The association between the use of fertility indicators and fecundability in a Danish preconception cohort.

EMM and ASDL designed the study. EMM, HTS, EEH, LAW and KJR planned and initiated the cohort. ESP analysed the data, and all authors interpreted the results. ESP drafted the manuscript. All authors approved the final version of the manuscript and are accountable for all aspects of the work.

In this study of Danish females attempting pregnancy, the use of any FI was associated with increased fecundability after adjustment for potential confounders, while associations for individual FIs varied in strength and precision. Cervical fluid monitoring and urinary ovulation testing were associated with the highest probabilities of pregnancy. Overall, the strengths of this study are a low risk of selection bias 31 and an objective based on a clear physiological mechanism of probable causality. The study participants were self‐selected and self‐enrolled in the study via the Internet, but this should not cause selection bias (i.e. when selection into a study depends on both exposure and outcome) since all comparisons were made within the study population. The participants had no information about their outcome (conception) at study entry. 32 Further, we included only participants trying to conceive for ≤6 cycles at baseline to reduce the influence of selection bias or behaviour changes that might have occurred because of extended pregnancy attempt time. 26 The limitations include the fact that participants who spent longer trying to conceive were more likely to use an FI and attenuation bias during follow‐up, which would both contribute to likely underestimation of the association. In addition, we do not have direct information about the impact of FI use on the timing of intercourse relative to the fertile window since we collected information on average intercourse frequency only. However, the findings consistently indicate increased fecundability from using FIs, except for BBT. Through adjustments for physical activity, BMI, multivitamin intake and smoking, we sought to control the influence of health behaviour. Some residual confounding remains since these adjustments are not a perfect measure of health behaviour. 33 , 34 Nonetheless, estimates of fecundability did not change notably after adjustment for the included indicators despite each being associated with fecundability. 35 , 36 , 37 , 38 Despite excluding participants with attempt times ≥6 cycles, the analysis restricted to ≤2 cycles of attempt at study entry and two cycles of follow‐up yielded a stronger association between FI use and fecundability than the primary analysis. There are two likely explanations: (1) differences in FI use by attempt time and (2) attenuation bias. First, participants might increase efforts to enhance their fecundity if they do not conceive promptly. Therefore, those with longer attempt times at entry may use FIs more frequently. This difference could attenuate the association with fecundability because participants using FIs would, on average, have lower fecundity. The analysis restricted to ≤2 cycles of attempt and two cycles of follow‐up should limit this potential bias, meaning estimates from this analysis are less likely than those from the main analyses to reflect the possible influence of participants with lower fecundity being more likely to use FIs. Second, attenuation bias cannot be ruled out. This occurs when heterogeneity in inherent fecundity causes the most fecund persons to deplete from the population at risk of pregnancy more rapidly than those who are least fecund. 39 In this study, FI users had higher pregnancy rates and thus are depleted more rapidly than non‐users. This leads to a diminution of the positive association during follow‐up, as the FI‐exposed group shifts to include a greater proportion of lower fecundity participants, in comparison to the unexposed group. A comparison of bias in fecundity studies indicated that a decrease in follow‐up time reduces attenuation bias, 39 which is also suggested in this study. Whether the increase in observed fecundability in the time‐restricted analysis is due to an increase in the use of FIs with a longer time trying to conceive and/or attenuation bias over time, we expect the time‐restricted results to more closely reflect the unbiased gain in fecundability in the short term. Another limitation of this study is that FI use could be misclassified in the baseline assessment, compared with the follow‐up, since some non‐users at baseline might start using an FI to increase conception chances. 