Acute kidney injury during daptomycin versus vancomycin treatment in cardiovascular critically ill patients: a propensity score matched analysis

This is a retrospective observational study with a propensity score adjustment to reduce the bias of selection for a comparative analysis between two antibacterial treatments used in routine care. The ethics committees (“CPP Sud Méditerranée IV”, N°Q-2015-05-03 and the “CCTIRS”, N°15.670) approved the study. The requirement for informed patient consent was waived since the nature of the study was retrospective. The data collection was authorized by the “CNIL” (N°DR-2015 − 643).

From January 2010 to December 2012, patients who were admitted to a cardiothoracic and vascular surgical ICU and who received either daptomycin (DAP group) or vancomycin (VAN group), identified from the delivery drug list of the hospital pharmacy, were screened for analysis.

The following criteria were required for inclusion: (i) Patient older than 18 years; (ii) Suspected or proven cardiac, vascular or profound surgical site infection with Gram-positive cocci (GPC) methicillin-resistant (MR) strains (including probabilistic treatment for patients with acquisition of MR risk factors); (iii) Treatment duration greater than or equal to 48 h (at least 2 doses of daptomycin administered or 2 days of vancomycin infusion); (iv) Antibiotic treatment started in peri-operative (from 48 h before the onset of surgery) or in postoperative period (during ICU stay). Patients with prophylaxis indication of antibiotics, kidney disease on chronic dialysis or acute onset of RRT before initiation of DAP or VAN treatment, or staphylococcus pneumonia were excluded.

The indication for a probabilistic or documented treatment with broad spectrum antibiotics against GPC followed institutional protocols and national recommendations for the specific population of patients admitted to ICU with sepsis or septic shock in the context of cardiovascular surgery or cardiogenic shock. Infectious endocarditis means infection on native or prosthetic valve or pacemaker leads. Surgical site infection is an infection located on the cutdown tissue or profound structures, i.e. sternal infection or mediastinitis. Ventricular assist device (VAD) provides a long-term mechanical circulatory support and only profound device-related infections occurring during the postoperative period of the implantation were considered in this study (excluding driveline or delayed infections). Vascular graft infection requires usually a probabilistic treatment after surgical samples during graft replacement, which is a complex and haemorrhagic surgery. Catheter related infections were identified by bacteraemia and positive culture of the removed catheter with the same strain and were aggressively treated in a population with cardiac or vascular prosthetic materials. The other conditions to treat were an anterior mediastinitis or a septic shock in waiting of documentation. Cardiovascular foreign body infection was defined by an infection related to intra-cardiac devices or prothesis or vascular graft with artificial tissue with development of bacterial biofilm.

Daptomycin was administered at a dose of 8 mg/kg in thirty-minutes intravenous infusion every 24 h in patients without severe impairment of kidney function or every 48 h in case of GFR below 30 ml/min/m². The creatine-kinase (CK) level was measured before the initiation of DAP and at least once a week to assess the occurrence of muscular toxicity defined by an increase of CK up to 3-fold the upper superior limit without any evidence of member ischaemia.

Vancomycin intravenous treatment was initiated by a loading dose of 30 mg/kg in 1 h and followed by a continuous maintenance infusion dosing between 15 and 30 mg/kg/d. The VAN dose was adapted to achieve a target serum vancomycin steady-state concentration of 20–30 mg/L assessed by a daily pharmacologic monitoring (therapeutic drug monitoring).

Demographic data, and main clinical characteristics were collected. Renal function was assessed by daily measurement of creatinine and urinary output monitoring in ICU. The need for RRT was left to the decision of the physician in charge of the patient, mainly because of threatening metabolic disorders (metabolic acidosis, hyperkalaemia) or fluid overload. The AKI stage was established during the first 7 days and 14 days after the initiation of the studied treatment.

The factors associated to renal function impairment before the antibiotic treatment were collected for the calculation of the propensity score: severity score as the Simplified Acute Physiology Score II (SAPS II) at ICU admission and the Sequential Organ Failure Assessment (SOFA) score at initiation of treatment, history of chronic renal failure (creatinine clearance below 50 ml/min), creatinine and AKI before treatment initiation (defined as baseline), circulatory shock related to sepsis or heart failure (defined as systolic arterial pressure below 90 mmHg and refractory to fluid challenge with evidence of end-organ hypoperfusion like oliguria, alteration of mental status or high plasma lactate), need of cardiopulmonary bypass for surgery (within 48 h around the antibiotic initiation), use of other nephrotoxic agents (iodine contrast, aminoglycoside, ciclosporin), documentation of GPC bacteraemia.

The primary endpoint was the incidence of AKI within the first week of treatment as defined by KDIGO consensus [14].

The secondary endpoints were chosen to better characterize AKI and describe the outcome. The incidence of AKI within the first 14 days of treatment and the maximal decrease of GFR (estimated by CKD-EPI formula) during treatment were also reported. Besides KDIGO classification, severe renal failure was estimated as stage 2 or 3 of the AKI or a reduction in glomerular filtration rate (GFR) of more than 50%, and from incidence and duration of RRT. Clinical treatment failure, defined by either persistent positive cultures, worsening of clinical status, death due to initial infection, or relapse after the end of treatment was assessed in case of documented GPC infection. Incidence of premature discontinuation of treatment was defined when treatment was stopped because of adverse event or clinical failure except death. Mortality rate was calculated at day 28 and month 6.

