- Email: [email protected]
A randomized trial of three oral contraceptives: Comparison of bleeding patterns by contraceptive types and steroid levels
A randomized trial of three oral contraceptives: Comparison of bleeding patterns by contraceptive types and steroid levels Walid A. Saleh, MD," Ronald T. Burkman, MD," Howard A. Zacur, MD, PhD,b Allyn W. Kimball, PhD,b Peter Kwiterovich, MD,b and William K. Bell, MDb Detroit, Michigan, and Baltimore, Maryland OBJECTIVE: Our purpose was to determine the relationship between bioavailability of contraceptive steroids and bleeding patterns. STUDY DESIGN: A randomized clinical trial evaluated 192 women on 50 fL9 of ethinyl estradiol and 1.0 mg of norethindrone (OC1), 35 fLg ethinyl estradiol and 1.0 mg of norethindrone (OC2), and 35 fLg ethinyl estradiol and 0.5 mg norethindrone (OC3) over nine cycles. RESULTS: Intermenstrual bleeding rates were higher for OC3 when compared with OC1 (p = 0.01). The number of intermenstrual bleeding days was highest for OC3 (p = 0.001) and higher for OC2 when compared with OC1 (p < 0.006). The onset of withdrawal bl~eding occurred faster in OC3 patients (p < 0.02). Bioavailability of both contraceptive steroids as measured by baseline values and 1-hour slopes did not correlate with bleeding patterns at 3, 6, and 9 months of use. CONCLUSION: These data suggest that differences in biologic responses associated with pill use cannot be explained solely on the basis of these particular hormone measurements. (AM J OBSTET GVNECOL 1993;168:1740-7.)
Key words: Ethinyl estradiol, norethindrone, oral contraceptives, pharmacokinetics
A number of clinical investigations have examined the side effects and beneficial effects of oral contraceptives. The main cause of failure in oral contraceptive use in noncompliance, and the principal method-related reason for noncompliance is discontinuation because of side effects. I. 2 Breakthrough bleeding has often been described as the most common side effect of oral contraceptives. 2 • 3 It is thought that although oral contraceptives with lower steroid contents have better health benefits they may be associated with a higher incidence of some minor side effects such as poor cycle control. As in the case of cross-over clinical trials from a standard-dose pill to a low-dose pill, there is still uncertainty as to whether changes in the rates of side effects are true responses to a change in the dose or
From the Department of Gynecology and Obstetrics, Henry Ford Hospital," and the Departments of Gynecology and Obstetrics, Biostatistics, Pediatrics, and Medicine, The Johns Hopkins University.' Supported by Contraceptive Evaluation Branch, National Institute of Child Health and Human Development contract No. N01-HD32816. Presented at the Sixtieth Annual Meeting of The Central Association of Obstetricians and Gynecologists, Chicago, Illinois, October 15-17, 1992. Reprint requests: Walid A. Saleh, MD, Department of ObstetricsGynecology, Henry Ford Hospital, Detroit, MI 48202. Copyright © 1993 by Mosby-Year Book, Inc. 0002-9378/93 $1.00 + .20 6/6/46194
1740
type of estrogen or progestogen. 4 • 5 In addition, the use of different data collection techniques, oral contraceptive preparations, and definitions of intermenstrual bleeding have prevented direct comparisons among different studies. 5 • 7 An accurate study of the clinical response of the endometrium to estrogens and progestins should include the factor of bioavailability. Although radioimmunoassay methods make such investigations possible, it has been ignored in most clinical trials. The purpose of this study was to investigate the incidence and pattern of vaginal bleeding in three groups of women prescribed three different oral contraceptives and to attempt correlation between those biologic responses and plasma steroid bioavailability. Material and methods From Oct. 1, 1985, through Oct. 15, 1987, healthy, young women potentially eligible to take oral contraceptives were evaluated at the Research Laboratory of the Department of Gynecology and Obstetrics of the Johns Hopkins Medical Institutions. The National Institutes of Health-funded study was originally performed to establish a relationship between oral contraceptives and changes in lipid and lipoprotein levels. A detailed description of the material and methods has been provided previously.H. 9 Subjects were separated into one of three oral contraceptive groups by comput-
Volume 168, Number 6, Part I Am J Obstet Gynecol
er-generated restricted randomization. The final study groups after randomization and withdrawals were composed as follows: 50 f.Lg of ethinyl estradiol plus 1.0 mg of norethindrone (OC I), 67 women; 35 f.Lg of ethinyl estradiol plus 1.0 mg of norethindrone (OC2), 61 women; 35 f.Lg of ethinyl estradiol plus 0.5 mg of norethindrone (OC3), 64 women. During the third, sixth, and ninth cycles patients were scheduled for a visit on day 21 of the pill cycle. Ethinyl estradiol and norethindrone concentrations were measured ill plasma with a radioimmunoassay procedure described previously.'o The O-hour sample was timed to occur approximately 24 hours after ingestion of the pill for day 20. Subjects then took the tablet for day 21 and were again sampled at 1 hour after ingestion of the tablet. The initial slope was calculated with this I-hour level value minus the O-hour level over the actual time interval. Prior studies by our group had noted that O-hour value and I-hour slope provide a reasonable correlation with lipid levels."' " The ethinyl estradiol assay had a sensitivity of 12 pg/ml; within-assay variation was 10.6% and between-assay variation was 18.7%. Sensitivity for the norethindrone assay was 15 pg/ml; within-assay variation was 11.6% and between-assay variation was 12.5%. An extensive standardized interview was administered to all participants during each visit. A bleeding day was recorded if any bleeding took place during that day. Intermenstrual bleeding and withdrawal bleeding were defined as any bleeding occurring during days I through 21 and 22 through 28 of the cycle, respectively. There was no attempt to quantitate the amount of bleeding. Amenorrhea was defined as no withdrawal bleeding before the start of the next pill in a cycle with no intermenstrual bleeding. We omitted cycles where pills had been forgotten by the patient, as indicated by menstrual calendars. A linear regression analysis was used for comparison of slopes and bleeding days. Statistical methods also included X2 and the Student t test. Results
A total of 316 women enrolled in the study. Sixty-one percent (192) completed the full 9-month study and are included in the subsequent analyses. As previously described, eligible women not completing the full study cycle did not vary significantly from the women completing the study except for weighing on the average 2.4 kg less
Saleh et al.
