Pre-pregnancy migraine diagnosis, medication use, and spontaneous abortion: a prospective cohort study

We used data from Pregnancy Study Online (PRESTO), an ongoing online prospective study of the effects of lifestyle, dietary, and medical factors on fertility that began in June 2013. Study methods have been described in detail elsewhere [24]. Briefly, eligible participants self-identified as female, are 21–45 years of age, residents of the United States or Canada, and not using contraception or receiving fertility treatment. Participants complete an online baseline questionnaire eliciting information on demographics, lifestyle, and medical and reproductive histories. They then complete follow-up questionnaires every 8 weeks for up to 12 months to ascertain pregnancy status. Participants who report a conception are invited to complete additional questionnaires in early pregnancy (~ 8 weeks of gestation) and late pregnancy (~ 32 weeks of gestation).

We defined those with a history of migraine as participants with a reported diagnosis of migraine or use of a medication to treat migraine before conception (estimated as 14 days after the date of last menstrual period, LMP). We ascertained migraine history and medication use from self-administered online questionnaires at baseline and during follow-up. At baseline, participants reported whether they had ever been diagnosed with “migraine headaches” and, if so, how many migraines they had experienced in the past 4 weeks (none, one, two, more than two). On baseline and bimonthly follow-up questionnaires, participants reported migraine medication use in the 4 weeks prior to each questionnaire (yes/no), and medication use frequency (i.e., when experiencing symptoms, daily, or daily plus occasional extra use due to symptoms). Participants who completed the questionnaire after April 2016 also reported the name of their most recent migraine medication, if any. We also identified individuals taking migraine medications from a write-in field for pain medication. Participants who reported migraine as the indication for their use of pain relievers, or who provided the name of pain-relieving medication in the migraine medication field, were classified as users of pain-relieving medication for migraine.

We ascertained changes in the use of migraine medication on bimonthly follow-up questionnaires. In our analyses, we prioritized data on migraine medication use from the most recent follow-up questionnaire before estimated conception, as this was deemed to be the most etiologically relevant period of exposure. As we sent preconception questionnaires to participants every 8 weeks, the time between exposure assessment (questionnaire completion) and estimated conception date varied among participants. A timeline of exposure assessment relative to conception is presented in Fig. 2. Recent (previous 4 weeks) migraine medication use was assessed 0–8 weeks before conception for 75% of participants, after conception (before pregnancy detection) for 10% of participants, and more than 8 weeks before conception (participant skipped questionnaire(s)) for 15% of participants.

We classified responses by active ingredient (anticonvulsant, antidepressant, acetaminophen, aspirin, beta blocker, calcium channel blocker, calcitonin gene-related peptide (CGRP) antagonist, combination analgesic/caffeine, muscle relaxant, non-aspirin non-steroidal anti-inflammatory drugs (NSAIDs), opioid, triptan, steroid, other, unknown). We classified pain medications by active ingredient (acetaminophen, aspirin, non-aspirin NSAIDs, and opioid). Pain medications were not mutually exclusive. We classified combination analgesic-caffeine medications by their specific pain-relieving ingredient. We additionally adjusted the pain medication analyses for the participant-reported number of pain medication pills used in the previous month.

We further classified participants who reported medication use on their most recent follow-up questionnaire before conception into three mutually exclusive categories, according to type of medication: prescription migraine prophylaxis medication, prescription migraine treatment, and over-the-counter (OTC) migraine medication. Medication type may indicate migraine severity, with use of a daily preventive medication indicating higher migraine severity, followed by episodic prescription treatment and over-the-counter treatment. We categorized participants who reported use of multiple medication types into the highest hypothesized severity group indicated by their responses.

We defined SAB as pregnancy loss before 20 completed weeks of gestation, captured by questionnaire responses of “miscarriage”, “chemical pregnancy”, or “blighted ovum”. On follow-up questionnaires, participants reported the dates and results of home pregnancy tests, including negative and positive tests. Pregnant participants were asked how they detected their pregnancy (home pregnancy test, urine test at doctor’s appointment, etc.). Participants reported pregnancy losses, including date of loss, on bimonthly follow-up questionnaires and early and late pregnancy questionnaires. We calculated gestational weeks from the pregnancy LMP date.

