Success of Insulin Infusion Transitions in Moderate to Severe Diabetic Ketoacidosis With Transition Anion Gap of Less Than or Equal to 12 mEq/L Versus Greater Than 12 mEq/L

Abstract

Background: Patients with diabetic ketoacidosis (DKA) are transitioned from intravenous (IV) to subcutaneous (SQ) insulin upon DKA resolution. Although an anion gap (AG) ≤12 mEq/L is recommended before transition to SQ insulin, there are limited data to support this threshold. Objective: To compare the rates of successful transitions to SQ insulin in patients with DKA with an AG ≤ 12 mEq/L versus > 12 mEq/L. Methods: Retrospective cohort study of adult critically ill patients with moderate to severe DKA between September 2019 and December 2022. The primary outcome was the success of insulin transition between patients transitioned with an AG ≤ 12 mEq/L and those transitioned with an AG > 12 mEq/L. Transition was considered successful if the AG did not increase above the value at transition at 24 hours and insulin infusion was not restarted. Secondary outcomes include the individual components of the primary outcome and ICU length of stay (LOS); safety outcomes included hypoglycemia and electrolyte derangements. Results: In total, 92 patients were included, with 43 patients transitioned at AG ≤ 12 mEq/L and 49 patients transitioned at AG > 12 mEq/L. Transition was unsuccessful in 3 patients (7%) with AG ≤ 12 mEq/L and 2 patients (4%) with AG > 12 mEq/L (P = .66). There was no difference in the incidence of the individual components of this outcome between groups or in safety outcomes. Conclusion: This retrospective study showed no difference in success of insulin transition between the groups. Larger studies are needed to evaluate the impact of treatment characteristics on transition success and patient outcomes.

Background

Diabetic ketoacidosis (DKA) is a hyperglycemic crisis for which intensive care is warranted.^1,2 DKA is characterized by uncontrolled hyperglycemia, metabolic acidosis, and increased ketone production, primarily due to depleted insulin stores. Compensatory responses include the release of counter regulatory hormones, such as glucagon, cortisol, and catecholamines. Consequently, through lipolysis, 2 ketone bodies, beta-hydroxybutyrate (BHB) and acetoacetate, contribute to the development of an anion gap (AG) metabolic acidosis. Moreover, excessive hyperglycemia leads to glycosuria and osmotic diuresis, inducing a state of hyperosmolarity, dehydration, and impaired renal function.^1,2

Due to these metabolic derangements, the management of DKA centers on fluid replacement, insulin therapy, and electrolyte supplementation. ¹ Fluid resuscitation should be initiated immediately using crystalloid fluids, with guidelines recommending correction of estimated deficits within the first 24 hours. Next, intravenous (IV) regular insulin is preferred over subcutaneous (SQ) insulin due to the ability to titrate based on blood glucose (BG) measurements. ³ Finally, due to severe depletion of electrolytes, potassium, magnesium, and phosphate should be closely monitored, with supplementation ordered as clinically indicated.

Upon resolution of DKA, patients are transitioned from IV insulin to SQ insulin. Guidelines from the American Diabetes Association (ADA) define resolution of DKA when BG falls below 200 mg/dL and 2 of the following parameters are met: serum bicarbonate ≥ 15 mEq/L, venous pH > 7.3, and AG ≤ 12 mEq/L. ¹ As both bicarbonate and AG are generally reported on a basic metabolic panel, this provides an accessible method of assessing for DKA resolution. Notably, the current literature demonstrates substantial variability defining DKA resolution, with some trials utilizing an AG ≤ 12 mEq/L, whereas others do not reference transition AG and utilize parameters such as BG, pH, and bicarbonate.^{4 -6} Clinicians may consider transitioning to SQ insulin when AG remains above 12 mEq/L as IV insulin infusions carry risks of hypoglycemia and electrolyte derangements, and may require an ICU level of care. However, no trials to date have analyzed the success rates of insulin transition or patient outcomes when patients are transitioned with an AG ≤ 12 mEq/L versus an AG > 12 mEq/L.

