The emergence of extensively drug-resistant Gram-negative bacteria (XDR-GNB) has become a major global health challenge because of the rapid resistance transmission, increasing treatment difficulties, and association with high morbidity and mortality rates (Borer et al., 2009; Cassini et al., 2019). Polymyxins including colistin and polymyxin B (PMB) have excellent antibacterial activity against the majority of XDR-GNB as reported by recent surveillance surveys (Diekema et al., 2019; Xi et al., 2022) and have been revived as the last-resort therapeutic opinion for XDR-GNB infections (Katip et al., 2020; Katip et al., 2021; Liu et al., 2023). PMB is an active drug without requiring conversion to colistin after its administration, whereas colistin must be administered as an inactive prodrug, colistin methanesulfonate, to then be converted to colistin in vivo, which could be slow and incomplete. Therefore, PMB may be the preferred polymyxin for XDR-GNB infections. However, PMB has a narrow therapeutic window and high occurrence of nephrotoxicity, the most commonly reported adverse effect of PMB (Rigatto et al., 2016; Tsuji et al., 2019). Because the plasma concentration of PMB is closely related to its antibacterial efficacy and nephrotoxic effect, the area under the concentration-time curve over 24 h at steady state (AUC_{ss, 0–24 h}) of 50–100 mg·h/L was recommended for the therapeutic drug concentration, which is equivalent to an average plasma concentration at steady state (C_{ss, avg}) of 2–4 mg/L (Tsuji et al., 2019). If the AUC_{ss, 0–24 h} of PMB is more than 100 mg·h/L, the incidence and severity of nephrotoxicity increase significantly (Wang et al., 2021a; Yang et al., 2022).

PMB has been reported to accumulate in large quantities in the kidney, much more than in other organs, such as the heart, lungs, and brain. In the kidney, PMB accumulates mainly in the proximal tubule after intravenous injection, as demonstrated by animal model studies (Manchandani et al., 2015; Yun et al., 2015). Considering that PMB is poorly recovered in urine (less than 1%) (Zavascki et al., 2008; Yun et al., 2015), it may be filtered from the glomerulus and reabsorbed in the proximal tubule. Hence, the accumulation of PMB in the renal proximal tubule has been proposed as the possible main mechanism of PMB-related nephrotoxicity (Yun et al., 2015; Manchandani et al., 2017).

The elderly population, as a special group, has different characteristics from the young and middle-aged population in terms of disease progression and drug metabolism, distribution, and clearance. Elderly patients are more prone to develop multiple organ dysfunctions when infected than young patients. A large epidemiological survey has demonstrated that the incidence of sepsis increases with age, with the incidence in patients over 85 years being 100 times higher than in children (Angus et al., 2001), and nearly 65% of septic patients are elderly (Martin et al., 2006). Additionally, advanced age has been shown to be an independent risk factor for sepsis-associated acute kidney injury (AKI) (Poston et al., 2019). Meanwhile, the physiological changes in elderly subjects, such as the reduced glomerular filtration rate and lower tubular reabsorption capacity, may result in increased PMB exposure to the renal tubular cells that increase the occurrence of nephrotoxicity (Manchandani et al., 2017). Several studies have confirmed that older patients have significantly increased risk of polymyxins-related nephrotoxicity than younger patients (Wagenlehner et al., 2021; Yang et al., 2022).

The current international guideline recommends that the dose of PMB should be calculated based on the patient’s weight, irrespective of age (Tsuji et al., 2019). Several population pharmacokinetics (PK) studies have reported that PMB clearance is independent of age in critically ill patients or patients with cystic fibrosis (Sandri et al., 2013; Avedissian et al., 2018; Manchandani et al., 2018; Wang et al., 2020), whereas one study has found that age is significantly related with the volume of distribution of PMB (Liang et al., 2023). Whether elderly patients have higher PMB exposure than young critically ill patients and different clearance and volume of distribution from young patients given the standard dose has remained unknown until now. Therefore, we conducted this prospective observational clinical study aimed to compare PMB exposure in elderly patients and young critically ill patients and to explore potential covariates of PMB PK in critically ill patients.

This PK study has made a significant contribution to understanding how to optimize the use of PMB in clinical practice in critically ill patients. To our knowledge, this is the first study to date comparing PMB exposure and other individual PK parameters in critically ill patients of different ages given the standard dosing of PMB. Age was not shown to be correlated with clearance and volume of distribution of PMB in critically ill patients in this study, consistent with many previously published studies (Sandri et al., 2013; Avedissian et al., 2018; Manchandani et al., 2018; Wang et al., 2020; Wang et al., 2021a; Yu et al., 2021; Surovoy et al., 2022; Yang et al., 2022). However, we did find that the elderly patients had longer half-life time than the young critically ill patients and age was positively correlated with half-life time of PMB. Half-life time is an important factor influencing the AUC_{ss, 0–24 h} and was shown to be positively correlated with AUC_{ss, 0–24 h} (r = 0.688, p < 0.001) in this study. Although the mean AUC_{ss, 0–24 h} was not significantly different between the elderly and the young patients, the age of the patients with AUC_{ss, 0–24 h} more than 100 mg·h/L was higher than that of patients with AUC_{ss, 0–24 h} less than 50 mg·h/L, with marginal difference. These results indicate that elderly patients may have more delayed PMB clearance and metabolism than young patients and trend toward PMB overdose given the standard dosage.

