Structured Abstract
Objective:
To assess the effect of randomization to FertilityFriend.com (FF), a mobile computing fertility-tracking app, on fecundability.
Design:
Parallel non-blinded randomized controlled trial nested within Pregnancy Study Online (PRESTO), a North American preconception cohort.
Subjects:
Female participants aged 21–45 years attempting conception for ≤6 menstrual cycles at enrollment (2013–2019).
Intervention:
Randomization (1:1) of 5,532 participants to receive a premium FF subscription.
Main Outcome Measures:
Fecundability (per-cycle probability of conception). Participants completed bimonthly follow-up questionnaires until pregnancy or a censoring event, whichever came first. We first performed an intent-to-treat analysis of the effect of FF randomization on fecundability. In secondary analyses, we used a per-protocol approach that accounted for adherence in each trial arm. In both analyses, we used proportional probabilities regression models to estimate fecundability ratios (FR) and 95% confidence intervals (CI) comparing those randomized vs. not randomized, and applied inverse probability weights to account for loss-to-follow-up (intent-to-treat and per-protocol analyses) and adherence (per-protocol analyses only).
Results:
Using life-table methods, 64% of the 2,775 participants randomized to FF and 63% of the 2,767 participants not randomized to FF conceived during 12 cycles; these respective percentages were each 70% among those with 0–1 cycles of attempt time at enrollment. Of those randomized to FF, 72% were defined as adherent (68% of observed menstrual cycles). In intent-to-treat analyses, there was no appreciable association overall (FR=0.97, 95% CI: 0.90–1.04) or within strata of pregnancy attempt time at enrollment, age, education, or other characteristics. In per-protocol analyses, we observed little association overall (FR=1.06, 95% CI: 0.99–1.14), but weak-to-moderate positive associations among participants who had longer attempt times at enrollment (FR=1.15, 95% CI: 0.98–1.35 for 3–4 cycles; 1.14, 95% CI: 0.87–1.48 for 5–6 cycles), were aged <25 years (FR=1.29, 95% CI: 1.01–1.66), had ≤12 years of education (FR=1.32, 95% CI: 0.92–1.89), or were non-users of hormonal contraception within 3 months before enrollment (FR=1.10, 95% CI: 1.02–1.19).
Conclusion:
No appreciable associations were observed in intent-to-treat analyses. In secondary per-protocol analyses that accounted for adherence, randomization to FF was associated with slightly greater fecundability among selected subgroups of participants; however, these results are susceptible to unmeasured confounding.
Keywords: fecundability, fertility, time-to-pregnancy, preconception, randomized trial
Capsule:
No appreciable associations were observed in intent-to-treat analyses. In secondary per-protocol analyses that accounted for adherence, randomization to FertilityFriend.com was associated with slightly greater fecundability among selected subgroups of participants.
Introduction
In recent years, there has been a substantial increase in the use of mobile computing applications (“apps”) to track menstrual cycles and record fertility signs (1, 2). In 2020, an estimated 50 million individuals worldwide used menstrual-tracking apps (3). Many apps estimate the days when intercourse is most likely to lead to conception (i.e., fertile window) (4); however, the extent to which use of such apps influences fecundability—the per-cycle probability of conception—is unclear (5, 6). In an observational study, we reported previously that self-reported use of some fertility apps—particularly those that track cycle days, cervical fluid, cervix position, basal body temperature (BBT), and urine luteinizing hormone (LH)—was associated with 12–20% greater fecundability (7). These results are consistent with data from most (8–13) but not all (14) trials and cohort studies, including ours (12), which indicate that use of fertility indicators to identify the fertile window increases fecundability. They also agree with results from a 2020 randomized trial in which use of the ClearBlue E1C/LH test app to time intercourse within the fertile window increased the likelihood of conceiving within two menstrual cycles (test arm: 36.2% vs. control arm: 28.6%) (13). To date, however, no randomized trial has evaluated the effect of using a fertility app that tracks multiple fertility indicators on fecundability. Observational studies of this association are prone to selection bias and confounding because fertility app users may differ with respect to underlying fecundity and other socio-demographic characteristics (e.g., age, education, income, and fertility awareness).
In this study, we conducted a parallel non-blinded randomized controlled trial to assess the extent to which randomization of a free premium subscription to FertilityFriend.com (FF), a popular mobile computing app that permits users to track their menstrual cycles and fertility signs, was associated with fecundability.
Methods
Study design and population
This trial was nested within Pregnancy Study Online (PRESTO), an ongoing web-based preconception cohort study of pregnancy planners. Study methods have been described in detail previously (15). Eligible participants identify as female and are aged 21–45 years, reside in the United States or Canada, and are attempting to conceive without the use of fertility treatment at cohort entry. Participants complete a comprehensive baseline questionnaire with items on socio-demographics, behaviors, and reproductive and medical histories, and medication use. Participants complete follow-up questionnaires every 8 weeks for up to 12 months to ascertain pregnancy status and update any factors that may have changed over time.
Fertility Friend (FF)
Fertility Friend (FF) is a mobile app integrated with an online web-based platform for users to track their fertility awareness indicators. It provides a graphical summary of each menstrual cycle, with the ability for the user to enter daily observations for basal body temperature, urine LH and/or estrogen testing, cervical mucus observations, vaginal bleeding observations, and intercourse. The app database applies algorithms based on these features to identify the fertile window, focused on the most fecund (fertile) days, which precede and include the estimated day of ovulation (16). It identifies the days of the estimated fertile window for the user prospectively on its interface. It also allows the user to output graphical summaries of each cycle, or long-term summaries over many cycles (e.g., means, minimum, maximum of cycle lengths, day of ovulation, luteal lengths), as pdf documents, which the user can then give to a clinician for review.
Randomization scheme
PRESTO participants who enrolled between June 2013 through March 2019 were considered for inclusion in this trial. All participants who reported never use of FF at baseline and never had an FF account associated with their email address were eligible for randomization. Eligible participants were randomized immediately after enrollment (defined as completion of the baseline questionnaire) by an independent programmer who implemented a simple randomization scheme via computer algorithm using the Mersenne Twister random number generator (17). The procedure randomized 8,397 participants with 50% probability to receive a free premium FF subscription (N=4,162) or the standard study procedures only (N=4,235). All participants had the opportunity to contribute the full 12 months of follow-up after enrollment.
