The Effect of Insulin Infusion Upon Protein Metabolism in Neonates on Extracorporeal Life Support

Abstract

Objective:

Critically ill neonates on extracorporeal life support (ECLS) demonstrate elevated rates of protein breakdown that, in turn, are associated with increased morbidity and mortality. This study sought to determine if the administration of the anabolic hormone insulin improved net protein balance in neonates on ECLS.

Methods:

Twelve parenterally fed neonates, on ECLS, were enrolled in a randomized, prospective, crossover trial. Subjects were administered a hyperinsulinemic euglycemic clamp and a control saline infusion. Protein metabolism was quantified using ring-D5-phenylyalanine and ring-D2-tyrosine stable isotopic infusions. Statistical comparisons were made by paired sample t tests (significance at P < 0.05).

Results:

Serum insulin concentration increased 20-fold during insulin infusion compared with saline infusion control (P < 0.0001). Protein breakdown was significantly decreased during insulin infusion compared with controls (7.98 ± 1.82 vs. 6.89 ± 1.03 g/kg per day; P < 0.05). Serum amino acid concentrations were significantly decreased by insulin infusion (28,450 ± 9270 vs. 20,830 ± 8110 μmol/L; P < 0.02). Insulin administration tended to decrease protein synthesis (9.58 ± 2.10 g/kg per day vs. 8.60 ± 1.20; P = 0.05). For the whole cohort, insulin only slightly improved net protein balance (protein synthesis minus protein breakdown) (1.60 ± 0.80 vs. 1.71 ± 0.89 g/kg per day; P = 0.08). In neonates receiving ≥2 g/kg per day of dietary amino acids insulin significantly improved net protein balance (2.17 ± 0.34 vs. 2.40 ± 0.26 g/kg per day; P < 0.01).

Conclusions:

Insulin effectively decreases protein breakdown in critically ill neonates on ECLS. However, this is associated with a significant reduction in plasma amino acids and a trend toward decreased protein synthesis. Insulin administration significantly improves net protein balance only in those ECLS neonates in whom adequate dietary protein is provided.

A hyperinsulinemic euglycemic clamp was administered in a randomized, prospective, crossover, clinical trial to reduce protein catabolism in critically ill infants on ECLS. In those patients with adequate parenteral amino acid supplementation, protein balance was significantly improved by the administration of insulin. Net protein balance was measured by stable isotopic infusion.

Critically ill surgical neonates, requiring extracorporeal life support (ECLS), exhibit extremely high rates of protein catabolism even when compared with other hospitalized infants.^1–5 This protein loss is the hallmark of the metabolic stress response, and its extent is largely governed by the severity of illness.^6–11 If net protein loss persists in hospitalized patients, it is associated with increased morbidity and mortality.^12,13 Limiting protein degradation and maximizing protein accretion are thus important; however, it is of even more concern in neonates and infants because of their restricted protein reserves^14–16 and their requirement for growth and development.

Stable isotopic studies undertaken in critically ill neonates on ECLS demonstrate a net protein balance of −2.3 ± 0.8 g/kg per day^17,18 (mean ± SD) while normal full term neonates manifest protein accrual.^19,20 Although both protein breakdown and protein synthesis increase in neonates on ECLS, it is the former that predominates.¹ Thus, the search for novel means to optimize protein balance, primarily through a reduction in protein breakdown, is of metabolic interest and clinical relevance in ECLS patients.

Various anabolic agents have been used in clinical studies in critically ill adults to promote net protein accretion during periods of metabolic stress, including growth hormone (GH),^21,22 insulin-like growth factor-1 (IGF-I),²³ and testosterone analogs.^24,25 All have shown, at best, modest success, and some have proven dangerous.

Insulin is a potent anabolic hormone that has been used experimentally and therapeutically in humans for over 80 years,²⁶ and its actions are well studied. The administration of insulin, utilizing a hyperinsulinemic euglycemic clamp,²⁷ has been shown to decrease protein breakdown in normal adults, burned patients, and preterm neonates.^9,28–39 Further, insulin infusion, with a goal of maintaining euglycemia, is associated with a significant mortality reduction in critically ill adults.⁴⁰

Using stable isotopic techniques and a hyperinsulinemic euglycemic clamp, this study sought to determine if the administration of insulin improved net protein balance in neonates and infants on ECLS. Given the direct effects of IGF-I and GH on protein balance,^41–43 as well as the metabolic effects of cortisol and thyroid hormone concentrations in critically ill patients,^44–46 these hormones were also followed to detect any effect of insulin on their concentrations. Insulin-like growth factor binding protein 1 (IGF-BP1), secreted constitutively by the liver, is elevated in hepatic insulin resistance and is suppressed with adequate insulin effect in the liver; hence, it was measured here as a surrogate marker for hepatic insulin resistance.

