Extracorporeal membrane oxygenation (ECMO) is used mainly in neonates to treat a variety of cardiorespiratory problems such as meconium aspiration syndrome, congenital diaphragmatic hernia, persistent pulmonary hypertension of the newborn, and sepsis/pneumonia 1.

The ECMO circuit, like the cardiopulmonary bypass (CPB) circuit, triggers an important inflammatory reaction and is clinically associated with the so-called capillary leakage syndrome, resulting in intravascular hypovolaemia and renal hypoperfusion 2. Consequently, in the initial phase (in the first 24–48 hours) the ECMO patient becomes usually increasingly oedematous. Diuretics, especially loop diuretics such as furosemide, are therefore the mainstay in the enhancement of diuresis to mobilize fluid excess. Furosemide is often used as a continuous infusion in patients treated with ECMO, based upon the observations in infants after CPB surgery 3456.

We recently made an inventory of furosemide regimens used in neonates treated with ECMO and concluded that continuous intravenous furosemide was frequently used, but the used regimens varied widely in continuous doses and additional intermittent doses 7. Although adequate urine output was achieved within 24 hours with all regimens, the used furosemide regimens might not be the optimal regimen for neonates treated with ECMO. In an accompanying editorial it was suggested that development of more standardized and efficacious dosing regimens would be preferable 8.

Since ECMO and CPB result in fluid overload, at least partially based on the same pathophysiology, it seems reasonable to assume that pharmacokinetic/pharmacodynamic (PK/PD) models developed for infants following cardiac surgery might also be applicable for neonates treated with ECMO 9. We therefore conducted a prospective exploratory study in neonates treated with ECMO to evaluate a suggested furosemide regimen that was initially developed for infants after CPB surgery. The regimen consisted of a continuous furosemide infusion at a rate of 0.2 mg/kg per hour that was preceded by a loading bolus. The aim was to achieve a urine output of 6 ml/kg per hour. The main objectives of the study were to establish the efficacy of such a regimen and also to document serum furosemide concentrations to rule out ototoxic levels.

In the present article we report the findings of the proposed furosemide regimen in neonates treated with venoarterial ECMO in our unit.

The study was performed at the paediatric surgical intensive care unit of the Sophia Children's Hospital of Erasmus Medical Centre in Rotterdam, The Netherlands. The study protocol was approved by the Committee on Medical Ethics of the Erasmus Medical Centre and was conducted according to the principles of the Declaration of Helsinki. Parental written informed consent was obtained for all patients.

Since the observation that continuous intravenous furosemide might be superior to intermittent administrations in infants after CPB surgery, the use of continuous furosemide infusion has increasingly be documented in patients following CPB surgery 345615. Based upon the observations in infants after CPB surgery, the use of continuous intravenous furosemide in neonates treated with ECMO is increasing.

We recently evaluated furosemide regimens used in neonates treated with ECMO in our unit and concluded that continuous intravenous furosemide was frequently used in neonates (78%) treated with ECMO 7. The furosemide regimens used varied widely, in continuous doses and in additional intermittent doses. Although all used regimens achieved adequate urine output within 24 hours, the use of additional furosemide bolus injections suggests that the regimens might not be the optimal for neonates treated with ECMO, and therefore dosing regimens should be developed 7.

Since ECMO and CPB are 'comparable' procedures, the developed PK/PD model for infants after CPB surgery might also be applicable for neonates treated with ECMO 9. There are, however, obvious differences between ECMO and CPB: in the time of exposure to the procedure, and thereby the presence of the 'circuit' with an ongoing inflammatory reaction, in the underlying illness and in the age of the patients. We therefore conducted a prospective exploratory study in neonates treated with ECMO to evaluate a suggested furosemide regimen developed for infants after CPB surgery. The results suggest that the used regimen was effective and well tolerated in neonates treated with ECMO.

Continuous intravenous furosemide was started in all patients at a rate of 0.2 mg/kg per hour and was preceded by a loading bolus of 1 mg/kg. The furosemide dose was adapted according to urine output. The dose was decreased from the first day to the second day of the study, from 0.17 ± 0.06 mg/kg per hour to 0.08 ± 0.04 mg/kg per hour. The furosemide doses used in neonates treated with ECMO (0.17 ± 0.06, 0.08 ± 0.04, and 0.12 ± 0.07 mg/kg per hour) were lower than the doses used in infants after CPB surgery (0.22 ± 0.06, 0.25 ± 0.10, and 0.22 ± 0.11 mg/kg per hour) over the first day, second day, and third day of furosemide therapy, respectively 16.

