Evaluation of Dose-Fractionated Polymyxin B on Acute Kidney Injury Using a Translational In Vivo Rat Model

We investigated dose-fractionated polymyxin B (PB) on acute kidney injury (AKI). PB at 12 mg of drug/kg of body weight per day (once, twice, and thrice daily) was administered in rats over 72 h. The thrice-daily group demonstrated the highest KIM-1 increase (P = 0.018) versus that of the controls (P = 0.99) and histopathological damage (P = 0.013). A three-compartment model best described the data (bias, 0.

ABSTRACT

We investigated dose-fractionated polymyxin B (PB) on acute kidney injury (AKI). PB at 12 mg of drug/kg of body weight per day (once, twice, and thrice daily) was administered in rats over 72 h. The thrice-daily group demonstrated the highest KIM-1 increase (P = 0.018) versus that of the controls (P = 0.99) and histopathological damage (P = 0.013). A three-compartment model best described the data (bias, 0.129 mg/liter; imprecision, 0.729 mg²/liter²; R², 0.652,). Area under the concentration-time curve at 24 h (AUC₂₄) values were similar (P = 0.87). The thrice-daily dosing scheme resulted in the most PB-associated AKI in a rat model.

TEXT

The widespread use of broad-spectrum antimicrobial agents has led to an increasing rate of resistant infections and associated morbidity and mortality (1–4). Although agents in the pipeline are promising, there is a prudent need to maximize the clinical efficacy and safety of currently available antibiotics (5). As a result of the paucity of active antibiotics, the polymyxins remain last-line options despite their associated nephrotoxicity (6–16). While the various mechanisms of polymyxin toxicity are being elucidated, dosing strategies that mitigate toxicity remain poorly defined. Toxicity thresholds for the plasma 24-h area under the curve (AUC₂₄) of polymyxin B and colistin have recently been highlighted, yet approaches to minimize nephrotoxicity risk resulted in mixed outcomes (16–21). More specifically, it remains unclear whether dividing the total daily dose of polymyxins into fractions (e.g., giving twice or thrice daily) can circumvent kidney injury. This study aimed to examine the impact of dose-fractionated polymyxin B (PB) on acute kidney injury (AKI) in a preclinical model and describe the polymyxin B pharmacokinetic (PK) profile (22, 23).

Male Sprague-Dawley rats were divided into three experimental groups (Fig. 1) as follows: once daily (QD), twice daily (BID), and thrice daily (TID) and received subcutaneous clinical grade PB. PB was fixed at 12 mg of drug/kg of body weight per day (allometrically scaled) for experimental groups for 72 h (24, 25). Control groups received equal volumes of normal saline (NS) as per QD protocol. Blood sampling strategies were similar in frequency and volume across groups (Fig. 2) (26, 27). Plasma creatinine content and PB1 were quantified (see the supplemental material) (28). Urine samples were analyzed for biomarkers (KIM-1, IP-10, TIMP-1, CLN, and OPN) (27). Histopathological examination of kidney tissues (n = 32) was performed (23, 29, 30). Midwestern University Institutional Animal Care and Use Committee approved this study (protocol 2677). PK models were fitted using the nonparametric adaptive grid (NPAG) algorithm within the Pmetrics (version 1.5.2) package (3) for R (3, 31). The best-fit PK model is described in the supplemental material. Best fit models were selected according to −2 log-likelihood (−2LL), Akaike’s information criterion (AIC), and rule of parsimony. Goodness of fit of the competing models was evaluated by observed versus predicted plots, bias and imprecision, and visual examination of individual concentration-time profiles. Bayesian posterior predicted concentrations identified maximum concentration of drug in serum (C_max) and minimum concentration of drug in serum (C_min); AUC₂₄ was calculated across 24-h intervals using noncompartmental analysis (3). Analysis of variance (with Geisser and Greenhouse epsilon hat correction method accounting for subjects, treatments, and repeated measures over time) or mixed-effects models were utilized to compare urinary biomarkers, plasma creatinine, and PK indices between study groups (GraphPad Prism 8.2.1, La Jolla, CA). Ordinal logistical regressions were performed on histopathological scores (Stata 13, College Station, TX) (32). All tests were two-tailed with an α level of 0.05.