MATERIALS AND METHODS

Subjects

Infants were recruited from the 2 centers in New England that provide ECLS support: Children's Hospital Boston and the Massachusetts General Hospital. The Institutional Review Boards at both centers approved the study design. Subjects were eligible for recruitment if they were cannulated on ECLS at less than 1 year of age and were considered hemodynamically stable on ECLS by the treating attending physician. Exclusion criteria included: significant intracranial hemorrhage, a known fatal chromosomal anomaly, irreversible cardiopulmonary disease, renal insufficiency (serum creatinine concentration >1.5 mg/dL), hepatic failure (prothrombin time greater than 20 seconds before cannulation), and concomitant treatment with insulin, growth hormone, or thyroid hormone.

After a child met criteria for recruitment, the parents were approached initially by a treating clinician, and then if permission was granted, by a study physician. Informed consent was obtained by a study physician.

Study Design

The study protocol was performed on 2 consecutive days. Upon enrollment, subjects were selected randomly as to the order in which they received insulin or control infusions. Randomization was performed by a computer-generated algorithm using an allocation ratio of 1:1 at each study center in a permuted blocks design with sealed envelopes. In this way, subjects underwent either the hyperinsulinemic euglycemic clamp on day 1 of the study followed by saline (control) infusion on day 2 or the converse (Fig. 1). Each subject therefore served as his or her own control during the study period. Randomization was incorporated to minimize any potential systematic difference in critical illness between days of the study. An identical stable isotope infusion protocol was used on both days of the study.

FIGURE 1. The schematic outline shows a prospective, crossover trial with a randomization of insulin infusion followed by saline control or the converse. Identical stable isotopic infusions are performed on the insulin day and the saline control day.

Hyperinsulinemic Euglycemic Clamp

The hyperinsulinemic euglycemic clamp was performed according to the published methods of Defronzo et al²⁷ with modifications made for neonatal application. Ten minutes before the start of the stable isotope protocol, a 50% glucose infusion was initiated at a rate of 10 mg/kg per minute. This glucose infusion was a supplement to the glucose in the parenteral nutrition (PN) and served to maintain euglycemia during insulin administration. PN was continued throughout the study. Macronutrient nutritional targets were communicated to the treating physicians with a desired protein goal of 2–3 g/kg per day and a caloric allotment of 90 kcal/kg per day.

The insulin solution was prepared by adding 0.75 mL of the patient's blood to the insulin mixture to prevent binding of insulin to the syringe and tubing.⁴⁷ The insulin pool was then primed with a bolus dose of U-100 regular insulin (Humulin, Lilly, Indianapolis, IN) followed by a constant infusion of 120 mU/m ² per minute insulin (approximately equal to 0.6 U/kg per hour and 10 mU/kg per minute). Plasma glucose levels were monitored at the bedside every 5 to 10 minutes throughout the hyperinsulinemic euglycemic clamp using arterial blood and the glucose dehydrogenase reaction in duplicate (HemoCue B-Glucose Analyzers, HemoCue AB, Angelholm, Sweden). A normal blood glucose level of 100 mg/dL was targeted by varying the rate of the glucose infusion. Serum potassium and arterial blood gas measurements were monitored every 30 to 60 minutes throughout the insulin infusion. After the 4-hour stable isotope protocol, the insulin infusion was discontinued immediately, and the supplemental glucose solution was weaned off over the subsequent several hours. The relationship between insulin infusion and the stable isotope administration is outlined in Figure 2.

FIGURE 2. These data show the hyperinsulinemic euglycemic clamp and isotopic infusion for one of the subjects studied. Insulin infusion and concomitant glucose administration result in a plateau in glucose concentrations by 2 hours. The primed stable isotope and insulin infusion continue for 4 hours. Isotopic steady state coincides with the fourth hour of the study.

