Cannabis smoking, tobacco cigarette smoking, and adenomyosis risk

Most participants were white and non-Hispanic, and a majority were ages 40–49 years at the reference date ( Table 1 ). Cases tended to have an earlier age at menarche (<10 years) and a history of three or more pregnancies compared to hysterectomy and population controls. Although cases were similar to hysterectomy controls with regard to educational attainment, body mass index (BMI), and menopausal status at the reference date, cases tended to have lower educational attainment (≤high school diploma/general equivalency diploma (GED)), higher BMI (≥25.0 kg/m 2 ), and be premenopausal compared to population controls. Uterine fibroids were the most common pathology finding among cases (67%) and hysterectomy controls (72%) ( Supplemental Table 1 ). The distribution of demographic characteristics was similar between study participants enrolled in the study and those with available abstracted medical record data ( Supplemental Table 2 ). History of ever smoking cannabis was common among study participants (62% cases, 61% hysterectomy controls, 60% population controls) ( Table 2 ). At the reference date, few participants reported currently smoking cannabis and the years since last cannabis smoking among past users was similar by case status ( Supplemental Table 3 ). We did not observe an association between ever smoking cannabis and adenomyosis risk (cases vs. hysterectomy controls: OR 0.9, 95% CI: 0.6–1.3; cases vs. population controls: OR 1.0, 95% CI: 0.7–1.4). Similarly, no association was observed between joint-years of cannabis smoking and adenomyosis. Ever smoking tobacco cigarettes was also frequently reported (53% cases, 45% hysterectomy controls, 43% population controls) ( Table 3 ). At the reference date, 20% of cases, 16% of hysterectomy controls and 13% of population controls were current smokers. The distribution of years between age at reference date and age at last smoking for former smokers was similar by case status ( Supplemental Table 3 ). Our data suggested a modest association between ever smoking cigarettes (vs. never) and adenomyosis, comparing cases to hysterectomy controls (OR 1.3, 95% CI: 0.9–1.9) and to population controls (OR 1.2, 95% CI: 0.8–1.8). However, our data suggested a 50% increased odds of adenomyosis with smoking >15 pack-years (vs. never smoking) comparing cases and hysterectomy controls (OR 1.5, 95% CI: 0.9–2.6, P trend =0.135) and a 40% increased adenomyosis odds with smoking >5-<15 pack-years (OR 1.4, 95% CI: 0.8–2.4, P trend =0.136) comparing cases to population controls. In the sensitivity analyses, we observed similar results to that of the main analyses after additionally adjusting for gravidity ( Supplemental Table 4 ), considering cumulative cannabis and cigarette smoking before age 35 years ( Supplemental Table 5 ), and restricting population controls to those who would allow hysterectomy if warranted ( Supplemental Table 6 ). Although, the associations between cigarette smoking and adenomyosis were attenuated in the latter two sensitivity analyses. After additionally adjusting for alcohol consumption, our data suggested increased adenomyosis odds with cannabis smoking >0.004–0.2 joint-years (vs. 0 joint-years) comparing cases to hysterectomy controls (OR 1.3, 95% CI: 0.7–2.2) and cannabis smoking >0.2 joint-years (vs. 0 joint-years) comparing cases to population controls (OR 1.3, 95% CI: 0.8–2.2) ( Supplemental Table 7 ). In the exploratory analysis of individual use and co-use of cannabis and tobacco cigarette smoking, we observed a strong association between history of only ever smoking tobacco cigarettes and adenomyosis (cases vs hysterectomy controls: OR 2.5, 95% CI: 1.2–5.0; cases vs. population controls: OR 1.5, 95% CI: 0.8–2.8) ( Supplemental Table 8 ). Those with a history of smoking both cannabis and cigarettes, in general, had lower median pack-years of cigarette smoking compared to those who only ever smoked cigarettes ( Supplemental Table 9 ). We observed a high frequency of ever smoking cannabis among cigarette smokers (hysterectomy controls: 83%, population controls 88%), raising the concern that adjustment for joint-years of cannabis smoking in analyses of cigarette smoking and adenomyosis risk may result in over adjustment. In a post hoc analysis we repeated the analyses without adjusting for cannabis smoking. The magnitude of the associations between ever cigarette smoking, pack-years of smoking, and adenomyosis were slightly stronger ( Supplemental Table 10 ). For example, we observed 40% increased odds of adenomyosis with ever cigarette smoking (cases vs. hysterectomy controls: OR of 1.4, 95% CI: 1.0–2.0; cases vs. population controls OR 1.4, 95% CI: 1.0–1.9).