FIG 1.

Allocation and design of experimental groups. QD, once daily; BID, twice daily; TID, thrice daily; a total of 3, 6, and 9 doses for QD/control, BID, and TID groups, respectively, over 72 hours. Route of administration was subcutaneous. *, Envigo, Indianapolis, IN.

FIG 2.

Schematic of study design for each study group. Day 0 included surgical cannulation of catheter for blood sampling and recovery period. A maximum of 16 PK samples during the 3-day period with no more than 8 samples in a single day. Each sample was replaced with an equal volume of NS to maintain euvolemia. Day 4 included terminal sampling and terminal anesthesia. The left kidney was fixed in 10% formalin for histopathology examination, and the right kidney was stored at −80°C. Animals were maintained in plastic cages on a 12-h light and 12-h dark cycle; food and water were freely accessible at all times. *, Kidney image from Open Michigan (https://commons.wikimedia.org/wiki/File:Kidney_Cross_Section.png).

All 32 animals weighed between 291.1 and 321.1 g. The QD group had significantly lower urine volume compared to the BID group on day −1 (P = 0.01). All experimental groups produced significantly more urine on study days versus the control, except the BID group on day 1 (P = 0.23) (Fig. 3). Tables 1 and 2 and Fig. 4 describe PK models and population parameters. Means for AUC₂₄ (P = 0.15) and C_min (P = 0.19) for QD, BID, and TID groups were similar over the study period while mean C_max differed (P = 0.0063) (Fig. 5). Although the QD group exhibited higher AUC₂₄ than the TID group (P = 0.003) on day 1, no significant effects were observed for overall exposure versus time (P = 0.87). Only the QD group had a lower mean C_min than that of the BID group (P < 0.0001) and TID group (P < 0.0001) during the first 24 h. Compared to that of the TID group, QD and BID groups showed significantly higher C_max during the first 24 h (P = 0.0083 and P = 0.049, respectively). The QD group exhibited the highest C_max on days 2 and 3 (P = 0.0025 and P = 0.0017, respectively). KIM-1 (P < 0.0001), OPN (P = 0.029), IP-10 (P = 0.046), and TIMP-1 (P < 0.0001) exhibited significant treatment effect over the study period while plasma creatinine did not (P = 0.18) (Fig. 6). As a representative biomarker, KIM-1 rose rapidly on days 1 and 2 for experimental groups. The TID group experienced the largest KIM-1 increase (mean increase [95% confidence interval {CI}] day 1 from day −1, 4.44 [0.89, 8.00] ng/ml; P = 0.018) compared to that of the control (0.03 [−0.42, 0.49] ng/ml; P = 0.99). Nonsignificant increases were observed with QD (P = 0.06) and BID (P = 0.06) groups. Further, the TID group exhibited a significant KIM-1 increase on day 2 (mean increase [95% CI], 2.44 [1.22, 3.67] ng/ml; P = 0.0013), and a decrease was observed on day 3 (2.39 [−0.053, 4.83] ng/ml; P = 0.055). Mean KIM-1 changes did not differ on days 2 and 3 between QD and BID groups. Similar trends and significant treatment and time effects were also observed for OPN (P = 0.029), IP-10 (P = 0.046), and TIMP-1 (P < 0.0001) but not CLN (P = 0.093). Significant histopathological damage was observed in the TID group (median score, 2; P = 0.013) (Table 3; see also Fig. S1 to S5 in the supplemental material).

FIG 3.

Mean 24-h urine volume between experimental and control groups. Presurgery mean (standard deviation [SD]) urine volumes were 5.2 (1.7), 8.6 (2.2). 6.4 (2.3), and 5.8 (2.2) ml for QD, BID, TID, and control groups, respectively.

TABLE 1.

