Objective. The optimal fluid management for diabetic ketoacidosis (DKA) is uncertain. In an effort to simplify DKA therapy, we revised the treatment protocol in our institution to use a simpler method of calculating fluid needs, use fluids with higher sodium concentration, and allow glucose concentration to be adjusted easily. We performed a retrospective study to determine the effects of these revisions.
Design. We compared patients treated with traditional and revised protocols (∼220 and ∼300 patients, respectively, over consecutive 2.75-year intervals). Sixty patient records were randomly selected from the first group (30 treated with each of 2 protocol versions) and 30 from the second group. Biochemical and clinical parameters were analyzed.
Results. Patients selected for detailed analysis were similar in demographics and initial laboratory measurements. Patients treated under the revised fluid protocol received less total fluid, needed fewer intravenous fluid changes, were treated at less cost, and resolved acidosis more rapidly than patients treated under the original protocols. The rate of cerebral edema (0.3%–0.5%) was unchanged.
Conclusion. A DKA protocol that necessitates less fluid delivery and fewer calculations simplifies therapy and is associated with more rapid correction of acidosis.
- DKA =
- diabetic ketoacidosis •
- IV =
- intravenous •
- NS =
- normal saline •
- D =
- dextrose •
- ICU =
- intensive care unit •
- IDDM =
- insulin-dependent diabetes mellitus •
- DCCT =
- Diabetes Control and Complications Trial
Diabetic ketoacidosis (DKA) is a major source of morbidity and mortality in children and adolescents with type 1 diabetes mellitus. One of the most serious complications of DKA is cerebral edema, which occurs in as many as 3% of children with DKA and accounts for 30% of DKA deaths and 20% of overall childhood diabetes mortality.1 ,2 DKA management includes correcting hyperglycemia, dehydration, and electrolyte disturbances using intravenous (IV) fluids and IV insulin,3 but some retrospective studies have suggested that cerebral edema is associated with rates of fluid administration exceeding 4 L/m2/24 hours4 or >50 mL/kg over the first 4 hours of treatment.5 Although there have been few systematic comparisons of different fluid regimens,1the most recent editions of the widely used pediatrics textbooks and therapy manuals6–9 have decreased the amount of fluid recommended for DKA treatment (Table 1). Nevertheless, depending on the number and size of initial fluid bolus infusions, the estimated degree of dehydration, and the maintenance and fluid replacement rates chosen, it is possible to exceed the suggested safe limit of 4 L/m2/24 hours.
Failure of nonspecialist medical staff to follow formal DKA management protocols is associated with increased morbidity and mortality.10 Therefore, protocols that minimize fluid calculations and are easy to implement might be expected to reduce the risk of DKA complications.
We revised our DKA protocol several years ago to provide a more isotonic saline solution with less total fluid delivery than many traditional protocols and to require fewer fluid calculations and IV fluid changes during DKA therapy. Because of the dearth of data comparing current protocols with traditional protocols that provide more fluids, we performed a retrospective study to determine the effects of our protocol revision on actual fluid administration, complication rates, biochemical parameters, and time to resolution of acidosis.
The protocol used to manage DKA in patients at Children's Medical Center of Dallas was formally revised in July 1997. Patients treated before and after this date will be referred to as group 1 and group 2, respectively.
On presentation to the emergency department, all patients received a 20-mL/kg bolus infusion of 0.9% NaCl (normal saline [NS], 150 mmol/L of Na+) over 30 to 45 minutes. This was repeated if necessary to maintain adequate peripheral perfusion, defined as normal peripheral pulses and normal capillary refill time (<3 seconds). After completion of bolus infusions, patients from both treatment groups received regular human insulin in a premixed solution (0.2 U/mL in 0.9% NaCl) at a rate of 0.1 U/kg/hour IV. An initial IV insulin bolus was not administered.
Fluid requirements were calculated differently in the 2 groups (Table 1). In group 1, the fluid deficit was calculated by multiplying the percentage of dehydration (7%–10%, determined clinically on initial presentation) by the patient's weight (in kilograms).11This fluid deficit was added to 1.5 times the maintenance rate (based on admission weight) to determine the patient's total fluid requirement. Half of the total required fluid was ordered over the first 12 hours of treatment and the remaining 50% over the next 24 hours. In group 2, total fluids were delivered at 2.5 times the maintenance rate regardless of the degree of dehydration. Fluids were decreased to 1 to 1.5 times the maintenance rate after 24 hours of treatment (or earlier if acidosis resolved) until urine ketones were negative. In both groups, patients usually were changed to a subcutaneous insulin regimen and allowed to eat and drink ad libitum at the first meal time after resolution of acidosis, defined as a venous pH ≥7.30.
