Type 2 diabetes mellitus (T2DM) is a common metabolic disorder that currently affects approximately 415 million people worldwide, and the prevalence is increasing (International Diabetes Federation, 2015). Patients with T2DM have an increased risk of dementia (Koekkoek et al., 2015), but the neuropathophysiological mechanisms underlying T2DM-related cognitive impairment are not entirely clear. In addition to studies that only rely on neuropsychological performance and experiments based on molecular biology, neuroimaging techniques provide important clues about brain structure, function and metabolism for documenting neurological involvement in T2DM.

Because resting-state functional magnetic resonance imaging (fMRI) is noninvasive and task-free, it has been widely used to investigate the pathogenesis of various neuropsychiatric disorders. Previous resting-state fMRI studies have identified the subtle brain changes located in widespread cortical and subcortical regions that are observed in T2DM patients, but these studies have reported relatively inconsistent results. For instance, several researchers observed reduced neural activity in the cerebellum of T2DM patients (Wang et al., 2014; Peng et al., 2016). In contrast, another study found increased activity in this region (Xia et al., 2013). Furthermore, only one study showed diminished activity in the precentral gyrus (Zhou et al., 2014). These reported discrepancies can potentially be attributed to the limited sample size, variable clinical demographics and use of different methods across studies. Thus, an urgent need exists for a meta-analysis to provide concordant results concerning the role of resting-state anomalies as common markers in T2DM patients.

As a coordinate-based meta-analytic method for meta-analyses of neuroimaging literature, activation likelihood estimation (ALE) can identify brain locations showing a consistent response across multiple experiments. This approach is based on the collection of peak coordinates from each study included in the meta-analysis rather than the input of raw images (Turkeltaub et al., 2002; Laird et al., 2005; Eickhoff et al., 2009). To date, the ALE technique has been successfully applied to neuroimaging studies of various neurological and psychiatric disorders (Shao et al., 2015; Weng et al., 2015). Hence, our present study aimed to quantitatively conduct a meta-analysis using the ALE algorithm to determine the resting-state brain anomalies underlying T2DM.

We identified nine eligible studies (Xia et al., 2013, 2015a,b; Cui et al., 2014, 2015, 2016; Wang et al., 2014; Zhou et al., 2014; Peng et al., 2016) for the meta-analysis according to the search criteria mentioned above. In one study, the analysis was performed based on two different subgroups of T2DM patients who were then compared with the same healthy control group (Peng et al., 2016). In another study, the analysis was performed with two different methodologies (Wang et al., 2014). Therefore, we treated these studies as unique reports, with each patient subgroup included independently in the meta-analysis; therefore, a total of 11 datasets were ultimately included in the meta-analysis. Figure 1 shows the details of the literature search strategy and data extraction process. The clinical and demographic data of the participants from all recruited studies are presented in Table 1. The patients and controls from each study were generally matched by age, gender and education.

As illustrated in Figure 2, a total of 73 peak foci were reported in this meta-analysis. Compared to healthy controls, T2DM patients had widespread reduced resting-sate neural activity in the whole brain, including the bilateral lingual gyrus, left postcentral gyrus, right inferior temporal gyrus, right cerebellar culmen, right insula and right posterior cingulate cortex (PCC). However, enhancements were also observed in the right precuneus and left superior frontal gyrus. Table 2 displays the coordinates of the cluster maxima.

The current study is the first meta-analysis to explore resting-state brain abnormalities in T2DM patients. By analyzing nine eligible studies, this meta-analysis identified consistent regions of resting-state activity brain anomalies across almost the entire brain of T2DM patients, including the frontal, parietal, sensorimotor, temporal, occipital and insular cortices. Therefore, T2DM-related neuropsychiatric impairment is believed to be correlated with diffuse aberrant resting-state brain activity involving both cortical and subcortical structures, and these findings will allow for a more comprehensive understanding of neuropathological mechanisms of this disorder.

Consistent with previous brain perfusion studies without resultant coordinates (Novak et al., 2006), parietal and occipital regions were found to be significantly altered in this meta-analysis. As a key region of the visual cortex, the occipital gyrus is the dominant region that exhibited remarkable reduction of resting-state neural activity in T2DM patients, consistent with the idea that visuospatial dysfunction is a common manifestation in T2DM patients (Moran et al., 2013). Worse visuospatial performance has been observed in T2DM patients than in their matched controls (Bangen et al., 2015), which could be due to the deficits in this area. Furthermore, the postcentral gyrus is known to be the main sensory receptive region for the sense of touch, proprioception, pain and temperature, which could contribute to common diabetic limb pain and temperature sensation decline (Cauda et al., 2009). In fact, impaired sensorimotor function has been described in T2DM patients with peripheral neuropathy (Menz et al., 2004). Nevertheless, the cause-and-effect relationship between deactivation in the postcentral gyrus and diabetic neuropathy cannot be explained at this stage because all the included studies were cross-sectional.

