Clinical Outcomes Associated with Polymyxin B Dose in Patients with Bloodstream Infections Due to Carbapenem-Resistant Gram-Negative Rods

Abstract

There is significant variation in the use of polymyxin B (PMB), and optimal dosing has not been defined. The purpose of this retrospective study was to evaluate the relationship between PMB dose and clinical outcomes. We included patients with bloodstream infections (BSIs) due to carbapenem-resistant Gram-negative rods who received ≥48 h of intravenous PMB. The objective was to evaluate the association between PMB dose and 30-day mortality, clinical cure at day 7, and development of acute kidney injury (AKI). A total of 151 BSIs were included. The overall 30-day mortality was 37.8% (54 of 151), and the median PMB dosage was 1.3 mg/kg (of total body weight)/day. Receipt of PMB dosages of <1.3 mg/kg/day was significantly associated with 30-day mortality (46.5% versus 26.3%; P = 0.02), and this association persisted in multivariable analysis (odds ratio [OR] = 1.58; 95% confidence interval [CI] = 1.05 to 1.81; P = 0.04). Eighty-two percent of patients who received PMB dosages of <1.3 mg/kg/day had baseline renal impairment. Clinical cure at day 7 was not significantly different between dosing groups. AKI was more common in patients receiving PMB dosages of ≥250 mg/day (66.7% versus 32.0%; P = 0.03), and this association persisted in multivariable analysis (OR = 4.32; 95% CI = 1.15 to 16.25; P = 0.03). PMB dosages of <1.3 mg/kg/day were administered primarily to patients with renal impairment, and this dosing was independently associated with 30-day mortality. However, dosages of ≥250 mg/day were independently associated with AKI. These data support the use of PMB without dose reduction in the setting of renal impairment.

INTRODUCTION

Over the last decade, carbapenem-resistant Gram-negative rods (CRGNRs) have emerged as important health care-associated pathogens that are associated with significant morbidity and mortality (1, 2). Due to broad antimicrobial resistance among CRGNRs, clinicians have been increasingly forced to rely upon polymyxins for the treatment of infections caused by these organisms (3–5). The polymyxins, colistin and polymyxin B, came into use in the 1960s, but due to high rates of nephrotoxicity and neurotoxicity, they were replaced with less toxic alternatives (6, 7). Unfortunately due to the early development of polymyxins and a subsequent lack of use in a clinical setting, there is a paucity of information describing optimal dosing (6).

Several recent studies have provided a better description of polymyxin B pharmacokinetics (PK) and pharmacodynamics (PD). As with colistin, the bactericidal activity of polymyxin B is concentration dependent and appears to be best correlated with the area under the concentration-time curve over 24 h in steady state divided by the MIC (AUC/MIC) (8). However, unlike with colistin, polymyxin B elimination is minimally affected by creatinine clearance, and dose reduction in the setting of renal impairment may result in subtherapeutic serum drug concentrations and decreased efficacy (9–11). In clinical practice, polymyxin B doses have traditionally been reduced in the setting of renal insufficiency, and many clinicians are reluctant to use higher doses due to concerns about nephrotoxicity.

Despite these recent advancements in our understanding of the PK and PD of polymyxins, few studies have reported clinical outcomes in patients receiving these agents, and most evaluated colistin (12–14). Due to differences in PK and PD parameters, these results may not apply to polymyxin B (15). A previous study demonstrated an independent association between polymyxin B dosages of ≥200 mg/day and decreased in-hospital mortality, despite an increase in rates of severe kidney injury (16). However, in that study, only a small fraction of patients had CRGNR bacteremia, and weight-based polymyxin B dosing was not assessed. We therefore conducted a study of patients with CRGNR bacteremia to correlate weight-based polymyxin B dosing regimens to clinical outcomes such as mortality, clinical cure, and acute kidney injury (AKI).

MATERIALS AND METHODS

Study design.

This retrospective cohort study evaluated all adult patients with a bloodstream infection (BSI) due to carbapenem-resistant Enterobacteriaceae (CRE), Pseudomonas aeruginosa, or Acinetobacter baumannii from January 2006 until December 2013 at two large, academic, tertiary care medical centers in New York, NY. Patients were included if they were treated with polymyxin B for ≥48 h within 5 days of the first positive blood culture. Exclusion criteria consisted of age of <18 years, receipt of treatment for prior qualifying BSIs within the previous 30 days, or polymicrobial bacteremia within 5 days of the initial CRGNR-positive blood culture, with the exception of a single isolation of coagulase-negative staphylococci.