The compositions of IV fluids also differed between the 2 groups. After initial fluid bolus infusions, patients in group 1 received 0.45% NaCl (1/2 NS, 75 mmol/L Na+) whereas patients in group 2 received 0.675% NaCl (3/4 NS, 115.5 mmol/L Na+). The amount of KCl and K2PO4 used in each group depended on initial serum levels of K+, PO4−, and Ca2+ but usually totaled 40 mmol/L of K+. Patients received K+ only after voiding and confirmation of a serum K+ level <5.5 mEq/L. K+concentrations in fluids were increased if the patient became hypokalemic.
In both treatment groups, if the serum or capillary blood glucose concentration decreased by 350 mg/dL in <6 hours or by >100 mg/dL/hour or when it reached <300 mg/dL, glucose (dextrose, D) was added to the fluid regimen. In group 1, 2 methods of doing this were used depending on the attending physician. In group 1A, the initial 0.45% NaCl solution was discontinued and replaced with an identical solution containing an appropriate amount of glucose to provide a 4:1 glucose (measured in grams) to insulin (in units) ratio (5–12.5 g/dL glucose, D5–D12.5), and this was changed as necessary to control the level and rate of decrease of serum glucose.11 In group 1B, 10 g/dL of glucose (D10) was added to a separate solution that was otherwise identical to the initial fluid. The rate of infusion of each of the 2 solutions was varied as necessary to control the level and rate of decrease of serum glucose, with both the insulin and total fluid delivery remaining constant. Therefore, 3 separate IV solutions including the insulin solution (3-bag protocol) were needed for such patients rather than 2 bags. In group 2, the use of 3 solutions was mandated for all patients.
The protocols used for all patients specified that fluid administration rates should be decreased to maintenance or less if impending cerebral edema was suspected based on signs and symptoms such as headache, changes in sensorium, bradycardia, ophthalmoplegia, or rapidly falling serum sodium.
Almost all patients admitted with DKA in our institution are initially evaluated and treated in the emergency department. The majority are admitted to a regular hospital floor for additional management when stable (usually defined as normal level of consciousness with stable vital signs and a serum pH ≥7.20). Patients who are obtunded, have severe acidosis (initial pH <7.00), are hemodynamically unstable, or are very young (<3 years) often are admitted to the intensive care unit (ICU).
Our hospital serves as a teaching facility for several different programs in addition to pediatrics. Therefore, depending on chance and venue, a patient in DKA might be cared for by residents in pediatrics, family practice, or emergency medicine or (in the ICU) by a pediatric ICU fellow or an anesthesia resident. Pediatric residents and ICU fellows are given a single lecture on DKA management in July of each year; others are referred to written protocols.
The protocol revision in July 1997 was endorsed by all pediatric endocrinology attending physicians and by the division directors of pediatric emergency medicine and pediatric critical care. The change in protocol was accomplished by distributing the revised written protocol and changing the content of the July lecture on management.
Laboratory measurements were monitored in the same time frame for both groups. Each patient's weight, vital signs (qh), bedside meter blood glucose (qh), venous blood gas (q2 hours if pH <7.2, q4 hours otherwise), β-hydroxybutyrate, electrolytes, glucose, calcium, phosphorus, magnesium (all q4 hours), and urinary ketones (all voids) were measured on presentation and at intervals thereafter.
Charts were reviewed under an institutional review board–approved protocol. Patients with insulin-dependent diabetes mellitus (IDDM) who received DKA therapy under a traditional fluid protocol11 (group 1) were identified from a list of patients at Children's Medical Center of Dallas who had discharge diagnoses of “diabetic ketosis and/or ketoacidosis” and admission dates from September 1, 1994 to June 30, 1997, whereas patients treated under the revised fluid protocol (group 2) were identified from a list of patients admitted from July 1, 1997 to March 31, 2000. However, charts selected for additional review excluded a 1-year period before and after July 1, 1997 to avoid confounding the analysis with possibly increased attention paid to DKA management by physicians around the time of the protocol revision. A total of 111 randomly selected charts from September 1, 1994 to June 30, 1996 were reviewed to identify patients with an initial pH <7.30. Two groups of 30 patients each were selected corresponding to groups 1A and 1B (ie, patients treated with the 2-bag and 3-bag protocols, respectively). Similarly, 48 charts from patients treated after July 1, 1998 were reviewed to identify 30 patients with an initial pH <7.30.
