Introduction

Personality is defined as a set of psychological qualities related to feelings, thoughts and behaviours [1]. Many theories of personality structure have been proposed but the most widely known are the Psychobiological Model [2], the Big-Five Model [3] and Eysenck’s Three-Factor Model [4]. These models share the idea that personality can be described in terms of several factors but differ from each other in the number and in the formal characteristics of such factors, and in the specific tools needed to assess them. The Psychobiological Model (PM) [2] explored the seven basic personality dimensions of temperament (i.e. novelty seeking, harm avoidance, reward dependence and persistence) and character (i.e. self-directedness, cooperativeness and self-transcendence). The Big-Five Model [3] proposes that five main dimensions characterize personality: extraversion, agreeableness, conscientiousness, neuroticism and openness. The Eysenck’s Three- Factor Model (TFM) [4] model was centred on three dimensions: extraversion, neuroticism and psychoticism (for more information, see Supplementary Material 1, A). The three models are the most widely accepted personality theory in the scientific community and are commonly used in the research and study of personality in psychology [5].

Early studies reported that migraineurs’ personality is characterized by orderliness, perfectionism, inflexibility and a tendency to react excessively to problems, which, in turn, could lead to attacks of migraine [6]. Later on, studies comparing personality traits in migraineurs and non-migraineurs reported mixed results. For instance, some studies investigating the personality profile according to the psychobiological model (PM) reported a higher level of harm avoidance in migraineurs [7,8,9,10], whereas one study did not observe differences between migraineurs and non-migraineurs [11] and another reported the opposite pattern [12]. Similarly, some studies reported higher levels of Persistence in migraineurs with respect to non-migraineurs [7, 9, 10], but some studies did not observe differences between the two groups [11, 12]. Analogous inconsistencies have been reported for self-directedness, whereas more consistent findings (no difference between migraineurs and non-migraineurs) have been reported for other psychological traits included in the PM [7,8,9,10,11,12]. When the personality profile was explored by questionnaires assessing the psychological dimensions proposed by the Eysenck’s Three-Factor Model (TFM), migraineurs tended to show higher scores than non-migraineurs in neuroticism [13,14,15,16], whereas mixed results have been reported about levels of extraversion and psychoticism [13,14,15]. Finally, only two studies [17, 18] evaluated the personality traits according to the Big-Five Model and found high level of neuroticism, conscientiousness and agreeableness in migraineurs compared to non-migraineurs. Assessing personality traits and psychosomatic mechanisms of migraine patients might allow developing non-pharmacological treatment, to make them less vulnerable to stress and more skilled in coping with pain. Unfortunately, previous findings on the personality traits in migraineurs are heterogeneous and did not reveal a well-established personality profile. On this basis, we performed two separate meta-analyses on primary studies investigating personality traits according to current psychological models in migraineurs and non-migraineurs. We explored the possible influence of demographic and clinical aspects on each personality trait in migraineurs, as suggested in a recent review [19].

Some studies have revealed a significant association between depression and migraine [20] but, until now, no study has explored the possible crucial role of the personality traits in modulating the relationship between depression and migraine. Therefore, we performed a meta-analysis on studies focusing on the depression and personality traits in migraineurs.

Methods

Search strategy and study eligibility criteria

A systematic literature search was performed up to March 2019 using PsycInfo (PROQUEST), PubMed and Scopus entering the following search terms: “personality” or “temperament” or “neurot*” or “psychoticism” or “negative emotionality” or “extraversion” or “introversion” or “openness to experience” or “cognitive rigidity” or “rigidity” or “agreeableness” or “conscientiousness” or “impulsiv*” or “novelty seeking” or “harm avoidance” or “reward dependence” or “persistence” and “Migraine”. This search was supplemented by hand searches of reference lists cited in the original and review articles. Two independent observers (F.G., C.B.) evaluated the results and resolved any disagreement by discussion or with recourse to a third arbitrator (G.S.). A primary study was included in the meta-analysis if it (i.) was published in peer-reviewed journal in English from 1980 to 2019, (ii) provided results about comparison on personality traits between patients with a diagnosis of migraine (i.e. migraine without or with aura) according to clinical criteria [21] and individuals without migraine or headache and (iii) reported statistical results (mean, standard deviation, standard difference, p value) about comparisons on personality traits between migraineurs and non-migraineurs. We excluded conference proceedings, letters to the editor, commentaries, theses, studies performed on animals and single cases studies. The data presented in more than one publication were used in their primary version (first publication). When two or more studies included patients from the same sample, we selected the primary study with the highest number of patients.