The primary objective of the study was to determine whether the median polymyxin B dosage (in milligrams per kilogram [of total body weight]/day) over the first 7 days of polymyxin B therapy was independently associated with 30-day mortality. We evaluated polymyxin B dosing as a continuous variable and using various weight-based dose cutoffs. Time to 30-day mortality was determined starting from the onset of infection, defined as the time of collection of the first CRGNR-positive blood culture resulting in the use of polymyxin B. Over the study period, prescribers variably used ideal, adjusted, or total body weight (TBW) when selecting dose regimens. For consistency, the polymyxin B dose calculations presented in the discussion of our results were all based on TBW. Many patients received dose reduction for renal impairment in accordance with an institution-specific dosing algorithm that recommended 1.5 mg/kg/day in patients with a creatinine clearance (CL_CR) between 30 and 50 ml/min or 1 to 1.5 mg/kg every 2 to 3 days in those with a CL_CR of <30 ml/min and in those receiving renal replacement therapy (RRT).

Secondary objectives included clinical cure at day 7 and the development of AKI by day 7 of polymyxin B therapy. Clinical cure was defined as a composite outcome of survival and resolution of all signs and symptoms of infection by day 7 of polymyxin B therapy. AKI was defined as an increase in serum creatinine 1.5 times the value at polymyxin B initiation or the initiation of RRT by day 7 of polymyxin B treatment, as defined to by the Risk, Injury, Failure, Loss, and End-stage kidney (RIFLE) classification system (17). Patients receiving RRT at polymyxin B initiation were excluded from the AKI analysis.

Data points collected at the time of admission included patient demographics, the Charlson comorbidity index (CCI) (18), and presence of immunosuppression (defined as the receipt of chemotherapeutic agents within 90 days or receipt of corticosteroids at doses of >5 mg/day prednisone equivalents or other immunosuppressive agents for at least 14 days in the past 30 days). Additional data points collected at the time of BSI and over the course of therapy included presumed source of BSI (as documented by the treating physician), source control, the Pitt bacteremia score (PBS) (19), concomitant use of antibiotics and potentially nephrotoxic agents, height and weight, use of RRT, the presence of renal impairment at polymyxin B initiation, stay in the intensive care unit (ICU), the presence of septic shock (20), and date of death. Source control was defined as the removal of urinary catheters or central venous catheters, or drainage of intra-abdominal collections, where applicable, within 7 days of index pathogen isolation. Renal impairment was defined as CL_CR of <50 ml/min or the use of RRT. CL_CR was calculated using the Cockcroft-Gault equation using actual or adjusted body weight (21). Active agents were defined as antibiotics to which the organism tested susceptible in vitro, using Clinical and Laboratory Standards Institute (CLSI) interpretive breakpoints (22). We also recorded the use of polymyxin B loading doses, defined as an initial dose of ≥2.5 mg/kg. Development of AKI was assessed by collecting serum creatinine at polymyxin B initiation and serum creatinine peak during the first 7 days of therapy, as well as by determining whether RRT was initiated during the first 7 days of polymyxin B treatment.

Microbiology.

All isolates were identified by the microbiology laboratories located within the respective study centers. MIC testing methods varied during the study period but included broth microdilution (Thermo Scientific) or Etest (bioMérieux) for polymyxin B and Vitek 2 (bioMérieux), Etest, disk diffusion, or broth microdilution for β-lactam antibiotics and other active agents. In accordance with current CLSI breakpoints for the Enterobacteriaceae, carbapenem resistance was defined as a meropenem MIC of ≥4 mg/liter (22).

Statistics.

Statistical analyses were conducted using IBM SPSS Statistics, version 21 (SPSS, Chicago, IL). Variables were compared using χ² test or Fisher's exact test for categorical variables as appropriate and the Mann-Whitney U test for continuous variables. All tests were two-tailed, and a P value of <0.05 was considered significant. Baseline differences between groups and variables with P values of <0.1 on univariable analysis were considered for inclusion into two multivariable logistic regression models identifying factors associated with 30-day mortality and AKI, respectively. Kaplan-Meier survival estimation was also used to evaluate 30-day mortality.

