The only way that we can find the answer to the question of what threshold blood sugar we should use to treat babies with low blood sugars is a prospective RCT, and Behold! Look! Lo! How say you? (van Kempen A, et al. Lower versus Traditional Treatment Threshold for Neonatal Hypoglycemia. N Engl J Med. 2020;382(6):534-44) in the hypoEXIT trial 2024 “at risk” babies (late preterms 35 to 37 weeks, SGA <10%le, LGA >90%le or infants of diabetic mothers) were monitored and 698 of them developed a plasma glucose between 2.6 and 2.0 mmol/L; they were then randomized to one of 2 protocols, the first, the low threshold group, treated the infants by increasing their glucose intake (with feeds or IV glucose) whenever the plasma glucose was less than 2.0, the second treated the infants in the same way, but used a threshold of 2.6 mmol/L. The same threshold was used during the entire 48 hours of the trial.
The primary outcome of the trial was the developmental status at 18 months of age, using the BSID-3 cognitive and motor scales, and the trial was designed as a non-inferiority trial to determine whether the lower threshold was not inferior, looking for a difference of 0.5 SD, and was powered for the 4 at-risk groups individually. Around 85% of the babies came back for follow-up, which is pretty good, I think, for a trial in full-term babies that only lasted 48 hours: Bravo!
As you can see from the main result for the entire sample, there was absolutely no hint of a whisper of a difference between the groups. When you look at each of the strata there is also absolutely nothing there.
In case you think those mean scores of slightly over 100 disprove my previous comment about these at-risk groups having overall poorer scores than their not at-risk peers, these are the BSID-3, and in a society which is somewhat similar to Holland (where hypoEXIT was done), in Australia, term babies at 24 months had a mean cognitive score of over 108, and a mean motor score of 118, Anderson PJ, et al. Underestimation of Developmental Delay by the New Bayley-III Scale. Arch Pediatr Adolesc Med. 2010;164(4):352-6) when excluding babies under 37 weeks or under 2.5 kg; which suggests that these scores are indeed lower than population means.
After the initial randomization, the babies in the low threshold group were more likely to have a plasma glucose < 2.0 mmol/L (10% of them vs 5%), and were more likely to have recurrent episodes < 2.6 mmol/L; 9% compared to 2% went < 2.6 four or more times. The numbers affected with multiple episodes were too small to really affect the developmental outcome of the groups overall.
This is the only reliable information in the medical literature about which glucose threshold should be used to increase glucose intake.
Of note, the HypoEXIT study used a threshold for entry to the study of 2.6 mmol/L of plasma glucose, and added, in a revision of the protocol, that if whole blood glucose is used then the values for entry into the trial were between 1.7 and 2.2 mmol/L, because of the difference between whole blood and plasma values. Maybe this is why the proportion of at-risk infants in HypoEXIT who were eligible was less than the 1/2 in Harris’s study, more like 1/3; HypoEXIT, in fact, required a plasma glucose which was lower than the equivalent blood glucose in the Harris study.
Most of the guidelines, as I mentioned in part 2, have somewhat ignored this distinction, which makes a major difference to which babies we treat and how often they will be considered to be hypoglycaemic.
A whole blood glucose threshold of 2.6 is equivalent to a plasma glucose of about 3.0 mmol/L. A whole blood glucose of 2.0 is a plasma glucose of around 2.3 mmol/L.
Changing an approach to therapy based on only one study should be carefully considered, but, for early neonatal transitional hypoglycaemia, there are no previous data to compare it to. There is no previous prospective controlled data to suggest that screening the more than 30% of newborn infants who are considered to be at-risk and treating them at a threshold of 2.6 mmol/L is of any value to them compared to treating at a lower threshold. Among at-risk infants the proportion of babies who have at least one glucose under is about 50%; so by following many guidelines we label over 15% of newborn infants as being hypoglycaemic, and they then receive an intervention.
To return to where I started a few thousand words ago, the CPS guideline has been revised, although the text has changed only a little, the algorithm that accompanies it no longer has the lower threshold for intervention and retesting at 2 hours (2.0 mmol/L) that was in the previous version; subsequent tests were treated with a threshold of 2.6 mmol/L.
The new guidelines will mean that thousands of babies in Canada, if they followed the old algorithm, would simply have continued to have repeat tests; with the new algorithm, they will be labelled as being hypoglycemic and receive intervention. The usual intervention will simply be oral glucose gel, but retesting is then required after the gel. The threshold is also based on blood glucose rather than plasma glucose, which is likely to mean that about 50% of screened babies will be labelled hypoglycemic, whereas if plasma glucose was used it would be more like 30%. Screening, of course, causes pain, and intervention might possibly actually increase the risk of poor outcomes, as suggested by the increased problems noted in McKinlay’s study among those who had a more rapid correction of a low glucose.
Based on the only reliable data regarding different thresholds that we now have, I think that intervention during the 1st 48 hours of life can be safely withheld until the plasma glucose is < 2.0 mmol/L (blood glucose < 1.7 mmol/L), on one occasion. If plasma glucose is <2.6 on several occasions (blood glucose <2.2), or for a prolonged period, we do not have enough information to suggest that it is safe, and further intervention and testing may be warranted. We need more studies like HypoEXIT, with even larger sample sizes; for such a common phenomenon it si vital that we figure this out, once and for all.
When considering the confusion between plasma and whole blood glucose, have you also considered the impact of the difference between capillary and venous levels of glucose. , as many of us take samples as we put an IV in. We use istat machines and the manufacturer suggests the difference could be up to 0.38 mmol/L lower in venous samples. This is clinically significant when we are talking discriminating need for treatment between 1.7-2.0.
This is an important issue as the hematocrit in a capillary sample is generally higher, increasing the error/difference between capillary and venous blood samples.
I understood that most modern POC glucometers se a correction factor so that although they measure whole blood glucose, they report the « plasma glucose equivalent ». But most machines will under correct by using a factor of ~1.2 based on average adult hemaotocrit (see Clinical Chemistry DOrazio Clin-Chem-2005-51-1573-1576 51:9 1573–1576 (2005)). Neonates generally have a higher hematocrit so there will still be some error. Does anybody do further correction in practice? There is at least one POC machine (?ACCU-CHEK Aviva) available which measures the Hct with the whole blood glucose in order to give the appropriate correction in order to report the PGL. However I believe that nearly all POC machines are liable to error.
Pingback: Not COVID-19. Who is at risk for hypoglycaemia? | Neonatal Research
Pingback: Glucose screening and treatment in the newborn; what now? | Neonatal Research
Pingback: We need to rethink neonatal hypoglycaemia | Neonatal Research