Dani Hall. Diabetic Ketoacidosis, Don't Forget the Bubbles, 2020. Available at:
Maisie is 2 years old. Apart from a few coughs and colds, she is usually a very well, happy little girl. She’s been a bit poorly for the last 48 hours – a bit off colour, off her food, lethargic and just not her usual cheeky self.
She’s been drinking though and has had good wet nappies. In triage she has a runny nose and slight cough. She is pretty tachycardic and tachypnoeic and doesn’t look well and so she’s moved to majors.
Maisie is put on a monitor and immediately you can see that her respiratory rate is elevated at 40 breaths per minute with saturations of 97% in air. Her heart rate is also elevated at 150 beats per minute with a normal blood pressure of 105/65. Her capillary refill time is 3 seconds peripherally and she is afebrile.
Her heart sounds are normal, her chest is clear and her abdomen is soft although mildly tender throughout.
The only objective thing you have is the tachypnoea with a bit of a runny nose. You wonder if she has viral induced wheeze and is just too tight for the wheeze to be audible so you prescribe salbutamol and review after 10 minutes. But that’s made no difference to her respiratory rate, and her chest is still completely clear.
Things just don’t add up. She’s holding her tummy – perhaps her tachypnoea and tachycardia are secondary to pain. Her abdomen remains soft with no guarding, but she doesn’t like you palpating it. Could this be appendicitis? Or even worse an intussusception? You speak to the paediatric surgeon who asks you to cannulate Maisie and send some blood. They’ll be down to review her shortly.
You cannulate Maisie and take a venous gas. The results seem to take an age. And your heart sinks when you see them…
She’s acidotic at 7.15 and it looks metabolic with a bicarbonate of 13.9 and base deficit of minus 8.7. Her lactate is 2.9 and her glucose is very high at 29.5. You run a drop of Maisie’s blood through the bedside ketone monitor. Her blood beta-hydroxybutyrate is. 5.1.
This is diabetic ketoacidosis.
DKA can be really difficult to diagnose in toddlers
Classically children with DKA present with polyuria and polydipsia with abdominal pain, nausea, and vomiting. This can progress to dehydration, weakness, and in severe cases, they may be lethargic. Blood tests show a raised white cell count as a physiological response to the raised stress hormones, cortisol and catecholamines, and so are not a reliable indicator of infection. But, and this is a big but, an infection can precipitate DKA so it is important it is considered.
Although ketoacidosis may result in a classic ‘pear drop/acetone’ smell to the breath, not everyone has the chemoreceptors to detect it.
Common misdiagnoses include dehydration secondary to infection or respiratory presentations. Ketoacids stimulate the respiratory centre resulting in rapid fast breathing – Kussmaul breathing – blowing off carbon dioxide to compensate for the metabolic acidosis. It may also present as an acute abdomen as abdominal pain and ileus can result from hypokalaemia, acidosis, and poor gut perfusion. If opioids are given for pain this can suppress the Kussmaul breathing leading to worsening acidosis
The BSPED (2020), and ISPAD definitions of DKA are:
acidotic with a bicarbonate of <15 mmol/l or a pH <7.3 and ketones of >3.0 mmol per litre
However… just when you starting thinking it was easy… children with known diabetes may develop DKA with normal glucose and so you must keep a high index of suspicion and check pH and ketones in an unwell child with diabetes.
Lack of insulin leads to rising blood glucose levels in the bloodstream. However, this glucose is not transported into the cells and so the body needs to produce an alternative energy source for cellular activity. Three processes occur:
- muscle is broken down to mobilise amino acids which are then used to create glucose (catabolism)
- fat is broken down to produce glycol and ketones (lipolysis)
- the liver uses lactate, glycol and amino acids to create more glucose (gluconeogenesis)
High ketone levels lead to metabolic acidosis.
Hyperglycaemia leads to glucose spilling into the urine (glycosuria). These glucose molecules exert an osmotic pull, dragging water, cations, and anions such as phosphate, potassium, and sodium into the urine (osmotic diuresis). The child becomes dehydrated with physiologically low levels of potassium and phosphate. Ketones also spill into the urine (ketonuria) in preference to chloride, which is retained in the plasma, leading to a worsening chloride-driven acidosis.
