Pediatr Nephrol
DOI 10.1007/s00467-016-3471-9
ORIGINAL ARTICLE
Previous aminoglycoside use and acute kidney injury risk
in non-critically ill children
Jeremy Andrew Saban 1 & Michael Pizzi 1 & Jillian Caldwell 1 & Ana Palijan 1 &
Michael Zappitelli 1
Received: 21 December 2015 / Accepted: 18 July 2016
# IPNA 2016
Abstract
Objectives Aminoglycosides (AG) are a group of bactericidal
antibiotics with nephrotoxic effects that are commonly used in
the treatment of hospitialized children. We have examined
previous AG treatment as a risk factor for acute kidney injury
(AKI) during current AG treatment.
Study design We performed a retrospective cohort study of
children ranging in age from 1 month to 18 years who were
treated with AG between October 2008 and April 2012 at
Montreals Childrens Hospital. Children for whom no serum
creatinine data (SCr) were available and those with baseline
renal disease were excluded from the analysis. Main expo-
sures were prior AG use (number and hours of prior treat-
ments) and time since last AG treatment. The main outcome
was AKI, defined on the basis of the Kidney Disease:
Improving Global Outcomes guidelines. Logistic regression
was used to examine exposureoutcome associations.
Results AG treatments episodes with Stage 1, 2, and 3 AKI,
respectively, were associated with a median of 98 [interquar-
tile range (IQR) 339], 231 (IQR 688), and 111 (IQR 505) h of
prior AG treatment, respectively, versus non-AKI (median 0,
IQR 54 h) (p < 0.0001). AKI episodes were associated with a
mean ( standard deviation) of 1.5 1.8 AG treatments in the
previous 6 months, versus 0.9 1.6 AG treatments for non-
AKI. The number of AG-treatment days during the preceding
6 months [adjusted odds ratio (adjOR) 1.04, 95 % confidence
interval (CI) 1.031.06; p < 0.001], younger age (adjOR 0.96,
95 % CI 0.930.99; p = 0.009), admission to hematology
* Michael Zappitelli
mzaprdr@yahoo.ca
1
Division of Nephrology, Department of Pediatrics, Montreal
Childrens Hospital, McGill University Health Centre, 2300 Tupper,
Room E-213, Montreal, Quebec, Canada
oncology department (adjOR 3.88, 95 % CI 2.176.96;
p < 0.001), and tobramycin use (adjOR 1.77, 95 % CI 1.04
3.02; p = 0.04) were independently associated with AKI.
Episodes with Stage 1 and 2 AKI were associated with fewer
days since last treatment compared to non-AKI treatment
(p < 0.02 and p < 0.005, respectively; Mann-Whitney test).
Conclusions Based on these results, prior AG treatment is a
risk factor for AKI and should be considered when dosing and
monitoring hospitalized children being treated with AG.
Keywords Nephrotoxic . Risk factors . Acute renal failure .
Antibiotics . Hospitalization
Introduction
Acute kidney injury (AKI) is a common occurrence in hospi-
talized children and is associated with increased mortality and
morbidity [15]. Nephrotoxic medication use is one of the
most common causes of pediatric AKI, accounting for up to
16 % of hospital-acquired AKI [68]. A recent study showed
that AKI induced by nephrotoxic agents was associated with
residual kidney damage 6 months later [9]. Aminoglycosides
(AG) are very effective against Gram-negative bacteria but are
one of the most common nephrotoxins used to treat children
and are well known to cause AKI [1012]. They are freely
filtered at the glomerulus and taken up by proximal tubular
cells through megalin receptors [13, 14]. AG accumulation in
lysosomes causes cell damage and may ultimately lead to cell
death [15]. One study has documented elevated levels of in-
jury proteins in the proximal tubule during AG therapy [16],
and AG-induced AKI has been shown in children to be asso-
ciated with increased length of hospital stay and healthcare
utilization [11, 17].

In a recent meta-analysis, use of AG, amongst other factors,
was associated with long-term renal abnormalities in children
treated for cancer [18]. It has been postulated that repeated AG
treatment contributes to chronic nephropathy [16, 19, 20],
however this issue remains unresolved. The extent to which
prior AG use is an AKI risk factor during a current AG treat-
ment is unknown. Understanding this risk could allow for a
simple, predictive factor used to identify hospitalized children
who should be monitored more closely or in whom AG use
should be avoided.
