Dementia is the most common neurodegenerative disease in the world, and the reported incidence and prevalence are around 8.6/1,000 person-years and 5.1%, respectively (Ponjoan et al., 2019). The classification of dementia depends on the neuropathological signs, but such features are usually overlapping and presenting as the spectrum (Raz et al., 2016). Demented patients have poor memory, impaired executive function, psycho-behavioral problems, and higher morbidity and mortality (Wolters and Ikram, 2018). Due to worldwide aging, the dementia population is gradually increasing. However, dementia remains incurable at present times. Only early intervention seems to delay dementia progression and prevent disability. Nevertheless, there is still a lack of standard guidelines in clinical practice on how to identify patients at risk of dementia early.

In addition to environmental and socioeconomic status (Jia et al., 2020), genetics is also an important risk factor for dementia. Apolipoprotein E (ApoE) polymorphism is the most susceptible genetic factor for late-onset Alzheimer’s disease (AD), which is located on chromosome 19, composed of 299 amino acids (Mahley and Rall, 2000). The amino acids at positions 112 and 158 determine three alleles (ε2, ε3, and ε4) (Mahley and Rall, 2000). The ApoE genotypes present differential effects on lipid-binding ability, amyloid-β (A-β) aggregation and clearance, and the susceptibility of the blood-brain barrier (Liu et al., 2013), and are linked to a neurodegenerative process (Mahley, 2016), including AD, vascular dementia, and mixed dementia (Liu et al., 2012). The heterozygous ε4 allele increased the AD risk by 2.6–3.2 times (Farrer et al., 1997; Houlden et al., 1998; Harold et al., 2009), while the homozygous ε4 allele further increased the risk by 14.9 times in Caucasians (Farrer et al., 1997). The ε4 allele effect seemed to possess ethnic differences, presenting 2.7–3.5 times (Lai et al., 2003) and 5 times (Zheng et al., 2016) higher dementia risk in the Chinese population, respectively. Besides, it is unclear how age modifies the risk related to ApoE polymorphism. Only one western cohort study demonstrated the reverse U-shaped dementia risk curve with the peak age at 70–75 years in both ε4 homozygotes and heterozygotes (Bonham et al., 2016). However, whether the Asian population has the same risk curve of ApoE polymorphism has not been disclosed.

Dementia subjects carry around two to eight chronic medical illnesses (Sanderson et al., 2002). Headache (Wang et al., 2018), ischemic stroke (Vijayan and Reddy, 2016), coronary artery disease (CAD) (Deckers et al., 2017), depression (Bennett and Thomas, 2014), sleep disturbance (Shi et al., 2018), fibromyalgia (Tzeng et al., 2018), and epilepsy (Sen et al., 2018) have been described as common comorbidities of dementia. In Taiwan’s cross-sectional study (Chen et al., 2017), diabetes mellitus (DM), cerebrovascular disease, liver cirrhosis, and asthma were found to be highly associated with cognitive decline. These chronic diseases seemed to have a bidirectional relationship with dementia because they might share similar risk factors, biological pathways, and environmental backgrounds. Besides, the ApoE genotype also affects the susceptibility to develop these illnesses (Afroze et al., 2016; Konialis et al., 2016; Burns et al., 2020). Oxidative stress, neuroinflammation, vascular risk factors, and blood-brain barrier breakdown are responsible for the development of both dementia and associated comorbidities (Raz et al., 2016). The relationship among ApoE, dementia, and comorbidities seems to be complex. How comorbidities interact with and influence cognitive function in the timeline has not been well-established. The role of these diseases, acting as a prodrome or as a predictor of dementia, remains unclear.

Therefore, we used a 3-year longitudinal case-control study to explore the interaction between ApoE polymorphism and clinical presentations. This is the first study that combined genetic and clinical information to identify the riskiest comorbidities to predict cognitive decline. Thus, the occurrence of such diseases in the subjects can be considered a red flag or a marker to inform clinicians to screen the cognitive function within a certain age population.

This hospital-based case-control study demonstrated the impact of specific diseases/events and the ApoE polymorphism on dementia occurrence. This is the first study to explore how clinical and genetic factors affect dementia risk individually and collectively. We combined the ε4 allele and comorbidities as factors to establish a model to predict the occurrence of dementia within 3 years. Therefore, it allows us to screen for dementia and provide early intervention once these risky comorbidities occur in ε4 carriers.