If participants were lost to follow-up after reporting a pregnancy, we attempted to determine the pregnancy outcome by contacting them via alternative channels. These included email, phone, searching social media and baby registries, using data shared by participants from fertility tracking apps, and linking participant data to birth registries in states with the highest numbers of PRESTO participants (MA, CA, NY, FL, MI, TX, PA, and OH). If we were able to contact participants, we asked them to provide information on pregnancy status, including date of pregnancy loss, if applicable. If we identified a birth in a birth registry that corresponded to an LMP date during the study period, we assumed there was no pregnancy loss. If we were unable to contact participants, we censored them at completion of their last questionnaire.

We ascertained values of covariates from information collected on the baseline questionnaire. Covariates included maternal age, partner age, education, body mass index, physical activity, smoking, alcohol intake, caffeine intake, polycystic ovary syndrome (PCOS), endometriosis, irregular menstrual cycles, history of anxiety or depression, stress (Perceived Stress Scale [PSS]--10 score) [25], and depressive symptoms (Major Depression Inventory [MDI] score) [26].

From June 2013 through September 2021, 15,319 eligible female participants enrolled and completed the baseline questionnaire. We excluded 3910 participants (26%) who completed the baseline questionnaire but were lost to follow-up before identification of pregnancy, and 3519 (23%) who did not conceive within 12 months of follow-up (Fig. 1). Participants who were lost to follow-up tended to be younger, had lower educational attainment and household income, and reported higher levels of stress than individuals who were not lost to follow-up. Loss to follow-up was similar among participants with and without a diagnosis of migraine at baseline (26% of each group). The total analytic sample was 7890 participants (Fig. 1).

We used descriptive statistics to examine associations between migraine diagnosis, medication use, and baseline covariates. We used multiple imputation to impute missing data on covariates applying fully conditional specification methods [27]. We created 20 imputed datasets and statistically combined coefficient and standard error estimates across datasets. No participants were missing data on migraine status. As pain medication indication and specific migraine medication were not added to the questionnaire until 2016, this information was missing for many medication users. We limited pain medication analyses to individuals who provided an indication (n = 1683). Missingness for all covariates was < 1%, except for physical activity (1.2%) and household income (2.7%). For 1% of participants, we imputed missing gestational weeks at pregnancy loss or censoring events.

We estimated gestational weeks at SAB based on participant-reported weeks of pregnancy duration. When this information was unavailable, we estimated gestational weeks at SAB based on the pregnancy end date and pregnancy due date, using the following formula: pregnancy end date-(pregnancy due date-280)/7. When no pregnancy due date was available, we estimated gestational weeks at SAB using an alternate formula: (pregnancy end date-LMP date)/7. We applied the Andersen-Gill data structure [28] with one observation per week of gestation, starting at gestational week of pregnancy detection (when available) or 4 weeks (i.e., median gestation at first pregnancy detection in the cohort) and ending at SAB, 20 weeks of gestation, or other censoring event (e.g., ectopic pregnancy or induced abortion, study withdrawal, or completion of last questionnaire if lost to follow up), whichever came first.

We used Cox proportional hazards regression with gestational weeks as the time scale to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between migraine and SAB [29]. We selected confounders based on a literature review and a directed acyclic graph. In the final models we adjusted for the following baseline confounders: age (< 25, 25–29, 30–34, ≥35 years), partner age (continuous), education (< 16 vs. ≥16 years), body mass index (BMI, kg/m², continuous), current smoker at baseline (yes/no), alcohol intake (0, 1–6, 7–13, ≥14 drinks per week), caffeine intake (mg/day, continuous), recent irregular menstrual cycles (yes/no), physical activity (total metabolic equivalents of activity per week [30, 31], continuous), stress (perceived stress scale (PSS-10) score [25], continuous), depressive symptoms (Major Depression Inventory (MDI) [26] score, continuous), and diagnoses of PCOS, endometriosis, anxiety, or depression (yes/no).

We examined the association between any reported migraine (ever vs. never) and migraine frequency (0, 1, 2, > 2) in the 4 weeks before baseline and SAB. Based on the most recent follow-up questionnaire before the estimated date of conception, we examined the following exposures: use of any migraine medication in the last 4 weeks (yes, no), use of specific medication types (prescription preventive medication, prescription treatment medication, over-the-counter medication), and use of migraine medications containing specific ingredients (analgesic(s) and caffeine, triptans). We compared these data to responses from individuals who did not report migraine at baseline or during follow-up. We also subcategorized early and late SAB (< 8 vs. ≥8 weeks), to determine whether any potential association between migraine and SAB was stronger for early SAB. We hypothesized that preconception exposures would contribute to an increase in early, rather than late, SAB [23, 32].

