Introduction

Linezolid is an oxazolidinone antibiotic used to treat infectious diseases caused by drug-resistant gram-positive bacteria, such as methicillin-resistant Staphylococcus aureus and vancomycin-resistant Enterococci. Linezolid inhibits bacterial protein synthesis by binding to ribosomal RNA (30S and 50S ribosomal subunits) [1]. This unique mechanism prevents cross-resistance to existing antimicrobial agents of other classes [2]. However, the major treatment-related adverse event of linezolid therapy is hematological toxicity, mostly thrombocytopenia, which leads to treatment discontinuation and failure [3,4,5]. Generally, linezolid and its primary metabolites are excreted via non-renal (approximately 65%) and renal mechanisms [6]; therefore, dose adjustment is not required in patients with decreased renal function (DRF) [2, 7, 8]. However, recent studies have revealed that during linezolid therapy, the incidence of treatment-related hematological toxicity is significantly higher in patients with DRF than in those with normal renal function [9,10,11,12,13].

To avoid hematological toxicity, some studies have suggested that linezolid dose optimization based on its plasma concentration may be effective [14,15,16]. The pharmacokinetic (PK)/pharmacodynamic parameter of linezolid associated with effectiveness is the area under the concentration curve (AUC)/minimum inhibitory concentration [17, 18]. However, details of the concentrations and PK parameters associated with the safety evaluation of linezolid have not been clarified. In general, the trough concentration or AUC is used to evaluate the safety of antimicrobials. Although association of the trough concentration or AUC with the safety of linezolid has been frequently reported, it is unclear whether trough concentration or AUC is a suitable PK parameter for safety evaluation; furthermore, the appropriate range has yet to be determined. Systematic reviews and meta-analyses have recommended using vancomycin for safety monitoring cases with an AUC of 400–600 mg × h/L [19, 20]. However, no systematic review or meta-analysis has explored the concentrations or PK indices associated with linezolid safety.

Therefore, this meta-analysis aimed to determine whether hematological toxicity has a high incidence in patients with DRF. To avoid adverse events, we also performed a systematic review to evaluate linezolid’s monitoring parameters and ranges.

Methods

Search strategies

Search strategy for the evaluation of linezolid-associated hematotoxicity in patients with DRF

PubMed, Web of Sciences, Cochrane Register of Controlled Trials, and ClinicalTrials.gov databases were searched for relevant studies published before June 24, 2022. Two of four reviewers (MA, CI, RS, and TN) independently searched databases for literature using the following research terms: “linezolid,” “renal,” “kidney,” “thrombocytopenia,” “anemia,” “neutropenia,” “myelosuppression,” “leucopenia,” and “hematotoxicity.” The publication language was limited to English, and there was no restriction on the publication year. Duplicate articles were excluded.

Search strategy for the evaluation of linezolid monitoring and ranges

We similarly searched PubMed, Web of Sciences, Cochrane Register of Controlled Trials, and ClinicalTrials.gov databases for relevant studies published before June 24, 2022. Two of the four reviewers (MA, CI, RS, and TN) independently searched for literature using the following research terms: “linezolid,” “monitoring,” “area under the curve,” “trough,” and “therapeutic drug monitoring.” The publication language was limited to English, and there was no restriction on the publication year. Duplicate articles were excluded from the study.

Study selection

Study selection for the evaluation of linezolid-associated hematotoxicity in patients with DRF

Two of the four reviewers (XL, MA, SO, and RS) independently screened the extracted literature. A study was considered eligible for evaluation in this meta-analysis provided that it met the following inclusion criteria: (1) the study included patients with and without DRF; (2) the study included patients who received linezolid treatment; and (3) the study revealed outcomes corresponding to hematotoxicity (thrombocytopenia, anemia, neutropenia, myelosuppression, and leukopenia). Studies that met the following criteria were excluded: (1) studies involving cells or animal models; and (2) case reports, case series, or reviews.

Study selection for the evaluation of linezolid monitoring and ranges

Two of the four reviewers (XL, MA, SO, and TN) independently screened the literature. A study was considered eligible for evaluation in this systematic review provided that it met the following inclusion criteria: (1) the study revealed the AUC or trough values of patients; (2) the study included patients who received treatment with linezolid; and (3) the study revealed the outcomes of thrombocytopenia.

Data extraction

Data extraction for the evaluation of linezolid-associated hematotoxicity in patients with DRF

Two of the four reviewers (XL, SO, CI, and RS) independently extracted data from the studies. The study period, study design, country of the study, age and weight of the patients, definition of hematotoxicity, definition of DRF, and patients with and without DRF (patients with or without hematotoxicity were counted separately) were extracted according to the predefined eligibility criteria.

Data extraction for the evaluation of linezolid monitoring and ranges

Two of the four reviewers (XL, SO, CI, and RA) independently extracted data from the studies. The study period, study design, country of study, age of the patients, and AUC or trough values were extracted.

