Is a Priming Dose of Insulin Necessary in a Low-Dose Insulin Protocol for the Treatment of Diabetic Ketoacidosis?

Our criteria for the diagnosis of DKA included plasma glucose >250 mg/dl, HCO₃⁻ <15 mEq/l, pH <7.3, and moderate ketones in the urine or blood by the nitroprusside method. A total of 37 patients were prospectively randomly assigned to three groups as follows: 1) the load group consisted of 12 patients who received a loading dose of insulin at 0.07 unit/kg body weight followed by an infusion of 0.07 unit · kg body weight⁻¹ · h⁻¹; 2) the no load group consisted of 12 patients who received only an intravenous infusion of insulin at a dose of 0.07 unit · kg body weight⁻¹ · h⁻¹ without a loading dose; or 3) the twice no load group consisted of 13 patients who received an infusion of insulin at a dose of 0.14 unit · kg body weight⁻¹ · h⁻¹ without a bolus dose. Responses to therapy were based on the time to achieve a glucose level of ≤250 mg/dl, bicarbonate of ≥15 mEq/l, and pH ≥7.3. All other modes of therapy (hydrating solution, KCl, and HCO₃⁻) were similar in the three groups and followed the recommendations of the first ADA position statement (6). All patients were seen in the emergency department of the Regional Medical Center at Memphis. Upon arrival, history was obtained and a physical examination was performed. Urine ketones (by nitroprusside) plus glucose by fingerstick and blood gas measurements were obtained immediately. Once a suspected diagnosis of DKA was made, the initial laboratory blood samples were obtained for basal metabolic parameters and complete blood count with differential; urinalysis was performed for blood and urine ketones by both the nitroprusside and enzymatic methods. Patients then received 1 liter of 0.9% NaCl per hour until laboratory results were available. After laboratory confirmation, the intravenous fluids were adjusted, depending on the glucose and corrected serum sodium value, to ½ N saline or dextrose with 1 N or ½ N saline. With confirmation of the diagnosis of DKA and an acceptable potassium level (>3.3 mEq/l), the patients were then transferred to the Clinical Research Center (CRC) of the University of Tennessee Bowld Hospital after the research study was described to the patients or their next of kin and the consent form, which was approved by the University of Tennessee Health Science Center of Memphis Institutional Review Board, was signed. On arrival at the CRC, the treatment and care of patients was managed by the DKA team, including the medical housestaff on rotation for the service, following an established protocol with consultation of the staff endocrinologist. Hourly blood glucose measurements and blood samples for free insulin were obtained at 0, 5, 15, 30, 45, 60, 120, 240, and 480 min. Blood for measurement of other hormones and ketone bodies, blood gases, and metabolic profile; urinalysis for ketones; and other parameters were obtained. Random assignment of patients to different insulin regimens was done after arrival at the CRC. Random assignment was not based on sex or ethnicity. Because a priming dose of insulin adds an additional risk of hypokalemia, potassium levels were assessed and corrected on an hourly basis for the first few hours of therapy until glucose reached the predetermined goal of <250 mg/dl (i.e., ∼4 h) and every 2 h thereafter.

Laboratory studies included complete blood cell count, blood gas measurements, and routine blood chemistry values measured by standard assays in the clinical chemistry laboratory. In addition, plasma free insulin, C-peptide, β-hydroxybutyrate, acetoacetate, pyruvate, lactate, cortisol, and FFAs were assayed. Glucose was measured by a glucose oxidase method using a Beckman glucose analyzer. Plasma free insulin and C-peptide were measured after polyethylene glycol precipitation by a double-antibody radioimmunoassay method (14). Cortisol, FFAs, and ketone bodies were determined by previously described methods (2,15).

Overall mean comparisons for means and slopes for the three groups (load, no load, and twice no load) were made with one-way ANOVA using the Clinfo and SAS (SAS Institute, Cary, NC) software packages. These were followed by pairwise comparisons of averages and regression estimates of the slopes by observing the overlap/nonoverlap of the 95% confidence limits for all of the estimates (means and slopes). If any 95% confidence limits for an estimate overlapped, then no statistical significance was declared at P = 0.05. If there was no overlap due to variability, then statistical significance was declared at P < 0.05. This significance and nonsignificance is denoted by superscript letters (as in Table 4). Thus, groups having superscripts in common are not significant at P = 0.05, whereas those having no superscripts in common are significant at P < 0.05.

Table 1 shows the baseline characteristics of patients in the three treatment groups, with the following findings: 1) none of the patients were obese, 2) the majority of the patients were African Americans and were receiving insulin therapy, 3) omission of insulin (47%) and infection (33%) were the major precipitating factors for DKA, and 4) within the load group there were twice as many men as women, within the no load group there were three times as many women as men, and within the twice no load group men and women were equally distributed. However, in all of these groups, the majority of subjects were African American.

