Effect of the treatment of Type 2 diabetes mellitus on the development of cognitive impairment and dementia

Description of the condition

Diabetes mellitus can be divided into two major categories. Type 1 diabetes mellitus is an autoimmune disorder in which the insulin‐producing beta cells in the pancreas are destroyed. It typically begins in childhood or early adulthood (Atkinson 1994), although it can occur in adults as well (ADA 2017a). Type 2 diabetes mellitus is characterised by insulin resistance in peripheral tissues and progressive dysfunction of beta pancreatic cells (Kahn 2014). It usually affects adults, its prevalence increasing with age.

Diabetes may be diagnosed based either on the fasting plasma glucose (FPG) or the plasma glucose two hours after a 75 g oral glucose tolerance test (2‐h PG) — the plasma glucose criteria; or on the level of glycated haemoglobin in the blood — the HbA1c criteria (ADA 2017a).

There are a number of serious complications associated with diabetes. These have traditionally been divided into microvascular complications (retinopathy, neuropathy, and nephropathy) and macrovascular complications (cardiovascular and cerebrovascular disease) (Wright 2016).

Cognition is a term used to describe a set of higher cerebral functions, including memory, language, executive function, attention, perception and social behaviour. Dementia is a condition of cognitive impairment, usually progressive and irreversible, which involves a significant decline from a previous level of function and which interferes with independence in everyday activities (DSM‐5 2013). The greatest risk factor for dementia is increasing age. The most common diseases that lead to dementia are Alzheimer's disease (AD) and cerebrovascular diseases (Morris 2003).

In recent years, important epidemiological evidence has been gathered regarding the relationship between diabetes, cognitive dysfunction and dementia. No evidence of a substantial long‐term decline in cognitive function has been found in a large group of patients with Type 1 diabetes followed for an average of 18 years (DCCT/EDIC 2007), although this remains somewhat controversial (Kodl 2008). Cognitive changes have been described in Type 2 diabetes, affecting mainly learning and memory, mental flexibility and mental speed (Kodl 2008). People with Type 2 diabetes mellitus have been reported to have a risk of developing dementia 1.5 to 2 times higher than people without diabetes (Ott 1999; Gregg 2000; Cukierman 2005). A more recent meta‐analysis by Cheng 2012 found the relative risk of developing dementia in people with diabetes versus those without diabetes to be 1.46 for dementia due to Alzheimer's disease and 2.48 for vascular dementia. Diabetes mellitus and pre‐diabetes have also been found to accelerate the rate of progression from mild cognitive impairment to dementia (Xu 2010). Patients with Type 2 diabetes and established dementia have been found to be less likely to receive adequate diabetic care compared with those who have diabetes alone (Bunn 2014). The relationship between diabetes and dementia is likely to be complex and multifactorial.

Patients with Type 2 diabetes mellitus often have other cardiovascular risk factors and conditions, including high blood pressure, dyslipidaemia, cardiovascular disease, and cerebrovascular diseases that may contribute to cognitive impairment (Prince 1996; Kilander 1998; Rockwood 2002). Macrovascular disease, both peripheral and cerebral, is associated with dementia risk in patients with Type 2 diabetes mellitus (Bruce 2008). Cerebrovascular disease underlies vascular and mixed dementia and it has also been suggested that it triggers a series of events that lead to Alzheimer's disease (De la Torre 2000; Kalaria 2000; Shi 2000).

Patients with diabetes may develop microvascular disease, manifested as diabetic nephropathy and retinopathy. Albuminuria, an indicator of nephropathy, is associated with accelerated cognitive impairment (Barzilay 2013). It has also been shown that diabetic retinopathy is associated with dementia risk in patients with Type 2 diabetes mellitus (Exalto 2014).

Diabetes has a bidirectional association with depression, which is associated with dementia (Hugo 2014; Tabák 2014).

Insulin receptors are widely distributed in the brain, especially on neurons in areas concerned with memory and learning, such as the hippocampus and parts of the cerebral cortex (Wickelgren 1998). It is possible that insulin disorders may affect the utilisation of glucose in these areas. Type 2 diabetes also causes extensive disturbances in brain insulin and insulin‐like growth factor signalling mechanisms, which could promote molecular, biochemical, and histopathological lesions in AD (de la Monte 2008).

