Contraceptive discontinuation is a significant cause of unplanned pregnancy. Oral contraceptives are central to any discussion of contraceptive discontinuation, since more women use the pill than any other reversible method,1 and because pill discontinuation rates are higher than those for any other female method with similar efficacy.2 Estimates of oral contraceptive discontinuation vary widely (30–60% at one year), but levels are consistently found to be elevated among young and socioeconomically disadvantaged women.3-6 The fertility-related costs of discontinuation are clear: In a study of 1,657 oral contraceptive adopters, more than 80% of those who discontinued the method and were potentially at risk of unintended pregnancy either failed to adopt another method or switched to a less effective one.7 An estimated 20% of annual unintended pregnancies in the United States are due to discontinuation of oral contraceptives.8
One main reason that women give for discontinuing pill use is method side effects, including bleeding irregularities, weight gain, mood changes and headaches.7,9 In an analysis of data from the 2002 National Survey of Family Growth, Moreau, Cleland and Trussell found that 29% of oral contraceptive users discontinued the pill because of method dissatisfaction, and 65% reported side effects as a reason for dissatisfaction.9 Other reasons are method-related logistics (women report difficulty using, obtaining or paying for pills), lack of immediate need for contraception (women are no longer having sex or they desire pregnancy) and lack of partner satisfaction with pill use.7,9
Although the effects of method-related factors on oral contraceptive discontinuation within a current interval of use are well documented,9 little information is available about whether women’s oral contraceptive experiences affect their experiences and continuation in subsequent intervals of use. This question is important because some women who discontinue the pill use it again at a later time.4,6,7,10
In this exploratory study, we empirically address this issue, using a sample of low-income, minority women who were restarting pill use. We examine the association between pill restarters’ main reason for prior discontinuation and their experiences on the pill, feelings about the method as they resume use and early experiences on it once they restart. We focus on early experiences in the current interval because early discontinuation is common and often is attributable to method-related factors.7
We distinguish among three primary reasons for prior oral contraceptive discontinuation: side effects, logistical reasons and fertility-related reasons. Side effects are negative changes to one’s body, mood or health associated with method use. Logistical reasons reflect the interplay between attributes of the pill and users’ life circumstances or capacities; they include having trouble remembering to take the pill, running out of pills, being fearful that someone will find the pills and having no money to purchase pills. Fertility-related reasons reflect that users no longer need the pill because they are or want to become pregnant, or because they are not having sex. Side effects are the reasons most closely tied to method attributes; fertility-related reasons, least closely.
We hypothesize that the more closely tied to method attributes a woman’s reason for previous pill discontinuation was, the more likely she will be to report negative experiences with the method during past use and after restarting. In short, we expect the most negative experiences and shortest durations of use among those who discontinued because of side effects, followed by those who stopped for logistical reasons and then women who discontinued for fertility reasons.
Data for this study were collected as part of a multicenter randomized trial of a novel method of initiating oral contraceptives. The trial was conducted between March 2003 and February 2005 in family planning and adolescent clinics at three university centers: Emory University in Atlanta, Mount Sinai Medical Center in New York City and the University of Texas Southwestern Medical Center in Dallas.
Young women seeking oral contraceptives at the recruitment clinics were the target population. After completion of clinical care, clinicians referred women who requested oral contraceptives to study staff for eligibility screening. Women were eligible for the study if they were younger than 25, intended to have sex in the next 30 days, had a negative pregnancy test on the day of study enrollment, had no medical contraindications for pill use, had not used depot medroxyprogesterone acetate in the past 12 months and did not desire a pregnancy within the next six months. After giving written informed consent, participants completed a baseline interview that covered their demographic characteristics; sexual, contraceptive and reproductive history; and contact information. Follow-up interviews were conducted by telephone at three and six months, and included specific questions regarding oral contraceptive continuation and reasons for discontinuation. Our analyses are based on the baseline and three-month follow-up data.