The PK/PD model for diuretic therapy with furosemide in infants after CPB suggested that doses between 0.2 and 0.3 mg/kg per hour, preceded by a loading bolus, would result in a urine production of 6 ml/kg per hour 9. Based upon our observational study, which indicated that relatively low doses of continuous furosemide were used, we decided to use the lowest dose suggested by the model. The rational for the loading bolus was based on the simulated urine production profiles generated with the use of different furosemide regimens and on the observed effects of the loading bolus in the retrospective study 79.

In the retrospective study, positive effects of the 'loading' bolus were observed, although not statistically significant, in the urine output in the first 24 hours and in the time to reach the desired urine output of 6 ml/kg per hour 7. Also, no additional furosemide bolus injections were administered during the continuous infusion to the patients who received a bolus prior to the continuous infusion. These observed effects might suggest that one loading bolus might be sufficient to overwhelm the effects of the ECMO circuit.

The data from the present study suggest that the starting dose was too high, as indicated by the urine output exceeding the target urine output in the first 24 hours. Although a full understanding of this phenomenon is hard to reach, it seems logical to assume that contributing factors might be the ECMO circuit, the renal function of the patients, and the age of the patients 17181920212223. The patients treated with ECMO were younger (median 3 days) than the patients after CPB surgery (median 12 weeks), and therefore by definition had a less mature renal function, which leads to a decreased renal clearance of furosemide.

The renal function (median creatinine 30 μmol/l) was normal for age in the ECMO patients, whereas (transient) renal failure (median creatinine 95 μmol/l) was observed in the majority of the patients after CPB surgery 1216. Therefore it can be hypothesized that the acute renal failure observed in the patients after CPB surgery had a major impact on renal clearance, which is most closely related with drug response, since furosemide is excreted renally and only acts after reaching the tubular lumen 24252627. This hypothesis might explain why higher doses were needed in the patients after CPB surgery. In addition, phase II reactions are better developed in infants and, as a result, the percentage of furosemide glucuronide will be higher 23. Less unchanged furosemide can therefore be assumed available to interact with the furosemide receptor in the infants included in the cardiac surgery study, and higher doses are consequently needed to reach the same furosemide excretion rate 2526. This assumption might clarify why higher doses were required in the patients after CPB surgery.

On the other hand, the lower continuous furosemide doses after the loading bolus used in the ECMO patients might be explained by the effects of the ECMO circuit 1718. The observed increased volume of distribution in our patients was in accordance with the values reported in the literature 17. Wells and colleagues reported that the steady-state volume of distribution and the elimination half-life of the loop diuretic, bumetanide, in term neonates treated with ECMO were increased compared with values in premature and term neonates without ECMO, while the plasma clearance was similar for both groups 17.

The increased volume of distribution is not only due to the addition of a large exogenous blood volume for priming of the circuit, but is also caused by the possible absorption of furosemide onto the ECMO circuit components 1828. Scala and coworkers performed an in vitro analysis to identify loss of furosemide in the ECMO circuit and observed a reduction of 63–87% in the serum furosemide concentration over a 4-hour period. The loss of drug was most pronounced in the first 30 minutes 29. Since the continuous infusion was started at the time of the bolus injection, and as only furosemide samples were taken during the continuous infusion, we could not estimate the furosemide clearance in our patients.

Mehta and colleagues recently published research on the potential sequestration of drugs to the ECMO circuit. In vivo experiments showed that there was a significant drug loss in crystalloid-primed circuits as well as in blood-primed circuits. For instance, the loss of analgetics ranged from 17% for morphine to 87–100% for fentanyl depending on the type of circuit 30. In addition, our own group described a decreased clearance of morphine during the first 10 days of ECMO in neonates and infants treated with venoarterial ECMO compared with patients after noncardiac major surgery 3132.

