Prepregnancy Migraine, Migraine Phenotype, and Risk of Adverse Pregnancy Outcomes

Abstract

Background and Objective

Migraine is a highly prevalent neurovascular disorder among reproductive-aged women. Whether migraine history and migraine phenotype might serve as clinically useful markers of obstetric risk is not clear. The primary objective of this study was to examine associations of prepregnancy migraine and migraine phenotype with risks of adverse pregnancy outcomes.

Methods

We estimated associations of self-reported physician-diagnosed migraine and migraine phenotype with adverse pregnancy outcomes in the prospective Nurses' Health Study II (1989–2009). Log-binomial and log-Poisson models with generalized estimating equations were used to estimate relative risks (RRs) and 95% CIs for gestational diabetes mellitus (GDM), preeclampsia, gestational hypertension, preterm delivery, and low birthweight.

Results

The analysis included 30,555 incident pregnancies after cohort enrollment among 19,694 participants without a history of cardiovascular disease, diabetes, or cancer. After adjusting for age, adiposity, and other health and behavioral factors, prepregnancy migraine (11%) was associated with higher risks of preterm delivery (RR = 1.17; 95% CI = 1.05–1.30), gestational hypertension (RR = 1.28; 95% CI = 1.11–1.48), and preeclampsia (RR = 1.40; 95% CI = 1.19–1.65) compared with no migraine. Migraine was not associated with low birthweight (RR = 0.99; 95% CI = 0.85–1.16) or GDM (RR = 1.05; 95% CI = 0.91–1.22). Risk of preeclampsia was somewhat higher among participants with migraine with aura (RR vs no migraine = 1.51; 95% CI = 1.22–1.88) than migraine without aura (RR vs no migraine = 1.30; 95% CI = 1.04–1.61; p-heterogeneity = 0.32), whereas other outcomes were similar by migraine phenotype. Participants with migraine who reported regular prepregnancy aspirin use had lower risks of preterm delivery (<2×/week RR = 1.24; 95% CI = 1.11–1.38; ≥2×/week RR = 0.55; 95% CI = 0.35–0.86; p-interaction < 0.01) and preeclampsia (<2×/week RR = 1.48; 95% CI = 1.25–1.75; ≥2×/week RR = 1.10; 95% CI = 0.62–1.96; p-interaction = 0.39); however, power for these stratified analyses was limited.

Discussion

Migraine history, and to a lesser extent migraine phenotype, appear to be important considerations in obstetric risk assessment and management. Future research should determine whether aspirin prophylaxis may be beneficial for preventing adverse pregnancy outcomes among pregnant individuals with a history of migraine.

Migraine is a disabling neurovascular disorder characterized by moderate to severe headache with photosensitivity, phonosensitivity, nausea, vomiting, and/or aversion to physical activity that affects 14% of adults older than 18 years.^1,2 Among individuals with migraine, 30% experience aura, which are transient neurologic symptoms (e.g., visual, sensory, or language disturbances) that precede headache onset.³ Women are 2 to 3 times more likely than men to experience migraine in their lifetime, and the 3-month prevalence of migraine is highest (23.5%) among women aged 18–44 years.² In women, incidence of migraine with aura peaks around 12 and 13 years of age, whereas peak incidence of migraine without aura occurs between ages 14 and 17.⁴

Despite remarkable sex differences in migraine prevalence and a disproportionate effect on women of reproductive age, the relationship between migraine and pregnancy outcomes remains poorly understood. Mechanistically, migraine and adverse pregnancy outcomes—particularly hypertensive disorders in pregnancy (HDP), including gestational hypertension and preeclampsia—may be linked through shared pathophysiologic characteristics such as inflammation,^5,6 endothelial dysfunction,^7,8 and platelet activation.^9,10 Individuals with migraine have a higher prevalence of cardiovascular risk factors that are associated with adverse pregnancy outcomes, including obesity,¹¹ hyperlipidemia,¹² and chronic hypertension,¹³ which may confer greater obstetric risks.^14-16 Moreover, adverse pregnancy outcomes and migraine, particularly migraine with aura, are each consistently associated with higher risks of coronary heart disease and ischemic stroke,^17,18 suggesting shared etiology.

