OBJECTIVE: The optimal treatment of neonatal hyperglycemia is unclear. The aim of this trial was to determine whether tight glycemic control with insulin improves growth in hyperglycemic preterm infants, without increasing the incidence of hypoglycemia.
METHODS: Randomized, controlled, nonblinded trial of 88 infants born at <30 weeks’ gestation or <1500 g who developed hyperglycemia (2 consecutive blood glucose concentrations (BGC) >8.5 mmol/L, 4 hours apart) and were randomly assigned to tight glycemic control with insulin (target BGC 4–6 mmol/L, “tight” group) or standard practice (restrictive guidelines for starting insulin, target BGC 8–10 mmol/L, “control” group). The primary outcome was linear growth rate to 36 weeks’ postmenstrual age.
RESULTS: Eighty-eight infants were randomly assigned (tight group n = 43; control group n = 45). Infants in the tight group had a lesser lower leg growth rate (P < .05), but greater head circumference growth (P < .0005) and greater weight gain (P < .001) to 36 weeks’ postmenstrual age than control infants. Tight group infants had lower daily BGC (median [interquartile range] 5.7 [4.8–6.7] vs 6.5 [5.1–8.2] mmol/L, P < .001) and greater incidence of hypoglycemia (BGC <2.6 mmol/L) (25/43 vs 12/45, P < .01) than controls. There were no significant differences in nutritional intake, or in the incidences of mortality or morbidity.
CONCLUSIONS: Tight glycemic control with insulin in hyperglycemic preterm infants increases weight gain and head growth, but at the expense of reduced linear growth and increased risk of hypoglycemia. The balance of risks and benefits of insulin treatment in hyperglycemic preterm neonates remains uncertain.
- BGC —
- blood glucose concentrations
- IGF-1 —
- insulin-like growth factor 1
- OFC —
- occipitofrontal head circumference
- PMA —
- postmenstrual age
- SGA —
- small for gestational age
What’s Known on This Subject:
Insulin is commonly used to treat neonatal hyperglycemia, but there are few data to support its use. Tight glycemic control with insulin improves outcome in diabetic patients, but it is not known whether it is effective in hyperglycemic preterm infants.
What This Study Adds:
Tight glycemic control with insulin in hyperglycemic preterm neonates decreases the rate of linear growth despite increased weight and occipitofrontal head circumference gain and increases the risk of hypoglycemia. Insulin may not be a safe and effective treatment in hyperglycemic preterm neonates.
Hyperglycemia is common in preterm infants, with an estimated incidence of >50% in extremely low birth weight infants.1 Although it is associated with increased morbidity (intraventricular hemorrhage1 and retinopathy of prematurity2) and mortality, whether this is a causal relationship or a reflection of illness severity is unknown.3 Thus, it remains uncertain whether hyperglycemia should be treated and, if so, how this is best done.4 Neonatal hyperglycemia may be managed by tolerating high blood glucose concentrations (BGC), by administering insulin, or by decreasing the carbohydrate intake.5 Two randomized trials have compared insulin infusion with decreased glucose intake in hyperglycemic extremely low birth weight infants in the first 2 weeks after birth.6,7 Both found that insulin treatment increased glucose and energy intake, and one reported an increase in short-term weight gain and a reduction in sepsis, but neither reported long-term outcomes.
Tight glycemic control with insulin has been shown to decrease morbidity8,9 and mortality10 in patients with diabetes. However, there is conflicting evidence about whether tight glycemic control in hyperglycemic adults in intensive care is beneficial or harmful, with a recent meta-analysis showing no effect on mortality, but a substantially increased risk of hypoglycemia.11 Intensive insulin therapy in infants and children, mostly after cardiac surgery, decreased mortality, secondary infection, and length of stay in intensive care, but increased the incidence of hypoglycemia.12 Conversely, early elective insulin replacement therapy in very low birth weight infants increased mortality at day 28, although not by the expected date of delivery, and also increased the incidence of hypoglycemia.13 These findings are of concern, because recurrent hypoglycemia has been associated with adverse neurodevelopmental outcome in preterm infants.14
Poor postnatal weight gain and head growth, common in preterm infants,15 also have been associated with adverse neurodevelopmental outcomes in preterm infants.16 Insulin is a potent growth factor in the fetus,17 and tight glycemic control with insulin increases weight gain in diabetic patients.18 The aim of this trial was to determine whether tight glycemic control with insulin improves growth in hyperglycemic preterm infants without increasing the incidence of hypoglycemia.
