• Neonatology
• Infant Feeding

In 1989, I commenced a research project to investigate the processes of metabolic adaptation after birth. One aspect was to investigate the endocrine and metabolic responses to neonatal hypoglycaemia.1 ,2 When notified out of hours of admission of a ‘hypoglycaemic’ baby to a neonatal unit I would rush from meals and family commitments, access stores of dry ice in a remote outhouse of the hospital, and conduct fiddly centrifuging, pipetting and snap freezing of tiny blood samples. The study sample size was perhaps larger than it should have been as the diagnostic cut-off of 2.6 mmol/l had found its way into use (more of that below). Many babies on my arrival seemed surprisingly well and were physically resisting all efforts to site nasogastric tubes and intravenous cannulae, and were returning to the sender the volumes of formula milk that were given to them with the best of intentions. Imagine my dismay, especially on the part of the babies and their mothers, when laboratory measurements demonstrated that for many babies, the admission glucose was not confirmed as being in the hypoglycaemic range and that the initial near-patient testing strip had proved to be an inaccurate measurement (again more of that below). Or a baby admitted to the neonatal unit for another reason had a blood glucose measurement very shortly after birth, at the time of the universal nadir described below and on the basis of this single measurement the baby was labelled as hypoglycaemic, while if left to his own devices or with normal neonatal unit care, his blood glucose would have subsequently risen.

Further, as the study elapsed and following discussion with experts such as the late Marvin Cornblath, it became evident that many babies did not have a pathological condition, rather they were simply doing what babies do in the process of adaptation after birth, mounting a metabolic response to a physiological fall in blood glucose.3 ^– 5 For such babies, to ascribe a pathological diagnosis and commence invasive treatment for a normal postnatal process equates to sending 1-day-old babies for ligation of persistent ductus arterteriosus or carrying out exchange transfusions for moderate physiological jaundice.

The flood of litigation that this ‘pathology’ unleashed has been well recognised on both sides of the Atlantic, to the extent that Dr Cornblath in a lecture at Neonatal Hot Topics in Washington DC in 2000 noted that we had become human litogens—a litogen having previously been described as ‘a drug that does not cause malformations, but does cause lawsuits.’.6 He followed this by stating ‘There comes a time to give up your principles and do what's right!’

I will not rehearse in detail the transition between fetal and neonatal nutrition and metabolism; this is covered in many standard texts and reviews, including those referenced below. To summarise, almost all babies have a sharp fall in blood glucose in the hours after birth and then a slow rise from a nadir which occurs in the first 12 h. For some this is slower than others and then for some blood glucose levels dip again until full milk feeding is established (the intakes of breast milk are low even for healthy and hungry babies in the first 3 days). However as with other potentially precarious postnatal adaptations, there is a backup plan and the healthy baby responds to falls in blood glucose by mobilising and using alternative fuels to glucose. Thus, healthy babies are unlikely to manifest clinical signs or sustain brain injury during this physiological postnatal fall of blood glucose.

However, there are groups of babies who are at risk of more profound or prolonged fall in blood glucose, or more importantly a failure to mount the normal metabolic responses to this fall. At-risk groups include babies with hyperinsulinism (eg, after poor control of maternal diabetes in pregnancy), those with intrauterine growth restriction, preterm babies and those with other pathologies such as infection or inborn errors of metabolism. Untreated, the low blood glucose levels in the absence of alternative fuels to glucose will cause clinical signs and, in the extreme case, brain injury.

So the first challenge, in terms of definition, is the description and diagnosis of the pathological condition. The description of the condition we are concerned about is not ‘a low blood glucose measurement’. Rather, the description should be ‘impaired metabolic adaptation.’ As described above, looking only at blood glucose levels will ascribe a pathological diagnosis and inappropriate interventions to too many babies. Normal metabolic adaptation is a diagnosis of exclusion and those assessing a baby with a low blood glucose measurement must assure themselves of an absence of risk factors and clinical signs.