Description of the intervention

Treatment for a patient with Type 2 diabetes should be individualised, taking into account the patient's characteristics and comorbidities (NICE 2016; AACE/ACE 2017; ADA 2017b). International clinical practice guidelines (CPGs) and consensus standards vary, but all agree that the first line of treatment should include lifestyle changes, such as exercise and weight loss. Counselling may be provided to support these goals (NICE 2016; AACE/ACE 2017; ADA 2017c). Additional pharmacological treatment may also be offered at the time of diagnosis (AACE/ACE 2017); and is recommended if lifestyle recommendations do not achieve glycaemic goals (NICE 2016; ADA 2017b).

There are many different drugs available to treat Type 2 diabetes, including metformin, sulphonylureas, thiazolidinediones, dipeptidyl‐peptidase 4 (DPP‐4) inhibitors, sodium‐glucose cotransporter 2 (SGLT2) inhibitors, glucagon‐like peptide 1 (GLP‐1) agonists and insulin (NICE 2016; AACE/ACE 2017; ADA 2017b). Depending on the level of HbA1c, treatment may start with a single agent, commonly metformin, or a combination of drugs. The treatment becomes more complex if satisfactory glycaemic control is not achieved over time. Important factors influencing the choice of drug treatment relate to the overall health condition of the patient, especially considering the risk of hypoglycaemia, comorbidities, functional status, remaining life expectancy, renal disease, and macrovascular complications (NICE 2016; AACE/ACE 2017; ADA 2017b); and the presence and degree of cognitive impairment or dementia (ADA 2017d).

In recent years, intensive treatment strategies have been proposed to reduce the risk of micro‐ and macrovascular complications. However, the more intensive the treatment, the greater the risk of adverse effects such as hypoglycaemia and higher mortality. Therefore, intensive treatment might be not suitable in those at high risk of treatment complications (ADA 2017d; ADA 2017e).

How the intervention might work

Different drug treatments may have differential effects on dementia risk due to different actions on pathophysiological processes at the molecular level or different effects on other risk factors for cognitive decline. Hyperglycaemia is a risk factor for dementia (Crane 2013). Haemoglobin A1c levels reflect average glycaemia over several months. High levels have a strong predictive value for general diabetic complications and are related to the risk of developing cognitive impairment (Yaffe 2006). Hence, glycaemic control seems to be related to cognitive dysfunction in patients with Type 2 diabetes and tight glycaemic control may lower the risk of cognitive decline and dementia. On the other hand severe hypoglycaemia, which can be caused by pharmacological treatment of diabetes, may have permanent neurological consequences which may accelerate dementia. There are studies demonstrating that episodes of serious hypoglycaemia which require hospitalisation or visits to the emergency room are associated with a higher risk of dementia, and a greater number of episodes increases the risk (Whitmer 2009).

Why it is important to do this review

There is a very high level of interest in any interventions which could prevent or delay the onset of dementia. Since diabetes is a risk factor for dementia and is very common in the older population at risk of dementia, its treatment offers considerable potential for prevention. This review aims to identify and synthesise high‐quality evidence to determine whether different management strategies for Type 2 diabetes have any effect on cognitive function and the incidence of dementia. Given the prevalence of both conditions, even small effects of a particular management strategy could have consequences for public health.

Criteria for considering studies for this review

Search methods for identification of studies

Data collection and analysis

We tried to contact the trial investigators whenever data were missing, unclear, or if we believed we had detected a possible error in the published data. If this was unsuccessful, we described the uncertainties. We did not impute any missing data.

Description of studies

Details of all included and excluded studies are provided in the Characteristics of included studies and Characteristics of excluded studies tables.

Risk of bias in included studies

We considered none of the included studies to have an overall low risk of bias (Characteristics of included studies; Figure 2; Figure 3).

Effects of interventions

See: Table 1; Table 2; Table 3

See 'Summary of findings' tables for the comparisons of: 1) glibenclamide (glyburide) versus repaglinide; 2) intensive glycaemic control versus standard glycaemic control; and 3) rosiglitazone plus metformin versus glibenclamide (glyburide) plus metformin.

There is no 'Summary of findings' table for the comparison of metformin with placebo because we were unable to extract data from the study concerned (Guo 2014).