This was a randomized, controlled, nonblinded study that took place from July 2005 until October 2008 at the level III NICU at National Women’s Health, Auckland, New Zealand. Infants were eligible if they were born at <30 weeks’ gestation or <1500 g birth weight and became hyperglycemic (2 consecutive BGC >8.5 mmol/L at least 4 hours apart). Exclusion criteria were hyperglycemia because of iatrogenic overdose of glucose, major congenital malformation, or judged to be dying. Ethnicity was self-reported by parents, and was assessed because of the possible effects of ethnicity on neonatal hyperglycemia7 and growth.19 Approval was obtained from the Northern X ethics committee, and written informed consent was obtained from a parent of each child. A data safety monitoring committee reviewed all serious adverse events (death, BGC <1.5 mmol/L) and mortality and morbidity data at 6-month intervals.
The randomization schedule was computer-generated independent of the study team or clinicians, with concealed allocation, stratified in random blocks of 4 and 6 by gender and weight for gestational age (small for gestational age [SGA] or not SGA). Infants were allocated to tight glycemic control (“tight” group) or standard management according to the neonatal unit guidelines (“control” group) as soon as possible after they met the criteria for hyperglycemia by clinicians entering eligibility details into the randomization computer. The investigators collecting outcome data were not involved in the clinical care of the infants. Infants completed the trial when they reached 36 weeks’ postmenstrual age (PMA).
Insulin (Actrapid, Novo Nordisk, Bagsværd, Denmark) was given by continuous intravenous infusion, with solutions allowed to preadsorb to the tubing for 1 hour before use. Infants randomly assigned to the tight group were immediately commenced on insulin at 0.05 U/kg per hour and the infusion rate titrated to maintain BGC between 4 and 6 mmol/L. Infants randomly assigned to the control group were only treated with insulin if they met all of the following criteria: BGC >10 mmol/L or persistent glycosuria >2+; tolerating <100 kcal/kg per day; >72 hours of age, and not acutely stressed. If these infants were treated with insulin, the infusion started at 0.05 U/kg per hour, and was titrated to maintain BGC between 8 and 10 mmol/L.
Lower leg length was measured twice weekly by using a neonatal knemometer (Force Technology, Copenhagen, Denmark), a highly accurate and sensitive method with a coefficient of variation of 0.31% (<1 day of growth of the lower leg).20 The first 3 measurements were discarded and then the mean was taken of the subsequent 10 measurements. Each infant was measured by 1 of 2 investigators. Weight was measured as per the unit protocol either on alternate days or twice weekly. Occipitofrontal head circumference (OFC) was measured by tape measure weekly.
Myocardial hypertrophy was assessed within 2 days of enrolment into the study, 2 weeks and 4 weeks after enrolment, and again at 36 weeks’ PMA. The left ventricular posterior wall and interventricular septum thicknesses were assessed in end diastole by M-mode echocardiography at the level of the mitral valve leaflets by using a parasternal short-axis view (HDI 5000, 7.5–11 MHz probe, Philips Healthcare, Best, Netherlands).21
Blood was taken for analysis of plasma cortisol, insulin, and insulin-like growth factor 1 (IGF-1) concentrations on the day of randomization, and for insulin and IGF-1 also at 7 and 14 days after randomization, and at 36 weeks’ PMA. Plasma cortisol and insulin concentrations were measured on an Azsym system autoanalyzer (Abbott Laboratories, Abbott Park, IL), and IGF-1 concentrations by radioimmunoassay.22
BGCs were measured as clinically indicated by using a glucose oxidase method (ABL 700, Radiometer Ltd, Copenhagen, Denmark), and urine was tested for glycosuria by using urine test strips (Multistix reagent strips, Bayer Healthcare, Auckland, New Zealand). All BGCs, episodes of hypoglycemia (BGC <2.6 mmol/L) and glycosuria ≥2+ from randomization until 36 weeks’ PMA were included in the analysis.
