Introduction
A diminished ability to perceive the onset of hypoglycaemia occurs in 17–25% of people with type 1 diabetes [
1,
2]. Impaired awareness of hypoglycaemia (IAH) is a major risk factor for severe hypoglycaemia (SH), defined as an event requiring external assistance, and increases the risk of SH sixfold [
1,
3]. Cognitive decline may be a complication of longstanding type 1 diabetes [
4], and several cognitive domains seem to be affected [
4‐
7]. Cognitive dysfunction may therefore contribute to suboptimal diabetes management, including the avoidance and treatment of hypoglycaemia. In support of this hypothesis, some adults with IAH do not modify their behaviour to prevent or avoid hypoglycaemia [
3], and fail to adhere to recommended therapeutic measures [
8].
Recurrent exposure to hypoglycaemia is strongly implicated in the pathogenesis of IAH [
9]. A putative association between IAH and impaired cognitive function may therefore exist, since both could be the consequence of recurrent SH, the frequency of which is promoted by IAH [
1,
3]. Alternatively, for people with type 1 diabetes who have premorbid cognitive dysfunction, self-management may be suboptimal, thereby increasing the risk of IAH. Furthermore, the development of IAH and impaired cognitive function may have a common predisposing factor. If IAH is associated with premorbid cognitive dysfunction, impairment should involve several cognitive domains. Alternatively, if cognitive impairment in people with IAH is caused by recurrent SH, then we would expect cerebral functions dependent on brain regions that are vulnerable to hypoglycaemia to be diminished.
A causal association between recurrent SH and cognitive impairment in adults with type 1 diabetes is unproven. Anecdotal reports have described memory loss following SH [
10‐
12], and cross-sectional studies have demonstrated impairment of several cognitive domains in adults with a history of SH [
13‐
16]. However, the Epidemiology of Diabetes Interventions and Complications (EDIC) study (the follow-up to the DCCT) and a smaller Swedish prospective study both found that recurrent SH had little or no adverse effect on cognition in adults with type 1 diabetes [
17,
18], a conclusion supported by a meta-analysis [
5]. A more recent meta-analysis concluded that reduced memory and executive function are associated with SH [
7], which people with IAH experience at a much higher frequency than was recorded in the DCCT/EDIC study. SH may cause localised neuronal death within the hippocampus and cerebral cortex, and in white matter, as demonstrated histologically and in vivo with MRI after SH in animals and humans [
11,
19‐
21]. It is therefore plausible that recurrent SH could compromise cognitive functions that are dependent on brain regions particularly sensitive to neuroglycopenia.
Three previous studies in the early 1990s found a possible association between IAH and cognitive impairments, including memory impairment, selective attention and a trend towards reduced intelligence quotient. These investigators hypothesised that the impairments resulted from frequent exposure to SH, as experienced by people with IAH [
22‐
24]. However, putative associations between IAH, recurrent hypoglycaemia and cognitive dysfunction have remained unresolved.
The aim of the present study was to compare cognitive function in people with type 1 diabetes who had established IAH, with those in whom hypoglycaemia awareness remained intact. For this purpose, tests of verbal memory, object-location memory, pattern separation, working memory, information processing speed and executive function, including planning, were applied. Optimal cognitive function depends on interaction within networks of brain regions. For learning, memory, and pattern separation abilities, the most central structure for normal functioning is the hippocampus [
25,
26], while executive functions, working memory and information processing speed depend on frontal and parietal cortices and their connectivity [
27,
28]. The intention was to test cognitive abilities that depend on brain regions susceptible to damage during hypoglycaemia [
11,
21] and cognitive abilities that are recognised to be impaired in patients with type 1 diabetes [
4,
5]. Finally, because many people with IAH do not modify their behaviour to avoid SH [
3], exemplified by some failing to measure their blood glucose in relation to driving [
29,
30], executive functions that include planning ability [
31] and pattern separation, which can affect a person’s ability to identify a hypoglycaemic episode, were assessed. A secondary aim was to assess whether cognitive function in participants with IAH is related to their historical SH burden.
