A “Simplified” DKA algorithm and its rationalization

DKA Algorithm Lucid Chart



Diabetic Ketoacidosis is a life-threatening condition with the possibility of cerebral edema that occurs in type 1 Diabetes Mellitus (T1DM) and occasionally in type 2 DM (T2DM). For many it is a difficult process to manage with many moving targets, time frames, and life-threatening consequences if mistakes are made. No wonder these people go to the ICU. They need 1:1 care. But they all start out in the ED and we HAVE to make this diagnosis. That’s why on this deep dive I want to go into the literature behind the algorithms. This way you will see why I wrote up my own Simplified DKA algorithm “A 3-pronged approach” (see above). The mantra governing this “sweet” DKA treatment algorithm is “DKA took a bit to happen, so you should take a bit to treat it” (aka don’t rush DKA!)

Just as with any diagnosis that has been around since the dawn of time, the literature for DKA is…well nearly non-existent. I mean it’s bad. Yes, there are a lot of little studies, but seriously shouldn’t there be some big RCT’s? I mean look at cardiology literature those guys have thousands of patients enrolled! It’s so bad that Tran1 in her article entitled “Review of Evidence for Adult Diabetic Ketoacidosis Management Protocols” concluded “there is a major deficiency of strong evidence for the optimal management of DKA” and that “all components of DKA management would benefit from prospective RCTs “. Tell us how you really feel, why don’t ya! Despite this paucity of the literature we will go through what we can.


Oh, geez are you going to tell me there is controversy in DIAGNOSING DKA? No…no, no, no…well yes.  As a side note, I love America but I was born in Canada, so every once in a while, I think “God save the queen” because occasionally those Brits and Canucks have a good idea. This might just be the case in the diagnostic criteria of DKA.  The table below, adapted from US guidelines 2,  shows how “we the people” define DKA:

DKA Diagnostic criteria

How do they do it in the UK 3? Well it’s literally a mnemonic of the letters DKA. Oh, come on! Dammit, that’s good! Even I can remember that its:

D – “Dextrose”: 200 mg/dL or previous history of DM

K – “Ketones”: Serum ketones >3 mmol/L or >2+ on urine ketone stick

A – “Acidosis”: pH <7.3, Bicarb <15 mmol/L, Gap (no cutoffs listed)

Well “Ello Gov’nor”! I hate to admit it but that’s good. Come one ADA! Just adopt that one… too easy! You may also notice that this definition is a bit more conservative and uses a cutoff lower than that of the U.S (250 mg/dL). Obviously, this increases capture rate (sensitivity) of DKA and may also enhance identification of euglycemic DKA.  Euglycemic DKA is thought to occur in 10% of patients with DKA3. This can occur with certain medications such as the sodium-glucose cotransporter inhibitor, Invokana [canagliflozin] 4.  This can also occur in pregnancy. Pregnancy is a state of insulin resistance thus insulin requirement progressively rises throughout pregnancy explaining the higher incidence of DKA in the second and third trimesters 5. In addition to euglycemia other conditions that make DKA difficult to diagnose include:

  1. Conditions that increase serum bicarbonate (e.g. vomiting)
    1. Can cause a mixed acid/base disorder so pH is not as low
  2. Causes of significant osmotic diuresis (excessive protein intake)
    1. Loss of ketoacids may lead to a normal gap
  3. AKA which can cause elevated β-hydroxybutyrate level
    1. See section on differential diagnosis.

Take home point:  Screen all patients with point of care glucose testing >200 mg/dL for DKA. Know the definition of DKA. Consider euglycemia and other co-morbidities that make DKA more difficult to diagnose.


Obviously, there are a lot of different ways to do the same thing.  It would be great to have a non-invasive method of screening for DKA. End-tidal CO2 has been postulated as one such way to rule out DKA. Sadly, there just isn’t enough literature yet. One such trial of 21 DKA patients, showed that an ETCO2 > 35 mmHg was 100% sensitive. In the table below I’ve listed the studies (an “half-asked” look at the literature on the topic, so I may have missed some). If you look at that table there really aren’t any consistencies at all in the value of ETCO2 used, the correlation with bicarb or pH and so on. I’d like to see a large study that finds a value of ETCO2 that is 100% sensitive. Possibly, with shared decision making, using a value greater than 36 mmHg might be reasonable, assuming easy access to ETCO2. This may be an option but without larger trials I’m not sure this idea is ready hang my hat on!