Existing evidence is insufficient to determine whether migraine history and migraine phenotype may serve as clinically useful markers of obstetric risk. Some epidemiologic studies, though not all, have reported positive associations of migraine and adverse pregnancy outcomes including gestational diabetes mellitus (GDM), HDP, preterm delivery, and low birthweight.^19-34 Most of the existing literature is limited to case-control studies prone to selection and recall biases, or retrospective registry-based studies that defined migraine status through migraine-related hospitalization or migraine-specific prescriptions.^{22,23,25,27-33} A few, mostly small, prospective studies reported variable associations of migraine or migraine comorbidities with HDP,^19,21,24,34 preterm delivery (overall, spontaneous, and medically indicated),^20,21,34 and low birthweight,²¹ but incompletely adjusted for important prepregnancy confounders such as body mass index (BMI), analgesic use, chronic hypertension, and/or oral contraceptive use. Finally, the relationship between migraine with aura, the phenotype most strongly associated with vascular risk, and adverse pregnancy outcomes has not yet been examined in a prospective study. To address these gaps, we evaluated associations of prepregnancy physician-diagnosed migraine and migraine phenotype with gestational hypertension, preeclampsia, preterm delivery, low birthweight, and GDM among 30,555 incident pregnancies in the Nurses' Health Study II (NHSII), a prospective cohort study of US nurses.

Methods

Study Population

The NHSII is a prospective cohort study of 116,429 US female registered nurses aged 25–42 years in 1989 when they responded to a baseline questionnaire. Updated information on medical conditions and behavioral factors has been collected through biennial questionnaires since 1989, for which the cumulative follow-up rate has exceeded 90%.

Standard Protocol Approvals, Registrations, and Patient Consents

The NHSII protocol was approved by the institutional review board of Mass General Brigham, Boston, MA. Informed consent was obtained from all participants.

Assessment of Migraine

Participants indicated on the baseline questionnaire in 1989 whether they had ever received a physician diagnosis of migraine and when they received the diagnosis (response options: before 1980; 1980–1984; or 1985–present). On the 1993 questionnaire, participants indicated if they received a physician diagnosis of migraine since June 1, 1991, and when the diagnosis occurred (response options: before June 1, 1991; June 1991–May 1993; or after June 1, 1993). Migraine was assessed similarly on the 1995 questionnaire. Prepregnancy migraine status was classified according to the nurse's response on the questionnaire proximal to but preceding the pregnancy. Migraine exposure status was cumulative such that all pregnancies after a migraine diagnosis were considered migraine-exposed. On the 2007 questionnaire, participants were asked if they had ever experienced a migraine headache (response options: yes or no) and whether they had aura (response options: yes or no). We classified participants with prepregnancy migraine and who responded affirmatively to having ever experienced aura with a migraine as having migraine with aura. Although the assessment of migraine was not validated in the NHSII, previous research in a comparable population suggests that agreement between self-reported physician-diagnosed migraine and ICHD-II criteria is >87%.³⁵

