Diabetes Keto Acidosis

Cognitive Function Following Diabetic Ketoacidosis in Children With New-Onset or Previously Diagnosed Type 1 Diabetes Diabetes Care 2020;43:2768–2775 | https://doi.org/10.2337/dc20-0187

OBJECTIVE

This study assessed whether a single diabetic ketoacidosis (DKA) episode is associated with cognitive declines in children with newly diagnosed type 1 diabetes and whether the same is true in children who had previously been diagnosed after accounting for variations in glycemic control and other relevant factors.

RESEARCH DESIGN AND METHODS

Weprospectivelyenrolled758children,6–18yearsold,whopresentedwithDKAina randomized multisite clinical trial evaluating intravenous fluid protocols for DKA treatment. DKA was moderate/severe in 430 children and mild in 328 children. A total of 392 children with DKA had new onset of type 1 diabetes, and the rest were previously diagnosed. Neurocognitive assessment occurred 2–6 months after the DKA episode. A comparison group of 376 children with type 1 diabetes, but no DKA exposure, was also enrolled.

RESULTS

Among all patients, moderate/severe DKA was associated with lower intelligence quotient (IQ) (b 5 20.12, P < 0.001), item-color recall (b 5 20.08, P 5 0.010), and forward digit span (b 5 20.06, P 5 0.04). Among newly diagnosed patients, moderate/severe DKA was associated with lower item-color recall (b 5 20.08, P 5 0.04). Among previously diagnosed patients, repeated DKA exposure and higher HbA1c were independently associated with lower IQ (b 5 20.10 and b 5 20.09, respectively, P < 0.01) and higher HbA1c was associated with lower item-color recall (b 5 20.10, P 5 0.007) after hypoglycemia, diabetes duration, and socioeconomic status were accounted for.

CONCLUSIONS

A single DKA episode is associated with subtle memory declines soon after type 1 diabetes diagnosis. Sizable IQ declines are detectable in children with known diabetes, suggesting that DKA effects may be exacerbated in children with chronic exposure to hyperglycemia.

Diabetic ketoacidosis (DKA) is a common complication of type 1 diabetes (1). Of children with new onset of type 1 diabetes, 25–40% present with DKA (2). In established patients, DKA may be the consequence of poor adherence to insulin regimens, illness, or malfunction of diabetes care equipment (e.g., insulin pumps) (2). Brain injury has long been recognized as an uncommon, but serious, complication of

1Department of Psychology, University of Cal- ifornia, Davis, Davis, CA 2Center for Mind and Brain, University of Cal- ifornia, Davis, Davis, CA 3Department of Emergency Medicine, UC Davis Health, UC Davis School of Medicine, Sacramento, CA 4Department of Pediatrics, UC Davis Health, UC Davis School of Medicine, Sacramento, CA 5Division of Emergency Medicine, Department of Pediatrics, Children’s Hospital Colorado, Univer- sity of Colorado School of Medicine, University of Colorado Denver, Aurora, CO 6Division of Emergency Medicine, Department of Pediatrics, Children’s Hospital of Philadelphia, Perelman School of Medicine, University of Penn- sylvania, Philadelphia, PA 7Department of Pediatrics, University of Utah School of Medicine, Salt Lake City, UT 8Division of Emergency Medicine, Department of Pediatrics, Nationwide Children’s Hospital, The Ohio State University College of Medicine, Co- lumbus, OH 9Departments of Emergency Medicine and Pedi- atrics, Rhode Island Hospital, The Warren Alpert Medical School, Brown University, Providence, RI

Simona Ghetti,1,2 Nathan Kuppermann,3,4

Arleta Rewers,5 Sage R. Myers,6

Jeff E. Schunk,7 Michael J. Stoner,8

Aris Garro,9 Kimberly S. Quayle,10

Kathleen M. Brown,11 Jennifer L. Trainor,12

Leah Tzimenatos,3 Andrew D. DePiero,13

Julie K. McManemy,14 Lise E. Nigrovic,15

Maria Y. Kwok,16 Clinton S. Perry III,2,17

Cody S. Olsen,7 T. Charles Casper,7 and

Nicole S. Glaser,4 for the Pediatric

Emergency Care Applied Research Network

(PECARN) DKA FLUID Study Group*

2768 Diabetes Care Volume 43, November 2020

P A T H O P H Y S IO LO

G Y /C O M P LI C A T IO N S

DKA (3,4). Recent studies, however, sug- gest that subtle brain injury occurs even among children with no obvious neuro- logical symptoms during DKA (5,6). Re- ports document that DKA is associated with alterations in memory, attention, verbal intelligence quotient (IQ) (7–12), and changes in brain microstructure (10–12). Despite this evidence, several ques-

tions remain. First, it is unclear whether a single DKA episode results in lasting cognitive declines that are detectable shortly after the DKA episode in children with new onset of type 1 diabetes. Most studies reporting lower cognitive func- tioning after a single DKA episode were retrospective, included children previ- ously diagnosed with type 1 diabetes, or documented deficits in comparison withcontrolsubjectswith nodiagnosisof type 1 diabetes (7–9). One small pro- spective study of newly diagnosed pa- tients found that those who experienced DKA(compared with a sampleof children with type 1 diabetes and no histories of DKA) showed subtle mental status and memory deficits between 48 h and 5 days after DKA (12). These group differ- ences no longer persisted, however, 1 month and 6 months after diagnosis (12). Among children who experienced DKA, the severity of the DKA episode and alterations in brain microstructure de- tected during DKA were associated with worse cognitive functioning 6 months after diagnosis (12). It is possible that variations in DKA severity predict cogni- tive functioning only among children who experienced DKA, without resulting in overall differences between children with and without DKA histories. How- ever, these findings suggest that large- scale studies are necessary to fully assess the neurocognitive effects of DKA in

children with newly diagnosed type 1 diabetes.

