Ctrough of 11 mg/ L using a trapezoidal method to calculate AUC. 35 For pediatric patients, monitoring of vancomycin Ctroughs is a recommendation stated in the summary of product characteristics and by several professional societies. 3 During a study where vancomycin TDM was performed and 7935 vancomycin concentrations were obtained, the median Ctrough increased from 10.9 to 13.7 mg/L, 36 which agrees with the recommendations published by the Infectious Disease Society of America. 2 These data suggest that vancomycin TDM is commonly performed in pediatric patients, and the majority of abnormal Ctroughs are associated with appropriate modifications of the dosing regimen. 36 Nevertheless, vancomycin TDM practices are reported to be highly variable in children admitted to pediatric hospitals. 37 The frequency with which serum vancomycin concentrations were monitored in children increased after the publication of the adult guidelines. This fact made some authors claim that the development of pediatric consensus guidelines is needed to optimize patient care and resource utilization. 37 Pediatric PPK models for vancomycin with Bayesian estimation can be used to reliably predict vancomycin exposure in children: the use AUC instead of Ctroughs, alone, can maximally optimize vancomycin administration in children. 41 Compared with one sample, the two samples sampling strategy improved accuracy and precision in estimating and predicting future AUCs. 40 Overweight and obese pediatric patients may have elevated initial vancomycin Ctroughs when empiric dosing is based on total body weight. This fact should make TDM mandatory in children. 38 In pediatric cancer patients, a Vd of 34.7 L was reported and clearance values that were correlated with body weight, tumor disease, and cyclosporine co-administration. 39 Based on simulation results, dose (mg/kg) should be individualized based on body weight and cyclosporine co-administration. 39 The "PPK Model review" in pediatric and neonatal patients found five and seven studies, respectively. From these, three were selected in pediatric and six in neonates. The excluded studies were T a b l e 5 .   those that were not population models or were not developed using nonlinear modeling of mixed effects methods (Table 7).

(a) With cystic fibrosis
No reference concerning pediatric patients with CF was found in the three international guidelines evaluated.
The "PPK Model review" revealed only two studies concerning this population subgroup. One of these showed that vancomycin dosing of 60 mg/kg/day does not reliably achieve a vancomycin Ctrough of 15-20 mg/L in pediatric patients with CF. 50 The second also reported that younger CF patients may require higher vancomycin doses. 51 The PPK model found for this specific subpopulation is described in Table 7.

(b) Neonates
In neonates, vancomycin is the first choice for late-onset sepsis treatment. However, prescribing the right dose and dosing regimen remains a challenge in neonatal intensive care units. 52 The high degree of pharmacokinetic variability in neonates makes TDM essential to ensure adequate therapeutic exposure 53 and prevent adverse renal outcomes. 54 When using TDM in neonates the basic rules apply. However, additional factors should also be taken into consideration. First, due to both pharmacokinetic variability and nonpharmacokinetic factors, the correlation between doses and concentration is poor, but can be overcome using more complex, validated dosing regimens. Second, the time to reach steady-state is increased, especially when no loading dose is used and, therefore, TDM sampling timing is of utmost importance in neonates. Third, the target concentration may be uncertain. Finally, because of differences in matrix composition (eg, protein, bilirubin), assay-related inaccuracies may differ in neonates. 55 With currently recommended vancomycin dosing, the therapeutic target of AUC/MIC > 400 is achieved only by 25% of neonates. 56 Most of Ctrough in neonates achieved using two published dosing regimens did not reach the 10 mg/L. 57 Table 7 describes the studies proposing PPK models for neonates' patients found in the "PPK Model review".

| Elderly patients
The selected international guidelines also do not present specifications for this subpopulation. In the "PPK Model review" three studies were found concerning elderly patients.
The recommended target range of 15-20 mg/L for vancomycin Ctrough seems to be acceptable for controlling vancomycin exposure, although a value of approximately 11 mg/L was found as optimal and safer in elderly patients. 67 Efficacy of vancomycin was associated with area under the trough level (AUTL), a novel pharmacokinetic parameter. 68 Determining the target AUTL or Ctrough may enhance the efficacy of vancomycin therapy in elderly patients with MRSA pneumonia. 68 Given that nephrotoxicity may increase with a Ctrough >15 mg/L, this level should not be exceeded in this subpopulation.

