Dementia is one of the most prevalent neurodegenerative diseases globally (Cao et al., 2020). In 2020, over 50 million people had dementia worldwide, and this number is expected to rise to 152 million by 2050 (Barbarino et al., 2020). Dementia is a leading cause of morbidity, disability, and mortality (Witthaus et al., 1999; Noale et al., 2003; Todd et al., 2013). Alzheimer’s disease (AD) was the sixth greatest cause of death in the United States in 2017 (Kramarow and Tejada-Vera, 2019). Moreover, the total annual global cost is estimated to be the US $2 trillion by 2030 (Barbarino et al., 2020). In this regard, the risk factors such as biological- and lifestyle behavior-related factors are being added by new evidence support consistently (Livingston et al., 2020). Recent studies have indicated that dementia is associated with gut microbiota (Angelucci et al., 2019; Liu et al., 2019; Askarova et al., 2020; Łuc et al., 2021).

Gut microbiota consists of the aggregate of all microorganisms that affect essential human functions (Sekirov et al., 2010; Lange et al., 2016). Specifically, the gut microbiota–brain axis is a bidirectional communication network between the central nervous system (CNS) and the enteric nervous system (ENS) that maintains the homeostasis of the CNS, gastrointestinal, and microbial systems (Carabotti et al., 2015; Kowalski and Mulak, 2019; Morais et al., 2021). In the presence of dysbiosis, the products of pathogenic microorganisms can cross the blood–brain barrier (BBB), and peripheral immune cells activate glial cells and the neuroinflammatory pathway (Doifode et al., 2020). This interplay induces amyloid-beta burden and tau accumulation, triggering neurodegeneration (Sadigh-Eteghad et al., 2015; Doifode et al., 2020). Therefore, the dysbiosis of gut microbiota caused by antibiotics is considered to play an important role to enhance neurodegenerative disease severity (Askarova et al., 2020; Morais et al., 2021).

Previous studies about the relationship between antibiotics and dementia are insufficient. The gut microbiome is highly affected by an unhealthy lifestyle (such as smoking status, irregular physical activity, unhealthy diet, immoderate alcohol consumption, and sleep deprivation) and its significant changes have been reported in patients with AD (Lourida et al., 2019; Askarova et al., 2020). In particular, the changes in the intestinal microbiota were found in elderly patients after antibiotic therapy (Askarova et al., 2020). Moreover, the fecal microbial diversity of AD patients was markedly distinct from individuals with amnestic mild cognitive impairment or healthy controls (Liu et al., 2019). A recent study suggested that gut microbiota changes induced by the use of antibiotics such as streptozotocin and ampicillin can cause AD, whereas broad-spectrum antibiotics including gentamicin and vancomycin may be utilized to treat AD depending on their microbiome target (Zhu et al., 2020). Therefore, a better epidemiological comprehension of the relationship between antibiotics and dementia is needed. In particular, South Korea has a high rate of antibiotic usage, with 31.7 defined daily doses (DDD) per thousand people in 2016, compared to the OECD average of 23.7 (OECD, 2020).

This population-based retrospective cohort study was conducted to assess the association between antibiotic exposure and the risk of dementia incidence using the Korean National Health Insurance Service-Health Screening Cohort (NHIS-HEALS) database.

The baseline characteristics of participants are shown in Table 1. A total of 313,161 individuals were analyzed, with a mean age of 54.9 and a male distribution of 57.3%. The number of incident cases was 5,386 for overall dementia, 4,126 for AD, and 694 for VD. Compared to antibiotic non-users, antibiotic users tended to be older, smoke less, drink less alcohol, have lower income, and have higher CCI values. Among antibiotic users, long-term users tended to be older, smoke less, drink less alcohol, have higher CCI values, and have more antidepressant use.