1741
cally different among various preparations. For the 59 patients eliminated from the study the incidence of the intermenstrual bleeding was 24%. There was no evidence that women lost to follow-up ceased method use because they were unwilling to tolerate their bleeding pattern. The incidence of amenorrhea, dysmenorrhea, nausea, headache, depressed mood, breast tenderness, and acne was similar for all three preparations averaging 0.5%, 7%, 16%, 26.5%, 16.6%, and 44.3%, respectively. A higher percentage of patients had intermenstrual bleeding (intermenstrual bleeding rates) on the lowest dosage of norethindrone and ethinyl estradiol (OC3) when compared with patients on OC I during each of the nine cycles
1742 Saleh et al.
June 1993 ..\m J Obstet Gynecol
80 •
0C1
o
0C2
•
0C3
8
9
70 60 CD
~
-=
50
.j ~
40 30 20 10
3
2
1
0
4
5
7
6
Months Fig. 1. Intermenstrual bleeding (1MB) rates. Cycle 9 was excluded because of low numbers.
100
80 60 40 20 O~~~~DF~~~~~~--~~~~~~~~~~~~~~
o
i m
~"#.
28
56
84
112
140
168
196
224
252
100
80 60
40 20 O~~~~~~~~~~~~~~~~~~~~~~~
o
28
56
84
112
140
168
196
224
252
100
80 60 40 20 o~~~~~~~~~~~~~~~~~~~~~~~~
o
28
56
84
112
140
168
196
Fig. 2. Daily percentages of patients with intermenstrual bleeding.
224
DAYS
252
Volume 168, Number 6, Part I Am J Obstet Gynecol
Saleh et al.
1743
100
80
2'
1
OC1 ---0--
60
0C2 0C3
,a
.r:.
'j
40
~
20
12
15
18
24
21
27
Day
Fig. 3. Average daily bleeding rates per day of cycle.
II)
>-
6
C"CS
• •
"0
'0
A E :::l
c:
5
OC1
~
4
~
CC3
3 2
o 2
3
4
5
6
7
8
9
months
Fig. 4. Number of intermenstrual bleeding days per cycle.
OCI and OC2 and between OC2 and OC3 (Fig. 5). However, compared with women on OC3, women on OC 1 had significantly more days of withdrawal bleeding during the first 3 months of use r.p < 0.041). For all patients the number of bleeding days between days 22 and 28 remained relatively constant over nine cycles with a mean ± SE number of days of 3.9 ± 0.1. The number of days elapsed between the last active pill day and the onset of withdrawal bleeding were analyzed in cycles that did not have intermenstrual bleeding. Overall, the mean ± SE number of days to withdrawal was 1.3 ± 0.1, 1.1 ± 0.1, and 0.9 ± 0.1 for OCl, OC2, and OC3, respectively. Overall, women on OC3 withdrew faster than those on OCI r.p = 0.001) or OC2 r.p = 0.018). The difference between OCI and OC2 was not statistically significant.
Bioavailability of both contraceptive steroids as measured by baseline levels and I-hour slopes (Table I) was correlated with bleeding patterns at 3, 6, and 9 months of oral contraceptive use. In general, higher ethinyl estradiol slopes were associated with higher norethindrone slopes at 3, 6, and 9 months r.p < 0.004). Baseline values and I-hour slopes of ethinyl estradiol and norethindrone showed no correlation with the number of intermenstrual bleeding days, days of withdrawal bleeding, or number of days elapsed before withdrawal bleeding. Comment
Cross-cultural studies of patterns and perceptions of menstruation show that the number of days of bleeding is the primary measure ofvolume."· 12 Consequently, for
1744 Saleh et al.
June 1993 Am J Obstet Gyneco1
5
4
3 en
;;-
o
2
o
2
3
4
5
7
6
8
9
Months
Fig. 5. Number of withdrawal bleeding days per cycle.
