Studies have discussed the efficacy and safety of polymyxin B in European and American patients [13, 14], but the efficacy and nephrotoxicity of polymyxin B in elderly infected patients have not been reported. A total of 215 cases of sepsis caused by CRO infection in the elderly were analyzed retrospectively. The total effective rate was 50.7% (109/215), and the total bacterial clearance rate was 44.2% (95/215). The overall incidence of AKI in this study was 37.2% (80/215), the recovery rate of AKI was 35% (28/80), and the incidence of polymyxin B-related AKI ranged from 10.2% to 37.2% (22/215–80/215). Gomes EC et al. [15] found that 31 (33%) of 94 patients with polymyxin B-related AKI in Brazil recovered renal function within 1 year, which is similar to the results of this study. Although our subjects were elderly patients, our study and the study in Brazil have shown that intravenous polymyxin B can improve bacterial clearance and clinical efficiency in infected patients. The clearance of polymyxin B does not depend on glomerular filtration, but the mechanism of its nephrotoxicity is to increase the permeability of the renal tubular epithelial cell membrane, resulting in the inflow of cations, anions and water, resulting in cell swelling and cell lysis [16]. Mohammad et al. [17] detected that the concentration of polymyxin B in human renal tubular cells is approximately 4760 times higher than the extracellular concentration, indicating that the reabsorption of polymyxin B in renal tubules leads to abnormal accumulation in cells and direct damage to tubular cells. This process may explain how polymyxin B causes damage to the kidney when it is cleared as a nonrenal drug. Previous studies reported that the incidence of polymyxin B-induced nephrotoxicity ranged from 18.2% to 46.1% [18, 19]. However, different definitions of nephrotoxicity were applied, and a broad polymyxin B dose was administered, which may explain the different results in previous studies. The incidence of polymyxin B-related nephrotoxicity reported by Kady et al. [20] is lower than that observed by us (23.1% and 37.2%), which is related to their use of the RIFLE standard to define AKI, while we used the KDIGO standard. The subjects were elderly patients who were more prone to renal injury caused by infection and drug induction, which may explain the high incidence of nephrotoxicity.

Infection can cause acute renal injury in patients but can be reversed by appropriate antibiotic treatment [21]. Studies have shown that patients with existing renal dysfunction before polymyxin B treatment may be at higher risk for polymyxin B nephrotoxicity, although it is unclear whether this renal dysfunction is caused by a baseline or severe infection [22]. In this study, 22 cases of AKI occurred in the effective treatment group, which was considered to be caused by polymyxin B nephrotoxicity. The total number of cases was 215, and the AKI incidence was 10.2% (22/215) in the effective treatment group. However, in the ineffective treatment group, acute renal injury was first considered to be related to the aggravation of infection, but the nephrotoxicity of polymyxin B could not be completely excluded. Therefore, the nephrotoxicity of polymyxin B in this study was in the range of 10.2–37.2%. Without sufficient anti-infective treatment and good anti-infective effects, it may be difficult to distinguish polymyxin B-related AKI from infection-induced AKI. For the incidence of AKI, the higher the APACHE II score, the greater the norepinephrine dose, the higher the polymyxin B dose and the longer the duration of PB treatment resulted in a higher incidence of AKI. The incidence of AKI in the polymyxin B noninitial treatment group, treatment ineffective group, bacteria nonclearance group and death group was significantly higher than that in the polymyxin B initial treatment group, treatment effective group, bacteria clearance group and survival group (P < 0.05). The recovery rate of AKI in the effective treatment group was higher than that in the ineffective group, indicating that the more serious the condition of elderly patients was, the more likely they were to have AKI with high-dose and long-term polymyxin B. The high bacterial clearance, effective treatment group and initial treatment group had good infection control, which could prevent AKI caused by infection. In this study, why was there renal function improvement in the treatment ineffective group? It may be that renal perfusion is guaranteed or that other causes of renal damage were corrected. Compared with the group with acute renal injury before PB treatment, the no-AKI incidence in the group with no renal damage before PB treatment was significantly higher than that of chronic renal damage before PB treatment (P < 0.01); when comparing the AKI incidence, the effective group was significantly lower than the ineffective group (P < 0.05). It showed that effective treatment was very important, which not only reduced the 28-d all-cause mortality and the incidence of AKI but also improved the recovery rate of AKI. Compared with the acute renal damage group before PB treatment, acute renal injury was more likely to be aggravated on the basis of previous chronic renal damage because simple acute renal damage can restore renal function through anti-infection treatment or ensuring renal perfusion, but chronic renal damage is difficult to correct. When polymyxin B must be used, it may make things worse and cause damage to the kidney again. To ensure the anti-infective effect and renal function, if the deterioration of renal function cannot be reversed by anti-infective treatment after polymyxin B treatment, CRRT treatment can be used to improve the recovery rate of AKI and improve the prognosis.