We hypothesized that prior AG increases the risk for AKI
during a current AG treatment episode. We therefore evaluat-
ed, in non-critically ill hospitalized children treated with AG,
the association between AG use in the 6 months prior to a
current AG treatment and an AKI during that current AG
treatment episode.
Patients and methods
Design, setting, and patient selection
This was a retrospective cohort study of children admitted to
non-critical care wards of the Montreal Childrens Hospital
(Canada) who were treated with AG between October 2008
and April 2012. Throughout this period, the Pharmacy depart-
ment provided us with the names of patients on non-critical
care wards who were started on AG treatment, for the purpose
of identifying and recruiting patients into another, observa-
tional study to validate new AKI biomarkers. All data on
identified patients were recorded and used to perform this
initial retrospective evaluation. Inclusion criterion for the anal-
ysis reported here was any confirmed AG treatment.
Exclusion criteria were age of <1 month or >18 years at the
start of AG treatment, presence of prior renal disease (deter-
mined during the initial studys screening process by commu-
nication with the medical team and chart review), admission to
hospital for a primary renal condition (e.g., urinary infection),
and any admission to a critical care unit during the current
admission (higher AKI risk due to several non-AG factors).
Patients were included more than once in the study if they had
a separate AG treatment episode. At analysis, treatments for
which no serum creatinine (SCr) were available (required for
AKI determination) were excluded.
Data collection
The original screening database contained the following var-
iables, as acquired from the Pharmacy department, medical
care team or via chart review: date of birth/age at treatment
start, gender, reason for admission, hospital ward and service
of admission (surgical, medical, hematologyoncology), pres-
ence of previous renal problems (e.g., tubulopathy, nephritis,
Pediatr Nephrol
chronic kidney disease), AG treatment startstop date, and
time and dosage details.
For this study, we also retrospectively collected SCr values
required for AKI status determination, using the electronic
hospital laboratory system. Baseline SCr (bSCr) was taken to
be the lowest SCr measurement in the 3 months before the
current AG treatment [21, 22]. When no bSCr measurement
was available, we estimated it using the median normative
values for age and gender, specific to our center [21, 23].
Our hospital electronic laboratory database did not include
height measurements, therefore we could not estimate bSCr
using estimated glomerular filtration rate (eGFR) equations, as
recommended in the Kidney Disease: Improving Global
Outcomes (KDIGO) guidelines [22]. Peak SCr during AG
treatment was taken to be the highest SCr from the first AG
treatment day to 5 days after treatment end. We also collected
start and end dates and times as well as dosage details for all
AG treatments which occurred in the 6 months before the first
treatment in our screening database.
Exposure definitions
There were two main exposures:
(1) Prior AG treatment, expressed as:
a) number of separate (by 72 h) AG treatments in the pre-
vious 6 months.
b) number of days of AG treatment in the 6 months before
the current AG treatment (calculated by dividing number
of hours of past AG treatment by 24).
(2) Number of days since last AG treatment.
Outcome: AKI
The definition of AKI used in the study was based on the
definition of AKI in the KDIGO AKI guideline [22]. Stage
1 AKI was a rise in SCr level to 1.5- to <2-fold baseline SCr
within 7 days, or by an increase in SCr of 0.3 mg/dl
(26.5 mol/L) from baseline within 48 h; Stage 2 AKI was
an increase in SCr level to 2.0 to <3-fold baseline SCr; Stage 3
AKI was an increase in SCr to 3-fold baseline SCr, or an
increase in SCr to 4.0 mg/dl (353.6 mol/L). We were
unable to evaluate the Stage 3 KDIGO AKI criterion of any
decrease in eGFR to <35 ml/min/1.73 m2 because no height
data were available. We were also unable to fully evaluate the
Stage 1 requirement for timing of SCr rise (50 % within
7 days or 26.5 mol/L within 48 h); however, we verified
that in the majority of AKI cases included in our study, peak
SCr occurred within 7 days of AG treatment start (Table 1).
