Irregular Night Shifts Linked to Migraines and the Idea of a Shift Work Migraine Disorder with Body Clock-Based Solutions

Our research question followed the PECO (Population, Exposure, Comparison, Outcome) (19) format: general working population (P); irregular night shift work or rotating night shifts with ≥ 5 nights/month (E), defined by not remaining on the same schedule for ≥2 weeks; regular non-rotating night/evening or permanent night shifts with <5 nights/month (C); and migraine/headache prevalence (O), assessed via odds ratios from observational studies.

Studies were eligible if they included adult workers (≥18 years) from general populations (any occupation/industry; excluding clinical/non-working groups like patients/retirees). Exposure required frequent irregular/rotating night/evening shifts with schedule changes (34, 36); comparisons needed fixed non-rotating night/evening schedules. Evening/night shifts were grouped due to temporal overlap (e.g., 9:00 p.m.–midnight) and shared health effects.

While there’s overlap in the hours [9:00 p.m. to midnight are included in both evening and night shifts (33, 46, 57)], both evening and night shifts are categorized together in these studies due to shared characteristics and health implications. Both shift types occur during peak melatonin production and sleep propensity, contributing to similar health risks (72). Including both captures the broader impact of night shift work schedules on circadian disruption, despite overlapping hours. Outcomes of interest were the prevalence of migraine or severe headache, defined by clinical criteria (e.g., International Classification of Headache Disorders/ICHD-based questionnaire for migraine), medical interview, or self-report. We included studies reporting migraine prevalence or associations with shift work, even if they encompassed other headache subtypes [e.g., tension-type headache (TTH), chronic daily headache (CDH), or medication-overuse headache (MOH)]; we retained those with extractable migraine-specific data but also included severe headache or CDH-focused studies without explicit migraine subgroups as proxies for high-burden migraine-like presentations.

Studies were required to provide odds ratios (ORs) to quantify the association between shift work and migraine or headache prevalence, either directly reported ORs or data allowing for OR calculation (e.g., prevalence rates, contingency tables). Eligible study designs included observational studies, specifically cross-sectional studies, case–control studies, and cohort studies (prospective or retrospective), published in peer-reviewed journals or grey literature (e.g., theses, conference proceedings) with sufficient data, in any language (with translation available if needed), and with no date restrictions, given continuous knowledge updates. The 2015 NHIS report (46) was included in our systematic review as it provided relevant data on night shift work and migraine, aligning with our study’s focus.

Exclusion criteria were applied to maintain focus and quality. Studies were excluded if they exclusively involved non-workers (e.g., students, unemployed individuals, retirees), were limited to pediatric populations (<18 years) or specific clinical cohorts (e.g., only migraine patients without a working context), lacked clear definitions of night shift work (e.g., no mention of timing or irregularity), did not differentiate shift work (e.g., combining day and night shifts without separate analysis), or focused solely on day shift work without night shift comparison. Studies without a comparison group (e.g., case series or descriptive studies with no control) or those comparing night shift work to irrelevant groups (e.g., unemployed individuals) rather than the specified comparators were also excluded. Additionally, studies were excluded if they did not report migraine or headaches as outcomes (e.g., focusing only on sleep disorders or fatigue) reported only headache subtypes other than migraine (e.g., tension-type or cluster headaches), unless headaches were non-subtyped, or provided only qualitative outcomes (e.g., no prevalence or OR data). Studies not providing odds ratios or sufficient data to derive them (e.g., only p-values or narrative results), using experimental or non-associational metrics (e.g., means without prevalence), or employing non-observational designs (e.g., randomized controlled trials, laboratory-based experiments) were excluded. Reviews, editorials, or opinion pieces without original data were also excluded, though their reference lists were screened for eligible studies. Studies with insufficient detail on shift work patterns (e.g., no frequency or quick return data) were excluded. Including studies where ORs could be calculated from prevalence or contingency tables (not just reported) broadened the pool of eligible studies without sacrificing rigor. All reviewers used these criteria to screen titles and abstracts, then full texts, resolving discrepancies through consensus. The criteria ensured that studies aligned with the question’s focus on the impact of irregular night shifts on migraine and headaches, while excluding irrelevant or low-quality data.

