Association between SARS-CoV-2 vaccination and healthcare contacts for menstrual disturbance and bleeding in women before and after menopause: nationwide, register based cohort study.

Menstrual disturbances such as excessive, frequent, and irregular menstruation or absent, scant, and rare menstruation have been reported in association with SARS-CoV-2 vaccines. The US Vaccine Adverse Event Reporting System, the UK Medicines and Healthcare Products Regulatory Agency’s Yellow Card surveillance scheme, and the Swedish Medical Products Agency have received many reports of menstrual disturbance after SARS-CoV-2 vaccination via their respective pharmacovigilance systems. 1 2 3 Several studies on self-reported menstruation cycles after SARS-CoV-2 vaccination, from survey data and a menstrual cycle tracking app, indicate changes in menstruation cycles. 4 5 6 7 8 9 A link between SARS-CoV-2 vaccination and menstrual disturbance has also been widely discussed on social media. 10 However, menstrual cycles vary naturally and minor menstrual disturbances are generally not considered to be of clinical importance. Changes can, however, generate considerable distress in the affected women, especially during a mass vaccination campaign when concerns are raised about adverse reactions that might not yet be well characterised. 11 The Pharmacovigilance Risk Assessment Committee of the European Medicines Agency has recommended listing heavy menstrual bleeding as a side effect of unknown frequency in the product information for the SARS-CoV-2 mRNA vaccines. The recommendation follows a review of the available evidence, including cases reported during clinical trials, cases spontaneously reported in Eudravigilance, and findings from the medical literature. 12 Previously, investigations researched concerns about menstrual disturbances from other vaccines (eg, against human papillomavirus), but no such association was established. 13 14 15 Pharmacovigilance systems relying on self-reporting are useful for identifying potential safety signals but not suited for quantifying the frequency of health event occurrence or estimating the strength of the potential association. To characterise and quantify suspected adverse effects of SARS-CoV-2 vaccines, outside what is detected in clinical trials, individual level data from large observational studies are needed. 16 In a nationwide cohort study in Sweden, we evaluated the risks of menstrual disturbance and bleeding after SARS-CoV-2 vaccination in women who were before or after menopause. High quality data from nationwide registers enabled us to evaluate the risk by vaccine product and vaccination dose number.

In total, 2 946 448 girls and women aged 12-74 years were included in the vaccination analyses. Of these, 2 580 007 (87.6%) received at least one SARS-CoV-2 vaccination before the end of follow-up on 28 February 2022. Among the vaccinated, 1 652 472 (64.0%) of 2 580 007 women received three doses, but this proportion varied by age ( fig 1 ). Participants’ demographics and medical history are presented in table 1 and table S2 by vaccine status. Women can contribute with person-time to more than one vaccine status group. Cumulative proportion of vaccine uptake (up to four doses) in different age groups, between 1 January 2020 and 28 February 2022, among women in a Swedish population cohort. Vaccination started on 27 December 2020 for the oldest age group and patients at highest risk More than 99% of menstrual disturbance (19 329/19 443 cases in the National Patient Register) or bleeding disorder diagnoses (9370/9407 cases) in the overall study population were from specialist outpatient care. In the subpopulation where primary care data were available (n=1 156 260, approximately 40% of the Swedish female population), about 11% (666/6207 cases) of the diagnoses reflecting premenopausal and postmenopausal bleeding, and 19% (2119/11 344 cases) of diagnoses of menstrual disturbance were recorded in primary healthcare. Crude annual rates of the outcomes during the study period of 2015-22 were of similar magnitude (supplement table S3). For the analyses of menstruation or bleeding disorders after a positive SARS-CoV-2 test, 754 991 (25.7%) of 2 942 544 women tested positive for SARS-CoV-2 during the study period of 1 August 2020 to 26 December 2020. Adjusted hazard ratio comparing the risk for postmenopausal bleeding after vaccination with any dose compared with unvaccinated periods was 1.12 (95% confidence interval 1.00 to 1.25) in the one to seven days risk window and 1.14 (1.06 to 1.23) in the 8-90 days risk window ( table 2 , supplement figure S1). The impact of adjustment for covariates was modest. The highest risks were observed after the third dose, both in the one to seven days risk window (1.28 (1.01 to 1.62)) and the 8-90 days risk window (1.25 (1.04 to 1.50)). The precision of these estimates was overall good. The results from the subpopulation with primary care data showed similar pattern to the main analyses, but with generally lower risk estimates ( table 3 , supplement figure S2). After restriction to women without prior hormone treatment, increased risks were observed after the third dose in both risk windows, with slightly higher estimates than in the main analyses ( table 2 , supplement table S4). The strongest association was reported in the third dose in the one to seven days risk window (1.48 (1.12 to 1.94)); similar risks