On the alternate day of the study, as determined by prior randomization, an intravenous infusion of normal saline was initiated along with the stable isotope protocol utilizing the same infusion rate as calculated above for insulin administration.

A 2.0-mL blood sample was obtained at the end of the hyperinsulinemic euglycemic clamp and saline infusion to measure serum concentrations of insulin, insulin-like growth factor binding protein 1 (IGFBP-1), cortisol, thyroxine (T₄), triiodothyronine (T₃), insulin-like growth factor I (IGF-I), leptin, lactic acid, and C-reactive protein (CRP).

Stable Isotope Infusion Protocol

Identical stable isotope infusion protocols were carried out on each of 2 consecutive days. A 1.0-mL baseline blood sample was obtained to document the baseline enrichments of phenylalanine and tyrosine. The patients were then administered an intravenous infusion, over 4 hours, of l-[²H₅]phenylalanine, with a prime of 6 μmol/kg and a continuous rate of 4 μmol/kg per hour (²H₅-isotopic enrichment, 98%; Cambridge Isotope Laboratory, Andover, MA), as well as a 4-hour tyrosine infusion consisting of a prime of 0.5 μmol/kg l-[²H₄]tyrosine and continuous delivery of l-[²H₂]tyrosine at 1 μmol/kg per hour (²H₄ and ²H₂ isotopic enrichments, 98%; Cambridge Isotope Laboratory).

After 200 minutes of l-[²H₅]phenylalanine, and l-[²H₂]tyrosine infusions, 3 blood samples were obtained at 20-minute intervals to determine isotopic enrichments of ²H₅-phenylalanine, ²H₄-tyrosine, and ²H₂-tyrosine. A calibrated syringe pump with an in-line 0.22-μm filter (Medex, Inc., model 2010, Duluth, GA) was used for the studies.

All blood samples were obtained from a previously placed arterial catheter and transferred into prechilled tubes. They were immediately centrifuged at 3000 rpm for 10 minutes, and the plasma stored at −80°C until analyzed. Syringes containing isotope infusate were weighed before and after infusion to precisely quantify the total volume delivered. All stable isotope infusates were prepared using aseptic technique and tested for pyrogenicity and sterility prior to administration. Two additional 0.5-mL blood samples were obtained immediately before and at the end of the stable isotope infusions, on both insulin and saline days, to determine serum amino acid concentrations.

Sample Analyses

Concentrations of the tyrosine infusates, phenylalanine infusates, and serum amino acids were measured using ion exchange chromatography on a Hitachi 8800 high-pressure amino acid analyzer (Hitachi, Tokyo, Japan). Plasma concentrations of insulin, IGF-I, cortisol, T₄, and T₃ were measured by chemiluminometric immunoassay. Plasma concentrations of IGFBP-1 and leptin were measured by radioimmunoassay. Plasma concentrations of CRP were measured using the immuno-turbidimetric method (917 Chemical Analyzer, Roche Diagnostics). Plasma lactate concentrations were measured using enzymatic calorimetry with lactate oxidase and 4-aminoantipyrine (Roche Diagnostics). Blood gas analyses were performed on the Bayer Rapidlab 860 analyzer (Bayer, Tarrytown, NY).

The stable isotopic models and analyses used in this study have been previously described in detail.2,18,48,49 The measurement of whole body protein turnover using the l-[²H₅]phenylalanine technique⁴⁹ was modified by the addition of an l-[²H₄]tyrosine prime so that isotopic plateau would be achieved at 200 minutes of the infusion.

Statistical Analyses

Subject characteristics, metabolic data, and clinical laboratory results are expressed as mean ± SD (range). Graphical representations of primary outcome data are presented as bars, where the bar represents the mean and the error bars indicates the standard deviation. Comparisons of protein kinetic and laboratory data between insulin and control days of the study were performed using paired t tests. Subgroups were assessed by 2-sample t test. The study was designed with a power of 0.8. Statistical significance was set at P < 0.05.