Assessment of Adverse Pregnancy Outcomes

On the baseline questionnaire in 1989, participants indicated whether they were currently pregnant and each age at which they completed pregnancies lasting 6 months or more and pregnancies lasting less than 6 months (including spontaneous and elective abortions). Incident pregnancies occurring after 1989 were reported on each biennial questionnaire. On the 2009 questionnaire, participants provided a comprehensive reproductive history that included detailed information on each pregnancy during their lifetime, including year of pregnancy end, pregnancy outcome, gestation length, birthweight, and pregnancy complications. Pregnancy outcomes included singleton live birth, multiple birth, miscarriage or stillbirth, tubal or ectopic pregnancy, or induced abortion. Gestation length was reported in categories (response options: <8, 8–11, 12–19, 20–27, 28–31, 32–36, 37–39, 40–42, ≥43 weeks), as was birthweight (response options: <5, 5–5.4, 5.5–6.9, 7–8.4, 8.5–9.9, ≥10 lbs). Diagnoses of GDM, preeclampsia, and gestational hypertension were based on the 2009 self-report. Preterm delivery was defined as birth before 37 weeks' gestation, and moderate and very preterm delivery were defined as birth between 32 and 37 and <32 weeks, respectively. Those who reported pregnancies on prior questionnaires but were missing pregnancy outcome information in 2009 were excluded (<4.9% of pregnancies). Pregnancies missing age or year of completion (<0.3%) and those in individuals with a prior diagnosis of cardiovascular disease, type II diabetes, or cancer (other than melanoma skin cancer) were further excluded from the analytic sample (<1.9%) to reduce potential confounding by these conditions.

Previous validation studies in the NHSII suggest high validity of self-reported pregnancy outcomes compared with medical records (GDM: 94% confirmation; preeclampsia: 89% confirmation; preterm delivery: 81% sensitivity (r = 0.75 for three-category variable [term, moderate preterm, very preterm]); birthweight: r = 0.74).^36-39

Assessment of Covariates

In 1989, participants reported demographic, behavioral, and health characteristics including self-reported race and ethnicity, current height and weight, cigarette smoking, age at menarche, oral contraceptive use, parity (defined as the number of pregnancies lasting 6 months or longer), history of infertility (defined as more than 12 months trying to conceive without success), and history of physician-diagnosed conditions (e.g., chronic hypertension, type 2 diabetes, cardiovascular disease, and cancer). Updated information on weight, physician-diagnosed conditions, smoking status, oral contraceptive use, infertility diagnosis, and parity was collected on each biennial questionnaire. BMI for each questionnaire cycle was calculated using nurses' self-reported height and weight. Information on physical activity was collected in 1991, 1997, 2001, and 2005 through questions regarding average time spent per week participating in specific activities, which were used to calculate metabolic equivalent of task (MET)-hours per week.⁴⁰ Alcohol intake was assessed in 1991 and every 4 years thereafter using a validated food frequency questionnaire.^41-43 The use of analgesic medications (i.e., aspirin, ibuprofen, acetaminophen, and COX-2 inhibitors) was queried in 1989 and 1993, and on each subsequent questionnaire. Information on antihypertensive medications (i.e., thiazide and furosemide diuretics, calcium channel blockers, beta-blockers, ACE-inhibitors, and “other” antihypertensive medications) was collected biennially. The use of tricyclic antidepressants was assessed on the 1993, 1997, and 1999 questionnaires.

Statistical Analysis

Age-standardized baseline characteristics of participants were estimated according to prepregnancy migraine status in 1989. Relative risks (RRs) and 95% CIs were estimated using log-binomial regression models or log-Poisson models in instances of model nonconvergence. The unit of analysis was pregnancies, and generalized estimating equations with an exchangeable working correlation structure accounted for within-person variability across multiple pregnancies contributed to the analysis per participant. All analyses were limited to pregnancies ending in live birth or stillbirth (i.e., pregnancies lasting ≥20 weeks).

Models were adjusted for variables identified a priori as potential confounders, including age at pregnancy (continuous), age at menarche (≤11, 12–13, ≥14 years), race/ethnicity (non-Hispanic White, other), and updated prepregnancy BMI (<18.5, 18.5–24.9, 25.0–29.9, >30 kg/m²), chronic hypertension diagnosis (yes/no), alcohol consumption (none, <5, ≥5 g/d), physical activity (<3, 3−<18, 18−<42, ≥42 MET-hrs/wk), smoking status (never, past, current), analgesic use (<2, ≥2 times/wk), ever oral contraceptive use (ever/never), infertility diagnosis (yes/no), and parity (nulliparous, 1, ≥2 pregnancies lasting >6 m). Updated covariates were assigned from the biennial questionnaires proximal to but preceding each pregnancy. In primary analyses, we assigned missing values (<14% for any single covariate) to the largest category of each covariate.

