Intensive Care Medicine

, Volume 39, Issue 5, pp 951–957

Incidence of milrinone blood levels outside the therapeutic range and their relevance in children after cardiac surgery for congenital heart disease


    • Department of PediatricsUniversity of Alberta
  • Ari R. Joffe
    • Department of PediatricsUniversity of Alberta
  • Ambikaipakan Senthilselvan
    • Department of Public Health SciencesUniversity of Alberta
  • Demetrios J. Kutsogiannis
    • Division of Critical Care MedicineUniversity of Alberta
    • Critical Care Medicine and PediatricsUniversity of Toronto School of Public Health
Pediatric Original

DOI: 10.1007/s00134-013-2858-3

Cite this article as:
Garcia Guerra, G., Joffe, A.R., Senthilselvan, A. et al. Intensive Care Med (2013) 39: 951. doi:10.1007/s00134-013-2858-3



To evaluate whether variability in milrinone blood levels (MBL) occurs during administration to critically ill children after surgical repair of congenital heart disease, and the clinical relevance of this variability.


Prospective cohort study conducted in the pediatric intensive care unit of a tertiary care teaching and referral hospital. MBL were measured at three time periods after starting milrinone infusion (9–12, 18–24, 40–48 h) and at the end of the infusion. MBL were categorized as within (100–300 ng/ml) or outside the therapeutic range. Low cardiac output syndrome was defined by elevation of either lactate (>2 mmol/l) or arteriovenous oxygen difference (>30 %). Five other clinical outcomes were evaluated. Regression analyses evaluated the relationships between MBL and outcomes.


Sixty-three patients were included with a total of 220 MBL. Quantification of MBL was by high-performance liquid chromatography. Overall, 114 (52 %) MBL were outside the therapeutic range: 78 (36 %) subtherapeutic, and 36 (16 %) supratherapeutic. Repeated-measures analysis found a significant association between supratherapeutic MBL and low cardiac output syndrome (p = 0.02), and supratherapeutic MBL were associated with arterial–central venous oxygen saturation difference >30 % at time 3 (p = 0.007).


In this cohort, nontherapeutic MBL were common. Further investigation of milrinone dosing recommendations may improve the postoperative outcomes of children.


MilrinoneCongenital heart diseaseChildCritical carePharmacology


Milrinone is a drug administered to improve cardio-circulatory function in one-third of critically ill children admitted to Canadian intensive care units [1]. There is limited information about the safety or effectiveness of milrinone. Most information about milrinone pharmacokinetics and pharmacodynamics in pediatrics comes from a single company-sponsored clinical trial [2]. Milrinone’s mechanism of action is related to inhibition of phosphodiesterase (PDE) type III, which leads to increased level of cardiac intracellular cyclic adenosine monophosphate (cAMP) [36]. This increased cAMP activates protein kinases that lead to phosphorylation of functional proteins, increase in extracellular calcium influx, and increased calcium uptake by the sarcoplasmic reticulum [79]. More calcium enters the cell during each cardiac cycle and more calcium is released from the sarcoplasmic reticulum, activating the contractile proteins [7]. This leads to a positive inotropic effect, along with an acceleration of relaxation and shortening of the time of contraction [9]. Inhibition of PDE III within vascular smooth muscles induces accumulation of intracellular cAMP, cyclic guanosine monophosphate (cGMP), with relaxation of systemic and pulmonary vasculature [10].

Milrinone is a first-line therapy after surgery for congenital heart disease (CHD), and a second-line therapy in children with vasoconstricted septic shock [2, 11, 12]. Milrinone inotropic and vasodilator effects are significantly dose related and concentration dependent [13, 14]. A therapeutic range has been identified for milrinone blood levels (MBL) [1519]. However, identification of patients with MBL outside this range is not straightforward. First, clinical evaluation cannot reliably distinguish the signs of milrinone excess (toxicity) or lack of efficacy (underdosing) from underlying critical illness. Second, prediction of levels from pharmacologic formulae may be inaccurate. The occurrence and clinical impact of MBL outside the therapeutic range in critically ill children are unknown. We studied the incidence of MBL outside the therapeutic range in pediatric patients who received milrinone infusions after surgical repair of CHD. As a secondary objective we explored the association of MBL outside the therapeutic range and signs of toxicity and lack of efficacy.

