Carbapenem Therapeutic Drug Monitoring in Critically Ill Adult Patients and Clinical Outcomes: A Systematic Review with Meta-Analysis

Drug monitoring is one strategy of antibiotic stewardship to face antimicrobial resistance. This strategy could have a determinant role in critically ill patients treated with carbapenems to overcome pharmacokinetic variability, reduce the risk of subtherapeutic dosage or toxicity, and reduce the risks inherent to treatment. However, the effectiveness of therapeutic drug monitoring (TDM) is unknown. This paper aims to identify TDM effectiveness in critically ill patients treated with carbapenems. English and ClinicalTrials.gov databases were searched to identify relevant studies evaluating carbapenem TDM. Randomized controlled trials (RCTs) and comparative cohort studies were selected for inclusion if they compared carbapenem TDM to standard care in adult critically ill or sepsis/septic shock patients. The primary outcome was mortality. Secondary outcomes included morbidity, clinical cure, microbiological eradication, antimicrobial resistance, drug-related side effects, and achievement of target plasma concentrations. Overall, performing carbapenem TDM was not associated with a decrease in mortality. However, it could be evidence for a relationship with clinical cure as well as target attainment. Some studies found favorable outcomes related to clinical and microbiological responses, such as lower procalcitonin levels at the end of the monitored therapy compared to standard care. For the primary and secondary outcomes analyzed, strong evidence was not identified, which could be due to the size, risk of bias, and design of selected studies.


Introduction
In the treatment of infectious diseases, the therapeutic index (TI) of antimicrobials can show significant pharmacokinetic (PK) variability. It is of great importance to know the concentration of the antibiotic, which must be higher than the minimum effective concentration (MEC) for a favorable clinical outcome for the patient [1] to avoid antibioticresistant bacteria [2]. The measurement of the concentration of drugs in fluids such as plasma, serum, or blood in patients at specific intervals is called therapeutic drug monitoring (TDM) [3,4]. PK (study of drug processes throughout the body, such as absorption, distribution, metabolism, and excretion) and pharmacodynamic (PD) (drug transformation processes at the site of action and its pharmacological effect) can vary [5] to adapt the drug dose and improve efficacy and/or reduce toxicity. The measurement of drug concentrations allows the dose to be adjusted to an adequate therapeutic range [6]. There are several indications to perform TDM, such as drugs with a narrow therapeutic index (NTI), high variability or unpredictability between drug dose and plasma concentration, situations where there is knowledge of the clinical effect and/or toxicity related to concentration, In line with the above-mentioned, the alteration of PK parameters has implications in the clinic, as does the administration of an adequate dose of antibiotics, making a case for the implementation of routine TDM [24]. However, the practice and benefits of carbapenem TDM in critically ill patients is not clear [25]. The aim of this review is to identify the efficacy of TDM in these types of patients in terms of clinical and bioanalysis results, and as a strategy to reduce antimicrobial resistance.

Search Results
The reviewers screened 383 articles selected from electronic databases according to the search criteria described in Table S1. After removing duplicates and articles outside the scope of the study, 107 studies were eligible based on the assessment of titles and abstracts. A further 102 studies were excluded based on the review of full-text articles because of their population, intervention, or the absence of a control group. One study was also excluded because it contained preliminary results of a randomized controlled trial (RCTs) published later. Five studies were included, two of them were RCTs and three were retrospective cohort studies ( Figure 1).

