Introduction

Both infertility (i.e., the inability to become pregnant after 1–2 years of trying) and unintended fertility are common, yet there is limited research that acknowledges the possibility that people may experience both infertility and unintended fertility within their reproductive life course [1]. Despite concurrently high rates of infertility and total fertility in population-level estimates, the relationship between experiences with infertility and subsequent intended and unintended fertility have seldom been investigated longitudinally within individuals. Understanding fertility experiences after infertility provides a holistic and dynamic view of the reproductive life course, improving upon existing work that focuses on singular reproductive events or experiences.

Coexisting high rates of infertility and unintended pregnancy are evident in Malawi, where approximately 20% of women report infertility (“have you ever tried to conceive a pregnancy for two years or longer without conceiving in that time?”) [2] or difficulty becoming pregnant (“whether an individual reports ever experiencing a difficult time in getting pregnant”) [3] and 41% of pregnancies are considered unintended [4]. Much remains unknown about the fertility experiences of women who become pregnant after experiencing a period of infertility or subfecundity, especially in lower-resource settings. Moreover, studies suggest that women who have experienced infertility may be less likely to use contraception than women who have not experienced infertility [5–7]. If women who have experienced infertility decide they do not want to become pregnant, there may be an opportunity to intentionally include them in family planning programs to support all women in reaching their reproductive goals.

Studies that examine the incidence of pregnancy after an experience of infertility are often focused on the clinical phenomenon of becoming pregnant after a one- or two-year period of exposure to unprotected sex. While the majority (79–92%) of fecund women will become pregnant within one year of attempting to become pregnant [8, 9], studies in clinical settings have found that around half of women who met the clinical definition for infertility–no pregnancy after one year–eventually became pregnant without infertility treatment [8, 10–12]. These studies demonstrate the potential to experience a period of infertility and still be able to conceive a pregnancy at some point within the reproductive life course. Women who become pregnant after meeting the definition for experiencing infertility are sometimes referred to as having resolved infertility or are referred to as “truly fertile.” Unlike sterility, which is not addressed in our paper, infertility refers to a specific period, not necessarily a permanent state; thus, it is possible that unintended fertility could follow a period of infertility. In this study, we refer to all women who have ever met the definition of infertility reflected in the survey item (having tried to conceive for two years without conceiving in that time) as having ever experienced infertility.

No studies, to our knowledge, explicitly examine pregnancy desires among women who become pregnant after experiencing infertility. There is a common assumption that women who experience infertility want to become pregnant, even though this may not be an accurate or enduring assumption [13]. Fertility preferences are dynamic [14], perhaps even among women and couples who once tried to become pregnant for a 1–2-year period and were not able to. Understanding the potential for women to have an unintended pregnancy after experiencing infertility presents challenges because it requires retrospective or longitudinal data that ideally spans the reproductive life course to understand the temporality of infertility and unintended fertility, as well as possible linkages between the two experiences.

Studying infertility and unintended fertility requires nuanced and time-specific measures of both phenomena. Moreover, both infertility and unintended fertility are difficult to measure [14–16]. Thus, few studies have examined the occurrence of unintended pregnancy after experiencing infertility, although one study found that having a “later-than-expected” birth—potentially due to fertility problems—may be associated with increased odds of a subsequent mistimed birth, likely due to the assumption that it would take longer to conceive based on their previous experience [17].

A deeper understanding of fertility after experiencing infertility is critical to ensuring that family planning programs address the needs of women and couples, both while they are experiencing difficulty becoming pregnant, as well as after (if they eventually conceive). A proposed mechanism for the relationship between experienced infertility and unintended pregnancy is shown in Fig 1, informed by constructs within the Health Belief Model [18]. Ever experiencing infertility may lower one’s perceived risk of becoming pregnant, contributing to contraceptive non-use, which in turn may lead to unintended pregnancy. Of course, women who experience infertility may want to become pregnant (and, indeed, may only report infertility if they want to become pregnant), which would potentially influence both contraceptive use and pregnancy [13]. Other women may be infertile, but because they are not actively trying to become pregnant, they may be unaware or unlikely to be counted as infertile in some measures [13]. Additionally, women who have experienced infertility may eventually report that they do not desire to become pregnant as a way of resolving the cognitive dissonance between what one wants and what one believes they can realistically achieve [19]. Indeed, infertility has been associated with lower fertility intentions in a U.S. study [20].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Analytic framework examining ever experienced infertility, pregnancy, and unintended pregnancy¹.

