Interventions for treatment of neonatal hyperglycemia in very low birth weight infants
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
Primary Objective:
To assess the effects of interventions for treating neonatal hyperglycemia in the very low birth weight neonate receiving total or partial parenteral nutrition. Specific interventions to be reviewed are:
(i) restriction in the rate of parenteral glucose infusion, with or without any increase in the provision of parenteral lipids or amino acids (compared with no restriction in the rate of parenteral glucose infusion)
(ii) insulin infusion (compared with no restriction in the rate of parenteral glucose infusion)
(iii) insulin infusion (compared with restriction in the rate of parenteral glucose infusion)
Secondary Objectives:
Data permitting, sub‐group analyses will be carried out according to population:
(a) Birth weight 500 ‐ 749 g; 750 ‐ 999 g; 1000 ‐ 1499 g
(b) Criterion for hyperglycemia at study entry: 8.3 ‐ 13.8 mM/L (150 ‐ 250 mg/dL); > 13.8 mM/L (> 250 mg/dL)
(c) Associated morbidity at study entry: high/low morbidity score or as reported by author
Background
Description of the condition
In the very low birth weight neonate (VLBW, < 1500 g at birth), an elevated blood glucose concentration occurs frequently, especially during the first weeks after birth when glucose is administered as part of parenteral nutrition. The extremely low birth weight neonate (ELBW, < 1000 g at birth) is especially vulnerable to this occurrence. The increasing survival of this population and the use of early and aggressive parenteral nutrition have resulted in an increased frequency of hyperglycemia in the NICU. Hays 2006 reported that among 82 ELBW neonates studied in the first week after birth, the average daily prevalence of infants with at least one measurement of blood glucose concentration above 8.3 mM/L (> 150 mg/dL) was 57%; above a threshold of 13.9 mM/L (> 250 mg/dL), it was 32%. The risk for hyperglycemia is inversely related to gestational age and birth weight (Blanco 2006; Falcão 1998) and increases with the severity of accompanying illnesses in this population (Louik 1985).
There is no uniform agreement about the definition for hyperglycemia in the VLBW or ELBW neonate. Irrespective of the precise definition, adverse clinical outcomes have been associated with an elevated neonatal blood glucose concentration in these neonates. Adverse outcomes include death (Hays 2006; Kao 2006; Heimann 2007), intraventricular hemorrhage (IVH) grades 3 and 4 (Hays 2006), late onset bacterial infection (Kao 2006), fungal infection (Rowen 1995; Manzoni 2006), retinopathy of prematurity (ROP) (Garg 2003; Blanco 2006; Ertl 2006) and necrotizing enterocolitis (NEC) (Kao 2006). The threshold for hyperglycemia used in these analyses varied: > 8.3 mM/L (> 150 mg/dL) (Hays 2006; Blanco 2006; Heimann 2007), > 8.5 mM/L (> 153 mg/dL) (Ertl 2006), > 10.0 mM/L (> 180 mg/dL) (Kao 2006) and > 12.0 mM/L (> 216 mg/dL) (Manzoni 2006). Several of these studies noted a particularly strong association with adverse outcomes when the hyperglycemia was prolonged (Hays 2006; Kao 2006; Heimann 2007; Rowen 1995). Length of hospital stay (LOS) was reported to be greater when hyperglycemia was sustained, especially when combined with an increased oxygen requirement, which could be related to bronchopulmonary dysplasia (Hays 2006).
The mechanisms of the hyperglycemia that is commonly seen in the parenterally fed VLBW/ELBW infant in the early neonatal period are probably multifactorial. In order to provide energy and support growth until full enteral feeding is established, these babies are usually given a parenteral nutrition regimen which delivers glucose at a rate substantially higher than the endogenous glucose production rate of 4 ‐ 7 mg/kg/min that was reported for VLBW infants by Bier 1977. Unlike the non‐pregnant and non‐diabetic adult, the preterm neonate does not completely suppress endogenous glucose production in response to a parenteral glucose infusion administered at a rate comparable to the rate of endogenous glucose production (Cowett 1983; Sunehag 1994). In comparison to older infants and children, the VLBW/ELBW neonate may also have a reduced tolerance for intravenous glucose administration due to a limited amount of insulin‐dependent tissue (specifically fat and muscle), and a limited insulin secretory response to glucose (Mitanchez‐Mokhtari 2004).
