Factors Affecting Pharmacokinetics and Dosing Strategies of Polymyxin B in Multidrug-resistant Gram-negative Infected Patients with Renal Impairment: A Systematic Review with Expert Opinion

Abstract

Background and Objective

Multidrug-resistant (MDR) Gram-negative infections pose significant therapeutic challenges, with limited evidence guiding optimal polymyxin B dosing in patients with renal impairment. This systematic review aimed to identify key covariates influencing polymyxin B pharmacokinetic and inform dosing strategies in this population.

Methods

A comprehensive search of PubMed, Scopus, Embase, and Cochrane library was conducted from inception to May 2025. Population–pharmacokinetic (pop-PK) studies evaluating polymyxin B in adult patients in adult patients with renal impairment, with or without renal replacement therapy (RRT) were included. Study quality was assessed using a validated critical appraisal of clinical pharmacokinetic studies tool (CACPK).

Results

Eleven pop-PK studies (603 patients) were included. Considerable interindividual variability in clearance and volume of distribution was noted. Creatinine clearance (CrCL), RRT modality, and disease severity were key covariates influencing drug exposure. Convective RRT modalities (CVVH, CVVHDF) were associated with increased clearance and lower exposure, whereas diffusive modalities (CVVHD) showed variable effects. However, the clinical relevance of CrCL-based dose adjustment alone remains uncertain and should be interpreted alongside other clinical factors.

Conclusion

Polymyxin B dosing in renal impairment requires an individualized approach integrating renal function, RRT modality, and disease severity. While a loading dose of 150 mg is appropriate, maintenance dosing should be guided by clinical context, with higher doses in convective RRT and cautious dosing in diffusive modalities. Therapeutic drug monitoring (TDM), where available, is recommended to optimize efficacy and minimize toxicity.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s13318-026-01021-1.

Key Points

Polymyxin B pharmacokinetics in renal impairment show high interindividual variability, influenced by creatinine clearance (CrCL), disease severity and renal replacement therapy (RRT) modality.

Convective RRT modalities are associated with increased drug clearance requiring higher doses, whereas diffusive modalities show variable exposure requiring cautious dosing.

Individualized dosing supported by therapeutic drug monitoring (TDM) is essential to balance efficacy and toxicity in this population.

Introduction

Despite renewed interest in polymyxins for treating multi-drug resistant (MDR) Gram-negative infections, optimal dosing in patients with renal impairment remains unclear [1]. The United States Food and Drug Administration recommends 15,000–25,000 IU/kg/day (1.5–2.5 mg/kg/day) for patients with normal kidney function, with dose adjustments suggested for renal impairment. However, these guidelines lack robust evidence [2]. In contrast, international consensus guidelines do not recommend dose adjustments in renal impaired patients [3], as no consistent correlation between polymyxin B clearance and renal function has been demonstrated [4, 5]. Most supporting studies investigated doses below 2.5 mg/kg/day, limiting applicability to higher conventional dosing (2.5–3 mg/kg/day) [5–9]. Consequently, existing evidence does not fully support current recommendations of polymyxin B dosing [1].

The incidence of polymyxin-associated acute kidney injury (AKI) varies widely, ranging from 17% to 51%, depending on clinical definitions and the formulations used [10]. This high rate of nephrotoxicity impacts patient outcomes and carries economic implications [10]. Studies suggest a strong association between inappropriate polymyxin B dosing and renal impairment [11]. The absence of uniform dosing strategies results in unpredictable pharmacokinetics, complicating therapeutic management in patients with impaired renal function [12]. In extracorporeal therapies, factors such as continuous renal replacement therapy (CRRT) techniques, protein binding, dialysis dose, and polymyxin B’s physicochemical properties (e.g., molecular size, lipophilicity) significantly impact drug clearance [13]. Without a standardized dosing protocol, overexposure in renal-impaired patients increases nephrotoxicity risk, while underdosing compromises efficacy and may lead to treatment failure [14].

Few population–pharmacokinetic (pop-PK) studies have focused on optimizing polymyxin B dosing in renal impairment or those undergoing renal replacement therapy (RRT) [4, 5, 15–22]. Consensus guidelines target an area under the concentration–time curve across 24 h at steady state (AUC_SS,24h) of 50–100 mg h/L, and an average steady-state concentration (Css.avg) of ~ 2 mg/L to balance efficacy and risk of AKI due to polymyxins [3, 23, 24]. However, significant interindividual variability in polymyxin B pharmacokinetics, especially among patients receiving CRRT, like continuous veno-venous hemodiafiltration (CVVHDF), poses challenges in achieving these therapeutic targets by impacting drug clearance [25].

With the growing reliance on polymyxin B amid rising antimicrobial resistance, and the complex interplay between its pharmacokinetics, drug exposure, and clinical outcome, it becomes crucial to identify significant covariates that impact the optimal dosing strategies of polymyxin B. Considering current evidence, establishing a better-informed uniform dosing strategies for polymyxin B in renal-impaired patients and those on RRT is essential. This systematic review presents an overview of significant covariates affecting the pharmacokinetics of polymyxin B in this patient population. The expert opinion section explores the impact of different covariates on the pharmacokinetics of polymyxin B and its dosing recommendations, offering practical guidance for clinicians and pharmacists in managing polymyxin B therapy for patients with renal impairment or on RRT techniques.

Methods

Protocol Registration

The study protocol was registered in PROSPERO with registration number CRD42021294141 [26]. We used a participants, intervention, comparator, outcomes, and study design (PICOS) framework to establish inclusion and exclusion criteria. The review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines [27].

Eligibility Criteria and Selection of Studies

We considered all published retrospective and prospective pop-PK studies reporting the factors affecting pharmacokinetics of polymyxin B, which included patients with renal impairment or on RRT modalities. Studies including adult renal-impaired MDR Gram-negative infected patients, with or without RRT techniques, receiving intravenous polymyxin B for a duration ≥ 48 h (once the drug achieves steady-state) were considered. Outcome of interest was to assess the impact of covariates on pharmacokinetic parameters including clearance (CL), volume of distribution (V_d) and half-life (t_½). Studies that did not include patients with renal impairment, without sufficient data, narrative reviews, and protocols were excluded. Renal impairment was defined according to the criteria used in individual studies, most commonly based on creatinine clearance (CrCL). However, standardized classifications such as kidney disease: Improving Global Outcomes (KDIGO)-defined AKI or chronic kidney disease (CKD) stages were not uniformly reported across studies, limiting the ability to perform stratified analyses. A minimum duration of ≥ 48 h of polymyxin B was required to ensure that pharmacokinetic sampling reflected near steady-state conditions. However, we acknowledge that, in patients receiving CRRT, steady-state assumptions may not always be valid due to dynamic extracorporeal clearance.

Sources of Information

We performed a thorough literature search from inception to May 2025, using relevant MeSH terms and free text keywords of polymyxins and pharmacokinetics across databases such as PubMed/Medline, Scopus, Embase and the Cochrane Library. We also performed a bibliographic search of included studies to identify additional relevant studies. Only studies published in the English language were included. Detailed search strategies for each database can be found in Supplementary table S1.

Study Selection and Data Collection Process

We conducted a title and abstract screening, followed by a full-text assessment based on predefined inclusion and exclusion criteria. Collected data included the following details. Study characteristics: study ID, country, study design, study period, population, sample size; patient characteristics: age, male gender, weight, severity score, number of patients with renal impairment, patients on RRT, dialysis parameters, mean/median CrCL; polymyxin B dosing characteristics: duration of therapy, dose, formulation of drug and infusion time, and adverse effects; analysis and sampling strategy: analytical method, sampling strategy, plasma concentration and half-life of the drug, plasma protein binding; modeling parameters: pharmacokinetic model, software and structure, candidate and significant covariates, AUC in a steady state, and at 0–24 h, V_d, clearance, interindividual and residual variability, internal and external validation. Two reviewers (A.K.R., V.S.) independently performed the study selection and data extraction processes, with any discrepancies resolved through consensus or discussion with another member of the research team (G.T.). In cases of missing or incompletely reported pharmacokinetic data, no imputation was performed and analyses were based on available reported data. Study authors were not contacted for additional information, as the review focused on published aggregate data. No restrictions were applied on publication date, and all eligible studies meeting the predefined criteria were considered.

