eTable. Multinomial Logistic Regression Showing Adjusted Odds Ratios of the Factors Associated With Acute Kidney Injury in Children ≤18 Years with Type 1 Diabetes Presenting With Diabetic Ketoacidosis at British Columbia Children’s Hospital, Canada
Customize your JAMA Network experience by selecting one or more topics from the list below.
Hursh BE, Ronsley R, Islam N, Mammen C, Panagiotopoulos C. Acute Kidney Injury in Children With Type 1 Diabetes Hospitalized for Diabetic Ketoacidosis. JAMA Pediatr. 2017;171(5):e170020. doi:10.1001/jamapediatrics.2017.0020
What proportion of pediatric patients with type 1 diabetes who present in diabetic ketoacidosis develop acute kidney injury, and what are the associated risk factors?
In this medical record review of 165 children with type 1 diabetes who were hospitalized for diabetic ketoacidosis, 106 (64.2%) met the criteria for acute kidney injury. Serum bicarbonate level less than 10 mEq/L and an elevated heart rate were found to be associated with an increased risk of severe acute kidney injury.
Children in diabetic ketoacidosis are at high risk for acute kidney injury, suggesting that clinicians should consider acute kidney injury as a frequent complication in this population.
Acute kidney injury (AKI) in children is associated with poor short-term and long-term health outcomes; however, the frequency of AKI in children hospitalized for diabetic ketoacidosis (DKA) has not been previously examined.
To determine the proportion of children hospitalized for DKA who develop AKI and to identify the associated clinical and biochemical markers of AKI.
Design, Setting, and Participants
This medical record review of all DKA admissions from September 1, 2008, through December 31, 2013, was conducted at British Columbia Children’s Hospital, the tertiary pediatric hospital in British Columbia, Canada. Children aged 18 years or younger with type 1 diabetes and DKA and with complete medical records available for data analysis were included (n = 165). All data collection occurred between September 8, 2014, and June 26, 2015. Data analysis took place from August 25, 2015, to June 8, 2016.
Main Outcomes and Measures
Acute kidney injury was defined using Kidney Disease/Improving Global Outcomes serum creatinine criteria. Multinomial logistic regression was used to identify potential factors associated with AKI.
Of the 165 children hospitalized for DKA, 106 (64.2%) developed AKI (AKI stage 1, 37 [34.9%]; AKI stage 2, 48 [45.3%]; and AKI stage 3, 21 [19.8%]). Two children required hemodialysis. In the adjusted multinomial logistic regression model, a serum bicarbonate level less than 10 mEq/L (compared with ≥10 mEq/L) was associated with a 5-fold increase in the odds of severe (stage 2 or 3) AKI (adjusted odds ratio [aOR], 5.22; 95% CI, 1.35-20.22). Each increase of 5 beats/min in initial heart rate was associated with a 22% increase in the odds of severe AKI (aOR, 1.22; 95% CI, 1.07-1.39). Initial corrected sodium level of 145 mEq/L or greater (compared with 135-144 mEq/L) was associated with a 3-fold increase in the odds of mild (stage 1) AKI (aOR, 3.29; 95% CI, 1.25-8.66). There were no cases of mortality in patients with or without AKI.
Conclusions and Relevance
This study is the first to date to document that a high proportion of children hospitalized for DKA develop AKI. Acute kidney injury was associated with markers of volume depletion and severe acidosis. Acute kidney injury is concerning because it is associated with increased morbidity and mortality as well as increased risk of chronic renal disease, a finding that is especially relevant among children who are already at risk for diabetic nephropathy. Strategies are needed to improve the diagnosis, management, and follow-up of AKI in children with type 1 diabetes.