This study aims to compare the rates of successful transitions from IV to SQ insulin in patients with DKA with an AG ≤ 12 mEq/L versus an AG > 12 mEq/L. Additionally, this study evaluates patient-centered outcomes associated with treatment of DKA, including ICU and hospital LOS, hypoglycemia, and electrolyte abnormalities.

Methods

Study Design and Sample

This was a single-center, retrospective cohort study of patients admitted to the medical ICU of a 914-bed tertiary care institution in north Texas. The study was approved by and conducted in compliance with the health system’s Institutional Review Board. Trial data were collected by the study investigators. Data points between September 1, 2019 and December 31, 2022 were included.

Eligible patients were 18 years and older with a diagnosis of moderate to severe DKA, as defined by the ADA guidelines. ¹ Required laboratory parameters were an initial BG of 250 mg/dL or more, arterial pH below 7.24, serum bicarbonate below 15 mEq/L, AG over 12 mEq/L, and positive BHB. All AG values collected were corrected for the patients’ serum albumin and rounded to integer values. Additionally, initiation of IV insulin infusion was required for inclusion.

Patients were not eligible for inclusion if any of the following criteria were met: those younger than 18 years, transferred from an outside hospital with prior receipt of IV insulin, received IV insulin for a diagnosis other than DKA, received enteral nutrition prior to DKA resolution, history of end-stage renal disease (ESRD) on hemodialysis (HD) or with progression to renal replacement therapy during DKA treatment, lactic acid above 4 mmol/l for over 24 hours, persistent AG metabolic acidosis from toxic overdoses, transitioned from DKA IV insulin infusion order to a non-DKA insulin infusion order, transition from IV to SQ insulin without at least 1 hour of overlap, BG greater than 250 mg/dL at the time of transition, left against medical advice (AMA) within 24 hours of transition or prior to transfer from the ICU, not transitioned to SQ insulin or expired prior to transition, and pregnant patients.

Within this institution’s order set for DKA management, treatment goals include optimizing volume status, evaluating precipitating factors, and correcting hyperglycemia, ketoacidosis, and electrolyte abnormalities. Prior to starting IV insulin infusion, clinicians must order IV fluids and correct potassium to at least 3.3 mEq/L. The initial IV insulin rate is 0.1 units/kg/hour, with a “standard” insulin sensitivity factor (ISF) of 0.04. Compared to low ISF, standard ISF prompts more aggressive correction of BG values via the insulin calculator integrated into the electronic health record. Choice of ISF type is at the provider’s discretion; for patients that may be sensitive to insulin, the provider is prompted to consider using a low ISF of 0.02. Point-of-care BG readings are ordered hourly, with adjustments made to the IV insulin rate based on the sequential changes in BG. When patients’ BG falls below 400 mg/dL, the calculated insulin rate is obtained using the following formula: (BG – 60) ×ISF. Dextrose-containing fluids are ordered to be initiated once BG falls below 250 mg/dL. Electrolytes are monitored every 4 hours. Patients receive continuous IV insulin until DKA resolution occurs, at which point the patient is transitioned to SQ insulin utilizing a basal/bolus treatment approach. This order set describes DKA resolution as a combination of the 3 parameters: AG is closed, serum bicarbonate level is normalized, and BG is < 200 to 250 mg/dL. It should be noted that this protocol does not include specific cut-off values for AG and serum bicarbonate to signify DKA resolution and these are chosen at provider discretion.

Other treatment characteristics evaluated included use of steroids and specific fluid types used. Steroid use was defined as receipt of greater than or equal to prednisone 20 mg per day or an equivalent steroid dose during DKA treatment. Fluid type evaluated the use of 0.9% sodium chloride (NS), Lactated Ringer’s (LR), or other solutions for both initial boluses and maintenance fluids.