Critical Appraisal

We utilized the validated critical appraisal of clinical pharmacokinetic studies tool (CACPK) [28] to appraise the included studies. This tool is specifically designed to evaluate pharmacokinetic and pop-PK studies, addressing key methodological domains such as sampling design, analytical methods, model development and validation. Conventional risk of bias assessment tools developed for clinical or observational studies may not adequately capture these pharmacometrics specific aspects and therefore were not applied in this review. The CACPK tool comprises of 21 items divided into five sections, each evaluated based on the methodological rigor of the study with responses of “Yes”, “No”, “I do not know”, or “Not Applicable”. A score of 13 or above was deemed high quality, scores between 12 and 13 were considered fair to moderate, and scores below 12 were classified as poor[29, 30]. Two reviewers (A.K.R., V.S.) conducted the quality assessment independently, and any discrepancies were resolved through discussion with third reviewer (G.T.). The use of a domain-specific appraisal tool enhances the methodological relevance and interpretability of the quality assessment in pharmacokinetic research.

Results

Study Selection

A total of 12,703 records were initially identified through database search. The titles and abstracts screening yielded, 421 studies for full-text assessment, from which eleven articles were included in the systematic review. The PRISMA flow diagram illustrating the process of study selection is shown in Fig. 1.

Fig. 1.

Flow diagram illustrating the systematic process of study identification, screening, eligibility assessment, and inclusion for the present review. Adapted from Page et al. [27]

Critical Appraisal of the Included Studies

The quality of all 11 pop-PK studies included in this review was deemed high (> 14). The average score of the included studies was 16. However, several studies lacked detailed reporting on sampling sites, pharmacokinetic model validation methods, [4, 5, 15, 17, 20, 21], and explicit presentation of pharmacokinetic equations[4, 5, 15, 19–21], which should be considered when interpreting the robustness and reproducibility of these models. While these reporting limitations primarily affect transparency, they may introduce some uncertainty in the interpretation and generalizability of the findings. The critical appraisal of the included studies is provided in Supplementary Table S2.

Characteristics of Included Studies

All 11 studies [4, 5, 15–22, 31] were published between 2013 and 2024, with the majority from China (n = 7) [16, 18–22, 31], followed by Brazil (n = 2) [4, 15] and the United states of America (n = 2) [5, 15]. Of these, nine studies were prospective [4, 5, 15–19, 21, 31] and two were retrospective [20, 22] in nature. All the included studies used polymyxin B for infections due to Klebsiella pneumoniae, Acinetobacter baumannii, Enterobacterales, and Pseudomonas aeruginosa. Sample sizes ranged from 24 to 142 patients. Detailed characteristics of the studies are provided in Table 1.

Table 1.

Characteristics of included population pharmacokinetic studies evaluating polymyxin B in sepsis patients with renal impairment

Study ID; Country	Study design	Study period	Population	Sample size	Key findings
Wang et al., 2024; China	Single center prospective	April 2018 to March 2022	Critically ill patients with sepsis or septic shock with multidrug-resistant Gram-negative bacterial infections	83 patients, 466 samples	Fluid management as well as renal function are essential factors affecting polymyxin B PK among sepsis patients.
Hanafin et al., 2023; U.S.A, Singapore, Brazil	Multi-center prospective	March 2016 to June 2021	Hospitalized patients with definitive or empirical therapy for sepsis	142 patients, 681 plasma samples	For patients undergoing CVVHDF, higher doses of polymyxin B may be necessary, along with close monitoring of drug levels to ensure treatment.
Pi et al., 2023; China	Prospective	July 2021 to March 2022	Critically ill patients with sepsis, septic shock or severe infection	30 patients, 212 samples	Dosing should be tailored to the patient’s body weight and the specific CRRT modality. Re-evaluation of dosing adjustment is necessary for patients on CRRT.
Surovoy et al., 2023; Russia	Prospective	May 2021 to Feb 2022	Sepsis patients with varying degree of renal function with secondary pneumonia or BSI due to carbapenem resistant strains	37 patients	Patients with renal insufficiency experience reduced polymyxin B clearance, increasing the risk of toxicity. For those on CVVHD, a fixed dose of 100 mg twice daily is recommended.
Luo et al., 2022; China	Prospective	Jan 2021 to Nov 2022	Critically ill patients with confirmed or suspected infection due to carbapenem resistant organisms	63 patients, 189 samples Model dataset: 49 patients, 147 samples	In patients receiving CRRT, polymyxin B dosage should not be reduced. Instead, adjustments should be based on renal function or SOFA scores.
Wang et al., 2022; China	Multicenter prospective	April 2018 to June 2021	Critically ill patients with multidrug-resistant Gram-negative bacteria on CVVH	53 patients, 275 samples	Patients on CVVH may require higher doses of polymyxin B, with dose adjustments guided by therapeutic drug monitoring.
Li et al., 2021; China	Prospective	NA	Renal transplant patients	50 patients, 151 plasma sample	Renal function significantly affects polymyxin B clearance, necessitating adjustments in the dosage regimen.
Wang et al., 2021; China	Retrospective	April 2018 to March 2020	Patients with multidrug resistant Gram-negative bacterial infection with or without renal insufficiency	70 patients, 462 samples NRF-37 (CLCR≥ 80ml/min) RI (CLCR< 80 ml/min)−33	Renal dosing adjustment of polymyxin B may not be necessary
Yu et al., 2020; China	Retrospective	June 2018 to May 2019	Patients with various renal function	32 patients, 112 samples	Renal function significantly influences polymyxin B clearance, necessitating dosage adjustments based on CrCL in patients with renal insufficiency.
Thamlikitkul et al., 2017; Thailand, U.S.A	Prospective, multicenter, open-labelled study	October 2014 to June 2016	Patients with renal insufficiency with suspected or documented infections	19 patients, NRF-5 RI-14	Polymyxin B dosing with renal insufficiency patients’ needs to be re-examined
Sandri et al., 2013; Brazil, Australia	Multicenter study	March 2011 to Jan 2012	Critically ill patients	24 patients,	Polymyxin B dosage should be based on total body weight rather than renal function.

AKI acute kidney injury, CL_CR creatinine clearance, CRRT continuous renal replacement therapy, CVVHDF continuous veno-venous hemodiafiltration, CVVH continuous veno-venous hemodiafiltration, CVVHD continuous veno-venous hematolysis, MD maintenance dose, NRF normal renal function, RI renal impairment, SOFA sequential organ failure assessment

Characteristics of the Included Patients

Populations among the included studies considered the patients who are critically ill, [4, 16, 18, 19, 31] with sepsis, [16, 31] hospitalized, [15] non-critically ill sepsis, [17] on empirical or definite therapy with polymyxin B for MDR Gram-negative infections [5, 20, 24], and renal transplant patients [21]. Mean age of the patients ranged from 43.5 to 68.9 years, with a majority of male patients (n = 411). The criteria for renal dysfunction were based on author’s discretion from each study. Most studies included patients who were on RRT (n = 7). including patients undergoing various CRRT [4, 21] techniques, such as continuous veno-venous hemofiltration (CVVH) [16, 18, 19], continuous veno-venous hematolysis (CVVHD) [15–17], CVVHDF [15, 16], sustained low efficiency daily dialysis (SLEDD) [15], and hemodialysis (HD) [15]. Detailed patient characteristics are provided in Table 2.

Table 2.

