Factors affecting time to pregnancy

Abstract

BACKGROUND: Both lifestyle factors and occupational and environmental factors have been suggested to affect the female reproductive system. In the present study, the separate and joint effects of several such factors are investigated. METHODS: Information on time to pregnancy (TTP) was available for 1578 women randomly selected from the general Swedish population. The information was collected retrospectively by using self-administered questionnaires. By means of logistic regression of survival data, fecundability odds ratios were determined for many factors. Multivariate models were used to determine which factors had the most impact on TTP. RESULTS: Several lifestyle factors were found to associate with TTP. However, only use of oral contraceptives prior to attempting to conceive, menstrual cycle length, age at conception and parity remained in the multivariate models. Together, these factors explained 14% of the variance in TTP. Excluding first and second month conceptions, only age at conception and menstrual cycle length remained in the multivariate models, together explaining only 8% of the variance in TTP. CONCLUSIONS: Although information on several factors was available, the multivariate model explained only a small fraction of the variation in the observed time to pregnancies. Furthermore, female biological factors seemed more important predictors of TTP than lifestyle factors.

Introduction

The female reproductive system seems to be very vulnerable to outside interference: lifestyle-related factors, such as alcohol consumption (Wulff et al., 1999; Juhl et al., 2001; Juhl et al., 2003; Sallmen et al., 2003) and smoking habits (Spinelli et al., 1997; Curtis et al., 1999; Taskinen et al., 1999; Thonneau et al., 1999; Wulff et al., 1999; Abell et al., 2000; Jensen et al., 2000; Sallmen et al., 2003; Hassan and Killick, 2004b), have been shown to affect fertility by means of delayed conception and increased risk of miscarriage, whereas psychological stress has been found to disturb menstrual function (Fenster et al., 1999; Hatch et al., 1999; Hjollund et al., 1999), as well as decrease fertility (Hjollund et al., 1999) and increase the risk of miscarriage (Bashour and Abdul Salam, 2001). Furthermore, various work-related factors such as number of working hours (McDonald et al., 1986; McDonald et al., 1988b; Hatch et al., 1997; Tuntiseranee et al., 1998; El-Metwalli et al., 2001), physically demanding work (McDonald et al., 1986; El-Metwalli et al., 2001), prolonged standing (McDonald et al., 1988a; Henriksen et al., 1995) and shift/night work (Ahlborg et al., 1996; Bisanti et al., 1996) seem to interfere with both the female reproductive system and the growth of the fetus.

By assessing the time from cessation of contraceptive use to conception, i.e. the time to pregnancy (TTP), one can achieve a sensible measure of a couple’s fertility (Baird et al., 1986; Joffe, 1989). It has been found that this information can be collected retrospectively with a good validity (Joffe, 1989) by using self-administered questionnaires (Zielhuis et al., 1992).

The purpose of the current study was to investigate which factors affect fertility, measured as TTP, in a group of randomly selected Swedish women. This was done without any preconceived idea of a primary potential predictor of TTP.

Materials and methods

Study population and questionnaire

In order to perform a study on reproductive effects among Swedish female hairdressers, a sample of age-matched women from the general Swedish population were selected as controls (Axmon et al., unpublished data). Of 5299 controls born in 1960, or later, who were sent a self-administered questionnaire regarding TTP in 2000, 2860 (54%) participated in the current study. The questionnaires were designed in such a way that factors which applied to the pregnancy in general [e.g. use of oral contraceptives (OCs) prior to trying to conceive] were asked about only once and related to the time prior to when the woman started her attempts to become pregnant. Other factors that may change over time (such as smoking habits) were determined for each 3-month period during the year prior to conception. From women who never succeeded in conceiving, or were currently trying to conceive, information was collected for the period during which they tried to conceive.

The study was approved of by the Ethics Committee at Lund University, Sweden.

Outcome

The outcome used was the TTP for the woman’s first planned pregnancy. This was assessed using a line of questions: (1) Did you become pregnant the first month of trying? (2) If no, did you become pregnant the second month of trying? And (3) If no, in which month did you become pregnant? The median recall time was 8 years (range 1–25 years). Women who had never been pregnant were asked if they were, or at some point had been, trying to conceive, and if so, for how long. These trying times were included in the analyses as censored TTPs (see also statistics).

Investigated risk factors

The factors investigated for a potential influence on fertility are presented in Table I. For women whose TTP was shorter than 12 months, the information corresponding to the time when they started their attempts to become pregnant was used. For women whose TTP was longer than 12 months, the information concerning the tenth to twelfth months prior to conception was used.