RESULTS

A total of 338 CRGNR blood isolates were identified during the study period in patients who had received at least one dose of polymyxin B. Of these, 187 were excluded from the study: 90 (48.1%) were from patients who had polymicrobial BSIs, 58 (31.0%) were from patients who were treated with polymyxin B for <48 h, 26 (13.9%) occurred within 30 days of a prior qualifying BSI, 7 (3.7%) were from patients who were initiated on polymyxin B >5 days after the isolation of the first positive blood culture, and 6 (3.2%) were from patients of <18 years of age. One hundred fifty-one episodes of bacteremia were included in the final analysis. One hundred two of the affected patients (67.6%) were in an ICU at the time of the first positive blood culture, and 55 (36.4%) presented with septic shock. The most common CRGNR pathogens were Klebsiella pneumoniae (n = 92 [60.9%]), Acinetobacter baumannii (n = 32 [21.2%]), and Pseudomonas aeruginosa (n = 17 [11.3%]). The primary outcome of 30-day mortality occurred in 54 patients (37.8%), and 96 (63.6%) achieved the secondary outcome of clinical cure by day 7. The polymyxin B MIC₅₀ and MIC₉₀ for the study population were 1 mg/liter and 2 mg/liter, respectively.

The median duration of polymyxin B therapy was 14.0 days (interquartile range [IQR] = 8.0, 21.5), with a median dose of 1.3 mg/kg/day (IQR = 0.9, 2.0). No significant differences with regard to effectiveness or safety were found when dose was examined as a continuous variable. However, a post hoc analysis demonstrated that patients treated with polymyxin B dosages less than the median (<1.3 mg/kg/day) experienced significantly higher 30-day mortality than those treated with dosages of ≥1.3 mg/kg/day (33 of 71 [46.5%] versus 21 of 80 [26.5%]; P = 0.02). There was no significant difference in the rate of clinical cure at day 7 between those who received dosages of <1.3 mg/kg/day and those who received higher dosages (44 of 71 [62.0%] versus 52 of 80 [65.0%]; P = 0.70). Of 71 patients who received doses less than the median, 58 (81.7%) met predefined criteria for baseline renal impairment. Several other dose cutoffs were examined, but no significant differences in 30-day mortality or AKI were observed (Table 1).

TABLE 1.

Description of trends in 30-day mortality and acute kidney injury stratified by polymyxin B dosing groups

Dosing range	Value fora:
	30-day mortality	Acute kidney injury
mg/kg/day
<1.3	33/71 (46.5)	13/36 (36.1)
1.3–2.4	16/61 (26.2)	21/54 (38.9)
≥2.5	5/19 (26.3)	5/19 (26.3)
mg/day
<150	39/99 (39.4)	21/60 (35.0)
150–249	11/40 (27.5)	10/37 (27.0)
≥250	4/12 (33.3)	8/12 (66.7)

^aValues are presented as no./subgroup (percent).

In addition to polymyxin B, 127 patients (84.1%) received combination therapy. β-Lactam antibiotics were administered concomitantly in 82 patients (54.3%) and were the most commonly used combination agents. Seventy-six patients (50.3%) received one or more additional active agents (sometimes with a β-lactam agent). The most frequently used β-lactam antibiotics were meropenem and cefepime. Almost all patients who received these agents were dosed with the renally adjusted equivalent of meropenem at 500 mg every 6 h or cefepime at 2 g every 8 h. Meropenem was not considered to be an active therapy because this traditional dosing regimen would not have provided adequate coverage for organisms with meropenem MICs of ≥4 mg/liter. Cefepime was also not considered active therapy because all patients that were included and received cefepime had isolates that were resistant per current CLSI breakpoints (22). Other agents used in combination included aminoglycosides, tigecycline, minocycline, and fluoroquinolones. Use of these various combination strategies was not different between survivors and nonsurvivors. Only 19 patients (12.6%) received polymyxin B dosages of ≥2.5 mg/kg/day. When examined separately, the 30-day mortality and clinical cure at day 7 for this dosing group were 26.3% and 63.2%, respectively.

