Fluid regimens for diabetic ketoacidosis in children and adolescents: systematic review and meta-analysis

Abstract

Background

Diabetic ketoacidosis (DKA), a life-threatening complication of type 1 diabetes, requires prompt fluid resuscitation to restore circulatory volume and correct metabolic abnormalities. However, the optimal fluid type, volume, and tonicity remain unclear due to concerns regarding cerebral injury and electrolyte imbalances.

Methods

We searched PubMed, CINAHL, Cochrane, Scopus, Clinicaltrials.gov, and WHO International Clinical Trials Registry Platform without date restrictions for randomized controlled trials (RCTs) evaluating fluid regimens in children with DKA. Risk of bias was assessed using the Cochrane RoB-2 tool, and certainty of evidence was assessed using the Grading of Recommendations Assessment, Development, and Evaluation approach. Meta-analysis was conducted using RevMan 5.4.

Results

The review included 10 RCTs (n = 2291). Children receiving a 20 mL/kg bolus followed by 0.9% saline exclusively for rehydration may have a higher risk of adverse events (AEs) than those receiving 10 mL/kg (risk ratio [RR] = 1.11, 95% CI = 1.01-1.23, low certainty of evidence). However, in children given 0.9% saline exclusively for rehydration, there was a significantly lower risk of hypoglycemia in the 20 mL/kg bolus group compared to the 10 mL/kg group (RR = 0.78, 95% CI = 0.62-0.99, high certainty). For rehydration fluid, compared to 0.45% saline, 0.9% saline (48-72 hours) was associated with a higher risk of hypoglycemia (RR = 1.35, 95% CI = 1.06-1.72, high certainty) and AEs (RR = 1.18, 95% CI = 1.05-1.31, low certainty). Over 36 hours, 0.9% saline may be associated with more AEs (RR = 1.15, 95% CI = 1.03-1.27, low certainty).

Conclusion

Smaller bolus volumes may be safer than larger ones, and prolonged use of 0.9% saline may increase the risk of hypoglycemia and AEs, although the certainty of evidence was low.

An estimated 1.2 million children and adolescents under the age of 20 live with type 1 diabetes mellitus (DM), with an additional 149 500 newly diagnosed cases per year [1]. A common complication typically occurring in people with type 1 DM is diabetic ketoacidosis (DKA), a metabolic derangement characterized by hyperglycemia, acidosis, and ketonemia [2]. Among children with type 1 DM, poorly managed DKA can be a potential cause of mortality [3]. The mortality rate in DKA varies from 0.15% to 0.31% in high-income countries and from 3.4% to 13.4% in low- and middle-income countries (LMICs) [4]. A key concern that contributes to the morbidity and mortality of DKA is cerebral injury, secondary to electrolyte abnormalities and the buildup of blood urea nitrogen levels, which are more common in the pediatric population. It occurs in approximately 1% of all pediatric DKA episodes and accounts for 20% to 40% of diabetes-related deaths in children [5, 6].

Fluid resuscitation followed by insulin administration is the cornerstone of DKA management. The immediate goal is to correct dehydration and restore circulating volume, thereby correcting electrolyte imbalances and clearing glucose and ketones from the blood [7]. Fluid management in DKA has long been approached with caution due to the risk of cerebral injury. Guidelines on fluid therapy vary considerably for initial fluid resuscitation and ongoing rehydration [8]. Since the publication of the Pediatric Emergency Care Applied Research Network (PECARN) FLUID trial, the approach to these treatments has shifted toward a more liberal strategy. However, consensus on the specifics remains lacking. According to the 2022 International Society for Pediatric and Adolescent Diabetes (ISPAD) guidelines, children in shock should receive a bolus dose of Ringer's lactate or normal saline at 20 mL/kg rapidly, followed by intravenous (IV) maintenance fluids given per hour, calculated on the severity of dehydration. For children not in shock, a bolus of Ringer's lactate or normal saline at 10 to 20 mL/kg should be administered over 20 to 30 minutes [9].

In contrast, guidelines for the integrated management of common childhood illnesses (IMCI) [10] and its updated guidance in the Pediatric Emergency Triage, Assessment and Treatment (ETAT) by the World Health Organization (WHO) [11] recommend that children in shock should receive a bolus dose of Ringer's lactate or normal saline at 10 to 20 mL/kg over 30 minutes to 1 hour, followed by IV maintenance fluids (2.5-4.0 mL/kg per hour). For children not in shock, a bolus of Ringer's lactate or normal saline at 10 to 20 mL/kg should be administered as rapidly as possible [11]. This cautious administration is crucial, as dehydration may lead to organ failure, while rapid rehydration may result in pulmonary and cerebral edema.

Although the IMCI and ETAT guidelines provide comprehensive clinical recommendations for outpatient and inpatient care for sick children, neither provides explicit recommendations for children and adolescents with type 1 DM, particularly those presenting with DKA. This systematic review was therefore commissioned by WHO to inform the development of consolidated guidelines on the management of common childhood illnesses. Comparing the fluids recommended in these guidelines to those used in other studies on DKA fluid management may help identify the optimal fluid regimen for pediatric DKA. This systematic review thus aims to synthesize current evidence on optimal fluid regimens for children and adolescents (1-19 years) with confirmed DKA, both with and without signs of shock.

Methods

The protocol for this review was registered with the International Prospective Register of Systematic Reviews (PROSPERO: CRD42024568723). The systematic review adhered to the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to identify, appraise, and synthesize relevant studies.

Eligibility criteria

We included randomized controlled trials (RCTs) evaluating various fluid regimens for the management of DKA in children and adolescents aged 19 years or younger. Detailed eligibility criteria for this review are given in Table 1.

Table 1.

Eligibility criteria

Inclusion criteria:

Population: Children and adolescents with a confirmed diagnosis of DKA, with or without signs of shock, ranging from 1 to 19 years, and with type 1 DM only

Setting: Low-, middle-, or high-income countries

Study design: RCTs

Type of intervention: Studies with current guidelines on fluid management from 2013 guidelines and ETAT guideline, ie, children with shock must receive IV fluids of 10-20 mL/kg body weight of isotonic fluids (crystalloids) over 30-60 minutes.

