Neurological Consequences ofDiabetic Ketoacidosis at InitialPresentation of Type 1 Diabetesin a Prospective Cohort Studyof ChildrenDiabetes Care 2014;37:1554–1562 | DOI: 10.2337/dc13-1904

OBJECTIVE

To investigate the impact of new-onset diabetic ketoacidosis (DKA) during child-hood on brain morphology and function.

RESEARCH DESIGN AND METHODS

Patients aged 6–18 years with and without DKA at diagnosis were studied at fourtime points: <48 h, 5 days, 28 days, and 6 months postdiagnosis. Patients under-went magnetic resonance imaging (MRI) and spectroscopy with cognitive assess-ment at each time point. Relationships between clinical characteristics atpresentation and MRI and neurologic outcomes were examined using multiplelinear regression, repeated-measures, and ANCOVA analyses.

RESULTS

Thirty-six DKA and 59 non-DKA patients were recruited between 2004 and 2009.With DKA, cerebral white matter showed the greatest alterations with increasedtotal white matter volume and higher mean diffusivity in the frontal, temporal,and parietal white matter. Total white matter volume decreased over the first6months. For graymatter in DKA patients, total volumewas lower at baseline andincreased over 6 months. Lower levels of N-acetylaspartate were noted at base-line in the frontal graymatter and basal ganglia. Mental state scores were lower atbaseline and at 5 days. Of note, although changes in total and regional brainvolumes over the first 5 days resolved, they were associated with poorer delayedmemory recall and poorer sustained and divided attention at 6 months. Age attime of presentation and pH level were predictors of neuroimaging and functionaloutcomes.

CONCLUSIONS

DKA at type 1 diabetes diagnosis results in morphologic and functional brainchanges. These changes are associated with adverse neurocognitive outcomesin the medium term.

The incidence of childhood-onset type 1 diabetes varies from 0.1 to 57.6 per 100,000and is increasing worldwide (1). Long-term cognitive consequences of type 1 diabetesand associated fluctuations in glycemia during childhood and adolescence are well docu-mented (2). Strategies to prevent or ameliorate these adverse outcomes require an

1Department of Endocrinology and Diabetes,Royal Children’s Hospital, Murdoch Children’sResearch Institute, University of Melbourne,Melbourne, VIC, Australia2Department of Psychology, Royal Children’sHospital, Murdoch Children’s Research Institute,University of Melbourne, Melbourne, VIC,Australia3Washington University, St. Louis, MO4Queensland University of Technology, Brisbane,QLD, Australia5Department of Neurology, Royal Children’sHospital, Murdoch Children’s Research Insti-tute, University of Melbourne, Melbourne, VIC,Australia

Corresponding author: Fergus J. Cameron,fergus.cameron@rch.org.au.

Received 12 August 2013 and accepted 15February 2014.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc13-1904/-/DC1.

A slide set summarizing this article is availableonline.

© 2014 by the American Diabetes Association.See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.

Fergus J. Cameron,1 Shannon E. Scratch,1

Caroline Nadebaum,1

Elisabeth A. Northam,2 Ildiko Koves,1

Juliet Jennings,1 Kristina Finney,3

Jeffrey J. Neil,3 R. Mark Wellard,4

Mark Mackay,5 and Terrie E. Inder, 3 on

behalf of the DKA Brain Injury Study Group

1554 Diabetes Care Volume 37, June 2014

CLINCARE/ED

UCATION/N

UTR

ITION/PSY

CHOSO

CIAL

understanding of the nature and timing ofthe neurological insults associated with di-abetic dysglycemia.The most severe acute diabetes-

related central nervous system compli-cation in type 1 diabetes is cerebraledema associated with diabetic ketoaci-dosis (DKA) (2), with 10–25% of affectedchildren experiencing chronic centralnervous system morbidity (3). Althoughthe frequency of DKA at diagnosis is rel-atively high (15–70% depending on ageand geographic region), fulminate clini-cal cerebral edema in this context is rel-atively rare, with an incidence rate of0.5–0.9% (4); hence, documented braininjury is also rare. The more frequent,milder alterations in brain function andchemistry in newly diagnosed DKA havebeen assumed to be transitory with nor-malization after metabolic stabilization(5–8), but this assumption has not beenformally tested. We hypothesized thatjust as there exists a continuum of clin-ical and subclinical cerebral edema (9),there also is a continuum of brain injuryin DKA and that brain injury outside thecontext of florid cerebral edema is morecommon than previously recognized.The purpose of this study in children

with new-onset DKA was twofold. Thefirst aim was to document acute alter-ations in magnetic resonance imaging(MRI) measures of cerebral structureand cognitive function and correlatethese with clinical parameters. The sec-ond aim was to study the evolution ofthese alterations in brain structure andcognitive functioning over the first6 months.