A bivariate analysis comparing those who survived to day 30 and those who did not is summarized in Table 2. In addition to being more likely to have received a polymyxin B dosage of ≥1.3 mg/kg/day over the first 7 days, survivors were significantly younger than nonsurvivors (60 years [IQR = 48, 74] versus 70 years [IQR = 55, 76]; P = 0.04) and more likely to have a lower Pitt bacteremia score (3.0 [IQR = 1.0, 6.0] versus 5.0 ([IQR = 3.0, 7.0]; P < 0.001) and a lower Charlson comorbidity index (3.0 [IQR = 2.0, 4.0] versus 4.0 [IQR = 2.3, 5, 8]; P = 0.03). Survivors also had a significantly longer time from first positive blood culture to start of active therapy than did nonsurvivors (2 days versus 1 day; P = 0.01) and were more likely to be immunosuppressed (51.5% versus 29.6%; P = 0.02). There was no association observed between polymyxin MIC and clinical outcomes. In a multivariable logistic regression model (Table 3), a polymyxin B dosage of <1.3 mg/kg/day over the first 7 days was independently associated with 30-day mortality (odds ratio [OR] = 1.58; 95% confidence interval [CI] = 1.05 to 1.81; P = 0.04). Pitt bacteremia score, time from first positive blood culture to start of active therapy, immunosuppression, and intra-abdominal source were also independently associated with 30-day mortality. Age and Charlson comorbidity index were not independent predictors of mortality. Finally, Kaplan-Meier estimation showed a significant decrease in mortality for patients who received ≥1.3 mg/kg/day (P = 0.008) (Fig. 1).

TABLE 2.

Characteristics associated with 30-day mortality among patients who received intravenous polymyxin B^a

Parameter	Value for:		P
	Survivors (n = 97)	Nonsurvivors (n = 54)
Demographic factors
Female	40 (41.2)	26 (48.2)	0.52
Age, yrs	60 (48, 74)	70 (55, 76)	0.04
Total body wt, kg	74 (63, 87)	80 (66, 92)	0.25
Charlson comorbidity index	3 (2, 4)	4 (2, 6)	0.03
Immunosuppressed	50 (51.5)	16 (29.6)	0.02
Hospital length of stay prior to first positive blood culture, days	15 (3, 49)	24 (12, 48)	0.07
Isolated pathogen
Klebsiella pneumoniae	60 (61.9)	32 (59.3)	0.89
Acinetobacter baumannii	21 (21.6)	11 (20.4)	1.00
Pseudomonas aeruginosa	9 (9.3)	8 (14.8)	0.45
Other Enterobacteriaceaeb	7 (7.2)	3 (5.6)	0.74
Presumed source of infection
Respiratory	19 (19.6)	15 (27.8)	0.34
Intra-abdominal	16 (16.5)	16 (29.6)	0.09
Genitourinary	17 (17.5)	5 (9.3)	0.26
Central line associated	11 (11.3)	2 (3.7)	0.14
Gut translocation	8 (8.2)	4 (7.4)	1.00
Skin and soft tissue	6 (6.2)	2 (3.7)	0.71
Multiplec	11 (11.3)	9 (16.7)	0.50
Unknown	9 (9.3)	1 (1.9)	0.10
Source control	26 (26.8)	15 (27.8)	1.00
Characteristics at polymyxin B initiation
Pitt bacteremia score	3 (1, 6)	5 (3, 7)	<0.001
Serum creatinine, mg/dl	1.5 (0.8, 2.1)	1.1 (0.9, 2.0)	0.35
Renal impairment	47 (48.5)	30 (55.6)	0.51
CLCR of <50 ml/min	22 (22.7)	13 (24.1)	0.84
IHD	13 (13.4)	5 (9.3)	0.62
CRRT	12 (12.4)	12 (22.2)	0.11
ICU	62 (63.9)	40 (74.1)	0.27
Septic shock	32 (33.0)	23 (42.6)	0.32
Time from first positive blood culture to start of active therapy, days	2.0 (0.2, 2.3)	1.0 (0.0, 1.8)	0.02
Polymyxin B MIC of ≤1 mg/liter (MIC50)	66 (68.8)	36 (66.7)	0.94
Polymyxin B MIC of ≤2 mg/liter (MIC90)	87 (90.6)	49 (90.7)	1.00
Meropenem MIC of 4–8 mg/liter	11 (13.6)	3 (6.3)	0.32
Details of polymyxin B therapy over the first 7 days
Dosage, mg/day	123 (78, 190)	112 (75, 155)	0.35
Dosage, mg/kg/day	1.5 (1.1, 2.4)	1.3 (0.9, 2.0)	0.10
Dosage of <1.3 mg/kg/day	38 (39.2)	33 (61.1)	0.02
Load of ≥2.5 mg/kg as first dose	21 (21.6)	13 (24.1)	0.89
Monotherapy	15 (15.5)	9 (16.7)	1.00
Combination with 1 or more active agents	46 (47.4)	30 (55.6)	0.43
Concomitant β-lactam	58 (59.8)	24 (44.4)	0.10