Maisie is dehydrated because of the osmotic diuresis exerted by the glucose in her urine. And she’s acidotic because of the ketones circulating in her bloodstream. The question is, what are we going to do about this?
The complications of DKA have incredibly high mortality and morbidity, so we’re going to start here.
Cerebral oedema occurs in approximately 1% of children with DKA. While relatively rare, it can have devastating consequences with mortality of approximately 25%. In fact, more than half of all diabetes-related deaths in children are caused by cerebral injury.
Cerebral oedema usually occurs within the first four to 12 hours of starting treatment for DKA, suggesting that it’s the treatment itself that precipitates cerebral oedema.
Risk factors for cerebral oedema in DKA can be split into two groups.
The first group is characterised by children who have a longer duration of symptoms and are therefore more severely dehydrated at presentation. Younger children, particularly toddlers, and children who present in DKA without a previous diagnosis of diabetes mellitus, fall into this group, most likely because they have been in DKA for a long period of time before the diagnosis is made.
The second group is children who develop cerebral oedema because of the treatment they have received. Giving insulin within the first hour of treatment increases the risk of cerebral oedema – the theory is that the usually inactive sodium-hydrogen ion exchange pump is activated by the double hit of high intracellular hydrogen ions early in treatment while children are more acidotic plus insulin crossing the leaky blood-brain-barrier. The exchange pump transports sodium into the intracellular fluid, which then drags water with it due to its osmotic effect, leading to cerebral oedema.
Giving bicarbonate also increases the risk of cerebral oedema. The physiological mechanism of this is unclear but there is some thought that giving bicarbonate to correct acidosis can worsen tissue hypoxia due to effects on 2,3-DPG in erythrocytes (remember acidosis shifts the oxygen dissociation curve to the right, increasing the affinity of haemoglobin and oxygen) or that giving bicarbonate may lead to preferential movement of carbon dioxide across the blood-brain barrier, both of which will promote acidosis and poor oxygen offload in the CSF. However, whatever the cause, it is clear from a systematic review published in 2011 by Chua et al that giving bicarbonate to children with DKA is linked with increased rates of cerebral oedema. The guidance, therefore, mandates that bicarbonate should not be used routinely to correct acidosis. Fluids and insulin will do that by improving skin perfusion and reducing ketosis. Only give bicarbonate if the acidosis is resulting in reduced cardiac function, and then give very carefully…
So, with that in mind, how are we going to treat Maisie?
The initial management of a child with DKA follows the principles laid out in APLS: ABC resuscitation.
If a child is obtunded and not protecting their own airway then they should be intubated because of the risk of airway obstruction. However, intubation in DKA is risky… both sedation and the resultant hypercarbia can cause cerebral herniation. Central lines are also risky in these children because of the increased risk of thrombosis. Only use them if absolutely necessary and remove them as soon as possible.
Luckily, Maisie is maintaining her own airway, her GCS is 15 and she is not obtunded. The airway is not a problem for her.
After managing the airway and breathing we move onto circulation. So the question is: should we give Maisie a fluid bolus?
This is a big question. We are taught that children with cardiovascular compromise should receive fluid boluses to support their circulation. But assessing cardiovascular compromise in children with DKA can be very challenging. Clinical evaluation of hydration and shock is very difficult in children with DKA. Acidosis drives tachycardia and reduces peripheral skin perfusion.
Koves et al set out to look at this by studying a group of 37 children under 18 presenting with DKA.
Emergency Department doctors recorded heart rate, respiratory rate, blood pressure, cool peripheries, capillary refill time, skin turgor, the presence or absence of sunken eyes and dry mucous membranes to provide a clinical estimate of dehydration. A second emergency department doctor, blinded to the clinical interpretations of the primary doctor, was asked to review the patient before treatment and record their assessment of the same clinical variables. There was a good clinical correlation between the two assessments. Following admission, the children’s weights were measured daily until discharge and percentage dehydration was calculated from the weight gain from admission to discharge.