We examined associations of migraine phenotype with each adverse pregnancy outcome using a 3-level categorical variable (no migraine, migraine without aura, migraine with aura). We formally tested for heterogeneity by migraine phenotype using the contrast test method. We additionally examined whether associations of migraine with preterm delivery, gestational hypertension, and preeclampsia may be modified by prepregnancy aspirin use (<2×/week vs ≥2×/week), given its overall effectiveness in reducing risks of preterm delivery and HDP.⁴⁴ To determine whether associations of migraine with adverse pregnancy outcomes might vary according to established obstetric risk factors, we conducted analyses stratified by age (<35 years vs ≥35 years), BMI (<25 vs ≥25 kg/m²), and parity (primiparous vs multiparous). p-Interactions were estimated from 2-sided Wald tests of cross-product terms in multivariable models.

Since adverse pregnancy outcomes more frequently affect multiple gestations, we conducted sensitivity analyses restricted to singleton pregnancies. To assess potential residual confounding by antihypertensive medications and tricyclic antidepressant medications, which are commonly prescribed to treat or prevent migraine, we also conducted sensitivity analyses excluding pregnancies with any history of using these medications. Finally, to test whether our main results were robust to different methods of handling missing data, we conducted sensitivity analyses using multiple imputation by chained equations. In doing so, we generated 5 imputed data sets, analyzed each data set individually, and combined the parameter estimates and associated standard errors across the 5 data sets. All analyses were conducted using SAS v9.4 software.

Data Availability

Any data not published within the article will be shared at the request of other qualified investigators for purposes of replicating procedures and results. Further information, including the procedures for obtaining and accessing data from the Nurses' Health Studies II, is described at nurseshealthstudy.org/researchers (email: nhsaccess@channing.harvard.edu).

Results

Among the 19,694 participants in the analytic sample, 2,234 (11.3%) reported a history of physician-diagnosed migraine in 1989, with 1,078 (5.5%) of individuals classified as having migraine with aura and 1,156 (5.9%) classified as having migraine without aura. At baseline, individuals with migraine had higher physical activity levels and were more likely to report a history of infertility, overweight/obesity, and oral contraceptive use than those without migraine. Individuals with migraine more frequently reported current use of antihypertensive medications and analgesics, particularly aspirin, compared with nonmigraine individuals (Table 1).

Table 1.

Age-Standardized Baseline Characteristics of Women With Incident Pregnancies (n = 19,694) in the Nurses' Health Study II (NHSII) by Prepregnancy Physician-Diagnosed Migraine in 1989

In fully adjusted models, prepregnancy physician-diagnosed history of migraine was associated with greater risks of preterm delivery (RR = 1.17; 95% CI = 1.05–1.30), gestational hypertension (RR = 1.28; 95% CI = 1.11–1.48), and preeclampsia (RR = 1.40; 95% CI = 1.19–1.65) compared with no history of migraine (Table 2). Chronic hypertension, followed by analgesic use, and BMI, accounted for most of the modest differences between age-adjusted and multivariable-adjusted estimates. In analyses separately examining preterm and term preeclampsia, associations were strongest for term preeclampsia (RR = 1.65; 95% CI = 1.36–1.98). Prepregnancy migraine was more strongly associated with moderate preterm delivery (32–37 weeks' gestation; RR = 1.20; 95% CI = 1.07–1.35) than very preterm delivery (<32 weeks' gestation; RR = 0.97; 95% CI = 0.71–1.34). The positive association for preterm delivery persisted for normotensive preterm delivery (RR = 1.19; 95% CI = 1.06–1.34). By contrast, prepregnancy migraine was not associated with GDM (RR = 1.05; 95% CI = 0.91–1.22) or low birthweight (RR = 0.99; 95% CI = 0.85–1.16).