The age of the woman when she was included in the study (i.e. answered the questionnaire) was established. Since all women were included at the same time, age at inclusion corresponds to year of birth. Information was also collected on the woman’s age at first menstruation, her highest level of education (elementary school, high school or college/university), her age at conception, the parity of the first planned pregnancy and whether or not she performed heavy lifts prior to the pregnancy. No definition of ‘heavy lifts’ was given, but this relates to the woman’s own perception of her situation. The use of contraceptive method prior to attempting to become pregnant was determined as OC/intrauterine device or other contraceptive method. No information was collected concerning the type of OC used. The women were asked about the length of their menstrual cycle prior to conception. Since all pregnancies inquired about were planned, it was understood that the women were not using OC at this time point. Some of the women claimed to have very short menstrual cycles (4% stated that their menstrual cycles were shorter than 14 days). It was assumed that these women had mistaken menses duration for menstrual cycle length, and their menstrual cycle length was set to missing.

For smoking habits (both the woman’s own and her partner’s), a yes/no question was first asked to establish whether the person inquired about was a smoker. For smokers, a follow-up question then established the number of cigarettes smoked per day. Consumption of alcoholic beverages (yes/no) was determined for both the woman and her partner.

The women were asked whether they used any medication, vitamins or food supplements regularly prior to, or during pregnancy, and if so, which product(s). A physician went through the list of product names and classified them as either medication or vitamin/minerals/iron supplementation. To further investigate the possible relation between use of medication and TTP, the Anatomical Therapeutic Chemical (ATC) (http://www.whocc.no/atcddd) classification system was used to categorize the different drugs. Each woman was defined as a user or non-user in each category. Unfortunately, it was not possible to make a distinction between medication/vitamin use prior to and medication/vitamin use during the pregnancy.

Employment status was determined as part-time employed, full-time employed (in Sweden, this normally means 40 h per week, although some of the women included in the present study actually worked more than this), self-employed, housewife, student, unemployed and other. Women who were full-time employed were grouped with those who were self-employed, using the other categories as reference. Women in the workforce were asked if they had irregular working hours or worked night shift. The term irregular working hours was not defined in the questionnaire but relates to how the woman perceives her own situation. We also asked these women whether they performed their work in a standing position, and if so, for how many hours per day. Furthermore, information was collected on self-perceived work-related stress and workplace smoking (yes/no).

All pairwise correlations (Spearman’s) among the different factors in Table I were checked. The woman’s alcohol consumption was found to correlate with that of her partner (r_S = 0.70). None of the other correlations exceeded 0.5.

Non-respondents

The year of birth was known for respondents as well as non-respondents. Furthermore, data from the Swedish Medical Birth Registry were available for women who gave birth between 1973 and 1994 (although smoking habits were not recorded until 1982), enabling comparison between non-respondents and respondents with respect to smoking as well as several reproductive outcomes.

There was a lower fraction of non-smokers among the non-respondents (63%) than among the respondents (74%). The non-respondents were more likely to have given birth to a growth-retarded infant (5.2 versus 2.7%) but less likely to have given birth to a stillborn infant (0.3 versus 1.1%). The differences regarding birth year, having given birth between 1973 and 1994, number of children, ever giving birth to a low birth infant (<2500 g) or an infant with malformation(s) were minor or none (data not shown).

Statistics

Since TTP describes the time to an event, Cox regression has been suggested to be a suitable method for comparing populations with different exposures. However, since ties (events which happens at the same time) have no underlying order in TTP data, the Cox regression model translates into the Cox’s model for discrete-time data, which likelihood equation is similar to that for the proportional odds model. Thus, the model can be applied using logistic regression to a database of months (cycles), estimating the per month (cycle) relative odds for pregnancy (Baird et al., 1986). Furthermore, analyses of simulated TTP data have shown that the results estimated by logistic regression are similar to those estimated by Cox regression (Axmon, 2003). Thus, to estimate the association between the different factors on one hand and TTP on the other, logistic regression was applied to a database of months, estimating the per month relative odds of pregnancy. By using this method, the calculated risk estimate is an odds ratio (OR). The interpretation of this OR is the same as that of the fecundability ratio (FR), i.e. the monthly conception rate among exposed compared with that among the unexposed. Thus, to make the results easier to interpret and compare with other studies, we will refer to our risk estimates as FRs. The FR describes the change in fertility when moving from one level of a factor to the next, e.g. FR = 0.87 indicates a 15% longer TTP compared with the reference group (1/0.87 = 1.15). For continuous data, the FR gives the change in fertility for each unit of increase in the continuous variable (e.g. being 1 year older). For categorical data, the FR gives the change in fertility for each category compared with the reference category. For yes/no variables, the no-category was consistently used as reference category.

