We measured the dependent variable, adverse reproductive health outcomes, by asking participants, “did you experience any of the following during or after your most recent exposure to a chemical agent used by law enforcement such as tear gas/canisters, smoke bombs/grenades, or pepper spray/mace?” Respondents who answered, “I experienced unexpected or early bleeding/period/menstruation” or “I experienced late or delayed bleeding/period/menstruation” or reported “uterine cramping” or “breast tenderness” were coded as having experienced an adverse reproductive outcome. We coded those who answered, “I did not experience any unusual or unexpected disruptions to my menstrual cycle/bleeding/period after being exposed” or reported no menstrual symptoms after exposure as not having experienced an adverse reproductive outcome.

Participants reported the number of days of exposure to chemical agents when asked, “How many days were you most recently exposed to chemical agents used by law enforcement/ police, such as tear gas/canisters, smoke bombs/grenades, or pepper spray/mace?” Most recently is used in this question because a previous question assessed lifetime exposure. We used quartiles to create a categorical variable of exposure, given that some people reported between 1 and 152 exposures and the distribution was quite right skewed. This variable was coded as follows Q1 “1 day,” Q2 “2 days” Q3 “more than 2 days but less than 5 days” Q4 “5 days or more.” We coded demographic controls as follows: Respondents’ self-reported age was categorized into four groups, 18–22 years, 23–29 years, 30–39 years, and 40 years or older. Gender identity was self-reported and respondents could identify as 1 “Female/cis-woman,” 2 “Male/cis-man” 3 “Genderqueer or non-binary,” 4 “Transgender female,” 5 “Transgender male,” or 6 “Other.” Because our sample was limited to people with a uterus, those who identified as 2 “Male/cis-man” and 4 “Transgender female,” were not eligible in this study. Due to small cell sizes, we had to create a dichotomous gender identity variable with 0 “female/cis-woman: and 1 “genderqueer, transgender female, or other.” Because a majority of the sample was white and sample sizes were small across other racial groups, we grouped race into a dichotomous racialized variable of 1 “white” and 2 “racialized.” “Racialized” included any person who identified as Black, American Indian/Alaska Native, Asian, Native Hawaiian/Pacific Islander, Other, Hispanic/Latinx, and Middle Eastern and North African.

The 9 census divisions (19) were collapsed into 4 regions in alignment with US Census categorization: 1 “Northeast (Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont, New Jersey, New York, and Pennsylvania)” 2 “Midwest (Illinois, Indiana, Michigan, Ohio, Wisconsin, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, and South Dakota), 3 “South (Delaware; Florida; Georgia; Maryland; North Carolina; South Carolina; Virginia; Washington, D.C., West Virginia, Alabama, Kentucky, Mississippi, Tennessee, Arkansas, Louisiana, Oklahoma, and Texas), and 4 “West (Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming, Alaska, California, Hawaii, Oregon, and Washington).

We coded Marital status dichotomously as 0 “Not married” which included single, divorced, separated, and widowed, and 1 “Married” which included people who reported being married or partnered. Combined family income was self-reported and coded into five categories: 1 “Less than $25 K,” 2 “$25 K or greater but less than $50 K,” 3 “$50 K or greater but less than $75 k,” 4 “$75 k or greater but less than $100 k,” 5 “$100 k or greater.”

We coded health behaviors and conditions as follows: Respondents self-reported their menstrual pattern as 1 “Regular,” 2 “Irregular,” 3 “I do not get my period,” and 4 “Other.” Those who reported they did not get their period were excluded from this analysis due to the nature of the research question and to assess the role of stress on menstrual experience. We measured Hormone use dichotomously, self-reported by participants. To control for other reproductive health conditions, we created a variable for previous reproductive health diagnosis. We asked participants if they have ever been diagnosed with any of the following conditions: polycystic ovary syndrome, premenstrual dysphoric disorder, endometriosis, ovarian insufficiency, menopausal symptoms, overactive thyroid (hyperthyroidism) or underactive thyroid (hypothyroidism), or uterine fibroids. If they checked yes for any they were coded as 1 “Previous reproductive health diagnosis;” if none were checked they were coded as 0 “No previous reproductive health diagnosis.” Lastly, to determine if they have a menstrual stress response history, we asked participants “In the past, have you noticed changes in your bleeding/menstrual cycle/period when you've experienced high amounts of stress?”. If they said “Yes, I bleed earlier than normal,” “Yes, I bleed later than normal,” “Yes, I bleed heavier than usual,” or “Yes, I bleed less than normal”, we coded these participants as 1 “History of menstrual stress response.” If they responded “No, I don't notice any change” or “I’m unsure if I experience changes,” we coded them as 0 “No history of menstrual stress response”.

