Moreau, Caroline MD, PhD1,2; Trussell, James PhD2; Gilbert, Fabien MA1; Bajos, Nathalie PhD3; Bouyer, Jean PhD1
About 100 million women worldwide use oral contraceptive pills (OCPs),1,2 which have been the most popular form of reversible contraception used in industrial countries for several decades. Since their first introduction in the early 1960s, concerns over oral contraceptive–related adverse effects have led to substantial changes in their composition. Three approaches have been used to reduce these effects: a gradual decrease in the estrogen dosage, introduction of new progestin components, and new sequences of administration. Today health care providers may choose among a wide variety of OCPs, which differ according to the dosage of estrogen (ranging from 15 to 50 mcg per day), the type of progestin (generally classified in three generations), the sequence of administration (either monophasic or multiphasic), and the number of active pills per cycle. Based on theoretical improvements in safety profiles from reduced estrogen and progestin dosage, providers have increasingly favored the prescription of the lowest estrogen dose formulations (20 mcg and the recently approved 15-mcg pills) combined with third-generation progestins (desogestrel, norgestimate), which are claimed to be better tolerated than the second-generation progestin (levonorgestrel) because of their reduced androgenic effects. However, no clear evidence supports either the choice of third-generation over second-generation pills3 or the use of 15- to 20-mcg pills rather than pills with higher doses of estrogen.4 In fact, a difficult balance between cycle control and adverse effects needs to be struck, because adverse effects are found to decrease pill compliance and to increase the likelihood of early discontinuation of use.5 An additional concern comes from the differences in cost of OCPs, because third-generation pills are generally more expensive than second-generation pills. This financial gap is amplified in France, where third-generation pills, unlike second-generation pills, are not reimbursed by the state health insurance plan.
Most information on OCP adverse effects comes from clinical trials, which compare a limited number of OCP formulations during a short period of use. However, these trials may fail to capture the typical conditions of OCP use and, in particular, how long and consistently the methods are used in the general population.
In this study, we investigate the frequency of several reported symptoms, according to the type of OCP used, among French women. We also examine differences in the reporting of symptoms according to the duration of use of the pill.
PARTICIPANTS AND METHODS
We report here analyses using data from a national cohort study (Cocon survey) exploring contraceptive practices and abortion in France. The methodology of this study has been described in detail elsewhere.6 Here, we will present only the main features.
A two-stage stratified probability sampling method was used to identify 2,863 French-speaking women of reproductive age (18–44 years) living in France. An initial sample of 14,704 French households, including at least one eligible woman (aged 18–44 years), was randomly selected from the telephone directory in 2000. One eligible woman per household was then randomly selected to answer the telephone questionnaire. The response rate among eligible women was 74.6%. The sampling procedure included a second phase designed to overrepresent women who had had an abortion or an unintended pregnancy in the 5 years before the survey. All women who met these criteria were selected (sampling proportion, 100%; n=1,034), whereas only a fraction of the other women were randomly selected (sampling proportion, 19%; n=1,829). Women received no financial compensation. The sample obtained was weighted to reflect the sampling design and the main social demographic composition (age, marital status, professional activity, level of education) of French women in the 1999 census. The percentages shown in the tables are weighted values, but the Ns (the number of women actually interviewed) are unweighted.
After the first interview in 2000 on entry into the cohort, women agreeing to participate were followed up once per year for 4 years (2001–2004) to investigate any changes in contraceptive practices that had occurred since the previous interview. Between 2000 and 2004, the survey sample fell in size by 45% from 2,863 to 1,568. The attrition rate was highest the first year of follow-up (23%) and fell to 10% in the following 3 years. In all, 2,210 women answered the follow-up questionnaire in 2001, 1,911 in 2002, 1,724 in 2003, and 1,568 in 2004. Although this substantial reduction in the sample size affects the precision of the statistical analysis, the attrition of the cohort studied between 2000 and 2002 was not found to suffer from selection bias on the variables of interest (contraceptive histories and current patterns of use).7 In particular, there were no differences in the proportion of women lost to follow-up, either by type of oral contraceptive pills or by frequency of reported symptoms.