We were unable to extract any efficacy data from two of the seven included studies. Araki 2012 did not report the MMSE scores at the end of the follow‐up; the authors were contacted twice to request the data for inclusion in this systematic review, but no response was received. It was not possible to extract the data we needed from Guo 2014 because the results were reported only in an ANOVA format and no data could be extracted from the figures. Similarly, the authors were contacted twice to request the data, but no response was received. Zimering 2016 did not report measures of dispersion (e.g. standard deviation) for its cognitive outcomes and so we were not able to include this efficacy data in the meta‐analyses. We were able to extract data related to safety from only three studies of the seven studies included (ADVANCE 2008; Launer 2011; Zimering 2016).

Summary of main results

We included seven studies of treatments recommended by international clinical practice guidelines (CPGs) for the management of Type 2 diabetes. Objective tests of cognitive function were used in all seven studies, but two studies did not report useable data. Consequently, data were available for three comparisons: 1) repaglinide versus glibenclamide (glyburide) (one study); 2) standard versus intensive glycaemic control (three studies); and 3) rosiglitazone plus metformin versus glibenclamide (glyburide) plus metformin (one study).

1. Repaglinide versus glibenclamide (glyburide): we identified one study with 156 older people (age range 60 to 78 years) with Type 2 diabetes. Using this study in our analysis, we found low‐quality evidence of a small advantage on global cognitive function measured with the MMSE for glibenclamide (glyburide) over repaglinide after 12 months of treatment. However, we identified no difference in a composite score of attention and executive functions. The difference in MMSE score was 0.9 points. No Minimal Clinically Important Difference (MCID) for prevention studies has been established, but in participants without cognitive impairment at baseline the MMSE has a 'ceiling effect' (Spencer 2013), and the FDA considers any statistically significant difference in cognitive outcomes in a prevention trial to be of interest (FDA 2013). Hence a difference of 0.9 MMSE points, although smaller than is often considered clinically important in dementia treatment trials, may be important in this context.

2. Intensive versus standard glycaemic control: we found four studies with 16,081 participants, but only two studies with 13,934 participants had extractable efficacy data and three studies with 14,908 participants had extractable data on adverse effects. We found moderate‐quality evidence to suggest that there is probably little or no difference between intensive and standard treatment regimes on global cognitive function measured with the MMSE. We also found moderate‐quality evidence of no effect on several cognitive subdomains. We found low‐quality evidence from one study suggesting that there may be no difference in the incidence of dementia; moderate‐quality evidence from two studies that patients receiving an intensive treatment strategy are more likely to experience at least one severe hypoglycaemic episode; and low‐quality evidence from three studies that the treatment strategies may not be associated with any difference in all‐cause mortality.

3. Rosiglitazone plus metformin versus glibenclamide (glyburide) plus metformin: we found only one study with 145 participants. We found low‐quality evidence to suggest that there may be little or no difference in cognitive speed or working memory, but that there may be a small advantage for rosiglitazone plus metformin over glibenclamide (glyburide) plus metformin on episodic memory.

Overall completeness and applicability of evidence

The studies identified for inclusion in this review were insufficient to address all of its objectives.

We found data on only two comparisons of different drug regimes. Both were small and relatively short studies and both used glibenclamide (glyburide) in one study arm. This is a long‐acting sulphonylurea associated with a high risk of hypoglycaemia and is contraindicated in those over 65 years old (AGS 2012; ADA 2017d). Therefore, we believe that these studies are not helpful to clinicians who are treating older people, in whom Type 2 diabetes is most common.

More data were available to compare intensive with standard glycaemic control. For this comparison we could include data from three much larger studies. One study focused on people older than 55 years with a high risk for cardiovascular disease events (Launer 2011). The intensive glycaemic control group in this study had a target HbA1c of less than 6.0% and the standard control group had a target HbA1c between 7.0% and 7.9%. Since this study was conducted, the target HbA1c in current clinical practice guidelines (CPGs) has been reduced to less than 6.5% (NICE 2016; AACE/ACE 2017) or less than 7% (ADA 2017e). The standard control group in the trial has the HbA1c now expected for people who are frail, disabled, or have multiple comorbidities, although trial participants did not have these characteristics (ADA 2017d). Therefore the findings in this trial might not be very applicable to current practice. The second study, ADVANCE 2008, focused on patients aged over 55 years at the start of the study (mean 66 years) and with a history of major macrovascular or microvascular disease or at least one other risk factor for vascular disease. The intensive glycaemic group was treated with gliclazide plus other treatments to achieve a HbA1c of 6.5% or less, while the standard group was treated on the basis of local guidelines. This intensive glycaemic control is similar to the goal of current CPGs (ADA 2017c; NICE 2016); hence the results of this study might be more useful to answer our clinical question, especially considering the long follow‐up of five years. However, although the study evaluated global cognition and mortality, the incidence of dementia or MCI was not an outcome. The third study in this comparison, Zimering 2016, focused on participants from the Veteran Affairs Medical Centers in US who were older than 41 years and who were non‐responsive to a maximum dose of at least one oral agent or daily insulin injections, or both. A limitation of this study in the context of our clinical question is the average age of only 60 years at the start of the study, so that the study population was at low risk of cognitive decline over the five‐year follow‐up period.