RESEARCH DESIGN AND METHODS

We conducted a prospective, longitudi-nal cohort study at the Royal Children’sHospital (RCH), Melbourne, Australia,between June 2004 and October 2009in children aged 6–18 years with a newdiagnosis of type 1 diabetes. The diabe-tes clinic at RCH provides servicesto ;75% of children with diabetes inMelbourne, a large (population of 4.17million), sociodemographically, and eth-nically diverse region. The clinic popula-tion comprised 1,650 children andadolescents aged up to 19 years, with2–3 children being diagnosed eachweek. Study participants were recruitedsequentially as they presented to RCHat diagnosis. Participants underwentMRI, magnetic resonance spectroscopy

(MRS), and cognitive evaluation at thefollowing four time points: Baseline waswithin 48 h, day 5 between 5 and 7 days,day 28 between 26–30 days, and6months between 25 and 27weeks afterpresentation. Exclusion criteria were pre-existing neurological abnormality, evi-dence of established brain abnormalityon MRI, clinical signs of cerebral edema(headache with obtundation [impairedsensorium] or deterioration in GlasgowComa Scale score, bradycardia with hy-pertension), presence of metal-wired or-thodontic braces, and inability to undergoMRI without general anesthesia. Of 205eligible children, 95 were enrolled in thestudy (36 with DKA [defined as serum pH,7.30] and 59 without DKA). Clinical sta-tus of participants (DKA vs. non-DKA) wasclassified unambiguously according toparameters measured biochemically in aNational Association of Testing Authorities,Australia–accredited laboratory. SiemensRAPIDLab 1260 with a measurement ofuncertainty of60.02 was used to measurepH level. All participants were treatedper standard hospital protocol (http://www.rch.org.au/clinicalguide/guideline_index/Diabetes_Mellitus). The study wasapproved by the RCH Human ResearchEthics Committee, and written informedconsent was obtained from participantsand their guardians.

Measures

Clinical and Demographic Information

Clinical information collected at diagno-sis were age, sex, weight, height, bloodpressure, pulse rate, and serum biochem-ical measures (pH, CO2, HCO3, base ex-cess, glucose, osmolality, sodium, urea).Percent dehydration was calculated as[(discharge weight2 admission weight)/discharge weight] 3 100. Dischargeweight was measured between 3 and5 days after admission. Socioeconomicstatus (SES) was rated according to postalarea using the Australian Bureau of Sta-tistics Socio-Economic Indexes for Areas(10). Parents also provided informationon their child’s developmental and schoolhistory (birth complications, motor andlanguage milestones, academic perfor-mance, need for remedial classes, tutor-ing, or enrollment in special educationsetting).

The following data were collected ateach assessment: height, weight, epi-sodes of severe hypoglycemia, episodesof recurrent DKA, comorbidities, insulin

dose and regimen, and glycated hemo-globin. Severe hypoglycemia was de-fined as hypoglycemia with loss ofconsciousness and/or seizure or alteredconscious state that required therapy.DKA was defined in the same way(serum pH ,7.30) as at presentation. In-sulin dose–adjusted A1C (IDAA1C) levelswere calculated as an index of degreeof residual endogenous insulin secretion(11).

MRI

Brain imagingwas performed at four timepoints using a Siemens Trio 3T scanner.Images were acquired using the standardquadrature head coil. Transverse, T2-weighted images were acquired using aturbo spin echo sequence (repetitiontime [TR]/echo time [TE] 6,400/93 ms,echo train length 17, slice thickness 3.5mm, in-plane resolution 0.51 3 0.51mm). T1-weighted images were acquiredusing an magnetization-prepared rapidacquisition with gradient echo sequencewith contiguous slices (TR/TE/inversiontime 1,900/2.19/900 ms, flip angle 98,contiguous slices 1 mm3 isotropic spatialresolution). Diffusion tensor imaging datawere acquired with a spin echo echo-planar imaging sequence using one oftwo image parameter sets. The first wasTR/TE 6,000/97 ms, slice thickness 3 mm,in-plane resolution 1.5 3 1.5 mm, andb amplitude 1,000 3 1023 s/mm2 with20 directions. The second was TR/TE4,000/89 ms, slice thickness 4 mm, in-plane resolution 1.7 3 1.7 mm, andb amplitude 1,000 3 1023 s/mm2 with24 directions plus five b = 0 acquisitions.

Cortical and volumetric segmentationwere performed with FreeSurfer 4.4(http://surfer.nmr.mgh.harvard.edu)using the T1-weighted images. Volumeswere determined for both cortical andsubcortical regions (12). Parcellated brainvolumes were combined into larger re-gions matching the regions for diffusionanalysis. These included frontal, tempo-ral, and parietal regional volumes (Sup-plementary Table 1) and were expressedas a proportion of the total brain volume(TBV) to normalize for age-related varia-tions in brain size.

Mean diffusivity (MD) and anisotropywere measured. MD is a measure of tis-sue water displacement. In the clinicalsetting, MD values are typically used todetect brain injury, with a reduction inMD associated with acute and subacute

care.diabetesjournals.org Cameron and Associates 1555

injury such as stroke. Increases in MDvalues reflect increased tissue fluid con-tent and/or more subtle cellular injuryor loss. Diffusion anisotropy is a mea-sure of the directional variation of waterdisplacement, with values greater inhighly ordered structures such as mye-linated white matter. Anisotropy valuesmay consistently change in the oppositedirection of MD. It is likely that this in-verse relationship is caused by tissueswelling, with the greater spacing ofmembrane boundaries associated withtissue swelling leading to an increasein MD and decrease in anisotropy. As aresult of this close coupling betweenMD and anisotropy, the anisotropy val-ues are relatively uninformative. Thus,the results and discussion in this articlefocus on MD, and anisotropy is not dis-cussed in any detail.Raw data from the diffusion acquisi-

tions were phased using a Bayesian pro-cedure (13). The images were aligned tothe corresponding T2-weighted imagedata set. Diffusion encodings corruptedby participant motion were automaticallyidentified and removed. Parametric mapswere generated forMD and fractional an-isotropy. Parameter sampling was per-formed with Analyze version 9.0 software(Mayo Clinic, Rochester, MN). A data ana-lyst blinded to subject status placed re-gions of interest (ROIs) in the gray andwhite matter of the frontal, temporal,and parietal lobes.