Diabetic ketoacidosis (DKA) is a severe complication of diabetes in children caused by a state of insulin deficiency. It is defined by hyperglycemia, metabolic acidosis, and the production of ketoacids, and hyperglycemia leads to an osmotic diuresis that causes volume depletion. It may occur at the initial presentation of newly diagnosed type 1 diabetes, or in a child with preexisting type 1 diabetes in times of illness or insulin omission. Diabetic ketoacidosis is the leading cause of hospitalization, morbidity, and mortality in children with type 1 diabetes,1-5 and cerebral edema is the most serious complication that leads to morbidity and mortality.6 To mitigate the risk of cerebral edema, pediatric-specific DKA protocols have been developed and frequently recommend conservative fluid administration at presentation. The DKA protocol at the British Columbia Children’s Hospital (BCCH), Vancouver, Canada,7 is used provincewide for all pediatric patients in British Columbia and is noted to be conservative regarding initial fluid management.8
Acute kidney injury (AKI), previously referred to as acute renal failure, is a common event in hospitalized children and implies a sudden worsening of the kidney’s ability to function. The clinical manifestations of pediatric AKI range from a mild increase in serum creatinine to anuric renal failure that requires dialysis. The most common risk factor for pediatric AKI is prerenal disease or volume-responsive AKI, which is caused by hypovolemia and reduced renal perfusion. If a prerenal insult is severe or prolonged, the injury can result in structural damage to the renal parenchyma, a condition known as acute tubular necrosis. Despite the marked intravascular volume depletion that occurs in DKA and subsequent cautious fluid-rehydration strategies, AKI in children with DKA has not been systematically studied. Only 2 case reports9,10 have been published to date indicating that severe kidney injury occurs in the pediatric DKA setting. Murdoch and colleagues9 reported on 2 girls, aged 13 and 14 years, with DKA complicated by anuric renal failure requiring dialysis. Both girls had biopsy evidence of acute tubular necrosis. More recently, Al-Matrafi and colleagues10 reported on a 12-year-old girl with severe DKA who developed anuric renal failure requiring hemodialysis. These reports are particularly concerning in view of the evolving body of literature indicating that AKI in children is associated not only with increased morbidity and mortality but also an increased risk of chronic kidney disease.11
The primary objective of this study was to determine the proportion of children with type 1 diabetes, hospitalized for DKA at BCCH over 5 years, who developed AKI. We hypothesized that, because DKA is associated with both volume depletion and conservative fluid administration upon presentation, these children are potentially at high risk for AKI, above the level of risk expected by the rare reported cases in the literature. As a secondary objective, we sought to explore whether clinical signs and laboratory parameters of volume depletion and acidosis are associated with an increased risk for AKI in these patients.
This study was conducted at BCCH, the only tertiary pediatric hospital serving British Columbia, Canada. Cases of DKA were identified using the Internal Classification of Diseases, Tenth Revision (ICD-10) codes for diabetes (E10-14) for the period September 1, 2008, through December 31, 2013. Patients were eligible for inclusion in the study if they were aged 18 years or younger and had confirmed DKA (blood glucose level ≥200 mg/dL, pH ≤7.3 or bicarbonate level ≤15 mEq/L, and elevation of serum or urine ketones). (To convert glucose level to millimoles per liter, multiply by 0.0555; to convert serum bicarbonate level to millimoles per liter, multiply by 1.) Cases with incomplete medical documentation were excluded. The University of British Columbia Clinical Research Ethics Board reviewed and approved the study protocol and waived the requirement for patient consent. All data collection occurred between September 8, 2014, and June 26, 2015. Data analysis took place from August 25, 2015, to June 8, 2016.
Paper and electronic records were reviewed. Medical history, current illness, and hospital admission were captured. Clinical and biochemical assessments, including initial estimated level of volume depletion and fluid resuscitation, were recorded. Serum creatinine level was measured at BCCH using the enzymatic Ortho Diagnostics Vitros CREA method (Ortho-Clinical Diagnostics Inc). Initial physical examination data, including height, weight, blood pressure, and heart rate, were recorded.
