There is no consensus among practitioners on recommendations for using a back-up contraceptive method when oral contraceptives (OC) are initiated. Most recommendations on the initial efficacy of the OC formulation use start day as a determinant. A 1993 analysis of OC package inserts from nine manufacturers describes three start days for the first pack of pills: day 1 of menses, day 5 of menses or the Sunday after the start of the menstrual cycle.1 However, current package inserts offer a choice between a day 1 start and a Sunday start. Women are advised to use no back-up contraception with a day 1 start and back-up for 7 days with a Sunday start. Conflicting opinions and advice are potentially confusing to both clinicians and OC users. Of new users of OC, an estimated 6.9% experience failure and 32% discontinue use for method-related reasons in the first year of use.2 Inconsistent instructions might contribute to poor compliance, as patients may perceive that correct and effective pill-taking is complicated. In addition, it is unclear whether women who opt for a Sunday start require use of a back-up method.
Although multiple studies have analyzed ovarian activity and ovulation rates during routine contraceptive use3,4 and with missed pills,5 OC action in the initial cycle of use is not well characterized. Delaying OC initiation may place patients at higher risk for pregnancy. Oral contraceptives started on day 1 and day 6 of the menstrual cycle in female monkeys indicates that selection of a dominant follicle can occur with delay of initiation to day 6.6 A nonrandomized study indicated that 1 of 11 (9%) women starting OC on day 1 of the menstrual cycle had a follicle larger than 10 mm compared with 7 of 11 (64%) women starting on day 5 of the menstrual cycle.7 Importantly, in view of patient pill-taking errors, it is unknown if a delayed start can contribute to decreased efficacy and justifies use of a back-up method.
We assessed ovarian follicle formation and subsequent ovulation in women starting OC use on menstrual cycle day 1, 4, or 7. We planned to characterize the way in which OCs affect ovulation in the first month of use to develop consistent use instructions and improve patient compliance.
MATERIALS AND METHODS
We performed a prospective single-center randomized single-masked study in the Clinical Research Center of the Magee-Womens Research Institute between December 1997 and April 2000. The study protocol was approved by the institutional review board of Magee-Womens Hospital, and informed consent was obtained from all volunteers before entry into the study.
Women 18 to 38 years of age who wished to initiate OC use were eligible for enrollment if they had regular menstrual cycles (every 26 to 30 days), were within 35% of ideal body weight, had both ovaries visible on vaginal ultrasonography, and were willing to be abstinent or use a barrier method of contraception during the study. If a participant had a previous pregnancy, she must have had one spontaneous menses and be at least 60 days from the end of the pregnancy on the day of study drug administration. Women were excluded if they had: a history of or current thrombophlebitis or thromboembolic disorder; a known or suspected clotting disorder; cerebrovascular or coronary artery disease; myocardial infarction; breast cancer; known or suspected estrogen-dependent neoplasia; undiagnosed abnormal genital bleeding; cholestatic jaundice of pregnancy or jaundice with previous OC use; hepatic adenoma or carcinoma; known neuro-vascular lesion of the eye or serious visual disturbance; hypertension at baseline examination; migraine headaches; recent Papanicolaou smear with high-grade squamous intraepithelial lesion; a history of polycystic ovary syndrome; current use of rifampin, ampicillin, tetracycline, griseofulvin, carbamazepine, phenytoin, phenobarbitol, hypnotics, or any other drug that may affect oral contraceptive metabolism; use of an injectable contraceptive (ie, depomedroxyprogesterone acetate) within 12 months of the initial study visit; use of contraceptive implants or OC pills within 60 days of study, current breast-feeding; previous oophorectomy; known hypersensitivity to estrogens or progestins; serious adverse experience with previous OC use; malabsorption due to medical condition or surgery; known or suspected drug or alcohol abuse; use of drugs mandating in the labeling simultaneous use of effective contraception; or age 35 years or older and smoking.
At the first of five visits, medical and menstrual histories were obtained, and physical examination and a high-sensitivity urine pregnancy test were performed. Volunteers were randomly assigned to begin the OC on menstrual cycle day 1 (group 1), day 4 (group 2), or day 7 (group 3). Each oral contraceptive package contained 21 days of ethinyl estradiol (30 μg)/norgestrel (300 μg) and 7 days of inert pills (Lo/Ovral; Wyeth-Ayerst Laboratories, St. Davids, PA). Randomization was performed according to random permuted blocks of nine by using a random-number table to assure equal groups. The study group was written on a card and sealed in consecutively numbered envelopes. A person unrelated to the study performed the randomization sequence and prepared the envelopes. The group was assigned by opening the next sequentially numbered sealed opaque envelope.