Table 2.

Updated Prepregnancy Physician-Diagnosed Migraine and Risk of Adverse Pregnancy Outcomes Among Incident Pregnancies, Nurses' Health Study II (NHSII)

In analyses examining migraine phenotype, compared with no migraine, migraine with aura was somewhat more strongly associated with risk of preeclampsia (RR = 1.51; 95% CI = 1.22–1.88) than migraine without aura (RR = 1.30; 95% CI = 1.04–1.61), although the test for heterogeneity was not significant (p = 0.32) (Table 3). Migraine with aura and without aura were similarly associated with greater risks of preterm delivery and gestational hypertension, compared with no migraine. Neither migraine with nor without aura were associated gestational diabetes or low birthweight.

Table 3.

Updated Prepregnancy Physician-Diagnosed Migraine and Risk of Adverse Pregnancy Outcomes Among Incident Pregnancies, by Migraine Phenotype, Nurses' Health Study II (NHSII)^a

Prepregnancy aspirin use modified the association of prepregnancy migraine with preterm delivery (<2×/week RR = 1.24; 95% CI = 1.11–1.38; ≥2×/week RR = 0.55; 95% CI = 0.35–0.86; p-interaction < 0.01) (Table 4). There were also qualitative differences in the association between prepregnancy migraine and preeclampsia by aspirin use (<2×/week RR = 1.48; 95% CI = 1.25–1.75; ≥2×/week RR = 1.10; 95% CI = 0.62–1.96; p-interaction = 0.39); however, power for this stratified analysis was limited.

Table 4.

Updated Prepregnancy Physician-Diagnosed Migraine and Risk of Adverse Pregnancy Outcomes Among Incident Pregnancies, Stratified by Prepregnancy Aspirin Use, Nurses' Health Study II (NHSII)

Associations of prepregnancy migraine and adverse pregnancy outcomes were not modified by age at pregnancy (eTable 1, links.lww.com/WNL/C588). Stronger associations for prepregnancy migraine and risk of preterm delivery were evident among those with prepregnancy BMI ≥25 kg/m² (RR = 1.41; 95% CI = 1.18–1.69) than those with BMI <25 kg/m² (RR = 1.06; 95% CI = 0.92–1.21; p-interaction = 0.02) but not for other adverse pregnancy outcomes (p-interactions > 0.05). There was also evidence of effect modification of the migraine-low birthweight association by parity, with lower risk among primiparous participants (RR = 0.78; 95% CI = 0.61–0.99) and higher risk among multiparous participants (RR = 1.18; 95% CI = 0.97–1.42; p-interaction < 0.01).

Further adjustment for AHEI diet score in multivariable models did not materially change effect estimates for any adverse pregnancy outcome (data not shown). Estimates from sensitivity analyses restricted to singleton pregnancies and to individuals without a history of antihypertensive or tricyclic antidepressant medication use did not meaningfully differ from those reported in the main analyses (eTable 2; eTable 3, links.lww.com/WNL/C588). Sensitivity analyses that used multiple imputation to address missing data also produced estimates consistent with those from the main analyses (eTable 4).