Outcomes

The primary outcome of this study was to evaluate the success of transition from IV to SQ insulin between patients at an AG ≤ 12 mEq/L compared to an AG > 12 mEq/L. AG was corrected for albumin levels throughout the study, using the following equation: corrected AG = AG + [2.5 × (4 – albumin in g/dL)]. Insulin transition was deemed unsuccessful if both of the following occurred: the AG at 24 hours post-transition increased to a value higher than the AG at the time of transition, and IV insulin infusion was re-initiated.

Secondary outcomes evaluated the incidence of the individual components of the primary outcomes between groups and ICU LOS. Additionally, regression analyses regarding the association of factors with AG re-opening and IV insulin re-initiation were included in secondary outcomes. Safety outcomes compared the incidence of hypoglycemia and electrolyte disturbances, specifically hypokalemia and hypophosphatemia. Hypoglycemia was defined as BG < 70 mg/dL, hypokalemia as potassium < 3.3 mEq/L, and hypophosphatemia as phosphorus < 1.5 mg/dL.

Statistical Analysis

Data entry was performed using Microsoft Excel. Descriptive statistics were used to summarize all demographic and clinical characteristics. Chi-square tests were performed for categorical data, and Wilcoxon rank sum test for continuous variables. Logistic regression analysis was used to evaluate the association between treatment variables and AG re-opening and insulin re-initiation. The Hosmer-Lemeshow goodness of fit test was used for regression analyses. Negative binomial generalized linear regression was used to evaluate the relationship of variables on ICU LOS. A P-value of less than.05 was considered to be statistically significant.

Results

Patients

A total of 236 patients were evaluated for inclusion in this study, and 144 patients were excluded (Figure 1). The most common reasons for exclusion were an inadequate overlap time between IV and SQ insulin, transition from a DKA protocol insulin infusion to a non-DKA insulin infusion, leaving AMA within 24 hours of transition or prior to transfer from ICU, and history of ESRD on HD or progression to dialysis during DKA treatment (Table 1).

Figure 1.

Patient selection.

Table 1.

Reasons for Patient Exclusion.

Exclusion criteria	Number of patients
Inadequate overlap time between IV and SQ insulin	43
Transition to non-DKA insulin infusion	41
Left hospital AMA within 24 h of admission	22
History of ESRD on HD or progression to HD during admission	19
Lactate > 4 mmol/l for 24 h	6
BG > 250 mg/dL on transition	4
Did not transition to SQ insulin	4
Pregnancy	2
Received insulin infusion for a diagnosis other than DKA	2
Transferred from outside hospital on insulin infusion	1

In the group with transition AG ≤ 12 mEq/L, the average AG at transition was 10.2 mEq/L. In the groups with transition AG > 12 mEq/L, the average transition AG was 14.5 mEq/L. There were no significant differences between baseline characteristics of the 2 groups (Table 2). Both groups had comparable acuity of illness based on predicted mortality and expected LOS. These severity of illness indicators were calculated utilizing a proprietary prediction model integrated within the institution’s electronic health record. Moreover, DKA severity was analyzed using initial lab values of AG, pH, and bicarbonate, per the ADA guidelines. A majority of patients in both groups met the criteria for severe DKA.

Table 2.

Baseline Characteristics.

Variable*	Transition AG ≤ 12 mEq/L N = 43	Transition AG > 12 mEq/L N = 49	P-value
Male gender, N (%)	24 (55.8)	28 (57.1)	1.00
Age, years (IQR)	41 (26, 57)	44 (29, 55)	.43
Total body weight, kg (IQR)	63.5 (56.3, 77.4)	74.5 (56.5, 88.0)	.14
Body mass index, kg/m2 (IQR)	22.7 (18.9, 27.3)	25.1 (21.6, 30.6)	.11
DM type 2, N (%)	21 (48.8)	28 (57.1)	.62
Hemoglobin A1c, % (IQR)	13.1 (10.6, 14)	12 (10.6, 13.7)	.40
History of CKD	2 (4.7)	1 (2.0)	.90
Predicted Mortality Score, % (IQR)	1.24 (0.5, 2.7)	1.54 (0.5, 6.1)	.23
Predicted LOS Score, days (IQR)	3.1 (2.9, 3.6)	3.2 (2.9, 4.4)	.49
Severe DKA, N (%)	27 (62.8)	27 (55.1)	.59