Baseline demographic, clinical and renal characteristics of patients included in the studies

Study ID	Age (years)	Male (n)	Weight (kg)	Severity score	No. of patients on RRT	Dialysis parameters	Patients with renal dysfunction	CrCL (ml/min)
Wang et al., 2024	52 (18–94)	67	65 (45–110)	SOFA: 8 (3: 17) APACHE II: 18 (6–35)	NA	NA	20	87.5 (17.3–309.7)
Hanafin et al., 2023	65 (19–94)	91	71 (32.5–130)	APACHE II– 17 (0–41)	CVVHD: 3 CVVHDF: 13 HD: 2 SLED: 6	NA	24	55.2 (8.58–322)
Pi et al., 2023	55 (21–82)	24	63.5 (40–90)	APACHE II: 22 (12–34) SOFA: 10 (2–19)	CVVHDF: 8 CVVH: 10 CVVHD: 2	Blood flow rate: 150–200 mL/min Dialysate rate or ultrafiltration rate: 2–3 L/h	20	104.3 ± 69.93
Surovoy et al., 2023	NRF: 63.3 ± 3 RI: 68.9 ± 3 CVVHD: 68.7 ± 3.4	28	NRF: 98 ± 4.2 RI: 87.7 ± 3.6 CVVHD: 94 ± 3.5	SOFA–CVVHD: 12 ± 0.4 RI: 7.3 ± 0.8	CVVHD: 11	Dialysis flow rate: 2720 ± 180 Ultrafiltration rate: 123 ± 32 Dialysis dose: 29 ± 1.8	24	NRF: 112 ± 7 RI: 48.9 ± 6 CVVHD: NA
Luo et al., 2022	65 (27–93)	33	60 (41–91)	APACHE II: 20 (7–35) SOFA score: 6 (0–22)	CVVHF: 11	Blood flow: 150–250 ml/min Dialysate flow rate: 2–3 L/h	11	83.69 (13.48–197.84)
Wang et al., 2022	53.53 ± 17.95	29	66.94 ± 15.31	SOFA score: 10.47 ± 3.26	CVVH: 53	Filtration time: 41.0 ± 12.76 Blood flow rate: 160.0 ± 23.24 Post-filter rate: 1.81 ± 0.32 Fluid removal: 203.68 ± 77.04	53	51.11 ± 34.86
Li et al., 2021	43.5 (18–66)	32	ABW: 57.8 ± 12.4	NA	CRRT: 11	NA	46	22.2 (4.29–90.7)
Wang et al., 2021	NRF: 47.3 ± 17.7; RI: 54.2 ± 17.5	NRF: 31 RI: 27	NRF: 68.6 ± 11.6 RI: 66.9 ± 11.1	NA	NA	NA	33	NRF:123.3 (81.6–315.2) RI: 42.0 (15.6–77.6)
Yu et al., 2020	63.63 (12.92)	26	61.73 ± 11.77	NA	NA	NA	20	86.58 ± 53 CrCL<20: 3 20≤ CrCL<60: 6 60≤ CrCL<90: 11 90≤CrCL: 12
ThamlikitkulV et al., 2017	NRF: 61.0 ± 9.4 RI: 62.8 ± 14.3	NRF: 3 RI: 7	NRF: 73.6 ± 25.7 RI: 54.8 ± 10.4	NA	NA	NA	14	NRF: 90.0 ± 12.5 RI: 40.8 ± 21.8
Sandri et al., 2013	61.5 (21–87)	13	TBW: 62.5 (41.250) LBW: 46 (29–99)	APACHE II:21.5 (10–29)	CRRT: 2	NA	2	33 (10–143)

ABW absolute body weight, AKI acute kidney injury, APACHE II acute physiology and chronic health evaluation, CL_CR creatinine clearance, CRRT continuous renal replacement therapy, CVVHDF continuous veno-venous hemodiafiltration, CVVH- continuous veno-venous hemodiafiltration, CVVHD continuous veno-venous hematolysis, MD Maintenance dose, NRF normal renal function, RI renal impairment, SOFA sequential organ failure assessment, TBW total body weight

Characteristics of Polymyxin B Dosing

Five studies [5, 18, 20, 21, 31] reported the duration of polymyxin B therapy, with median duration ranging from 9 to 14 days. Patients were included after 48 h of initiating polymyxin B therapy. Polymyxin B infusion time ranged between 1 and 6 h. The median maintenance dose of polymyxin B varied between 2 and 3 mg/kg/day. Detailed information on polymyxin B dosing characteristics is presented in Table 3.

Table 3.

Polymyxin B dosing regimens, treatment duration, and adverse events reported in the included studies

Study ID	Formulation of drug and infusion time	Polymyxin B dose	Polymyxin B duration (days)	Adverse effects
Wang et al., 2024	IV Polymyxin B ≥ 96 h, 1 (0.5–2) h infusion	LD: 150 (60–200) MD: 150 (60–200)	14 (4–44)	NA
Hanafin et al., 2023	IV Polymyxin B ≥48 h, 1–6 h infusion	200 mg (80–400) 2.56 mg/kg (1.33–6.00) BD and OD dosing	NA	NA
Pi et al., 2023	IV polymyxin B ≥ 48 h	LD: 150 mg (100–200): 20 patients MD: 75 mg (50–100): 30 patients BD dosing	NA	NA
Surovoy et al., 2023	IV polymyxin B ≥ 48 h	LD: 200–300mg MD: 100–150mg NRF: 2.8 (2.3–4) mg/kg/day RI: 2.6 (2–3.34) CVVHD: 2.5 (2–3) BD dosing	NA	NA
Luo et al., 2022	IV Polymyxin ≥ 48 h	LD: 1.73 (1.25–2.73) MD: 2.22 (1.52–3.00) BD dosing	12 (4.28)	NA
Wang et al., 2022	IV Polymyxin B ≥ 48 h	LD: 100–200 mg MD: 50–100 mg; 1.94 ± 0.53 mg/kg BD dosing	NA	NA
Li et al., 2021	IV Polymyxin B ≥ 48 h; 1–2 hrs infusion	MD: 50 mg q12h: 20 40 mg q12h: 22 Other doses: 8	9 (5–52)	Nephrotoxicity: 5 (risk:4; injury: 1) Neurotoxicity: 31 (facial and peripheral paresthesia: 29; seizures: 3; dizziness: 3) Dermatitis: 17 Pigmentation: 7
Wang et al., 2021	IV Polymyxin B ≥ 72 h; ≥1 h	Daily dose/body weight: 2.0 ± 0.5	NRF: 14 (4–54) RI: 12 (4–44)	AKI: 23 (risk: 11; injury: 7; failure: 5) NRF: 15 RI: 8
Yu et al., 2020	IV Polymyxin B ≥ 48 h	100–200 mg/day Daily dose: 136.6 ± 25.48 Dose/weight: 2.26 ± 0.49 BD dosing	NA	NA
Thamlikitkul et al., 2017	IV Polymyxin B ≥ 48 h, intermittent infusions	Dose/weight: NRF: 2.2 ± 0.2 RI: 1.9 ± 0.3 BD and OD dosing	NRF: 12.4 ± 2.2 RI: 12.6 ± 6.2	NA
Sandri et al., 2013	IV Polymyxin B ≥ 48 h; short term infusion (1–4 hrs)	Dose/weight: 0.45–3.38 BD and OD dosing	NA	NA

AKI acute kidney injury, CRRT continuous renal replacement therapy, CVVHDF continuous veno-venous hemodiafiltration, CVVH continuous veno-venous hemodiafiltration, CVVHD continuous veno-venous hematolysis, MD Maintenance dose, NRF normal renal function, q12h dosing every 12 h, RI renal impairment

Analysis and Sampling Strategy of Polymyxin B

Among the seven studies [4, 15–17, 19–21] that reported polymyxin B plasma concentrations and AUC levels, substantial variability was observed in the drug’s half-life across different patient populations and RRT modalities. In patients with normal renal function, polymyxin B exhibited a half-life of approximately 26.2 h, consistent with prior pharmacokinetic data [17]. In contrast, patients with renal impairment not undergoing dialysis had a shorter reported half-life of ~ 13.6 h [15]. Although seemingly counterintuitive, this finding may reflect differences in study design, sampling windows, or altered pharmacokinetic behavior due to impaired renal function without dialysis-mediated clearance. Potential factors include tissue distribution, altered protein binding or compensatory tubular reabsorption in non-dialyzed renally impaired patients.

