Carbamazepine coadministration with an oral contraceptive: effects on steroid pharmacokinetics, ovulation, and bleeding.

Some anti-epileptic drugs (AEDs) induce the hepatic p450 enzyme system thereby accelerating metabolism of contraceptive steroids. The interaction between AEDs and hormonal contraception is particularly important for women with epilepsy; their pregnancies must be planned in order to optimize seizure control and minimize teratogenic risk. Since the 1970s, clinicians have implicated AEDs in oral contraceptive (OC) failure. Case reports documented pregnancies and breakthrough bleeding during use of AEDs and OCs ( Kenyon, 1972 ; Back et al., 1988 ) and pharmacokinetic studies documented decreased levels of contraceptive steroids ( Back et al., 1980 ). Case reports, however, could not establish the drug interaction as causal to bleeding or pregnancy because breakthrough bleeding is common among all women and often leads to OC discontinuation. We conducted this study to determine if the pharmacokinetic interaction between a low-dose OC and carbamazepine (CBZ) affects the pharmacodynamic processes of breakthrough bleeding and ovulation.

The study was approved by the Columbia University Medical Center Institutional Review Board with oversight by a data safety and monitoring committee. Written informed consent was obtained from all participants. This was a double-blind, two-period crossover study. Participants took the OC for four cycles. Additionally, for the first two cycles participants took either CBZ 600 milligrams daily or a matching placebo and then crossed over to the alternate treatment for the second two cycles. The treatment order was random. During the second cycle of each treatment, participants underwent twice weekly transvaginal sonograms to measure ovarian activity and weekly venipuncture to measure serum progesterone. The pharmacokinetic portion of the study was conducted twice, in the third week of the second cycle of each treatment. Healthy women were recruited via paper postings and newspaper advertisements and received up to 600 U.S. dollars for their participation, divided over study visits. We included women ages 18 to 35 years with regular menstrual cycles and a body mass index of 18 to 30. We excluded women with infertility, current breast feeding, recent depomedroxyprogesterone acetate use, recent pregnancy, ovarian cysts, endometriosis, oophorectomy, smoking >10 cigarettes a day, contraindications to OCs, taking medications known to affect the cytochrome p450 system or with contraindications to CBZ by screening laboratory testing. The Columbia University Medical Center research pharmacy prepared carbamazepine and matching placebo tablets. Originally, we increased the treatment dose from 200mgs to 600mgs over four days. The first two participants discontinued due to side effects; thereafter, we increased the dose over two weeks. During informed consent, participants reviewed CBZ side effects and teratogenic effects; this was highlighted given possible decreased OC efficacy due to the drug interaction. Participants agreed to use condoms. Participants took a low-dose monophasic OC containing 20 micrograms of ethinyl estradiol and 100 micrograms of levonorgestrel (Aviane TM , Barr Laboratories, Inc.) for 21 days followed by 7 days of inactive pills. We monitored electrolyte levels, AST/ALT, bilirubin, creatinine, hemoglobin, white cell and platelet indices, and CBZ levels at the start and end of the second treatment cycle. A CBZ level above 12 mcgs/ml was considered toxic, below 4 mcgs/ml sub-therapeutic. The randomization sequence was generated by the research pharmacy using a concealed block size, held by staff uninvolved with the trial and concealed from participants and investigators until completion of data collection and analysis. We assessed treatment compliance with a medication diary, reports of missed pills to study staff and CBZ levels. Participants recorded spotting or bleeding daily. For purposes of analysis, we combined bleeding and spotting days. We defined breakthrough bleeding as occurring on cycle days 1–21. We compared bleeding during the second and fourth months (the second month of placebo or CBZ) to isolate breakthrough bleeding related to a drug interaction. We collected venous blood samples at 0, ½ , 1, 1 ½ , 2, 3, 4, 8, 12, 16, and 24 hours following OC administration in the General Clinical Research Center at Columbia University Medical Center. Specimens clotted at room temperature, were separated by centrifugation and serum was stored at −80C. Serum levonorgestrel and EE levels were quantified by radioimmunoassays (RIAs) ( Stanczyk et al., 1975 ; Price et al., 1993 ). Prior to RIA, each analyte was extracted with ethyl acetate:hexane (3:2). The EE assay requires removal of interfering unconjugated metabolites by Celite column partition chromatography. A highly specific antiserum was used with an iodinated radioligand in each RIA. Separation of free from antiserum-bound LNG or EE was achieved with a second antibody. The sensitivities of these RIAs are 0.05 ng/ml for LNG and 15 pg/ml for EE. Intra-assay and inter-assay coefficients of variation are 4.4% and 8.9% for the LNG RIA and 6.9% and 11% for the EE RIA. The assays were performed in one batch. Study staff performed transvaginal ultrasounds twice weekly to identify any follicle-like structures and to follow their growth; this ultrasound schedule is usual in contraceptive studies evaluating ovulation. A mean diameter was estimated as the average of the largest follicle-like structure (FLS) in two perpendicular planes. Participants underwent weekly venipuncture to measure serum progesterone. Serum was separated by centrifugation after clotting for 30 minutes at room temperature. The reproductive endocrine laboratory at Columbia University Medical Center performed progesterone assays using a chemiluminescent immunoassay and the Immulite analyzer (Siemens, Medical Solutions Diagnostics, Malvern, PA). We considered a progesterone level of ≥ 3 ng/ml indicative of ovulation ( Grimes et al., 1994 ). We did not determine an a priori sample size for this exploratory study. We could not estimate expected rates of ovulation or breakthrough bleeding from existing literature. Previous studies examining the drug interaction between OCs and CBZ included ten or fewer participants; we sought a sample of 12 women. For the PK analysis, serum concentration values of LNG and EE for each participant were fitted using a non-compartmental approach. We recorded individual values of the maximum concentration (Cmax), the time to maximum concentration (Tmax), and the trough concentration at 24 hours (Cmin). We calculated the area under the curve (AUC) from 0 to 24 hours using linear trapezoidal approximation. Data analyses compared AUC (0–24 hours), Cmax and Cmin for EE and LNG during exposure to CBZ versus placebo. We log transformed AUC and Cmax values then compared the means using a paired t test. We used Mc Nemar’s test for paired data to compare three outcomes during CBZ versus placebo exposure: the proportions of participants ovulating, the proportions experiencing breakthrough bleeding and the proportions with LNG measurements below the level of detection. We used the Wilcoxon signed ranks test to compare the median number of bleeding days during CBZ versus placebo exposure.