Discussion

In this prospective study of 30,555 incident pregnancies among 19,694 participants free of cardiovascular disease, type II diabetes, and cancer, self-reported physician-diagnosed migraine before pregnancy was associated with 17%, 28%, and 40% higher risks of preterm delivery, gestational hypertension, and preeclampsia, respectively, compared with no migraine. Migraine was more strongly associated with term than preterm preeclampsia. The higher risk of preterm delivery was driven by moderate rather than very preterm delivery and persisted when excluding preterm deliveries complicated by HDP. By contrast, prepregnancy migraine was not associated with risks of GDM or low birthweight. Migraine with aura was slightly more strongly associated with preeclampsia than migraine without aura, whereas risks of gestational hypertension and preterm delivery were similar by migraine phenotype. Consistent with previous literature showing a general beneficial effect of aspirin on preterm delivery,⁴⁴ individuals with migraine who reported prepregnancy regular aspirin use (≥2×/week) had 45% lower risk of preterm delivery than those without migraine who reported regular aspirin use, although the number of preterm deliveries in this category was small and should be cautiously interpreted. We also observed a suggestive, but not statically significant, lower risk of preeclampsia associated with migraine accompanied by regular aspirin use. The migraine-preterm delivery relationship was modified by BMI, whereby positive associations were limited to individuals with prepregnancy overweight and obesity. Importantly, that associations of prepregnancy migraine with gestational hypertension and preeclampsia did not vary by age, BMI, or parity, which suggests that migraine may portend heightened risk of hypertensive disorders of pregnancy in pregnancies otherwise considered to be low risk by these established obstetric risk factors.

Previous epidemiologic literature, largely limited to case-control and retrospective registry studies, suggests 1.2–6.7-fold higher odds of preeclampsia and gestational hypertension with migraine.^{21-25, 27-29, 31, 33, 34} The observed 28% and 40% higher risks of gestational hypertension and preeclampsia, respectively, associated with migraine in our data are compatible with previously reported estimates but extend the literature in several notable ways. Our findings of particularly high risk of term preeclampsia associated with migraine are novel because previous studies did not cross-classify preterm delivery and preeclampsia. Furthermore, previous prospective studies were smaller and incompletely adjusted for prepregnancy confounders such as BMI and oral contraceptive use, which could result in bias.^21,24,34 Our substantially larger sample size, combined with extensive prepregnancy covariate data, allowed for more complete adjustment for important confounders and evaluation of potential effect modification by cardiovascular and obstetric risk factors, which few studies have examined. Moreover, migraine status in the NHSII was ascertained by self-reported physician diagnosis, in contrast to Danish and Taiwanese register studies that identified migraine cases by migraine-related hospitalizations or prescription medications.^22,33 As migraine is typically diagnosed and managed in the primary care setting, migraine-related hospitalizations likely capture only the most severe cases, which may introduce bias and preclude generalizability to those with mild-moderate migraine. These register-based studies are also prone to unmeasured or residual confounding because data on prepregnancy health characteristics (e.g., BMI, analgesic use, oral contraceptive use) were not consistently adjusted for.^22,33 Nevertheless, our estimates are most similar to those reported in register studies, suggesting a relatively small effect of these potential sources of bias on HDP findings.

The availability of migraine aura information in the NHSII, though reported after the completion of most pregnancies in the cohort, allowed us to examine whether migraine phenotype is differently associated with adverse pregnancy outcomes, which no prospective studies have addressed. To our knowledge, only 1 case-control study separately examined migraine phenotype; they reported similarly elevated odds of preeclampsia for migraine with aura (OR = 3.7; 95% CI = 2.1–6.4) and without aura (OR = 3.8; 95% CI = 1.7–8.9), compared with no migraine.²³ Our finding of elevated risks of gestational hypertension and preeclampsia for migraine with and without aura suggests that migraine, irrespective of migraine phenotype, is associated with an increased risk of HDP. Migraine with aura was slightly more strongly associated with preeclampsia than migraine without aura in our data, but these findings should be confirmed in a prospective cohort with phenotype data collected before pregnancy.