The process for identifying, screening, and selecting studies followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The search strategy was tailored to the research question and the inclusion/exclusion criteria, specifying the databases, keywords, and time frame used to identify observational studies examining the association between irregular or rotating night shift work and the prevalence of migraine or non-subtyped headaches among workers, as measured by odds ratios or calculable odds ratios (e.g., from prevalence or contingency tables). The databases searched included PubMed and Web of Science. The search used a combination of controlled vocabulary (e.g., MeSH terms in PubMed) and free-text keywords to capture the population, exposure, comparison, outcome, and study design elements, grouped by concept and combined using Boolean operators (AND, OR, NOT). For the population (workers), terms included controlled terms like “Workers” [MeSH], “Occupational Groups” [MeSH], “Employment” [MeSH], and free-text terms such as worker*, employee*, “working population,” “labor force,” occupation*, staff, and personnel. For exposure (irregular or rotating night shift work), controlled terms included “Shift Work Schedule” [MeSH], “Work Schedule Tolerance” [MeSH], “Circadian Rhythm” [MeSH], and free-text terms like “night shift*,” “shift work,” “rotating shift*,” “irregular shift*,” “quick return*,” “short recovery,” “frequent shift*,” “shift change*,” and “night work.” Comparison terms (less frequent shifts, regular shifts, non-night work) included free-text terms such as “regular shift*,” “fixed shift*,” “non-rotating shift*,” “day shift*,” “permanent schedule*,” “non-night shift*,” and “standard work hours.” For the outcome (migraine and non-subtyped headaches), controlled terms were “Migraine Disorders” [MeSH], “Headache” [MeSH], “Headache Disorders” [MeSH], and free-text terms included migraine*, headache*, cephalalgia, “head pain,” and “cranial pain,” with a note that subtypes like “tension-type” or “cluster” would be excluded unless part of broader headache data. Measurement and study design terms (odds ratios, relative risks, observational studies) included controlled terms like “Odds Ratio” [MeSH], “Relative Risk” [MeSH], “Observational Study” [MeSH], “Cross-Sectional Studies” [MeSH], “Case–Control Studies” [MeSH], “Cohort Studies” [MeSH], and free-text terms such as “odds ratio*,” OR, prevalence, association, “cross-sectional,” “case–control,” cohort, observational, and epidemiology*.

An example search string for PubMed was as follows: (“Workers” [MeSH Terms] OR “Occupational Groups” [MeSH Terms] OR “Employment” [MeSH Terms] OR worker* OR employee* OR “working population” OR occupation*) AND (“Shift Work Schedule” [MeSH Terms] OR “Work Schedule Tolerance” [MeSH Terms] OR “night shift*” OR “shift work” OR “rotating shift*” OR “irregular shift*” OR “quick return*” OR “night work”) AND (“Migraine Disorders” [MeSH Terms] OR “Headache” [MeSH Terms] OR migraine* OR headache* OR cephalalgia) AND (“Odds Ratio” [MeSH Terms] OR “relative risk*” OR “hazard ratio*” OR “Observational Study” [MeSH Terms] OR “Cross-Sectional Studies” [MeSH Terms] OR “Case–Control Studies” [MeSH Terms] OR “Cohort Studies” [MeSH Terms] OR “odds ratio*” OR prevalence OR “cross-sectional” OR cohort OR observational). A similar structure was adapted for Web of Science: TS = ((worker* OR employee* OR “working population” OR occupation*) AND (“night shift*” OR “shift work” OR “rotating shift*” OR “irregular shift*” OR “quick return*” OR “night work”) AND (migraine* OR headache* OR cephalalgia) AND (“odds ratio*” OR “relative risk*” OR “hazard ratio*” OR prevalence OR “cross-sectional” OR “case–control” OR cohort OR observational)). Filters in PubMed included study type (Observational Study, Cross-Sectional Studies, Case–Control Studies, Cohort Studies where applicable), language (no restriction, translations sought if needed), and species (humans), while Web of Science filters included document type (Article, Conference Proceeding, Early Access), research area (Medicine, Public Environmental Occupational Health, Neurosciences, Epidemiology if applicable), and language (no restriction). There was no date restriction, with searches including all available records up to the current date (March 27, 2025), reflecting the lack of a cutoff in the inclusion criteria and the continuous knowledge update capability, which ensures comprehensive coverage of both historical and recent studies. Search strings, run dates, and result counts were recorded for transparency (Figure 1, PRISMA flow diagram). Reference lists of included studies and relevant reviews were manually searched for additional eligible studies. This strategy strikes a balance between sensitivity (capturing relevant studies) and specificity (focusing on the question’s scope). We prioritized ORs for consistency across study designs, although cohort studies often report relative risks or hazard ratios. Therefore, we included terms for these metrics in our search and calculated ORs where possible.

