Introduction

Dementia is a progressive disease that impairs cognitive abilities and behavior, significantly disrupting daily living. According to the World Health Organization, nearly 9.9 million people develop dementia globally each year, and with improved life expectancy, this figure is expected to rise. This increase drives up costs for governments, communities, and families, reduces economic productivity, and accounts for 11.9% of the years lived with disability1. Dementia of the Alzheimer type (DAT), characterized by the accumulation of amyloid plaques and neurofibrillary tangles in the brain leading to neurodegeneration, is the most common cause of dementia. The second most common cause is vascular dementia (VaD), which involves the contribution of vascular pathology to dementia and includes subtypes such as subcortical vascular dementia and multi-infarct dementia2.

The relationship between cancer and dementia has been a topic of interest for decades. Several population cohort studies and their meta-analyses consistently indicated a lower risk of DAT in various cancer types, including lung, renal, colorectal cancers, and lymphoma3,4,5,6,7,8,9. Moreover, studies using two-sample Mendelian randomization suggested causal relationships between cancer and DAT10,11. VaD has also been shown to have an inverse association with cancer in population studies, although the number of studies is limited12,13.

Based on these findings, it has been suggested that anticancer drugs may have protective effects on dementia. Although cancer and dementia appear to exhibit opposing cellular mechanisms, they share common pathways, including dysregulated cell cycles, reduced autophagy, increased inflammation, and chronic oxidative stress14,15,16,17,18. Consequently, anticancer drugs that modulate cancer pathophysiology have been examined as factors that may influence dementia risk. For example, a few population studies have found that chemotherapy, as an unclassified group, decreased the incidence of Alzheimer's disease and VaD in colorectal cancer19 and in breast cancer patients20. Other population studies of androgen deprivation therapy in prostate cancer patients have shown inconsistent results regarding the risk of dementia21,22,23. Also, some anticancer drugs, such as masitinib and pexidartinib, have been tested in clinical trials for the treatment of DAT based on their modulation effect on inflammatory cells24,25.

Despite the potential of anticancer drugs as dementia therapies, there is still a lack of large-scale longitudinal population studies on comparing associations between the incidence of dementia and comprehensive anticancer drugs with various mechanisms of action. Furthermore, examining both DAT and VaD in a single study may provide deeper insights into the underlying mechanisms of effective anticancer drugs on dementia. In response to these limitations, the present study aimed to investigate the incidence of dementia (including DAT and VaD) in patients treated with different anticancer drugs within a large nationwide cohort.

Methods

Study design and population

We performed a retrospective population-based cohort study using the Korean National Health Insurance Service (K-NHIS) database, which covers nearly the entire South Korean population. This study included men and women aged 65 years and older who were diagnosed with cancer and were administered anticancer drugs between January 1, 2008, and December 31, 2018 (N = 116,506). We excluded patients who died within six months of cancer diagnosis (N = 4715), were diagnosed with a brain tumor (N = 958), or had a history of dementia before receiving anticancer drugs (N = 6116). After these exclusions, the final number of participants for analysis was 105,283 (Fig. 1). The cohort entry date was defined as the initiation of any anticancer drugs. The follow-up window extended from the cohort entry date to the incident dementia (event date), the end of the study period (December 31, 2018), or death (Fig. 2).

Figure 1
figure 1

Selection of study participants. Cancer patients aged 65 years and older who received anticancer drugs between January 1, 2008 and December 31, 2018 were identified from the Korean National Health Insurance Service (K-NHIS) database (N = 116,506). Exclusions included patients who died within six months of cancer diagnosis (N = 4715), those diagnosed with a brain tumor (N = 958), or with a history of dementia before receiving anticancer drugs (N = 6116). The final analytic cohort comprised 105,283 participants.

Full size image
Figure 2
figure 2

Diagram of study design. Cohort entry occurred at the initiation of any anticancer drug from January 1, 2008. The follow-up period continued until the incident dementia (event date), the end of the study period (December 31, 2018), or death.

Full size image

Measurements

The K-NHIS database contains claims data for inpatient and outpatient visits, procedures, and prescriptions coded using the International Classification of Diseases, 10th Revision (ICD-10), and the Korean Drug and Anatomical Therapeutic Chemical Codes [10]. The K-NHIS routinely audits claims, and the data are considered reliable and have been used in several peer-reviewed publications26. The K-NHIS cohort includes mortality data provided by Statistics Korea27.