1Constructs in grey are not examined in this study.

https://doi.org/10.1371/journal.pgph.0001646.g001

This study aims to answer two main questions: (1) independent of pregnancy desire, are women who ever experienced infertility, here defined as ever trying to conceive for two years or longer without conceiving in that time, less likely to become pregnant in the future than women who have not experienced infertility? And, (2) among women who become pregnant, are those who have experienced infertility more likely to report that their pregnancy was unintended compared to women who had not experienced infertility?

Methods

Variables

Dependent variables. Any new pregnancy after Wave 1 was calculated by subtracting the number of lifetime pregnancies women reported at Wave 1 from the number they reported at each subsequent wave. This variable was then recoded as binary (1 indicated that they ever reported a new pregnancy after Wave 1 and 0 indicated that they reported no new pregnancies after Wave 1). The outcome of new pregnancy was examined in both samples.

Number of pregnancies was chosen rather than number of living children because living children may fluctuate due to child mortality. Thus, reporting fewer living children after Wave 1 is possible and cannot be assumed to be a reporting error [22]. Complete birth histories were not available.

Retrospective pregnancy planning was measured using the London Measure of Unplanned Pregnancy (LMUP) among women who had at least one pregnancy between Waves 1–5 (the pregnancy sample). The authors of the LMUP use the terms ‘unintended’ and ‘unplanned’ interchangeably [23]. This six-item scale was developed by a research team who conducted qualitative research and psychometrically tested and validated the scale in a sample of women in western Europe (Cronbach’s alpha = 0.97; test/retest reliability = 0.97) [23]. It was later validated in Malawi (Cronbach’s alpha = 0.78; rest/retest reliability = 0.80) [24]. The scale includes items measuring “expressed intentions, desire for motherhood, contraceptive use, pre-conceptual preparations, personal circumstances/timing, and partner influences” [23]. An English translation of the Chichewa version of the LMUP used in Malawi, along with how each response option is scored, is included in S3 Table. As suggested by the authors of the LMUP scale, we constructed the LMUP as a continuous variable, with scores ranging from 0 (least planned/intended) to 12 (most planned/intended) [23].

In the present study, all six LMUP items were correlated and the entire scale yielded a Cronbach’s Alpha of 0.90 (not shown). Item six (pre-conception preparations) was the least correlated with other items, but all six items loaded onto a single factor, and thus all six items were retained.

Independent variables.

All independent variables were measured at Wave 1 (2014–2015). The main independent variable was ever experienced infertility (have you ever tried to conceive a pregnancy for two years or longer without conceiving in that time?). We examined additional sociodemographic covariates typically associated with incidence of pregnancy and/or pregnancy intentions, including relationship status, age, number of pregnancies and number of living children, and education [21, 22, 25] (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Independent variables, response options (measured at Wave 1).

https://doi.org/10.1371/journal.pgph.0001646.t001

Results

Pregnancy incidence sample

Table 2 shows characteristics of women who had and had not ever experienced infertility reported at Wave 1. Overall, 65% of women became pregnant after Wave 1; the proportion was the same for women who had experienced infertility (65%) and for women who had not experienced infertility (65%) (Table 2). Women who reported infertility at Wave 1 were more likely to have ever experienced a sexually transmitted infection (STI) than women who did not report infertility (16% vs. 9%; p<0.05) and had fewer years of education on average (4.3 vs. 5.4 years; p<0.001). Notably, there were no meaningful differences by age, number of pregnancies, or number of living children at Wave 1 between women who did and did not report infertility (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Wave 1 characteristics of women by ever experiencing infertility at Wave 1 (pregnancy incidence sample; N = 776)1.

https://doi.org/10.1371/journal.pgph.0001646.t002

Table 3 shows characteristics of women who did and did not report a new pregnancy after Wave 1. Approximately 20% of both women who did and did not report a pregnancy after Wave 1 had ever experienced infertility as reported in Wave 1. Women who became pregnant after Wave 1 were younger than women who did not (26 vs. 28 years as of Wave 1; p<0.001), had experienced fewer pregnancies by Wave 1 (2.6 vs. 3.9; p<0.001), and had fewer living children (2.2 vs. 3.0; p<0.001) at Wave 1. Additionally, women who reported that they wanted a/nother child at Wave 1 were more likely to report becoming pregnant after Wave 1 (77% vs. 59%; p<0.001) (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Wave 1 characteristics of women who reported a new pregnancy after Wave 1 (pregnancy incidence sample; N = 776)1.

