Keywords extracorporeal membrane oxygenation - ECMO - pediatric - kidney - pharmacokinetics
- intensive care
Introduction
Extracorporeal membrane oxygenation (ECMO) is a lifesaving organ support in critically
ill patients to support or replace cardiac and respiratory function. The pharmacokinetics
of medications on ECMO can be severely altered because of the increased distribution
volume, sequestration to tubing and oxygenator, hemolysis, reduced clearance, and
inflammatory responses.[1 ]
[2 ] Vancomycin is a frequently used antibiotic in children on ECMO due to their high
risk of infection with gram-positive cocci, especially when children have open chest
cannulation and delayed sternal closure.[3 ] Vancomycin is renally eliminated, but kidney function can be critically impaired
on ECMO. A recent review described a high percentage of target non-attainment with
standard dosing regimens.[4 ] Published investigations have reported very few children given vancomycin on ECMO,
or they involved no control group of critically ill children. We hypothesize children
on ECMO require a different vancomycin dose than children without ECMO support. Our
aim was to determine vancomycin's necessary dosage adjustment under ECMO compared
to critically ill children with delayed sternal closure and hence similar disease
severity to extrapolate the influence of ECMO support more accurately on the pharmacokinetics.
Methods
We conducted an 8-year, single-center, retrospective cohort study at our pediatric
cardiac intensive care unit (ICU) in a tertiary referral heart center on children
(0–18 years) undergoing ECMO support. Critically ill children with delayed sternal
closure after cardiac bypass surgery served as the control group (non-ECMO). Data
were collected from digital patient charts.
The reason for vancomycin administration was prophylaxis of mediastinitis, surgical
site infection, and sepsis. Vancomycin serum levels were collected as trough levels
before the next administered dose. The first level was obtained before the third dose
and in reduced renal function earlier. Trough levels of 10 to 20 μg/ml were considered
in the target range as this was the standard operating procedure on this unit. These
levels were predefined by the fact that at our pediatric cardiac ICU, we had only
21 cases of methicillin-resistant Staphylococcus aureus out of 238 overall detected Staphylococcus and a minimal inhibitory concentration (MIC) of >1 mg/l only in 43/238 cases. Vancomycin
serum levels were measured via immunoassay by Cobas 8000 modular analyzer (Roche Diagnostics
Deutschland GmbH, Mannheim, Germany). The vancomycin dose was calculated as the total
daily dose per kilogram of body weight. The starting vancomycin dose in both cohorts
for neonates was (day 0–7 of age) 10 to 15 mg/kg per dose every 12 hours, for neonates
(day >7) 10 to 15 mg/kg per dose every 8 hours, for children after the neonatal period
40 mg/kg per day in three single doses. Administration interval was adjusted according
to age, renal function, and serum levels at 6, 8, 12, or 24 hours. Vancomycin doses
and levels as well as laboratory parameters included in the calculations were of the
time on ECMO in the ECMO group, or the open chest time in the non-ECMO group. Laboratory
parameters like serum urea, creatinine, alanine aminotransferase (ALAT), cholinesterase,
D-dimers, N-terminal fragment brain natriuretic peptide (NT-proBNP), and troponin
T were measured repeatedly during ECMO or open chest.
This retrospective data analysis was approved by the ethical review committee (Freiburg
University 23-1000-S1-retro) and in accordance with the ethical review committee,
the request for parental consent was waived. Data are presented as median and its
interquartile range (IQR). Boxplots depict the median, hinges the IQR, and whiskers
the largest value up to 1.5*IQR from the hinge and beyond outlier points. Statistical
analysis with Wilcoxon's rank-sum test was performed for not normally distributed
variables. p -values <0.05 were considered statistically significant. RStudio 2021.09.1 was used
for statistical analysis and graph creation.
Results
Patient enrollment and exclusions are listed in [Fig. 1 ]. We included a total of 85 courses on ECMO in 82 children and 99 non-ECMO courses
in 92 children receiving vancomycin ([Table 1 ]). Our patients were mainly on venoarterial ECMO (93%), with the majority undergoing
open chest cannulation postcardiotomy if they could not be safely weaned from cardiopulmonary
bypass (86%). The remaining patients were cannulated via neck or femoral cannulation
not postcardiotomy (14%). All 85 ECMO courses were on new-generation rotary ECMO blood
pumps: 77 on Medos Deltastream DP3 (Fresenius, Bad Homburg, Germany), 7 on Sorin Revolution
Centrifugal pump (Livanova, London, United Kingdom), and 1 on CentriMag Blood Pump
(Abbott, Illinois, United States). Depending on the calculated cardiac output, we
used the following oxygenators: Medos Hilite 800 LT (Fresenius, Bad Homburg, Germany),
Newborn A.L.One 14 days (Eurosets S.r.l., Medolla, Italy), or Medos Hilite 2400 LT
(Fresenius, Bad Homburg, Germany). Both groups presented similar body weight (4.6 kg
in ECMO vs. 3.9 kg in non-ECMO) although the ECMO group children were slightly older
(144 vs. 63 days). There were no significant differences between the ECMO and non-ECMO
groups in surgical duration, cardiopulmonary bypass time, aortic cross-clamp time,
or lowest body temperature during surgery. However, the amount of blood transfusions—including
red blood cell concentrates, fresh frozen plasma, and platelet concentrates—was significantly
higher in the ECMO group (p < 0.001 for all). For all patients' characteristics, see [Table 1 ].
