Selection of studies
A total of 246 studies were imported for initial screening. Abstracts were initially screened by two blinded reviewers. Each reviewer screened each abstract, and studies that were agreed on by both reviewers advanced to full text screening. 77 full text studies were reviewed by the two blinded reviewers. Each reviewer screened each full text article and studies deemed appropriate by both reviewers were included in the final review. After excluding any nonclinical studies as well as studies that did not directly study estrogen, 19 studies were included in the final review. This process is summarized in the PRISMA flow diagram in Figure 1. A summary of findings can be seen in Table 2.
The 19 studies included in our review consisted of 3 randomized controlled trials (RCT), 3 cross sectional studies, 3 prospective cohort studies, 2 open-label clinical trials, 4 crossover studies (2 of which were double-blind placebo-controlled), and 4 survey/diary-based studies consisting of 37,139 total patients. The patients in these studies included 1,646 menstruating women, 35,394 postmenopausal women, 50 male-to-female transgender women, and 49 pregnant women. Further details of patient characteristics are summarized in Table 1. Of the 19 studies, 12 studies investigated the effect of estrogen withdrawal on migraine occurrence and concluded that the withdrawal of estrogen during menses is a key factor in the precipitation of migraines, with the majority of these studies focusing on menstrual migraines in premenopausal women.4, 7-17 In particular, two studies found that priming with estrogen followed by a drop in serum estradiol levels below 45-50 pg/mL increased the risk of migraine precipitation.13, 17
Three studies assessed the effects of hormone replacement therapy (HRT) on migraine in postmenopausal women with results varying by estradiol dosing, but generally concluding that estrogen replacement increased the incidence of migraine.18-20 Three other studies found that women with a history of migraine had an increased sensitivity to physiologic fluctuations in estradiol levels.16, 21, 22 Lastly, four survey/diary-based studies suggested that migraines most often occurred perimenstrually and were more common in male-to-female transgender patients on HRT when compared to the general population.6, 10-12 These results suggest that estrogen is a major hormone implicated in migraine pathogenesis and that physiologic withdrawal of estrogen during menses likely plays a role in this condition. However, one study followed pregnant women with a history of migraine and found that migraine frequency did not change significantly as pregnancy progressed.23
The Initial Link Between Estrogen and Migraines: Estrogen Withdrawal Hypothesis
Estrogen’s association to migraines was first demonstrated by Somerville in 1972, providing for the first time an explanation for menstrual migraines.17 Somerville found that he was able to delay menstrual migraine attacks by up to 9 days by treating participants with supplemental estrogen. He noted that migraines were particularly triggered when estradiol levels fell below 45-50 pg/mL during the perimenstrual period. Somerville concluded that estrogen plays a large role in the precipitation of menstrual migraines and that estrogen withdrawal during menses is the primary trigger. Although these claims inspired decades of research on the topic, Somerville’s study only consisted of 6 subjects and was prone to confounding bias due to some participants being close to the age of menopause.
The phenomenon of estrogen withdrawal was later studied by Lichten in 1996 in 28 postmenopausal women taking HRT and with a history of severe menstrual migraines prior to menopause.13 Lichten administered a one-time intramuscular injection of 5 mg depo-estradiol cypionate and serially tracked serum estradiol levels until all participants experienced a migraine. He found that all participants in the experimental group experienced a migraine 18 +/- 4 days after administration of the injection and that the average serum estradiol level at the time of migraine was between 45 and 50 pg/mL, as Somerville had found in his study. However, no participants in the control group (no prior history of migraine) experienced a migraine during the course of the study. Lichten concluded that is likely a genetic component in migraine pathogenesis and that a drop in estradiol levels to below 50 pg/mL after a period of priming with higher levels can be a trigger for migraine. Like the Somerville study, we assigned this study a very low GRADE score because of the limited sample size and confounding effects of ongoing HRT by participants.
Estrogen Withdrawal Hypothesis—Supporting Studies
In support of Somerville findings, we found six studies in our review that investigated the role of estrogen withdrawal in the precipitation of menstrual migraine. In 1986, de Lignieres conducted a double-blind placebo-controlled crossover study based on Somerville’s initial project and also concluded that physiologic withdrawal of estrogen precipitates migraine and supplementing estrogen can prevent migraine.9 De Lignieres administered 1.5 mg of percutaneous estradiol 48 hours prior to expected migraine onset in menstruating women with a history of consistent menstrual migraine in their last 12 menstrual cycles and found that only 31% of participants experienced a menstrual migraine after treatment compared to women in the placebo group. While this was consistent with Somerville’s work and improved on the study design, it was limited by the small sample size of 20 women and advanced age of participants (average age 42.5).