MRS

Single-voxel, water-saturated spectrawere acquired using a standard pointresolved spectroscopy sequence. ROIsincluded the left-side frontal white mat-ter, frontal gray matter, and left-sidebasal ganglia, each previously identifiedas vulnerable (7,14). The ROIs werepositioned using T1-weighted scout im-ages. Basal ganglia ROIs were positionedover the lentiform nuclei; frontal lobewhite matter ROIs were positionedabove the anterior tip of the left lateralventricle; and frontal lobe gray matterROIs were placed in the same plane asthe white matter voxel, anterior to andat the level of the corpus callosum, withincingulate gray matter. Acquisitions com-prised80 transients recordedwith a TE/TRof 30/3,000 ms. The voxel placed in thebasal ganglia was 203 203 10mm3, andthe frontal lobe voxels were 25 3 20 315 mm3. All measurements were made

with the same Siemens 3T scannerusing a quadrature head coil. Metaboliteconcentrations are presented relative tocreatine concentration. The values forN-acetylaspartate (NAA) reported hereinclude the NAA-glutamate componentand are expressed as a ratio relativeto creatine (reported as NAA). NAA isconsidered amarker of neuronal density.Myoinositol concentration is an endoge-nous osmolyte considered to be amarker of gliosis. Tissue metabolite con-tent was determined using LCModel (15)with a basis set of 21 metabolites pre-pared using reference solutions.

Cognition

Mental State. At day 1, tests were limitedto those that could be administered atthe bedside to unwell children, whichwere the Mental State School-YearsScreening Test for the Evaluation ofMental Status (SYSTEMS) (16) and apaired associate verbal learning task(17). The SYSTEMS is a child-appropriatealternative to the adult Mini-MentalState Examination (18) on which it isbased. It was developed in a pediatrichospital to measure subtle changes inmental state and cognition in the con-text of both acute and chronic illness orbrain injury. The SYSTEMS was adminis-tered at days 1 and 5 after admission.Responses were classified using age-specific cutoff scores (16).Memory. Verbal memory was assessed ateach time point using a paired associatelearning task where children learnednovel word pairs and recalled them aftera delay (17). Alternative forms of thetest were used to minimize practice ef-fects. The number of words remem-bered after a delay was reported.Attention. Four tasks (Sky Search [fo-cused attention], Sky Search Dual Task[auditory/visual divided attention],Score Dual Task [auditory/auditory di-vided attention], and Walk/Don’t Walk[sustained attention/impulsivity]) fromthe Test of Everyday Attention for Chil-dren (19) were administered at day 5,day 28, and 6 months. Age- and sex-standardized scores were reported.

Statistical AnalysisAll data (Supplementary Tables 2–4)were analyzed using SPSS version 20(IBM Corporation) statistical software.Groups (DKA and non-DKA) were com-pared using ANOVA and ANCOVA (con-tinuous variables) and x2 (categorical

variables) analyses. Age and sex wereadded as covariates in all analyses in-volving nonage- and nonsex-adjustedvariables. Change in neurologic out-come over time was investigated usingrepeated-measures ANCOVA. Relation-ships between clinical characteristicsat presentation and neurologic out-comes were examined using multiplelinear regression analyses with age andsex entered as additional predictors. Re-gression diagnostics were performed,and no multicollinearity among indepen-dent variableswas identified. Data for thismodel are also presented in terms of per-centage of variance accounted for by theindividual predictor (sr2). Proportionatechanges in brain measures at presenta-tion were calculated as (measured valueat day 5 2 measured value at baseline)/value at day 5. Proportionate volumechanges were correlated with cognitivevariables at 6 months. All analyses weretwo-sided, and a Bonferroni correctionwas applied, resulting in an a level of0.01.

RESULTS

Participant characteristics are describedin Table 1. DKA/non-DKA participantnumbers were 36/59, 29/46, 27/46,and 25/44 at the four time points of,48 h, day 5, day 28, and 6 months,respectively. There were no differencesbetween the DKA and non-DKA groupsin age, duration of symptoms, SES, orpreexisting premorbid need for schoolassistance/remediation. After diagnosis,the groups did not differ in frequency ofsevere hypoglycemia, recurrent DKA,thyroid or celiac disease, insulin regi-mens, or glycated hemoglobin (raw orIDAA1C).

MRI and MRS Measures

Total Cerebral Volumes

Total supratentorial brain volumes didnot differ with DKA at any time point(Fig. 1A) or change over time in eithergroup.

White Matter

Relative total white matter volume wasgreater in the DKA group at baselinecompared with the non-DKA group (F =7.43, P = 0.008) (Fig. 1D) and decreasedover 6months (F = 25.72, P, 0.001). Forregional analysis, relative frontal whitematter volume was greater in the DKAgroup at baseline (F = 10.84, P = 0.002,Fig. 2A) and decreased over 6 months

1556 Impact of Newly Diagnosed Diabetes Upon Brain Diabetes Care Volume 37, June 2014

(F = 16.56, P = 0.001). A decrease in rel-ative parietal white matter volume alsooccurred over 6 months in the DKA grouponly (F = 9.19, P = 0.008). For diffusion

measures, within the DKA group, whitematter MD at baseline was higher in thefrontal lobe (F = 18.78, P , 0.001) (Fig.2B), temporal lobe (F = 18.12, P, 0.001),

and parietal lobe (F = 13.42, P , 0.001).Reductions in white matter MD occurredover 6 months with DKA in the frontal (F =11.08, P = 0.006) (Fig. 2B), temporal