Institutional review boards of the clinic centers and Columbia University (the coordinating center) reviewed and approved the study. Details about the larger study’s methodology and sample are available elsewhere.11
•Prior pill experiences
At the baseline interview, all previous oral contraceptive users were asked why they had ended their last use interval. Interviewers entered their open-ended responses into the following precoded categories: side effects, logistical reasons, fertility reasons, partner-related reasons, fear of the pill and other reasons. Women who reported more than one reason were asked for the main reason they had stopped taking the pill. For the bivariate analyses, the main reason was recoded into a trichotomous variable distinguishing among side effects, logistical reasons and fertility-related reasons. Eight women did not report any of these reasons and were omitted from these analyses. For the multivariate analysis, we created a summary measure that distinguished between women who mentioned side effects as a reason for discontinuation (whether it was their main reason or not) and those who did not cite this as a reason.
It is worth noting that among the fertility-related reasons for pill discontinuation, pregnancy represents a very different scenario from desire to become pregnant or lack of partner. Women who became pregnant while using the pill experienced a method failure, whereas women citing these other fertility-related reasons did not. We could not omit women who experienced method failures from the analysis, because we were unable to distinguish which specific fertility-related reason a woman offered. However, oral contraceptive failure rates are low (0.1–1.25% with perfect use and 3–5% with typical use12); thus, this is unlikely to have any substantive effect on the analysis.
We used a Likert-scale question to assess overall satisfaction with the pill; possible responses (coded 1–4) were “very unsatisfied,”“somewhat unsatisfied,”“somewhat satisfied” and “very satisfied.” We also created a dichotomous measure summarizing whether women were satisfied with the pill.
•Experiences with restarting
Level of commitment to restarting the pill was measured at the baseline interview, before participants were given the pill. Women were asked if they had been considering the pill when they came to the clinic (yes=1, no=0), if they had been considering any other method when they came to the clinic (yes=1, no=0) and how sure they had been that they wanted the pill when they came to the clinic (1=very sure, 0=somewhat sure or not sure at all).
Current pill satisfaction was measured at the three-month follow-up interview using the same question and measures used for prior pill satisfaction. This variable was measured only for women who actually took the pill during the study period.
Continuation at the three-month interview was based on questions that probed whether women had started the pill, how many pill packs they had completed, whether they had taken the pill in the previous week (unless it was the last week of the cycle) and whether they had used the pill for the entire three-month period. The variable was coded 1 for women who had never started the pill or who had completed less than one pack (short-term users), 2 for women who had completed more than one pack but were not using the pill in the week before the interview (intermediate-term users), and 3 for women who had used the pill continuously through the interval (continuous users). We could not construct a continuous measure of time to discontinuation because of extensive problems with data on the date of discontinuation.
To assess side effects, we asked women whether they had experienced changes in each of four areas since starting the pill (headaches, weight, mood and sexual satisfaction), and whether the changes were good, neutral or bad. Our variable indicating side effects in the current interval was a scale that summed the number of negative changes that women experienced.
Main reason for pill discontinuation was measured for women who stopped using oral contraceptives before the three-month interview, and was coded as side effects, logistical reasons or fertility-related reasons.
Age was measured as a continuous variable. Race and ethnicity was measured as a trichotomous variable distinguishing among Latinas, blacks and women of other races. Women were first asked if they were Latina, and then all non-Latinas were asked with which racial group they most identified (black, white, Asian or another group). Because the vast majority of women in the sample (94%) were Latina or black, we combined the women of other races into one residual category.
A dichotomous measure of economic hardship was used to assess whether a respondent had had any of the following experiences in the past six months: been on public assistance, gone to bed hungry, not had enough money to live comfortably or been unable to get money she needed for an emergency.
Given the relatively young age of the sample, parity was coded as a trichotomous variable distinguishing among women who had had no, one, or two or more births. A woman’s prior success at controlling her fertility was measured by the proportion of her prior pregnancies that were unplanned (as calculated from her reports of her number of pregnancies and number of unplanned pregnancies).
Bivariate analyses used cross-tabulations and chi-square statistics or one-way analysis of variance, depending on whether the outcome measure was categorical or continuous. Multivariate models of continuation at three months were estimated with multinomial logistic regression, since the early discontinuation outcome variable was trichotomous.
Women in the clinical trial sample were young and socioeconomically disadvantaged (Table 1). Their mean age was 19, and 94% were black or Latina; 54% had experienced some form of economic hardship in the past six months. Seventy-eight percent had at least one child. The level of contraceptive need was high; 63% of those who had previously been pregnant reported that all of their prior pregnancies were unplanned.
Of the 1,716 women in the clinical trial sample who were initiating oral contraceptive use, 695 had used the method previously. Pill restarters and new pill users did not differ on race and ethnicity or on the proportion of prior pregnancies that had been unplanned, but they differed on the other demographic characteristics. Restarters were somewhat older and were more likely to have experienced instances of economic hardship in the past six months. The relationship between prior pill use and parity was curvilinear. Pill restarters were more likely than new starters to have had either no births or two or more births, and were substantially less likely to have had one birth.
In all, 687 pill restarters provided a main reason for their prior discontinuation; of these, 70% cited method-related reasons—17% side effects and 53% logistical problems (Table 2). Twenty-two percent cited fertility-related reasons, and the remainder cited other reasons. Interestingly, the rank ordering of reasons was the same when women were allowed to offer multiple reasons as when they were asked to identify their main reason for prior pill discontinuation. In both cases, logistical factors were the most common, fertility-related reasons were next and side effects followed closely.
Some 628 pill restarters said that their main reason for discontinuation in the prior interval was side effects, logistics or fertility-related factors; of these, women citing side effects reported the most negative experiences with the method (Table 3). Post-hoc Scheffe comparisons indicate that women reporting side effects had the lowest mean level of satisfaction with the pill during their prior interval (2.4, compared with 3.7–3.8) and were the least likely to say that they had been satisfied with the method (49%, compared with 95–98%).
Women who had previously discontinued oral contraceptive use because of side effects also were less firmly committed to choosing the pill at baseline than were other restarters. On the day they came to the clinic, these women were the least likely to have been considering the pill (81% vs. 92–95%) or to have been sure that they wanted the pill (73% vs. 88–94%), and they were the most likely to have been considering another method (33%, compared with 19% of those in each of the other groups).
The number of side effects in the current use interval was marginally higher among women who had previously discontinued the pill because of side effects than among those who had stopped for other reasons. Current satisfaction with the pill was significantly lower in this group than in the others (mean, 3.1, compared with 3.5–3.6). Likewise, the proportion who were satisfied with the method was lowest among those who had previously discontinued use because of side effects (76% vs. 89–93%).
Pill continuation at the three-month follow-up was lowest among those who had previously discontinued because of side effects (52%, compared with 65–70% among those citing another main reason for discontinuation), and the proportion reporting the shortest duration of use (less than one cycle) was highest in this group (17%, compared with 6–8%).
In a final set of bivariate analyses, we examined the relationship between the main reason for pill discontinuation in the prior and current use intervals among the 215 women who had not used the pill continuously during the study period (not shown). Because only 22 women had discontinued for fertility-related reasons within the first three months of current use, we collapsed the main reason variable to distinguish between those discontinuing because of side effects (81) and those doing so for logistical or fertility-related reasons (108). The findings do not indicate stability in the main reason for stopping pill use. Women who had ended their prior use interval because of side effects were no more likely than others to say that side effects were the main reason they ended this one (57% vs. 46%).
Because the bivariate analyses indicated that current pill experiences were similar for women who had previously discontinued for logistical reasons and those who had stopped for fertility-related reasons, our multinomial logistic regression models used the dichotomous measure of prior reason for discontinuation, distinguishing between side effects and all other reasons. Age, race and ethnicity, economic hardship and parity were included because they were associated with the main reason for discontinuation in the previous use interval (not shown). Furthermore, because the bivariate analyses suggested that discontinuation is related to satisfaction with the pill in the prior interval and to number of side effects in the current interval, we estimated one model that added these variables. The models were estimated on oral contraceptive continuation in the three- and the six-month follow-up intervals, and the results were the same for both. For the sake of parsimony, we present the three-month data only.
