Epidemiology/Health Services/Psychosocial Research ORIGINAL ARTICLE

Relationships Between and Cognitive Performance Among Adults With Type 1 and Type 2

DANIEL J. COX, PHD ANTHONY MCCALL, MD, PHD test cognitive functioning at 14.5 and 16 BORIS P. KOVATCHEV, PHD KEVIN J. GRIMM, MA mmol/l in adults with type 2 diabetes. LINDA A. GONDER-FREDERICK, PHD WILLIAM L. CLARKE, MD During hyperglycemia, significant dis- KENT H. SUMMERS, PHD ruptions occurred in the performance of complex tests of cognitive functioning, such as four-choice reaction time. How- ever, other investigators (5,6) were un- OBJECTIVE — Hyperglycemia is a common event among patients with type 1 and type 2 able to detect decay in cognitive-motor diabetes. While the cognitive-motor slowing associated with is well documented, the acute effects of hyperglycemia have not been studied extensively, despite patients’ reports of performance on selected neuropsycho- negative effects. This study prospectively and objectively assessed the effects of hyperglycemia on logical tests during hyperglycemia. cognitive-motor functioning in subjects’ natural environment. Contrary to hypoglycemia and its associated neuroglycopenia, a major RESEARCH DESIGN AND METHODS — Study 1 investigated 105 adults with type 1 barrier to the investigation of the effects of diabetes (mean age 37 years and mean duration of diabetes 20 years), study 2 investigated 36 hyperglycemia on cognitive-motor per- adults with type 2 diabetes (mean age 50 years and mean duration of diabetes 10 years), and formance is the absence of a clear physi- study 3 investigated 91 adults with (mean age 39 years and mean duration of ological mechanism that explains how diabetes 20 years). Subjects used a hand-held computer for 70 trials over 4 weeks, which hyperglycemia negatively influences required them to complete various cognitive-motor tasks and then measure and enter their current blood reading. functioning. However, research suggests several possible mechanisms (6). RESULTS — Hyperglycemia (blood glucose Ͼ15 mmol/l) was associated with slowing of all Blood-brain barrier microvascular dys- cognitive performance tests (P Ͻ 0.02) and an increased number of mental subtraction errors for function may occur as a result of transient both type 1 and type 2 diabetic subjects. The effects of hyperglycemia were highly individualized, hyperglycemia (7). Altered synthesis or impacting ϳ50% of the subjects. reuptake of monoamine neurotransmit- ters as a result of altered precursor avail- CONCLUSIONS — Acute hyperglycemia is not a benign event for many individuals with ability to the brain or changes in diabetes, but it is associated with mild cognitive dysfunction. availability to the brain are other possible Diabetes Care 28:71–77, 2005 explanations (8,9). Complex effects on peptide neurotransmitters may be pro- duced by uncontrolled diabetes (10,11). Any one of these mechanisms alone may atients with diabetes often report a blood glucose level of 16.7 mmol/l but be insufficient, and several of these mech- acute and transient cognitive dis- were unable to replicate this subsequently anisms could be additive. At this point, no ruptions associated with hypergly- using an auditory reaction-time task (2). P conclusion can be drawn about any spe- cemia. The impact of such effects could Davis et al. (3) reported that blood glu- influence quality of life and daily func- cose in the 20- to 30-mmol/l range was cific mechanism(s) responsible for possi- tioning, as well as indicate cues to aid pa- associated with a 9.5% reduction in type 1 ble short-term cognitive dysfunction. tients in better recognizing the presence diabetic children’s performance IQ. Per- However, hyperglycemia’s disruptive ef- of hyperglycemia. Holmes et al. (1) re- formance IQ was worsened in 67% of the fects on cognitive-motor functioning ported significant slowing of visual reac- children studied. Summerfield et al. (4) must first be verified, and then exploring tion time during a hospital clamp study at used a hyperinsulinic glucose clamp to possible physiological mechanisms will be justified. ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● The specific hypotheses tested in this From the University of Virginia Health System, Charlottesville, Virginia. study are 1) hyperglycemia is associated Address correspondence and reprint requests to Daniel J. Cox, Center for Behavioral Medicine Research, with cognitive-motor dysfunctions, 2) hy- Box 800-223, University of Virginia Health System, Charlottesville, VA, 22908. E-mail: [email protected]. perglycemia disrupts cognitive-motor Received for publication 8 March 2004 and accepted in revised form 21 September 2004. functioning in adults with either type 1 or K.H.S. is currently affiliated with Purdue University School of Pharmacy, West Lafayette, Indiana. Abbreviations: HHC, hand-held computer; PSAT, Paced Serial Addition Test; SMBG, self-monitoring of type 2 diabetes, and 3) cognitive-motor blood glucose. disruptions associated with hyperglyce- A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion mia are individualized. The last hypothe- factors for many substances. sis is based on our previous findings that © 2005 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby symptoms, as well as cognitive motor dis- marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ruptions, associated with hypoglycemia

