Glucose variability and mood in adults with diabetes: A systematic review

Abstract Aims To systematically review the literature regarding the association between glucose variability (GV) and mood in adults with diabetes, appraise the used methods and make suggestions for future research. Methods A systematic review of literature published up to May 2019 was performed. Abstracts and full texts were screened independently in duplicate. Experimental and observational studies reporting the association between GV and mood in adults with type 1 diabetes or type 2 diabetes were evaluated. A descriptive analysis of the extracted data was conducted, along with a quality assessment. Results Out of the 2.316 studies screened, eight studies met our criteria. Studies used a variety of measures and metrics to determine GV and mood. Four studies used continuous glucose monitoring (CGM). An association between GV and mood was found in four studies when correlating either postprandial glucose rate of increase with current mood or multiday GV with mood measured retrospectively. The other four studies did not find any association. Conclusions There is no clear empirical support for a link between GV and mood in adults with type 1 and type 2 diabetes. More rigorous research is warranted using CGM and ecological momentary assessment of mood to assess if and under what conditions an association between GV and mood exists.


| INTRODUC TI ON
The association between mood and glucose variability (GV) in persons with diabetes has been a topic of interest since the 1930s. 1 Stress and negative mood have been assumed to explain unpredictable and extreme blood glucose fluctuations often referred to as "brittle diabetes". 2 In the early 1980s, the attention shifted to the opposite direction, that is the effect of "diabetic instability" on psychological problems. 3 To date, when investigating the association between different static glucose levels on mood, experimental research in healthy volunteers showed no consistent effect, 4 while some studies in persons with diabetes suggest that both hyperglycaemia and hypoglycaemia can induce negative mood states, including anxiety, sadness and agitation. [5][6][7] Also, self-monitoring of glucose values can elicit strong emotional responses, often negative and related to a sense of failure. 8 When investigating the link between the dynamics in glucose levels and mood, blinded continuous glucose monitoring (CGM) technology provides the opportunity to observe the association between GV and mood, as noted by Rausch et al a decade ago. 9 It is important to note that some metrics of GV strongly correlate with mean glucose. 10 However, the relationship between mean glucose and mood does not capture the daily emotional impact of glucose excursions. A better understanding of this association might help to reduce the uncertainty around the interrelationship between one's blood glucose level and mood, which has been identified as one of the most frequently endorsed problem areas by both people with type 1 and type 2 diabetes. 11 Moreover, new diabetes medications and diabetes technologies can help to achieve less glucose variability and more "time in range". 12 With increasing uptake of CGM use in research, a literature overview can help to enhance our understanding of the potential psychological benefits of improved glucose stability for persons with diabetes.
In this systematic review, we aim to give an overview of the existing literature regarding the association between GV and mood in adults with diabetes mellitus. Furthermore, we discuss the strengths and weaknesses of the methods used to examine this association and make suggestions for future research.

| Data sources and searches
A literature search was performed based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. 13 PubMed, EMBASE and PsycINFO (EBSCO) databases were searched using the following terms (including synonyms and closely related words) as index terms or free-text words: "Diabetes Mellitus", "Blood glucose variability" and "Mood" to identify studies that examined the association between glucose variability and mood in adults with type 1 and type 2 diabetes (for full search, see Appendix S1).
References of included studies and relevant reviews were checked for additional relevant articles. The initial search was performed in July 2018 and updated in May 2019. Covidence software was used to manage the screening process. 14

| Study results and study selection
The literature search generated a total of 3.049 references: 944 in PubMed, 1.786 in Embase.com and 319 in PsycINFO. After removing duplicates of references that were selected from more than one database, 2.316 references remained. The flow chart of the search and selection process is presented in Figure 1.
Peer-reviewed studies published in English that examined the association between glucose variability and mood in adults with diabetes were included. Reviews, conference abstracts, qualitative studies, editorials and case report forms were excluded. Inclusion criteria were as follows: observational or experimental research designs; assessing glucose variability (eg "fluctuation", "instability" and "glucose rate of change"); and assessment of mood (eg emotion, well-being and affect) (see Appendix S2 for detailed inclusion criteria).
Study titles and abstracts were screened, and subsequently, full texts were reviewed for inclusion in duplicate by seven reviewers (CR, LTM, MdW, FJS, THW, FR and AB) independently. Three reviewers (CR, LTM and MdW) discussed conflicts until agreement was reached.

| Data extraction and quality assessment
Data were extracted independently and in duplicate by three reviewers (CR, LTM and MdW) including study design, country of

Novelty statement
What is already known • There is a long-standing interest in the association between glucose variability (GV) and mood in persons with diabetes.
• Empirical evidence regarding this association has not been systematically reviewed.

What this study has found
• Four of the eight included studies used continuous glucose monitoring (CGM).
• A significant association was found between a higher rate of postprandial glucose increase and more negative mood symptoms.
• No other evident patterns between GV and mood emerged.
• Higher quality experimental and observational studies are needed using CGM and ecological momentary assessment.