Among the studied risk factors, the ε4 allele of the ApoE genotype is the persistently significant factor to predict dementia throughout all age groups. Our ε4 allele carriers had around two times higher risk for all-cause dementia within 3 years compared with ε3/ε3 carriers, which seemed to be relatively lower compared with previous case-control studies. The OR for subjects carrying at least one ε4 allele was 3.84 for late-onset AD and 1.70 for vascular dementia in Han Chinese (Yang et al., 2001). In Caucasians, the odds ratio for AD was 3.2 in ε3/ε4 and 14.9 in ε4/ε4 compared with the ε3/ε3 allele (Farrer et al., 1997). In contrast to such studies, our non-demented controls were selected, enrolling age- and gender-matched subjects, instead of only non-demented cases. This method avoided the confounding effect of age since the dementia group might be older.

After age stratification, the risk of dementia in ε3/ε4 carriers only emerged after 65 years. However, the ε4/ε4 carriers had a persistently higher dementia risk in all age groups. This phenomenon suggested the additive effect of the ε4 allele, for which the number of ε4 alleles is correlated with the dementia risk and the onset age (Liu et al., 2013). Although ApoE polymorphism had been told to show extensive influence from the age of 40–90 years (Farrer et al., 1997), the strength of the risk seemed to change with age. Instead of a reverse U-shaped risk curve for dementia observed in the western literature (Bonham et al., 2016), our heterozygous and homozygous ε4 carriers demonstrated separate trends of the age-dependent risk. The ε3/ε4 carriers presented a slowly upgoing line. In contrast, we found that the dementia risk seemed to reach the peak at the age of 65–75 years in ε4/ε4 carriers, and then the genetic effect declined after 75 years. However, because only 19 dementia cases were diagnosed in our ε4/ε4 carriers, we could not confirm whether the presence of the broken line in homozygous ε4 subjects was due to a smaller sample size. The ε3/ε4 and the ε4/ε4 had different routes for dementia occurrence in Han Chinese. The strategy of risk assessment for dementia screening could not be the same for these two groups. The ε4 homozygotes need more active surveillance and earlier prevention.

The dementia subjects were reported to carry around 2.4 comorbidities (Schubert et al., 2006), and we found 0.6 comorbidities occurred within 3 years of dementia being diagnosed. Among the listed comorbidities, CVA, sleep disorders, and functional gastrointestinal disorders seemed to be the important risk factors to predict dementia occurrence within 3 years.

Occurrence of CVA or stroke was found to have 1.7–5.0 times of risk for cognitive decline. Stroke has been considered the major risk factor for vascular dementia and AD for a long time (Vijayan and Reddy, 2016). The risk of post-ischemic stroke dementia was 1.7–3.8 (Desmond et al., 2002), and the risk was much higher in ε4 homozygotes (Pendlebury et al., 2020). The incidence of new-onset dementia gradually increased after the index stroke, reporting from 7% in the first year to 10% in the third year, and 23% in the tenth year (Zhu et al., 2000). Stroke leads to intracranial hypoxia, provoking oxidative stress (Slevin et al., 2015), inflammation, and microRNA alteration, which then facilitate and accelerate A-β protein accumulation (Vijayan and Reddy, 2016), and finally, dementia pathogenesis (Brenowitz et al., 2021). We found that the post-stroke dementia risk is highest in the relatively younger group, which appeared in the reverse direction of the stroke incidence and prevalence that increases with age (Sedova et al., 2021). We postulated that the younger brain is more susceptible to the ischemic insult, which might be related to a more fulminant inflammatory response. ApoE ε4 also disturbs lipid homeostasis in astrocytes and microglia and then leads to blood-brain barrier failure in stroke patients (Duong et al., 2021). Furthermore, the ε4 allele is also related to a younger stroke age (Lagging et al., 2019) and a higher ischemic stroke risk (Konialis et al., 2016). We found that the ε4 allele and stroke would individually increase the dementia risk, but this risk was not summative when both factors were simultaneously present. Therefore, when subjects had the first event of a stroke, especially at a younger age, it is necessary to keep an eye on the subtle signs of dementia occurrence.

Sleep disturbance showed a 2.5–4.3-times higher risk for future cognitive decline while it was reported to increase the risk of dementia by 50–80% in previous studies (Shi et al., 2018; Brenowitz et al., 2021). On the molecular level, abnormal sleep duration is associated with amyloid-β accumulation (Winer et al., 2021) and sleep-disordered breathing increased tau deposition (Bubu et al., 2019). The dementia risk was also relatively higher in subjects aged less than 65 years in our study, reflecting that the younger brain is vulnerable to such abnormal protein accumulation. However, there was no evidence that ApoE genotypes have differential effects on sleep disorders (Palpatzis et al., 2021). This phenomenon explained our finding in Figure 1 that subjects who had only sleep disorders presented with a significantly higher dementia risk regardless of carrying the ε4 allele. Therefore, sleep disorder itself seems to jeopardize dementia risk more than the ApoE genotype does. Thus, sleep disturbance might be a potentially modifiable risk factor to prevent dementia, even if there has been insufficient evidence to confirm it yet.

Gastrointestinal dysfunctions, such as constipation and delayed gastric emptying, are associated with a 2.4–3.0-times higher dementia risk throughout all age groups in our study. Gut microbiota alteration occurred several years before the appearance of cognitive decline (Li et al., 2019), which indicated that the subjects had been exposed to systemic disturbances related to metabolites of the gut microbiota, such as lipopolysaccharide and trimethylamine. They influence the neurological system by affecting neurotransmitters (Zhuang et al., 2020), such as serotonin and GABA, and increasing amyloid-β deposition (Nagpal et al., 2019). Consequently, the theory of the gut-brain axis had been proposed in many neurodegenerative disorders (Martin et al., 2018). However, whether probiotic treatment will prevent dementia development after a change in the intestinal ecology is still unknown (Santiago and Potashkin, 2021). Besides, ApoE polymorphism is also associated with the compositions of the gut microbiome (Parikh et al., 2020). This phenomenon is also reflected in our observation. When ε4 allele carriers presented any functional gastrointestinal disorder, their risk for developing dementia within 3 years further increased, exceeding those having any one of the two factors. Among comorbidities, functional gastrointestinal disorders might act as an important red flag sign in ε4 allele carriers. Accordingly, even though gastrointestinal symptoms are usually non-specific, they might be used as a biomarker for early detection and prediction of cognitive decline in ε4 allele carriers in clinical practice.

Cerebrovascular accident, sleep disorders, functional gastrointestinal dysfunctions, and fibromyalgia seem to be the predictors of dementia with variable risk based on age, suggesting the susceptibility of the brain changes with age. Although we could not confirm the cause-effect relationship between dementia development and comorbidity occurrence, the interaction of ApoE polymorphism and these clinical illnesses certainly modifies the dementia risk. Combining the genetic and clinical information, we have the odds to early detect cognitive decline and optimize interventions when the subjects first present with a specific illness at a particular age and carry a specific ApoE allele.

There were some limitations in our study. First, the diagnosis of dementia and associated diseases was based on the medical records only from our hospital. We could miss the diagnoses which had been made in other hospitals. Second, the illness with a minor severity, such as headaches or gastrointestinal discomforts, could also be ignored if the subjects did not search for medical assistance. Underestimation of the risk related to these illnesses might be possible. Besides, instead of further classifying dementia subtypes, such as vascular dementia and AD, we took all-cause cognitive decline as the single group for analysis. Furthermore, our sample size was not large enough because only 600 dementia cases were identified. It made the statistical model unstable. We had the preservation to conclude the interaction between the ApoE genotype and comorbidities, as well as the dementia risk related to individual comorbidities in three age groups.

Our study demonstrated that dementia patients not only had a higher chance to carry the ε4 allele, but also had a higher prevalence of many chronic illnesses within 3 years before the cognitive decline, including hypertension, headache, epilepsy, CVA, CAD, sleep disorders, psychiatric disorders, functional gastrointestinal disorders, and fibromyalgia. Although ApoE polymorphism seems to be the riskiest factor for dementia, the effects of the number of ε4 alleles differ. The ε4 homozygotes presented a persistently high risk for dementia in all age groups, while the risk only emerged after 65 years in ε3/ε4 subjects. Among these comorbidities, the occurrence of CVA, sleep disorders, functional gastrointestinal disorders, and fibromyalgia seemed to be predictive of cognitive decline within 3 years. Besides, functional gastrointestinal disorders might be an important predicting factor for dementia occurrence in ε4 allele carriers.

The clinical data presented in this study are available on request from the corresponding authors. The genetic data from the Taiwan Precision Medicine Initiative are not publicly available because some access restrictions may apply to the data underlying the findings. The data used in this study cannot be made available in the manuscript, the supplemental files, or in a public repository due to the Personal Information Protection Act executed by Taiwan’s government, starting in 2012. Requests for data can be sent as a formal proposal to obtain approval from the ethics review committee of the appropriate governmental department in Taiwan.

The studies involving human participants were reviewed and approved by the Institutional Review Board of Taichung Veterans General Hospital (TCVGH) (CE20316A). The patients/participants provided their written informed consent to participate in this study.

HT, C-HL, Y-MC, and T-HH: work conception and study design. HT and W-JL: data acquisition and collection. T-HS and C-SL: conducting NGS and genomic data analysis. HT and L-SC: clinical data analysis. C-HL and Y-MC: interpretation of data. HT and T-HH: drafting the work. Y-YL and T-HH: revising the work for valuable intellectual content. HT, Y-YL, and T-HH: final approval of the version. All authors contributed to the article and approved the submitted version.