Introduction
Alzheimer’s disease (AD) stands as the primary contributor to dementia, marked by the accumulation of amyloid-β (Aβ) peptides and neurofibrillary tangles. Reference Jia, Quan and Fu1–Reference Livingston, Sommerlad and Orgeta3 The protracted progression of AD elicits a substantial burden of disability, engendering significant economic pressures on society in its entirety, as well as the families directly impacted by this ailment. Reference Livingston, Sommerlad and Orgeta3,Reference Frankish and Horton4 It’s worth noting that AD is a multifactorial and complex disease, and the lack of effective treatments makes it particularly important to explore modifiable risk factors for AD and develop effective prevention and treatment strategies. Reference Zhang, Chen and Deng5
Migraine is the most prevalent disabling neurological disorder, affecting over 15% of the world’s population, which characterized by recurrent headaches and can be classified into two subtypes based on symptoms, including migraine with aura and migraine without aura. Reference Fang, Zhang, Zhou, Wu, Ji and Zou6–Reference Gazerani9 Headache is an important dementia risk. Reference Hagen, Stordal, Linde, Steiner, Zwart and Stovner10,Reference Qu, Yang, Yao, Sun and Chen11 Several epidemiological studies also had found that migraine may be related to the AD risk. Reference Morton, St John and Tyas12,Reference Hurh, Jeong, Kim, Jang, Park and Jang13 However, it is important to note that the conclusions have been inconsistent. Reference Daghlas, Rist and Chasman14,Reference Kim, Ha, Park, Han and Baek15 Establishing the link between migraine and AD may provide evidence and new strategies for interventions and delaying subsequent cognitive decline.
We hypothesized that patients with migraine may have a higher risk of developing AD compared to individuals without migraine. Due to previous studies exploring this association have been limited by small sample sizes, cross-sectional designs and confounding factors. Reference Lee, Gibbons and Lee16 Importantly, these studies have not taken into account the causal relationship between migraine and AD in the context of genetic susceptibility. Reference Bowden, Davey Smith, Haycock and Burgess17–Reference Burgess, Butterworth and Thompson20 Thus, novel research methodologies are necessary to attain a comprehensive understanding of the intricate association between these two conditions. Therefore, in this study, we sought to extensively dissect the genetic and phenotypic relationships between migraine and AD by using the UK Biobank and bidirectional Mendelian randomizations (MR) approach, which may provide new strategies for the treatment and clinical prevention of AD.
Materials and methods
Data source and study population
The UK Biobank indeed is a large population-based prospective cohort study that includes over 500,000 participants, who were initially recruited between 2006 and 2010 from 22 assessment centers. Reference Fry, Littlejohns and Sudlow21 Ethical approval was granted by the North West-Haydock Research Ethics Committee (REC reference:16/NW/0274). Reference Sudlow, Gallacher and Allen22 All participants in the study provided informed consent and underwent a baseline screening process, during which individuals self-reporting cognitive impairment or those diagnosed with all-cause dementia through hospital records were excluded. Moreover, only participants who exhibited AD outcome during follow-up were considered for analysis, with those who developed AD before experiencing migraine symptoms being excluded. Additionally, individuals who did not report any headache symptoms at baseline or follow-up were included as control subjects. A schematic diagram representing the study design is presented in Figure 1.
Migraine
The diagnosis of migraine is based on information derived from self-reported conditions, primary healthcare data, hospital admission records and death registry records. Migraine diagnosis is determined using ICD-10 codes, as well as reading codes (2nd edition and Clinical Terms Version 3). Reference Madjedi, Stuart and Chua23 Participants previously diagnosed with AD are excluded from the diagnosis of migraine.
Ascertainment of AD
In the UK Biobank cohort study, AD was determined through algorithmic methods incorporating mortality register date, hospital inpatient records, self-reported data. Reference Lourida, Hannon and Littlejohns24,Reference Wilkinson, Schnier and Bush25 The specific codes from the International Classification of Diseases–Tenth Revision (ICD-10) employed for defining AD can be found in Supplemental Table 1. The follow-up for all outcomes encompassed the period up to December 31, 2021.