We did not evaluate KDIGO urine output criteria, since the
Pediatr Nephrol

Table 1 Comparison of
characteristics of acute kidney Patient characteristics No AKI (n = 399) AKI (n = 284) p
injury versus non-acute kidney
injury treatment episodes Gender (male) 195 (48.9 %) 142 (50 %)
Age (years) at AG treatment start 8.9 7.8 7.6 5.3 <0.05
Total hours on AG, current treatment 166.1 157.4 196.4 180.2 <0.05
Peak SCr occurring within 7 days of treatment start 345 (86.5 %) 223 (78.5 %) <0.05
Length of hospital stay (h) 34.4 74.0 51.8 88.2 <0.05
Warda
Pediatric (medical)) 121 (30.4 %) 41 (14.4 %) <0.001
Surgical 129 (32.4 %) 27 (9.5 %)
Hematology-Oncology 148 (37.2 %) 216 (76.1 %)
Main admission reason
Febrile neutropenia 79 (19.8 %) 112 (39.4 %) <0.001
Other 320 (80.2 %) 172 (60.6 %)
AG drug used
Gentamycin 149 (37.3 %) 36 (12.7 %) <0.001b
Tobramycin 237 (59.4 %) 240 (84.5 %)
Amikacin 13 (3.3 %) 8 (2.8 %)

Continuous data is expressed as the mean standard deviation (SD); categorical data are expressed as frequency
(number) with the percentages (for each column) given in parenthesis
AKI, Acute kidney injury; AG, aminoglycosides; Scr, serum creatinine,
a
Patient characteristics do not add up to 399 in the No-AKI group due to lack of data on one patient.
b
Significance at p < 0.001 from the Pearson Chi-square analysis of the relation between AKI and the 3-category
drug name. The overall association was significant. Further analysis revealed that patients receiving tobramycin
had a significantly higher AKI incidence than those not receiving tobramycin (see text)

data were not available from our database as all subjects were outcome of interest (i.e., AKI) and allows for comparison of
admitted to non-critical care wards where reliable urine output different models (by comparing the AUCs between models;
data were not available. for example, predicting AKI using a predictive model with
versus without a variable of interest). Specifically, we sought
to compare predictive models for AKI which did versus did
Statistical analysis
not include the exposure variables of interest, as a way of
evaluating how past AG exposure might be a marker which
Continuous variables were compared between two groups
adds to predicting AKI risk early in an AG treatment. Model-
using the MannWhitney or paired t tests and for multiple
derived AUCs were compared using the DeLong method [24].
groups using the KruskalWallis or one-way analysis of var-
Several sensitivity analyses were performed on multivari-
iance tests (depending on distribution). Chi-square or Fishers
ate analyses to evaluate robustness of the results: (1) only one
exact tests were used to compare categorical variables be-
AG treatment episode per patient was included (last episode in
tween groups.
the database); (2) only AG treatments associated with a known
Logistic regression, in which unadjusted and adjusted odds
bSCr measurement were included; (3) only AG treatments
ratios [adjOR and 95 % confidence interval (CI)] were report-
with 6 complete months of prior observation were included.
ed, was used to evaluate the association of exposure variables
Analyses were performed using Stata 10 statistical soft-
with AKI. In multivariate analysis, ORs were adjusted for
ware package (Stata Corp., College Station, TX).
other variables associated with AKI or decided upon a priori.
Collinearity between variables was evaluated; highly collinear
variables were excluded from the model (e.g., febrile neutro-
penia diagnosis was highly collinear with other variables and Results
was dropped).