^aCategorical variables are presented as no. (percent), and continuous variables are presented as median (interquartile range). Abbreviations: CL_CR, estimated creatinine clearance; IHD, intermittent hemodialysis; CRRT, continuous renal replacement therapy; ICU, intensive care unit.

^bOther Enterobacteriaceae represented here include Escherichia coli, Enterobacter cloacae, and Klebsiella oxytoca.

^cPatients with multiple sources primarily included those with combinations of presumed pulmonary, genitourinary, and/or intra-abdominal infections.

TABLE 3.

Multivariable analyses of characteristics associated with 30-day mortality

Parameter	OR (95% CI)a	P
Age, yrs	1.64 (0.73–3.70)	0.23
Charlson comorbidity index	1.07 (0.91–1.25)	0.42
Immunosuppression	0.36 (0.15–0.84)	0.02
Intra-abdominal source	3.67 (1.38–9.77)	0.009
Time from first positive blood culture to start of active therapy (each day)	0.79 (0.63–0.98)	0.03
Pitt bacteremia score	1.22 (1.05–1.41)	0.01
Polymyxin B dosage of <1.3 mg/kg/day	1.58 (1.05–1.81)	0.04

^aHosmer-Lemeshow χ² test = 6.95 (P = 0.58). Abbreviations: OR, odds ratio; CI, confidence interval.

FIG 1.

Kaplan-Meier 30-day survival curve comparing patients who received <1.3 mg/kg of polymyxin B to those who received ≥1.3 mg/kg.

After exclusion of 42 patients who were receiving RRT at the time of polymyxin B initiation, 109 patients were evaluated for the development of AKI. Within this group, 39 patients (35.8%) experienced AKI by day 7 of polymyxin B therapy. A bivariate analysis of those who developed AKI is summarized in Table 4. Polymyxin B dosage stratified by weight (milligrams per kilogram per day) and the use of a loading dose were not different between groups. Development of AKI was associated with female sex and concomitant aminoglycoside therapy. Administration of other nephrotoxins was not associated with AKI. AKI occurred more frequently in those who received polymyxin B dosages of ≥250 mg/day than in those who received lower dosages (8 of 12 [66.7%] versus 31 of 97 [32.0%]; p = 0.03). In the multivariable logistic regression model (Table 5), independent associations with the development of AKI were found with a polymyxin B dosage of ≥250 mg/day (OR = 4.32; 95% CI = 1.15 to 16.25; P = 0.03), female sex (OR = 2.44; 95% CI = 1.05 to 5.68; P = 0.04), and concomitant aminoglycoside therapy (OR = 3.19; 95% CI = 1.11 to 9.23; P = 0.03).

TABLE 4.

Characteristics potentially associated with acute kidney injury at day 7 of polymyxin B therapy