There was no agreement between assessed and measured dehydration. There was a tendency to overestimate dehydration in children with <6% measured dehydration and underestimate in children >6% dehydrated.
It’s a tricky business and these same parameters clearly won’t be of use in estimating shock in these children.
A true assessment of shock in DKA should rely on assessment on blood pressure measurements and peripheral pulse volume. So that doesn’t really help us. Maisie’s blood pressure and pulse volumes are normal so she’s not shocked. But she clearly is dehydrated.
BSPED (2020) uses pH and bicarbonate to classify the severity of DKA:
- pH 7.2–7.29 or bicarbonate <15 mmol/l is mild DKA with 5% dehydration
- pH 7.1–7.19 or bicarbonate <10 mmol/l is moderate DKA with 7% dehydration
- pH < 7.1 or bicarbonate <5 mmol/l is severe DKA with 10% dehydration
So, are we going to give her a fluid bolus? Let’s turn to the guidelines…
BSPED 2020 states:
Any child in DKA presenting with shock (as per the APLS definition of tachycardia and prolonged capillary refill time) should receive a 20 ml/kg bolus of 0.9% saline over 15 minutes. Let’s call this a ‘resuscitation bolus’.
Further 10 ml/kg boluses may be given if required up to a total of 40 ml/kg. Then add inotropes if the child remains shocked.
Boluses given to treat shock should NOT be subtracted from the calculated fluid deficit.
All children with DKA, whether mild, moderate or severe, who require IV fluids should receive an initial 10 ml/kg bolus over 60 minutes. Let’s call this a ‘rehydration bolus’. This bolus SHOULD be subtracted from the calculated fluid deficit.
The American Academy of Pediatrics, agrees that all children with DKA should have a bolus of 10ml/kg over 30 minutes to an hour. If a child is critically unwell with hypovolaemic shock, then additional boluses of 20ml/kg of 0.9% saline should be given.
Australian guidelines vary depending on region – from no routine fluid boluses to 10-20 ml/kg 0.9% saline for the sickest. Some say subtract fluid boluses from rehydration calculations, others don’t. There is no clear consensus.
So why is there so much international variation?
Traditionally, we have been warned about the danger of causing cerebral oedema in children with DKA by giving them too much fluid, reducing serum osmolality and literally flooding the brain. This is based, on the most part, by an old paper that showed an association between large volume fluid resuscitation in DKA and cerebral oedema. Note the word association, not causation.
Fluid management in DKA
Dogma has been to restrict fluids in paediatric DKA. It is widely thought that the rapid administration of intravenous fluids reduces serum osmolality, resulting in cerebral oedema. Guidelines traditionally have, therefore, advised slow fluid replacement using isotonic fluids as using hypotonic fluids was thought to cause further drops in osmolality.
And the evidence seemed to support this. Retrospective reviews showed better outcomes in children with DKA who received less fluid.
- only an association had been demonstrated, not causality.
- and it is reasonable to suspect a confounder in that those with more severe DKA could be expected to be both at higher risk of cerebral oedema and more likely to receive large volumes of fluid resuscitation based on their clinical presentation.
And then along came this paper by Kupperman et al published in the New England Journal of Medicine in 2018, which has shifted thinking a bit, as well as causing some controversy…
Lead authors Nate Kupperman and Nicole Glaser suggested the causal effect of fluid resuscitation and cerebral oedema was a myth in Glaser’s 2001 retrospective case-control study that gave us the list of risk factors for cerebral oedema in DKA.
Kupperman’s team wanted to look specifically at the relationship between fluids and cerebral oedema (defined in the study as a drop in GCS, or longer-term evidence of neurological injury defined as a drop in IQ or short-term memory difficulties 2-6 months later) in DKA in children. They looked at 1255 children with DKA presenting to 13 hospitals in the States over a 9 year period, which, because 101 children presented twice, equated to 1389 episodes of DKA. Children were excluded if their GCS was less than 12, or if they had already received significant DKA management prior to assessment. 289 were withdrawn by the treating physician. The mean age was 11. It’s important to think about all of this as these exclusion criteria mean that the very sick and the very young, two groups who are at significantly increased risk of cerebral oedema, were probably lost in this cohort.