Study Characteristics
Five studies were included according to their study design and relation with the aims of this study. Two of the studies were single-center RCTs and the other three were singlecenter retrospective cohort studies. All selected studies included meropenem as the main carbapenem of interest, however, Fournier also included imipenem and ertapenem [26]. The studies were partially blinded and used chromatography analysis to establish antibiotic plasma concentrations. De Waele's RCT had a unique combined antibiotic exposure (meropenem and piperacillin/tazobactam) without availability of reliable absolute frequencies, therefore, it was not considered for the meta-analysis [27].
For the primary outcome, a total of four out of five studies included mortality, although just three of them (observational studies) were included to perform the meta-analysis, with a total of 448 patients, with 248 assigned to the treatment guided by TDM and 200 assigned to the standard care group [28][29][30]. Only two of the secondary outcomes from observational studies could be performed with meta-analysis, which were ICU length of stay and microbiological eradication, with a total of 294 and 198 patients, respectively ( Figure 2). For the rest of the secondary outcomes, a descriptive summary was performed because of their intervention/exposition variability or high risk of bias. The description of studies selected, and clinical outcomes are shown in Tables 1 and 2 respectively. ntibiotics 2021, 10, x FOR PEER REVIEW 4 of 1

Study Characteristics
Five studies were included according to their study design and relation with the aim of this study. Two of the studies were single-center RCTs and the other three were single center retrospective cohort studies. All selected studies included meropenem as the main carbapenem of interest, however, Fournier also included imipenem and ertapenem [26] The studies were partially blinded and used chromatography analysis to establish antibi otic plasma concentrations. De Waele's RCT had a unique combined antibiotic exposure (meropenem and piperacillin/tazobactam) without availability of reliable absolute fre quencies, therefore, it was not considered for the meta-analysis [27].
For the primary outcome, a total of four out of five studies included mortality, alt hough just three of them (observational studies) were included to perform the meta-anal ysis, with a total of 448 patients, with 248 assigned to the treatment guided by TDM and 200 assigned to the standard care group [28][29][30]. Only two of the secondary outcome from observational studies could be performed with meta-analysis, which were ICU length of stay and microbiological eradication, with a total of 294 and 198 patients, respec tively ( Figure 2). For the rest of the secondary outcomes, a descriptive summary was per formed because of their intervention/exposition variability or high risk of bias. The de scription of studies selected, and clinical outcomes are shown in Table 1 and Table 2 re spectively.     hospital destination (discharge from ICU, interhospital transfer and mortality), demographic variables n/a n: number of participants; AB: antibiotic; ↑: increase; ↓: decrease; PK: pharmacokinetics; n/a: not available; MEM: meropenem; PTZ: piperacillin/tazobactam; TDM: therapeutic drug monitoring; fT >MIC : time fraction above of the minimum inhibitory concentration; C max : maximum concentration; ICU: intensive care unit; C min : minimum concentration; UHPLC-MS/MS: ultra-high performance liquid chromatography-tandem mass spectrometry; PK/PD: pharmacokinetics/pharmacodynamics; PCT: procalcitonin; LD: licensed dose; HD: high dose; SD: standard deviation, HPLC: high-performance liquid chromatography Table 2. Clinical Outcome.

Clinical Outcome
For combined effect in the primary clinical outcome (in-hospital mortality), 41 of the 248 patients (16.53%) died in the TDM cohort and 46 of 200 patients (23%) died in the standard care group. The pooled relative risk of in-hospital mortality did not show a statistically significant difference between the groups (0.75 [95% confidence interval 0.49 to 1.13]), although Meyer's research showed differences that favor TDM (Figure 2) [29]. In the secondary clinical outcomes, the pooled relative risk of the ICU length of stay and microbiological eradication between groups were statistically non-significant with a mean difference of −31.38 [95% confidence interval −180.96 to 4.56] and a relative risk of 1.23 [95% confidence interval 0.86 to 1.76], respectively (Figure 2). For the rest of secondary outcomes only a systematic review was performed ( Table 2).
As for achieving clinical cure, Aldaz et al. found a statistically significant reduction in procalcitonin (at least 80%) in the TDM cohort compared with the control (71% versus 53%, p-value equal to 0.02) [28]. For target attainment, studies were heterogenous in their measurements of pharmacokinetic parameters, however, both De Waele et al. and McDonald et al. measured 100% fT >MIC , with a statistically significant higher attainment of target levels in the TDM group in the former (94.7% versus 68.4% with p-value of 0.045), which also reported a statistically significant difference in the 100% fT >4MIC between groups (57.9% versus 15.8% with p-value equal to 0.007) [27,30].
De Waele et al. and Aldaz et al. included antimicrobial persistence (antimicrobial resistance was not measured in any study), with statistically non-significant differences between groups [27,28]. Adverse reactions were another evaluated outcome, with Aldaz and McDonald observing no significant differences between groups for various systems and organs [28,30]. Finally, hospital re-admission was evaluated in a cohort study (Aldaz) which found a small, but non-significant reduction in the patients re-admitted in the group that received therapeutic drug monitoring, compared to the control group (6.49% versus 9.09% with p-value of 0.54) [28].

