The glycemic load of meals, cognition and mood in middle and older aged adults with differences in glucose tolerance: A randomized trial

Open AccessPublished:April 23, 2014DOI:https://doi.org/10.1016/j.clnme.2014.04.003

      Summary

      Background & aims

      In younger individuals a lower rather than a higher glycemic load (GL) meal has been shown to benefit cognition and mood; however the data in older and middle aged adults is limited and conflicting. One explanation is that the GL of a meal may interact with pre-existing glucose tolerance (GT) or the tendency for blood glucose to drop to low levels. The present study therefore considered the interaction between the GL of a meal, individual differences in glucose tolerance and the tendency to develop low blood glucose.

      Methods

      155 healthy older and middle aged adults, aged 45–80 years, were randomly assigned to receive either a glucose, sucrose or isomaltulose (Palatinose™) based breakfast. The meals were of identical macronutrient composition and differed only in GL. Mood and cognitive performance was assessed 30, 105 and 195 min after breakfast.

      Results

      Those with better but not poorer GT had better mood, episodic and working memory after a lower rather than a higher GL meal. The tendency to develop low blood glucose had little effect on adults' response to meals differing in GL.

      Conclusions

      Although it has been assumed previously that those with poor glucose tolerance would benefit from a low GL meal, it was in fact those with better tolerance who responded positively. (Clinical Trials Identifier: Clinicaltrials.gov NCT01842022).

      Keywords

      1. Introduction

      An inability to control the level of postprandial blood glucose has been related to cognitive decline, particularly in older adults [
      • Ryan C.M.
      • Geckle M.
      Why is learning and memory dysfunction in type 2 diabetes limited to older adults?.
      ]: those with glucose intolerance have poorer cognitive performance, most notably episodic memory [
      • Lamport D.J.
      • Lawton C.L.
      • Mansfield M.W.
      • Dye L.
      Impairments in glucose tolerance can have a negative impact on cognitive function: a systematic research review.
      ]. In addition, reactive hypoglycemia is associated with temporary decrements in performance and mood [
      • Taylor L.A.
      • Rachman S.J.
      The effects of blood sugar level changes on cognitive function, affective state, and somatic symptoms.
      ]. Therefore meals that mimic better glucose tolerance, that is those with a lower glycemic load (GL), have been suggested to facilitate memory and maintain mood. The present study therefore examined the interaction between the GL of meals and pre-existing glucose tolerance.
      The benefits to cognition of consuming a low GL meal has been demonstrated in children [
      • Benton D.
      • Maconie A.
      • Williams C.
      The influence of the glycaemic load of breakfast on the behaviour of children in school.
      ], adolescents [
      • Cooper S.B.
      • Bandelow S.
      • Nute M.L.
      • Morris J.G.
      • Nevill M.E.
      Breakfast glycaemic index and cognitive function in adolescent school children.
      ] and young adults [
      • Nabb S.
      • Benton D.
      The influence on cognition of the interaction between the macro-nutrient content of breakfast and glucose tolerance.
      ], however, the evidence with older and middle aged samples is limited and inconsistent. One possibility is that the response to a meal depends on pre-existing individual differences in glucose tolerance (GT). Glucose tolerance becomes poorer with age and it has been argued that meals with a lower rather than a higher GL may “reduce the cognitive differences between better and worse glucoregulators” [
      • Lamport D.J.
      • Lawton C.L.
      • Mansfield M.W.
      • Dye L.
      Impairments in glucose tolerance can have a negative impact on cognitive function: a systematic research review.
      ]. Thus the hypothesis presently tested was that those with poorer rather than better GT would differentially respond to meals differing in GL.
      To date studies of the interaction between an individual's glucose tolerance and the GL of a meal have produced inconsistent findings. Studies of older adults with impaired GT have reported that eating a low rather than a high GL meal either improved [
      • Papanikolaou Y.
      • Palmer H.
      • Binns M.A.
      • Jenkins D.J.
      • Greenwood C.E.
      Better cognitive performance following a low-glycaemic-index compared with a high-glycaemic-index carbohydrate meal in adults with type 2 diabetes.
      ] or has no effect on cognition [
      • Kaplan R.J.
      • Greenwood C.E.
      • Winocur G.
      • Wolever T.M.
      Cognitive performance is associated with glucose regulation in healthy elderly persons and can be enhanced with glucose and dietary carbohydrates.
      ]. Conversely, cognitive improvements have been noted following a high GL drink, with the suggestion that the response was greater in those with poorer GT [
      • Messier C.
      • Tsiakas M.
      • Gagnon M.
      • Desrochers A.
      • Awad N.
      Effect of age and glucoregulation on cognitive performance.
      ].
      An important consideration is the time scale over which effects were examined, a dimension that has varied considerably. Although there may or may not be a transient benefit from consuming a high GL within the first hour, high GL meals may generate a hypoglycemic reaction two to three hours later that disrupts cognitive functioning. Alternatively, a lower GL meal that releases glucose more slowly may not confer an immediate benefit but may prevent a decline later on. The finding that a low GL meal did not benefit the memory of young adults up to ninety minutes after consumption, but it did after three and a half hours [
      • Benton D.
      • Ruffin M.P.
      • Lassel T.
      • Nabb S.
      • Messaoudi M.
      • Vinoy S.
      • et al.
      The delivery rate of dietary carbohydrates affects cognitive performance in both rats and humans.
      ], illustrates the need to examine the prolonged response to a meal.
      In addition, any susceptibility to post-prandial hypoglycemia will influence the relationship between the GL of a meal and subsequent cognitive performance. When blood glucose levels drop to low values the brain is deprived of glucose resulting in neuroglycopenia, cognitive dysfunction and poorer mood [
      • McCrimmon R.J.
      Update in the CNS response to hypoglycemia.
      ]. For these reasons those with a tendency to develop low levels of blood glucose were distinguished and their response to meals differing in GL considered.
      To test these hypotheses the present study used various sugars to create meals of identical macro-nutrient composition that differed in their glycemic response. Isomaltulose (Palatinose™) is a disaccharide made from sucrose by the enzymatic rearrangement of the alpha 1,2 linkage between glucose and fructose, to an alpha 1,6 linkage [
      • Lina B.A.
      • Jonker D.
      • Kozianowski G.
      Isomaltulose (palatinose): a review of biological and toxicological studies.
      ]. It provides the same energy as sucrose but has a glycemic index (GI) of 32 rather than 67. It is completely but slowly digested and absorbed in the small intestine, leading to a continuous and prolonged delivery of blood glucose. The present study therefore contrasted the influence of meals sweetened with glucose, sucrose or isomaltulose. The aim was to establish whether the slow release of glucose from isomaltulose, rather than glucose or sucrose sweetened meals, beneficially affected the cognition of older and middle aged adults. In addition, how these sugars interacted with pre-existing differences in glucose tolerance was considered.

