|Home | About | Journals | Submit | Contact Us | Français|
The Canadian guidelines recommend blood glucose (BG) screening starting at 2 h of age in asymptomatic ‘at-risk’ babies (including small-for-gestational-age [SGA] and large-for-gestational-age [LGA] infants), with intervention cut-offs of 1.8 mmol/L and 2.6 mmol/L. The present study reviews and audits this practice in full-term newborn populations.
A literature review meta-analyzed BG values in appropriate-for-gestational age (AGA) term newborns to establish normal 1 h, 2 h and 3 h values. A clinical review audited screening of ‘at-risk’ SGA and LGA term newborns, evaluating both clinical burden and validity.
The review included six studies, although none clearly defined the plasma glucose standard. The pooled mean (plasma) BG level in AGA babies 2 h of age was 3.35 mmol/L (SD=0.77), significantly higher than 1 h levels (3.01 mmol/L, SD=0.96). In the audit, 78 SGA and 142 LGA babies each had an average of 6.0 and 4.7 BG tests, respectively. The mean 2 h BG levels for SGA (3.42 mmol/L, SD=1.02) and LGA (3.31 mmol/L, SD=0.66) babies did not differ significantly from the AGA pooled mean. Receiver operating characteristic curves showed that 2 h BG levels in LGA and SGA babies predicted later hypoglycemia (defined as a BG level lower than 2.6 mmol/L), but sensitivities and specificities were poor.
Published 2 h BG levels for AGA babies are higher than 1 h values and are similar to audited 2 h levels in SGA and LGA babies. Clinically, 2 h levels are predictive of later hypoglycemia but may require repeat BG testing. Audit is an important tool to validate national guidelines, to minimize their burden and to maximize their utility.
Selon les lignes directrices canadiennes, il faut entreprendre le dépistage de la glycémie à compter de deux heures de vie chez les bébés asymptomatiques vulnérables (y compris les nourrissons petits par rapport à leur âge gestationnel [PAG] et gros par rapport à leur âge gestationnel [GAG]), les limites d’intervention s’élevant à 1,8 mmol/L et 2,6 mmol/L. La présente étude contient l’analyse et la vérification de cette pratique au sein de populations de nourrissons à terme.
Une analyse bibliographique a permis de procéder à une méta-analyse des valeurs de glycémie chez des nouveau-nés à terme au poids adéquat par rapport à leur âge gestationnel (AAG), afin de déterminer les valeurs normales à une heure, deux heures et trois heures de vie. Une analyse clinique a permis de vérifier le dépistage des nouveau-nés à terme vulnérables PAG et GAG, ainsi que d’évaluer le fardeau clinique et la validité des résultats.
L’analyse contenait six études, mais aucune ne définissait clairement la norme de glucose plasmatique. La glycémie regroupée (plasmatique) moyenne des bébés AAG à deux heures de vie s’élevait à 3,35 mmol/L (ÉT=0,77), ce qui est considérablement plus élevé que la glycémie à une heure (3,01 mmol/L, ÉT=0,96). Dans le cadre de la vérification, 78 bébés PAG et 142 bébés GAG ont obtenu des glycémies moyennes de 6,0 et 4,7, respectivement. Les glycémies moyennes à deux heures de vie chez les bébés PAG (3,42 mmol/L, ÉT=1,02) et GAG (3,31 mmol/L, ÉT=0,66) ne différaient pas de manière significative des glycémies regroupées moyennes des bébés AAG. Les courbes de fonction d’efficacité du récepteur ont démontré que la glycémie à deux heures de vie chez les bébés GAG et PAG était indicatrice d’une hypoglycémie (définie comme une glycémie inférieure à 2,6 mmol/L) ultérieure, mais la sensibilité et la spécificité étaient faibles.