26 However, only 5%–6% of participants started or stopped using an FI during follow‐up, indicating low levels of behaviour change. Further, the analyses allowing for time‐varying exposure status showed slightly attenuated associations for all FIs and an unchanged association for any FI, indicating little influence of misclassification caused by reported behaviour change. The results correspond to those of Stanford et al. 17 in a comparable study in North America, although some associations with individual FIs were inconsistent. This study investigated a similar research question in a different study population, where the demographic characteristics of participants and FI use patterns differed from the North American population. Furthermore, access to healthcare services, including reproductive assistance if needed, in the two populations differ, which may affect attitudes towards and intensity of trying to conceive in the preconception period. Surprisingly, this study did not find a positive association between measuring BBT and fecundability, unlike the study by Stanford et al. This difference might be explained by a lower user prevalence of this indicator in our cohort (3% vs. 21%), although this alone does not explain why measuring BBT was associated with lower fecundability in this study. Exactly how BBT is used could affect the association. If BBT is used purely retrospectively to time intercourse, it would be counterproductive, because the BBT rise is usually a postovulatory event. 3 , 10 However, retrospectively BBT‐identified ovulation timing can be used to forecast the timing of ovulation in future cycles, which may be a more effective use of BBT. 40 In a related article, Stanford et al. 12 found that use of smartphone apps was associated with a 12%–20% higher fecundability compared with no app use, while this study found a modest and imprecise association. This discrepancy might be explained by the vastly lower prevalence of app use in our cohort (17% vs. 72%), whereas some participants of the study by Stanford et al. 12 were randomised to a free app subscription after baseline enrolment. If the ability to accurately identify the fertile window differs between the types of apps used in the two populations, this could have contributed to the difference in estimates. Information about the specific apps used by the Danish cohort was not collected. We found a noticeably higher fecundability among users of cervical fluid monitoring compared with participants not using any FI. The potential increase in fecundability from monitoring cervical fluid is physiologically plausible, and has been supported in several observational studies. 6 , 8 , 9 , 10 , 41 , 42 However, a prior small randomised trial failed to confirm the impact of monitoring cervical fluid on time‐to‐pregnancy, suggesting that additional research may still be needed. 43 Feeling ovulation was associated with increased fecundability after adjustment for some potential confounders, but not after additional adjustment for reproductive history and gynaecological factors. Forty‐seven per cent of participants who reported feeling ovulation were parous at baseline, a proportion larger than in any other group of FI users. Together, these findings suggest that the association between feeling ovulation and fecundability was driven by reproductive history, because those who feel ovulation more frequently had proven reproductive potential. Further, ‘feeling ovulation’ may mean different things to different individuals. 44 Our results do not support using feeling ovulation as a tool for optimising fertility. A notable difference between types of FIs is the mechanism of identifying the fertile window or peak fertility. Cervical fluid monitoring and urinary ovulation testing are biological methods based on physiological observations for each menstrual cycle, expressing the effects of current hormonal levels in the body that help create the fertile window, particularly, oestrogen and luteinising hormone. 3 , 4 In contrast, counting days in the menstrual cycle and some fertility apps are statistical methods that rely on characteristics of previous menstrual cycles, and thus do not necessarily reflect the current cycle. In this study, the biological methods were associated with higher fecundability than the statistical methods, aligning with previous findings, 8 , 10 , 17 especially among those with less than fully regular cycles. Among females of normal fecundity, there is substantial variability in the day of ovulation and the fertile window. 45 , 46 , 47 Therefore, statistical methods may estimate the fertile window incorrectly if they do not adequately account for individual cycle length or variation, leading to worsened chances of achieving pregnancy since intercourse may be timed outside of the fertile peak.