      2. Methods

      2.1 Subjects

      One hundred and fifty five adults, aged 45 years or over, who satisfied the inclusion criteria, were recruited following an appeal in the local media and the placing of posters. Exclusion criteria included anyone with type 1 or type 2 diabetes, chronic liver disease, or gastrointestinal problems that may interfere with absorption, for example Crohn's disease. Participants who were intolerant of any of the breakfast ingredients were also excluded as were those with a current diagnosis of a mood disorder or dementia. Eyesight and hearing was normal or corrected to normal. Of the subjects 16.1% were taking drugs to control blood pressure, 3.8% thyroxine, 3.2% statins and 2.7% asthma controlling drugs (Table 1). The incidence of the taking of medication did not differ between groups. All those recruited finished the trial. The procedure was followed with the approval of the local ethics committee and only after the participants had given written informed consent. All testing took place in the laboratories of Swansea University between September 2011 and April 2013.
      Table 1Demographic data for the four glucoregulatory groups. Data are mean (SE) or number of participants. Adults with better GT and LBG above baseline were significantly younger than those with poorer GT and LBG above baseline. Similarly, adults with better GT and LBG below baseline were significantly younger than those with poorer GT and LBG above baseline.
      Poorer GT

      LBG above baseline
      Poorer GT

      LBG below baseline
      Better GT

      LBG above baseline
      Better GT

      LBG below baseline
      p
      BMI28.1(0.8)29.5(1.6)25.0(1.4)26.8(1.0)ns
      Gender
       M20171011ns
       F4726158ns
      Age58(11.4)*,**56.2(9.6)53.0(10.7)*54.2(10.8)***p < 0.01

      **p < 0.005
      Thyroxine5/672/434/252/19ns
      Asthma (inhaler)5/674/430/250/19ns
      Statins8/671/431/251/19ns
      ACE inhibitor15/672/433/253/19ns
      Beta blocker2/670/430/250/19ns

      2.2 Procedure – day 1

      All subjects were treated identically on the first day when they became familiar with the test battery and took an oral glucose tolerance test (OGTT). After fasting over-night 1.75 g of glucose per kilogram of body weight, to a maximum dose of 75 g, was consumed. Every thirty minutes, for two and a half hours, blood glucose was monitored from finger pricks using an ExacTech sensor (Medisense Britain Limited) that using an enzymic method, coupled with microelectronic measurement, which has been shown to be accurate [
      • Matthews D.R.
      • Holman R.R.
      • Bown E.
      • Steenson J.
      • Watson A.
      • Hughes S.
      • et al.
      Pen-sized digital 30-second blood glucose meter.
      ]. Subjects were also given a chance to practise the test battery (outlined below).

      2.3 Procedure – day 2

      On a second testing day, after again fasting overnight, subjects completed a baseline mood measure and in a parallel design they randomly, and under a double-blind procedure, consumed one of three test meals (0845-0900). The random sequence was computer generated by DB after which participants were allocated to the indicated condition by HY. The subjects and all experimenters who met the subjects were blind as to the nature of the meals consumed. Subjects were allocated to the condition indicated by randomization in the order they were recruited.
      After breakfast the test battery was taken on three occasions 0930–1000; 1045–1130; 1200–1230, that corresponded to 30, 105 and 195 min after ingestion. Different but comparable versions of each test were used at each session. Participants consumed the entire breakfast and did not consume any other food throughout the morning, although water was consumed if requested. Each test battery included the tests in the same order: mood, immediate recall of words, serial sevens, reaction times, sustained attention, delayed recall of word list, mood.

      2.4 Breakfast

      Each meal consisted of two slices of whole meal toast (60 kcal) topped with reduced sugar jam, 100 g of plain low fat yogurt that was sweetened with either 15 g of glucose, sugar or isomaltulose. Participants also received an orange flavored drink that was sweetened with 25 g of one of the three sugars. Table 2 presents the nutritional composition of the three test meals that were identical in terms of macro-nutrients and appearance but produced a GL of 24.3, 34.9 or 45.4, for the isomaltulose, sucrose and glucose based breakfasts.
      Table 2The macro-nutrient content of the experimental meals.
      Higher GLMedium GLLower GL
      Energy

      (Kcal)
      Carb

      (g)
      Pro

      (g)
      Fat

      (g)
      GIGLEnergy

      (Kcal)
      Carb

      (g)
      Pro

      (g)
      Fat

      (g)
      GIGLEnergy

      (Kcal)
      Carb

      (g)
      Pro

      (g)
      Fat

      (g)
      GIGL
      2 slices wholemeal bread.11019.45.01.26913.311019.45.01.26913.311019.45.01.26913.3
      Low carbohydrate jam112.40.20.00**112.40.20.00**112.40.20.00**
      Low calorie yoghurt 100 g416.04.30.25191.14416.04.30.25191.14416.04.30.25191.14
      15 g Glucose60150010015
      15 g Sucrose6015006710.0
      15 g Isomaltulose601500324.8
      Drink 25 g Glucose64250010025
      Drink 25 g Sucrose6425006716.75
      Drink 25 g Isomaltulose64250032.8
      Total27556.49.31.4554.427556.49.31.4541.2527556.49.31.4527.2
      * The low carbohydrate jam was specially made for the study from strawberries sweetened with sucralose, an artificial sweetener that is not broken down by the body so it offered no calories. The GI of the jam is unknown so the jam is not included in the summated nutritional data. However, the GI of strawberries is 40, so the jam that was eaten by all subjects might be expected to have added about 1 to each of the three total GLs.