RESULTS

Twelve infants (7 girls and 5 boys) on ECLS were studied (Table 1). There were no significant differences noted between the patients randomized to receive insulin on the first day of the study as compared with the group receiving insulin on the second day of the study. The primary diagnoses for the cohort were congenital diaphragmatic hernia (n = 4), persistent pulmonary hypertension of the newborn (n = 2), meconium aspiration syndrome (n = 1), and infants with congenital cardiac anomalies (n = 5). No subjects were exposed to maternal corticosteroids before delivery. Eleven of the patients were placed on venoarterial ECLS and one was treated with venovenous ECLS. No study was conducted within 48 hours of cannulation or surgical intervention. There were also no significant differences in the nutritional support for the children on the insulin day compared with the saline day (Table 2). However, 6 of 12 subjects received less than the minimum goal of 2 g/kg per day of protein supplementation via PN. No subjects received enteral nutrition during the study period.

TABLE 1. Patient Characteristics (n = 12)

TABLE 2. Nutritional Intake From PN (n = 12)

At steady state, the hyperinsulinemic euglycemic clamp increased serum insulin levels almost 20-fold as compared with saline infusion (447 ± 116 vs. 23 ± 23 μU/mL; P < 0.0001) (Fig. 3). Glycemic control was maintained throughout the insulin and saline infusions (105 ± 26 mg/dL vs.106 ± 9 mg/dL; P = 0.94). Mild hypokalemia was noted during the insulin infusion (3.0 ± 0.3 mmol/L). The insulin infusion resulted in a very slight acidemia when compared with the saline control (pH 7.39 ± 0.05 vs. pH 7.44 ± 0.05; P < 0.05). Serum concentrations of lactic acid were increased by insulin infusion as compared with saline infusion (1.56 ± 1.14 vs. 0.83 ± 0.40 mmol/L; P < 0.05).

FIGURE 3. The average insulin concentrations were 23 ± 23 μU/mL during the saline (control) infusion and 447 ± 116 μU/mL during the insulin infusion (P < 0.0001).

The 4-hour primed continuous infusion of l-[²H₅]phenylalanine achieved isotopic steady state for the last hour of the study. The coefficient of variation of L-[²H₅]phenylalanine enrichment at plateau was 6.0%. The average mole % excess (MPE) of l-[²H₅]phenylalanine was 9.1 ± 0.9% (for all 24 infusions). The plateau in isotopic enrichment evident in l-[²H₅]phenylalanine was mirrored by the l-[²H₄]tyrosine pool with a mean MPE of 2.1 ± 0.9% (for all 24 infusions) (Fig. 4).

FIGURE 4. The 4-hour primed continuous infusion of l-[²H₅]phenylalanine achieved isotopic steady state for the last hour of the study. The coefficient of variation of L-[²H₅]phenylalanine enrichment at plateau was 6.0%. The average mole % excess (MPE) of l-[²H₅]phenylalanine was 9.1 ± 0.9% (first panel). The plateau in isotopic enrichment evident in l-[²H₅]phenylalanine was mirrored by the l-[²H₄]tyrosine pool with a mean MPE of 2.1 ± 0.9% (second panel).

Endogenous phenylalanine flux was significantly suppressed by insulin infusion as compared with saline control (93 ± 21 vs. 80 ± 12 μmol/kg per hour; P < 0.05). This translated into a reduction in whole body protein breakdown from 7.98 ± 1.82 to 6.89 ± 1.03 g/kg per day (P < 0.05). Serum amino acid concentrations were significantly decreased by insulin infusion (28,450 ± 9270 vs. 20,830 ± 8110 μmol/L; P < 0.02). This included a reduction in the concentration of all of the essential amino acids. The flux of phenylalanine going to protein synthesis was also decreased by insulin infusion as compared with saline controls (112 ± 25 vs. 100 ± 14 μmol/kg per hour; P < 0.05). Protein synthesis declined in the insulin administration group as compared with the saline control (8.60 ± 1.20 vs. 9.58 ± 2.10 g/kg per day; P = 0.05). Isotopic protein balance (protein synthesis minus protein breakdown), for the cohort as a whole (n = 12), was only slightly increased by insulin infusion as compared with the saline control (1.71 ± 0.89 g/kg per day vs. 1.60 ± 0.80 g/kg per day; P = 0.09). These data are summarized in Figure 5.