The association between cumulative antibiotic prescription days and the risk of developing overall dementia, AD, and VD is depicted in Figure 2 and Supplementary Table S1. Compared to antibiotic non-users, antibiotic users, including users of 1–30, 31–90, 91 or more cumulative prescription days, had an increased risk of overall dementia (aHR, 1.09; 95% CI, 1.01–1.17 for 1–30 days, aHR, 1.23; 95% CI, 1.12–1.36 for 31–90 days, and aHR, 1.44; 95% CI, 1.19–1.74 for 91 or more days), AD (aHR, 1.12; 95% CI, 1.03–1.22 for 1–30 days, aHR, 1.25; 95% CI, 1.12–1.40 for 31–90 days, and aHR, 1.46; 95% CI, 1.17–1.81 for 91, or more days), and VD (aHR, 0.93; 95% CI, 0.77–1.12 for 1–30 days, aHR, 1.24; 95% CI, 0.96–1.59 for 31–90 days, and aHR, 1.38; 95% CI, 0.83–2.30 for 91 or more days) in a duration-responsive manner. There was a significant trend of risk increase for overall dementia and AD upon increasing levels of cumulative days of antibiotic prescription from non-users to users. The second model, additionally adjusted for infectious diseases, also showed a clear duration-responsive relationship between cumulative antibiotic prescription days with overall dementia (aHR, 1.13; 95% CI, 1.04–1.22 for 31–90 days and aHR, 1.32; 95% CI, 1.10–1.58 for 91 or more days), AD (aHR, 1.11; 95% CI, 1.02–1.22 for 31–90 days and aHR, 1.29; 95% CI, 1.05–1.59 for 91 or more days), and VD (aHR, 1.33; 95% CI, 1.08–1.65 for 31–90 days and aHR, 1.50; 95% CI, 0.92–2.44 for 91 or more days). The aHR for VD in antibiotic users for 91 or more days was not significant, but the trend for the increase of VD risk with increasing cumulative prescription days appeared significant (p for trend = 0.003). Therefore, a significant trend of risk increase was observed for overall dementia, AD, and VD upon increasing levels of cumulative prescription from short-term to long-term users.

Table 2 reports the stratified analysis according to subgroups of covariates for the association of antibiotic use with dementia. There was no interaction term by age, sex, BMI, smoking status, alcohol consumption, physical activity, household income, CCI, and antidepressant use.

In Table 3, we also conducted analyses by excluding 1-, 2-, and 3-year latent periods. The aHR for overall dementia in 91 or more days of antibiotic use was 1.53 (95% CI, 1.26–1.87) compared to the non-user group in the 3-year wash-out period analysis. Moreover, the aHR for AD in 91 or more days of antibiotic use was 1.56 (95% CI, 1.24–1.96). After shifting the index date from 1 January 2006 to 1 January 2007, those with antibiotic users for 91 or more days had a higher risk of overall dementia (aHR, 1.29; 95% CI, 1.12–1.49) and AD (aHR, 1.34; 95% CI, 1.13–1.58) compared to antibiotic non-users. Those with antibiotic users for 91 or more days had a higher risk of overall dementia (aHR, 1.29; 95% CI, 1.14–1.45) and AD (aHR, 1.27; 95% CI, 1.10–1.45) compared to those with antibiotic users for 1–30 days.

The effect of the number of antibiotic classes prescribed on dementia incidence is demonstrated in Table 4. Using the non-user group as a reference, we revealed a positive association between the number of antibiotic classes and dementia including overall dementia and AD (p for trend <0.001). Furthermore, Supplementary Table S2 shows dementia risk to cephalosporin exposure. Both overall dementia (aHR, 1.14; 95% CI, 1.02–1.28) and AD (aHR, 1.17; 95% CI, 1.03–1.33) risks were notably higher in cephalosporin users compared to non-users. Supplementary Table S3 and Supplementary Table S4 show examples of each antibiotic class and the sources of infection based on ICD-10 codes.

Supplementary Tables S5, S6 depict the descriptive characteristics of the study population and the risk for dementia according to antibiotic exposure after propensity score matching, respectively. Antibiotic users had a higher risk of overall dementia and AD than non-users.

In this nationally representative cohort study, antibiotic use was associated with the risk of dementia incidence. There was a cumulative duration-dependent relationship between dementia incidences with cumulative days of antibiotic exposure. The results were significant after adjustments for all covariates. Furthermore, the results of the propensity score matching analysis verified our primary findings. To our knowledge, this is the first study to indicate that longer cumulative days of antibiotic exposure could lead to significantly higher dementia risk.

