Electronic Cigarettes and Fecundability

Results From a Prospective Preconception Cohort Study

Alyssa F. Harlow; Elizabeth E. Hatch; Amelia K. Wesselink; Kenneth J. Rothman; Lauren A. Wise


Am J Epidemiol. 2021;190(3):353-361. 

In This Article


Study Design and Sample

Pregnancy Study Online (PRESTO) is an ongoing Web-based preconception prospective cohort study of pregnancy planners, described previously.[22] Eligible participants include women aged 21–45 years who are residents of the United States or Canada, not using fertility treatments or contraception at study entry, in a stable relationship with a male partner, and are actively trying to get pregnant. At baseline, women report demographic information, medical history, and lifestyle factors, and they are invited to complete the National Cancer Institute's Dietary Health Questionnaire II, a Web-based food frequency questionnaire.[23] Women then complete bimonthly follow-up surveys for up to 12 months, reporting on pregnancy status and factors that might change over follow-up. PRESTO was approved by the Boston University Medical Campus Institutional Review Board. All participants provided informed consent.

Questions on e-cigarettes were added to the baseline questionnaire on June 22, 2017. From that date through January 2, 2020, 5,971 eligible women completed the baseline questionnaire. We excluded women who had last menstrual period (LMP) dates >6 months before baseline and those with missing or implausible LMP dates (n = 91), those with no prospective LMP dates during follow-up (n = 14), and those trying to conceive for >6 menstrual cycles at baseline (to limit the possibility of reverse causation from women changing their behaviors in response to difficulty conceiving; n = 1,280). The final study population included 4,586 women.

Exposure Assessment: E-cigarette use

E-cigarette use was measured on the baseline and follow-up questionnaires. At baseline, participants were asked, "Have you ever used e-cigarettes, e-hookahs, vaping pens, personal vaporizers, or any other battery-powered device that simulates smoking?" Those who reported "yes" were asked how many milliliters of liquid they currently vape per day, and whether the device they used contains nicotine. At each follow-up assessment, participants reported any e-cigarette use in the previous 4 weeks.

Covariate Assessment

At baseline, we ascertained information on potential risk factors for fecundability, including age,[24] education,[25,26] annual household income,[26] race/ethnicity,[25,26] body mass index,[27] caffeine intake,[28] alcohol- and marijuana-use frequency,[29,30] secondhand and in-utero cigarette smoke exposure,[9] sleep duration,[31] daily use of multivitamins and/or folic acid,[32] hours of work per week,[31] Perceived Stress Scale score,[33] physician-diagnosed depression/anxiety,[34] depression symptoms via the Major Depression Inventory,[34] intercourse frequency, and parity. Women additionally reported whether they were doing anything to improve chances of pregnancy (e.g., ovulation testing, basal body temperature) and last contraception method used. A measure of overall diet quality,[35] the Healthy Eating Index, was calculated based on the food frequency questionnaire data.

Combustible cigarette smoking status and history were ascertained from baseline and follow-up questionnaires. Participants were classified as current regular smokers (smoked ≥1 cigarette/day), current occasional smokers (smoked <1 cigarette/day), past smokers (smoked cigarettes regularly for ≥6 months in lifetime), or never smokers. We calculated cumulative pack-years of cigarette smoking by multiplying smoking duration by intensity (packs smoked per day). Never smokers (n = 3,432) and occasional smokers who were never regular smokers (n = 48) were assigned a pack-year value of zero.

Time to Pregnancy

On each follow-up questionnaire, participants reported their most recent LMP date and whether they conceived or had any pregnancy losses since the previous questionnaire. Among nonrespondents, we obtained pregnancy information via phone interviews, online baby announcements/baby registries, fertility tracking software data, and birth registry linkage in selected states (California, Florida, Massachusetts, Michigan, Ohio, Pennsylvania, and Texas).

Time to pregnancy was calculated using baseline and follow-up questionnaire data. At baseline, participants reported the number of menstrual cycles they had been attempting to conceive, their LMP date, and the length of their usual menstrual cycle. Women were considered to have irregular cycles if they reported not being able to "predict from one menstrual period to the next about when the next menstrual period would start." For women with irregular cycles, we estimated cycle length using LMP data ascertained during follow-up. Time to pregnancy was calculated in discrete cycles as follows: menstrual cycles of attempt time at baseline + ((LMP date from the most recent follow-up assessment − baseline questionnaire date)/usual menstrual cycle length) + 1.