To assess confounding by indication, we compared SAB risk among users of pain medications for migraine or other indications with non-users of pain medications in the previous 4 weeks. We adjusted for potential confounding variables listed above, as well as for the total number of pain medication pills the participant reported using in the previous 4 weeks (for any indication, as a proxy for medication dose). We also adjusted results for use of medication to treat urinary tract infections, pelvic inflammatory disease, hypertension, thyroid disorders, diabetes, fibroids, PCOS, depression, anxiety, hay fever, endometriosis, acid reflux, or use of antibiotics reported on the same questionnaire (yes to any, no).

Table 1 shows the distribution of covariates at baseline by preconception migraine status and preconception migraine medication status. There was little difference in demographic characteristics between migraineurs and non-migraineurs. Migraineurs who took medication for migraine had slightly lower educational attainment and household income than non-migraineurs. Those taking migraine medication also were more likely to have a history of SAB than participants without migraine;32% of migraineurs who took medication, 26% of migraineurs who did not take medication, and 25% of non-migraineurs reported a history of SAB. Participants with migraine were more likely than those without migraine to report a history of infertility and diagnoses of PCOS, endometriosis, or anxiety disorders, regardless of medication use. Although many lifestyle factors measured at baseline are common migraine triggers (alcohol use, stress, physical activity, caffeine), migraineurs did not have an appreciably different prevalence of these triggers compared to non-migraineurs.

A preconception history of migraine was not appreciably associated with SAB risk (HR = 1.03, 95% CI: 0.91–1.16) (Table 2). Results were similar for early and late SABs. Migraine frequency also was not appreciably associated with SAB risk: HRs for 0, 1, 2, and > 2 migraine episodes in the 4 weeks before baseline (compared with no history of migraine) were 1.03 (95% CI: 0.86–1.23), 0.93 (95% CI: 0.74–1.16), 1.10 (95% CI: 0.82–1.37), and 1.14 (95% CI: 0.86–1.49), respectively.

Approximately half (n = 807, 48%) of the 1683 migraineurs reported medication use at baseline or during preconception follow-up. Migraineurs who reported no preconception migraine medication use did not have a higher risk of SAB than non-migraineurs (HR = 0.96, 95% CI: 0.82–1.13) (Table 3). Migraineurs who used medication in the past 4 weeks on the most recent questionnaire before conception had a slightly higher SAB risk (HR = 1.14, 95% CI: 0.96–1.36). Daily use of migraine medication was associated with an increased risk of SAB (HR = 1.38, 95% CI: 0.81–2.35), whereas less than daily use of migraine medication was not (HR = 1.09, 95% CI: 0.90–1.32).

While use of prescription migraine prophylaxis medication, indicative of the most severe cases, was associated with increased risk of SAB (HR = 1.43, 95% CI: 0.72–2.84), there were only 11 SABs among exposed participants. The HRs for use of prescription (episodic) or over-the-counter episodic medication were 1.04 (95% CI: 0.71–1.53) and 1.16 (95% CI: 0.93–1.44), respectively. Due to the variety of write-in migraine medications reported, we could analyze only the association between the two most frequently-reported specific active ingredients and SAB. The HR for use of analgesic/caffeine medication was 1.42 (95% CI: 0.99–2.04) and the HR for use of triptans was 1.30 (95% CI: 0.83–2.07).

Table 4 shows HRs for the association between migraine medication use and SAB risk, stratified by the timing of exposure assessment. HRs were similar when the exposure was assessed post-conception (HR = 1.06, 95% CI: 0.61–1.82), 0–4 weeks before conception (HR = 1.13, 95% CI: 0.88–1.43), and 5–8 weeks before conception (HR = 1.14, 95% CI: 0.81–1.61). The association between migraine medication use and SAB was strongest when exposure was assessed >8 weeks before conception (HR = 1.43, 95% CI: 0.87–2.37).

Most participants (n = 5362, 68%) reported pain medication use on their most recent follow-up questionnaire before conception, and 7% reported migraine as the indication for using pain medication. The HR for use of pain medication for indications other than migraine was 0.99 (95% CI: 0.84–1.17), while the HR for use of pain medication for migraine was 1.15, (95% CI: 0.84–1.47) (Table 5). Generally, use of specific pain medications for migraine or other indications was not meaningfully associated with SAB risk. Use of aspirin for migraine was associated with a slight increased risk of SAB (HR = 1.21, 95% CI: 0.82–1.79), while use of aspirin for other indications was associated with a decreased risk of SAB (HR = 0.76, 95% CI: 0.42–1.38), compared with no pain medication use.