FIGURE 5. Calculations based upon the phenylalanine to tyrosine conversion model reveal that insulin infusion significantly decreases protein degradation as compared with saline control (6.89 ± 1.03 vs. 7.98 ± 1.82 g/kg per day; P < 0.05), decreases protein synthesis compared with saline control (8.60 ± 1.20 vs. 9.58 ± 2.10 g/kg per day; P = 0.05), and only marginally improves protein balance compared with saline control (1.71 ± 0.89 g/kg per day vs.1.60 ± 0.80 g/kg per day; P = 0.09). These data represent the whole cohort studied (n = 12).

Protein balance was also analyzed with respect to a preset cutoff of adequacy of amino acid nutrition (2 g/kg per day). In those subjects (n = 6) who received ≥2 g/kg per day of amino acids via PN, insulin infusion significantly improved net protein balance as compared with saline control (2.40 ± 0.26 vs. 2.17 g/kg per day; P < 0.01), a 10% improvement. No significant benefit of insulin infusion was seen in those patients (n = 6) receiving <2 g/kg per day of amino acids as compared with saline control (1.02 ± 0.73 vs. 1.03 ± 0.71 g/kg per day; P = 0.8). The isotopic data for this subgroup analysis are presented in Figure 6.

FIGURE 6. Calculations of net protein balance are based upon the phenylalanine to tyrosine conversion model. Insulin infusion in the subgroup (n = 6) receiving inadequate amino acid parenteral nutrition (<2 g/kg per day) results in an insignificant improvement in net protein balance as compared with saline control (1.02 ± 0.73 vs. 1.03 ± 0.71 g/kg per day; P = 0.8). In contrast, insulin infusion in the subgroup (n = 6) receiving adequate amino acid parenteral nutrition (≥2 g/kg per day) results in a significant improvement in net protein balance as compared with saline control (2.40 ± 0.26 vs. 2.17 ± 0.34 g/kg per day; P < 0.01). It should also be noted that baseline (saline control) net protein balance is 37% greater in the subgroup receiving adequate amino acid parenteral nutrition compared with the inadequate amino acid parenteral nutrition cohort.

The 2 subgroups (amino acid inadequate vs. amino acid adequate) were not significantly different with respect to ECLS support (120 ± 48 vs. 100 ± 13 mL/kg/h; P = 0.39) and CRP concentrations (5.32 ± 6.47 vs. 3.56 ± 3.53 mg/L; P = 0.59). Average amino acid allotments were 1.16 ± 0.67 g/kg per day for the amino acid inadequate group and 2.50 ± 0.40 g/kg per day for the amino acid adequate cohort (P < 0.01).

The amino acid inadequate cohort (n = 6) encompassed 5 cardiac patients and 1 child with persistent pulmonary hypertension. The postoperative cardiac surgical patients were severely fluid restricted at the time of the study, and this may have led to a forced limitation of amino acid delivery. In a multivariate regression model taking into account postoperative cardiac status, however, amino acid intake was the main predictor of insulin effect on protein balance (P = 0.001). The amino acid adequate group (n = 6) consisted of 4 children with congenital diaphragmatic hernia, 1 with persistent pulmonary hypertension, and 1 with meconium aspiration syndrome.

IGF-BP1 was suppressed by 25% during insulin administration (249 ± 501 vs. 329 ± 557 ng/mL; P = 0.01). Plasma concentrations of leptin, IGF-I, cortisol, T₄, and T₃ were statistically unchanged in response to the insulin infusion.

DISCUSSION

This study tested the hypothesis that the administration of insulin improves net protein balance in a cohort of critically ill surgical neonates and infants on ECLS. Patients with cardiorespiratory failure, requiring ECLS, have been noted to have profoundly elevated rates of net protein breakdown^1–5 that are sometimes not amenable to amelioration by nutritional support alone.¹⁸ An anabolic hormonal intervention was thus deemed metabolically and clinically reasonable. Insulin was chosen because of its well-studied metabolic effects, defined safety profile, short half-life, and intravenous administration route. The use of a hyperinsulinemic euglycemic clamp²⁷ permitted the administration of large quantities of insulin without engendering hypoglycemia. In a previous pilot study of ECLS patients, we have demonstrated that insulin administered by means of a hyperinsulinemic euglycemic clamp is associated with a marked decrease in protein breakdown.⁵⁰ However, the methodology employed in the preliminary work did not allow for the measurement of protein synthesis hence insulin's effect upon actual protein balance remained unclear.⁵⁰

The phenylalanine to tyrosine conversion model used in the current study has the major advantage that protein breakdown, protein synthesis, and the net protein balance can be determined without the need for CO₂ collection.^48,49 This is particularly germane to the ECLS setting as exogenous CO₂“sweep gas” is routinely administered to avoid respiratory alkalosis. This can confound accurate determination of endogenous CO₂ production unless the “sweep gas” is turned off, which in turn can result in marked alkalemia.