Heterogeneity of Studies
For the primary clinical outcome (in-hospital mortality), heterogeneity among studies was not observed (I 2 = 11%; χ 2 = 2.25; p-value = 0.32) (Figure 2). The heterogeneity for secondary outcomes (ICU length of stay and microbiological eradication) is described in each forest plot (Figure 2).

Publication Bias, Risk of Bias, and Quality of Evidence
We did not test for the presence of publication bias for any outcome because there were less than 10 studies. For the selected RCTs we found a high risk of bias, mainly in the blinding processes, incomplete outcome data and selective reporting of information (Figures 3 and 4) [26,27]. For observational studies, Aldaz's had a low risk of bias and the remaining studies received a moderate risk of bias score, with blinding mechanisms and follow-up considered the major problems (Table 3) [28][29][30].      A study can be awarded a maximum of one star ( ) for each item except comparability, for which a study can be awarded a maximum of two stars ( ).

Discussion
To our knowledge, this is the first systematic review and meta-analysis to assess the clinical outcomes of performing carbapenem TDM. TDM is an underused practice and its application has been limited to a few antibiotics with PK features that increase the risk of clinical failure and toxicity, like vancomycin and aminoglycosides [31,32]. β-Lactam and, specifically, carbapenem TDM have not been widely investigated because of the wide TI associated with these antibiotics. However, some populations would benefit from performing TDM because of their intra and interindividual PK variability; the unpredictable PK in critically ill patients is the main reason why TDM could optimize antibiotic treatment, for β-lactams and especially carbapenems despite their wide TI, leading either to reduced antibiotic resistance and adverse effects or enhanced clinical or microbiological cure. Although, evidence of the relationship between implementation of TDM and improved outcomes is, according to our findings, substantially less solid.
For the critically ill population in particular, optimizing therapy ensuring an adequate antibiotic exposure can be challenging, because sepsis itself can induce multiple organ dysfunction that leads to a reduction of antibiotic clearance [33], increasing the likelihood of toxicity [33] and reducing the exposure, making it more difficult to achieve the PK/PD target. According to Blot et al., some other hemodynamic changes that occur in critically ill patients which alter PD are homeostatic disturbance, endothelial dysfunction, capillary leak, decreased plasma protein concentrations, and extreme body weight changes [34]. Some PK changes in this population include increased Vd, augmented renal clearance, hypoalbuminemia (which alters the unbound fraction of the drug), and reduced bacterial susceptibility [35]. Therefore, prescription of standard doses is likely to result in subtherapeutic concentrations [34], increasing the likelihood of therapeutic failure.
Furthermore, dangerous adverse effects resulting from antibiotic use could be avoided with the use of TDM [3], particularly, by maintaining plasma levels below the threshold of the minimum toxic concentration (MTC), also called maximum safe concentration (MSC), which is when unacceptable adverse effects begin to take place [36]. In the case of carbapenems, neurotoxicity-chiefly in the form of seizures-is particularly concerning when patients are overdosed relative to weight or renal function [37].
We selected mortality as our primary outcome. As shown in Table 2, TDM was not associated with inferior mortality rates. The lack of efficacy of TDM for improving mortality rates may relate to some inconsistencies found in the selected studies, including study design, number of participants, procedures, and methodological disparities. Only the study by Meyer et al. had statistically significant results that support this outcome [29]. It is likely that new experiments as well as bigger studies could show major differences. Because meropenem has a post-antibiotic effect, prolonged follow-up durations may be necessary to detect TDM-related benefits in terms of mortality. However, most of the included studies, including both RCTs, did not assess outcomes after hospital discharge, and there are no studies that measure mortality rates in a significant period.
According to the meta-analysis performed in this review, TDM does not significantly affect the ICU length of stay of patients. This result could be due to the design of the studies and the size of both the TDM cohort and control groups in the studies analyzed. However, a multi-center RCT, the DOLPHIN trial, is underway, designed to assess the efficacy and cost-effectiveness of model-based TDM of β-lactam and fluoroquinolones, their primary outcome is the ICU length of stay [38]. This systematic review is intended to be updated when the DOlPHIN trial shows its first results. On the other hand, microbiological eradication and TDM did not show a direct relationship in our study. However, studies such as those by Bricheux et al. have found failure occurred in a large proportion of patients whose dose was not increased, though the difference between the groups were not statistically significant [39].