Data extraction and coding

Data extracted and coded from the primary articles included (i) characteristics of the publication (e.g. authors, journal status, year of publication, journal) and (ii) characteristics of the sample (e.g. total sample size, gender, duration and frequency of headache, subjective pain rating). To perform a meta-regression where depression was the outcome and personality trait was the moderator, we selected primary studies which provided mean score of depression questionnaire and mean score of personality traits in patients with migraine. We defined as outcomes seven dimensions put forward by PM, including both temperament (i.e. novelty seeking, harm avoidance, reward dependence and persistence) and character (i.e. self-directedness, cooperativeness and self-transcendence) and three dimensions included in the TFM (i.e. neuroticism, psychoticism and extraversion). When a study evaluated personality traits by questionnaires not specifically developed on the basis of the PM or TFM, we decided to use the dimensions considered theoretically associated to each personality model.

Statistical analyses

We synthesized study data using meta-analytic methods (ProMeta 3; Intenovi, 2015). Meta-analytic method was used to synthesize study data: we computed the effect sizes (ES) from data reported in the primary studies (e.g. means and standard deviations; p values) using Hedges’ g unbiased approach (like the Cohen d statistic). Negative values of the Hedges’ g indicated that migraineurs had lower scores than non-migraineurs on each outcome. Conventionally, values of Hedges’ g < 0.20 indicate small effects, values of about 0.50 moderate effects, and values of about 0.80 large effects. For each ES, 95% confidence interval, variance, standard error and statistical significance were computed. Moreover, ES across studies were pooled for obtaining an overall effect size with the inverse-variance method. We used the random-effects model since it is a conservative approach useful to account for different sources of variation among studies and to generalize the meta-analytic finding beyond the studies included here. Q and I2 statistics index were computed to assess the heterogeneity among the studies. A significant Q value indicates a lack of homogeneity of findings among studies; the proportion of observed variance that reflects real differences in ES was estimated by I2. A value of 25, 50 and 75% was considered as low, moderate and high, respectively. Sensitivity analyses were performed to check the stability of study findings, computing how the overall ES would change removing one study at a time. To explore the publication bias, we applied the funnel plot, a scatter plot of the ES estimated from individual studies against a measure of their precision (e.g. their standard errors). To evaluate the funnel plot more reliably, we employed the Egger’s regression method, which to statistically test the asymmetry of the funnel plot, with non-significant results indicative of absence of publication bias. Moreover, we applied the trim and fill procedure, an iterative non-parametric statistical technique, which evaluates the effect of potential data censoring on the result of the meta-analyses. In this method, the absence of publication bias is indicated by zero trimmed studies, or in the presence of trimmed studies, by a trivial difference between the observed and the estimated ES. To further explain heterogeneity across studies and to explore the possible influence of demographic (age at evaluation, gender, years of schooling) and clinical variables (i.e. duration and frequency of headache, subjective pain rating) on each outcome, we performed several meta-regressions. Meta-regressions were also performed to evaluate which personality trait moderate the severity of depression in migraine. The references for methodological procedures were reported in Supplementary Material 1, C. Statistical analyses were conducted with the meta-analytic software ProMeta 3.0.

Results

Study selection

Figure 1 shows the flow diagram based on PRISMA statement. The initial search identified 3777 articles; after removing duplicates, we obtained 2611 articles. After full-text assessment, 181 studies were considered eligible. According to inclusion and exclusion criteria, we included eight studies [7,8,9,10,11,12, 22, 23] in the meta-analysis on the personality defined according to PM and 10 studies [13,14,15,16,17,18, 22,23,24,25] in the meta-analysis on personality defined according to TFM. Moreover, we included a study by Karmakar et al. [18] where a self-reported provider diagnosis (SRPD) of migraine was used. As three studies [14, 17, 24] provided mean and standard deviations of migraineurs with and without aura as two separate groups, these were collapsed into one group by calculating the pooled means and standard deviations from statistics provided in each paper. As for the association between personality traits and the type of the migraine, among the eight studies included in the meta-analysis on the personality defined according to PM, four studies specified that the sample included migraineurs without aura [8, 10, 22, 23] whereas the remaining four studies did not specify the type of migraine [7, 9, 11, 12]. Out of 10 studies included in the meta-analysis on the personality defined according to TFM, five included migraineurs without aura [14, 17, 22,23,24], three included migraineurs with aura [14, 17, 24], whereas the remaining five studies did not specify the type of migraine [14,15,16, 18, 25]. Characteristics of primary studies included in the meta-analysis are reported in Table 1.