DIABETES CARE, VOLUME 28, NUMBER 1, JANUARY 2005 71 Hyperglycemia cognitive-motor dysregulation

Table 1—Demographic variables for the three study groups

Study 1 Study 2 Study 3 Age (years) 37.5 Ϯ 0.9 (23–59) 50 Ϯ 11 (28–75) 39.4 Ϯ 10.4 (25–61) Disease duration (years) 19.7 Ϯ 9.9 (2–46) 10 Ϯ 9 (1–35) 20.2 Ϯ 10.7 (1–52) Mean blood glucose (mmol/l) 9.6 Ϯ 1.8 (4.7–16.9) 9.7 Ϯ 3.7 (5.9–23.9) 9.3 Ϯ 5.1 (1.1–33.3) Ϯ Ϯ Ϯ HbA1c (%) 8.6 2.3 (6.3–16.9)* 6.9 1 (7–14) 7.6 1.2 (4.0–12.6) BMI (kg/m2) NA 34 Ϯ 10 (19–65) 25.4 Ϯ 4.4 (17.7–41.2) % female 62 42 57 % using insulin 100 58 100 Ϯ Data are means SD (range) unless otherwise noted. *HbA1c estimate based on HbA1 assay. are individualized (12–14) in terms of came to mind. Then, the subject was pre- Study 2 subjects glycemic threshold for occurrence and in- sented with 10 mental subtraction prob- Thirty-four adults with type 2 diabetes dividual vulnerability. lems that involved subtracting a were recruited through newspaper and randomly generated single-digit number television announcements in Central Vir- RESEARCH DESIGN AND from a three-digit number with answers ginia. The subject group consisted of 15 METHODS entered on the HHC number pad. Finally, women and 19 men, with a mean (Ϯ SD) 15 four-choice reaction-time trials oc- age of 50 Ϯ 11 years and a mean diabetes Study 1 subjects curred in which the number 1, 3, 7, or 9 duration of 10 Ϯ 9 years (Table 1). A total of 105 adults with type 1 diabetes was randomly presented on the screen Procedures were identical to those of were recruited through three different and the subject pressed the correspond- study 1, except for the following differ- ϭ sites: the University of Virginia (n 33), ing number on the keypad as quickly as ences. The Handspring Visor Platinum ϭ Vanderbilt University (n 39), and the possible. The dependent variables were served as the HHC. Because the reaction- ϭ Joslin Diabetes Center (n 33). Because the number of “A” words recalled, time time test did not have a significant hyper- subjects were recruited to participate in to complete 10 mental subtractions, as glycemic group effect in study 1, in study the Blood Glucose Awareness Training well as the number of errors, and time to 2 it was replaced with two levels of the study (15), and not for the expressed pur- complete the 15 reaction-time trials. Paced Serial Addition Test (PSAT). The pose of investigating the acute effects of Next, the HHC prompted the subject PSAT presented a sequence of single-digit hyperglycemia, subjects and research as- to measure and enter his/her blood glu- numbers, and the subject entered the sum sistants collecting the data were blind to cose reading from the One Touch meter of each pair of sequential numbers. Levels the hypotheses. The subject group con- after completing the cognitive-motor 1 and 2 presented the numbers at 4- and sisted of 65 women and 40 men with a tests. To assess reliability, the 1-month 2-s intervals, respectively. The dependent mean (Ϯ SD) age of 37.5 Ϯ 9.0 years, a variables were the number of correct ad- HHC data collection was repeated 5 diabetes duration of 19.7 Ϯ 9.9 years, and ditions on the faster and slower PSAT. months later, generating an average of an HbA1 of 10.4 Ϯ 2.3% (equivalent to an 82.5 valid HHC trials per subject. HbA of 8.6%) (Table 1). Study 3 subjects 1c Three precautions were taken to en- In groups of four to eight subjects, the Ninety-one adults with type 1 diabetes study was described, informed consent courage and monitor valid data entry, i.e., were recruited through newspaper and was secured, and subjects were taught to cognitive-motor testing before SMBG and television announcements in Central use and demonstrated competency in the thus blind to actual blood glucose level. Virginia to participate in a study investi- use of One Touch memory meters (Lifes- First, the HHC prompted subjects with gating bio-behavioral precursors to hypo- can, Milpitas, CA) and the Psion 250 the message “No blood sample yet” as glycemia. The subject group consisted of hand-held computer (HHC). During the soon as it was turned on. Second, the 52 women and 39 men, with a mean (Ϯ subsequent month, subjects were in- HHC tracked elapsed time between the SD) age of 39.4 Ϯ 10.4 years and a mean structed to use their HHC immediately prompt “Measure your blood glucose” diabetes duration of 20.2 Ϯ 10.7 years before routine self-monitoring of blood and the entry of that SMBG reading. Since (Table 1). glucose (SMBG), three to four times each at least 45 s were required for a subject to Procedures were identical to those in day, for a total of 50 trials. On each HHC lance a finger, collect a blood sample, and study 2, except only the subtraction test trial, data were collected on cognitive- analyze blood glucose level with the One was used in order to reduce subject bur- motor performance and then blood glu- Touch meter, any readings entered in den. cose level. First, the HHC presented and Ͻ45 s were considered invalid. Third, the Data analysis. The data were analyzed at recorded the subject’s performance on blood glucose readings entered into the two levels: across subject at a group level three cognitive-motor tests. In a psy- HHC were compared with memory meter and within subject for individual subjects. chomotor task, the subject was instructed data to ensure accuracy of the SMBG re- To evaluate group effects, blood glucose to think of as many words as possible that sults. We have used these procedures ex- readings were categorized into five differ- start with the letter “A” in 30 s and to tap tensively in our investigations of ent ranges that contained a similar num- the “enter” key each time an “A” word hypoglycemia (12,16,17). ber of readings across all subjects (Table