The primary outcome variable, AKI, was defined by the Kidney Disease/Improving Global Outcomes (KDIGO) serum creatinine criteria.12 Because no study participants had available baseline serum creatinine values prior to admission, we used an estimated glomerular filtration rate (GFR) of 120 mL/min/1.73 m2 to calculate an expected baseline creatinine level (EBC) using the Schwartz estimating equation.13 A GFR of 120 mL/min/1.73 m2 was selected on the basis of previously established standards in pediatric AKI studies.14-16 However, with normal GFR values13 ranging from 90 to 120 mL/min/1.73 m2, a baseline GFR of 90 mL/min/1.73 m2 was also used in a sensitivity analysis to calculate the most conservative estimates of AKI. Stage 0 (no AKI) was defined as occurring if all creatinine values were less than 1.5 times the EBC. Stage 1 AKI occurred if a creatinine value was 1.5 to less than 2 times the EBC. Stage 2 AKI occurred if a creatinine value was 2 to less than 3 times the EBC. Stage 3 AKI occurred if a creatinine value was 3 or more times the EBC. KDIGO AKI urine output criteria were not used, because recording of hourly urine output rates was inconsistent between cases. Corrected sodium (Na) was calculated using this formula7:
Corrected Na = [(Glucose (mg/dL)/18) − 5.6] × 0.36 + Serum Na.
A clinical estimation of volume depletion was documented as mild, moderate, or severe by using the clinical categories on BCCH’s DKA protocol for infants and children.
Continuous variables were presented as median and interquartile range (IQR), while categorical variables were presented as number and percentage. For the regression analysis, AKI (the primary outcome variable) was classified as no AKI (stage 0), mild AKI (stage 1), and severe AKI (stages 2 and 3). Stages 2 and 3 AKI were combined into a severe AKI category because these stages represent more substantial tubular injury and were previously shown to be associated with poorer outcomes, including increased mortality and hospital length of stay, in other pediatric populations.15,17-19 Multinomial logistic regression (using no AKI as the reference category) was used to identify potential factors associated with AKI. Variables were selected a priori on the basis of clinical importance and included age and sex as well as initial values for clinically estimated volume depletion, initial heart rate, and serum bicarbonate, corrected serum sodium, and hematocrit levels. Note that bicarbonate level was categorized using the accepted cut point between mild and moderate to severe DKA (10 mEq/L).20 Corrected sodium level was categorized as low (≤134 mEq/L), normal (135-144 mEq/L), or elevated (≥145 mEq/L) because both hyponatremia and hypernatremia can occur with kidney injury. (To convert sodium level to millimoles per liter, multiply by 1.) Hematocrit level was categorized as either elevated or not elevated, with a cut point of 45% chosen to reflect substantial hemoconcentration. (To convert hematocrit level to a proportion of 1, multiply by 0.01.) Variables that were significant at P < .10 in the unadjusted analysis were considered for the multivariable model. Age and sex were added to the multivariable model irrespective of their significance in the unadjusted analysis. The effect measures were reported as odds ratio (OR) and corresponding 95% CI. All the analyses were conducted in SAS/STAT software version 9.4 (SAS Institute Inc).21 All tests were 2-sided at a significance level of P = .05.
From September 1, 2008, to December 31, 2013, a total of 211 children were hospitalized for DKA at BCCH. Of these, 165 admissions met the inclusion criteria for analysis. Reasons for exclusion included the following: patient with type 1 diabetes was admitted for a non-DKA reason (n = 2), DKA criteria were not fully met (n = 8), medical record was not available (n = 3), medical documentation was incomplete (n = 31), and patient had neonatal diabetes (n = 2).