Second, reports of DKA-related cogni- tive deficits in children previously diag- nosed with type 1 diabetes raise the question of whether these deficits may be more extensive or become more ro- bust over time if DKA occurs in the setting of long-standing hyperglycemia (9). In a longitudinal study of children previously diagnosed with type 1 diabetes (36% with one DKA episode), DKA exposure was associated with lower IQ 18 months later (13). However, these studies have not been able to fully account for important correlatesofDKAoccurrenceandseverity, including glycemic variability (e.g., hyper- glycemia, hypoglycemia) (9), which are also associated with cognitive outcomes, and socioeconomic status (SES), which is generally associated with access to care and cognitive outcomes (14).

In the current study, we examined whether the severity of a single DKA episode was associated with cognitive deficits in children newly diagnosed with type1diabetes.Inaddition,weexamined whether DKA severity of a single episode and/or repeated episodes of DKA are associated with more general cognitive deficits in children with previously di- agnosed type 1 diabetes, after account- ing for measures of glycemic control, such as HbA1c and severe hypoglycemic events, duration of type 1 diabetes, and SES.

RESEARCH DESIGN AND METHODS

Children with type 1 diabetes between the ages of 6 and 18 years were recruited from sites participating in the Pediatric Emer- gency Care Applied Research Network (PECARN) Fluid Therapies Under Investiga- tion in DKA (FLUID) randomized controlled trial (15,16). The study was approved by the

InstitutionalReviewBoardateachsite,and written informed consent was obtained from each patient’s parents or guardians.

ThePECARNFLUIDtrialcomparedfour fluidrehydrationprotocolscapturingvar- iations in protocols currently used in the U.S. totreat DKA inchildren(15). DKA was defined by 1) blood glucose concentration .300 mg/dL and 2) venous pH ,7.25 or serum bicarbonate concentration ,15 mmol/L.Exclusioncriteriahavepreviously been described (16) and included condi- tions unrelated to DKA that affect mental status or cognitive abilities and/or sub- stantial treatment for DKA prior to trans- fer to the study centers. All patients’ medical records were reviewed by the child’s primary endocrinologist to deter- mine whether there were previous epi- sodes of DKA or severe hypoglycemia. Parents and guardians were also queried about any hospital admissions occurring at hospitals other than the study site. If any such episodes were recalled, the child’s endocrinologist reviewed these episodes with the family to verify that the admission was for DKA. For the pres- ent report, children were excluded if they experienced clinically apparent DKA- related brain injury (15) or could not return within 6 months for cognitive test- ing. Finally, children younger than 6 years of age were also excluded because the cognitive testing procedures used in the current study were not appropriate for younger children.

Children with type 1 diabetes, but no histories of DKA, were recruited from the pediatric diabetes clinics at the partici- pating PECARN centers. Absence of DKA at diagnosis and between diagnosis and follow-up, and episodes of severe hypo- glycemia, were verified through medical record review and confirmed by the patient’s endocrinologist and family.

10Division of Emergency Medicine, Department of Pediatrics, St. Louis Children’s Hospital, Wash- ington University School of Medicine in St. Louis, St. Louis, MO 11Division of Emergency Medicine, Department of Pediatrics, Children’s National Medical Center, The School of Medicine & Health Sciences, The George Washington University, Washington, DC 12Division of Emergency Medicine, Department of Pediatrics, Ann and Robert H. Lurie Children’s Hospital of Chicago, Northwestern University Feinberg School of Medicine, Chicago, IL 13Division of Emergency Medicine, Nemours/Alfred I. duPont Hospital for Children, Sidney Kimmel Medical College,ThomasJeffersonUniversity,Philadelphia,PA

14Division of Emergency Medicine, Department of Pediatrics, Texas Children’s Hospital, Baylor College of Medicine, Houston, TX 15Division of Emergency Medicine, Department of Pediatrics, BostonChildren’s Hospital, Harvard Medical School, Boston, MA 16Division of Emergency Medicine, Department of Pediatrics, New York Presbyterian Morgan Stanley Children’s Hospital, Vagelos College of Physicians and Surgeons, Columbia University, New York, NY 17Department of Psychology, Tufts University, Medford, MA

Corresponding author: Simona Ghetti, sghetti@ ucdavis.edu

Received 26 January 2020 and accepted 18 August 2020

This article contains supplementary material online at https://doi.org/10.2337/figshare.12824396.

*A list of members of the Pediatric Emergency Care Applied Research Network (PECARN) DKA FLUID Study Group can be found in the supple- mentary material online.

© 2020 by the American Diabetes Association. Readersmayuse this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More infor- mationisavailableathttps://www.diabetesjournals .org/content/license.

care.diabetesjournals.org Ghetti and Associates 2769

Recruitment was targeted to maintain approximately equal proportions of pa- tients with new onset of type 1 diabetes and patients with known diagnoses of type 1 diabetes in the DKA and non-DKA groups. Patients with DKA returned between

2 and 6 months after hospital discharge for neurocognitive assessment. Patients with type 1 diabetes but no DKA histories were scheduled at their earliest conve- nience. The assessment for both the DKA and non-DKA groups was resched- uled if children had either hypoglycemia (glucose ,70 mg/dL) or hyperglycemia (glucose.350mg/dL)withketosis(urine dipstick with moderate or large ketones) at the time of the visit.

Outcomes A comprehensive neurocognitive assess- ment included the following: 1) color and spatial memory tasks evaluating long- term memory for pictures and associa- tion with the correct color background or spatial location with which children were initially presented (range 0–1); 2) Wechsler Abbreviated Scale of Intelli- gence, yielding an IQ score (range 65– 160) (17); and 3) digit span forward and backward task (range 0–18), evaluating short-term and working memory. Higher scorescorrespondto betterperformance. We used these measures as in our previous research (7,15,18) (see supplementary materials for detailed description).

Statistical Analyses We described the characteristics of pa- tients with new onset of type 1 diabetes and those previously diagnosed with and without histories of DKA using relative frequencies for categorical characteris- tics and means and SDs for continuous characteristics. We tested for differences as a function of DKA status using likeli- hood ratio tests and ANOVA tests for categorical and continuous characteris- tics, respectively. We used x2 and Wil- coxon rank sum tests to determine whether participants with DKA histories who completed the neurocognitive as- sessments differed from those who were lost to follow-up. To include all patients with observed

outcomesintheanalyses,weusedchained regression equations for imputation of missing data (19). Ten imputations were used, and results were combined using standard methods (20–22) (Supplementary

Table 1). Multiple imputation was used for predictors so that all patients with outcomes were included in the statisti- cal analyses; imputed values were not used foroutcome measures. Patients with moderate/severe DKA and mild DKA and without DKA histories were compared on all of the outcome measures with use of mixed linear models. All models included random terms for the enrolling clinical center.