Screen Shot 2018-07-01 at 4.35.17 PMGETTING GAS-SY: ARTERIAL BLOOD GAS (ABG) VS. VENOUS (VBG)

So, if a pH is required to make the diagnosis, then we probably need a blood gas. Can we do a VBG instead of an ABG? Afterall, it just can’t feel good poking a needle into a radial artery. The US guidelines6 allow for venous pH in those “without heart, lung, or kidney disease”.  How good is the correlation between venous and arterial pH? It seems pretty darn good. There are multiple studies that all show similar findings. This 2014 systematic review of 1747 acidotic patients from 15 studies showed a that the venous pH was only 0.03 (95%: ±0.004), lower than the arterial pH7. This observational prospective trial of 200 DKA patients showed similar results with a venous pH being 0.015 (95%: ±0.006) lower than arterial pH8. The same study also showed that the arterial pH (when drawn simultaneously to venous) only changed the treatment or disposition in 5/200 patients or 2.5% of the time8.

Unfortunately, the remaining components aren’t that great but do you really need those components to manage DKA? Probably not. I will say that this systematic review by Bloom found in 392 patients a venous pCO2 of < 45 mmHg had a 100% sensitivity and negative predictive value for an arterial pCO2 to be not greater than 50 mmHg 9. That is, if the vbg shows normocapnia then so will the abg.

Take home point: The venous pH will be 0.01 lower than the arterial pH and a pCO2 of < 45 mm Hg will have a normal arterial pCO2


            This is probably is probably the most contentious area of DKA. Its the age old balanced fluids (e.g LR) vs normal saline (NS) debate. I’m guessing the answer is probably that the fluid type doesn’t matter that much. For now, I think the discussion of what fluid to use in the critically ill patient will be reserved for another blog. However, there are a few caveats and quid pro quo’s that I’d like to discuss briefly (“You keep using that word. I do not think it means what you think it means …”).

            As Tran points out in her review the ideal solution to use is “unclear” and most of the studies in DKA are pretty small1. Probably the biggest concern with using Normal Saline is a hyperchloremic metabolic acidosis. Additionally, there is a likely sub-clinical change in renal function that has yet to result in any meaningful clinical outcome such as ICU days or dialysis days. Recently, however, there has been a large trial published specifically looking at normal saline in DKA. This landmark NEJM trial by Kupperman compared 0.9%NS to 0.45% NS in 1389 events of DKA 10. While he found no clinical differences in the primary end point (neurologic outcomes) he did find some statistically significant electrolyte differences in the form of hypophosphatemia, hyperchloremia and hypocalcemia. However, none of these abnormalities changed the speed of recovery from DKA as measured by time to resolution of the DKA (~14 hrs) or time to discharge (~46 hrs). For this reason, in my “3-pronged approach” to DKA, I elected to go with 0.45% NS for everything except boluses since there is no difference in recovery time and less laboratory abnormalities.

Take home point: Use NS to bolus then switch to 0.45%NS after. There does not appear to be any benefit at this time of “balanced fluids” vs 0.45%NS


            Typically, the total body water (TBW) deficit in DKA is thought to be around a 10% loss and is, therefore, the volume replaced. However, I think this is also up for debate. In 1989 Adrogue11 looked at “large volumes” (1000 ml/hr over 4 hours) vs “small volumes” (500 ml/hr over 4 hours) of fluid replaced. They found NO difference in resolution of DKA but did show less cost, less fluid infused and less hyperchloremia with the smaller volumes.  Similarly, Kupperman10 also looked at large (10% TBW loss) vs small (5%TBW loss) replaced fluid volumes. They too found less hyperchloremia in smaller volumes and no differences in outcomes or resolution of DKA. For this reason, in my “3-pronged approach” to DKA I elected to go with bolusing 10 ml/kg as well as calculating and replacing only a 5% TBW deficit since there is no difference in recovery time, while allowing for a benefit of less cost and less laboratory abnormalities. Obviously, the patient in shock will require appropriate fluid boluses to whatever endpoint of resuscitation the clinician uses. However, this is not the case for most DKA.

Take home point: Replace less fluid unless there is SHOCK. Use a TBW deficit of 5% and use 10 ml/kg boluses initially unless there is SHOCK.


            As is the case in most of medicine we have persistent dogma that has more to do with “one time I saw this happen so that must be the case” versus a true cause-effect relationship. While the concern for cerebral edema is real in DKA, it is unlikely a result of the rate of fluid administration. In fact, this isn’t a new finding. In 2002 Azzopardi12 looked at 10 adults with DKA and treated them with isotonic or hypotonic solutions and then performed a CT scan before treatment and then at 12 hrs, 24 hrs and 6 months after initiation of treatment. They did not find clinical cerebral edema in either group. In 2013 Glaser13 showed that in 10 children randomized to get slow or fast infusions of fluid. They performed diffusion weighted MRI during and after treatment to see if there was vasogenic cerebral edema. They showed no MRI differences in children with DKA, regardless of rate of infusion. In 2018, Kupperman10 showed no difference in neurologic outcomes in 1389 events of DKA whether they were randomized to one of four groups with varying rates of infusion and normal vs half-normal saline.