were also observed in the subpopulation with primary care data, most obviously for the second dose in the 8-90 days risk window (supplement table S4). Exclusion of women with prior coagulation disorders did not change the results notably compared with the main analyses ( table 2 , supplement table S5). Hazard ratios (HR) with 95% confidence interval (CI) for menstruation disorders after each dose in one to seven days and 8-90 days risk windows, among women in a Swedish population cohort Crude model included no covariates. Full model included age, country of birth, employed as a healthcare worker, marital status, education, and health seeking behaviours during 2018-19 (ie, no. of primary care visits, number of specialist outpatient visits, and days of inpatient stay), and prior comorbidities and treatments listed in supplement table S1. Hazard ratios (HR) with 95% confidence interval (CI) for menstrual disturbance and bleeding after each dose in one to seven days and 8-90 days risk windows in the subpopulation with primary care data (Stockholm region and Västra Götaland region, approximately 40% of total population), among women in a Swedish population cohort Crude model included no covariates. Full model included age, country of birth, employed as a healthcare worker, marital status, education, and health seeking behaviours during 2018-19 (ie, no. of primary care visits, no. of specialist outpatient visits, and days of inpatient stay), and prior comorbidities and treatments listed in supplement table S1. Analyses of associations from the full model with individual vaccine products suggested an increased risk of 41% (one to seven days) and 23% (8-90 days) with BNT162b2 after the third dose, as well as 14% increased risk during the 8-90 days risk window after the second dose ( table 4 , supplement figure S3). No increased risk was observed after the first dose with BNT162b2. For mRNA-1273, risk increased by 33% after the third dose in the 8-90 days risk window. The risk estimates for mRNA-1273 and ChAdOx1 nCoV-19 were overall imprecise ( table 4 , supplement figure S3). Hazard ratios (HR) with 95% confidence interval (CI) for postmenopausal bleeding after each dose in one to seven days and 8-90 days risk windows, stratified by vaccine product, among women in a Swedish population cohort Crude model included no covariates. Full model included age, country of birth, employed as a healthcare worker, marital status, education, and health seeking behaviours during 2018-19 (ie, no. of primary care visits, no. of specialist outpatient visits, and days of inpatient stay), and prior comorbidities and treatments listed in supplement table S1. The adjusted hazard ratio for menstrual disturbance after vaccination with any dose compared with unvaccinated periods was 1.13 (95% confidence interval 1.04 to 1.23) in the one to seven days risk window and 1.06 (1.01 to 1.11) in the 8-90 days risk window. Adjustment for covariates strongly attenuated or almost completely removed the weak associations noted in the dose specific crude analyses ( table 2 , supplement figure S1). The strongest adjusted association observed was a 26% increased risk of menstrual disturbance among women aged 12-49 years in the one to seven days risk window (1.26 (1.11 to 1.42)) after the first dose. The precision of these estimates was good overall. The results from the subpopulation with primary care data were largely similar to the main analyses ( table 3 , supplement figure S2). Similarly, product specific risk estimates were largely consistent with the overall risk estimates (table S6, supplement figure S3). The adjusted hazard ratio for premenopausal bleeding after vaccination with any dose compared with unvaccinated periods was 1.08 (95% confidence interval 0.90 to 1.30) in the one to seven days risk windows and 1.01 (0.91 to 1.12) for the 8-90 days risk windows. Adjustment for covariates almost completely removed the associations reported in the crude analyses ( table 2 , supplement figure S1). The estimates were more imprecise compared with the other outcomes because of fewer observed events. The strongest associations observed, although not significant, were a 14% increased risk in the one to seven days risk window both after the first dose (1.14 (0.86 to 1.50)) and the third dose (1.14 (0.77 to 1.70)). No increased risk was observed after the second dose (0.96 (0.71 to 1.30)) in the corresponding risk window. Again, similar results were observed in the subpopulation with primary care data but with even wider confidence intervals ( table 3 , supplement figure S4). Product specific risk estimates did not show any clearly increased risks and were very imprecise (table S7, supplement figure S3). In supplement table S8, we show menstruation disorders in the subpopulation with primary care data after each dose within the risk window at days seven, 28, or 90. The risk for the three outcomes was reduced during the first seven days after a positive test (supplement table S9). However, within 90 days, the risk weakly increased, most notably for postmenopausal bleeding (hazard ratio 1.28 (95% confidence interval 0.88 to 1.86)) and premenopausal bleeding (1.45 (0.91 to 2.32)). Of note, the number of cases of premenopausal bleeding was very low. A similar pattern was observed in the subpopulation with primary care data, with slightly higher point estimates for postmenopausal bleeding (supplement table S10).