The results shown in this review are not sufficient to demonstrate that meropenem TDM helps reach a defined PK/PD target. The only study that had statistically significant results regarding performing TDM and antimicrobial exposure was the RCT by De Waele et al., which had important biases [27]. Some studies demonstrate that reaching PK/PD targets is related to better clinical outcomes. In a retrospective cohort study, achieving the ideal PK/PD target in other β-lactams such as cefepime and ceftazidime demonstrated greater clinical success and bacteriological eradication; the PK/PD target established was 100% fT >Mic , being compared with a group with fT >MIC lower than 100% [40]. Another study that demonstrated better clinical outcomes related to achieving PK/PD targets was the defining antibiotic levels in intensive care unit patients (DALI) study, which was a prospective, multinational PK point-prevalence study and showed that the group that did not achieve the Pk/PD target of 50% fT >MIC were more than 30% less likely to have a positive clinical outcome (odds ratio [OR], 0.68; p = 0.009) [41]. This also demonstrated that increasing 50% fT >MIC and 100% fT >MIC ratios was associated with positive clinical results (OR, 1.02 and 1.56, respectively; p < 0.03), with significant impact on sickness severity status [41]. Another retrospective study showed a significant correlation with successful clinical outcomes related to achieving a different PK/PD target, using the steady stationary concentration divided by the minimum inhibitory concentration of the isolated bacteria (C ss /MIC ratio) instead of percentage of time above the MIC (OR = 12.250, 95% CI 1.268-118.361; p = 0.03) [42].
High-dose continuous-infusion optimized by TDM may represent a useful mechanism when a carbapenemase producing microorganisms is present [42]. However, more high-quality studies are required to approach this setting. In our research, antimicrobial resistance outcome was not studied directly, and only antimicrobial persistence was reported in two studies. Similarly, adverse reactions by systems and patient readmission did not get positive results for the quality of the selected research.
A pharmacoeconomic analysis related to TDM should include assessing the incidence of drug-induced adverse reactions, reduction in total length of hospitalization, cure rates, mortality rates, and cost savings associated with monitoring plasma levels of drugs. Unfortunately, very few articles have reviewed the pharmacoeconomic impact of this practice and, to our knowledge, there are no studies related specifically to TDM of carbapenems and cost-effectiveness. Although, for other drugs, TDM practice remains cost-effective through improved clinical outcomes, like biologic therapies in inflammatory bowel disease through reducing dosing and improving disease control [43] or digoxin through toxicity reduction [44].
Regarding antibiotics, there are few studies demonstrating significant cost reductions when TDM is performed, especially in aminoglycosides therapy. Bootman et al. performed a cost-effective analysis of burn patients, demonstrating a saving of $6689/patient in direct costs when patient's gentamicin therapy was monitored by a clinical pharmacokinetic service [45]. Another study performed by Burton et al. found a reduction in length of stay and a potential reduction of costs when aminoglycosides TDM was put into practice [46]. Improvement of quality of patient care, reduction of patient length of stay, reduction of mortality, and earlier attainment of therapeutic drug concentrations in intervention groups compared to control groups have demonstrated potential costs savings [47]. Nonetheless, further studies are needed to demonstrate that TDM is a cost-effective practice [48].
Finally, our meta-analysis has several limitations. First, only a few RCTs met the criteria to be included in our analysis. As such, more RCTs are needed to identify true differences between groups. Second, participants and healthcare staff were aware of the group assignments in all included studies, and selective reporting and follow-up were problematic, possibly resulting in performance bias. Lastly, many of the secondary outcomes had different ways of being measured, therefore, it was not possible to do effective comparisons. It is imperative to carry out higher quality research to approach carbapenem TDM effectiveness in critically ill patients.