Women called the investigators at the onset of their menses and were reminded when to start taking their OC. They returned for evaluation within 3 days of menstrual cycle days 7, 14, 21, and 28. Follicular development was measured by using vaginal ultrasonography (GE RT3200 Advantage II or GE Pro Series; General Electric, Milwaukee, WI). All ultrasonographic examinations were performed by one of the three physician-authors (JLS, MDC, or HCP). Although intersonographer or intrasonographer variation was not measured, two of the physicians were present for at least the first 10 evaluations by each physician. The physicians were blinded with respect to start day and were not allowed to review the participant's chart in its entirety until the completion of the study. Blinding was further ensured by providing the physicians with a data sheet to complete on each participant that was physically separate from her chart at the time of the visit. In addition, participants were reminded at each visit, before meeting with the physician, not to tell the physician any information about start day. Maximum follicular diameter was assessed at each visit; each follicle 5 mm or larger was measured in two perpendicular planes. The larger of these two diameters was used as the maximum follicle size for the observation.
Serum progesterone levels were measured on menstrual cycle days 21 and 28. Evidence of ovulation was defined as a serum progesterone level greater than 3 ng/mL. Serum progesterone levels were determined in duplicate by using a radioimmunoassay (Diagnostic Products Corporation, Los Angeles, CA) with a minimal detection level of 0.06 nmol/L. Samples were batched and analyzed in a single run. The coefficient of variation averaged 2.3% across the entire assay range, with values of 1.9% at a concentration of 75 nmol/L and 3.0% at 5 nmol/L.
All participants were asked to keep daily diaries of pill compliance, bleeding, adverse events, and concomitant medications and to return the empty pill pack. Pill-taking errors were considered minor if the participant missed one pill and took it the next day, if she missed placebo pills, or if she started the regimen one day late. Major errors included missing more than one consecutive pill or incorrectly taking placebo pills instead of active pills.
Sample size estimates were based on published data suggesting that if women are instructed to improperly take oral contraceptives in the first 5–17 days of the cycle, break-through ovulation rates are 29%.5 In contrast, ovulation rates in users of monophasic pills have been calculated to be as low as 0%.4 Thus, we assumed ovulation rates of 0%, 15%, and 30% in groups 1, 2, and 3 respectively. Sample size calculations were made by using the study planning software nQuery Advisor Release 4.0 (Statistical Solutions; Saugus, MA) and were based on a χ2 test of equal proportions in three groups. At a 0.05 level of significance (two-tailed) and power of 80%, 84 total participants were required to complete the study (defined as having at least three ultrasonographic evaluations). We initially estimated that approximately 10% of women would not complete the study and calculated an enrollment of 91 women. However, because of the higher than expected attrition rate, we amended the protocol to increase the total number of women enrolled to 130.
The primary outcome measures were maximum follicle-like structures and ovulation. The groups were compared according to median maximum follicle diameter and percentages of maximum follicle diameters larger than 13 mm; these cut-off values were selected because follicle size must generally exceed 13 mm before ovulation can occur.8 Ovulation rates and the grouped percentages of maximum follicle size larger than 13 mm were compared by using a χ2 test for linear trend. Median follicle size was compared among groups by using the Kruskal–Wallis test, since the follicle sizes on days 7, 14, 21, and 28 were not normally distributed. All statistical analyses were performed by using SPSS statistical software, release 10.0 (Chicago, IL).
One hundred thirty healthy women participated in the study (Table 1). Eighty-five women who met study criteria and underwent at least three ultrasonographic examinations were included in the analysis. Inclusion of the available data for the three women who were lost to follow-up and the three women who missed three or more pills did not change any outcomes. The women in each group were similar in all demographic characteristics and tended to be in their early twenties, white, and nulliparous (Table 2). Pill-taking errors in the 85 participants were minor (Table 3).
The median maximum follicle size was 9.0 mm, 9.0 mm, and 13.0 mm for start day 1, 4, and 7, respectively (P < .001). A follicle-like structure that reached a maximum diameter larger than 13 mm was observed in three of 29 (10.3%), five of 29 (17.2%), and 12 of 27 (44.4%) women in groups 1, 2, and 3, respectively (P = .003). When we compared the medians of the groups for follicle measurements on day 7, 14, 21, and 28, we observed a significant difference in increasing follicle size on day 7 (6.5 mm, 7.0 mm, and 11.0 mm) and day 14 (6.0 mm, 6.0 mm, and 9.0 mm) of the cycle in groups 1, 2, and 3, respectively. Although the comparison among the three groups on day 28 was statistically significant, this difference is likely to not be clinically significant because most of the follicles were smaller than 10 mm (Figure 1).
Since more subjects from group 3 made minor pill-taking errors (Table 3), we reanalyzed the data without the 13 participants who started the regimen 1 day late or missed pills. The median maximum follicle size was 9.0 mm, 9.0 mm, and 12.0 mm for start days 1, 4, and 7, respectively (P = .001). In addition, a follicle-like structure that reached a maximum diameter larger than 13 mm was observed in three of 26 (11.5%), four of 26 (15.4%), and 7 of 20 (35.0%) women in groups 1, 2 and 3, respectively (P = .055). Although these groups are small, the lack of statistical significance at a P value of .055 suggests that even without the minor pill-taking errors (which were more common in group 3) the difference in follicular development is still significant.