Description of the intervention
Interventions that might be used to treat hyperglycemia in the parenterally fed VLBW/ELBW neonate include:
i) restricting the rate of parenteral glucose infusion, with or without any change in the other energy components of the infusate
ii) treating with insulin, primarily by intravenous infusion
How the intervention might work
Reducing the rate of parenteral glucose infusion:Cowett 1979 reported that 10 VLBW neonates who received a three hour infusion of glucose at 14.0 mg/kg/min all developed hyperglycemia, defined as a plasma glucose concentration > 8.3 mM/L (>150 mg/dL). When glucose was infused in 16 infants at a rate of 11.2 mg/kg/min, half of them developed hyperglycemia; but among nine infants infused at a rate of 8.1 mg/kg/min, none developed hyperglycemia. In a retrospective survey of 86 ELBW neonates in one NICU over a one‐year period, Cowett 1997 reported a direct linear association between plasma glucose concentration and the glucose infusion rate. These observations can be interpreted to suggest that simply reducing the glucose infusion rate may be effective in treating hyperglycemia. However, the nutritional disadvantages need also to be considered. In the early postnatal period, many VLBW/ELBW infants cannot be fed enterally, and must rely on parenteral feeding, either in whole or in part, in order to grow. The parenterally fed neonate needs an energy intake of 80 ‐ 90 kcal/kg/day in order to grow at the intrauterine rate (Heird 1992; Zlotkin 1981). The energy requirement for adequate growth may not be met at a reduced parenteral glucose intake, depending on the amount of energy provided by the other components of the parenteral feeding regimen.
Sunehag 1999 showed that when the glucose infusion rate is reduced, the VLBW neonate can use part of the energy supplied by non‐carbohydrate sources, including glycerol and amino acids, to sustain the blood glucose concentration. It is possible that a reduction in the glucose infusion rate, even if partially offset by an increase in the administration of lipid and amino acids, might allow some reduction in blood glucose concentration while reducing the risk of energy deficiency. Nevertheless, early and/or rapid advancement of lipid infusion in the preterm infant should be carefully monitored for potential side effects, including lipid intolerance with lipemia and hypertriglyceridemia, increased free bilirubin concentration, impaired pulmonary function and interference with immune and platelet function (Thureen 1999). Similar concerns apply for amino acids, in response to which the neonate might develop metabolic acidosis, hyperammonemia and/or azotemia.
Treating with insulin: Insulin acts on cells throughout the body to stimulate uptake, utilization and storage of glucose. In particular, insulin stimulates the liver to store glucose in the form of glycogen, and facilitates the entry of glucose into muscle and adipose tissue. However, there appear to be important developmental differences between VLBW infants and older subjects in regard to the role of insulin in regulating blood glucose concentration. In proportion to body weight, the VLBW neonate has a relatively small mass of muscle and fat, but a larger brain. Insulin facilitates the import of glucose into muscle and fat, via the glucose transporter GLUT 4, but has no role in the uptake of glucose by the brain. Since the brain is the major glucose‐utilizing organ, these considerations suggest that insulin may have a relatively restricted role in regulating glucose utilization in the VLBW neonate. Moreover, Mitanchez‐Mokhtari 2004 reported that extremely preterm infants with hyperglycemia during the first week of life had a very high concentration of proinsulin, although insulin concentration did not differ significantly from that noted in normoglycemic controls. This suggested that proinsulin processing to mature insulin was partially defective. Mitanchez‐Mokhtari 2004 also found that the hyperglycemic neonate responded to exogenous insulin infusion, but needed higher doses to achieve normoglycemia. These data suggest the possibility of insulin resistance.
There have been several observational studies, principally case series, suggesting that exogenous insulin administration can safely reduce glucose concentration in the hyperglycemic VLBW or ELBW neonate (Vaucher 1982; Binder 1989; Thabet 2003). Usually, a starting insulin infusion rate of 0.02 to 0.10 units/kg/h is administered. The insulin is typically administered in a small volume of isotonic saline delivered by an infusion pump (Ostertag 1986). It has been recommended to first flush the intravenous tubing with the insulin solution in order to saturate binding sites on the tubing so that the insulin does not adhere (Fuloria 1998) or to administer the insulin with a protein solution of 0.25% salt poor albumin (Cowett 1987). Frequent glucose monitoring, with adjustments of the rates of insulin and/or glucose infusion, is required. With careful monitoring and dosage adjustments, hypoglycemia occurred very infrequently in these studies.