Initially, models were applied to all patients and fit to item-color and item- spacememory tasks, IQ, and forward and backward digit span scores separately. The focal independent variable was DKA status,whichwasbasedonvenouspH(or serum bicarbonate concentration in pa- tients without a measured pH at pre- sentation) inpatientswith DKA and was a numeric scale defined as 2, moderate/ severe (pH #7.19 or bicarbonate con- centration #9 mmol/L); 1, mild (pH between 7.20 and 7.25 or bicarbonate between 10 and 15 mmol/L) (23); or 0, no DKA history (pH .7.25 or bicarbonate .15 mmol/L for patients with new onset of type 1 diabetes and no recorded DKA episodes for patients with established type 1 diabetes). We also included sex, SES (measured by maternal parental education on a numeric scale: 0 5 high school/GED or less, 1 5 some college/vocational school, 2 5 college degree or more), diabetes history (new onset vs. previously diagnosed), and his- tory of severe hypoglycemic episodes (one or more vs. none). Age was included in models for item-color memory, item- space memory, and digit span scores but not IQ, which is normalized by age. In- dependent variables were tested for an association with use of type III F tests and a significance level of 0.05. We calculated 95% CIs for standardized re- gression parameters.

We subsequently applied these mod- els separately to patients with new onset and those previously diagnosed with type 1 diabetes. We report results only for outcomes showing associations with DKA status in the initial models including all patients. In models pertaining to pre- viously diagnosed patients, we addition- ally included 12-month average HbA1c, diabetes duration, and whether DKA had previously been experienced as indepen- dent variables. Again, F tests were used to identify significant associations, and 95% CIs were calculated for standardized

regression coefficients. Analyses were performed using SAS software (version 9.4; SAS Institute, Cary, NC).

In exploratory analyses restricted to children with new onset of type 1 di- abetes, we examined the correlation be- tween pH at the time of diagnosis and IQ using Spearman correlation coefficients. In other exploratory analyses restricted to previously diagnosed patients, we in- cluded 1) the interaction between DKA severity and SES in the regression models described earlier, given differences in SES across DKA status groups in these pa- tients, and the possibility that variability in SES was confounded with or obscured DKA status effects, and 2) the interaction between DKA exposure and age at onset of type 1 diabetes to explore whether the effects of DKA were stronger in children with onset of type 1 diabetes in early childhood.

RESULTS

A description of patient enrollment can be found in a flow diagram (Supplementary Fig. 1). Between February 2011 and Sep- tember 2016, 1,249 children with DKA were enrolled in the PECARN FLUID trial and were eligible for neurocognitive follow- up; of these, 901 (72.1%) returned for follow-up, and 758 (84.1%) of patients who returned were retained for the pres- ent report based on eligibility criteria. Of these children (Table 1), 430 (56.7%) ex- perienced moderate or severe DKA and 328 mild DKA (43.3%); 392 (51.7%) had new onset of diabetes, and 366 were previously diagnosed. Among previously diagnosed patients, 270 (73.8%) had at least one DKA episode prior to participat- ing.ChildrenwithDKAexposurewhowere lost to follow-up were older, more likely to be previously diagnosed, and more likely to have higher mean HbA1c levels than those who completed the neurocognitive assessments (Supplementary Table 2). In addition, we assessed 376 children with type 1 diabetes and no DKA histories; 199 (52.9%) of these had new onset of type 1 diabetes.

Among newly diagnosed patients, there were no significant differences in demographic variables as a function of DKA status (Table 1). Among the children with previously diagnosed type 1 diabe- tes,differencesasafunctionofDKAstatus were found in several demographic and diabetes-related variables (Table 1).

2770 Pediatric Diabetic Ketoacidosis and Cognition Diabetes Care Volume 43, November 2020

The first analysis included all partic- ipants (i.e., both newly diagnosed and previously diagnosed patients) and com- paredcognitiveoutcomesin childrenwith moderate or severe DKA, mild DKA, or no DKA history. Additional variables in the model included age at testing, sex, SES,

experience of severe hypoglycemia, and whetherchildren werenewly orpreviously diagnosed. DKA status was associated with lower scores in overall IQ, memory for item-color associations, and forward digit span recall (Table 2). Backward digit span recallandmemoryforitem-spacerelations

did not differ between moderate/severe DKA, mild DKA, and non-DKA groups (Supplementary Table 3). Children with pre- viously diagnosed type 1 diabetes showed lower performance compared with chil- dren with new onset in the item-color associations and overall IQ (Table 2).

Table 1—Baseline characteristics of the enrolled patients by DKA status

Non-DKA Mild DKA Moderate/severe DKA P*(N 5 376) (N 5 328) (N 5 430)

New onset, N (%) 199 (52.9) 198 (60.4) 194 (45.1) ,0.001 Male, N (%) 108 (54.3) 94 (47.5) 92 (47.4) 0.29 Age at testing, mean (SD) 11.3 (3.16) 11.5 (2.79) 11.6 (2.85) 0.52 Age at onset of diabetes, mean (SD) 10.7 (3.17) 10.8 (2.79) 10.9 (2.89) 0.90 Glucose at neurocognitive testing, mean (SD) 164.4 (68.78) 163.0 (73.57) 167.0 (77.93) 0.86 Any hypoglycemic episodes prior to testing, N (%) 0 (0.0) 19 (9.6) 11 (5.7) ** SES, N (%)† 0.74 High school/GED or less 48 (24.1) 47 (23.9) 57 (29.4) Some college/vocational school 54 (27.0) 52 (26.3) 50 (25.8) College degree or more 97 (48.9) 98 (49.7) 87 (44.8)

