Introduction

Migraine has been reported as the sixth most common pathogenesis of disability [1]. In addition, migraine has been reported as the second most common factor associated with disability-adjusted life years worldwide by the Global Burden of Disease study [1]. Migraine is one of the most common pain disorders and its prevalence affects up to 25% of women and 9.4% of men worldwide [2]. In addition, Pompili et al. [3] has systematically documented a strong bidirectional association between migraine and psychiatric disorders. Moreover, the relationship between migraine and psychopathology resulting in enhanced psychosocial impairment has often been clinically discussed rather than systematically studied [4,5,6]. These studies supported the importance of this emergent research in the field.

Subjective cognitive decline is not unusual in migraine. Although cognitive impairment is identified as the core symptom of migraine, a large amount of migraine patients complain of cognitive impairment, especially deficits in attention and memory. However, previous studies revealed inconsistent results regarding association between migraine and cognitive impairment. Indeed, some studies reported that migraine is associated with a lower cognitive function during both interictal [7] or ictal [8, 9] periods. Wen et al. reported that migraine patients tend to score higher in cognition tests than non-migraine subjects [10]. Conversely, some studies did not show any difference in cognitive function between migraine patients and non-migraine subjects [11,12,13]. In addition, previous studies found inconsistent results regarding the association between migraine and risk of dementia [14,15,16]. We hypothesized that migraine patients showed lower general cognitive function, language, visuospatial function, attention, executive function and memory, compared to no migraine group. Additionally, we hypothesized that migraine was significantly associated with risk of dementia. The study aimed to make a meta-analysis exploring comparison result in different types of cognitive function between migraine patients and non-migraine subjects. In addition, meta-analysis was made to explore the association between migraine and risk of dementia.

Methods

The study was made according to the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) guideline [17]. The PRISMA Checklist is included in the Table 1.

Table 1 PRISMA 2009 Checklist
Full size table

Search strategy

We searched for articles published before June 2022 in the following databases: PubMed, Web of Science, SCOPUS, EMBASE, EBSCO, PROQUEST, ScienceDirect and Cochrane Database of Systematic Reviews. Included studies explored association between migraine and cognitive impairment. We used the following search terms: (“migraine” OR “hemicrania” OR “cephalagra”) AND (“cognitive impairment” OR “cognitive deficit” OR “dementia” OR “Alzheimer’s disease”).

Inclusion and exclusion criteria

We included studies investigating the association between migraine and cognitive impairment. We excluded studies according to the following exclusion criteria: 1) Studies which did not provide sufficient information regarding cognitive function in both migraine and healthy controls (HCs); 2) Studies which did not provide sufficient information for odds ratios (ORs) in case-control studies or relative risks (RRs) in cohort studies and their 95% confidence intervals (CIs); 3) Meta-analyses, reviews and case-reports.

Data extraction

We extracted the following data from finally included studies: Author, publication year, study type, type of migraine, study location, sample size, mean age of patients, gender of patients, disease duration of migraine, attack frequency of migraine, duration of migraine attack, pain intensity, explored cognitive functions, adjusted variables, follow-up time and results.