To determine whether the new protocol influenced rates of fluid administration, biochemical parameters, or time to correction of DKA, 30 charts each from groups 1A, 1B, and 2 were examined. For any given parameter being examined by Student's t test, this sample size has >90% power to detect a difference between means of 0.85 times the standard deviation at a significance level of 0.05 (for total fluid delivered, this corresponds to ∼1.2 L). Charts were analyzed for ≥24 hours from the time patients initially were seen in the emergency department. We recorded profile data, biochemical data, and composition, rate, and duration of administration of each IV fluid. Fluid and laboratory costs for all treatment groups were estimated from current (February 2001) Children's Medical Center inpatient price lists.
Differences in profiles of, biochemical data of, and fluid delivered to the patients from the 3 groups were evaluated using unpaired Student'st test. Differences in the number of IV solution changes for each patient were evaluated using Kruskal-Wallis (all 3 groups) or Mann-Whitney (comparisons of any 2 groups) tests. Differences between the groups for race, sex, number of patients admitted to the ICU, and number of patients presenting with new onset of IDDM were evaluated using chi-square tests. All data were analyzed with StatView 4.5 (Abacus Concepts, Inc, Berkeley, CA, 1996), withP < .05 (2-sided) considered statistically significant.
In a 2.75-year period after institution of this protocol, there were 502 admissions to Children's Medical Center of Dallas with a discharge diagnosis of “diabetic ketosis and/or ketoacidosis.” Review of 48 randomly selected charts showed that 30 (63%) of these patients presented with an initial pH <7.30. There were 363 patients admitted with the same diagnosis in a 2.75-year period before this protocol was instituted. Assuming that ∼60% of patients admitted to our hospital with this diagnosis have pH values <7.30, we estimate that ∼220 patients were treated with the traditional protocol during the first period (group 1), and ∼300 patients were treated with the revised protocol during the second period (group 2).
There were no statistically significant differences between groups in profile data or initial biochemical characteristics (Table 2).
Impact on Management: 2-Bag Versus 3-Bag Protocols
When we revised our DKA protocol, we specified the use of a 3-bag system consisting of an insulin solution and 2 bags of electrolyte solutions that were identical except that 1 contained 10% dextrose, with the goal of varying total fluid rate and glucose infusion rate independently. Although this was mandated for all patients in July 1997, it was ordered on approximately half of patients in 1995–1997, permitting direct—albeit retrospective and nonrandomized—comparison of 2-bag and 3-bag protocols in which rates and composition of administered fluids were otherwise identical (ie, groups 1A and 1B, respectively). In fact, we found that these subgroups did not differ in total fluid administration, time to resolve acidosis, or any biochemical parameters, although the number of fluid changes ordered per patient was reduced by ∼23%; this resulted solely from a decreased number of orders to change the dextrose concentration (P = .008). By comparison, a similar protocol reduced fluid bags per DKA admission by 44% in a recently published study.12
Because of the minimal differences between groups 1A and 1B, they were pooled for most of the statistical analyses in the remaining discussion.
Impact on Management: Changes in Fluid Rate and Sodium Concentration
Our protocol revision was intended to minimize inadvertent administration of excessive amounts of free water and thus to reduce the risk of cerebral edema. Previous studies had suggested that fluid administration >4 L/m2/24 hours (2.7 times typical maintenance fluid requirements of 1.5 L/m2/24 hours) is associated with a greater risk of cerebral edema.4 Because most physicians are able to calculate maintenance fluid requirements, we reasoned that specifying a rate slightly less than the 4 L/m2/24 hour limit—2.5 times the maintenance rate—would ensure provision of adequate but not excessive fluids.
As expected, mean total IV fluid administration over the first 24 hours was lower in patients treated with the new protocol (5.3 vs 4.1 L/m2, P < .0001; Table 3). Although one might have expected acidosis to resolve more slowly under these circumstances, the time to resolve acidosis also was shorter (16.7 hours vs 12.6 hours,P = .01). There was also a nonsignificant decrease in mean length of hospital stay for DKA after the new protocol was instituted (2.86 ± 0.20 vs 2.61 ± 0.11 days,P = .22).
Based on current charges for the various IV fluid formulations and laboratory determinations, there were no significant differences between groups 1A and 1B in IV fluid or laboratory charges per patient, but the new protocol reduced IV fluid costs from $1060 to $776 (P = .0006) per patient and laboratory charges from $2752 to $2001 (P = .005) over the first 24 hours of treatment. Thus average total costs per patient were reduced by $1036 (P = .0009). At our current rate of ∼130 patients per year treated for DKA, the protocol revision should save ∼$130 000 annually in IV fluid and laboratory charges.