72 DIABETES CARE, VOLUME 28, NUMBER 1, JANUARY 2005 Cox and Associates

Table 2—BG categories used in group data analysis pared between the two datasets collected at months 1 and 6. The test-retest corre- ϭ BG events/category lation of each of these variables was r Blood glucose Blood glucose range 0.60–0.63 (P Ͻ 0.001), and there were category (mmol/l) Study 1 Study 2 Study 3 no significant differences in test-retest Ͻ means. Similarly, test-retest HbA1s were 1 6.1 8 1,456 599 968 ϭ Ͻ 28Ͻ 10 1,017 472 865 highly correlated (r 0.85, P 0.001). 310Ͻ 12.2 1,237 322 803 The test-retest performance of cognitive- Ͻ motor tests was highly correlated, (r ϭ 4 12.2 15 1,178 221 777 Ͻ 5 Ͼ15 1,238 284 864 0.80–0.96, P 0.001 in each case). Therefore, the data were considered reli- able over time and collapsed across the 2). This allowed assessment of cognitive- individual test for cognitive dysfunction two time samples. motor performance in discrete blood glu- (see APPENDIX). For each subject and for cose ranges at a group level. Subjects’ each test, we first computed the perfor- Group effects of hyperglycemia responses were analyzed using ANOVA. mance mean score and SD during eugly- In study 1, hyperglycemia (blood glucose Since each subject contributed several cemia (blood glucose 6–8 mmol/l) and Ͼ15 mmol/l) was associated with slower entries in each blood glucose range, the then computed its deviation for each hy- performance on the psychomotor task, as assumption of independence of the obser- perglycemic trial (blood glucose Ͼ15 reflected by fewer A words retrieved in vations may be violated, resulting in an mmol/l) as the number of SDs from eugly- 30-s intervals (F ϭ 5.08, P ϭ 0.0004) artificially higher number of degrees of cemic performance. In other words, the (Fig. 1). Similarly, hyperglycemia slowed freedom in the F tests. To avoid potential deviation of each test at hyperglycemia mental subtraction speed (F ϭ 3.71, P ϭ exaggeration of the group effects and to was presented as a Z score using norma- 0.005) and increased subtraction errors account for parallel tests, the significance tive euglycemic data. For each subject a (F ϭ 2.72, P ϭ 0.02). However, hyper- level was lowered to P Ͻ 0.01. cognitive test was considered significantly glycemia did not significantly slow choice Each subject’s individual data were disrupted during hyperglycemia if these Z reaction time (F ϭ 2.08, P ϭ 0.09). analyzed to determine how many of the scores were significantly higher than zero In study 2, hyperglycemia (blood glu- cognitive-motor tasks were significantly at P Ͻ 0.05. The statistical reasoning be- cose Ͼ15 mmol/l) was associated with impaired with elevated blood . hind this test is presented in the APPENDIX. slower performance on the psychomotor To evaluate an individual subject’s perfor- task, as reflected by fewer A words re- mance/blood glucose relationships, RESULTS trieved in 30-s intervals (F ϭ 6.38, P ϭ ANOVAs were inappropriate because of 0.001). Similarly, hyperglycemia slowed too few samples in each blood glucose Test-retest reliability mental subtraction speed (F ϭ 4.82, P Ͻ category and correlation analysis was in- In study 1 the average blood glucose, the 0.001) and increased subtraction errors appropriate because of its assumption of percentage of readings Ͼ8.9 mmol/l and (F ϭ 3.62, P Ͻ 0.01). Performance on linearity. Consequently, we designed an Ͼ11.1 mmol/l, respectively, were com- PSAT-1 (F ϭ 3.74, P ϭ 0.02) and PSAT-2

Figure 1—Study 1 mean Ϯ SEM error bars for performance variables for different blood glucose categories (mmol/l) and ANOVA P levels for type 1 diabetic subjects.

DIABETES CARE, VOLUME 28, NUMBER 1, JANUARY 2005 73 Hyperglycemia cognitive-motor dysregulation

Figure 2—Study 2 mean Ϯ SEM error bars for performance variables for different blood glucose categories (mmol/l) and ANOVA P levels for type 2 diabetic subjects.

(F ϭ 4.83, P Ͻ 0.001) was disrupted, ease, sex, glycosylated hemoglobin, per- mia was only mildly related to number of with more addition errors (Fig. 2). centage of SMBG readings Ͼ15 mmol/l, cognitive tests impaired during hypergly- In study 3, hyperglycemia (blood glu- and mean cognitive-motor performance cemia, as reflected by both higher HbA1c cose Ͼ15) was associated with slower during euglycemia and , as de- readings and greater percentage of SMBG mental subtraction time (F ϭ 7.73, P Ͻ fined by the Beck Depression Inventory. readings Ͼ15 mmol/l. Similarly, there 0.0001) and increased number of errors Greater routine exposure to hyperglyce- was a small relationship indicating that (F ϭ 3.87, P ϭ 0.003).