Outcome analysis

Outcome analysis for the evaluation of linezolid-associated hematotoxicity in patients with DRF

The primary outcome was the incidence rate of hematotoxicity. The rate of hematotoxicity was defined according to each study’s definition. Subgroup analysis was performed according to the classification of hematotoxicity, including thrombocytopenia, anemia, pancytopenia, and myelosuppression.

Outcome analysis for the evaluation of linezolid monitoring and ranges

The primary outcome was the incidence of thrombocytopenia determined according to AUC24 (calculated by AUC12 if unavailable) and Cmin (minimum blood plasma concentration) in children and adults.

Assessment of the risk of bias

Two of the four reviewers (XL, SO, CI, and RA) independently assessed the risk of bias based on Cochrane Collaboration (Risk Of Bias In Non-Randomized Studies of Interventions, ROBINS-I) [21]. Discrepancies were resolved by discussion or consultation with the third reviewer (YE).

Assessment of quality of evidence

The GRADE handbook was used to rate the grade quality of the meta-analysis [22]. GRADE specifies that the quality of the evidence can be classified into four categories according to the corresponding evaluation criteria: (1) high (⊕⊕⊕⊕); (2) moderate (⊕⊕⊕⊖); (3) low (⊕⊕⊖⊖); and (4) very low (⊕⊖⊖⊖).

Analysis of the results and statistical analyses

The Review Manager for Windows (RevMan, Version 5.4, Copenhagen: The Nordic Cochrane Centre, The Collaboration, 2020) was used for data analysis and the preparation of forest plots. We used random-effects model for pooling study results. We calculated odds ratios (OR) with 95% confidence intervals (CIs) for discrete variables. To assess heterogeneity, I2 was calculated. Finally, funnel plots were constructed to assess potential publication bias.

Protocol registration

The present study was not registered with Prospero or elsewhere.

Results

Search results

In the database search for the evaluation of linezolid-associated hematotoxicity, 1213 articles were screened after duplicates were extracted (Fig. 1A). Twenty-five articles [9,10,11,12,13, 23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42] were included for the evaluation of linezolid-associated hematotoxicity.

Fig. 1
figure 1

Flow chart of the study selection. Flow chart of A meta-analysis of hematotoxicity associated with linezolid, and B systematic review of hematotoxicity associated with the linezolid area under the concentration curve or Cmin (minimum blood plasma concentration)

Full size image

In the database search for the evaluation of linezolid monitoring and ranges, 1087 articles were screened after exclusion of duplicates (Fig. 1B). Twenty-seven articles [16, 23, 25, 43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66] were included in the evaluation of linezolid monitoring strategies.

Characteristics

The characteristics of the 25 studies included in the meta-analysis for evaluating linezolid-associated hematotoxicity are shown in Table 1. These studies included 3831 patients, 1240 of whom had DRF. The definitions of DRF and hematotoxicity in each study are shown in Table 1. Most studies were conducted in Asian countries (16 of 25 studies). Twenty-three studies were retrospective, and two studies [25, 37] were prospective studies with a small number of cases conducted in Japan. Thrombocytopenia, anemia, pancytopenia, and reduction in neutrophils corresponded to hematotoxicity.

Table 1 Characteristics of the studies included in the meta-analysis
Full size table

The characteristics of the 27 systematically reviewed studies are shown in Tables 2, 3, 4 and 5. Tables 2 and 3 show studies that evaluated the incidence of thrombocytopenia associated with AUC values in children and adults, respectively. In the analysis of AUC values associated with thrombocytopenia, two studies involved children (Table 2), and 15 studies involved adults (Table 3). A total of 230 patients (including eight children) were included in the analysis. All studies analyzing AUC values associated with thrombocytopenia in children were prospective studies. Of the 15 adult studies, two were retrospective studies, while 12 were prospective studies, on the analysis of AUC values associated with thrombocytopenia in adults. The National Institute of Allergy and Infectious Diseases (NIAID) study in 2018 was a clinical trial.

Table 2 Characteristics of the studies included in the systematic review about AUC (children)
Full size table
Table 3 Characteristics of the studies included in the systematic review AUC (adults)
Full size table
Table 4 Characteristics of the studies included in the systematic review about Cmin (children)
Full size table
Table 5 Characteristics of the studies included in the systematic review about Cmin (adults)
Full size table

Tables 4 and 5 list studies that evaluated the incidence of thrombocytopenia associated with Cmin in children and adults, respectively. In the analysis of Cmin associated with thrombocytopenia, three studies included children (Table 4), and 17 studies included adults (Table 5). Two of the three studies were prospective in the analysis of Cmin associated with thrombocytopenia in children. Twelve of the 14 studies were prospective studies that analyzed Cmin associated with thrombocytopenia in adults.