In patients receiving CVVHDF, the half-life was reduced to approximately 6.9 h, suggesting enhanced drug clearance via combined convective and diffusive mechanisms [15]. Conversely, in those receiving CVVHD, the half-life was around 24.6 h, comparable to patients with preserved renal function [17] indicating a lesser effect of diffusive clearance alone. However, in CVVHD patients with underlying renal impairment, the half-life was prolonged to 36.4 h, underscoring the complex interplay between baseline renal function, dialysis modality and extracorporeal drug clearance [17]. A summary of the analytical methods, sampling strategy, plasma concentration, plasma protein binding and half-life of polymyxin B is given in Table 4.

Table 4.

Analytical methods, sampling strategies, pharmacokinetic parameters and exposure metrics of polymyxin B in the included studies

Study ID	Analytical method	Plasma protein binding	Sampling strategy	Plasma concentration and half-life	AUC parameters (mg h/L)	Equation
Wang et al., 2024	UPLC-MS/MS	NA	Intensive (6 samples)	NA	NA	Vci (L) = tvVc ($\frac{\mathit{DFI}}{4.23}$)dVcdDFI x exp (ƞVc) Vpi (L) = tvVp x exp (ƞVp) Cli (L/h) = tvCL ($\frac{\mathit{CrCL}}{87.46}$)dCLdCrCL ($\frac{\mathit{DFI}}{4.23}$)dCLdDFI x exp (ƞCL) Qi (L/h) = tvQ x exp (ƞQ)
Hanafin et al., 2023	UPLC-MS/MS	64.6% (39.4–86)	Sparse (4–5 samples)	T½: CVVHDF patients: 6.92 (5.7–24.8) Non-RRT patients: 13.6 (5.54–32.1)	AUC0–24 All patients: 112 ± 57.2 Non-RRT patients: 118 ± 56.8 RRT patients: 75.1 ± 45.6 CVVHDF patients: 29.9	CLi = 1.19. ($\frac{\mathit{CrCli}}{55.2}$) 0.243. $e^{\eta \text{i}}$ CL Vc,i = 11.8. ($\frac{\mathit{BWi}}{71}$)1.$e^{\eta \text{i}}Vc$
Pi et al., 2023	HPLC-MS/MS	NA	Intensive (8 samples)	T½: All patients: 5.29 (0.72–21.54) CSS,avg: All patients: 1.48 (0.34–6.33) Cmax: All patients: 4.59 (1.61–14.53) Cmin: (20 patients): 1.08 (0.17–4.73)	AUCSS,0–12 −30 patients: 17.73 (4.10–76.01) CRRT patients: 10.79 Non-RRT patients: 42.11	CL (L/h) = 1.672 x (weight/63.817) ^2.031 x exp [1.114 x (CRRT = 1, not CRRT = 0)] x exp (ȠCL) Q (L/h) = 7.009 x (weight/63.817) ^ 3.408 x exp (ȠQ) V (L) = 7.857 x exp (ȠV) V2 (L) = 12.668 x exp (ȠV2)
Surovoy et al., 2023	ELISA technique	NA	Intensive (6–8 samples)	CSS,avg NRF: 2.8 ± 0.4 RI: 3.7 ± 0.2 CVVHD: 4.6 ± 0.4 Cmax NRF: 5.8 ± 0.6 RI: 5.9 ± 0.4 CVVHD: 8.1 ± 0.9 T1/2β: NRF: 26.2 ± 7.6 RI: 36.4 ± 6.2 CVVHD: 24.6 ± 5.4	AUC0–24 h NRF: 67.3 ± 9.8 RI: 87 ± 5.8 CVVHD: 110.6 ± 10.3	Varying degrees of renal function without CVVHD: CL (L/h) = 2.27 x ($\frac{\mathit{GFR}}{72.7})$ 0.53 x eȠCL V1 (L) = 20.8 x eȠV1 Q (L/h) = 24.3 x e eȠQ V2 (L) = 90 x $(\frac{\mathit{Albumin}}{25.4}$)−3.4 x eȠV2 CVVHD with anuria: CL (L/h) = 1.87 x ($\frac{\mathit{Albumin}}{25.05})$ 1.7 x eȠCL V1 (L) = 19.79 x eȠV1 Q (L/h) = 8.52 x e eȠQ V2 (L) = 50.56 x $(\frac{\mathit{Albumin}}{25.05}$)−2.18 x eȠV2
Luo et al., 2022	HPLC-MS/MS	NA	Sparse (3 samples)	NA	NA	CL (L/h) = ((1-CRRT) x 1.5 x ($\frac{\mathit{CLCR}}{83.69}$)0.506 + CRRT x 1.95) x exp (0.0447 x (SOFA −6) x exp (ȠCL) V1 (L) = 11.7 Q (L/h) = 1.34 V2 (L) = 17.9 x exp (ȠV2)
Wang et al., 2022	UPLC-TMS	NA	Intensive (6–8 samples)	CSSmin: CVVH: 0.54 ±  0.32 Off CVVH: 2.03 ± 1.10	AUCSS,24 h CVVH:27.94 ± 10.92 Off CVVH: 77.89 ± 35.66	NA
Li et al., 2021	HPLC-MS/MS	NA	Sparse (1–6 samples)	0.44 mg/L to 8.15 mg/L	AUC0–24 h: 74.60 ± 17.81	CL (L/h) = 1.18 x (CrCL)/22.2)0.14*exp (ȠCL) V (L) = 12.09*exp (ȠV)
Wang et al., 2021	LC-MS/MS	NA	Sparse (4–7 samples)	CSS,avg : All patients: 3.33 ± 2.20 (0.85–10.55) μg/ml	AUCSS,24 h : NRF:68.63 ± 43.43; RI: 93.04 ± 59.52 AKI: 108.66 ± 70.10 No AKI: 66.18 ± 34.79	NA
Yu et al., 2020	HPLC-MS/MS	NA	Sparse (2–4 samples)	NA	NA	CL (L/h) = 1.59 x ($\frac{\mathit{CrCL}}{80}$)0.408 V (L)= 20.5
Thamlikitkul et al., 2017	UPLC-TMS	NA	Sparse (4 samples)	NA	AUC0–24 h NRF: 63.5 ± 16.6 RI- 56.0 ± 17.5	NA
Sandri et al., 2013	LC-MS/MS	NA	Intensive (8 samples)	CSS,avg : All patients- 2.79 ± 0.90 (0.68–4.88)	AUC0–24 h : All patients 66.9 ± 21.6 (16.4–117)	NA

AKI acute kidney injury, CL clearance, CRRT continuous renal replacement therapy, CVVHDF continuous veno-venous hemodiafiltration, CVVH continuous veno-venous hemodiafiltration, CVVHD continuous veno-venous hematolysis, ELISA enzyme-linked immunosorbent assay, HPLC high performance liquid chromatography, LC-MS/MS liquid chromatography, MD maintenance dose, NRF normal renal function, RI renal impairment, UPLC-TMS ultra performance liquid chromatography, V central volume of distribution, V2 peripheral volume of distribution

Population Pharmacokinetic Modeling

Non-linear mixed effect modeling (NLMEM) was the approach used in 10 of the 11 included studies [4, 15–22, 31]. Eight studies [4, 15–20, 31] employed a two-compartment model and three studies [5, 21, 22] utilized a one-compartment model. Monte Carlo dose simulation was conducted in ten studies [4, 15–22]. A summary of pharmacokinetic parameters of the model is provided in Table 5.

Table 5.

