Methods

Therapeutic agents. In the original study, patients were randomized to receive daptomycin monotherapy or standard therapy with either vancomycin (1 g every 12 h with appropriate dose adjustment) or an antistaphylococcal penicillin (nafcillin, oxacillin, or flucloxacillin; 2 g every 4 h). Choice of standard therapy was based on the susceptibility of the causative organism to methicillin. All patients in the standard therapy group and patients with a high likelihood of left-side endocarditis who received daptomycin were to receive initial low-dose gentamicin (1 mg/kg every 8 h, with appropriate dose adjustment) for the first 4 days of therapy. These medications, including initial low-dose gentamicin, are considered study medications in this article. In addition, medications given to patients before enrollment into the study were reviewed to determine which patients received initial low-dose gentamicin <2 days before randomization. The institutional review board at each site approved the protocol, and all patients or their authorized representatives provided written informed consent.

Definitions. Renal impairment clinical adverse events recorded by the investigators included interstitial nephritis, toxic nephropathy, acute prerenal failure, acute or chronic renal failure, renal impairment, or renal tubular necrosis. Creatinine clearance was calculated using the Cockcroft-Gault equation [8]. In addition, the glomerular filtration rate was calculated using the Modification of Diet in Renal Disease equation [9]. Analyses presented use the Cockcroft-Gault equation.

A clinically significant decrease in creatinine clearance was defined as having a decrease in creatinine clearance to <50 mL/min if the baseline creatinine clearance was ⩾50 mL/min or having a decrease in creatinine clearance of ⩾10 mL/min if the baseline creatinine clearance was <50 mL/min [6]. A sustained decrease in creatinine clearance was defined as having ⩾2 sequential measurements of decreased creatinine clearance. This was determined for both 25% and 50% reductions from baseline creatinine clearance.

Collection of laboratory data. Laboratory data, including serum creatinine level, were obtained on days 1, 4, and 7 and then weekly until the end of therapy, as well as at the 6-week posttherapy safety visit. All analyses of renal function included laboratory data collected from baseline through the end of treatment with study agents. Patients had to have a baseline and at least 1 postbaseline creatinine value to be deemed evaluable and to be included in the analyses evaluating laboratory evidence of renal dysfunction.

We conducted 2 separate sets of analyses. First, we assessed the occurrence of adverse events and changes in renal function in the original randomized study arms. Second, to better characterize the impact of initial low-dose gentamicin on renal function, we conducted analyses in which patients were classified as having received initial low-dose gentamicin or not, regardless of whether the gentamicin was received as a study medication or <2 days before study enrollment. These analyses included comparison of the occurrence of and time to a clinically significant decrease in creatinine clearance, the percentage of patients experiencing a sustained 25% or 50% reduction in creatinine clearance, and an assessment of the relationship between renal function and gentamicin dose and duration. We conducted a multivariate analysis to determine the impact of risk factors in the development of clinically significant decreases in creatinine clearance.

Statistical analysis. Categorical data were analyzed using Fisher's exact test, and continuous data were analyzed using analysis of covariance for continuous data. We assessed time to a clinically significant decrease in creatinine clearance using the Kaplan-Meier method. Logistic regression was performed using a forward stepwise procedure. Variables with P<.10 in the univariate analysis were candidates for the multivariable analysis. All predictors were checked for colinearity and confounding. We evaluated effect modification among variables by testing appropriate interaction terms for statistical significance. The final regression models were analyzed for overfitting by the bootstrap method (1000 bootstrap samples of all the data were used). All tests were 2-sided; adjustments were not made for multiple analyses. We conducted all analyses using SAS software, version 9.1 (SAS Institute).

Results

Patient characteristics. Baseline characteristics were comparable in patients who did and did not receive initial low-dose gentamicin (table 1) and across the original randomized antibiotic treatment groups, as detailed elsewhere [6]. In the original randomized groups, baseline creatinine levels were similar among patients receiving daptomycin, vancomycin, or antistaphylococcal penicillins (mean, 1.09, 1.18, and 1.13 mg/dL, respectively; P=.35). Baseline creatinine clearance was also similar among these 3 groups (median, 86, 77, and 87 mL/min, respectively; P=.25). Similar numbers of patients in all 3 groups had diabetes mellitus, and similar numbers received therapy with concomitant nephrotoxic agents.

