We conducted a systematic search of all studies designed cohort and case-control studies addressing the association between migraine and breast cancer. A comprehensive literature search of numerous electronic databases including PubMed, Scopus, ScienceDirect and the Cochrane library was done up until February 1^st, 2015.

A search strategy was conducted using the following search terms in the titles, abstract or key words: migraine, migraine disorder, breast, breast cancer, case-control studies, cohort studies, and observational studies. The citations and references listed in retrieved articles were also reviewed to identify additional related studies.

Two authors reviewed the retrieved studies independently to identify eligible studies by using the following criteria: 1) Study design: cohort or case-control study, 2) the breast cancer patients were confirmed by medical records 3) the migraine status was confirmed by medical records, 4) the numbers of case and control (in case-control studies) and exposed and non-exposed groups (in cohort studies) were reported or the relevant data was available to calculate the odds ratio (OR) or risk ratio(RR). Women with breast cancer were included regardless of age, race, menopausal and marital status.

Two independent authors read full-text articles and the following information was extracted: name of the first author, publication year and location of study conduction, study design (cohort or case-control study), number of cases and controls or exposed and non-exposed, age at migraine diagnosis, histological subtype (ductal and lobular), and hormone (estrogen (ER) and progesterone (PR)) receptor status (ER+/PR+, ER+/PR−, ER−/PR−). The quality of studies was assessed according to the STROBE statement and the studies according with STROBE criteria were defined as high-quality studies [17] .

The pooled ORs with 95% confidence interval (CI) estimating the association between migraine and breast cancer were obtained using the random effect model. Cochran’s Q test was used to identify the heterogeneity of results across studies and it was quantified using the I² statistic. Q statistic with p-value <0.10 or I² statistic >50% was considered as significant heterogeneity across studies. The between-study variance was estimated using tau-squared (t² or Tau²) statistic [18] . We used Egger’s linear regression test [19] to investigate publication bias (p < 0.05 set as significant level). Meta-analysis was performed by the comprehensive meta-analysis software version 2.0. The PRISMA statement was used as a guide in the reporting of this study [20] .

Table 1 presents the characteristics of the studies included in the meta-analysis. Of 652 retrieved studies, we

CR: Cancer Registry at the oncology center of Isfahan University of Medical Sciences, Iran; SEER: Surveillance Epidemiology and End Results cancer registry; CSS: Cancer Surveillance System; WHI: Women’s Health Initiative; WHS: Women’s Health Study.

eventually included 6 population-based studies in the meta-analysis including two cohort and four case-control studies involving 130812 and 14396 people, respectively. Figure 1 presents the flowchart of the literature search process.

Table 2 illustrates the summary effect sizes of breast cancer risk and migraine history by age at migraine diagnosis, histological subtype and hormone receptor status.

Summary results of six population-based studies (Figure 2) indicated that having any history of migraine was inversely associated with breast cancer risk (OR = 0.77; 95% CI: 0.64, 0.92).

We stratified the migraine history by the age at diagnosis (Figure 3). Pooled result from three studies showed that both age at migraine diagnosis < 20 years and over 20 years decreased the risk of breast cancer (OR = 0.66, 95% CI 0.48, 0.90 and OR = 0.68, 95% CI 0.47, 0.99, respectively). That means the women who reported a clinical diagnosis of migraine in age <20 or >20 years were significantly at lower risk of breast cancer.

The summary results of associations between migraine history and breast cancer risk showed that the association does not vary by histological type compared to non-migraineur women (Figure 4). That means women with a history of migraine had similarly reduced risks of both ductal and lobular carcinomas (OR = 0.84, 95% CI 0.70 - 1.01 and OR = 0.79, 95% CI 0.70, 0.90, respectively). Test for histological subgroup differences was not significant (p = 0.585). Overall, women with migraine history had a 19% reduced risk of breast cancer regardless of histological subgroup.

Pooled result showed that women with migraine history had a 15% reduced risk of breast cancer regardless of hormone receptor status (OR = 0.85, 95% CI 0.77, 0.94). Figure 5 shows the forest plot of OR estimates of association between migraine and breast cancer risk by hormone receptor status. We did not observe significant differences with regard to hormone receptor status (p = 0.253). On the other hand, migraineur women had a lower risk of ER+/PR+ (OR = 0.80, 95% CI 0.68, 0.93), ER+/PR− (OR = 0.82, 95% CI 0.67, 0.99), and ER−/PR− breast cancer (OR =0.97, 95% CI 0.80, 1.17).

Significant heterogeneity between studies was noted in the studies. The results of Q-test showed that the studies

were significantly heterogeneous (p < 0.001). The I² and tau statistics were 88.5% and 0.04, respectively (Figure 2). Of six included studies assessed the effect of migraine on breast cancer risk, five studies reported negative associations (four out of which were statistically significant) and one study reported non-significant positive association. The heterogeneity between studies with related statistics is shown in Table 2. We couldn’t assess the publication bias using the funnel plot, because the number of included studies in the meta-analysis was relatively small. No publication bias was detected by Egger’s regression (p = 0.647).