Population pharmacokinetic model characteristics, covariate analysis and parameters estimates of polymyxin B in patients with renal impairment

Study ID	Modeling and software	Model structure	Candidate covariates	Significant covariates	Parameter estimates		Inter-patient variability (ω2)	Residual variability	Internal validation	External validation
					V (L) CL (L/h)
Wang et al., 2024	NLMEM (FOCE-ELS); Phoenix NLME	Two-compartment model with linear elimination	Age, gender, weight, septic shock, mechanical ventilation, APACHE II score, SOFA score, albumin	CL: CrCL and DFI; Central compartment volume: DFI	tvVc: 10.05 tvVp: 18.78	tvCL: 1.97 tvQ: 2.99	Vc: 18.26 Vp: 29.87 CL: 16.06 Q: 47.38	Proportional error: 0.1034	GOF plots, PC–VPC, bootstrap	No
Hanafin et al., 2023	NLMEM (FOCE); NONMEM	Two-compartment disposition model with first-order linear elimination	BW, CrCl, CVVHDF	CL: CrCL, CVVHDF V: BW	V(L): 11.8 V2(L): 7.92	CL(L/h): 1.19 Q(L/h): 1.98 CL: CVVHDF patients: 2.58 (2.04–3.84) Non-RRT patients: 1.17 (0.49–4.24)	CL: 45.7 V: 62.7	Additive error: 0.593 Proportional error: 7.67	Bootstrap	No
Pi et al., 2023	NLMEM (FOCE-ELS); Phoenix NLME	Two-compartment model	Age, gender, TBW, height, TBIL, DBIL, ALB, TP, BUN, lactate, Scr, CVH and 24-h fluid intake and output, CRRT and Crcl	CL: CRRT, BW Q: BW	tvV(L): 7.857 tvV2(L): 12.668	tvCL(L/h): 1.672 tvQ(L/h): 7.009 CL (CRRT): 5.15 (1.17–15.68) CL (non-CRRT): 1.66 (0.71–4.38)	Exponential model V: 0.387 V2: 0.645 Cl: 0.282 Q: 0.356	Additive error	GOF, VPC, Bootstrap	No
Surovoy et al., 2023	NLMEM (SAEM); Monolix	Two-compartment model	Age, TBW, BIL, ALB, SOFA score, GFR, dialysis and ultrafiltration rate	CL and V2: Log-transformed albumin level and GFR	Without CVVHD: V(L): 20.8 V2(L): 90 With CVVHD: V(L): 19.79 V2(L): 50.56 V: RI:20.3 ± 0.3 V: CVVHD: 19.6 ± 0.2 V2: RI:103.9 ± 15.6 CVVHD: 56 ± 7.1	Without CVVHD: CL(L/h): 2.77 Q (L/h): 24.3 With CVVHD: CL(L/h): 1.87 Q(L/h): 8.52 CL: RI: 2.1 ± 0.1 CL: CVVHD: 2 ± 0.23 Q: RI: 27.7 ± 3.4 Q: CVVHD: 14.6 ± 4.6	Without CVVHD: CL: 0.35 V1: 0.18 Q: 0.6 V2: 0.6 With CVVHD: V: 0.14 V2: 0.053 CL: 0.24 Q: 1.64	Proportional error Without CVVHD: 0.19 With CVVHD: 0.12	No	No
Luo et al., 2022	NLMEM; NONMEM	Two-compartment model	Age, gender, TBW, APACHE II, SOFA, ALT, AST, ALP, GGT, TBIL, TP, ALB, BUN, UA, Scr, Crcl, and CRRT	CL: CRRT, CLCR, SOFA score and BW	V(L): 11.7 V2(L): 17.9	CL(L/h): 1.5 Q(L/h): 1.34	CL: 0.0598 V2: 0.854	Proportional error: 0.0589	GOF, Bootstrap	Yes (n = 14)
Wang et al., 2022	NLMEM (FOCE-ELS); Phoenix NLME	Two-compartment model	Age, TBW, gender, CRRT, SOFA, ALT, AST, TBIL, BUN, SCr, UA, TP, ALB, Crcl, eGFR, dialysis volume, blood flow and ultrafiltration rate	CL, V2, Q: CRRT	tvV(L): 15 tvV2(L): 6.54	tvCL(L/h): 1.95 tvQ(L/h): 2.28	V: 0.13 V2: 1.38 CL: 0.17 Q: 0.51	Proportional error: 0.1303	GOF, PC–VPC, Bootstrap	No
Li et al., 2021	NLMEM (FOCE-ELS); Phoenix NLME	One-compartment model with 1st order elimination	Age, TBW, ALT, AST, TBIL, DBIL, ALB, BUN, UA and Crcl	CL: CLCR	V (L): 12.09	CL(L/h): 1.18	V: 0.04 CL: 0.06	Proportional error	GOF, PC–VPC, Bootstrap	No
Wang et al., 2021	NLMEM (FOCE-ELS), Phoenix NLME	Two-compartmental model	Age, sex, BW, CLCR	No significant covariates	tvV(L): NRF: 6.87 RI: 6.98 tvV2(L): NRF: 11.97 RI: 10.57	tvCL(L/h): NRF: 2.19 RI: 1.58 tvQ(L/h): NRF: 13.83 RI: 10.28	V: NRF: 0.78 RI:0.38 CL: NRF:0.22 RI: 0.26 V2: NRF: 0.32 RI: 0.74 Q: NRF: 0.68 RI: NA	Proportional error: NRF: 0.13 RI: 0.10	GOF, PC–VPC, Bootstrap	No
Yu et al., 2020	NLMEM (FOCE-ELS), NONMEM	One-compartment model with linear elimination	Age, sex, TBW, CrCL, serum albumin	CL: CLCR	V(L): 20.5	tvCL(L/h): 1.59	BSV CL: 13	Proportional error: 40.5	PVC–VPC, normalized prediction distribution error (NPDE) test, bootstrap	No
Thamlikitkul et al., 2017	ADAPT 5	One-compartment linear model with zero-order input	NA	NA	NA	CL(L/h): NRF: 2.5 ± 0.4 RI: 2.0 ± 0.6	NA	NA	NA	NA
Sandri et al., 2013	NLMEM; S-ADAPT	Two-compartmental disposition model with linear elimination	TBW, LBW, gender, age, Crcl, ALB, and APACHE II	NA	V(L): 6.35 V2(L): 22.3	CL(L/h): 1.87 Q(L/h):9.86 CVVHD CL in 2 patients were 0.26/L and 0.37/L	BSV CL: 32.4 V: 73.3 V2: 70.1 Q: 50.4	Proportional error	GOF	No

AKI acute kidney injury, BSV between subject variability, BW body weight, CV coefficient of variation, CL clearance, CL_CR creatinine clearance, CVVHD continuous veno-venous hemodialysis, DFI daily fluid intake, FOCE-ELS first-order conditional estimates method using least squares estimation, NLMEM non-linear mixed effect modelling, NRF normal renal function, Q Intercompartmental clearance, SAEM stochastic approximation expectation–maximization algorithm, SOFA sequential organ failure assessment, tvV typical value of central compartment distribution volume, tvV₂ typical value of peripheral compartment distribution volume, V central volume of distribution, V2 peripheral volume of distribution

Covariates Affecting Pharmacokinetics of Polymyxin B

Significant covariates affecting polymyxin B clearance included CrCL [15, 17, 18, 21, 22, 31], CRRT technique [16, 18] such as CVVHDF [15], body weight [16, 18], fluid overload [31], logarithmic transformed albumin level and sequential organ failure assessment (SOFA) score [18]. Recent studies with larger patient sample size and severe renal impairment confirmed CrCL significant impact on polymyxin B clearance [21], unlike earlier studies showing a non-significant effect [4]. CVVHDF [15] and CVVH [19] also significantly influenced polymyxin B clearance.

Body weight [15, 16] was a key covariate for V_d in the central compartment, but was not significant in renal transplant patients [21]. While consensus guidelines recommend weight-adjusted dosing, a recent study found no correlation between body weight and polymyxin B clearance [3]. A pop-PK study in sepsis patients showed an inverse relationship between polymyxin B concentration and daily fluid intake or fluid overload [31].

Log-transformed albumin level and GFR [17] affected the V_d in the peripheral compartment, CVVH significantly impacted the peripheral V_d, total and intercompartmental clearance [19], and SOFA score also influenced polymyxin B clearance, indicating the need for maintenance dose adjustments based on severity of sepsis/disease [18]. Table 5 elaborates on the candidate covariates tested in each study.

Effect of RRT Techniques on AUC_ss24h Levels of Polymyxin B

Among the included studies, AUC at steady state (AUC_ss) was plotted against dose and various RRT modalities or renal impairment. A comparison of mean AUC_ss24h of polymyxin B showed that patients on CVVH[19] and CVVHDF[15] had mean AUC_ss24h levels below the standard 50–100 mg h/L range, while those on CVVHD[17] and without RRT[15] had AUC_ss24h levels above the standard range. This highlights variability in drug exposure across RRT techniques and dosing differences among studies. Figure 2 presents a bar chart illustrating study-state AUC levels of polymyxin B with the indicated dose.