Individual effects of hyperglycemia Table 3 —Individual effects and correlates of cognitive disruptions during hyperglycemia In study 1 (Table 3) the percentage of type 1 diabetic subjects who demonstrated sig- Study 1 Study 2 Study 3 nificant (P Ͻ 0.01) disruptions in 0, 1, 2, and 3 of any of the cognitive-motor tasks Mean no. of SMBGs Ͼ15 mmol/l 17.8 19.4 17.9 with hyperglycemia (blood glucose Ͼ15) Number of tests significantly impaired: was 45, 24, 16, and 11%, respectively. % subjects with impaired performance Concerning specific tests and mental sub- 0 45% 46% 41% traction time and errors, psychomotor 1 24% 33% 48% task and choice reaction time were signif- 2 16% 8% 12% icantly affected for 21, 20, 20, and 28%, 3 11% 13% respectively, of the individual subjects % subjects with specific tests significantly (Table 3). The standardized effect size affected by hyperglycemia (APPENDIX) for those type 1 diabetes sub- Subtractions, speed 21% 38% 19% jects significantly affected was 3.2, 2.7, Subtractions, errors 20% 21% 52% 2.4, and 3.0, respectively, for subtraction A words 20% 21% time, subtraction errors, A words, and re- Reaction time 28% action time. PSAT 1 4% A series of exploratory correlations PSAT 2 4% and t tests were conducted between po- Relationship with no. of cognitive tests tential predictor variables and the num- impaired ber of cognitive-motor dysfunctions per Duration r ϭϪ0.36* ϭ ϭϪ ϭ subject to investigate possible factors con- HbA1c r 0.20† r 0.30* r 0.34‡ tributing to individual differences in sen- % SMBGs Ͼ15 mmol/l r ϭ 0.20* r ϭϪ0.29† r ϭ 0.39‡ sitivity to hyperglycemia (Table 3). The Mean euglycemic performance r ϭϪ0.18† r ϭϪ0.38* r ϭϪ0.19* predictor variables were duration of dis- *P Ͻ 0.05, †P Ͻ 0.1, ‡P Ͻ 0.01.