Note. AG = anion gap; CKD = chronic kidney disease; DKA = diabetic ketoacidosis; DM = diabetes mellitus; IQR = interquartile ratio; LOS = length of stay. *All continuous variables are expressed as the median value, followed by the IQR.

Treatment Characteristics and Transition Characteristics

Treatment characteristics included the initial insulin bolus and rate, ISF, fluid type, steroid use, and initial laboratory values (Table 3). There was a higher initial BHB in the transition AG ≤ 12 mEq/L group. There were no other significant differences between groups for these parameters. Patients in the transition AG ≤ 12 mEq/L group had a significantly lower transition BG than patients in the transition AG > 12 mEq/L (Table 4). AG at 24 hours after transition was also significantly lower in the transition AG ≤ 12 mEq/L group. The duration of insulin infusion was not different between groups.

Table 3.

Initial Treatment Characteristics.

Variable*	Transition AG ≤ 12 mEq/L N = 43	Transition AG > 12 mEq/L N = 49	P-value
Insulin bolus used, N (%)	13 (30.2)	9 (18.4)	.28
Initial insulin rate, units/hour (IQR)	6.8 (5.7, 7.7)	7.6 (5.7, 8.6)	.23
Low ISF, N (%)	26 (60.5)	33 (67.4)	.64
Bolus fluid type, N (%)	NS: 27 (62.8) LR: 16 (37.2)	NS: 28 (57.1) LR: 21 (42.9)	.74
Maintenance fluid type, N (%)	NS: 31 (72.1) LR: 10 (23.3) Other: 2 (4.7)	NS: 29 (59.2) LR: 17 (34.7) Other: 3 (6.1)	.43
Steroid use, N (%)	0 (0)	4 (8.2)	.16
Initial BG, mg/dL (IQR)	688 (463, 869)	733 (541, 917)	.21
Initial pH (IQR)	7.05 (6.94, 7.11)	7.04 (6.96, 7.14)	.89
Initial AG, mEq/L (IQR)	31 (27, 36.5)	34 (29, 38)	.36
Initial HCO3, mEq/L (IQR)	5 (5, 7)	5 (5, 7)	.86
Initial BHB (IQR)	5.7 (5.3, 6)	5.1 (4.6, 5.8)	.0043

Note. AG = anion gap; BG = blood glucose; BHB = beta hydroxybutyrate; HCO₃ = serum bicarbonate; IQR = interquartile ratio; LR = Lactated Ringers; NS = normal saline. *All continuous variables are expressed as the median value, followed by the IQR.

Table 4.

Transition Characteristics.

Variable*	Transition AG ≤ 12 mEq/L N = 43	Transition AG > 12mEq/L N = 49	P-value
Transition BG, mg/dL (IQR)	152 (120, 186)	160 (141, 204)	.04
Transition HCO3, mEq/L (IQR)	20 (18, 22)	19 (17, 21)	.25
Insulin infusion duration, hours (IQR)	26 (21, 31)	25 (19, 40)	.91
Long-acting insulin dose, units (IQR)	25 (15, 31)	25 (20, 30)	.58
IV and SQ overlap time, hours (IQR)	2 (1, 2)	2 (1, 2)	.71
AG at 24 h, mEq/L (IQR)	12 (10, 13)	15 (12, 16)	.0001
HCO3 at 24 h, mEq/L (IQR)	22 (20, 25)	21 (19, 24)	.15
Albumin at 24 h, g/dL (IQR)	3.2 (2.6, 3.8)	3.1 (2.6, 3.6)	.80
Peak BG at 24 h, mg/dL (IQR)	268 (231, 317)	296 (223, 319)	.93

Note. AG = anion gap; BG = blood glucose; HCO₃ = serum bicarbonate; IQR = interquartile ratio; IV = intravenous; LR = Lactated Ringers; NS = normal saline; SQ = subcutaneous. *All continuous variables are expressed as the median value, followed by the IQR.