Immediately after randomization, investigators and participants were made aware of the group to which participants were randomized. Participants assigned to receive FF had to click on a link embedded within the invitation email to access the subscription. Other than the standard information found in the FF app, no additional information was provided to participants about the app, nor was there any stated requirement or additional incentive to use the app. In the app, participants could record day-specific data on menstrual flow (none, spotting, light, medium, heavy), BBT, cervical fluid consistency (dry, sticky, creamy, watery, egg white), sexual intercourse (no vs. yes; if yes, AM, PM, or both), secondary signs (e.g., cervical position), and results from testing devices (ovulation predictor kit for urine luteinizing hormone (OPK); progesterone test; ferning microscope; and pregnancy test). Using a secure password-protected server, study investigators downloaded de-identified FF app data from PRESTO participants who were randomized to receive the premium FF subscription.
The study protocol was approved by the Institutional Review Board at Boston University Medical Campus, and online informed consent was obtained from all participants. There was no separate consent form for the trial. In the main study’s consent form, all participants were informed that a random subset of never users of FF would be offered a premium subscription after enrollment. The trial was not registered with clinicaltrials.gov because it did not involve randomization of a medical treatment, supplement, or device.
Exclusions
From June 2013 through March 2019 (i.e., time period when FF was randomized), 10,405 participants completed the baseline questionnaire. Of these, 2,008 participants were not eligible for randomization because they were former or current users of FF. Of the remaining 8,397 participants, we randomized 4,162 to FF and 4,235 to the standard protocol. As part of our standard fecundability analyses, we excluded 120 participants with missing/implausible last menstrual period (LMP) data (defined as having baseline LMP >6 months before the baseline questionnaire completion date or any time after the baseline questionnaire completion date), and 1,158 participants who were pregnant at study entry as identified using dates of LMP and first positive pregnancy test (47% randomized and 53% not randomized to FF). We then excluded 1,577 participants who had been attempting pregnancy for >6 cycles at enrollment, to reduce the potential for reverse causation bias. We also anticipated that the effect among participants who had been attempting pregnancy for >6 cycles at enrollment might differ from the effect among those attempting for ≤6 cycles. Our results should thus be interpreted as relevant primarily to this latter group. The final analytic population comprised 5,542 participants: 2,775 randomized to FF and 2,767 not randomized to FF (Supplemental Figure 1).
Assessment of time-to-pregnancy
We estimated time-to-pregnancy (TTP) using data from the baseline and follow-up questionnaires. At baseline, participants reported their LMP date, usual cycle length (only for those with regular cycles), and the number of cycles they had attempted conception. On each follow-up questionnaire, participants reported their most recent LMP date, whether they had conceived since the previous questionnaire, and the method of pregnancy confirmation. In these analyses, we accepted confirmation of pregnancy via positive home pregnancy test, blood test from doctor’s office, or ultrasound. Among those with irregular cycles, defined as those who reported not being able to predict from one menstrual period to the next about when their next menstrual period would start (when not using hormonal contraception), we estimated cycle length based on date of LMP at baseline and prospectively-reported LMP dates during follow-up. We estimated intervening LMP dates between questionnaires by subtracting most recent cycle length or typical cycle length from the LMP reported on follow-up questionnaires and repeated this process until the calculated LMP was within 15 days of the LMP reported on the previous questionnaire. We then calculated TTP as follows: menstrual cycles of attempt time at study entry + total number of self-reported and calculated LMPs between baseline and end of observation.
Assessment of covariates
At baseline, we collected demographic and clinical information, including age, height, weight, relationship duration, marital status, race/ethnicity, income, education, hours of sleep per night, parity, gravidity, vitamin use, caffeine intake, smoking status, marijuana use, alcohol consumption, sugar-sweetened soda intake in the past month, total metabolic equivalents of task (METs) from vigorous and moderate exercise in the past week (18), depressive symptoms via the Major Depression Inventory (19), the 10-item Perceived Stress Scale (PSS) in the past month (20), average intercourse frequency in the past month, contraception history, and infertility. We calculated body mass index (BMI) as weight (kilograms) divided by height (meters) squared.
On the baseline and follow-up questionnaires, we asked participants whether they were currently doing anything to improve their chances of conception (e.g., recording BBT, monitoring cervical mucus, using OPKs) and if they used “any software program and/or web-based or phone “app” to record menstrual cycle data and/or fertility signs?” If they responded yes, they were asked to write the name of the program or app using a free text response.
Data analysis
Couples that did not conceive within 12 cycles of attempted conception were censored at 12 cycles, the time after which couples typically seek infertility treatment (21). Couples contributed menstrual cycles to the analysis from enrollment until reported pregnancy (49.4%) or a censoring event (initiation of fertility treatment: 7.5%; cessation of pregnancy attempts: 3.3%; loss to follow-up: 18.7%; or 12 cycles of attempt: 21.1%), whichever came first. Participants who were and were not lost-to-follow-up were similar with respect to baseline age, last method of contraception used, parity, sleep duration, and depressive symptoms, but differed according to randomization to FF or free home pregnancy tests, BMI, educational attainment, income, marital status, geographic region of residence, use of prenatal supplements, infertility history, pregnancy attempt time at enrollment, and male partner participation (Supplemental Table 1).
We used life-table methods to compute the percentage of couples that conceived during follow-up, after accounting for censoring (22). To account for variation in attempt times at study entry (range: 0–6 cycles) and to reduce bias due to left truncation, we analyzed observed cycles using the Andersen-Gill data structure (i.e., we outputted one observation per menstrual cycle at risk, excluding earlier cycles that were unobserved) (23). For example, if a participant enrolled in the study after having tried to conceive for 2 cycles and was followed for an additional 4 observed cycles until reported pregnancy, that participant would contribute a total of 4 observed cycles starting at cycle 3. We used proportional probabilities regression models (24) to estimate fecundability ratios (FRs) and 95% confidence intervals (CIs) for the association between trial arm and fecundability. The FR represents the ratio of fecundability in each group compared with the reference category. Our models included cycle-specific indicator variables for “menstrual cycle at risk” to account for the decline in fecundability in the study population over time (24).
We first performed an intent-to-treat analysis to compare fecundability across the trial arms. In these analyses, those who showed no evidence of adherence to their assigned intervention (e.g., randomized to FF but showed no evidence of using the app) were analyzed in their assigned group. We then performed a separate per-protocol analysis in which we sought to analyze the relationship between initiating use of FF and fecundability. Exposure was defined as adherence to the assigned protocol during or before the estimated follicular phase of a given menstrual cycle: [(date LMPn), (date LMPn+1 −14 days)]. We defined adherence for those not randomized to FF as the continued absence of any reported FF use on a follow-up questionnaire. We defined adherence for those randomized to FF as initiation of FF use, beginning on the earlier of two dates:
Date participant first clicked on the invitation link to access the FF app (68% of all menstrual cycles) or the date of earliest reported data in FF app (40% of all cycles), whichever occurred later (because those who logged into the app could have entered data retrospectively in calendar time).
Date of first reported FF use on a follow-up questionnaire (13% of all cycles).
Once a participant moved from “non-adherent” to “adherent” in the FF arm or “adherent” to “non-adherent” in the control arm (i.e., user of FF in both scenarios), they remained coded in their latter status for all subsequent observed menstrual cycles. The per-protocol analyses discarded observed menstrual cycles for which participants were not adherent to their assigned protocol and applied weights to account for non-adherence. In each cycle of follow-up, we created weights for the probability of adherence based on potential determinants of FF use (Supplemental Table 2). We conducted sensitivity analyses in which we 1) redefined adherence as “not using any fertility app” in the control arm and 2) redefined adherence as “use of any fertility app” in the intervention arm and as “not using any fertility app” in the control arm.
In both intention-to-treat and per-protocol analyses, we derived and applied weights to account for bias due to differential loss-to-follow-up. Briefly, we used inverse probability weighting to create a pseudo-population of participants who, had they not been lost to follow-up, would have contributed more complete follow-up to the study. Using data from all participants enrolled at the start of follow-up, we developed logistic regression models for the probability of continuing in the study at each follow-up cycle, conditional on remaining uncensored at the previous follow-up. The model contained a set of variables, some of which were time-varying, hypothesized to predict loss to follow-up (Supplemental Table 1). We fit separate logistic regression models that included only time-invariant variables as independent variables. We computed stabilized weights by dividing the predicted probability of loss to follow-up from the second model (containing time-invariant variables only) by the predicted probability of loss to follow-up from the first model (time-varying and time-invariant variables) and multiplying by stabilized weights from previous cycles. The resulting weights were inversely proportional to the probability of remaining under study at each cycle. We then applied these weights to our analyses at each time point (i.e., follow-up cycle). We used a similar approach to account for non-adherence (Supplemental Table 2). In the per-protocol analyses only, we multiplied the weights for adherence with the weights for loss-to-follow-up.
In both analyses, we ran additional models in which we discarded all pregnancies and person-time that occurred in each participant’s first contributed menstrual cycle of observation. We reasoned that it would take at least one cycle for those randomized to FF to use the software in a way that could have meaningfully influenced fecundability. Finally, we also stratified by factors that could potentially modify the association between FF use and fecundability: attempt time at entry, age, education, parity, infertility history, recency of hormonal contraceptive use, and non-use of fertility app at baseline.
Missingness for covariates ranged from <0.1% (prior pregnancy, history of subfertility, caffeine use, and history of anxiety) to 3.4% for household income. We used the fully conditional specification (FCS) method to multiply impute missing data for exposures, covariates, and pregnancy status (25). To reduce selection bias from differential loss to follow-up, we assigned one cycle of follow-up for the 13% of participants with no data from follow-up questionnaires (N=726) and then imputed their pregnancy status (yes vs. no) via multiple imputation. The imputation models included >100 covariates. To ensure validity without compromising computing efficiency (26), we created twenty imputed datasets using SAS PROC MI and then combined coefficient and standard error estimates from the datasets using SAS PROC MIANALYZE (27). We used linear regression to impute continuous variables, the discriminant function method to impute nominal categorical variables, and logistic regression to impute dichotomous and ordinal categorical variables. We examined the trace plots for continuous variables and compared the frequency distribution of dichotomous and categorical variables before and after imputation to ensure the quality of the imputation. Analyses were performed using SAS software version 9.4 (27).
Results
Table 1 presents demographic, reproductive, and behavioral characteristics of the participants at baseline, stratified by randomization status. The median age of participants was 30 years (interquartile range: 27–33 years). The majority of participants were married (89%) and college-educated (70%). Fewer than 10% reported a history of infertility; 32% were parous; 76% reported taking folic acid, multivitamins, or prenatal supplements; 32% had a BMI ≥30 kg/m2; and nearly 40% had used a hormonal method as their most recent form of contraception.
Table 1.