74 DIABETES CARE, VOLUME 28, NUMBER 1, JANUARY 2005 Cox and Associates worse performance during euglycemia CONCLUSIONS — Consistent with nitive-motor functioning occurs. This was associated with greater decay during some other investigations using labora- range would appear to be between 4 and hyperglycemia. tory methodologies (1,3,4), this field 15 mmol/l. This might be considered an The percentage of type 2 diabetic sub- study assessing both type 1 and type 2 intuitive finding, since other physiologi- jects in study 2 (Table 3) who had zero to diabetic subjects’ cognitive-motor func- cal parameters, such as blood pressure, four cognitive-motor disruptions signifi- tioning during daily routines found sig- body temperature, water concentration, cantly associated with hyperglycemia was nificant cognitive dysfunction during and body weight all have a range within 46, 33, 8, and 13%, respectively. Mental hyperglycemia among some subjects. In- which the body functions optimally. subtraction time and errors, psychomotor spection of the figures suggests that this is These findings have potential clinical task, PSAT-1, and PSAT-2 were signifi- not a linear relationship but that there meaning for individuals living with diabe- cantly slowed for 38, 21, 21, 4, and 4% of may be a threshold around 15 mmol/l, tes. In this study, hyperglycemia resulted the subjects, respectively (Table 3). The when cognitive-motor function begins to in increased errors and slower responses standardized effect size for those type 1 be affected. The generalizability of these when performing basic verbal and math- diabetic subjects significantly affected findings is enhanced by the use of three ematical tasks, which are important in nu- was 2.9, 2.3, 2.7, 1.6, and 1.8, respec- different subject samples with both type 1 merous daily functions, such as balancing tively, for subtraction time, subtraction and type 2 diabetes, multiple cognitive checkbooks, calculating insulin dosing, errors, A words, PSAT-1, and PSAT-2. tasks assessing both verbal and mathe- and school and work performance. Fur- A similar series of exploratory corre- matical skills, and repeated samples per ther research is needed to determine the lations and t tests were conducted as in subject, with subjects from multiple cen- impact of hyperglycemia-related cogni- study 1 (with the addition of BMI) to ters and blind to the hypotheses being tive disruption on the daily lives and func- identify factors contributing to individual tested. An interesting observation is that tioning of individuals with type 1 and differences in sensitivity to hyperglyce- while hypoglycemia has been consistently type 2 diabetes. Also implicit is that ex- mia. Subjects with a shorter duration of associated with initial cognitive slowing, cessive carbohydrate loading before ex- disease, less exposure to hyperglycemia, but not typically an increase in errors ams or other cognitive-sensitive tasks and worse performance during euglyce- (1,2), hyperglycemia appears to be re- intended to avoid the negative conse- mia were more likely to have more cogni- lated to either cognitive slowing or in- quences of hypoglycemia may in fact be creased errors (of the affected subjects, counter productive, if such actions lead to tive tests impaired during hyperglycemia only ϳ25% demonstrated both a signifi- hyperglycemia. Instead, optimal cogni- (Table 3). cant slowing and errors). This was con- tive functioning would be anticipated The percentage of type 1 diabetic sub- firmed in the group