In the model including only the demographic covariates, previous discontinuation of pill use due to side effects was predictive of early discontinuation in the current use interval (Table 4, page 142). Compared with other women restarting pill use, those who had stopped because of side effects were more likely to be short-term, rather than continuous, users (odds ratio, 3.8). They also were more likely to discontinue at an intermediate duration of use than they were to use continuously (1.9). Finally, they were more likely to be short-term than intermediate-term users (2.0).
In the model that included prior method satisfaction and current side effects, women who had previously discontinued oral contraceptive use because of side effects remained significantly more likely than other restarters to be in the shortest use interval rather than the longest one (odds ratio, 4.5). They also were more likely to be in the shortest interval rather than the intermediate interval (2.5) However, they were only marginally more likely than other restarters to be in the intermediate range. The number of current side effects was positively associated with short-term rather than continuous use (1.6) and with an intermediate duration rather than continuous use (1.9), but not with short-term rather than intermediate use.
Despite the pill’s high level of effectiveness, popularity and ease of use, high rates of rapid discontinuation have been an ongoing reality of its use. Because women can discontinue oral contraceptive use without a visit to a provider who can help them to identify an alternative contraceptive method, discontinuation poses particular challenges for the reduction of unintended pregnancy rates. Women who stop using the pill do not necessarily switch to another method in a timely fashion, and this provides a window for unintended pregnancies. As a result, any information that would help providers identify women who may be candidates for early pill discontinuation is important.
This study explored one potential “risk” factor for early pill discontinuation: women’s prior experiences on the pill. From a counseling perspective, this factor is important because it is relevant for large numbers of pill starters and is relatively easy to ask women about. Our results confirm that women who had previously discontinued pill use because of side effects were at higher risk for dissatisfaction and early discontinuation in a subsequent pill use interval than those who had previously discontinued for logistical or fertility-related reasons. Before resuming pill use, these women may well benefit from extra counseling both about the full spectrum of contraceptive options and about the management of method side effects.
We had expected both women who had discontinued the pill because of side effects and those who had stopped for logistical reasons to be at higher risk for method dissatisfaction and early discontinuation in a subsequent use interval than those who had stopped for fertility-related reasons. Our expectations held for the first group, but not for the second. Perhaps this reflects that as women’s life situations change, the logistical requirements of oral contraceptive use may become less problematic for them. For example, a woman’s ability to pay for the pill can vary with changes in her economic and insurance status; a woman can decide to build in strategies to help her remember to take a pill each day or to return to the clinic for a new prescription.
We add one important caveat to these findings. Some studies have raised questions about the meaning of women’s reports that side effects were the main reason for their contraceptive discontinuation. Kalmuss and colleagues found that while menstrual side effects were the main reason that women gave for discontinuing the contraceptive implant within six months of insertion, those who had had the implant removed were no more likely to report any of the assessed menstrual irregularities than were continuing users.13 Westhoff and colleagues found that while women who discontinued oral contraceptive use within six months of initiation were more likely to report negative physical changes in that interval, they were no more likely to report that those changes were caused by the pill than by other factors.14 These data challenge a straightforward interpretation of women’s reports that side effects were their main reason for contraceptive discontinuation.
Examination of the actual role of method side effects in oral contraceptive discontinuation awaits further study. This was a secondary analysis; the data we used were not collected to address the current research question. While some key variables could be assessed through the available data, others could not. For example, we did not have a side effect profile from women’s last interval of pill use. We also did not have pill restarters’ direct comparisons of their current and previous pill experiences and satisfaction. Further research is needed that would include these variables over time.
These findings have concrete and easily implemented programmatic implications that may help to reduce early pill discontinuation and unintended pregnancies among women who stop using the method. Specifically, routine counseling with women who want to start oral contraceptives should probe their experiences with the pill and their reasons for prior discontinuation. Women citing side effects should be offered individualized counseling to explore and address their risk factors for early discontinuation, as well as information about method choices that may pose less of a risk of side effects.