Les valeurs de glycémie publiées chez les bébés AAG à deux heures de vie sont plus élevées que celles à une heure de vie et similaires à celles vérifiées à deux heures de vie chez les bébés PAG et GAG. D’un point de vie clinique, la glycémie à deux heures de vie est indicatrice d’une hypoglycémie ultérieure, mais elle peut exiger des lectures répétées. La vérification est un outil important pour valider les lignes directrices nationales afin d’en réduire le fardeau au minimum et d’en accroître l’utilité au maximum.
In December 2004, the Canadian Paediatric Society (CPS) published guidelines (1) for the screening of newborns who were considered to be ‘at risk’ for neonatal hypoglycemia. The thresholds for hypoglycemia were drawn from three sources – so-called ‘normal’ values derived from appropriate-for-gestational-age (AGA) term babies, neurophysiological evidence from assumed hypoglycemic babies and follow-up data from preterm infants with transient hypoglycemia. The resulting guidelines were pragmatic, based on evidence that was limited in both level and quality, and recommended further study. With this caveat, the guidelines recommended that AGA newborns did not need routine screening of blood glucose (BG) levels. They also recommended that infants considered to be ‘at risk’ should be screened at 2 h of age and every 3 h to 6 h thereafter until the risk was minimized (a grade C recommendation, based on level 2 to 4 evidence of limited quality). Infants of diabetic mothers (IDMs), preterm infants (younger than 37 weeks’ gestational age), and those who were small for gestational age (SGA – weighing less than the 10th percentile) or large for gestational age (LGA – weighing more than the 90th percentile) were considered to be at risk.
The CPS algorithm organized interventions using the 2 h BG level and two initial intervention thresholds (1.8 mmol/L and 2.6 mmol/L). At-risk newborns with a BG level lower than 1.8 mmol/L would receive intravenous glucose therapy, whereas those between 1.8 mmol/L and 2.5 mmol/L (inclusive) would receive supplemental feeding and augmented surveillance. The CPS acknowledged that studies of healthy AGA term babies without risk factors showed that BG levels of 1.8 mmol/L to 2.5 mmol/L lay within the normal range at this age, reflecting the natural trough that occurs after birth. A valid concern was the number of healthy term babies who would require intervention.
The audit evaluates the CPS guidelines using two approaches: first, through a literature review of the normal ranges for BG levels in healthy AGA newborns in the first few hours of age; and second, through a clinical review (ie, a comprehensive chart audit) of SGA and LGA term babies. The audit evaluates the burden of the CPS guidelines on an asymptomatic population, as well as the predictive value of a 2 h BG level in babies at risk for subsequent hypoglycemia (defined as a BG level lower than 2.6 mmol/L). The study objective was to rationalize the recommended practice of recurrent testing of at-risk babies with potentially normal BG levels.
Articles written in English, French and German were retained from a PubMed (Medline) and Embase search (from inception to June 2008) using the following terms: ‘infant, newborn’, ‘blood glucose’, ‘plasma glucose’, ‘serum glucose’, ‘normal’, ‘term’ and ‘appropriate-for-gestation’. The search included all trials, reviews, clinical practice guidelines, follow-up studies and meta-analyses. No unpublished or abstract-only papers were included.
Studies using laboratory standard BG testing in healthy AGA term babies up to and including 3 h of age were included if they specified method(s) of analysis, time(s) of sampling, mean values and SDs (or variances or standard errors). Two of five authors of studies missing these fields responded to the request for further data and were included in the analysis.
Acceptable sampling methods included plasma, serum or whole blood obtained from venous, arterial or capillary samples. For the purposes of the present review, the term ‘BG’ was used generically to describe all the former methods of sampling and processing. One hour, 2 h and 3 h plasma glucose values were abstracted and, when encountered, whole blood values were adjusted to plasma glucose values using a correction factor of 1.135 (2,3). Studies (4) using bedside monitoring (strips and metres) were excluded due to concerns regarding quality control and reliability. Studies including SGA and LGA newborns and IDMs, or mothers with significant perinatal complications such as toxemia, anemia or hypertension, were also excluded. Both longitudinal and cross-sectional studies were included.