Cognitive tests included

According to previous studies [18], neuropsychological examinations were divided into 6 cognitive domains: (1) general cognitive function, (2) language, (3) visuospatial function, (4) attention function, (5) executive function, (6) memory function. General cognitive function was evaluated by the Mini-mental state examination (MMSE) and Montreal cognitive assessment (MoCA). Language function was assessed by Fluency test (phonemic fluency test and verbal fluency test) and Mill hill vocabulary test part A. Visuospatial function was assessed by Rey-Osterrieth complex figure test (ROCFT) and Clock drawing test (CDT). Attention function was evaluated by Trail making test (TMT)-A, Digital Symbol Substitution Test (DSST), Letter digit substitution test (LDST) and Stroop color and word test (SCWT) A, B. Executive function was assessed by Digital span test (DST)-backward, TMT-B, SCWT C and Semantic similarity test. We identified and recorded the mean value and standard deviation (SD) of raw scores of each neuropsychological test. Higher raw scores indicate better cognitive function on almost all the cognitive tests. However, the TMT (A and B) and SCWT (A, B and C) present an exception, as there is a reversed interpretation for the raw scores (where longer time indicates poorer performance). For this reason, the TMT (A and B) and SCWT (A, B and C) scores of the study have been reversed, so that higher scores indicate better performance. The mean value and SD of cognitive scores in migraine and no migraine groups were standardized and reported in relation to the mean value in no migraine groups. Then, each cognitive domain’s standardized score was determined by averaging the standardized scores of relevant tests. Risk of dementia was measured by calculating the incidences of dementia.

Meta-analysis

We used STATA 12.0 software to compute results. Standardized mean values and SD of cognitive function associated scores were computed. In addition, ORs/RRs and their CIs were computed. We used Q test and I2 to evaluate heterogeneities between included studies. We used random effects models to compute results. We used subgroup studies (for different ethnicities and study types) to explore the source of the heterogeneity. We used meta-regression analysis to investigate the source of heterogeneity. We used sensitivity analysis to assess the study stabilization. We used Begg’s test, Egger’s test and funnel plot to assess publication bias.

Results

Study characteristics

Supplementary Fig. 1 showed the inclusion and exclusion process. Tables 2 and 3 showed study characteristics. N = 22 studies [7, 10,11,12,13, 19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35] (including 3295 migraine patients) investigated cognitive function in both migraine and HCs. These studies included N = 4 cohort studies and N = 18 cross-sectional studies. N = 11 studies [14,15,16, 36,37,38,39,40,41,42,43] included N = 3 case-control studies [14, 16, 36] (including 12,871 dementia patients and 56,365 no dementia participants) and N = 8 cohort studies [15, 37,38,39,40,41,42,43] (including 47,942 migraine patients and 190,024 HCs) investigated the association between migraine and risk of dementia.

Table 2 Characteristics of studies regarding comparison in various cognition between migraine group and no migraine group
Full size table
Table 3 Characteristics of studies regarding association between migraine and risk of dementia
Full size table

meta-analysis results

Comparison in general cognitive function

Meta-analysis showed a lower general cognitive function in migraine group, compared to no migraine group with a random effects model (standard mean difference (SMD) = − 0.40, 95% CI = − 0.66 to − 0.15, I2 = 92.8%, p < 0.001, Fig. 1). Subgroup analysis showed no significant difference in general cognitive function between migraine group and no migraine group in Caucasian, whereas migraine group showed a lower general cognitive function in migraine group, compared to no migraine group in Asian (Supplementary Table 1 and Supplementary Fig. 2. A). Subgroup analysis showed a lower general cognitive function in migraine group, compared to no migraine group in cross-sectional studies (Supplementary Table 2 and Supplementary Fig. 2. B). Meta-regression analysis showed that age of migraine, gender of migraine, disease duration of migraine, attack frequency of migraine and pain intensity were not responsible for the heterogeneity across studies (Supplementary Table 3). Sensitivity analysis indicated no changes in the direction of effect when anyone study was excluded (Supplementary Fig. 3. A). Begg’s test, Egger’s tests and funnel plots indicated a significant risk of publication bias (Supplementary Table 4 and Supplementary Fig. 3. B).