Previous studies showed that antibiotics can induce alterations in the gut microbiota, leading to AD (Angelucci et al., 2019; Liu et al., 2019). In a Chinese cohort, it has been demonstrated that the diversity of intestinal microbiota was significantly lower in AD patients (Liu et al., 2019). Enterobacteriaceae was enriched in AD patients compared with amnestic mild cognitive impairment patients and healthy control groups (Liu et al., 2019). In our study, antibiotic use was significantly correlated with dementia, and this could have been caused by the change in gut microbiota. Another previous study demonstrated that unhealthy lifestyle aspects can affect the gut microbiome (Pushpanathan et al., 2019). It has been suggested that a sedentary lifestyle can increase pro-inflammatory bacteria and the levels of Aβ in the brain, gradually causing a higher risk of AD (Askarova et al., 2020). Therefore, as lifestyle behavior could be a confounding factor, we included representative factors of lifestyle, such as smoking status (Anstey et al., 2007; Chen et al., 2009), alcohol consumption (Chen et al., 2009), and physical activity (Chen et al., 2009) as covariates. Our results of the stratified analyses on subgroups of these factors were consistent in all three subgroups, showing that lifestyle behaviors did not influence the effect of antibiotics on dementia.

Dementia is a chronic neurodegenerative disease with clinical characteristics such as slow onset and gradual progression (Arvanitakis et al., 2019; Dementia, 2019). In particular, dementia risk is linked to numerous diseases such as infections and head injury (van der Flier and Scheltens, 2005; Chen et al., 2009). Therefore, there is a possibility that antibiotics were prescribed due to the prognostic symptoms before the diagnosis of dementia. For instance, during the clinical course of dementia, the prevalence of dysphagia is high (Easterling and Robbins, 2008; Suh et al., 2009). It is likely to predispose individuals to aspiration pneumonia by dysphagia (Easterling and Robbins, 2008) and this may cause antibiotic prescription. In this case, antibiotics would have been prescribed for dementia, and it is difficult to assert that antibiotics induced dementia. Therefore, we considered CCI as a confounding factor and there was no significant interaction in a stratified analysis. Furthermore, we additionally adjusted for infectious diseases for analyses in the antibiotic user group considering the possibility of an association between infection and dementia incidence (Dunn et al., 2005). We used only one main principal diagnosis for infectious diseases to conduct accurate analyses. Thus, our results show that antibiotic prescription is an independent risk factor for future dementia risk.

Multiple mechanisms of the association between antibiotics and dementia risk have been postulated. First, the gut microbiota–brain axis is considered a crucial way that mediates the relationship between gut dysbiosis by antibiotics and dementia. A recent study has suggested that antibiotics can cause neuroinflammation and neurodegeneration through the gut microbiota–brain axis, favoring AD or worsening its course (Angelucci et al., 2019). Another study suggested the role of microbiota in the physiological and biological basis of neurodegenerative disorders in the gut–brain axis (Dinan and Cryan, 2017). Antibiotic exposure leads to short- and/or long-term changes in the gut microbiota such as reduced colonization resistance against pathogens, loss of diversity (Lange et al., 2016; Angelucci et al., 2019), and changes in brain chemistry (Bienenstock et al., 2015; Angelucci et al., 2019). The accumulation of antibiotics fosters bacterial resistance, and the human microbiome has become a reservoir of resistance genes (Francino, 2016). This causes prolonged dysbiosis of gut microbiota, explaining the increased risk of dementia due to long-term antibiotic use (Francino, 2016). Therefore, prolonged antibiotic use may cause a profound change in the gut–brain axis profoundly. Microbial exudates and amyloid-enhancing factors caused by impaired gut permeability and microbial dysbiosis enter the circulatory system, and this induces an abnormal level of pro-inflammatory cytokines (Doifode et al., 2020). These signals can cross the BBB, and peripheral immune cells activate glial cells and the neuroinflammatory pathways (Doifode et al., 2020). This process triggers tau accumulation and amyloid-beta burden that causes neurodegeneration (Doifode et al., 2020). Amyloid in the brain can activate microglia and astrocytes, resulting in inflammation (Doifode et al., 2020). Second, it has been demonstrated that overall antibiotics can cause neurotoxicity (Bhattacharyya et al., 2014), leading to encephalopathy (Bhattacharyya et al., 2016). As encephalopathy can present various symptoms including memory loss and dementia, this finding can partly explain the link between antibiotics and dementia. In particular, it is known that β-lactams such as penicillins and cephalosporins impede inhibitory neurotransmission at the ligand-gated ion channel γ-aminobutyric acid class A receptor (GABA_AR), causing central excitotoxicity (Bhattacharyya et al., 2016). In our study, only cephalosporins among antibiotic classes were associated with a higher risk of dementia. In this regard, it is known that cephalosporins can induce nephrotoxicity in larger doses (Yılmaz and Özcengiz, 2017). It is becoming explicit that kidney diseases are affected by changes in the gut microbiota which produces excessive amounts of uremic toxins such as indoxyl sulfate but fewer renoprotective metabolites (Mahmoodpoor et al., 2017; Yang et al., 2018). Furthermore, in a cohort study with 3,207 participants, chronic kidney disease was significantly associated with dementia (Sasaki et al., 2011). Thus, gut microbiota dysbiosis by antibiotic exposure directly or indirectly may affect the incidence of dementia.