The temporal lobe has been linked to memory, verbal fluency, language processing and speech production, which are impaired in T2DM patients (McCrimmon et al., 2012). T2DM patients showed an abnormal pattern in the inferior temporal gyrus in the current study, which is also consistent with a previous voxel-based morphometry (VBM) study that revealed that T2DM patients exhibited gray matter atrophy in the right temporal lobe (Chen et al., 2012). Although it has been assumed that functional activity changes could precede or lead to subtle brain structural abnormalities, the relationship between functional and structural alterations should be further clarified by longitudinal studies. Several hypothesis-driven MRI studies have also demonstrated the vulnerability of the temporal lobe in T2DM patients. By seeding the PCC during a resting-state fMRI scan, two different research groups reported decreased functional connectivity within the temporal lobe (Musen et al., 2012; Chen et al., 2014). Similarly, by seeding the bilateral hippocampus, abnormal connections within the temporal lobe have also been found. A subtle effect of T2DM on cortical atrophy has also been observed in this area (Brundel et al., 2010). These aforementioned findings provide evidence of disrupted temporal functioning in T2DM patients and suggest that temporal dysfunction should be evaluated when cognitive decline occurs in T2DM patients.

Reduced resting-state brain activity in the cerebellum and insula may also contribute to cognitive deficits in T2DM patients. The cerebellum controls not only motor coordination but also advanced cognitive function (Marklund et al., 2007). Given that the salience network has been found to be impaired in T2DM patients (Chen et al., 2015), and the insula acts as an integral area in the salience network (Menon and Uddin, 2010), decreased insula activity could lead to impaired generation of appropriate behavioral responses to salient stimuli in these patients. Interestingly, according to the results, the right hemisphere of T2DM patients is more seriously impaired than the left hemisphere. Likewise, the effects of T2DM on the cortical surface and volume were only significantly pronounced in the right hemisphere (Brundel et al., 2010). The right hemisphere has been determined to be dominant over the left in geometric analysis and visual-spatial perceptual functioning, including the perception of distance, direction, shape, position, orientation and the detection of complex and hidden figures (Joseph, 1988); therefore, visual-spatial perceptual disturbances, which have been described in T2DM-related cognitive decline, could arise due to right hemisphere impairment. Coincidentally, according to some results extracted from the studies included in this meta-analysis, performance on one neuropsychological test that reflects visual-spatial function, the Rey-Osterrieth Complex Figure Test, has been observed to be correlated with the neural activity in the right brain region (Xia et al., 2015a). Nonetheless, because most studies in this field observed brain alterations on both sides in T2DM patients, the lateralization of impairment of brain activity in T2DM patients requires further investigation.

An additional finding of the present meta-analysis was hyperactivity in the frontoparietal cortex. The precuneus is involved in visuospatial function and working memory, and the frontal lobe is considered as a critical region involved in executive function and attention that subserves cognitive and emotional processing (Radua et al., 2012). These enhanced activities could act in a compensatory or recruitment capacity in T2DM patients such that their cognitive performance could be retained to some extent.

Several limitations should be noted in this meta-analysis. First, the number of included studies was relatively small. Second, ALE can provide excellent control of false positive results, but false negatives are more difficult to avoid (Radua et al., 2012). Finally, heterogeneity of the included studies, such as the demographics of the patients and different imaging modalities that represented different aspects of resting-state abnormalities, could not be entirely excluded. For example, ALFF and ReHo are relevant to the intensity and temporal synchronization of regional spontaneous neuronal activity in the whole brain, respectively (Biswal et al., 1995; Zang et al., 2004). ICA decomposes the entire resting-state BOLD data set into spatially distributed components and automatically identifies meaningful brain networks (Mantini et al., 2007), and ASL perfusion MRI measures regional cerebral blood flow and reflects regional brain metabolism and neural activity (Detre and Alsop, 1999). Despite these differences, various analytic approaches might be complementary to each other and provide more comprehensive information. Similar patterns of resting-state anomalies could also be detected by different analytic modalities. For instance, reduced resting-state activity in the occipital lobe was found in most of our included studies, regardless of the imaging modality.

In summary, this meta-analysis demonstrated robust resting-state brain anomalies in T2DM patients compared to healthy controls that were widespread in the frontal, parietal, sensorimotor, temporal, occipital and insular cortex. Further large, multicenter, resting-state fMRI studies are required to comprehensively investigate whether this abnormal activity pattern is a valuable diagnostic and prognostic biomarker for T2DM-related neuropsychiatric impairment.

WX and Y-CC performed the analysis and wrote the manuscript. JM designed the meta-analysis and revised the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.