Diabetes mellitus (DM) has been reported to be a potential risk factor for cognitive impairment and dementia including Alzheimer’s disease (AD) (Koekkoek et al., 2015; Hu et al., 2020; Noguchi-Shinohara et al., 2022). Some human and preclinical studies have supported the idea that DM may cause cognitive impairment and dementia by impairing brain structure and metabolism, e.g., amyloid β (Aβ), tau, and glucose metabolism, as well as insulin signals and cerebrovascular dysfunction (Jayaraman and Pike, 2014; Gonzalez-Reyes et al., 2016; Shinohara and Sato, 2017; Sutherland et al., 2017; Beeri and Bendlin, 2020; Rasool et al., 2021). In particular, midlife or early-onset DM may hasten cognitive impairment and dementia in later life and could be more stronger associated with an increased risk of all dementia and AD dementia than late-onset DM (Mayor, 2014; Harvard Health Letter, 2015; Sadahiro et al., 2019; Wang and Liu, 2021).

Apolipoprotein E ε4 allele (APOE4) is the strongest risk gene for AD (Yamazaki et al., 2019) and affect the major pathway by which APOE4 increases the risk of AD by driving earlier and more abundant Aβ pathology in the brains of APOE4-positive older adults (Morris et al., 2010; Yamazaki et al., 2019). In addition, APOE4 may cause cognitive impairment and dementia by reducing functions in multiple brain homeostatic pathways, such as lipid transport, synaptic integrity and plasticity, glucose metabolism, and cerebrovascular function (Lowe et al., 2014; Yamazaki et al., 2019; Jeon et al., 2020).

Taken together, both major risk factors, i.e., DM and APOE4, may affect AD and cognitive dysfunction by sharing common pathways (Liu et al., 2013; Zhao et al., 2017; Shinohara et al., 2020). However, the interaction between DM and APOE4 as contributing to cognitive dysfunction and dementia has not yet been fully clarified. A cohort study showed a significant interaction between a DM history and the APOE genotype, in which any DM history affected the cognitive decline in APOE4-negative older adults, but not in APOE4-positive older adults (Shinohara et al., 2020). In contrast, a cross-sectional study on middle-aged to older adults revealed the modifying effect of the presence of APOE4 on the association between current DM and poor cognitive performance (Dore et al., 2009). Furthermore, a longitudinal study showed that APOE4 modifies the association between midlife DM and poor cognitive decline (Bangen et al., 2013). These conflicting results may be due to a difference in the DM period, i.e., midlife or late-life DM. Nevertheless, no study has yet investigated the interaction between DM and APOE4 in contributing to poor cognition and AD or dementia according to the presence of DM during the midlife and late-life period, respectively. Therefore, we examined the association between midlife and late-life DM and global cognition in non-demented older adults. We also explored the moderating effect of APOE4 on these associations. Additionally, we investigated the moderating effect of APOE4 on the associations between other vascular risks and global cognition for comparative purpose.

This study of older adults found that midlife DM worsened global cognition, i.e., the TS, which was moderated by the effect of APOE4 on the association between midlife DM and decreased global cognition. Previous studies have suggested that DM and APOE4 are major risk factors for AD and related cognitive decline, and there are many reports that each has a significant association with cognitive decline (Morris et al., 2010; Liu et al., 2013; Jayaraman and Pike, 2014; Lowe et al., 2014; Koekkoek et al., 2015; Gonzalez-Reyes et al., 2016; Shinohara and Sato, 2017; Sutherland et al., 2017; Zhao et al., 2017; Yamazaki et al., 2019; Beeri and Bendlin, 2020; Hu et al., 2020; Jeon et al., 2020; Shinohara et al., 2020; Rasool et al., 2021; Noguchi-Shinohara et al., 2022). The coexistence of these two risk factors is thought to have a more devastating effect on AD and related cognitive decline, but few studies have examined the interaction between the two. In addition, midlife and late-life DM may have different effects on cognitive function (Mayor, 2014; Harvard Health Letter, 2015; Sadahiro et al., 2019; Wang and Liu, 2021) in addition to the moderation effect of APOE4. Nevertheless, no study has examined DM, APOE4, and the DM interaction by dividing the exposure to DM into middle age and older age, respectively.