Why it is important to do this review
The association of neonatal hyperglycemia with adverse clinical outcomes in the VLBW neonate, described previously in observational studies, does not necessarily mean that the hyperglycemia causes these sequelae. It is possible that it is the sicker neonate, who is at higher risk for adverse clinical outcomes, who is more prone to hyperglycemia, e.g. because of stress or metabolic injury to vital organs, especially the brain. It is also possible that it is the hyperglycemia per se that is a cause of adverse clinical outcomes, especially so if hyperglycemia is severe enough to cause significant plasma hyperosmolality and accompanying fluid shifts from the intracellular to the extracellular fluid compartment. This is of particular concern with respect to the risk of cerebral bleeding (Finberg 1967). There is also some evidence of an adverse effect of severe hyperglycemia on neurologic outcome following cerebral ischemia in both animals and adult humans, as well as the neonate (Efron 2003). Thus, in the VLBW/ELBW neonate with hyperglycemia, it is important to determine the benefits and risks of treating the hyperglycemia by either reducing the rate of glucose infusion, or by administering exogenous insulin. We propose to systematically review the evidence from randomized controlled trials of these interventions in this population. We are not aware of any previous systematic review of randomized trials of treatments for hyperglycemia in the VLBW/ELBW neonate. Raney et al recently reported a systematic review, including both randomized trials and observational studies, of insulin infusion for the treatment of VLBW neonates with hyperglycemia (Raney 2008).
Objectives
Primary Objective:
To assess the effects of interventions for treating neonatal hyperglycemia in the very low birth weight neonate receiving total or partial parenteral nutrition. Specific interventions to be reviewed are:
(i) restriction in the rate of parenteral glucose infusion, with or without any increase in the provision of parenteral lipids or amino acids (compared with no restriction in the rate of parenteral glucose infusion)
(ii) insulin infusion (compared with no restriction in the rate of parenteral glucose infusion)
(iii) insulin infusion (compared with restriction in the rate of parenteral glucose infusion)
Secondary Objectives:
Data permitting, sub‐group analyses will be carried out according to population:
(a) Birth weight 500 ‐ 749 g; 750 ‐ 999 g; 1000 ‐ 1499 g
(b) Criterion for hyperglycemia at study entry: 8.3 ‐ 13.8 mM/L (150 ‐ 250 mg/dL); > 13.8 mM/L (> 250 mg/dL)
(c) Associated morbidity at study entry: high/low morbidity score or as reported by author
Methods
Criteria for considering studies for this review
Types of studies
Randomized or quasi‐randomized trials, parallel groups, that randomize individual patients. Randomized cross‐over trials will be excluded. Unpublished trials or trials reported only in abstract form will be eligible.
Types of participants
Birth weight < 1500 g or gestational age < 32 weeks, postnatal age up to 28 days, on full or partial parenteral feeding, and with documented hyperglycemia, defined as whole blood or plasma glucose concentration > 8.3 mM/L (> 150 mg/dL).
Types of interventions
i) Restriction in the rate of parenteral glucose infusion, with or without any increase in the provision of parenteral lipids or amino acids
Restriction in the rate of glucose infusion is defined as a stepwise or one‐time decrease resulting in either a 25% or greater total decrease, or a decrease to a rate of less than 7 mg/kg/min, or other restriction as defined by the authors. Duration of intervention as defined by the authors. An eligible increase in the provision of parenteral lipids or amino acids is any increase up to but not exceeding that required to replace the deficit in energy intake caused by the reduction in the rate of glucose infusion.
The comparison is no restriction in glucose infusion rate. No restriction is defined as a decrease of less than 10%, or other adjustment as defined by author.
ii) Insulin infusion (any dose, any dose adjustment protocol, any duration)
The comparison is no restriction in glucose infusion rate, as defined above.
iii) Insulin infusion (any dose, any dose adjustment protocol, any duration)
The comparison is restriction in glucose infusion rate, as defined above.