Our study had several limitations. First, the study population was confined to participants who underwent a national health examination. There is a possibility of selection bias that participants in a bad health have been excluded. Second, the operational definition of dementia was based on the claim data. Therefore, it has a validation problem about the real diagnoses although we defined the outcome by diagnosis and prescription concurrently. Thus, future studies with medical chart reviews are required. Another limitation concerns the follow-up duration of 8 years, which may not be long enough to fully disclose the effects of antibiotics on the risk of dementia. As dementia usually develops slowly with gradual cognitive decline (Arvanitakis et al., 2019; Dementia, 2019), future studies with a longer follow-up period are required to fully ascertain the conclusions of this research. Fourth, antibiotic exposure was defined by prescription. There is a possibility that some participants did not take antibiotics on the actual prescription. Moreover, information about lifestyle behaviors was collected from a self-reported survey. Thus, there is a possibility that the survey data had insufficient and unclear information. Furthermore, we could not consider the education level, which may be a primary risk factor (Chen et al., 2009), due to the lack of information in the database. Instead, we adjusted for the income level, which may be a surrogate marker of education. Furthermore, information about the exact changes in gut microbiota by antibiotics could not be investigated. In future studies, it is necessary to analyze its changes at both clinical and experimental levels considering factors that can change the gut microbiota. Finally, this study was retrospective cohort research. Therefore, this study does not prove causation but rather indicates an association between antibiotic exposure and dementia risk, even though we attempted to control for possible confounders and biases.

Despite these limitations, our study’s strengths include the ability to assess the association of antibiotic exposure with dementia risk in older adults. First, the NHIS-HEALS database was based on the medical claim records collected at the nationally representative level. It contains individual-level information about demographic data, health examinations, diagnoses, and prescriptions. Furthermore, a large study population of 313,161 may strengthen the generalization of our results. Second, we conducted the association between antibiotic exposure and dementia incidence by regarding various risk factors such as age, household income, CCI, and infectious diseases. Third, we conducted sensitivity analysis and propensity score matching analysis to determine the robustness of our findings. Analyses that considered latent periods to address reverse causality also showed consistent results.

Antibiotic exposure appears to be associated with greater dementia incidence in our cohort study. There was a cumulative duration-dependent association between antibiotic exposure and dementia risk. However, it is possible that individuals who were long-term prescribed antibiotics had a poor health status, which increases their vulnerability to dementia. In other words, antibiotic exposure might be a proxy for dementia risk factors that we could not consider in this study, not a direct cause of dementia. This study suggests that antibiotic exposure could be one of several risk factors for dementia. Ultimately, future prospective and epidemiological studies with a broader study population that investigate the association between changes in gut microbiota by antibiotic exposure and dementia incidence are needed to validate the interpretation of our findings.