The area under the receiver operating characteristics curve Study cohort characteristics
(AUC, C-statistic) resulting from the multivariate logistic re-
gression models to predict AKI (Bpredictive models^) was We were informed of 1709 AG treatments in 1245 individuals.
calculated. The predictive models AUC provides an overall Of these, 1271 treatments met the eligibility criteria (reasons
measure of how well a multivariate model predicts the for exclusion: renal diagnosis for admission, pre-existing

significant renal disease, intensive care unit admission, age
<1 month). Of the 1271 eligible treatment episodes, 588 were
excluded due to the lack of SCr measurements during treat-
ment, leaving a study population of 683 treatment episodes in
366 individuals. AG treatments excluded due to the lack of
SCr data during treatment had a mean [median] treatment
duration of 75 [55] h versus 179 [145] h for treatments where
a SCr measurement was available. Of the 683 treatment epi-
sodes included in the analysis, 317 were associated with at
least one AG treatment in the 6-month period preceding the
current treatment episode. AKI occurred during 284 of the 683
(42 %) AG episodes (Stage 1: 134 [19.6 % of total]; Stage 2:
107 [15.7 %]; Stage 3: 43 [6.3 %]). Baseline SCr data were
available for 617 AG treatment episodes. Table 1 presents
cohort characteristics by AKI status. AKI episodes were asso-
ciated with younger age, being on the hematologyoncology
service, and diagnosis of febrile neutropenia (see Table 1 for
details). AG treatment episodes associated with tobramycin
use (often used to treat febrile neutropenia) were also more
likely to be associated with AKI development compared with
episodes associated with gentamicin/amikacin use (50 % AKI
for tobramycin episodes vs. 21 % AKI for gentamicin/
amikacin episodes; p < 0.001). The hospital stay of patients
with AG treatment episodes associated with AKI was approx-
imately 1.5-fold longer than that of patients with AG episodes
without AKI (Table 1).
Prior AG use and AKI risk
Aminoglycoside treatment episodes with the development of
any AKI were significantly associated with a higher number
of past AG treatment episodes in the prior 6 months (Table 2).
Table 2 Association of past aminoglycoside use with acute kidney injury during a current aminoglycoside treatment episode.
Measures of past AG use Total treatment episodes
No AKI
(n = 399)
Number of AG treatments in preceding 6 months
1, n (%) 139 (34.8 %)
2, n (%) 79 (19.8 %)
Median (IQR) mean 0 (1) 0.9
Number of AG-treatment days in preceding
6 months
Median (IQR) mean 0 (2.3) 3.9
Number of days since last AG treatment episode
Median (IQR) Mean 50 (127) 114
IQR, Interquartile range; AG, aminoglycoside; AKI, acute kidney injury
*, **Significant difference at *p < 0.05 and **p < 0.001 between non-AKI vs. AKI treatment episodes. Significant difference at p < 0.05 across four
groups (non-AKI, Stages 1, 2 and 3 AKI; KruskalWallis test for continuous data and Chi-square test for categorical data). Tests for differences between
individual groups are described in text.
Pediatr Nephrol
Episodes with Stages 1, 2, and 3 AKI were each associated
with a significantly higher number of AG treatment episodes
in the past 6 months compared to non-AKI episodes (all
p < 0.001 for Stages 13 AKI vs. no-AKI; see Table 2 for
details). There was no statistically significant difference in
number of prior AG treatments between the different AKI
severity stages (all p > 0.05). AKI episodes were also associ-
ated with a significantly higher number of AG-treatment days
in the previous 6 months (Table 2). Again, episodes with
Stages 1, 2, or 3 AKI were each associated with a significantly
higher number of AG-treatment days in the prior 6 months
compared to non-AKI episodes (p < 0.0001 for all AKI stage
comparisons vs. no-AKI). Comparison of the number of AG-
treatment days in the prior 6 months according to AKI stage
revealed that the number of AG-treatment days was only sta-
tistically significant for Stage 2 AKI episodes versus Stage 1
AKI episodes (p = 0.002; Table 2).