      3. Test battery

      3.1 Episodic memory – word list recall

      Episodic memory, the type of long-term memory involved in storing contextual knowledge and experiences, declines with age [
      • Verhaeghen P.
      • Salthouse T.A.
      Meta-analyses of age-cognition relations in adulthood: estimates of linear and nonlinear age effects and structural models.
      ], has been shown susceptible to nutritional interventions [
      • Young H.
      • Benton D.
      Caffeine can decrease subjective energy depending on the vehicle with which it is consumed and when it is measured.
      ], and can be measured by recalling a word list. Using the MRC Psycholinguistic Database four lists of thirty words were constructed, matched for the number of syllables, image-ability and the frequency with which they occur in English. Using a recorder words were presented at a rate of one word every two seconds. Immediately after presentation, as many words as possible were written down (immediate or short-term recall). Approximately 25 min later, after completion of the other tasks, subjects were again asked to recall these words (delayed or long-term recall).

      3.2 Semantic memory – verbal fluency

      Semantic memory reflects knowledge that is unrelated to specific experiences and the meaning of words and knowledge. Verbal fluency was assessed as a measure of semantic memory. In a one minute period subjects were asked to provide as many words as possible beginning with a particular letter of the alphabet [
      • Lezak M.D.
      Neuropsychological assessment.
      ]: words beginning with C, F and L on day one (practice session) and P, R and W on day two (experimental session). In each set of letters words beginning with the first letter occur frequently, with the second letter less frequently and with the third letter even less often. The scores excluded proper nouns such as people's names, place names or the same word with a different suffix. The test was only administered during the third testing session of day two as four comparable versions could not be created.

      3.3 Working memory – serial sevens

      Working memory is a type of short term memory that temporarily holds and manipulates information: it declines with age [
      • Verhaeghen P.
      • Salthouse T.A.
      Meta-analyses of age-cognition relations in adulthood: estimates of linear and nonlinear age effects and structural models.
      ] and can be assessed using the serial sevens test. A computerized version of the task was used in which subjects were required, from a starting number between 800 and 999, to say whether a second given number was seven less than the previously observed number. Participant responses to 28 sequences were recorded and the number of errors made and the times to respond were reported.

      3.4 Vigilance – sustained attention

      A computer generated a series of digits at the rate of 100 digits per minute (Rapid Information Processing Task). Subjects pressed the space bar when they detected three consecutive odd or consecutive even digits. Eight target sequences were presented every minute for a total of five minutes. Following the presentation of the third consecutive digit, 3000 ms were allowed for a correct response to be made. Responses made at any other time were recorded as errors. A minimum of five and a maximum of thirty digits separated any two target sequences. This task has previously been shown to be sensitive to GL manipulations [
      • Nabb S.
      • Benton D.
      The effect of the interaction between glucose tolerance and breakfasts varying in carbohydrate and fibre on mood and cognition.
      ].

      3.5 Reaction times

      The reaction time procedure was based on that of Jensen [
      • Jensen A.R.
      Individual differences in the Hick paradigm.
      ]. On a panel eight lamps were arranged in a semicircle, each five and a half inches from a central button (the home key). The index finger was placed on the home key. Within one to two seconds an auditory warning signal sounded and after a random interval of one to four seconds one of the lamps illuminated. The subject then extinguished the light by depressing a button directly below the lamp, using the finger initially on the home key. All subjects completed a practice session of twenty trials using all eight lamps. Simple reaction times were then measured for twenty trials using one lamp. Choice reaction times were then measured over three sets of twenty trials when one of two, four or eight lamps could potentially illuminate. A decision time was the time taken to lift the finger from the home key. Movement time was the time from the hand leaving the home key to pressing the button under the illuminated light.

      3.6 Mood

      Subject were asked to report how they felt “at this moment” using visual analog scales with pairs of adjectives at the ends of 100 mm lines; Composed/Anxious; Hostile/Agreeable; Elated/Depressed; Unsure/Confident; Energetic/Tired; Confused/Clearheaded [
      • McNair D.M.
      • Lorr M.
      • Droppleman L.F.
      Revised manual for the profile of mood states.
      ]. Mood was measured both before and after each of the three testing sessions.

      3.7 Glucoregulatory profile

      Given that those with poorer glucose tolerance (GT) are more likely to display mild cognitive decline, and that low blood glucose is also disruptive [
      • Lamport D.J.
      • Lawton C.L.
      • Mansfield M.W.
      • Dye L.
      Impairments in glucose tolerance can have a negative impact on cognitive function: a systematic research review.
      ], the sample was distinguished depending on the results of the day one OGTT. Subjects were divided into four groups according to their blood glucose levels after 120 and 150 min.
      To allow a sufficient sample size, if their blood glucose after 120 min was 7.0 mmol/l or higher they were considered to have poorer GT; if their blood glucose was less than 7 mmol/l they were considered to have better GT. Although the clinical definition for glucose intolerance is blood glucose >7.8 mmol/L 2 h after an OGTT, how such thresholds relate to changes in cognition is currently unknown [
      • Cherbuin N.
      • Anstey K.
      Could ignoring higher blood sugar levels in the normal range in nondiabetics compromise cerebral health?.
      ]. Indeed a recent study found that blood glucose levels that are raised, but still within a ‘normal’ range, as defined by the WHO, are associated with lower grey/white matter volumes in otherwise cognitively healthy older adults [
      • Mortby M.
      • Janke A.
      • Anstey K.
      • Sachdev P.
      • Cherbuin N.
      High “Normal” blood glucose is associated with decreased brain volume and cognitive performance in the 60s: the PATH through Life Study.
      ]. 7.0 mmol/L was chosen in the present study as this allowed reasonable sample sizes per group to carry out the analysis.
      The lowest blood glucose value (LBG) was determined and individuals were divided into two further groups, according to whether at any point during the test they either fell or not fall below the fasting blood glucose value. Those whose blood glucose fell below fasting values all had capillary glucose levels <5.0 mmol/l by the end of the test (after 150 min). Importantly, 50% fell below 4.0 mmol/l and 30% had glucose levels <3.6 mmol.
      By classifying subjects depending on whether they had poorer or better GT, and whether the LBG did or did not fall below the baseline value, four Glucoregulatory groups were created: better GT and LBG above baseline (N = 25); better GT and LBG below baseline (N = 19); poorer GT and LBG above baseline (N = 67); poorer GT and LBG below baseline (N = 43). The glucose profiles of the four groups on the first day of testing are illustrated in Fig. 1. It is apparent that four distinct patterns resulted that were used when analyzing the findings from the second day of testing.
      Figure thumbnail gr1
      Fig. 1Oral glucose tolerance profiles of four groups of subjects. The data are mean blood glucose values as mmol/l for four groups defined in terms of poorer and better glucose tolerance (above and below 7 mmol/dl at 120 min) and either staying above or falling below baseline values. Poorer glucose tolerance/Above baseline. Poorer glucose tolerance/Below baseline. Better glucose tolerance/Below baseline. Better glucose tolerance/Above baseline.