Table I. Mean steroid parameters by oral contraceptive type and time 50 pg ethinyl estradiol 1.0 mg norethindrone
+ Ethinyl estradiol Baseline o hr Initial Slope 3 rno o hr Initial Slope 6 rno o hr Initial Slope 9 rno o hr Initial Norethindrone Baseline o hr Initial Slope 3 rno o hr Initial Slope 6 rno o hr Initial Slope 9 rno o hr Initial
35 pg ethinyl estradiol
+ 1.0 mg norethindrone
35 pg ethinyl estradiol 0.5 mg norethindrone
+
All pills combined
23.0 (1.0) 89.5 (6.0)
22.9 (0.7) 66.6 (5.9)
23.0 (0.8) 57.5 (4.0)
23.0 (0.5) 71.6 (3.3)
48.0 (2.7) 126.9 (9.0)
40.6 (2.5)' 101.9 (6.9)
43.1 (3.3) 90.6 (6.6)
44.0 (1.7) 106.9 (4.6)
53.2 (3.1) 124.0 (8.8)
48.0 (3.6) 101.9 (6.9)
41.9 (3.0) 91.2 (6.2)
47.8 (1.9) 105.9 (4.6)
51.6 (3.4) 127.0 (7.5)
47.9 (3.3) 103.1 (6.9)
42.1 (2.8) 89.0 (5.9)
47.2 (1.8) 106.8 (4.1)
0.15 (0.03) 3.65 (0.26)
0.12 (0.00) 3.61 (0.30)
0.14 (0.01) 1.88 (0.13)
0.14 (0.01) 3.05 (0.15)
2.25 (0.20) 5.79 (0.38)
1.86 (0.13) 5.87 (0.37)
1.27 (0.12) 3.72 (0.20)
1.80 (0.09) 5.12 (0.20)
2.42 (0.19) 6.27 (0.44)
2.18 (0.21) 6.26 (0.44)
1.23 (0.10) 3.86 (0.21)
1.95 (0.11) 5.45 (0.23)
2.10 (0.18) 6.77 (0.45)
2.08 (0.17) 5.92 (0.41)
1.16 (0.08) 3.94 (0.21)
1.78 (0.09) 5.55 (0.23)
data collection and analysis we have analyzed one type of menstrual disturbance, the duration of bleeding. The similarity found between intermenstrual bleeding rates obtained from interviews alone and number of days obtained from menstrual calendars suggest that in our population the number of days was an important component of the suggestive complaint. In our study the incidence of intermenstrual bleeding appears to be above the range reported for most
monophasic oral contraceptives!' 5. 13 The study design accounts for some of these differences as women with a history of previous menstrual irregularities were excluded from the study. In agreement with previous studies, our data suggest that cycle control improves after a few months of use and that oral contraceptives with lower total steroid dose have an increase in cyclic irregularities. 5 • 13 The difference in intermenstrual bleeding days between OC 1, OC2, and OC3 suggests
Saleh et al.
Volume 168, Number 6, Part 1 Am J Obstet Gynecol
that intermenstrual bleeding is related to both the progestogen and the estrogen content of birth control pills. The withdrawal period, on the other hand, seems less dependent on the dose of either contraceptive steroid. As expected, the oral contraceptive with the lowest total steroid dose provided the least endometrial stability. On this pill withdrawal bleeding occurred earlier than on the other preparations. For women on the lowest progestogen dose the pattern of monthly bleeding assumed an expected biphasic pattern with intermenstrual bleeding rates peaking between days 9 to 20. 5 For women on the the oral contraceptive with a higher dosage of either steroid, intermenstrual bleeding occurred in continuation with the withdrawal period. Clinically, this may represent a less disturbing side effect than midcycle bleeding. The actual difference in the number of bleeding days, although statistically significant, may not be clinically significant to the user. Because of lack of availability, a preparation containing 50 fLg ethinyl estradiol and 0.5 mg norethindrone was not included. Thus a more detailed analysis of doseresponse was not possible. Advances in the development of oral contraceptives have resulted in steroid doses approaching threshold levels of good cycle control. Understanding the relationship between clinical performance and pharmacokinetics of contraceptive steroids becomes extremely important when one's goal is to decrease nuisance side effects while maintaining or increasing health benefits. This paper presents an approach to examining the relationship between contraceptive steroid levels at day 21 of the pill cycle and bleeding patterns. Baseline values and I-hour slopes had been found previously to provide a reasonable correlation with lipid levels and to be proportional to steroid dosages. 8 , 9 Yet, for the preparations used in this study, no significant relationship between bleeding patterns and baseline or I-hour slopes of contraceptives steroids was apparent. In agreement with other authors, there is a large intersubject variation in the serum levels of both ethinyl estradiol and norethindrone. 10. 14. 15 Steroids differ in absorption, distribution, first-pass effect, and elimination. The picture is further complicated by interpatient and interpopulation variations in plasma kinetics of contraceptive steroids. The results of this study suggest a need for further investigation of the pharmacokinetics of contraceptive steroids and the occurrence of undesired side effects. An attempt at controlling for other factors such as sex hormone-binding globulin capacity might shed more light on the relationship between steroid levels and bleeding patterns. We showed that differences in biologic responses associated with pill use cannot be explained solely on the basis of baseline values and I-hour slopes when determined at day 21 of the pill cycle.