https://doi.org/10.1371/journal.pgph.0001646.t003

Next, we present results from logistic regressions estimating odds of reporting a new pregnancy after Wave 1. In the unadjusted model, ever experienced infertility was not associated with reporting a pregnancy after Wave 1 (OR = 0.99; 95% CI: 0.69–1.44), and this relationship was similar in the adjusted model (AOR = 0.89; 95% CI: 0.60–1.34) (Table 4). The 95% confidence intervals for older age groups, number of living children, desire for a/another child, and years in the cohort did not include 1.0 in the unadjusted models (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Odds of reporting a new pregnancy after Wave 1 (N = 776).

https://doi.org/10.1371/journal.pgph.0001646.t004

In the adjusted model, number of living children at Wave 1 was associated with lower odds of reporting a pregnancy (AOR = 0.77; 95% CI: 0.65–0.90). Desire for a/another child at Wave 1 (AOR: 1.59; 95% CI: 1.06–2.39) and years in the cohort (AOR: 1.51; 95% CI: 1.26–1.81) were both associated with higher odds of reporting a pregnancy (Table 4).

In survival analyses examining the relationship between infertility history and time-to-conception during the follow-up period, we find no noteworthy differences between women who had experienced infertility and those who had not, supporting our results from the logistic regression approach (S4 Table and S1 Fig).

Pregnancy intentions/planning

Among the 375 women who became pregnant between Waves 1 and 5, we examined the six items that comprise the LMUP scale individually (S5 Table) and as a composite continuous measure, with higher scores indicating the pregnancy was more intended. Overall, pregnancies in the cohort tended to be intended. The average LMUP score among the sample was 9.13 (median: 10; range: 0–12), and this did not differ between women who experienced infertility and those who did not. When we examined individual items, we found that the majority (91%) of women were not using contraception when they became pregnant (item 1). Approximately 80% of women reported that their pregnancy happened at the right time (item 2), that they intended to get pregnant (item 3), that they wanted to have a baby (item 4), and that they agreed with their partner to have a baby before becoming pregnant (item 5). Finally, 35% of women engaged in a pro-active behavior prior to becoming pregnant (e.g., eating healthily, seeking advice) (item 6). There were no significant differences in any of the individual LMUP items by whether or not a woman reported ever experiencing infertility at Wave 1.

Examining the relationships between participant characteristics and mean LMUP score, we found that women who wanted a/another child at Wave 1 tended to have higher LMUP scores than women who did not report wanting a/another child at Wave 1 (9.4 vs. 8.4; p<0.05) (Table 5). We also found that women who had experienced infertility tended to have slightly lower LMUP scores (i.e., had slightly lower pregnancy intendedness) than women who had not experienced infertility (8.8 vs. 9.2), but this difference was not statistically significant (Table 5). Women in older age groups tended to score slightly lower on the LMUP than younger woman (e.g., the mean LMUP for women aged 25–29 years was 9.7 and the mean for women aged 35–40 years was 8.5), but this difference was also not statistically significant. Given lack of statistical significance and relatively large standard deviations, indicating a lack of meaningful variation in LMUP scores by experienced infertility and most other sociodemographic characteristics, we did not conduct a multivariable linear regression predicting the LMUP score.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Mean LMUP score among women who became pregnant by Wave 5 (N = 375).

https://doi.org/10.1371/journal.pgph.0001646.t005

Discussion

This study found that experiencing infertility, measured here by not becoming pregnant after two years of trying, may not indicate sustained infertility or permanent sterility. Two-thirds of women in this study population became pregnant during the follow up period, regardless of whether they reported at Wave 1 ever having experienced infertility. We also found no meaningful differences in the retrospective consideration of pregnancy intendedness between women who did and did not report ever having experienced infertility at Wave 1.