Data Sources and Searches
We performed a systematic review as per the guidelines established by the Cochrane  15 December 2020. We used a population, intervention, comparators, and outcomes (PICO) format question [42] for the search strategy.
Search strategy associated the Medical Subject Heading (MeSH) and keywords linked to carbapenems with terms for critical illness, septic shock, sepsis, therapeutic drug monitoring, antibiotic treatment outcome, clinical outcomes, and antimicrobial drug resistance (full search strategy is in Table S1 and Table S2). Two authors (J.B. and S.L.) independently scrutinized full texts to include those who meet the selection criteria. When the two independent reviewers did not agree, a third reviewer (R.B.) made the selection.

Search Eligibility Criteria
We considered RCTs and comparative retrospective or prospective cohort studies as preferred study designs for inclusion. Suitable studies compared carbapenem TDM to standard care in adult critically ill or sepsis/septic shock patients. Selected research had to address at least one of the following outcomes of effectiveness determined by the authors: mortality (primary), morbidity, clinical cure, microbiological eradication, antimicrobial resistance, drug-related side effects, and adverse reactions, and achievement of target plasma concentrations (secondary). We excluded cross-sectional studies, intervention, or observational studies without a control group and research carried out in the pediatric population.

Data Extraction and Assessment of Methodological Quality
We collected all the papers found in a free web application Rayyan QCRI ® , where two authors (J.B. and S.L.) separately scrutinized full texts for inclusion and remove duplicates. Three authors (Y.F., H.L., and N.D.) independently assessed randomized controlled trials (RCTs) for risk of bias adopting the Cochrane risk of bias tool [40] and observational studies applying the Newcastle-Ottawa scale (NOS) [43]. RCTs were acknowledged as having a low risk of bias (all features graded as low risk), high risk of bias (one or more items classified as high risk), or unclear risk of bias (one or more items rated as unclear risk of bias and no items within the high risk). Observational studies were classified using NOS as low risk of bias (7-9 stars), moderate risk of bias (4-5 stars), and high risk of bias (0-3 stars). Disparities were resolved by consensus with another author as required.

Data Synthesis
We performed a systematic and descriptive review and did a meta-analysis of those outcomes with more than one research and the same study design (RCTs and observational studies separately). We employed Mantel-Haenszel and inverse-variance, and DerSimonian and Laird models within the random-effects model, for continuous and dichotomous outcomes, respectively [44]. Merged estimates were expressed within 95% confidence intervals (CIs), as mean differences (MDs) for continuous outcomes and as relative risks (RRs) for dichotomous outcomes. A two-sided p-value < 0.05 was interpreted as statistically significant. To evaluate heterogeneity, we performed the Chi-square (χ 2 ) test, with significance defined as p-value < 0.05, and the I 2 statistic (≥50% was interpreted as severe heterogeneity) [45]. We did not assess publication bias because of the reduced number of studies identified. Analyses were done with Review Manager (RevMan version 5.4, The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark, 2014).

Conclusions
This systematic review and meta-analysis did not find a significant association between carbapenem TDM and favorable clinical outcomes, including mortality, reduced ICU stay, microbiological, or clinical cure, possibly because of limited evidence. Higher quality longitudinal studies are required to establish TDM-guided therapy effectiveness in the critically ill patient setting.