What are the clinical implications of the study
• Increasing use of sensor technology in routine care will increase insight into glucose over time.

| Data synthesis
Given the heterogeneity of measurements and study designs, a standard meta-analysis was not feasible. We therefore conducted a descriptive analysis of the collected data. For data synthesis, we grouped the studies using a two-dimensional map based on the time frame of the measurement of GV (within 1 day, ie intraday, vs. more than 1 day, ie multiday, on the X-axis) and mood (momentary vs. retrospective on the Y-axis), resulting in four quadrants.

| Study characteristics
A flowchart of the study selection process is shown in Figure 1.

| Quality assessment of included studies
Using the NIH Quality Assessment Tool, the quality of seven studies [18][19][20][21][22][23]25 was judged to be fair, and one study 24 was judged to be of poor quality (see Table 2). Appendix S3 gives an overview of the quality assessment of each study.

| Glucose variability measures
Three studies used the CGM device, called Medtronic MiniMed (CGMS; Medtronic MiniMed), which they blinded for the study participants and allowed only the retrospective analysis of glucose values, 19,24,25 and four studies used SMBG 18,21-23 to determine GV, using various metrics and time windows as indicators of GV. One study used a glucose/insulin infusion procedure allowing for continuous measurement of blood glucose similar to CGM. 20 Five studies measured intraday GV, 19,20,[22][23][24] two studies measured GV over multiple days, 18,21 and one study captured intraday GV as well as over one week. 25 The GV metrics used as described by Siegelaar et al 15 were the standard deviation (SD), with time windows ranging from 10 hours up to four weeks, 18,21,24,25 coefficient of variation (CV) 19 and continuous overall net glycaemic action (CONGA). 24 Other GV metrics reported by the authors were intraday change in blood glucose to indicate GV, such as blood glucose rate of change 19,20,22,23 and a newly introduced measure called CGM "energy". 24

| Outcomes of association between glucose variability and mood
The association between GV and mood was examined in three ways, as presented in Figure 2. First, the association between intraday GV and momentary mood was assessed in five studies 19,20,[22][23][24] (Figure 2, and proved strongest at one hour postmeal. 23 No correlation was observed between postprandial blood glucose rate of changes and positive (energetic) mood symptoms. 22 Gonder-Frederick et al 20 Table 2.  15 However, the optimal time period between the tested GV window and the correlated current mood rating, or change in mood rating in the previously mentioned GV window, has yet to be determined. It would seem that studies linking GV and momentary mood ratings should at least cover multiple  in type 1 diabetes. 29 The impact of GV on mood can be assumed to be a function of experiencing extreme glucose excursions, that is amplitude and the frequency of oscillations. Research in type 1 diabetes has established profound effects of severe hypoglycaemia on mood states that may persist over time. 7,30,31 Similarly, acute hyperglycaemia might alter mood in type 2 diabetes, only above a certain glycaemic threshold. 32,33 As to the direction of the relationship between GV and mood, almost all studies examined whether GV was a predictor of subsequent mood changes, but reversed causality cannot be excluded.

HbA1c in mmol/mol; % (mean ± SD (range))
Wagner et al 25 indeed assessed whether mood was a predictor of subsequent GV, but found no evidence for this direction. It is important to note that none of the reviewed studies used time-series statistical analysis to model the relationship between GV and mood using temporal data. As suggested by Rausch, 9 this approach is necessary to accurately determine the direction of the relationship between glucose variability and mood.
More work needs to be done to understand potential mechanisms underlying an association between GV and mood. while depression is characterized by low mood variability. 35 Persons can also differ in terms of interoceptive (bodily) awareness, including impaired hypoglycaemia awareness. Other possible moderators of the link between GV and mood include trait anxiety 36 and sleep 37,38 that are associated with instable glucose levels as well as poor emotional well-being.
Another phenomenon that could hypothetically alter the relationship between GV and mood is impaired cardiovascular autonomic modulation, 39,40 as is the case in cardiovascular autonomic neuropathy, one of the complications of diabetes. 41 Understanding interindividual differences in emotional reactivity to GV could help predict which persons with diabetes could profit most from more stable blood glucose levels in terms of their emotional health.
To further improve the quality of research in this field, standardiza- In conclusion, based on this systematic review of eight studies no firm conclusions can be drawn with regard to the association between GV and mood in adults with type 1 and type 2 diabetes.
More and higher quality experimental and observational studies with larger populations over a longer period of time are needed. New technologies, such as blinded CGM and EMA mobile applications, are promising to assess this association more precisely, addressing a question that is perceived to be of high importance from the perspective of persons with diabetes.

ACK N OWLED G EM ENTS
None.

CO N FLI C T O F I NTE R E S T
No conflicts of interest. were involved in drafting the manuscript and revising it critically, read and approved the final manuscript and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

E TH I C S S TATEM ENT
Since this review summarizes and informs already published studies, ethical approval is not applicable.

DATA AVA I L A B I L I T Y S TAT E M E N T
Supporting data about the search details, the inclusion and exclu-