Table 1—Patient characteristics

Non-DKA group (n = 59) DKA group (n = 36) P value

Age (years) 11.52 (6.17–16.16) 11.52 (6.05–17.87) 0.99

Male 54.2 47.2 0.5

Duration of polyuria/polydypsia (days) 23.41 (2–168) 19.46 (1–56) 0.4

SES (percentile) 61.51 (5–99) 60.58 (4–100) 0.99

Requiring school assistance 15.8 16.7 .0.99

At presentationLowest pH 7.37 (7.30–7.44) 7.12 (6.72–7.29) ,0.001Highest blood glucose (mmol/L) 25.77 (13.00–45.10) 31.00 (16.00–81.00) 0.03Highest corrected serum sodium (mmol/L) 139.69 (134–146) 147.99 (124–198) ,0.001Lowest corrected serum sodium (mmol/L) 136.00 (133–141) 142.56 (128–169) 0.075Highest urea (mmol/L) 5.15 (2–8) 6.21 (3–13) 0.006Serum osmolality (mmol/L) 298.39 (285–318) 312.80 (264–394) ,0.001Percent dehydration* NA 7.84 (0–14) NA

At 6-month follow-upPatientswith one ormore episodes of severe hypoglycemia 1.7 2.8 0.3Patients with one or more episodes of subsequent DKA 0 0 d

Percentage of patients onTwo insulin injections/day 78 (n = 46) 74 (n = 26) 0.7Four insulin injections/day 22 (n = 13) 26 (n = 9)

Total daily insulin dose per body weight (units/kg/day) 0.7 (0–1.4) 0.8 (0.3–1.7) 0.016HbA1c (%) 7.52 (5–12) 7.40 (5–9) 0.617HbA1c (mmol/mol) 59 (31–108) 57 (31–75)IDAA1C 10.2 (5.6–15.7) 10.6 (6.9–14.8) 0.205

Data are mean (range) or % unless otherwise indicated. NA, not applicable. *Calculated as (weight at discharge2weight at presentation)/weight atdischarge.

Figure 1—Group differences (DKA vs. non-DKA) for TBVs. Data are presented as the estimated mean and 95% CI at each time point. A: Totalsupratentorial brain volumes. B: Total supratentorial white matter volume. C: Total supratentorial gray matter volume; *P = 0.014. D: Total corticalwhite matter volume/TBV; *P = 0.008. E: Total cortical gray matter volume/TBV; *P = 0.008.

care.diabetesjournals.org Cameron and Associates 1557

(F = 42.20, P , 0.001), and parietal (F =16.63, P = 0.002) lobes. Fractional anisot-ropy values were highly correlated withMD values in all areas/analyses, changingin the opposite direction of MD. Frontalwhite matter NAA levels were lower withDKA at 6 months (F = 8.51, P = 0.005) (Fig.2C). There was a trend for an increase infrontal white matter NAA levels over6 months in the non-DKA group (F =6.186, P = 0.02) but no change in theDKA group.

Gray Matter

Relative total cortical gray matter vol-ume was lower with DKA on day 1 (F =7.64, P = 0.008) (Fig. 1E) and increasedover 6 months (F = 10.18, P = 0.005).Regionally, the relative temporal corti-cal gray matter volume was lower withDKA on day 1 (F = 8.66, P = 0.005), with atrend to increase over 6 months (F =6.22, P = 0.02). For diffusion, parietallobe gray matter MD was higher withDKA at baseline (F = 10.2, P = 0.002).Gray matter MD in the frontal lobes de-creased over 6 months with DKA (F =50.23, P , 0.001) (Fig. 2E) and trendedlower at 6 months compared with thenon-DKA group (F = 5.80, P = 0.02). Forspectroscopy, frontal gray matter NAAlevels were lower at baseline in theDKA group (F = 8.73, P = 0.005) (Fig. 2F),

whereas myoinositol levels decreasedover 6 months (F = 14.827, P = 0.002).

Hippocampus and Basal Ganglia

Relative hippocampal and basal gangliavolumes did not differ with DKA at anytime point. There was no change in ei-ther volume over 6 months. In the left-side basal ganglia, NAA levels trendedlower at baseline (F = 5.24, P = 0.03)andday 5 (F = 7.29 andP = 0.01)withDKA.

CognitionThe mental state score was lower in theDKA group at baseline (F = 10.14, P =0.002) (Fig. 3A), with both groups show-ing similar improvement by day 5 (DKA:F = 28.77, P, 0.001; non-DKA: F = 21.42,P , 0.001). Memory scores were lowerin the DKA group at baseline (F = 14.58,P , 0.001) but improved by 6 months(F = 8.22, P = 0.009) (Fig. 3B). No differ-ences in memory and divided or sus-tained attention were noted with DKAat 6 months. Focused attention im-proved between day 5 and 6 months inthe non-DKA group (F = 13.73, P = 0.001)(Fig. 3C) but not in the DKA group.

Clinical Parameters on Admission andMRI and Cognitive OutcomesLowest pH level and age were the stron-gest predictors of MRI/MRS and cogni-tive measures. Lower baseline pH level

was positively related with change ingray matter volume by day 5 (sr2 =7.51%, P = 0.006); negatively associatedwith baseline white matter diffusivity inthe frontal (sr2 = 41.3%, P, 0.001), tem-poral (sr2 = 33.03%, P, 0.001), and pari-etal (sr2 = 24.07%, P, 0.001) regions; butpositively associated with 6-month fron-tal gray matter diffusivity (sr2 = 11.9%, P =0.008). There were positive associationsbetween baseline pH level with baselinefrontal gray matter NAA levels (sr2 =12.8%, P = 0.005) and baseline memory(sr2 = 17.47%, P , 0.001) such that thelower the pH level, the lower the NAAlevel and poorer the memory score.