To avoid interference from medical treatment for infertility, TTP was censored at 12 months (n = 82) (Baird et al., 1986). All women who stated that their pregnancy indeed was a result of medical treatment (n = 69) were included in the analyses under the assumption that their TTP was longer than 12 months (i.e. censored at 12 months), since this is normally the earliest time point when medical treatment is introduced. The TTP for women who currently were (n = 65), or previously had been (n = 18), trying to conceive were censored at 12 months (n = 35) or at the time they filled out the questionnaire/stopped trying to conceive if this represented less than 12 months of trying time (n = 48). Twelve women who stated that they did not conceive during the first 2 months of trying (i.e. had checked the ‘no’ boxes for achieving pregnancy in the first or second month), but not provided a TTP, were included in the analyses but censored at 2 months.

Even though each explanatory variable was considered as a single factor that could affect the outcome variable (TTP), it is obvious that some of the explanatory variables might have acted as confounders for each other. We therefore aimed to find a multivariate model for each outcome. This model was established by first choosing the variable with the lowest P-value, and then adding the other variables, one by one, if their P-value in the univariate model was less than 0.20, and their P-value in the multivariate model was less than 0.10. The Nagelkerke R² (Nagelkerke, 1991) was used to describe the proportion of the total variance explained by the multivariate models. Since some of the variables were applicable only to women in the workforce, separate models were evaluated for these women. Furthermore, on the basis of the results found for use of OC (see below), we also performed subgroup analyses excluding first and second month conceptions.

Although 1557 women provided a TTP, the number of women included in the different analyses differed due to incompletely answered questionnaires. However, in the main multivariate analyses, the number of women included were 1232 and 918 for all women and women in the workforce, respectively.

Results

The median TTP was 2 months (95%, range 1–27 months; Figure 1). Increasing age at conception increased the TTP (FR = 0.95, 95% confidence interval (CI): 0.93, 0.96; the FR represents the change in fertility compared with women who are 1 year younger; Table I) as did menstrual cycle length (FR = 0.96, 95% CI: 0.94, 0.98; the FR describes the change in fertility associated with a 1 day increase in menstrual cycle length) and nulliparity (FR = 0.81, 95% CI 0.70, 0.94). Use of OC prior to trying to conceive increased overall TTP (FR = 0.72, 95% CI: 0.62, 0.83), but no effect was found when excluding women who became pregnant the first or second month of trying (FR = 0.89, 95% CI: 0.71, 1.12).

Alcohol consumption increased the TTP (FR = 0.83, 95% CI: 0.72, 0.95; Table I) as did a regular use of medication (FR = 0.67, 95% CI: 0.53, 0.84). When using the ATC classification system, the effect on TTP from drug use was found mainly among non-steroidal anti-inflammatory drugs (FR = 0.32, 95% CI: 0.12, 0.90) and nervous system drugs (FR = 0.38, 95% CI: 0.19, 0.76).

TTP was prolonged also among women who had full-time work compared with those employed part-time or without employment (FR = 0.82, 95% CI: 0.71, 0.95). Among women in the workforce (including those with part-time work), TTP was associated with working night shift (FR = 0.74, 95% CI: 0.55, 0.98) and self-perceived work-related stress (FR = 0.78, 95% CI: 0.67, 0.91).

Both multivariate models included use of OC, menstrual cycle length, age at conception and parity (Table II). According to Nagelkerke R², the model for all women explained 14% of the total variance. In the subgroup of women who were in the workforce, the final model explained 12% of the total variance.

Excluding first and second month conceptions, use of OC, nulliparity, full-time work and alcohol consumption no longer qualified for inclusion in multivariate models for all women, since their P-values in univariate models were above 0.20. The final model in this subgroup analysis included age at conception and menstrual cycle length (Table II). The multivariate model for the subgroup of women in the workforce consisted of the same variables. Both models explained only 8% of the total variance.

Discussion

In the present study, several factors were found to have an impact on TTP. However, only use of OC, menstrual cycle length, age at conception and parity remained in multivariate models. Furthermore, when excluding first and second month conceptions, only age at conception and menstrual cycle length were found to be significant predictors of TTP.

Fertility was measured using TTP. It has been found that this is a useful tool when assessing reproductive effects (Baird et al., 1986; Joffe, 1989), and that data can be collected retrospectively by using self-administered questionnaires, even when the recall times are as long as 14 years (Joffe, 1989; Zielhuis et al., 1992). A major critique concerning retrospective TTP studies has been the exclusion of women who never conceived. However, a strength of the present study is that we also included women who had tried to conceive but so far had failed to do so.