The Eunice Kennedy Shriver National Institute of Child Health and Human Development (R21‐HD050264, R01‐HD086742 and R01‐HD060690) and the Danish Medical Research Council (271‐07‐0338) supported the establishment of the cohorts.

We used data from two preconception cohorts; SnartGravid.dk (Soon Pregnant, SG) 21 and its successor, SnartForældre.dk (Soon Parents, SF). During 2007–2011, SG comprised participants who self‐identified as female. When male partners began to be included in 2011, the study name was changed to SnartForældre.dk. As of January 2024, the SF cohort continues to enrol participants. Eligible females are 18–49 years old, residing in Denmark, trying to conceive with a male partner ≥18 years old and not receiving fertility treatments. Participants complete an online screening questionnaire and, if eligible, an extensive baseline questionnaire. Follow‐up questionnaires are sent to female participants every 2 months for up to 12 months until reported pregnancy, cessation of pregnancy attempt, study withdrawal or initiation of fertility treatment, whichever comes first. FI use was ascertained at baseline with the question, ‘Do you or your partner do anything to identify the time during your menstrual cycle when you have the greatest chance of becoming pregnant?’ Following a positive response, participants selected from multiple predefined FIs in a drop‐down menu or submitted a free text response, grouped with the predefined categories when possible or otherwise categorised as ‘other’ ( n  = 77). Predefined FIs were counting days from the last menstrual period, urinary ovulation testing, cervical fluid monitoring, measuring BBT, feeling when ovulation occurs (not further specified in the questionnaire) and using smartphone applications that track the menstrual cycle. The study's outcome was fecundability, measured using time‐to‐pregnancy (TTP) data. Fecundability is the average per‐cycle probability of achieving pregnancy among non‐contracepting couples engaged in regular intercourse. Pregnancy was reported as a positive home pregnancy test or physician‐confirmed blood or urine test. The number of cycles at risk was calculated from self‐reported cycle length by participants with regular cycles and calculated for participants with irregular cycles based on the reported last menstrual period (LMP) date at baseline and LMP dates during follow‐up. TTP was measured in cycles rather than months to represent one pregnancy chance and was calculated with the following formula 22 : TTP = Cycles of attempt at baseline + Most recent LMP date − Date of baseline questionnaire Usual cycle length + 1 This calculation accounts for left truncation due to delayed entry of participants who had attempted pregnancy for up to six cycles at baseline. Participants who had attempted pregnancy for six cycles or more were excluded from this study. Based on the literature, covariates were selected using a directed acyclic graph (see Figure  1 ). Covariates were reported at baseline and included age, length of education/vocational training, household income, 23 body mass index, current smoking, multivitamin intake, physical activity, intercourse frequency, parity, menstrual cycle regularity, last method of contraception, gynaecological diseases (polycystic ovaries (PCO)/polycystic ovary syndrome (PCOS), endometriosis, fibromas), history of infertility (before enrolment) and pregnancy attempt time at study entry. Directed acyclic graph of the expected relationship between fertility indicator use and fecundability. We used proportional probabilities regression models to estimate fecundability ratios (FRs) and 95% confidence intervals (95% CIs) for the association between FI use (vs. non‐use) and fecundability. 24 , 25 Participants contributed cycles of observation from study entry until they reported a pregnancy or one of the following events occurred: initiation of fertility treatment, study withdrawal, cessation of pregnancy attempt, loss to follow‐up or 12 months of follow‐up. We investigated the use of any FI and specific FIs in relation to fecundability. Three models were used to describe the association: (1) a simple model adjusted for the number of FIs when analysing FI types, (2) a general model adjusted for age, household income, duration of vocational training, body mass index (BMI), physical activity, smoking and multivitamin intake and (3) a reproduction‐specific model adjusted for reproductive history (parity, previous infertility and attempt time at study entry) and gynaecological factors (menstrual cycle regularity, PCO/PCOS, endometriosis and fibromas) in addition to the covariates used in model 2. In stratified analyses, we assessed the potential effect measure modification by educational attainment 23 as a proxy for health literacy, which could affect the ability to use the FI as intended. To investigate if the effect of exposure varied with length of attempt and follow‐up time, 26 we performed an analysis restricted to participants with an attempt time of ≤2 cycles at study entry, and we limited follow‐up to the next two cycles. Additionally, we performed analyses with time‐varying exposures, where the use of FIs was reassessed and updated every two cycles. Life table methods were used to estimate the pregnancy rate when accounting for censoring events. Statistical analyses were performed using SAS Software (version 9.4, SAS Institute Inc, Cary, NC). We used multiple imputation by fully conditional specification 27 to account for missing data 28 , 29 , 30 in instances where participants provided incomplete responses to the baseline questionnaire. Binary and nominal variables were imputed using logistic regression, and normally distributed continuous variables were imputed using linear regression. Covariates were imputed in order of least to most missing data. The final dataset for analysis consisted of 10 imputed datasets combined into parameter estimates and confidence intervals using Rubin's rule. 28 Of the included participants, 1813 were lost to follow‐up after the baseline questionnaire. We assigned these participants one cycle of follow‐up and multiply imputed their pregnancy status based on the characteristics of the remaining participants. This study was registered with Aarhus University (record number 2016‐051‐000001, #431). Ethical approval was not required as the study only included questionnaire data.