Discussion
By employing an extensive cognitive test battery and validated methods to assess hypoglycaemia awareness in well-matched participants with type 1 diabetes, the present study demonstrated that adults with type 1 diabetes who have IAH have modestly impaired cognitive performance compared with people with NAH, thus adding further evidence to previous reports on this topic [
22‐
24].
The IAH group exhibited significant impairment in pattern separation abilities in comparison with the NAH group, as well as on supplementary analyses of planning function (the Tower test illegal moves). Pattern separation is critical for accurate memory: decreased pattern separation ability contributes to interference among memories and convergence of similar episodes into a generalised representation rather than distinct memories [
25]. It is possible that people with IAH have a diminished ability to distinguish cues that are specifically associated with hypoglycaemia and hence are unable to take appropriate action to avoid SH. Executive function measured with the Tower test in the present study assesses planning ability and, as such, a person’s capacity to adjust behaviour to current and future demands and goals [
31]. The present results suggest that planning ability may be restricted in people with IAH and might underlie the observation that many people with IAH do not modify their behaviour to prevent hypoglycaemia [
3] or adhere to prescribed therapy [
8].
In the IAH group, significant impairments were observed in the learning, memory and pattern separation tests, all of which rely on the integrity of the hippocampus, a brain structure vulnerable to neuroglycopenic injury [
11,
19,
20]. In people with type 1 diabetes, learning and memory seem to be largely unaffected [
5,
17], although two studies have shown memory impairment in people with recurrent SH [
24,
38]. In the present study, participants with IAH exhibited both learning difficulties and impaired delayed recall in the Verbal memory test. An IAH-specific learning deficit was also evident in the Objects in grid test, which is an object-location memory and one-trial learning test. Hence, the impairment in memory and learning in those with IAH was generalised, pertaining both to words heard and objects seen. The difference between the IAH and NAH groups in the Verbal memory test is similar to the difference observed after 7 years of ageing in a middle-aged non-diabetic population [
39]. The deficits observed in the present study are subtle and unlikely to be apparent to individuals in the performance of everyday tasks. However, the present findings suggest that adults with type 1 diabetes who have developed IAH may have a reduced cognitive reserve compared with those with NAH, which may render them more susceptible to experiencing subsequent cognitive decline and associated educational and occupational challenges.
These findings suggest that frequent exposure to SH, as experienced by people with IAH, may underlie the observed cognitive impairments. However, causation cannot be determined from cross-sectional data. The lack of correlation between the frequency of SH episodes and cognitive test results may indicate that the observed association between IAH and cognitive deficits did not result from exposure to SH; instead, it might be explained by inaccurate recall of SH episodes, since it is known that retrospective estimation of hypoglycaemia is vulnerable to recall bias [
40]. An association was found between the number of invalid tests and the number of SH episodes since diabetes onset in participants with IAH, which supports the hypothesis that recurrent SH may promote cognitive impairment.
The participants’ premorbid cognitive function was not assessed, and it is therefore not possible to establish whether (1) the cognitive impairment associated with IAH had resulted from recurrent exposure to SH, (2) premorbid cognitive impairment per se predisposed the individual to develop IAH, or (3) another common predisposing factor led to the simultaneous development of IAH and cognitive impairment. Since the IAH group did not exhibit impairment across all cognitive domains, but had significant impairments in tests of learning and memory that are associated with brain regions vulnerable to neuroglycopenia [
11,
19,
20], the present results support a role for recurrent SH in the pathogenesis of IAH [
9].
The awareness status of a few of the participants had changed between the cross-sectional study of 2011 and the present study. This is consistent with the dynamic nature of the IAH syndrome: awareness status may fluctuate and may even be restored by avoidance of hypoglycaemia [
41]. When excluding the five participants whose hypoglycaemia awareness status had changed, group differences became more evident, thus demonstrating that persistent IAH status was most negatively associated with cognitive deficiency. Participants with IAH tended to have experienced more SH overall compared with participants with NAH and recorded more asymptomatic hypoglycaemia during the month preceding the study, consistent with the recognised characteristics of the IAH syndrome.