Outcomes

There was no significant difference in the rates of unsuccessful insulin transition between the 2 groups (Table 5). When individual parameters of this outcome were analyzed, there was also no difference between groups. Very few patients had re-initiation of insulin infusion, with larger numbers of patients experiencing an increase in AG at 24 hours after transition.

Table 5.

Outcomes.

Variable*	Transition AG ≤ 12 mEq/L N = 43	Transition AG > 12 mEq/L N = 49	P-value
Primary outcome
AG reopened at 24 h AND insulin re-initiated, N (%)	3 (6.97)	2 (4.08)	.66
Secondary outcomes
Insulin re-initiated, N (%)	4 (9.30)	4 (8.16)	1.00
AG reopened at 24 h, N (%)	24 (55.81)	20 (40.82)	.21
ICU LOS, days (IQR)	3 (2, 3)	3 (2, 4)	.31
Hospital LOS, days (IQR)	4 (3.5, 6)	5 (4, 8)	.03
Safety outcomes
Hypoglycemia events, N (%)	0 (0, 1)	0 (0, 1)	.41
Hypokalemia events, N (%)	1 (0, 4)	2 (1, 5)	.14
Hypophosphatemia events, N (%)	1 (0, 3)	1 (0, 2)	.56

Note. AG = anion gap; ICU = intensive care unit; IQR = interquartile ratio; LOS = length of stay. *All continuous variables are expressed as the median value, followed by the IQR.

To further analyze the factors influencing re-opening of AG at 24 hours after transition and insulin infusion re-initiation, regression models were utilized. Within the logistic regression model for re-opening of AG at 24 hours, factors that were significantly associated were the initial bolus fluid type and bicarbonate at 24 hours (Table 6). Use of NS, rather than LR, was associated with AG re-opening. A lower bicarbonate value at 24 hours was also associated with AG re-opening. Within the logistic regression analysis of factors impacting insulin re-initiation, utilization of standard ISF was associated with higher re-initiation rates (Table 7). The Hosmer-Lemeshow goodness-of-fit test for these logistic regression models resulted with P-values greater than .5, indicating that the data was a good fit for the model.

Table 6.

Logistic Regression Analysis for Impact of Factors on AG Re-opening.

Variable	Odds Ratio (CI)	P-value
Body mass index (kg/m2)	1.05 (1, 1.11)	.06
Hemoglobin A1c (%)	1.04 (0.88, 1.22)	.67
Bolus fluid type	0.35 (0.14, 0.83)	.02
Maintenance fluid type	0.92 (0.46, 1.83)	.81
DKA severity (%)	0.68 (0.29, 1.56)	.36
Long-acting insulin dose (units)	1.01 (0.98, 1.05)	.42
IV and SQ overlap time (hours)	1 (0.89, 1.13)	.99
Standard ISF	0.86 (0.37, 2.03)	.73
HCO3 at 24 h (mEq/L)	0.8 (0.7, 0.91)	.0008
Peak BG at 24 h (mg/dL)	1 (1, 1.01)	.31

Note. BG = blood glucose DKA = diabetic ketoacidosis; GOF = goodness-of-fit; HCO₃ = serum bicarbonate; ISF = insulin sensitivity factor; IV = intravenous; SQ = subcutaneous.

Table 7.

Logistic Regression Analysis for Impact of Factors on Insulin Re-initiation.