Outcome analysis for the evaluation of linezolid-associated hematotoxicity in patients with DRF

Twenty-three retrospective studies and two prospective studies with 1240 patients with DRF and 2591 patients without DRF were enrolled in the meta-analysis. Compared with patients without DRF, patients with DRF had a significantly higher incidence of hematotoxicity (OR = 2.37; 95% CI: 1.93–2.90; p < 0.001; I2 = 33%) (Fig. 2).

Fig. 2
figure 2

Forest plot of the hematotoxicity associated with linezolid treatment with or without decreased renal function. Vertical line indicates no significant differences between the groups. Diamond shapes and horizontal lines indicate odds ratios and 95% confidence intervals, respectively. Squares indicate point estimates and the size of each square indicates the weight of each study

Full size image

We also conducted a subgroup analysis based on the classification of hematotoxicity. The incidences of thrombocytopenia (OR = 2.45; 95% CI: 1.95–3.09; p < 0.001; I2 = 36%) and anemia (OR = 2.31; 95% CI: 1.27–4.21; p = 0.006; I2 = 29%) were significantly higher in patients with DRF than in those without DRF (Fig. 3A and C). However, no significant differences were observed in the incidence of pancytopenia (OR = 1.41; 95% CI: 0.10–20.72; p = 0.80, I2 = 65%) in patients with and without DRF (Fig. 3B).

Fig. 3
figure 3

Forest plot of the subgroup analysis of the hematotoxicity classification associated with linezolid treatment with or without decreased renal function. Vertical line indicates no significant differences between the groups. Diamond shapes and horizontal lines indicate odds ratios and 95% confidence intervals, respectively. Squares indicate point estimates and the size of each square indicates the weight of each study. Subgroup analysis of A anemia; B pancytopenia; and C thrombocytopenia

Full size image

Outcome analysis for AUC values and the incidence of thrombocytopenia

No confidential threshold values for the development of thrombocytopenia were found in AUC values for children or adults (Tables 2 and 3). Only four studies reported the AUC values for patients with thrombocytopenia, and the values were 180.5 [44] 243 [49], 280.74 [16], and 175.0 or 345.8 [66] mg × h/L. Thrombocytopenia did not occur when the mean or median AUC24 (calculated by AUC12 if it was not available) was within 95.2–328.3 mg × h/L in adults (Table 3).

Outcome analysis for Cmin and the incidence of thrombocytopenia

Twelve studies reported the incidence of thrombocytopenia. In the analysis for children, two studies revealed the incidence of thrombocytopenia, and the Cmin values of thrombocytopenia and non-thrombocytopenia were 4.7–7.17 and 0.1–4.6 μg/mL, respectively. One patient with a Cmin value of 4.7 μg/mL received high-dose methotrexate in combination treatment. In the adult analysis, 10 studies revealed the incidence of thrombocytopenia, and the Cmin values of thrombocytopenia and non-thrombocytopenia were 4.28–67.7 and 0.2–5.8 μg/mL, respectively. In seven studies, Cmin for patients without thrombocytopenia was not determined. Except for a Cmin of 4.28 μg/mL, thrombocytopenia occurred at Cmin values of > 6–7 μg/mL.

Publication bias

Funnel plots of the incidence of hematotoxicity are shown in Fig. 4. The funnel plots were symmetric and did not suggest the presence of publication bias in favor of a positive study for all outcomes.

Fig. 4
figure 4

Publication bias plot of 16 trials of linezolid-associated hematotoxicity in patients with decreased renal function

Full size image

Assessment of the risk of bias

The results of the assessment of the risk of bias are presented in Figs. S1 and S2. A high risk of confounding bias was found in the study by Hiraki et al. [25]. Information regarding selection bias was unavailable for most studies; few studies identified bias issues. No problems in intervention bias were identified, and moderate missing data bias was identified in the study by Choi 2019. Three studies [30, 33, 40] had a moderate risk of measurement of outcome bias. No information was available for deviation from the intended intervention and reporting biases.

Quality of the evidence

The results of the quality evaluation according to the GRADE guideline are shown in Table 6. This meta-analysis consisted primarily of observational studies, so there was a low initial rating. Some problems in the risk of bias downgraded the quality of evidence by one level, while a large magnitude of effect upgraded the quality of evidence by one level. The low final grade of the evidence shows that our confidence in the effect estimate is limited.

Table 6 GRADE assessment of meta-analysis
Full size table

Discussion

In this meta-analysis of retrospective and prospective studies, the incidence of hematotoxicity was significantly higher in patients with DRF than in those without. In addition, subgroup analysis revealed a significant difference in the incidence of thrombocytopenia and anemia, but there was no significant difference in the incidence of pancytopenia (Fig. 3A–C). These results suggest that linezolid should be cautiously administered in patients with DRF while monitoring for hematotoxicity, especially thrombocytopenia and anemia.