We used data from a case-control study of adenomyosis conducted among 18- to 59-year-old enrollees of a large integrated healthcare system in western Washington State, Kaiser Permanente Washington, formerly known as Group Health. When the case-control study was conducted, (cis)women-related terminology was used and data on gender was not collected. Thus, we use gender-inclusive language when describing the study population [ 22 ]. Participant recruitment and enrollment in the original case-control study is described in Appendix 1 . Cases were enrollees with a first-time diagnosis of adenomyosis between April 1, 2001, and March 31, 2006. Cases were identified by reviewing Group Health databases for hospitalizations, inpatient and outpatient surgery, and medical visits with the International Classification of Disease (ICD) 9 th Revision diagnostic code 617.0 “endometriosis of uterus”. The presence of adenomyosis in the uterine specimen from hysterectomy was confirmed by medical record review of the pathology report. Two control groups were employed. Hysterectomy controls were enrollees who had undergone a hysterectomy for a non-cancerous indication during the same period cases were diagnosed and no adenomyosis was observed in the uterine specimen per pathology report. Population controls were those enrolled in Group Health at some point between April 1, 2001 and March 31, 2006, without a history of diagnosed adenomyosis, and with an intact uterus. Population controls were randomly selected from computerized Group Health enrollment databases and frequency matched to adenomyosis cases in 5-year age-groups. All participants were required to have been enrolled in Group Health for at least six months before the reference date. The reference date was the date of first visit at Group Health for symptoms leading to adenomyosis diagnosis in cases and hysterectomy in hysterectomy controls. In the cases and hysterectomy controls, the most common indications for hysterectomy were abnormal uterine bleeding and pain ( Supplemental Table 1 ). Population controls were assigned reference dates that corresponded with the distribution of reference dates in cases. A total of 402 cases, 241 hysterectomy controls, and 354 population controls completed the informed consent process and were enrolled. Of these participants, 397 cases, 239 hysterectomy controls, and 353 population controls completed the in-person structured interview (described below). In the present analysis, we restricted the study population to those for whom medical records were available and were abstracted ( Appendix 1 ): 386 cases, 233 hysterectomy controls, and 323 populations, respectively comprising 97%, 97%, and 92% of the enrolled study participants who completed the study interview. The primary study activity of the case-control study was an in-person structured interview. Participants were asked about a range of many topics, including reproductive and pregnancy history, as well as demographic and lifestyle characteristics that occurred before the reference date. The interview was conducted by a trained female interviewer, using a life event calendar to assist participant recall of health history, including smoking history. Relating history to key life events, including dates of marriages, divorces, pregnancies, job changes, and residential moves, enhances recall [ 23 , 24 ]. The present analyses used deidentified data and were deemed not to include human subjects by the Human Research Protection Program at Michigan State University. During the structured interview, each participant was asked “Before the reference date, did you ever smoke marijuana or hashish?”. If a participant responded yes, additional questions elicited the ages at which marijuana or hashish smoking started and stopped (if not smoked for one year or longer) as well as the average frequency of smoking and the average number of inhalations. Since marijuana smoking can change over time, participants were asked about additional periods when they started regularly smoking marijuana or hashish again, including ages when they started and stopped and average smoking frequency and inhalations. Using this information, we created a binary variable on ever smoking cannabis (no, yes). We estimated lifetime cumulative cannabis smoking in joint-years. We defined joint-years as the number of joints smoked per day times the number of years a person smoked [ 25 ]. For each period reported (maximum number of periods reported was four), we converted the frequency of cannabis smoking to number of times per day, multiplied by the number of inhalations, and divided the resulting quantity by 12; approximately 12 inhalations equate to one joint. We multiplied the number of joints smoked per day by years in the period and summed the joint-years across periods. Study participants who reported never smoking cannabis were assigned the value 0 joint-years. We categorized joint-years into tertiles using the population control distribution (0, >0–0.004, >0.004–0.2, and >0.2 joint-years). To put the joint-years categories into context, 0.2 joint-years, for example, can be interpreted