Statistical Analysis

Participants contributed menstrual cycles of attempt time until pregnancy or censoring, whichever came first. Censoring events included loss to follow-up, no longer attempting to conceive, initiation of fertility treatment, or 12 cycles of attempt time. We compared age-standardized baseline characteristics across baseline e-cigarette use. We computed the proportion of women who conceived over follow-up using life-table methods to account for censoring. We analyzed observed cycles only and utilized an Anderson-Gill data structure to generate time-varying exposure and covariates and account for left truncation due to variation in pregnancy attempt times at study entry (ranging from 0–6 cycles).[36] We estimated fecundability ratios and 95% confidence intervals for the association between e-cigarette use and time to pregnancy by fitting proportional probabilities regression models.[14] Fecundability ratios represent the ratio of average per-cycle probability of conception in each e-cigarette category compared with never e-cigarette use.

We operationalized e-cigarette exposure as ever versus never use and as current, former, and never use. Current use was defined as having ever used an e-cigarette and currently vaping >0 mL of e-liquid/day, or currently using a device containing nicotine (e-cigarettes with no nicotine might still contain other potentially harmful chemicals and metals). In a sensitivity analysis, we excluded women who currently used a device not containing nicotine. Former users had ever used an e-cigarette but did not meet the current-use definition. We classified baseline current e-cigarette users by intensity of use, distinguishing between vaping <3 mL/day and ≥3 mL/day (the average daily dose in some prior studies).[37,38] We analyzed baseline-only and time-varying e-cigarette exposure.

To adjust for confounding, we included covariates in our model that were potential confounders (risk factors for impaired fecundability that are associated with e-cigarette use, excluding causal intermediates). Certain variables were associated with e-cigarette use in our data but were not associated with time to pregnancy in prior PRESTO studies (e.g., in-utero cigarette exposure,[9] caffeine intake,[39] marijuana use,[30] anxiety diagnosis[34]). To preserve precision and avoid model convergence problems, we did not adjust for these covariates. Final models adjusted for age, annual household income (in $: <50,000, 50,000–99,000, 100,000–149,000, ≥150,000), education (up to some high school, high-school diploma or equivalency, some college, college degree, graduate degree), baseline smoking status (current regular smoker, current occasional smoker, former smoker, never smoker), pack-years of cigarette smoking (intensity × duration), weekly alcohol intake (drinks/week: <1, 1–6, 7–13, ≥14), intercourse frequency (times/week: <1, 1, 2–3, ≥4), using methods to improve pregnancy chances (yes/no), body mass index, Major Depression Inventory score (continuous), multivitamin/prenatal supplement use (yes/no), Healthy Eating Index score (continuous), and parity. If parity is a marker of underlying fertility, adjusting for parity might induce bias.[40,41] We therefore examined models with and without parity adjustment.

In secondary analyses, we created mutually exclusive joint-exposure categories based on time-varying current use of both e-cigarettes and combustible cigarettes, classifying participants as exclusive e-cigarette users, exclusive cigarette smokers, dual users of e-cigarettes and cigarettes, and current nonusers of either product (referent). We then further expanded categories to differentiate former and never use, with women who never used e-cigarettes or cigarettes as the referent. We replicated our primary analysis among never cigarette smokers, because this analysis is less likely to be affected by residual confounding by cigarette smoking. In these analyses, we adjusted for the same covariates mentioned above, except for baseline cigarette smoking status.

In additional models, we stratified by age (<30 vs. ≥30 years at baseline) and pregnancy attempt time at study entry (<3 vs. ≥3 cycles) to assess potential for reverse causation (i.e., subfertile women might quit cigarette smoking and start using e-cigarettes). In an additional sensitivity analysis, we reclassified conceptions resulting in pregnancy loss as no conception to examine the association between e-cigarette use and time to viable pregnancy.

We used multiple imputation with 5 imputed data sets to impute missing data on exposure, outcome, and covariates. Women who completed no follow-up questionnaires (n = 1,159) were assigned 1 cycle of observation, with pregnancy status imputed at that cycle. No variable was missing information for more than 1% of participants, except for income (3.3%), hours worked per week (3.5%), and Healthy Eating Index score (43.2%). Analyses were performed using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina).