Introduction

Severe Gram-positive bacteria infections, especially if due to methicillin-resistant Staphylococcus aureus (MRSA), represent a major nosocomial problem associated with high morbidity and mortality [1]. MRSA infections are responsible for more than one tenth of all cases of septic shock [2] and can be an independent determinant of in-hospital mortality among critically ill patients [3]. Vancomycin remains one of the first antibiotic choices for the treatment of severe MRSA infections [4]. However, several studies suggested that standard vancomycin regimens were inadequate to efficiently treat more resistant strains of MRSA, especially those with a minimum inhibitory concentration (MIC) for the drug above 1 μg/ml [5]. Recent recommendations for intermittent infusion have, thus, highlighted the importance of achieving serum trough vancomycin concentrations of 15–20 μg/ml when treating MRSA infections in critically ill patients [4].

A key concern during vancomycin therapy is the development of acute kidney injury (AKI), with an incidence as high as 40 % in some studies [6, 7]. Several retrospective reports have suggested that the risk of AKI is increased when initial vancomycin doses exceed 4 g/day or when drug concentrations exceed 15 μg/ml [8–11]. A recent randomized clinical trial also showed that vancomycin was associated with a greater incidence of AKI when compared to linezolid in the treatment of MRSA pneumonia [12]. Nevertheless, these studies did not specifically focus on critically ill patients, in whom the pharmacokinetics of vancomycin can be significantly altered, resulting in inadequate drug concentrations [13], and who are at high risk of developing AKI [14]. In this setting, one effective approach to maximizing the antimicrobial activity of vancomycin is to use a continuous infusion [15]. Although this approach has not been shown to improve clinical efficacy in MRSA infections when compared to standard intermittent regimens [16], continuous infusion may achieve therapeutic vancomycin concentrations more rapidly, may facilitate dose adjustment, and may reduce variability in drug levels [16]. A continuous infusion may also decrease the risk of AKI when compared to standard repeated infusions [17], although few data are available on the potential nephrotoxicity of continuous vancomycin infusions in critically ill patients [18]. The objectives of this study were, therefore, to evaluate the incidence of AKI during a continuous infusion of vancomycin in septic patients and to identify the risk factors and time course (early vs. late) of the development of AKI in this population, as well as the long-term consequences of continuous vancomycin infusions on renal function.

Methods

Patients and data collection

We reviewed a database of adult (>18 years of age) patients who were treated with a continuous infusion of vancomycin, either as monotherapy or combined with other antimicrobials, in our multidisciplinary Department of Intensive Care, over a two-year period (from 1st of January 2008 to the 31st of December 2009). Patients were included in the analysis if they: (a) had a diagnosis of sepsis, according to standard criteria [19]; (b) had received vancomycin therapy for more than 48 h; (c) had undergone daily drug monitoring; (d) had a total body weight (TBW) noted in the data file; (e) were not being treated with continuous renal replacement therapy (CRRT) or conventional hemodialysis (HD) at the time of initiation of vancomycin therapy. We excluded patients who had received vancomycin within the week prior to the present episode, those with end-stage renal failure chronically treated with HD, and those with missing data on serum creatinine (sCr), urine output, and/or concomitant administration of nephrotoxic agents. The study was approved by the Ethics Committee, which waived the need for informed consent in view of its retrospective nature.

We collected demographics, pre-existing chronic diseases, admission diagnosis, biological, and microbiological data for all patients. The severity of illness is assessed routinely on admission using the Acute Physiology and Chronic Health Evaluation (APACHE) II score [20] and the Sequential Organ Failure Assessment (SOFA) [21] score. SOFA scores are also routinely calculated daily during the intensive care unit (ICU) stay. sCr levels at the start of and during vancomycin therapy were recorded, as was the creatinine clearance (CrCl) on the first day of therapy, calculated from the 24-h urine collection and normalized to the body surface area (BSA). Treatment with vasoactive drugs and mechanical ventilation during vancomycin therapy was recorded, as was the length of ICU stay and overall ICU mortality.

Vancomycin therapy

In our department, continuous infusion is the standard mode of administration for vancomycin for ICU patients when treating suspected or documented infections due to MRSA or other Gram-positive bacteria. Serum concentrations of vancomycin were determined by particle enhanced turbidimetric inhibition immunoassay (Dimension Xpand, Siemens Healthcare Diagnostics, Newark, DE). The limit of quantification and total imprecision of the assay were 0.8 mg/ml and <5 %, respectively. Blood samples (3 ml) were taken every day at 8 a.m. and sent immediately to the central laboratory. The exact sampling time was recorded by the nursing staff in the patient data monitoring system (Picis Critical Care Manager, Picis Inc., Wakefield, MA).

Vancomycin was administered as a 15 mg/kg loading dose over 60–90 min (maximum rate: 1 g/h), followed by a 20–30 mg/kg daily continuous infusion calculated using TBW, according to the CrCl, as previously described [11]. The aim was to reach serum concentrations between 20 and 30 μg/ml. Thus, if the serum vancomycin concentration was less than 20 μg/ml, an additional bolus of 500 mg was given, followed by an increase in the daily dose by 500–1,000 mg. If the concentration was greater than 30 μg/ml, the continuous infusion was discontinued for at least 4 h and the daily dose reduced by 500–1,000 mg per day.

Vancomycin concentrations were routinely measured every day at 8 a.m. in all patients receiving vancomycin continuous infusion. We specifically considered: drug concentration on day 1; mean vancomycin concentration over the first 3 days of therapy (C mean, or the first 2 days of therapy if the drug was administered for only 48 h); and maximal drug concentrations over the entire vancomycin therapy (C max).

Definitions

The development of AKI was defined as a daily urine output <0.5 ml/kg/h and/or an increase in the sCr level of at least 0.3 mg/dl or a 1.5–2 times increase from baseline (i.e., the first day of vancomycin therapy), whichever was greater, on at least two consecutive days during and in the 72 h after vancomycin discontinuation. We considered only the daily urine output, as this measure may be less influenced by the effects of specific therapies, such as volume loading and diuretic administration, compared to partial urine outputs over 6–12 h. We also recorded any potentially nephrotoxic agents which were administered concomitantly, including angiotensin-converting enzyme inhibitors (ACEIs), angiotensin II receptor blockers (ARBs), non-steroidal anti-inflammatory drugs (NSAIDs), aminoglycosides, amphotericin B, colistin, and immunosuppressant agents (e.g., calcineurin inhibitors, such as ciclosporin and tacrolimus). Exposure to intravenous contrast media was considered to be potentially nephrotoxic only if given within 5 days before the start of vancomycin therapy and/or during drug infusion.

AKI was considered as ‘early’ if it developed within the first 2 days of therapy and ‘late’ if it occurred thereafter. The development of severe AKI was defined as the need for CRRT or HD during vancomycin therapy and within 3 days after drug discontinuation. To evaluate the long-term effects on renal function, we recorded the need for CRRT/HD and sCr concentrations at 1 month (±7 days) after the initiation of therapy.

Statistical analysis

Statistical analyses were performed using the SPSS 18.0 for Windows NT software package (SPSS Inc., 2004, Chicago, IL). Descriptive statistics were computed for all study variables. A Kolmogorov–Smirnov test was used, and histograms and normal–quantile plots were examined to verify the normality of distribution of continuous variables. Discrete variables were expressed as counts (percentage) and continuous variables as mean ± standard deviation (SD) or median [25th–75th percentiles]. Demographics and clinical differences between groups (AKI vs. no AKI; early AKI vs. late AKI) were assessed using a Chi-square test, Fisher’s exact test, Student’s t-test, or Mann–Whitney U-test, as appropriate. Multivariable logistic regression analysis with AKI development during vancomycin therapy as the dependent variable was performed in all patients; only variables associated with a higher risk of inadequate concentrations (p < 0.2) on a univariate basis were introduced in the multivariate model. Collinearity between variables was excluded prior to modeling; the influence that individual observations could exert on the coefficients was also evaluated. Odds ratios (ORs) with 95 % confidence intervals (CIs) were computed. Variables considered in the analysis were demographic variables, co-morbidities, APACHE II score on admission, SOFA scores on days 1 and 2 of therapy, type of admission (medical or surgical), source of infection, mechanical ventilation, administration of vasopressor agents, vancomycin doses (loading and daily dose; mean dose over the first 3 days of therapy, Dose mean), vancomycin concentrations (on day 1, C mean and C max), and CrCl. The same analysis was then performed for early and late AKI, separately. The discriminative ability of variables identified by multivariate analysis to predict the development of AKI was evaluated using receiver operating characteristic (ROC) curves with the corresponding area under the curve (AUC). The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were also computed. A p-value < 0.05 was considered as being statistically significant.

Results

During the study period, 207 patients met the inclusion criteria (Fig. 1); their characteristics are shown in Table 1. The lung (n = 103, 50 %) was the most common source of infection, followed by the abdomen (n = 44, 21 %) and the urine (n = 27, 13 %). Gram-positive organisms were identified in 112 patients (65 %), including methicillin-sensitive S. aureus (MSSA) in 30 (27 %), MRSA in 22 (20 %), coagulase-negative Staphylococcus ssp. in 30 (27 %), Enterococcus ssp. in 21 (19 %), and others in 9 (8 %).

Fig. 1
figure 1

Flow chart of the study. sCr serum creatinine

Full size image
Table 1 Characteristics of the study population, with regard to the development of acute kidney injury (AKI) during vancomycin therapy (n = 207, no missing values)
Full size table

Fifty patients (24 %) developed AKI during vancomycin therapy. These patients more frequently had diabetes and cirrhosis and were more severely ill than the other patients, as shown by higher APACHE II and SOFA scores on admission (Table 1). In addition, patients developing AKI were more frequently exposed to other concomitant nephrotoxics and had lower CrCl values on the first day of therapy than those who did not develop AKI. Finally, patients who developed AKI had a longer duration of therapy, higher vancomycin concentrations on day 1, and higher C mean values than those who did not develop AKI. AKI developed in 2/34 (6 %) patients with C mean < 15 μg/ml, in 5/47 (11 %) patients with C mean 15–20 μg/ml, in 18/60 (30 %) patients with C mean 20–25 μg/ml, in 15/39 (38 %) patients with C mean 25–30 μg/ml, and in 10/27 (37 %) patients with C mean > 30 μg/ml (Fig. 2). Patients with AKI had a longer ICU stay and higher mortality rate (40 vs. 18 %) than the other patients.

Fig. 2
figure 2

Proportion of patients developing early and late acute kidney injury (AKI) during vancomycin therapy, according to different ranges of mean vancomycin concentrations during the first 3 days of therapy

Full size image

AKI developed early in 28 patients (56 %) and late in 22 (44 %). Patients with early AKI had higher APACHE II and SOFA scores on admission, higher vancomycin concentrations on day 1, and higher C mean values, as well as lower CrCl values on admission; they also received less vancomycin than those developing late AKI (Table 1). Among the patients developing early AKI, the C mean values were above 25 μg/ml in 18 (64 %), compared to 7 patients (31 %) among those with late AKI (p = 0.02; Fig. 2).

Nine patients (18 %) with AKI required CRRT/HD during vancomycin therapy, whereas only 3 patients (2 %) without AKI during vancomycin therapy eventually needed CRRT/HD (9, 11, and 13 days after vancomycin discontinuation, respectively) (p < 0.01). Among the survivors (n = 153), those who developed AKI (n = 30) had a higher sCr at 1 month after the onset of therapy than those without AKI, despite a similar sCr on admission (Fig. 3). Among the nine patients with AKI who required CRRT/HD, five died and one still needed CRRT/HD at 1 month after therapy.

Fig. 3
figure 3

Serum creatinine (sCr) concentrations at 1 month after the onset of therapy between patients with acute kidney injury (AKI) during vancomycin therapy and those without AKI (*p = 0.04)

Full size image

In the multivariable logistic regression analysis, the duration of vancomycin therapy and C mean were independently associated with the occurrence of AKI (Table 2). The AUC to predict the development of AKI was 0.70 (95 % CI, 0.62–0.77) for C mean and 0.71 (95 % CI, 0.63–0.78) for the combination of C mean and duration of therapy (Fig. 4). A C mean > 30 μg/ml enabled the prediction of AKI during vancomycin therapy with a sensitivity of 22 %, a specificity of 90 %, a PPV of 40 %, and an NPV of 79 %. Particularly, the effects of different thresholds of combined C mean and duration of therapy on the occurrence of AKI are shown in Fig. 5. In addition, high C mean and low Dose mean, as well as bacteremia, were independently associated with the occurrence of early AKI, whereas duration of therapy and diabetes were independently associated with the occurrence of late AKI (Table 2).

Table 2 Multivariable logistic regression analysis for the development of acute kidney injury (AKI) during continuous infusion of vancomycin (n = 207, no missing values)
Full size table
Fig. 4
figure 4

Receiver operating characteristic (ROC) curves constructed to study the ability of mean vancomycin concentrations (C mean, gray dotted line) or C mean and duration of therapy (black line) to predict the development of acute kidney injury (AKI) during vancomycin therapy. The area under the ROC curve is 0.70 [95 % confidence interval (CI), 0.62–0.77] for C mean and 0.71 (95 % CI, 0.63–0.78) for C mean and duration of therapy. A C mean > 30 μg/ml enabled the prediction of AKI during vancomycin therapy with a sensitivity of 22 %, a specificity of 90 %, a positive predictive value (PPV) of 40 %, and an negative predictive value (NPV) of 79 %

Full size image
Fig. 5
figure 5

Occurrence of acute kidney injury (AKI) in relation to different ranges of C mean and duration of therapy

Full size image

Discussion

We have shown that AKI developed in one-fourth of critically ill patients treated with a continuous infusion of vancomycin. More than half the episodes of AKI occurred within 48 h of therapy. AKI was more common in more severely ill patients. As expected, the development of AKI resulted in a greater need for CRRT/HD and in higher sCr at 1 month after the initiation of therapy than in patients without AKI. Finally, C mean and duration of therapy were the strongest variables for predicting AKI during vancomycin therapy, with C mean being particularly associated with early AKI and duration of therapy with late AKI.

AKI is a common complication in critically ill patients, with incidence ranging from 10 to 25 % in the various cohorts, and is uniformly associated with a poor outcome [14]. Although AKI has been recognized as one of the most severe complications associated with vancomycin administration, the mechanisms, including tubular epithelial cells necrosis, oxidative and inflammatory stress, and mitochondrial damage [22, 23], have not been clearly identified. In humans, vancomycin has been reported to induce tubulo-interstitial necrosis or nephritis; moreover, the development of AKI increases if combined with other drugs, such as aminoglycosides [24, 25]. Furthermore, several prospective studies have reported that vancomycin was associated with a higher incidence of AKI than other anti-MRSA drugs, such as linezolid, tigecycline, or daptomycin [12, 26, 27].

The studies evaluating the risk of AKI during vancomycin therapy in different populations, including critically ill patients, have reported incidences ranging from 7 to 40 % (Table 3). However, our findings have some important differences when compared to previous articles. First, we did not find any relationship between daily vancomycin dose and the occurrence of AKI. Lodise et al. [8] and Jeffres et al. [10] reported that daily doses of vancomycin exceeding 4 g or 30 mg/kg were predictive of the development of AKI, especially if associated with elevated body weight and low CrCl. However, one cannot exclude that, in these studies, patients with a greater severity of illness were more likely to receive more aggressive vancomycin regimes, explaining the harmful effects of high initial doses. In our patients, early AKI patients received even lower vancomycin doses than patients without AKI, because of dose adaption to initially lower CrCl, and we observed no cases of renal dysfunction among those patients receiving doses higher than 4 g/day (data not shown).