Between August 2007 and 9 February 2023, 16,229 respondents completed the baseline questionnaire. We excluded 44 participants who reported a pregnancy with an LMP date before baseline, 31 participants who withdrew their consent, 470 respondents who reported last menstrual period dates that were implausible or >6 months before study entry, 1244 who reported neither a menstrual period or pregnancy during follow‐up, 2806 who had been trying to conceive for >6 cycles at entry, 79 not yet up for follow‐up and 227 repeat enrolments, leaving a final study population of 11,328 participants (Figure  2 ). Flow chart describing study inclusion. The baseline characteristics of participants using FIs are shown in Table  1 . Forty‐seven participants (0.41%) had missing data for their FI use. Of the 11,328 participants, 63.3% reported using at least one FI, and 69.7% of users used more than one FI. A descriptive analysis of exposure status over the follow‐up time showed that 5%–6% of participants either started or stopped using an FI during this period. Among baseline FI users compared with non‐users, a larger proportion were 30 years or older, had higher household incomes and had high educational attainment. Reproductive history varied between users and non‐users; more users than non‐users had previously experienced infertility, were parous and had regular menstrual cycles. Among those who had attempted pregnancy for >3 cycles, a higher proportion used an FI at baseline, compared with those who had attempted pregnancy for ≤3 cycles. Baseline characteristics by fertility indicator use ( n  = 11,328). Abbreviations: BBT, basal body temperature; FI, fertility indicator; PCO/PCOS, polycystic ovaries/polycystic ovary syndrome. Complete as many or more METs of physical activity per week compared with the population median of 45.2 METs per week. The most frequently used FI was counting days from the last menstrual period while measuring BBT was used least often. Participants used cervical fluid monitoring, urinary ovulation testing, feeling ovulation and fertility apps with similar frequencies. A large percentage of participants who reported feeling ovulation were parous compared with those who did not report feeling ovulation. A large proportion of participants who used ovulation tests had attempt times of >3 cycles at entry, compared with those who did not use ovulation tests. The life table estimate of the pregnancy rate during 12 follow‐up cycles was 84.8% when accounting for censoring. Separate life table estimates were 86.3% for FI users and 82.5% for non‐users. In an unadjusted analysis, the FR for any FI use compared with no FI use was 1.17 (95% CI 1.12, 1.23) (Table  2 , Model 1). The association attenuated slightly after adjustment for age, socio‐economic position (household income and vocational training duration) and health indicators (current smoking, BMI, physical activity and multivitamin intake). Additional adjustments for reproductive history and gynaecological factors did not affect this estimate appreciably. Cervical fluid monitoring was associated with the highest FR compared with no FI use, followed by urinary ovulation testing, counting days from the last menstrual period and fertility apps (Table  2 , Model 3). Feeling ovulation was associated with increased fecundability before adjustment for reproductive history, but the estimate attenuated after additional adjustment for parity, previous infertility, menstrual cycle regularity, PCO/PCOS, endometriosis and fibromas (Table  2 , Models 2 and 3). Measuring BBT compared with no FI use was associated with a lower fecundability in all models, although this estimate was imprecise, with few participants using BBT. Every additional FI used was associated with a smaller increase in fecundability, although the marginal gain decreased (Table  S1 ). Sensitivity analysis showed no meaningful effect measure modification by educational attainment (Table  S2 ). Association between fertility indicator use and fecundability. Note : Model 1: Crude estimates for no FI use and any FI use, adjusted for number of FIs for individual FI‐types. Model 2: Adjusted for model 1 variables, age, household income, vocational/academic training, BMI, multivitamin intake, current smoking and physical activity. Model 3: Adjusted for model 2 variables, parity, previous infertility, menstrual cycle regularity, last type of contraception, endometriosis, fibromas and PCO/PCOS. Abbreviations: BBT, basal body temperature; CI, confidence interval; FI, fertility indicator; FR, fecundability ratio. When we restricted the analyses to participants with an attempt time of ≤2 cycles at study entry and allowed for only two cycles of follow‐up, the association between FI use and fecundability increased in magnitude for any FI and for each individual FI (Table  3 , Model 3). The analyses with time‐varying exposures showed attenuated associations between individual FIs and fecundability before and after covariate adjustment. In contrast, the association for any FI remained unchanged from that of the primary analysis (Table  4 ). Association between fertility indicator use and fecundability restricted to ≤2 cycles of attempt time and two cycles of follow‐up. Note : Model 3: Adjusted for age, household income, vocational training, BMI, multivitamin intake, current smoking, physical activity, parity, previous infertility, menstrual cycle regularity, last type of contraception, endometriosis, fibromas and PCO/PCOS. Individual FI types are adjusted for number of FIs used. Abbreviations: BBT, basal body temperature; CI, confidence interval; FI, fertility indicator; FR, fecundability ratio. Association between time‐varying fertility indicator use and fecundability. Note : Model 1: Crude estimates for no FI use and any FI use, adjusted for number of FIs for individual FI‐types. Model 3: Adjusted for model 1, age, household income, vocational training, BMI, multivitamin intake, current smoking, physical activity, parity, previous infertility, menstrual cycle regularity, last type of contraception, endometriosis, fibromas and PCO/PCOS. Abbreviations: BBT, basal body temperature; CI, confidence interval; FI, fertility indicator; FR, fecundability ratio.

