Diabetes Ther (2021) 12:765–779 https://doi.org/10.1007/s13300-021-01009-x ORIGINAL RESEARCH Impact of Intensive Glucose Control on Brain Health: Meta-Analysis of Cumulative Data from 16,584 Patients with Type 2 Diabetes Mellitus Xingyao Tang . Marly A. Cardoso . Jinkui Yang . Jian-Bo Zhou . Rafael Simó Received: November 1, 2020 / Accepted: January 23, 2021 / Published online: February 6, 2021 Ó The Author(s) 2021 ABSTRACT versus standard care on brain health, including structural abnormalities of the brain (atrophy, Introduction: Despite growing evidence that white matter hyperintensities, lacunar infarc- type 2 diabetes is associated with dementia, the tion, and cerebral microbleeds), cognitive dys- question of whether intensive glucose control function, and risk of dementia. can prevent or arrest cognitive decline remains Methods: We searched the PubMed and unanswered. In the analysis reported here, we Embase databases, the Web of Science website, explored the effect of intensive glucose control and the Clinicaltrial.gov registry for studies published in English prior to July 2020. Only studies with a randomized controlled trial (RCT) X. Tang design were considered. We analyzed structural Beijing Tongren Hospital, Capital Medical University, Beijing, China abnormalities of the brain (atrophy, white matter hyperintensities, lacunar infarction, and J. Yang  J.-B. Zhou (&) cerebral microbleeds), cognitive function (cog- Department of Endocrinology, Beijing Tongren nitive impairment, executive function, mem- Hospital, Capital Medical University, Beijing, China e-mail: jbzhou@ccmu.edu.cn ory, attention, and information-processing speed), and dementia (Alzheimer’s disease, vas- M. A. Cardoso cular dementia, and mixed dementia). Department of Nutrition, School of Public Health, University of Sao Paulo, Sao Paulo, Brazil Results: Six studies (5 different RCTs) with 16,584 participants were included in this meta- R. Simó analysis. One study that compared structural Endocrinology and Nutrition Department, Hospital Universitari Vall d’Hebron, Universitat Autònoma changes between groups receiving intensive de Barcelona, Passeig de la Vall d’Hebron, Barcelona, versus conventional glucose control measures Spain reported non-significant results. The results of the five studies, comprising four cohorts, indi- R. Simó Diabetes and Metabolism Research Unit, Vall cated a significantly poorer decline in cognitive d’Hebron Institut de Recerca (VHIR), Universitat function in the intensive glucose control group Autònoma de Barcelona, Passeig de la Vall (b - 0.03, 95% confidence interval [CI] - 0.05 d’Hebron, Barcelona, Spain to - 0.02) than in the conventional glucose R. Simó control group. Further subgroup analysis Centro de Investigación Biomédica en Red de showed a significant difference in the change in Diabetes y Enfermedades Metabólicas Asociadas cognitive performance in composite cognitive (CIBERDEM), Instituto de Salud Carlos III (ISCIII), Madrid, Spain function (b - 0.03, 95% CI - 0.05 to - 0.01) 766 Diabetes Ther (2021) 12:765–779 and memory (b - 0.13, 95% CI - 0.25 to DIGITAL FEATURES - 0.02). One trial evaluated the prevalence of cognitive impairment and dementia between This article is published with digital features, groups receiving intensive and conventional including a summary slide, to facilitate under- glucose control, respectively, and the differ- standing of the article. To view digital features ences were insignificant. for this article go to https://doi.org/10.6084/ Conclusion: This meta-analysis suggests that m9.figshare.13614653. intensive glucose control in patients with type 2 diabetes can slow down cognitive decline, especially the decline in composite cognition INTRODUCTION and memory function. However, further studies are necessary to confirm the impact of strict Type 2 diabetes is a significant health challenge glucose control on structural abnormalities in worldwide, and the prevalence of diabetes in the brain and the risk of dementia. adults is predicted to increase to 10.4% by 2040 [1]. Dementia and diabetes frequently coexist Keywords: Brain health; Diabetes; Intensive [2], with the prevalence of dementia being glucose control approximately 14% among people with diabetes aged more than 80 years [3]. Moreover, diabetes is a risk factor for cognitive dysfunction that can Key Summary Points range from subtle diabetes-associated cognitive decline to pre-dementia and dementia [4]. The Why carry out this study? potential mechanisms of type 2 diabetes Despite growing evidence that type 2 impacting brain health might be related to diabetes is associated with dementia, dysfunction of the inflammatory system [5], whether intensive glucose control can oxidative stress [6], and endothelial dysfunction prevent or arrest the cognitive decline [7], all of which are induced by both insulin remains unknown. resistance and elevated blood glucose levels. Previously published results show that fluctua- This analysis was carried out with the aim to tions in blood glucose level have an effect on further explore the