data analyses of all with optimal blood glucose control. The jects in study 3 (Table 3) who had zero to three studies, where subtraction errors in- immediate negative consequences of hy- two disruptions significantly associated creased when blood glucose exceeded poglycemia inherently encourage some with hyperglycemia was 41, 48, and 12%, 15 mmol/l. individuals to keep their blood glucose respectively. Subtraction speed and/or er- While the group results demonstrate Ͼ4 mmol/l. The possibility that cognitive rors were significantly affected for 19 and that hyperglycemia can have a negative dysfunction may occur when blood glu- 52% of the subjects (Table 3). The stan- impact on cognitive performance, these coses exceed 15 mmol/l could be similarly dardized effect size for those type 1 dia- effects were highly individualized. Ap- motivating for some patients to avoid betic subjects significantly affected was proximately 55% of the subjects across all hyperglycemia and to achieve tighter 3.8 and 2.3, respectively, for subtraction three studies demonstrated such effects, blood glucose control. Further, if pa- time and subtraction errors. similar to the 67% reported by Davis et al. tients could be trained to both recog- A comparable series of exploratory (3). Exploratory analyses of individual nize and accurately interpret these dis- correlations and t tests were conducted as differences for both type 1 diabetes stud- ruptions in cognitive performance, they in studies 1 and 2 to determine what fac- ies indicated a mild relationship between may better recognize hyperglycemia tors might predict individual sensitivity to greater routine exposures to hyperglyce- and more quickly take action to treat it hyperglycemia. Similar to study 1, more mia, as reflected by the percentage of (15). routine exposure to hyperglycemia and Ͼ SMBG readings 15 mmol/l, HbA1c, and These findings are inconsistent with poorer performance during euglycemia more cognitive tests affected by hypergly- some laboratory studies (2,5,6,20) that was associated with greater effects of cemia. This is consistent with the results did not find effects of hyperglycemia on hyperglycemia on cognitive functioning of studies demonstrating that greater ex- cognitive functioning. The differences (Table 3). posure to hypoglycemia is associated with may be attributed to differences in meth- An overall logistic regression was pre- greater cognitive impairments during hy- odologies. The negative laboratory stud- formed combining subjects from all three poglycemia (16) and is inconsistent with ies performed one neuropsychological studies to predict those who did and did adaptive theories speculating that indi- test during a single artificially induced not demonstrate cognitive sensitivity to viduals may accommodate to extreme episode of hyperglycemia and typically hyperglycemia. This correctly classified blood glucose levels. with a small subject sample. The current 66% of the cases (␹2 ϭ 16.83, P ϭ 0.002) These findings, along with a consis- studies had larger subject samples and incorporating only two predictor vari- tent literature demonstrating cognitive multiple testing per subject during hyper- ables: percentage of blood glucoses Ͼ15 deficits with hypoglycemia (3,14,18,19), glycemia within naturalistic conditions. mmol/l and baseline performance of sub- suggest that there is a homeostatic neuro- Since only approximately half of the sub- traction errors glycemic range within which optimal cog- jects are anticipated to demonstrate cog-