Fewer studies have examined migraine and preterm delivery, and results are conflicting.^{20-22,26,33,34} Our observed 17% increased risk of preterm delivery among individuals with prepregnancy migraine is most consistent in direction and magnitude with the Danish and Taiwanese register studies,^22,33 whereas 2 smaller prospective studies reported stronger positive associations^26,34 and 2 others reported no association with overall,²¹ spontaneous, or medically indicated preterm delivery.²⁰ As with migraine and HDP, these studies lacked information on important prepregnancy confounders, a limitation more fully addressed by our analysis, and which may explain the higher magnitudes of association observed previously. Moreover, these studies did not separately examine risk of preterm delivery among pregnancies not affected by HDP, which contribute to ∼15% of preterm deliveries in the United States. Our finding of increased risk of normotensive preterm delivery directly fills this gap, indicating that migraine confers higher risk of preterm delivery independent of HDP. Owing to our large sample size, we were also able to examine risk of preterm delivery by gestational age (moderate vs very preterm) and found that the overall positive association was driven by moderate preterm delivery. Given that very preterm delivery more frequently results from infection,⁴⁵ the higher risk observed only for migraine and moderate preterm delivery is consistent with an underlying vascular pathology linking migraine and preterm delivery. An additional contribution of our study included the ability to stratify by potential modifying obstetric risk factors, which revealed stronger associations for migraine and preterm delivery among participants with overweight/obesity.

Recent recommendations support the use of low-dose aspirin (81 mg/d) after 12 weeks' gestation to prevent preeclampsia, preterm delivery, small for gestational age, and perinatal mortality among pregnant women at high risk; however, migraine history is currently not included among indicating risk factors.⁴⁴ We observed a 45% lower risk of preterm delivery for individuals with migraine who self-reported aspirin use >2×/week before pregnancy, which suggests that aspirin use may also be beneficial for preventing preterm delivery among those with migraine. Although migraine-aspirin use interactions were not statistically significant for preeclampsia, effect estimates were suggestive of an aspirin benefit for those with migraine. We were unable to examine finer categories of aspirin frequency and dosage in this cohort, and we lacked information on the indication for aspirin use. Additional studies, likely clinical trials, will be needed to determine the optimal dosage and timing of aspirin initiation to prevent preterm delivery and possibly preeclampsia among pregnant individuals with migraine.

Few studies have examined migraine and risk of low birthweight or GDM. The overall lack of association observed for migraine and low birthweight here is consistent with 1 small prospective study in Italy²¹ but conflicts with the 2 register-based studies, which reported modest ORs of 1.14 and 1.16.^22,33 Discordant findings may relate in part to the underlying distribution of effect modifiers in each population (e.g., parity) or greater migraine severity in register studies that classified migraine status by hospitalizations and/or medication use. Our findings of no association between migraine and GDM agree with a single register study,³³ which together suggest that migraine history and phenotype are unlikely to be a useful risk marker for GDM.

Migraine and adverse pregnancy outcomes, including preterm delivery and HDP, are associated with an unfavorable cardiometabolic risk profile,^17,18 which suggests overlapping pathophysiology. Cardiovascular risk factors including chronic hypertension and obesity were slightly more prevalent among those with migraine at baseline in our sample; however, careful adjustment for these and other confounders minimally attenuated positive associations, implying involvement of other subclinical factors. For example, C-reactive protein, a putative marker of systemic inflammation, is elevated in individuals with migraine⁴⁶ and is also implicated in preterm delivery and HDP.^47,48 Migraine is also associated with thrombogenic susceptibility,⁴⁹ and recent data suggest that preconception platelet factor 4, a specific marker of platelet activation, is positively associated with placenta-mediated adverse pregnancy outcomes, particularly HDP.⁹ Future research elucidating the underlying pathways between migraine and adverse pregnancy outcomes is clearly warranted.