Fig. 2.

Steady-state area under the concentration-time curve of polymyxin B in patients with normal renal function, renal impairment, and different renal replacement therapy modalities. Reported AUCss,24h values from included studies are presented according to renal function category and renal replacement therapy modality. Bar colors represent the respective patient groups and renal replacement therapy modalities. Data labels indicate reported AUC values. This figure presents the estimated steady-state area under the concentration–time curve (AUCss, 24ℎ) of polymyxin B reported across included studies in patients with normal renal function, renal impairment, and those receiving different renal replacement therapy (RRT) modalities. AUCss area under the concentration–time curve at steady state, CRRT continuous renal replacement therapy, CVVHDF continuous veno-venous hemodiafiltration, CVVH continuous veno-venous hemodiafiltration, CVVHD continuous veno-venous hematolysis, IHD Intermittent hemodialysis, SLEDD sustained low efficiency dialysis, Non-RRT not on any renal replacement therapy, RI renal impairment, NRF non-renal failure, RTP renal transplant patients. Dashed horizontal lines indicate the recommended polymyxin B target exposure range (AUCss, 24 h 50–100 mg h/L)

Dosing Strategies Based on Significant Covariates Affecting CL and Vd of Polymyxin B

• i. Requirement of Loading Dose: Most studies emphasized the need for a loading dose to achieve therapeutic concentrations, particularly for bacteria with minimum inhibitory concentration (MICs)≥ 2 mg/L. Li et al.[21] recommended a 75-mg loading dose followed by a 50-mg twice daily (BD) maintenance dose, effective in approximately 90% of patients but inadequate for MICs ≥ 2 mg/L. Sandri et al[4]. observed reduced drug exposure on the first day without a loading dose compared to steady-state.
• ii. Renal Insufficiency Patients not on any RRT: Studies reported differing AUC_SS0-24h values between normal and impaired renal function. Yu et al.[22] recommended a 33% dose reduction for moderate renal insufficiency (30 ≤ CrCL < 60 mL/min), while Wang et al.[31] suggested dose reductions for CrCL < 40 ml/min. Others found no renal function-based adjustments necessary.[4, 20] Sandri et al.,[4] advised 3 mg/kg/day with a loading dose for severe infections (MIC ≤ 2 mg/L) and up to 2.5 mg/kg/day for MICs ≤ 1 mg/L.
• iii. Patients on RRT Techniques: Hanafin et al.,[15] found a 50% lower median steady-state AUC_0–24h in CVVHDF patients, irrespective of dose-normalized AUC_0–24h. They recommended a 2.5-mg/kg loading dose followed by 1.5-mg/kg BD for CVVHDF, while 95% target attainment was achieved with maintenance doses of 1.25–1.5 mg/kg, 75 mg, or 100 mg BD for infections with MICs < 2 mg/L in non-RRT or other modalities. Luo et al.[18] reported that 75–100 mg BD was sufficient for CRRT patients with SOFA scores ≤11 and MICs ≤ 1 mg/L, but no regimen (maintenance doses of 30 mg, 40 mg, 50 mg, 75 mg or 100 mg BD) worked for SOFA scores 11–12 and MICs ≥ 2 mg/L.

  Surovoy et al.[17] recommended a fixed maintenance dose of 100 mg BD with a 200-mg loading dose for CVVHD patients. Wang et al.[19] suggested therapy–modality-based regimens, with 2.5 mg/kg loading on CVVH and 1.5 mg/kg off CVVH, adjusted for weight. While effective for MICs < 0.25 ml/L in CVVH, none were ideal for MIC > 1mg/L.[19]

• iv. Renal Transplant Patients Requiring CRRT: Li et al.[21] proposed 50-mg BD maintenance doses to reach target AUC_SS0-24h, though insufficient for high SOFA scores (≥ 1 mg/L).

Discussion

This systematic review has examined covariates influencing polymyxin B pharmacokinetics in patients with renal impairment, including those receiving RRT. The findings underscore the complexity of dosing polymyxin B in this population due to high interindividual variability in drug clearance and exposure. Key determinants include creatinine clearance, body weight, severity of illness, and the type of RRT modality received. These factors necessitate an individualized dosing approach supported by therapeutic drug monitoring (TDM) where available.

Impact of Renal Function (CrCL)

Recent pop-PK studies present conflicting evidence regarding the influence of CrCL on polymyxin B clearance. While earlier investigations reported minimal or no effect [4, 20], more recent studies have identified CrCL as a significant covariate, especially in patients with severe renal impairment (CrCL < 40 mL/min) [17, 21, 22, 31]. For instance, Li et al. demonstrated a strong CrCL clearance relationship in renal transplant recipients [21]; however, the extrapolation of these results to the broader population must be approached with caution due to altered physiology in transplant patients.

Moreover, Wang et al. noted differences in clearance and AUC between patients with and without renal impairment but concluded that the two-fold variation did not justify routine dose adjustments [20]. These discrepancies between statistical significance and clinical impact highlight the need for outcome-driven studies to establish dosing thresholds for renal function-based dose adjustment.

As a result, while CrCL is an important determinant of polymyxin B clearance, it should not be used in isolation for dose adjustment and must be interpreted in conjunction with factors such as RRT modality, disease severity, and patient-specific characteristics. An important limitation in interpreting the impact of renal function on polymyxin B pharmacokinetics is the lack of a standardized classification of renal impairment across studies. Most investigations relied on CrCL thresholds rather than internationally accepted criteria such as KDIGO staging for AKI or CKD. This heterogeneity limits comparability between studies and restricts the ability to derive stage-specific dosing recommendations. Future studies should adopt uniform renal function classifications to enable more clinically meaningful stratification and to improve dosing precision. They should also explore integrating CrCL with additional clinical parameters, such as fluid status, concurrent medications, and organ support modalities, to optimize dosing

Importantly, the clinical significance of pharmacokinetic variability associated with covariates such as CrCL should be interpreted in the context of pharmacokinetic–pharmacodynamic targets, particularly the AUC/MIC ratio. While several studies demonstrated a statistically significant relationship between CrCL and polymyxin B clearance, the magnitude of this effect may not consistently translate into clinically meaningful differences in target attainment, especially for pathogens with low MIC values. Conversely, in infections caused by organisms with higher MICs (≥ 2 mg/L), even modest reductions in drug exposure due to increased clearance may result in suboptimal AUC/MIC ratios and therapeutic failure. Therefore, assessment of covariate effects should extend beyond statistical significance to include their impact on pharmacodynamic target attainment. This highlights the importance of integrating MIC data and, where feasible, TDM to guide individualized dosing strategies in patients with varying renal function.

Fixed Versus Weight-Based Dosing

The appropriate dosing strategy for polymyxin B—whether weight-based or fixed—remains a matter of ongoing debate. Sandri et al. initially recommended weight-based dosing (2.5–3 mg/kg/day based on total body weight) to achieve pharmacodynamic targets, [4] a strategy adopted by consensus guidelines [3]. However, Hanafin et al., in a more recent prospective analysis incorporating data from Sandri et al., proposed fixed loading and maintenance doses (e.g., 200 mg loading, 100 mg q12h maintenance), citing reduced variability in exposure across diverse patient populations [15]. Their findings highlight that, while total body weight may influence V_d, it does not consistently predict clearance in critically ill or renally impaired patients. Thus, fixed dosing with clinical refinements based on renal function, disease severity, and TDM may offer a more pragmatic and consistent approach in real-world settings.