as approximately one inhalation per day for two years, nearly 2.5 inhalations per day for one year, or nearly 5 inhalations per day for half of a year. Data was only collected on the inhalation of cannabis and not on other forms of use. We reviewed the interview data for any information on cannabis use or THC for chronic pelvic pain or menstrual pain that may indicate reverse causation (i.e., that chronic pelvic pain associated with adenomyosis leads to cannabis use). Cannabis use or THC was not reported when participants were asked about treatments taken for pelvic pain when not menstruating. When asked what medications participants took if they experienced menstrual pain, only two cases and one hysterectomy control specified using cannabis or THC to manage this symptom. Interview data on whether a participant had smoked a total of 100 cigarettes or more in their lifetime was used to identify participants who had ever smoked cigarettes (no, yes). If the participant responded yes, additional questions inquired about ages when cigarettes were first smoked regularly and stopped (if stopped for one year or longer), and, when applicable, the ages when smoking cigarettes started again and stopped. For each smoking episode (highest number reported was 6 episodes), data was collected on the average number of cigarettes smoked per day. We estimated the lifetime cumulative cigarette smoking in pack-years, defined as the number of packs smoked per day times the number of years a person smoked. To estimate the number of packs per day, we divided the number of cigarettes smoked per day by 20. We multiplied this quantity by years in a smoking episode and summed the smoking pack-years across episodes. Study participants who reported never smoking cigarettes were assigned the value 0 pack-years. We categorized pack-years as 0, >0–5, >5–15, >15 pack-years, using categories reported in a large consortium detecting an association between cigarette smoking and another gynecologic condition (ovarian cancer) [ 26 ]. We used unconditional logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the associations between cannabis smoking, tobacco cigarette smoking, and adenomyosis risk. We compared adenomyosis cases to hysterectomy controls and population controls in separate analyses. All analyses were adjusted for age at reference date (20–39, 40–44, 45–49, 50–59 years), age at menarche (≤10, 11, 12, 13, ≥14 years), and education level (≤high school, some college, college graduate, post-graduate). For the analyses of cannabis smoking, we also adjusted for cigarette smoking pack-years (0, >0–5, >5–15, >15 pack-years); analyses of tobacco cigarette smoking were additionally adjusted for cannabis smoking joint-years (0, >0–0.004, >0.004–0.2, and >0.2 joint-years). Analyses comparing cases to population controls were further adjusted for reference year (continuous). To test the trend across categories of cannabis smoking joint-years and cigarette smoking pack-years, we created a continuous variable assigning the median category values to participants in each category and included this variable in the adjusted logistic regression model. We used available data in the analyses; with the exception of joint-years with 11% missing, <1% of exposure and covariate data were missing. We conducted several sensitivity analyses. First, we repeated the analyses additionally adjusting for gravidity (0, 1, 2, 3, 4, ≥5 pregnancies), a factor consistently associated with increased adenomyosis risk in prior studies [ 27 – 31 ]. We did not adjust for gravidity in the main analyses since the timing of pregnancies may not precede the initiation of smoking of cannabis or cigarettes and may be on the pathway between exposure and outcome. Second, we repeated the analyses considering only cumulative exposure to cannabis and tobacco cigarette smoking before age 35 as young adulthood may be a critical window for disease risk. Third, we repeated our analysis restricting population controls to those who responded that they would “probably” or “definitely” allow a hysterectomy to be performed if they developed severe menstrual bleeding, severe menstrual pain, or severe pelvic pain every month for six months or more and hysterectomy was recommended (n=168). By imposing this restriction, the population controls may more closely represent the underlying population that gave rise to cases. Fourth, we repeated the main analyses additionally adjusting for alcohol consumption, defined as ever consumption of ≥1 alcoholic beverage per month for ≥6 months (no, yes). Alcohol consumption was not adjusted for in the main analyses given its strong correlation with cannabis and tobacco cigarette smoking (among population controls, 96% of ever tobacco cigarette smokers had ever consumed alcohol and 97% of ever cannabis smokers had ever consumed alcohol). We conducted an exploratory analysis to evaluate individual use and co-use of cannabis and tobacco cigarette smoking (never smoked, only ever smoked cannabis, only ever smoked cigarettes, ever smoked both). The analyses were conducted using SAS Version 9.4 (SAS Institute Inc., Cary, NC).

