A Systematic Review

Management of Primary Headaches During Pregnancy, Postpartum, and Breastfeeding

Ian J. Saldanha MBBS, MPH, PhD; Wangnan Cao PhD; Monika Reddy Bhuma BDS, MPH; Kristin J. Konnyu PhD; Gaelen P. Adam MLIS, MPH; Shivani Mehta BA; Andrew R. Zullo PharmD, PhD; Kenneth K. Chen MD; Julie L. Roth MD; Ethan M. Balk MD, MPH


Headache. 2021;61(1):11-43. 

In This Article


We used established methodology for conducting SRs as outlined in the AHRQ Methods Guide for Effectiveness and Comparative Effectiveness Reviews.[7] We developed the research questions and protocol after in-depth discussions with a panel of key informants and technical experts. The prospectively registered review protocol is available through PROSPERO (registration number: CRD42020158310).

Search Strategy

We conducted two parallel searches: (a) for primary studies evaluating benefits and harms of various pharmacologic and non-pharmacologic interventions in pregnant, postpartum, or breastfeeding women with primary headache (these studies provide "direct evidence" to manage headaches), and (b) for SRs evaluating harms (not benefits) of various pharmacologic and non-pharmacologic interventions in women in the same phases but regardless of indication for the intervention (these SRs provide "indirect evidence" with regards to use for primary headache). The latter search was conducted because some interventions of interest, such as antiepileptics and NSAIDs, are most commonly used during pregnancy for non-primary headache-related indications.

Appendix A details the search strategies used in this SR. For direct evidence, we searched for published studies in Medline (via PubMed), Embase, the Cochrane Central Register of Clinical Trials, and CINAHL, and for unpublished studies in ClinicalTrials.gov. The search included controlled-vocabulary terms and free-text words related to pregnancy, postpartum, breastfeeding, headache, migraine, tension headache, cluster headache, other TACs, and the various interventions of interest. For indirect evidence, we searched Medline (via PubMed), the Cochrane Database of SRs, and Epistemonikos; these searches did not include terms for headaches. No date or language restrictions were applied. The searches were independently peer reviewed. All searches were current as of June 5, 2020. We also scanned the reference lists of available clinical practice guidelines and SRs for potentially eligible studies.

Study and SR Selection

Nine investigators screened each title and abstract independently in duplicate using the online program abstrackr (https://abstrackr.cebm.brown.edu/). All accepted citations were retrieved and reviewed in full text. All discrepancies were resolved through discussion.

For direct evidence, we included studies of women who were pregnant, attempting to become pregnant, postpartum (defined as up to 12 months post-delivery), or breastfeeding (for any length of time) with a history of primary headache (for prevention) or with acute attacks of primary headache (for treatment). We considered any pharmacologic and non-pharmacologic intervention for prevention or treatment (Appendix B). Appendix C lists our outcomes of interest and indicates the ones we prioritized for assessment of strength of evidence (SoE). Outcome prioritization was based on consultation with the panel of key informants and technical experts. Studies could be of any sample size and design, including randomized controlled trials (RCTs), prospective or retrospective nonrandomized comparative studies (NRCSs), single-group studies, and N-of-1 trials.

For indirect evidence, we included SRs of studies assessing harms of the same interventions (described in Appendix B) in women who were pregnant, attempting to become pregnant, postpartum, or breastfeeding, regardless of indication for the intervention. SRs had to fulfill each of four minimum criteria that we established a priori: (a) specified eligibility criteria for including studies, (b) conducted a comprehensive search (defined as searched at least two electronic databases and searched for unpublished studies through at least one source), (c) assessed risk of bias (RoB) in included studies using any instrument, and (d) used appropriate methods for meta-analysis, if conducted.

Data Extraction, Study RoB Assessment, and SR Quality Assessment

For each study, one investigator assessed RoB and extracted data into the Systematic Review Data Repository (https://srdr.ahrq.gov). All extractions were verified independently by a second investigator. To assess RoB in primary studies, we implemented questions from the Cochrane Risk of Bias Tool,[8] the Risk of Bias in Nonrandomized Studies (ROBINS-I) Tool,[9] and the National Heart, Lung, and Blood Institute (NHLBI) Quality Assessment Tool.[10] We assessed the quality of the SRs using specific items from AMSTAR 2 (A Measurement Tool to Assess SRs, Version 2).[11]

Strength of Evidence Assessment

We graded the SoE as per the AHRQ Methods Guide.[7] As part of this multidimensional assessment, we considered RoB, consistency across studies, precision, and directness. For each prioritized outcome within each intervention comparison, we assigned a SoE rating of high, moderate, low, or insufficient. To determine each SoE, we began with a high rating and then iteratively downgraded for each domain (RoB, consistency, precision, and directness) for which a problem was identified. Grades of high, moderate, and low SoE indicate the degree of confidence we have that the estimate lies close to the true effect; a rating of insufficient suggests that, based on the available evidence, we are unable to estimate an effect or we have no (or very low) confidence in the estimate of effect.[7]