If we agree to the more wordy but meaningful description, how do we make a diagnosis, taking into account that the various alternative fuels to glucose cannot easily be measured? The process must first of all identify those babies at risk of failure or delay in metabolic adaptation, and then monitor blood glucose levels and clinical signs. Or for a baby not previously thought to be at risk but who presents with any concerning clinical signs, to ensure prompt and accurate blood glucose measurement is performed.

Clinical experience and data from studies and case reports indicate that when neonatal hypoglycaemia results in clinical signs or brain injury, the temporal evolution is as follows:

• low blood glucose levels are found but the baby does not have clinical signs as the baby is still able to draw on alternative fuel stores, for example, glycogen and fat. This could be defined as biochemical hypoglycaemia with adequate metabolic adaptation.

• if untreated, the baby exhausts alternative fuel stores and develops subtle clinical signs which are not specific only to hypoglycaemia (eg, irritability or lethargy or poor feeding) but hypoglycaemia is not damaging at this stage. This is the onset of impaired metabolic adaptation.

• if untreated, the baby develops obvious and severe clinical signs (eg, fits, coma) but may escape damage if treated very promptly. Metabolic adaptation has failed.

• if not treated sufficiently soon after onset of clinical signs, hypoglycaemia becomes damaging and in severe cases results in cardiorespiratory arrest.

The time period for this process will be highly variable. For example, with severe hyperinsulinism, metabolic adaptation is completely suppressed and the baby is very vulnerable to a sustained postnatal fall in blood glucose. For the moderately preterm baby who does not attain sufficient milk intake, the process (if left unmanaged) will be more gradual as the baby's reduced fat stores become exhausted. Therefore, the duration of low blood glucose cannot be woven into the definition. Assiduous clinical monitoring of those at risk and a low index of suspicion for a previously well baby presenting with unexpected clinical signs are key to management in order to prevent brain injury.

There have been swings of opinion and controversy over the years regarding the numerical definition of neonatal hypoglycaemia. This arises from the clinical scientific world setting itself an unrealistic task in an area which contains so many variables. Attempts to arrive at a single definition (usually to one decimal place in the context of devices which cannot measure accurately to ±0.5 mmol/l) have been reviewed elsewhere.4 ,7 The greatest concern was the widespread adoption of a single numerical value based on two published papers whose results are now generally considered not to justify the conclusion that a level of below 2.6 mmol/l should be used to define neonatal hypoglycaemia. As described above, this numerical definition generally leads to overdiagnosis and treatment, but more worryingly with inaccurate devices would give false reassurance regarding babies for whom 2.6 mmol/l is too low should metabolic adaptation fail. Despite widespread knowledge of the concern regarding inaccuracy of near-patient testing devices, their use remains widespread.

As a result of the controversy regarding numerical definition, a group of interested clinicians on both sides of the Atlantic evaluated the evidence base and, finding this lacking, wrote a consensus paper in 2000.4 Groups of experts since then have reviewed the evidence base in the light of more recent publications and have presented no arguments to move from the pragmatic consensus of 2000.5 ,8 ^– 10

The suggested blood glucose thresholds for intervention are:

• any single value <1.0 mmol/l

• baby with abnormal clinical signs—single value <2.5 mmol/l

• baby at risk of impaired metabolic adaptation but without physical signs—<2.0 mmol/l and remaining <2.0 mmol/l at next measurement

The paper sets higher therapeutic goals than the blood glucose levels for intervention, especially for babies though to have significant hyperinsulinism.

In summary, if ‘neonatal hypoglycaemia’ continues in use as a short hand diagnostic term it should be accurately defined as ‘a persistently low blood glucose level, measured with an accurate device, in a baby at risk of impaired metabolic adaptation but with no abnormal clinical signs; or a single low blood glucose level in a baby presenting with abnormal clinical signs.’ There has been no recent evidence to argue against the pragmatic operational threshold values described above.

We are trained as clinicians to make assessments and decisions to avoid harm arising from the underlying condition but also to avoid iatrogenic harm, such as the effects of separation of mother and baby.11 Numerical definitions of hypoglycaemia and use of algorithms cannot replace these clinical skills.