Types of outcome measures
Primary outcomes
1. All‐cause mortality up to 36 weeks postmenstrual age or as defined by authors
2. Neurodevelopmental impairment, defined as presence of one of the following: cerebral palsy, MDI or PDI < 70, blindness or deafness, assessed either between 18 and 24 months postmenstrual age, or the latest assessment up to 24 months postmenstrual age
3. Severe IVH, defined as grade 3 or 4 by Papile classification
4. Incidence of ROP: a) any stage; b) requiring treatment
5. Proportion of neonates with one or more episodes of bacterial sepsis, defined as a positive culture for bacteria in blood, urine or cerebrospinal fluid up to 36 weeks postmenstrual age or as reported by authors
6. Proportion of neonates with one or more episodes of fungal sepsis, defined as a positive culture for fungus in blood, urine or cerebrospinal fluid up to 36 weeks postmenstrual age or as reported by authors
Secondary outcomes
1. Time to resolve hyperglycemia, defined as hours to achieve plasma or blood glucose concentration < 8.3 mM/L (< 150 mg/dL), or as defined by authors
2. Number of episodes of recurrent hyperglycemia per patient and/or proportion of neonates having one or more episodes; defined as plasma or blood glucose concentration > 8.3 mM/L (> 150 mg/dL) or as defined by authors
3. Number of episodes of hypoglycemia per patient, and/or proportion of neonates having one or more episodes; defined as plasma or blood glucose concentration < 2.5 mM/L (< 45 mg/dL) or as defined by authors
4. Caloric intake, assessed as kcal/kg/d, in one week period, or longer or as reported by the authors
5. Incidence of necrotizing enterocolitis (NEC), defined as stage 2 or above by the Bell classification
6. Duration of mechanical ventilation
7. Incidence of chronic lung disease (CLD) at 36 weeks postmenstrual age (Jobe 2001)
8. Length of hospital stay, defined as number of days until discharge home or as reported by authors
Search methods for identification of studies
The standard search strategy of the Neonatal Review Group, as outlined in the Cochrane Library, will be used. The following sources will be searched for eligible reports in any language:
Electronic searches
Electronic databases to be searched will include: The Cochrane Central Register of Controlled Trials (CENTRAL). The most recent disk issue will be used. MEDLINE (1966 ‐ July 2008), EMBASE (1980 ‐ July 2008) and CINAHL (1982 ‐ July 2008).
The search string for searching CENTRAL, and MEDLINE via PubMed, will include the following terms: (glucose/administration and dosage OR glucose infusion OR parenteral glucose OR intravenous glucose OR energy intake OR insulin/administration and dosage) AND (hyperglycemia OR hyperglycemic OR glucose intolerance OR glucose intolerant) AND (infant, very low birth weight OR very low birth weight OR VLBW OR extremely low birth weight OR ELBW OR preterm). To limit to clinical trials, we will use the maximum sensitivity methodologic filter for questions of therapy, implemented in PubMed Clinical Queries (Haynes 2005).
We will use a similar search string for searching EMBASE and CINAHL via Ovid, adapting the search terms to the structured vocabulary and syntax required for those databases; and adapting the limits according to what is available in those databases.
Searching other resources
Abstracts presented in the past three years at the annual meetings of The European Society for Paediatric Research and The Society for Pediatric Research will be searched from the journal Pediatric Research and abstracts on line.
On‐going trials will be searched at the following websites: clinicaltrials.gov and controlled‐trials.com
Using Web of Science, we will do a citation search of Collins (Collins JW, Hoppe M, Brown K, Edidin DV, Padbury J, Ogata ES. A controlled trial of insulin infusion and parenteral nutrition in extremely low birth weight infants with glucose intolerance. The Journal of Pediatrics 1991; 118:921‐927), and Meetze (Meetze W, Bowsher R, Compton J, Moorehead H. Hyperglycemia in extremely low birth weight infants. Biology of the Neonate 1998; 74: 214‐221).
Data collection and analysis
Selection of studies
The titles and abstracts of reports which are detected by the described search strategies will be assessed independently by two review authors (MB and JS) to determine their eligibility for inclusion in this review. Eligibility for inclusion will be judged according to the criteria listed under Criteria for considering studies. If there is uncertainty as regards inclusion/exclusion, the full report will be obtained in order to make a decision re eligibility. An attempt will be made to resolve any disagreement by discussion. Unresolved disagreements will be referred to RC for arbitration.