1745
More detailed measurements over 24 hours or measurements at other times during the pill cycle may be necessary to further explore this relationship. REFERENCES 1. Wheble AM, Street P, Wheble SM. Contraception: failure in practice. Br J Fam Plann 1987;13:40-5. 2. De Lia JE, Emery MG. Clinical pharmacology and common minor side effects of oral contraceptives. Clin Obstet GynecoI1981;24:879-92. 3. Neel EU, Litt IF, Jay MS. Side effects and compliance with low- and conventional-dose oral contraceptives among adolescents. J Adolesc Health Care 1987;8:327-9. 4. Edelman DA, Kothenbeutel R, Levinski MJ, et al. An assessment of the side effects of switching from one oral contraceptive to another. Int J Gynaecol Obstet 1976; 14: 397-400. 5. Edgren RA, Nelson JH, Gordon RT, Keifer WS. Bleeding patterns with low-dose, monophasic oral contraceptives. Contraception 1989;40:285-97. 6. Trussell J, Kost K. Contraceptive failure in the United States: a critical review of the literature. Stud Fam Plann 1987;18:237-83. 7. Talwar PP, Berger GS. A prospective, randomized study of oral contraceptives: the effect of study design on reported rates of symptoms. Contraception 1979;20:329-37. 8. Burkman RT, Zacur HA, Kimball AW, et al. Oral contraceptives and lipids and lipoproteins. I. Variations in the mean levels by oral contraceptive type. Contraception 1989;40:553-61. 9. Burkman RT, Zacur HA, Kimball AW, et al. Oral contraceptives and lipids and lipoproteins. II. Relationship to plasma steroid levels and outlier status. Contraception 1989;40:676-89. 10. Zacur HA, Linkins S, Chang V, Smith B, Kimball AW, Burkman R. Ethinyl estradiol and norethindrone radioimmunoassay following Sephadex LH-20 column chromatography. Clin Chern Acta 1991;204:209-15. 11, Snowden R, Christian B, eds. Patterns and perceptions of menstruation. London: Croom Hel, 1983. 12. Belsey EM. The association between vaginal bleeding patterns and reason for discontinuation of contraceptive use. Contraception 1988;38:207-25. 13. Hill GA, Wheeler JM. Incidence of breakthrough bleeding during oral contraceptive therapy. J Reprod Med 1991; 36:334-9. 14. Goldzieher JW. Pharmacology of contraceptive steroids: a brief review. AM J OBSTET GYNECOL 1989;160:1260-4. 5. Back DJ, Breckenridge AM, Crawford FE, et al. Kinetics of norethindrone in women. I. Radioimmunoassay and concentrations during multiple dosing. Clin Pharmacol Ther 1978;24:439-47.
Discussion DR. RUDI ANSBACHER, Ann Arbor, Michigan. Although birth control pills were introduced for clinical use 32 years ago, there is still a paucity of information concerning how the steroids contained therein are metabolized or how side effects of these medications are correlated with the steroid content. This study, which emphasizes bleeding patterns of patients on three different formulations of oral contraceptives, sheds little light on what causes the differences from one patient to another. The authors state that there was no correlation between baseline or I-hour slopes of contraceptive steroids and intermenstrual bleeding. However, this is based on one blood sample obtained just before ingestion of the pill on day 21 of
1746 Saleh et al.
the cycle (approximately 24 hours after ingestion of the previous pill) and a second blood sample taken 1 hour after ingestion of the pill on the same day. Helton et al. I. 2 in the mid-1970s were the first to note the marked variability in an individual's metabolism of the active hormones, especially the estrogenic component, contained in oral contraceptives. In 1989 Goldzieher" emphasized that the blood levels from a given oral dose of ethinyl estradiol and norethindrone vary widely from patient to patient because of the partial inactivation of the steroids on their first pass through the enterohepatic circulation. Therefore wide ranges in bioavailability can occur in an individual and between individuals. In a 1992 review article Zacur and Stewart' summarized the information currently known about the mechanism involved in the metabolism of the oral contraceptive pill steroids and the effects that other drugs and medications may have on subsequent steroid levels. However, there are still many gaps in our knowledge. Ansbacher' in 1991 stated that the use of a single day test period for comparing low-dose contraceptive agents could not be extrapolated to the entire 28-day cycle, because the individual patient variation in the absorption of the active ingredients and the latter's subsequent metabolic breakdown cannot be assured. In addition, mean patient data were not sufficient to ensure similar bioavailability of a product in a particular patient. In the study we are reviewing today the bioavailability data are not given for individual patients but for the entire group, and these data are expressed as means. Were there differences noted in individual patients, were some patients outliers, and if there were outliers, how did these affect the mean data for the entire group? Can one really accept data from hormonal assays that have a between-assay variation of 18.7% for ethinyl estradiol and 12.5% for norethindrone? Correlation of hormonal levels with the status of the endometrium would enable us to properly adjust dosage to correct the bothersome symptom of intermenstrual bleeding in individual patients. Admittedly, this would be most costly, but without so doing we will be left with the intermenstrual bleeding patterns described in this study. REFERENCES 1. Helton ED, Williams MC, Goldzieher ]W. Human urinary and liver conjugates of 17 a-ethynylestradiol. Steroids 1976; 27:851-67. 2. Helton ED, Simmons R, Meltz ML, Goldzieher ]W. Variability in conjugates of ethynylestradiol produced by in vitro incubation of human liver tissue. Contraception 1977;16: 256-60. 3. Goldzieher ]W. Pharmacology of contraceptive steroids: a briefreview. AM] OBSTET GYNECOL 1989;160:1260-4. 4. Zacur HA, Stewart D. New concepts in oral contraceptive pill use. Curr Opin Obstet Gynecol 1992;4:365-71. 5. Ansbacher R. Interchangeability oflow-dose oral contraceptives. Are current bioequivalent testing measures adequate to ensure therapeutic equivalency? Contraception 1991;43: 139-47.