Variable	Odds Ratio (CI)	P-value
Body mass index (kg/m2)	1.01 (0.92, 1.10)	.89
Hemoglobin A1c (%)	1.17 (0.88, 1.56)	.28
Bolus fluid type	0.88 (0.2, 3.94)	.87
Maintenance fluid type	1.33 (0.42, 4.16)	.63
DKA severity	0.84 (0.19, 3.75)	.82
Long-acting insulin dose (units)	1.02 (0.96, 1.07)	.59
IV and SQ overlap time (hours)	1.01 (0.82, 1.23)	.96
Standard ISF	15.62 (1.83, 133.45)	.01
AG at 24 h (mEq/L)	1.1 (0.9, 1.35)	.33
HCO3 at 24 h (mEq/L)	0.9 (0.75, 1.09)	.28
Peak BG at 24 h (mg/dL)	1 (0.99, 1.01)	.37

Note. BG = blood glucose DKA = diabetic ketoacidosis; GOF = goodness-of-fit; HCO₃ = serum bicarbonate; ISF = insulin sensitivity factor; IV = intravenous; SQ = subcutaneous.

ICU LOS was similar between groups, but hospital LOS was significantly shorter in the transition AG ≤ 12 mEq/L group. Safety outcomes included incidences of hypoglycemia and electrolyte derangements (Table 5). The incidence of these events was overall low, with no significant difference between groups.

Discussion

This study was the first to analyze the success of IV to SQ insulin transition when utilizing a transition AG of ≤ 12 mEq/L compared to >12 mEq/L. This study found that the incidence of unsuccessful insulin transition was not different between groups. These findings remained consistent when assessing the individual composite components, as very few patients required re-initiation of the DKA treatment protocol. Throughout our analyses, several key factors were notable and warrant further discussion, and these include the ICU LOS, impact of fluid types, and incidence of hypoglycemia and electrolyte derangements, and impact of ISF type.

ICU LOS represents an important outcome for both patients and healthcare providers. While definitions of DKA resolution and timing of transition theoretically may impact LOS, this study did not reveal a significant difference between groups. Several factors may have limited this study’s ability to accurately assess ICU LOS. This study measured ICU LOS based on when patients were physically transferred out of the ICU and their location was updated within the electronic health record. However, there were likely instances in which patients were no longer under the care of ICU providers and not receiving ICU level of care, but physically remained in the ICU due to bed availability, potentially leading to a falsely extended LOS. Moreover, this study did not evaluate the specific trigger for DKA or the incidence of mechanical ventilation, both of which are likely to impact LOS. Future trials may be necessary to elucidate the impact of these factors on patient disposition during DKA management. However, this study showed that ICU LOS was not prolonged by transitioning at an earlier point, as re-initiation of the DKA protocol was low in both groups. Additionally, the hospital and ICU LOS reported in this study is similar to other studies evaluating DKA management.

The next factor warranting further discussion is the impact of fluid types on DKA management and resolution. As uncontrolled hyperglycemia leads to severe dehydration within DKA, all patients should receive IV fluids at presentation to restore intravascular volume. The choice of fluid, specifically balanced crystalloids versus normal saline, has been the topic of several previous trials. In 2020, Self et al and colleagues completed a post-hoc analysis of 2 cluster randomized clinical trials, comprising 172 adults. This analysis revealed that balanced crystalloids resulted in more rapid resolution of DKA, as well as shorter time to insulin infusion discontinuation, compared to 0.9% sodium chloride, or normal saline. ⁷ A large systematic review and meta-analysis of these trials performed by Alghamdi et al and colleagues showed that the use of normal saline, compared with balanced crystalloids, may be associated with longer time to DKA resolution, higher post-resuscitation chloride levels, lower post-resuscitation bicarbonate levels, and longer hospital stay. ⁸ Finally, in a prospective, randomized study including 45 patients, Mahler et al and colleagues showed that resuscitation with balanced solutions led to lower serum chloride levels and higher serum bicarbonate levels than patients receiving NS, preventing the development of hyperchloremic metabolic acidosis. ⁹ The current study adds to the available literature that appears to favor the use of balanced crystalloids, as the use of normal saline was associated with a higher incidence of AG re-opening at 24 hours after transition in the logistic regression model. Factors that were not evaluated within this study were the rates of acute kidney injury on presentation and at DKA resolution, as well as association of fluid type with time to IV insulin discontinuation. Clinicians may choose one fluid preferentially based on comorbidities, initial lab values, or personal preference. The current body of evidence appears to support the use of balanced crystalloids over normal saline, but more trials may be needed to elucidate the effects on ICU and hospital LOS within this patient population.