Parameter	Value for patients witha:		P
	No AKI (n = 70)	AKI (n = 39)
Demographic factors
Female	27 (38.6)	24 (61.5)	0.04
Age, yrs	62 (52, 76)	64 (48, 74)	0.65
Charlson comorbidity index	3 (1, 5)	4 (2, 4)	0.27
Pitt bacteremia score	3 (2, 6)	3 (1, 5)	0.61
30-day mortality	20 (28.6)	17 (43.6)	0.17
Clinical cure at 7 days	49 (70.0)	23 (60.0)	0.24
Details of polymyxin B dosing regimens
Dosage, mg/day	129 (99, 190)	139 (112, 212)	0.13
Dosage, mg/kg/day	1.8 (1.3, 2.5)	2.0 (1.5, 2.6)	0.26
Dosage of ≥250 mg/day	4 (5.7)	8 (20.5)	0.03
Load of ≥2.5 mg/kg as first dose	10 (14.3)	9 (23.1)	0.30
Use of concomitant nephrotoxins
Vancomycin	26 (37.1)	19 (48.7)	0.33
Aminoglycosides	8 (11.4)	11 (28.2)	0.05
Calcineurin inhibitors	13 (18.6)	4 (10.3)	0.29
Intravenous contrast	8 (11.4)	7 (17.9)	0.51
Amphotericin B	2 (2.9)	0 (0.0)	0.54
Any nephrotoxin	41 (58.6)	30 (76.9)	0.09
Nephrotoxins other than aminoglycosides	39 (55.7)	25 (64.1)	0.52

^aCategorical variables are presented as no. (percent), and continuous variables are presented as median (interquartile range).

TABLE 5.

Multivariable analyses of characteristics associated with acute kidney injury at day 7 of polymyxin B therapy

Parameter	OR (95% CI)a	P
Sex, female	2.44 (1.05–5.68)	0.04
Concomitant aminoglycoside therapy	3.19 (1.11–9.23)	0.03
Dose of ≥250 mg/day	4.32 (1.15–16.25)	0.03

^aHosmer-Lemeshow χ² test = 0.11 (P = 0.99). Abbreviations: OR, odds ratio; CI, confidence interval.

DISCUSSION

To our knowledge, this retrospective study is the first to find an independent association between polymyxin B dosage stratified by weight and 30-day mortality. Polymyxin B dosages of <1.3 mg/kg/day (the median dosage used in our patients), Pitt bacteremia score, time to start of active therapy, immunosuppression, and intra-abdominal source were found to be independent predictors of 30-day mortality. Interestingly, immunosuppression and longer time to active antibiotic therapy were associated with decreased mortality. Two prior studies conducted at our institution evaluated immunocompromised patients with bacteremia due to CRE and found that death often occurred within 3 to 4 days of BSI onset (23, 24). This early mortality may have resulted in many of the more severely ill immunocompromised patients being excluded because they did not receive 48 h of polymyxin B. Those with longer time to active antibiotic therapy were less likely to have a history of multidrug-resistant organisms and may have required less complicated treatment regimens.

Our primary findings are consistent with the results of a previous study conducted by Elias and colleagues that evaluated 276 patients with infections due to Acinetobacter baumannii and Pseudomonas aeruginosa (16). They demonstrated that polymyxin B dosages of <200 mg/day were independently associated with increased in-hospital mortality (adjusted odds ratio [aOR] = 4.07; 95% CI = 2.22 to 7.46). An earlier retrospective study conducted by Oliveira and colleagues examined 41 patients with infections due to A. baumannii and found a 30-day mortality rate of 61% (25). Combination therapy with β-lactams was utilized in relatively few patients in each of these studies (16, 25).

The present study has several key differences from the preceding studies. First, we evaluated weight-based polymyxin B dosing regimens. Additionally, the above-mentioned studies included patients with pathogens isolated from any source (16, 25). By including only patients with bacteremia, there was less difficulty in differentiating colonization from infection. Third, the majority of our patients were infected with CRE, whereas in the above-mentioned studies, only patients infected with A. baumannii and P. aeruginosa were examined. Notably, we did not find differences in outcomes between patients infected with CRE and those infected with A. baumannii and P. aeruginosa. Finally, the majority of patients in our study received concomitant therapy with β-lactam antibiotics, particularly carbapenems, whereas the majority of patients in the study by Elias and colleagues received polymyxin B monotherapy (16). Although isolates in our study were carbapenem resistant, and none had MICs that could potentially be treated with carbapenems given our dosing regimens, it is possible that their use contributed to the lower overall mortality (37.8%) observed in our study population compared to other studies evaluating outcomes related to polymyxin B (16, 25).

