Extended-Interval Aminoglycoside Administration for Children: A Meta-analysis

Search Strategy

We searched Embase and PubMed (from January 1966 to September 2003) using the following key words: (aminoglycoside, amikacin, gentamicin, tobramycin, netilmicin, isepamicin, kanamycin, or sisomycin) and (extended interval aminoglycoside administration, EIAA, extended interval dosing, extended interval, single daily, single dose, once a day, once daily, once-daily, once, or daily) and (children, child, childhood, pediatric, newborn, neonate, infant, or infantile). We also searched the Cochrane Controlled Trials Registry (last search, September 2003). Reference lists of the eligible articles and pertinent reviews were also scrutinized for potential relevant, randomized, controlled trials.

Selection Criteria

We included only randomized, controlled trials in which ODD was compared with MDD administration with similar total daily doses of the aminoglycoside. We allowed up to 25% dissimilarity in average daily doses between the compared arms. We considered only trials involving pediatric patients (upper age limit: 20 years) and trials involving both adults and children that provided separate data for the pediatric population. We included only studies with parenteral (intravenous or intramuscular) administration of aminoglycosides.

Data Extracted and Outcomes

From each study we extracted the following data: author, year of publication, patient age range, clinical setting, aminoglycoside type, total daily dose, dosing intervals in the MDD arm, concurrent use of other antibiotics, definitions of clinical and bacteriologic failure, nephrotoxicity and ototoxicity, number of randomized patients per arm, time point for outcome evaluation, and events per arm for efficacy and toxicity outcomes. We also noted whether information on the mode of randomization, allocation concealment, and blinding was presented.

We considered the following outcomes: 1) clinical or microbiologic failure, as defined in each study (when both were available, clinical failure data were preferred over microbiologic failure data); 2) clinical failure; 3) microbiologic failure; 4) primary nephrotoxicity outcome, ie, any increase in serum creatinine levels or decrease in creatinine clearance, with thresholds as defined in each study; 5) secondary nephrotoxicity outcomes, ie, urinary excretion of proteins (retinol-binding protein, β₂-microglobulin, Clara cell protein [P1], microalbumin, N-acetyl-β-d-glucosaminidase, alkaline phosphatase, alanine aminopeptidase, or γ-glutamyltransferase) or phospholipids; and 6) ototoxicity, based on pure-tone audiometry, brainstem auditory evoked responses, or otoacoustic emissions for neonates and infants, vestibular testing, clinical impressions, or any method. All of the efficacy and toxicity outcomes were evaluated at the end of therapy. When no data were available for the end of therapy, we used data for the time point closest to the end of therapy. All data were extracted independently in duplicate. Discrepancies were resolved by review with a third investigator, and consensus was reached for all items.

Statistical Analyses

We report pooled estimates of efficacy and toxicity outcomes. Study-specific risk ratios were used as the measure of choice. Between-study heterogeneity in risk ratios was appraised with the Q statistic.³² Between-study heterogeneity means that the differences in the results of various studies are not attributable to chance alone. The Q test is based on the χ² distribution, and it provides a measure of the sum of the squared differences between the results observed and the results expected in each study, under the assumption that each study estimates the same treatment effect. In the absence of between-study heterogeneity, risk ratios were synthesized across studies with fixed-effects models.³³ Fixed-effects models were also more appropriate in this meta-analysis, because for most studies and outcomes the number of events was either 0 or very small. For studies with 0 events, risk ratios were entered in the calculations by adding 0.5 to the cells of the 2 × 2 table. Fixed-effects analyses are robust with respect to 0 counts and small numbers. Subgroup analyses were performed according to the type of clinical setting and the type of aminoglycoside. Effect sizes in subgroups were compared by using general variance models. Finally, we examined whether larger trials differed from smaller trials in their results, using the test described by Begg and Mazumdar,³⁴ which examines the τ rank correlation between the natural logarithm of the risk ratio and the weight of a study. Analyses were performed with SPSS 11.0 (SPSS Inc, Chicago, IL), StatXact 3.0 (Cytel Corp, Boston, MA), and Meta-analyst (Joseph Lau, Boston, MA) software. All P values were 2-tailed.

Identification of Eligible Studies and Study Characteristics

Twenty-four eligible studies^{21–31,35–47} published between 1991 and 2003 were identified (Table 1). Aminoglycosides were used in different clinical settings (neonatal intensive care unit: 6 studies; cystic fibrosis: 3 studies; cancer: 5 studies; urinary tract infections: 4 studies; diverse infectious indications: 5 studies; pediatric intensive care unit: 1 study).

Aminoglycosides used in the eligible trials included amikacin (9 studies), gentamicin (11 studies), tobramycin (2 studies), netilmicin (2 studies), and tobramycin or netilmicin (1 study) (Table 1). The dosing varied among studies (amikacin 15–20 mg/kg per day; gentamicin 4–7.5 mg/kg per day; tobramycin 8–15 mg/kg per day; netilmicin 5–10 mg/kg per day). Eighteen trials used exactly the same aminoglycoside dose in the ODD arm and in the MDD arm. However, in 5 cases^{21,23,27,28,47} slightly higher aminoglycoside dose was used in the MDD arm; and in 1 case³⁶ a slightly lower dose was used in the MDD arm. Most of the children included in the meta-analysis received short-term aminoglycosides (up to 10 days' duration), although some children in at least 11 trials^{22,24,27,29,36–38,42,44–46} were treated for over 10 days.

In 20 trials additional antibiotics were concurrently administered to patients in both arms. In 4 of them concurrently administered antibiotics were systematically different in the 2 arms. These trials were excluded from all efficacy analyses.