            So what are some risk factors for the progression in children to cerebral edema? Glaser and Kupperman 14 found in 2001 that children with DKA vs random controls are at higher risk for cerebral edema if they have a low paCO2, a high BUN or are treated with bicarbonate.

Take home point: Cerebral edema is likely associated with bicarbonate and lower paCO2. The rate and fluid are probably not a cause of cerebral edema but I’m sure using less fluid overall is a good idea.


            You will notice in my Simplified DKA algorithm “A 3-pronged approach” there are two commonly included elements missing. The first is the purposeful omission of bicarbonate supplementation (the second is omission of phosphate replacement). We just said from above that bicarb treatment is a risk factor for cerebral edema. Subsequently, there are those that will make the argument that in very low pH <6.9 bicarbonate may have a role. The ADA guidelines2 suggest “Because severe acidosis may lead to [sic] a numerous adverse vascular effects, it is recommended that adult patients with a pH 6.9…receive…sodium bicarbonate…until the venous pH is 7.0…”. I disagree with this, but don’t just take my word for it. The UK guidelines3 also recommend against the use of bicarbonate replacement “with the rationale that fluid and insulin replacement alone will be sufficient to raise pH”. This is also demonstrated in several studies. This study looked at 147 pediatric DKA patients including 9 with a pH less than 7 and one with a pH of 6.73! Not only did they find no benefit, they found longer hospitalizations in the group given bicarb. In this 1986 RCT of 20 adults with a pH from 6.9 to 7.14, bicarbonate therapy did not affect the speed of recovery in DKA. There seems to be a theme here and that’s enough for me. I might have to wear a redcoat from now on when I’m treating DKA but it seems worth it!

Take home point: There is no role for bicarbonate therapy in DKA even with a pH <7.0.


            As mentioned above phosphate is the other element that is absent from my Simplified DKA algorithm “A 3-pronged approach”. The ADA guideline recommends, in my opinion, a complex phosphate replacement. However, they only recommend it in patients with cardiac dysfunction, anemia, respiratory depression, or patients with phosphate levels <3.2mmol/L 2. The review by Tran states that “Prospective randomised studies demonstrated that phosphate replacement offers no improvement to DKA outcomes1. The NICE guidelines recommend to “not generally use phosphate replacement in the management of DKA in adults.15”.

Take home point: It’s highly unlikely you will need to replace phosphate. If you do, it should only be when <3.2 mmol/L or if respiratory depression. There is plenty of time to look this up and do it later.


            Notice how much we’ve discussed and HAVEN’T even spoken of insulin yet? Despite total-body potassium depletion, mild-to-moderate hyperkalemia is common in patients with hyperglycemic crises2. Since understanding this is critical I want to break this down into simple terms. This will be anything but a biochemistry lecture. Once you give insulin therapy you help get rid of the ketones and thus the acid in the blood. Less acid means you are further driving potassium into the cells. [Remember in hyperkalemia we give bicarb, a base (aka less acid) to drive K into the cells, same principal] If the K is already super low (<3.3) you can’t give insulin yet so you have to hold ‘the base” (insulin) so you don’t push more K into the cell and cause a dangerously low potassium. Therefore, small amounts of K are added until the K is >3.3. If the k is 3.3 to 5.3 then all you need to do is to maintain that k balance so add 20-30 mEq to each liter of maintenance fluid. If K is >5.3 then it will come down on its own just by treating the acidosis. PEARL: Make sure urinary flow is appropriate in DKA otherwise you can’t urinate out the K that is moving around.

Take home point: Insulin therapy will LOWER the serum potassium. Don’t give insulin if its already low because you will cause a dangerously low K. Try to keep the K between 3.3 and 5.3. Monitor for good urine output


So lastly, (like I said “no rush”) is insulin. I don’t think there is much to say about this. Regular insulin is used because of its low cost. A bolus does not change time to resolution of DKA, length of hospital stay, hypokalemia, or other complications including death 1.  In 2008, a small prospective randomized trial found that an initial bolus of insulin avoided the need for supplemental insulin doses if the insulin infusion rate was at least 0.14 units/ kg/h 16.


