Wednesday, May 7, 2014

Four DKA Pearls


I have a confession to make: I love treating DKA.   It’s satisfying to take a patient from severe acidosis, electrolytic disarray, and hypovolemia to normal physiology during an ICU shift.   Although it's usually straightforward, there are some pitfalls and a few tricks that may help your patients improve faster.  

Pearl #1: Avoid normal saline

A common phenomenon observed when starting a DKA resuscitation with normal saline (NS) is worsening of the patient’s acidosis with decreasing bicarbonate levels (example below).   This occurs despite an improvement in the anion gap, and is explained by a hyperchloremic metabolic acidosis caused by bolusing with NS.   This could be a real problem for a patient whose initial bicarbonate level is extremely low.1   A while ago I made the switch from NS to lactated ringers (LR) for resuscitation of DKA patients, and have not observed this phenomenon when using LR.  

Example of the effect of normal saline resuscitation during the initial phase of DKA resuscitation.  This patient received approximately 3 liters normal saline between admission labs and the next set of labs as well as an insulin infusion, all textbook management per American Diabetes Association guidelines.   The anion gap decreased from 33 mEq/L to 30 mEq/L, indicating improvement of ketoacidosis.   However, the bicarbonate decreased from 8 mEq/L to 5 mEq/L due to a hyperchloremic metabolic acidosis caused by the normal saline.    Note the increase in chloride over four hours.   Failure of the potassium to decrease significantly despite insulin infusion may reflect potassium shifting out of the cells in response to the hyperchloremic metabolic acidosis.

There is only one randomized controlled trial comparing NS to LR for resuscitation in DKA (Zyl et al, 2011).   These authors found a trend towards faster improvement in pH when using LR compared to NS (p = 0.076).   They also found that patients in the NS group experienced a decrease in average serum bicarbonate during the first hour of treatment (from 8.86 to 8.21 mEq/L), whereas patients in the LR group experienced an increase in average serum bicarbonate during the first hour of treatment (from 7.71 mEq/L to 8.83 mEq/L).   Although the authors concluded that this was a negative study, the data suggests an advantage of using LR in correcting the acidosis.  

There is better data supporting the use of Plasmalyte in DKA.   Plasmalyte is a balanced crystalloid with nearly the same pH effects as LR (more about Plasmalyte here).   Rinaldo Bellomo’s research group in Australia performed a retrospective analysis of patients with DKA who received predominantly NS or Plasmalyte (Chua et al 2012).   Patients who received plasmalyte had more rapid improvement in their serum bicarbonate.   There was no difference in the strong ion gap (an index of ketoacids, similar to the anion gap), indicating that the higher bicarbonate in the plasmalyte group was due to avoidance of a NS-induced hyperchloremic acidosis.   Mahler et al 2011 performed a prospective randomized controlled study of NS vs. Plasmalyte which also showed higher bicarbonate levels and less hyperchloremia in the Plasmalyte group.   Note that there are other proprietary crystalloids available with a composition nearly identical to Plasmalyte (i.e. Normosol).

Bottom line?   Normal saline induces a hyperchloremic acidosis which drops bicarbonate levels in the initial phase of DKA resuscitation, and is probably not the ideal fluid to use.   LR, Plasmalyte, or Normosol are better choices.2  

Pearl #2: Use bicarbonate to treat hyperchloremic, non-anion-gap acidosis.

The usual debate about bicarbonate is regarding the initial treatment of the patient when the pH is extremely low.   This has already been debated extensively and I’m not going to go there.

DKA patients often develop a non-anion-gap, hyperchloremic metabolic acidosis.   This may occur due to gradual development of DKA with urinary excretion of ketoacid (which then cannot be converted to bicarbonate) and/or initial resuscitation with NS.   In these cases the anion gap closes but the bicarbonate level remains frustratingly low.   