Materials and methods

We conducted a prospective observational study of children aged ≤2 years, having milrinone infusion started after surgical repair for CHD with cardiopulmonary bypass (CPB). Eligible subjects were consecutive pediatric patients admitted to the pediatric intensive care unit (PICU) from November 2007 to March 2009 at the Stollery Children’s Hospital. Patients with any of the following were excluded: no CPB used, prematurity (<36 weeks postconceptual age), weight <2 kg, or mechanical circulatory support while on milrinone infusion. The study was approved by the University of Alberta Health Ethics Research Board. Written consent from each patient’s parent/legal guardian was obtained. The Stollery Children’s Hospital is the CHD surgical referral center for five western provinces and two territories in Canada: Alberta, Saskatchewan, Northwest Territories, British Columbia, Nunavut, Manitoba, and Yukon. Milrinone dosing was at the discretion of the attending physician. In patients receiving a loading dose, this was administered in the operating room.

Four blood samples (0.5 ml/sample) were taken for measurement of MBL in each patient, at the following pragmatic time points after starting the postoperative milrinone infusion: [time 1] 9–12 h, [time 2] 18–24 h, [time 3] 40–48 h, and [time 4] end infusion. Samples in the first 24 h after initiating infusion were intended to ensure data were collected around the nadir of myocardial function at 9–12 h post-CPB [20]. Samples were separated by a minimum of 6 h, to provide opportunity for levels to change (>2 half-lives) [2, 15, 2123]. The samples were drawn at the same time as other tests, coordinated to occur with routine assessments of arterial lactate and arterial and central venous oxygen saturation.

To quantify the MBL a high-performance liquid chromatography (HPLC) assay was used [24]. A HPLC system and an isocratic HPLC technique with a C18 3 micron column, 150 × 4.6 mm, from Varian Inc. (Palo Alto, CA) were used. The daily standard curve was obtained using milrinone, 1,6-dihydro-2-methyl-6-oxo-[3,4′-bipyridine]-5-carbonitrile lactate (PRIMACOR®, Bridgewater, NJ), at 20–1,500 ng/ml. Whole blood was centrifuged on site, and plasma was obtained. Plasma and infusion samples were initially stored at −10 °C for 24–48 h and then at −70 °C until analysis was performed. The sample was vortexed, centrifuged, evaporated, and then reconstituted on mobile phase. Detection was performed with ultraviolet light at 340 nm. The lower limit of detection was 5 ng/ml. Within-day coefficient of variation was <10 % (r2 > 0.99). Previous investigators have used 0.6–4 ml blood [2, 21, 25]. We used 0.5 ml blood per sample to minimize blood extraction.

The MBL were categorized as within the therapeutic range (100–300 ng/ml) or outside the therapeutic range; levels <100 ng/ml were considered subtherapeutic, and levels >300 ng/ml were considered supratherapeutic [15, 17, 19]. We explored the association between subtherapeutic MBL and potential signs of lack of milrinone efficacy, and supratherapeutic MBL and potential signs of milrinone toxicity. An important goal of milrinone administration is to prevent (or mitigate) the low cardiac output syndrome (LCOS) that often develops 9–12 h after cardiac surgical intervention [2]. Therefore, we recorded signs of LCOS as our measure of lack of milrinone efficacy. Given the limited reliability of clinical examination to predict cardiac output, commonly described objective measures were used to define LCOS: arterial lactate level >2 mmol/l and/or arterial–central venous oxygen saturation difference >30 % (AVO2 >30 %) [2, 11, 2632]. Toxicity associated with milrinone is similar to toxicity of other PDE inhibitors [3336]. Seizures, feed intolerance, thrombocytopenia, hypotension, and arrhythmias were considered potential signs of milrinone toxicity [25, 37]. The occurrence of tachycardia and/or hypotension could be signs of LCOS and/or signs of milrinone toxicity [19, 38]. We defined tachycardia as heart rate above the normal limit for age, and hypotension as systolic or mean arterial blood pressure less than the normal lower limit for age (“Appendix”) [39]. Thrombocytopenia was defined as platelet count <50 × 109/l. The inotrope score was calculated by assigning one point for each mcg/kg/min of dopamine or dobutamine, and 10 points for each 0.1 µg/kg/min of epinephrine or norepinephrine [22]. None of the patients received vasopressin and/or phenylephrine, therefore modifications of the inotrope score were not used. Renal dysfunction was defined by pRIFLE criteria using an estimated preoperative CrCl of 100 ml/min/1.73 m2 and the average CrCl measured for the first 48 h [40].