There are many possible explanations for the lack of difference in subsequent incidence of pregnancy between women who ever experienced infertility and those who had not. First, we consider how those who reported infertility at Wave 1 differed from those who did not report infertility. Notably, the 20% of women who reported infertility at Wave 1 had largely demonstrated their ability to become pregnant, indicated by the fact that nearly all (92%) of the women who reported ever experiencing infertility reported that they had been pregnant at least once. Almost as many (86%) had at least one living child. Neither number of pregnancies nor number of living children at Wave 1 differed meaningfully between those who had ever experienced infertility and those who had not. A limitation of this study is that the measure of infertility does not specify when a woman experienced a period of two years trying to become pregnant without conceiving in that time. It is possible that women experienced infertility prior to their first pregnancy and had since been able to conceive one or more pregnancies, and it is also possible that women experienced infertility when trying to conceive a second or higher order pregnancy, or that they were currently experiencing infertility at the time of the Wave 1 survey. Because this timing was not specified, we could not control for possible differences in reports of a pregnancy by timing or duration of experienced infertility.

Some of our findings may also be driven by the make-up of our sample, as only two women in the analytic sample reported zero pregnancies throughout the observation period. Women who were excluded from the analytic sample because they lacked follow up data were younger, less likely to be married, and more likely to be nulliparous at Wave 1 than women in the analytic sample. It is possible that these women were less likely to become pregnant than those with follow up data, but are not captured because they relocated outside of the UTHA cohort area for marriage, perhaps because of divorce or remarriage, which we know may be a consequences of experiencing infertility [27]. In comparing our sample to the DHS, we found that our sample of women was both younger on average and less likely to be nulliparous compared to the DHS sample of women. The DHS does not include a direct question about infertility. Thus, we cannot make any direct comparisons regarding how infertility may differ in our sample.

Although the two-year definition of infertility used in this study surpasses the threshold required for a clinical diagnosis of infertility and the WHO definition of infertility (one year without conceiving) [28], the data indicate that some women may have experienced infertility episodically. A possible explanation for episodic infertility may be the presence of active sexually transmitted infections (STIs) that are treated before causing clinical Pelvic Inflammatory Disease (PID), or permanent scarring to a woman’s reproductive system [29]. While STI progression may lead to PID and infertility, women with STIs that are treated before clinical PID develops may experience reduced fecundity, or a period where they try unsuccessfully to become pregnant, but are ultimately able to conceive once treated for the STI [30]. Trends in our data support this: at Wave 1, women who reported that they had experienced infertility were more likely to report that they had ever experienced an STI than women who did not report infertility. Integrating a the full spectrum of sexual and reproductive health services, such as HIV/STI prevention, screening, and treatment, along with family planning, infertility, and maternal health, may have a variety of positive impacts on reproductive health and well-being, including infertility prevention [31].

The rate of pregnancy following a report of ever experiencing infertility in our study is similar to what has been found in a study in New Zealand where, among cases of untreated infertility or subfertility in women, approximately half spontaneously conceived within three years [11]. Given the lack of available infertility treatment in Malawi and in similar low resource settings [32–34], it is likely that women who reported infertility and then conceived a pregnancy did so without biomedical treatment for infertility, although we cannot confirm this because use of treatment was not measured.

Infertility is typically experienced at the couple-level, but the measure used in this study does not distinguish between whether a woman was reporting a history of male or female infertility (and, in many cases, she may not know). There are challenges to being able to understand the biological source or cause of infertility within a couple outside of a clinical study setting, but we know that women may be held responsible for infertility due to social and gender norms [35]. A recent review (2020) found that 21% of infertility is attributable to male factors alone and an additional 20% is attributable to both female and male factors in the Africa region [36]. In our study population, the origin of infertility may not be known, but previous research suggests that women and men sometimes attribute infertility to a lack of compatibility between partners [35]. Divorce and re-marriage are relatively common [27]; thus, women may have “resolved” infertility through re-partnering. The present study does not account for possible changes in partnerships during the study period, as these data were not available.

Infertility measures may also be vulnerable to recall bias. Recalling past experiences using a two-year time marker may not be salient in this population. Another study using UTHA cohort data suggested that recalling the timing of specific events may be inaccurate in this low-numeracy population [37]. It is possible that women reported trying for two years to become pregnant, but, in fact, they had tried for a shorter amount of time. Qualitative evidence from this study population indicates that women may consider themselves infertile, and be perceived by others as being infertile, after a much shorter period of time due to social pressures and norms around childbearing [35].

The measure of infertility we used, and, indeed, nearly all self-reported infertility measures, also implies that women will readily differentiate between periods of time when they ‘tried’ to become pregnant and when they did not ‘try.’ We know that pregnancy intentions can change over relatively short periods of time [14] and that ambivalence around pregnancy is common in Malawi [21]; thus, women may have responded based on an overall, generalized assessment of a two-year period in their life, even if they were not trying to get pregnant (or exposed to pregnancy) for the entire duration of two years. Although the infertility question specifies exposure to pregnancy by asking if the woman “tried” to become pregnant, the concept of trying for pregnancy is highly subjective and relies on women accurately recalling and reporting many different components of sexual behavior.

Additionally, contraceptive use is excluded from this analysis because there was no measure of continuous contraceptive use in the data. However, given the commonality of contraceptive use in the cohort overall [38], it is likely that women’s exposure to the possibility of becoming pregnant was less than the time presumed in the analysis (i.e., the length of time a woman persisted in the cohort). Although many methods of contraception are vulnerable to imperfect use or intentionally inconsistent use (e.g., injectables) [39], some women likely used these methods efficaciously or used methods of contraception with close to perfect efficacy (e.g., the implant) [40]. These women may have been almost entirely invulnerable to pregnancy for part of or even for the entirety of the follow up period; however, we are unable to assess this from the available data. Thus, contracepting women are still included in the denominator. To assess the potential effects of sterilization, women who were sterilized after Wave 1 were examined, finding that more than half reported a new pregnancy before they were sterilized, justifying inclusion of women who were sterilized during the follow up period in the denominator.

Given the assumptions and likelihood that not everyone in the denominator was exposed to equal risk of becoming pregnant, it is notable that 65% of women still reported a pregnancy after Wave 1. With better measures of pregnancy exposure (e.g., longitudinal measures of contraceptive use and sexual frequency), it is possible that the pregnancy rate would have been even higher by removing from the denominator women who were not susceptible to pregnancy or adjusting for the length of exposure to the risk of pregnancy. There is also a strong possibility that pregnancies were under-reported in the data due to early miscarriage. Upwards of 20–30% of pregnancies result in a miscarriage [41, 42] and women may have a more difficult time recalling attributes of a pregnancy that did not result in a live birth [43]. It is also possible that some pregnancies were miscarried before the woman knew she was pregnant; these pregnancies would not be reported, contributing to a possible under-estimation of the number of women who became pregnant in the cohort.

Differential exposure to pregnancy due to contraceptive use between the women who had and had not experienced infertility could also potentially contribute to our finding of no difference between reporting a subsequent pregnancy in these groups. However, further analyses comparing desire for a/another child, which may be a proxy for contraceptive non-use, shows that similar proportions of women who had and had not experienced infertility wanted a/another child. This may suggest similar exposure to pregnancy throughout the study period between these groups.

An additional consideration is the definition of infertility that we used–two years without conceiving–which differs from the clinical and WHO standard definition of one year without conceiving [28]. This may have impacted our findings in two ways. First, it could mean that our results are conservative. Some respondents may have not conceived after one year, but conceived between one and two years (and thus, they would have met the clinical definition for infertility, but not be counted in our measure). A second implication of our measure is that it may be subject to greater recall bias than a one-year definition, as it is more difficult to recall a longer period of time [37].

In this study, we found that the LMUP was skewed toward intended pregnancies and women rarely reported the middle response options on the LMUP survey items (e.g., pregnancy happened at not quite the right time). The LMUP findings here differed from a previous application of the LMUP in Malawi, which observed more nuance in the measure [24]. Hall (2013) found that the median LMUP score was 6.0 (range = 0–12) among the 125 women surveyed, which is lower than the median score of 10 in the present study. However, Hall (2013) examined the LMUP among a smaller population of women in Malawi (N = 125) who were, on average, younger than the women in the present study (mean age of 24 vs. 27 years in the present study) and less likely to be married (80% vs. 93% in the present study). In the Malawian context, where childbearing is expected within marriage [27], it is likely that unmarried women have less intended pregnancies. Additionally, 71% of women in the present study reported prospectively that they wanted a/another child at Wave 1, which may contribute to the overall finding of high levels of pregnancy intendedness reported retrospectively.

Supporting information