Baseline characteristics of participants by randomization status, 2013–2019, PRESTO
| Randomized to FF | ||
|---|---|---|
|
| ||
| Characteristica | Yes (N=2,775) | No (N=2,767) |
|
| ||
| Age at baseline, years (mean) | 29.8 | 29.9 |
| Race/ethnicity (%) | ||
| Non-Hispanic, White | 82.6 | 82.6 |
| Hispanic | 6.5 | 7.6 |
| Non-Hispanic mixed/other race | 5.0 | 4.5 |
| Non-Hispanic Black | 3.3 | 3.6 |
| Non-Hispanic Asian | 2.6 | 1.8 |
| Education (%) | ||
| ≤High school | 6.3 | 5.9 |
| Some college | 24.0 | 24.8 |
| College degree | 34.3 | 33.0 |
| Graduate school | 35.3 | 36.4 |
| Annual household income (%) | ||
| <$50,000 | 23.1 | 23.5 |
| $50,000–$99,999 | 38.0 | 39.0 |
| $100,000–$149,999 | 23.6 | 23.8 |
| ≥$150,000 | 15.4 | 13.7 |
| Body mass index, kg/m2 (%) | ||
| <25 | 43.2 | 42.5 |
| 25–29 | 24.2 | 25.1 |
| 30–34 | 13.9 | 13.2 |
| ≥35 | 18.7 | 19.3 |
| Last method of contraception - hormonal (%) | 39.6 | 39.7 |
| History of infertility (%) | 9.1 | 9.7 |
| History of miscarriage (%) | 24.4 | 24.3 |
| Sleep duration, hours (%) | ||
| <7 | 26.9 | 27.6 |
| 7–8 | 66.5 | 66.0 |
| ≥9 | 6.6 | 6.4 |
| Major Depression Inventory score (%) | ||
| <20 | 82.9 | 82.1 |
| 20–24 | 7.1 | 7.7 |
| 25–29 | 4.8 | 4.5 |
| ≥30 | 5.3 | 5.7 |
| History of sexually transmitted infections (%) | 14.4 | 16.4 |
| History of polycystic ovary syndrome (%) | 9.1 | 9.2 |
| History of uterine leiomyomata (%) | 2.2 | 2.2 |
| History of endometriosis (%) | 3.8 | 2.5 |
| Parous (%) | 32.0 | 31.4 |
| Not married (%) | 10.4 | 10.9 |
| Geographic region of residence (%) | ||
| U.S. Northeast | 21.9 | 22.8 |
| U.S. South | 23.6 | 25.1 |
| U.S. Midwest | 21.2 | 21.0 |
| U.S. West | 16.0 | 15.2 |
| Canada | 17.3 | 15.8 |
| Unemployed (%) | 3.9 | 4.6 |
| Hours/week of work (%) | ||
| <30 | 25.1 | 27.2 |
| 30–49 | 64.9 | 63.8 |
| ≥50 | 10.0 | 9.0 |
| Had Pap test in past 3 years (%) | 91.8 | 92.8 |
| History of abnormal Pap test (%) | 33.0 | 32.6 |
| Current smoker (%) | 11.9 | 12.3 |
| Daily use of folic acid/multivitamins (%) | 75.7 | 76.6 |
| Intercourse frequency, times/week (mean) | 2.3 | 2.3 |
| Doing something to improve chances (%) | 72.7 | 72.9 |
| Sugar-sweetened soda intake, drinks/wk (%) | ||
| ≤1 | 74.1 | 73.8 |
| 2–6 | 18.5 | 18.9 |
| ≥7 | 7.4 | 7.3 |
| Alcohol intake, drinks/week (%) | ||
| 0 | 31.2 | 30.1 |
| 1–6 | 54.3 | 55.1 |
| ≥7 | 14.5 | 14.8 |
| Perceived stress scale score (%) | ||
| <10 | 12.0 | 11.3 |
| 10–14 | 27.6 | 28.8 |
| 15–19 | 30.8 | 31.5 |
| 20–24 | 20.3 | 18.3 |
| ≥25 | 9.3 | 10.0 |
| Physical activity, hours/week (%) | ||
| <5 | 13.4 | 14.7 |
| 5–9 | 20.4 | 20.9 |
| 10–14 | 33.6 | 32.8 |
| ≥15 | 32.6 | 31.6 |
| Current marijuana use (%) | 13.7 | 14.0 |
| Use other fertility tracking app (%) | 68.3 | 69.0 |
| Randomized to receive home pregnancy tests (%) | 12.1 | 12.4 |
| Completed dietary questionnaire (%) | 64.2 | 61.3 |
| Invited male partner (%) | 52.3 | 54.0 |
| Attempt time at study entry, cycles (mean) | 2.1 | 2.1 |
All characteristics except for age are standardized to baseline age of cohort.
Of the 2,775 participants randomized to FF, 1,990 (72%) of participants and 9,567 (68%) of cycles were defined as adherent. Of the 2,767 participants not randomized to FF, 2,766 (99.9%) of participants and 13,199 (97%) of cycles were defined as adherent. Figure 1 shows descriptive data among participants assigned to the FF arm and who entered data into the FF app at any point during follow-up [n=1,102 (40%)]. Median time to first FF app use was 0 days (interquartile range: 0–4 days). Among participants who entered the start and end date of at least one cycle into FF, on average 52% of the days in a cycle were logged (interquartile range: 23%–93%). Among the 1,102 (40%) participants who entered some data into the FF app during follow-up (5,956 total cycles), 1074 participants (97% of FF users; 4,021 cycles; 68% of total cycles) entered data on menstrual bleeding dates; 897 participants entered data on intercourse (81% of FF users; 3,147 cycles; 53% of total cycles); 793 participants entered data on cervical fluid (72% of FF users; 2409 cycles; 40% of total cycles); 573 participants entered data on BBT (52% of FF users; 1839 cycles; 31% of total cycles); 490 participants entered data on OPK (44% of FF users; 1459 cycles; 25% of total cycles); and 184 participants entered data on cervical position/openness/texture (17% of FF users; 412 cycles; 7% of total cycles).
Figure 1.
Use of app features among those randomized to FF who entered data into app, 2013–2020, PRESTO
Abbreviations: FF=FertilityFriend.com, OPK=ovulation predictor kit (urine LH), BBT=basal body temperature, Cervix Other=position/openness/texture.
Notes: Dark shaded bar displays percentage of participants randomized to FF who entered data into FF app at any point during follow-up (unit of analysis: participant). Light shaded bar displays the percentage of menstrual cycles during which participants randomized to FF entered data into FF app (unit of analysis: menstrual cycle).
Overall, 81.3% of participants in the analytic cohort completed follow-up (i.e., conceived or reached another study endpoint). Using life-table methods, we identified pregnancy during 12 cycles of follow-up for 64% of the 2,775 participants randomized to FF and 63% of the 2,767 participants not randomized to FF. When restricting to participants with 0–1 cycles of attempt time at enrollment, those respective cumulative pregnancy rates were 70% in each arm.
Table 2 reports FRs from the intent-to-treat analysis, stratified by potential effect measure modifiers. The overall FR comparing those who were and were not randomized to FF was 0.97 (95% CI: 0.90–1.04); results were identical after removing the first cycle of observation contributed by each participant. While there was a slightly elevated but imprecise FR among participants aged <25 years (FR=1.18, 95% CI: 0.91–1.53), we did not observe any consistent patterns of association when we stratified results by attempt time at cohort entry, parity, history of infertility, cycle regularity, or education. Results were slightly inverse among non-users of a different fertility app at baseline (FR=0.89, 95% CI: 0.78–1.01). There was little evidence of an association among non-users of hormonal contraception within 3 months before study enrollment.
Table 2.