Parenteral nutrition was commenced within the first day after birth, and small enteral feeds with breast milk were usually started within the first 2 days. The unit parenteral nutrition guideline was changed in January 2007 for reasons unrelated to the trial, aiming to decrease fluid and increase protein intake. All parenteral and enteral nutrition given from birth until 36 weeks’ PMA was recorded. Nutrient intake from breast milk was calculated by using local data.23
The primary outcome was growth rate from randomization to 36 weeks’ PMA, measured by knemometry. The secondary outcomes were sepsis, myocardial hypertrophy, plasma insulin, IGF-1 and cortisol concentrations, and the incidence of hypoglycemia. SGA was defined as a birth weight less than the 10th percentile (z score < −1.28). Sepsis was defined as a positive culture (bacterial or fungal) from a sterile space (blood, cerebrospinal fluid, urine [catheter or suprapubic specimen only]) which required antibiotic or antifungal treatment of ≥5 days. Chronic lung disease was defined as requiring oxygen or respiratory support at 36 weeks’ PMA. Patent ductus arteriosus was defined as a ductus requiring treatment with indomethacin or ligation. Full enteral feeding was defined as the first day without parenteral feeding on which the volume of enteral feeds was ≥150 mL/kg per day.
Based on previous data on lower leg growth from our neonatal unit,24 we estimated we needed 44 infants in each group to detect a 5% difference between groups in growth rate from birth to 36 weeks’ PMA, with 80% power and a 5% level of significance (2-tailed).
Parametric data were compared between groups with paired and unpaired Student t test as appropriate, and are presented as mean (SEM). Nonparametric data were log transformed if possible or analyzed by the use of the Mann-Whitney U test and are presented as median (interquartile range). Plasma insulin and IGF-1 concentrations were analyzed by factorial analysis of variance, to include all data including those from infants who died. Categorical data were analyzed by the χ2 test.
Mortality, morbidity, and growth data were adjusted for birth weight by logistic regression. Weight and anthropometric measurements were expressed as z scores.25 Nutritional data were analyzed by repeated-measures analysis of variance up to 5 weeks after birth. Knemometry, weight gain, and OFC data were analyzed by multivariate linear regression. Variables forced to enter into a stepwise regression were: days since insulin was first started; whether intravenous insulin had been administered since the last measurement; randomization group; day after randomization (day after birth for weight analysis); randomization group × day after randomization interaction; before/after parenteral nutrition protocol change, and before/after parenteral nutrition protocol change × day after randomization interaction. Variables entered with an f value to enter of 0.25 were: gender; gestation; birth weight z score; ethnicity; calorie intake; prerandomization sepsis; multiple birth, and postnatal steroids. Variables forced to enter and those which met the predetermined f value were then entered into the standard least squares regression model. An overall analysis was performed initially; if there was a significant difference between the groups, then the data were analyzed week by week or fortnightly by using the same linear regression model, and a P value of <.01 or <.005 was taken as significant to account for repeated testing over 5 or 10 periods, respectively.
Data were analyzed by using JMP (SAS Institute Inc, Cary, NC) on an intention-to-treat basis.
Eighty-eight infants were randomly assigned (Fig 1). Parents of 8 infants, all in the tight group, withdrew them from the trial, most because of the need for intravenous access and frequent heel-prick blood sampling. However, all outcome data were collected for 4 of these. Baseline demographics in the 2 groups were similar (Table 1); however, infants in the tight group were, on average, 100 g lighter at birth and had a smaller OFC than infants in the control group.
All but 1 infant in the tight group (1 protocol violation), and two-thirds of infants in the control group, were treated with insulin (Table 2). There was no difference in myocardial thickness between the groups (Fig 2). The lower-leg growth rate from randomization to 36 weeks’ PMA was less, but rate of weight gain and OFC growth were greater, in the tight group (Fig 3). Thus, by 6 weeks of age, adjusted leg length was 8% less and adjusted weight 4% greater in infants randomly assigned to the tight group in comparison with controls. At 36 weeks’ PMA there were no differences between groups in weight, length, or OFC (Table 3). However, the decrease in z score for OFC from birth to 36 weeks’ PMA was 0.5 SD less in the tight group.
Infants in the tight group received insulin for almost 4 times longer than infants in the control group, and received a higher total and mean daily dose of insulin. The daily maximum, mean and minimum BGCs were all lower in the tight group (overall mean difference −0.7 mmol/L [95% confidence interval −0.6 to −0.9 mmol/L], difference while on insulin −2.4 mmol/L [−2.1 to −2.7 mmol/L]). Infants in the tight group also had fewer days with glycosuria ≥2+ (Table 2).
Twice as many infants in the tight group as in the control group experienced at least 1 episode of hypoglycemia, and more also experienced recurrent hypoglycemia (Table 3). There were no significant differences between groups in mortality or neonatal morbidity. There was no difference between groups in nutritional intake (Fig 4), or the time to full enteral feeds (Table 3).