One hour, 2 h and 3 h results (means and variances) were pooled into three respective groups; pooling was performed by weighting for sample size. Comparisons of means and variances were performed within and among groups using PEPI version 4.0 (5).
Between July 10, 2005, and April 25, 2006, a chart review at the Janeway Children’s Health and Rehabilitation Centre (St John’s, Newfoundland and Labrador) identified the number of SGA and LGA term babies (37+0 to 40+6 weeks’ gestation) and their BG results. IDMs, due to their small number, and preterm babies, due to their variation in clinical symptoms, were not included. During this time, the CPS guidelines for BG screening were institutional policy. A positive screen was defined as an initial (2 h) BG level lower than 1.8 mmol/L, or a subsequent level lower than 2.6 mmol/L.
A clinical audit was performed using a near-patient, quality-controlled methodology. BG levels were measured using a Radiometer ABL700 analyzer (Radiometer A/S, Denmark) – a desktop device requiring a sample size of approximately 0.1 mL of heparinized capillary blood. This uses a direct-reading glucose electrode measuring plasma glucose in whole blood; it is quality-controlled down to a reading of 1.0 mmol/L and reliable to a reading of 0.1 mmol/L. The BG level was reported according to the plasma glucose standard as required by the International Federation of Clinical Chemistry (Milan, Italy) (6). Point-of-care reflectance or strip devices were not used for newborns in this institution.
The following details were abstracted for all SGA and LGA term newborns (defined as weighing less than the 10th percentile and more than the 90th percentile, respectively, according to Canadian data from Kramer et al ): gestational age, birth weight, date and time of birth, BG levels and times of sampling. Babies weighing more than the 90th percentile were further subdivided into 90th to 95th percentiles and more than the 95th percentiles to compare these two populations. Breastfeeding initiation rates in St John’s were approximately 55% during this time (K Aziz, personal communication). Standard practice was to encourage feeding all healthy ‘at-risk’ babies before their first BG screen.
Receiver operating characteristic (ROC) curves were used to determine whether the 2 h BG level predicted a later episode of hypoglycemia on subsequent testing (defined as a BG level lower than 2.6 mmol/L). ROC curves outline how sensitivity of a test might increase as the BG cut-off increases, including more babies in the definition of ‘hypoglycemia’; however, specificity will decrease as more normal babies are labelled as ‘hypoglycemic’. The arbitrary target was to estimate a BG level that identified more than 90% of babies who would subsequently have a BG level lower than 2.6 mmol/L. This analysis was performed using the ‘Analyse-it’ General and Laboratory Statistics package version 1.73 (STATCON, Germany) plug-in for Excel (Microsoft Corporation, USA). The present study was conducted with the approval of the Human Investigation Committee of Memorial University and Eastern Health (St John’s).
There were no prospective, randomized, controlled studies that evaluated thresholds for intervention.
Twelve studies were found that specifically targeted BG levels at 1 h to 3 h of age. No studies accurately described a plasma glucose standard reporting methodology as recommended by the International Federation of Clinical Chemistry, so inferences were made by the authors (KA and JC) from the methods sections of each study. Five authors were contacted regarding missing data and two responded (8,9). Subsequently, six studies were excluded: three for missing data, two for the inability to verify sample processing methodology and one because babies with low sugar levels were treated before 1 h of age (Table 1).
Six observational studies representing 12 specific samples at 1 h, 2 h or 3 h met selection criteria (Table 2), describing either prospective or retrospective, and longitudinal or cross-sectional studies of AGA term newborns in hospitals.
The pooled mean BG values from these studies at 1 h, 2 h and 3 h (with pooled SD, sample size and number of studies) were 3.01 mmol/L (SD=0.96, n=262, six cohorts), 3.35 mmol/L (SD=0.77, n=186, three cohorts) and 3.25 mmol/L (SD=0.96, n=301, three cohorts), respectively (Table 3). Using multiple comparisons of means (PEPI version 4.0), there were significant differences among means within groups at both 1 h and 3 h (Table 3). The pooled means at 1 h differed significantly among groups from both 2 h and 3 h (P<0.001), whereas no significant difference was found among 2 h and 3 h pooled means. Pooled variances at 1 h and 3 h were significantly greater than at 2 h.