The occurrence of any potential signs of milrinone toxicity and or lack of efficacy was recorded every 2 h for the first 72 h of milrinone infusion. The presence of these signs was recorded as one or more measurements of tachycardia, hypotension, wide pulse pressure, high lactate, AVO2 >30 %, arrhythmia, or LCOS. The platelet count was evaluated daily for the first 3 days of milrinone infusion.

Statistical analysis

Descriptive methods were used to characterize the study population. Summary measurements including proportions, means, median, and interquartile range (IQR) were used as appropriate. Descriptive methods and binomial exact method were used to calculate the proportion of levels outside the therapeutic range and their 95 % confidence intervals (CI). The results are expressed for each time period and as “overall” for all periods combined (time 1 to time 4). The association between MBL categories (subtherapeutic and/or supratherapeutic) and the dependent variables (milrinone toxicity/lack of efficacy) was tested using multivariable logistic regression. The analysis was performed separately for each time period in which MBL were drawn in the first 48 h of milrinone infusion (time 1, 2, and 3).

The following independent variables were included in univariate analysis: age, type of surgical repair (single ventricle versus biventricular), CPB time, Pediatric Risk of Mortality (PRISM) III, MBL category (subtherapeutic, therapeutic, supratherapeutic), and use of nitroprusside. Variables for multiple logistic regression were chosen using purposeful selection methods. Variables significant at p-value <0.1 on univariate analysis, clinically relevant, or potential confounders were considered. MBL category and age were retained in all models. Continuous independent variables were tested for linearity. Variables not meeting the linearity assumption were converted into categorical variables according to their distribution as follows: age <1.5 or ≥1.5 months, PRISM score <7 or ≥7, cardiopulmonary bypass <76 or ≥76 min. Once the final model was obtained, collinearity and interaction effects between the independent variables were tested. The goodness of fit of the final model was assessed using the Hosmer–Lemeshow test.

Repeated-measures analysis of the first three time periods evaluated relationships between MBL and the seven clinical outcomes. MBL were represented in two ways. First, MBL were described as subtherapeutic, therapeutic or supratherapeutic. Second, the values of the measured MBL were modified to represent the absolute difference from the mid-point of the therapeutic range (200 ng/ml). Thus, a MBL of 50 ng/ml was represented as a variation of 150 (=200−50) ng/ml, and a MBL of 500 ng/ml was represented as 300 (=500−200) ng/ml. An exchangeable correlation structure was used for all analyses.

Statistical analysis was performed using STATA (Stata Statistical Software, release 10, 2007; StataCorp LP, College Station, TX), and repeated-measures analyses were performed using SAS 9.1 (SAS Institute, Cary, NC). p-value <0.05 was considered statistically significant.


Between November 2007 and March 2009 all the parents/legal guardians of consecutive children who were eligible for the study were approached for enrollment when a research nurse was available. Of 100 eligible children, 65 were enrolled, and 35 were not enrolled (17 declined consent, and 18 parents/guardians were not available for consent). Sixty-five children were enrolled in the study. Two patients were excluded after enrollment: one child was found to be <2 kg, and one returned from surgery on mechanical circulatory support. In total, 220 MBL were obtained from 63 evaluable patients.