Almost all trials were performed in single countries (United States: 3 studies; Mexico: 1 study; Europe: 5 studies; Australia: 2 studies; Israel: 2 studies; Africa: 5 studies; Asia: 4 studies). There were only 2 international trials.^36,38 The largest study considered 412 patient-episodes,³⁸ and only 7 trials^{22,23,36–38,45,47} included >100 children (or patient-episodes).

The exact definitions of clinical failure, bacteriologic failure, nephrotoxicity, and ototoxicity in each study are detailed in appendices available upon request, along with the number of events per arm. Of the 24 eligible trials, 6 studies^{21,29,31,37,42,45} reported in detail the mode of randomization and 3 studies^22,36,41 reported enough details of designs that ensured allocation concealment. Two studies^25,28 were double-blind, 6 studies^{22,30,36–39} were stated to be open-label, and 16 studies did not report on blinding status.

Efficacy

Efficacy analysis of all trials^{23,24,26,30,31,37,39,40,43–47} indicating either clinical or microbiologic failures demonstrated pooled failure rates of 4.6% (23 of 501 cases) in the ODD arms and 6.9% (34 of 494 cases) in the MDD arms (Table 2). The fixed-effects risk ratio was 0.71 (95% confidence interval [CI]: 0.45–1.11; P = .13) (Fig 1), and there was no significant between-study heterogeneity. There was a suggestion of a significant benefit with ODD versus MDD in trials using amikacin, whereas this was not observed for trials using other antibiotics (Table 2).

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Meta-analysis of combined clinical or microbiologic failures with ODD versus MDD. Each study is indicated by author name, year, type of aminoglycoside (A: amikacin; G: gentamicin; T: tobramycin; N: netilmicin), point estimate, and 95% CI for the risk ratio. Also shown is the pooled fixed-effects risk ratio. The continuous vertical line passes through the risk ratio of 1 (no effect), and the discontinuous vertical line passes through the summary estimate. The upper 95% CI values are not shown for many studies, because they are very high, but they can be inferred from the lower 95% CI values (the upper and lower CI values are symmetric around the point estimate).

The pooled clinical failure rates were 6.7% (22 of 330 cases) in the ODD arms and 10.4% (34 of 327 cases) in the MDD arms (Table 2). The fixed-effects risk ratio was 0.67 (95% CI: 0.42–1.07; P = .09). There was no between-study heterogeneity. The majority of clinical failures occurred in 2 trials,^31,45 ie, 1 trial involving critically ill children in the pediatric intensive care unit of a South African university hospital⁴⁵ and 1 trial involving children 1 month to 8 years of age with suspected or proven severe bacterial infections in a university hospital in Zimbabwe.³¹ In the other 8 trials^{24,26,37,39,40,43,44,46} with pertinent data, clinical failure rates ranged from 0 to 11% in the ODD arms and from 0 to 10% in the MDD arms and there was no major difference in the compared regimens.

The pooled microbiologic failure rates were 1.8% (5 of 283 cases) with ODD and 4.0% (11 of 275 cases) with MDD (Table 2). The fixed-effects risk ratio was 0.51 (95% CI: 0.22–1.18; P = .1). There was no between-study heterogeneity.

Nephrotoxicity

The pooled primary nephrotoxicity outcome rates were 1.6% (15 of 955 cases) in the ODD arms and 1.6% (15 of 923 cases) in the MDD arms (Table 3). The fixed-effects risk ratio was 0.97 (95% CI: 0.55–1.69; P = .90) (Fig 2). The pooled secondary nephrotoxicity outcome rates were 4.4% (3 of 69 cases) in the ODD arms and 15.9% (11 of 69 cases) in the MDD arms (Table 3), suggesting a statistically significant superiority of ODD. The fixed-effects risk ratio was 0.33 (95% CI: 0.12–0.89; P = .03). Results were consistent across types of clinical settings and aminoglycosides.

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

Meta-analysis of primary nephrotoxicity with ODD versus MDD. Each study is indicated by author name, year, type of aminoglycoside (A: amikacin; G: gentamicin; T: tobramycin; N: netilmicin), point estimate, and 95% CI for the risk ratio. Also shown is the pooled fixed-effects risk ratio. The continuous vertical line passes through the risk ratio of 1 (no effect), and the discontinuous vertical line passes through the summary estimate. The upper 95% CI values are not shown for many studies, because they are very high, but they can be inferred from the lower 95% CI values (the upper and lower CI values are symmetric around the point estimate).

Ototoxicity

The pooled ototoxicity rates for studies that provided auditory testing results^{21,23,25,35–40,42,44,46,47} were 2.3% (10 of 436 cases) in the ODD arms and 2.0% (8 of 406 cases) in the MDD arms. The fixed-effects risk ratio was 1.06 (95% CI: 0.51–2.19; P = .92) (Table 3). In studies that provided clinical vestibular function testing results,^22,36 no toxicity was documented among 209 patients given ODD and 206 patients given MDD (Table 3). Studies noting only the clinical impression of hearing impairment^26,27,45 also failed to identify any toxicity (ODD: n = 114; MDD: n = 114) (Table 3).

Subgroup and Bias Analyses

We detected no statistically significant differences between ODD and MDD, in any of the examined subgroups (neonatal intensive care unit, cystic fibrosis, cancer, or urinary tract infection), with respect to combined clinical or microbiologic failure outcomes, primary nephrotoxicity outcomes, or ototoxicity (based on auditory testing) (Table 3), when sufficient data were available. There was no significant relationship between the effect size (risk ratio) and the trial size for any of the outcomes (data not shown).