DIABETES CARE, VOLUME 28, NUMBER 1, JANUARY 2005 75 Hyperglycemia cognitive-motor dysregulation nitive affect during hyperglycemia, future 2) For each subject, each cognitive rier. Brain Res Brain Res Rev 23:210–218, studies might benefit from large sample test, and each hyperglycemic reading 1997 sizes, with multiple testing during hyper- (e.g., when blood glucose is Ͼ15 mmol/ 8. Wurtman RJ: Effects of their nutrient glycemia and a selected subject sample of l), the Z score of that cognitive test is com- precursors on the synthesis and release of individuals who perform poorly during puted as the number of SDs away from the serotonin, the catecholamines, and acetyl- euglycemia. test’s mean euglycemic performance. choline: implications for behavioral disor- A limitation of these data are that, al- 3) For each subject and each cogni- ders. Clin Neuropharmacol 11 (Suppl. 1): S187–S193, 1988 though they were replicated across tasks, tive test (CT) the average hyperglycemic Z 9. Figlewicz DP, Brot MD, McCall AL, Szot diseases, and sites, they still represent ob- score (ZCT) is computed together with the P: Diabetes causes differential changes servational findings and not experimental number of hyperglycemic readings (nh). in CNS noradrenergic and dopaminergic manipulations. These findings also do not 4) Criterion for significance: the CT in the rat: a molecular study. help to understand the physiological is considered significant for hypergly- Brain Res 736:54–60, 1996 ␨ ϭ 1/2 mechanisms underlying such hypergly- cemia if the product CT nh ZCT 10. Sipols AJ, Baskin DG, Schwartz MW: cemic effects on brain function. However, is Ͼ1.28. Effect of intracerebroventricular insulin some laboratory manipulations of hyper- Statistical background of the test. Un- infusion on diabetic hyperphagia and glycemia have confirmed these findings der the null hypothesis that a cognitive hypothalamic neuropeptide gene expres- (1,4). As detailed in the introduction, test is not elevated during hyperglycemia, sion. Diabetes 44:147–151, 1995 there are several proposed underlying 11. Havel PJ, Hahn TM, Sindelar DK, Baskin the Z score ZCT would have a central nor- physiological mechanisms, but these will mal distribution (with a mean of 0 and SD DG, Dallman MF, Weigle DS, Schwartz require laboratory studies to determine MW: Effects of streptozotocin-induced of 1). Thus, the average Z score of nh ob- which are responsible for the observed diabetes and insulin treatment on the servations would have a normal distribu- hypothalamic melanocortin system and cognitive disruptions during hyperglyce- 1/2 tion of 0 and an SD of 1/(nh ). It follows muscle uncoupling protein 3 expression mia. In addition, since these studies only ␨ ϭ 1/2 that CT nh ZCT would have a nor- in rats. Diabetes 49:244–252, 2000 assessed adults with diabetes, further nat- mal distribution with a mean of 0 and an 12. Cox DJ, Gonder-Frederick LA, Antoun B, uralistic studies are needed to determine ␨ Ͼ SD of 1. Therefore, if CT is 1.28 (central Cryer PE, Clarke WL: Perceived symp- whether hyperglycemia similarly affects normal distribution quintile correspond- toms in the recognition of hypoglycemia routine cognitive-motor performance in ing to a probability of 0.9), the null hy- [see comments]. Diabetes Care 16:519– pediatric populations. pothesis must be rejected at a significance 527, 1993 level of 0.1. 