Fig. 1
figure 1

Flowchart of the selection process of the primary studies (PRISMA)

Full size image
Table 1 Characteristics of primary studies included in the meta-analysis
Full size table

Personality profile according to PM

Novelty seeking

The ES (= − 0.16; 95% CI = − 0.40–0.08) was not significant and there was no publication bias. The heterogeneity among the studies was significant and moderate. After removing Wang et al.’s study [22], the ES (= − 0.09; 95% CI = − 0.31–0.13) remained not significant and the heterogeneity became not significant (Table 2; Supplementary Material Figure 2A).

Table 2 Summary of meta-analytic results of the following personality domains: novelty seeking, harm avoidance, reward dependence, perseverance/persistence, self-directedness, cooperativeness, self-transcendence
Full size table

Harm avoidance

The ES (= 0.45; 95% CI = 0.07–0.83) was significant and moderate: migraineurs scored higher than non-migraineurs. There was no publication bias, whereas the heterogeneity across the studies was high. After removing Park et al.’s study [8], ES (= 0.31; 95% CI = 0.01–0.60) remained significant and the heterogeneity decreased (Table 2; Supplementary Material Figure 2B).

Reward dependence

The ES (= 0.01; 95% CI = −0.15-0.18) was not significant, without publication bias and heterogeneity across the studies (Table 2; Supplementary Material Figure 2C).

Persistence

The ES (= 0.25; 95% CI = − 0.11–0.61) was not significant without publication bias. The heterogeneity across the studies was significant and high (I2 = 78%). After removing Boz et al.’s study [12], the ES (= 0.37; 95% CI = 0.01–0.72) became significant and moderate and the heterogeneity decreased slightly (I2 = 71%) (Table 2; Supplementary Material Figure 2D).

Self-directedness

The ES (= − 0.16; 95% CI = − 0.53–0.21) was not significant, without publication bias, but the heterogeneity was high. After removing Sanchez-Roman et al.’s the study [9], the ES (= − 0.33; 95% CI = −0.57 to − 0.09) became significant and moderate whereas the heterogeneity decreased (Table 2; Supplementary Material Figure 3A).

Cooperativeness

The ES (= 0.11; 95% CI = − 0.25–0.48) was not significant. There was no publication bias; the heterogeneity was significant and high. After removing Sanchez-Roman et al.’s the study [9], the ES (= − 0.05; 95% CI = −0.28–0.17) remained not significant and the heterogeneity became not significant (Table 2; Supplementary Material Figure 3B).

Self-transcendence

The ES (= − 0.04; 95% CI = − 0.25–0.17) was not significant, without publication bias or the heterogeneity across the studies (Table 2; Supplementary Material Figure 3C).

Comparison between migraineurs with and without aura

A subgroup analysis to identify possible differences on personality traits between the two groups was not performed as only 4/8 studies specified the type of migraine.

Personality profile according to TFM

Neuroticism

The ES (= 0.56; 95% CI = 0.34–0.78) was significant and moderate: migraineurs scored higher than non-migraineurs. There was no publication bias, whereas the heterogeneity across the studies was high. After removing Brandt et al.’s study [13], the ES (= 0.47; 95% CI = 0.32–0.63) remained significant and moderate but the heterogeneity decreased (Table 3; Supplementary Material Figure 4A).

Table 3 Summary of meta-analytic results of the following personality domains: Neuroticism, Psychoticism, Extraversion, Lie
Full size table

Psychoticism

The ES (= − 0.04; 95% CI = −0.39–0.30) was not significant, without publication bias; the heterogeneity across the studies was significant and high (Table 3; Supplementary Material Figure 4B).

Extraversion

The ES (= − 0.08; 95% CI = −0.14 to − 0.03) was significant but very low: migraineurs scored lower than non-migraineurs. No publication bias and heterogeneity across studies were found (Table 3; Supplementary Material Figure 4C).