Why should we care?   Some evidence suggests that hyperchloremic acidosis may cause renal vasoconstriction and worsen renal function.    Many DKA protocols suggest delaying initiation of subcutaneous insulin until the bicarbonate is above 18 mEq/L, so a hyperchloremic acidosis may delay transition off the insulin infusion.   My bias is that patients feel better and breathe easier when their bicarbonate is normalized so they don’t need to maintain a compensatory respiratory alkalosis.   

Predicted Final Bicarb level = Bicarbonate level + Ketone level

Ketone level Anion gap – 10 = (Na – Cl – Bicarb) - 10

Predicted Final Bicarb Level  Na – Cl – 10

During a DKA resuscitation, keep an eye on the predicted final bicarb level as calculated above.   This is an estimate of where the bicarbonate level will end up once the anion gap closes and all ketoacid is converted to bicarbonate.   If the predicted final bicarb level is significantly low (i.e., <18 mEq/L), the patient is likely to wind up with a hyperchloremic acidosis at the end of the resuscitation.   In such cases, I think it’s reasonable gradually administer some isotonic bicarbonate toward the end of the resuscitation.3   Bicarbonate may delay clearance of ketoacidosis, so it’s best to do this when the ketoacidosis is nearly resolved.    To estimate the amount of bicarbonate needed, the predicted final bicarb level may be used to calculate the bicarbonate deficit.   For more discussion of this strategy of pH-guided fluid resuscitation, see the last blog.

Gradually infusing some isotonic bicarbonate toward the end of the resuscitation is generally safe.   If the isotonic bicarbonate is constituted in D5W, it may cause a mild increase in the glucose level.   Occasionally this increase in glucose may cause the insulin drip to be titrated up, which can be beneficial in terms of speeding the resolution of ketoacidosis.  However, care should be taken to avoid hypoglycemia if the insulin drip remains at a high rate.   Bicarbonate may decrease the potassium level due to intracellular shifting, but this generally isn’t an issue as potassium is already being monitored and aggressively repleted.  

To be painfully clear, I’m not advocating for the use of bicarbonate to treat ketoacidosis, but rather to treat non-anion-gap, hyperchloremic acidosis if this is significant.   The benefits of treating a hyperchloremic acidosis are debatable.   If you want the patient to have a normal anion gap and normal bicarbonate level at the end of your shift, the above approach is effective.  

Pearl #3: Avoid intubation if possible.  

When approaching a critically ill patient, securing the airway is often an initial consideration.   However, in DKA this is fraught with hazard and often destabilizes the patient.   With the exception of a patient who has truly developed respiratory muscle fatigue (and lost the ability to generate a compensatory respiratory alkalosis), intubation will typically worsen the patient.  

First, the act of intubation is dangerous.   Sick DKA patients typically have extremely low bicarbonate levels, for which they are compensating with a respiratory alkalosis (for example, consider a patient with bicarbonate of 2 mEq/L, PaCO2 15 mm, and pH 6.75).   If there is difficulty intubating this patient and the PaCO2 rises to, say, 60mm then the pH will fall to 6.15.   Approaches to mitigate this risk include attempting to mask-ventilate the patient in the peri-intubation period or awake intubation.   It should be noted that DKA patients may develop gastroparesis and are at risk for aspiration.   In summary, even if the patient has an anatomically normal airway, the physiology of DKA makes this a dangerous procedure.

Second, if the patient was generating a reasonable compensatory alkalosis then the PaCO2  is usually higher on the ventilator than prior to intubation.   The ET tube adds resistance to the respiratory circuit.   If the patient is passive on the ventilator, the ventilator only provides active inhalation with passive exhalation (compared to the patient prior to intubation who was actively inhaling and exhaling).   In practice it is generally impossible to achieve the same level of respiratory alkalosis on the ventilator that a strong non-intubated patient can generate.  