      3.8 Statistical analysis

      The data were examined using appropriate analysis of co-variance design with performance on the first day of testing as the covariate. Typically they took the form of Meal (Glucose, Sucrose, Isomaltulose) × Time (30, 105, 195 min) × Glucoregulatory group (Poorer GT and LBG below baseline/Better GT and LBG below baseline/Poorer GT and LBG above baseline/Better GT and LBG above baseline). When interactions were significant appropriate post hoc tests were conducted to determine the nature of the interaction.

      4. Results

      Given the wide age range used in the present study preliminary analysis tested whether age moderated the influence of GL on cognition. Age was considered by distinguishing those 60 years and below from those 61 years and above. This was chosen as the cut off because it had been argued that there may be a critical period, which begins at around 60 years, during which glucose intolerance has the greatest impact on cognition [
      • Biessels G.J.
      • Deary I.J.
      • Ryan C.M.
      Cognition and diabetes: a lifespan perspective.
      ] and after which age related cognitive decline becomes more apparent [
      • Verhaeghen P.
      • Salthouse T.A.
      Meta-analyses of age-cognition relations in adulthood: estimates of linear and nonlinear age effects and structural models.
      ]. Adults aged 61 or over remembered fewer words (F(1,147) = 20.682, p < 0.0001), had slower working memory during the final test session (195 min after breakfast) (F(2,284) = 4.133, p < 0.02) and had slower decision times (F(1,147) = 7.385, p < 0.0001). However, age did not interact with the GL of breakfast to influence cognition and therefore was not considered further.
      It has previously been reported that gender may influence a subjects responses to the GL of a meal [
      • Mahoney C.R.
      • Taylor H.A.
      • Kanarek R.B.
      • Samuel P.
      Effect of breakfast composition on cognitive processes in elementary school children.
      ], therefore preliminary analysis tested for any interactions between gender and GL. Gender significantly predicted memory (F(1,157) = 4.378, p < 0.05); females has better memory than males (M 9.4(0.4) F 10.6(0.3)). However, there were no interactions between gender and the GL of breakfast.
      BMI ranged from 19.5 to 30.3 therefore, given such a wide range, it was considered whether BMI moderated the effect of GL on cognition. BMI was considered by distinguishing those with a BMI greater than 25.0 from those 25.0 or under. Those with a BMI greater then 25.0 had a poorer memory than those with a BMI 25.0 or less (F(1,147) = 5.319, p < 0.006), however BMI did not influence participant responses to the GL of breakfast.

      4.1 Mood

      On the second day, mood was measured at baseline and before and after each of the three testing sessions. Initially baseline values were checked to ensure groups were equivalent: the interaction Meal × Glucoregulatory group was not significant (F(6,140) = 1.025, ns) and neither were the main effects of Meal (F(2,140) = 1.603, ns) or Glucoregulatory group (F(3,140) = 0.939, ns). To determine the effect of the type of meal on mood change from baseline, scores were calculated by subtracting the mood ratings at baseline from subsequent scores. A positive score indicated an improvement in mood and a negative score indicated a decline in mood from baseline. These scores were then entered in the analysis Meal × Time (30, 105, 195 min) × Glucoregulatory Group × Before/after testing.
      Initially the six mood dimensions were considered independently, resulting in one main effect of the influence of the type of sugar. With the Agreeable – Hostile dimension there was a main effect of Meal (F(2,140) = 2.88, p < 0.05). Those who consumed isomaltulose (Palatinose™) were more agreeable than those who consumed glucose (p < 0.03). In addition sucrose rather than glucose consumption increased ratings of being agreeable (p < 0.03). There were no significant differences between those who had eaten sucrose or isomaltulose (Fig. 2).
      Figure thumbnail gr2
      Fig. 2The effect of Meal on agreeableness. Data are mean (standard error) for decline in VAS ratings of agreeableness across the morning. Adults that consumed isomaltulose were more agreeable (less decline) than those who consumed glucose (p < 0.03). In addition, adults who ate sucrose were more agreeable than those that consumed glucose (p < 0.03).
      There was also a series of significant interactions involving the type of meal, for example, with Clearheaded – Confused (F(12,280) = 2.33, p < 0.007) and Elated – depressed (F(12,280) = 2.65, p < 0.04) the Meal × Glucoregulatory group × Time interactions reached statistical significance. The effects of the type of meal were, however, similar and rather than reporting a series of three way interactions for the six mood dimensions, for brevity and clarity all six were added to produce an overall mood score that illustrates the effects that were observed. The interaction Meal × Glucoregulatory group × Time was statistically significant (F(12,280) = 2.177 p < 0.01; Fig. 3) and is representative of the types of effects found with specific mood dimensions. Post hoc tests found a series of differences in the response to meal.
      Figure thumbnail gr3
      Fig. 3The influence of type of meal on mood in the four glucoregulatory groups. Data are mean change in mood from baseline. Adults with better GT and LBG above baseline had better moods if they ate isomaltulose than if they ate glucose after 105 (p < 0.03) and 195 min (p < 0.03). In addition, adults with better GT and LBG above baseline had better moods if they ate isomaltulose than if they ate sucrose after 105 min (p < 0.003) and 195 min (p < 0.001). Conversely, adults with poorer GT and LBG above baseline had better moods, after 30 min, if they ate glucose rather than isomaltulose (p < 0.05). Adults with poorer GT and LBG below baseline, who ate sucrose, reported better moods, after 105 min, than those who ate glucose. The nature of the meal consumed did not affect the mood of adults with better GT and LBG below baseline.

      4.1.1 Better GT and LBG above baseline

      A profile of better GT with no LBG was associated with differential responses to the type of meal. After both 105 (p < 0.003) and 195 min (p < 0.001) having consumed sucrose was associated with a better mood than after a glucose containing meal. The consumption of isomaltulose, rather than sucrose, resulted after 105 min in a better mood (p < 0.01). Similarly, after 195 min the isomaltulose rather than glucose meal resulted in a more positive mood (p < 0.03).

      4.1.2 Better GT and LBG below baseline

      In those with better GT and LBG below baseline mood did not depend on the meal that had been eaten.