Fig. 1
figure 1

Forest plots regarding comparison in general cognitive function between migraine group and no migraine group. Abbreviations: CI, confidence interval; SMD, standard mean difference

Full size image

Comparison in language function

Meta-analysis showed a lower language function in migraine group, compared to no migraine group with a random effects model (SMD = − 0.14, 95% CI = − 0.27 to − 0.00, I2 = 65.1%, p = 0.001, Fig. 2). Subgroup analysis showed no significant difference in language function between migraine group and no migraine group in Caucasian (Supplementary Table 1 and Supplementary Fig. 4. A). Subgroup analysis showed a lower language function in migraine group, compared to no migraine group in cross-sectional studies (Supplementary Table 2 and Supplementary Fig. 4. B). Meta-regression analysis showed that age of migraine was responsible for the heterogeneity across studies, whereas gender of migraine, disease duration of migraine, attack frequency of migraine, duration of migraine attack and pain intensity were not responsible for the heterogeneity across studies (Supplementary Table 3). Sensitivity analysis indicated no changes in the direction of effect when anyone study was excluded (Supplementary Fig. 5. A). Begg’s test, Egger’s tests and funnel plots indicated no significant risk of publication bias (Supplementary Table 4 and Supplementary Fig. 5. B).

Fig. 2
figure 2

Forest plots regarding comparison in language function between migraine group and no migraine group. Abbreviations: CI, confidence interval; SMD, standard mean difference

Full size image

Comparison in visuospatial function

In addition, meta-analysis showed no significant difference in visuospatial function between migraine group and no migraine group with a random effects model (SMD = − 0.23, 95% CI = − 0.53 to 0.08, I2 = 56.1%, p = 0.077, Fig. 3). Subgroup analysis showed a lower visuospatial function in migraine group, compared to no migraine group in Caucasian (Supplementary Table 1 and Supplementary Fig. 6). Meta-regression analysis showed that age of migraine, gender of migraine, disease duration of migraine and attack frequency of migraine were not responsible for the heterogeneity across studies (Supplementary Table 3).

Fig. 3
figure 3

Forest plots regarding comparison in visuospatial function between migraine group and no migraine group. Abbreviations: CI, confidence interval; SMD, standard mean difference

Full size image

Comparison in attention function

However, no significant difference in attention between migraine group and no migraine group with random effects models (SMD = − 0.01, 95% CI = − 0.10 to 0.08, I2 = 52.6%, p = 0.002, Fig. 4). Subgroup analysis showed no significant difference in attention between migraine group and no migraine group in Caucasian (Supplementary Table 1 and Supplementary Fig. 7. A). Subgroup analysis showed no significant difference in attention between migraine group and no migraine group in cross-sectional studies (Supplementary Table 1 and Supplementary Fig. 7. B). Meta-regression analysis showed that age of migraine and gender of migraine were responsible for the heterogeneity across studies, whereas disease duration of migraine, attack frequency of migraine, duration of migraine attack and pain intensity were not responsible for the heterogeneity across studies (Supplementary Table 3). Sensitivity analysis indicated no changes in the direction of effect when anyone study was excluded (Supplementary Fig. 8. A). Begg’s test, Egger’s tests and funnel plots indicated no significant risk of publication bias (Supplementary Table 4 and Supplementary Fig. 8. B).

Fig. 4
figure 4

Forest plots regarding comparison in attention between migraine group and no migraine group. Abbreviations: CI, confidence interval; SMD, standard mean difference

Full size image

Comparison in executive function

Meta-analysis showed no significant difference in executive function between migraine group and no migraine group with random effects models (SMD = − 0.05, 95% CI = − 0.16 to 0.05, I2 = 54.7%, p = 0.001, Fig. 5). Subgroup analysis showed no significant difference in executive function between migraine group and no migraine group in Caucasian (Supplementary Table 1 and Supplementary Fig. 9. A). Subgroup analysis showed no significant difference in executive function between migraine group and no migraine group in cross-sectional and cohort studies (Supplementary Table 1 and Supplementary Fig. 9. B). Meta-regression analysis showed that gender of migraine was responsible for the heterogeneity across studies, whereas age of migraine, disease duration of migraine, attack frequency of migraine, duration of migraine attack and pain intensity were not responsible for the heterogeneity across studies (Supplementary Table 3). Sensitivity analysis indicated no changes in the direction of effect when anyone study was excluded (Supplementary Fig. 10. A). Begg’s test, Egger’s tests and funnel plots indicated no significant risk of publication bias (Supplementary Table 4 and Supplementary Fig. 10. B).