impact of intensive vascular complications [8, 9], which also versus conventional glucose control on impacts brain health. Thus, appropriate blood brain health. glucose control is vital for preserving brain What was learned from the study? health in patients with diabetes. A number of cohort studies have focused on The results suggest that intensive glucose the differential impact of intensive and con- control in patients with type 2 diabetes can ventional blood glucose control on brain slow down cognitive decline, especially the health, including structural abnormalities in decline in composite cognition and memory the brain [10], cognitive dysfunction, and function. dementia [11]. Authors have suggested that cognitive functioning in patients with type 2 However, further studies are necessary to diabetes might benefit from several months of confirm the impact of strict glucose control improved glycemic control [12, 13]. A recent on structural abnormalities in the brain and meta-analysis indicated that people with type 2 the risk of dementia. diabetes performed better in processing speed and executive function but more poorly in memory and attention when on intensive glu- cose control regimes [14]. However, this analy- sis assessed the performance of the participants only at the end of the study. Diabetes Ther (2021) 12:765–779 767 The aim of the meta-analysis reported here with type 2 diabetes; (3) at least one group of was to further explore the impact of intensive patients received intensive therapy and the versus conventional glucose control on brain other received conventional therapy to control health. Based on the previous studies, we blood glucose levels; (4) clear and consistent focused not only on cognitive performance at definitions of glucose level, brain structure the end of the study but also on cognitive abnormalities, and/or cognitive function or any change during the follow-up years, structural type of dementia; and (5) data available on abnormalities of the brain, and the risk of comparisons of brain health between the study dementia, as reported in randomized controlled groups receiving intensive and conventional studies. glucose control therapy with more than a 1-year follow-up. Exclusion criteria were: (1) reviews, case reports, animal studies, or letter to the METHODS editor; (2) articles that did not clearly define clinical outcomes; (3) the author could not Study Inclusion provide valid data when contacted; and (4) duplicated data. The PubMed and Embase databases, the Web of The definition of ‘‘intensive therapy’’ was Science website, and the Clinicaltrial.gov reg- achieving glycated hemoglobin (HbA1c) of \ istry were searched for relevant studies pub- 7% or a fasting glucose level of \ 130 mg/dL. lished up to July 2020. Search terms used were The format and content of the interventions ‘‘diabetes mellitus,’’ ‘‘hypoglycemic therapy’’ could vary. For the current analysis, if articles (including antidiabetic drugs, physical activity, provided data from the same cohort, we selec- and diet therapy), ‘‘brain structural abnormali- ted the cohort with the longest follow-up, lar- ties’’ (including brain atrophy, white matter gest sample size, and most comprehensive hyperintensities, lacunar infarction, and cere- results. We analyzed the differences in brain bral microbleeds), ‘‘cognitive function’’ (in- structural abnormalities, cognitive dysfunction, cluding cognitive impairment, executive and prevalence of dementia between the group function, memory, attention, and information- receiving conventional glucose control and the processing speed), and ‘‘dementia’’ (including one receiving intensive glucose control. Alzheimer’s disease, vascular dementia, and mixed dementia), used alone and in combina- Data Extraction and Quality Assessment tion. Further pertinent articles were supple- mented by inspecting the references of the The literature search was performed in July selected articles. 2020. Two investigators (XYT, JBZ) indepen- This report was conducted according to the dently extracted the data from the included Meta-analysis Of Observational Studies in Epi- studies, following identical standards on the demiology [15] and the Preferred Reporting study cohort, follow-up years, population Items for Systematic Reviews and Meta-Analysis characteristics, and outcomes. The risk of bias in [16]. This article is based on previously con- selecting the included studies was assessed ducted studies and does not contain any new using the Newcastle–Ottawa Quality Assess- studies with human participants or animals ment Scale (NOS) criteria [17]. Following the performed by any of the authors. NOS guidelines, we rated the quality of the studies on the basis of selection, comparability, Inclusion and Exclusion Criteria and outcome. In case of disagreement, the two investigators discussed the study with the other We included studies that met the following authors to arrive at a consensus. criteria: (1) original articles based on random- Composite cognitive function was defined ized controlled trials (RCTs) published in Eng- using the scores of the Mini-Mental State lish; (2) study population comprised patients Examination (MMSE) [18] or the 768 Diabetes Ther (2021) 12:765–779 Comprehensive Assessment and Referral Evalu- due to duplication. The remaining 5632 articles ation-Diagnostic Scale [19]; memory was asses- were subjected to title and abstract screening, sed using digit span scores [20] or Rey Auditory resulting in a further 5185 publications being Verbal Learning Test scores [21]; attention and excluded as they were reviews, letters or con- executive function were assessed using the ference abstracts, and case reports. The scores of the Trail Making Test [22] or Stroop remaining 447 articles were eligible for full-text Test [23]; and information-processing speed was review and data assessment. Of these, 441 arti- assessed using the Digit symbol substitution test cles were excluded for lack of relevant infor- scores [24]. Cognitive impairment was defined mation or because they were based on animal as a decline in cognitive function (decrease in studies. Ultimately, a total of six studies the MMSE score by at least 3 points, compared [10, 11, 26–29] with 16,584 participants met the to the baseline score) [11]. inclusion criteria. The mean follow-up duration ranged from 3.5 to 10 years. The mean age of Statistical Analysis participants in the studies included in the cur- rent meta-analysis was 65.6 years at the initia- Heterogeneity between the studies was evaluated tion of the studies, and the proportion of using the I2 metric, and the variance between women was 40.8% (Fig. 1). Table 1 shows the studies was evaluated using the Tau-squared characteristics of the six included studies. (Tau2) metric. Random-effects models were selected if I2 [ 50%, and fixed-effects models Quality Assessment were chosen if I2 B 0%. Data were obtained directly from the studies. Pooled differences in The NOS evaluation tool was used to assess the cognitive change during follow-up between the quality of the included studies. All quality groups receiving intensive glucose control and assessment scores fell in the range of 8 or 9, those receiving conventional glucose control indicating high quality (Table 2). were performed using the b metric and the 95% confidence interval (CI). For the Stroop Test, a Structural Changes in the Brain Between positive change in the value represents a wors- Groups Receiving Intensive ening score; hence, we used the opposite number and Conventional Glucose Control of the results. To identify additional potential factors impacting the bias, we also conducted a Only one cohort [10] study compared structural subgroup and sensitivity analysis. Sensitivity changes in the brain between groups receiving analyses were conducted to assess the influence intensive and conventional glucose control, of a single result on the pooled estimate. Egger’s respectively. The Action to Control Cardiovas- asymmetry test was used to evaluate potential cular Risk in Diabetes Memory in Diabetes publication bias [25]. p values were two-tailed, (ACCORD MIND) trial, a sub-cohort of the and p \ 0.05 was considered to be statistically ACCORD trial, comprised 614 patients with significant. The statistical analyses were per- brain outcomes assessed by magnetic resonance formed using STATA version 12.0 (Stata Corp., imaging during 80 months of follow-up. The College Station, TX, USA). results showed that both the intensive and conventional glucose control groups had a RESULTS decrease in total brain volume and an increase in abnormal white matter volume. Compared to Literature Search Outcomes and Validity the conventional glucose control group, the Assessment intensive glucose control group had a signifi- cantly slower decrease in total brain volume (4.6 cm3; 95% CI 2.0–7.3, p = 0.0006) at The search strategy identified 8852 potentially 40 months; however, there was no significant relevant reports, of which 3220 were excluded difference in the decrease in total brain volume Diabetes Ther (2021) 12:765–779 769 Fig. 1 PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram summarizing the screening process between the groups at 80 months (0.2 cm3; 95% Change in Cognitive Function CI - 3.4–3.8, p = 0.91). The increase in log- in the Intensive and Conventional transformed abnormal white matter volume at Glucose Control Groups 40 months was significantly quicker in the intensive glucose control groups than that in Five studies [10, 26–29] comprising four the conventional glucose control groups (0.10; cohorts, including the Informatics in Diabetes 95% CI 0.02–0.18, p = 0.014). However, at Education and Telemedicine Study, the Anglo- 80 months, there was no significant difference Danish–Dutch Study of Intensive Treatment in between the groups (0.05; 95% CI - 0.07 to People with Screen-Detected Diabetes in Pri- 0.17, p = 0.41). mary Care, the Veterans Affairs Diabetes Trial 770 Table 1 Characteristics of studies included in the meta-analysis on the effects of intensive and conventional glucose control on brain health First Year Country Triala Duration of Sample (n) Age, mean Follow- Female Intensive glucose goal author of diabetes (SD) up study (years)lb Launer 2011 USA ACCORD Intensive: 9 Total: 1977 Intensive: 62.3 40 months Intensive: HbA1c: \ 6% [26] MIND (5–14) Intensive: (5.7) years 47.5% Conventional: 1469 Conventional: Conventional: 9 (5–15) Conventional: 62.7 (5.9) 45.8% 1508 years Luchsinger 2011 USA IDEATel Intensive: 10.9 Total: 2169 Intensive: 70.6 3.5 years Intensive: HbA1c: \ 7% [27] (9.4) Intensive: (6.6) years 61.8% Conventional: 1093 Conventional: Conventional: 10.7 (9.0) Conventional: 70.4 (6.8) 60.2% 1076 years Koekkoek 2012 The ADDITION – Total: 183 Intensive: 59.3 10 years Intensive: HbA1c: \ 53 mmol/mol [28] Netherlands Intensive: 97 (5.6) years 42.3% (7.0%) Conventional: Conventional: Conventional: 86 59.5 (5.3) 35.7% years Zimering 2016 USA VADT 11.4 (7.5) Total: 1764 60.4 (8.7) years 5.6 years 4% HbA1c: \ 6.0% [29] Intensive: 877 Conventional: 887 Murray 2017 USA ACCORD 10.3 (7.0) Total: 1328 62.1 (5.5) years 80 months 43.10% HbA1c: \ 6.5% [10] MIND Intensive: 684 Conventional: 644 Diabetes Ther (2021) 12:765–779 Diabetes Ther (2021) 12:765–779 771 (VADT), and the ACCORD MIND trial, descri- Study, ADDITION the Anglo–Danish–Dutch Study of Intensive Treatment in People with Screen-Detected Diabetes in Primary Care, VADT the Veterans Affairs ACCORD MIND Action to Control Cardiovascular Risk in Diabetes Memory in Diabetes, IDEATel the Informatics in Diabetes Education and Telemedicine Intensive glucose goal bed the difference in the change in cognitive function between the groups receiving inten- sive glucose control and those receiving con- HbA1c: \ 6.5% ventional glucose control. Since the cohorts in two studies were identical (ACCORDION MIND; [10, 26]), we selected the study with the longer Diabetes Trial, ADVANCE the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation follow-up period [10]. A total of 2751 patients who received intensive glucose control treat- ment and 2693 patients who received conven- Conventional: tional glucose control treatment were included Intensive: 42.6% in this analysis. With the conventional glucose 42.3% Female control group as a reference, the pooled results indicated a statistically significant slower decline in cognitive function in the intensive Follow- glucose control group (b - 0.03, 95% CI - 0.05 5 years Values are presented as the mean with the range in parentheses or as the mean with the SD in parentheses to - 0.02). Further subgroup analysis showed a up significantly slower decline in the cognitive performance composite (b - 0.03, 95% CI Conventional: 66 (6) years Intensive: 66 - 0.05 to - 0.01) and memory (b - 0.13, 95% Age, mean (6) years CI - 0.25 to - 0.02) in the intensive glucose control group; however, no difference in infor- (SD) mation-processing speed (b 0.09, 95% CI - 0.06 to 0.25) or attention and executive function (b Conventional: Total: 11,140 - 0.14, 95% CI - 0.30–0.03) was seen between Sample (n) the groups (Fig. 2). Intensive: The results of the meta-regression analyses 5571 5569 suggested that differences in cognitive change between the two groups were not associated with sample size, number of follow-up years, Conventional: Intensive: 7.9 Duration of mean age, and NOS scores (p [ 0.05). After 8.0 (6.4) (years)lb diabetes excluding studies one by one, the outcomes of (6.3) the sensitivity analysis indicated that the results were stable. In the analysis of publication bias, the p value of the Egger test was [ 0.05 ADVANCE (p = 0.74) for the difference in cognitive change between the intensive and conventional glu- Triala cose control groups. The funnel figure of these studies showed no publication bias, having an asymmetrical inverted distribution, which is 2018 20 countries consistent with the results of the Egger test. Year Country (Fig. 3). SD Standard deviation Difference in the Risk of Dementia Table 1 continued Between the Intensive and Conventional Glucose Control Groups author of Kirkman Only one cohort [11], the Action in Diabetes [11] study First and Vascular Disease: Preterax and Diamicron b a 772 Table 2 Risk of bias assessment (Newcastle–Ottawa Quality Assessment Scale criteria) of studies on the association between intensive and conventional glucose control on brain health included in the meta-analysis First Selection Comparability Outcome Quality author of Representativeness Selection Ascertainment Demonstration Comparability Ascertainment Was Adequacy score study of the exposed of the of exposure that the outcome of cohorts of outcome follow-up of follow- cohort non- of interest was based on the long up of exposed not present at the design or the enough for cohorts cohort start of the study analysis outcomes to occur? Launer * * * * ** * * * 9 [26] Luchsinger * * * * ** * – * 8 [27] Koekkoek – * * * ** * * * 8 [28] Zimering * * * * ** * * * 9 [29] Murray * * * * ** * * * 9 [10] Kirkman * * * * ** * * * 9 [11] The Newcastle-Ottawa Scale consists of 8 items with 3 subscales. The total maximum score of these 3 subsets is 9. A study which scores C 7 is a high-quality study Diabetes Ther (2021) 12:765–779 Diabetes Ther (2021) 12:765–779 773 Fig. 2 Difference in cognitive function change between random effects estimate of the subgroup analysis studies. groups receiving intensive glucose control and those Weights are from the fixed-effects analysis, and %Weight is receiving conventional glucose control. CI Confidence the weight assigned to each study, based on the inverse of interval. I-squared is the variation in effect estimates the within- and between-study variance. The size of the attributable to heterogeneity, Overall refers to the pooled gray boxes around the point estimates reflects the weight fixed effect estimate of all studies, and Subtotal is the fixed assigned to each study Modified Release Controlled Evaluation 95% CI 0.87–1.85) between the two groups. (ADVANCE) trial, measured the cognitive Further analysis of relative risk reduction also decline and risk of dementia between intensive showed non-significant outcomes between the and conventional glucose control groups. The two groups in terms of cognitive impairment trial included 5571 patients in the intensive (2%; 95% CI - 7–11%) and risk of dementia control group and 5569 in the conventional (- 27%; 95% CI - 86–13%). control group with 8 years of follow-up. How- ever, there was no significant difference in the cognitive impairment (risk ratio [RR] 0.98, 95% CI 0.90–1.07) and risk of dementia (RR 1.27, 774 Diabetes Ther (2021) 12:765–779 Fig. 3 Funnel plot of differences in impact of intensive and conventional glucose control on cognitive function change. SE Standard error DISCUSSION on brain structural abnormalities and risk of dementia are needed. A total of six studies involving five cohorts were A recent meta-analysis that included only included in our meta-analysis. We compared three studies on type 2 diabetes compared the the impact of intensive and conventional glu- difference in the impact of intensive and con- cose control on brain health, including brain ventional glucose control on cognitive perfor- structural abnormalities, cognitive change, and mance among patients with diabetes [14]. The the risk of dementia. According to the pooled pooled analysis of these three studies, including results, intensive glucose control can slow down 3200 individuals with type 2 diabetes, showed the cognitive function—which declines with the benefits of intensive glucose control therapy age—in terms of composite performance and on information-processing speed and executive memory, but not in terms of information-pro- functions, but no difference was observed on cessing speed or attention and executive func- attention and memory. However, the sample tion. Although the impact of intensive glucose size was small, and the analysis focused only on control on composite performance showed sta- cognitive performance at the end of the follow- tistical significance, the rate of decline was up period. The current meta-analysis had a lar- minimal; hence, the results should be inter- ger sample size and analyzed the difference in preted with caution. Therefore, more rigorous impact between intensive and conventional randomized controlled studies on the impact of glucose control on cognitive change during the intensive versus conventional glucose control follow-up period, brain structural Diabetes Ther (2021) 12:765–779 775 abnormalities, and risk of dementia. We found However, a higher mortality and frequency of that only one RCT reported a difference in total hypoglycemic events, both of which are threats brain volume decline and an increase in to brain health, have been reported in patients abnormal white matter volume between the treated with intensive glucose control measures two groups; another study mentioned the risk [38]. Evidence has shown that insulin resistance of dementia. might damage the brain structure, even in those There is evidence that in comparison to without clinical diabetes [39], indicating that conventional glucose control, intensive glucose abnormal functioning of insulin signaling treatment measures increase the risk of severe rather than blood glucose levels affect the cog- hypoglycemia [30]. Overall mortality in the nitive impairment in patients with type 2 dia- ACCORD and VADT studies was higher in the betes. Furthermore, the Diabetes Heart Study intensive glucose control group than in the Mind provided neuroimaging data and revealed conventional glucose control group, while in that the presence or absence of type 2 dia- the ADVANCE study, the overall mortality was betes—but not HbA1c level—was indepen- lower in the intensive than in the conventional dently associated with reduced white matter treatment group [30]. Severe hypoglycemia is volume and reduced gray and white matter considered a risk factor for cognitive impair- fractional anisotropy [40]. Thus, the influence ment [31] and mortality [32]. Nevertheless, the of tighter glucose control on brain health (brain current analysis showed that intensive glucose structural abnormalities, cognitive decline, and control slowed the cognitive