Recently there have been numerous publications and discussions about whether VBGs can replace ABGs in DKA. The growing consensus is that VBGs are indeed adequate. Eliminating painful, time-consuming arterial blood draws is a huge step in the right direction. However, the ABG vs. VBG debate overlooks a larger point: neither ABG nor VBG measurements are usually helpful.
It is widely recommended to routinely obtain an ABG or VBG, for example by both American and British guidelines. Why? Is it helping our patients, or is it something that we do out of a sense of habit or obligation?
Diagnosis of DKA: Blood gas doesn’t help
These are the diagnostic criteria for DKA from the America Diabetes Association. They utilize either pH or bicarbonate in a redundant fashion to quantify the severity of acidosis. It is unclear what independent information the pH adds beyond what is provided by the bicarbonate.
Practically speaking, the blood gas doesn’t help diagnose DKA. This diagnosis should be based on analysis of the metabolic derangements in the acid-base status (e.g. anion gap, beta-hydroxybutyrate level). The addition of a blood gas to serum chemistries only adds information about the respiratory status, which does not help determine if the patient has ketoacidosis.
Management: Does the pH help?
It is debatable whether knowing or attempting to directly “treat” the pH is helpful. The pH will often be very low, usually lower than would be expected by looking at the patient. This may induce panic. However, it is actually a useful reminder that acidemia itself doesn't necessarily cause instability (e.g. healthy young rowers may experience lactic acidosis with a pH <7 during athletic exertion; Volianitis 2001).
A question often arises regarding whether bicarbonate should be used for patients with severe acidemia. There is no evidence to support for this practice, which may be harmful (British Guidelines)(1). For profoundly acidemic patients, my approach is occasionally to intensify the treatment of the DKA: give more fluid, more dextrose, and increase the rate of the insulin infusion (2). This makes more sense to me, because sicker patients may be more insulin-resistant. However, there isn't evidence to support this approach either. Overall, it's possible that the best approach would be simply not to check the pH.
Suppose that you would like to know the severity of the acidemia, perhaps to determine patient disposition (e.g. ward vs. ICU)? pH correlates fairly well with the serum bicarbonate. Thus, as discussed above, the serum bicarbonate can be used as a convenient index of acidosis severity.
Management: Does the pCO2 help?
Metabolic acidosis should induce a compensatory respiratory alkalosis (hyperventilation). In severe DKA, there may be concern that the patient will tire out and lose the ability to hyperventilate. Such fatigue would lead to profound acidemia.0
The most useful piece of information contained in a blood gas is probably the pCO2, which shows how well ventilation is compensating for the metabolic acidosis. This might theoretically allow detection of patients with inadequate respiratory compensation, who might theoretically benefit from ventilatory support. Realistically, how much will pCO2 affect management? This hinges entirely on the management strategy for supporting ventilation in DKA.
Management strategy for ventilatory support in DKA
Admitting that this is a data-free zone, high-flow nasal cannula (HFNC) is usually my preferred approach to noninvasive support in DKA (3). HFNC may reduce the work of breathing by decreasing the respiratory dead space and thereby improving the efficiency of ventilation. This may be a reasonable option in patients with severe acidosis (e.g. bicarbonate <10 mEq/L) when there is a concern that the patient may develop diaphragmatic fatigue over time.
When in doubt about whether the patient should receive HFNC vs. no support, there is little to be lost from a trial of HFNC. Patients who are struggling may feel better with the HFNC, whereas patients who don’t really need the support may be irritated by it.
It is generally wise to avoid intubation in DKA if possible. Otherwise, the decision to intubate in DKA is similar to the decision to intubate in general. This is primarily a clinical decision based on physical examination, clinical scenario, and anticipated course. Patients usually require intubation for issues other than the DKA itself (e.g. airway protection prior to emergent surgery).
Checking VBGs: Searching for a unicorn?
Above is my general approach to determining the appropriate level of ventilatory support (4). Usually a VBG is unnecessary. The only real utility of a VBG would be detection of occult ventilatory failure (referred to here as a unicorn). Specifically, a “unicorn” would be a patient who:
- has severe metabolic acidosis (e.g. bicarbonate <10 mEq/L)
- is failing to mount a compensatory respiratory alkalosis (e.g. pCO2 > 30 mm)
- does not look so ill that they obviously require intubation
- based upon consideration of the blood gas, does require intubation
Despite years of searching I haven't yet found a unicorn. The closest I've come were a few patients who received opioids for abdominal pain, which blunted their compensatory respiratory alkalosis (creating a scary blood gas, but they were fine). Nevertheless, I plan to continue to search for unicorns by using VBGs in selected patients when there is a concern that intubation may be needed.
Rock Bottom Principle of DKA: Even if checking the pCO2 does alter management, this is unlikely to improve the outcome
It is currently impossible to know whether checking pCO2 improves outcomes in DKA. The data simply doesn't exist. We can search for patients whose pCO2 is higher than we would like, and attempt to decrease their pCO2 with various modes of ventilatory support. Are we improving these patients' outcomes, or would they recover regardless of ventilatory support?
In general DKA patients almost always improve with appropriate treatment, often rapidly. Thus, nearly all patients will be their sickest (rock bottom) upon arrival. This means that if the patient is initially clinically OK and an abnormality in the VBG is “missed,” it probably doesn't matter. Within hours they will very likely be doing better and their blood gas will have improved (5).
This is not intended to belittle the value of careful clinical monitoring (e.g. trends in vital signs, examination, and chemistry panel). Occasionally underlying conditions may be missed or patients may require more aggressive therapy to improve. The key is probably the clinical trajectory: whether the patient is improving is more important than any particular VBG.
Evidence: Is blood gas analysis beneficial?