The survival curves of the AKI group and non-AKI group showed that the survival time of the non-AKI group was significantly longer than that of the AKI group (P < 0.001), and the mortality of patients without AKI was 37.55% lower than that of AKI, indicating that patients without polymyxin B-related AKI had a better prognosis, long survival time and low mortality. Multicenter studies showed that the mortality of patients with polymyxin B-related AKI was 17% higher than that of patients without AKI [23], and our study showed that the mortality of patients with AKI was 37.55% higher than that of patients without AKI, which was significantly higher than the above research results. This is related to the fact that our selected subjects were elderly patients. Once AKI occurs, the prognosis is extremely poor. For the AKI recovery rate, Gomes’s research shows that the recovery rate of renal function in patients with polymyxin B-related AKI is 33% [15]; most AKI caused by polymyxin is reversible, in part because of the high reabsorption rate of the kidney, especially the proximal renal tubular cells in the renal cortex; and this is due to the saturable active form of Polymyxin B [24]. Our subgroup analysis showed that the recovery rate of stage I AKI was significantly higher than that of stage III AKI (47.1% vs. 12.5%). The higher the norepinephrine dose group, polymyxin B dose 2.0 mg/kg/d group and treatment duration 8–14 d group, the higher the recovery rate of AKI. The recovery rates of AKI in the polymyxin B initial treatment group, effective treatment group and survival group were significantly higher than those in the polymyxin B noninitial treatment group, ineffective treatment group and death group (P < 0.05). Appropriate duration and dose of polymyxin B, initial treatment and effective treatment can improve the recovery rate of AKI. This suggests that once polymyxin B-related AKI occurs in elderly patients with infection, it is necessary to actively treat the primary disease to improve the prognosis of patients. It may also be of certain significance to improve the prognosis by correcting possible reversible risk factors, avoiding aggravation of AKI, and protecting or reversing renal function to the greatest extent.

