O R T R E P O R T

Effect of Long-Term Glycemic Control onCognitive FunctionNADINA B. LINCOLN, PHDRACHAEL M. FALEIRO, BSCCATHERINE KELLY, BSC

BERNADETTE A. KIRK, RGNWILLIAM J.JEFFCOATE, FRCP

OBJECTIVE — To investigate the relationship between recurrent hypoglycemia and cogni-tive impairment in insulin-dependent diabetic patients.

RESEARCH DESIGN A N D METHODS— Seventy patients who were diagnosed as di-abetic at age 18 years or older, were under 55 years old, and had no condition likely to affectcognitive abilities were recruited from a diabetic register. Patients were interviewed to obtaininformation on the frequency of major and minor hypoglycemia. Their cognitive abilities wereassessed on tests of premorbid intelligence, current intelligence, reaction time, concentration,memory, and information processing.

RESULTS — There was a significant correlation between the apparent decline in intelligence,expressed as the discrepancy between the estimated premorbid and the actual performanceintelligence quotient, and the frequency of major hypoglycemic attacks (rs = —0.30; P < 0.01).Comparison of patients with and without recurrent hypoglycemia showed few significant dif-ferences in cognitive ability.

CONCLUSIONS — Results support previous work that suggests that major hypoglycemicattacks have a significant effect on some aspects of cognitive function, but the clinical importanceof this finding remains to be determined.

The main objective of medical man-agement of 1DDM is the achieve-ment of normal blood glucose con-

centrations, usually through the use ofintensified regimens for insulin injec-tions. However, it is well recognized thatthe use of such regimens increases the riskof severe hypoglycemia, and in the Diabe-tes Control and Complications Trial,there was a threefold increase in the inci-dence of disabling attacks in the intensivetreatment group (1). This is made worseby the fact that close control may reducethe warning symptoms of hypoglycemia(2). Not only does disabling hypoglyce-mia carry with it an immediate risk of ma-jor injury, but there is increasing evidencethat repeated episodes may cause irre-versible brain damage.

Langan et al. (3), in a study of 100

insulin-dependent subjects, found thegreatest discrepancy between the currentintelligence level and estimated premor-bid intelligence level in those insulin-dependent diabetic patients who had thegreatest number of hypoglycemic attacks.Group comparisons between individualswith a history of five or more episodes ofsevere hypoglycemia and those with nosuch episodes were significant. A subse-quent comparison of results from the 100diabetic patients in the Langan et al. studywith 100 closely matched nondiabeticcontrol group has indicated that the dia-betic patients did not differ from controlsubjects in premorbid intelligence, al-though they did differ in current intelli-gence (4). Given the potential importanceof these findings, we have sought to rep-licate them.

From the Stroke Research Unit (N.B.L., R.M.F., C.K.) and the Edwards Lane Diabetes Centre (B.A.K., W.J.J.),City Hospital, Nottingham, U.K.

Address correspondence and reprint requests to N.B. Lincoln, MD, Dept. of Psychology, University ofNottingham, University Park, Nottingham, NG7 2RD, U.K. E-mail: nbl@psyc.nott.ac.uk.

Received for publication 14 August 1995 and accepted in revised form 11 January 1996.1Q, intelligence quotient; NART, National Adult Reading Test; PASAT, Paced Auditory Serial Addition

Task; WA1S-R, Wechsler Adult Intelligence Scale-Revised.

RESEARCH DESIGN ANDMETHODS — Patients under the careof a single diabetologist (W.J.J.) identifiedfrom the Nottingham computerized reg-ister of diabetic patients were consideredfor inclusion if they had been treated withinsulin for > 1 year, were diagnosed at age18 or older, were under age 55 years, hadno history of cerebrovascular disease, hadno significant head injury or epilepsy,were taking no drugs likely to affect cog-nitive function (e.g. psychotropic drugs,j3-adrenergic blockers, and glucocorti-coids), had no history of ethanol abuse,had at least one eye with a corrected visualacuity of 6/9 or better, and were preparedto measure their own blood glucose atleast ten times each month. The study wasapproved by the Ethical Committee ofNottingham City Hospital, and all whoparticipated gave written informed con-sent.