In this large population-based study of nearly three million women, we observed weak but reasonably precise associations between SARS-CoV-2 vaccination and healthcare contacts for postmenopausal bleeding. Increased risk was observed after the second and third dose in the 8-90 days risk window and was of similar size in the one to seven days risk window after a third dose. This pattern is somewhat unexpected for a causal association. Analyses of associations with individual vaccine products and risk of postmenopausal bleeding provided results that suggest an increased risk with BNT162b2 and mRNA-1273 after the third dose, but suggest a less clear association with ChAdOx1 nCoV-19. For menstrual disturbance, adjustment for covariates almost completely removed the associations found after vaccination in the crude estimates, and only a weak association remained after the first dose, limited to the one to seven days risk window. Considering the characteristics of this condition, and that the change is measured based on encounters with specialist healthcare in this study, a causal effect limited to this risk window is unlikely. The number of healthcare contacts for the outcome of premenstrual bleeding were fewer, and risk estimates after vaccination consequently more imprecise. The risk was also notably attenuated by adjustment for covariates and overall did not support an association with SARS-CoV-2 vaccination. The risk of the three outcomes did not substantially increase after covid-19, although point estimates for postmenopausal bleeding were increased in the 90 day risk window after infection. The main strengths of our study include the population-based cohort design, large sample size, near complete follow-up, and independent ascertainment of data for SARS-CoV-2 vaccinations and healthcare contacts from nationwide registers with mandatory reporting, in a setting with a universal, tax financed healthcare system. We have adjusted for socioeconomic factors, previous healthcare use, and for several specific medical conditions, including diagnosis of obesity and chronic obstructive pulmonary disease. We have no direct information on ease of access to healthcare, body mass index, or smoking. With a possible exception for postmenopausal bleeding, healthcare contacts for menstrual disorders might have a modest sensitivity. Also, we have no information on whether the healthcare contact was a planned or an acute visit. Time from first symptoms to healthcare contact is probably longer for women with menstrual disturbances and bleeding before menopause than for women after menopausal who have bleeding. Use of the date of healthcare contact for these conditions does not mean that the date of onset of the condition is analysed. The time between onset, start of symptoms, and date of healthcare contact might thus be considerable, making the interpretation of effect of different risk windows challenging. Hence, we might also, especially for menstrual disturbances and premenopausal bleeding, catch some prevalent (before exposure) cases, especially in the one to seven day time window analysed. We are unable to acquire the point in time when a woman enters menopause. Hence, we rely on the physician using the correct codes from the International Statistical Classification of Diseases and Related Health Problems for defining premenopausal bleeding or postmenopausal bleeding. Reverse causation, where women get vaccinated before a planned healthcare contact, is also an issue. Also, women with an ongoing covid-19 infection will probably cancel or postpone planned or semi-acute healthcare contacts. The concern for an association between SARS-CoV-2 vaccination and menstrual or bleeding disturbances in women has been triggered by the large number of spontaneous case reports related to such conditions. 1 2 3 10 Also, several studies on self-reported menstruation cycles changes after SARS-CoV-2 vaccination have been published. 4 5 6 7 8 9 27 28 The European Medicines Agency has recommended that heavy menstrual bleeding should be acknowledged as a side effect of both SARS-CoV-2 mRNA vaccines. 12 However, European Medicines Agency considered that the available data do not support causal association between SARS-CoV-2 mRNA vaccines and absence of menstruation. 12 The results from the present study are not necessarily contradictive of this labelling, which was mainly based on self-reported survey data and spontaneous case reports. This type of data can be prone to recall bias. Self-reporting might also obtain events that normally would not result in a healthcare contact but might still be sufficiently disturbing to be relevant for the affected women. Self-reporting, as well as health seeking behaviour, can be stimulated by media attention. 10 29 To the best of our knowledge, no previous large observational study has assessed an association between SARS-CoV-2 vaccination and healthcare contacts for menstrual or bleeding disorders using independent ascertainment of both exposure and outcome. No clear and specific mechanistic explanation allows for this type of association or supports a general such association with vaccines. 30 31 An unspecific activation of the immune system might trigger menstruation effects. 11 Two studies based on self-reported data have reported some associations between human papillomavirus vaccines and menstruation effects. 13 14 However, a large population-based study found no association between human papillomavirus vaccination and primary ovarian insufficiency. 15 Menstrual effects are not labelled in any of the influenza vaccines or hepatitis A or B vaccines used in the European Union at present. 