A weakness of the present study is the low response rate. The non-respondent analyses found the respondents and non-respondents to be comparable with respect to several reproductive outcomes. However, the non-smokers were over-represented among the respondents. It is fairly well documented that smoking is associated with fertility (Practice Committee of the American Society for Reproductive Medicine, 2004). Thus, the results may be affected by a selection bias with respect to smoking habits.

The personal experience of the authors is that self-administered questionnaires on reproductive outcomes will supply a low response rate. In designing the questionnaire for the present study, we therefore tried to avoid questions that we suspected would lower the response rate even more. Such a question was that on frequency of sexual intercourse. However, frequency of intercourse may partly be considered an effect of certain lifestyle factors, such as stress and working hours, and thus an intermediate effect rather than a biological background determinant. Thus, lack of information on frequency of sexual intercourse should not pose a problem in the present study.

For the women who had yet to become pregnant, information was collected for the entire period when they tried to conceive. This could be several years, and thus these data may be biased. However, the fraction of women who had not yet conceived was very low (5%), and the impact on the fecundability estimates should therefore be slight.

OCs inhibit folliculogenesis by a central suppressive action on the release of gonadotropins. A detailed study of 40 long-term users of low-dose OCs, aged 22–36 years, showed that during the 7-day pill-free interval the hypothalamic–pituitary axis seems to recover completely from the steroidogenic feedback effects of the preceding pill cycles (Hemrika et al., 1993). These physiological results, supporting that the feedback effects wear off rapidly, give some circumstantial evidence that the fecundity will not be affected after ending OC use. However, a recent study of almost 3000 pregnant women showed a modest significant decline in fecundity after stopping the use of long-term combined OC as compared with condom users (averagely 3 months longer TTP) (Hassan and Killick, 2004a). This effect was more evident after long-term use in older women and in women who were overweight or who had menstrual disturbances, i.e. those with potentially compromised ovarian function. No such effect was present after the use of progesterone-only pills. In contrast, in another study, prolonged use of OC was associated with improved fecundity (Farrow et al., 2002). The interpretation of the results from this latter study is, however, hampered by a potential selection bias (exclusion of unintentional pregnancies). The results of the present study give some support for the fact that prior long-term OC use might have a modest reversible detrimental effect on fecundity.

We were not surprised to find that fecundability was reduced with increasing age, since this must be considered a well-known fact, which also has been supported in recent studies (Dunson et al., 2002; Gnoth et al., 2003). Although Jensen et al. (2000) found that fertility increased with age, they also argued that this finding most likely was due to bias.

Women who are not nulliparous constitute a selected group with respect to fertility. Thus, it was not unexpected to find that non-nulliparity would be associated with decreased TTP. Nevertheless, even though adjusting for parity as a potential confounder has become routine in studies of TTP, we have failed to find any previous study designed to investigate fertility in relation to parity.

A mechanistic linkage between prolonged menstrual cycle length and decreased fertility has recently been proposed (Greb et al., 2005). About 20% of women are homozygous for a single-nucleotide polymorphism in exon 10 of the FSH receptor, leading to higher ovarian threshold to FSH, decreased negative feedback of luteal secretion to the pituitary during the intercycle transition and longer menstrual cycles. To the best of our knowledge, there are no prior studies designed to investigate the relation between menstrual cycle length and TTP. However, an effect of the study design might partly also have contributed to our finding of prolonged menstrual cycles as a risk factor for prolonged TTP. Since a woman can conceive only once in each menstrual cycle, the proper way to measure TTP is of course in cycles. However, the majority of studies investigating TTP have, as have we, measured TTP in months. For women with regular menstrual cycles of 28 days, the effect of the choice of unit would only have a very slight impact on the effect estimates. However, for women with long menstrual cycles, measuring TTP in months would overestimate the true TTP. Thus, it is possible that our finding with respect to menstrual cycle length is in part because of the way we measured TTP.

In conclusion, although information on several explanatory factors was available, the multivariate models presented explained only a relatively small fraction of the variation in the TTP. Furthermore, female biological factors, such as age and menstrual cycle length, were more important predictors of TTP than lifestyle factors such as smoking habits and working hours. Thus, male factors (Bonde et al., 1998) as well as yet unidentified female factors are likely to contribute to the variation in couple fertility measured as TTP.

Acknowledgements

This work was financed by grants from the Swedish Council for Medical Research, the Swedish Council for Working Life and Social Research, the Medical Faculty at Lund University and Region Skåne.

References