Covid-19 vaccination and menstrual cycle length in the Apple Women’s Health Study

The study investigates whether COVID-19 vaccination is associated with changes in menstrual cycle length, addressing widespread anecdotal reports and early studies suggesting temporary menstrual alterations after vaccination. Despite established vaccine safety and no evidence of impacts on fertility, concerns persisted due to self-reported menstrual changes to surveillance systems (e.g., VAERS, MHRA) and media reports. The research aims to quantify the onset, magnitude, timing (by menstrual phase), and persistence of any changes in menstrual cycle length following vaccination using data from the Apple Women’s Health Study (AWHS). Understanding these associations can inform clinical counseling and public health communication regarding vaccination.

Prior research has noted temporary menstrual changes post COVID-19 vaccination, including increased cycle length, heavier bleeding, longer periods, irregularities, and breakthrough bleeding. Surveillance systems (VAERS in the U.S. and the UK MHRA Yellow Card) received thousands of reports of menstrual disturbances after vaccination. However, vaccines are considered safe, with no evidence of detrimental effects on fertility from laboratory, clinical trial, or observational data. Earlier cohort studies (e.g., Edelman et al., Alvergne et al.) observed small increases in cycle length during vaccination cycles. This study builds on that literature by using longitudinal, prospectively tracked menstrual data and by examining phase-specific timing effects and persistence of changes across subsequent cycles.

Design and population: The Apple Women’s Health Study (AWHS) is a prospective digital cohort in the U.S., recruiting participants beginning November 2019 via the Apple Research app. Eligibility included being assigned female at birth, having menstruated at least once, residing in the U.S., age ≥18 years (state-specific minima applied), English literacy, sole iPhone and iCloud account user, and informed consent. The study was IRB-approved (Advarra PR0000375) and registered (NCT04169859). Participants were surveyed monthly about menstrual health. Vaccination survey: A COVID-19 Vaccine Update survey (first deployed September 2021; re-administered quarterly) collected vaccination status, dates of doses, vaccine type (Pfizer-BioNTech, Moderna, Johnson & Johnson/Janssen (J&J), or other), and 48-hour post-dose symptoms. Implausible vaccination dates (before December 2020 or after February 2022, or second dose before first) led to exclusion. Cycle tracking and definitions: Menstrual data were logged via Apple Health Cycle Tracking or compatible third-party apps, including historical entries up to 2 years pre-enrollment. Spotting was excluded. A cycle was defined from the first day of menstrual flow through subsequent days of flow, followed by at least two days with no tracked flow. Cycles shorter than 7 days or longer than 90 days were excluded as non-natural or likely artifact. Additional artifact filtering used participant-specific thresholds informed by prior work. Each cycle in vaccinated participants was classified as: pre-vaccination (entire cycle before first dose), first dose (cycle of first mRNA dose), second dose (cycle of second mRNA dose), J&J dose (cycle of J&J single dose), and post-vaccination cycles (numbered sequentially after completion of the primary series: second mRNA dose or single J&J dose). Analyses of persistence were limited to the first four post-vaccination cycles (to avoid booster confounding). Timing within the menstrual cycle was approximated: luteal phase defined as the 14 days before the start of the next cycle, and follicular phase as the remainder. Participants receiving two doses in a single cycle were excluded from main analyses due to definitional constraints on cycle length; such scenarios were evaluated in sensitivity analyses. Covariates: Potential confounders included age (categorized), race/ethnicity (collapsed into broader categories due to small cell sizes), body mass index (BMI: underweight, normal, overweight, obese), and calendar time/seasonality. BMI came from self-reported height/weight at enrollment or annual updates. Missingness was handled via missing categories where relevant. Statistical analysis: Primary analyses used linear regression with participant fixed effects to estimate within-participant differences in mean cycle length (MCL) between pre-vaccination cycles and vaccination/post-vaccination cycles, adjusting for age, BMI, and seasonality (month-year). Participant ID fixed effects controlled all time-invariant between-participant confounders. The probability of a long cycle (≥38 days, per FIGO) was assessed with conditional logistic regression yielding odds ratios (ORs) versus pre-vaccination cycles. Phase-specific effects compared vaccination in follicular vs luteal phases using differences between regression coefficients. Secondary between-participant analyses used linear mixed-effects models (random intercepts) for MCL and GEEs for long-cycle odds comparing vaccinated vs unvaccinated participants. Sensitivity analyses included: restricting to participants with ≥3 tracked cycles (and, for vaccinated, ≥2 pre-dose cycles plus ≥1 vaccination/post-vaccination cycle), restricting to average cycle length 24–38 days, complete case analysis (excluding cycles with missing covariates), restricting to participants reporting never testing positive for COVID-19, restricting to cycles logged before vaccine survey completion, and inclusion of cycles with two doses in a single cycle or cycles between doses. Bonferroni-adjusted confidence intervals were presented for the seven primary associations. Analyses were conducted in Python 3.6.