June 1993 Am J Obstet Gynecol
DR. FRANK W. LING, Memphis, Tennessee. Complications of oral contraceptives, such as breakthrough bleeding, are commonly referred to as only "minor risks" or "side effects." Nevertheless, it is these annoyances that are the primary reason for noncompliance. Minimizing or eliminating these side effects could help patients more effectively use the contraceptive products available. Twenty percent is a commonly quoted incidence of breakthrough bleeding for oral contraceptive users, with higher rates occurring with low-dose preparations. The optimal combination oral contraceptive agent will, for a given individual, maximize contraception while minimizing untoward side effects. In a study originally designed to evaluate the relationship between oral contraceptives and changes in lipoprotein and lipid levels, the authors report bleeding patterns of 192 women on three different pills. The authors also attempted to correlate these biologic responses with the bioavailability of plasma steroids. Both goals of the study have been at least partially achieved, because the authors have successfully described bleeding patterns in these three groups and determined that there is no correlation of either baseline value or I-hour slopes of ethinyl estradiol and norethindrone at day 21 of the cycle with the amount or type of bleeding observed. Several questions regarding definitions and methods do remain, however. For example, the authors state that women on higher dosages of either steroid had intermenstrual bleeding in continuation with the withdrawal period. This seems to contradict Fig. 5, in which the number of withdrawal bleeding days per cycle for OC 1 and OC2 did not exceed 5 days during any of the nine cycles. For a woman to have intermenstrual bleeding continuous with withdrawal bleeding, according to the authors' definitions, withdrawal bleeding would have to have occurred through day 28 and into day 1 of the subsequent cycle. Should this day 1 bleeding really be considered "intermenstrual"? Clarification of this point would be helpful. Also, because there was no attempt to quantitate the amount of bleeding on any given day, could the authors speculate on their observations regarding this very crucial issue? Because patients will often accept a small amount of spotting but will not accept a large amount of bleeding, could this lack of quantification of bleeding seriously affect the interpretation and clinical use of these data? Also, because the number of intermenstrual bleeding days was fewer than 3 days per month over the 9-month period, even for the OC3 users, was this felt to be a significant problem by the subjects when interviewed? It was further stated that there was no evidence that patients were lost to follow-up or stopped participation because of an inability to tolerate their bleeding pattern. It was interesting to note that the other side effects such as amenorrhea, dysmenorrhea, nausea, headache, depressed mood, breast tenderness, and acne had frequencies ranging from 0.5% to 44.3%. Were these the reasons cited for discontinuing the pill, and if so, are there similar frequencies of these side effects in the
Volume 168, Number 6, Part 1 Am J Obstet Gynecol
patients who did complete the study? If these side effects were not the reason the patients dropped from the study, I would be interested in knowing why they did, I would be particularly interested in knowing which side effects had an incidence of 26.5% and 44.3%. In Fig. 1, in comparing intermenstrual bleeding rates in each cycle among the three preparations, the authors state that the ninth cycle data were discarded because of limited numbers. Because the study described 192 patients who completed 9 months, this loss of data in the final month of the study is confusing. I noted that the number of days that elapsed between the last active pill day and the onset of withdrawal bleeding was analyzed in only those cycles in which there had not been intermenstrual bleeding. Why was this specific exclusion used? Because intermenstrual bleeding could include even a small amount of spotting early in the cycle, the authors should explain what they feel the relationship might be between intermenstrual bleeding early in the cycle with withdrawal bleeding at the end of the cycle, which would warrant exclusion of these cycles. The authors point out that their incidence of intermenstrual bleeding appears above the range usually reported for monophasic pills and may be caused by their failure to exclude women with a previous history of menstrual irregularities. Because a previous history of menstrual irregularity was not controlled for, were other factors which might affect menstrual disturbance or oral contraceptive metabolism similarly not controlled? A final consideration is whether there were assurances as to the timing of ingestion of the pill on a daily basis. Because the authors excluded cycles in which a pill was missed, should a similar adjustment have been made if a pill had been taken at a markedly different time of the day? DR. SALEH (Closing). Although much of the clinical information presented in this study relative to pill dosage and vaginal bleeding are known by most clinicians, a significant goal was to attempt to correlate clinical behavior of birth control pills and steroid bioavailability. We felt this was a unique opportunity to study the subject in depth by means of a randomized clinical trial. We hoped to relate bleeding patterns to contraceptive sex steroid levels for the time period and samples available to us. Unfortunately, as discussed, we were unable to demonstrate a significant relationship. As pointed out by Dr. Ansbacher, I agree that multiple samples drawn at different time intervals would have been helpful. However, because of cost restraints, the two time periods were chosen. They were initially
Saleh et al.