Using z-scores, the estimated BG values for the 5th and 10th percentiles at 2 h were 2.1 mmol/L and 2.4 mmol/L, respectively (1.7 mmol/L and 2.1 mmol/L at 3 h). At 2 h of age, 1.8 mmol/L and 2.6 mmol/L approximated to the 2nd and 16th percentiles respectively, and the 6th and 23rd percentiles at 3 h of age (Table 3).
During the nine-month study period (between July 10, 2005, and April 25, 2006), 1805 babies were born at the Janeway Children’s Health and Rehabilitation Centre. Of 397 term and preterm babies who were eligible for glucose testing, 374 (95%) newborns were screened. Of the 233 screened babies who were 37+0 to 40+6 weeks’ gestation, 85 (36%) were SGA and 148 (64%) were LGA. The mean birth weight for SGA term babies was 2584 g (SD=1047) and for LGA term babies was 4253 g (SD=1711).
For SGA and LGA term newborns, respectively, an average of 6.0 (SD=3.9) and 4.8 (SD=2.9) BG screening tests were conducted per baby. Initial testing times averaged 2 h 44 min for SGA and 3 h 4 min for LGA babies; the average final sampling occurred at 25 h 14 min and 19 h 40 min of age, respectively. Furthermore, the average frequency of blood collection was every 4 h 30 min for SGA babies and every 4 h 2 min for LGA babies.
Initial BG levels for SGA and LGA term babies were 3.42 mmol/L (SD=1.02) and 3.31 mmol/L (SD=0.66), respectively, with four (5%) being lower than 1.8 mmol/L in SGA and one (1%) being lower than 1.8 mmol/L in LGA babies. These mean values do not differ significantly from each other, or from the literature-derived pooled means at 2 h and 3 h (Table 3). The CPS cut-off of 2.6 mmol/L identified 21% of SGA and 14% of LGA babies as requiring further monitoring or intervention. Twenty-one SGA babies and 24 LGA babies (26% and 17%, respectively) subsequently had levels lower than 2.6 mmol/L. Of 233 screened babies, a total of 57 (24%) babies had BG levels that, according to the CPS algorithm, would lead to supplementation or intravenous therapy.
ROC curves (Figures 1 and and2)2) showed that initial BG sampling was predictive of subsequent hypoglycemia (defined by the CPS as lower than 2.6 mmol/L) in both populations (SGA, P<0.05, area under the curve 66%; LGA, P<0.0001, area under the curve 77%). In the LGA cohort, an initial BG threshold of 3.5 mmol/L had a sensitivity of 92% and specificity of 47% for predicting subsequent hypoglycemia (Figure 2). No similar point of inflection was clinically relevant for SGA babies (Figure 1). However, an initial BG threshold of 4.1 mmol/L was required in the SGA population to obtain a sensitivity of 91% and a specificity of 20% for predicting subsequent hypoglycemia.
The mean BG level of babies between the 90th and 95th percentiles for birthweight was significantly higher than that of babies weighing more than the 95th percentile (3.52 mmol/L, SD=0.56 versus 3.20 mmol/L, SD=0.68; P<0.01).
To our knowledge, this is the first publication to comprehensively evaluate the Canadian BG screening guidelines in clinical practice. Our literature review confirmed the normal ranges for healthy AGA term babies at 2 h and 3 h of age; and our clinical audit found these to be similar to healthy SGA and LGA term babies. The CPS protocol could be followed in practice. We found that the Canadian guidelines required approximately one in eight term babies to be screened, and one in 30 to be treated. We also found that although initial BG sampling predicted later hypoglycemia, ‘at-risk’ babies were still exposed to numerous, repeated blood sampling. Finally, we discovered that LGA babies may not be a homogeneous population, with larger LGA babies having significantly lower glucose levels.