Covariates
In this study, we adjusted for several potential confounding factors by including multiple covariates. Sociodemographic variables considered for adjustment included age, sex (female/male), race (white/nonwhite), education level (high school or below/college or above) and Townsend deprivation index (which reflects social deprivation status and was categorized into low/medium/high levels). The APOE ε4 status was assessed using genetic database information, while tobacco use was self-reported by participants. Health conditions such as hypertension and diabetes were ascertained through self-reports and electronic medical records. Additionally, we considered various health-related factors including body mass index (BMI), hypertension, diabetes, stroke and coronary heart disease (CHD). The presence of CHD was identified through a 12-lead resting electrocardiogram recording coded according to the Minnesota system, and further confirmed by linkage to the HES database using the ICD-10 codes I20-I25. Similarly, for stroke, we utilized the same approach of linking our data to the HES database and identifying cases using the ICD-10 codes I60-I64. All of these factors were assessed at the baseline. Detailed definitions and assessments of covariates can be found in Supplemental Table 2.
MR analysis
The bidirectional MR study was conducted to investigate the causal relationship between migraine and genetic susceptibility to AD. The overview of research was shown in Supplement Figure S1. The present study utilized a large-sample cohort of the European population, encompassing data obtained from a publicly available genome-wide association study (GWAS) dataset. The summary data on migraine was derived from the largest genome-wide meta-analysis, which comprised 102,084 migraine cases and 771,257 European ancestry controls. Reference Hautakangas, Winsvold and Ruotsalainen26 For AD, summary data from the recent GWAS on AD from the Alzheimer’s Disease and Dementia Consortium were utilized, which comprised 85,934 European ancestry cases and 401,577 controls. Reference Bellenguez, Kucukali and Jansen27 The main analysis employed a random effects inverse variance weighted model as the primary criterion, with supplementary analyses using MR-Egger regression, weighted median and maximum likelihood estimation. Moreover, sensitivity analyses, including tests for heterogeneity and horizontal pleiotropy, were conducted using Cochran’s Q, MR-Egger intercept and MR-PRESSO tests to ensure robustness of the conclusions. In addition, when migraine was used as the exposure and AD as the outcome, the measure of effect was odds ratio (OR) and its 95% CI; conversely, the effect measure was β and its 95% CI. The MR analysis was conducted using R version 4.2.1 and the “Two Sample MR” (version 0.5.6) and “MR-PRESSO” (version 1.0) packages in R software. Reference Hemani, Zheng and Elsworth28,Reference Verbanck, Chen, Neale and Do29 Significance was determined at a two-sided P-value less than 0.05.
Statistical analyses
Baseline characteristics were compared using one-way analysis of variance and chi-square tests. Categorical variables were presented as numbers and percentages, while continuous variables were presented as means and standard deviations.
The primary aim of our analysis was to investigate the association between migraine and incident AD. We employed Cox proportional hazard models to estimate hazard ratios (HRs) and confidence intervals (CIs) for migraine (positive vs. negative). Person-years from baseline to AD diagnosis, death or loss to follow-up, whichever occurred first, were used as the measure of time. The assumption of proportional hazards was assessed by including an exposure-time interaction term in the model.
Our analysis comprised several steps. In Model 1, we adjusted for covariates including year of birth, sex, race, smoking status, alcohol-drinking status and education level. Model 2 additionally considered the competing risk of mortality and included adjustments for year of birth, sex, race, education, BMI, current smoking, current drinking, hypertension, diabetes, coronary heart disease, stroke and APOE ε4 status.
To validate our findings, subgroup analyses were performed based on factors such as sex (female/male), Townsend deprivation index (high/medium/low), education level (high school or below/college or above), smoking status (ever/never smoker), alcohol-drinking status (current/non-current drinker), BMI and APOE ε4 status (carrier/non-carrier). Furthermore, we employed a propensity score methodology to select matched controls from the pool of participants without migraine. This approach accounted for various demographic and health-related factors, including age, sex, race, education, BMI, current smoking, current alcohol consumption, hypertension, diabetes, coronary heart disease, stroke and APOE ε4 status. Subsequently, within each of the four age-of-onset groups, we examined the association between migraine and AD by comparing the cases to their respective matched control subjects.
