In the same, neonatal, issue of the Formerly Prestigious New England Journal of Medicine as my recent post about inhaled steroids, is a fascinating cohort study of serial blood glucose monitoring in 404 at-risk term infants (the CHYLD study). The infants were initially enrolled in two studies, one of them examining the effects of hypoglycemia on the EEG and the other being the “sugar babies” study, an RCT looking at the use of oral glucose gel as treatment for hypoglycemia. About 75% of them had continuous subcutaneous glucose monitoring started at a median of 4 hours of age. The definition of hypoglycemia used for the primary analysis was a glucose less than 2.6 mmol/L (multiply by 18 for old-fashioned folks= 47 mg/dl).
In both of the initial studies, the protocol was designed to monitor the blood glucose at 1 hour of age then before feeds every 3 to 4 hours. In both studies the protocols were designed to intervene and attempt to maintain a blood sugar above 2.6. Babies from both of those other studies were enrolled in a follow-up program which examined them at 2 years of age with Bayley scales (version 3) and tests of executive function.
They defined 2 major outcomes,
Neurosensory impairment was defined as any of the following findings: developmental delay (BSID-III cognitive or language composite score of <85), motor impairment (BSID-III motor composite score of <85), cerebral palsy, hearing impairment (requiring hearing aids), or blindness (≥1.4 logMAR [log10 of the minimal angle of resolution] in both eyes).
Processing difficulty was defined as either a motion coherence threshold or an executive-function score that was more than 1.5 SD from the mean, indicating performance in the worst 7% of the cohort.
The investigators found no evidence of an adverse effect of low blood sugars at 2.6 mmol/L threshold, (nor at lower thresholds, of 2.3, 2.0 or 1.7, at least as far as I can see from the figures provided in the supplementary appendix). When they looked at the data from the continuous subcutaneous monitoring, there was also no convincing evidence of an adverse effect of low blood sugars, even though a quarter of the babies with low blood sugar actually had hypoglycemia that lasted for 5 hours of more when they examined the data from the continuous monitoring (which they call interstitial glucose in the article).
In fact there was some evidence that babies who had an early blood sugar above 3.0 mmol/L actually had worse outcomes.
This makes me feel good, (when I first typed that it came out “this makes me feel god”, I hope that isn’t a Freudian mis-type!) because the statement that the Fetus and Newborn Committee of the CPS came out with, while I was chair, suggested that a first blood glucose of 1.8 (at 2 hours of age) was acceptable (although it needed following). And that eventually blood sugars should be maintained over 2.6. This was based on relatively limited data, which suggested that the lower limit (or rather the 5th percentile) of blood glucose in healthy full-term breast fed infants with no other risk factors was about 1.8 at 2 hours of age. Of course basing a recommendation on a statistical norm in a healthy population, and then applying them to an at-risk population as a goal of therapy is not necessarily correct.
So maybe we haven’t done any harm with those recommendations, at least, and we might have reduced some over-interpretation and over-treatment of a usually benign phenomenon, which looks like it may be benign even in at-risk babies.
The real problem with all this comes from another very recent publication which suggests that transient hypoglycemia might be harmful. Kaiser JR, et al. Association between transient newborn hypoglycemia and fourth-grade achievement test proficiency: A population-based study. JAMA Pediatrics. 2015;169(10):913-21. It is a study with a different design, different analysis, and different conclusions. In this study from a hospital in Arkansas they had a universal glucose screening policy. The first screen was done at 1 to 3 hours of age, and the “usual practice” was to treat if the glucose was <2 mmol/L. In this study they compared school outcomes between children who had 1 blood glucose <2, <2.25 or <2.5. Children whose neonatal blood sugar was still below the particular threshold on the second blood sugar were eliminated.
They started out with nearly 2000 babies (of 23 to 42 weeks gestation) who had blood sugar results, and were able to confidently match about 1400 of them with school achievement scores.
The babies who were hypoglycemic, below each of the thresholds, were more premature, lower birth weight, and more likely to be twins and more likely to be SGA. their mothers were more likely to be diabetic, have pregnancy complications, and a cesarean delivery, they were also more likely to have not completed high school. All of these variables were then put in a logistic regression, which finally showed an association between having a single blood sugar below each of the thresholds and having lower proficiency on the math or literacy tests.
I’m really not sure about these data, the hypoglycemic babies had so many different characteristics than the non-hypoglycemic, that I am uncertain if logistic regression can really correct for all of them.
What to do with all this, I don’t know if we will ever have an RCT to answer the questions surrounding early hypoglycemia in the newborn. The CHYLD study suggests that at-risk babies, if treated in a protocol designed to try and get the blood sugar over 2.6 mmol/L (47 mg/dl) don’t have any adverse effects of blood sugars getting as low as 1.7 mmol/L. The Kaiser et al paper suggests that in a heterogenous and unselected group of newborns, with a usual practice of trying to get the blood sugar above 2, there is an adverse association of having just one blood sugar under 2.5.
My take on this is that we could continue to follow the CPS statement recommendations for now. I think there is too much risk of residual bias in the Kaiser data to change practice and aim for higher blood sugars, but I must say I am not certain about that, and I don’t know how we will even answer the question. We need more data, it is easy to always say. I think more cohort study data as high quality as the CHYLD study might be able to give us clearer answers. Although the sample size of CHYLD was good, more data with larger samples, and including some lower-risk babies would help. I guess I could dare to suggest an RCT? If we randomly compared at-risk babies at one hour after birth, with 2 different thresholds for intervention, and an effective but non-invasive intervention, and followed them for long enough to see a clinically significant effect, then we could get the answers we need. Any takers?