Children were randomized into 4 groups. All patients in both groups received IV insulin at 0.1u/kg/hr. Dextrose was added to the saline solution when blood glucose dropped to 11.1 to 16.7 mmol/l.
Children were randomized into 4 groups:
- FAST rehydration with 0.45% sodium chloride
- FAST rehydration with 0.9% sodium chloride
- SLOW rehydration with 0.45% sodium chloride
- SLOW rehydration with 0.9% sodium chloride
In short, Kupperman’s team found no difference between the groups. There was no significant difference in GCS, or longer-term evidence of neurological injury. The endpoint that many of us are most concerned about, clinically apparent brain injury (deterioration in neurological status requiring hyperosmolar therapy or endotracheal intubation or resulting in death) was a secondary outcome, presumably due to its rarity and hence difficulty in showing statistically significant differences between groups. But again, there was no significant difference between groups.
There was a 0.9% rate of brain injury overall and it didn’t matter which type of fluids or how fast. Patients were more likely to get hyperchloraemic acidosis in the 0.9% NaCl group but this is of debatable clinical significance.
The evidence does not seem to support our traditionally cautious approach to DKA. The speed of IV fluids does not seem to be the cause of brain injury in DKA. But… and this is a big but… don’t forget the youngest and sickest patients weren’t included. All we can probably really conclude is that children who are not in the at-risk group for cerebral oedema are probably more resilient to higher volumes of fluids delivered at faster rates.
Ok… back to Maisie. How are we going to manage her fluids
Well, again it depends where in the world Maisie presents.
BSPED 2020 advises to calculate maintenance fluids the same way as they’re normally calculated for children in the UK:
- 100 ml/kg/day for the first 10kg
- plus 50ml/kg/day for each kg between 10 and 20kg
- plus 20ml/kg/day for each kg above 20kg
(A maximum weight of 80kg should be used for fluid calculations)
The International Society for Pediatric and Adolescent Diabetes guidance is as follows:
- Shock is rare in DKA but if present should be treated with 20ml/kg fluid boluses, repeated as necessary to achieve tissue perfusion.
- Give all children a 10ml/kg bolus over an hour to rehydrate them.
- Calculate maintenance fluids in the normal way using the simplified Holliday-Segar formula.
- Replace rehydration fluids over 24-48 hours, using clinical signs of dehydration to estimate the degree of dehydration. 2 or 3 signs would constitute to 5% dehydration, more signs would equate to 7% dehydration and weak pulses, hypotension or oliguria would indicate the child is 10% dehydrated.
Once you’ve navigated the quagmire of fluid management in DKA, you need to think about adding electrolytes. Remember, glucose molecules in the urine exert an osmotic pull, dragging water, cations, and anions such as phosphate, potassium, and sodium into the urine: the child becomes dehydrated with physiologically low levels of the electrolytes potassium and phosphate.
Always assume whole body potassium depletion in DKA. This is compounded by the treatment you give which causes potassium to move intracellularly. Replace potassium as soon as the patient has urine outpatient and labs confirm the child is not hyperkalaemic.
An ECG will give you clues about clinically significant hypokalaemia:
- Prominent U waves (an extra positive deflection at the end of the T wave)
- Flat or biphasic T waves
- ST-segment depression
- Prolonged PR interval
Although phosphate is lost in the urine as part of the osmotic diuresis in DKA, prospective studies involving relatively small numbers of subjects and with limited statistical power have not shown clinical benefit from phosphate replacement. Administrating phosphate can be dangerous, by causing calcium levels to drop. However, symptomatic severe hypophosphataemia, when serum phosphate levels drop below 1 mg/dL with an ensuing metabolic encephalopathy or depressed cardiorespiratory function, can be dangerous, albeit very rare. A sensible approach is to monitor phosphate levels alongside regular potassium level checks and, if a child is hypophosphataemic and symptomatic, replace phosphate whilst carefully monitoring serum calcium levels.