            With any case of a metabolic acidosis one should keep a broad differential of its causes. One differential to consider is alcoholic ketoacidosis (AKA). Sometimes the two can be difficult to distinguish. This study compared 12 patients with DKA and 12 with AKA. They found on average the following serum value similarities in DKA vs. AKA (respectively):  bicarb [11 vs 11 mEql/L], venous pH [7.18 vs 7.28], anion gap (30 vs 30), Sodium (134 vs 139 mEq/L), beta-hydroxybutyrate (7 vs 6 mmol/L). They found on average the following serum value differences in DKA vs. AKA (respectively): glucose (578 vs 118), acetoacetate (2.6 vs 1.1), beta-hydroxybutyrate to acetoacetate ratio (3:1 vs 7:1), and lactate (1.6 vs 3 mmol/L) 17. Thus, the biggest differentiator was the ketone ratio of 3:1 in DKA and 7:1 in AKA and not the presence of beta-hydroxybutyrate.


            That’s sums up the understanding and evidence of DKA. Any protocol you use is better than none1 so make sure you have a protocol easily available and read it over BEFORE you have a case of DKA. Until next time…

DKA Algorithm Lucid Chart






  1. Tran, T. T. T. et al. Review of Evidence for Adult Diabetic Ketoacidosis Management Protocols. Front Endocrinol (Lausanne) 8, 106 (2017).
  2. Kitabchi, A. E., Umpierrez, G. E., Miles, J. M. & Fisher, J. N. Hyperglycemic crises in adult patients with diabetes. Diabetes Care 32, 1335-1343 (2009).
  3. Dhatariya, K. K. & Vellanki, P. Treatment of Diabetic Ketoacidosis (DKA)/Hyperglycemic Hyperosmolar State (HHS): Novel Advances in the Management of Hyperglycemic Crises (UK Versus USA). Curr Diab Rep 17, 33 (2017).
  4. Kelmenson, D. A. et al. Euglycemic diabetic ketoacidosis with prolonged glucosuria associated with the sodium-glucose cotransporter-2 canagliflozin. Journal of investigative medicine high impact case reports 5, 2324709617712736 (2017).
  5. Kamalakannan, D. Diabetic ketoacidosis in pregnancy. Postgraduate Medical Journal 79, 454-457 (2003).
  6. Kitabchi, A. E. et al. Management of hyperglycemic crises in patients with diabetes. Diabetes Care 24, 131-153 (2001).
  7. Byrne, A. L. et al. Peripheral venous and arterial blood gas analysis in adults: are they comparable? A systematic review and meta-analysis. Respirology 19, 168-175 (2014).
  8. Ma, O. J., Rush, M. D., Godfrey, M. M. & Gaddis, G. Arterial blood gas results rarely influence emergency physician management of patients with suspected diabetic ketoacidosis. Academic Emergency Medicine 10, 836-841 (2003).
  9. Bloom, B. M., Grundlingh, J., Bestwick, J. P. & Harris, T. The role of venous blood gas in the emergency department: a systematic review and meta-analysis. Eur J Emerg Med 21, 81-88 (2014).
  10. Kuppermann, N. et al. Clinical Trial of Fluid Infusion Rates for Pediatric Diabetic Ketoacidosis. N Engl J Med 378, 2275-2287 (2018).
  11. Adrogué, H. J., Barrero, J. & Eknoyan, G. Salutary effects of modest fluid replacement in the treatment of adults with diabetic ketoacidosis: use in patients without extreme volume deficit. Jama 262, 2108-2113 (1989).
  12. Azzopardi, J., Gatt, A., Zammit, A. & Alberti, G. Lack of evidence of cerebral oedema in adults treated for diabetic ketoacidosis with fluids of different tonicity. Diabetes Res Clin Pract 57, 87-92 (2002).
  13. Glaser, N. S. et al. Subclinical cerebral edema in children with diabetic ketoacidosis randomized to 2 different rehydration protocols. Pediatrics 131, e73-80 (2013).
  14. Glaser, N. 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 344, 264-269 (2001).
  15. Amiel, S. A., Pursey, N., Higgins, B., Dawoud, D. & Guideline, D. G. Diagnosis and management of type 1 diabetes in adults: summary of updated NICE guidance. BMJ 351, h4188 (2015).
  16. Kitabchi, A. E., Murphy, M. B., Spencer, J., Matteri, R. & Karas, J. Is a priming dose of insulin necessary in a low-dose insulin protocol for the treatment of diabetic ketoacidosis. Diabetes Care 31, 2081-2085 (2008).
  17. Umpierrez, G. E. et al. Differences in metabolic and hormonal milieu in diabetic- and alcohol-induced ketoacidosis. J Crit Care 15, 52-59 (2000).

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.