Lie

The ES (= − 0.19; 95% CI = − 0.45–0.07) was not significant, without publication bias. The heterogeneity across the studies was high and significant. After removing Brandt et al.’s study [13], the ES (= − 0–09; 95% CI = −0.33–0.15) remained not significant and the heterogeneity became not significant (Table 3; Supplementary Material Figure 4D).

Comparison between migraineurs with and without aura

A subgroup analysis on five studies [14, 17, 22,23,24] revealed no difference between the two groups on neuroticism (Q = 0.14, df = 1, p = 0.710) and extraversion (Q = 0.17, df = 1, p = 0.682) (Supplementary Material Figure 5).

Moderator analysis

Age moderated the level of neuroticism: higher levels were found in younger than in older migraineurs (B = − 0.04, p = 0.038). Sex had no significant effect on each trait. We could not assess the possible effect of education, frequency and duration of attacks; pain intensity on the relationship between personality traits and migraine since there were not at least 10 samples to 1 covariate as suggested by Borenstein et al. [26].

Personality traits as moderators of depression in migraineurs

Studies that provided mean score of depression and mean score of personality traits in patients with migraine were reported in Table 4. Among the personality traits, only neuroticism moderated the severity of depressive symptomatology (B = 10.46, p = 0.034). Other personality traits did not influence the severity of the depression.

Table 4 Characteristics of primary studies that provided mean score of depression and personality traits in patients with migraine
Full size table

Discussion

Personality profile associated with migraine

The present meta-analytic study focused on personality traits associated with migraine. When personality was conceptualized according to PM, we found a higher level of both harm avoidance and persistence and a lower level of self-directedness in migraineurs with respect to non-migraineurs. No significant difference was found between migraineurs and non-migraineurs on the remaining dimensions of temperament and character. When we took into account the TFM, a higher level of neuroticism and a lower level of extraversion were significantly related to migraine. Only neuroticism moderated the severity of depressive symptomatology in migraineurs. A higher level of harm avoidance seems thus to be a distinctive personality trait of migraineurs [7,8,9,10]. Since in PM harm avoidance is characterized by behavioural inhibition, excessive fear/worry, our findings suggested that migraineurs are prone to have a pessimistic apprehension in anticipation of future problems and to show passive avoidant behaviours such as fear of uncertainty and rapid fatigability [2]. Only two studies did not report any difference between migraineurs and non-migraineurs on this personality trait, but the features of their sample might have biased these reports. Boz et al. [12] enrolled a sample of migraineurs smaller and older (n = 51; age = 28.24) than the sample of non-migraineurs (n = 82; age 25.84), whereas Nylander et al. [11] explored the personality traits in a small sample of 26 individuals belonging to the same family with a dominant autosomal inheritance pattern for migraine. On this basis, we would conclude that high level of harm avoidance is a very consistent finding in the migraineurs. The present meta-analysis cannot address neurobiological issues, but the finding of a consistent association between high level of harm avoidance and migraine might suggest that both conditions share common neurobiological mechanisms. In particular, it has been suggested that both harm avoidance and migraine might depend on a dysfunction of serotonergic transmission [8]. Indeed, high brain serotonin levels and lower 5-HT4 receptor binding within neocortex were found in migraineurs between attacks, whereas low serotonin levels occur typically during a migraine attack [27, 28]. Moreover, high level of harm avoidance seems to be associated with altered serotoninergic activity in dorsal raphe nuclei [29]. High level of persistence was another personality feature consistently associated with migraine [7, 9, 10]. According to PM, high levels of persistence characterize eager, ambitious, determined, resolute individuals who tend to persevere despite frustration and fatigue; this trait expresses the tendency to maintain unrewarded behaviours and correlate with high rigidity and obsessiveness [2]. From this perspective, migraineurs might be characterized by a tendency to be vulnerable to stress and a higher risk of developing depressive and anxiety symptoms when exposed to stressful events [30]. Moreover, this trait seems to influence the clinical course of migraine, as it might favour progression of episodic migraine to the chronic form, and/or the tendency to develop overuse of symptomatic medications [31, 32]. On neurobiological grounds, persistence has been related to altered glutamatergic systems [2], which have been involved in migraine pathophysiology [33]. The present meta-analysis would thus encourage future studies addressing the possible links between persistence and migraine. Among modifiable traits (characters) included in PM, only self-directedness differed in migraineurs and non-migraineurs and was lower in migraineurs. This finding, together with the high level of persistence, would confirm that patients suffering from migraine have poor coping skills and are vulnerable to stress. In the only study reporting conflicting results, Sanchez-Roman et al. [9] compared migraineurs with non-migraine individuals affected by chronic pain conditions, a sample characterized by very low levels of self-directedness [34]. Moreover, two studies [11, 12] did not find differences in self-directedness between migraineurs and control individuals, but, as specified above, the results of these papers could be biased by sample-related factors. Moreover, it is worth mentioning that Boz et al. [12] did observe a specific significant difference in one single subscale of the self-directedness character. Low levels of self-directedness and high levels of harm avoidance have been associated with altered serotonergic activity [35], an issue deserving to be investigated. Our meta-analysis of studies on the personality profile according to TFM revealed a higher level of neuroticism and a lower level of extraversion in migraineurs with respect to non-migraineurs. The strong relationship between migraine and high levels of neuroticism is in keeping with the finding of high level of harm avoidance described above, as the two personality traits overlap at least partially [36]. This finding would confirm that migraineurs might be less able to endure migraine pain, and, possibly, less able to cope with that pain [37]. Moreover, we found that high level of neuroticism was associated with more severe depressive symptoms supporting that the neuroticism is related to the tendency to experience negative emotions [3] and could be a predictor of development of depressive symptoms in migraineurs. Therefore, these findings suggested the clinical relevance of an early evaluation of personality profile in patients even early phase of disease to identify patients with maladaptive coping abilities and therefore more at risk for depression [38]. Our moderator analysis showed that age moderated the degree of association between neuroticism and migraine: younger subjects tend to have higher level of neuroticism than older ones. This finding confirmed the idea of cross-sectional age differences in this personality trait [39]. Extraversion is thought to be related to the desire of being with others as opposed to being alone and pursuing solitary activities [4]. The observed low level of extraversion might suggest that migraineurs can be prone to experience less positive life events such as fulfilling social interactions and to show a maladaptive response to stress and a general feeling of malaise, which is one of the most common trigger for migraine attacks [40]. As five studies investigated the TFM personality traits specifically associated with MwA [14, 17, 24] or MwoA [14, 17, 22, 23], we could perform a subgroup analysis. Our findings did not show differences on neuroticism and extraversion personality traits between the two groups. However, this finding should be considered cautiously due to the few studies comparing MwA and MwoA patients on the personality profile [14, 17, 22,23,24] and needs to be addressed in future studies.