Receipt of antibiotics. Sixty-three patients received antistaphylococcal penicillins as a study drug: 27 (43%) received only an antistaphylococcal penicillin for the duration of the study, 35 (56%) received initial empirical vancomycin treatment (most for only 1 or 2 days), with a change to an antistaphylococcal penicillin once the isolate was determined to be MSSA, and 1 patient (2%) infected with MRSA inappropriately received an antistaphylococcal penicillin for 9 days before being switched to vancomycin.

Most patients randomized to standard therapy received treatment with initial low-dose gentamicin as a study medication (108 [93%] of 116), including 49 (92%) of 53 patients who received vancomycin and 59 (94%) of 63 patients who received an antistaphylococcal penicillin (table 2). In contrast, only 1 (1%) of the 120 patients assigned to receive daptomycin received initial low-dose gentamicin as study medication. The median length of gentamicin administration was 5 days (range, 1–7 days) in patients treated with vancomycin and 4 days (range, 2–12 days) in patients treated with antistaphylococcal penicillins.

Overall, 43 (18%) of 236 patients received initial low-dose gentamicin before study enrollment (21 in the daptomycin arm and 22 in the standard therapy arm). None of the 21 patients receiving daptomycin received initial low-dose gentamicin as a study medication, whereas all the patients in the standard therapy arm did. Thus, a total of 130 patients received any initial low-dose gentamicin (table 2), 122 of whom were evaluable for analysis. One hundred two evaluable patients were in the standard therapy arm and received a median of 5 days of initial low-dose gentamicin (range, 1–13 days) at a mean daily dose of 2.1 mg/kg. Twenty evaluable patients were in the daptomycin arm and received a median of 2 days of initial low-dose gentamicin (range, 1–12 days) at a mean daily dose of 3.1 mg/kg; 19 of these patients received gentamicin only before study enrollment. Of all patients receiving initial low-dose gentamicin, only 3 patients received a total exposure of >4 mg/kg of gentamicin per day. Of 106 patients who did not receive any gentamicin, 100 patients were evaluable for analysis (93 in the daptomycin arm and 7 in the standard therapy arm).

Evidence of renal dysfunction in the original randomized treatment arms. Renal impairment adverse events were more commonly reported in the standard therapy arm than in the daptomycin arm, regardless of comparator agent (table 3). Eight (7%) of 120 patients who received daptomycin experienced at least 1 renal adverse event, compared with 10 (19%) of 53 patients who received vancomycin and 11 (17%) of 63 patients who received antistaphylococcal penicillins. The difference in rates of renal impairment adverse events was most pronounced in elderly patients and patients with diabetes (table 3).

The mean serum creatinine measurements among patients in the standard therapy group were higher than those in the daptomycin group during treatment with the study drug (figure 1). The significant changes from baseline creatinine level occurred early in therapy for patients receiving antistaphylococcal penicillins (days 4 and 7) and later in therapy for patients receiving vancomycin (days 14, 21, and 28).

Clinically significant decreases in creatinine clearance were more common in the standard therapy arm relative to the daptomycin arm (24% vs. 8%, P=.002) (table 3) and were similarly distributed among patients who received an antistaphylococcal penicillin (16 [25%] of 63) and vancomycin (10 [22%] of 46). Again, patients in the standard therapy arm who were ⩾65 years old and patients with diabetes were more likely to experience a clinically significant decrease in creatinine clearance than were those in the daptomycin arm (table 3).