A prospective, randomized, crossover trial was designed where each subject acted as his or her own control (Fig. 1). Identical stable isotopic infusions were administered on consecutive days under the conditions of a hyperinsulinemic euglycemic clamp and saline infusion control. Patients were randomized to receive either insulin or saline the first day to obviate any systematic bias. The characteristics of the patients and their nutritional support regimens during saline and insulin infusion days are outlined in Tables 1 and 2, respectively. These data reveal a relatively young cohort (12.7 ± 24.5 days), with a mean body weight of 3.3 ± 0.96 kg, which received very similar parenteral nutritional support during the intervention and control phases of the study.

The hyperinsulinemic euglycemic clamp studies were associated with no major untoward effects. Euglycemia was well maintained throughout the insulin and saline infusions. Insulin administration was not associated with any statistically significant alterations in the concentration of leptin, IGF-I, cortisol, T₄, or T₃. IGF-BP1 decreased by 25% during hyperinsulinemia, suggesting that the insulin dose used in this clamp study (120 mU/m² per minute) did overcome any hepatic insulin resistance. IGF-BP1 is produced constitutively in the liver, is increased during periods of insulin resistance such as critical illness,⁵¹ and is suppressed during increased hepatic insulin effect.⁵² As anticipated, approximately half of the subjects had asymptomatic mild hypokalemia during the insulin portion of the study that was corrected with intermittent potassium administration. Insulin infusion as compared with saline control also caused a statistically significant, but clinically quite modest, increase in lactate concentration, and a statistically significant but clinically even less impressive alteration in acid-base balance.

In the cohort of ECLS newborns studied, the hyperinsulinemic euglycemic clamp raised serum insulin concentrations almost 20-fold over saline infusion control levels. This infusion of insulin was associated with a 14% reduction in whole body protein breakdown compared with control. It was also linked with a 27% reduction in plasma amino acid concentration, including a decrease in the concentrations of all of the essential amino acids. Presumably, the decrement in protein breakdown resulted in a diminution of amino acids entering the free amino acid pool. It is important to note that most of amino acids (98%) reside in whole body proteins, not in the free state. Hence, the decline in protein synthesis found under the condition of insulin administration, as compared with the saline infusion, may be explained by the decreased free amino acids available for protein production. Because of the diminution of protein synthesis, isotopic protein balance (protein synthesis minus protein breakdown), for the group as a whole, was only slightly augmented by insulin infusion as compared with the saline control.

A recent study of the short-term application of a hyperinsulinemic euglycemic clamp to the leg of healthy controls demonstrates an analogous reduction in muscle protein synthesis and underscores the potential importance of maintaining amino acid availability during insulin infusion.⁵³ In normal volunteers, insulin induces a significant decrease in whole body proteolysis, as measured by stable isotopic techniques, during a hyperinsulinemic euglycemic clamp.29,32,33,38 Two investigations that did not provide adequate amino acids during the protocol also found a decrease in whole-body protein synthesis, and this seemed to be related directly to an drop in serum amino acid concentrations induced by insulin's reduction of proteolysis.^38,54 In contrast, the maintenance of normal amino acid concentrations during a hyperinsulinemic euglycemic clamp actually appears to stimulate forearm muscle protein synthesis.³⁴

For the purposes of the current study, our aim was to administer parenteral amino acids at 2 to 3 g/kg per day. Although all clinicians were aware of this goal, concerns regarding potential fluid overload precluded the administration of sufficient PN in half the patients studied. Future similar studies of the efficacy of insulin in enhancing protein balance should exclude all subjects receiving less than 2 g/kg per day of dietary protein.