Many couples have trouble conceiving and upwards of 20% can be categorised as infertile at some point in their lives after attempting pregnancy unsuccessfully for 12 months. 1 Some causes of infertility, such as tubal factors and anovulation, ultimately require surgical or medical intervention. 2 However, a modifiable cause of failure to conceive is the mistiming of intercourse. 2 Identification of ovulation and the surrounding fertile window is relevant to improving chances of conception by enabling the timing of sexual intercourse around peak fertility. 3 To identify this window, fertility indicators (FIs) can be employed, such as counting days since the last menstrual period, urinary ovulation testing, monitoring cervical fluid and measuring basal body temperature (BBT). The accuracy of the self‐identified fertile peak relative to ultrasonography‐confirmed ovulation varies among methods, 4 with urinary hormonal arrays and cervical fluid monitoring consistently found to be the most accurate. 5 , 6 , 7 , 8 , 9 , 10 Calendar methods are based on statistical averages and do not necessarily reflect the current menstrual cycle. 3 BBT measurements are difficult to interpret prospectively since the temperature rise is an event that indicates retrospectively when ovulation has occurred. 7 , 10 The latter two methods are more imprecise among persons with menstrual cycle irregularity. Recently, smartphone applications have been introduced, some of which use adaptive calendar tracking or other FIs to predict ovulation and peak fertility prospectively. 11 These methods have resulted in 12%–20% increased fecundability compared with no FI use. 12 Previous research on the timing of intercourse has primarily evaluated FIs as contraception methods, thus capturing unintended pregnancies among couples intending to avoid conception. 13 , 14 , 15 , 16 Studies on FIs as fertility‐promoting tools are scarce. 8 , 12 , 17 Furthermore, few studies have connected the use of FIs to actual fecundability measured as achieved pregnancies rather than theoretical fecundability regarding confirmed ovulation. 5 , 7 , 9 , 10 , 18 , 19 A systematic review from 2015 emphasised the lack of studies and ranked the quality of existing evidence as low or very low. 20 In 2019, a North American preconception cohort study found that the use of any FI was associated with 25% increased fecundability. 17 Evidence backing the use of FIs is necessary for clinicians to advise on non‐interventional fertility assistance, as well as empowering couples to maximise their chances of conceiving and reduce avoidable healthcare spending. 20 This study investigated whether the use of FIs was associated with fecundability based on a preconception cohort of Danish couples trying to conceive without fertility treatments. 3 , 4 Additionally, we compared types of FIs in relation to fecundability.

In this preconception cohort of Danish females, the use of any fertility indicator was associated with a 14% increase in the probability of conception per menstrual cycle among users compared with non‐users. Fecundability varied across types of fertility indicators: cervical fluid monitoring and urinary ovulation testing were associated with the greatest increases in fecundability.

The authors declare no conflicts of interest.

Table S1. Table S2.