Factors influencing the distribution, metabolism, and excretion of a drug can increase its half-life time. We speculate a few possible explanations for why elderly patients have longer half-life time than young patients. Firstly, in comparison with the young patients, the elderly patients had lower TBW, which was shown to be positively correlated with PMB clearance in this study. Consistently, three studies have demonstrated a significant positive correlation of TBW with clearance or volume of distribution of PMB (Sandri et al., 2013; Miglis et al., 2018; Crass et al., 2021). Our study also found TBW was negatively correlated with AUC_{ss, 0–24 h} while the AUC is dependent on the dose and clearance of a drug, indicating that the TBW affects the AUC mainly through influencing PMB clearance. For example, the prescribed PMB dose was calculated by TBW in this study. If a patient has low TBW, the given dose of PMB would decrease, and PMB clearance would decrease while the AUC_{ss, 0–24 h} would increase, according to our study. Only when a decrease in clearance is greater than the decrease in the given dose, an increase in AUC could be observed. However, the relationship between TBW and clearance and volume of distribution of PMB was not certain. In contrast to our study, many PK studies did not find TBW as a covariate of the PMB clearance and volume of distribution (Avedissian et al., 2018; Kubin et al., 2018; Manchandani et al., 2018; Wang et al., 2020; Wang et al., 2021a; Wang et al., 2021b; Li et al., 2021; Yu et al., 2021), and weight-based dosage regimens are being challenged based on these findings (Liu et al., 2023). Therefore, such discrepancy could be caused by several factors, Such as the difference in sample size, with a range of 9–70 (Sandri et al., 2013; Avedissian et al., 2018; Kubin et al., 2018; Manchandani et al., 2018; Miglis et al., 2018; Wang et al., 2020; Wang et al., 2021a; Wang et al., 2021b; Crass et al., 2021; Li et al., 2021; Yu et al., 2021), different compartment models, and different dosing regimens used in various studies. For instance, PMB was administered with loading doses first or directly administered with maintenance doses ranging from 0.34 to 3.45 mg/kg/day (Sandri et al., 2013; Avedissian et al., 2018; Kubin et al., 2018; Manchandani et al., 2018; Miglis et al., 2018; Wang et al., 2020; Wang et al., 2021a; Wang et al., 2021b; Crass et al., 2021; Yu et al., 2021) or with the fixed dose of 40–100 mg twice a day in these studies (Wang et al., 2020; Wang et al., 2021a; Crass et al., 2021; Li et al., 2021; Chen et al., 2022). The study population was also different, including critically ill patients and patients with cystic fibrosis, renal transplantation, extracorporeal membrane oxygenation, or different renal function (Avedissian et al., 2018; Wang et al., 2021a; Crass et al., 2021; Li et al., 2021; Yu et al., 2021; Surovoy et al., 2022), and some studies did not mention the disease severity of the enrolled patients (Avedissian et al., 2018; Kubin et al., 2018; Manchandani et al., 2018; Miglis et al., 2018; Wang et al., 2020; Wang et al., 2021a; Crass et al., 2021; Li et al., 2021). The pharmacokinetics of antibiotics are more profoundly altered in critically ill patients than in general patients and healthy volunteers due to many pathophysiological changes during acute critical illness such as capillary leak, edema, reduced serum albumin concentration, altered renal blood flow, etc. (Gonçalves-Pereira et al., 2021).

Secondly, the elderly patients had higher Charlson comorbidity scores than the young patients in this study. This means the elderly patients had more co-morbidities and more medications for their co-morbidities and tended to have a reduced renal glomerular filtration rate and hepatic metabolism. However, the metabolism and excretion of PMB remain unclear. Generally, PMB is mainly eliminated by non-renal pathways as a very low urine recovery of PMB was found after intravenous injection (Zavascki et al., 2008; Sandri et al., 2013). Hepatic biliary excretion could be one of the non-renal routes of PMB elimination because all four major PMB components were detected in bile over 4 h after intravenous injection (Manchandani et al., 2015). Due to the unclear metabolism routes, we cannot identify which type of medications or which underlying co-morbidities could interfere with the metabolism and excretion of PMB.