Survey responses were eligible for analysis if respondents reported tear gas exposure in the United States between in 2020 or 2021 and completed the full survey without missing data. We used STATA (20) to generate descriptive statistics for the study sample including the distribution of participant characteristics with percentages, means, and standard errors. We stratified these results by those participants reporting no adverse reproductive health outcomes and those reporting adverse reproductive health outcomes to understand how different characteristics are distributed across the outcome of interest. We conducted chi-square tests to determine if a significant difference in distribution existed between groups. Next, we created three nested logistic regression models starting with a bivariate logistic model between our key predictor—number of days of exposure to chemical agents—and our dichotomous outcome of reporting or not reporting adverse reproductive health. We then performed multivariable logistic regression with sociodemographic predictors (e.g., gender identity, combined family income) added in Model 2, and health predictors (e.g., hormone use) added in Model 3. We additionally ran a fourth model to test for moderation of a history of menstrual stress response on number of days of exposure with results provided in Supplemental Materials, as no significant results were found.

This study presents a clear association between exposure to chemical agents used by police and reported adverse reproductive health outcomes, which included uterine cramping (69%), early menstrual bleeding (55%), breast tenderness (30%), and delayed menstrual bleeding (19%). Numbers of days of exposure to tear gas, menstrual stress response history, income, hormone use, and region of exposure were significantly associated with reproductive health outcomes related to tear gas exposure. Number of days of exposure had a significant positive associations with experiencing an adverse reproductive health outcome as participants who reported adverse reproductive health outcomes. This finding confirmed our hypothesis and, while the magnitude of each adverse reproductive health outcome was not quantified in this analysis, the significant association and increased days of exposure may indicate a crude dose-response pattern which has been observed for other health outcomes previously in tear gas exposure studies (10).

While only 56% of participants reported having a menstrual cycle that was historically disrupted during times of high stress, the association between a menstrual stress response history and reporting adverse reproductive health outcomes following tear gas exposure was significant and confirmed our hypothesis. To our knowledge, this study is the first to assess the role of menstrual stress response history as a potential mediating factor in the association with chemical agent exposure. Our findings suggest several possibilities. Participants with a menstrual stress response history may have experienced adverse reproductive outcomes based on the stress and trauma of police violence experienced during protesting, regardless of tear gas deployment, which is consistent with the research literature on reproductive health impacts of police violence (21). However, our results show that tear gas exposure is still associated with these adverse reproductive health events even after controlling for menstrual stress response, suggesting that the stress experienced during protesting alone cannot explain the reported impacts on reproductive health. Another possibility is that participants with underlying irregular menstrual responses during stress may be more vulnerable to the potential reproductive health effects of tear gas exposure. This possibility may also help explain our finding that having a previous reproductive health diagnosis and underlying reproductive health conditions were also significantly associated with adverse reproductive health outcomes following tear gas exposure. These findings suggest that people with reproductive health conditions may be more vulnerable to the impacts of tear gas.

The strength of the association found with hormone use (OR = 37.74, CI: 22.22, 64.10) is likely related to the large volume of total participants (91%) who reported hormone use, such as hormonal contraceptives or gender-affirming hormone treatment, at the time of most recent tear gas exposure. These treatments are often prescribed to regulate irregular menstrual cycles and/or suppress menstruation altogether. As such, we might expect exogenous hormones to be protective against potential menstrual irregularities related to tear gas exposure due to their menstrual stabilizing and suppressing mechanisms of action. We posit that our high odds ratio indicating the opposite finding is due to overrepresentation of participant hormone use in the total sample, likely related to the recruitment methods that required social media connections to Planned Parenthood and may have included Planned Parenthood patients prescribed hormones. We theorize people who have very regular menses due to hormone use are more likely to notice a change.