Fig. 1 Patient enrollment in the ECMO and non-ECMO groups. ECMO, extracorporeal membrane
oxygenation.
Table 1
Patients' characteristics
Characteristics median (interquartile range) or n (%)
ECMO
non-ECMO
p -value
Treatment courses
85
99
Patients
82
92
Age, days
144 (10–940)
63 (7–296)
0.048
Body weight, kg
4.6 (3.4–11.5)
3.9 (3.4–7.2)
0.133
Days on ECMO or open chest
4 (2–6)
3 (2–4)
0.001
VA ECMO/VV ECMO
79 (93%)/6 (7%)
–
–
Surgery
Duration, minutes
455 (152–550)
391 (300–479)
0.340
Cardiopulmonary bypass time, minutes
218 (0–282)
197 (150–236)
0.506
Aortic cross-clamp time, minutes
70 (0–128)
82 (47–134)
0.062
Nadir temperature, °C
29.9 (25.0–33.8)
29.7 (25.5–33.7)
0.903
Blood transfusions, mL/kg/day
Red blood cell concentrates
16.0 (7.4–36.0)
0 (0–5.5)
<0.001
Fresh frozen plasma
6.9 (2.0–21.7)
3.0 (0–9.6)
<0.001
Platelet concentrates
10.7 (4.4–22.0)
0 (0–2.9)
<0.001
Renal failure[a ]
Day 0
18 (21%)
13 (13%)
0.148
Last day of treatment
32 (38%)
21 (21%)
0.014
GFR at day 0, mL/min/1.73 m2
55.5 (35.7–77.7)
51.9 (39.4–78.9)
0.575
Renal replacement therapy
19 (22%)
9 (9%)
0.013
Hemodialysis
7 (8%)
0
0.004
Peritoneal dialysis
15 (18%)
9 (9%)
0.087
In-hospital deaths
40 (47%)
8 (8%)
<0.001
Time to death, days
8 (5–17)
8 (6–18)
0.220
Overall vancomycin dose, mg/kg/day
33.3 (19.1–44.4)
38.5 (29.4–44.1)
<0.001
Age 0–7 days, mg/kg/day
29.0 (21.3–35.1)
30.8 (28.6–39.0)
0.355
Age >7 days, mg/kg/day
33.1 (24.4–46.1)
40.3 (33.0–44.7)
0.044
GFR normal at day 0, mg/kg/day
37.5 (21.9–46.9)
38.2 (29.4–43.5)
0.299
GFR reduced at day 0, mg/kg/day
22.5 (13.8–33.3)
42.1 (30.4–47.6)
<0.001
Vancomycin dose on day X
Day 0, mg/kg/day
39.1 (33.3–44.3)
38.8 (30.3–43.5)
0.631
Day 3, mg/kg/day
28.7 (13.3–42.7)
33.7 (28.9–43.1)
0.071
Vancomycin level, μg/mL
11.2 (7.1–16.4)
11.5 (8.9–16.7)
0,055
Serum urea, mg/dL
42 (27–69)
32 (23–49)
<0.001
Creatinine, mg/dL
0.54 (0.34–0.85)
0.53 (0.40–0.70)
0.552
ALAT, U/L
36 (20–68)
24 (16–41)
<0.001
D-dimer, mg/L
4.6 (1.7–14.7)
1.9 (0.6–4.7)
<0.001
Abbreviations: ALAT, alanine aminotransferase; ECMO, extracorporeal membrane oxygenation;
GFR, glomerular filtration rate; VA, venoatrial; VV, venovenous.
Characteristics of ECMO and non-ECMO groups. Data are presented as median (interquartile
range) or n (%).
a Defined by glomerular filtration rate less than age-related normal range calculated
by revised Schwartz equation.
Death while on ECMO occurred during nine ECMO courses (11%). ECMO weaning failed with
death within ≤24 hours after ECMO explantation in 13 courses (15%). Patients were
not put back on ECMO if the prognosis was truly dire and we had the parents' consent.