      4.1.3 Poorer GT and LBG below baseline

      Those with the poorer GT and LBG profile again responded to the nature of sugar consumed. After 105 min, having eaten the glucose containing meal resulted in a poorer mood than when sucrose (p < 0.04) was eaten with a trend for difference between glucose and isomaltulose (p < 0.07).

      4.1.4 Poorer GT and LBG above baseline

      When those with poorer GT and LBG above baseline were considered those that ate glucose rather than isomaltulose (p < 0.05) reported better moods after 30 min. There was a similar trend for those who ate glucose rather then sucrose (p < 0.07) to report better moods after 30 min but this just missed significance.

      4.2 Episodic memory

      The episodic memory data were analyzed using a four-way ANOVA; Meal (Glucose, Sucrose, Isomaltulose) × Short or Long term memory (ST/LT) × Time (30, 105, 195 min) × Glucoregulatory Group. There was a significant Meal × Glucoregulatory group × Time interaction (F(12,280) = 2.23, p < 0.01; Fig. 4).
      Figure thumbnail gr4
      Fig. 4The influence of type of meal on memory in the four glucoregulatory groups. Data are mean number of words remembered. Adults with better GT and LBG above baseline remembered more words after 105 (p < 0.03) and 195 min (p < 0.05) if they had eaten isomaltulose than if they had consumed glucose. Adults with better GT and LBG above baseline after 15 min (p < 0.02), 105 min (p < 0.04) and 195 min (p < 0.03), remembered more words if they had eaten isomaltulose rather than sucrose Conversely, memory after 15 min in adults with poorer GT and LBG above baseline was better after glucose rather than isomaltulose. Adults with better GT and LBG below baseline who had eaten glucose had poorer memory after 195 min compared to those who had eaten sucrose (p < 0.01) or isomaltulose (p < 0.004). The number of words remembered by those with poorer GT and LBG below baseline did not depend on the type of meal they had been eaten.

      4.2.1 Better GT and LBG above baseline

      When those with better GT and LBG above the baseline were considered those who ate isomaltulose had better memories after 30 min (p < 0.02), 105 min (p < 0.04) and 195 min (p < 0.03) compared to those who had eaten sucrose. Those who consumed isomaltulose also had better memories compared to those who ate glucose after 105 (p < 0.03) and 195 min (p < 0.05), but not earlier (Fig. 4).

      4.2.2 Better GT and LBG below baseline

      When those with better GT and LBG below the baseline were examined, those that ate glucose had poorer memory after 195 min compared to those that had eaten sucrose (p < 0.01) and isomaltulose (p < 0.004). There were, however, no effects of meal after 30 or 105 min (Fig. 4).

      4.2.3 Poorer GT and LBG above baseline

      With the group with poorer GT, and LBG above the baseline, the consumption of glucose rather than isomaltulose resulted in better memories after 30 min (p < 0.02), although not during the two later testing sessions. There were no differences between those who consumed sucrose and those who ate isomaltulose at any time. Similarly there were no differences in those eating glucose or sucrose (Fig. 4).

      4.2.4 Poorer GT and LBG below baseline

      The response to meal did not differ in those with poorer GT and LBG that fell below the baseline.

      4.3 Working memory

      4.3.1 Accuracy

      When the number of incorrect responses were considered the Meal (Glucose, Sucrose, Isomaltulose) × Time (30, 105, 195 min) × Glucoregulatory Group interaction was not significant (F(12,264) = 0.62, ns). However, the interaction Meal × Glucoregulatory Group approached significance (F(6,264) = 1.98, p < 0.06). Older and middle aged adults with better GT and LBG above the baseline made more errors after 180 min if they ate glucose rather than isomaltulose (p < 0.05). There was a similar pattern of results after 105 min (6.2(7.1) vs 1.8(0.7)) and after 30 min (6.7(5.1) vs 2.5(2.1)), although these effects just missed statistical significance. The GL of breakfast did not influence working memory accuracy in any of the other three glucoregulatory groups.

      4.3.2 Reaction times

      With the speed of reaction the Meal (Glucose, Sucrose, Isomaltulose) × Time (30, 105, 195 min) × Glucoregulatory Group was not significant (F(12,264) = 1.10, n.s.).

      4.4 Semantic memory

      With the semantic memory measure the Meal × Glucoregulatory Group interaction was not significant (F(6,140) = 1.62, n.s.).

      4.5 Reaction times

      With decision times the interaction Meal × Time (30, 105, 195 min) × Glucoregulatory group × Number of lamps [
      • Ryan C.M.
      • Geckle M.
      Why is learning and memory dysfunction in type 2 diabetes limited to older adults?.
      ,
      • Lamport D.J.
      • Lawton C.L.
      • Mansfield M.W.
      • Dye L.
      Impairments in glucose tolerance can have a negative impact on cognitive function: a systematic research review.
      ,
      • Benton D.
      • Maconie A.
      • Williams C.
      The influence of the glycaemic load of breakfast on the behaviour of children in school.
      ,
      • Kaplan R.J.
      • Greenwood C.E.
      • Winocur G.
      • Wolever T.M.
      Cognitive performance is associated with glucose regulation in healthy elderly persons and can be enhanced with glucose and dietary carbohydrates.
      ] was non-significant (F(36,804) = 0.70, n.s.), however, the interaction Meal × Time reached significance (F(4,268) = 2.71, p < 0.03), although post hoc tests failed to find any significant differences between meals. Movement times proved similarly non-significant: Meal × Time (30, 105, 195 min) × Glucoregulatory group × Number of lamps [
      • Ryan C.M.
      • Geckle M.
      Why is learning and memory dysfunction in type 2 diabetes limited to older adults?.
      ,
      • Lamport D.J.
      • Lawton C.L.
      • Mansfield M.W.
      • Dye L.
      Impairments in glucose tolerance can have a negative impact on cognitive function: a systematic research review.
      ,
      • Benton D.
      • Maconie A.
      • Williams C.
      The influence of the glycaemic load of breakfast on the behaviour of children in school.
      ,
      • Kaplan R.J.
      • Greenwood C.E.
      • Winocur G.
      • Wolever T.M.
      Cognitive performance is associated with glucose regulation in healthy elderly persons and can be enhanced with glucose and dietary carbohydrates.
      ] (F(36,804) = 1.02, n.s.).