Considerations for Loading Dose and MIC

Loading doses are crucial for achieving therapeutic concentrations, especially in infections caused by organisms with MIC ≥ 2 mg/L. However, this must be balanced against the risk of nephrotoxicity. Luo et al. showed that, while a loading dose of 2.5 mg/kg followed by 1.5–2 mg/kg every 12 h improved target attainment, it also raised the probability of toxicity in patients with renal dysfunction or on CRRT [18]. Similarly, Hanafin et al. found it challenging to achieve target AUCs in high-MIC scenarios without exceeding nephrotoxic thresholds [15]. These findings reinforce the importance of individualized loading doses guided by pathogen susceptibility, renal reserve, and RRT status. In patients with high MICs, combining intravenous polymyxin B with other agents or using adjunctive inhaled formulations may be safer than aggressive monotherapy escalation.

Effect of RRT Modalities on Clearance

RRT type significantly influences polymyxin B pharmacokinetics. Convective therapies such as CVVH and CVVHDF enhance drug clearance due to their ability to remove larger molecules via filtration and adsorption [15, 16, 32]. In contrast, diffusive modalities like CVVHD or SLEDD typically have a less pronounced effect due to the drugs’s large molecular size and high protein binding [17, 25]. Notably, even within similar RRT modalities, drug exposure can vary based on filter type, flow rates, and patient-specific factors. For example, patients receiving CVVHDF demonstrated reduced dose-normalized AUCs and shorter half-lives compared to non-dialysis patients [15]. Meanwhile, some patients on CVVHD with underlying renal impairment exhibited prolonged half-lives up to 36 h, [17] underscoring the complex interplay between baseline renal function and dialysis characteristics. Moreover, although polymyxin B undergoes minimal renal excretion, tubular reabsorption in non-dialyzed patients may prolong drug exposure and contribute to toxicity. In CRRT patients, however, extracorporeal removal bypasses this reabsorption, leading to increased clearance. There findings support the need for dose adjustments in CRRT to avoid therapeutic failure or toxicity [18, 33, 34].

In addition to RRT modality, variability in dialysis parameters such as blood flow rate, dialysate and ultrafiltration rates, and membrane characteristics may significantly influence polymyxin B clearance. However, these parameters were inconsistently reported across studies, limiting the ability to quantitatively assess their impact. As a result, the observed variability in drug exposure across RRT modalities may partly reflect differences in dialysis settings rather than modality alone. Future studies incorporating standardized reporting and sensitivity analyses of dialysis parameters are needed to better define their contribution to pharmacokinetic variability.

Role of Disease Severity

SOFA score was identified as a significant covariate affecting polymyxin B clearance in one pop-PK study [18]. The association likely reflects augmented renal clearance and capillary leak in severely ill patients. However, this finding has not been validated in other studies. As a composite score, SOFA was not developed to guide antimicrobial dosing, and its use in this context should be approached cautiously. Although higher SOFA scores may suggest a need for more aggressive dosing, they should be considered alongside other clinical indicators. Until more evidence emerges, TDM-guided adjustments remain preferable over SOFA-based strategies.

This systematic review highlights the multifactorial determinants of polymyxin B pharmacokinetics in renally impaired and critically ill patients. CrCL and RRT modality are key contributors to variability in drug exposure. While weight-based dosing may offer initial guidance, fixed dosing adjusted for clinical status and supported by TDM may provide more consistent therapeutic outcomes. Importantly, aggressive loading in high-MIC infections must be balanced with toxicity risk, and RRT-specific dose modifications remain essential for optimizing treatment efficacy and safety.

Expert Opinion

Polymyxin B remains a critical last-resort agent for MDR Gram-negative infections, yet dosing in patients with renal impairment—particularly those receiving RRT—poses considerable challenges.[4] Although existing guidelines do not mandate routine dose adjustments based on renal function,[3] accumulating pharmacokinetic data and clinical variability support a shift toward individualized dosing strategies informed by patient condition, pharmacodynamic targets, and, when available, TDM. These recommendations are intended to complement existing guideline-based approaches and should not be considered a substitute for established clinical guidelines.

Pharmacodynamic Targets and Variability

The pharmacodynamic efficacy of polymyxin B is best described by free-drug AUC/MIC, with a proposed target of 50–100 mg h/L and steady-state concentration near 2 mg/L [3, 4, 24]. However, these targets are primarily derived from preclinical models and may not fully capture real-world variability, especially in critically ill or renally impaired populations.

Polymyxin B is moderately protein-bound (~ 50–60%), and alterations in protein binding can increase the unbound drug fraction. While this may enhance bacterial killing, it may also increase the nephrotoxicity risk. Unfortunately, most pop-PK studies report only total concentrations, potentially overestimating target attainment in patients with altered protein binding states [13, 14]. An additional consideration is the distinction between total and unbound polymyxin B concentrations. As only the unbound fraction is V_d active, reliance on total plasma concentrations may not accurately reflect pharmacodynamic target attainment. This is particularly relevant in critically ill patients, where hypoalbuminemia, systemic inflammation, and altered protein binding can increase the unbound fraction, potentially enhancing both efficacy and toxicity. Despite this, most pop-PK studies report total drug concentrations and data on unbound polymyxin B exposure remain limited. Consequently, current PK-PD targets based on total concentrations may either overestimate or underestimate true pharmacologically active exposure in certain patient populations. Future studies evaluating unbound drug concentrations and their relationship with clinical outcomes are essential to refine dosing strategies.

Other contributors to interindividual variability include fluid overload, capillary leak, altered perfusion, and extracorporeal support. For instance, Wang et al. found that fluid status significantly influenced polymyxin B exposure in septic patients [31]. Variability in RRT modalities such as membrane type, flow rates, and timing of drug administration further complicates dosing.

Dosing in RRT: A Modality-Specific Approach

Polymyxin B pharmacokinetics differ significantly across RRT modalities. Convective techniques like CVVH and CVVHDF generally enhance clearance more than diffusive techniques such as CVVHD, due to mechanisms like filtration and membrane adsorption.[4, 15, 32] CrCL should be considered as a supportive parameters within a broader individualized dosing framework rather than a sole determinant of dose adjustment.

• CVVH and CVVHDF: These patients often require higher doses (e.g., ≥ 150 mg loading and ≥ 100 mg every 12 h maintenance) due to enhanced clearance.

• CVVHD: Despite being a diffusive modality with lesser clearance potential, clinical data remain inconclusive. Surovoy et al. found exposures similar to those in patients with preserved renal function and recommended a fixed 100 mg every 12 h.[17] However, this study had a small sample size and did not evaluate toxicity or outcomes.

In contrast, our expert opinion suggests a cautious, individualized dosing approach in patients receiving CVVHD. While Surovoy et al. reported adequate exposure with a maintenance doses of 100 mg every 12 h, the evidence is limited by small sample size and lack of outcome-based validation. Given the variability in dialysis parameters, residual renal function, and altered protein binding in critically ill patients, lower initial maintenance doses (e.g., 50 mg BD) may be considered in selected high-risk patients to minimize potential toxicity. However, in clinically stable patients or those with high-MIC infections, escalation to 100 mg every 12 h may be appropriate, provided close clinical and, where available, TDM is ensured.

Several other factors which could influence this decision are (1) the diffusive nature of CVVHD flow rates, and session duration, (2) the moderate protein binding of polymyxin B introduces uncertainty regarding the amount available for removal, and (3) most CVVHD patients in real-world settings also have varying degrees of residual renal function, hypoalbuminemia, and organ dysfunction that may increase drug exposure and risk of nephrotoxicity.

Therefore, until more robust data are available, our recommendations reflect a conservative, individualized dosing approach that prioritizes safety, particularly in patients with fluctuating renal function or limited access to TDM. Nevertheless, in clinically stable patients with high MIC infections, a higher dose (up to 100 mg BD) may still be warranted, as supported by Surovoy et al., provided close monitoring is feasible.

Given the uncertainty, we recommend a conservative dose of 50 mg every 12 h in CVVHD patients, particularly those with residual renal function or hypoalbuminemia. In stable patients with high-MIC infections, escalation to 100 mg every 12 h may be considered with close clinical monitoring.

Timing Around Dialysis

In patients receiving intermittent hemodialysis (IHD) or SLEDD, the timing of polymyxin B administration is critical. Administering the drug after dialysis minimizes extracorporeal clearance and ensures adequate post-dialysis plasma concentrations [19, 35]. Although polymyxin B is primarily non-renally eliminated, high-flux dialysis membranes and convective modalities may still remove appreciable drug amounts via filtration or adsorption [13, 25].