Patients were interviewed by ei-ther a trained diabetes specialist nurse or aconsultant diabetologist. They were toldthat the object of the study was to seek acorrelation between psychological func-tioning and quality of blood glucose con-trol. Specific reference to hypoglycemiawas omitted to avoid selective reporting.During the course of the interview, thepatients completed a questionnaire inwhich they were asked to recall the fre-quency with which they had suffered bothmajor and minor episodes of hypoglyce-mia. Minor hypoglycemia was defined asbeing symptomatic but easily aborted,while major attacks required the interven-tion of another person. Patients wereasked to recall the frequency of such at-tacks in the preceding 8 weeks, the pre-ceding year, and over their lifetime. Ca-sual blood glucose was measured at thetime of the assessment.

An assistant psychologist who wasunaware of the clinical information as-sessed the patients by their performanceon the National Adult Reading Test(NART), as a measure of premorbid intel-ligence quotient (IQ) (5); the WechslerAdult Intelligence Scale-Revised (WAIS-R)short form, consisting of the Comprehen-sion, Vocabulary, Block Design, and Ob-ject Assembly subtests, to measure cur-

656 DIABETES CARE, VOLUME 19, NUMBER 6, JUNE 1996

Lincoln and Associates

Table 1—Relationship between glycemic control and intellectual function

NART predicted IQWAIS-R

Verbal IQPerformance IQFull scale IQ

NART-WAIS-R discrepancyReaction time

MeanSD

PASAT (4 s)PASAT (2 s)

Frequency of severehypoglycemic (rs)

- 0 5

- 1 1- 1 8- 1 3- 2 2 *

3350T

- 1 3- 0 4

8 weeks

10

02- 0 2- 0 1

05

0615

- 0 205

Severe

Year

0-s)

- 0 7

- 0 6- 0 5- 0 6

01

34

42*

- 0 512

Number of episodes of hypoglycemia

Lifetime

- 0 1

- 0 3- 0 5- 0 2

10

2642*

- 0 501

8 weeks

0010

11181201

-40*- 3 7 *

23*17

Mild

Year

(';)

01

04100204

- 3 1

- 2 01214

Lifetime

('00b

0215

•0311

- 3 4

- 2 01515

Coefficients are given without decimal points. For reaction time, sample size was 20 because of a recording error. ''Significant at ± 0.05; "^significant at P < 0.01.

rent IQ (6); four-choice serial reactiontime test (7); the Paced Auditory SerialAddition Task (PASAT), to assess concen-tration and speed of information process-ing (8); and the Adult Memory and Infor-mation Processing Battery Form A orF'orm B (9). The discrepancy in IQ wascalculated by subtracting the perfor-mance IQ on the WAIS-R from the pre-dicted performance IQ, with a correctionof — 7 to reduce the overestimate of theNART, as used by Langan et al. (3).

RESULTS — Of the patients who metthe inclusion criteria, 200 were asked toparticipate. Of the 87 who agreed to takepart, 17 withdrew or failed to attend forinitial testing, leaving 70 who were re-cruited. Age ranged from 24 to 54 years(38.9 ± 7.9 years [mean ± SD]) and 45were men. The majority of patients hadno episodes of severe hypoglycemia (89%in the previous 8 weeks, 76% in the pre-vious year, and 49% in a lifetime), butmild hypoglycemia was relatively com-mon. As a group, they were of averagepremorbid intelligence (NART 108.1 ±9.3, range 90-123) with a distributionslightly above that of the general popula-tion (z = 3.2;P<0.01). Current intelli-gence was consistent with the estimatedpremorbid level, with performance IQ(110.6 ± 16.9, range 75-147) beinghigher than verbal IQ (103.3 ± 15.4,range 67-140). Memory and informationprocessing mean scores were within the

average range and consistent with the in-telligence level of the group. Spearmanrank correlation coefficients were calcu-lated between each of the measures of di-abetes control and each of the psycholog-ical test scores. These are summarized inTable 1. There were few significant corre-lations. There was a significant correlationbetween the IQ discrepancy and the re-ported frequency of severe hypoglycemiabut not with the estimated number ofepisodes of hypoglycemia. To checkwhether this was a reflection of differ-ences in the ranges of the two measures,the correlation between current IQ andfrequency of severe hypoglycemia wascalculated controlling for premorbid IQ (r= 0.20; P - 0.07, which is not statisti-cally significant).