Plasma IGF-1 concentrations increased over time (P < .0001), and plasma insulin concentrations decreased over time (P < .0001), but there were no differences in plasma insulin or IGF-1 concentrations between groups (Table 2).
This is the first reported randomized trial of tight glycemic control with insulin in hyperglycemic preterm neonates. Intravenous insulin therapy is commonly used in hyperglycemic preterm infants,26 with some units targeting tight glycemic control, but there are few data to support its use or to indicate the optimal glycemic range. We found that tight glycemic control is achievable with pragmatic bedside titration of the insulin dose. However, this resulted in an increased risk of hypoglycemia, an effect also seen in other trials of tight glycemic control.27 Moreover, although there was greater weight gain and OFC growth, there was reduced linear growth, suggesting an increase in fat mass rather than lean mass.
Previous randomized trials have shown increased short-term growth in preterm infants treated with insulin.6,28 However, the comparison group in these trials had decreased carbohydrate intake, which likely resulted in reduced growth rate. In the current trial, where nutritional intakes were similar in the 2 groups, hyperglycemic infants treated with insulin to maintain tight glycemic control had a greater rate of weight gain. This is unlikely to be due to the direct effects of insulin, because plasma insulin concentrations were similar in both groups. However, the tight group also experienced less glycosuria and may thus have retained a greater proportion of the available carbohydrate. Glycosuria may result in the potential risk of osmotic diuresis, but a preterm infant may also lose significant amounts of energy (2.9–5.8 kcal/kg per day) from glycosuria of only 2+. This potentially may have contributed to poorer weight gain and OFC growth in the control group.
It is uncertain why the infants in the tight group had slower linear growth. It is possible that the greater weight gain in these infants was primarily due to an increase in fat mass rather than muscle mass or bone growth. This is of concern, because preterm infants have a higher proportion of intra-abdominal adipose tissue at term-corrected age,29 and insulin resistance in childhood30 and young adulthood.31 Our data suggest that insulin treatment to maintain tight glycemic control in preterm infants may further aggravate their cardiovascular risk in later life.
Offspring of pregnancies complicated by diabetes have been shown to have both interventricular and ventricular wall hypertrophy on echocardiograms performed in utero32 and in the neonatal period.33 It is probable that the cardiac hypertrophy is related to hyperglycemia and hyperinsulinemia in the fetus, because the severity of the hypertrophy has been related to the degree of glycemic control in pregnancy.33 However, it is not known if hyperglycemia in preterm infants causes ventricular hypertrophy, or if treatment with insulin in these infants would alter this. We found that tight glycemic control with insulin did not change the interventricular septal or ventricular wall thickness. Although our data were similar to reference data from preterm infants at 28 days of age34 without a euglycemic control group, it is difficult to speculate on the effect of neonatal hyperglycemia itself on these parameters.
A reported association between hyperglycemia in the first 24 hours and reduced cerebral white matter volume on MRI at term suggests that preterm infants with neonatal hyperglycemia may be at increased risk of neurodevelopmental delay.35 It is encouraging that infants in the tight group had greater OFC growth. While it is possible that other tissues such as subcutaneous fat may have contributed to the OFC, there is a strong association between OFC and brain volume, 36,37 and greater OFC growth has been associated with an improvement in neurodevelopment.37 Nevertheless, any benefit in head growth from tight glycemic control would need to outweigh the potential increased risk of neurodevelopmental disability from increased hypoglycemia.14 Furthermore, neonatal hypoglycemia induced by insulin may be more detrimental than hypoglycemia secondary to low substrate availability, because insulin also affects fat and protein metabolism, preventing production of alternative cerebral fuels.38 Previous studies of insulin treatment in hyperglycemic preterm infants have not shown a marked increase in hypoglycemia.6,7 However, an increased rate of hypoglycemia with insulin use has been reported in studies of tight glycemic control in patients with diabetes39 and patients in intensive care.12,27
Stochastic modeling of BGC has been shown to be effective in determining the insulin infusion rates and may decrease the incidence of hypoglycemia.40 However, this approach requires frequent BGC measurements, and may not be practical for prolonged insulin use in preterm infants. Indeed, this was a major reason for parental withdrawal of infants from our study. Continuous subcutaneous glucose monitors have been shown to be safe and effective in preterm infants,41 and may assist in the safer provision of tight glycemic control, with less hypoglycemia, in the future.