Median age was 3 months and weight 4.5 kg; 31 (49 %) were male; 51 (80 %) had biventricular physiology postoperatively. Seventeen patients (27 %) were neonates, 42 (67 %) infants, and 4 (6 %) 1–2 years of age. Thirteen patients (21 %) were on milrinone infusion preoperatively, and 19 (30 %) received an intraoperative loading dose of median 50 µg/kg (IQR: 25–50 µg/kg). The median (IQR) milrinone dose was 0.5 (0.5–0.75) μg/kg/min. Milrinone doses were adjusted in the first 72 h in 40 (63 %) patients. There were 62 (98 %) survivors to PICU discharge; one child died several days after milrinone infusion was stopped (Table 1). Table 2 shows the different cardiac diagnosis of the cohort.
Table 1

Demographic and clinical characteristics of the study participants


N = 63

Age (months)

3.0 (0–6)

Male sex

31 (49.2 %)

Weight (kg)

4.5 (3.5–6.1)

Biventricular repair

51 (80.9 %)

Milrinone infusion preoperatively

13 (20.6 %)

Cardiopulmonary bypass time (min)

84.0 (67–122)

Milrinone loading dose

19 (30.1 %)

Milrinone dose (µg/kg/min)

0.5 (0.5–0.75)

Milrinone hours

92 (51.5–164.0)

PRISM score

9 (5–11)

Length of mechanical ventilation (days)

5 (3–7)

Length of stay (days)

6 (4–8)

pRIFLE (based on creatinine clearance)a

53 (84.1 %)


9 (16 %)


30 (52 %)


12 (21 %)

Renal replacement therapy

4 (6.3 %)

Inotrope score

4.5 (0–7)


62 (98.4 %)

Values are median (IQR) or n (%)

PRISM Pediatric Risk of Mortality III score on admission to PICU postoperatively

aCreatinine clearance was calculated using 12 h urine collection

Of the 220 MBL, 106 (48 %) were in the therapeutic range (95 % CI 41–54) and 114 (52 %) were outside the therapeutic range (95 % CI 45–58). Seventy-eight (36 %) were subtherapeutic (95 % CI 29–42), and 36 (16 %) were supratherapeutic (95 % CI 11–21) (Table 3). The proportion of MBL outside the therapeutic range at each time period is presented in Table 3. The median (IQR) of MBL from the minimum to the maximum levels within each patient was 82 (46–185) ng/ml, with a median (IQR) ratio (range/mean) of 59 % (36–83 %), suggesting significant within-patient variability in MBL over time.
Table 2

Cardiac diagnoses of study participants


n (%)

Ventricular septal defect

21 (33.3)

Atrial septal defect

2 (3.2)

Tetralogy of Fallot

11 (17.4)

Atrioventricular septal defect

5 (7.9)

Transposition of the great arteries

6 (9.5)

Total anomalous pulmonary venous connection

1 (1.6)

Hypoplastic left heart syndrome

8 (12.7)

Complex single ventricle anatomy

3 (4.7)

Pulmonary atresia

3 (4.7)


1 (1.6)

Heart transplant

1 (1.6)

Mitral valve repair

1 (1.6)

ALCAPA anomalous left coronary artery from the pulmonary artery

Table 3

Proportion of milrinone blood levels outside the therapeutic range by time period

Milrinone dose (mcg/kg/min)

Median (IQR)

Milrinone blood level category


Proportion (95 % CI) (%)

9–12 h (time 1)

 0.5 (0.5–0.75)



28 (17–40)

 0.5 (0.5–0.75)



52 (34–60)

 0.75 (0.75–0.75)



20 (10–31)

20–24 h (time 2)

 0.5 (0.5–0.5)



17 (8–29)

 0.5 (0.5–0.75)



57 (42–70)

 0.75 (0.5–0.75)



26 (15–40)

40–48 h (time 3)

 0.5 (0.5–0.5)



24 (13–38)

 0.5 (0.5–0.75)



60 (34–60)

 0.75 (0.5–0.75)



16 (7–29)

End of infusion (time 4)

 0.25 (0–0.5)



71 (57–82)

 0.25 (0.25–0.5)



25 (14–38)

 0.5 (0.5–0.5)



4 (4–12)