We enrolled twenty-five women from July 2005 to January 2006. Their characteristics are presented in Table 1 . Supplemental Figure 1 presents the flow of participants through the study. One participant was excluded before randomization because she did not menstruate. Twenty-four women were randomized. In the CBZ first group, the investigator removed one participant for anemia and another withdrew. In the placebo first group, one participant withdrew due to depression. Five women in the CBZ first group discontinued due to side effects in the first cycle (see below). Three women in the placebo first group discontinued in the first cycle; one was diagnosed with an unrelated leg sarcoma, one was lost to follow-up and one discontinued without giving a reason. One participant in the placebo first group discontinued after completing cycle two because of difficult intravenous access. One participant in each group was removed by the investigator after study completion due to lost samples and because CBZ was not detected (<0.5 micrograms/ml) on any routine checks. Seventeen women took at least one dose of CBZ. Of these, three reported no side effects. The 14 others reported 25 separate side effects; dizziness (6 participants), sleepiness (5), headaches (4) and nausea (2) were most common. One woman took a lower dose of CBZ (400mgs) during the second cycle of CBZ treatment due to side effects. Fourteen women took at least one dose of placebo, five reported no side effects. The others reported 15 separate side effects: nausea/GI upset (3), headaches (2), and dizziness (2) were the most common. Three participants had likely drug reactions during CBZ treatment. One developed fever and elevated liver function tests and white blood count; another developed a rash. Both resolved after CBZ discontinuation. No toxic levels of CBZ were detected. During placebo treatment, CBZ was undetectable in routine checks. During CBZ treatment, one sample level was 3.4 mcg/ml; all others were in the therapeutic range. We conducted 149 ultrasounds during 20 cycles in 10 women. In 12 cycles, 8 ultrasounds were completed. In five cycles, seven ultrasounds were completed; we omitted the final ultrasound due to absent follicular growth. In three cycles, six ultrasounds were completed due to missed visits. We sought to perform ultrasounds twice weekly. We were usually successful. In two participants, 10 days elapsed between ultrasounds. We drew 73 progesterone levels. In 14 of the 20 cycles, a progesterone level was drawn weekly. In five cycles, the final measurement was omitted after elevated progesterone occurred. In one cycle, progesterone was unmeasured for ten days. Twelve women completed both pK studies; ten were available for analysis (two lost). Table 2 . presents the pharmacokinetic parameters AUC and Cmax for each hormone by treatment condition. At 24 hours, LNG levels below the threshold of detection occurred in six of ten women during CBZ exposure versus one of ten women during placebo exposure (p<0.05) Progesterone levels indicating ovulation occurred in 5 of the CBZ cycles and 1 placebo cycle (p=0.06). In the CBZ cycles, we detected increased progesterone during the first week after the placebo OCs in 3 participants. There was no relationship between trough LNG levels and ovulation (data not shown). Treatment with CBZ had no effect on the size of maximum follicle-like structures compared to placebo (data not shown). All participants completed their bleeding diaries. Breakthrough bleeding was more frequent during CBZ exposure. The median number of bleeding days was seven (range 0-10) during the CBZ cycles compared to zero days (range 0-6) during the placebo cycles (p=0.04). More than two days of breakthrough bleeding occurred during seven of the ten CBZ cycles compared to two of the ten placebo cycles (p= 0.07).