Intent-to-treat analysis of randomization to FertilityFriend.com (FF) and fecundability, 2013–2020, PRESTO
| Pregnancies | Cycles | FR (95% CI)a | |
|---|---|---|---|
|
| |||
| All cycles | |||
| Randomized to FF | 1,373 | 14,033 | 0.97 (0.90–1.04) |
| Not randomized to FF | 1,357 | 13,661 | Reference |
| Omit first contributed cycle | |||
| Randomized to FF | 1,373 | 11,258 | 0.97 (0.90–1.04) |
| Not randomized to FF | 1,357 | 10,894 | Reference |
| Attempt time at entry <3 cycles | |||
| Randomized to FF | 1,038 | 9,580 | 0.95 (0.88–1.03) |
| Not randomized to FF | 1,040 | 9,219 | Reference |
| Attempt time at entry 3–4 cycles | |||
| Randomized to FF | 242 | 3,022 | 1.00 (0.84–1.18) |
| Not randomized to FF | 231 | 2,967 | Reference |
| Attempt time at entry 5–6 cycles | |||
| Randomized to FF | 93 | 1,431 | 1.09 (0.83–1.43) |
| Not randomized to FF | 86 | 1,475 | Reference |
| Age <25 years | |||
| Randomized to FF | 122 | 1,331 | 1.18 (0.91–1.53) |
| Not randomized to FF | 92 | 1,188 | Reference |
| Age 25–34 years | |||
| Randomized to FF | 1,105 | 10,597 | 0.95 (0.88–1.02) |
| Not randomized to FF | 1,118 | 10,353 | Reference |
| Age ≥35 years | |||
| Randomized to FF | 146 | 2,105 | 0.99 (0.79–1.22) |
| Not randomized to FF | 147 | 2,120 | Reference |
| Parous | |||
| Randomized to FF | 434 | 4,245 | 0.98 (0.86–1.11) |
| Not randomized to FF | 411 | 4,026 | Reference |
| Nulliparous | |||
| Randomized to FF | 939 | 9,788 | 0.96 (0.89–1.05) |
| Not randomized to FF | 946 | 9,635 | Reference |
| No history of infertility | |||
| Randomized to FF | 1,314 | 12,784 | 0.97 (0.91–1.05) |
| Not randomized to FF | 1,291 | 12,443 | Reference |
| History of infertility | |||
| Randomized to FF | 59 | 1,249 | 0.87 (0.62–1.21) |
| Not randomized to FF | 66 | 1,218 | Reference |
| Regular menstrual cycles | |||
| Randomized to FF | 1,181 | 11,477 | 0.97 (0.90–1.05) |
| Not randomized to FF | 1,171 | 11,235 | Reference |
| Irregular menstrual cycles | |||
| Randomized to FF | 192 | 2,556 | 0.97 (0.80–1.18) |
| Not randomized to FF | 186 | 2,426 | Reference |
| Education ≤12 years | |||
| Randomized to FF | 54 | 773 | 0.99 (0.67–1.45) |
| Not randomized to FF | 42 | 627 | Reference |
| Education 13–15 years | |||
| Randomized to FF | 257 | 3,314 | 0.95 (0.81–1.12) |
| Not randomized to FF | 259 | 3,257 | Reference |
| Education ≥16 years | |||
| Randomized to FF | 1,062 | 9,946 | 0.98 (0.90–1.06) |
| Not randomized to FF | 1,056 | 9,777 | Reference |
| Non-use of fertility app at baseline | |||
| Randomized to FF | 413 | 4,800 | 0.89 (0.78–1.01) |
| Not randomized to FF | 424 | 4,413 | Reference |
| Non-user of hormonal contraception within 3 months before baseline | |||
| Randomized to FF | 1,054 | 11,117 | 0.99 (0.91–1.07) |
| Not randomized to FF | 1,028 | 10,901 | Reference |
All models included cycle-specific indicator variables for “menstrual cycle at risk” to adjust for the decline in fecundability in the study population over time.
Table 3 presents FRs from the per-protocol analysis, which accounts for adherence, stratified by potential effect measure modifiers. The overall FR comparing those who did and did not initiate FF use according to their randomization assignment was 1.06 (95% CI: 0.99–1.14); results were null after removing the first cycle of adherence contributed by each participant (FR=0.99, 95% CI: 0.93–1.07). There were weak to moderate positive associations among participants with longer attempt times at enrollment (FR=1.15, 95% CI: 0.98–1.35 for 3–4 cycles; 1.14, 95% CI: 0.87–1.48 for 5–6 cycles), were aged <25 years (FR=1.29, 95% CI: 1.01–1.66), had ≤12 years of education (FR=1.32, 95% CI: 0.92–1.89), or were non-users of hormonal contraception within 3 months before enrollment (FR=1.10, 95% CI: 1.02–1.19). In addition, when we stratified results by attempt cycle, we did not observe any time trends in the estimated per-cycle FR (data not shown). Use of inverse probability weights to account for loss to follow-up did not make a large difference in the results of either the intent-to-treat or the per-protocol analyses (Supplemental Tables 3 and 4). Finally, when we re-ran the per-protocol analyses in which we 1) redefined adherence as “not using any fertility app” in the control arm or 2) redefined adherence as “use of any fertility app” in the intervention arm and as “not using any fertility app in the control arm, effect estimates were generally stronger than or similar to, respectively, the original per-protocol results (Supplemental Table 5).
Table 3.