Type of comparator/control: Studies with other fluid resuscitation regimen than the 2013 ETAT guidelines

Exclusion criteria:

Population: Studies with only highly selected groups,  such as neonates or children being treated with high-dose corticosteroids or receiving chemotherapy Studies with participants diagnosed with syndromic diabetes, neonatal diabetes, corticosteroid-induced diabetes, maturity-onset diabetes of the young, and gestational diabetes Animals Adults

Study design: Case reports, case series, cross-sectional, case–control, and cohort studies, opinions, editorials, conference abstracts, reviews, and systematic reviews.

Studies on fluid management of DKA and those that do not report clinical outcomes (eg, morbidity, mortality, adverse events and length of hospital stay) Studies that do not explicitly state the management of DKA among the target age group

Outcomes

The following outcomes relevant to fluid therapy in pediatric DKA were prespecified and included in our review:

1. Morbidity: This outcome included the following 3 conditions:

   1. Cerebral edema was defined using the Glasgow Coma Scale or by assessing deterioration in the neurological status.

   2. Hypoglycemia was defined according to a varying range of glucose levels.

   3. Hypokalemia was defined according to varying thresholds of serum potassium levels.

2. Overall morbidity was a composite outcome of cerebral edema, hypoglycemia, and hypokalemia.

3. All-cause mortality: the total number of deaths that occurred during the study period.

4. Hospital stay:

   1. Hospital stay was defined as the length of stay in the hospital in hours.

   2. Intensive care unit (ICU) stay was defined as the length of stay in the ICU in hours.

5. Adverse events: Adverse events were any complications that occurred during fluid therapy other than hypoglycemia, hypokalemia, and cerebral edema, as defined by the authors.

Search methods for the identification of studies

PubMed, CINAHL, Wiley Cochrane Library, and Scopus were used to identify relevant studies using the search strategy (Table S1) [12]. The search strategy was developed using free-text and MeSH terms to retrieve eligible studies. There was no date restriction applied. Only studies in English were included in this review. Clinical trial registries, including clinicaltrials.gov and WHO International Clinical Trials Registry Platform (ICTRP), as well as bibliographies of previous relevant systematic reviews and all included studies, were also searched to identify any relevant studies.

Data collection and analysis

Selection of studies

All records identified by the search were imported to Covidence for screening (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia). After deleting the duplicates, 2 authors independently reviewed the titles and abstracts of the studies. All studies meeting the eligibility criteria were considered for the full-text screening. The disagreements encountered at this stage were discussed among the 2 reviewers and resolved by contacting the third reviewer. A similar process was followed for full-text screening. The reasons for excluding studies at the full-text screening stage were recorded.

Data extraction and management

Data from studies fulfilling the inclusion criteria after full-text screening were extracted independently by 2 reviewers into a pre-formatted extraction sheet in Covidence. The extracted data was checked for accuracy by another reviewer, and any discrepancies were resolved through discussion and consultation with a third reviewer.

The authors of relevant studies were contacted to obtain missing or unpublished data. Furthermore, for studies where full-text articles were unavailable, the study authors were contacted and followed up to request the retrieval of the articles. Studies were marked as not retrieved after exhausting all possible options. A similar process was conducted for the conference abstracts, presentations, and studies identified through the registries.

Quality assessment of studies

The quality assessment of the included RCTs was conducted using the Cochrane Risk of Bias tool II (RoB-2) [13]. Two reviewers independently assessed the risk of bias for each study's outcome using this tool. Differences in assessment were resolved through discussion and the involvement of a third reviewer. Each potential source of bias was graded as low, high, or some concern. The following domains were assessed in the RoB-2 tool: randomization process (D1), deviation from intended intervention (D2), missing outcome data (D3), measurement of the outcome (D4), and selection of the reported result (D5).

Statistical analysis

The meta-analysis was conducted on Review Manager (RevMan) 5.4.1 software to quantitatively pool studies where appropriate. Meta-analysis was performed when outcomes were reported in at least 2 studies within the same comparison. For continuous outcomes, the mean difference (MD) and its 95% confidence interval (CI) were reported. For studies that did not report exact mean and SD, Hozo's method was used to calculate mean and SD from median, range, and interquartile range [14]. For dichotomous outcomes, the risk ratio (RR) and its 95% CI were calculated using the number of participants who experienced the outcome and the total number of participants randomized to each group. A random-effect model was used in the meta-analysis.

Forest plots were created for all the outcomes. The results of trials that provided data unsuitable for inclusion in pooled analyses were described in this review's textual content.

Assessment of heterogeneity

Statistical heterogeneity was assessed by I², and by visually inspecting forest plots to detect nonoverlapping CIs. Based on prior clinical knowledge, clinical and methodological heterogeneity was expected among the included studies. Had heterogeneity been found, an attempt would have been made to identify possible reasons for it through subgroup analysis. This was not possible due to insufficient data in the comparisons identified for meta-analysis.

Subgroup analysis

We conducted a subgroup analysis by type of rehydration fluid administered to the groups.

Certainty of evidence

The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach was used for rating the certainty of the body of evidence for each outcome from the included studies. A summary of findings tables were generated for all outcomes, presenting the quality of evidence for each outcome according to GRADE criteria [15]. It covered consideration of within-study risk of bias, directness of evidence, heterogeneity, precision of effect estimates, and risk of publication bias.

We considered studies with overall “some concerns” of bias by downgrading the certainty of the evidence by 1 level for risk of bias, and studies with a high risk of bias by downgrading by 2 levels. We considered the magnitude and direction of heterogeneity when deciding whether to downgrade our certainty in evidence for inconsistency. The line of no effect was defined as a minimal clinically significant difference for all outcomes. For imprecision, outcomes were rated down once if optimal information size (ie, 300 participants) was not met or the 95% CI crossed the line of no effect, and downgraded twice if both conditions were met. We did not downgrade for indirectness as all included studies directly addressed the research question and were conducted in a tertiary care setting. The quality of the evidence was then rated as “high,” “moderate,” “low,” or “very low” for each outcome.