The following data were extracted: first author, year of publication, country, sample size, male to female ratio, study design, type of comparison (rotating vs. non-rotating night shiftwork, frequent vs. infrequent irregular night shift work), and number of people with migraine or headache (where headache was not phenotyped) in the compared groups. Six authors participated in data extraction.

The Newcastle–Ottawa Scale (73) (by YW and MF) was used to assess the quality of each included article in the following three domains: selection of study groups, comparability of groups, and ascertainment of exposure or outcome.

A random-effects model was employed for the meta-analysis to account for between-study heterogeneity, which arises from differences in study settings, populations, and sample sizes. This model provides a more conservative estimate of the overall effect compared to a fixed-effects model. The inverse variance method with the DerSimonian-Laird estimator was used to weight studies based on their precision, accommodating the wide range of study sizes and heterogeneity effectively. In cases of sparse events, the robustness of this approach was considered, with the Mantel–Haenszel method as a potential alternative for sensitivity analyses to ensure stability in estimates. Odds ratios (ORs) were used as the effect measure instead of relative risk, as ORs provide consistent and comparable associations across observational studies, particularly cross-sectional designs. ORs are less sensitive to variations in baseline risks and are well-suited for studies where temporality or causality cannot be established. Data were synthesized using the online platform1 (74) to calculate effect sizes and assess heterogeneity. Meta-Essentials software (75) was utilized to perform statistical analyses, generate forest plots, conduct meta-regression, and evaluate publication bias. All procedures adhered to standard meta-analysis guidelines to ensure robust and reliable results. The study protocol was registered on PROSPERO (CRD420250654865) on 24 February 2025. Inter-rater reliability checks and duplicate screening were conducted by YW and EH to enhance methodological rigor.

Our results align with prior evidence that fixed night shift schedules are preferable to irregular or rotating schedules for minimizing circadian disruption (70). Irregular shifts, particularly those with rapid rotations [changing night shift schedules every 1–4 days (91)], quick returns [less than 11 h of rest between shifts (92)], and consecutive night shifts (>3 nights) (93, 94) desynchronize the SCN, the master circadian pacemaker, from peripheral clocks in organs such as the liver and gut (95). This desynchronization manifests as irregular sleep–wake cycles, and gastrointestinal disturbances, and may amplify sensory processing (3, 9, 96, 97) to migraine triggers such as bright light, loud sound, and dietary irregularities, though direct evidence is needed. Fixed schedules, by contrast, allow for gradual circadian adaptation for workers transitioning from day to night shifts, involving a 12-h schedule change, with the biological clock adjusting by approximately 1–2 h per day (60, 98). Based on this adaptation rate, we recommend maintaining consistent shift schedules for at least 12–14 days to allow sufficient time for entrainment of the 12-h shift, potentially reducing physiological stress and migraine risk, though further research is needed to confirm this effect.

Rotating shifts, when unavoidable, should follow a delay direction (morning → evening → night) rather than an advance direction (night → morning) (99–101). This aligns with the natural tendency of the human circadian clock to delay rather than advance, as demonstrated in a meta-analysis, which found lower rates of sleep disruption and mood disturbances with delay-rotated schedules (99–101). Additionally, minimizing consecutive night shifts—ideally to one per cycle—reduces cumulative sleep debt and cortisol dysregulation, both of which are implicated in migraine pathophysiology (102–106).

Avoiding quick returns (shifts with <11 h of rest) reduces fatigue and sleep disturbances, as short inter-shift intervals exacerbate circadian misalignment (92, 107). However, Katsifaraki et al. (108) found no association between sleep duration and headaches in shift workers (OR = 1.00, 95% CI: 0.97–1.02), suggesting headache triggers may involve circadian factors rather than sleep loss.