Previously diagnosed, N (%) 177 (47.1) 130 (39.6) 236 (54.9) Male, N (%) 87 (49.2) 58 (44.6) 106 (44.9) 0.64 Age at testing, mean (SD) 12.2 (3.18) 13.1 (2.83) 14.0 (2.56) ,0.001 Age at onset of diabetes, mean (SD) 8.3 (3.46) 6.7 (3.55) 7.8 (3.66) ,0.001 Duration of diabetes in years, mean (SD) 3.4 (3.33) 5.9 (3.43) 5.7 (3.33) ,0.001 Glucose at neurocognitive testing, mean (SD) 206.4 (87.30) 247.7 (103.07) 230.9 (97.37) ,0.001 Any hypoglycemic episodes prior to testing, N (%) 12 (6.9) 35 (27.1) 48 (20.4) ,0.001 One or more additional DKA episodes, N (%) 0 (0.0) 101 (77.5) 169 (71.5) ** 12-month average HbA1c result, mean (SD) 8.3 (1.34) 10.2 (1.91) 10.6 (1.75) ,0.001 SES, N (%)† 0.003 High school/GED or less 32 (18.0) 47 (36.4) 100 (42.5) Some college/vocational school 48 (27.2) 46 (35.6) 75 (31.6) College degree or more 97 (54.8) 36 (28.0) 61 (25.9)

*P values are from likelihood ratio x2 tests for categorical characteristics [those with N (%)] and ANOVA F tests for continuous characteristics [those with mean(SD)],usingmethodsforcombiningresultsacrossmultipleimputeddatasets.**Pvalueswerenotcalculatedduetodesign-imposedrelationshipswith DKA status. †Maternal education if documented. If not, imputed using paternal education and household income in a multinomial regression model.

Table 2—Regression models applied to all patients: standardized regression coefficients and 95% CIs*† and adjusted means*† and raw means by DKA status

Color task IQ Digit span recall: forward

Coefficient (95% CI) P Coefficient (95% CI) P Coefficient (95% CI) P

DKA status‡ 20.08 (20.13, 20.02) 0.010 20.12 (20.19, 20.06) ,0.001 20.06 (20.12, 20.00) 0.04

Male 20.05 (20.10, 0.00) 0.07 0.01 (20.04, 0.07) 0.67 20.00 (20.06, 0.05) 0.88

Age 0.47 (0.41, 0.53) ,0.001 Not estimated 0.41 (0.35, 0.47) ,0.001

Previously diagnosed# 20.07 (20.13, 20.01) 0.02 20.08 (20.14, 20.03) 0.004 20.05 (20.11, 0.01) 0.10

SES§ 0.10 (0.04, 0.16) ,0.001 0.32 (0.26, 0.38) ,0.001 0.12 (0.06, 0.17) ,0.001

Previous severe hypoglycemia 0.01 (20.05, 0.07) 0.72 20.03 (20.09, 0.03) 0.30 0.00 (20.05, 0.06) 0.96

Color task IQ Digit span recall: forward

DKA status Adjusted mean (SE) Mean (SD) Adjusted mean (SE) Mean (SD) Adjusted mean (SE) Mean (SD)

No DKA 0.50 (0.012) 0.50 (0.180) 104.6 (1.15) 107.3 (13.95) 8.5 (0.14) 8.6 (2.25)

Mild DKA 0.48 (0.010) 0.49 (0.170) 102.6 (1.02) 103.7 (13.46) 8.4 (0.12) 8.2 (2.11)

Moderate/severe DKA 0.47 (0.011) 0.49 (0.168) 100.7 (1.12) 101.2 (12.31) 8.2 (0.13) 8.5 (2.06)

*Regressioncoefficientsarestandardizedandrepresentthemeanchangeinoutcomeassociatedwitha1-SDchangeinacontinuousfactororthechange associated with that level (male, previously diagnosed, previous severe hypoglycemia) for binary variables. Note that the estimated change for the reference value (female, new onset, no hypoglycemia) is 21 times the coefficient. Adjusted means average over sex, previous diagnosis, and previous severe hypoglycemia and assume age 5 12 and SES 5 1. †Random effect of site included in each model. ‡DKA status is on a numeric scale: 0 5 no DKA, 15mildDKA,25moderate/severeDKA.#Previouslydiagnosed:05newlydiagnosedpatients and1 5previouslydiagnosedpatients,allowingforthe comparison of these two groups. §SES is indicated by maternal education and is on a numeric scale: 0 5 high school/GED or less, 1 5 some college/ vocational school, 2 5 college degree or more; when data were missing, income and paternal education were used to impute SES.

care.diabetesjournals.org Ghetti and Associates 2771

The next set of analyses examined per- formance separately for patients with new onset of type 1 diabetes and those pre- viously diagnosed. We limited these anal- yses to the outcomes that showed an overall effect of DKA status in the previous analysis, namely, item-color memory, IQ, and forward digit span.

Patients With Newly Diagnosed Type 1 Diabetes The regression models including newly diagnosed patients revealed significant differences as a function of DKA status at presentation for item-color association memory but not for IQ and forward digit span (Table 3). In exploratory analyses restricted to patients who presented with DKA, lower venous pH was associ- ated with lower IQ (r 5 0.18, P 5 0.001).

This correlation retained its significance when we controlled for sex, SES, and glucose level at diagnosis of DKA and at testing (r 5 0.16, P 5 0.003).