The correlation between increased polymyxin B dose and improved clinical effectiveness is further supported by the findings of several PK/PD studies. These studies suggest that current dosing recommendations may provide insufficient serum drug levels and suboptimal AUC/MIC ratios (8, 9, 11). In fact, Sandri and colleagues demonstrated that dosing regimens of 2.5 to 3 mg/kg/day may be needed to reliably achieve optimal bactericidal activity (9). They also found that polymyxin B elimination is minimally affected by decreased creatinine clearance and questioned the practice of dose adjustment in the setting of renal impairment. In our study, baseline renal impairment was the most common reason for polymyxin B dose adjustment and the receipt of doses less than the median. Although receipt of doses less than the median was independently associated with 30-day mortality, the findings for baseline renal impairment were not different between survivors and nonsurvivors, adding support to the findings of the PK/PD studies described above.

Another key finding of the present study was the increased development of AKI in patients who received polymyxin B dosages of ≥250 mg/day, occurring in 35.8% of patients. Two prior studies have found that patients who received dosages of ≥200 mg/day had a significantly higher risk of severe renal impairment (16, 26). A third study found that increasing dosages between 150 and 199 mg/day were associated with AKI and that this association was irrespective of weight and suggests a total daily dose toxicity threshold (27). Our study adds support to these findings by demonstrating that the risk of nephrotoxicity with polymyxin B is increased with higher total daily doses. We hypothesize that total daily dose, as opposed to weight-based dose, correlates best with the amount of drug accumulation in the proximal tubular cells of the kidney, which is thought to be the primary mechanism of polymyxin B-induced AKI (28). We also found that the concomitant use of aminoglycosides increases this risk, a finding that was not obtained in other studies. In our study, the rate of AKI at 7 days was 35.8%, which is comparable to the findings of three other studies, which found rates of AKI at any time during polymyxin B treatment to be 21.1% to 60% (29–31). In all of these studies, a large percentage of patients received concomitant nephrotoxins and any number of factors may have accounted for AKI development. We evaluated only the first 7 days of therapy, rather than the entire course, to limit the number of confounding factors that may have contributed to AKI development.

Our study is associated with several important limitations. First, this study was not a randomized comparison of different polymyxin B dosing regimens. Thus, differences in outcomes may have been related to factors other than polymyxin B dosing. In order to minimize this confounding, we conducted robust multivariable analyses and demonstrated that polymyxin B dose was independently associated with mortality and nephrotoxicity. By including only bloodstream isolates, the applicability of our results may be difficult to extrapolate to other sites of infection. Perhaps most importantly, many patients in our study received dosages much lower than the proposed 2.5 to 3 mg/kg/day (9). However, because these weight-based dose cutoffs have not been well studied in a clinical setting, we examined multiple potential dose cutoffs in our analysis. We found that 30-day mortality in this higher dosing group (≥2.5 mg/kg/day) was not markedly different compared to all patients who received dosages of ≥1.3 mg/kg/day (26.3% versus 26.5%). Our findings are limited by the fact that we made multiple comparisons to identify statistically significant dose cutoffs. We do not believe that a statistical adjustment for multiple comparisons is feasible given the limited numbers of patients in the study with the outcomes of interest. Furthermore, the selected dose cutoff in the mortality analysis primarily reflected patients who received lower doses because of renal insufficiency. The relatively poor outcomes in these patients are consistent with pharmacokinetic data that would predict suboptimal exposures in patients who receive these dose adjustments (9). With regard to treatment strategy, although the use of combination therapies was not different between groups, the types of regimens utilized varied greatly and may have also had an impact on clinical outcomes such as 30-day mortality and AKI.

In conclusion, the administration of polymyxin B dosages of <1.3 mg/kg/day was independently associated with increased 30-day mortality. Due to the infrequent use of the recently suggested dosage regimens of 2.5 to 3 mg/kg/day in our study, the therapeutic benefit and safety of such regimens in the setting of combination therapy could not be adequately evaluated. Development of AKI by day 7 of polymyxin B therapy was found to have an independent association with dosages of ≥250 mg/day. The benefits of higher doses and the increased risk of AKI should be weighed against the high mortality associated with CRGNR infections in each patient. Our findings provide additional support to suggest that polymyxin B dose reduction in the setting of renal impairment may result in decreased efficacy and higher 30-day mortality. Further advances in optimization of polymyxin B dosing would benefit from evaluation in larger trials, particularly with patients receiving higher total daily doses.