All statistical analyses were conducted using R 4.2.1, and a two-sided P-value < 0.05 was considered statistically significant.
Results
Baseline characteristics
Table 1 demonstrates the baseline characteristics of the study population, stratified by AD status. The population included 404,318 participants, and during the mean follow-up time of 12.31 years, for 22,558 migraine patients, with a mean age of 54.95 (±0.05) years, and a mean BMI of 27.43 (±0.01); for non-migraine individuals, with a mean age of 56.21 (±0.01) years, and a mean BMI of 27.24 (±0.03); During the mean follow-up time of 12.31 years. Additionally, we analyzed the differences in other baseline characteristics between migraine and non-migraine patients.
SD = standard deviation. Data are presented as n (%) and mean (SD). The p values are derived using Student’s t test, Mann-Whitney U test or X 2 test.
Association between migraine and AD risk among patients with migraine
As shown in Table 2 and after adjusting for multiple factors, migraine patients had a significantly increased risk of developing AD compared to non-migraine patients, with a multivariable-adjusted hazard ratio (HR) of 2.34, 95% confidence interval (CI) of 2.01–2.74, P < 0.001; Additionally, using propensity score matching, one control participant was randomly selected for each migraine patient from the pool of subjects without migraine, we separately analyzed the association between migraine and AD in both the group of migraine patients and their matched control group participants. After propensity score matching, a total of 45,116 migraine participants and 45,116 matched control participants were included in this analysis. As shown in Table 2, after adjusting for multiple factors, migraine patients had a significantly higher risk of developing AD compared to non-migraine patients (HR = 1.85, 95%CI = 1,68–2.05, P < 0.001).
AD = Alzheimer’s disease; CI = confidence intervals; HR = hazard ratio. Values are HR (95% CI), unless otherwise indicated.
a Adjusted for covariates including year of birth, sex, race and education.
b Adjusted for competing risk of deaths and covariates including year of birth, sex, race, education, BMI, current smoking, current drinking, hypertension, diabetes, coronary heart disease, stroke and ApoE4 carriers.
Causal relationship between migraine and AD
In the analysis of the causal relationship between migraine and AD. Initially, 59 SNPs were obtained as IVs after excluding palindromic SNPs and SNPs related to confounding factors. The F-statistic scores of all these selected SNPs were over 10, indicating that the strength of the instruments was robust. Using the IVW-based method, the MR analysis demonstrated significant correlation between genetically determined migraine and AD [odds ratio (OR) = 2.315; 95% confidence interval (CI) = 1.029–5.234; P = 0.002] (Table 3 and Fig. 2a). In addition, there was no evidence of heterogeneity (MR-Egger regression: Q = 71.05, p = 0.099; IVW model: Q = 71.16, p = 0.114) and horizontal pleiotropy (MR-Egger intercept and MR-PRESSO tests, all p > 0.05). Moreover, the leave-one-out sensitivity analysis further confirmed that no single SNP was driving the causal effect, indicating that this study is stable (Fig. 3a).
OR = odds ratio; CI = confidence interval; MR = Mendelian randomization; SNPs = single nucleotide polymorphisms; NA = not available; AD = Alzheimer’s disease; SE = standard error.