Although insulin effectively decreases protein breakdown in critically ill neonates on ECLS, this causes a significant reduction in plasma amino acids and, hence, an associated decline in protein synthesis. Therefore, insulin infusion improves net protein balance only in those ECLS neonates in whom adequate dietary amino acids are provided. The dose of insulin used in this study is a pharmacologic dose. Prior to being implemented in the clinical setting, a dose-response study is needed to identify the lowest insulin dose at which protein balance is still significantly enhanced. Further validation within the framework of a long-term insulin infusion protocol is also indicated.

ACKNOWLEDGMENTS

The authors thank Mr. Jamin Alexander for his efforts in the preparation of this manuscript.

Discussions

Dr. Kevin P. Lally (Houston, Texas): Dr. Jaksic, I would like to congratulate you and your colleagues on a nicely performed clinical research study. Dr. Jaksic and colleagues have previously demonstrated surprisingly high levels of protein catabolism in infants treated with extracorporeal membrane oxygenation. This study extends that observation and looks at a potential way to blunt this response. I have a couple of questions, though.

While insulin infusion increases net protein balance with appropriate protein intake, a good response is seen with adequate protein alone. Is there a significant enough benefit with insulin to warrant its use?

Secondly, is there a benefit to insulin infusion to maintain a set glucose level as is commonly performed in adult ICUs as opposed to a set insulin level?

Similarly, why was this level of insulin infusion used as opposed to different or lower levels of insulin?

Fourth, the CRP level suggests ongoing metabolic stress. But what happens on days 6 or 8? That is, is protein catabolism ongoing throughout the entire ECMO course?

Lastly, ECMO is a time-limited problem, or is treating a time-limited problem, and while prolonged protein depletion is bad, is 4 to 6 days a significant enough problem?

Dr. Tom Jaksic (Boston, Massachusetts): The administration of insulin, under the conditions of this experiment, did result in a 10% improvement in net protein balance. This may indeed be physiologically relevant. If one is going to use this type of intervention in clinical practice, it will probably be in the realm of hyperglycemic neonates. In our own ICU we found that 80% of surgical neonates are intermittently hyperglycemic. Thus, we do intend to do a trial where we just treat hyperglycemic neonates with insulin, and concomitantly we will take a look at their protein balance, amongst other outcome variables, to see if there is an improvement.

As for the dose of insulin, we chose 0.6 units per kilogram per hour, which is quite a large dose, based on 3 burn studies that have been done. Two of those 3 showed a reduction in protein breakdown, one did not. The one that did not had a low rate of insulin infusion. So we went to a higher rate. There have been 2 pilot studies in neonates, one of them ours, to show the feasibility of this approach. They seemed to indicate that this was safe. So in brief, we took the highest dose that we could within the safety parameters that we were comfortable with.

Catabolism from extracorporeal life support is not really due to the circuit itself. We have studied these patients 3 weeks after they have come off extracorporeal life support and they maintain catabolism although it is considerably ameliorated over the acute phase. So it does persist over time. A child on extracorporeal life support for 14 days may lose approximately 15% of their lean body mass. A loss of 30% of lean body mass it is usually fatal. Thus a relatively short period of inadequate nutrition is potentially deleterious to a patient on extracorporeal life support.

Dr. Murray F. Brennan (New York, New York): My experience with hyperinsulinemic and euglycemic clamps is in adults, so I apologize if my question is inappropriate. I didn't understand how you were able to give that degree of insulin infusion and yet the glucose infusion rate was the same. Can you explain that?

Dr. Tom Jaksic (Boston, Massachusetts): The insulin infusion rate in your adults was the same?

Dr. Murray F. Brennan (New York, New York): No, no. If I gave that degree of insulin infusion then the 2 groups would vary widely in the amount of glucose I would have to give to maintain the same glucose. And you suggested they were the same in both groups, or did I misunderstand?

Dr. Tom Jaksic (Boston, Massachusetts): No, I didn't mean to suggest that. The neonates, when they were given the saline portion of the infusion, maintained euglycemia with the TPN alone. Variable additional amounts of glucose were indeed needed for neonates given high dose insulin.

Dr. Murray F. Brennan (New York, New York): Finally, just 2 insights. With such great emphasis on normoglycemia in the adult patients and critical care patients, people forget that all of those patients, like yours, also receive TPN. So maintaining the blood glucose alone by hyperinsulinemia is not the solution. It is maintaining it, as your paper beautifully showed, in babies with the appropriate nutrient infusion.