Per-protocol analysis of randomization to and actual use of FertilityFriend.com (FF) and fecundability, 2013–2020, PRESTO
| Pregnancies | Cycles | FR (95% CI)a | |
|---|---|---|---|
|
| |||
| All cycles | |||
| Randomized to FF | 1,059 | 9,567 | 1.06 (0.99–1.14) |
| Not randomized to FF | 1,301 | 13,199 | Reference |
| Omits first contributed cycle | |||
| Randomized to FF | 895 | 7,571 | 0.99 (0.92–1.06) |
| Not randomized to FF | 1,301 | 10,433 | Reference |
| Attempt time at entry <3 cycles | |||
| Randomized to FF | 800 | 6,664 | 1.00 (0.92–1.08) |
| Not randomized to FF | 997 | 8,864 | Reference |
| Attempt time at entry 3–4 cycles | |||
| Randomized to FF | 190 | 1,952 | 1.15 (0.98–1.35) |
| Not randomized to FF | 220 | 2,881 | Reference |
| Attempt time at entry 5–6 cycles | |||
| Randomized to FF | 69 | 951 | 1.14 (0.87–1.48) |
| Not randomized to FF | 84 | 1,454 | Reference |
| Age <25 years | |||
| Randomized to FF | 95 | 905 | 1.29 (1.01–1.66) |
| Not randomized to FF | 92 | 1,182 | Reference |
| Age 25–34 years | |||
| Randomized to FF | 858 | 7,326 | 1.04 (0.96–1.12) |
| Not randomized to FF | 1,068 | 9,957 | Reference |
| Age ≥35 years | |||
| Randomized to FF | 106 | 1,336 | 1.10 (0.89–1.36) |
| Not randomized to FF | 141 | 2,060 | Reference |
| Parous | |||
| Randomized to FF | 316 | 2,731 | 1.06 (0.94–1.20) |
| Not randomized to FF | 398 | 3,941 | Reference |
| Nulliparous | |||
| Randomized to FF | 743 | 6,836 | 1.06 (0.98–1.15) |
| Not randomized to FF | 903 | 9,258 | Reference |
| No history of infertility | |||
| Randomized to FF | 1,012 | 8,749 | 1.07 (0.99–1.14) |
| Not randomized to FF | 1,237 | 11,999 | Reference |
| History of infertility | |||
| Randomized to FF | 47 | 818 | 0.99 (0.72–1.37) |
| Not randomized to FF | 64 | 1,200 | Reference |
| Regular menstrual cycles | |||
| Randomized to FF | 905 | 7,718 | 1.07 (0.99–1.15) |
| Not randomized to FF | 1,121 | 10,824 | Reference |
| Irregular menstrual cycles | |||
| Randomized to FF | 154 | 1,849 | 1.06 (0.88–1.28) |
| Not randomized to FF | 180 | 2,375 | Reference |
| Education ≤12 years | |||
| Randomized to FF | 42 | 437 | 1.32 (0.92–1.89) |
| Not randomized to FF | 41 | 626 | Reference |
| Education 13–15 years | |||
| Randomized to FF | 190 | 2,224 | 1.01 (0.86–1.19) |
| Not randomized to FF | 252 | 3,205 | Reference |
| Education ≥16 years | |||
| Randomized to FF | 827 | 6,906 | 1.07 (0.99–1.15) |
| Not randomized to FF | 1,008 | 9,368 | Reference |
| Non-use of fertility app at baseline | |||
| Randomized to FF | 313 | 3,259 | 0.94 (0.83–1.06) |
| Not randomized to FF | 411 | 4,282 | Reference |
| Non-user of hormonal contraception within 3 months before baseline | |||
| Randomized to FF | 809 | 7,467 | 1.10 (1.02–1.19) |
| Not randomized to FF | 984 | 10,517 | Reference |
All models included cycle-specific indicator variables for “menstrual cycle at risk” to adjust for the decline in fecundability in the study population over time.
Discussion
In this randomized controlled trial conducted within a North American prospective cohort of pregnancy planners, we did not find any appreciable association between randomization to FertilityFriend.com (FF), a popular fertility tracking app, and fecundability in intent-to-treat analyses. Low adherence in the intervention arm likely contributed to these null findings, thereby limiting our ability to rely on the intent-to-treat analyses for causal inference. The most reasonable explanation for low adherence in the intervention arm was that 69% of participants were already using other fertility apps at entry into the study and participants were likely satisfied with their current apps. Participants may have also already collected several cycles’ worth of data and did not want to transfer their data to a new app, even if FF was perceived to be of better quality.
In secondary per-protocol analyses that accounted for adherence, we found evidence for a weak positive association between initiating use of FF and fecundability. In the latter analyses, we observed slightly stronger associations between FF use and fecundability among participants who were younger, had lower attained levels of education, had longer pregnancy attempt times at study enrollment, and who were non-users of hormonal contraception within the 3 months before study enrollment.
The analysis of randomized exposure is an improvement over previous studies of observational data because users of fertility-tracking apps may be different from non-users in ways that are difficult to measure. For example, users of fertility-tracking apps may be more likely to be older and more educated than non-users, and have greater awareness of their fertility signs. On the other hand, users of fertility-tracking apps may also have a higher prevalence of fertility problems that resulted in app use. We reasoned that we would still have potential for bias in our analyses if non-users of apps at baseline who experienced difficulties conceiving during follow-up were more likely to initiate app use over time, potentially biasing FRs downward. Accounting for adherence in the per-protocol analyses helped address this concern. In addition, when we stratified the per-protocol estimates by attempt cycle, we did not observe any time trends in the estimated per-cycle FR.
Our definition of adherence was limited in that, for some participants, we did not know the precise date during which data were first entered into the FF app. We relied on data from date of first click of the invitation link to the app, and we used that or any later date of actual FF data entry or reported first use on a follow-up questionnaire as the date of adherence. Data on app use were self-reported on all questionnaires, in addition to being downloaded directly from the app itself for participants randomized to receive FF. Thus, for the FF users who were not randomized to the FF intervention arm, we did not have any comparable information about the features of FF use. As a result, our trial could not assess the impact of perfect, consistent, or continuous use of FF on fecundability and had to instead estimate the effect of initiation of use only. Sensitivity analyses in which we varied the definition of adherence in the control arm (i.e., no use of any fertility app) or both the intervention arm (i.e., use of any fertility app) and control arm (i.e., no use of any fertility app) showed generally similar or stronger results. Some studies have documented inconsistent recording of data among >50% of users of similar apps (16, 28). Sporadic app use would be expected to attenuate results if more consistent use improves fecundability. While our protocol included allocation concealment, which was achieved by having an independent computer programmer code and implement the randomization scheme, it did not include participant or investigator “blinding” in randomizing the intervention. Stratification of data by multiple factors may have introduced potential for chance findings. Finally, the prospective assessment of FF use was based on questionnaires completed every 2 months, which would not capture more frequent changes; this could have resulted in non-differential misclassification of FF use obtained outside of the app itself.