13 14 15 We observed weak and inconsistent associations between SARS-CoV-2 vaccination and healthcare contacts for postmenopausal bleeding, and even less consistent for menstrual disturbance, and premenstrual bleeding. Extensive adjustment for confounding attenuated most risk estimates. The patterns of association are not consistent with a causal effect. These findings do not provide any substantial support for a causal association between SARS-CoV-2 vaccination and healthcare contacts related to menstrual or bleeding disorders. Large numbers of spontaneous case reports report menstrual disturbance after SARS-CoV-2 vaccination Studies that used self-reported data indicate menstrual cycle changes after SARS-CoV-2 vaccination No evidence of an increased risk of healthcare contacts for menstrual disturbances or before menopausal bleeding in a cohort of nearly three million women using independent ascertainment of both SARS-CoV-2 vaccination and healthcare contacts Postmenopausal bleeding and contacts with healthcare had a weak association, but with a pattern that is not expected for a hypothesised underlying causal association between vaccines and postmenopausal bleeding

For all individuals, we linked data from Swedish national and regional registers as an analysis within the RECOVAC (register-based large-scale national population study to monitor SARS-CoV-2 vaccination effectiveness and safety) study, which is within the larger project of SCIFI-PEARL (Swedish Covid-19 Investigation for Future Insights—a Population Epidemiology Approach using Register Linkage), described in detail elsewhere. 17 A complete medical history from 1 January 2015 was obtained from the national patient register and drug history for prescription drugs from 1 January2018 from the national prescribed drug register. 18 19 History of cancer was obtained from the national cancer register. 20 Sociodemographic data including education, family situation, income, and occupation data from 2015 were obtained from Statistics Sweden. 21 Information about pregnancy was obtained from the national medical birth register. Information about older patients living at special care facilities or receiving home care services was obtained from the register of social service interventions for the elderly and the disabled. 22 Vaccination data, including vaccine product, dose number, and date of vaccination, were obtained from the national vaccination register. 23 Positive results from SARS-CoV-2 polymerase chain reaction tests were identified from SmiNet, the national register of notifiable communicable diseases. 24 We obtained diagnoses of menstrual disturbance and bleeding in women before or after menopause from healthcare contacts registered as outpatient specialist visits or inpatient stays from the national patient register. The risk of having any diagnosis of menstrual disturbance, bleeding before and after menopause after contact with a healthcare service is hereafter referred to as risk of menstruation disorders. In Sweden, women with gynaecological issues will often, especially in urban areas, turn directly to gynaecological specialist care. However, in a subpopulation, we were also able to include information on primary care visits. Thus, for women living in the two largest metropolitan areas (Stockholm region and Västra Götaland region), diagnoses were additionally obtained from regional primary healthcare registers. The date and cause of death were obtained from the register of the total population and the national cause of death register. 25 26 The study included all women aged 12-74 years who were residing in Sweden on 1 January 2018 (to ensure previous comorbidities are accounted for), and still resident in the country on 27 December 2020, when the SARS-CoV-2 vaccine campaign started in Sweden. Data for sex was taken from information in the registry rather than from patient reported gender. The exclusion criteria were women living at special care facilities (5927 women (0.15% of those aged 12-74)) until 31 December 2020, and individuals who were pregnant or had a history of any menstruation disorders, breast cancer, cancer of the female genital organs, or who underwent a hysterectomy between 1 January 2015 (the maximum period of stored history from the register data) and 26 December 2020. The study period was from 27 December 2020 to 28 February 2022. Exposure variables were each dose of any vaccine, and several different risk periods were applied. In the main analyses, we used two mutually exclusive risk periods, one to seven days and 8-90 days after vaccination. The first seven days were deemed to be a negative control period. The time needed for an unknown pathological mechanism to manifest need to be considered, the symptoms then develop to become sufficiently worrying for the woman to seek medical attention, and the healthcare system them provides an appointment or admission, which results in a diagnosis. For menstrual disturbances, a woman is unlikely to notice any effects and be able to get an acute appointment within the first week. As menstrual cycles are around 28 days, we anticipated that a women would be delayed in deciding to seek medical attention for any disturbances. Hence, the 90 day window allows for two cycles and an additional month for the individual to get an appointment, including a potential additional interval in getting an appointment with a gynaecologist. We assessed the risk of menstruation disorders in each risk period after the administration date of the first, second, and third dose with any vaccine. Stratified analyses were performed for three specific vaccine brands used in Sweden, BNT162b2 (Pfizer-BioNTech), mRNA-1273 (Moderna), and ChAdOx1 nCoV-19 (AZD1222) (AstraZeneca). In sensitivity analyses, we also estimated risk with follow-ups at days seven, 28, and 90 starting the day after exposure date. We performed our main analyses for the whole study population and additional analyses in a subpopulation living in the two largest metropolitan areas (Stockholm region and Västra Götaland region), where regional primary healthcare data were also available. To contextualise the results, we also estimated the risk for menstruation disorders after a SARS-CoV-2 infection in women who were not vaccinated. The study period for this analysis was from 1 August 2020 (when full-scale testing was implemented in Sweden) to 26 December 2020 (when vaccinations started). In this analysis, we included all female individuals aged 12-74 years who were residing in Sweden on 1 January 2018 and 1 August 2020, who were not pregnant, did not live in nursing home on 1 August 2020, and did not have the previously mentioned comorbidities within five years before August 2020. We also studied the risk of menstruation disorders during follow-up at days seven, 28, and 90 after the first positive test result of SARS-CoV-2 infection. We studied three different menstruation disorders in different restricted age ranges (defined to include premenopausal or postmenopausal women when relevant for the respective outcomes). We identified only incident cases by using the first recording of a primary diagnosis, according to the Swedish clinical modification of the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10-SE), in one of the registers to define the outcome. Hence, outcomes were based on a healthcare contact (admission to hospital or visit) where a physician registered any of the diagnoses under study. All healthcare contacts with any of the diagnoses under study were included in the analyses including primary care, regardless of whether these contacts were related to a physician or other healthcare worker visit. The date of diagnosis was regarded as a proxy for date of onset because we have no means to assess the true start of symptoms. We studied postmenopausal bleeding in women of 45-74 years, using ICD-10-SE code N95.0. We also studied menstrual disturbance in women aged 12-49 years, using ICD-10-SE codes N91 and N92. Additionally, we studied premenopausal bleeding in women aged 12-49 years, using ICD-10-SE codes N93.8 and N93.9. For the Stockholm region and Västra Götaland region, the ICD-10-SE-P (for primary care) code N93 was additionally used. Covariates included in the full models were age (cubic spline with four knots), country of birth (Sweden/other countries), employed as a healthcare worker (yes/no), marital status (married/not married), education (primary, secondary, tertiary, undetermined), number of primary care visits, number of specialist outpatient visits, and days of inpatient stay, during 2018-19, as well as prior comorbidities and treatments (each yes/no; listed in supplement table S1, directed acyclic graphs, supplement DAG S1, and supplement DAG S2) ( table 1 , supplement table S2). Distribution of characteristics related to demographics and medical history, by vaccine status. All women were unvaccinated at the baseline, and they can contribute with person-time to more than one vaccine status group Data are number (percentage), unless otherwise stated. IQR=interquartile range; NSAIDs=Non-steroidal anti-inflammatory drugs. Cox proportional hazards models with time varying exposure were used, where each woman’s follow-up time was divided according to her vaccination status (unvaccinated, first dose, second dose, and third dose), and then at each risk window (one to seven days and 8-90 days after each dose in the main analyses, and within days seven, 28, or 90 in the sensitivity analyses). Each individual was followed up from 27 December 2020 until the earliest of the outcome of interest, end of each risk window, or a censoring event (defined as receiving a second, third, or fourth dose of any vaccine, emigration, death, or end of study on 28 February 2022). An individual contributed person-time as unvaccinated until the first vaccination. After each vaccination dose, individuals contributed person-time in each corresponding risk window of interest (ie, exposed risk time). We also restricted analyses to the subpopulation where primary care data were available. Additionally, we performed sensitivity analyses limited to women without previous hormone treatment, and in women without a prior diagnosis of coagulation disease or a filled prescription for anticoagulants. In the complementary analyses, to assess the risk for menstruation disorders at days seven, 28, or 90 after a covid-19 infection in unvaccinated women, each woman’s follow-up time was divided according to covid-19 infection status (no infection period and period after first positive SARS-CoV-2 test) and then at each risk window (within days 7, 28, or 90 after first positive test). Each woman was followed up from 1 August 2020 until the earliest of the outcome of interest, end of each risk window, or a censoring event (ie, emigration, death, or end of study on 26 December 2020). Hazard ratios with 95% confidence intervals were estimated from Cox models. We report results from a crude model without any adjustment for covariates, and a full model adjusted for all covariates listed previously. Patients were not directly involved in the study. However, the rationale for the study was around 8000 (November 2022) reports of suspected adverse drug reactions regarding menstrual disturbances that were reported to the Swedish Medical Products Agency. Approximately 90% of the suspected adverse drug reactions were reported by consumers.