Back to Maisie. We’ve managed her fluids according to our local guidelines. But how can we tell if we’ve got the balance right?
We can’t monitor urine output as Maisie is going to be polyuric anyway because of the osmotic effect of glycosuria. Her capillary refill time will be prolonged because she’s acidotic and therefore skin perfusion will be reduced.
And her serum sodium and osmolality won’t be reliable indicators of fluid balance because of the effect of plasma glucose on her electrolytes. On top of this, her kidneys will preferentially excrete chloride from any saline and potassium chloride over ketones so there’s limited value to monitoring the anion gap because it doesn’t differentiate between hyperchloraemia or ketones. Instead, we should measure Maisie’s corrected sodium.
Because of its osmotic effect, glucose drags water with it into the intravascular compartment diluting the other osmols – 1mmol rise in glucose will drop sodium and chloride by 1mmol/L. If Maisie’s glucose goes up by 1 the other osmols will go down by 1. If glucose goes down by 1 the other osmols will go up by 1.
The corrected sodium must rise with therapy at a rate of 0.5-1 mmol/h
- Falling corrected sodium means too much water gain: we’ve been overzealous the fluids.
- A rapidly rising corrected sodium means too much water loss: we’ve been too fluid restrictive.
The Evelina London South Thames Retrieval Service has a great corrected sodium calculator on their website. You plug in her numbers – her initial sodium was 148 with glucose of 29.5, giving her a corrected sodium of 157.6. A couple of hours have passed and her latest gas shows that her glucose has come down to 24.5 – great – and her sodium has improved slightly to 144. You press calculate…
… and your heart sinks as you see her second corrected sodium has fallen by 6 points to 151.6 as you know that the corrected sodium must rise with treatment.
You go back to Maisie’s bedside to review her.
Maisie has dropped her GCS to 12 (E3, V4, M5). This is incredibly worrying – her GCS was 15 when you last checked on her. You move her round to resus and ask your nurse to grab some hypertonic saline.
Clinical features of cerebral oedema
- Slowing heart rate
- Rising blood pressure
- Focal neurology such as cranial nerve palsies
- Falling oxygen saturations
- Change in neurological status including restlessness, irritability, drowsiness, confusion, incontinence
Treat cerebral oedema with either hypertonic saline or mannitol.
Calculate your dose of hypertonic saline or mannitol before you need it and know where it’s kept. If a child has an acute deterioration, treat it.
Mannitol is an osmotic diuretic and can be given at 0.5 – 1 g/kg over 10-15 minutes. The effects should be apparent after 15 minutes. Mannitol lasts about 2 hours and can be repeated at this point if needed.
Hypertonic saline is a good alternative to mannitol or can be used after mannitol if a second agent is needed.
Don’t forget other neuroprotective measures like elevating the head of the bed to 30 degrees and intubation if concern regarding airway protection.
If there’s no improvement in GCS, do a CT, but not until the child is stable. CT is used to identify any potential lesion that would warrant neurosurgery – intracranial haemorrhage, or a lesion that would warrant anticoagulation such as thrombosis.
Hypertonic saline or mannitol?
DeCourcey et al (2013) conducted a retrospective cohort study over a 10 year period to see whether the increase in the use of hypertonic saline had had any effect on mortality in DKA. They looked at over 43,000 children under the age of 19 with DKA presenting to 41 children’s hospitals in America and found that the use of hypertonic saline replaced mannitol as the most commonly used agent in many of the participating hospitals. Controversially, their data suggested that hypertonic saline may not have benefits over mannitol and may be associated with a higher mortality rate.
However, this does remain controversial, with a counter-argument published as a letter to the editor a few months later arguing that (1) the fact that mortality from cerebral oedema in DKA had decreased by 83% over the same time period that use of hypertonic saline had increased, along with (2) the fact that DeCourcey’s paper only found a statistically significant difference in mortality between hypertonic saline and mannitol once age and race were removed from analyses (two factors that, themselves, have a significant influence on mortality in DKA-related cerebral oedema), meant that we shouldn’t be rushing to conclude that hypertonic saline is less safe than mannitol in the treatment of cerebral oedema.