Geographic region emerged as significantly associated with reproductive health outcomes. Participants exposed in the Southern United States were more likely to report adverse reproductive health outcomes (2). While the majority of participants in this study were white and not directly oppressed by structural racism, they are still impacted by the intersections of structural racism and structural violence through increased police force and weaponized responses that are specifically heightened at protests for racial justice. Police violence is in and of itself is a manifestation of structural racism (22, 23).

Several pathophysiological pathways point to biological plausibility between tear gas and menstrual disruptions, including it metabolizing into cyanide and subsequent hypoxia and adrenal and thyroid dysregulation, indicating potential endocrine disruption (24). Additionally, CS gas is known to activate pain receptors (25) that are also involved in pain related to dysmenorrhea (26, 27), endometriosis (27, 28), and other uterine conditions (29), helping to explain uterine cramping reported by participants.

This study had several limitations. First, sample selection and recruitment methods presented opportunities for selection bias due to exclusive digital recruitment through social media and email listservs, targeting those with online connections to Planned Parenthood and researchers and limiting generalizability. Further, response bias may have impacted our data as people who felt most impacted by tear gas exposure may have been more likely to respond to the survey and our racially homogenous sample may indicate hesitancy by people of color to participate in health care research due to long-standing histories of medical and institutional racism. As this study relied on self-report, bias could be introduced without additional data sources like medical records to confirm outcomes. Importantly, we cannot completely disentangle the impacts of the chemical exposure from the psychological factors.

The retrospective and cross-sectional design also impact study strength. Given the uncontrolled context of law enforcement tear gas use, quantifying exposure and distinguishing between the types of chemical agents used is challenging. Participants differed in frequency and duration of exposures, ranging from repeat exposures over several months to spill-over exposure in their residential areas. This study was not designed to establish a direct causation between tear gas exposure and adverse reproductive health outcomes, as such a study would be unethical.

Despite these limitations, this study systematically collected nationwide data on tear gas exposure and related adverse reproductive health outcomes during racial justice protests in the United States, making significant contributions to the field. Our community-based sampling, while introducing potential bias compared to medical records, also allowed us to recruit participants who would or could not access health care, increasing our reach and response and including valuable health equity analyses in our study. Our study also collected robust data on participants’ menstrual and reproductive health and histories that proved significant in our models and allowed us to identify important factors for further investigation.

This study establishes an association between tear gas exposure and adverse reproductive health outcomes and strongly supports adding adverse reproductive health outcomes to the growing list of safety concerns about the use of tear gas on the public. The implications of these findings for reproductive, perinatal, maternal, and infant safety are particularly concerning, as this study identifies potential populations who may be more vulnerable to health impacts of tear gas, including pregnant people. Of note, our sample included 19 participants who reported being pregnant at the time of exposure to tear gas, 10 of whom reported a subsequent pregnancy loss (52%), a much higher rate than the expected miscarriage rate of 26% for all pregnancies (30). Given the small sample size, we cannot draw significant conclusions about tear gas exposure and pregnancy outcomes, however this finding raises numerous questions about the risks that tear gas may pose to pregnant people, and to short and long-term fertility.

We believe the evidence can only be strengthened by further studies on the risks of tear gas exposures, including confirming potential pathophysiological mechanisms for reproductive harm; in-depth analysis of the associations between adverse reproductive health outcomes, tear gas exposure, menstrual stress, hormone use and underlying reproductive health conditions; and surveillance studies that help to systematically identify actual chemical agents and the deployment procedures/conditions used by law enforcement to better describe exposure. While more evidence can help fill in scientific gaps, ultimately the onus of proving the safety of tear gas resides with manufacturers and end-users of these chemical weapons, including law enforcement and government agencies. In the absence of such evidence and reassurances and in the presence of evidence suggesting adverse effects, the precautionary principle proves relevant, and the indiscriminate use of tear gas by law enforcement warrants serious inquiry and reconsideration from policymakers. Meanwhile, increased transparency about and surveillance of chemical agents and conditions for use by law enforcement is called for to protect the public.