Table 3 Summary of the most important articles, including adult intensive care unit (ICU) patients, which evaluated the incidence of acute kidney injury (AKI) during vancomycin therapy
Full size table

Second, we found that C mean was independently associated with the development of AKI, but only for early-onset AKI. In a retrospective analysis, Lodise et al. [28] reported that each increase in the initial trough drug concentration was associated with an OR of 1.13 for AKI occurrence. In other studies, vancomycin trough levels greater than 15–20 μg/ml were also identified as a significant predictor of AKI during therapy [8–10, 28, 29]. In contrast, Minejima et al. [30] found that the strongest determinants of AKI included a prior episode of AKI, a history of malignancy, and baseline renal function, whereas vancomycin trough levels were not significantly related to the development of AKI. As vancomycin is almost entirely eliminated by the kidneys, drug clearance varies linearly with CrCl [6]. As such, an alteration in renal function from any cause will increase vancomycin concentrations and it is very difficult to evaluate whether elevated drug levels are responsible for renal damage or only represent a marker of decreased glomerular filtration rate (GFR). We found that patients who developed AKI early were more severely ill and had a worse renal function at baseline; renal deterioration was related more to the underlying sepsis, as suggested by the predictive role of bacteremia on early AKI, rather than to the drug itself. Few studies have evaluated the occurrence of early AKI during vancomycin therapy; Lodise et al. [8] found that one-sixth of patients developing renal dysfunction during vancomycin therapy had early AKI and that these patients also had higher severity scores and risk of renal injury.

Third, the duration of therapy was an independent predictor of AKI, especially for the late onset of renal dysfunction. These data are in line with previous studies, which also reported, for intermittent infusion and non-ICU populations, that the onset of AKI ranged from 4 to 8 days from the start of therapy and the duration of drug exposure could influence the development of AKI [8, 28, 31]. As we also found that diabetes was independently associated with late AKI, it is possible that pre-existing renal dysfunction, such as in chronic diabetic nephropathy, could be a predisposing factor for AKI when prolonged administration of vancomycin is used. Spapen et al. [18] also found that patients with AKI during vancomycin therapy were more likely to be diabetic than patients without AKI (79 vs. 54 %; p = 0.01).

Fourth, the concomitant administration of nephrotoxic agents did not contribute to the development of AKI in the multivariable analysis. This is in contrast with other studies on vancomycin-treated patients, which showed that aminoglycosides [8, 32, 33] or loop diuretics [10, 33] were associated with AKI, especially if drug trough concentrations exceeded 15 μg/ml. It is possible that the influence of concomitant nephrotoxic agents on AKI in our cohort was blunted by the high severity of illness [8, 10, 33].

Fifth, the high incidence of AKI we observed, when compared to previous studies, was probably related to the criteria we used, which considered an increase of at least 0.3 mg/dl of sCr from the initial value as the cut-off to define renal dysfunction [14, 18, 30]. Previous studies often used a ≥50 % change in sCr or CrCl or an increase in sCr of at least 0.5 mg/dl from baseline values to identify AKI (Table 3). If we had used this latter criterion, only 39 of our patients (19 %) would have been defined as developing AKI (data not shown). The use of more strict criteria to define AKI [4] may potentially improve the care of these patients by facilitating early detection of renal dysfunction [30] and, hence, earlier therapy. Definitions need to be considered when comparing the occurrence of AKI among different studies.