Clinical phase III trials have reported a 2.4% incidence of thrombocytopenia in patients receiving linezolid therapy [67]. In our meta-analysis, the incidence of thrombocytopenia in patients with and without DRF ranged between 28.9 and 78.6% (except for the study by Hiraki et al. [25]) and 10.5 and 42.9%, respectively, which were higher than those previously reported. Nearly all the patients included in this meta-analysis were from Asian countries, such as Japan, China, and Korea, and had lower body weights than those of individuals from Western countries. Previously, lower body weight was considered a risk factor for thrombocytopenia [23]. Generally, linezolid was administered twice daily (600 mg × 2) and the dose was not adjusted by body weight. A comparison of the median weights among the groups that received linezolid treatment showed that the median weight was 80 kg when the AUC was 95.2 mg × h/L [53] and 58.3 kg when the AUC was 291.6 mg × h/L [45]. The difference in AUC values may be accounted for by the difference in the dose per body weight. Additionally, advanced age [68] and the duration of administration [69] are also considered risk factors; therefore, this difference in the patients’ backgrounds may explain the higher incidence of hematotoxicity.

A major reason for the higher incidence of thrombocytopenia in patients with DRF than in patients without DRF is the delayed excretion of linezolid and increased blood linezolid concentrations. Approximately 30% of linezolid is excreted by the kidneys of patients with normal renal function [70]. Furthermore, Matsumoto et al. evaluated the clearance of linezolid with renal function and reported a correlation between linezolid and creatinine clearance or blood urea nitrogen [69]. Therefore, we hypothesized that linezolid overexposure or higher Cmin is associated with decreased renal function [59, 71].

In this meta-analysis, no significant differences were observed in the incidence of pancytopenia. This result does not indicate the absence of a relationship between DRF and the incidence of pancytopenia, as the number of cases included in the systematic review was notably smaller than that of thrombocytopenia. In addition, many studies have focused on thrombocytopenia, which occurs most frequently among the different forms of hematotoxicity (Sheldon et al. 2003 [5];). Therefore, it might have been easier to identify significant differences in thrombocytopenia. If more studies on pancytopenia are published in the future, significant differences in the incidence of pancytopenia will be found.

The incidence of thrombocytopenia was higher when the Cmin of linezolid exceeded 6–7 μg/mL (Tables 4 and 5). Previous studies revealed the efficacy and safety ranges of linezolid trough values as 2–8 μg/mL [15, 16, 62, 72], 3.6–8.2 μg/mL [61], and 2–7 μg/mL [73]. In this study, we conducted a systematic review of the incidence of thrombocytopenia and Cmin in children and adults, as determined by the extracted Cmin threshold; the incidence of thrombocytopenia was higher when the Cmin exceeded 6–7 μg/mL. However, this systematic review could not determine the clinically relevant threshold of linezolid in terms of the AUC (Tables 2 and 3). Matsumoto et al. reported a strong correlation between AUC and trough concentrations [61]. Only four studies reported the AUC values for patients with thrombocytopenia in this study.

Further studies are required to determine the target AUC that correlates with thrombocytopenia. However, it is difficult to measure the AUC in clinical settings; therefore, Cmin may be a surrogate index of AUC in clinical practice. Consequently, we believe that therapeutic drug monitoring should be performed for linezolid administration from the perspective of safety and that the dose should be controlled to achieve a target trough value of < 6–7 μg/mL.

The previous meta-analysis showed that impaired renal function was associated with an increased risk of linezolid-induced thrombocytopenia [74]. Based on this knowledge, finding an association between hematotoxicity and patients with DRF, we classified hematotoxicity and performed a subgroup analysis, which showed that thrombocytopenia and anemia were significantly higher in patients with DRF than in those without DRF. We also conducted a systematic review and determined that hematotoxicity was higher when Cmin exceeded 6–7 μg/mL. This finding is a strength of the current study. To our knowledge, this study is the first systematic review to explore the association of Cmin with linezolid safety. This result may serve as an indication for the implementation of therapeutic drug monitoring and provide insights for further clinical trials.

This study had several limitations. First, most of the analyzed studies were observational studies. Therefore, the patient characteristics and study designs contained various types of bias, hindering their results’ generalizability. Second, the definitions of thrombocytopenia were different in these studies. Third, the estimation method of AUC differed in each study. This might have led to a misunderstanding of our results. However, this analysis did not clarify the target AUC due to the limited number of studies.

Conclusion

Decreased renal function correlates with an increased risk of thrombocytopenia and anemia due to overexposure. To maximize the efficacy and minimize the toxicity of linezolid, therapeutic drug monitoring should be recommended, using evidence-based thresholds in patients on long-term linezolid treatment or in patients with DRF.