We addressed two Key Questions and one Contextual Question

KQ 1: What are the (comparative) benefits and harms of interventions to prevent attacks of primary headache in women who have a history of primary headache and are pregnant (or attempting to become pregnant), postpartum, or breastfeeding?

KQ 1a. Do the (comparative) benefits and harms vary by phase (i.e., preconception, first trimester of pregnancy, second trimester of pregnancy, third trimester of pregnancy, postpartum, breastfeeding)?

KQ 1b. Do the (comparative) benefits and harms vary by type of primary headache (i.e., migraine, tension headache, cluster headache, and other trigeminal autonomic cephalgias)?

KQ 2: What are the (comparative) benefits and harms of interventions to treat acute attacks of primary headache in women who are pregnant (or attempting to become pregnant), postpartum, or breastfeeding?

KQ 2a. Do the (comparative) benefits and harms vary by phase (i.e., preconception, first trimester of pregnancy, second trimester of pregnancy, third trimester of pregnancy, postpartum, breastfeeding)?

KQ 2b. Do the (comparative) benefits and harms vary by type of primary headache (i.e., migraine, tension headache, cluster headache, and other trigeminal autonomic cephalgias)?

Contextual Question: What is the available evidence concerning levels in maternal serum/blood, fetal/child serum/blood, breast milk, amniotic fluid, meconium, cord blood, or child urine of drugs used to prevent or treat attacks of primary headache in women who are pregnant (or attempting to become pregnant), postpartum, or breastfeeding?

Findings in Relationship to What is Already Known

This section provides information on how this AHRQ review relates to other systematic reviews of and clinical guidelines for the prevention and treatment of headache among those who are pregnant.

Related review findings

1. Negro A, Delaruelle Z, Ivanova TA, et al. Headache and pregnancy: a systematic review. The journal of headache and pain. 2017 Oct 19;18(1):106. doi: 10.1186/s10194-017-0816-0. PMID: 29052046
2. Pearce CF, Hansen WF. Headache and neurological disease in pregnancy. Clinical obstetrics and gynecology. 2012 Sep;55(3):810-28. doi: 10.1097/GRF.0b013e31825d7b68. PMID: 22828113 .
3. Allais G, Chiarle G, Sinigaglia S, et al. Migraine during pregnancy and in the puerperium. Neurological sciences : official journal of the Italian Neurological Society and of the Italian Society of Clinical Neurophysiology. 2019 Mar 18. doi: 10.1007/s10072-019-03792-9.PMID: 30880362
4. Robbins MS. Headache in Pregnancy. Continuum (Minneap Minn). 2018 Aug;24(4, Headache):1092-107. doi: 10.1212/con.0000000000000642. PMID: 30074551 .
5. Silberstein SD. Migraine and pregnancy. Neurologic clinics. 1997 Feb;15(1):209-31. doi: 10.1016/s0733-8619(05)70305-4. PMID: 9058407
6. Burch R. Headache in Pregnancy and the Puerperium. Neurologic clinics. 2019 Feb;37(1):31-51. doi: 10.1016/j.ncl.2018.09.004. PMID: 30470274 .

Existing Guidelines:

1. ACOG guidelines on perinatal care
2. American Headache Society (AHS) guidelines on the management of headache PMID 27432623, PMID 27300483, PMID 25600718
3. Canadian Headache Society (CHS) guidelines on the management of headache PMID: 23968886

The findings of this SR are intended to be used in development of ACOG clinical guidance.

Journal articles from this report:

1. Management of primary headaches during pregnancy, postpartum, and breastfeeding: A systematic review PMID 33433020

Implications for Clinical Practice

Although we used both direct and indirect evidence to inform our conclusions in this SR, we emphasize that the direct evidence is sparse. There is surprisingly little directly useful evidence for guiding clinical practice for women with primary headache who are pregnant (or attempting to become pregnant), postpartum, or breastfeeding. This paucity of information also applies to medications more recently approved for migraine by the Food and Drug Administration, such as CGRP inhibitors, noninvasive neuromodulation devices, and botulinum toxin, and some that are available over the counter, such as acetaminophen. Our a priori approach to examining harms of relevant medications in this review was restricted to harms reported in (1) primary studies of women with primary headache who were pregnant (or attempting to become pregnant), postpartum, or breastfeeding, and (2) existing SRs that met minimal quality criteria in this population of women regardless of indication. Thus, as discussed above, some harms, such as the risks of NSAIDs when used after 32 weeks of gestation, may not have been found in our search of these two sources of evidence.