The multiple logistic regression analysis included number
of AG treatment days in the previous 6 months, age, gender,
service of treatment (indicator variables for medical and he-
matologyoncology service, with surgical service as the ref-
erence variable), and tobramycin versus non-tobramycin treat-
ment (amikacin grouped with gentamycin due to low num-
bers). AKI development during a current treatment was asso-
ciated with number of AG-treatment days in the preceding 6
months (adjOR 1.04, 95 % CI 1.031.06, p < 0.001), younger
age (adjOR 0.96, 95 % CI 0.930.99, p = 0.009), being on the
hematologyoncology service (vs. surgical) (adjOR 3.88,
95 % CI 2.176.96, p < 0.001), and treatment with tobramycin
(adjOR 1.77, 95 % CI 1.043.02, p = 0.04). The AUC from
the AKI prediction model using all these variables was 0.77
(95 % CI 0.740.81). Adding prior AG treatment exposure
Treatment episodes according to AKI stage
Any AKI AKI Stage 1 AKI Stage 2 AKI Stage 3
(n = 284) (n = 134) (n = 107) (n = 43)
161 (56.7 %)** 76 (56.7 %) 63 (58.9 %) 22 (51.2 %)
113 (39.8 %)** 52 (38.8 %) 44 (41.1 %) 17 (39.5 %)
1 (2) 1.5 ** 1 (2) 1.4 1 (2) 1.6 1 (3) 1.7
5.4 (18.5)14.1** 4.1 (14.1)10.0 9.6 (28.7) 19.6 4.6 (21.1) 13.4
27 (51) 55* 28 (44) 59 23 (50) 47 37 (86) 65
Pediatr Nephrol
(AG days in previous 6 months) to the model resulted in a
statistically significant increase in the model AUC (from 0.72
without this variable to 0.77 with this variable; p < 0.001).
Similar multiple logistic regression analyses were per-
formed, but these included number of past AG treatment ep-
isodes (expressed as 2 past treatment episodes for ease of
interpretation) as the main exposure variable, instead of past
AG-treatment days. AKI during the current AG treatment was
associated with the presence of 2 past AG treatment episodes
(adjOR 1.81, 95 % CI 1.242.63; p = 0.002), independent of
other variables. Younger age and being on the hematology
oncology service were similarly associated with higher AKI
risk, as described above. The AUC from the AKI prediction
model using all of these variables was 0.74 (95 % CI 0.70
0.78), which was statistically significantly higher than the
model not including the variable of 2 past AG treatment
episodes (AUC 0.72 vs. 0.74; p = 0.02).
Time since last AG treatment and AKI risk
Acute kidney injury treatment episodes were found to be as-
sociated with fewer days since the last AG treatment, as
shown in Table 2 together with the number of days since the
last AG treatment across all AKI severity strata. Episodes with
Stage 1 and 2 AKI were associated with statistically signifi-
cantly fewer days since last treatment than non-AKI episodes
(p < 0.02 and p < 0.005, respectively, MannWhitney test);
other group comparisons were not significant. On multivariate
analyses (including only episodes associated with a prior AG
treatment), fewer days since last AG treatment was associated
with higher AKI risk independent of other variables (adjOR
0.996, 95 % 0.9930.998). The clinical predictive model in-
cluding number of days since last treatment and other clinical
variables had an AUC of 0.65 (95 % CI 0.590.71) to predict
AKI (vs. AUC 0.60, 95 % CI 0.54-0.67 for the model that did
not include days since last treatment; p < 0.05 for difference
between AUCs).
Sensitivity analyses
When only one AG treatment episode per individual was in-
cluded (the last treatment, n = 366), results were extremely
similar. AKI development was independently associated with
AG-treatment days in the previous 6 months (adjOR 1.03,
95 % CI 1.011.05, p = 0.004), presence of 2 past AG treat-
ments (adjOR 4.73, 95 % CI 2.1310.49, p < 0.001), and few-
er days since last AG treatment (adjOR 0.995, 95 % CI 0.991
0.999). When only those treatment episodes with a known
bSCr (n = 617 ) were included, the results were almost identi-
cal (not shown). The multivariate analysis results were also
extremely similar in magnitude and direction when episodes
with <6 months of prior observation were excluded, except
that the association of AKI with presence of 2 past AG
treatments was no longer statistically significant in the adjust-
ed analyses (adjOR 1.57, 95 % CI 0.972.63, p = 0.09).
Discussion
We have shown that AG treatment in the previous 6 months of
a current AG treatment, whether expressed as prior number of
hours of AG treatment, number of treatments, or time since
last treatment, is a clinical risk factor for AKI in hospitalized
children. Our findings were robust under different sensitivity
analyses, and there was a pattern of increasing AG exposure
being associated with more severe AKI. This is one of the only
studies that has explicitly examined this issue.