Our literature review, with lower BG levels at 1 h than at later times, suggests that the timing of early BG screening affects screening and treatment thresholds. The lack of difference between 2 h and 3 h samples allows for some flexibility in sampling times. The CPS thresholds of 1.8 mmol/L and 2.6 mmol/L approximate to the 10th and 33rd percentiles at 1 h, but, at the most, the 6th and 23rd percentiles at 2 h to 3 h. The clinical inference is that screening for hypoglycemia at 1 h of age may lead to more false-positive diagnoses and, therefore, unnecessary interventions, unless the intervention threshold at this age is lowered in keeping with population norms.
There were several limitations to the literature review. First, we included both cross-sectional and longitudinal studies. Longitudinal studies follow the same baby at 1 h, 2 h and 3 h of age, whereas cross-sectional data sample from different populations at each of these times. This discrepancy makes it impossible to analyze data by repeated measures, and leads to heterogeneity within individual time periods. Ideally, population norms should be derived from longitudinal data.
Second, we included studies from babies who were either bottle and/or breastfed, or not fed at all. If feeding practices affect BG levels, they may explain some of the significant differences within the same time samples. It was also not clear in most studies whether glucose sampling occurred before or after feeding – another potential confounder. Screening thresholds may need modifying based on feeding practices.
Third, methods for obtaining BG and subsequent analysis varied. Although we excluded studies that used bedside reflectance metres, and only selected those that apparently used robust laboratory methods, we can only have limited confidence that sampling and processing techniques did not influence our results. None of the studies reviewed specifically addressed this issue to our satisfaction; consequently, we made inferences regarding sampling methodology. Confounding factors may have included the source of blood (venous versus capillary), delays in transportation to the laboratory and the discrepancies among measurements of whole blood, serum and plasma glucose. Future studies on BG levels in newborns must be clear on the nomenclature as well as the use of a plasma glucose standard (6).
Accepting the above limitations, it appears that the timing of BG sampling impacts mean levels, and that delaying sampling until 2 h of age may avoid the natural trough that occurs after birth. However, a caveat would be the potential delay in the diagnosis of hypoglycemia in an at-risk baby. This concern may be overstated in asymptomatic term babies – the CPS threshold in asymptomatic preterm and SGA newborns stemmed from observations of low BG levels for prolonged periods of time (longer than 24 h) (12). There is a case for earlier screening in IDMs because they become hypoglycemic well before 30 min of age. On the other hand, in the asymptomatic LGA population, no harm has been demonstrated from glucose levels above 1.5 mmol/L.
In following these guidelines, 5% of at-risk babies were not screened. No explanation was found for this omission, but one might speculate that these babies were not identified as being at risk. Twelve per cent of term babies born in our institution were exposed to screening, and one-quarter of these infants required supplemental feeding or intravenous glucose for neonatal hypoglycemia. As a result, following the CPS guidelines resulted in 3% of healthy term babies being identified as requiring intervention (supplemental feeding or intravenous therapy). A mean of five to six blood tests were taken from each at-risk baby, up to an average of approximately one day of age, representing a substantial burden. Given the CPS recommendation that LGA babies do not require screening beyond 12 h, these durations and frequencies may have been excessive in larger babies; clinical review of local practice is, therefore, warranted, and, perhaps, this recommendation could be clearer on the algorithm.
Screening, although recommended at 2 h, occurred closer to 3 h in SGA and LGA term babies. The reasons for this are not clear, and local processes need to be evaluated to discover them. Some of this variance is mitigated by the finding that BG levels in this population did not differ from published 2 h and 3 h BG levels in AGA term populations. A noteworthy aspect of our review was the normal practice of feeding babies before performing the BG screen, whether by breast or bottle, and the avoidance of early screening. We believe that this approach to the at-risk newborn facilitated normal transition in otherwise normal babies, but may have resulted in an acceptable delay in initial testing.