      4.6 Vigilance

      Again with the reaction times of the vigilance task the interaction Meal × Time (30, 105, 195 min) × Glucoregulatory group × Minute (1_5 min of test) was not significant (F(48,1016) = 0.85, n.s.). Although the interaction Time × Minute × Group did reach significance (F(24,1016) = 1.95, p < 0.004) this appeared to reflect a chance variation in a one cell. When accuracy during the vigilance task was considered again the meal was without effect (Meal × Time (30, 105, 195 min) × Glucoregulatory Group × Minute; (F(48,1016) = 1.06, ns).

      4.7 Summary

      • The influence of meals differing in GL depended on pre-existing individual differences in glucose tolerance.
      • Although it has been assumed previously that it was those with poor glucose tolerance that would benefit from a low GL meal, it was those with better glucose tolerance who responded to a meal containing isomaltulose (Palatinose™) with better mood and memory.
      • It has been suggested that a tendency to develop low levels of blood glucose will disrupt neural functioning, a phenomenon that can be reduced by consuming low GL meals. However, no evidence of this mechanism was found as there was no clear interaction between the GL of breakfast and the pre-existing tendency for blood glucose to fall to low values.

      5. Discussion

      The major finding was that on a number of occasions individual glucoregulatory status interacted with the type of meal that was consumed. Interestingly, older and middle aged adults with better GT and no tendency for blood glucose to fall below baseline value, that is those who were best able to manage their blood glucose levels, benefited most from consuming a lower rather than a higher GL meal (Fig. 2, Fig. 3, Fig. 4). In particular, those who ate the isomaltulose (Palatinose™) based meal had both a better episodic and working memory and a better mood. These effects are consistent with the literature that suggests that a lower rather than a higher GL meal may improve cognitive performance and mood in both children [
      • Benton D.
      • Maconie A.
      • Williams C.
      The influence of the glycaemic load of breakfast on the behaviour of children in school.
      ] and adolescents [
      • Cooper S.B.
      • Bandelow S.
      • Nute M.L.
      • Morris J.G.
      • Nevill M.E.
      Breakfast glycaemic index and cognitive function in adolescent school children.
      ]. Given their young age it is reasonable to assume that such populations would have relatively better GT compared to older adults. It may be the case that the inconsistencies in the literature dealing with older adults reflect a failure to establish the prior ability to regulate blood glucose levels.
      From this perspective it may not be surprising that the literature addressing the effects of GL on mental performance in older adults, with or without poor GT, is limited and the findings inconsistent. On one hand Papanikolaou et al. [
      • Papanikolaou Y.
      • Palmer H.
      • Binns M.A.
      • Jenkins D.J.
      • Greenwood C.E.
      Better cognitive performance following a low-glycaemic-index compared with a high-glycaemic-index carbohydrate meal in adults with type 2 diabetes.
      ] reported in older diabetics that a lower GL (50 g pasta) rather than a higher GL meal (50 g white bread), improved cognition sixty to one hundred and forty minutes after consumption. However, it was unclear whether poor glycemic state interacted with the nature of the meals as there was no healthy control group. Also the majority of subjects were taking oral hypoglycemic medication (metformin or sulphonylureas) that are known to influence insulin secretion, insulin resistance and gluconeogenesis and thus may interact with the GL of a meal. Similarly, Nabb and Benton [
      • Nabb S.
      • Benton D.
      The influence on cognition of the interaction between the macro-nutrient content of breakfast and glucose tolerance.
      ] found that meals that were low in carbohydrates and high in fiber resulted in better memory thirty and one hundred and five minutes after consumption; an effect that was only observed among those with higher fasting glucose levels. However, when interpreting these findings one must consider that although impaired fasting glucose and impaired glucose tolerance are correlated they are not synonymous and represent different underlying pathologies. On the other hand, Lamport et al. [
      • Lamport D.J.
      • Dye L.
      • Mansfield M.W.
      • Lawton C.L.
      Acute glycaemic load breakfast manipulations do not attenuate cognitive impairments in adults with type 2 diabetes.
      ] reported that neither diabetics, nor older adults with better GT, benefited from a low (toast and yogurt) rather than high GL meal (glucose drink) after thirty or one hundred and twenty minutes. However, it is not clear if glucose tolerance was measured in the control group and thus some of the adults considered ‘healthy’ may actually have had poorer GT, as defined in the present study.
      Although in the present study a low GL meal did not generally influence cognition or mood in older and middle aged adults with poorer GT, it is interesting that on a number of occasions they selectively benefitted from a higher GL meal during the immediate postprandial period. Specifically, those with poorer GT had better memory if they had consumed glucose rather than sucrose or isomaltulose based meals. Similar short term effects have been reported previously. For example Messier et al. [
      • Messier C.
      • Tsiakas M.
      • Gagnon M.
      • Desrochers A.
      • Awad N.
      Effect of age and glucoregulation on cognitive performance.
      ] found that drinking 50 g of glucose, rather than a placebo, attenuated the deficits observed in older adults with poor GT, and reduced the difference in memory between better and poorer gluco-regulators in a period up to forty-five minutes after consumption. Similarly Kaplan et al. [
      • Kaplan R.J.
      • Greenwood C.E.
      • Winocur G.
      • Wolever T.M.
      Cognitive performance is associated with glucose regulation in healthy elderly persons and can be enhanced with glucose and dietary carbohydrates.
      ] gave older adults meals that provided 50 g of carbohydrate that differed in the speed of absorption (glucose drink, potato and barley). Although those with poorer glucose regulation had greater improvements in memory after all three meals (tested up to one hundred and five minutes) improvements were similar in all carbohydrate treatments. Interestingly, a study by Craft et al. [
      • Craft S.
      • Murphy C.G.
      • Wemstrom J.
      Glucose effects on complex memory and nonmemory tasks: the influence of age, sex, and glucoregulatory response.
      ] found that drinking 50 g of glucose, rather than a placebo, improved memory in older men with better but not poorer GT. However, glucose also improved the memory of younger men with poorer GT but in contrast decreased the performance of younger men with better GT. What is interesting is that older men with better GT were comparable to younger men with poorer GT; thus there could be an optimal range of GT within which glucose has a facilitatory effect. However, in these studies there was substantial variation in the definition used to classify those having better or poorer GT such that a range of mechanisms could have differentially interacted with a meal.