Previous studies have clearly shown that there is an impor-
tant association between AG use and AKI [10, 11, 17, 25];
however these studies did not specifically examine prior AG
use as a potential risk factor. Oliveira et al. showed that the
concurrent administration of general nephrotoxic medications
to adults admitted to the intensive care unit was an indepen-
dent risk factor for AG-induced AKI [26]. These authors high-
light the importance of considering other concurrent risk fac-
tors (i.e., other nephrotoxins) when evaluating risk for AKI
caused by AG. However, we have shown that an evaluation of
past nephrotoxin use (specifically, AG) may also help to
identify patients at high risk for AG-induced AKI at the initi-
ation of AG treatment. To our knowledge, this is the first
pediatric study to specifically focus on the patients history
of AG use, a risk factor of AG-induced AKI which is relative-
ly simple to evaluate. Although we were able to control for a
number of potential confounding variables present during the
current AG treatment (e.g., main diagnosis, treatment loca-
tion), we were not able to control for a number of potentially
crucial confounders, including past and current use of other
nephrotoxins, history of past AKI events, and history of non-
AG-associated prior hospitalizations. Thus, we cannot con-
clude that past AG exposure itself leads to chronic renal tubu-
lar damage, increasing the risk for AKI. However, our find-
ings do strongly support that a prior history of AG use is a
marker of risk for developing AKI during a current AG treat-
ment episode. Predictive markers of AKI risk which are
evaluable early in AG treatment could potentially be useful
to risk stratify patients for more intensive AG level and renal
function monitoring, to facilitate decision-making regarding
early AG cessation with early signs of AKI, and to evaluate
the riskbenefit balance when considering AG use. Our AUC
analysis suggests that including knowledge on past AG use (in
particular number of days of past AG use) enhances the ability
to predict AKI risk, over other easily evaluable factors, includ-
ing age and diagnosis. Future studies should evaluate whether
the relation between past AG use and AG-induced AKI is
independent of other factors, in particular past AKI episodes,
use of other nephrotoxins, and other patient characteristics, as

the results of such studies would provide stronger support for
the causative relation between past AG use and AKI risk
Although our study was open to any patients initiated on
AG therapy, about one-half of our study subjects were being
treated for febrile neutropenia in the setting of an oncologic
illness. First, this may limit the generalizability of our study
findings to all pediatric patients, and it is possible that in other
more specific diagnoses (e.g., cystic fibrosis, patients treated
empirically post-operatively), our findings may differ.
However, our study also highlights that many patients treated
for febrile neutropenia are still receiving nephrotoxic AGs,
despite their high risk for AKI. More research is needed to
identify the extent to which using AGs adds benefit to febrile
neutropenia therapy, with the possibility to stop using AGs in
this setting taken into consideration. Again, this study cannot
be used as evidence that AG causes AKI during a current
episode, but the results may serve as a stimulus to question
the use of AGs in high-risk patients with febrile neutropenia.
In addition to the limitations described above (lack of data on
other nephrotoxic exposures and other patient factors, majority of
data from the febrile neutropenia population), our study had other
limitations. First, it was retrospective, with associated design
limitations. However, the only data collected retrospectively were
SCr values and past AG treatment details, which are not subjec-
tive variables and should be of high accuracy. We did not eval-
uate the exact timing of the SCr rise, as suggested by the KDIGO
definition; however, 86.5 % and 78.5 % of patients in the No-
AKI and AKI groups had a peak SCr within 7 days of treatment
start (Table 1), suggesting that the first AKI likely occurred be-
fore then. Had the study been prospective, we might have been
able to ensure daily SCr monitoring in all patients; it is likely that
AKI was underestimated in our study population. The lack of
complete SCr data during AG treatment also led to 588 patients
being excluded due to a lack of on-treatment SCr data. However,
these patients were shown to have very short treatment periods,
likely representing very low-risk patients. It is also possible that
some patients were treated with AG in other centers during the
study period, which would mean that we would not have
completely captured our exposure variable. Although it is diffi-
cult to be certain, this most likely would have led to dampening
of our results, rather than biasing towards stronger associations.