The present analysis is based on data collected at each interview from 2000 to 2004. The study sample comprises all current OCP episodes (at the time of each interview) for which women had specified the brand name of their pill. Oral contraceptive pill episodes after a pregnancy or a gynecological procedure (surgery of the breast, fallopian tubes, uterus, or ovaries) in the year before the interview were excluded from the analysis because of the importance of the gynecological symptoms related to these events. Of the initial sample of 2,863 women in the cohort, 1,620 reported using OCPs at least once at the time of the interviews. These women reported a total of 4,267 OCP episodes, of which 4,214 had a specified brand name. We further excluded 523 OCP episodes that occurred in the 12 months after a gynecological procedure or a pregnancy. The final sample comprises 3,691 OCP episodes reported by 1,458 women. The number of episodes per woman ranged from one to five: 29% of women contributed one episode, 23% contributed two, 19% contributed three, 16% contributed four, and the remaining 13% contributed five episodes.
The first telephone questionnaire in 2000 had a mean duration of 40 minutes. Data were collected on the sociodemographic characteristics of the women and their reproductive and contraceptive histories. Several questions related more specifically to the use of OCPs at the time of the survey (brand name, duration of use, compliance with the regimen). The questionnaire also included a health module, specifically exploring gynecological disorders as well as a list of perceived symptoms (including recent weight gain, breast tenderness, irregular menstrual cycle), which have been described in the literature as possible hormonal adverse effects.
The follow-up questionnaires (2001–2004) took a mean duration of 20 minutes to complete. They were designed to document all contraceptive changes over the past 12 months. The questions regarding women's perceived symptoms in the 2000 questionnaire were repeated in each follow-up interview.
By collecting the brand name of the OCPs used at the time of each interview, we were able to identify the progestin component and the estrogen dosage of each OCP episode. Using this information, we defined two classifications of OCPs. The first classification distinguishes eight types of OCP according to the generation of the progestin component, the dosage of estrogen for third-generation pills, and the dosage of progestins for progestin-only pills. Oral contraceptive pills containing the newest progestin component (drospirenone) were excluded from this classification because of an insufficient number of episodes (n=51). The second classification distinguishes OCPs by dosage and sequence of administration of estrogens and by dosage of progestins for progestin-only pills. The distribution of OCPs according to these two classifications is reported in Table 1 and Table 2.
Each year, women were asked to report a list of symptoms by asking them if the following statements applied to their current situation: Have you recently gained weight? Do you often have nausea or want to vomit? Do you often have breast pain? Do you have no more or almost no more periods? Do you often have periods that last more than 1 week or breakthrough bleeding? Do you often have painful periods? Do you often have heavy periods? Do you often have swollen legs? (This last question was added only in the follow-up questionnaires from 2001 to 2004)
The response to each of these questions was used to construct eight binary variables for the analysis. Some of these symptoms (such as bleeding patterns) are clearly related to the intake of OCPs, whereas others are not (weight gain, swollen legs). However, these last symptoms may be perceived as potential OCP adverse effects in the general population and contribute to early pill discontinuation. In this study, we were interested not only in the physiological effects of the different types of OCPs but also in the women's personal experiences using these contraceptives.
We used logistic regression models to assess the odds of reporting a symptom according to the type of OCP used. Given that one woman could contribute to as many as five episodes, we introduced an intraclass correlation term at the individual level. Two series of eight models (one for each symptom) were estimated by using the first and second classification of OCPs. Each model was adjusted for the same set of potential confounders previously described in the literature and found to be associated (with P≤.20) either with the type of OCP or with the frequency of reported symptoms or both. The following confounders, measured at the time of each interview, were included in the models: age (18–24, 25–34, 35 or more years), prior pregnancy (yes/no), body mass index (less than 18.5, 18.5–25, 25 or more), marital status (married/cohabiting/single), professional activity (employed/unemployed/nonactive), level of education (less than high school graduation or high school graduation or more), income (high/medium/low), daily cigarette consumption (0, 1–9, or 10 or more cigarettes), type of regular gynecological follow-up (none/general practitioner/gynecologist), sexually transmitted infection in the last 12 months (yes/no), duration of use of the pill (less than 6 months, 7–12 months, more than 12 months), pill compliance (missed pills in the last month/no missed pills), prior change of pills because of adverse effects (yes/no), and year of interview.