The final noteworthy factor was the incidence of hypoglycemia and electrolyte derangements between groups. IV insulin is labeled as a high-alert medication by the Institute for Safe Medication Practices (ISMP). ¹⁰ Insulin carries a risk for hypoglycemia, defined as BG below 70 mg/dL, which can have severe consequences if not promptly treated. Additionally, insulin leads to intracellular shifting of electrolytes, conferring a risk for electrolyte abnormalities including hypokalemia, hypomagnesemia, and hypophosphatemia. Institutional protocols for DKA typically require hourly BG checks during IV insulin infusion, as well as serial basic metabolic panels to assess for electrolyte abnormalities and replete as indicated. Theoretically, patients who are transitioned from an IV insulin infusion at an earlier point may have a reduced risk for these metabolic derangements. However, this study did not reveal significant differences in occurrences of electrolyte derangements or hypoglycemia. It should be noted that there was not a significant difference between groups in the duration of insulin infusion, despite one group transitioning at a higher AG. Future studies may be helpful to assess whether duration of insulin infusion is associated with a larger risk for hypoglycemia and electrolyte abnormalities.

It should be noted that there was a significant association between utilization of a standard ISF type and re-initiation of IV insulin. This appeared contrary to the expected finding, as we would predict that using a higher ISF, resulting in more intensive BG control, would result in lower incidence of IV insulin re-initiation. We postulated that this finding may be due to faster resolution of hyperglycemia when standard ISF is utilized, potentially leading to premature discontinuation of IV insulin, followed by rebound hyperglycemia and insulin re-initiation. However, further studies evaluating the impact of using various ISF types are needed to elucidate this finding.

There are several study limitations that should be considered. First, due to the retrospective design, selection and/or information bias may have occurred. Due to lack of or inaccurate data, physiologic information or clinical outcomes may have therefore been incomplete, limiting the conclusions that can be drawn. This was a single-center study, and institution-specific practices may limit external validity. While strict exclusion criteria reduced confounding factors, this also resulted in a large number of excluded patients. Additionally, the precipitating factors for each patient’s DKA were not able to be recorded, as many cases appeared to be multifactorial, and documentation of these factors was not always completed. Finally, due to the lack of previous data published on this topic, a power calculation could not be completed to determine the sample size needed to detect a difference between groups.

This study had important strengths. First, this study adds to the body of evidence surrounding DKA management and fills a current literature gap surrounding optimal transition AG. Strenuous exclusion criteria allowed the authors to eliminate several potential confounding factors and strengthened the internal validity of the study. All AG values were corrected for patient’s current albumin levels, which allowed for an accurate evaluation of this laboratory value. This study evaluates patient-oriented outcomes, such as ICU LOS and incidence of adverse effects from IV insulin. Finally, the use of logistic regression models to evaluate the impact of pertinent factors on key outcomes was a strength of this study.

Conclusion and Relevance

In adult critically ill patients undergoing treatment for DKA, there was not a significant difference in rates of unsuccessful transition when comparing a transition AG of ≤ 12 mEq/L and a transition AG > 12 mEq/L. Future studies are warranted to explore the impact of treatment characteristics on insulin transition.