Sample: 9,652 participants contributed 128,904 menstrual cycles (median 10 cycles per participant, IQR 4–22); 88% were vaccinated. Among vaccinated participants: 55% received Pfizer-BioNTech, 37% Moderna, and 8% J&J. Within-participant differences in mean cycle length (vs pre-vaccination cycles): - mRNA first dose cycle: +0.50 days (95% CI: 0.21, 0.78) - mRNA second dose cycle: +0.39 days (95% CI: 0.11, 0.67) - J&J dose cycle: +1.26 days (95% CI: 0.45, 2.07) - Post-vaccination cycles: returned to average pre-vaccination length across subsequent cycles. Phase-specific timing effects (difference vs pre-vaccination): - Follicular phase vaccination: • First mRNA dose: +0.97 days (95% CI: 0.53, 1.42) • Second mRNA dose: +1.43 days (95% CI: 1.06, 1.80) • J&J dose: +2.27 days (95% CI: 1.04, 3.50) - Luteal phase vaccination: • First mRNA dose: +0.21 days (95% CI: -0.14, 0.57) • Second mRNA dose: -0.97 days (95% CI: -1.39, -0.55) • J&J dose: +0.39 days (95% CI: -0.75, 1.53) Long cycle (≥38 days) odds during vaccination cycles (vs pre-vaccination) indicated an increase that did not persist post-vaccination; reported ORs included: - J&J dose cycle: OR 2.16 (95% CI: 1.16, 4.03) - mRNA dose cycles: modest, not consistently statistically significant increases (e.g., first dose OR 1.13 [0.87, 1.48]; second dose OR 1.08 [0.83, 1.43]). Sensitivity analyses: Findings were robust across restrictions. When two doses occurred within a single cycle, that cycle was on average +4.96 days (95% CI: 4.42, 5.50) longer than pre-vaccination cycles; cycles between doses averaged 1.94 days shorter (95% CI: -2.65, -1.23), patterns influenced by structural constraints on cycle length definition in these scenarios.

The study shows that COVID-19 vaccination is associated with small, temporary increases in menstrual cycle length during vaccination cycles, with effects concentrated when doses occur in the follicular phase. Cycles typically returned to pre-vaccination length within 1–2 cycles. The probability of experiencing a clinically long cycle (≥38 days) increased during vaccination cycles but did not persist across subsequent cycles. A second mRNA dose administered in the luteal phase was associated with a shorter cycle, underscoring the importance of timing relative to menstrual phase. The magnitude of observed changes falls well within normal intra-individual variability (<8 days difference between shortest and longest cycles is considered normal by FIGO), suggesting limited clinical impact for most individuals. Potential mechanisms may involve immune-driven inflammation affecting hypothalamic–pituitary–ovarian axis signaling, prolonging follicular recruitment (leading to longer cycles) or altering endometrial stability in the luteal phase (shortening cycles). Overall, results support vaccine safety regarding menstrual timing and provide nuanced information for counseling: any changes are minor and transient.

COVID-19 vaccination is associated with a very small, time-limited increase in menstrual cycle length, particularly when doses are received in the follicular phase; cycles typically normalize within one to two cycles post-vaccination. The findings align with prior reports and indicate that such menstrual changes should not discourage vaccination. Future research should investigate biological mechanisms (e.g., inflammatory pathways and HPO axis responses), characterize potential susceptible subgroups (e.g., those with preexisting menstrual or endocrine conditions), assess outcomes beyond cycle length (e.g., flow characteristics), validate menstrual phase with hormonal measures, and further evaluate scenarios with two doses within a single cycle or short inter-dose intervals.

- Self-reported vaccination details and menstrual tracking may introduce measurement error; limited information on user tracking behavior could affect accuracy, though fixed-effects models mitigate between-participant biases. - Analysis focused on mean cycle length and could not identify subpopulations (e.g., with preexisting conditions) who might experience larger effects. - Potential length-time bias: individuals might be more likely to receive vaccination during longer cycles, possibly biasing associations toward increased length. - Menstrual phase was inferred (luteal defined as 14 days before next cycle) without hormonal confirmation, potentially impacting phase-specific estimates. - Exclusion of cycles with two doses in a single cycle and cycles between doses from main analyses; these scenarios have definitional constraints on cycle length that complicate interpretation. - Limited data on other menstrual characteristics (e.g., flow, symptoms); hormone levels were not measured. - Residual confounding by unmeasured time-varying factors cannot be entirely excluded, though analyses adjusted for age, BMI, seasonality and included participant fixed effects. - Generalizability may be limited to U.S. iPhone users; results may not extend to all populations. - Some participants’ infection status was incomplete; however, restricting to those never testing positive for COVID-19 yielded consistent results.