| Obese patients
The  and obese pediatric patients may have elevated initial vancomycin Ctroughs when empiric dosing is based on ABW and, therefore, TDM should be mandatory for this specific subpopulation. 38 Vancomycin TDM showed that underdosing and overdosing occur more often and effective levels are less often achieved, in obese patients. TDM might be of special importance, in obese patients. 70 The majority of these patients present subtherapeutic concentrations, which increases the risk of treatment failure and bacterial resistance.
Further studies are needed to determine the optimal dosing strategy in morbidly obese patients, ie with more than 100 kg and at least 140% of their ideal body weight. 69 Calculating individual pharmacokinetic parameters using equations may be a valid tool for dosing vancomycin in obese patients with renal insufficiency. 71 TDM has been correlated with pharmacokinetic/pharmacodynamic optimization for vancomycin in the obese population with skin and soft tissue infections, and should be used in these cases. 72 Using two serum vancomycin concentrations significantly improves subsequent target Ctrough attainment in the obese population. 73

| Patients with impaired renal function
The JPN guideline 3 summarizes evidence and present specific recommendations for the subpopulation of patients with impaired renal function ( Table 8).
The Ctroughs of 62.9% patients with high creatinine clearance (Clcr) were <10 mg/L. 74 Since augmented renal clearance was significantly associated with subtherapeutic vancomycin concentrations, it was necessary to devise adjusted dosage regimens for these patients, based on Clcr values. 74 The "PPK Model review" in patients with impaired renal function found seven studies: two general, three on hemodialysis, one on peritoneal dialysis and one on continuous hemofiltration patient).
Four of those studies were excluded as they were not population models or they were developed without using nonlinear modeling of mixed effects methods. The remaining three studies proposing PPK for patients with impaired renal are described in Table 9.

(a) Hemodialysis patients
The JPN guideline 3 summarizes evidence and presents recommendations for the subpopulation of patients undergoing hemodialysis (Table 10).
There is considerable variation in vancomycin pharmacokinetics in patients undergoing hemodialysis. 78 Attention must be paid to the reliability of several empiric dosing recommendations derived from small pharmacokinetic studies in heterogeneous populations. Followup TDM is suggested as essential to ensure that concentrations remain within the target range in these patients. 78 Pharmacokinetic variables of prolonged distribution phase, T a b l e 1 0 . Guidelines review of patients receiving renal replacement therapy

Question Answer (LoE/GoR) Guideline
Patients receiving hemodialysis 1. Initial dose of 15-25 mg/kg (as actual body weight) is recommended. As an initial dose of 15 mg/kg may not be adequate to achieve recommended trough concentrations, experts recommend that a loading dose of 20-25 mg/kg should be administered. (C/1) 2. As a greater amount of vancomycin is removed during hemodyalisis, doses of 500 mg (7.5-10 mg/kg) after each dialysis treatment are given as maintenance doses. (C/1) 3. Weekly vancomycin dosing results in subtherapeutic serum levels and should be abandoned in a high-flux setting. (D/2) 4. Achievement of a steady-state concentration is delayed. Although there is no evidence concerning the timing of TDM, the committee recommend that TDM is performed within 1 week after the start of therapy. (C/1) 5. There is no consensus concerning the necessity of follow-up TDM in whom the dosage regimen was not altered. (Unresolved issue) 6. Blood samples for TDM should be drawn before dialysis treatment. Because of the rebound phenomenon, trough levels immediately after the completion of hemodialysis do not reflect the exact drug concentrations of patients. (C/1) 7. Although the maintenance of trough concentrations of <20 mg/L is desirable, there is no consensus concerning the concentrations causing adverse events. (Unresolved issue)

2013, JPN 3
Patients receiving continuous renal replacement therapy 1. An initial dose of 15-20 mg/kg (as actual body weight) is generally administered. Some experts recommend higher loading dose is required to achieve target trough concentrations. (C/1) 2. As a great amount of vancomycin is removed during continuous venovenous hemodiafiltration doses of 500 mg (7.5-10 mg/kg) are given every 24 h as maintenance doses. It is recommended to adjust the doses according to the result of TDM. (C/1) 3. In patients with residual renal function in whom the main purpose of this therapy is removal of several mediators that cause detrimental effects during sepsis, increased vancomycin dosing may be required  Table 9 describes the PPK models developed for hemodyalisis patients found in the "PPK Model review".