      Although the mechanisms mediating the effect of GT on cognitive decline are yet to be elucidated, Convit [
      • Convit A.
      Links between cognitive impairment in insulin resistance: an explanatory model.
      ] proposed that individuals with glucose intolerance have poorer cognitive performance because they have endothelial dysfunction and are therefore less able to transport glucose across the blood brain barrier. It is interesting that our sample of older and middle aged adults with poorer GT benefitted in the short-term from a higher rather than a lower GL. It is possible that those with impaired GT may require higher blood glucose and insulin responses to drive glucose utilization by the brain. Although the transport of glucose into the brain is not thought to be insulin dependent, insulin may act as a vasodilator [
      • Ghasemi R.
      • Haeri A.
      • Dargahi L.
      • Mohamed Z.
      • Ahmadiani A.
      Insulin in the brain: sources, localization and functions.
      ] and thus increased insulin levels (as a result of a glucose load) may improve endothelial functioning and indirectly aid transport of glucose across the blood brain barrier.
      A further finding from the present study was that the tendency to develop lower levels of blood glucose did not moderate the effect of glycemic load on cognitive performance. If anything the tendency to develop an LBG below baseline values seemed to reduce the benefits observed in those with better GT. For example, those with better GT and LBG above baseline values had better memory throughout the entire morning after eating isomaltulose. However, those with better GT, and the tendency for LBG to fall below baseline, only benefitted from a lower GL during the late postprandial period (195 min).
      We previously found that a tendency to develop low blood glucose levels benefitted those with poor glucose tolerance [
      • Young H.
      • Benton D.
      The nature of the control of blood glucose in those with poorer glucose tolerance influences mood and cognition.
      ]. Similarly, it is well documented that some insulin treated diabetics after re-current hypoglycemic episodes develop ‘hypoglycemia unawareness’. This not only involves a reduction in sympathetic nervous system symptoms [
      • Cryer P.E.
      Mechanisms of hypoglycemia-associated autonomic failure and its component syndromes in diabetes.
      ] but also a reduction in cognitive dysfunction during subsequent hypoglycemic episodes [
      • Fruehwald-Schultes B.
      • Born J.
      • Kern W.
      • Peters A.
      • Fehm H.L.
      Adaptation of cognitive function to hypoglycemia in healthy men.
      ]. The mechanisms involved in this process are not yet determined, however, there is evidence that brain adaptation can result from exposure to hypoglycemia. For example, in animals that previously had experienced hypoglycemia, hippocampal glucose concentrations were subsequently higher under normal glycemic conditions [
      • McNay E.C.
      • Williamson A.
      • McCrimmon R.J.
      • Sherwin R.S.
      Cognitive and neural hippocampal effects of long-term moderate recurrent hypoglycemia.
      ]. In healthy humans' experimental exposure to recurrent hypoglycemia resulted in the preservation of cognitive performance and an increased uptake of glucose by the brain when on a later occasion low glucose levels were again experienced [
      • Boyle P.J.
      • Nagy R.J.
      • O'Connor A.M.
      • Kempers S.F.
      • Yeo R.A.
      • Qualls C.
      Adaptation in brain glucose uptake following recurrent hypoglycemia.
      ]. It is possible that the older and middle aged adults in our sample who had a tendency for LBG to be below the baseline did not benefit from the lower GL meal, with its sustained supply of glucose from the periphery, because they had already developed compensatory mechanisms (i.e. increased brain interstitial glucose levels) to deal with a transient shortage in peripheral energy supply. Again, clearly more research is required.
      It is inevitable that this type of analysis produces groups of different sample size, raising the question as to whether the results reflected differences in the consequent statistical power. If as an example memory after 195 min is considered, the largest differences resulted in those with better GT and LBG that remained above baseline. The differences in this group required groups of 14 to have 0.8 power at the p < 0.05 level. As such we can be confident of the non-significant finding of those with poorer GT and LBG below the baseline, as the latter group had more than fifty percent more subjects. There were fewer subjects in both the groups where LBG fell below baseline, albeit that there were still significant differences between those consuming the different meals when it was combined with better GT. The nature of the meal did not influence those with poorer GT and LBG below baseline and logically the possibility cannot be excluded that a larger sample size might have resulted in statistical significance although the similarity of the means gives no support for such a suggestion.
      Given the association between age and glucoregulatory status that was observed during the preliminary analysis, a reasonable question would be does age, rather than GT per se, moderate the effect of GL on cognition? There was no evidence that this was the case; although adults aged 61 or above had poorer episodic and working memory and slower decision times than those 60 and younger, there were no interactions with the GL of breakfast. Thus in the present study it was GT, rather than age, that was the critical moderator of the effect of GL on cognition.
      In conclusion, a lower rather than a higher GL meal improved cognition and mood in healthy older and middle aged adults with better GT. These effects were strongest during the late postprandial period (105–195 min post meal). A lower GL meal did not benefit older and middle aged adults with poorer GT or interact with the tendency to develop low blood glucose. It is possible that defects in the brain glucose metabolism/transport may help explain why older and middle aged adults with poorer GT are unable to benefit from a lower GL meal in the same way as healthy older and middle aged adults. There is a need to further consider the nature of the problems that underlie glucose intolerance related cognitive decline, in the hope that such findings will help in the development of interventions that aid different sections of the population. It is similarly unclear to what extent any benefits will be lasting or alternatively whether the body will adapt with repeated presentations of a particular type of meal. Finally the novelty of the present findings, that were not predictable, suggests that they should be replicated before any recommendation is made of their use.