Age at presentation was positively re-lated to diffusion measures at baseline infrontal and parietal gray matterMD (sr2 =22.84%, P, 0.001, and sr2 = 21.87%, P,0.001, respectively) and negatively asso-ciated with baseline frontal and temporalwhite matter MD (sr2 = 10.36%, P ,0.001, and sr2 = 11.46%,P = 0.001, respec-tively). Age at presentation was positivelyassociated with 6-month frontal (sr2 =16.0%, P = 0.002) and parietal (sr2 =17.1%, P = 0.003) gray matter MD.

MRI Measures in the First Week and6-Month Cognitive OutcomesThe reduction between baseline and day5 in total white matter volume (an

Figure 2—Group differences (DKA vs. non-DKA) for frontal lobe volumes, diffusivity, and spectroscopy. Data are presented as the estimated meanand 95% CI at each time point. A: Frontal lobe white matter relative volume; *P = 0.002. B: Frontal lobe white matter diffusivity; *P , 0.001.C: Frontal lobe white matter NAA spectroscopy; *P = 0.005. D: Frontal lobe gray matter relative volume. E: Frontal lobe gray matter diffusivity; *P =0.013. F: Frontal lobe gray matter NAA spectroscopy; *P = 0.005.

1558 Impact of Newly Diagnosed Diabetes Upon Brain Diabetes Care Volume 37, June 2014

indirect measure of swelling) was nega-tively associated with performance ondual-modality divided (r = 20.46, P =0.002) and sustained (r = 20.40, P =0.007) attention at 6 months. Frontalwhitematter volume reductionwas neg-atively associated with dual-modalitydivided attention (r = 20.43, P = 0.004)at 6 months, temporal white matter vol-ume reduction was negatively associ-ated with memory (r = 20.49, P =0.0009) at 6 months, and parietal whitematter volume reduction was negativelyassociated with sustained attention (r =20.47, P = 0.002) at 6 months. Parietalgray matter volume increase was nega-tively associated with sustained atten-tion (r = 20.45, P = 0.002).

CONCLUSIONS

This study highlights the common na-ture of transient focal cerebral edemaand associated impaired mental stateat presentation with new-onset type 1diabetes in children. We demonstratethat alterations occur most markedly incerebral white matter, particularly inthe frontal lobes, and are most promi-nent in the youngest children with themost dramatic acidemia. Given thesomewhat arbitrary nature of a serum

pH ,7.30 as the cut point to defineDKA, we undertook regression analysesusing pH as a continuous variable. Wefound that although changes in MRImeasures over the first week after diag-nosis resolved, these early brain changeswere associated with persisting alter-ations in attention andmemory 6monthslater. Children with DKA did not differ inage, sex, SES, premorbid need for schoolassistance/remediation, or postdiagnosisclinical trajectory. Earlier diagnosis oftype 1 diabetes in children may avoidthe complication of DKA and the neuro-logical consequences documented in thisstudy and is worthy of a major publichealth initiative.

This study used several MRI techni-ques to define cerebral structure andbiochemistry. All regional brain volumedata were normalized to account for dif-ferences in absolute volume related tovarying participant age. Thus, the mea-sures from baseline showed a relativeincrease in white matter volume and de-crease in gray matter volume.With DKA,white matter volume expansion wasgreatest within the frontal whitematter,but a more widespread reduction in pa-rietal and frontal white matter volumewas seen over the 6 months following

diagnosis. Diffusion values were notnormalized because they are compara-tively stable throughout childhood (20).Diffusion values in the DKA group werehigh on day 1 throughout all white mat-ter regions evaluated but returned tonon-DKA levels by day 5. We postulatethat increases in both volume and diffu-sivity within the white matter on initialpresentation were due to increased tis-sue water, and this is supported bystrong correlations between the reduc-tion in volume and decrease in diffusiv-ity. It is notable that diffusivity appearedto be more sensitive than volume be-cause an increase in white matter vol-ume was detected only for frontal whitematter. This finding also suggests agreater vulnerability for the frontalwhite matter to DKA, which may rep-resent a maturational vulnerability be-cause the frontal white matter remainsimmature at this period of cerebraldevelopment.

We propose two explanations, whichare not necessarily mutually exclusive,for the increased volume and altereddiffusion that is acutely prominent inthe white matter: 1) the osmotic effectscaused by relatively rapid restoration ofblood glucose (and osmolarity) in the

Figure 3—Group differences (DKA vs. non-DKA) for cognitive outcomes. Data are presented as the estimated mean and 95% CI at each time point.A: Mental state (SYSTEMS) score; *P = 0.002. B: Delayed memory raw score; *P = 0.002. C: Focused attention (Sky Search) scaled score. D: Dual-modality divided attention (Sky Search DT) scaled score. E: Single-modality divided attention (Score DT) scaled score. F: Sustained attention/impulsivity (Walk/Don’t Walk) scaled score. DT, dual task.