In this first study examining cannabis smoking and risk of adenomyosis, we observed a null association. Our result conflicts with the hypothesis that smoking cannabis may decrease adenomyosis risk by lowering the production of estrogen, a hormone central to the development of adenomyosis [ 32 ]. Although the etiology of adenomyosis remains enigmatic, a unifying mechanism postulates that injury at the endometrial-myometrial junction leads to a series of events involving inflammation and local estrogen production that perpetuate the injury and contribute to endometrial cell migration into the myometrium and disease establishment [ 33 , 34 ]. A prior study of rhesus monkeys that evaluated the daily administration of THC, the main psychoactive component of cannabis, on the menstrual cycle observed the lack of expected rise in estrogen during the follicular phase [ 12 ]. It has been proposed that THC may suppress the release of hormones from the hypothalamus which stimulate ovarian estradiol production [ 10 ]. However, consistent with our results, an in vivo study found that THC does not compete for estrogen receptors in rhesus monkey or human uteri [ 11 ]. Additionally, a small study of women (n=16) which measured serum estradiol levels two hours before cannabis administration and up to three hours after smoking a cigarette containing 1 gram of marijuana observed no changes in estradiol concentrations [ 35 ]. Our data suggested an association between tobacco cigarette smoking and increased adenomyosis risk. The observed positive association also conflicts with our hypothesis that tobacco cigarette smoking decreases the risk of adenomyosis based on data suggesting the anti-estrogen effects of this exposure. Cigarette smoke and tobacco leaves contain aromatase inhibitors that suppress aromatase enzymes, resulting in a decrease in estrogen production [ 36 ]. Smoking cigarettes can also result in the C2-hydroxylation of estradiol, producing irreversible estradiol products with only minimal estrogenic activity that are cleared from circulation [ 37 ]. An in vitro study suggested that components in tobacco cigarettes can bind to estrogen receptors, thus not allowing estradiol to exhibit its effect [ 38 ]. In support of the positive association we observed, the risk of adenomyosis could be increased with inhalation of any one of the 4800 contaminants in tobacco cigarette smoke [ 39 ]. One such contaminant present in tobacco cigarettes, the toxic metal cadmium, accumulates in the body, and blood cadmium levels have been observed in previous studies to be related to an increase in estrogen levels [ 40 ]. Furthermore, the observed increased adenomyosis risk with tobacco cigarette smoking persisted in our sensitivity analyses investigating potential sources of bias. The association between tobacco cigarette smoking and increased adenomyosis risk was observed in our analyses comparing cases to hysterectomy controls as well as to population controls. Prior studies conducted among hysterectomy patients have reported mixed results, with some studies reporting inverse, positive and no associations [ 17 – 21 , 31 , 41 ]. Given that those studies had similar distributions of demographic characteristics, such as menarche age and BMI, the discrepancy of results across studies may be due to bias introduced by the selection of hysterectomy controls in a manner related to the exposure or by the approach used to ascertain smoking data. Most prior studies only evaluated tobacco cigarette smoking status at the time of hysterectomy or adenomyosis diagnosis [ 16 , 19 , 21 ]. Another study that employed two control groups, similar to our study, observed no association between cigarette smoking and adenomyosis risk [ 31 ]. Although that study was similarly conducted among Group Health enrollees with a similar distribution of demographic characteristics, cigarette smoking information was collected by medical chart abstraction that is susceptible to exposure misclassification, limiting the ability of that study to detect an association [ 31 , 42 ]. The present study benefitted from the extensive data collected on history of cannabis and tobacco cigarette smoking. This allowed us to evaluate several aspects of exposure, including ever use, cumulative lifetime exposure, and cumulative exposure in early adulthood. Furthermore, we were able to evaluate co-use of both tobacco cigarettes and cannabis and each exposure individually in an exploratory analysis. Our results suggest that tobacco cigarette smoking contributes to increased adenomyosis risk for those with a history of ever smoking of both tobacco cigarettes and cannabis. Another strength of our study was the novel study design employing two control groups. The selection of controls can be challenging when cases of adenomyosis are ascertained using the gold standard for diagnosis, histopathologic examination of the uterus after hysterectomy [ 7 ]. The comparison group of patients undergoing hysterectomy allows for the confirmation of the absence of adenomyosis and are a group similar to cases on factors related to willingness to allow hysterectomy. However, hysterectomy controls are not sampled from the underlying population that gave rise to cases; sampling from the base population is critical to ensuring that non-cases are selected independent of exposure, a key principle for the design of a valid case-control study [ 8 , 43 , 44 ]. To mitigate the concern regarding the use of hysterectomy controls, we also employed population controls that were selected from the same health plan population as the cases. Although this sampling approach is independent of exposure, confounding is still possible due to factors associated with willingness to undergo hysterectomy among cases. To evaluate this source of confounding, we were able to conduct