Article Text

Download PDFPDF

Moderate hypoglycaemia obliterates working memory in humans with and without insulin treated diabetes
  1. I J Deary1,
  2. A J Sommerfield2,
  3. V McAulay2,
  4. B M Frier2
  1. 1Department of Psychology, University of Edinburgh, 7 George Square, Edinburgh EH8 9JZ, Scotland, UK
  2. 2Department of Diabetes, Royal Infirmary of Edinburgh
  1. Correspondence to:
    Professor Ian J Deary;

Statistics from

Moderate hypoglycaemia is common in people with insulin treated diabetes, and some mental functions deteriorate during this state. Working memory is a crucial cognitive process, necessary for many everyday tasks, but its integrity during hypoglycaemia is not known. We examined the effect of moderate hypoglycaemia on working memory in 32 young adults (16 with type 1 (insulin dependent) diabetes). Mean performance on working memory was reduced almost to levels of chance. During hypoglycaemia, therefore, this important mental ability is almost obliterated though the subjects remain fully conscious.

Working memory is the mental capacity simultaneously to store and manipulate information.1 It supports human capabilities in composing spoken language and in negotiating the environment. Individual differences in working memory correlate very highly with measures of reasoning and general intelligence.2 Glucose is the brain’s principal fuel, and mental function deteriorates when arterial plasma glucose falls below 3.0 mmol/l. Mental functions that are complex and performed under time pressure are particularly affected, the usual finding being a modest reduction in performance.3 To date, working memory has not been studied during hypoglycaemia. Here we examine the effect of controlled, moderate hypoglycaemia on performance of a challenging working memory task.

Working memory performance was studied in 32 young adults. Sixteen (nine men, seven women) were healthy, non-diabetic subjects aged between 26 and 34 years (mean (SD), 29.6 (1.7) years); a further 16 (nine men, seven women) were otherwise healthy individuals with type 1 diabetes, aged between 20 and 38 years (28.5 (5.4) years). Those with diabetes had a mean HbA1c of 8.3 (0.5)%. The mean number of correct items in the national adult reading test was 36.9 (6.1) for the non-diabetic group and 34.6 (5.6) for the diabetic group. Thus both groups comprised people of above average cognitive ability, with an overall mean intelligence quotient of about 113. They were examined under conditions of euglycaemia (4.5 mmol/l) and hypoglycaemia (2.5 mmol/l). The hyperinsulinaemic glucose clamp technique was used to achieve precise control of arterialised plasma glucose.

Permission to conduct the study was given by Lothian Health medical research ethics committee. Written, informed consent was obtained from all subjects.

The experiment had a repeated measures, counterbalanced design with glycaemic condition as a within subjects factor (repeated measure); between subjects factors were whether or not subjects had diabetes and the order of conditions (euglycaemia-hypoglycaemia and hypoglycaemia-euglycaemia). Data, presented as mean (SD), were analysed using general linear modelling (repeated measures) on SPSS version 10.1. There were no significant order effects.

The “four term order” working memory test is from Kyllonen’s well validated cognitive battery of tests which is based on a model of human information processing.4 The task involves storing rules and stimuli in working memory and mentally manipulating the order of the stimuli so that they fit the rules. For each item of the task, subjects listen to three rules read aloud by the experimenter and then choose which of eight response options correctly accords with the rules. For example, the rules in one item are: “The animals come after the furniture; the horse comes before the dog; the desk does not come before the rug.” Eight response alternatives are then displayed on a card. The subject would score correctly by choosing the sequence “rug-desk-horse-dog.” Twenty four different items were used in each study condition. Subjects also performed the digit symbol test from the Wechsler adult intelligence scales–revised, and the trail making B test from the Halstead Reitan neuropsychological test battery. Performance on these latter two tasks is known to deteriorate during hypoglycaemia.

As expected, mean scores on the digit symbol and the trail making B tasks were significantly poorer during moderate hypoglycaemia (table 1). The hypoglycaemia induced deterioration involved a fall in the mean scores overall, with considerable overlap in scores between the two conditions. The variance of these two tests did not differ significantly between the euglycaemia and hypoglycaemia conditions (Pitman’s test for variance difference in non-independent measures, NS). Scores on the four term order task declined from a mean of 12.4 (4.3) during euglycaemia to 3.7 (2.3) during hypoglycaemia (p < 0.0001). Two further aspects of this result are notable. First, a score of 3.0 represents chance responding on the four term order test, which is close to the mean score during hypoglycaemia. Second, the variance in the four term order test scores fell significantly from hypoglycaemia to euglycaemia (p < 0.001, using Pitman’s test; table 1). Thus a range of individual differences in working memory scores during euglycaemia—from chance to near perfect performance—was reduced to near uniform chance responding during hypoglycaemia. There was no significant association in the diabetes group between HbA1c and the decline in working memory induced by hypoglycaemia. There was no difference in the effect of hypoglycaemia between the diabetic and non-diabetic groups on any of the three mental tests (that is, there was no group by condition interaction for any test).

Unlike other mental functions, working memory—as indexed by the four term order task—was almost obliterated during moderate hypoglycaemia in young adult humans, with and without type 1 diabetes. Working memory is a pillar of human mental performance, its variance in humans being almost indistinguishable from differences in reasoning and general intelligence.2 It is likely that, whereas other working memory tasks might also deteriorate during hypoglycaemia, those that were less complex than the four term order task would not be reduced to chance levels of performance. Animal studies suggest that performing difficult working memory tasks makes especially high demands on brain extracellular glucose, specifically in the hippocampus.5 However, the comparability of human and animal studies of working memory is limited because of the different tasks used, and the different meaning that the concept of working memory may have when applied to human versus other species. The hippocampus and other medial temporal lobe structures and the frontal lobes are especially sensitive to the effects of neuroglycopenia.6 A possible basis for the special vulnerability of working memory to the effects of hypoglycaemia is therefore the interaction of local brain energy demands and the regional density of the cerebral glucose transporters GLUT1 and GLUT3. The degree of hypoglycaemia induced here, which often occurs in people with type 1 (insulin dependent) diabetes, does not affect conscious level. Though many mental tasks show some deterioration in this state, it is remarkable to find a core mental function that the brain is almost totally unable to support.

Table 1

Working memory and other mental test scores during euglycaemia and hypoglycaemia (n=32)


AJS and VMcA were funded by Eli Lilly during this research.



  • Competing interests: none declared.

  • IJD is the recipient of a Royal Society-Wolfson Research merit award.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.