Table 2 shows biochemical data on admission and at complete recovery from DKA in the three groups. Although there were no statistically significant differences among any of the parameters in the groups on admission, cortisol values are lower in the no load group than in the other two groups. It is of interest that all the subjects in this group are African Americans, the majority are women, and the free insulin level in this group is also about twice that of the other two groups. However, this counter-regulatory hormone, similar to that in previous reports (5,16,17), is still twofold higher than the baseline value. All measured parameters reached their near-normal values at recovery, with total ketone bodies reaching ∼3.5 mmol/l. Of interest was the FFA level, which at recovery was about twice the normal value but consistent with elevated levels of FFA in diabetic patients without DKA (16). The mean anion gap at recovery was <12.

Figure 1 demonstrates the kinetics of intravenous insulin after infusion in the three arms. This figure shows that infusion of insulin of 0.07 (no load) or 0.14 unit · kg⁻¹ · h⁻¹ (twice no load) without an initial bolus resulted in peaks of 66 and 202 μU/ml, respectively, whereas the use of an intravenous bolus of 0.07 unit/kg followed by a 0.07 unit · kg⁻¹ · h⁻¹ infusion resulted in peak insulin of ∼460 μU/ml at 5 min, which reached a plateau level of ∼86 μU/ml in 30 min. It is also of interest that plateau levels of insulin for both the load and no load groups are not significantly different from 60 to 120 min. However, the twice no load group (0.14 unit · kg⁻¹ · h⁻¹, which is ∼10 units/h of insulin in a 70-kg patient) maintained higher levels of insulin during 30–120 min of infusion than those of the load or no load group.

Table 3 demonstrates the outcome recovery data for DKA, depicting the time that predetermined values were reached for glucose, HCO₃⁻, and pH. The results suggested no significant differences in these values for the three doses of insulin, but 5 of 12 patients in the no load group required additional insulin for control of their blood glucose. The length of hospitalization was not significantly different among the groups, there were no complications in any of the groups, and no deaths occurred in this cohort. No hypokalemia was noted in any treatment group. The potassium nadirs were 4.3, 4.2, and 3.7 mEq/l in the load, no load, and twice no load groups, respectively (Fig. 2).

Table 4 shows estimates of the slopes for glucose, venous pH, ketones, FFAs, bicarbonate, and cortisol during the first 8 h of therapy in the three insulin therapy groups. The response of serum bicarbonate and cortisol showed no differences among the treatment groups. However, the slopes for glucose, venous pH, ketones, and FFAs varied among treatment groups.

Although numerous studies have effectively used low-dose insulin in the treatment of DKA (3,9,18), there are no data from a prospective randomized study evaluating an insulin bolus (0.07 U/kg) compared with no bolus in conjunction with an infusion of low-dose insulin (0.07 vs. 0.14 unit · kg⁻¹ · h⁻¹). Previous studies have demonstrated that although 0.1 unit/h of insulin in patients with DKA can effectively inhibit lipolysis, this dose will only suppress hepatic gluconeogenesis by 50% (11).

We had previously shown that an intermediate dose of insulin of 0.1 unit · kg body weight⁻¹ · h⁻¹ (equivalent of 7 units/h of insulin in a 70-kg patient) by intravenous infusion was much more effective than a dose by the subcutaneous or intramuscular route (4). In these studies we used a priming dose. In addition, in a selected case report we have used as little as 6 units/h for a patient with modest DKA with further maintenance of euglycemia by 1–2 units/h (14). However, a bolus dose of insulin preceded the low-dose infusion (19). Furthermore, a priming dose has not been used for pediatric patients with DKA (10) and has not been recommended in the pediatric age-group in the latest ADA consensus report (12). Deleterious outcomes with the use of bolus insulin in this group may include hypokalemia as well as possible cerebral edema (12).

In summary, our study suggests that the use of a bolus or priming dose of insulin is not necessary when an adequate continuous insulin infusion such as 0.14 unit · kg⁻¹ · h⁻¹ (or 10 units/h in a 70-kg patient) is used. However, a dose of 0.07 unit · kg⁻¹ · h⁻¹ without priming is not adequate to obtain desired changes in glucose without supplemental doses of insulin.

This study was supported in part by U.S. Public Health Service Grant RR00211 (to the General Clinical Research Center) and by Grant DK07405 and Training Grant AM070884 from the National Institutes of Health.

The assistance of Drs. Dale Childress and James Peyton in assessment of some of the data during their training is greatly appreciated. The authors appreciate the secretarial assistance of Brenda Scott on this manuscript.