Table 1 highlights the clinical characteristics for the entire study population and for each AKI severity stage on admission. Overall, the admissions for DKA were characterized by a slight preponderance of females (89 [53.9%]) with a median age of 10.6 years (interquartile range, 5.1-13.8 years). Over half of the admissions (90 [54.5%]) were transfers from other hospitals within British Columbia, and about one-quarter of cases (40 [24.2%]) required intensive care unit (ICU) care. Three-quarters of cases (125 [75.8%]) were newly diagnosed with type 1 diabetes during the hospitalization.
One hundred six patients (64.2%) developed AKI during the admission. Of those with AKI, 37 (34.9%), 48 (45.3%), and 21 (19.8%) reached a maximum AKI stage of 1, 2, and 3, respectively. The proportion of AKI in children with DKA was higher in those admitted to the pediatric ICU (34/40 [85%]) compared with those who received care on the general pediatrics ward (72/125 [57.6%]). Of the 106 patients, 105 (99.1%) developed AKI within 24 hours of hospitalization. Starting with their initial AKI severity staging within the first 24 hours, 6 cases subsequently increased in AKI severity from 24 to 48 hours of hospitalization (cases increased in AKI severity from stage 0 to 1 [1 case], 1 to 2 [2 cases], and 2 to 3 [3 cases]), and 1 additional case increased in severity from stage 2 to 3 AKI at 48 to 72 hours of hospitalization. All cases had reached their maximum documented AKI stage by 72 hours. Finally, 2 patients required hemodialysis for indications of fluid overload or oliguria during their hospitalization. The first was a 13-year-old girl who underwent dialysis for 3 days but still had stage 3 AKI at discharge. The next follow-up measurement of her creatinine level 9 months later had normalized with no evidence of hypertension or albuminuria noted in the nephrology clinic. The second patient was an 11-year-old girl who was treated with hemodialysis for 2 days and then transferred to another hospital with no follow-up records available. No cases of cerebral edema or deaths were documented.
Fifty-four of 106 cases (50.9%) had documented resolution of their AKI by 72 hours of hospitalization. Another 4 cases (3.8%) had documented resolution by 96 hours. Two patients (1.9%) were transferred to other centers with ongoing AKI; therefore, we were unable to document AKI resolution; notably, 1 of these 2 patients underwent dialysis at BCCH. Furthermore, 2 cases (1.9%), 1 of whom also had been receiving dialysis, still had AKI at the time of discharge, but a follow-up visit to the nephrology clinic was arranged as an outpatient. Finally, 44 cases (41.5%) did not have documentation of AKI resolution prior to discharge, and nephrology clinic follow-up was not arranged. Of these cases, 22 (50%) had severe AKI at their last measured creatinine level. Of the 44 cases discharged without documented AKI resolution, the median (IQR) peak and discharge serum creatinine levels were 0.92 mg/dL (IQR, 0.79-1.02) at peak and 0.89 mg/dL (IQR, 0.79-0.95) at discharge for those with stage 1 AKI at discharge; 1.18 mg/dL (IQR, 1.10-1.36) at peak and 1.18 mg/dL (IQR, 1.1-1.3) at discharge for those with stage 2 AKI at discharge; and 1.31 mg/dL (IQR, 1.12-1.36) at peak and 1.31 mg/dL (IQR, 1.12-1.36) at discharge for those with stage 3 AKI at discharge. (To convert creatinine to micromoles per liter, multiply by 88.4.)
Table 2 highlights the unadjusted and adjusted analysis using a multinomial logistic regression model. All variables except age and sex were statistically significant in the unadjusted analysis, especially when comparing no AKI with severe AKI. In the adjusted analysis, a lower serum bicarbonate level of less than 10 mEq/L (compared with ≥10 mEq/L) was associated with a 5-fold increase in the odds of severe AKI (adjusted odds ratio [aOR], 5.22; 95% CI, 1.35-20.22), whereas each incremental increase of 5 beats/min in initial heart rate was associated with a 22% increase in the odds of severe AKI (aOR, 1.22; 95% CI, 1.07-1.39). Initial corrected sodium of 145 mEq/L or greater (compared with 135-144 mEq/L) was associated with a 3-fold increase in the odds of mild AKI (aOR, 3.29; 95% CI, 1.25-8.66).