Fig. 5
figure 5

Forest plots regarding comparison in executive function between migraine group and no migraine group. Abbreviations: CI, confidence interval; SMD, standard mean difference

Full size image

Comparison in memory function

Meta-analysis showed no significant difference in memory between migraine group and no migraine group with random effects models (SMD = − 0.14, 95% CI = − 0.30 to 0.03, I2 = 82.5%, p < 0.001, Fig. 6). Subgroup analysis showed no significant difference in memory between migraine group and no migraine group in Caucasian, whereas migraine group showed a lower memory function, compared to no migraine group in Asian (Supplementary Table 1 and Supplementary Fig. 11. A). Subgroup analysis showed no significant difference in memory between migraine group and no migraine group in cross-sectional studies (Supplementary Table 1 and Supplementary Fig. 11. B). Meta-regression analysis showed that age of migraine, gender of migraine, disease duration of migraine, attack frequency of migraine and duration of migraine were not responsible for the heterogeneity across studies (Supplementary Table 3). Sensitivity analysis indicated no changes in the direction of effect when anyone study was excluded (Supplementary Fig. 12. A). Begg’s test, Egger’s tests and funnel plots indicated no significant risk of publication bias (Supplementary Table 4 and Supplementary Fig. 12. B).

Fig. 6
figure 6

Forest plots regarding comparison in memory between migraine group and no migraine group. Abbreviations: CI, confidence interval; SMD, standard mean difference

Full size image

Association between migraine and risk of dementia

The meta-analysis showed a significant association between migraine and risk of dementia with a random effects model (OR/RR = 1.30, 95% CI = 1.11 to 1.52, I2 = 83.5%, p < 0.001, Fig. 7). Subgroup analysis showed no significant association between migraine and risk of dementia in Caucasian, whereas a significant association between migraine and risk of dementia was showed in Asian (Supplementary Table 1 and Supplementary Fig. 13. A). Subgroup analysis showed a significant association between migraine and risk of dementia in cohort studies (Supplementary Table 1 and Supplementary Fig. 13. B). Meta-regression analysis showed that age of migraine and gender of migraine were not responsible for the heterogeneity across studies (Supplementary Table 3).

Fig. 7
figure 7

Forest plots regarding association between migraine and risk of dementia. Abbreviations: CI, confidence interval; OR, odds ratio; RR, relative risk

Full size image

However, the study showed no significant association between migraine without aura (MWoA) and risk of dementia with a random effects model (OR/RR = 1.03, 95% CI 0.89 to 1.19, I2 = 0.0%, p = 0.453, Supplementary Fig. 14. A). In addition, the study showed significant associations between migraine and risk of vascular dementia (VaD), Alzheimer’s disease (AD) with random effects models (VaD: OR/RR = 1.84, 95% CI 1.18 to 2.88, I2 = 0.0%, p = 0.423, Supplementary Fig. 14. B; AD: OR/RR = 2.60, 95% CI 1.51 to 4.48, I2 = 43.8%, p = 0.169, Supplementary Fig. 14. C).

Discussion

Meta-analysis showed lower general cognitive function and language function in migraine group, compared to no migraine group, whereas the study showed no significant difference in visuospatial function, attention, executive function and memory between migraine group and no migraine group. In addition, the meta-analysis showed a significant association between migraine and risk of dementia.