Strengths of this study include its prospective design and large sample size, minimal loss to follow-up (<10%), and data on numerous prepregnancy confounders and lifetime pregnancy history. Nevertheless, our study also has limitations. First, the prevalence of overall migraine in the NHSII is lower than among women in the general population (11% vs 24%); specifically, the prevalence of migraine without aura here is lower than among women in the general population (6% vs 16%), while the prevalence of migraine with aura in our analytic sample is similar to general population estimates in women (6% vs 7%).^2,50 These discrepancies are likely explained by the fact that migraine status was defined as self-reported physician diagnosis in the NHSII and that individuals experiencing migraine without aura may be less likely to seek a physician diagnosis than those experiencing migraine without aura. As such, our questionnaire assessment of migraine, particularly migraine with aura, is likely highly specific but may be less sensitive to milder migraine. This exposure misclassification may underestimate the true RRs. In addition, migraine phenotype was queried for the first time on the 2007 questionnaire, after nearly all pregnancies in the NHSII cohort were completed. Reverse causation may have affected migraine phenotype estimates if, for example, participants with prepregnancy migraine who experienced preeclampsia were more likely to develop new-onset aura during or after the preeclamptic pregnancy. Pregnancy outcomes were recalled up to 10 years after delivery; thus, recall error of pregnancy outcomes could produce estimates that are most likely biased toward the null. It is also possible that individuals with migraine, particularly those with migraine with aura, more frequently interacted with medical providers during their pregnancy, potentially providing greater opportunity for the diagnosis of specific APOs (i.e., HDP and GDM). However, if surveillance bias arising from this scenario explained the observed positive associations between migraine and HDP, we would expect to see a similar positive association with GDM. The lack of association observed for GDM suggests that this scenario is unlikely to explain our findings. We also lacked information on subtypes of preterm delivery (i.e., spontaneous vs iatrogenic and spontaneous preterm labor vs preterm premature rupture of membranes), which should be addressed in future work. Residual confounding arising from unmeasured (e.g., genetic factors or migraine-specific medications) or imprecisely measured variables could partly explain our results, although age-adjusted estimates were similar to those from multivariable models. Finally, the NHSII is comprised of predominantly non-Hispanic White individuals with relatively high socioeconomic status and health literacy levels. As such, generalizability to other racial/ethnic groups and to individuals with lower socioeconomic status and health literacy levels may be limited; however, it seems unlikely that the associations would biologically differ by these factors.

Prepregnancy migraine is associated with greater risks of adverse pregnancy outcomes, including preterm delivery, gestational hypertension, and preeclampsia, independent of established obstetric and cardiovascular risk factors. Migraine history, and to a lesser extent migraine phenotype, constitute clinically meaningful markers of obstetric risk, particularly for adverse pregnancy outcomes with underlying vascular pathology. Future research should confirm whether aspirin may reduce risk of preterm delivery and preeclampsia among pregnant individuals with migraine, examine additional migraine features (e.g., attack frequency) and preterm delivery subtypes, and identify the mechanisms that underlie these associations.

Glossary

BMI

body mass index

GDM

gestational diabetes mellitus

HDP

hypertensive disorders in pregnancy

ICHD-II

International Classification of Headache Disorders-II

MET

metabolic equivalent of task

NHSII

Nurses' Health Study II

RR

relative risks

Appendix. Authors

Study Funding

This work was supported by the National Institutes of Health (U01 CA176726 and U01 HL145386). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funders had no role in considering the study design or in the collection, analysis, interpretation of data, writing of the report, or decision to submit the article for publication.

Disclosure

A.C. Purdue-Smithe reports no disclosures. J.J. Stuart reports support from the US National Institutes of Health unrelated to this work. L.V. Farland reports support from the US National Institutes of Health unrelated to this work. A.M. Harriott reports support from the US National Institutes of Health unrelated to this work, an authorship agreement with Abbvie unrelated to the present study, honoraria from the Headache Cooperative of New England, participation in the DSMB for the MindfulRAVANS, and leadership roles at the American Academy of Neurology, American Headache Society, and Headache Cooperative of New England. J.H. Kang reports support from the US National Institutes of Health and Pfizer, Inc. unrelated to this work. J.W. Rich-Edwards reports support from the US National Institutes of Health unrelated to this work. K.M. Rexrode reports support from the US National Institutes of Health unrelated to this work and a leadership role at the American Heart Association. Go to Neurology.org/N for full disclosures.