Ma et al. 2003, Acad Emerg Med: Arterial blood gas results rarely influence emergency physician management of patients with suspected diabetic ketoacidosis
.This was a prospective observational study involving 195 patients with suspected DKA (based on hyperglycemia, ketonuria, and clinical features). Treating physicians wrote down their management and disposition plans without knowing the ABG results. Changes based on the ABG were as follows:
Knowing the ABG altered the diagnosis in two cases. However, in no case did the ABG affect the diagnosis of whether or not the patient had DKA. Note that an ABG shouldn't be needed to make a diagnosis of pneumonia: history, physical exam, pulse oximetry, chest radiography, and thoracic ultrasonography are likely more helpful.
In five cases, unexpectedly low or high pH affected the intensity of insulin therapy or disposition. However, this same information was actually contained in the bicarbonate concentration: patients with “unexpectedly low” pH had severe metabolic acidosis (bicarbonate <10 mEq/L) whereas patients with “unexpectedly high” pH had mild metabolic acidosis (bicarbonate >18 mEq/L). Thus, greater attention to the bicarbonate concentration may substitute for the pH.
In only one DKA patient did an ABG lead to changes in respiratory therapy (borderline hypoxemia with pO2 of 62mm prompted initiation of supplemental oxygen). A second patient was found to be hypercapneic due to obesity hypoventilation syndrome, but notably this patient did not actually have DKA.
Overall, in this series of 195 patients with suspected DKA (of whom 48 had DKA), routine ABG had a very little impact on patient management. ABG added little information that would not have been available from thoughtful evaluation including serum chemistries, history and physical examination, and additional tests as clinically warranted (e.g. chest X-ray, thoracic ultrasonography).
Hutten-Cazpski 2013, Can J Rural Med: How can you manage DKA without an ABG?
This letter describes one physician's transition from a teaching hospital to a small rural hospital without an ABG analyzer. Initially he was upset, but he quickly realized that he could manage just fine without blood gas analysis. Although blood gas analysis has been widely accepted in medicine, there is little evidence that it improves outcomes. It is often possible to manage patients based on their clinical examination instead of their ABG: might this actually improve care?
Perhaps experience taught me to pay better attention to the respiratory rate, urinary ketones, or the gestalt of how the patient looked, and I became a better doctor for it.
Don't be an ABG wanker
In honor of the late Dr. John Hinds, an ABG wanker is someone who focuses excessively on the blood gas, often while ignoring the patient or common sense. For example, an ABG wanker might order q2hr ABGs for a DKA patient to monitor their respiratory status, rather than looking in on them occasionally to examine them and ask how they feel (6). Delaying clearly necessary interventions such as transfer or intubation to obtain an ABG is another common offense. Although ABG wankers are primarily fixated in ABGs, a similar phenomenon can occur with VBGs.
- Guidelines recommend checking an ABG or VBG in all patients with DKA. This practice is not evidence-based and should be abandoned.
- ABG or VBG provides little information about whether or not the patient has DKA (beyond what is already known from the serum chemistries).
- Rather than pH, serum bicarbonate may be used to gauge the severity of acidosis.
- There is no evidence that detecting or reacting to a very low pH is helpful.
- Decisions about the level of ventilatory support that a patient needs can almost always be made on a clinical basis. When in doubt, close attention to the patient with serial examination is often a sound approach. Focus on the patient, not the blood gas.
- VBG might be helpful in cases where it is unclear whether the patient requires intubation, or if there is a significant underlying respiratory disease (e.g. COPD or obesity hypoventilation syndrome)..
- I have previously recommended the use of isotonic bicarbonate towards the end of a DKA resuscitation, with a goal of gently correcting a residual non-anion-gap metabolic acidosis if one is present. This is explained further here (pearl #2). This practice should not be confused with the use of bicarbonate as the response to a scary low pH in the beginning of a DKA resuscitation, with a goal of somehow masking an anion-gap metabolic acidosis.
- Volume depletion will cause the body to compensate by producing cortisol and catecholamines, which in turn reduce the sensitivity to insulin. Thus, allowing a patient to remain under-resuscitated may delay their recovery from DKA. These patients often have severe volume depletion and might benefit from aggressive goal-directed therapy (e.g. volume resuscitation until the IVC is reasonably full on ultrasonography). With regards to insulin therapy, there isn't anything magical about 0.1 mg/kg/hr of insulin infusion – perhaps a patient with profound acidosis would benefit from a higher rate? Overall, instead of blindly administering bicarbonate it may be better to think hard about the patient, what might be driving their DKA, and how their DKA management can be optimized.
- Why use HFNC and not BiPAP? There is no evidence to support the use of either in DKA. BiPAP is fine, particularly in a patient with underlying COPD. My preference is usually HFNC for two reasons. First, patients often have diabetic gastroparesis and/or gastroenteritis and may be at increased risk of vomiting, a concern with BiPAP. Second, it may be harder to access the patient’s respiratory and neurologic status with BiPAP on.
- Please note that this is purely a management algorithm. A patient with dyspnea may also merit diagnostic evaluation with a chest radiograph, ultrasound, etc.
- There are some noteworthy exceptions to the rock bottom principle. First, if the DKA is due to a rapidly progressive underlying problem (e.g. septic shock), deterioration could occur despite perfect management. The second, and by far the most common exception, is resuscitation with large volumes of normal saline. If a patient presents with severe acidemia, bolusing them with several liters of normal saline will worsen their acidemia. Resuscitation with a balanced crystalloid should avoid this problem.
- In the spirit of full disclosure, I have done similarly foolish things during my training. Thus, I do not believe that being an ABG wanker is necessarily a permanent character flaw, but instead it can be a temporary phase that many practitioners pass through as they grow (eventually, after obtaining lots of ABGs, it becomes clear when they are actually useful).