The furosemide loading bolus especially seems to compensate for the increased volume of distribution. Since the effects of furosemide are dependent on renal function, the apparent need for lower continuous furosemide dose might be explained by the absence of impaired renal function, and consequently the increased renal clearance, in the patients on ECMO compared with the patients post CPB surgery 24.

We previously noticed that additional loop diuretics were needed in approximately 40% of the patients on ECMO therapy during the continuous furosemide infusion 7. In the present study no additional loop diuretics were needed, demonstrating that furosemide monotherapy is highly effective, which is a considerable advantage.

The total administered furosemide dose in the current study was substantial higher on the first day (4.97 mg/kg per 24 hours) than the dose used in our retrospective study (1.92 mg/kg per 24 hours). The respective doses were slightly lower on the second day and third day (1.63 mg/kg per 24 hours and 1.50 mg/kg per 24 hours in the present study compared with 1.92 mg/kg per 24 hours and 2.0 mg/kg per 24 hours in the retrospective study). The cumulative furosemide doses over the three study days, however, were comparable between the two studies. The cumulative furosemide dose in the current study showed less variation in dose 7. Importantly serum furosemide levels remained far below the commonly accepted safety level for ototoxicity (50 μg/ml) 33.

To obtain an acceptable fluid balance with a maintenance fluid of 120–140 ml/kg per 24 hours, the target urine production is set at 6 ml/kg per hour in our unit. In all patients studied, the target urine production of 6 ml/kg per hour was obtained a median 7 hours after the start of the continuous infusion. This is considerable faster than in our retrospective study in which the target urine production was reached in median 24 hours. The rapid attainment of the target urine may be explained by the initial higher infusion rate and the loading bolus.

The observed variability in urine output was small (4.1–8.8 ml/kg per hour) throughout the entire observation period – although it was striking that in one patient, despite administration of a high dose of furosemide, the urine output remained low, if not negligible, for a period of approximately 33 hours. We could not identify an obvious cause for this. In our retrospective study in which the patients received additional intermittent furosemide bolus injections, the variability in urine output was 0.7–16.1 ml/kg per hour during the study period. This is in accordance with studies in infants after CPB surgery, where less variance in urine output was observed with continuous administration compared with intermittent furosemide administration 345. This suggests that strict protocols for diuretic therapies reduce variability in patients' response. It is probable that a tailored PK/PD model for furosemide therapy in neonates treated with ECMO may further optimize diuretic therapy for these critically ill neonates.

The obtained fluid balances were approximately zero for all three study days, although with substantial variability. The forced diuresis was well tolerated, as shown by the stable haemodynamic parameters and by the reduction of the vasopressor score.

Hypochloraemic metabolic alkalosis is a well-known side effect of furosemide therapy. A tendency for metabolic alkalosis was observed in two patients after approximately 48 hours of furosemide therapy. Since hypochloraemia was present in one patient, furosemide therapy was most probably the cause of the metabolic alkalosis. We have no explanation, however, for the metabolic alkalosis in the other patient, after contraction alkalosis and prerenal failure were excluded, and no increased use of inotropic drugs was present. This aspect should be recognized in the ongoing development and testing of a PK/PD model including more patients.

The evaluated furosemide regimen of 0.2 mg/kg per hour preceded by a loading of 1 mg/kg is an effective means to obtain rapid and sufficient diuresis without cardiovascular instability in neonates treated with ECMO with a relatively low interpatient variability in urine production. The present exploratory study, however, suggests that for neonates on ECMO the proposed furosemide regimen as used in infants after CPB is using furosemide doses for the continuous infusion that are too high. A PK/PD model should therefore be developed for neonates on ECMO, identifying factors such as the circuit age, renal function and albumin that influence drug disposition during ECMO.

• The furosemide regimen, proposed by the PK/PD model developed for infants after CPB surgery, is too high for neonates treated with ECMO.

• A PK/PD model should be developed for neonates on ECMO, identifying the factors that influence drug disposition during ECMO.

CPB = cardiopulmonary bypass; ECMO = extracorporeal membrane oxygenation; PK/PD = pharmacokinetic/pharmacodynamic.

The authors declare that they have no competing interests.

MMJvdV and JB designed the study, evaluated the data, and wrote the manuscript. JdH analysed the furosemide samples, EW and DT were involved with patient management.