Results

The search of electronic databases, conducted on 22 June 2024, yielded 638 citations (Fig. 1). In total, 528 records remained after removing all duplicates. After screening titles and abstracts, we identified 39 citations for full-text review. Overall, 29 studies were excluded, including 3 studies that used language other than English, 8 that were protocol/trial registries of studies already included, and 8 for which the full text could not be retrieved. In these 8 studies, 3 were trial registries in which the trial status was listed as terminated.

Figure 1.

Study flow diagram.

Study characteristics

A total of 10 studies were included in this systematic review as shown in Table 2. All studies were RCTs. Of these 10 studies, 4 were conducted in the United States [16, 18, 20, 21], 5 in India [17, 19, 22-24], and 1 in Australia [25].

Table 2.

Study characteristics

Study	Country	Blinding	Age range	Gender intervention (%)	Gender control (%)	Sample size	Outcomes reported
Bakes et al (2016) [16]	USA	Open label	0-18 years	F = 72 M = 28	F = 48 M = 52	50	Morbidity (cerebral edema), hospital stay, adverse events
Dhochak et al (2018) [17]	India	Open label	0-12 years	F = 53 M = 46	F = 40 M = 60	30	Morbidity (cerebral edema, hypoglycemia, hypokalemia)
Glaser et al (2013) [18]	USA	Single blinding	8-18 years	F = 62 M = 38	F = 40 M = 60	18 (20 DKA episodes)	Morbidity (cerebral edema, hypoglycemia)
Kola et al (2022) [19]	India	Open label	0-18 years	F = 35 M = 65	F = 32.5 M = 67.5	40	Morbidity (cerebral edema), adverse events
Kuppermann et al (2018) [20]	USA	Single blinding	0-18 years	F = 53.3 M = 46.7		1255 (1389 DKA episodes)	Morbidity (cerebral edema, hypoglycemia, hypokalemia), mortality, hospital stay, adverse events
Rewers et al (2021) [21]	USA	Single blinding	0-18 years	F = 53.5 M = 46.5		667 (714 DKA episodes)	Adverse events
Shafi and Kumar (2018) [22]	India	Open label	0-18 years	F = 57.5 M = 42.5		40	Morbidity (cerebral edema), mortality, adverse events
Singhal et al (2024) [23]	India	Triple blinding	6 months-18 years	F = 72 M = 28	F = 52 M = 48	50	Morbidity (cerebral injury, hypokalemia), mortality, hospital stay, adverse events
Williams et al (2020) [24]	India	Double blinding	1 month-12 years	F = 47.1 M = 52.9	F = 53.1 M = 46.9	64 (66 DKA episodes)	Morbidity (cerebral edema, hypoglycemia, hypokalemia), mortality, hospital stay, adverse events
Yung et al (2017) [25]	Australia	Double blinding	10-15 yearsa	F = 39 M = 61	F = 38 M = 62	77	Morbidity (cerebral edema), hospital stay, adverse events

^a Interquartile range.

Participant characteristics

The 10 studies included 2291 participants, aged 0 to 18 years. Six studies [16, 19-23] included participants aged ≤18 years and 2 studies [17, 24] included participants aged ≤12 years. Glaser et al [18] included participants between 8 and 18 years of age, whereas Yung et al [25] did not specify an age range in their inclusion criteria.

Diagnosis of DKA

All included studies specified diagnostic criteria using varying DKA thresholds. Three studies [18, 20, 21] defined DKA as serum glucose >300 mg/dL, venous pH <7.25, or serum bicarbonate <15 mEq/L, and a positive urine ketone test. Three studies [17, 19, 25] defined DKA as blood glucose >200 mg/dL, venous pH<7.3 and/or bicarbonate <15 mEq/L, ketonemia, or ketonuria. The remaining studies [16, 22-24] used other varying biochemical criteria to define DKA as outlined in Table S2 [12].

Interventions

All included trials compared different fluid resuscitation regimens, either as initial boluses or rehydration fluids, for the treatment of DKA in children and adolescents (Table 3). Six studies used 0.9% saline as a rehydration fluid over 48 hours in one group and other fluids in the other group. Other fluids included balanced crystalloids such as Plasma-Lyte [24], Hartmann's Solution [25], and Ringer's lactate [23], 0.45% saline [20, 21] and oral rehydration solution [19]. The assumed fluid deficit was based on DKA severity [19, 23, 25] and 5% of body weight [20, 21] while a deficit of 6.5% to 10% was assumed in 1 study [24]. The bolus dose of 0.9% saline was similar between groups and was administered at 10 mL/kg [20, 21], 20 mL/kg [23, 24], and 10 to 30 mL/kg [25]. Moreover, 2 studies [20, 21] administered 0.9% saline over 36 hours as rehydration fluid to one group and 0.45% saline to the other group. The assumed fluid deficit was 10% of body weight, with half replaced in the first 12 hours and the remaining half over the subsequent 24 hours. A bolus dose of 10 mL/kg 0.9% saline, given twice, was the same in both groups.

Table 3.