Limitations

The present meta-analysis is characterized by some limitations: we could not investigate the possible moderator effect of relevant clinical variables such as duration, drugs, frequency, pain intensity and type (i.e. chronic versus episodic) on the relationship between personality traits and migraine because these clinical aspects were not reported in enough primary studies. Moreover, we could not exclude that personality profile of migraineurs evidenced in the current meta-analysis was a consequence of the disease since personality profile has never been investigated before the onset of the migraine. Finally, since we included only cross-sectional studies, we could not investigate the causal role of the personality traits on the onset of migraine.

Clinical implications

Despite the abovementioned limitations, our results might have relevant clinical implications. Some personality traits could negatively influence long-term history of migraine and the response to drug treatment [41]. The meta-analytic evidence of a specific personality migraine reinforces the relevance of psychological evaluation to obtain prognostic information and to possibly identify patients at high risk of development of psychopathological disturbances such as depression and anxiety. Assessing personality traits can also help in planning non-pharmacological therapies such as psychoeducational interventions focused on personality traits, i.e. self-directedness and neuroticism. As migraineurs often show vulnerability to psychological distress, and reduced success in personal, social, cognitive and spiritual development because of maladaptive coping styles [38], might follow psychoeducational interventions aimed at enhancing self-directedness and neuroticism to improve their well-being. Finally, specific personality traits, especially neuroticism, could influence the Quality of Life (QoL) [42] and the type of coping strategies used to overcome unexpected troubles and tolerate stressful events. On the basis of these considerations, a well-established personality profile associated with migraine might allow developing psychological interventions to improve the exploitation of adaptive coping strategies and to early identify patients at risk of developing psychopathological disturbances such as depression.