Given the paucity of information regarding estimates of the effectiveness of various interventions to prevent or treat primary headaches, decision makers will need to rely on the evidence from the general population. This is especially true for interventions for prevention of primary headaches and for treatments of primary headaches other than migraine, for which we found limited studies among pregnant, postpartum, or breastfeeding women. Ideally, high SoE information from studies of pregnant women with migraine (or other types of primary headache) would inform this decision, but there is generally sparse information for this population. Thus, decision makers are left to extrapolate from studies of pregnant women with mostly other conditions regarding the safety of the interventions, especially regarding potential harms to the fetus or infant. However, the risks involved in using the same drugs for treatment of other conditions (e.g., seizures, psychosis, depression) may not translate well to the risks for women with primary headaches because the underlying risks of fetal/child adverse effects may differ. Treatment doses and durations often differ by indication. Concomitant drug use (and thus drug-drug interactions) are likely to differ. Psychosocial behaviors, such as smoking, substance use, and caffeine intake, may also differ. Nevertheless, clinicians, patients, and policymakers are left with the options of making treatment decisions based on: (1) extrapolating information about harms from studies where these drugs were used for other indications (not included as direct evidence in this SR, but summarized as indirect evidence through examination of published SRs); and/or (2) depending on what is known about the levels of the various drugs in maternal serum/blood or that are transmitted to the fetus through amniotic fluid and/or cord blood, or to the infant through breast milk.

We did not consider levels of various drugs in maternal serum/blood or in other fluids transmitted to the fetus/infant as relevant outcomes in our SR. However, especially given the absence of studies examining interventions for primary headache in breastfeeding women, we recognize that decision makers may be interested in drug levels in various fluids. We found that the most complete resource with data about levels of the various drugs is the Drugs and Lactation Database (LactMed®, available at https://toxnet.nlm.nih.gov/newtoxnet/lactmed.htm). Because the data in LactMed are readily available and are being continually updated, we refer the interested reader there for current information on specific drugs of interest. LactMed contains extensive information about the levels of drugs that are of interest to the current SR. This information is reported in LactMed for specific drugs, often at the level of individual studies and/or specific body fluids. Data are often reported for individual (deidentified) women at various time-points. While LactMed is frequently used by clinicians and sometimes incorporated into discussions with patients about potential risks, it should be noted that for many drugs, the association between levels of drug exposure in body fluids and harms, either short- or long-term, is not well established. In other words, drug levels are, at best, intermediate outcomes and may not correlate well with harms to the offspring.

We encourage clinicians to inform patients about the limitations of existing research on interventions for preventing or treating primary headaches during pregnancy, postpartum, or breastfeeding phases. Given the limitations of the evidence, the patient's values and preferences and the clinician's expertise and experience are even more important. A related important aspect that should be considered is the severity of the primary headache, which could shift the balance between benefits and harms of a given intervention, or comparative benefits and harms between a set of interventions, under consideration.

Implications for Research

The sparseness of the direct evidence addressing the interventions addressed in this comprehensive SR is striking. It does not imply, however, that interventions are not beneficial or harmful. Because interventions may affect pregnant women (and their offspring) differently than non-pregnant individuals, there is an important and urgent research need for direct evidence in pregnancy, postpartum, or breastfeeding phases.

Research is needed both for pharmacologic and nonpharmacologic interventions. Triptans, the most studied classes of pharmacologic interventions, were discussed by the included studies only in the context of their harms. Currently, their use in clinical practice appears to be based on their effectiveness in nonpregnant populations; the findings of the current SR suggests the absence of evidence in pregnant women. We found low strength of evidence that combination metoclopramide and diphenhydramine was more effective and no more harmful than codeine when used for treating migraine or tension headache. In an era of heightened concern about opioid use, the evidence suggests that it is possible that this combination is a viable alternative to codeine for pregnant women experiencing migraine or tension headache.