In 1993, Cachrimanidou’s multicenter RCT of 300 patients at family planning clinics in Sweden found that changing the oral contraceptive (OCP) dosing regimen from a traditional 3 weeks on, 1 week off schedule to a 9 weeks on, 1 week off schedule reduced the frequency of migraine from 17.3% to 9.7% (p < 0.01).15 We considered this study to be high GRADE due to the large sample size and RCT study design. Cachrimanidou’s study was bolstered in 2006 when MacGregor found that daily administration of percutaneous estradiol gel (1.5 mg) from 6 days prior to menses until day 2 of menses resulted in a 22% reduction of migraine frequency (RR 0.78; 95% CI, 0.62-0.99, p = 0.04).14 However, MacGregor also found that there was a 40% increase in migraine frequency 5 days after discontinuing the intervention (RR 1.40; 95% CI, 1.03-1.92, p = 0.03). Cachrimanidou’s and MacGregor’s findings indicated that preventing estrogen withdrawal prior to menses can delay estrogen withdrawal migraine.
Calhoun conducted two studies that also supported the role of estrogen withdrawal in migraine pathogenesis.7 In 2004, Calhoun followed 20 women with a history of menstrual migraine and showed that estrogen replacement with 0.9 mg conjugated equine estrogens during the placebo week of OCP dosing significantly reduced migraines experienced per month by 76% (7.6 vs. 1.6 headache days per month) offering evidence that preventing the withdrawal of estrogen can reduce migraine incidence.7 Calhoun’s group continued their efforts in 2008 investigating a prospective cohort of 229 menstruating women with intractable menstrual migraines showed that using ethinyl estradiol-containing oral contraceptives to prevent premenstrual estradiol decline by more 10 mcg reduced migraine days per month by 55.8% (p < 0.001) for women that experienced a resolution of their menstrual migraine.4 However, 36% of women in the study continued to experience persistent menstrual migraines with no reduction in frequency.
In 2003, Martin conducted a RCT of 21 menstruating women that were treated with GnRH agonist therapy to simulate medical oophorectomy.16 The experimental group was given estrogen add-back therapy to maintain serum estradiol at > 50 pg/mL (reported in prior studies as the threshold for triggering menstrual migraine) and was found to have 33.7% reduction in headache index (95% CI, 3.0-64.4%). However, Martin also found that participants in the experimental group had a 45% increase in headache index during days 1 and 2 of the study, which he attributed to the rise in serum estradiol immediately after estrogen add-back therapy was initiated. Martin concluded that both estrogen withdrawal and rise can precipitate migraines, and that patients with a history of menstrual migraine are very sensitive to changes in serum estradiol levels.
Studies Investigating Fluctuations of Estrogen on Migraines
Martin was not the first author to suggest that fluctuations in estrogen may contribute to migraine pathogenesis. In 1995, Lichten’s group investigated 29 menstruating women with medically intractable menstrual migraines who were treated for two months with 3.75 mg depo-leuprolide acetate to effectively induce medical menopause.21 Of the 29 women treated, 17 remained migraine-free at two months and reported an average 50% improvement in headache index. These 17 women were then treated for 10 more months with transdermal estradiol add-back therapy to maintain estradiol levels at a steady value. At the one-year mark, 14 women remained migraine-free. This study was able to conclude that ovarian hormones play a key role in migraine pathogenesis, and that reducing physiologic fluctuations in estrogen can help reduce the incidence of migraine. In a nearly identical study in 1997, Murray treated five menstruating women with intractable menstrual migraines with the same dose of depo-leuprolide acetate for 10 months with 0.1 mg daily transdermal ethinyl estradiol added back from month five onwards.22 Murray found that participants reported a 74% decrease in headache index after being treated with only GnRH analog therapy, and this increased to 80% after estrogen add-back therapy at five months. Murray also concluded that minimizing fluctuations in estrogen can reduce migraine incidence.
However, Amir’s study in 2005 contradicted Murray’s initial finding that treatment with only GnRH analog therapy reduced headache index (a metric for headache frequency, severity, and duration).8 Amir studied 98 women undergoing in vitro fertilization with GnRH agonists and found that use of GnRH agonists to reduce serum estradiol levels was associated with a 28.6% increased incidence of migraine (95% CI, 19.7-37.5%). In our assessment of study quality, we gave Murray’s findings a very low GRADE score due to the limited sample size and lack of statistical analysis. On the other hand, Amir’s study was given a moderate GRADE score because of the study design, sample size, and narrow confidence interval. Overall, these studies suggest that physiologic fluctuations in estrogen may play a role in migraine pathogenesis, but complete elimination of estrogen may precipitate migraines.