The ROC curves suggest that, for as many as 50% of LGA babies, one blood sample may have been enough to exclude the majority of later hypoglycemia (ie, if the BG level exceeded approximately the 50th percentile for the population). This is consistent with the observation that hypoglycemia in this population usually occurs in the first few hours of life (14).
Prediction of subsequent hypoglycemia using initial BG levels is less reliable in SGA babies; in this population, a normal initial BG level may still be followed by a level of less than 2.6 mmol/L (ie, poor sensitivity and specificity for subsequent hypoglycemia). This finding may reflect differing mechanisms of glucose transition in smaller babies, with occasional, unpredictable and later presentation of neonatal hypoglycemia (14). Perhaps the SGA population should be screened less frequently than their larger counterparts, but for a longer period of time – ie, until feeding is established, which is also recommended by the CPS guidelines.
Even more interesting is the observation that babies between the 90th and 95th percentiles appear to have different glucose responses following birth when compared with their larger counterparts (greater than the 95th percentile). There is some debate regarding the need to screen LGA babies (15–17). Our results fuel this debate. Similarly, one might ask whether the 10th percentile is an appropriate cutoff for screening SGA babies; given our smaller numbers, we refrained from analyzing these data. Unfortunately, our study was not powered or designed to answer either of these questions adequately.
Our clinical audit was performed using a near-patient, quality-controlled methodology that is now available in many blood gas machines. Given the wide variation in BG levels in at-risk newborns, there should be concerns regarding the risks and benefits of using less reliable techniques to assess glucose level (such as reflectance metres and glucose strips). Methods with poor accuracy or reliability will increase both false-positive and false-negative diagnoses of hypoglycemia, and raise the number and frequency of repeated tests. We believe that newer, near-patient technologies will improve the value of screening, as well as facilitate larger audits such as our own.
There is an important caveat when developing glucose screening policies. Although data from this study suggest potential areas for modification of existing guidelines, these observations do not have sufficient power to detect occasional cases of severe neonatal hypoglycemia, particularly in those who are undiagnosed IDMs. It is, therefore, important to confirm our findings using a larger cohort of newborns. With these reservations, we hypothesize that screening LGA babies could be limited to a single 2 h BG test, only repeated (once) if the result is lower than 3.5 mmol/L. We question the need to screen babies between the 90th and 95th percentiles, suggesting that this be reserved for jurisdictions that do not screen for gestational diabetes in pregnancy. Our data also support published observations that SGA babies may present with later hypoglycemia (14), hence the poor sensitivity and specificity of early BG testing.
Our clinical review was not designed to test the validity of the 2.6 mmol/L threshold for intervention, which would label approximately one in five babies as hypoglycemic. It is, however, this threshold that determines the proportion of babies who will be treated ‘unphysiologically’ with supplemental feeds or intravenous infusions. Observational studies, such as ours, may improve the efficiency of the screening process but, ultimately, the benefit of BG screening and the appropriate thresholds for intervention, will only be confirmed using a prospective experimental design. The important clinical outcome measure would be long-term neurodevelopmental outcomes of babies subjected to screening, perhaps randomly assigned to different intervention thresholds or screening algorithms.
The audit of national guidelines is important to the process of guideline development; although our audit interrogates many aspects of the CPS guidelines, our data do not allow us to draw definitive conclusions about the value of screening. That being said, the current CPS guidelines for the screening of at-risk newborns functioned in a clinical setting. Our suggested modifications require evaluation in larger populations of LGA and SGA babies, and are likely to considerably diminish the burden of glucose screening. There is clearly a need for prospective studies with longer-term follow-up to resolve these important issues.
CONTRIBUTIONS: Dr Croke performed the chart review, data collection, participated in statistical analysis and drafted the manuscript. Dr Sullivan helped perform the chart review, data collection and statistical analysis. Dr Ryan-Drover and Dr Randell participated in the design of the study. Dr Andrews participated in the design of the study and manuscript preparation. Dr Aziz participated in the design of the study, statistical analysis, manuscript preparation, reanalyzed the data following reviewers’ comments and drafted the final manuscript.