Cancer was defined as the presence of a cancer-specific insurance claims code (V193 code), and its specific type was defined according to a C-code from an ICD-10 code. From the data on prescription claims, available groups of anticancer drugs were platinum, antimetabolites, antibiotics (anti-cancer), alkylating agents, microtubule inhibitors, topoisomerase inhibitors, molecular targeted therapy, biologic therapy, and others such as hormonal agents (see Supplementary Table S1 for details). While the other groups of anticancer drugs consist of drugs with relatively homogenous mechanisms, the molecular targeted therapy group exhibits distinctive mechanisms depending on the pathways they target. Therefore, the molecular targeted therapies were further subdivided by their involved signaling pathways: epidermal growth factor receptor (EGFR) inhibitors, multikinase inhibitors, proteasome inhibitors, BCR/ABL inhibitors, mammalian target of rapamycin (mTOR) inhibitors, anaplastic lymphoma kinase (ALK) inhibitors, Janus kinase (JAK) inhibitors, cyclin-dependent kinase (CDK) inhibitors, BRAF inhibitors and others. The incidence of dementia was defined using ICD-10 codes for dementia: DAT (F00, G30), VaD (F01), and unspecified dementia (F03). For covariates, we included age, sex, income (under medical aid, < 30th percentile, 30th–70th percentile, > 70th percentile, or unknown), residential area (metropolitan or rural), and the Charlson Comorbidity Index score at baseline. The Charlson Comorbidity Index scores were calculated from claims data defined using ICD-10 codes.

Statistical analysis

The population characteristics at baseline were summarized as frequencies with proportions for categorical variables or as means with standard deviations for continuous variables. The Kaplan–Meier method was used to depict the differences in cumulative incidence between groups. The incidence rate was calculated as the number of events per 1000 person-years. For the association between anticancer drugs and dementia, Cox proportional hazards regression models were used to calculate the hazard ratios (HRs) with 95% confidence intervals (CIs) for each group of anticancer drugs versus all other groups (for example, group A vs. group B + C + D). The proportionality of the hazards was confirmed by visual inspection of log-minus-log plots and Schoenfeld residuals. Two sets of HRs were calculated: one set adjusted only for age, and another set adjusted for all covariates including age, sex, income, residential area, and the Charlson Comorbidity Index score at baseline. In the sensitivity analysis, to account for competing risks due to mortality, we fitted a proportional sub-distribution hazards regression model28 with death as the competing event.

All p-values were two-tailed, and a p-value of < 0.05 was considered statistically significant. Analyses were performed using SAS Visual Analytics (SAS Institute Inc., USA) or R 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria).

Ethics approval and consent to participate

The Institutional Review Board of Samsung Medical Center approved this study and waived the requirement for informed consent because we accessed only de-identified and previously collected data (IRB number: SMC 201912064). All processes and analyses were performed in accordance with relevant guidelines and regulations.

Results

Among all cancer patients who received anticancer drugs, the mean age (standard deviation) was 71.67 (4.96) years, and 64.4% were female. Platinum (39.0%) was the most common regimen, followed by antimetabolites (30.5%) and antibiotics (21.3%) (Table 1). During the follow-up period (mean = 2.3 years, max = 11.8 years), 5102 patients developed dementia after receiving anticancer drugs. Table 2 shows that antimetabolites (HR = 0.93; 95% CI 0.88–0.99) and molecular targeted therapy (HR = 0.67; 95% CI 0.57–0.79) were associated with a lower incidence of dementia compared to other anticancer drugs. Kaplan–Meier curves for these two groups of anticancer drugs are depicted in Fig. 3A, B. Analysis by subtypes of dementia revealed that antimetabolites (HR = 0.91; 95% CI 0.84–0.97) and molecular targeted therapy (HR = 0.60; 95% CI 0.49–0.74) were associated with a lower risk of DAT compared to the other anticancer drugs. Both antimetabolites and molecular targeted therapy showed insignificant associations with the incidence of VaD. The results of the competing risk analyses were consistent (data not shown).

Table 1 Baseline characteristics of the study population.
Full size table
Table 2 Hazard ratios (95% CI) for dementia incidence by groups of anticancer drugs.
Full size table
Figure 3
figure 3

Kaplan–Meier curves for the incidence of dementia by anticancer drugs. (A) Cumulative incidence of dementia in patients treated with antimetabolites compared to that of all other anticancer drugs. (B) Cumulative incidence of dementia in patients treated with molecular targeted therapies compared to that of all other anticancer drugs. Among molecular targeted therapies, (C) EGFR inhibitors and (D) multikinase inhibitors show lower cumulative incidence compared to all other anticancer drugs.