Impact on Biochemical Parameters
Patients in DKA usually are depleted of both Na+ and K+, which must be replaced along with free water to prevent hyponatremia and hypokalemia from developing during fluid resuscitation. Because the revised protocol provides lower fluid volumes, the Na+concentration in the IV fluids was increased to maintain the rate of Na+ replacement. As expected, there were no significant changes in total Na+ delivered with use of the revised protocol, and serum Na+concentrations decreased by similar amounts (∼5 mmol/L) with both protocols (Table 3). There were also no significant changes in frequency or degree of hypokalemia. Four of the 14 changes in IV fluids in the 30 patients from group 2 were made to treat hypokalemia, and each such patient initially presented with borderline hypokalemia (range 3.5–3.9 mEq/L). Therefore, the risk of hypokalemia might be further decreased by anticipatory increases in K+concentration in IV fluids in patients with an initial serum K+ concentration <4.0.
Patients in group 2 had slightly greater increases in serum chloride concentration with treatment and smaller increases in bicarbonate concentration, but these differences were not statistically significant and did not prolong the time to resolve acidosis. Hyperchloremia might be reduced by substituting another anion for part of the chloride in the IV fluids. Lactate (eg, as Ringer's lactate solution) and acetate have occasionally been used as components of fluids used in DKA, but we are unaware of extensive data on safety and efficacy of these anions in treating this condition.
Impact on Rate of Cerebral Edema
Complications were rare both before and after institution of the revised protocol. There was 1 case of cerebral edema during each period (incidence of 0.5 and 0.3%, respectively); for purposes of this study, we defined cerebral edema as signs of increased intracranial pressure warranting and responding to treatment with mannitol. These patients were boys 17 and 14 months old with new-onset diabetes and severe acidosis (initial pH 6.87 and 7.03, respectively) who had both initially been misdiagnosed by family members as having asthma and had been treated with albuterol nebulizers. Both needed intubation, neuromuscular blockade, and intracranial pressure monitoring; both children recovered completely. In addition, during the 5.5-year study period there was 1 case of mild cerebral edema (necessitating mannitol but not intubation or intracranial pressure monitoring) in a 4-year-old girl with new-onset diabetes and ketosis (pH 7.34) who was not treated with intravenous insulin; this patient also recovered completely. Thus the overall rate of cerebral edema during the study period was 3/865 (0.3%).
There were no deaths in group 1 (∼220 patients) and 1 in group 2 (∼300 patients); this was a child with new-onset IDDM who was already in cardiorespiratory arrest on arrival in the emergency department.
Incidence of DKA
The number of patients in DKA currently admitted to our hospital (∼130/year) is larger than in other published studies cited to support current management recommendations.2 ,5 13–15This is a consequence of our large diabetes patient population of ∼1000 children aged <1–18 years. Because ∼40% of our DKA admissions are new onset (Table 2), we estimate that there are ∼80 admissions per year of established patients for DKA, or a rate of ∼80 admissions/1000 patients/year. We are unaware of extensive data from the United States on hospitalization rates for children with diabetes, but our experience compares favorably with a prospective study of children in the United Kingdom2 that reported a rate of 190 admissions/1000 patients/year.
Risk of hospitalization may be related to degree of metabolic control. The mean hemoglobin AIC for our entire practice is 8.2 ± 1.4%, which is comparable to the levels achieved by adolescents in the Diabetes Control and Complications Trial (DCCT)16 and to levels (7.9%–8.2%) for adult patients from this study 4 years after they were returned to the care of their regular physicians.17 Because the DCCT was a 6.5-year trial of intensive diabetes management, participants in this study were highly selected for motivation and compliance. Therefore, we suspect that it will be difficult to improve the general level of control in our patients with our current management, which implements recommendations of the DCCT.16 In contrast, patients with established IDDM admitted to our hospital with DKA had a mean glycohemoglobin of 16.5% ± 3.2%, suggesting that they complied poorly with diabetes management (our hospital laboratory's glycohemoglobin values typically are 1 percentage unit higher than the hemoglobin AIC test used in our outpatient office).
Whereas ∼40% of our DKA patients have new-onset diabetes, ∼30% of our patients with new-onset diabetes (total of ∼180/year) present in DKA. It may also be difficult to reduce this percentage significantly, particularly with delays in treatment resulting from parents failing to seek medical attention in a timely manner (eg, both cases of severe cerebral edema in our study were in children who were misdiagnosed by caregivers). Increased sensitivity by parents and physicians to the prodromal signs of polyuria and polydipsia would be most helpful, but these signs often are insidious and difficult to recognize except in retrospect.
