Late-onset sepsis occurs in approximately 20% of preterm neonates and remains a major cause of mortality and morbidity.1 The most commonly reported pathogens are coagulase-negative staphylococci (CoNS).2 The increased incidence of late-onset sepsis due to methicillin-resistant staphylococci in neonatal intensive care units (NICUs) has led to an increased use of vancomycin.3^–5 Vancomycin is a time-dependent antibiotic, and its killing efficiency is highly correlated with the duration of bacterial exposure to the antibiotic rather than being concentration dependent. A continuous infusion is the best way to keep above the minimal inhibitory concentration (MIC).6^–8 Furthermore, it decreases the risk of shock related to rapid infusion and repeated access to the central venous line (CVL).9^–11 However, few data are available about the best method for administering vancomycin to preterm infants and it remains controversial.12^–18

In adults treated with continuous vancomycin infusion, the target concentrations have been reported to have reached faster, with lower costs and less variability between patients than in patients receiving intermittent infusion.19 In infants, bactericidal efficacy has been reported only for intermittent infusion (74–85%).20

In neonates, we had observed very low serum vancomycin concentrations and inconsistent bactericidal efficacy when using complex schedules previously proposed for continuous vancomycin infusion.16 This study aimed to evaluate the impact of a new simplified dosage schedule on vancomycin serum concentrations and bactericidal results.

METHODS

Study infants

Preterm neonates (<34 weeks) suspected of having Gram-positive infection, and born or transferred to our tertiary neonatal unit between June 2002 and December 2005, were enrolled in a prospective, cohort study. Neonates with a history of prior allergy to vancomycin, necrotising enterocolitis, or severe malformative pathology were not included. In all, we evaluated 145 neonates (73 in group 1 and 72 in group 2; see results for definitions of the groups).

Dosage schedule of antibiotics

At time zero (T0), when late-onset sepsis was suspected and prior to antibiotics, we collected blood for culture and assessed creatinine levels. All infants received a constant rate infusion of vancomycin (Vancomycin, Merck, Lyon, France), as it remains stable at 23°C,21 diluted in a 5% dextrose solution to obtain a final concentration of 5 mg/l. It was infused separately from the parenteral nutrition in the central venous line (CVL) or a peripheral catheter. The dose started was 25 mg/kg/day or 15 mg/kg/day (period 1: June 2002 to May 2004) and 20 mg/kg/day or 30 mg/kg/day (period 2: May 2004 to December 2005) depending on whether serum creatinine was below or above 90 μmol/l, respectively. We did not use a loading dose because the serum vancomycin plateau can be obtained very rapidly.22 The concentration-dependent antibiotic amikacin (Amiklin, Merck, Lyon, France) was administered by slow intravenous infusion (20 min) separately from vancomycin, using a previously recommended dosage,23 for the first 48 h of vancomycin treatment. According to the routine protocol of our unit, heparin was infused at 100 units/kg/day,24 25 when local inflammatory signs appeared on the CVL.

Every 48 h after treatment onset, blood was sampled for culture, serum vancomycin and serum creatinine. Antibiotics were withdrawn if the blood culture was negative. If infection was confirmed, vancomycin was continued at a dosage dependent on serum vancomycin, as measured by linked immunoassay. The required therapeutic range for vancomycin was 10–25 mg/l. The lower limit was based on data that vancomycin time-kill kinetics are optimal when serum vancomycin is three or four times the Staphylococcus MIC which is 1–4 mg/l.7 26^–28 The upper limit was set based on previous reports,19 even though no renal side effects have been described under 40 mg/l or ototoxicity under 80 mg/l.29 30 Vancomycin dosage was reduced or augmented by 5 mg/kg/day, depending on whether serum vancomycin was above 25 mg/l or below 10 mg/l, respectively. If serum vancomycin was in the therapeutic range, the same dosage was continued.

If blood culture performed at T2 (96 h after treatment onset) was positive, the CVL was removed and vancomycin dosage adapted when necessary. In all cases of proved infection, vancomycin was continued for 8 days after the first negative blood culture.

Statistical analysis

Continuous variables were expressed as the mean (SD) or as the median and interquartile range if their distribution was skewed. Categorical variables were expressed as frequency and proportion. We determined the risk factors associated with serum vancomycin being measured outside the required range at T1 (48 h after beginning antibiotic treatment) and with the failure of antibiotic treatment (blood culture positive at T1). Differences between the two groups were determined using the Student t test or the Mann–Witney U test for continuous variables, and the Fisher exact test (when the expected frequency was <5) or the Pearson χ² test for categorical variables.

Multivariate analysis of bacteriological efficacy was undertaken, including all covariates with a p value <0.20 in univariate analysis. The final model included only those variables with a p value <0.05. The strength of association between covariates and outcome was measured by odds ratio (OR) with estimated 95% confidence intervals. We calculated the Hosmer–Lemeshow goodness-of-fit statistic. All tests were two tailed and statistical significance was defined as p<0.05. Statistical analysis was performed with SAS software, version 8.2 (SAS Institute Inc, Cary, North Carolina, USA).

RESULTS

Table 1 presents the differences between the two groups at birth. Group 1 received vancomycin 15 mg/kg/day or 25 mg/kg/day depending on whether serum creatinine was > or ⩽90 μmol/l. Group 2 received vancomycin 20 mg/kg/day or 30 mg/kg/day depending on whether serum creatinine was > or ⩽90 μmol/l. At T0, birth weight was below –2 SD31 in 41/145 patients (28%); this proportion was lower in group 1 than in group 2 (12/73, 16% vs 29/72, 40%, p<0.05). At the onset of treatment, most neonates had a CVL (128/145, 88%). A similar percentage of neonates was treated with catecholamines, heparin and diuretics in the two groups.