A previous randomized controlled trial of elective insulin therapy in the first week after birth in preterm infants who were not hyperglycemic at enrolment resulted in an increased incidence of hypoglycemia and in increased mortality at 28 days of age, despite the decreased incidence of hyperglycemia.13 That trial found no evidence of benefit with insulin replacement therapy. The current trial, which only randomly assigned hyperglycemic preterm infants, also did not find a beneficial effect of insulin therapy. The results of these trials emphasize the lack of evidence for the efficacy and safety of intravenous insulin in preterm infants.
It was not feasible to blind the intervention in this study, because clinical staff needed to titrate the insulin dose to achieve the targeted BGC range. However, it is unlikely that this has substantially influenced our results, because the outcome data were collected by investigators not involved in the clinical care of the infants. The increased number of BGC measurements in the tight group may have resulted in an ascertainment bias for hypoglycemia. However, infants in the control group received insulin for fewer hours and, although they had more BGC measurements per hour of treatment than infants in the tight group, they therefore had fewer BGC measurements in total. Hypoglycemia is unusual beyond the first few days of life in preterm infants who are on stable nutrition and not receiving insulin. Thus, it is unlikely that episodes of hypoglycemia were undetected in control infants.
Insulin therapy has become widespread in neonatal nurseries26 with very few data to support its use4 and side effects that are potentially harmful. This trial found that maintaining tight glycemic control with insulin in hyperglycemic preterm infants resulted in a greater weight gain and OFC growth, but less linear growth, suggesting an increase in fat rather than lean mass. Moreover, this treatment also increased the incidence of potentially harmful hypoglycemia. Our data suggest that insulin infusion may not be a safe and effective treatment in hyperglycemic preterm neonates.
We acknowledge the members of the Data Safety Monitoring Committee: Dr David Knight (Mater Mothers' Hospital) and Dr Philip Weston (Waikato Hospital). We thank also Coila Bevan, Tineke Crawford, Dr Carl Kuschel, and Christine Keven for valuable contributions to this study.
- Accepted December 13, 2011.
- Address correspondence to Jane M. Alsweiler, MBChB, FRACP, PhD, Newborn Services, Level 9, Support Building, Auckland City Hospital, Private Bag 92024, Auckland Mail Centre, Auckland 1142, New Zealand. E-mail:
This trial has been registered with the Australian Clinical Trials Registry (identifier 12606000270516).
Drs Alsweiler, Harding, and Bloomfield conceived and designed the study; Drs Alsweiler and Bloomfield collected and compiled the data; Drs Alsweiler, Harding, and Bloomfield analyzed the data; Dr Alsweiler wrote the initial draft of the manuscript; and Drs Alsweiler, Harding, and Bloomfield critically revised the paper and contributed to discussion.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: This trial was funded by grants from the Auckland Medical Research Foundation, the Maurice and Phyllis Paykel Trust, and the Starship Foundation.
- Hays SP,
- Smith EOB,
- Sunehag AL
- Bottino M,
- Cowett RM,
- Sinclair JC
- Griesdale DEG,
- de Souza RJ,
- van Dam RM,
- et al
- Lucas A,
- Morley R,
- Cole TJ
- Cole T,
- Statnikov Y,
- Santhakumaran S,
- Pan H,
- Modi N
- Ehrenkranz RA,
- Dusick AM,
- Vohr BR,
- Wright LL,
- Wrage LA,
- Poole WK
- The Diabetes Control and Complications Trial Research Group
- Gibson AT,
- Pearse RG,
- Wales JK
- Mader J
- ↵FoodWorks Professional Edition [computer program]. Version 3.02. Auckland, New Zealand: Build 540, Xyris Software; 1998–2003
- Dalziel SR,
- Parag V,
- Rodgers A,
- Harding JE
- Skelton R,
- Gill AB,
- Parsons JM
- ↵Cheong JL, Hunt RW, Anderson PJ, et al. Head growth in preterm infants: correlation with magnetic resonance imaging and neurodevelopmental outcome. Pediatrics. 2008;121(6). Available at: www.pediatrics.org/cgi/content/full/121/6/e1534
- White NH,
- Cleary PA,
- Dahms W,
- Goldstein D,
- Malone J,
- Tamborlane WV,
- Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) Research Group
- Plank J,
- Blaha J,
- Cordingley J,
- et al
- Copyright © 2012 by the American Academy of Pediatrics