Conclusions
The overall results of our study make a strong case for con-
sidering prior AG treatment when evaluating children treated
with AG. Along with strategies suggested in recent studies to
standardize SCr monitoring during nephrotoxic therapies [9,
27, 28], this could lead to better identification of at-risk chil-
dren and, potentially, a reduction in AG-associated AKI.
Future larger, epidemiologic studies should aim at better quan-
tification of the burden of AG-associated AKI in children in
Pediatr Nephrol
order to obtain an improved picture of the potential impact on
patient outcomes of reducing AKI due to AG treatment.
Acknowledgments We thank the Pharmacy and Medical Records de-
partments and Miss Melissa Piccioni, MSc, of the Montreal Childrens
Hospital for their significant contribution to the data collected for this
study.
Compliance with ethical standards The study was approved by our
institutional research ethics board, which waived the need for informed
consent.
Funding sources Fonds de Recherche du QubecSant; Research
Institute of the McGill University Health Centre. Dr. Zappitelli was a
recipient of salary awards from the Fonds de Recherche du Qubec
Sant and from the Kidney Research Scientist Core Education and
National Training (KRESCENT) program during the course of this study.
Financial disclosure The authors have no financial relationships to
disclose.
Conflict of interest There are no conflicts of interests to disclose.
References
1. Schneider J, Khemani R, Grushkin C, Bart R (2010) Serum creat-
inine as stratified in the RIFLE score for acute kidney injury is
associated with mortality and length of stay for children in the
pediatric intensive care unit. Crit Care Med 38:933939
2. Plotz FB, Bouma AB, van Wijk JA, Kneyber MC, Bokenkamp A
(2008) Pediatric acute kidney injury in the ICU: an independent
evaluation of pRIFLE criteria. Intensive Care Med 34:17131717
3. Zappitelli M, Krawczeski CD, Devarajan P, Wang Z, Sint K,
Thiessen-Philbrook H, Li S, Bennett MR, Ma Q, Shlipak MG,
Garg AX, Parikh CR, consortium T-A (2011) Early postoperative
serum cystatin C predicts severe acute kidney injury following
pediatric cardiac surgery. Kidney Int 80:655662
4. Li S, Krawczeski CD, Zappitelli M, Devarajan P, Thiessen-
Philbrook H, Coca SG, Kim RW, Parikh CR, Consortium T-A
(2011) Incidence, risk factors, and outcomes of acute kidney injury
after pediatric cardiac surgery: a prospective multicenter study. Crit
Care Med 39:14931499
5. Akcan-Arikan A, Zappitelli M, Loftis LL, Washburn KK, Jefferson
LS, Goldstein SL (2007) Modified RIFLE criteria in critically ill
children with acute kidney injury. Kidney Int 71:10281035
6. Hui-Stickle S, Brewer ED, Goldstein SL (2005) Pediatric ARF
epidemiology at a tertiary care center from 1999 to 2001. Am J
Kidney Dis 45:96101
7. Moffett BS, Goldstein SL (2011) Acute kidney injury and increas-
ing nephrotoxic-medication exposure in noncritically-ill children.
Clin J Am Soc Nephrol 6:856863
8. Misurac JM, Knoderer CA, Leiser JD, Nailescu C, Wilson AC,
Andreoli SP (2013) Nonsteroidal anti-inflammatory drugs are an
important cause of acute kidney injury in children. J Pediatr 162:
11531159
9. Menon S, Kirkendall ES, Nguyen H, Goldstein SL (2014) Acute
kidney injury associated with high nephrotoxic medication expo-
sure leads to chronic kidney disease after 6 months. J Pediatr 165
(3):522527.e2
Pediatr Nephrol
10. Tamma PD, Turnbull AE, Harris AD, Milstone AM, Hsu AJ,
Cosgrove SE (2013) Less is more: combination antibiotic therapy
for the treatment of gram-negative bacteremia in pediatric patients.