care.diabetesjournals.org Cameron and Associates 1559

face of slower loss of idiogenic osmolesin axons and myelin, leading to cellswelling, and/or 2) the breakdown ofthe blood-brain barrier, leading to ex-travasation of fluid into the whitematter interstitium with extracellular(vasogenic) edema. These disturbancesin brain water balance can be viewed asbeing analogous to those associatedwith extrapontine myelinolysis, whichis typically found in conjunction withfluid/electrolyte disturbances and hasbeen associated with diabetes (21,22).The large U.K. case-control series ofDKA-associated cerebral edema did notshow an association with any changes inblood glucose after therapy (6). Further-more, in a series of youth aged 6–17years with new-onset and establisheddiabetes and DKA, two magnetic reso-nance (MR) scans performed within12–24 h and 36–72 h after insulin ther-apy showed that themajority of patientshad a 2–5% increase in whole-braindiffusivity between the two time points(23). There was an associated decreasein mean transit time and no change inrelative cerebral blood volume, suggest-ing a vasogenic process. The presentfindings are also consistent with otherstudies that suggested brain water shiftsassociated with DKA (9), including in-creased diffusion in the frontal whitematter in the first hours of management(14). Further evidence of white matterpathology was demonstrated in twoatypical comatose pediatric DKA patientswithout cerebral edema in whom subcor-tical white matter microhemorrhage andinflammation were noted (24).Additional insight is provided from

the MRS. Acutely in the DKA cohort,frontal gray matter demonstrated a re-duction in NAA levels, which recovered.This may reflect neuronal dysfunction,emphasizing neuronal impact fromDKA. Over the first 6 months after diag-nosis, the greatest alterations on spec-troscopy were reductions in frontalwhite matter NAA levels. Although atypical maturational increase in NAA oc-curs in children without ketoacidosis(25), this was not seen in the DKA group.Reductions in frontal white matter andbasal ganglia NAA levels have been as-sociated previously with impaired cog-nition (26) and DKA (27–29). Lactate waspresent in ;10% of both DKA and non-DKA participants in any voxel. This lowincidence of elevated lactate level may

relate to the delay in imaging after initi-ation of therapy.

Two cross-sectional retrospectivestudies highlighted a negative impactof ketoacidosis on cognition in school-aged children with existing diabetes(30,31). For the current study, the strengthof examining patients with new-onset di-abetes has allowed the window of dysgly-cemia to be simplified to a single period ofhyperglycemia with or without one epi-sode of ketoacidosis. Of note, the presentfindings replicate a similar unidimensionalmodel of cognitive performance in rats af-ter one episode of DKA (32). Anothercross-sectional MR study investigatingthe effects of severe hyperglycemia asso-ciated with ketosis with or withoutacidemia in youth aged 9–22 years with amean duration of 9.5 years showed that ahistory of one or more episodes of severehyperglycemia was associated with in-creased diffusivity in the superior parietallobule and hippocampus (33).

In relation to clinical risk factors, thedegree of acidosis and younger age ap-peared to be the greatest risk factorsfor alterations in cerebral structure.A greater degree of cognitive impair-ment in younger children with type 1diabetes has been reported previously(2), but these studies have included chil-dren exposed to various aspects of dia-betes dysglycemia, including hypoglycemia.Specific studies of DKA in children havebeen limited thus far to cerebral edema.In these studies, the risk of cerebral edemawas associated only with elevated serumurea level and low PACO2 (34). In childrenand adults with DKA without cerebraledema, the degree of acidosis was the de-terminant of impaired conscious state(6,35).

Finally, cerebral volume changes inthe frontal, temporal, and parietal re-gions in the first week after diagnosiswere associated with lower attentionand memory scores 6 months later,suggesting that functional informationprocessing difficulties persist after res-olution of tissue water increases in ce-rebral white matter. These findings havenot been reported to date but are con-sistent with the growing concern overacademic performance in children withdiabetes (2).

This study has several limitations tonote. First, the lack of premorbid data onduration of preadmission hyperglycemia/acidemia, MR brain volumes, diffusivity,

spectroscopy, and cognition is unavoid-able. Currently, no screening programspractically identify children and adoles-cents destined to develop type 1 diabetesin low-risk populations. If there were suchscreening programs, it would be unethicalto allow DKA to develop. However, giventhe fact that the DKA and non-DKA groupsdid not differ on any demographic varia-bles or previous need for school assis-tance/remediation and that variations inbrain morphology aligned temporally andstatistically to the episode of acidosis, wesurmise that the likelihood of significantpremorbid group differences in brainstructure and function is low. The secondlimitation relates to the period of hyper-glycemia and DKA before admission. Noclinical tool can retrospectively define du-ration of hyperglycemia alone; however, abaseline HbA1c may have indicated ameasure of degree and duration of pre-admission hyperglycemia. Given the sever-ity of the clinical phenotype of DKA, anyvariation in duration would be a matter ofhours rather than days and is to some ex-tent reflected by the degree of acidosis,but exact duration remains unknown. Athird limitation relates to the short follow-up period of 6 months. However, the lon-ger the follow-up period, the greater thelikelihood of postdiagnosis glycemicevents contributing to brain injury. Wehad to balance the need for allowingfor recovery with avoiding additional clin-ical confounders; thus, we chose 6monthsas the follow-up period to provide the bestpicture of the neurological sequelae of theinitial DKA insult. A fourth limitation re-lates to the fact that only three-quartersof the participants completed all data col-lection, which may introduce some poten-tial bias to the later evaluations. Finally,although the purpose of this study wasto compare newly diagnosed diabetic pa-tientswith andwithoutDKA, a nondiabeticcontrol group may have been helpful forcomparative purposes because normativedata from some of theMRmeasures (e.g.,MD) are scant (33,36). To overcome a lackof MR normative data, analyses were ad-justed for age and/or sex. The volumetricanalyses were also expressed as a pro-portion of TBV to normalize for age-related variations in brain size. Diffusionparameters appear to be relatively con-stant over the age range studied (33), sowe believed it reasonable to compare par-ticipants with one another. Nevertheless,we also looked at diffusion parameters

1560 Impact of Newly Diagnosed Diabetes Upon Brain Diabetes Care Volume 37, June 2014

as a fractional change from baseline in afashion analogous to that used for thevolume data. This had no effect on thefinal results, so we have simply reportednonnormalizeddiffusionparameter values.