LCOS occurred in 58 (92 %), lactate >2 mmol/l in 48 (76 %), AVO2 >30 % in 38 (60 %), arrhythmia in 13 (20 %), hypotension in 47 (74 %), and tachycardia in 42 (66 %) at any point postoperatively while on milrinone infusion (Table 4). No patient developed feed intolerance and/or seizures, and one patient developed thrombocytopenia.
Table 4

Results of repeated-measures analysis of milrinone blood levels and clinical outcomes


Milrinone level category






Absolute variation from 200 ng/ml


(95 % CI)



(95 % CI)



(95 % CI)



56 (92)

−0.46 (−1.27, 0.34)


−0.91 (−1.77, −0.05)


−0.003 (−0.006, –0.0004)


High lactate

48 (76)

−0.31 (−1.14, 0.51)


−0.40 (−1.21, 0.41)


0.0003 (−0.002, 0.002)


AVO2 >30 %

38 (60)

−0.12 (−0.90, 0.66)


−0.83 (−1.73, 0.06)


−0.003 (−0.006, –0.0004)



13 (20)

−0.07 (−1.06, 0.91)


0.29 (−0.37, 0.98)


0.003 (0.001, 0.005)



47 (74)

0.58 (−0.25, 1.43)


−0.62 (−1.39, 0.13)


−0.0005 (−0.003, 0.002)



42 (66)

−0.06 (−0.88, 0.74)


−0.11 (−1.03, 0.79)


−0.0006 (−0.004, 0.002)


When analyzed by time period, there was a statistically significant association between AVO2 >30 % and supratherapeutic MBL only at time 3, after adjusting for age [odds ratio (OR) 15.7, 95 % CI 2.11–116.90, p = 0.007]. In the remaining analyses by time period we found no significant association between hypotension, tachycardia, high lactate levels, AVO2 >30 %, or LCOS and subtherapeutic or supratherapeutic MBL. There was no statistically significant association between milrinone infusion started preoperatively or loading dose, and the MBL or outcomes. Correlations between milrinone infusion dose and MBL were low (<0.3), and hence not clinically significant.

In repeated-measures analysis we explored the association of clinical outcomes with MBL category (Table 4). There was a statistically significant association between supratherapeutic MBL and LCOS [regression coefficient (β-coeff.): −0.91, 95 % CI −1.77 to −0.05, p = 0.03]. We also explored the association of clinical outcomes with the absolute variation of MBL from the mid-point of 200 ng/ml (variation = 200 – MBL), and found a statistically significant association with LCOS (β-coeff. −0.003, 95 % CI –0.006 to −0.0004, p = 0.02), AVO2 >30 % (β-coeff: −0.003, 95 % CI −0.006 to −0.0004, p = 0.02), and arrhythmia (β-coeff: 0.003, 95 % CI 0.001 to 0.005, p = 0.006).


We performed a prospective observational study of milrinone used for clinical indications in children after surgery for CHD. The main findings of our study are the following. First, despite using milrinone infusions at the recommended doses [2], MBL outside the therapeutic range were common in our cohort of children. At each of four time periods, almost 50 % of MBL were outside the therapeutic range. Second, subtherapeutic MBL were common, 17–71 % of MBL. The high proportion of subtherapeutic MBL at time 4 (71 %) probably reflects common practice in our pediatric intensive care unit where milrinone dose is gradually “weaned” when the patient’s hemodynamic status has improved. Third, although milrinone doses did not exceed 0.75 µg/kg/min, 26 % of MBL were supratherapeutic at 24 h after starting the infusion. After adjusting for age, a significant association between supratherapeutic MBL at time 3 and AVO2 >30 % was found (p value 0.007). Fourth, repeated-measures analysis showed a significant association between LCOS and supratherapeutic MBL (p value 0.02). Fifth, despite relatively constant dosing we found significant within-patient MBL variability; variation from a level of 200 ng/ml was associated with LCOS, AVO2 >30 %, and arrhythmias.