The eTable in the Supplement contains a sensitivity analysis using an estimated GFR of 90 mL/min/1.73 m2 for calculating the expected baseline creatinine. Based on this lower estimated GFR cutoff, 69 children (41.8%) met the criteria for AKI. Of these children with AKI, 37 (53.6%) had mild and 32 (46.4%) had severe AKI. In multinomial regression analysis based on these estimates, a serum bicarbonate of less than 10 mEq/L was associated with an increase in the odds of severe AKI, as in the main analysis, but this time with more than 10-fold increase in the odds of severe AKI (aOR, 10.95; 95% CI, 1.10-108.52). Each incremental increase of 5 beats/min in initial heart rate was associated with an 18% increase in the odds of mild AKI (aOR, 1.18; 95% CI, 1.03-1.35) but not severe AKI (aOR, 1.16; 95% CI, 0.99-1.35), while initial corrected sodium did not reach statistical significance.
To our knowledge, this is the first study to report on the proportion of children hospitalized with DKA who develop AKI. Of note, we documented that 64% of patients met the criteria for AKI during their hospitalization for DKA. Even with the use of the most conservative estimate of baseline renal function, the proportion of children with AKI was still high at 42% in the sensitivity analysis. Of further concern was that 65% of patients with AKI met the criteria for severe AKI (stage 2 or 3), suggesting that a large number of children with DKA may experience intrinsic tubular injury rather than a milder prerenal or volume-responsive injury. Finally, our findings also suggest an association between the severity of acidosis and volume depletion and the risk for severe AKI.
Two retrospective studies22,23 report AKI in adults with DKA. Woodrow and colleagues22 reviewed 1661 cases of acute renal failure seen over a 36-year period and identified 14 cases directly associated with an episode of DKA in adults with type 1 diabetes. These patients had a 50% mortality rate, and all of those identified as having AKI in the final 10 years of the study required ICU management.22 Orban and colleagues23 subsequently reviewed 94 DKA admissions in a medical ICU and found a 50% AKI incidence.
No epidemiologic studies on AKI in children with DKA exist for comparison, but 2 large studies of 3396 and 2106 pediatric ICU admissions revealed AKI rates in pediatric ICU populations of 10% and 17.9%, respectively.24,25 In contrast, in our study the proportion of AKI in children with DKA admitted to the ICU was 85%. We postulate that the high AKI rate in ICU hospitalizations for DKA compared with the broad ICU population is associated with the severe intravascular depletion inherent in more severe cases of DKA. This theory is supported by our study findings that clinical markers of volume depletion, specifically, elevated heart rate and corrected sodium level at presentation to the hospital, were both associated with more severe AKI. In comparison, Orban and colleagues23 found no difference in serum sodium level among patients with and without AKI; however, sodium values in this study were not corrected for serum glucose level.
In addition to severity of volume depletion, more severe acidosis was also associated with severe AKI in our study. A serum bicarbonate level of less than 10 mEq/L on presentation was associated with a 5-fold increased odds for the development of severe AKI compared with a serum bicarbonate level of 10 mEq/L or higher. Whether severe acute acidosis has direct physiological effects, including renal vasoconstriction that causes more severe AKI in this population, is not known. Orban and colleagues23 found no statistically significant difference in serum bicarbonate or serum pH in adults with DKA with and without AKI; however, their selection criteria were for patients admitted to the ICU for severe DKA, and the entire group had a median bicarbonate level of 6.5 mEq/L, which indicates that nearly all study participants had a bicarbonate level more severe than the at-risk level identified in our study (ie, bicarbonate <10 mEq/L).