Patients With Previously Diagnosed Type 1 Diabetes The models examining previously diag- nosed patients revealed that DKA status was not related to any of the cognitive outcomes but that greater number of previous DKA episodes and higher HbA1c were independently and negatively re- lated to IQ (Table 3). HbA1c was also negatively related to item-color memory (Table 3). However, DKA status was as- sociated with both mean HbA1c and SES among previously diagnosed patients in bivariable analyses (Table 1), limiting our ability to isolate the unique effects of

DKA among patients with poor glycemic control. Furthermore, the associations between DKA status and SES among previously diagnosed patients (Table 1) and SES and neurocognitive outcomes (Tables 2 and 3) raise the question of whether preexisting differences among patient groups may account for or ob- scure differences as a function of DKA status. To address this question, we conducted exploratory analyses in which an interaction term involving DKA status and SES was added to the regression model for previously diagnosed patients. These analyses revealed a significant in- teractioneffect(Supplementary Table4), such that DKA status was associated with lower IQ among children with higher SES (Fig. 1), whereas patients with lower SES exhibited lower IQ regardless of DKA

Table 3—Regression models applied separately to patients with new-onset and previously diagnosed diabetes

New-onset patients

Color task IQ Digit span recall: forward

Coefficient (95% CI) P Coefficient (95% CI) P Coefficient (95% CI) P

DKA status‡ 20.08 (20.16, 20.00) 0.04 20.06 (20.14, 0.02) 0.13 20.03 (20.11, 0.05) 0.44

Male 20.06 (20.13, 0.02) 0.14 0.02 (20.06, 0.09) 0.63 0.03 (20.04, 0.11) 0.37

Age 0.45 (0.37, 0.53) ,0.001 Not estimated 0.39 (0.31, 0.47) ,0.001

SES§ 0.11 (0.04, 0.19) 0.003 0.34 (0.26, 0.42) ,0.001 0.12 (0.04, 0.19) 0.003

Previous severe hypoglycemia 0.03 (20.08, 0.14) 0.61 20.07 (20.18, 0.04) 0.21 20.06 (20.17, 0.05) 0.28

Color task IQ Digit span recall: forward

DKA status Adjusted mean (SE) Mean (SD) Adjusted mean (SE) Mean (SD) Adjusted mean (SE) Mean (SD)

No DKA 0.52 (0.019) 0.49 (0.175) 103.9 (1.63) 106.9 (14.18) 8.3 (0.23) 8.5 (2.25)

Mild DKA 0.50 (0.016) 0.50 (0.159) 102.9 (1.43) 106.5 (13.15) 8.3 (0.20) 8.2 (2.06)

Moderate/severe DKA 0.48 (0.017) 0.47 (0.171) 101.9 (1.53) 103.9 (12.18) 8.2 (0.21) 8.4 (2.09)

Color task IQ Digit span recall: forward

Previously diagnosed patients Coefficient (95% CI) P Coefficient (95% CI) P Coefficient (95% CI) P

DKA status‡ 20.02 (20.12, 0.08) 0.66 20.08 (20.18, 0.02) 0.14 20.08 (20.18, 0.01) 0.10

Male 20.04 (20.12, 0.04) 0.36 20.00 (20.08, 0.07) 0.94 20.04 (20.12, 0.04) 0.30

Age 0.52 (0.42, 0.62) ,0.001 Not estimated 0.41 (0.32, 0.51) ,0.001

SES§ 0.06 (20.02, 0.15) 0.16 0.27 (0.18, 0.35) ,0.001 0.12 (0.04, 0.21) 0.005

Previous severe hypoglycemia 0.01 (20.06, 0.09) 0.77 0.03 (20.04, 0.10) 0.45 0.03 (20.04, 0.10) 0.47

Mean HbA1c 20.10 (20.17, 20.03) 0.007 20.09 (20.16, 20.02) 0.008 0.03 (20.04, 0.09) 0.45

Diabetes duration 20.00 (20.09, 0.09) .0.99 20.03 (20.12, 0.05) 0.42 0.02 (20.07, 0.11) 0.62

Previous DKA episodes 20.01 (20.09, 0.07) 0.80 20.10 (20.18, 20.03) 0.009 20.04 (20.12, 0.04) 0.35

Color task IQ Digit span recall: forward

DKA status Adjusted mean (SE) Mean (SD) Adjusted mean (SE) Mean (SD) Adjusted mean (SE) Mean (SD)

No DKA 0.47 (0.018) 0.51 (0.185) 103.0 (1.46) 107.6 (13.71) 8.5 (0.20) 8.7 (2.25)

Mild DKA 0.46 (0.013) 0.47 (0.185) 101.8 (1.14) 99.2 (12.78) 8.3 (0.14) 8.3 (2.18)

Moderate/severe DKA 0.46 (0.015) 0.51 (0.164) 100.6 (1.34) 98.9 (11.99) 8.0 (0.17) 8.6 (2.03)

Data shown are regression coefficients and 95% CIs and adjusted means and raw means as a function of DKA status (random effect of site included in each model). Regression coefficients are standardized and represent the mean change in outcome associated with a 1-SD change in a continuous factor or the changeassociatedwiththatlevel(male,previousseverehypoglycemia)forbinaryvariables.Notethattheestimatedchangeforthereferencevalue(female and no hypoglycemia) is 21 times the coefficient. Adjusted means are conditional on applicable covariates as follows: average sex and previous severe hypoglycemia,age512,SES51,meanHbA1c59.6%,diabetesduration55years,andonetotwopreviousDKAepisodes.‡DKAstatusisonanumericscale: 0 5 no DKA, 1 5 mild DKA, 2 5 moderate/severe DKA. §SES is indicated bymaternaleducation and is on a numeric scale: 0 5 high school/GED or less, 1 5 some college/vocational school, 2 5 college degree or more; when data were missing, income and paternal education were used to impute SES.

2772 Pediatric Diabetic Ketoacidosis and Cognition Diabetes Care Volume 43, November 2020

status. Finally, the broad range of age of onset of type 1 diabetes in this group and evidence that early age of onset may result in worse cognitive outcomes (7) motivatedtheexaminationofinteraction effects between age of onset and DKA exposure. We found an interaction effect (Supplementary Table 5) such that DKA exposure was associated with stronger decline in item-color memory if onset occurred prior to the child’s fourth birthday.

CONCLUSIONS

Since our initial report of memory de- clines following one single DKA episode (7), several studies have suggested that DKA is an important predictor of cogni- tive decline in children with type 1 diabetes, even in the absence of overt DKA-related brain injury (8–12). These studies, however, were retrospective, included small samples, or did not fully account for variability in glycemic control and other important variables, including SES (8–12). To our knowledge, this is the first large-scale, prospective study of DKA to differentiate associations between DKA status and cognition in children with new onset and children with a previous di- agnosis of type 1 diabetes. When both groups of patients were combined, we found lower performance in measures of short-term memory, long-term memory, and overall IQ among children who ex- perienced moderate/severe DKA. Analy- ses separating new onset from previously diagnosed patients provided a more nu- anced picture. Among newly diagnosed patients,

moderate/severe DKA was associated

with lower item-color memory. Greater acidosis was also associated with lower IQ in an analysis restricted to newly di- agnosed patients who experienced DKA. In addition, among newly diagnosed pa- tients, there were no differences in SES or in any other examined variable by DKA status (moderate/severe DKA, mild DKA, no DKA), suggesting that DKA-related effects are not due to preexisting differ- ences among these groups of newly di- agnosed patients. Instead, these findings suggest that DKA-related deficits are de- tectable soon after a single moderate/ severe DKA episode.