Interventions in the included studies

Study	Intervention initial bolus	Control initial bolus	Intervention rehydration fluid (fluid deficit + maintenance as calculated by the Holliday–Segar method)	Control rehydration fluid (fluid deficit + maintenance, Holliday–Segar)
Bakes et al (2016) [16]	20 mL/kg of 0.9% saline given in 1 hour	10 mL/kg of 0.9% saline given in 1 hour	0.9% saline given, later switched to 0.45% saline Rate: 0.675% saline + potassium at 1.5 times maintenance rate Total duration of 48 hours	0.9% saline given, later switched to 0.45% saline Rate: 0.675% saline + potassium at 1.25 times maintenance rate Total duration of 48 hours
Dhochak et al (2018) [17]	10-20 mL/kg of 0.9% saline	10-20 mL/kg of 0.9% saline	1 bag: bag 1 contained 0.9% saline (sodium 77 mEq/L, chloride 77 mEq/L) plus potassium (40 mEq/L)	2 bags: received IV fluids through 2 fluid bags, both containing similar concentrations of electrolytes (sodium 77 mEq/L, chloride 77 mEq/L, potassium 40 mEq/L) but differing concentrations of dextrose: one containing no dextrose and the other containing 12.5% dextrose
Glaser et al (2013) [18]	20 mL/kg of 0.9% saline given in 1 hour	10 mL/kg of 0.9% saline given in 1 hour	0.9% saline given. Switched to 0.45% saline Assumed fluid deficit: 10% of body weight. Two-thirds of the fluid deficit is replaced over the first 24 hours, and the remaining one-third over the next 24 hours	0.9% saline given. Switched to 0.45% saline Assumed fluid deficit: 7% of body weight Rate of deficit replacement: the fluid deficit is replaced evenly over 48 hours
Kola et al (2022) [19]	0.9% saline in 1 hour in the first hour	0.9% saline in 1 hour in the first hour	Oral rehydration (primarily included oral rehydration salts [ORS] and other clear fluids such as coconut water and plain water) given for 48 hours The fluid deficit was based on the severity of DKA: Moderate DKA: assumed dehydration of 5-7% Severe DKA: assumed dehydration of 7-10% Total duration is 48 hours	0.9% or 0.45% saline given for 48 hours The fluid deficit was based on the severity of DKA: Moderate DKA: assumed dehydration of 5-7%. Severe DKA: assumed dehydration of 7-10% Total duration is 48 hours
Kuppermann et al (2018) [20]	10 mL/kg of 0.9% saline in the first hour	10 mL/kg of 0.9% saline in the first hour	0.9% saline given Assumed fluid deficit: 5% of body weight Process for deficit replacement: the deficit is replaced evenly over 48 hours with maintenance fluids Total duration is 48 hours	0.45% saline given Assumed fluid deficit: 5% of body weight Process for deficit replacement: the deficit is replaced evenly over 48 hours with maintenance fluids Total duration is 48 hours
Kuppermann et al (2018) [20]	10 mL/kg of 0.9% saline 2 boluses given	10 mL/kg of 0.9% saline 2 boluses given	0.9% saline given Assumed fluid deficit: 10% of body weight Process for deficit replacement: half of the fluid (plus maintenance fluids) is replaced during the first 12 hours. The remaining fluid (plus maintenance fluids) is replaced over the subsequent 24 hours	0.45% saline given Assumed fluid deficit: 10% of body weight Process for deficit replacement: half of the fluid (plus maintenance fluids) is replaced during the first 12 hours. The remaining fluid (plus maintenance fluids) is replaced over the subsequent 24 hours
Rewers et al (2021) [21]	10 mL/kg of 0.9% saline in the first hour	10 mL/kg of 0.9% saline in the first hour	0.9% saline given Assumed fluid deficit: 5% of body weight Process for deficit replacement: the deficit is replaced evenly over 48 hours with maintenance fluids Total duration is 48 hours	0.45% saline given Assumed fluid deficit: 5% of body weight Process for deficit replacement: the deficit is replaced evenly over 48 hours with maintenance fluids Total duration is 48 hours
Rewers et al (2021) [21]	10 mL/kg of 0.9% saline 2 boluses given	10 mL/kg of 0.9% saline 2 boluses given	0.9% saline given Assumed fluid deficit: 10% of body weight Process for deficit replacement: half of the fluid deficit (plus maintenance fluids) is replaced during the first 12 hours. The remaining deficit (plus maintenance fluids) is replaced over the subsequent 24 hours	0.45% saline given Assumed fluid deficit: 10% of body weight. Process for deficit replacement: half of the fluid (plus maintenance fluids) is replaced during the first 12 hours. The remaining fluid (plus maintenance fluids) is replaced over the subsequent 24 hours
Shafi and Kumar (2018) [22]	20 mL/kg of hypertonic saline (3% NaCl) given in 1 hour	20 mL/kg of 0.9% saline given in 1 hour	0.9% saline switched to 0.45% saline Fluid deficit: 5% body weight Rehydration fluid given for 48 hours	0.9% saline switched to 0.45% saline Fluid deficit: 5% body weight Rehydration fluid given for 48 hours
Singhal et al (2024) [23]	10-20 mL/kg of 0.9% saline infused over 20-30 minutes to restore the peripheral circulation. If tissue perfusion is poor the initial fluid bolus volume should be 20 mL/kg	10-20 mL/kg of 0.9% saline infused over 20-30 minutes to restore the peripheral circulation. If tissue perfusion is poor the initial fluid bolus volume should be 20 mL/kg	Fluid therapy Ringer’s lactate was administered as an hourly infusion The fluid deficit was based on the severity of DKA: Moderate DKA: assumed dehydration of 5-7%. Severe DKA: assumed dehydration of 7-10% Total duration of 24-48 hours	Fluid therapy 0.9% saline was administered as an hourly infusion The fluid deficit was calculated based on the severity of DKA: Moderate DKA: assumed dehydration of 5-7% Severe DKA: assumed dehydration of 7-10% Total duration of 24-48 hours
Williams et al (2020 [24]	20 mL/kg of 0.9% saline given over an hour to patients only with shock (perfusion abnormalities with or without hypotension [blood pressure <5th centile for age])	20 mL/kg of 0.9% saline given over an hour to patients only with shock (perfusion abnormalities with or without hypotension [blood pressure <5th centile for age])	Plasma-Lyte given Switched later to 0.45% saline Volume deficit of 6.5-10% assumed Total duration of 48 hours	0.9% saline given Switched later to 0.45% saline Volume deficit 6.5-10% assumed Total duration of 48 hours
Yung et al (2017) [25]	10-30 mL/kg of 0.9% saline given to only hypovolemic patients as rapidly as possible	10-30 mL/kg of 0.9% saline given to only hypovolemic patients as rapidly as possible	Hartmann's solution for 12 hours Switched later to 0.45% saline Fluid deficit: 6% for moderate DKA 10% for severe DKA Fluids already administered were subtracted from the calculated deficit Total duration of 48 hours	0.9% saline for 12 hours Switched later to 0.45% saline Mean fluid deficit of 6% for moderate DKA 10% for severe DKA Fluids already administered were subtracted from the calculated deficit Total duration of 48 hours

For the initial bolus, 4 studies [16, 18, 20, 21] compared bolus volumes of 20 and 10 mL/kg of 0.9% saline. For the rehydration fluid, 2 of these studies [16, 18] used 0.9% saline, switching to 0.45% saline later in the protocol, while the other 2 [20, 21] administered 0.9% saline only. The assumed fluid deficit was 10% in the intervention group and 7% in the control group for Glaser et al [18], except for Kuppermann et al [20] and Rewers et al [21], who used 5% fluid deficit for the 10 mL/kg group. One study [22] used an initial bolus of 20 mL/kg of 0.9% saline for 1 hour in the intervention group, while the other group received 20 mL/kg of hypertonic saline (3% NaCl) over the same period. Rehydration fluids were administered for 48 hours in both groups: 0.9% saline for the first 4 hours, followed by 0.45% saline at a weight-based rate.

Another study [17] administered a bolus dose of 10 to 20 mL/kg of 0.9% saline. With a total fluid rate of 200 mL/h, patients in the first group initially received IV fluids with the desired dextrose concentration from a single bag, which was changed when the dextrose concentration changed. For rehydration, the second group received IV fluids through 2 bags, each containing similar electrolyte concentrations (sodium 77 mEq/L, chloride 77 mEq/L, and potassium 40 mEq/L) but differing dextrose concentrations: one containing no dextrose and the other containing 12.5% dextrose. The desired dextrose delivery concentration (0-12.5%) was achieved by adjusting the 2.5% rate to maintain blood glucose within 150 to 200 mg/dL.

The studies were grouped for synthesis according to the specific fluid exposure randomized (fluid type, volume, tonicity, and system of administration) and the phase of therapy (initial bolus vs rehydration).

Concomitant management strategies

There was uniformity between groups in the administration of insulin, glucose, or electrolytes for all studies included in the review. Insulin was administered as an IV infusion, with most using a standard rate of 0.1 U/kg/h. Potassium replacement was consistent across most studies [16, 17, 19, 22-24], in which 40 mEq/L of potassium was added to IV fluids. Other studies [18, 20, 21, 25] used varying potassium salt combinations, with monitoring intervals ranging from 2 to 4 hours. Moreover, hourly glucose monitoring was done in all studies to adjust dextrose concentrations in IV fluids to prevent hypoglycemia. Dextrose was introduced when blood glucose dropped below 200 to 300 mg/dL, with varying concentrations (5-12.5%) adjusted based on glucose levels. However, Dhochak et al [17] used a dual-bag system for flexible dextrose titration, while Yung et al [25] introduced glucose when blood glucose fell below 270 mg/dL or dropped rapidly (>90 mg/dL/hour). Further details are shown in Table S3 [12].

Risk of bias

For cerebral edema, 9 studies were assessed for RoB-2 (Fig. S1) [12]. Of these, 1 study had a high overall risk of bias, 7 had some concerns, and 1 had a low risk of bias. The study with a high risk of bias [18] showed low risk in domains of the randomization process, missing outcome data and measurement of outcome, some concerns in deviations from intended interventions, and high risk in the selection of reported results.

Of the 4 studies that evaluated hypoglycemia (Fig. S2) [12], Glaser et al [18] showed a high risk of bias, and 2 studies [17, 24] showed some concerns, while Kuppermann et al [20] had a low risk of bias. The study with a high risk of bias showed a low risk in the randomization process and missing outcome data, some concerns in deviations from intended interventions and selection of reported results, and a high risk in the measurement of the outcome domain.

Similarly, 4 studies reported hypokalemia (Fig. S3) [12]. Of these, 3 raised some concerns of bias [17, 23, 24], whereas 1 [20] showed a low risk of bias. For mortality (Fig. S4) [12], 2 studies [22, 24] had some concerns and 2 studies [20, 23] showed a low risk of bias.

Moreover, of the 5 studies that reported hospital stay, 4 [16, 20, 23, 24] had some concerns, whereas 1 [25] showed a low risk of bias (Fig. S5) [12].

For adverse events, 1 study was classified as high-risk [21], 5 [16, 19, 22-24] raised some concerns, and 2 [20, 25] showed a low risk of bias (Fig. S6) [12]. In the high-risk-of-bias study, low risk was found in the randomization process and missing outcome data, some concerns in deviations from intended interventions, and high risk in domains of measurement of outcome and selection of reported result.

Outcome results

Comparison 1: type of rehydration fluids: 0.9% saline vs other fluids over 48 to 72 hours

Morbidity

The definition of cerebral injury varied across studies that explicitly reported their outcome definitions (Table S4) [12]. Cerebral injury was reported in five studies [19, 20, 23-25], and the results were comparable (RR = 1.00, 95% CI = 0.61 to 1.63, P-value = 1.00, I² = 0%, studies = 5, n = 927, certainty of evidence = very low) (Fig. S7) [12]. For hypoglycemia, 2 studies [20, 24] reported the outcome. In the group receiving 0.9% saline, 115 out of 381 participants experienced hypoglycemia, compared to 84 out of 379 participants in the group receiving other fluids (RR = 1.35, 95% CI = 1.06 to 1.72, P-value = .01, I² = 0%, studies = 2, n = 760, certainty of evidence = high) (Fig. S8) [12].

Three studies [20, 23, 24] reported hypokalemia, but there was no significant difference between the 2 groups (RR = 1.36, 95% CI = 0.85 to 2.18, P-value = .20, I² = 38%, studies = 3, n = 810, certainty of evidence = moderate) (Fig. S9) [12]. Out of 1252 participants in the group receiving 0.9% saline, 232 had overall morbidity as compared to 184 out of 1245 participants receiving other fluids, with comparable risk between the 2 groups (RR = 1.35, 95% CI = 0.96 to 1.91, P-value = .09, I² = 19%, studies = 5, n = 2497, certainty of evidence = low) (Fig. S10) [12].

In the subgroup analysis by the fluid type, the results were comparable for cerebral injury, hypokalemia, and overall morbidity (Figs. S11-S13) [12]. For hypoglycemia, the risk was 35% higher in the 0.9% saline group (RR = 1.35, 95% CI = 1.06 to 1.72, P-value = .02, I² = not applicable, studies = 1, n = 694, certainty of evidence = high) but comparable for balanced crystalloids (Fig. S14) [12].

Mortality

One study [20] reported no mortality events in the groups, whereas the other studies [23, 24] reported deaths in the group receiving other fluids. However, the results were not significantly different (RR = 0.26, 95% CI = 0.03 to 2.31, P-value = .23, I² = 0%, studies = 3, n = 810, certainty of evidence = low) (Fig. S15) [12]. Subgroup analyses by fluid type in the comparator group showed comparable mortality risk between the 0.9% saline and control groups (Fig. S16) [12].

Hospital stay

Both hospital and ICU stay were reported in 3 studies. The MD between the 2 groups for length of stay was 37 hours, but the results were comparable (MD = −0.37, 95% CI = −4.83 to 4.08, P-value = .87, I² = 0%, studies = 3, n = 810, certainty of evidence = very low) (Fig. S17) [12]. For ICU stay, similar results were observed (MD = 0.54, 95% CI = −11.59 to 12.66, P-value = .93, I² = 80%, studies = 3, n = 193, certainty of evidence = very low) (Fig. S18) [12]. Subgroup analyses by fluid type in the comparator group also showed no difference between the 2 groups (Figs. S19 and S20) [12].

Adverse events

Six studies [19-21, 23-25] reported adverse events. In the 0.9% saline group, 631 of 4954 participants experienced adverse events, compared with 526 of 4909 in the other fluids group. Adverse events were 19% more likely to occur in the 0.9% saline group (RR = 1.19, 95% CI = 1.07 to 1.33, P-value = .001, I² = 0%, studies = 6, n = 9863, certainty of evidence = high) (Fig. S21) [12].

In the subgroup analysis by fluid type, adverse events were 18% more likely in the 0.9% saline group compared to the 0.45% saline group (RR = 1.18, 95% CI = 1.05 to 1.31, P-value = .004, I² = 0%, studies = 2, n = 9380, certainty of evidence = low). The risk was comparable in those receiving balanced crystalloids and oral fluids (Fig. S22) [12].

Acute kidney injury was observed in 3 studies [23-25]. Hyponatremia was reported in 2 studies [23, 25], while hypernatremia was reported in 2 studies [21, 23]. In addition, hyperchloremic acidosis was reported in 2 studies [20, 21].

One study [19] observed asymptomatic hyperchloremia in all study participants within the first 12 hours of fluid therapy initiation. However, hyperchloremia improved significantly over 24 to 48 hours in the oral group compared with the IV group. Singhal et al [23] also noted hyperchloremia, along with mechanical ventilation and need for inotropes.

Kuppermann et al [20] classified a range of common clinical adverse events, which included headache, oropharyngeal pain, pyrexia, abdominal pain, constipation, as well as non-neurological events such as hypophosphatemia, hypocalcemia, thrombosis, renal failure, pancreatitis, pulmonary edema, and cardiac arrhythmia.

Comparison 2: type of rehydration fluids: 0.9% saline vs 0.45% hypotonic saline over 36 hours

Morbidity

Cerebral injury, hypoglycemia, and hypokalemia were reported in 1 study [20], but the results were comparable between the 0.9% saline and 0.45% hypotonic saline groups (Figs. S23-S25) [12]. Overall morbidity occurred in 179 out of 1053 participants in the 0.9% saline group, compared to 165 out of 1032 participants in the 0.45% saline group as shown in Fig. S26 of the Supplementary File (RR = 1.06, 95% CI = 0.88 to 1.29, P = .53, I² = not applicable, studies = 1, n = 2085, certainty of evidence = moderate) [12].

Mortality

Mortality was reported in 1 study [20] in the 0.45% hypotonic saline group, but the results were not significant (RR = 0.33, 95% CI = 0.01 to 7.99, P = .49, I² = not applicable, studies = 1, n = 695, certainty of evidence = moderate) (Fig. S27) [12].

Hospital stay

One study [20] reported hospital stay (Fig. S28) [12]. However, the results were not statistically significant (MD = 0.43, 95% CI = −3.98 to 4.84, P = .85, I² = Not Applicable, studies = 1, n = 695, certainty of evidence = moderate).

Adverse events

Two studies [20, 21] reported adverse events. In the 0.9% saline group, 661 of 4749 participants experienced adverse events compared with 561 of 4642 in the 0.45% hypotonic saline group. Adverse events were 15% more likely to occur in the 0.9% saline group as shown in Fig. S29 of the Supplementary File (RR = 1.15, 95% CI = 1.03 to 1.27, P = .009, I² = 0%, studies = 2, n = 9391, certainty of evidence = low) [12].

Comparison 3: volume of initial bolus: 20 mL/kg of 0.9% saline vs 10 mL/kg of 0.9% saline

Morbidity

With an initial bolus of 20 mL/kg of 0.9% saline in one group and 10 mL/kg of 0.9% saline in the other group, the results were comparable for cerebral injury, hypoglycemia, and hypokalemia (Figs. S30-S32) [12]. Overall morbidity occurred in 182 out of 1098 participants in the intervention group and 203 out of 1092 participants in the control group (RR = 1.47, 95% CI = 0.25 to 8.57, P = .67, I² = 49%, studies = 3, n = 2190, certainty of evidence = very low) (Fig. S33) [12].

In the subgroup analysis by type of rehydration fluid, results were comparable for cerebral injury, hypokalemia, and overall morbidity (Figs. S34-S36) [12]. However, subgroup analysis by rehydration fluid type showed comparable effects on hypoglycemia. At the same time, the risk was higher in the 0.9% saline rehydration fluid group (Fig. S37) (RR = 0.78, 95% CI = 0.62 to 0.99, P = .04, I² = not applicable, studies = 1, n = 700, certainty of evidence = high) [12].

Hospital stay

Two studies [16, 20] reported hospital stay and the results were comparable between the 2 groups (Fig. S38) (MD = 2.07, 95% CI = −3.86 to 8.00, P = .49, I² = 54%, studies = 2, n = 750, certainty of evidence = very low) [12]. Subgroup analysis by rehydration fluid type also showed comparable effects between the 20 and 10 mL/kg bolus groups receiving 0.9% saline (Fig. S39) [12].

Adverse events

Three studies [16, 20, 21] reported adverse events. In the 20 mL/kg bolus group, 661 out of 4774 participants experienced adverse events, compared to 582 out of 4738 participants in the 10 mL/kg bolus group. The results suggested that adverse events were 11% more likely to occur in the group receiving a bolus of 0.9% saline at a volume of 20 mL/kg as shown in Fig. S40 (RR = 1.11, 95% CI = 1.01 to 1.23, P = .04, I² = 0%, studies = 3, n = 9512, certainty of evidence = low) [12].

Subgroup analysis further showed a significantly higher proportion of participants experiencing adverse events in the group that was administered 0.9% saline rehydration fluid (Fig. S41) [12].

Comparison 4: tonicity of initial bolus: 0.9% saline vs 3% hypertonic saline

This comparison included 1 study [22] where cerebral injury was reported in both groups receiving 0.9% saline and 3% hypertonic saline (Fig. S42) [12]. However, the results were not statistically significant, and heterogeneity was not applicable. There were no deaths or adverse events related to hyperchloremia in either group (Figs. S43 and S44) [12].

Comparison 5: system of administration: 1 vs 2 bags

Only 1 study [17] was included for a 1- vs 2-bag comparison of the system of administration. The study reported hypoglycemia in 4 of 15 participants who received 2 bags and 6 of 14 who received 1 bag. There was no significant difference between the 2 groups (P = .37) (Fig. S45) [12]. Hypokalemia was reported in 10 of 15 participants who received 2 bags and 9 of 14 participants who received 1 bag (Fig. S46) [12]. Subsequently, 14 out of 45 participants in the 2-bag group and 15 out of 42 participants in the 1-bag group had overall morbidity. However, the results suggested that there was no significant difference in overall morbidity between the group that received 1 bag of fluid and the group that received 2 bags of fluid (Fig. S47) [12].

Certainty of evidence

A summary of findings for all comparisons and subgroup analysis is shown in Tables S5-S14 [12].

Discussion

This systematic review aimed to evaluate the optimal fluid regimen for children with DKA, with or without shock, by synthesizing existing evidence and assessing its alignment with existing recommendations from the WHO guidelines for the management of common childhood illnesses [10, 11]. We found that only 1 study [24] administered a bolus of 20 mL/kg of 0.9% saline over 1 hour to children who presented with shock. However, the results were not stratified by shock status, and its effect on outcomes was not reported. The remaining included studies assessed fluid therapy for children with DKA more generally, without distinguishing between patients with and without shock. Studies were thus compared based on fluid type, volume, tonicity, and system of administration. By pooling studies that had comparable phases of fluid regimens, the meta-analysis in this review increased the precision of effect estimates for clinically relevant outcomes that individual studies were underpowered to assess.

Across all the studies included, 0.9% saline was the most used fluid, either alone or in combination with other solutions. Our findings showed that using 0.9% saline for ongoing rehydration over 48 to 72 hours was associated with a statistically significant increase in hypoglycemia and other adverse events compared with regimens that included either 0.45% saline or balanced crystalloids. Subgroup analysis identified 0.45% saline as a favorable fluid for reducing both hypoglycemia and adverse events. These results align with current international guidelines, such as those from the ISPAD, which recommend transitioning to hypotonic saline (eg, 0.45% saline) or balanced crystalloids for ongoing rehydration once initial volume deficits are corrected and blood glucose levels begin to normalize [26]. This strategy aims to prevent hyperchloremic acidosis and provide adequate free water replacement, thereby mitigating the risks of both hypoglycemia and cerebral edema. Our findings thus provide further evidence supporting the move away from prolonged use of isotonic 0.9% saline as the sole rehydration fluid. However, the certainty of the evidence for adverse events was low due to the high risk of bias in one of the included studies.

Underlying these findings are the specific physiological effects of different fluid tonicity and composition. In patients with DKA, high saline concentrations can worsen hyperchloremic metabolic acidosis and delay recovery of metabolic abnormalities. Large chloride loads from 0.9% saline blunt renal bicarbonate reabsorption and may reduce glomerular filtration, promoting fluid retention and worsening of acidosis [27]. In contrast, balanced crystalloids (Ringer's lactate or Plasma-Lyte) provide a lower chloride load and include metabolizable anions, which can buffer acidosis [27, 28]. The increased risk of hypoglycemia with 0.9% saline was supported by high-certainty evidence from a multicenter trial conducted in the United States [20]. However, local factors such as clinical protocols, patient characteristics, and baseline hypoglycemia risk may differ across settings. These contextual differences underline the need for further studies in diverse settings with a high burden of type 1 DM to confirm or refute these results. In clinical practice, therefore, judicious use of balanced crystalloids or mildly hypotonic saline may reduce the risk of hyperchloremia and secondary acidosis without compromising hemodynamic support.

Regarding the choice of bolus volume, our review highlighted that adverse events may be less common in children who received a 10 compared to 20 mL/kg isotonic bolus. In particular, this risk was observed in the group whose rehydration fluid was 0.9% saline throughout rehydration, but not in those whose fluid was transitioned to 0.45% saline. The risk of adverse events should be cautiously associated with bolus volume, as DKA is a dynamic process, and it is important to consider the stage of fluid therapy at which they occur. However, further analysis was not performed due to inconsistent reporting of the timing of adverse events relative to fluid changes. Because the evidence was found to be low due to the risk of bias, the findings remain inconclusive but align with a systematic review that assessed the safety of fluid volume rates in the treatment of pediatric DKA [29]. Future studies should incorporate a core outcome set with standardized definitions and precise reporting of event timing.

Moreover, when the type of rehydration fluid administered was analyzed for hypoglycemia, the risk was higher in children receiving a 10 mL/kg bolus of 0.9% saline without a subsequent switch to 0.45% saline. This high-certainty evidence offers an essential consideration for clinical practice. Current NICE guidelines recommend using 0.9% saline without added glucose for both rehydration and maintenance fluid until plasma glucose falls below 250 mg/dL, at which point fluids should be changed to 0.9% sodium chloride with 5% glucose [30]. Our finding reinforces the guideline's implicit need for timely glucose introduction and fluid transition once blood glucose levels drop. Therefore, prolonged use of 0.9% saline alone for the entire duration of rehydration therapy may contribute to an increased risk of adverse events.