For some interventions, although we concluded that studies provided insufficient evidence to make conclusions, these studies found a signal of potential effectiveness and/or safety that should be explored in future research. These include topiramate for prevention of primary headache, and acupuncture, thermal biofeedback, behavioral therapy, physical therapy, peripheral nerve blocks, and transcranial magnetic stimulation for treatment.

Because of the absence of studies addressing prevention or treatment of cluster headache and other TACs in pregnant women, researchers should also design studies that, either entirely or in part, enroll these patients. When enrolled as part of a larger study, subgroup-specific data for these types of primary headache, should be reported.

It is important that future studies either randomize patients (after considering the ethical issues in this population) to minimize selection bias, or report between-arm estimates of treatment effect that adequately account for important confounders, such as age and severity of headache attack (or of history of headaches). Studies should also, where feasible, conduct blinding of participants, care providers, and outcome assessors to minimize the likelihood of performance and detection biases. Given the concern regarding exposing the fetus to potentially harmful pharmacologic interventions, we recognize that RCTs will likely continue to be infrequent. As an alternative to randomization, when observational studies, such as those using patient registries, are conducted, they should be adequately designed and analyzed to compare treatments. Such analyses should appropriately account for differences between comparison groups of patients that are inherently different. Ideally, propensity score analyses (or similar rigorous techniques) should be used to adequately adjust for these differences. A propensity score analysis, for example, estimates the likelihood that each patient had one or the other intervention (conditional on their measured characteristics) and controls for this likelihood. These analyses generally require relatively large numbers of patients for whom there are granular data about risk factors for outcomes. Additionally, while registry data will likely continue to be important in identifying harms, researchers should report more details about disease severity as well as intervention doses, durations, and frequencies.

When reporting studies, it is also important that authors adhere to relevant reporting guidelines so that adequate details about the population, interventions (and comparators), and outcomes are clearly described.

Future studies should also evaluate other important maternal outcomes, such as headache-related symptoms (e.g., photosensitivity), quality of life, functional outcomes (e.g., impact on employment/school attendance), and patient satisfaction with intervention; adverse effects on breastfeeding, such as decreased milk supply; and some important fetal/child adverse outcomes. None of the studies included in this SR addressed these outcomes.

This review does not provide cost information.

Limitations

Strengths and Limitations of the Evidence Base

The limitations of the evidence we identified vastly outnumber its strengths. A major limitation is that, for most interventions, direct evidence about the effectiveness and/or harms in patients who are pregnant (or attempting to become pregnant), postpartum, or breastfeeding is sparse or absent. We did not identify any primary studies for entire classes of pharmacologic agents: analgesics/antipyretics, tricyclic antidepressants, beta blockers, calcium channel blockers, other antihypertensive medications, serotonin and norepinephrine reuptake inhibitors (SNRIs), benzodiazepines, central nervous system stimulants, muscle relaxants, N-methyl-D-aspartate (NMDA) receptor antagonists, calcitonin gene-related peptide (CGRP) inhibitors, mood-stabilizing agents, tetracyclic antidepressants, corticosteroids, butalbital-containing analgesics, sympathomimetic amines, topical anesthetics, antipsychotics, somatostatin analogs, and intravenous magnesium. Similarly, no primary studies addressed entire classes of nonpharmacologic agents: supplements, chemodenervation, and hydration therapy. However, as discussed above, some of these interventions were described in the indirect evidence.

Where evidence was identified, all studies included women exposed to the interventions (or comparators) during pregnancy; we did not find evidence in women attempting to become pregnant or when postpartum or breastfeeding (except for some studies of triptans [direct evidence] that compared treatments during versus before pregnancy and one SR [indirect evidence] that examined nonsteroidal anti-inflammatory drug (NSAID) use in the postpartum period). In terms of type of primary headache, most primary studies focused on migraine and some focused on tension headache, but none focused on cluster headache or other trigeminal autonomic cephalgias (TACs).