In the per-protocol analyses, we cannot rule out potential for unmeasured or residual confounding, whereby use of FF reflects behaviors that have not been fully accounted for by measured covariates. We did, however, account for many covariates known to be associated with fecundability, including age, income, education, parity, BMI, infertility history, and use of prenatal supplements. The intent-to-treat analyses, in which investigators randomized FF (in a 1:1 ratio) to participants, are robust to measured and unmeasured confounders at baseline. However, the intent-to-treat effect is expected to be closer to the null than the true effect of app use in this trial because adherence to FF use was low (29). Moreover, many participants were already using other fertility-tracking apps (69%), as the prevalence of app use among pregnancy planners tends to be higher than the general population. Despite the limitations of the per-protocol analysis, we believe the per-protocol effect asks a clinically relevant question: what is the impact on fecundability of initiating use of the FF app, had everyone adhered to their treatment assignment? Previous research has demonstrated that per-protocol effect estimates are of greatest interest to individuals for making treatment decisions when they intend to adhere (30). Since it is reasonable to expect that individuals attempting to conceive would know a priori whether they expect to initiate the use of a fertility app, the effect of initiation will likely be useful information. In comparison, while other usage questions, such as the impact of consistent or prolonged use, may also be of clinical interest, it may be less likely that an individual would know a priori whether they are likely to use a new app consistently.
PRESTO enrolled participants who were trying to conceive spontaneously. All participants completed self-administered questionnaires via the Internet. The study population overrepresents non-Hispanic White participants (>80%) with higher educational attainment and household income than the general population. For these reasons, the results of this study may not extend to the general population.
Previous observational studies of use of fertility-tracking apps and fertility indicators (7–9, 11, 12, 31), including our own study (7, 12), have found positive associations between the use of fertility indicators and fecundability. While a 2020 randomized trial in which use of an ovulation testing app to time intercourse within the fertile window increased the likelihood of conceiving within two menstrual cycles (13), other randomized trials of use of fertility indicators have produced inconsistent results, some positive and some null (10, 14, 32), likely reflecting difficulties in conducting trials among pregnancy planners. Our overall effect estimates for the association between FF use and fecundability in this trial are weaker than those reported in a prior publication based on an observational analysis from PRESTO, in which we found effect sizes of 20% (95% CI 1.13–1.28) for use of selected apps including FF and other similar apps at baseline vs. non-use; and 12% (95% CI 1.04–1.21) for time-varying use of selected apps vs. non-use (7). Nevertheless, our present results remain consistent with the possibility that app use may improve fecundability and reduce TTP for some subgroups. Our observation of stronger effect among younger individuals and those with lower educational attainment is not entirely surprising given that these individuals may have lower awareness of their fertility indicators (33). App usage may have led to substantial improvements in their knowledge about fertility signs and identification of the fertile window. We did not stratify by age in the prior report (7) and we are not aware of any previous publications that have done so, which limits our ability to corroborate the possibility of a differential effect by the user’s age. Similarly, those with longer pregnancy attempt times at enrollment may have also been less aware of their fertility signs, which may have contributed to their delays in conception before FF assignment. Stronger effects among non-users of hormonal contraception in the 3 months before enrollment are plausible because recent use of hormonal contraception temporarily reduces fecundity (34, 35). Thus, recent hormonal contraceptive use may have obscured any effect of app use on fecundability.
In conclusion, we found no appreciable association between randomization to FF and fecundability in intent-to-treat analyses. The high percentage of non-adherence in the intervention arm may have contributed to these null results. In secondary per-protocol analyses that accounted for adherence, initiating use of FF—a fertility-tracking app that allows the user to record multiple indicators (e.g., BBT, cervical fluid, or OPK testing)—was associated with a small increase in fecundability (shortening of TTP) among selected subgroups. Our study was not able to identify which FF features were most important, if any, in promoting fecundability. Results based on the per-protocol analyses are prone to unmeasured confounding and other sources of bias. Nevertheless, if the per-protocol results are causal, they might extend to a broader population of pregnancy planners, which is the population most likely to use a fertility-tracking app.
Supplementary Material
Attestation Statement:
We do not have any knowledge about whether participants in our study were concomitantly involved in other randomized trials. Because our trial did not involve randomization of a medical treatment, supplement, or device, if our participants were indeed concomitantly involved in other randomized trials, their involvement in other trials should not have introduced any concerns with respect to human subjects research.
Data from this particular analysis have not been previously published
Acknowledgements:
We acknowledge the contributions of PRESTO participants and staff. We thank Mr. Michael Bairos for technical support in developing the study’s web-based infrastructure.
Funding:
This research was supported by grants R21HD072326 and R01HD086742, from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, USA.
Footnotes
Disclosures: Dr. Wise reports in-kind donations from FertilityFriend.com (2013–2019) and Kindara.com (2019+) (free fertility apps for PRESTO participants), Swiss Precision Diagnostics (free home pregnancy tests for PRESTO participants), and Sandstone Diagnostics (discounted semen test kits for PRESTO participants) during the conduct of the study, and consulting fees from AbbVie.com for work unrelated to this paper. Dr. Stanford reports personal fees from Swiss Precision Diagnostics, outside the submitted work. Dr. Ncube is supported by National Institute on Minority Health and Health Disparities (NIMHD) Award Number K01-MD013911.
Trial Registration Number: Not applicable
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Data sharing:
We do not have permission from our participants to share individual-level data.
References
- 1.Worsfold L, Marriott L, Johnson S, Harper JC. Period tracker applications: What menstrual cycle information are they giving women? Womens Health (Lond). 2021;17:1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Schantz JS, Fernandez CSP, Anne Marie ZJ. Menstrual Cycle Tracking Applications and the Potential for Epidemiological Research: A Comprehensive Review of the Literature. Curr Epidemiol Rep. 2021;8:9–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Rosato D What Your Period Tracker App Knows About You. Consumer Reports. Available at: https://www.consumerreports.org/health-privacy/what-your-period-tracker-app-knows-about-you-a8701683935/. Accessed February 2, 2023.