Data extraction and management
For included studies, data will be extracted concerning study design, methodology, clinical features of the population, interventions and outcomes, and treatment effects, using specially designed data collection forms. For studies that were initially considered possibly eligible for inclusion but which were excluded after reading the full report, the reason for exclusion will be documented. All data will be extracted independently by two review authors (MB and JS), compared, and any discrepancies resolved by discussion or, if necessary, through contact with the primary investigators. Unresolved disagreements will be referred to RC for arbitration. Outcomes which are categorical will be expressed as the negative outcome, eg death, not survival. We will attempt to obtain data sets that are as complete as possible. We will request from the primary investigators any unreported data on study outcomes, if necessary. To the extent possible, we will extract outcome data on all patients randomized.
Assessment of risk of bias in included studies
The methodologic quality of each included trial will be assessed independently by two review authors, MB and JS, with any disagreement resolved by discussion. Each trial will be assessed for blinding of randomization, blinding of intervention(s), complete follow‐up, and blinding of outcome(s) measurement, using a 3‐category scale: yes, no, can't tell.
Measures of treatment effect
For individual trials, effect measures for categorical outcomes will include relative risk (RR), and absolute risk difference (RD), each with its 95% confidence interval (CI). For statistically significant effects, number needed to treat (NNT), or number needed to harm (NNH), will be calculated. For continuous outcomes, the effect measure will be mean difference (MD) or, if the scale of measurement differs across trials, standardized mean difference (SMD), each with its 95% CI. For any meta‐analyses (see below), for categorical outcomes the typical estimates of RR and RD, each with its 95% CI, will be calculated; and for continuous outcomes the weighted mean difference (WMD) or a summary estimate for SMD, each with its 95% CI, will be calculated.
Unit of analysis issues
We do not anticipate any unit of analysis issues. Cross‐over trials are not eligible. Cluster randomized trials are not expected in this field. For episodes of hypoglycemia, and episodes of recurrent hyperglycemia, we plan to analyse the data as number of episodes per neonate, and/or proportion of neonates having one or more episodes.
Dealing with missing data
If some outcome data remain missing despite our attempts to obtain complete outcome data, we will perform an available‐case analysis, based on the numbers of patients for whom outcome data are known. For primary outcomes, if there are instances of statistically significant effects but with missing data, we will perform a worst case/best case sensitivity analysis based on imputation, to test whether the effect is sustained, or overturned.
Assessment of heterogeneity
Before any meta‐analysis is done, we will judge whether there is sufficient similarity between the eligible studies in their design features and clinical features (population, interventions) to make pooling for meta‐analysis scientifically and clinically credible. If there is sufficient similarity, we will proceed to meta‐analysis. If not, the results of individual trials will be described separately.
The amount of heterogeneity of treatment effect across trials in a meta‐analysis will be estimated using the I‐squared statistic. Whether heterogeneity is statistically significant will be tested using the chi‐squared statistic. If substantial heterogeneity is present, its source(s) will be explored post facto, considering differences in design or clinical features of the trials.
Assessment of reporting biases
We do not anticipate a sufficient number of included trials in this field to permit assessment of possible publication bias and other biases using symmetry/asymmetry of funnel plots.
For included trials, we will explore possible selective reporting of study outcomes by comparing the primary and secondary outcomes in the reports with the primary and secondary outcomes nominated at trial registration, using the websites www.clinicaltrials.gov and www.controlled‐trials.com. If such discrepancies are found, we will contact the primary investigators to try to obtain missing outcome data on outcomes pre‐specified at trial registration.
Data synthesis
If meta‐analysis is judged to be appropriate, it will be done using RevMan 5, supplied by the Cochrane Collaboration. For estimates of typical relative risk and risk difference, we will use the Mantel‐Haenszel method. For measured quantities, we will use the inverse variance method. All meta‐analyses will be done using the fixed effect model. When meta‐analysis is judged to be inappropriate, individual trials will be analyzed and interpreted separately.
Subgroup analysis and investigation of heterogeneity
Data permitting, pre‐specified subgroup analyses will be done according to population subgroups as defined in Secondary Objectives. If data are not available to permit a pre‐planned subgroup analysis to be done, that will be stated as a result in this review. Any post‐facto subgroup analyses, e.g. to explore unanticipated sources of heterogeneity, will be labeled as such. It is not anticipated that a sufficiently large number of eligible trials will be available to permit exploration of heterogeneity of treatment effect using meta‐regression.
Sensitivity analysis
As noted previously, in the case of missing outcome data, a worst case/best case analysis is planned to test whether a significant result is sensitive to (ie overturned by) imputations according to worst/best case assumptions. There are no other pre‐planned sensitivity analyses.