In our study, co-administration of CBZ during OC use permitted more ovulation and breakthrough bleeding than observed in the same women during co-administration of a placebo. Previous pharmacokinetic studies established a clear relationship; CBZ use decreases levels of contraceptive steroids ( Crawford et al., 1990 ; Doose et al., 2003 ). Our study shows that this pharmacokinetic effect is paired with pharmacodynamic effects likely to increase the risk of pregnancy. In a clinical context, breakthrough bleeding may further increase pregnancy risk since bleeding is a common reason for OC discontinuation. Recruitment and retention proved challenging. Expected side effects of CBZ discouraged enrollment and when side effects occurred, most discontinued participation. The size of our study, while comparable or larger than similar PK studies, was still too small to show robust statistical differences between the placebo and CBZ groups for the outcomes of ovulation and bleeding. Our results, however, are not likely attributable to chance. The direction of our findings was consistent; the CBZ group had lower levels of both EE and LNG, more ovulations and more breakthrough bleeding than the placebo group. One strength of our study is the design. We included methods for detecting ovulation throughout the OC cycle, a diary to record bleeding patterns and maintained blinding throughout data collection and analysis. Most studies examining drug interactions, for CBZ or other enzyme inducing AEDs, lacked methods to ascertain ovulation ( Doose et al., 2003 ). A few included isolated measurements of FSH or progesterone; however, that approach is unlikely to detect unpredictable ovulations ( Sidhu et al., 2006 ). We identified no study including detailed methods to ascertain bleeding patterns. Our pharmacokinetic results support a meaningful increase in risk of pregnancy during CBZ use. Decreased levels of EE and LNG during CBZ persisted to the final trough level measurement (t24). Previous reports with CBZ also document this persistent difference ( Crawford et al., 1990 ; Doose et al., 2003 ). Studies from contraceptive injections and implants suggest trough hormone levels may be most important for pregnancy risk ( Rahimy et al., 1999 ; Sivin et al., 2001 ). We did not identify a relationship between trough levels of LNG and ovulation in this study, however, we had little power to do so. Progestins can provide contraceptive effects via cervical mucous changes even if ovulation occurs, however, such progestin only methods are administered continuously. The formulation studied here, as well as most OCs available today, include a pill free (placebo) interval. We studied CBZ because it is widely used by reproductive-age women for epilepsy, pain syndromes and psychiatric conditions ( Radley et al., 2006 ), and is a strong and rapid inducer of the P450 system known to decrease levels of contraceptive steroids ( Crawford et al.,1990 ; Doose et al., 2003 , Magnusson et al., 2008 ). We chose a daily dose of 600 milligrams as typical for epilepsy and likely to result in clinically important enzyme induction ( Crawford et al., 1990 ). Our results may not be generalizable to lower doses of CBZ or other p450 inducers. For example, lamotrigine use is associated with smaller decreases in levels of levonorgestrel, but no change in ethinyl estradiol levels ( Sidhu et al., 2006 ). We studied an OC containing low and commonly used doses of steroids. Two reports have examined the effect of CBZ on the pharmacokinetics of higher dose OCs. Those studies demonstrated similar decreases in EE, LNG and norethindrone, however, did not evaluate methods to detect ovulation or record bleeding ( Crawford et al., 1990 ; Doose et al., 2003 ). We were surprised by the high rate of ovulation during CBZ use. We planned comparisons of follicle growth as surrogates for risk of ovulation but in fact, ovulation itself occurred. The timing of ovulations, immediately after inactive pills, may account for why we did not detect a relationship between follicle growth and exposure to CBZ; some ovulations occurred immediately before ultrasounds in the study cycle. Contraception in women with epilepsy is critical but little is known about its use. One recent study among women with epilepsy attending clinics at an academic medical center (n=145) found that only half of those at risk of pregnancy reported use of highly effective contraception; of these, 28% used OCs ( Davis et al., 2008 ). In this same sample, most were unaware of the drug interactions between hormonal contraception and AEDs or the teratogenic potential of their own AED therapy ( Pack et al., 2009 ). Our findings confirm a long-standing clinical suspicion; a low-dose OC performs poorly, in terms of ovulation suppression and cycle control, during co-administration of a common dose of CBZ. Some clinicians recommend extended cycle regimens with OCs containing 50 micrograms of EE and a shortened pill-free interval. Further studies should explore whether such regimens prevent ovulation and provide a tolerable bleeding pattern.

Supplemental Figure 1. Flow of participants through the trial