      Conflict of interests

      Both authors declare that they have no conflict of interest.

      Acknowledgments

      The assistance of the volunteers is gratefully acknowledged. We are similarly grateful to the BENEO Group, Germany , for funding of study and supplying Palatinose™ and sugar sweetened foods. Beneo, a member of the Suedzucker Group, reserves the exclusive right to use the results and data for possible Health Claim requests. Both authors were responsible for the design, data analyses and drafting of the manuscript whereas HY was responsible for the collection of the data.

      References

        • Ryan C.M.
        • Geckle M.
        Why is learning and memory dysfunction in type 2 diabetes limited to older adults?.
        Diabetes Metab Res Rev. 2000; 16: 308-315
        • Lamport D.J.
        • Lawton C.L.
        • Mansfield M.W.
        • Dye L.
        Impairments in glucose tolerance can have a negative impact on cognitive function: a systematic research review.
        Neurosci Biobehav Rev. 2009; 33: 394-413
        • Taylor L.A.
        • Rachman S.J.
        The effects of blood sugar level changes on cognitive function, affective state, and somatic symptoms.
        J Behav Med. 1988; 11: 279-291
        • Benton D.
        • Maconie A.
        • Williams C.
        The influence of the glycaemic load of breakfast on the behaviour of children in school.
        Physiol Behav. 2007; 92: 717-724
        • Cooper S.B.
        • Bandelow S.
        • Nute M.L.
        • Morris J.G.
        • Nevill M.E.
        Breakfast glycaemic index and cognitive function in adolescent school children.
        Br J Nutr. 2012; 107: 1823-1832
        • Nabb S.
        • Benton D.
        The influence on cognition of the interaction between the macro-nutrient content of breakfast and glucose tolerance.
        Physiol Behav. 2006; 87: 16-23
        • Papanikolaou Y.
        • Palmer H.
        • Binns M.A.
        • Jenkins D.J.
        • Greenwood C.E.
        Better cognitive performance following a low-glycaemic-index compared with a high-glycaemic-index carbohydrate meal in adults with type 2 diabetes.
        Diabetologia. 2006; 49: 855-862
        • Kaplan R.J.
        • Greenwood C.E.
        • Winocur G.
        • Wolever T.M.
        Cognitive performance is associated with glucose regulation in healthy elderly persons and can be enhanced with glucose and dietary carbohydrates.
        Am J Clin Nutr. 2000; 72: 825-836
        • Messier C.
        • Tsiakas M.
        • Gagnon M.
        • Desrochers A.
        • Awad N.
        Effect of age and glucoregulation on cognitive performance.
        Neurobiol Aging. 2003; 24: 985-1003
        • Benton D.
        • Ruffin M.P.
        • Lassel T.
        • Nabb S.
        • Messaoudi M.
        • Vinoy S.
        • et al.
        The delivery rate of dietary carbohydrates affects cognitive performance in both rats and humans.
        Psychopharmacology (Berl). 2003; 166: 86-90
        • McCrimmon R.J.
        Update in the CNS response to hypoglycemia.
        J Clin Endocrinol Metab. 2012; 97: 1-8
        • Lina B.A.
        • Jonker D.
        • Kozianowski G.
        Isomaltulose (palatinose): a review of biological and toxicological studies.
        Food Chem Toxicol. 2002; 40: 1375-1381
        • Matthews D.R.
        • Holman R.R.
        • Bown E.
        • Steenson J.
        • Watson A.
        • Hughes S.
        • et al.
        Pen-sized digital 30-second blood glucose meter.
        Lancet. 1987; 1: 778-779
        • Verhaeghen P.
        • Salthouse T.A.
        Meta-analyses of age-cognition relations in adulthood: estimates of linear and nonlinear age effects and structural models.
        Psychol Bull. 1997; : 231-249
        • Young H.
        • Benton D.
        Caffeine can decrease subjective energy depending on the vehicle with which it is consumed and when it is measured.
        Psychopharmacology. 2013; 228: 243-254
        • Lezak M.D.
        Neuropsychological assessment.
        3rd ed. Oxford University Press, Oxford1995
        • Nabb S.
        • Benton D.
        The effect of the interaction between glucose tolerance and breakfasts varying in carbohydrate and fibre on mood and cognition.
        Nutr Neurosci. 2006; : 3-4
        • Jensen A.R.
        Individual differences in the Hick paradigm.
        in: Vernon P.A. Speed of information-processing and intelligence. Ablex, New Jersey1987: 101-175
        • McNair D.M.
        • Lorr M.
        • Droppleman L.F.
        Revised manual for the profile of mood states.
        Educational and Industrial Testing Services, San Diego, CA1992
        • Cherbuin N.
        • Anstey K.
        Could ignoring higher blood sugar levels in the normal range in nondiabetics compromise cerebral health?.
        Future Neurol. 2013; 8: 5-7
        • Mortby M.
        • Janke A.
        • Anstey K.
        • Sachdev P.
        • Cherbuin N.
        High “Normal” blood glucose is associated with decreased brain volume and cognitive performance in the 60s: the PATH through Life Study.
        PLoS One. 2013; 8 (e73697): 9
        • Biessels G.J.
        • Deary I.J.
        • Ryan C.M.
        Cognition and diabetes: a lifespan perspective.
        Lancet Neurol. 2008; 7: 184-190
        • Mahoney C.R.
        • Taylor H.A.
        • Kanarek R.B.
        • Samuel P.
        Effect of breakfast composition on cognitive processes in elementary school children.
        Physiol Behav. 2005; 85: 635-645
        • Lamport D.J.
        • Dye L.
        • Mansfield M.W.
        • Lawton C.L.
        Acute glycaemic load breakfast manipulations do not attenuate cognitive impairments in adults with type 2 diabetes.
        Clin Nutr. 2012; 4: 265-272
        • Craft S.
        • Murphy C.G.
        • Wemstrom J.
        Glucose effects on complex memory and nonmemory tasks: the influence of age, sex, and glucoregulatory response.
        Psychobiology. 1994; 22: 95-105
        • Convit A.
        Links between cognitive impairment in insulin resistance: an explanatory model.
        Neurobiol Aging. 2005; 26: 31-35
        • Ghasemi R.
        • Haeri A.
        • Dargahi L.
        • Mohamed Z.
        • Ahmadiani A.
        Insulin in the brain: sources, localization and functions.
        Mol Neurobiol. 2013; 47: 145-171
        • Young H.
        • Benton D.
        The nature of the control of blood glucose in those with poorer glucose tolerance influences mood and cognition.
        Metab Brain Dis. 2014; ([in press])https://doi.org/10.1007/s11011-014-9519-2
        • Cryer P.E.
        Mechanisms of hypoglycemia-associated autonomic failure and its component syndromes in diabetes.
        Diabetes. 2005; 54: 3592-3601
        • Fruehwald-Schultes B.
        • Born J.
        • Kern W.
        • Peters A.
        • Fehm H.L.
        Adaptation of cognitive function to hypoglycemia in healthy men.
        Diabetes Care. 2000; 23: 1059-1066
        • McNay E.C.
        • Williamson A.
        • McCrimmon R.J.
        • Sherwin R.S.
        Cognitive and neural hippocampal effects of long-term moderate recurrent hypoglycemia.
        Diabetes. 2006; 55: 1088-1095
        • Boyle P.J.
        • Nagy R.J.
        • O'Connor A.M.
        • Kempers S.F.
        • Yeo R.A.
        • Qualls C.
        Adaptation in brain glucose uptake following recurrent hypoglycemia.
        Proc Natl Acad Sci U S A. 1994; 91: 9352-9356