Using the same methodology, we further examined the differences in the frequency of symptoms reported by third-generation compared with second-generation pill users, 20 mcg or less compared with more than 20 mcg third-generation pills, and finally triphasic compared with monophasic pill users among women using 30–40 mcg estrogen. The statistical power to detect such differences depends on the frequency of each symptom. In case of rare events, such as nausea, breakthrough bleeding, or lower frequency of menstrual periods, we were able to detect the following odds ratios (ORs) with a statistical power of 80%: OR less than 0.4 or OR greater than 2.2 in the comparison of third-generation versus second-generation pills, OR less than 0.3 or OR greater than 2.9 in the comparison of 20 mcg or less versus more than 20-mcg pills and monophasic versus triphasic pills. For more frequent events, such as heavy bleeding, painful periods, or swollen legs, these same estimations were OR less than 0.6 or OR more than 1.7 for third-generation versus second-generation pills, OR less than 0.4 or OR more than 2.2 for 20 mcg or less versus more than 20-mcg pills, and OR less than 0.5 or OR more than 2.0 for monophasic versus triphasic pills.
Statistical analyses were performed with Stata SE8 (StataCorp, College Station, TX). The study received the approval of the relevant French government oversight agency (the Commission Nationale de l'Informatique et des Libertés).
On average, each year more than half the women reported at least one symptom from the list of eight symptoms explored. The most common symptoms reported were recent weight gain (25.2%), painful periods (20.7%), swollen legs (20.9%), and heavy menstrual bleeding (15.6%) (Table 3). Painful menstrual periods and breakthrough bleeding or menorrhagia were more frequently reported at the start of pill use (1 year or less) than for longer durations of use (Table 3).
The reporting of several symptoms varied by pill composition. According to the first pill classification (by generation of progestin and estrogen dosage for third-generation pills (Table 4), women using progestin-only pills were more likely than others to report irregular menstrual bleeding (breakthrough bleeding, amenorrhea or reduced frequency of menstrual periods). Likewise, women using an extended regimen of a third-generation pill containing very low estrogen dosage (15 mcg) were more likely than women using pills with higher estrogen dosage to report a reduced frequency of menstrual bleeding. The odds of reporting breakthrough bleeding or heavy menstrual periods were lower among first-generation pill users compared with second-generation pill users. After excluding the 61 women using 50 mcg estrogen pills from the analysis, results remain unchanged with the exception of the effect of first-generation pill use on the frequency of breakthrough bleeding that was no longer significant (OR 0.3, 95% confidence interval [CI] 0.1–1.1], P=.06).
When comparing second and third-generation pill users, we found only one significant difference: a lower frequency of menstrual periods was more likely to be reported by third-generation pill users than second-generation pill users (OR 2.3, 95% CI 1.3–4.1, P=.003). Moreover, a subanalysis among women using the same dosage of estrogen (30 mcg) in which we distinguished the type of third-generation progestin (desogestrel or gestodene) found few differences between groups. Thirty-microgram gestodene pill users were less likely than 30-mcg levonorgestrel pill users to report pain during menstrual periods (OR 0.4, 95% CI 0.1–1.0, P=.04), 30-mcg desogestrel pill users were less likely than 30-mcg levonorgestrel pill users to report a reduction in the frequency of menstrual periods (OR 0.2, 95% CI 0.03–0.6, P=.01), and conversely, more likely to report breast pain (OR 2.5, 95% CI 1.1–5.7, P=.02).
Finally, our analysis shows only one significant difference in the frequency of symptoms with a decrease in the dosage of estrogen for third-generation pill users: 20-mcg or less pills were associated with a lower frequency of menstrual periods compared with higher dosage pills (OR 5.2, 95% CI 2.3–11.4, P<.001).
Despite a smaller sample size (1,173 episodes [32.8% of all episodes] had a duration of use of 1 year or less), we found greater differences in the reporting of symptoms by pill composition for the first year of use than for longer durations of use. Thus, in addition to the differences described above (with the exception of the association between first-generation pill use and breakthrough bleeding, which was no longer significant), recent first-generation pill users were more likely than second-generation pill users to report a decrease in the frequency of menstrual periods (OR 5.1, 95% CI 1.5–16.7) or having swollen legs (OR 7.4, 95% CI 1.7–32.7), and conversely, less likely to report experiencing pain during menstrual periods (OR 0.1, 95% CI 0.01–0.6). In addition, the use of third-generation pills containing 20 mcg estrogen was associated with a higher frequency of nausea (OR 3.6, 95% CI 1.2–10.8) and the use of macro-progestin pills (Luthenyl or Lutheran, Tables 1 and 2) with a lower frequency of breast pain (OR 0.3, 95% CI 0.1–1.0) compared with second-generation pill use.