Clinical ImplicationsThis study demonstrates common andwidespread alterations in brain struc-ture and chemistry in new-onset type 1diabetes with ketoacidosis in a pediatricand adolescent cohort from a socioeco-nomically and ethnically diverse popula-tion. There are several implications thatarise from these findings. First is the im-perative to avoid ketoacidosis in chil-dren through improved public andprofessional awareness. Reductions inthe rates of ketoacidosis at diagnosisare possible, as has been demonstratedwith public health campaigns in Italy andAustralia (37,38). Second, any neuropro-tective strategy developed in the futuremust be prioritized at initial diagnosis.Finally, we should focus greater atten-tion on neuropsychological evaluationof children with diabetes by both per-forming a brief mental state examina-tion in all newly diagnosed patientsand providing cognitive follow-up. Itwould appear sensible for clinicians todefer educational activities in patientswith a suboptimal mental state for atleast 1–2 weeks. Brain injury should nolonger be considered a rare complica-tion of DKA. This study has shown thatit is both frequent and persistent.

Acknowledgments. The authors thank MichaelKean andMichael Ditchfield fromtheDepartmentof Medical Imaging at the Royal Children’s Hospi-tal,Melbourne, for assistancewith the acquisitionof the MR images.Funding. This study was funded by JDRF andthe Australian National Health and MedicalResearch Council (project grant 334354). T.E.I.is supported by the Doris Duke CharitableFoundation, the Eunice Kennedy Shriver Na-tional Institute of Child Health and HumanDevelopment (grant P30-HD-062171), and theWashington University Intellectual and Devel-opmental Disability Research Center.Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.Author Contributions. F.J.C., E.A.N., andM.M. contributed to the data collection andanalysis and writing of the manuscript. S.E.S.and C.N. contributed to the data collation, statisticalanalysis, and editing of the manuscript. J.J. and I.K.contributed to patient recruitment and clinicalmanagement of the participants. K.F., J.J.N., andT.E.I. contributed to the MRI analyses and inter-pretation and editing of the manuscript. R.M.W.contributed to theMRSanalyses and interpretation

and editing of the manuscript. F.J.C. is the guaran-tor of this work and, as such, had full access to allthe data in the study and takes responsibility forthe integrity of the data and the accuracy of thedata analysis.Prior Presentation. Parts of this study werepresented in abstract form at IPSAD 2011, MiamiBeach, FL, 19–22 October 2011 and IPSAD 2012,Istanbul, Turkey, 10–13 October 2012.

References1. Craig ME, Hattersley A, Donaghue KC. Defi-nition, epidemiology and classification of dia-betes in children and adolescents. PediatrDiabetes 2009;10(Suppl. 12):3–122. McCrimmon RJ, Ryan CM, Frier BM. Diabetesand cognitive dysfunction. Lancet 2012;379:2291–22993. Marcin JP, Glaser N, Barnett P, et al.; Amer-ican Academy of Pediatrics; The PediatricEmergency Medicine Collaborative ResearchCommittee. Factors associated with adverseoutcomes in children with diabetic ketoacidosis-related cerebral edema. J Pediatr 2002;141:793–7974. Wolfsdorf J, Craig ME, Daneman D, et al. Di-abetic ketoacidosis in children and adolescentswith diabetes. Pediatr Diabetes 2009;10(Suppl.12):118–1335. Edge JA, Jakes RW, Roy Y, et al. The UK case-control study of cerebral oedema complicatingdiabetic ketoacidosis in children. Diabetologia2006;49:2002–20096. Edge JA, Roy Y, Bergomi A, et al. Consciouslevel in children with diabetic ketoacidosis isrelated to severity of acidosis and not to bloodglucose concentration. Pediatr Diabetes 2006;7:11–157. Cameron FJ, Kean MJ, Wellard RM, WertherGA, Neil JJ, Inder TE. Insights into the acute ce-rebral metabolic changes associated with child-hood diabetes. Diabet Med 2005;22:648–6538. Hoffman WH, Casanova MF, Bauza JA,Passmore GG, Sekul EA. Computer analysis ofmagnetic resonance imaging of the brain inchildren and adolescents after treatment of di-abetic ketoacidosis. J Diabetes Complications1999;13:176–1819. Glaser NS, Wootton-Gorges SL, BuonocoreMH, et al. Frequency of sub-clinical cerebraledema in children with diabetic ketoacidosis.Pediatr Diabetes 2006;7:75–8010. Socio-Economic Indexes for Areas (SEIFA).Canberra, Australia, Australian Bureau of Statis-tics, 200611. Mortensen HB, Hougaard P, Swift P, et al.;Hvidoere Study Group on Childhood Diabetes.New definition for the partial remission periodin children and adolescents with type 1 diabe-tes. Diabetes Care 2009;32:1384–139012. Desikan RS, Segonne F, Fischl B, et al. Anautomated labeling system for subdividing thehuman cerebral cortex on MRI scans into gyralbased regions of interest. Neuroimage 2006;31:968–98013. Larry Bretthorst G. Automatic phasing ofMR images. Part II: voxel-wise phase estimation.J Magn Reson 2008;191:193–20114. Glaser NS, Marcin JP, Wootton-Gorges SL,et al. Correlation of clinical and biochemicalfindings with diabetic ketoacidosis-related cere-bral edema in children using magnetic resonance