Finally, a continuous infusion of vancomycin may result in less AKI than standard vancomycin administration [17], but most of the studies have only considered standard regimes. One recent retrospective study on 129 critically ill patients receiving a continuous infusion of vancomycin reported a 29 % incidence of AKI; in this study, serum vancomycin concentrations above 30 mg/ml were independently associated with the development of AKI [18]. However, this study took into consideration only the highest value of vancomycin concentrations over the period of treatment, which does not take into account the exposure of the kidney to toxic levels of vancomycin over time. In addition, no data on the long-term consequences of AKI were reported in this population. We observed that, among survivors, patients who developed AKI had a higher sCr at 1 month after the onset of vancomycin therapy, despite similar sCr values on admission. Previous studies have reported conflicting results; in three studies, resolution of the nephrotoxicity was seen in 70–80 % of patients and none of these patients required renal replacement therapy as a consequence of AKI [8, 10, 30]. However, in two other studies, CRRT was used in 6 % of patients receiving vancomycin and with abnormal renal function on ICU admission [34, 35]. The implementation of new antimicrobial agents with less nephrotoxicity (i.e., shift from vancomycin to linezolid) in critically ill surgical patients has been associated with a reduced need for CRRT/HD [36]. It remains unclear as to whether the shift to less nephrotoxic agents, such as linezolid [34], in these cases would also allow renal recovery. Interestingly, Teng et al. [37] reported that vancomycin was continued in 10 patients developing AKI, with 7 of the 10 patients showing no further decline in renal function and five of them eventually returning to baseline sCr at discharge.

This study has some limitations. First, continuous infusion of vancomycin regimens have been validated in patients with severe MRSA infections; however, several studies have suggested the need for higher doses in ICU patients [38, 39] and the effects of such regimens on AKI occurrence need to be further studied. Second, we specifically focused on continuous infusion of vancomycin, but it remains unclear as to whether the method of administration of vancomycin (intermittent vs. continuous infusion) is a major determinant for the development of AKI. Studies that have compared these two strategies have yielded conflicting results and current guidelines advise against the administration of vancomycin in continuous infusion, because no clinical benefit over intermittent dosing has been demonstrated and this strategy may not adequately optimize the antibacterial properties of the drug [4]. Third, a control group with the same disease severity would have been necessary in order to assess the real impact of vancomycin on renal function in this setting. In one study, AKI occurred in 17 % of patients receiving vancomycin and in 5 % of control patients [40]. Data from studies comparing vancomycin to other drugs suggest a higher incidence of AKI with vancomycin [12]. However, vancomycin was largely underdosed in the early phase of therapy and these studies analyzed only elderly subjects or did not focus on ICU patients, making it difficult to extrapolate these conclusions to a critically ill population. Fourth, sCr is not a sensitive biomarker for rapidly detecting mild changes in renal function and the use of urinary and/or serum biomarkers of AKI may help to improve the accuracy of AKI diagnosis in the future. Fifth, we did not report other adverse effects associated with vancomycin administration, such as ototoxicity, neutropenia, phlebitis, or toxic epidermal necrolysis, although their incidence is very rare [44, 45]. Fifth, we recorded the use of other potential nephrotoxics, but no information regarding the length of exposure for each of these agents was available. Finally, our study was retrospective and we may have underestimated or misclassified some important determinants of AKI in this setting, and the temporal relationship between drug concentrations elevation and the occurrence of AKI could not be precisely assessed. Also, considering the local expertise in using continuous infusion of vancomycin over a number of years and the monocentric data collection, we could not evaluate the extent to which our results could be generalized to other ICUs.

Conclusions

Using a continuous infusion vancomycin regimen, acute kidney injury (AKI) occurred in one-fourth of critically ill patients and was associated with an increased need for continuous renal replacement therapy (CRRT) or conventional hemodialysis (HD) and altered residual renal function at 1 month. Early AKI was associated with the occurrence of bacteremia and higher mean drug concentrations over the first 3 days of therapy, but was likely the result of sepsis and of a more severe clinical condition than direct drug toxicity. Late AKI was associated with the duration of drug exposure and may be enhanced by the presence of concomitant diseases responsible for chronic renal dysfunction, such as diabetes.