13. Davis EA, Keating B, Byrne GC, Russell M, Jones TW: Hypoglycemia: incidence and Acknowledgements— This research was clinical predictors in a large population- supported by National Institutes of Health References based sample of children and adoles- Grants R01 DK28288 and R01 DK 51562 and 1. Holmes CS, Hayford JT, Gonzalez JL, cents with IDDM. Diabetes Care 20:22–25, grants from Eli Lilly & Company (Indianapo- Weydert JA: A survey of cognitive func- 1997 lis, IN) and Lifescan (Milpitas, CA). tioning at difference glucose levels in dia- 14. Gonder-Frederick LA, Cox DJ, Driesen betic persons. Diabetes Care 6:180–185, NR, Ryan CM, Clarke WL: Individual dif- 1983 Appendix ferences in neurobehavioral disruption 2. Holmes CS: Metabolic control and audi- during mild and moderate hypoglycemia tory information processing at altered glu- in adults with IDDM. Diabetes 43:1407– Individual test for cognitive cose levels in insulin-dependent diabetes. 1412, 1994 dysfunction Brain Cogn 6:161–174, 1987 15. Cox DJ, Gonder-Frederick L, Polonsky We propose the following procedure as a 3. Davis EA, Soong SA, Byrne GC, Jones W, Schlundt D, Kovatchev B, Clarke W: within-subject test for significance of a TW: Acute hyperglycaemia impairs cog- Blood glucose awareness training (BGAT- cognitive-motor dysfunction (or symp- nitive function in children with IDDM. 2): long-term benefits. Diabetes Care 24: J Pediatr Endocrinol Metab 9:455–461, toms) with respect to a certain glycemic 637–642, 2001 1996 condition such as hypoglycemia or hyper- 16. Cox DJ, Gonder-Frederick LA, Kovatchev 4. Sommerfield AJ, Deary IJ, Frier BM: Acute BP, Young-Hyman DL, Donner TW, Ju- glycemia. The test is based on a set of cog- hyperglycaemia alters mood state and nitive-motor performance scores taken impairs cognitive performance in people lian DM, Clarke WL: Biopsychobehav- during that condition and a set of cogni- with type 2 diabetes (Abstract). Diabet ioral model of severe hypoglycemia. II. tive-motor performance scores during a Med 20:31, 2003 Understanding the risk of severe hypo- control condition, such as euglycemia. In 5. Draelos MT, Jacobson AM, Weinger K, glycemia. Diabetes Care 22:2018–2025, the example below we will elaborate the Widom B, Ryan CM, Finkelstein DM, 1999 procedure for screening of hyperglycemic Simonson DC: Cognitive function in 17. Kovatchev BP, Cox DJ, Gonder-Frederick LA, Schlundt D, Clarke WL: Stochastic cognitive-motor performance scores with patients with insulin-dependent diabetes mellitus during hyperglycemia and hypo- model of self-regulation decision making a control performance during euglycemia exemplified by decisions concerning hy- defined as blood glucose between 5 and glycemia. Am J Med 98:135–144, 1995 6. McCall AL, Figlewicz DP: How does dia- poglycemia. Health Psychol 17:277–284, 8.3 mmol/l (90–150 mg/dl). betes mellitus produce brain dysfunction? 1998 1) For each subject and each cogni- Diabetes Spectrum 10:25–32, 1997 18. Frier BM: Hypoglycaemia and cogni- tive test, mean and SD are computed from 7. Mooradian A: tive function in diabetes. Int J Clin Pract the performance scores during euglyce- complications of diabetes mellitus: a (Suppl.):30–37, 2001 mia. perspective from the blood-brain bar- 19. Ryan CM, Geckle M: Why is learning and

76 DIABETES CARE, VOLUME 28, NUMBER 1, JANUARY 2005 Cox and Associates

memory dysfunction in type 2 diabetes 20. Gschwend S, Ryan C, Atchison J, Ars- counterregulatory in adoles- limited to older adults? Diabetes Metab Res lanian S, Becker D: Effects of acute hy- cents with insulin-dependent diabetes Rev 16:308–315, 2000 perglycemia on mental efficiency and mellitus. J Pediatr 126:178–184, 1995

DIABETES CARE, VOLUME 28, NUMBER 1, JANUARY 2005 77