It is of great significance to analyze the risk factors for polymyxin B-related AKI. Patients with many high-risk factors can avoid the occurrence of AKI to a certain extent by avoiding polymyxin B. A retrospective analysis revealed that the daily dose of polymyxin B, the use of vasoactive drugs and the length of stay in the ICU were independent risk factors for polymyxin B-related AKI in ICU patients [25]. A meta-analysis showed that advanced age, high daily dose, underlying diseases such as diabetes, and associated nephrotoxic drugs were independent predictors of nephrotoxicity [26]. Animal experiments have shown that the pathogenesis of polymyxin B-related AKI is the reabsorption and deposition of polymyxin in renal proximal convoluted tubular cells; this then induces oxidative stress, cell cycle arrest and apoptosis [27, 28]. This cytotoxic damage is dose dependent. The PK study of XuBen et al. showed that different renal functions have different effects on the clearance rate of polymyxin B. Monte Carlo simulation was also carried out to evaluate the possibility of using lower doses of polymyxin B to achieve optimal exposure and reduce toxicity in patients with renal insufficiency [29]. Univariate analysis showed that high APACHE II score, chronic renal insufficiency, high dose and long duration of polymyxin B, high dose of norepinephrine, nonclearance of bacteria and ineffective treatment were predictors of AKI in elderly CRO patients treated with polymyxin B. The multivariate regression analysis that included the above factors showed that high APACHE II score, long duration of polymyxin B, chronic renal insufficiency and ineffective treatment were predictors of AKI in elderly CRO patients treated with polymyxin B. Further ROC curve analysis showed that the cutoff value of serum creatinine before polymyxin B was 73 mmol/L; that is, the sensitivity and specificity of predicting AKI in elderly CRO patients with a cutoff greater than 73 mmol/L were 97.5% and 83%, and the AUC was 0.931. When polymyxin B was used for more than 10.5 days, the sensitivity and specificity of predicting AKI were 86.3% and 67.4%, and the AUC was 0.792. The APACHE II score was greater than 14.5; the sensitivity and specificity of predicting AKI were 83.8% and 66.7%; and the AUC was 0.693. The above results do not mean that AKI will occur when the levels of polymyxin B are higher than the cutoff value, but the probability of occurrence is higher than that lower than the cutoff value. The above cutoff values are lower than we expected, which may be related to the serious condition and more complications of the included elderly patients, such as more hematology and severe patients, low immunity and easy damage to renal function. We can further analyze the cutoff values of different subgroups. Of course, we do not recommend that the dose of polymyxin B be reduced in elderly patients with renal dysfunction because treatment failure may increase the incidence rate and mortality risk. This study and Rigatto's multicenter prospective study confirm this [19].

The above results suggest that for elderly patients with severe disease, basic renal insufficiency, shock, and treatment with high-dose and long-term polymyxin B, clinicians should be particularly alert to the occurrence of AKI, early correction of shock, reduction of drug dose and use time. At present, the drug dose selection of polymyxin B is mostly calculated according to the actual weight of patients. For elderly patients with great changes in pathophysiological state and drug metabolism, the plasma protein binding rate of polymyxin B in general patients and severe patients is very different (50–70% vs. more than 90%) [22]. The above rough calculation method has difficulty meeting the needs of higher levels. Therefore, the consensus of Chinese experts on the clinical application of polymyxin recommends titrating drug dose adjustments with therapeutic drug monitoring (TDM) [30]. However, most TDM detection methods detect free + bound drugs in blood samples, which cannot directly reflect the concentration of free drugs that can exert efficacy in severe patients. More pharmacokinetic studies are urgently needed to meet the clinical needs, such as Target Concentration Intervention (TCI) [31]. The current Polymyxin B PK/PD parameters have poor operability in practical work. There is an urgent need for convenient and simple evaluation index standards to seek a more accurate dose‒response relationship between polymyxin B drug dose and AKI to reduce the occurrence of drug-related AKI.

This study was a multicenter retrospective study. In the future, a well-designed multicenter, prospective controlled trial is needed to study the effectiveness of polymyxin B in elderly patients with infection, its effect on renal function and the treatment of preventing renal damage. In this study, the weight of most critically ill patients was difficult to obtain accurately, and the dosage of polymyxin B was relatively rough. In the later stage, the TCI of polymyxin B can be used to guide individualized administration to further understand the relationship between drug dose and AKI in elderly patients.

In conclusion, rational use of polymyxin B is safe and effective in elderly patients with CRO infection. The high bacterial clearance, polymyxin B effective treatment and initial treatment were conducive to infection control, which can prevent AKI caused by infection and reduce the nephrotoxicity of polymyxin B. The appropriate course and dose of polymyxin B, initial treatment and effective outcome can improve the recovery rate of AKI. This provides a certain basis for the treatment of CRO-infected elderly patients with polymyxin B and has clinical guiding significance.