Therefore, we recommend:

• Post Dialysis Administration, especially on dialysis days, to maximize systemic exposure.

• A Supplemental 50-mg Post-dialysis Dose (for IHD or SLEDD), to restore therapeutic levels and prevent subtherapeutic troughs, particularly when TDM is unavailable or dialysis is conducted near a scheduled dose.

A 50-mg supplemental dose of polymyxin B post-dialysis is recommended to compensate for potential drug loss during IHD or SLEDD sessions. While polymyxin B is primarily non-renally cleared, studies have shown that certain dialysis membranes and prolonged diffusive or convective sessions can remove measurable amounts of the drug through filtration or adsorption. Administering a supplemental dose immediately after dialysis helps restore plasma concentrations and maintain therapeutic exposure throughout the dosing interval, especially when dialysis is performed close to a scheduled dose or in patients without access to TDM.

The Role of Combination Therapy

Combination therapy is preferred over monotherapy in critically ill patients, particularly when dealing with high-MIC pathogens. Evidence from in vitro synergy studies and clinical observations supports combining polymyxin B with agents such as tigecycline, aminoglycosides, or carbapenems. This approach may improve bacterial killing, reduce resistance development, and allow safer polymyxin exposure levels when monotherapy cannot achieve pharmacodynamic targets [10, 18].

RRT practices vary globally, with CVVH and CVVHD more commonly used in Australia, and IHD or SLEDD more frequently in Asian settings.[36] Despite the growing need for individualized polymyxin B dosing, TDM is not widely accessible, and validated assays for unbound drug measurement remain limited. In such context, clinical monitoring using surrogate markers like creatinine, albumin, and SOFA score becomes essential.

A practical dosing algorithm summarizing key considerations for polymyxin B use in patients with renal impairment and different RRT modalities is provided (Fig. 3), to aid clinical decision-making.

Fig. 3.

Suggested clinical decision algorithm for polymyxin B dosing in patients with renal impairment and renal replacement therapy. This algorithm integrates available population pharmacokinetic evidence and expert interpretation to guide dosing based on renal function, renal replacement therapy modality, and infection severity. Recommendations should be individualized based on patient-specific factors and, where available, supported by TDM. This algorithm is intended to complement, not replace, existing clinical guidelines., CVVHDF continuous veno-venous hemodiafiltration, CVVH continuous veno-venous hemodiafiltration, CVVHD continuous veno-venous hemodialysis, IHD intermittent hemodialysis, LD loading dose, MD maintenance dose, MDR multidrug resistance, MIC Minimum inhibitory concentration, RRT renal replacement therapy, SLEDD sustained low efficiency dialysis, SOFA sequential organ failure assessment, TDM therapeutic drug monitoring

Limitations

This systematic review has several limitations that should be considered when interpreting its findings. Firstly, the heterogeneity in patient characteristics and varying degrees of renal dysfunction makes it challenging to compare the impact of specific renal function levels on polymyxin B pharmacokinetics. Secondly, there was inconsistency in detailed reporting of dosing regimens and modification based on the renal function. The variability in reporting makes it difficult to establish clear relationships between dosing modifications by the studies and its influence in pharmacokinetic outcomes. Thirdly, non-uniform methods used to calculate and report AUC values across studies may introduce bias in the comparison of drug exposure. Variability in AUC estimation approaches, including differences in pharmacokinetic modeling techniques, sampling strategies and calculation methods (e.g., model-based vs. non-compartmental analysis), can lead to systematic differences in reported exposure metrics. This heterogeneity limits the reliability of cross-study comparisons and may affect the interpretation of pooled pharmacokinetic findings. Moreover, the lack of detailed and standardized reporting of dialysis parameters across studies precluded sensitivity analyses to evaluate their impact on polymyxin B clearance. Furthermore, the absence of standardized renal impairment classification (e.g., KDIGO-defined AKI/CKD stages) across studies precluded subgroup analyses and limits the applicability of findings to specific stages of renal dysfunction. An additional limitation is the limited correlation between pharmacokinetic findings and clinically relevant outcomes such as mortality, microbiological clearance or treatment success across the included studies. Most pop-PK analyses have focused on identifying covariates influencing drug exposure, without linking these parameters to pharmacodynamic target attainment or patient-centered outcomes. This gap restricts the ability to translate pharmacokinetic variability into definitive dosing recommendations. Future research should prioritize integrated PK-PD studies that correlate drug exposure with clinical and microbiological outcomes to better inform evidence-based dosing strategies.

Future Perspectives

There is a critical need to investigate the impact of different RRT techniques, including HD, SLEDD, and peritoneal dialysis on polymyxin B dosing. Comparative studies of CRRT modalities such as CVVHDF, CVVHD, and SLEDD could clarify their effects on drug clearance. Additionally, assessing RRT parameters like dialysis dose, membrane type, and blood flow rate would refine dosing recommendations.

Incorporating biomarkers such as serum albumin into dosing algorithms may enable more personalized treatment. The impact of fluid overload on polymyxin B pharmacokinetics, especially in CKD and critically ill patients, warrants further study to improve dosing accuracy. Advanced modeling techniques offer promising solutions. Physiologically-based pharmacokinetic modeling can enhance understanding of drug distribution, while Bayesian forecasting can optimize dosing with limited sampling. In-silico modeling could predict drug–protein interactions and genetic influences. These approaches may lead to more precise dosing strategies, ultimately improving outcomes and minimizing toxicity.

Conclusion

This systematic review highlights the complex and multifactorial nature of polymyxin B pharmacokinetics in patients with renal impairment, including those receiving various RRT. While CrCL emerged as a significant covariate, dose adjustment based solely on CrCL may be insufficient and should be integrated with other clinical factors to guide individualized dosing. RRT modality plays a more decisive role, with convective therapies (e.g., CVVH, CVVHDF) associated with enhanced clearance requiring higher maintenance doses, whereas diffusive techniques (e.g., CVVHD, SLEDD) may permit lower doses due to variable drug removal. Body weight affects V_d but not clearance, limiting the utility of weight-based dosing alone. Given the narrow therapeutic index of polymyxin B, individualized dosing based on renal function and dialysis parameters is crucial. TDM, when available, should guide optimization. Until more robust outcome data emerge, cautious, patient-tailored dosing remains the best strategy to maximize efficacy while minimizing nephrotoxicity in this vulnerable population.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Conceptualization: Asha K Rajan (AKR), Muhammed Rashid (MR), Girish Thunga (GT); Data curation: AKR; MR; Formal analysis: AKR, Vishal Shanbhag (VS); Funding acquisition: None; Investigation: AKR, MR, VS; Methodology: AKR, VS; Project administration: AKR, GT; Resources: GT; Software: AKR, MR; Supervision: GT, Souvik Chaudhuri (SC), Sohil Khan (SK), Ravindra Prabhu Attur (RPA); Validation: GT, VS, Hassan Mitwally (HM), Indu Ramachandra Rao (IRR), Bhuvana Krishna (BK); Visualisation: AKR, MR, Maria Ioana Onofrei (MIO), Vijayanarayana K (VK); Writing original draft: AKR, VS, MR, MIO, IRR, SK, SC, RPA, HM, BH, VK, Surulivel Rajan (SR), GT; Review & editing: AKR, VS, MR, GT; Access to the data and verification: AKR, MR, GT; Guarantor: GT. All authors confirm that they had full access to all the data in the study, accept responsibility to submit for publication, and have read and approved the final version of the manuscript.

Funding

Open access funding provided by Manipal Academy of Higher Education, Manipal. This research did not receive any grant from funding agencies in the public, commercial or not-for-profit sectors.

Data Availability

All the available data are within the manuscript and its supplementary information.

Code Availability

Not applicable.

Declarations

Conflict of Interest

The authors declare that they have no conflict of interest, financial or non-financial related to this work.

Ethical Approval

Ethics approval was not required, as this study is based on previously published data and does not involve a new human or animal participants.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.