Using the second classification, by dosage of estrogen and sequence of administration (monophasic, biphasic, or triphasic pills) (Table 5), results from the multivariate analysis show the same effect of low-dose estrogen pills (15 mcg) or progestin-only pills on the reporting of irregular menstrual periods described above. In contrast, women using the highest dosage of estrogen pills (50 mcg) were less likely to report breakthrough bleeding. They were also less likely to report breast pain. In addition, results indicate a higher tendency to report heavy menstrual bleeding among progestin-only minipill users (P=.06) and a lower likelihood of reporting pain during menstrual periods among women using triphasic 30–40 mcg estrogen pills, compared with users of 30–35 mcg estrogen monophasic pills.
Similar differences were found among recent progestin-only pill users and women using low-dose 15 mcg estrogen pills, with the exception that heavy menstrual bleeding was no longer associated with recent use of progestin-only minipills. Likewise, the associations described above for recent 50 mcg estrogen pill users (which only accounted for 18 episodes) and the lower frequency of pain during menstrual periods among triphasic pills users were no longer significant. Conversely, additional differences were found among recent 20 mcg estrogen pill users who were more likely than 30–35 mcg estrogen monophasic pills users to report nausea (OR 4.3, 95% CI 1.4–12.9) and recent weight gain (OR 1.8, 95% CI 1.0–3.0). This association was also found among recent progestin-only minipill users (OR 2.4, 95% CI 1.0–5.7). Thus, with the exception of the lower frequency of pain during menstrual periods, we found no difference in the reporting of symptoms according to the sequence of administration (monophasic, biphasic, or triphasic) among women using 30–40 mcg estrogen pills, who account for 71% of the sample. When we restricted this analysis to women using the same generation of progestin (monophasic versus triphasic third-generation 30–40 mcg estrogen pills), no differences remain (results not shown).
The change of OCP because of adverse effects was only available in the follow-up interviews (2001–2004). In these 4 last years, we found that 10.7% of OCP episodes were preceded by a change in OCP because of adverse effects in the last 12 months. When we restrict the multivariate analysis to the OCP episodes collected at follow-up interviews, we found no effect of the change of OCP because of adverse effects on the frequency of reported symptoms. Likewise, this variable had no effect on the reporting of symptoms among recent OCP episodes, with the exception of nausea, more frequently reported in case of a change of pill because of adverse effects in the preceding year.
This study shows that some symptoms are frequently reported by OCP users: one fourth reported recent weight gain (25.2%), one fifth reported painful periods (20.7%) or swollen legs (20.9%), and one sixth (15.6%) experienced heavy menstrual bleeding. As is commonly described, recent OCP users were significantly more likely to report breakthrough bleeding than long-term OCP users. They were also more likely to report pain during their menstrual cycle. This decrease in the reporting of menstrual symptoms by duration of use may reflect a physiological adaptation to the hormonal regimen and/or the selection of women who experience the least symptoms over time. The frequency of other symptoms did not vary by duration of OCP use.
The questions on perceived symptoms were not specifically related to the use of OCPs. Thus, these results should not be interpreted as a measure of the prevalence of OCP adverse effects. However, they do capture the differences in the frequency of perceived symptoms according to the type of OCP used, after controlling for a large number of potential confounders. One limitation of this comparison is the statistical power of the analysis, which is generally low in the case of rare events (nausea, breakthrough bleeding, or lower frequency of menstrual periods). This is particularly acute in the analysis of different estrogen dosages in third-generation pills.
Nevertheless, our study shows some variation in the symptoms reported among a wide variety of types of OCPs. Our results confirm those of previous clinical trials showing an increase the likelihood of abnormal bleeding patterns among progestin-only pill users,8 which persists over time. Although these bleeding irregularities do not threaten health, they are the most common reported reason for discontinuation, thus relegating these pills to a second-line contraceptive option.3
Our analysis shows a reduction in the frequency of menstrual periods among women using extended or continuous regimens (which also happen to be those who use progestin-only or very low-dose estrogen pills), compared with classic cyclic regimens. These findings add to the growing body of literature on continuous OCP regimens that are gaining wider popularity in developed countries as a means to avoid menstruation for medical or personal reasons.9,10,11 Further contraceptive benefits of continuous administration on menstruation-related symptoms (nausea and breast tenderness), discussed in a recent review of the literature,11 could not be examined in our study because the only extended pill regimens available in our survey have lower estrogen dosage than do the cyclic pills. Other variations in the sequence of administration of OCPs (monophasic or triphasic) had little impact on the frequency of symptoms reported. Thus, consistent with the conclusion of a recent Cochrane review,12 our results offer no evidence of an improvement in the bleeding patterns of women using triphasic pills compared with those of monophasic pill users.