We observed successful weaning entailing survival >24 hours after ECMO explantation
in 63 courses (74%). Unadjusted survival probability at the end of our 30-day follow-up
period was 59% in ECMO (95% CI: 49–70%) and 93% in the non-ECMO group (95% CI: 88–98%;
[Supplementary Fig. S1 ] [available in the online version only]). Adjustments for age, gender, ECMO support,
dialysis, and the total vancomycin dose demonstrated an association between survival
and greater age (p = 0.046), non-ECMO support (p < 0.001), and a higher overall vancomycin dose (p = 0.007). The proportion of renal failure defined by the glomerular filtration rate
(GFR) under the age-related normal range calculated by the revised Schwartz equation
tended to be higher but not significantly different on day 0 in the ECMO group ([Table 1 ]). This number rose until the ECMO group's last day of ECMO support and the last
day of the open chest in the non-ECMO group to 38% and 21% (p = 0.014), respectively. The ECMO group underwent renal replacement therapy (RRT)
more often (22% in ECMO vs. 9% in non-ECMO), as hemodialysis only in the ECMO group
(7 vs. 0 times), and as peritoneal dialysis more often in the ECMO group (15 vs. 9
times), but this did not differ significantly.
The total number of vancomycin administration days while on ECMO or while opened chest,
respectively, was 416 in the ECMO and 372 in the non-ECMO group. The ECMO group's
necessary total daily vancomycin dose was significantly lower than the non-ECMO group's
(33.3 vs. 38.5 mg/kg/d; [Fig. 2A ]). At treatment initiation on day 0, both groups' doses were similar (39.1 and 38.8 mg/kg/d,
respectively). During the later treatment course, doses required adjustment to maintain
the vancomycin serum trough level within the target range of 10 to 20 μg/ml. The total
number of measured vancomycin serum levels was 259 in the ECMO and 189 in the non-ECMO
group. Both groups' total vancomycin serum trough levels were similar and in the target
range (11.2 and 11.5 μg/ml, respectively; [Fig. 2B ]). To achieve that range, the ECMO group's daily vancomycin dose declined over the
treatment faster than the non-ECMO group ([Fig. 2C ]).
Fig. 2 Overall vancomycin doses (A ) and serum levels (B ) in ECMO and non-ECMO groups. Median vancomycin doses by day on treatment with linear
regression and patients' trajectories (C ). ECMO, extracorporeal membrane oxygenation.
A subgroup analysis of normal and reduced GFR at the start of treatment on day 0 showed
that if GFR was normal at the start, the total daily vancomycin dose over the whole
ECMO support period or delayed sternal closure period was similar between groups ([Fig. 3A ]). But if GFR fell below the age-related range, the ECMO group's vancomycin dose
required to maintain serum levels within the target range was significantly lower
(22.5 vs. 42.1 mg/kg/d; [Fig. 3B ]). Both groups revealed a similar correlation between higher serum urea concentrations
and lower required vancomycin doses ([Fig. 3C, D ]).
Fig. 3 Vancomycin dose in subgroups with normal GFR (calculated by revised Schwartz equation)
at the start of treatment on day 0 (A ) versus reduced GFR on day 0 (B ). Linear regression between serum urea and vancomycin dose for ECMO (C ) and non-ECMO group (D ). ECMO, extracorporeal membrane oxygenation; GFR, glomerular filtration rate.
The ECMO group's total serum urea, ALAT, and D-dimers were higher than the non-ECMO
group's ([Supplementary Fig. S2 ] [available in the online version only]), while creatinine, cholinesterase, troponin
T, and NT-proBNP did not differ significantly.
Discussion
In our study with a cohort including 85 ECMO courses, we compared the necessary vancomycin
doses between children on ECMO and critically ill children to highlight the effects
of ECMO support. Our cohorts, ECMO and non-ECMO groups differed slightly in age with
younger children in the non-ECMO group. However, because kidney function matures during
the first year of life, our ECMO group's lower vancomycin dose—although they were
slightly older children—might even lead to an underestimation of the dose discrepancy,
given the less mature renal function of the non-ECMO group while required higher vancomycin
dose.[5 ] Nevertheless, as there was no difference in body weight and length, the two groups
are comparable. Our non-ECMO group children with delayed sternal closure had undergone
heart–lung machine surgery; such children are, often hemodynamically unstable, in
an increased inflammatory condition, and are more prone to infections because of their
open chest and multiple catheters inserted.[6 ] The similar state of critical illness of both groups is also supported by similar
surgical duration, cardiopulmonary bypass time, aortic cross-clamp time, lowest body
temperature during surgery, levels of creatinine, cholinesterase, troponin T, and
NT-proBNP, which did not differ significantly. Hence, our non-ECMO group, being critically
ill children, is suitable for comparison to the ECMO group—thus highlighting the ECMO
circuit's impact and less the ECMO group's critically ill state.