Previous literature does not paint a coherent picture about whether a single DKA episode at the time of diagnosis of type 1 diabetes suffices to generate cognitive declines that are detectable soon after recovery. Most studies report- ing the effects of a single DKA episode were retrospective or included patients with known type 1 diabetes (7–9), mak- ing it impossible to address this question directly. In a relatively small sample of prospectivelyexaminednewlydiagnosed patients, those who presented with DKA exhibited cognitive declines within days of the DKA episode; however, these group differences did not persist after 1–6 months (12). Interestingly, DKA se- verity and alterations in brain white matter integrity observed during DKA were associated with lower cognitive performance after 6 months. Consistent with these results, we found that in children with new onset of type 1 di- abetes, moderate/severe DKA was asso- ciated with lower item-color memory, and lower pH in DKA patients was cor- related with lower IQ, 2–6 months after diagnosis. Future research using a more extensive IQ measure may provide a more precise characterization of cogni- tive declines in individual subtests of IQ.

In contrast to our findings, a recent population study of Danish children did not find overall differences between children with and without DKA exposure in performance on standardized achieve- ment tests (24). However, DKA was docu- mented in only 18% of the patients, resulting in a sample that was substan- tially smaller than that examined here. Furthermore, the severity of DKA was not documented in this population study. If DKA was largely prevented or treated while still mild, neurocognitive deficits may have been less likely to emerge in that sample.

Among children previously diagnosed with type 1 diabetes, our findings show that elevated HbA1c was associated with lower item-color memory; overall DKA exposure was associated with lower item-color memory only among those who were diagnosed prior to age 4 years, consistent with other reports of increased vulnerability following early diagnosis (7,25). Previous DKA exposure and ele- vated HbA1c were independently associ- ated with lower IQ. DKA severity in the episode targeted in our study was associ- ated with lower IQ when these other factors were accounted for only among children with higher SES; lower-SES pa- tients showed lower performance overall, which was not further impacted by DKA severity. This finding counters the hy- pothesis that the effects of DKA status entirely reflect preexisting group differ- ences including a greater probability of experiencing DKA (or severe DKA) in low- SES children, who are also more likely to exhibit lower cognitive performance. Our results suggest instead that the unique ef- fects of DKA severity of a recent episode may be obscured or subsumed by other associated variables or risk factors among previously diagnosed patients.

Other evidence (13,26) suggests that DKA severity in previously diagnosed pa- tients is associated with lower IQ 18 months later. In that research, however, glycemic variability and SES were not fully accounted for, preventing firm conclu- sions about the unique effects of DKA. The IQ declines associated with DKA over an 18-month period (13,26) highlight that these deficits may worsen over time, placing children at increased risk for ac- ademic difficulties or problems managing type 1 diabetes independently.

The examination of a large cohort al- lowed us to capture more complex asso- ciations among relevant variables. The strong associations of DKA severity, SES, and poor glycemic control underscore that measures of cumulative DKA expo- sure and glycemic control may capture much of the variance in behavioral per- formanceand that patientswith repeated DKA exposure and poorly controlled type 1 diabetes are at substantial risk of cog- nitive deficits.

Our findings confirm that subtle brain injuries may occur as a result of DKA, but theunderlyingmechanismisstillunclear. Previous hypotheses suggested that os- motic declines during DKA treatment

Figure 1—Interaction effect between DKA status and levels of SES among previously diagnosed patients. Estimates and 95% CIs of IQ average over other effects in a multivar- iable mixed linear regression model, which included sex, duration of diabetes, previous DKAepisodes,historyofseverehypoglycemia, HbA1c, and study site. GED, General Educa- tional Development.

care.diabetesjournals.org Ghetti and Associates 2773

might cause cerebral edema resulting in cerebral injury. These hypotheses were largely disproven in our recent random- ized controlled trial, which found that variations in intravenous fluid treatment regimens for DKA do not substantially affect neurocognitive outcomes in chil- dren(15).Thus, othermechanismsshould be considered. Severe DKA is associated withalterationsincerebralperfusionand systemic increases in levels of inflam- matory mediators (27–31). In addition, studies in a juvenile rat DKA model demonstrate that a single DKA episode causes a neuroinflammatory response that involves reactive astrogliosis, acti- vation of microglia in the hippocampus, and long-term cortical neuronal loss (31). These findings suggest that inflam- matory processes triggered by DKA might cause ongoing neuroinflamma- tion that leads to long-term cognitive decline. Future studies are necessary to determine whether events similar to those documented in the rat model are responsible for cognitive deficits in chil- dren with DKA exposure. Observed def- icits involve memory and IQ, which have been associated with hippocampal integ- rity (32). Thus, examination of the links between markers of neuroinflammation, structural changes in the hippocampus, and behavioral outcomes may be partic- ularly promising. The current study has some limita-

tions. First, as in most studies of children with type 1 diabetes, we have no mea- sures of cognitive functioning prior to diagnosis, raising the issue of whether preexisting cognitive differences be- tween groups could account for some of our observations. However, associations of cognitive outcomes with DKA severity among patients with new onset of di- abetes, and the different pattern of results observed in these patients com- pared with those with a previous diag- nosis of type 1 diabetes, attenuate this concern. Newly diagnosed and previ- ously diagnosed children were largely comparable in age, sex, and SES and were recruited at the same centers using the same exclusion criteria (e.g., exclu- sion of children with known learning disabilities). Nevertheless, a clearer un- derstanding of the effects of DKA might beachievedinfuturelongitudinalstudies if neurocognitive assessments begin at diagnosis and DKA episodes occur later in the course of disease. Second, children