Dr. Marshall Z. Schwartz (Philadelphia, Pennsylvania): Dr. Jaksic, congratulations on presenting a very nice study, which clearly was difficult to carry out. I have a few questions.

First, to follow up on Dr. Brennan's question, it is difficult to maintain glucose levels within a tight range using insulin infusion in much larger patients and thus, I was surprised to see how easy you were able to do this in such small patients and maintain a fairly steady state. I wonder if you could elaborate on how you were able to accomplish this?

The second question is, was there a rebound effect with significantly elevated serum glucose levels after you stopped the insulin infusion? The third question is, many of the patients that now are candidates for ECMO are septic, and we know that sepsis causes insulin insensitivity. I do not believe there were any patients in your series that were septic but could you speculate on how this would affect your results.

Dr. Tom Jaksic (Boston, Massachusetts): First with regard to sepsis, we did not study any patients with sepsis. However, septic young adults have been studied and they do tend to decrease their protein breakdown at appropriate levels of insulin.

Dr. Marshall Z. Schwartz (Philadelphia, Pennsylvania): How were you able to maintain the glucose levels so well? Because the experience is that it is not so easy.

Dr. Tom Jaksic (Boston, Massachusetts): We actually measured glucose every 5 to 10 minutes with a HemoCue device. We adjusted our glucose administration based on those results. What we have done subsequently is to customize for ECMO continuous glucose monitoring devices that are commercially available. And that is how we intend to do our insulin intervention trials in the long term. When we stopped the infusion, what happened was there was actually a tendency towards hypoglycemia as the insulin washes out of the circuit, and we maintained vigilance for a period of 4 hours where we gave intermittent small amounts of glucose to maintain euglycemia.

Dr. Dennis P. Lund (Madison, Wisconsin): ECMO is such a complex soup of interventions that we are doing to these patients. There is some data to suggest that these patients are often adrenally insufficient. Did you do anything to control for cortisol or ACTH levels during your study?

Dr. Tom Jaksic (Boston, Massachusetts): We did not give any of our patients cortisol and none of the patients had mothers whom received antenatal cortisol. We did measure their serum cortisol. We found no differences during the insulin day versus the saline day in terms of serum cortisol levels.

Dr. Palmer Q. Bessey (New York, New York): Congratulations, Dr. Jaksic. Good study. I have 2 quick questions.

One has to do with the technique of stable isotopes depends upon appearance and disappearance of the isotope from really the serum compartment. And that assumes, of course, that there is equilibrium with the intracellular pools. I wonder if you have any data on whether or not that really is affected and what the time course of that is.

And extending that a little bit, the question has to do with how quickly after you begin the insulin administration do you see these effects? And do you have any sense of how long they persist? I know it is a difficult thing to do and you may not have done it, but if you did it, if you continued the insulin infusion for a day or 2, would you have a tachyphylaxis effect or would it persist?

Dr. Tom Jaksic (Boston, Massachusetts): With regards to the kinetic model, this is a model that was first described by Clark and Bier in the early 1980s. It has been modified subsequently. And by using a D4 tyrosine prime we can attain isotopic plateau at 200 minutes, which is why we designed the study in this fashion. Isotopic plateau implies that there is at least isotopic equilibrium amongst the various pools. Unfortunately, for phenylalanine there is no intracellular marker.

As for the second portion of the question as to how long the insulin effect persists, the answer is that quite frankly we don't know. We are hopeful that chronic administration of insulin over the long term will have a similar beneficial effect at a much lower dose in hyperglycemic patients.

Dr. Basil A. Pruitt, Jr. (San Antonio, Texas): I do have a question about the metabolic rate. You said it was equal to that of a 60% burn. As is the case with burn patients, can you amplify the beneficial effects of insulin by reducing metabolic rate with propranolol?

Secondly, can you amplify the effect of insulin with any anabolic agent such as oxandrolone or insulin-like growth factor 1 plus insulin-like growth factor binding protein 3?

Dr. Tom Jaksic (Boston, Massachusetts): It is important to realize that neonates are quite different than adults in that they don't manifest an increase in resting energy expenditure with injury. What happens is that they stop growing. Hence, the energy that they normally use for growth is used to fuel the metabolic response to injury. So we don't really want to lower their metabolic rate. Insulin is the only anabolic hormone we have tested in this setting.