At the start of treatment on day 0, we noted slightly but not significantly more ECMO-group
patients with renal failure (21%) than in the non-ECMO group (13%). However, on the
last day of support, a significantly higher number of ECMO patients were suffering
from renal failure. This might be because these ECMO patients in such critical condition
may also have lost some kidney function because of ECMO's influence on it, for example,
through the absence of pulsatility.[7 ]
[8 ] They therefore also required renal replacement therapy (RRT) more often (22%) than
did non-ECMO children (9%). Note that only the ECMO group required hemodialysis. Their
observed greater occurrence of hemodialysis might be because it is more feasible on
ECMO because of the dialysis connection into the ECMO circuit. Furthermore, the RRT
indication on ECMO is not only acute kidney injury but also fluid overload caused
by capillary leakage, accompanied by increased mortality on ECMO.[9 ]
[10 ] Even though vancomycin is removed via continuous hemodialysis and a higher dose
is usually needed, the ECMO group's total required vancomycin dose was lower while
being more often on hemodialysis.[11 ] This fact reveals that the total effect of RRT on dosing is inferior.
The ECMO group's total required vancomycin dose was significantly lower than the non-ECMO
group. As ECMO has such a strong and varied impact on pharmacokinetics, dose adjustments
of medications in both directions have been observed.[4 ]
[12 ] The present study demonstrates a necessary reduction in the daily vancomycin dose
during the initial days of ECMO. Although critically ill children in the non-ECMO
group also needed a dose reduction, the ECMO group's required dose was even lower.
This evidence supports the assumption that there is no relevant absorption of vancomycin
in the circuit because of its low lipophilicity.[13 ] Hence, factors like impaired kidney function demand lowering the vancomycin dose,
but not exclusively, as in our subgroup with impaired kidney function, only those
children on ECMO needed it. This means that neither impaired kidney function without
ECMO nor ECMO alone requires reducing the vancomycin dose, but the coexistence of
ECMO and reduced kidney function does mandate lowering it. Supporting this, we detected
no significant difference between creatinine serum levels and similar GFR at day 0
in both groups. In line with ours and their small prospective study, An et al demonstrated
decreased vancomycin clearance in neonates on ECMO and necessary dosing adjustment.[14 ] Children on ECMO often suffer from acute kidney injury.[8 ] The vancomycin level and kidney function are interdependent. Vancomycin excretion
depends primarily on kidney function. Nephrotoxicity can occur at vancomycin trough
levels of >20 mg/l.[15 ] The nephrotoxicity risk is high, especially when additional nephrotoxic agents have
been given. Hence, it is even more essential to prevent overshooting vancomycin levels
in patients on ECMO in advance.[16 ] The novel finding of our study is the time-dependent dose reduction during the initial
days of ECMO support which should raise awareness for the uncritical application of
a loading dose at the start of vancomycin administration.
Prophylactic vancomycin administration is common in patients on ECMO and those with
delayed sternal closure.[3 ]
[17 ]
[18 ] However, due to the increasing development of antimicrobial resistance, careful
consideration regarding antibiotic stewardship is essential before its use. In our
study, vancomycin was often administered prophylactically according to protocol, starting
with the initiation of ECMO or with the delayed sternal closure.
According to the pediatric Extracorporeal Life Support Organization registry's report,
the rate of survival-to-hospital discharge in a neonatal cohort on cardiac ECMO was
45%, in a pediatric cohort 57%, and overall survival of 52%.[19 ] We observed 30-day survival of 59% in our ECMO cohort versus 93% in the critically
ill non-ECMO group. Moreover, our multivariable analysis revealed that a higher vancomycin
dose was associated with longer survival as dose reductions often become necessary
in extremely critical conditions.
Limitations of this study are due to its retrospective design, being unable to compare
the same severity of disease, thus we may have overestimated the true effect of ECMO
on vancomycin dosages because of our ECMO group's more severe illness and more impaired
kidney function. Furthermore, trough concentrations are not as good a surrogate parameter
for effective antimicrobial treatment and safety as is the area under the concentration–time
curve divided by the MIC; to do that, we would have needed sequential blood samples,
which were unavailable because of our study's retrospective design.[20 ]
Conclusion
In pediatric patients on ECMO, careful vancomycin dosing is crucial, especially when
renal function is reduced at ECMO initiation. Our data suggest that the recommendation
to start with a higher loading dose to account for the greater volume of distribution
should be avoided in these cases to prevent overshooting therapeutic levels and further
renal impairment. In children on ECMO with reduced GFR, early therapeutic drug monitoring,
even before reaching a steady state with trough levels before the second or third
vancomycin dose, should be considered to avoid toxicity.