decline and might dementia prevalence) should be considered have more obvious benefits on cognitive more cautiously. performance. Public Health Impact Potential Mechanisms With the rapid increase in aging populations Various mechanisms may have the potential to worldwide, diabetes [41] and dementia [42] contribute to the association between elevated have become an increasing burden for society. blood glucose levels, glucose variability, and Prevention and management of cognitive dys- brain health. Diabetes causes insulin resistance, function and dementia are essential, especially imbalance in brain energy metabolism, changes in those diagnosed with diabetes. The benefits in the regional cerebral blood flow, and osmotic of intensive glucose control in preventing effects on neurons, all of which damage brain microvascular complications, such as diabetic health [33]. Insulin is one of the molecules that retinopathy and diabetic kidney diseases, have provides neuroprotection by regulating tau been proven during a long-term follow-up study protein phosphorylation in the neurons and is [43]. The results of the current meta-analysis crucial for the health of neuronal networks [34]. indicate that intensive glucose control among Insulin resistance may cause tau binding to patients with type 2 diabetes patients might microtubules, thus triggering the pathogenesis slow down cognitive decline in composite and of cognitive decline and eventually leading to memory function. Better glucose control effec- dementia [35]. Intensive glucose control can tively prevents deterioration in brain health, preserve insulin secretion and functional roles, which can help to avoid tremendous economic which, in turn, improve neuronal functions. costs. However, hypoglycemia unawareness, Another possible mechanism is that poor blood significant comorbidities (risk factors of brain glucose control might lead to neuronal mito- health), and mortality [44] caused by intensive chondrial dysfunction and stress, DNA damage, glucose control treatment [45] suggest that and oxidative stress in endothelial cells [36, 37]. more caution is necessary for prioritizing Thus, stricter management of blood glucose intensive glucose control as a general rule, par- levels might decrease these risk factors, espe- ticularly in elderly subjects. Therefore, an cially in those diagnosed with type 2 diabetes. 776 Diabetes Ther (2021) 12:765–779 individualized approach seems the most rea- ACKNOWLEDGEMENTS sonable strategy. Strengths and Limitations of the Study Funding. This work was supported by the National Natural Science Foundation of China Although our meta-analysis considered the (No. 82070851, 81870556, 81930019), Beijing entire spectrum from pre-dementia to dementia Municipal Administration of Hospital’s Youth and extracted valuable information regarding Program (QML20170204), Excellent Talents in associations between intensive and conven- Dongcheng District of Beijing. Capital’s Funds tional glucose control with brain health among for Health Improvement and Research (2020-1- patients with diabetes, its essential strengths 1181). The Rapid Service Fee was funded by the and limitations merit mention. The strengths authors. include choosing the strictest study design (RCT), developing a comprehensive search Authorship. All named authors meet the strategy, and conducting a comprehensive International Committee of Medical Journal search. Some limitations warrant mention. The Editors (ICMJE) criteria for authorship for this first is that only a few studies focused on the article, take responsibility for the integrity of relationship between intensive and conven- the work as a whole, and have given their tional glucose control with brain structural approval for this version to be published. abnormalities and the risk of dementia. Second, Disclosures. The authors are solely respon- the lack of information on the adverse effects of sible for the design and conduct of this study, intensive glucose control, such as severe hypo- all study analyses, the drafting and editing of glycemia, limited further analysis [46]. Third, the manuscript, and its final contents. Xingyao some antidiabetic drugs had a different impact Tang, Marly A. Cardoso, Jinkui Yang, Jian-Bo on cognitive impairment and dementia [47]; for Zhou, and Rafael Simó have no disclosures to example, glucagon-like peptide 1 analogs and declare. sodium-dependent glucose transporter 2 inhi- bitors showed a protective effect [48, 49], while Compliance with Ethics Guidelines. This thiazolidinediones might have a negative article is based on previously conducted studies impact [50]. However, the current analysis on and does not contain any new studies with this aspect was limited due to the lack of data. human participants or animals performed by Fourth, the heterogeneity of the results; these any of the authors. might be due to differences in race, methods, and goals of intensive glucose control. Data Availability. All data generated or analyzed during this study are included in this published article/as supplementary