This is the first study to demonstrate that APOE4 has a significant moderating effect on the association between DM and cognitive decline in middle age, but not in older age. This may be because participants with midlife DM have a longer duration of exposure to DM and are more likely to exacerbate the progression than those with late-life DM, including late-onset DM. These findings are consistent with the results of previous studies regarding the stronger effect of early-onset DM on the risk of AD dementia and cognition than late-onset DM (Mayor, 2014; Harvard Health Letter, 2015; Sadahiro et al., 2019; Wang and Liu, 2021). Moreover, this midlife DM is likely to have an interacting or synergistic effect in APOE4-positive older adults who may be susceptible to AD or related cognitive decline. Contrary to the results of our study, a longitudinal study demonstrated that non-demented older adults aged 65 years and over who have late-life DM and APOE4 are at a substantially higher risk of cognitive decline (Haan et al., 1999). Although their findings did not confirm the presence or absence of DM in the middle age, it seems similar to our findings on the DM-APOE4 interaction for cognitive decline when the proportion of DM that starts in middle age, i.e., midlife DM, is high among older adults with late-life DM. A cross-sectional study with middle to older adults showed a modifying effect of the presence of APOE4 on the association between current DM and poor cognitive performance (Dore et al., 2009). Their study included a significant number of middle-aged as well as older adults, so it is likely that their findings were similar to ours due to the significant proportion of middle-aged onset DM. Taken together, these conflicting results may be due to differences in the DM exposure period, i.e., midlife and late-life DM, as well as the proportion of participants with midlife DM.

In the present study, this moderating effect was only seen in the APOE4-positive group, and not in the APOE4-negative group. The majority of the findings of previous studies support our results (Haan et al., 1999; Dore et al., 2009; Bangen et al., 2013). However, a cohort study suggested that this moderating effect appeared in the APOE4-negative group, and commented that such effects may be masked by the presence of APOE4, which itself increases the risk of vascular problems (Shinohara et al., 2020). The difference may be due to the DM exposure period and the accuracy of cognitive assessment scales. Unlike our study, they defined the presence of DM as any DM history and used relatively simple scales, such as the CDR or mini-mental state examination (MMSE). In contrast, we defined DM as midlife or late-life DM and used the TS of the CERAD, which has much broader cognitive domains and scores with a wider range than CDR or MMSE, to reflect minimal changes in cognition over the global domains. Other possibilities include differences in the ratio and combination of APOE2 and 3 alleles in the APOE4-negative group and differences in clinical expression according to race (Barnes and Bennett, 2015).

The precise mechanism by which midlife DM and APOE4 interact to increase AD and cognitive dysfunction is unclear. As one of the most probable hypotheses, that accumulation of advanced glycation end products (Vistoli et al., 2013), a biomarker implicated in the development and progression of oxidative-based diseases such as DM (Yamagishi et al., 2012) and AD (Sasaki et al., 1998; Grillo and Colombatto, 2008; Li et al., 2012), may play a role in neurodegeneration given the co-localization of advanced glycation end products in the Aβ and tau pathology in AD (Sasaki et al., 1998; Grillo and Colombatto, 2008; Li et al., 2012). In addition, APOE4 is associated with a threefold greater advanced glycation end products-binding activity compared to the APOE3 isoform indicating that advanced glycation end products caused by APOE4 interactions may contribute to the formation of Aβ and tau deposition (Li and Dickson, 1997). As other mechanisms, insulin-degrading enzyme (Kurochkin et al., 2018), insulin resistance (Sedzikowska and Szablewski, 2021), and acylated glucagon-like peptide 1 (GLP-1) analog have been proposed to link DM and AD, and the role of APOE4 between them is being investigated. In particular, liraglutide, a novel DM drug with a GLP-1 agonist action, is currently undergoing a clinical trial, i.e., the ELAD trial (Evaluating Liraglutide in Alzheimer’s Disease NCT01843075), and is being investigated for its benefits on changes in AD biomarkers and clinical measures in addition to age and APOE4 status (Femminella et al., 2020).

In this study, no interactions were detected between other vascular risks and APOE4 on global cognitive decline. A few studies have been published on the possible moderating effect of APOE4 on the association between other vascular risks [such as hypertension (Yasuno et al., 2012), atherosclerosis (Haan et al., 1999), and peripheral vascular disease (Haan et al., 1999)] and cognitive decline. As these vascular risks have the potential to interact with each other, systematic follow-up and repeat studies are needed to clarify the relationship.

To the best of our knowledge, this is the first study to show a moderating effect of APOE4 on the association between midlife DM and global cognitive decline. The present findings did not change even after controlling for potential confounders and late-life vascular risks. Our study had some limitations. First, as this was a cross-sectional study, causal relationships could not be inferred. Long-term prospective studies are needed. In addition, the small number of participants decreased statistical power, thus, contributing to the null result for the relationship between midlife DM, APOE4-positivity, and global cognition. A study with a larger sample size is required. Second, about 64% of the study participants were diagnosed with MCI, which may also raise some concerns about the accuracy of self-report for midlife-DM status. However, although participants with MCI have problems for their recent memory, their remote memory is very well-preserved (Leyhe et al., 2009). Therefore, it is not likely that participants with MCI reported their history for midlife DM status history more erroneously because the self-report for midlife DM status mainly depends on remote memory rather than recent memory. In addition, even when we controlled for the clinical diagnosis (CN vs. MCI) as an additional covariate in Model 3 (Tables 2, 3 and Supplementary Tables 2, 3), the results were still very similar. Additionally, the accuracy of the information of DM was verified by reliable informants or by reviewing available medical charts. Third, we did not statistically analyze the duration, severity, complications, or types of DM. However, we analyzed the interaction of DM with APOE4 on global cognition in middle age and older age, i.e., midlife DM and late-life DM, based on the evidence of two significant lifetime periods vulnerable to AD and related cognitive decline (Mayor, 2014; Harvard Health Letter, 2015; Koekkoek et al., 2015; Gottesman et al., 2017; Sadahiro et al., 2019; Hu et al., 2020; Wang and Liu, 2021; Noguchi-Shinohara et al., 2022). Nevertheless, there remains a possibility that treatment response (e.g., well-controlled DM or not) for presence and severity of hyperglycemia may affect our findings. We assessed fasting blood glucose levels, which may be considered as a biomarker for the presence and severity of hyperglycemia, and obtained the same results after controlled the fasting glucose level as an additional covariate (Supplementary Tables 4, 5). Lastly, we could not statistically analyze the dose effect of APOE4 for the moderating effect on DM and global cognition because only one subject with E4E4 was detected among all of the participants.

Our findings suggest that midlife DM increases the risk for AD and cognitive decline, and this risk is modulated by APOE4 status. To prevent AD and cognitive decline, physicians should check for the possible coexistence of midlife DM and APOE4-positive status.

The study data are not freely accessible because the IRB of the Hallym University Dongtan Sacred Heart Hospital prevents public sharing of such data for privacy reasons. However, the data are available on reasonable request after IRB approval. Requests for data access can be submitted to an independent administrative coordinator by e-mail (yoon4645@gmail.com).

The studies involving human participants were reviewed and approved by the Institutional Review Board of the Hallym University Dongtan Sacred Heart Hospital. The patients/participants provided their written informed consent to participate in this study.

JK conceived and designed the study, served as principal investigator, and supervised the study. BL, YC, G-HS, I-GC, JL, HK, JH, DY, and JK were involved in acquisition, or analysis and interpretation of the data and helped to draft the manuscript. JK, BL, YC, G-HS, I-GC, JL, JH, HK, and DY were major contributors in writing the manuscript and critically revising the manuscript for intellectual content. All authors read and approved the final manuscript.