We found no significant association between subtherapeutic MBL and signs of lack of milrinone efficacy. Nevertheless, on repeated-measures analysis, supratherapeutic MBL and absolute variation of MBL from midpoint were both associated with clinical signs of LCOS. The absence of a strong association between MBL outside the therapeutic range and signs of lack of efficacy and/or toxicity should be interpreted with caution. This could be due to the lack of association between the clinical outcomes representing efficacy and toxicity and the MBL category, or due to the lack of statistical power to detect these associations. Moreover, as dose escalation and reduction was not strictly protocolized, subtherapeutic MBL may have been encountered as each patient’s intrinsic cardiac function was improving and milrinone was being titrated off. Otherwise stated, the dose of milrinone used may have been confounded by the indication of its use. Also, other co-interventions or unmeasured confounders could have masked an association between MBL category and the clinical variables. Clinical outcomes were secondary outcomes, and the sample size was not calculated to answer these exploratory hypotheses. In addition, the low incidence of some of these outcomes could have contributed to the results, as could the fact that we explored these associations only in the first 3 days of milrinone infusion. None of the patients had seizures or feeding intolerance, and only one developed thrombocytopenia. A significant number of patients had no episodes of the other secondary outcomes, supporting the rationale for dichotomizing outcomes.

This study has major strengths. First, this is a prospective observational study that reflects common practice in the use of milrinone infusions in a major pediatric cardiac surgical program. Previous studies of milrinone pharmacokinetics have been performed in the context of protocolized milrinone dosing [2, 19, 24, 25]. These studies do not necessarily represent how milrinone has been routinely administered in intensive care units [1]. Second, this investigator-initiated study was conducted independently of a pharmaceutical company. Third, the study includes a reasonable sample of 63 children with a variety of diagnoses and ages representative of standard practice and thus applicable to other similar pediatric intensive care units. The PRIMACORP trial, which obtained similar MBL, did not include children with single ventricle physiology [2]. Fourth, we evaluated the clinical relevance of MBL outside the therapeutic range.

This study has some limitations. First, the study was conducted in a single center. Consequently, the results reflect local practices that may differ in other centers. Vogt et al. reported a range of milrinone doses being used in European units similar to ours [12]. However, milrinone loading doses were administered to 30 % of our patients compared with 40–100 % in other reports [1, 14, 18, 23]. This finding may explain subtherapeutic MBL at time 1, but does not explain MBL in the supratherapeutic range, nor subtherapeutic MBL at times 2 and 3. Second, we evaluated patients ≤2 years old. Extrapolation of our results to other populations should be made with caution. Third, the absolute number of patients in each MBL category was modest, suggesting the study had limited power to evaluate associations between MBL and clinical toxicity and efficacy. Fourth, clinical and laboratory evaluation of these patients cannot reliably distinguish between LCOS due to MBL outside the therapeutic range and LCOS due to their underlying condition. Fifth, the first 72 postoperative hours is a period of continuous change, associated with multiple interventions to support the circulation. Sicker patients receive milrinone for longer periods of time and are likely to have received more co-interventions to support the circulation. The impact of MBL is thus likely to be confounded by these multiple other factors. We did not determine whether milrinone might lessen severity of LCOS even if it did not prevent LCOS. Sixth, observational studies cannot conclude cause and effect, due to potential confounding. In this study, confounding by severity of illness, renal function, and other unmeasured variables may account for the association of MBL and LCOS. An interventional design would be required to better evaluate the clinical effect of MBL on patient outcomes.

These results raise the concern that dosing of milrinone infusions and/or prediction of MBL should be improved to increase the proportion of therapeutic MBL. Milrinone is administered with the goal of preventing LCOS in children having surgical repair of CHD, which usually presents 12–24 h postoperatively [2]. Although milrinone is intended to support the heart function of these children, almost 30 % of MBL are subtherapeutic and likely not significantly impacting cardiac output; a further 20 % of MBL are supratherapeutic and potentially toxic.


This prospective cohort study found that the use of milrinone may be suboptimal. Future protocolized studies should further investigate the potential clinical implications of nontherapeutic MBL to better inform milrinone use. If MBL cannot be predicted, developing real-time monitoring may help direct therapy.


Financial support was provided by the Department of Pediatrics (University of Alberta) and Pediatric Critical Care Associates.

Conflicts of interest


Copyright information

© Springer-Verlag Berlin Heidelberg and ESICM 2013