Impact of gentamicin on renal impairment. Overall, 27 (22%) of 122 patients who received any initial low-dose gentamicin versus 8 (8%) of 100 patients who did not receive any initial low-dose gentamicin experienced a clinically significant decrease in creatinine clearance (P=.005) (table 4). Patients with a baseline creatinine clearance >80 mL/min experienced a clinically significant decrease in creatinine clearance relatively infrequently (5 [7.2%] of 69 who received initial low-dose gentamicin vs. 2 [3.8%] of 53 who did not; P=.70). In contrast, those with a baseline creatinine clearance of 50–80 mL/min were more likely to develop a clinically significant decrease in creatinine clearance when exposed to initial low-dose gentamicin (15 [44.1%] of 34 vs. 4 [14.3%] of 28; P=.014). The percentage of patients who experienced either a sustained 50% or a sustained 25% decrease in creatinine clearance was also significantly higher in the group that received initial low-dose gentamicin (table 4).

Clinically significant decreases in creatinine clearance occurred earlier (P=.01, by Wilcoxon test) and were sustained through the end of therapy (P=.009, by log rank test) in patients who received initial low-dose gentamicin than in patients who received no gentamicin (figure 2). None of the 7 patients who received standard therapy alone experienced a decrease in renal function (figure 2).

We conducted a multivariate analysis to determine the impact of risk factors in the development of a clinically significant decrease in creatinine clearance in this cohort of patients who received any initial low-dose gentamicin. Age ⩾65 years and receipt of any initial low-dose gentamicin were independently associated with a clinically significant decrease in creatinine clearance (table 5). Although a slight statistical association was found between duration of gentamicin therapy and renal impairment (table 5), the clinical association between dose and duration of gentamicin and nephrotoxicity appeared minimal in this cohort (figure 3).

All analyses presented herein were repeated using Modification of Diet in Renal Disease equations to calculate the glomerular filtration rate. No important differences in the results were noted (data not shown).

Discussion

Our findings suggest that receipt of even a few days of initial low-dose gentamicin in combination with vancomycin or an antistaphylococcal penicillin in the treatment of S. aureus bacteremia and native valve endocarditis is frequently associated with renal dysfunction. These findings were consistent when different measures of renal dysfunction, such as increase in mean creatinine level, decrease and sustained decrease in creatinine clearance, and time to decrease in creatinine clearance, were evaluated. Although data are not available on the frequency with which physicians in practice add initial low-dose gentamicin to standard therapy in the treatment of S. aureus bacteremia and endocarditis, we suspect that it occurs commonly, given that 43 (18.2%) of 236 patients in this study had been prescribed initial low-dose gentamicin before study enrollment.

Several reports have demonstrated that, when vancomycin and standard doses of gentamicin are administered concomitantly, the incidence of nephrotoxicity is greater than when either drug is administered alone [10, 11]. No studies, however, have evaluated nephrotoxicity associated with the combination of initial low-dose gentamicin and vancomycin for the treatment of S. aureus bacteremia or endocarditis.

Furthermore, although nephrotoxicity associated with gentamicin has been observed in patients receiving 2-week courses of low-dose gentamicin in conjunction with antistaphylococcal penicillins, compared with patients receiving antistaphylococcal penicillins alone for the treatment of endocarditis [4, 12], it has not been reported in patients receiving short courses of initial low-dose gentamicin. The amounts and durations of gentamicin received by patients in this study were low, with a median exposure of only 4 days at a daily dose of 2–3 mg/kg, a schedule not expected to produce significant serum peaks or troughs. The decline in renal function seen was clinically relevant, with discontinuation of study medication use in 4 (3%) of 122 patients, sustained 25% decreases in creatinine clearance in 26 (21%) of 122 patients, and sustained 50% decreases in creatinine clearance in 7 (6%) of 122 patients. Although no patients died or required long-term hemodialysis as a result of renal dysfunction in the cohort, elevations in serum creatinine levels and decreased creatinine clearance can lead to additional laboratory and imaging studies, nephrology consultation, changes in medication and doses, prolonged hospitalization, and increased mortality [13–15]. In addition, renal dysfunction occurred frequently in this cohort, despite the fact that a large proportion of patients had uncomplicated bacteremia and right-side endocarditis—infections that are generally considered less severe. We did not find an important relationship between dose or duration of gentamicin and the incidence of renal impairment in this cohort. This may be because relatively few patients received higher doses or longer therapy or because of the tendency to stop gentamicin therapy if evidence of renal impairment developed.