No guidelines are yet to recommend mannitol over hypertonic saline. This seems to be one of those situations where a prospective study is needed to really answer the question of whether mannitol is superior, or at least non-inferior to hypertonic saline.
Back to Maisie. You give Maisie 3ml / kg 3% saline over 15 minutes and are relieved to see her wake up. You decrease her fluid prescription and thankfully from that point on her corrected sodium starts to slowly rise.
Finally, you’re ready to give Maisie some insulin. Insulin will control Maisie’s glucose and switch off ketosis, therefore improving her acidosis. Some departments use 0.1 units/kg/hr and some use 0.05 units/kg/hr. The question is, what is the optimal dose?
Nasllasamy’s team set out to compare the efficacy and safety of low-dose and standard-dose insulin infusions. They randomized 50 children under the age of 13 with DKA presenting over a 12 month period to receive insulin infused at either 0.05 units/kg/hr or 0.1 units/kg/h. They found that the rate of decrease in blood glucose and time to resolution of acidosis were similar in each group. There was no statistical difference in complication rates of hypokalaemia, hypoglycaemia or cerebral oedema.
This study suggests that low dose insulin is non-inferior to standard-dose insulin in managing children with DKA. It’s important to note that this was a non-inferiority trial and a larger study, powered to show superiority would be helpful. However, many units have been using lower doses of insulin at 0.05 units/kg/hr safely for some time and this study supports the use of lower dose insulin.
BSPED 2020 states:
Insulin can be given at 0.05units/kg/hr or 0.1 units/kg/hr, although ‘0.05 units/kg/hr would probably be sufficient in most cases except perhaps severe DKA’.
Children under 5 years should be given 0.05 units/kg/hr.
And so, as Maisie is young and therefore in the higher risk group of children with DKA, you opt to start her on insulin at 0.05 units/kg/hr.
In children who are already on long-acting insulin, BSPED 2020 states that it should be continued, or if they are newly diagnosed, they advise to consider starting long-acting subcutaneous insulin alongside intravenous insulin.<
Manage high-risk children in PICU
- pH <7.1
- young (under 2s or under 5s depending which guideline you read)
- cardiovascular shock
- corrected sodium >150 or <130
- hyper or hypokalaemia
- altered conscious state
- glucose >50
Maisie has multiple risk factors: she’s young and she developed clinically apparent cerebral oedema. You admit her to PICU where she makes stable progress and is discharged home 4 days later on a subcutaneous insulin regime.
Glaser N, Barnett P, McCaslin I, et al. Risk factors for cerebral edema in children with diabetic ketoacidosis. The Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics. N Engl J Med 2001; 344:264
Chua HR, Schneider A and Bellomo R. Bicarbonate in diabetic ketoacidosis – a systematic review. Ann Intensive Care. 2011; 1:23
Koves IH et al. The Accuracy of Clinical Assessment of Dehydration During Diabetic Ketoacidosis in Childhood. Diabetes Care 2004:27(10);2485-2487
Kuppermann N et al. Clinical Trial of Fluid Infusion Rates for Pediatric Diabetic Ketoacidosis. N Engl J Med. 2018;378:2275-87
DeCourcey et al. Increasing use of hypertonic saline over mannitol in the treatment of symptomatic cerebral edema in pediatric diabetic ketoacidosis: an 11-year retrospective analysis of mortality. Pediatr Crit Care Med. 2013; 14(7):694-700
Tasker RC, Burns J. Hypertonic saline therapy for cerebral edema in diabetic ketoacidosis: no change yet, please. Pediatr Crit Care Med. 2014;15(3):284-285
Nallasamy K et al. Low-Dose vs Standard-Dose Insulin in Pediatric Diabetic Ketoacidosis. A Randomized Clinical Trial. JAMA Pediatr. 2014; 168(11): 999 – 1005
ISPAD Clinical Practice Consensus Guidelines 2018: Diabetic ketoacidodis and the hyperglycaemic hyperosmlar state. Pediatric Diabetes 2018; 19 (Suppl. 27): 155–177