previously diagnosed with type 1 diabe- tes who had histories of DKA and poor glycemic control were more likely to be lost to follow-up than newly diagnosed patients and those with good glycemic control. Thus, our reported effects may underestimatecognitivealterationsinthe overall populationbecause those children lost to follow-up may be at higher risk for cognitive declines. Comparable DKA se- verity between patients who completed the study and those lost to follow-up, however, suggests that the reported ef- fects of DKA status are generalizable. Third,wereliedonmedicalrecordreviews and, in part, on parents’ or guardians’ recollections of previous DKA episodes and severe hypoglycemic events to clas- sify patients according to these variables. It is possible that some additional epi- sodesofDKA orseverehypoglycemiamay have been missed among previously di- agnosed patients. The relatively uncom- mon occurrence of severe hypoglycemia in this study and its association with other variables may have limited our ability to detect associations between severe hy- poglycemia and alterations in cognition. Fourth, among patientswithnew onset of type 1 diabetes, HbA1c levels were not routinely measured at diagnosis. Although HbA1c levels at diagnosis are not pre- dictive of future glycemic control, recent studies suggest that glycemic exposure prior to diagnosis may impact cognitive outcomes (8), and we were unable to account for this effect. Finally, our ex- clusion of children with learning disabil- ities prevents us from drawing conclusions about the effects of DKA in patients who may be even more vulnerable to alteration of cognitive development than typically developing children. Conversely, lack of inclusion of a control group of children withoutdiabetespreventsusfromdrawing conclusions about general consequences of type 1 diabetes.

Overall, our results suggest that DKA is associated with cognitive alterations in children with type 1 diabetes, in- cluding those with newly diagnosed type 1 diabetes. The neurocognitive effects of DKA in children with pre- existing type 1 diabetes should be evaluated in the context of additional variables, including repeated DKA ex- posure and glycemic control. Our re- sults emphasize the importance of prevention of DKA in children with known type 1 diabetes and of prompt

diagnosis during new onset of type 1 diabetes before the development of DKA.

Acknowledgments. The authors thank Marci Fjelstad and Amy Watson from the PECARN Data Coordinating Center for their assistance; the research coordinators in PECARN, without whom this trial would not have been possible; the clinicians in PECARN who enrolled children into this study; the members of the data and safety monitoring board (Roger Lewis [Depart- ment of Emergency Medicine, Harbor-UCLA Medical Center], Jeffrey Blumer [School of Med- icine, Case Western Reserve University], Andrew Bremer [National Institute of Child and Human Development], Thomas Cook [Department of Biostatistics and Medical Informatics, University of Wisconsin], and Beth Slomine [Department of Psychiatry, Johns Hopkins University School of Medicine]); and the members of the study out- come adjudication committee (Kathleen Meert [Department of Pediatrics, Children’s Hospital of Michigan], Jerry Zimmerman [Center for Clinical and Translational Research, Seattle Children’s Hospital], and Robert Hickey [Division of Pedi- atric Emergency Medicine, Children’s Hospital of Pittsburgh]). Funding. This study was supported by a grant from the Eunice Kennedy Shriver National In- stitute of Child Health and Human Development (U01HD062417). This project was also supported in part by the Health Resources and Services Administration, Maternal and Child Health Bu- reau, and Emergency Medical Services for Chil- dren Network Development Demonstration Program under cooperative agreement num- bers U03MC00008, U03MC00001, U03MC00003, U03MC00006, U03MC00007, U03MC22684, and U03MC22685. Thisinformationorcontentandconclusionsare

those of the authors and should not be construed as the official position or policy of, nor should any endorsements be inferred by, Health Resources and Services Administration, Health and Human Services, or the U.S. government. Duality of Interest. No potential conflicts of interest relevant to this article were reported. Author Contributions. S.G. conceived and de- signed the study, supervised training of study personnel and neurocognitive data collection, contributed to data analysis, drafted the initial manuscript, and approved the final manuscript. N.K. conceived and designed the study, obtained grant funding, supervised training of study per- sonnel, supervised patient enrollment and data abstraction, contributed to data analysis, con- tributed to the initial draft, and revised and approved the final manuscript. A.R. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. S.R.M. supervised patient enrollment and data abstraction, con- tributed to the study design, and revised and approved the final manuscript. J.E.S. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. M.J.S. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and

2774 Pediatric Diabetic Ketoacidosis and Cognition Diabetes Care Volume 43, November 2020

approved the final manuscript. A.G. supervised patient enrollment and data abstraction, contributed to study design, and revised and approved the final manuscript. K.S.Q. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. K.M.B. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. J.L.T. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. L.T. supervised patient enrollment and data abstraction, contributed to study design, and revised and approvedthefinalmanuscript.A.D.D.supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. J.K.M. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. L.E.N. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. M.Y.K. supervised patient enrollment and data abstraction, con- tributed to study design, and revised and ap- proved the final manuscript. C.S.P. trained study personnel, assisted in neurocognitive data col- lection and processing, and reviewed and ap- proved the final manuscript. C.S.O. conducted data analyses and drafted tables and figures for the final manuscript and revised and approved the final manuscript. T.C.C. supervised data analyses and drafting of tables and figures for the final manuscript and revised and approved the final manuscript. N.S.G. conceived and designed the study, obtained grant funding, supervised training of study personnel, supervised patient enrollment anddataabstraction, contributedtodata analysis, contributed to the initial draft, and revised and approved the final manuscript. C.S.O. and T.C.C. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

References 1. Sperling MA. Diabetic ketoacidosis. Pediatr Clin North Am 1984;31:591–610 2. Faich GA, Fishbein HA, Ellis SE. The epidemi- ology of diabetic acidosis: a population-based study. Am J Epidemiol 1983;117:551–558 3. GlaserN,BarnettP,McCaslinI,etal.;Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pedi- atrics. Risk factors for cerebral edema in children with diabetic ketoacidosis. N Engl J Med 2001; 344:264–269 4. Rosenbloom AL. Intracerebral crises during treatment of diabetic ketoacidosis. Diabetes Care 1990;13:22–33