Causal relationship between AD and migraine
In the analysis of the causal relationship between AD and migraine. Initially, 58 SNPs were obtained as IVs after excluding palindromic SNP and SNPs related to confounding factors. The F-statistic scores of all these selected SNPs were over 10, indicating that the strength of the instruments was robust. Using the IVW-based method, the MR analysis demonstrated no significant correlation between genetically determined AD and migraine (OR = 1.000; 95%CI = 0.999–1.006; P = 0.971), which was consistent with the results from other methods (Table 3 and Fig. 2b). Additionally, due to the heterogeneity (MR-Egger regression: Q = 76.88, p = 0.033; IVW model: Q = 77.01, p = 0.039), therefore, the random effects IVW model was used to minimize the effect of heterogeneity. There was no evidence of horizontal pleiotropy, which was detected using the MR-Egger intercept and MR-PRESSO tests (p > 0.05). The leave-one-out sensitivity analysis further confirmed that no single SNP was driving the causal effect, indicating that this study is stable (Fig. 3b).
Discussion
This study evaluated the association between migraine and AD using the UK Biobank cohort and found that migraine patients had a higher incidence of AD compared to non-migraine patients. Additionally, to the best of our knowledge, the present study firstly demonstrated that migraine patients performed an increased tendency toward genetic susceptibility to AD by using MR. Therefore, our findings had provided valuable insights into the potential causal relationship between these two conditions.
Several studies have shown associations between headache and dementia. Reference Qu, Yang, Yao, Sun and Chen11,Reference George, Folsom and Sharrett30,Reference Wang, Xu, Sun, Yu and Fan31 However, associations between migraine and AD development have been conflicting. The variability of outcomes across different studies may be a contributing factor to the inconsistent results. Reference Hurh, Jeong, Kim, Jang, Park and Jang13 A meta-analysis including four case-control studies found a significant negative correlation between migraine and AD. Reference Breteler, van Duijn and Chandra32 Conversely, other studies have shown that migraine significantly increases the risk of dementia, Reference Wang, Xu, Sun, Yu and Fan31,Reference Cermelli, Roveta and Giorgis33 but this effect may be limited to dementia subtypes, such as vascular dementia, Reference Hagen, Stordal, Linde, Steiner, Zwart and Stovner10 specific subgroups, such as women Reference Lee, Lim, Oh, Kong and Choi34 or to broader headache disorders or non-migraine cases. Reference Tzeng, Chung and Lin35 Additionally, prior cohort studies also confirmed that migraine patients may be associated with the higher risk of AD, Reference Jiang, Liang, Li, Yu and Dong36–Reference Wang, Wu, Wang, Chen and Wang39 which was consistent with our finding. It is important to note that after controlling for all potential confounding factors using propensity score matching, migraine patients showed an increased risk of AD compared to matched non-migraine controls, which was the main innovation of the current study in controlling confounding factors, providing more robust conclusions. Therefore, it should be closely monitored and screen cognitive function changes in migraine patients in order to detect cognitive impairment early and enable timely intervention to prevent or at least delay the onset and progression of AD.
It is important to note that MR is a statistical technique employed in epidemiology and genetics to ascertain causal relationships between risk factors and outcomes. Reference Sanderson, Richardson, Morris, Tilling and Davey Smith40,Reference Lawlor, Harbord, Sterne, Timpson and Davey Smith41 MR is based on the principles of Mendelian genetics, which describe how genetic variations are randomly allocated during meiosis. Reference Emdin, Khera and Kathiresan42 This method utilizes instrumental variables, particularly genetic variants such as single nucleotide polymorphisms (SNPs) associated with the relevant risk factor, to explore whether the selected risk factor has a causal impact on the outcome of interest. Reference Bowden and Holmes43 In the absence of randomized controlled trials (RCTs), MR studies represent an alternative strategy for causal inference as genetic variations are randomly allocated during meiosis, thereby introducing an additional layer of data compared to observational studies. Reference Burgess, Butterworth and Thompson20,Reference Holmes, Ala-Korpela and Smith44 Consequently, MR offers advantages over traditional observational research, reducing the risk of confounding and reverse causation, making it a preferred tool for investigating causal relationships in epidemiological research. Reference Lee and Lim45 The bidirectional MR results of this study demonstrate an increased genetic susceptibility to AD in individuals with migraine, further reinforcing the evidence for migraine as a risk factor for AD and providing valuable insights for the development of novel prevention and treatment strategies for AD.