At T1 (48 h after beginning antibiotic treatment), most serum vancomycin values (108/145, 75%) were within the targeted range, but 15% were below 10 mg/l and 10% were above 25 mg/l (table 2). The increased dosage during period 2 had a notable effect on serum vancomycin concentrations and the maximal value was 43 mg/l. Neonates with a concentration outside the therapeutic range had a significantly lower gestational age (p = 0.03), but similar postnatal age (p = 0.59), body weight at T0 (p = 0.39), patent ductus arteriosus (p = 0.74), and treatment with heparin (p = 0.46), catecholamines (p = 0.06) or diuretics (p = 0.06).

At T1, serum creatinine levels were similar in both groups (64 (50–85) μmol/l vs 63 (49–85) μmol/l, p = 0.34).

Blood culture was positive at T0 in 82/145 (57%) neonates: 80 CoNS, 1 Candida albicans, and 1 Escherichia coli. Almost all of these neonates (75/82, 91%) had a CVL. Among the 80 neonates infected with CoNS, 39 (49%) were in group 1. At T1, 71% (57/80) of neonates with proved CoNS infection had recovered. Bacteriological efficacy was similar with the two dosing levels as 69% (27/39) and 73% (30/41) in groups 1 and 2 (p<0.05), respectively, had recovered at that time. At T2, blood culture remained positive in only 17% (4/23) of patients. Finally, 4 days after the onset of antibiotics, negative blood cultures were observed in most cases (76/82, 93%). Multivariate logistic regression analysis showed that a low postnatal age and absence of heparin were positively associated with bacteriological efficacy (table 3).

DISCUSSION

Using a simplified dosage schedule for vancomycin, we observed relatively good bacteriological and clinical efficacy without excessive levels of serum vancomycin. Continuous-infusion vancomycin therapy was used for the first time by Barois et al32 in 13 children with Staphylococcus-related meningitis. Many studies report dosage schedules for intermittent vancomycin infusion in preterm or term newborns12^–15 33^–36 or infants,36^–38 but only two report a dosage schedule for continuous-infusion vancomycin in premature neonates.16 17 Using a complex schedule based on post-conceptional age, body weight and vancomycin clearance, Pawlotsky et al16 reported that the therapeutic range of serum vancomycin was obtained in 55% and 75% of patients without and with a loading dose, respectively. Using a new simplified dosage schedule, we obtained serum vancomycin levels within the therapeutic range in a higher proportion of patients (75%).

As expected, serum vancomycin levels were higher in the second period, but there were also more neonates with levels above the upper limit of 25 mg/l, with a maximal value of 43 mg/l. Nephrotoxicity is the most frequent side effect of vancomycin treatment, but data in neonates are not available. In our study, there was no significant effect on serum creatinine, probably because renal side effects have been reported for serum levels higher than 40 mg/l.30 It was not possible to evaluate ototoxicity, as larger samples are needed to properly evaluate this side effect. However, ototoxicity does not appear to be clearly related to serum vancomycin concentration; it has been reported in adults for concentrations above 80 mg/l.29

To our knowledge, there are no published data about the bactericidal efficacy of constant infusion in very low birthweight infants. We observed good bacteriological efficacy, as 71% of infected patients had recovered at T1 and a total of 92% after 96 h of treatment without removing the CVL. In our study, maintaining therapeutic serum vancomycin levels was not associated with bacteriological efficacy, in agreement with previously published studies. Although pharmacokinetic information in neonates is available, there is a general lack of high-quality data on the efficacy of dosing regimens.36^–39 Thus it is difficult to determine which serum vancomycin level is required for the successful treatment of infections. Clinical efficacy can usually be obtained if the trough concentration of vancomycin is sufficiently above the MIC.40^–42 Further studies of larger populations are needed to evaluate the relationship between serum vancomycin and bacteriological efficacy.

We found that neonates with a greater postnatal age had a higher risk of continuing infection after 48 h of vancomycin treatment. This could be related to the increasing clearance of vancomycin with postnatal age and maturation.43 Indeed, some dosage schedules propose the use of higher vancomycin dosages when postnatal age is above 7 or 14 days.18 44 45 Our results indicate that introducing postnatal age into the dosage schedule could provide further improvement in neonates.

In our unit, heparin was infused continuously with vancomycin in the same venous line since studies on their interaction have reported stability and efficacy of vancomycin with heparin in dialysis solutions and parenteral nutrition.46^–48 In this study, we found that the absence of heparin was associated with increased bacteriological efficacy. However, this result could be related to heparin being prescribed when a CVL was used for a long period, which is known to increase the risk of sepsis.

Considering that we obtained serum vancomycin concentrations within the required therapeutic range for most neonates and observed good bacteriological efficacy, we propose the following dosage in preterm neonates: 20 mg/kg/day of vancomycin when serum creatinine is >90 μmol/l and 30 mg/kg/day when it is ⩽90 μmol/l.

Further studies of larger groups are needed to confirm the bacteriological efficacy of continuously-infused vancomycin and whether considering postnatal age in the determination of initial dosage improves the efficacy of vancomycin in neonates.