The strengths of the present study include the application of two validated methods to determine hypoglycaemia awareness status [
42] and the use of an extensive battery of validated cognitive tests [
25,
31,
35]. In addition, the use of strict criteria for inclusion in the statistical analyses excluded participants with IAH with the greatest performance impairments: only participants with IAH with the best cognitive function were compared with participants with NAH. Thus, the observed group differences in cognition between participants with IAH and NAH probably represented the minimum difference. The similar demographic and disorder-specific characteristics in the IAH and NAH groups, as well as in those participants who declined participation, are further strengths of this study. In the Norwegian Diabetes Registry [
43], the average age, diabetes duration and HbA
1c level in people with type 1 diabetes was 41.8 years, 20.8 years and 8.0%, respectively, i.e. quite similar to the measures in the present study, which supports the generalisability of the present findings. Furthermore, the prevalence of microvascular complications and the level of educational attainment were similar in the IAH and NAH groups, and are therefore unlikely to have confounded the results.
The limitations of the study include the lack of measurement of participants’ premorbid cognitive function and the relatively modest sample size. While these may contribute to selection bias, there is no reason to believe that those eligible candidates who declined participation in the study had higher or lower cognitive abilities than people with type 1 diabetes in general. In addition, participants with NAH were chosen at random to reduce selection bias. Moreover, pre-test power analyses indicated that the proposed number of participants would be sufficient to yield clinically significant results.
It could be argued that participants should have been assessed using a continuous glucose monitoring system before commencing the study to identify asymptomatic biochemical hypoglycaemia that may influence cognitive function. Although cognition is impaired during hypoglycaemia and may remain abnormal for 40–75 min after hypoglycaemia has been treated [
34], people with IAH have been shown to be less affected by hypoglycaemia compared with people with NAH and to recover more quickly [
34]. As cognitive function is less affected by hypoglycaemia in people with IAH than those with NAH [
34], any unrecognised biochemical hypoglycaemia in participants before the study would have been more likely to result in poorer performance in those with NAH, and would therefore not explain the present findings. Hyperglycaemia has also been found to impair cognitive function [
44,
45], but an upper limit for the plasma glucose level was not specified before cognitive testing commenced. However, plasma glucose levels before and after testing were similar in the IAH and NAH groups (Table
1), with no association being found between elevated glucose levels and poorer cognitive performance.
As participants were not observed during cognitive testing, it is possible that they could have used aids when self-administering the cognitive tests, although they were instructed not to. The design of most tests made them impervious to attempts at cheating and the test platform did not allow individuals to redo tests. Based on the time stamps of keyboard strokes and the duration of each test session, it is very unlikely that any of the participants used aids while performing the tests.
The results of the present study are of considerable relevance to people with type 1 diabetes. The modest cognitive impairment observed in people with IAH may contribute to their increased risk of developing severe hypoglycaemia, and emphasises the necessity to reinforce structured education by using psychotherapeutic and behavioural therapies, and utilising diabetes technologies to avoid SH [
46]. It has been suggested that impaired cognition may underlie the resistance shown by some people with IAH to co-operate in interventions to restore awareness of hypoglycaemia [
47]. The present observations underline the value of including cognitive tests in intervention programmes to evaluate whether impaired cognitive ability may affect adherence to treatment and outcomes.
Acknowledgements
The authors thank the study nurses S. Salater and H. Bjøru, St Olavs Hospital, Trondheim University Hospital, Trondheim, Norway, for their excellent practical assistance. We are grateful to B. O. Åsvold, associate professor at the Department of Public Health and Nursing, Faculty of Medicine and Health Sciences, NTNU – Norwegian University of Science and Technology, Trondheim, Norway for performing the random selection of participants with NAH. We also thank C. Stark, University of California, Irvine, for supplying the images used in the Pattern separation task.
Part of the study was presented at the 51st EASD Annual Meeting, Stockholm, Sweden, 14–18 September 2015.