5. Wootton-GorgesSL,BuonocoreMH,Kuppermann N, et al. Cerebral proton magnetic resonance spectroscopy in children with diabetic ketoaci- dosis. AJNR Am J Neuroradiol 2007;28:895–899 6. Glaser NS, Wootton-Gorges SL, Buonocore MH, et al. Frequency of sub-clinical cerebral edema in children with diabetic ketoacidosis. Pediatr Diabetes 2006;7:75–80 7. GhettiS,LeeJK,SimsCE,DemasterDM,Glaser NS. Diabetic ketoacidosis and memory dysfunc- tion in children with type 1 diabetes. J Pediatr 2010;156:109–114 8. Semenkovich K, Bischoff A, Doty T, et al. Clinical presentation and memory function in youth with type 1 diabetes. Pediatr Diabetes 2016;17:492–499 9. Ryan CM, van Duinkerken E, Rosano C. Neuro- cognitive consequences of diabetes. Am Psychol 2016;71:563–576 10. Siller AF, Lugar H, Rutlin J, et al. Severity of clinical presentation in youth with type 1 di- abetes is associated with differences in brain structure. Pediatr Diabetes 2017;18:686–695 11. Mackay MT, Molesworth C, Northam EA, Inder TE, Cameron FJ; DKA Brain Injury Study Group. Diabetic ketoacidosis and electroenceph- alographic changes in newly diagnosed pediatric patients. Pediatr Diabetes 2016;17:244–248 12. Cameron FJ, Scratch SE, Nadebaum C, et al.; DKA Brain Injury Study Group. Neurological consequences of diabetic ketoacidosis at initial presentation of type 1 diabetes in a prospective cohort study of children. Diabetes Care 2014;37: 1554–1562 13. Cato MA, Mauras N, Mazaika P, et al.; Di- abetes Research in Children Network. Longitu- dinalevaluationofcognitivefunctioninginyoung children with type 1 diabetes over 18 months. J Int Neuropsychol Soc 2016;22:293–302 14. Bradley RH, Corwyn RF. Socioeconomic sta- tus and child development. Annu Rev Psychol 2002;53:371–399 15. Kuppermann N, Ghetti S, Schunk JE, et al. Clinical trial of fluid infusion rates for pediatric diabetic ketoacidosis. N Engl J Med 2018;378: 2275–2287 16. Glaser NS, Ghetti S, Casper TC, Dean JM, Kuppermann N; Pediatric Emergency Care Ap- plied Research Network (PECARN) DKA FLUID Study Group. Pediatric diabetic ketoacidosis, fluid therapy, and cerebral injury: the design of a factorial randomized controlled trial. Pediatr Diabetes 2013;14:435–446 17. Wechsler D. Wechsler Abbreviated Scale of Intelligence. San Antonio, TX, The Psychological Corporation, 1999 18. Perry CS III, Rapinett G, Glaser NS, Ghetti S. Hydration status moderates the effects of drink- ing water on children’s cognitive performance. Appetite 2015;95:520–527 19. Raghunathan TE, Lepkowski JM, van Hoewyk J, Solenberger P. A multivariate technique for multiply imputing missing values using a sequence

of regression Mmodels. Surv Methodol 2001;27: 85–95 20. Rubin DB. Multiple Imputation for Nonre- sponse in Surveys. New York, John Wiley & Sons, 1987 21. Schafer JL. Multiple imputation: a primer. Stat Methods Med Res 1999;8:3–15 22. Von Hippel PT. Regression with Missing Y’s: an improved strategy for analyzing multiply imputed data. Sociol Methodol 2007;37:83– 117 23. Wolfsdorf JI, Glaser N, Agus M, et al. ISPAD Clinical Practice Consensus Guidelines 2018: di- abeticketoacidosis and the hyperglycemic hyper- osmolar state. Pediatr Diabetes 2018;19(Suppl. 27): 155–177 24. Skipper N, Gaulke A, Sildorf SM, Eriksen TM, Nielsen NF, Svensson J. Association of type 1 diabetes with standardized test scores of Danish schoolchildren. JAMA 2019;321: 484–492 25. Lacy ME, Gilsanz P, Eng CW, Beeri MS, Karter AJ, Whitmer RA. Recurrent diabetic ketoacidosis and cognitive function among older adults with type 1 diabetes: findings from the Study of Longevity in Diabetes. BMJ Open Diabetes Res Care 2020;8:e001173 26. Aye T, Mazaika PK, MaurasN, et al.; Diabetes Research in Children Network (DirecNet) Study Group. Impact of early diabetic ketoacidosis on the developing brain. Diabetes Care 2019;42: 443–449 27. Hoffman WH, Passmore GG, Hannon DW, et al. Increased systemic Th17 cytokines are associated with diastolic dysfunction in children and adolescents with diabetic ketoacidosis. PLoS One 2013;8:e71905 28. Omatsu T, Cepinskas G, Clarson C, et al.; Canadian Critical Care Translational Biology Group. CXCL1/CXCL8 (GROa/IL-8) in human di- abetic ketoacidosis plasma facilitates leukocyte recruitment to cerebrovascular endothelium in vitro. Am J Physiol Endocrinol Metab 2014; 306:E1077–E1084 29. Stentz FB, Umpierrez GE, Cuervo R, Kitabchi AE. Proinflammatory cytokines, markers of car- diovascular risks, oxidative stress, and lipid per- oxidation in patients with hyperglycemic crises. Diabetes 2004;53:2079–2086 30. Garro A, Chodobski A, Szmydynger-Chodobska J, et al. Diabetic ketoacidosis results in elevation of plasma levels of matrix metalloproteinase-9 in children with type 1 diabetes. Pediatr Diabetes 2017;18:95–102 31. Lo W, O’Donnell M, Tancredi D, Orgain M, Glaser N. Diabetic ketoacidosis in juvenile rats is associated with reactive gliosis and activation of microglia in the hippocampus. Pediatr Diabetes 2016;17:127–139 32. Demaster DM, Ghetti S. Developmental differences in hippocampal and cortical contri- butions to episodic retrieval. Cortex 2013;49: 1482–1493

care.diabetesjournals.org Ghetti and Associates 2775