In a previous study [17], the incidence of renal dysfunction was estimated at 12% with daptomycin and 35% with standard therapy. However, a higher incidence of renal failure was expected in ICU population, as much as twice the incidence observed in the reference study [10, 19]. Therefore, we assumed that AKI may occur in 60% VAN patients, meaning that a reduction to 20% in DAP group would be significant for a risk at 5% and a power equal to 85%, provided 52 patients (at least 26 in both groups) were included.

Since treatments were not randomized, we used the propensity score method for primary endpoint analysis. For this, we included the covariates potentially related to treatment and outcome (chronic kidney disease, cardiopulmonary bypass, GPC bacteraemia, heart failure, and shock) in a multivariate logistic model explaining the choice of treatment. The propensity score was built from the selected significant covariates (chronic kidney disease, cardiopulmonary bypass, GPC bacteraemia) and was used in the second model as an adjustment covariate. After matching on the propensity score, we checked that the covariates were statistically balanced between the two treatment groups. Then, we modelled the effect of the treatment on the main endpoint (AKI within first week of treatment) by selecting the covariates associated to this result, with a threshold of significance to 15% in a univariate analysis, and by including in the multivariate logistic model the propensity score as an additional adjustment variable.

The calculations were done under R software [20] by the team EA2415 of the University of Montpellier.

During the 36-month period of inclusion, among the 143 patients on the pharmacy delivery list for DAP or VAN treatment, 72 patients (28 DAP and 44 VAN) were selected for the study following application of inclusion and exclusion criteria (Fig. 1). After matching to the propensity score, 16 patients of each group were paired. Baseline demographic and clinical characteristics before and after matching are summarized in Table 1. After propensity score matching, characteristics were balanced between the two groups except a significantly higher proportion of patients with heart failure in the DAP group (75% versus 25%, p = 0.012).

The reasons for ICU admission (detailed on Table 2) were different between DAP and VAN groups (p = 0.004) with more vascular surgery for VAN patients and more cardiologic indications for DAP patients but the rate of cardiac surgery was similar. The primary infectious diseases (reported on Table 2) were distributed slightly differently between groups with more vascular graft infection in the VAN group and more VAD-related infection in the DAP group (p = 0.045). Cardiovascular foreign body infection concerned 51% of patients without difference between DAP (n = 16; 57%) and VAN (n = 21; 48%), p = 0.436. A GPC infection was documented in 52 patients (72%), 26 in each group, including different strains of staphylococci isolated in 55% of infections (Table 2).

The incidence of AKI at day 7 (primary endpoint) was significantly higher with vancomycin (64%, n = 28) than with daptomycin (21.4%, n = 6), p = 0.001 respectively. Similar results were observed at day 14 (Fig. 2).

Factors associated with 7-day AKI in univariate analysis are reported in Table 3. The multivariate logistic regression model, adjusted to the propensity score (which includes chronic kidney disease, cardiopulmonary bypass and GPC bacteraemia covariates), showed that only vancomycin treatment remained significantly associated with the occurrence of AKI with an Odds Ratio of 4.42 (95%CI: 1.39 to 15.34), p = 0.014 (Table 3).

AKI was more severe in VAN group than in DAP patients as assessed by incidence of AKI stage 2 or 3 or GFR decrease of more than 50% and RRT requirement and duration (Table 3).

Clinical failure of initial antibiotic therapy occurred more frequently with vancomycin than with daptomycin, namely in 11 out of 26 VAN patients (42%) and 4 out of 26 DAP patients (15%), with GPC documented infection, p = 0.032). Premature treatment discontinuation was not significantly different between DAP and VAN groups (2 (7%) versus 11 (25%), respectively, p = 0.055). Several changes were related to vancomycin (11 cases / 44 vs 2 / 28, p = 0.055) including 6 cases of switch from vancomycin to daptomycin. Two failure cases of daptomycin treatment resulted in a switch to vancomycin (occurrence of secondary staphylococci pneumonia, acquisition of resistance).

The ICU and hospital lengths of stay were similar in both groups (Table 4). Mortality was high in this cohort, with no significant difference at month 6 between the VAN and DAP groups (Table 4). The survival curves were not significantly different on a Kaplan Meier analysis (Fig. 3).

Due to the retrospective design, selection bias and missing data may have influenced the results. The minimal inhibitory concentrations of vancomycin for MRSA strains were not available for most of the patients. Furthermore, no follow up after hospital discharge was organized. The inclusion period was chosen during the first 3 years of daptomycin use in our unit, since we observed an increase in its indication instead of vancomycin in our routine practice after this period. Although the sample of this study is limited and there is always a risk of hidden bias, the absence of a previous or expected randomized trial in this area allows us to retain this result. Despite these limitations, both groups were balanced with respect to baseline characteristics (especially for severity scores SAPS II and SOFA). The striking difference in AKI incidence between both groups, added to the adjustment with the propensity score, suggest these limitations cannot overrule the clinical relevance of the results.