1747
chosen on the basis of a pilot study including 58 patients followed with multiple samples over a 24-hour period. The I-hour slope was found to be a reliable indicator of steroid metabolic behavior when lipid levels were concerned. We know that ethinyl estradiol and norethindrone tend to peak within a few hours of oral administration and then progressively decrease afterward. We are aware that ethinyl estradiol and norethindrone concentrations vary markedly over a 24-hour period, and this may complicate analyses of steroid bioavailability and biologic responses. Bleeding may be related to the amount of variation rather than hormone level. Relative to bleeding in patients lost to follow-up, the main reason for patients discontinuing the study was relocation out of the Baltimore area. In those patients, the incidence of bleeding was 44% and was not statistically different from the study group. Relative to the pattern of bleeding in the withdrawal period, the number of bleeding days within the withdrawal period, or days 21 to 28, was higher for the higher dosage preparation only during the first 3 months of the study. Those who had the highest endometrial stability during the first part of the cycle tended to bleed more during the withdrawal period, whereas those on the lowest steroid dosage preparation bled more during the intermenstrual period, or days 1 to 21. In one graph patients were excluded from the ninth month. The intermenstrual bleeding rates were omitted because the last visit often occurred in the middle of the ninth month, making calculation of a rate for that month impossible. What were the factors in the patients' history that may have influenced the outcome? Patients selected for the study were between 18 and 35 years old, had regular menstrual cycles lasting between 25 and 35 days, and did not have any contraindication to birth control pills or previous endocrinologic or metabolic disorders. Body weights were within 115% of the ideal body weight. Patients did not breast-feed, become pregnant, or take steroid or birth control pills within 6 months of the study. Why were higher rates of bleeding found in this study? As stated, women with previous menstrual irregularities were excluded from the study. Comparison between different studies is difficult because of the difference in study design and because the definition of intermenstrual bleeding is different. We used a liberal definition, including all bleeding irrespective of the amount, whereas other authors have used bleeding enough to stain a pad or significant enough to be a problem to the patient.
Key words: Ethinyl estradiol, norethindrone, oral contraceptives, pharmacokinetics
A number of clinical investigations have examined the side effects and beneficial effects of oral contraceptives. The main cause of failure in oral contraceptive use in noncompliance, and the principal method-related reason for noncompliance is discontinuation because of side effects. I. 2 Breakthrough bleeding has often been described as the most common side effect of oral contraceptives. 2 • 3 It is thought that although oral contraceptives with lower steroid contents have better health benefits they may be associated with a higher incidence of some minor side effects such as poor cycle control. As in the case of cross-over clinical trials from a standard-dose pill to a low-dose pill, there is still uncertainty as to whether changes in the rates of side effects are true responses to a change in the dose or
From the Department of Gynecology and Obstetrics, Henry Ford Hospital," and the Departments of Gynecology and Obstetrics, Biostatistics, Pediatrics, and Medicine, The Johns Hopkins University.' Supported by Contraceptive Evaluation Branch, National Institute of Child Health and Human Development contract No. N01-HD32816. Presented at the Sixtieth Annual Meeting of The Central Association of Obstetricians and Gynecologists, Chicago, Illinois, October 15-17, 1992. Reprint requests: Walid A. Saleh, MD, Department of ObstetricsGynecology, Henry Ford Hospital, Detroit, MI 48202. Copyright © 1993 by Mosby-Year Book, Inc. 0002-9378/93 $1.00 + .20 6/6/46194
1740
type of estrogen or progestogen. 4 • 5 In addition, the use of different data collection techniques, oral contraceptive preparations, and definitions of intermenstrual bleeding have prevented direct comparisons among different studies. 5 • 7 An accurate study of the clinical response of the endometrium to estrogens and progestins should include the factor of bioavailability. Although radioimmunoassay methods make such investigations possible, it has been ignored in most clinical trials. The purpose of this study was to investigate the incidence and pattern of vaginal bleeding in three groups of women prescribed three different oral contraceptives and to attempt correlation between those biologic responses and plasma steroid bioavailability. Material and methods From Oct. 1, 1985, through Oct. 15, 1987, healthy, young women potentially eligible to take oral contraceptives were evaluated at the Research Laboratory of the Department of Gynecology and Obstetrics of the Johns Hopkins Medical Institutions. The National Institutes of Health-funded study was originally performed to establish a relationship between oral contraceptives and changes in lipid and lipoprotein levels. A detailed description of the material and methods has been provided previously.H. 9 Subjects were separated into one of three oral contraceptive groups by comput-