- 4.Favaro C, Pearson JT, Rowland SP, Jukic AM, Chelstowska M, Berglund Scherwitzl E, et al. Time to Pregnancy for Women Using a Fertility Awareness Based Mobile Application to Plan a Pregnancy. J Womens Health (Larchmt). 2021;30:1538–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Earle S, Marston HR, Hadley R, Banks D. Use of menstruation and fertility app trackers: a scoping review of the evidence. BMJ Sex Reprod Health. 2021;47:90–101. [DOI] [PubMed] [Google Scholar]
- 6.Duane M, Stanford JB, Porucznik CA, Vigil P. Fertility Awareness-Based Methods for Women’s Health and Family Planning. Front Med (Lausanne). 2022;9:858977. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Stanford JB, Willis SK, Hatch EE, Rothman KJ, Wise LA. Fecundability in relation to use of mobile computing apps to track the menstrual cycle. Hum Reprod. 2020;35:2245–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Evans-Hoeker E, Pritchard DA, Long DL, Herring AH, Stanford JB, Steiner AZ. Cervical mucus monitoring prevalence and associated fecundability in women trying to conceive. Fertil Steril. 2013;100:1033–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Mu Q, Fehring RJ. Efficacy of achieving pregnancy with fertility-focused intercourse. MCN The American journal of maternal child nursing. 2014;39:35–40. [DOI] [PubMed] [Google Scholar]
- 10.Robinson JE, Wakelin M, Ellis JE. Increased pregnancy rate with use of the Clearblue Easy Fertility Monitor. Fertil Steril. 2007;87:329–34. [DOI] [PubMed] [Google Scholar]
- 11.Tiplady S, Jones G, Campbell M, Johnson S, Ledger W. Home ovulation tests and stress in women trying to conceive: a randomized controlled trial. Hum Reprod. 2013;28:138–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Stanford JB, Willis SK, Hatch EE, Rothman KJ, Wise LA. Fecundability in relation to use of fertility awareness indicators in a North American preconception cohort study. Fertil Steril. 2019;112:892–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Johnson S, Stanford JB, Warren G, Bond S, Bench-Capon S, Zinaman MJ. Increased Likelihood of Pregnancy Using an App-Connected Ovulation Test System: A Randomized Controlled Trial. J Womens Health (Larchmt). 2020;29:84–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Stanford JB, Smith KR, Varner MW. Impact of instruction in the Creighton model fertilitycare system on time to pregnancy in couples of proven fecundity: results of a randomised trial. Paediatr Perinat Epidemiol. 2014;28:391–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Wise LA, Rothman KJ, Mikkelsen EM, Stanford JB, Wesselink AK, McKinnon C, et al. Design and Conduct of an Internet-Based Preconception Cohort Study in North America: Pregnancy Study Online. Paediatr Perinat Epidemiol. 2015;29:360–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Faust L, Bradley D, Landau E, Noddin K, Farland LV, Baron A, et al. Findings from a mobile application-based cohort are consistent with established knowledge of the menstrual cycle, fertile window, and conception. Fertil Steril. 2019;112:450–7 e3. [DOI] [PubMed] [Google Scholar]
- 17.Matsumoto M, Nishimura T. Mersenne twister: a 623-dimensionally equidistributed uniform pseudo-random number generator. ACM Trans Model Comput Simul. 1998;8:3–30. [Google Scholar]
- 18.Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz AM, Strath SJ, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc. 2000;32:S498–504. [DOI] [PubMed] [Google Scholar]
- 19.Olsen LR, Jensen DV, Noerholm V, Martiny K, Bech P. The internal and external validity of the Major Depression Inventory in measuring severity of depressive states. Psychol Med. 2003;33:351–6. [DOI] [PubMed] [Google Scholar]
- 20.Cohen S, Williamson G. Perceived stress in a probability sample of the United States In: Spacapan S, Oskamp S, editors. The social psychology of health: Claremont Symposium on applied social psychology. Newbury Park, CA: Sage; 1988. [Google Scholar]
- 21.Baird DD, Wilcox AJ, Weinberg CR. Use of time to pregnancy to study environmental exposures. Am J Epidemiol. 1986;124:470–80. [DOI] [PubMed] [Google Scholar]
- 22.Cox DR. Regression models and life-tables. R Stat Soc J (B). 1972;34:187–220. [Google Scholar]
- 23.Therneau TM, Grambsch PM. Modeling Survival Data: Extending the Cox Model. New York, NY (USA): Springer-Verlag; 2000. [Google Scholar]
- 24.Weinberg CR, Wilcox AJ, Baird DD. Reduced fecundability in women with prenatal exposure to cigarette smoking. Am J Epidemiol. 1989;129:1072–8. [DOI] [PubMed] [Google Scholar]
- 25.Zhou XH, Eckert GJ, Tierney WM. Multiple imputation in public health research. Stat Med. 2001;20:1541–9. [DOI] [PubMed] [Google Scholar]
- 26.Sterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.SAS. SAS Institute Inc. 2014. SAS/STAT® 9.4 User’s Guide. Cary, NC. Cary, NC: SAS Institute; 2014. [Google Scholar]
- 28.Symul L, Wac K, Hillard P, Salathe M. Assessment of Menstrual Health Status and Evolution through Mobile Apps for Fertility Awareness. NPJ Digit Med. 2019;2:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Murray EJ, Swanson SA, Hernán MA. Guidelines for estimating causal effects in pragmatic randomized trials (pre-print). arXiv. 2019. [Google Scholar]
- 30.Murray EJ, Caniglia EC, Swanson SA, Hernandez-Diaz S, Hernan MA. Patients and investigators prefer measures of absolute risk in subgroups for pragmatic randomized trials. J Clin Epidemiol. 2018;103:10–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Bigelow JL, Dunson DB, Stanford JB, Ecochard R, Gnoth C, Colombo B. Mucus observations in the fertile window: a better predictor of conception than timing of intercourse. Hum Reprod. 2004;19:889–92. [DOI] [PubMed] [Google Scholar]
- 32.Manders M, McLindon L, Schulze B, Beckmann MM, Kremer JA, Farquhar C. Timed intercourse for couples trying to conceive. Cochrane Database Syst Rev. 2015:CD011345. [DOI] [PubMed] [Google Scholar]
- 33.Lundsberg LS, Pal L, Gariepy AM, Xu X, Chu MC, Illuzzi JL. Knowledge, attitudes, and practices regarding conception and fertility: a population-based survey among reproductive-age United States women. Fertil Steril. 2014;101:767–74. [DOI] [PubMed] [Google Scholar]
- 34.Hassan MA, Killick SR. Is previous use of hormonal contraception associated with a detrimental effect on subsequent fecundity? Hum Reprod. 2004;19:344–51. [DOI] [PubMed] [Google Scholar]
- 35.Yland JJ, Bresnick KA, Hatch EE, Wesselink AK, Mikkelsen EM, Rothman KJ, et al. Pregravid contraceptive use and fecundability: prospective cohort study. BMJ. 2020;371:m3966. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
We do not have permission from our participants to share individual-level data.