Full size image

Among cancer patients treated with molecular targeted therapies (N = 3626), the mean age (standard deviation) was 73.11 (5.55) years, and 54.0% were female. EGFR inhibitors were the most common regimen (45.6%) (Supplementary Table S2). Table 3 shows that EGFR inhibitors (HR = 0.66; 95% CI 0.54–0.82) and multikinase inhibitors (HR = 0.50; 95% CI 0.31–0.80) were associated with a lower incidence of dementia compared to all other anticancer drugs. Kaplan–Meier curves for these two classes of molecular targeted therapies are shown in Fig. 3C, D. They were also associated with a lower incidence of DAT (EGFR inhibitors: HR = 0.49; 95% CI 0.27–0.89; multikinase inhibitors: HR = 0.49; 95% CI 0.27–0.89) compared to other anticancer drugs, but not with VaD. The outcomes were robust in competing risk analyses (data not shown).

Table 3 Hazard ratios (95% CI) for dementia incidence by groups of molecular targeted therapies.
Full size table

Discussion

This nationwide population-based cohort study investigated the association between anticancer drugs and the incidence of dementia in over 100,000 cancer patients over 10 years. We found that antimetabolites and molecular targeted therapy decreased the incidence of DAT but not VaD. Among molecular targeted therapies, EGFR inhibitors and multikinase inhibitors were associated with a reduced incidence of DAT. These findings suggest specific anticancer drugs could be repurposed for the treatment of DAT.

Our study has revealed that antimetabolites were associated with a decrease in the incidence of dementia, especially DAT. Methotrexate, a representative antimetabolite drug, has been extensively studied for its dementia-protective effects, particularly in patients with rheumatoid arthritis rather than those with cancer29,30. However, the precise protective mechanisms of antimetabolites against dementia remain elusive. One possible explanation may involve their anti-inflammatory effects, which could mitigate neuroinflammation, a significant contributor to DAT pathogenesis. Chronic brain inflammation promotes neurodegeneration through sustained immune responses, facilitating amyloid and tau pathology31. Several experimental studies proposed that anti-inflammatory effects of methotrexate may reduce the accumulation of amyloid beta, thereby reducing the risk of DAT32. However, the understanding of the individual effects and mechanisms of other antimetabolites on dementia remains limited, warranting further investigation.

Our findings also indicate a potential protective effect of molecular targeted therapy against the development of dementia. With the increasing use of molecular targeted therapies in cancer treatment, to the best of our knowledge, this is the first population-based study to explore the effects of various molecular targeted therapies on dementia. Among these therapies, we have shown that EGFR inhibitors and multikinase inhibitors were associated with a decreased incidence of DAT compared to other anticancer drugs. Considering that EGFR stimulation increases inflammation, autophagy, and amyloid beta accumulation33,34,35,36, EGFR inhibitors may have the potential to mitigate the risk of Alzheimer's disease. Furthermore, in vitro and in vivo studies indicated that multikinase inhibitors modulated Alzheimer's disease pathophysiology37,38,39, with regorafenib reported to inhibit amyloid beta aggregation40.

While our findings provide promising insights, several limitations should be considered. Diagnostic reliance on ICD-10 codes within claims databases, although a common practice, may not capture the complete clinical context, potentially introducing inaccuracies. Particularly, the lack of detailed clinical data such as neuroimaging, fluid biomarkers, or APOE genotype limits our ability to accurately differentiate DAT from other dementias. However, DAT diagnosis based on ICD-10 codes is informed by comprehensive clinical assessments, incorporating mini-mental state examination and clinical dementia rating scores by clinicians, supporting the reliability of our dataset. Another limitation is our inability to account for several potential confounders, such as treatment combinations, cancer types and stages, which may affect the associations reported. Despite this, the observed specificity of associations between certain anticancer drugs and DAT, but not VaD, suggests underlying biological mechanisms. This specificity not only supports a genuine protective effect against DAT but also underscores the biological plausibility of our findings despite these limitations. Additionally, our study did not fully explore the influence of socioeconomic factors, lifestyle choices, and genetic predispositions that could impact both cancer outcomes and dementia risk. Future research should incorporate these variables to provide a more comprehensive understanding of the observed associations.

In conclusion, our large-scale, nationwide, population-based cohort study utilizing the K-NHIS database sheds light on the potential of anticancer drugs, specifically antimetabolites and molecular targeted therapies, to modulate the risk of developing DAT. These findings lay the groundwork for future investigations to delve deeper into these associations, incorporating broader clinical datasets and more sophisticated analyses to unravel underlying mechanisms. Ultimately, our study contributes valuable insights to drug repurposing efforts, highlighting the intersection between oncology and neurology in the pursuit of effective dementia treatments.