diffusion-weighted imaging. J Pediatr 2008;153:541–54615. Provencher SW. Estimation of metaboliteconcentrations from localized in vivo protonNMR spectra. Magn Reson Med 1993;30:672–67916. Ouvrier R, Hendy J, Bornholt LJ, Black FH.The School-Years Screening Test for the Evalua-tion of Mental Status. Sydney, Australia, TheChildren’s Hospital at Westmead and The Uni-versity of Sydney, 200117. Nott P. The paired-associate learning sub-test of the Wechsler Memory Scale: Six newparallel forms. Brit J Clin Soc Psychol 1975;14:199–20118. FolsteinMF, Folstein SE,McHugh PR. “Mini-mental state”. A practical method for gradingthe cognitive state of patients for the clinician.J Psychiatr Res 1975;12:189–19819. Manly T, Anderson V, Nimmo-Smith I,Turner A, Watson P, Robertson IH. The differen-tial assessment of children’s attention: the Testof Everyday Attention for Children (TEA-Ch),normative sample and ADHD performance.J Child Psychol Psychiatry 2001;42:1065–108120. Mukherjee P, Miller JH, Shimony JS, et al.Diffusion-tensor MR imaging of gray and whitematter development during normal humanbrain maturation. AJNR Am J Neuroradiol2002;23:1445–145621. HoffmanWH, Siedlak SL, Wang Y, CastellaniRJ, Smith MA. Oxidative damage is present inthe fatal brain edema of diabetic ketoacidosis.Brain Res 2011;1369:194–20222. Ranger AM, Chaudhary N, Avery M, Fraser D.Central pontine and extrapontine myelinolysis inchildren: a review of 76 patients. J Child Neurol2012;27:1027–103723. Vavilala MS, Marro KI, Richards TL, et al.Change in mean transit time, apparent diffusioncoefficient, and cerebral blood volume duringpediatric diabetic ketoacidosis treatment. Pe-diatr Crit Care Med 2011;12:e344–e34924. Mahmud FH, Ramsay DA, Levin SD, SinghRN, Kotylak T, Fraser DD. Coma with diffusewhite matter hemorrhages in juvenile diabeticketoacidosis. Pediatrics 2007;120:e1540–e154625. Bluml S, Wisnowski JL, Nelson MD Jr, et al.Metabolic maturation of the human brain frombirth through adolescence: insights from in vivomagnetic resonance spectroscopy. Cereb Cortex2013;23:2944–295526. Filippi CG, Ulug AM, Deck MD, ZimmermanRD, Heier LA. Developmental delay in children:assessment with protonMR spectroscopy. AJNRAm J Neuroradiol 2002;23:882–88827. Glaser N, Yuen N, Anderson SE, Tancredi DJ,O’Donnell ME. Cerebral metabolic alterations inrats with diabetic ketoacidosis: effects of treat-ment with insulin and intravenous fluids andeffects of bumetanide. Diabetes 2010;59:702–70928. Wootton-Gorges SL, Buonocore MH,Kuppermann N, et al. Cerebral proton magneticresonance spectroscopy in children with dia-betic ketoacidosis. AJNR Am J Neuroradiol2007;28:895–89929. Wootton-Gorges SL, Buonocore MH,Caltagirone RA, Kuppermann N, Glaser NS. Pro-gressive decrease in N-acetylaspartate/creatineratio in a teenager with type 1 diabetes and re-peated episodes of ketoacidosis without clinically

care.diabetesjournals.org Cameron and Associates 1561

apparent cerebral edema: evidence for perma-nent brain injury. AJNR Am J Neuroradiol 2010;31:780–78130. Shehata G, Eltayeb A. Cognitive function andevent-related potentials in children with type 1diabetesmellitus. J ChildNeurol 2010;25:469–47431. Ghetti S, Lee JK, Sims CE, Demaster DM,Glaser NS. Diabetic ketoacidosis and memorydysfunction in children with type 1 diabetes.J Pediatr 2010;156:109–11432. Glaser N, Anderson S, Leong W, Tancredi D,O’Donnell M. Cognitive dysfunction associatedwith diabetic ketoacidosis in rats. Neurosci Lett2012;510:110–114

33. Antenor-Dorsey JA, Meyer E, Rutlin J, et al.White matter microstructural integrity in youthwith type 1 diabetes. Diabetes 2013;62:581–58934. Glaser N, Barnett P, McCaslin I, et al.; Pedi-atric Emergency Medicine Collaborative Re-search Committee of the American Academyof Pediatrics. Risk factors for cerebral edemain children with diabetic ketoacidosis. N Engl JMed 2001;344:264–26935. Nyenwe EA, Razavi LN, Kitabchi AE, KhanAN, Wan JY. Acidosis: the prime determinantof depressed sensorium in diabetic ketoacido-sis. Diabetes Care 2010;33:1837–1839

36. Aye T, Barnea-Goraly N, Ambler C, et al. Whitematter structural differences in young childrenwithtype 1 diabetes: a diffusion tensor imaging study.Diabetes Care 2012;35:2167–217337. Vanelli M, Scarabello C, Fainardi V. Avail-able tools for primary ketoacidosis preventionat diabetes diagnosis in children and adoles-cents. “The Parma campaign”. Acta Biomed2008;79:73–7838. King BR, Howard NJ, Verge CF, et al. A di-abetes awareness campaign prevents diabeticketoacidosis in children at their initial presenta-tion with type 1 diabetes. Pediatr Diabetes2012;13:647–651

1562 Impact of Newly Diagnosed Diabetes Upon Brain Diabetes Care Volume 37, June 2014