Model selection summary^a

Compartmental model	−2LL	OFV change	AIC	Bias (mg/liter)	Imprecision (mg2/liter2)	Bayesian R2
One	1,709	Ref	1,715	−0.48	0.95	0.10
Two	1,709	0	1,720	−0.05	0.90	0.20
Three	1,450	259	1,463	0.13	0.73	0.65

^aOne-compartment model served as the base model to derive two- and three-compartment models; the three-compartment model is the final model. −2LL, −2 log-likelihood; OFV, objective function value; AIC, Akaike’s information criterion.

TABLE 2.

Mean population parameters from the final model

Population parametera	Mean	SD
Ka (h−1)	0.290	0.460
Ke (h−1)	0.411	0.076
Vc (liter)	0.056	0.079
K23 (h−1)	6.214	11.430
K32 (h−1)	3.163	43.00

^aK_a, absorption rate constant; K_e, elimination rate constant; V_C, central volume of distribution; K₂₃, intercompartmental transfer rate (central to peripheral); K₃₂, intercompartmental transfer rate (peripheral to central).

FIG 4.

Goodness-of-fit plot for Bayesian observed versus predicted plasma PB concentrations utilizing the final three-compartment model. Bias was defined as mean weighted prediction error; imprecision was defined as bias-adjusted mean weighted squared prediction error.

FIG 5.

Violin plots of PK indices from the final best-fit model by days for AUC₂₄, C_max, and C_min.

FIG 6.

Plasma creatinine and urinary biomarkers.

TABLE 3.

Ordinal logistic regression of histopathology scorings by group^a

Group	Median score (range)	P value	95% CI
Control	1.5 (1–2)	Referent	Referent
QD	2 (1–3)	0.156	−0.61–3.77
BID	2 (1–3)	0.092	−0.34–4.50
TID	2 (2–3.5)	0.013	0.70–5.88

^aHistopathology examination was conducted by IDEXX BioAnalytics (Columbia, MO, USA); the final histopathology score for an individual animal was calculated based on the highest score from the anatomical structural segment.

We demonstrated that fractionating the PB dose into three daily aliquots resulted in the most AKI as measured by urinary biomarkers (KIM-1, OPN, IP-10, and TIMP-1) with significant differences within 24 h. These markers have demonstrated high sensitivity for kidney injury (33–35). Further, KIM-1 and OPN are directly linked to proximal tubular toxicity and are concordant with previous reports (17, 33, 36). Histopathological findings also indicated that TID dosing led to the most severe AKI. PK analyses demonstrated that the PB exposures were similar across experimental groups.

Contemporary clinical dosing recommendations for PB suggest two divided doses based on a weight-based total daily dose (18, 37). It has been suggested that PB-induced AKI could be minimized by optimizing dosing intervals, similar to that observed with aminoglycosides (5, 38). Wallace et al. utilized a rat model to explore this possibility for colistin methanesulfonate and observed that every 12-h dosing led to a greater number and severity of renal lesions compared to those of every 8-h dosing. They concluded that higher and less fractionated doses of colistin methanesulfonate resulted in more extensive renal damage (19). Abdelraouf et al. administered PB to rats at 20 mg/kg/day or 5 mg/kg every 6 h (17). In contrast to Wallace et al. (but similar to our data), they observed a lower rate of nephrotoxicity in the once-daily group while the split-dosage group experienced a quicker onset of nephrotoxicity (serum creatinine elevation). The authors suggested that a nonpassive, saturable uptake may be responsible for the higher rate of renal injury when PB was given repeatedly (versus daily). Most recently, Okoduwa et al. conducted a retrospective, propensity score-matched study on patients treated with once-daily or twice-daily PB (20). In contrast to our findings, a higher proportion of nephrotoxicity (using clinical criteria) was observed in the once-daily group than in the twice-daily group (47% versus 17%, respectively; P = 0.0005). Our preclinical data agree with those from Abdelraouf et al. that dose fractionation consistently led to more AKI. Our study has limitations, including the use of a preclinical model and a 72-h study window; however, we employed highly sensitive biomarkers and histopathology to define AKI.

To date, this is the first preclinical, dose fractionation study of PB with a rich PK sampling design that allowed PK estimates at an individual level, demonstrating that thrice daily dosing worsened AKI more than once-daily dosing. Further studies are warranted to explore PB exposure linked to toxicity while maximizing efficacy.