Several hypothetical mechanisms may underlie the link between migraine and AD risk. On the one hand, Oxidative stress and inflammatory responses have been identified as key risk factors for migraine attacks. Reference Heidari, Shojaei and Askari46,Reference Ramachandran47 Chronic stress activates the hypothalamic-pituitary-adrenal (HPA) axis, leading to the release of glucocorticoids and HPA axis dysregulation. Reference Herman, McKlveen and Ghosal48 Moreover, research evidence supports an association between HPA axis dysregulation and amyloid deposition, as well as synaptic plasticity disruption related to the progression of AD. Reference Chi, Yu, Tan and Tan49,Reference Saeedi and Rashidy-Pour50 On the other hand, the association between migraine and AD may also be influenced by genetic factors. Individuals with familial AD due to presenilin-1 mutations are more likely to suffer from migraine or recurrent headaches. Reference Ringman, Romano and Medina51 Additionally, studies had found that chromosomes 1 and 19 are associated with both migraine and AD. Reference Breslau and Rasmussen52 Further investigation of these or other genotypes may help to elucidate the association between migraine and AD, as well as identify high-risk individuals. However, the exact mechanisms still require more research to fully understand.
The strengths of our study are multifaceted. Firstly, the inclusion of a large-sample size enabled the precise identification of statistically significant associations between the onset of migraines and subsequent development of AD. Moreover, propensity score matching methodology was also employed, reducing confounding bias and ensuring the robustness of our findings. Additionally, the utilization of the UK Biobank algorithm-defined outcome provided a standardized and reliable approach to defining cases of AD, enhancing the validity of our results. Moreover, our use of MR analysis offered primary evidence for causative associations between migraines and AD. Overall, by diligently accounting for confounding variables through propensity score matching and leveraging precise data from a sizable sample, our study offers valuable insights into the relationship between migraines and subsequent AD.
However, our study has certain limitations. Firstly, its observational nature precludes the establishment of causal relationships. Moreover, the selection of the study population did not employ systematic sampling methods, potentially impacting the generalizability of our findings to other ethnic groups and the wider UK population. Therefore, caution must be exercised when extrapolating the results, as they may only apply to white individuals in the UK. Future investigations in diverse populations are necessary to validate our findings. Secondly, despite controlling for various potential confounders implicated in the pathogenesis of AD, residual confounding may persist due to unmeasured factors within the UK Biobank dataset. Thirdly, underdiagnosis and underreporting of AD in medical records could have introduced misclassification bias, affecting the estimation of associations. While the diagnostic accuracy of dementia is generally high, this limitation should not be overlooked. Additionally, we did not explore the potential of migraine medications in reducing the risk of AD, an important avenue for future research.
Conclusion
In conclusion, the present study concludes that migraine patients, compared to a matched control group, exhibit an increased risk of developing AD. Moreover, migraine patients exhibit an increased predisposition of genetic susceptibility to AD. These findings hold significant clinical value for early intervention and treatment of migraines to reduce the risk of AD. However, further research is needed to synthesize our findings and elucidate the underlying pathological and physiological mechanisms between migraine and AD.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/cjn.2024.35.
Acknowledgements
The author expresses his profound gratitude to the establishers and participants of the GWAS databases for their invaluable contributions, which facilitated the successful execution of this MR study.
Data availability statement
The authors acknowledge and appreciate the effort and involvement of the UK Biobank participants and staff in this study; Additionally, the datasets analyzed for this study can be found in the IEU Open GWAS project (mrcieu.ac.uk).
Author contributions
G-CF: Study concept, design, software and paper writing. C-C: Dissertation Revision. All authors read and approved the final manuscript.
Funding statement
None.
Competing interests
None.
Ethics approval
UK Biobank received ethical approval from the National Health Service North West Centre for Research Ethics Committee (Ref: 11/NW/ 0382); Additionally, the MR analysis used summary GWAS data publicly available from GWASs. All these GWAS summary data are publicly available, and all studies included were approved by relevant ethics committee. All participants signed written informed consent prior to participation.