A related limitation of the sparse evidence base is that the studies we identified did not report data for many of our outcomes of interest. Unreported or rarely reported maternal outcomes include headache-related symptoms (nausea/vomiting, photosensitivity, dizziness), quality of life, functional outcomes (impact on family life, impact on employment/school attendance, time spent managing disease), resource use, acceptability of intervention, satisfaction with intervention, and certain adverse effects (stroke, myocardial infarction, reduced breast milk production, and maternal symptoms related to withdrawal of medication). Unreported or rarely reported fetal/child adverse effects include breastfeeding outcomes (delayed initiation, cessation, reduced frequency, reduced volume), poor infant attachment/bonding, and neonatal signs related to withdrawal of medication. Relatedly, few studies reported on the long-term effects and harms of the interventions for mother or child.

Three limitations with the evidence base pertain specifically to the included nonrandomized comparative studies (NRCSs). First, few of the NRCSs reported adjusted between-arm effect sizes. In the absence of the individual patient data, we were unable to calculate adjusted effect sizes. While feasible in some instances (i.e., when arm-specific data were reported), we did not consider it appropriate to calculate unadjusted effect sizes because the populations of women in the treatment arms were generally dissimilar on one or more important confounders. The primary headache disorders result in the use of interventions (for prevention or treatment). While interventions can cause harms that were investigated in this SR, the underlying disorders themselves can cause some of the harms, irrespective of exposure to interventions. This issue can contribute to confounding. Moreover, nonrandomized studies are prone to unmeasured confounding, which can only be accounted for satisfactorily by well-conducted randomized controlled trials (RCTs). Second, triptans, which were the most studied classes of pharmacologic interventions, were discussed by the included studies only in the context of their harms. Currently, their use in clinical practice appears to be based on their effectiveness in nonpregnant populations; the findings of the current SR suggest the absence of evidence of their effectiveness in pregnant women. Third, none of the NRCSs reported information about the doses, durations, and routes of administration. This is likely because most of the NRCSs were registry-based studies that might not have had access to such information, but the absence of such information can weaken conclusions.

We assessed most of the primary studies at an overall high risk of bias. The main reasons were because of a high risk of serious confounding; because participants, care providers, and/or outcome assessors were not blinded; and because of incomplete outcome data. Furthermore, the participant eligibility criteria, interventions, and outcomes were often inadequately described.

Finally, the included SRs of harms of pharmacologic interventions (indirect evidence) reported limited information regarding the doses, timings, durations, and routes of administration during pregnancy. Consequently, the estimates of harms obtained from these SRs were restricted to use versus nonuse of specific drugs or drug classes. We were unable to make conclusions regarding relative harms of various doses, timings, durations, and routes of administration.

Strengths and Limitations of the Systematic Review Process

We followed contemporary standards for SRs, including multiple stakeholder engagement in KQ development and refinement and careful adherence to recommended methods for literature searching, screening, data extraction, risk of bias assessment, data (narrative) synthesis, and strength of evidence (SoE) assessment. In anticipation of a sparse evidence base, we were very inclusive in our eligibility criteria, especially in terms of study designs, including RCTs, NRCSs, and single-group studies of interventions for primary headaches in pregnancy (as direct evidence); SRs of harms of interventions in pregnancy regardless of indication (as indirect evidence); and case reports (as supplemental evidence).

For all interventions examined in this SR, the paucity of the evidence precluded us from being able to be conduct meta-analyses (either pairwise or network) or make definitive conclusions about treatment effectiveness or harms of the various interventions.

Despite our comprehensive search and approach to using indirect evidence to find harms from SRs regardless of indication, some well-accepted harms of treatment were not addressed. For example, we did not find a SR that fulfilled our minimum quality criteria and provided evidence for the association between indomethacin and increased risk of premature closure of the ductus arteriosus (despite indomethacin being an effective treatment to close a patent ductus arteriosus in neonates110). Searching for primary studies of harms (or benefits) of medications regardless of indication during pregnancy was beyond the scope of this review.

Applicability

In addition to the sparseness of the evidence discussed above, a few factors may limit the applicability of our findings. As discussed, the limited information about doses, durations, and frequencies of the interventions reported in the NRCSs (especially triptans) constrains our ability to make definitive conclusions about individual triptans.

The population in the studies included in this SR were varied in terms of the trimester and gestational age, which limits our ability to apply our findings specifically to different trimesters of pregnancy.

Most primary studies in this SR were conducted in the U.S., Canada, or Europe. Various contextual factors may impact the effectiveness of treatments. It is unclear to what extent the findings of this SR might apply outside of these high-income settings.