information CONCLUSIONS files. The current study suggests that intensive glu- Open Access. This article is licensed under a cose control in persons with type 2 diabetes can Creative Commons Attribution-NonCommer- slow down cognitive decline, especially the cial 4.0 International License, which permits decline in composite and memory function. any non-commercial use, sharing, adaptation, The impact of intensive glucose control on the distribution and reproduction in any medium brain structural abnormalities and risk of or format, as long as you give appropriate credit dementia needs further rigorously designed to the original author(s) and the source, provide studies to validate our findings. Also, replicating a link to the Creative Commons licence, and and validating these findings is warranted. indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons Diabetes Ther (2021) 12:765–779 777 licence, unless indicated otherwise in a credit 10. Murray AM, Hsu FC, Williamson JD, et al. line to the material. If material is not included ACCORDION MIND: results of the observational extension of the ACCORD MIND randomised trial. in the article’s Creative Commons licence and Diabetologia. 2017;60(1):69–80. your intended use is not permitted by statutory regulation or exceeds the permitted use, you 11. Kirkman MS, Mahmud H, Korytkowski MT. Inten- will need to obtain permission directly from the sive blood glucose control and vascular outcomes in patients with type 2 diabetes mellitus. Endocrinol copyright holder. To view a copy of this licence, Metab Clin N Am. 2018;47(1):81-?. visit http://creativecommons.org/licenses/by- nc/4.0/. 12. Naor M, Steingrüber HJ, Westhoff K, et al. Cogni- tive function in elderly non-insulin-dependent diabetic patients before and after inpatient treat- ment for metabolic control. J Diabetes Complicat. 1997;11(1):40–6. REFERENCES 13. Ryan CM, Freed MI, Rood JA, et al. Improving metabolic control leads to better working memory 1. Ogurtsova K, Da Rocha FJD, Huang Y, et al. IDF in adults with type 2 diabetes. Diabetes Care. diabetes atlas: global estimates for the prevalence of 2006;29(2):345–51. diabetes for 2015 and 2040. Diabetes Res Clin Pract. 2017;128:40–50. 14. Peñaherrera-Oviedo C, Moreno-Zambrano D, Pala- cios M, et al. Does intensive glucose control prevent 2. Bunn F, Burn AM, Goodman C, et al. Comorbidity cognitive decline in diabetes? A meta-analysis. Int J and dementia: a scoping review of the literature. Chronic Dis. 2015;2015:680104. BMC Med. 2014;12:192. 15. Stroup DF, Berlin JA, Morton SC, et al. Meta-anal- 3. Wild S, Roglic G, Green A, et al. Global prevalence ysis of observational studies in epidemiology: a of diabetes: estimates for the year 2000 and pro- proposal for reporting. Meta-analysis Of Observa- jections for 2030. Diabetes Care. 2004;27(5): tional Studies in Epidemiology (MOOSE) group. 1047–53. JAMA. 2000;283(15):2008–12. 4. Biessels GJ, Despa F. Cognitive decline and 16. Knobloch K, Yoon U, Vogt PM. Preferred reporting dementia in diabetes mellitus: mechanisms and items for systematic reviews and meta-analyses clinical implications. Nat Rev Endocrinol. (PRISMA) statement and publication bias. J Cran- 2018;14(10):591–604. iomaxillofac Surg. 2011;39(2):91–2. 5. Exalto LG, Whitmer RA, Kappele LJ, et al. An update 17. Stang A. Critical evaluation of the Newcastle-Ot- on type 2 diabetes, vascular dementia and Alzhei- tawa scale for the assessment of the quality of mer’s disease. Exp Gerontol. 2012;47(11):858–64. nonrandomized studies in meta-analyses. Eur J Epidemiol. 2010;25(9):603–5. 6. Chang CM, Hsieh CJ, Huang JC, et al. Acute and chronic fluctuations in blood glucose levels can 18. Folstein MF, Folstein SE, Mchugh PR. ‘‘Mini-mental increase oxidative stress in type 2 diabetes mellitus. state’’. A practical method for grading the cognitive Acta Diabetol. 2012;49(Suppl 1):S171–7. state of patients for the clinician. J Psychiatr Res. 1975;12(3):189–98. 7. Wardlaw JM, Smith C, Dichgans M. Mechanisms of sporadic cerebral small vessel disease: insights from 19. Gurland B, Kuriansky J, Sharpe L, et al. The Com- neuroimaging. Lancet Neurol. 2013;12(5):483–97. prehensive assessment and Referral Evaluation (CARE)—rationale, development and reliability. Int 8. Cavalot F, Petrelli A, Traversa M, et al. Postprandial J Aging Hum Dev. 1977;8(1):9–42. blood glucose is a stronger predictor of cardiovas- cular events than fasting blood glucose in type 2 20. Da W. Wechsler adult intelligence scale—revised diabetes mellitus, particularly in women: lessons UK. New York: Psychological Corporation; 1997. from the San Luigi Gonzaga Diabetes Study. J Clin Endocrinol Metab. 2006;91(3):813–9. 21. Schmidt M. Rey auditory verbal learning test: a handbook. Los Angeles: Western Psychological 9. Zoppini G, Verlato G, Targher G, et al. Variability of Services; 1996. body weight, pulse pressure and glycaemia strongly predict total mortality in elderly type 2 diabetic 22. Corrigan JD, Hinkeldey NS. Relationships between patients. The Verona Diabetes Study. Diabetes parts A and B of the Trail Making Test. J Clin Psy- Metab Res Rev. 2008;24(8):624–8. chol. 1987;43(4):402–9. 