Volume 168, Number 6, Part I Am J Obstet Gynecol
er-generated restricted randomization. The final study groups after randomization and withdrawals were composed as follows: 50 f.Lg of ethinyl estradiol plus 1.0 mg of norethindrone (OC I), 67 women; 35 f.Lg of ethinyl estradiol plus 1.0 mg of norethindrone (OC2), 61 women; 35 f.Lg of ethinyl estradiol plus 0.5 mg of norethindrone (OC3), 64 women. During the third, sixth, and ninth cycles patients were scheduled for a visit on day 21 of the pill cycle. Ethinyl estradiol and norethindrone concentrations were measured ill plasma with a radioimmunoassay procedure described previously.'o The O-hour sample was timed to occur approximately 24 hours after ingestion of the pill for day 20. Subjects then took the tablet for day 21 and were again sampled at 1 hour after ingestion of the tablet. The initial slope was calculated with this I-hour level value minus the O-hour level over the actual time interval. Prior studies by our group had noted that O-hour value and I-hour slope provide a reasonable correlation with lipid levels."' " The ethinyl estradiol assay had a sensitivity of 12 pg/ml; within-assay variation was 10.6% and between-assay variation was 18.7%. Sensitivity for the norethindrone assay was 15 pg/ml; within-assay variation was 11.6% and between-assay variation was 12.5%. An extensive standardized interview was administered to all participants during each visit. A bleeding day was recorded if any bleeding took place during that day. Intermenstrual bleeding and withdrawal bleeding were defined as any bleeding occurring during days I through 21 and 22 through 28 of the cycle, respectively. There was no attempt to quantitate the amount of bleeding. Amenorrhea was defined as no withdrawal bleeding before the start of the next pill in a cycle with no intermenstrual bleeding. We omitted cycles where pills had been forgotten by the patient, as indicated by menstrual calendars. A linear regression analysis was used for comparison of slopes and bleeding days. Statistical methods also included X2 and the Student t test. Results
A total of 316 women enrolled in the study. Sixty-one percent (192) completed the full 9-month study and are included in the subsequent analyses. As previously described, eligible women not completing the full study cycle did not vary significantly from the women completing the study except for weighing on the average 2.4 kg less
Saleh et al.
1741
cally different among various preparations. For the 59 patients eliminated from the study the incidence of the intermenstrual bleeding was 24%. There was no evidence that women lost to follow-up ceased method use because they were unwilling to tolerate their bleeding pattern. The incidence of amenorrhea, dysmenorrhea, nausea, headache, depressed mood, breast tenderness, and acne was similar for all three preparations averaging 0.5%, 7%, 16%, 26.5%, 16.6%, and 44.3%, respectively. A higher percentage of patients had intermenstrual bleeding (intermenstrual bleeding rates) on the lowest dosage of norethindrone and ethinyl estradiol (OC3) when compared with patients on OC I during each of the nine cycles
1742 Saleh et al.
June 1993 ..\m J Obstet Gynecol
80 •
0C1
o
0C2
•
0C3
8
9
70 60 CD
~
-=
50
.j ~
40 30 20 10
3
2
1
0
4
5
7
6
Months Fig. 1. Intermenstrual bleeding (1MB) rates. Cycle 9 was excluded because of low numbers.
100
80 60 40 20 O~~~~DF~~~~~~--~~~~~~~~~~~~~~
o
i m
~"#.
28
56
84
112
140
168
196
224
252
100
80 60
40 20 O~~~~~~~~~~~~~~~~~~~~~~~
o
28
56
84
112
140
168
196
224
252
100
80 60 40 20 o~~~~~~~~~~~~~~~~~~~~~~~~
o
28
56
84
112
140
168
196
Fig. 2. Daily percentages of patients with intermenstrual bleeding.
224
DAYS
252
Volume 168, Number 6, Part I Am J Obstet Gynecol
Saleh et al.
1743
100
80
2'
1
OC1 ---0--
60
0C2 0C3
,a
.r:.
'j
40
~
20
12
15
18
24
21
27
Day
Fig. 3. Average daily bleeding rates per day of cycle.
II)
>-
6
C"CS
• •
"0
'0
A E :::l
c:
5
OC1
~
4
~
CC3
3 2
o 2
3
4
5
6
7
8
9
months
Fig. 4. Number of intermenstrual bleeding days per cycle.
OCI and OC2 and between OC2 and OC3 (Fig. 5). However, compared with women on OC3, women on OC 1 had significantly more days of withdrawal bleeding during the first 3 months of use r.p < 0.041). For all patients the number of bleeding days between days 22 and 28 remained relatively constant over nine cycles with a mean ± SE number of days of 3.9 ± 0.1. The number of days elapsed between the last active pill day and the onset of withdrawal bleeding were analyzed in cycles that did not have intermenstrual bleeding. Overall, the mean ± SE number of days to withdrawal was 1.3 ± 0.1, 1.1 ± 0.1, and 0.9 ± 0.1 for OCl, OC2, and OC3, respectively. Overall, women on OC3 withdrew faster than those on OCI r.p = 0.001) or OC2 r.p = 0.018). The difference between OCI and OC2 was not statistically significant.