(b) Patients receiving continuous renal replacement therapy
The JPN guideline 3 summarizes evidence and presents recommendations for the subpopulation of patients receiving continuous renal replacement therapy (Table 10).
Extracorporeal clearance of drugs increased with higher-intensity continuous renal replacement therapy. This increase was significant for vancomycin. In these patients, there is great variability in antibiotic pharmacokinetics, which complicates an empirical approach to dosing and suggests the need for TDM. 81 CI produced more frequently therapeutic vancomycin levels and less frequently subtherapeutic levels compared to intermittent infusion. However, therapeutic levels were achieved infrequently by either dosing method. Given equivalent TDM costs and the lack of a clear clinical benefit, the role of CI remains to be defined, in spite of practical and theoretical advantages, particularly in burn patients. 82 (c) Patients receiving ambulatory peritoneal dialysis The JPN guideline 3 summarizes evidence and presents recommendations for this subpopulation (Table 10).
Clinical outcomes of gram-positive and culture-negative peritonitis were not associated with either the frequency or levels of serum vancomycin measurements in the first week of treatment when vancomycin is dosed according to International Society for Peritoneal Dialysis. 83 Table 9 describes the PPK models developed for this subpopulation.

| Burn patients
None of the evaluated international guidelines presents defined recommendations for this subpopulation. The literature review identified a study that shows that higher clearance and lower serum vancomycin concentrations in patients with severe burns may increase the risk of suboptimal bactericidal action and development of resistance, highlighting the need for dose individualization. 84 The "PPK Model review" found only one study in this subpopulation which is described in Table 11. a target Ctrough of 11.5 mg/mL for febrile neutropenia in patients with hematological malignancies. 86 The high-dose, once-daily vancomycin nomogram attained trough levels greater than 10 mg/L in only 21% of patients with leukemia and a substantial number of adverse drug reactions were observed leading to the nonrecommendation of such regimen for outpatient therapy. 87 The "PPK Model review" found three studies in hematological patients. One was excluded because it was not a population model and the remaining two are described in Table 12.

| Neurosurgery patients
Adult neurosurgical ICU patients showed a significantly elevated Clvan (0.104 ± 0.036 L/h/kg). 92 Augmented Clvan should be considered when determining vancomycin doses in neurosurgical patients. 91 Two dosing equations were derived to achieve optimal serum vancomycin concentrations for this subpopulation. 92 Further research using TDM in the management of CNS infections, in this setting, in addition to work defining plasma and cerebrospinal fluid (CSF) concentrations associated with antibacterial efficacy and toxicity is mandatory. 93 The "PPK Model review" in neurosurgical patients found only one study that is described in Table 13.

(a) with spinal medulla lesions
Vancomycin dose selection in patients with spinal cord injury (SCI) is challenging due to difficulties in accurately estimating renal function in these patients. 95 A recent study suggests that the use of the  [102] Cl, clearance of vancomycin; CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; N, sample size; V1, volume of central compartment; V2, volume of peripheric compartment.
pump was significantly lower than that in the matched cohort. 100 As a result of drug sequestration and increased Vd, the ECMO procedure might lead to a decrease in drug concentrations. Vancomycin concentration remained unchanged in the ex-vivo ECMO circuit primed with whole human blood. 101 The literature review of PPK models found only one study, described in Table 14.

(d) Vascular surgery
The target concentration (10-25 mg/L) was achieved in 81% of all samples collected in one study of vascular surgery patients. 103 All patients achieved target concentrations at one or more-time points.
The regimen employed provided appropriate concentrations at the time of intervention. No potentially toxic concentrations or adverse reactions to vancomycin were reported in patients undergoing vascular surgery. Vancomycin given as CI delivers adequate serum concentrations. 103

| FUTURE PERSPECTIVES
Despite the availability of consensus guideline recommendations, practices for dosing and monitoring of vancomycin are not universally applied. 104 This review has gathered additional evidence that TDM has clinical relevance in several patient subpopulations (neonates, pediatric and with renal impairment) but there is still lack of research concerning other subpopulations (neurosurgical, oncological, cystic fibrosis).
An updated review of PPK models for specific subpopulations was carried out and models have been summarized for future reference/research and TDM refinements. Currently, most of these models have not been prospectively validated and TDM methodologies adaptations for specific populations are still not consensual. The use of dose adjustment methodologies based on PPK models and Bayesian estimation of parameters seems to gather the higher scientific consensus.
In the future, well designed prospective studies should be carried out to demonstrate the relevance of TDM, validate PPK models in clinical settings and find consensual refinement adaptations of TDM methodologies for specific patient subpopulations using vancomycin.

STANDARDS
The authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest in the subject matter or materials discussed in this manuscript.

CONF LICT OF I NTEREST
The authors hereby confirm that they do not have any conflicts of interest to declare.