The present study compared various types of cognition between migraine group and no migraine group. Included studies showed contradictory results on association between migraine and cognitive impairment. Cross-sectional and cohort studies reported worse cognitive function in migraine patients [25] or no association [11,12,13, 23, 26], whereas some longitudinal studies showed reduced decline of the general cognitive function and executive function in migraine [22, 24]. In addition, Wen et al. [10] reported that migraineurs, particularly migraineurs with aura, tend to score higher in cognition tests than non-migraineurs. These inconsistencies might be caused by different methodological issues including different migraine assessment methods. In addition, clinical features (age, gender, types of migraine, disease duration of migraine, attack frequency of migraine, duration of migraine attack, pain intensity, follow-up duration, headache medication use, diet, sleep, or physical activity, et al.) might be the source of inconsistencies. The effect of age and gender on association between migraine and cognitive impairment has been verified by meta-regression in the present study. More large-scale cohort studies were essential to explore the association between migraine and cognitive impairment.

Up to now, the exact mechanism regarding association between migraine and cognitive impairment is still not fully understood. Recent studies provided information for alterations in brain functional reorganization of cognitive cerebral networks in migraine. These cognitive cerebral networks included default mode network (DMN) [44], executive control network (ECN) [45], visual network [46], et al. The DMN plays an important role in several cognitive processes, such as memory, problem solving and planning [47]. The ECN mainly includes the dorsolateral prefrontal cortex (DLPFC) and the posterior parietal cortex (PPC) [48]. The frontal lobe involves in regulating behavior, complex planning, and learning [49]. Visual processing speed is linked to functional connectivity between right frontoparietal and visual networks [50]. In addition, somatic pain can drive a person to focus on the pain and shift his attention from other cognitive tasks. These mechanisms might contribute to the association between migraine and cognitive impairment.

The meta-analysis showed a significant association between migraine and risk of dementia. The result is corresponding to a recent meta-analysis (including N = 9 observational studies) which demonstrated that migraine may be a risk factor for dementia, particularly VaD and AD [51]. Previous studies supported that some vascular risk factors of VaD (including hypertension, diabetes and stroke) could cause migraine [52, 53]. In addition, migraine showed more prevalent in white matter hyperintensities (WMH), which shows an increased risk of dementia both VaD and AD [54, 55], compared to HCs. However, only N = 3 studies explored the association between migraine and risk of VaD or AD. Thus, more studies were essential to explore the association between migraine and risk of VaD or AD.

The present meta-analysis showed high heterogeneity between studies investigating association between migraine and risk of dementia. The present study mainly included observational studies, which were both clinically and methodologically inhomogeneous. Thus, high heterogeneity is inevitable and not surprising. Subgroup analysis showed no significant association between migraine and risk of dementia in Caucasian, whereas a significant association between migraine and risk of dementia was showed in Asian. Different ethnicities might be the source of heterogeneity. In addition, other clinical features, such as age, gender, types of migraine, disease duration of migraine, attack frequency of migraine, duration of migraine attack, pain intensity and follow-up duration, might be also the source of heterogeneity. In the present study, we selected studies according to explicit inclusion and exclusion criteria to decrease heterogeneity. However, heterogeneity still exists.

There are some limitations in the study. Firstly, the heterogeneity across studies is unavoidable. The high heterogeneity might have an impact on the reliability of our results. The high heterogeneity might be caused by different methodological issues and clinical features. More large-scale cohort studies were essential to explore the association between migraine and cognitive impairment. Secondly, the study included limited number of studies exploring the association between migraine and risk of VaD or AD. More studies were essential to explore the association between migraine and risk of VaD or AD. Thirdly, some included studies were case-control designed, which might cause recall bias. The recall of migraine may be uncertain and may result in a wrong diagnosis of migraine.

Conclusion

In conclusion, the meta-analysis demonstrated lower general cognitive function and language function in migraine. In addition, migraine is associated with an increased risk of all-cause dementia, VaD and AD. These results suggest a significant association between migraine and cognitive impairment. Because of the association between migraine and cognitive impairment, neurological physician should be vigilant and effectively intervene in migraineurs with high risk factors of cognitive impairment to prevent the development of cognitive impairment.