The results of this investigation also suggest that other situations in which patients receive low-dose and/or short-course gentamicin, such as surgical prophylaxis in the penicillin-allergic patient, should undergo scrutiny for safety. Indeed, nephrotoxicity has been reported to occur in 17 (20%) of 87 patients who received a single 2-g dose of dicloxacillin and 240 mg of gentamicin as surgical prophylaxis for the repair of intertrochanteric hip fractures, compared with only 4 (5%) of 76 patients who received no prophylactic antibiotics [16].

Surprisingly, no differences were apparent in the incidence of renal dysfunction between patients receiving an antistaphylococcal penicillin and those receiving vancomycin. If either of these agents is given as monotherapy, the incidence of nephrotoxicity is generally low, suggesting that the addition of gentamicin plays a role in the nephrotoxicity that we observed [17–20]. Nephrotoxicity was also observed earlier in patients receiving antistaphylococcal penicillins, with a mean serum creatinine level peak on study day 7, compared with patients receiving vancomycin, in whom the mean serum creatinine level began to increase on day 7 and continued to increase throughout the study. This finding may relate to the earlier and more rapid lysis of S. aureus seen with β-lactam antibiotics relative to vancomycin, leading to earlier antigen-immune complex deposition. Alternatively, the finding may suggest that the synergistic nephrotoxicity seen with vancomycin and gentamicin is more sustained.

The impact of initial low-dose gentamicin on renal function appears to be greater in individuals with underlying but underappreciated renal dysfunction, such as persons aged ⩾65 years and those with diabetes. Unfortunately, S. aureus bacteremia and endocarditis now occur frequently in persons with these risk factors [5, 21, 22]. Other studies have also demonstrated an association between gentamicin and renal dysfunction in these populations, but generally not with the initial low-dose gentamicin used in this study [23–25].

This study has limitations. The data are from a randomized, controlled trial that was not designed to assess the effect of gentamicin on renal dysfunction. The secondary analyses presented in this study, however, provide a consistent picture suggestive of an association between gentamicin and renal dysfunction. Accounting for the impact of potentially nephrotoxic agents on individual patient outcomes is challenging because individual patients may respond to these agents differently. For example, patients may have concurrent conditions, such as volume depletion, that predispose them to nephrotoxicity, particularly in the context of concomitant administration of a nephrotoxic drug. However, these differences were to some degree controlled for through the randomization process and were assessed in part via the risk factor analysis. The open-label design of the original trial could have affected investigator reports of renal adverse events, but observations based on objective laboratory data are unlikely to be affected by this potential bias.

We were unable to assess the impact of the addition of initial low-dose gentamicin on the efficacy of therapy with antistaphylococcal penicillins or vancomycin, given that only 7 patients in the standard therapy group did not receive gentamicin. However, early studies examining initial low-dose gentamicin combination therapy in patients with MSSA bacteremia and endocarditis showed either no benefit [3, 26] or minimal benefit (reduction in the duration of bacteremia by 1 day) [4]. More recent studies also show no significant benefit to adding low-dose aminoglycosides for injection drug users with MSSA right-side endocarditis who are undergoing 2-week short-course therapy [12] or in sterilizing valves in patients with native valve S. aureus endocarditis undergoing surgery [27].

In conclusion, initial low-dose gentamicin appears to result in significant renal dysfunction when used with either vancomycin or antistaphylococcal penicillins for the treatment of S. aureus bacteremia and native valve endocarditis. Recent American Heart Association guidelines for the management of endocarditis state that the addition of gentamicin in the treatment of native valve MSSA endocarditis should be considered optional. On the basis of evidence of potential harm and the lack of evidence for clinically significant benefit, we recommend against the routine use of initial low-dose gentamicin in the management of most cases of S. aureus bacteremia and native valve endocarditis, particularly in elderly and diabetic patients and in patients with even mild baseline renal dysfunction.