The high rates of AKI in our study are concerning for both short-term and long-term adverse health outcomes. Pediatric AKI is associated with increased morbidity, including greater need for and duration of ventilation, increased length of stay, higher hospitalization costs, and increased mortality.16,24,26-32 In addition, for those who survive AKI, recent data suggest that AKI likely results in permanent renal damage; these findings have challenged the previous notion that AKI was a completely reversible event. In a recent systematic review that included 13 cohort studies of adults with AKI, a single episode of AKI was associated with an increased risk of developing chronic kidney disease, with a pooled adjusted hazard ratio of 8.8 (95% CI, 2.1-25.5).33 Several observational studies of children have also supported the link of AKI and long-term renal dysfunction, with chronic kidney disease incidence ranging from 10% to 69% following an episode of AKI.34-39
The patients with AKI in our study are of particular interest given that children with type 1 diabetes are already at risk for renal complications from diabetes.40 Although progression to macroalbuminuria and renal insufficiency in childhood is rare, diabetic nephropathy is a leading cause of end-stage renal disease in adults.40,41 For this reason, it is especially important to develop strategies to identify, monitor, and treat additional renal insults, such as those that may occur with DKA.
Of concern was that 44 patients (42%) with AKI did not have documented resolution of AKI prior to discharge or arrangements for follow-up in the nephrology clinic. Of note, the final AKI stage was severe for 50% of these children. These data highlight that AKI is underrecognized both because of a lack of awareness of AKI as a complication of DKA and because the serum creatinine level in pediatric patients must be interpreted in the context of the child’s age and height. It is crucial to develop or have in place systems that identify and monitor abnormal markers of renal function in this population. Furthermore, prospective longitudinal studies are needed to assess the effect of these AKI episodes on the trajectory of renal disease in children with diabetes.
This study should be interpreted within the context of its limitations. The main limitation of this study is its retrospective nature; therefore, the results rely on the accuracy and completeness of patient records. Furthermore, we did not have complete urine output data with which to compare both the creatinine and urine output criteria within the KDIGO AKI classification. In addition, as is common in pediatric populations, baseline serum creatinine values were not available, resulting in the need to calculate an estimated baseline value. Therefore, our results must be confirmed with prospective longitudinal studies. Nevertheless, this is the first large-scale study to report AKI in pediatric patients with DKA. Furthermore, our sample size is larger than those of the 2 existing studies22,23 assessing AKI in adults with DKA, and we were able to identify clinical factors that are associated with severe AKI in this pediatric population.
To our knowledge, this is the first study to document a high proportion of children with AKI among those hospitalized with DKA. Severe AKI was associated with worsening markers of volume depletion and acidosis. Overall, these data suggest that clinicians should consider AKI as a frequent complication that accompanies pediatric DKA and should be especially alert to its presence in severe presentations of DKA. Prospective longitudinal studies are urgently needed to better understand both the risk factors for and long-term implications of AKI in this population, especially because these children are already at risk for chronic kidney disease secondary to diabetic nephropathy over the long term.
Corresponding Author: Constadina Panagiotopoulos, MD, FRCPC, Endocrinology & Diabetes Unit, British Columbia Children’s Hospital, 4480 Oak St, ACB K4-213, Vancouver, BC V6H 3V4, Canada (firstname.lastname@example.org).
Accepted for Publication: January 4, 2017.
Published Online: March 13, 2017. doi:10.1001/jamapediatrics.2017.0020
Author Contributions: Drs Hursh and Ronsley share first authorship. Dr Panagiotopoulos had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hursh, Mammen, Panagiotopoulos.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Hursh, Ronsley, Islam, Panagiotopoulos.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Hursh, Ronsley, Islam.
Study supervision: Mammen, Panagiotopoulos.
Conflict of Interest Disclosures: None reported.
Additional Contributions: Claire Ronsley, BSc, Division of Endocrinology, British Columbia Children’s Hospital, assisted with the data collection for this study. Ms Ronsley did not receive compensation for her contribution.