JAMA Pediatr 167:903910
11. Zappitelli M, Moffett BS, Hyder A, Goldstein SL (2011) Acute
kidney injury in non-critically ill children treated with aminoglyco-
side antibiotics in a tertiary healthcare centre: a retrospective cohort
study. Nephrol Dial Transplant 26:144150
12. Smyth A, Lewis S, Bertenshaw C, Choonara I, McGaw J, Watson A
(2008) Case-control study of acute renal failure in patients with
cystic fibrosis in the UK. Thorax 63:532535
13. Mingeot-Leclercq MP, Tulkens PM (1999) Aminoglycosides:
nephrotoxicity. Antimicrob Agents Chemother 43:10031012
14. Swan SK (1997) Aminoglycoside nephrotoxicity. Semin Nephrol
17:2733
15. Olbricht CJ, Fink M, Gutjahr E (1991) Alterations in lysosomal
enzymes of the proximal tubule in gentamicin nephrotoxicity.
Kidney Int 39:639646
16. Glass S, Plant ND, Spencer DA (2005) The effects of intravenous
tobramycin on renal tubular function in children with cystic fibrosis.
J Cyst Fibros 4:221225
17. Khalili H, Bairami S, Kargar M (2013) Antibiotics induced acute
kidney injury: incidence, risk factors, onset time and outcome. Acta
Med Iran 51:871878
18. Knijnenburg SL, Mulder RL, Schouten-Van Meeteren AY,
Bokenkamp A, Blufpand H, van Dulmen-den Broeder E,
Veening MA, Kremer LC, Jaspers MW (2013) Early and late renal
adverse effects after potentially nephrotoxic treatment for child-
hood cancer. Cochrane Database Syst Rev 10:CD008944
19. Akbar A, Rees JH, Nyamugunduru G, English MW, Spencer DA,
Weller PH (1999) Aminoglycoside-associated hypomagnesaemia
in children with cystic fibrosis. Acta Paediatr 88:783785
20. Abramowsky CR, Swinehart GL (1982) The nephropathy of cystic
fibrosis: a human model of chronic nephrotoxicity. Hum Pathol 13:
934939
21. Alkandari O, Eddington KA, Hyder A, Gauvin F, Ducruet T,
Gottesman R, Phan V, Zappitelli M (2011) Acute kidney injury is
an independent risk factor for pediatric intensive care unit mortality,
longer length of stay and prolonged mechanical ventilation in crit-
ically ill children: a two-center retrospective cohort study. Crit Care
15(3):R146
22. Kidney Disease: Improving Global Outcomes (KDIGO) Acute
Kidney Injury Work Group (2012) KDIGO clinical practice guide-
line for acute kidney injury. Kidney Int 2:1138
23. Zappitelli M, Parikh CR, Akcan-Arikan A, Washburn KK, Moffett
BS, Goldstein SL (2008) Ascertainment and epidemiology of acute
kidney injury varies with definition interpretation. Clin J Am Soc
Nephrol 3:948954
24. DeLong ER, DeLong DM, Clarke-Pearson DL (1988) Comparing
the areas under two or more correlated receiver operating charac-
teristic curves: a nonparametric approach. Biometrics 44:837845
25. Pankhurst T, Mani D, Ray D, Jham S, Borrows R, Coleman JJ,
Rosser D, Ball S (2014) Acute kidney injury following unselected
emergency admission: role of the inflammatory response, medica-
tion and co-morbidity. Nephron Clin Pract 126:8189
26. Oliveira JF, Silva CA, Barbieri CD, Oliveira GM, Zanetta DM,
Burdmann EA (2009) Prevalence and risk factors for aminoglyco-
side nephrotoxicity in intensive care units. Antimicrob Agents
Chemother 53:28872891
27. Goldstein SL, Kirkendall E, Nguyen H, Schaffzin JK, Bucuvalas J,
Bracke T, Seid M, Ashby M, Foertmeyer N, Brunner L, Lesko A,
Barclay C, Lannon C, Muething S (2013) Electronic health record
identification of nephrotoxin exposure and associated acute kidney
injury. Pediatrics 132:e756e767
28. Kirkendall ES, Spires WL, Mottes TA, Schaffzin JK, Barclay C,
Goldstein SL (2014) Development and performance of electronic
acute kidney injury triggers to identify pediatric patients at risk for
nephrotoxic medication-associated harm. Appl Clin Inform 5:313
333