a sensitivity analysis restricting population controls to those who would “probably” or “definitely” allow a hysterectomy to be performed if warranted, and we observed associations similar to the main results. Our study had several limitations. First, because collection of exposure data relied on participant recall, smoking history could differ by case status. However, smoking history was comparably ascertained between cases and controls before a reference date and recall was aided by use of the calendar method. Second, the case-control study was conducted among individuals, most of whom entered adulthood in the late 1960s, 1970s, and early 1980s. Thus, the characterization of cannabis exposure via inhalation in this study population may not be generalizable to current practices. Available data suggest that the potency of cannabis has increased over the past couple of decades [ 45 ]. Moreover, THC is now available for oral consumption in a range of edible products, and in vaporizers, which results in the inhalation of cannabis in vapor form, rather than smoke [ 46 ]. Third, although 60–62% of participants reported ever smoking cannabis, underreporting of cannabis smoking is possible. However, the legal and medical landscape of cannabis use has substantially changed over the past several decades with both recreational and medical use currently being more widely accepted. A longitudinal study conducted in Washington State reported that approval of cannabis use in adults had increased after medical marijuana was legalized in the state in 1998 [ 47 ]. Fourth, due to cases being identified from a large, integrated healthcare system as part of the delivery of care, a common surgical and pathology protocol for the diagnosis of adenomyosis by surgeons and pathologists was not implemented. Fifth, we were not able to confirm the absence of adenomyosis among population controls. Although no data exist on the prevalence of undiagnosed adenomyosis, an estimation is possible. In the present study, 20% (66 of 323) of population controls reported experiencing pelvic pain when not menstruating. In a study of women with a medical indication for transvaginal ultrasonography, including pelvic pain, the prevalence of transvaginal ultrasound-detected adenomyosis was 21% [ 29 ]. It is likely, therefore, that less than 5% of a population-based control group with this symptom would have undiagnosed adenomyosis. Lastly, it is possible that cannabis is used to manage chronic pelvic pain, a major symptom of adenomyosis, as evidence for this practice is emerging [ 48 ]. The concern for reverse causation is minimized in the present study as only a few participants reported using cannabis to manage pelvic pain and we observed a null association with adenomyosis risk.

Our data from a novel case-control study of health plan enrollees suggested that tobacco cigarette smoking, but not cannabis smoking, was associated with an increased risk of adenomyosis. Given that adenomyosis can confer substantial morbidity, cigarette smoking remains common, and use of recreational and medical cannabis is becoming more widely accepted, further research is warranted to replicate our results.

Adenomyosis is a non-cancerous gynecologic condition characterized by the presence of ectopic endometrial glands and stroma within the myometrium [ 1 ]. Those with adenomyosis frequently report painful menses and heavy menstrual bleeding [ 2 ] and that these symptoms negatively affect quality of life [ 3 ]. This can be compounded by the definitive treatment of symptoms being hysterectomy and the medical costs incurred from the management of adenomyosis and associated chronic pelvic pain [ 4 – 6 ]. Despite its adverse impact, adenomyosis remains understudied. This is in part due to the challenge of selecting non-cases for a valid epidemiologic study when cases are identified by hysterectomy, the gold standard for adenomyosis diagnosis [ 7 ]. Most studies have been conducted using a comparison group of hysterectomy patients without adenomyosis. Although disease absence can be confirmed, the medical indication warranting hysterectomy may be related to exposure, resulting in selection bias and discrepant results across prior studies [ 8 ]. This can be addressed by using a population-based sampling approach of selecting individuals from the underlying population that gave rise to cases. However, this approach is not entirely free of bias as confounding may arise from differences in factors between cases and population controls related to the cases’ willingness to undergo hysterectomy. Given that estrogen is central to adenomyosis pathogenesis [ 9 ], exposures with estrogenic or anti-estrogenic properties may alter disease risk. One such exposure that has been shown to suppress hypothalamic release of gonadotropin-releasing hormone and decrease ovarian estrogen production is Δ 9 -tetrahydrocannabinol (THC) present in cannabis [ 10 – 13 ]. Despite the increase in medical and legal recreational cannabis use, no prior study to our knowledge has evaluated cannabis smoking in relation to adenomyosis [ 14 ]. Another common inhaled exposure, tobacco cigarette smoking, contains components such as nicotine, which has demonstrated anti-estrogenic effects in animal studies [ 15 ]. However, human observational studies investigating tobacco cigarette smoking and adenomyosis have reported discrepant results [ 16 – 21 ], likely due to the challenge in selecting non-cases. Thus, the purpose of the present analyses was to investigate the associations between cannabis smoking, tobacco cigarette smoking, and the risk of adenomyosis using data from a novel case-control study that employed both hysterectomy and population control groups. We hypothesize that both cannabis and tobacco smoking are associated with a decreased risk of adenomyosis.