Of the 20 women who developed follicle-like structures larger than 13 mm, those in group 2 or group 3 were not more likely than those in group 1 to have developed a follicle of this size earlier (P = .7). Similarly, if women who started OC use on day 1 were compared to a combined population of groups 2 and 3 (all women who started the pill after day 1), there was still no difference in how early a follicle-like structure larger than 13 mm developed. The lack of a statistical difference is most likely because relatively few women in group 1 (one on day 7, one on day 14, and one on day 21) developed follicles larger than 13 mm. In comparison, 10 of the 17 (58.8%) women in groups 2 and 3 who developed follicles of this size had done so by day 7.
All 85 participants had at least one progesterone level measured. Five values were missing: one day 21 and one day 28 level for one woman each in groups 1 and 3, and one day 21 value in group 2. Two, one, and zero women in groups 1, 2, and 3, respectively, had evidence of ovulation on the basis of serum progesterone level (P = .2) These three participants had follicle-like structures larger than 13 mm and had diary cards and empty pill packs that indicated correct OC use. Of the two women in group 1 who ovulated, one had follicle-like structures larger than 13 mm by day 14 and one had such structures day 21 (Figure 2). In both women, the follicle-like structure continued to grow from day 14 to day 21 and decreased slightly in size by day 28. Progesterone levels in both of these women were less than 3 ng/mL on day 21 and greater than 3 ng/mL on day 28. The woman in group 2 who ovulated had a follicle-like structure of 28 mm diameter on day 14 followed by a progesterone level greater than 3 ng/mL on day 21.
Despite high efficacy, OCs do not completely inhibit ovarian follicle development.9,10 In theory, delaying initiation of the first OC tablet from day 1 to day 7 may place patients at higher risk for pregnancy. We performed this study because clinicians have made recommendations to use back-up methods in the first month of OC use based on speculation that women may be more vulnerable to contraceptive failure in the first cycle of use. Ovarian activity in the first cycle of use may be more concerning than in future cycles, and OC failure may be closely linked to delaying start of OC use.
We hypothesized that ovulation and possible pregnancy can occur more readily when there is previous follicular recruitment and residual ovarian activity, as with new starters of OCs. Increasing the pill-free interval from 7 to 11 days in women taking OCs has been shown to allow more follicular development and estradiol production without leading to method failure.11 In addition, missing pills in specific sequences of four consecutive pills does not lead to ovarian follicle formation and ovulation.12 However, delaying the initiation of the first OC package up to 7 days differs from extending the pill-free interval and missing pill sequences. In the former case, there is residual ovarian activity because follicular recruitment begins in the previous cycle, and delaying a new start may create a dominant follicle capable of ovulation.
In our study, ovarian follicles appeared to be quiescent by day 21, even with a day 7 start. Therefore, given that pill-taking errors occur, it would be sufficient to use a back-up method for 2 weeks. However, two out of the three women with characteristics indicating ovulation started OCs on day 1 of their menstrual cycle and demonstrated maximum follicle size at day 21 (Figure 2). Therefore, in this scenario, use of a back-up method for 2–3 weeks may offer women a contraceptive advantage. Alternatively, despite the difference in ovarian activity, there was no significant difference in ovulation; thus, use of a back-up method could be argued to confer no added contraceptive benefit to one group over another.
Although this study indicates that there is significantly more ovarian follicle formation in women who delay the start of OC use until day 7, it did not prove that ovulation is increased in these women. Unexpectedly, two (6.9%) women with a day 1 start and one (3.6%) with a day 4 start had progesterone levels consistent with ovulation. Chowdhury et al5 demonstrated that women who miss up to two consecutive days in any part of the pill cycle may have an escape ovulation compared with perfectly compliant users, whereas the cervical mucus remained thick and unfavorable to sperm in all participants. With this in mind, other mechanisms of OC action ensure that ovulation alone does not lead to pregnancy in a compliant pill taker. However, these findings suggest that women who use OC imperfectly and also delay the start of OC may have decreased efficacy. Although no studies have evaluated compliance in the first cycle of OC use, studies using electronic pill dispensers have shown that two to three times more pills were forgotten than women reported.13 Of the 85 women in our study, 12% of women reported missing one pill and 3% of women reported starting their pills one day later than they were told (Table 3).
We found no difference in the incidence of ovulation when OC start was delayed to day 7. This finding does not support recommending use of a back-up method in the first cycle of OC use for women who start on day 1 through day 7. However, this study may not have enough power to detect a small difference; we would recommend a larger study before changing clinical practice. We believe that our findings justify a prospective randomized study to evaluate first cycle pregnancy rates in women starting OCs on days 1, 4, and 7.