Our meta-regression and subgroup meta-analysis of 10,503 individuals (45% female) reveal a significant sex disparity in night shift migraine risk, with females facing over twice the odds versus males. This aligns with women’s higher migraine prevalence, amplified by circadian disruption (50). Despite lower night shift participation (15.2% vs. 17.6% for males) (46), females bear greater shift work-migraine burden, underscoring the need to probe underlying mechanisms. This sex disparity may stem from inherent circadian differences, with females exhibiting a shorter free-running circadian period (24.2 vs. 24.5 h in males) (58, 109–111) that heightens vulnerability to night-shift desynchronization. Hormonal fluctuations (menstrual cycles, pregnancy, or menopause) exacerbate misalignment effects (112), while females show greater sensitivity to sleep deprivation and stress (113, 114), further compounded by caregiving duties (115, 116) and circadian-timed stress responses, as seen in female mice with exaggerated active-phase disruptions in clock genes (113, 114).

Our study links irregular night shifts to higher migraine burden, proposing “shift work migraine disorder” (SWMD) as a novel chronobiological subtype warranting validation. We hypothesize chronic circadian misalignment amplifies trigeminovascular activation (117) via neuroinflammation and oxidative stress (118–120), evidenced by higher migraine prevalence in night vs. day workers and irregular vs. fixed night workers. SWMD likely increases susceptibility to circadian triggers, necessitating chronobiology-focused management. SWMD’s migraine specificity (vs TTH) aligns with greater sleep disturbances in migraine (121, 122); the absent TTH link underscores circadian disruption’s preferential migraine impact. Longitudinal studies should validate SWMD’s pathophysiology for potential ICHD inclusion, with proposed criteria:

These criteria distinguish SWMD from subtypes (e.g., chronic or menstrual migraine) by emphasizing circadian triggers, akin to shift work sleep disorder (SWSD) (123), a recognized circadian rhythm sleep disorder. Though sleep disruption triggers migraine (106, 124), SWMD uniquely ties attacks to shift transitions, differing from temporary jet lag disorder (125) and mirroring SWSD’s chronicity (Table 2). Unique factors include occupational context and melatonin suppression (72, 126–128), supporting targeted interventions and policies. Fixed schedules (e.g., 14-day consistency) and chronotherapy could aid alignment, unlike easier jet lag recovery, but require testing. In a study of 2,762 participants, social jet lag affected 18.9%, more in headache sufferers (21% vs. 17%, p = 0.006) (129), but not migraine-specific (22.4% vs. 20.8%, p = 0.651) (129) – indicating broad circadian effects without migraine distinction. Conversely, occupational SWMD merits separate recognition. Longitudinal validation and targeted interventions are essential; the Phase 3 solriamfetol trial for SWSD [NCT06568367↗ (130)] – a wake-promoting agent targeting TAAR-1, dopamine, and norepinephrine pathways overlapping with migraine mechanisms—may offer translational insights for SWMD due to shared circadian disruptions, but only as a hypothesis pending dedicated, migraine-specific efficacy trials.

Recent within-person data from the 1,001 Nights cohort (131) bolster our SWMD proposal as a distinct chronobiological subtype, showing 31% higher headache prevalence on night shifts (aPR/adjusted prevalence ratio 1.31, 95% CI 1.13–1.52) persisting after adjustments for sleep, psychosocial stressors, and demands—echoing our meta-analysis OR of 1.61 for irregular shifts and emphasizing circadian misalignment over sleep disruption. Diary-based sleep controls (naps/latency, quality) uphold the association (aPR 1.31 from 1.33), affirming non-sleep-mediated hypothalamic-trigeminal disruption. Second-night peaks (aPR 1.54 vs. days, 95% CI 1.28–1.84) pre-adaptation refine phenotyping for early irregular exacerbations, aligning with our irregular/fixed split, endorsing slow rotations (≥5 days), and urging validation. Its female-dominant sample (n = 522, 14% migraine) improves our female predominance meta-regression (β = 0.70, p = 0.0003) generalizability via within-person design, advancing chronotherapy and ICHD-3 candidacy for preventable SWMD.