For a patient with mental status alteration due to the DKA, it may be best to avoid intubation as long as the patient is protecting their airway (noting that traditional criteria to evaluate airway protection such as absence of a gag reflex or GCS scoring are not supported by evidence).4   If mental status changes are due to DKA, improvement often occurs rapidly.   Ultimately this is a clinical decision which must be made at the bedside by an experienced physician.   If there is doubt, close observation and serial evaluation may be helpful.

Pearl #4: Continue long-acting basal insulin throughout the DKA resuscitation.

Many DKA patients already have an established home medication regimen involving basal long-acting insulin (i.e., insulin glargine or insulin detemir).   The traditional approach to managing DKA has been to hold the long-acting insulin until the patient has recovered and subsequently to overlap this with the insulin infusion for a few hours.    Recent British guidelines suggest continuing the patient’s long-acting insulin throughout treatment of DKA.   For example, if the patient has been noncompliant, a full daily dose of long-acting insulin could be given immediately upon admission to the hospital.   Administration of long-acting insulin doesn’t obviate the need for an insulin infusion, which should be administered as usual.  

Continuing the long-acting insulin has three advantages.   First, it protects against the classic error of shutting off the insulin infusion due to hypoglycemia, with subsequent worsening of the ketoacidosis.   If the insulin infusion gets turned off in a patient who has received their long-acting insulin, the patient will still have some basal insulin on board and their ketoacidosis probably won’t worsen.   Second, it speeds transition from the insulin infusion to subcutaneous insulin: there is no longer a need to extend the infusion in order to overlap with the subcutaneous insulin.   Finally, it reduces rebound hyperglycemia when the insulin infusion is stopped. 


Current DKA protocols using normal saline and delayed initiation of long-acting insulin are proven to work well.   With the possible exception of patients whose admission bicarbonate level is extremely low (who may be harmed by the initial drop in bicarbonate due to bolusing NS), the choice of fluid is unlikely to have any long-lasting effects.   However, newer approaches employing balanced crystalloids and continuation of long-acting insulin are likely to make patients improve more rapidly, reducing the time spent on insulin infusions and cutting healthcare costs.5  


(1) I have seen a few patients who presented with bicarbonate levels in the 7-8 mEq/L range, were bolused with normal saline, and subsequently had undetectably low bicarbonate levels on serum chemistry (the assay just reads “<5 mEq/L”).   These patients ultimately did fine, but they scared me.   I wonder if there are very occasional patients who present with profound acidosis and are pushed off the edge of the acid-base cliff by normal saline.  
            Another group of patients with DKA who could be significantly harmed by normal saline resuscitation are patients presenting with life-threatening hyperkalemia.   The inorganic metabolic acidosis induced by normal saline will tend to shift potassium out of cells, hindering efforts to reduce the potassium.   

(2) Addendum 1/26/15: A prior version of this blog included the statement that Plasmalyte has a SID of 49 and is therefore gently alkalinizing, possibly making it the ideal fluid selection for DKA.   I have subsequently realized that this is a myth (discussed here) - the sodium gluconate in Plasmalyte is incompletely metabolized and therefore Plasmalyte is not nearly as alkalinizing as predicted.   Plasmalyte may be very slightly more alkalinizing than LR, but it is questionable whether this difference is clinically significant.   I have subsequently revised this blog post accordingly.   Please accept my apologies.   Plasmalyte remains a good selection of fluid for resuscitation of DKA.   

(3) Isotonic bicarbonate is typically obtained by adding 150 mEq of bicarbonate (usually 150 ml of 8.4% bicarbonate) to a liter of D5W or sterile water.   In the US, bicarbonate typically comes in ampules of 50ml volume of 8.4% bicarbonate, so isotonic bicarbonate is constituted by adding three such ampules per liter.   

(4) With regards to the gag reflex, I agree with the Ron Wall’s Manual of Emergency Airway Management and MackWay-Jones et al 1999 that it is not useful.   Recent evidence (i.e. Duncan R et al 2009) has dispelled the dogma that a GCS of 8 or less mandates intubation in poisoned patients.   Given that DKA patients and intoxicated patients both suffer from short-term toxic/metabolic encephalopathy, this finding likely applies to DKA patients as well.