Large fluid boluses can lower serum osmolality, creating osmotic gradients that drive fluid into cells, which potentially causes cerebral edema. Although traditional teaching warns against rapid bolus and overhydration, which can precipitate cerebral injury in DKA [31], recent trials (including the large PECARN trial) found no increase in clinical brain injury with faster vs slower rehydration when insulin and fluids were adequately matched. Nevertheless, prudent fluid administration remains important. Physiologically, children with DKA are not truly hypovolemic by more than 7% to 8% of body weight, since osmotic forces preserve the intravascular volume [32, 33]. As a result, pediatric guidelines recommend only a modest 10 to 20 mL/kg isotonic bolus when perfusion is compromised and emphasize that deficits are often overestimated in this context. Our findings support this conservative approach: smaller boluses suffice to improve perfusion without risking overload that might precipitate electrolyte shifts and hypoglycemia. In practice, this means reserving large boluses for frank shock or dehydration and, otherwise, favoring gentle rehydration (eg, 10-20 mL/kg over an hour) with careful glucose monitoring.

We also compared the tonicity of both bolus and rehydration fluids and found a higher risk of adverse events with 0.9% saline compared to 0.45% saline. However, the certainty of evidence was low due to downgrading for risk of bias in outcome measurement and selective reporting in 1 of the 2 included studies. These findings are consistent with previous evidence, which suggests no significant differences in metabolic or neurologic outcomes between isotonic and hypotonic solutions [34].

With limited evidence, our review also found that the 2-bag system, which allows dynamic titration of dextrose concentration without stopping insulin or fluid, did not demonstrate clear benefits in clinical outcomes, including changes in blood glucose or bicarbonate levels, compared with the traditional 1-bag system. There were no significant differences between the two in rates of hypoglycemia, hypokalemia, or overall morbidity. This contrasts with the findings of Patino-Galarza et al [34], which suggested that the 2-bag system may hasten metabolic recovery. However, their conclusions were based on secondary outcomes, and the authors acknowledged the need for further studies to confirm these differences. As the 2-bag system is not a universally standardized practice in DKA management, our review identified its use in specific clinical settings, particularly within LMICs. Although guidelines do not uniformly endorse this innovation, its practical advantage, including reduced number of IV fluid bags used, ease of management by nursing staff, and lower cost of therapy, contributes to its feasibility and cost-effectiveness [35].

Overall, evidence on primary neurological outcomes remains reassuring: no fluid strategy demonstrated superiority in preventing cerebral edema, consistent with findings from a similar review [34]. These concordant results strengthen the conclusion that variations in fluid composition or rate do not materially alter the risk of DKA-related cerebral edema.

While this review provides a comprehensive synthesis of evidence from multiple RCTs, the overall certainty of the evidence varied. We found evidence for fluid administration recommendations from ETAT and its updated guidelines, but limited evidence and significant results from a single study make the generalizability and applicability of our findings challenging. Because most included studies were conducted in the United States and India, it is important to consider differences in health system capacities and patient populations when applying these findings to other regions. In addition, some outcomes were measured inconsistently across studies, limiting comparability and the ability to draw definitive conclusions. Due to inconsistent reporting and limited data availability, sensitivity analyses based on outcome definitions were not feasible. However, statistical heterogeneity across pooled analyses was generally low, which suggests that variability in outcome definitions was unlikely to have altered the direction of the effect. Furthermore, the GRADE approach indicated that the quality of evidence for most outcomes ranged from moderate to low. This further emphasizes the need for cautious interpretation of results.

Despite these limitations, this review highlights key gaps in the current evidence. First, high-quality RCTs or longitudinal studies are needed to confirm the potential benefits of balanced crystalloids. In addition, further research is warranted to establish the impact of the 2-bag system on pediatric patient outcomes in LMICs where resource optimization is critical. Third, standardizing outcome measures across studies will enhance comparability for future meta-analyses.

Implications for practice and research

This systematic review will inform the revision of guidelines on the management of childhood illness. The findings support cautious fluid resuscitation in pediatric DKA, favoring smaller boluses and balanced crystalloids or hypotonic saline when appropriate. As evidence on outcomes, like hypoglycemia and adverse events, remains limited and context-specific, high-quality trials, particularly in low-resource settings, are needed to reinforce and support adaptable protocols globally.

Conclusion

While our review provides insight into different fluid resuscitation strategies for pediatric DKA, the lack of subgroup analyses for children in shock remains a significant gap in literature. Future studies should specifically evaluate fluid regimens in children with shock, as current evidence does not allow for clear recommendations on this critical subgroup. Given the ongoing debate on optimal fluid management in pediatric DKA, well-powered, multicenter RCTs are essential to validate these findings and guide evidence-based clinical practice.

Funding

This work was funded by a grant (2024/1486209-1) from the World Health Organization (WHO) to the Centre of Excellence in Women and Child Health, Aga Khan University (AKU). Employees of WHO were responsible for determining the research question, the population, interventions, comparators, and outcomes of interest for the systematic review. WHO was not involved in the implementation of the systematic review, nor directly involved in the development of the study manuscript. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit the publication. Any views or opinions presented are solely those of the authors and do not necessarily represent those of the World Health Organization, unless otherwise specifically stated.

Author contributions

S.S. and S.A. conceived the study and initiated the study design. R.A. led the conduct of searches. R.A., A.I., and H.I. screened relevant papers. A.H. and A.I. conducted the data extraction. R.A., F.M., F.A., and H.I. conducted the data analysis. A.H., F.M., and R.A. drafted the manuscript. S.A., S.S., and EL critically reviewed the manuscript. All authors have read and agreed to the published version of the manuscript.

Disclosures

All authors have no conflict of interest and nothing to declare.

Data availability

Original data generated and analyzed during this study are included in this published article or in the data repositories listed in the References section.