The optimal PMB dosing regimen is controversial currently. Determination of clinical variables influencing PMB exposure and the enhancement of recognition of healthcare professionals worldwide through education is critical to optimize dosing regimens to improve patient outcomes. The surprising findings in this study are that the lower total body weight and higher baseline serum creatinine levels were the significant clinical factors associated with PMB overdose. Also, total body weight was the only variable with a significant correlation with PMB plasma concentration as shown in this study. The relationship between body weight and PMB exposure is an unexpected result that has not been reported previously and deserves further study with a large population for verification. As for higher baseline serum creatinine, in contrast with the current guideline, which recommends there is no need to adjust PMB dosage in patients with renal insufficiency (Tsuji et al., 2019), our result indicates the adjustment of PMB dosing is needed in patients with higher serum creatinine levels to minimize the risk of nephrotoxicity. Theoretically, the pathophysiological changes in renal impairment or age-related physiological changes such as decreased glomerular filtration rate and tubular absorption dysfunction could decrease PMB clearance and result in higher PMB exposure (Avedissian et al., 2019; Nie et al., 2022). Similarly to our study, Wang et al. found that the mean AUC_{ss, 24 h} of PMB in their renal insufficiency group was higher than in their normal renal function group, with marginal significance (Wang et al., 2021a). Moreover, some clinical studies have demonstrated that the patients with PMB-associated AKI had higher AUC_{ss, 0–24 h} (Xu et al., 2022; Yang et al., 2022), higher baseline serum creatinine levels, or lower baseline glomerular filtration rates compared to patients without PMB-associated nephrotoxicity (John et al., 2017; Xi et al., 2022; Xu et al., 2022). These results support the speculation that patients with renal dysfunction are more likely to develop higher PMB exposure and that the dosing regimen of PMB should be adjusted in patients with renal dysfunction. However, the relationship between renal function and PMB exposure is still under debate. Another study did not find that AUC_{ss, 0–24 h} or normalized AUC_{ss, 0–24 h} of PMB was significantly different between the patients with normal renal function and patients with renal insufficiency (Thamlikitkul et al., 2016).

The relationship between renal function and PMB clearance is also still under debate. Some population PK studies have demonstrated a significant positive correlation of CrCl with PMB clearance (Avedissian et al., 2018; Manchandani et al., 2018; Wang et al., 2020; Wang et al., 2021a; Li et al., 2021; Yu et al., 2021), which was not confirmed in other studies (Sandri et al., 2013; Miglis et al., 2018; Wang et al., 2021b; Crass et al., 2021; Surovoy et al., 2022; Wang et al., 2022; Liang et al., 2023). It is interesting to note that CrCl, but not other renal function markers such as serum creatinine levels, has been proven to be associated with PMB clearance in some studies (Avedissian et al., 2018; Manchandani et al., 2018; Wang et al., 2020; Wang et al., 2021a; Li et al., 2021; Yu et al., 2021). The CrCl was estimated using the Cockcroft-Gault formula, which is determined by age, TBW, sex, and serum creatinine levels in all the current PMB PK studies; however, the Cockcroft-Gault formula has been shown to be highly inaccurate for the assessment of renal function in critically ill patients (Bragadottir et al., 2013). Age and especially TBW were the possible covariates affecting PMB clearance shown by our group and other studies (Sandri et al., 2013; Miglis et al., 2018; Crass et al., 2021; Liang et al., 2023). Thus, it is possible that the significant correlation of CrCl with PMB clearance is influenced by age or body weight. Therefore, we cannot draw the conclusion that renal function is correlated with PMB clearance based on the findings of significant correlations of CrCl and PMB clearance. Furthermore, in the two studies about PMB exposure in patients with different renal function (Thamlikitkul et al., 2016; Wang et al., 2021a), renal insufficiency was diagnosed by estimating CrCl<80 mL/min using the Cockcroft-Gault formula. Whether the studied population was general patients or critically ill patients, and the proportion of critically ill patients in the enrolled patients were not mentioned in these two studies (Thamlikitkul et al., 2016; Wang et al., 2021b). Therefore, further studies aiming to explore the relationship between renal function and PMB clearance or PMB exposure should evaluate renal function using more accurate biochemical markers and match potential covariates including body weight, age, and study population.

Several limitations to our study must be considered. First, the small sample size. A larger sample size would enhance the robustness of our findings. Second, critically ill patients are highly heterogeneous and have more factors that may influence drug metabolism, distribution, and clearance compared with general patients. As such, it is unknown whether the results of this study can be generalized to the general patient population. However, critically ill patients are more vulnerable to XDR-GNB infection and develop severe infections with multiple organ dysfunctions more often than general patients. Therefore, critically ill patients are the population most worthy of study. Third, the difference in the association of drug exposure with clinical outcomes between the elderly and young critically ill patients was not compared in this study because the sample sizes in the two groups were too small to reach any definite conclusions.

In conclusion, PMB exposure at steady state was comparable between the elderly and the young critically ill patients. Elderly patients may have more delayed PMB clearance and metabolism than young patients. TBW was the only covariate affecting PMB clearance. High baseline serum creatinine levels and low TBW were associated with PMB overdose, indicating that the PMB dosing regimen needs to be adjusted in these patients to minimize the risk of nephrotoxicity. Further clinical studies with large sample sizes are urgently needed to investigate PMB PK in critically ill patients with different renal functions assessed using more accurate biochemical markers, different ages, and various body weights.