(5) Honestly I don’t really care much about health care costs, I’m more interested in getting patients better faster.   However, it’s hard to change our ways in the healthcare system, especially when it comes to protocols which have been around for a long time.   Perhaps if administrators and insurance companies realize that new approaches would save them money, this may spur change.   I would bet that a DKA protocol using only Normosol/Plasmalyte for volume resuscitation (with use of D10W infusions to provide dextrose and complete avoidance of normal saline) as well as initiation of long-acting insulin immediately on admission would reduce hospital and ICU length of stay, reducing costs.   


  1. Great article! I love my "DKAer's" also! Nothing more satisfying than fixing a messed-up pt all in one shift! Look forward to seeing what else you have to offer! Thanks!

  2. Thank you! Glad the blog was helpful.

  3. "The usual debate about bicarbonate is regarding the initial treatment of the patient when the pH is extremely low. This has already been debated extensively and I’m not going to go there".

    Can you give me a reference or article on this statement? Trying to grasp Bicarbonate infusions and when they are helpful/non-helpful. My understanding is that bicarbonate works by converting to CO2 in the liver and your body needs to blow of the additional CO2 --> therefore not as helpful unless patient is able to breath at a faster rate. Please help me out a bit on this.



  4. Great post as usual Josh!

    George, think about bicarb as an innocent bystander. ALl it does is maintain electroneutrality. The actual "ingredient" is the sodium, without chloride. We have a lot of H20, a lot of CO2 and carbonic anhydrase. We have an endless supply of "bicarb." The addition of Na without Cl (a very high SID!) is what corrects the hyperchloremic acidosis. Look up Peter Stewart and John Kellum's work on this. Google will lead you to Kellum's site which breaks Stewart's work down. In my opinion, infusions are only needed when you abused NS in the first place (SID zero solution).


  5. Sodium bicarb will increase extracellular pH while increasing intracellular pH--likely doesn't matter if you give bicarbonate or not. Never has been shown to be helpful in human studies (ie animal studies show benefit) -- it does decrease inotropy and vasoconstriction but I haven't found anything in 'humans' that showed it to decrease pressors, change EF, etc. Moreover, acidosis is likely beneficial in sick patients. I get that <6.8 can be a reason to worry but it shouldn't be b/c of the pH it should be b/c you need to reverse the cause of the low pH. sick --> low pH (not vice versa). I would argue that bicarbonate in DKA with hyperCl (that is otherwise HD stable---and most likely uneven if they are not )is likely wasteful and I haven't seen any literature to support more RRT in patients with hyperCl. Pearl 2 to me is a case of Euboxia. Again, pH is a marker of illness, not the cause of illness -- we overTx this 'marker' (as we do lactate) far too much.

    I do like the idea of adding a long acting insulin to my initial DKA mgmt. I'm actually go to start doing this from now on. Thanks for another great post!

    Here's my articles on bicarbonate therapy if anyone is interested:

  6. Does the above blog post apply to children too?
    The paeds DKA literature still seems to be clinging to NS, with its isotonicity being cited as important. Should we be using LR/Plasmalyte instead of NS in kids with DKA?

    1. Sorry, but this blog does not apply to pediatrics. I know nothing about pediatrics. This is a blog about controversies in adult critical care. From what I hear pediatric DKA is much more complicated than adult DKA and I would have a low threshold to consult a pediatric endocrinologist, sorry I can't be of further help.

  7. Houshyar. Effectiveness of Insulin Glargine on Recovery of Patients with Diabetic Ketoacidosis: A Randomized Controlled Trial. J Clin Diagn Res 2015;9(5):OC01–OC05

    Anecdotal, I have used this on a few patients since reading the above and your blog -- great results.!