An important point to note is that our analyses are not adjusted for the level of glycaemic control such as HbA1c, fasting plasma glucose (FPG), or postprandial plasma glucose. Abbatecola 2006 did adjust some analyses for haemoglobin A1c and coefficient of variation of FPG and found that the statistically significant differences in the composite score and MMSE were maintained. Ryan 2006 found some correlation between FPG and reduction in Paired Associates Learning (PAL) errors for both treatments (Ryan 2006).

The included studies had a wide range of durations. Three of the seven studies had a follow‐up for five years or more but only one of these had extractable data on cognition. Two studies had only six months of follow‐up. A prolonged follow‐up is necessary to detect those people who will progress to MCI or dementia over time, probably the most important cognitive outcomes in studies of this nature. The MMSE is not sensitive enough to detect small changes in cognition over short time periods and this is a limitation of those studies with short follow‐up.

Overall, the review is limited by the small number of included studies and limited range of comparisons. Few studies of the efficacy of treatments for Type 2 diabetes have included measures of cognition. Where cognition has been studied, the tools used have not been standardised, and each study has used scales which evaluate different cognitive domains. This makes it very difficult to compare the findings of these studies.

Quality of the evidence

We evaluated the quality of evidence according to the GRADE approach (Guyatt 2008). This considers study limitations, directness of the evidence, heterogeneity in the data, precision of effect estimates, and other reasons to downgrade the quality (including indications of publication bias). In this review, we evaluated the overall quality of evidence for all results to be low or moderate, meaning that our confidence in the results is limited. Reasons for rating down the quality of evidence were risk of bias in the included studies and imprecision of the results or small sample sizes.

Potential biases in the review process

It is likely that we identified all relevant studies due to the strengths of the search strategy. We were, however, unable to obtain the data we needed from study authors when it had not been included in the published papers.

We have excluded some studies which report cognitive effects of treatments of Type 2 diabetes by deciding to exclude interventions that are not recommended by international CPGs (Sotaniemi 1995; Qiang 2015). However, we consider this exclusion to be important in order to represent usual clinical practice.

Agreements and disagreements with other studies or reviews

The question whether Type 2 diabetes treatments could have some positive effect over cognition or decrease the incidence of dementia in randomised clinical trials has, to our knowledge, been evaluated only by one other meta‐analysis (Tuligenga 2015). Tuligenga's review focused on intensive glycaemic control versus standard glycaemic control and the search was up to date until October 2014 in MEDLINE and ALOIS. It included five studies, of which we included only two. The reason why we did not include the Cukierman‐Yaffe 2014 study is because this study was not only focused on Type 2 diabetes but included participants with impaired glucose tolerance and impaired fasting glucose. Data for Type 2 diabetes patients were not reported separately. We also excluded Weinstock 2011 because the intervention — telemedicine — is not an approved intervention by international CPGs for Type 2 diabetes; and Koekkoek 2012, which did not assess cognition at baseline. Tuligenga 2015 also found no differences in cognitive function between the intensive glycaemic control group and the standard glycaemic control group, in agreement with our results for this comparison.

Characteristics of included studies [ordered by study ID]

Characteristics of excluded studies [ordered by study ID]

Characteristics of ongoing studies [ordered by study ID]

Internal sources

• No sources of support supplied

External sources

• NIHR, UK.

  This update was supported by the National Institute for Health Research (NIHR), via a Cochrane Programme Grant to the Cochrane Dementia and Cognitive Improvement group. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the Systematic Reviews Programme, NIHR, National Health Service or the Department of Health