778 Diabetes Ther (2021) 12:765–779 23. Houx PJ, Jolles J, Vreeling FW. Stroop interference: 35. Hong M, Lee VM. Insulin and insulin-like growth aging effects assessed with the Stroop Color-Word factor-1 regulate tau phosphorylation in cultured Test. Exp Aging Res. 1993;19(3):209–24. human neurons. J Biol Chem. 1997;272(31): 19547–53. 24. Silverstein AB. Two- and four-subtest short forms of the Wechsler Adult Intelligence Scale-Revised. 36. Russo VC, Higgins S, Werther GA, et al. Effects of J Consult Clin Psychol. 1982;50(3):415–8. fluctuating glucose levels on neuronal cells in vitro. Neurochem Res. 2012;37(8):1768–82. 25. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. 37. Schisano B, Tripathi G, Mcgee K, et al. Glucose BMJ. 1997;315(7109):629–34. oscillations, more than constant high glucose, induce p53 activation and a metabolic memory in 26. Launer LJ, Miller ME, Williamson JD, et al. Effects of human endothelial cells. Diabetologia. 2011;54(5): intensive glucose lowering on brain structure and 1219–26. function in people with type 2 diabetes (ACCORD MIND): a randomised open-label substudy. Lancet 38. Jacobson AM, Musen G, Ryan CM, et al. Long-term Neurol. 2011;10(11):969–77. effect of diabetes and its treatment on cognitive function. N Engl J Med. 2007;356(18):1842–52. 27. Luchsinger JA, Palmas W, Teresi JA, et al. Improved diabetes control in the elderly delays global cogni- 39. Willette A, Xu G, Johnson S, et al. Insulin resis- tive decline. J Nutr Health Aging. 2011;15(6):445–9. tance, brain atrophy, and cognitive performance in late middle-aged adults. Diabetes Care. 2013;36(2): 28. Koekkoek PS, Ruis C, Van Den Donk M, et al. 443–9. Intensive multifactorial treatment and cognitive functioning in screen-detected type 2 diabetes—the 40. Raffield L, Cox A, Freedman B, et al. Analysis of the ADDITION-Netherlands study: a cluster-random- relationships between type 2 diabetes status, gly- ized trial. J Neurol Sci. 2012;314(1–2):71–7. cemic control, and neuroimaging measures in the Diabetes Heart Study Mind. Acta Diabetol. 29. Zimering M B, Knight J, Ge L, et al. Predictors of 2016;53(3):439–47. cognitive decline in older adult type 2 diabetes from the veterans affairs diabetes trial. Front 41. Nanayakkara G, Kariharan T, Wang L, et al. The Endocrinol (Lausanne). 2016;7:123. https://doi.org/ cardio-protective signaling and mechanisms of 10.3389/fendo.2016.00123 adiponectin. Am J Cardiovasc Dis. 2012;2(4): 253–66. 30. Schernthaner G. Diabetes and cardiovascular dis- ease: Is intensive glucose control beneficial or 42. Rizzi L, Rosset I, Roriz-Cruz M. Global epidemiology deadly? Lessons from ACCORD, ADVANCE, VADT, of dementia: Alzheimer’s and vascular types. UKPDS, PROactive, and NICE-SUGAR. Wien Med Biomed Res Int. 2014;2014:908915. Wochenschr. 1946;2010(160):8–19. 43. Holman RR, Paul SK, Bethel MA, et al. 10-year fol- 31. Mattishent K, Loke Y. Bi-directional interaction low-up of intensive glucose control in type 2 dia- between hypoglycaemia and cognitive impairment betes. N Engl J Med. 2008;359(15):1577–89. in elderly patients treated with glucose-lowering agents: a systematic review and meta-analysis. 44. Rawlings AM, Sharrett AR, Mosley TH, et al. Glu- Diabetes Obes Metab. 2016;18(2):135–41. cose peaks and the risk of dementia and 20-year cognitive decline. Diabetes Care. 2017;40(7): 32. International Hypoglycaemia Study Group. Hypo- 879–86. glycaemia, cardiovascular disease, and mortality in diabetes: epidemiology, pathogenesis, and man- 45. Terry T, Raravikar K, Chokrungvaranon N, et al. agement. Lancet Diabet Endocrinol. 2019, 7(5): Does aggressive glycemic control benefit 385–396. macrovascular and microvascular disease in type 2 diabetes? Insights from ACCORD, ADVANCE, and 33. Biessels GJ, Strachan MW, Visseren FL, et al. VADT. Curr Cardiol Rep. 2012;14(1):79–88. Dementia and cognitive decline in type 2 diabetes and prediabetic stages: towards targeted interven- 46. Whitmer RA, Karter AJ, Yaffe K, et al. Hypoglycemic tions. Lancet Diabetes Endocrinol. 2014;2(3): episodes and risk of dementia in older patients with 246–55. type 2 diabetes mellitus. JAMA. 2009;301(15): 1565–72. 34. Hölscher C. Novel dual GLP-1/GIP receptor agonists show neuroprotective effects in Alzheimer’s and 47. Zhou J, Tang X, Han M, et al. Impact of antidiabetic Parkinson’s disease models. Neuropharmacology. agents on dementia risk: a Bayesian network meta- 2018;136(Pt B):251–9. analysis. Metabolism. 2020;109:154265. Diabetes Ther (2021) 12:765–779 779 48. Bohlken J, Jacob L, Kostev K. Association between patients with type 2 diabetes: a nested case-control the use of antihyperglycemic drugs and dementia study. Eur J Endocrinol. 2019;181(5):499–507. risk: a case-control study. J Alzheimer’s Dis. 2018;66(2):725–32. 50. Cheng C, Lin C, Tsai Y, et al. Type 2 diabetes and antidiabetic medications in relation to dementia 49. Wium-Andersen I, Osler M, Jørgensen M, et al. diagnosis. J Gerontol. 2014;69(10):1299–305. Antidiabetic medication and risk of dementia in