Beyond shift scheduling, our findings support multifaceted interventions, including chronotherapy (e.g., timed light) and policies such as mandatory recovery days, alongside sleep hygiene education (cool/dark environment, stimulus control, and sleep restriction) – proven effective in a meta-analysis for shift workers (132). Further research is needed on circadian-migraine links to refine strategies (Table 3).

Dynamic lighting mimicking day-night cycles (high melanopic lux 250–300 during shifts, dim <50 lux post-shift) reduces fatigue/mood issues (133), potentially stabilizing rhythms to cut migraine frequency. Blue-blockers post-shift reduce melatonin suppression (128); early-shift bright melanopic light (460-480 nm) boosts alertness (128, 134). Light therapy glasses (461 nm) cut sleepiness post-first shift (p = 0.012) (135), blue-enriched light (17,000 K) improves subjective alertness (136), and meta-analyses confirm benefits (134). Pre-Tmin bright light (7,000–12,000 lux, 2 h pre-wake) accelerate phase delays (133), while post-Tmin darkness (e.g., blue-blockers on commutes) aid adaptation and could reduce migraine – subject to verification (133, 137).

Evening/nocturnal exercise (e.g., 19:00–22:00 or 00:30, moderate-high intensity) delays melatonin for better alertness/sleep (138–140). Time-restricted eating (10 h window post-shift) syncs metabolism, enhances cognition, cuts cardiometabolic risks (141–144). Cooler ambient temps (~23 °C) boost alertness, reduce discomfort, and adapt melatonin (145). Social interactions during/post-shift anchor rhythms (146–149), amplified by light strategies (150). Judicious caffeine (100-200 mg at onset) combats sleepiness (151) but risks daytime sleep disruption (152–154). Melatonin (0.5-3 mg, 1-2 h pre-sleep) aids re-entrainment (155, 156); a 2024 meta-analysis favors 4 mg 3 h pre-bed for optimization, with timing for phase shifts (morning delay, evening advance) (156, 157).

Sleep disturbances in 8 studies tie chronodisruption—through circadian misalignment—to migraine exacerbation, as irregular sleep rhythms heighten trigeminovascular activation, reinforcing our SWMD proposal. Incomplete sleep adjustments may still pose residual confounding in shift work-migraine associations, despite high NOS comparability (mean 9.6/10) indicating robust quality. The 1,001 Nights cohort (131) resolves this by confirming sleep non-mediation, with associations persisting (aPR 1.31) post-rigorous controls for sleep duration, quality, and timing. Chronic circadian misalignment from irregular night shifts in SWMD is hypothesized to disrupt the hypothalamic-trigeminovascular pathway, amplifying neurogenic inflammation and cortical spreading depression susceptibility, independent of sleep disruption—as evidenced by persistent associations post-sleep adjustment in cohort studies (131).

The moderate heterogeneity observed in our meta-analysis is scientifically reasonable, given the diversity of the included studies, i.e., different working populations, countries, settings, and time periods. A consistent direction of effect across most studies (as seen in the forest plot, with the majority to the right) supports the observed trend. Future trials should assess objective sleep assessments (158) to refine interventions. A key limitation is the heterogeneity in migraine diagnostic classifications across included studies—from rigorous ICHD-3 criteria or physician validation (n = 5) to unvalidated self-report (n = 9)—which may introduce misclassification bias by over-including non-migraine headaches or under-detecting subclinical cases, potentially yielding conservative pooled ORs. However, sensitivity analyses by validation type (OR 1.64 validated vs. 1.54 self-report, p = 0.83) and specificity (OR 1.40 migraine-specific vs. 2.53 severe headache, p = 0.002) confirm robustness, with the elevated severe headache estimate likely reflecting undiagnosed high-burden migraine cases more susceptible to circadian disruption, further supported by subgroup analyses showing significant shift work associations only for migraine (OR 1.40, 95% CI: 1.04–1.89) versus null for tension-type headache (OR 0.92, 95% CI: 0.71–1.19; p = 0.01 for difference), reinforcing the hypothalamic-trigeminovascular pathway’s selectivity.