Comparison of the 11 patientswith five or more severe hypoglycemicepisodes in a lifetime with the 35 patientswho reported none, using a Mann-Whitney U test, showed significant differ-ences at the 5% level on three scores of theAdult Memory and Information Process-ing Battery.

CONCLUSIONS — The results of thepresent study support Langan et al. (3) infinding a significant correlation betweenthe reported frequency of episodes of se-vere hypoglycemia and a discrepancy be-tween a premorbid IQ estimate and cur-rent IQ. However, there are problems that

indicate these findings should be inter-preted with caution.

Of the 200 individuals invited,35% were assessed. Because informationwas not collected on those who refused toparticipate, it was not possible to checkwhether the 70 who were studied were arepresentative sample. Care should be ex-ercised in extrapolating the results ofthese selected samples to the general pop-ulation of insulin-dependent diabeticspatients, and the design of any futurestudies should encompass this issue.

The main outcome measure wasthe discrepancy between estimated pre-morbid performance IQ as assessed by theNART (6) and current performance IQ asmeasured on a short form of the WAIS-R(7). This was used to ensure that any ob-served correlations did not simply reflectthe effects of general intelligence but weredue to cognitive decline. The two testsused, NART and WAIS-R, have differentranges of possible scores and differentstandard deviations, and it is recognizedthat discrepancies indicative of cognitivedecline are more likely to occur in thosewith lower premorbid intelligence. Whenthis aspect was controlled by calculatingthe relationship between current IQ andhypoglycemia partialling out the effects ofpremorbid IQ, then the correlation wasno longer significant but still close to theuncorrected correlation.

Apart from this single significantcorrelation, there were few others out of the

DIABETES CARE, VOLUME 19, NUMBER 6, JUNE 1996 657

Glycemic control and cognitive function

total number sought between measures ofglycemic control and intellectual function.In particular, there was no deficit in thePASAT in which performance is particu-larly sensitive to cognitive decline. It ispossible that allowing for the total num-ber of correlations calculated, those fewthat were significant could have been sosimply as a result of chance. However, theconsistency of findings with those of Lan-gan et al. (3) suggest that this is unlikely.

Although our results and those ofLangan et al. (3) indicate that repeatedmajor hypoglycemic attacks may have anadverse effect, its importance should notbe overestimated. In addition, it is not al-ways clear whether the cognitive impair-ment is an effect of severe hypoglycemiaor a cause. We believe that caution mustbe exercised before it can be inferred thateither minor or infrequent major hypo-

glycemic attacks have any clinically sig-nificant effect on cognitive function andthat carefully controlled prospective stud-ies are required.

References

1. The Diabetes Control and ComplicationsResearch Group: The effect of intensivetreatment of diabetes and the develop-ment and progression of long term com-plications in insulin dependent diabetesmellitus. N Engl J Med 329:977-986,1993

2. Heller S, Ward JD: Neurologic conse-quences of hypoglycaemia and patho-genic mechanisms involved in diabeticneuropathy. Curr Opin Neurol Neurosurg6:423-428, 1993

3. Langan SJ, Deary IJ, Hepburn DA, FrierBM: Cumulative cognitive impairmentfollowing recurrent severe hypoglycaemiain adult patients within insulin treated di-

abetes mellitus. Diabetologia 34:337-344,1991

4. Deary IJ, Crawford JR, Hepburn DA, Lan-gan SJ, Blackmore LM, Frier BM: Severehypoglycaemia and intelligence in adultpatients with insulin treated diabetes. Di-abetes 42:341-344, 1993

5. Nelson HE: National Adult Reading Test.Windsor, U.K., NFER-Nelson, 1982

6. Wechsler D: Wechsler Adult IntelligenceScale (Revised). New York, PsychologicalCorporation, 1981

7. Wilkinson RT, Houghton D: Portable fourchoice reaction time test with magnetictape memory. Behav Res Methods Instrum17:441-446, 1975

8. Gronwall D: Paced Auditory Serial Addi-tion Task: a measure of recovery fromconcussion. Percept Mot Skills 44:367-373,1977

9. Coughlan AK, Hollows SE: Adult Memoryand Information Processing Battery. Leeds,U.K., Coughlan, 1985

658 DIABETES CARE, VOLUME 19, NUMBER 6, JUNE 1996