We further found no evidence to suggest a decrease in the reporting of symptoms with decreasing estrogen dosages in women using combined OCPs. Comparing 20 mcg or less pills with higher estrogen dosage third-generation pills, we found only an increase in the reporting of lower frequency of periods among women using the lowest estrogen pill regimens. Although insufficient statistical power may affect these conclusions, they are nevertheless consistent with the conclusions of another recent review of the literature comparing 20 mcg or less pills with higher estrogen dosage pills.4 In France, as in many other industrialized countries, third-generation pills are increasingly favored over second-generation pills13 on the basis of the two following arguments: the reduction in the risk of estrogen adverse events (although 40% of third-generation pills prescribed in France carry the same dose of estrogens as in second-generation pills) and the improvement of progestogenic and androgenic proprieties theoretically resulting in better tolerance. Our findings cast doubt on this second hypothesis because we found no decrease in the frequency of reported symptoms between third- and second-generation pills in general and very few differences if we select only women using the same estrogen dosage and distinguish the type of progestin component in third-generation pills.
In our survey, menstrual nuisances such as painful or heavy periods were not uncommon, even among long-term third-generation pill users (14.6% reported heavy bleeding, 17.6% painful periods, and 7.8% lower frequency of periods after 1 year of method use). However, unlike in randomized trials, we cannot rule out the possibility of selective prescribing in our study. Indeed, because third-generation pills are promoted as being better-tolerated than second-generation pills, women who experience the most hormone-related symptoms may be more likely to be prescribed a third-generation OCP in the first place or to switch from second- to third-generation because of effects.14 This later selection is partially mitigated in our analysis by controlling for a change of pill because of adverse effects in the year preceding the interview. That change had no effect on the frequency of reported symptoms in the multivariate models, which suggests that the overall selection bias due to OCP switching is limited in our survey. Nonetheless, the Cocon survey, based on women's reports, cannot capture the reasoning that governs the type of OCP prescribed, and future research should also focus on physicians' prescribing attitudes. The absence of an overall improvement in the reporting of symptoms between second- and third-generation pill users is consistent with the conclusions of the recent Cochrane review, which indicated similar acceptability, reasons for discontinuation, and overall adverse effects between second- and third-generation pills, which were both preferred to first-generation pills. In one of the trials included in this review, however, fewer women reported intermenstrual bleeding when using gestodene compared with levonorgestrel.15 We found no such association in our survey, but found that women using 30 mcg gestodene pills reported less pain during their menstrual period than women using 30 mcg levonorgestrel pills. As highlighted in the conclusion of the Cochrane review, future research should focus on well-designed randomized trials to further our understanding of the independent effect of the type of progestin on OCP tolerance. Such an understanding of an effect, if any, of the combination of estrogen dosage, progestin component, and sequence of administration is essential to guide physicians' choice of the type of OCPs that best suits a woman's needs. However, beyond the choice of the right OCP formulation that theoretically decreases the likelihood of very rare and severe medical outcomes associated with low-dose estrogen pills, providers should also recognize the importance of counseling about OCP-related symptoms. A recent study among women using implants showed that those who anticipated possible adverse effects reported lesser bleeding irregularities than those who had not been counseled.16 Likewise, in a prospective study among 943 new OCP users in the United States, Rosenberg and colleagues17 found an association between the degree of satisfaction with the pill and higher ratings of patient-provider interaction.
In the absence of sufficient evidence-based data to support the existence of differences in the tolerance profile of low-dose combined pills currently available, future well-designed randomized trials are needed to guide providers in their choice of OCPs. Research is also needed to assess the effectiveness of counseling on the tolerance of OCPs, an intervention that may prove to be more rewarding than basing the choice of OCPs on their theoretical properties.