My colleague Dr. Marks and I spent a great deal of time and energy working on the recently published CPS statement on mangement of hypoglycemia which I shared in the post Its Here. New Advice From The CPS On Managing Hypoglycemia For Newborns At Risk. Shortly after the release of this statement I became aware of a soon to be released trial called the HypoEXIT trial. This trial has now been published in the New England Journal of Medicine with some fanfare; Lower versus Traditional Treatment Threshold for Neonatal Hypoglycemia.

Why Did We Need This Study?

For some time a recommendation has existed to keep blood sugars at or above 2.6 mmol/L at least for the first 48-72 hours. Afterwards the threshold as per our recommedation is to increase the target to 3.3 mmol/L or above. As the authors of the paper above pointed out, the evidence for these thresholds impacting long term neurodevelopmental outcome in a positive way is shaky and therefore they set out to examine two different thresholds of less than 2.0 mmol/L vs the traditional less than 2.6 mmol/L in a non-inferiority trial. They used a fairly standard Bayley test at 18 months to determine if there were any difference in outcomes between groups. I gather this had been presented in abstract form earlier and many were eagerly awaiting the publication as the scuttlebut was that there was no difference seen which would favour a change in practice to accept the lower threshold. As with many studies though the devil is in the details and did this study actually address this question properly?

The Answer

To be fair, I think the authors here had a great goal in mind. What if we have been targeting the wrong threshold? How many babies would be needlessly tested and retested for glucose levels that really don’t make much of a difference to outcome? Before delving into the results though it is important to really look at what population here was studied. In order to qualify for this prospective study you had to be at risk of hypoglycemia defined by these four subgroups; Late-preterm infants (gestational age from 35 to 37 weeks), newborns who were small (below the 10th percentile) or large (above the 90th percentile) for gestational age, and infants of mothers with diabetes. Not included here are mothers who were on medications such as labetolol and those whose mothers had hypertensive disorders and were small but not small enough as examples. Importantly, none of the infants in the study could have an initial blood glucose < 1.9 mmol/L as they would be classified as severe. So we are looking at patients at risk who have moderate hypoglycemia. Already the field has been narrowed a fair bit. The next thing you had to have to get in the study if you were at risk was a blood glucose from 3-24 hours of age <2.6 mmol/L. Those kids were then randomized to either use the lower or higher thresholds for determining intervention. The authors determined to show non-inferiority they would need 200 babies in each group of at risk newborns to prove no-difference.

What Did They Find?

The authors managed to recruit 689 infants out of 5958 potential infants that were considered. The limits they put above to create moderate risk excluded many so we are talking about 10% of those at risk with hypoglcyemia who entered the trial. The findings were as follows; “Bayley-III-NL scores were assessed in 287 of the 348 children (82.5%) in the lower-threshold group and in 295 of the 341 children (86.5%) in the traditional-threshold group. Cognitive and motor outcome scores were similar in the two groups (mean scores [±SE], 102.9±0.7 [cognitive] and 104.6±0.7 [motor] in the lower-threshold group and 102.2±0.7 [cognitive] and 104.9±0.7 [motor] in the traditional-threshold group). The prespecified inferiority limit was not crossed.”

So they didn’t find a difference. When you look into why though it becomes clear this was a tough study to really help answer the question. Here are the problems.

  1. They didn’t enroll the 800 babies they needed for their power calculation. Failing to find a difference may not mean as much if they didn’t get enough infants. Out of the 689 they also didn’t follow up everyone so in the end they only had 582 infants.
  2. The second issue has to do with the setting of the two thresholds. If you look at glucose at randomization they were almost equal at 41.4 mg/dL vs 41.2 mg/dL. In table three the big issue with the study becomes evident.

These kids were so low risk for recurrent hypoglycemia by excluding the more severe patients at entry that very few regardless of threshold actually became hypoglcyemic again. Roughly half of both groups never had another low. Of those that had another low blood glucose, 70% in the low group and 82% in the high group had either no or only one other episode after entry in the study. Looking at the previous work of Allan Lucas who demonstrated five or more low readings could impact long term outcomes, only 9% in the low and 2% of the high group had another low blood sugar,

3. Also from table 3 one can see that the number of patients with true severe hypoglcyemia which might be expected to impact long term outcome was also low at only 10% and 5% of the study population.

4. The authors did note in the secondary outcomes that there was a much higher need for IV therapy 21% in the high vs 6% in the low group. Moreover, bolus glucose was 12% vs 6%. This study though recruited patients before the adoption of glucose gel so we have to take the benefit here with a grain of salt since use of glucose gel would in the modern era lessen the effect.

5. Is the Bayley III really what we should be looking at to make such a decision about changing thresholds? As in the study by McKinlay et al, Association of Neonatal Glycemia With Neurodevelopmental Outcomes at 4.5 Years, outcomes later in life would be more meaningful.

Finally would you put your money where your mouth is?

Looking at this from a practical standpoint, in this at risk population, the vast majority would have at most one more episode of hypoglcyemia after randomization. You can’t avoid the needle poke to recheck a low glucose and with glucose gel the need for IV therapy is less. The question I guess is whether you would be confident enough in these results to be find for the first 48 hours with your own child having blood sugars of 2.1. 2.0, 2.2 etc. I suspect if many people were forced to answer that question with their own child they might based on this study alone feel more comfortable with a supplemental feed +/- glucose gel.

I suspect there is more to be told regarding this topic but for the time being I don’t think this one study is enough to move the needle and change our approach to hypoglcyemia thresholds.