Although retrospective studies suggest that excessive fluid administration rates are associated with an increased risk of cerebral edema, such studies might be confounded by the possibility that more ill-appearing or dehydrated patients (who might be at increased risk for cerebral edema in any case18) received more fluids. Therefore, studies directly comparing different fluid protocols would be a useful way to optimize fluid therapy for DKA by administering sufficient fluids to resolve dehydration and acidosis quickly while minimizing the risk of cerebral edema. Unfortunately, there has been a paucity of such studies. One study claimed a decrease in cerebral edema with use of more isotonic fluids and a decreased fluid rate,13 but a follow-up study showed an essentially unchanged rate of cerebral edema (4/231, 2%), albeit with no deaths or permanent sequelae.14
The incidence of cerebral edema (0.3%) in our center was lower than in that study or in several other published studies,5 ,18 ,19but it was similar to the rate reported in Australia.20Among our patients, the rate of cerebral edema was similar in patients receiving more fluids than are currently recommended (group 1, an average of 5.3 L/m2/24 hours), those receiving fluids at rates similar to those currently recommended (group 2, an average of 4.1 L/m2/24 hours), and those who were not in significant ketoacidosis and therefore were not treated with intravenous insulin infusions. Therefore, within the range of fluid administration rates we examined, our data do not support the idea that reducing these rates per se reduces the risk of cerebral edema. The same conclusion was reached in a recently published multicenter study.18
On the other hand, the data do show that a moderate reduction in IV fluid rate, combined with an increase in the sodium content of the fluids, reduces the number of fluid changes needed to treat this condition; moreover, these changes are associated with decreased time to resolve acidosis. We cannot easily explain the latter finding. Because this was a retrospective, nonrandomized study and groups 1 and 2 were treated in successive time periods, it is possible that the difference reflects subtle improvements in DKA management other than those embodied in the protocol. However, it seems unlikely that such improvements could have been instituted and then continued by residents or nurses without our knowledge considering the size and constant turnover of these groups in this large teaching hospital. For example, there were no differences between groups 1 and 2 in time elapsed before each patient was started on insulin (2.3 ± 1.6 vs 2.1 ± 0.8 hours, P = .37). The time to resolve acidosis is highly correlated (P < .001 by linear regression, data not shown) with the number of fluid changes ordered per patient, but it is difficult to distinguish cause and effect in such a correlation.
Determining whether additional reductions in fluid rates will decrease or increase duration of acidosis or risk of cerebral edema warrants additional study. Although it will be difficult to detect additional improvement in incidence of cerebral edema with changes in therapy in a single medical center, our current protocol has advantages in simplicity and flexibility that should reduce the chance for clinical error, the duration of metabolic acidosis, and consequently, the cost of therapy.
This work was supported by National Institutes of Health Grants T32 DK07307 (to E.I.F.) and R37 DK37867 (to P.C.W.).
We thank Drs Bryan Dickson, John Germak, James Marks, Grace Tannin, and Melissa Ham and the attending and resident staffs of the Emergency and Critical Care Divisions for implementing the protocol. We also thank Marilyn Cox, Ben Eckert, and Sherry Silmon for their invaluable assistance with data collection.
- Received December 20, 2000.
- Accepted March 9, 2001.
Reprint requests to (P.C.W.) Department of Pediatrics, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390-9063. E-mail:
Dr Felner is now at Tulane University School of Medicine, New Orleans, Louisiana.
- ↵Sperling MA. Diabetes mellitus in children. In: Behrman RE, Kliegman RM, Jenson HB, eds. Nelson Textbook of Pediatrics. 16th ed. Philadelphia, PA: WB Saunders; 2000:1767–1791
- ↵Plotnick LP. Type 1 (insulin-dependent) diabetes mellitus. In: McMillan JA, DeAngelis CD, Feigin RD, Warshaw JB, eds. Oski's Pediatrics: Principles and Practice. 3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 1999:1793–1803
- ↵White NH. Diabetes mellitus in children. In: Rudolph AM, Hoffman JIE, Rudolph CD, eds. Rudolph's Pediatrics. 20th ed. Stamford, CT: Appleton & Lange; 1996:1803–1827
- ↵Raboudi N, Levitsky LL. Diabetes mellitus type 1. In: Burg FD, Ingelfinger JR, Wald ER, Polin RA, eds. Gellis & Kagan's Current Pediatric Therapy. 16th ed. Philadelphia, PA: WB Saunders; 1999:761–772
- ↵Marks JF, Gonzalez JL. Diabetic ketoacidosis. In: Levin DL, Morriss FC, eds. Essentials of Pediatric Intensive Care. 2nd ed. New York, NY: Quality Medical Publishing; 1997:565–570
- Copyright © 2001 American Academy of Pediatrics