Survey Studies Investigating Menstrual Related Migraines
Four survey-based studies were included in our review, and all were found to support Somerville’s estrogen withdrawal theory. However, these were graded to be low-tier evidence due to possible reporting biases. Mattsson’s study surveyed 728 women and concluded that 75% of these women reported migraines occurring within –2 to +3 days of menstruation.10 However, this study surveyed patients ranging in age from 40 to 74, suggesting a significant risk of reporting and recall bias, particularly for postmenopausal women being asked about their prior menstrual symptoms. Similarly, Stewart surveyed 81 menstruating women between ages 18-55 that self-recorded their migraines and menstrual cycles over 98 days.11 Stewart noted an increased incidence of migraine perimenstrually - days 0 and 1 had OR of 2.04 (95% CI, 1.49-2.81) and days -1 and -2 had OR of 1.80 (95% CI, 1.40-2.30).
Johannes’ survey studied 28 menstruating women age 22-29 via 4-month self-reported diaries and found that these women reported a higher incidence of migraine during the first 3 days of menses compared to the rest of the cycle (OR 1.66; 95% CI, 1.21-2.26).12 Though Johannes’ findings were supported statistically, it was given a low GRADE score because of the study design and risk of reporting bias. In a unique study, Pringsheim surveyed 50 male-to-female transgenders taking antiandrogen and estrogen therapy and had a history of migraine.6 Pringsheim found that the prevalence of migraines in male-to-female transgenders (26%) was significantly higher than cis-gender males (7.5%) in the population (p < 0.05), but was not significantly different from the prevalence of migraines in cis-gender females (25%). These findings suggested that estrogen might be a causative factor in migraine pathogenesis.
Estrogen’s Effects on Migraines in Postmenopausal Women
As menopause is a natural state of estrogen deficiency, postmenopausal women have been studied to investigate the effects of estrogen and the use of HRT on migraines. In 2003, Misakian published a cross-sectional study of 17,107 postmenopausal women and investigated how the use of HRT affects migraines.18 Misakian found that HRT use significantly increased the risk of experiencing a migraine in postmenopausal women (13% vs 9%, p < 0.001). This type of study has been conducted by other researchers as well, yet has yielded varying results. Brandes conducted a comprehensive cross-sectional study of 18,221 postmenopausal women and stratified the results into three groups based on the dose of HRT being taken - low (< 0.3 mg/day), intermediate (0.625 mg/day), and high (> 0.9 mg/day).5 Brandes reported that the intermediate dose group had a significantly lower risk of migraine frequency (OR 1.28; 95% CI, 1.10-1.48, p = 0.001) than the low (OR 2.00; 95% CI, 1.51-2.65, p=0.001) and high (OR 1.72; 95% CI, 1.39-2.13, p = 0.002) groups compared to the general population. We considered both of these studies to be high GRADE evidence because of their large sample sizes and efforts to minimize confounding bias, however it should be noted that both studies utilized patients from the Women’s Health Study and that there is likely a large overlap in the patients used in both studies.
Like Brandes and Misakian, Facchinetti also studied the effects of HRT on migraine pathogenesis in postmenopausal women.20 Facchinetti conducted a RCT of 38 postmenopausal women and stratified the results into three groups depending on the amount of HRT administered. In contrast to Brandes’ study, which showed a decrease in migraine frequency with 0.625 mg/day estradiol dosing, all three HRT regimens in Facchinetti’s study significantly increased migraine attack frequency (2.2 days per month vs 3.8, p < 0.001), (3.4 days per month vs 4.9, p < 0.001), (3.4 days per month vs 5.6, p < 0.001) over the course of 6 months, and patients also reported increased severity of headache after starting therapy. Notably, two of the three dosing regimens used by Facchinetti were significantly higher than the doses used by Brandes. Although the results between the two studies differed, we still considered the Facchinetti study to be moderate GRADE because it was a RCT with strong statistics that was limited only by small sample size. The conflicting data on HRT in postmenopausal women, especially compared to the estrogen withdrawal hypothesis in menstruating women, highlights the complexity of estrogen’s role in the pathogenesis in migraine headaches. Overall, these findings suggesting that postmenopausal women with a history of migraine are sensitive to increases in estrogen caused by HRT despite being in a low estrogen state.
Evidence Quality / GRADE
Using the GRADE framework, we assessed 3 studies to be high GRADE, 6 as moderate, 6 as low, and 4 as very low. The 3 high GRADE studies were a RCT15 and two cross sectional studies5, 18 which had very large sample sizes and narrow confidence intervals in their statistical analysis. The majority of our included studies were rated as moderate or low mainly due to smaller sample size or the presence of confounding bias. Lastly, the 4 studies rates as very low were survey/diary-based6, 10-12 and had a high risk of confounding and recall bias.