Bioavailability of both contraceptive steroids as measured by baseline levels and I-hour slopes (Table I) was correlated with bleeding patterns at 3, 6, and 9 months of oral contraceptive use. In general, higher ethinyl estradiol slopes were associated with higher norethindrone slopes at 3, 6, and 9 months r.p < 0.004). Baseline values and I-hour slopes of ethinyl estradiol and norethindrone showed no correlation with the number of intermenstrual bleeding days, days of withdrawal bleeding, or number of days elapsed before withdrawal bleeding. Comment
Cross-cultural studies of patterns and perceptions of menstruation show that the number of days of bleeding is the primary measure ofvolume."· 12 Consequently, for
1744 Saleh et al.
June 1993 Am J Obstet Gyneco1
5
4
3 en
;;-
o
2
o
2
3
4
5
7
6
8
9
Months
Fig. 5. Number of withdrawal bleeding days per cycle.
Table I. Mean steroid parameters by oral contraceptive type and time 50 pg ethinyl estradiol 1.0 mg norethindrone
+ Ethinyl estradiol Baseline o hr Initial Slope 3 rno o hr Initial Slope 6 rno o hr Initial Slope 9 rno o hr Initial Norethindrone Baseline o hr Initial Slope 3 rno o hr Initial Slope 6 rno o hr Initial Slope 9 rno o hr Initial
35 pg ethinyl estradiol
+ 1.0 mg norethindrone
35 pg ethinyl estradiol 0.5 mg norethindrone
+
All pills combined
23.0 (1.0) 89.5 (6.0)
22.9 (0.7) 66.6 (5.9)
23.0 (0.8) 57.5 (4.0)
23.0 (0.5) 71.6 (3.3)
48.0 (2.7) 126.9 (9.0)
40.6 (2.5)' 101.9 (6.9)
43.1 (3.3) 90.6 (6.6)
44.0 (1.7) 106.9 (4.6)
53.2 (3.1) 124.0 (8.8)
48.0 (3.6) 101.9 (6.9)
41.9 (3.0) 91.2 (6.2)
47.8 (1.9) 105.9 (4.6)
51.6 (3.4) 127.0 (7.5)
47.9 (3.3) 103.1 (6.9)
42.1 (2.8) 89.0 (5.9)
47.2 (1.8) 106.8 (4.1)
0.15 (0.03) 3.65 (0.26)
0.12 (0.00) 3.61 (0.30)
0.14 (0.01) 1.88 (0.13)
0.14 (0.01) 3.05 (0.15)
2.25 (0.20) 5.79 (0.38)
1.86 (0.13) 5.87 (0.37)
1.27 (0.12) 3.72 (0.20)
1.80 (0.09) 5.12 (0.20)
2.42 (0.19) 6.27 (0.44)
2.18 (0.21) 6.26 (0.44)
1.23 (0.10) 3.86 (0.21)
1.95 (0.11) 5.45 (0.23)
2.10 (0.18) 6.77 (0.45)
2.08 (0.17) 5.92 (0.41)
1.16 (0.08) 3.94 (0.21)
1.78 (0.09) 5.55 (0.23)
data collection and analysis we have analyzed one type of menstrual disturbance, the duration of bleeding. The similarity found between intermenstrual bleeding rates obtained from interviews alone and number of days obtained from menstrual calendars suggest that in our population the number of days was an important component of the suggestive complaint. In our study the incidence of intermenstrual bleeding appears to be above the range reported for most
monophasic oral contraceptives!' 5. 13 The study design accounts for some of these differences as women with a history of previous menstrual irregularities were excluded from the study. In agreement with previous studies, our data suggest that cycle control improves after a few months of use and that oral contraceptives with lower total steroid dose have an increase in cyclic irregularities. 5 • 13 The difference in intermenstrual bleeding days between OC 1, OC2, and OC3 suggests
Saleh et al.
Volume 168, Number 6, Part 1 Am J Obstet Gynecol
that intermenstrual bleeding is related to both the progestogen and the estrogen content of birth control pills. The withdrawal period, on the other hand, seems less dependent on the dose of either contraceptive steroid. As expected, the oral contraceptive with the lowest total steroid dose provided the least endometrial stability. On this pill withdrawal bleeding occurred earlier than on the other preparations. For women on the lowest progestogen dose the pattern of monthly bleeding assumed an expected biphasic pattern with intermenstrual bleeding rates peaking between days 9 to 20. 5 For women on the the oral contraceptive with a higher dosage of either steroid, intermenstrual bleeding occurred in continuation with the withdrawal period. Clinically, this may represent a less disturbing side effect than midcycle bleeding. The actual difference in the number of bleeding days, although statistically significant, may not be clinically significant to the user. Because of lack of availability, a preparation containing 50 fLg ethinyl estradiol and 0.5 mg norethindrone was not included. Thus a more detailed analysis of doseresponse was not possible. Advances in the development of oral contraceptives have resulted in steroid doses approaching threshold levels of good cycle control. Understanding the relationship between clinical performance and pharmacokinetics of contraceptive steroids becomes extremely important when one's goal is to decrease nuisance side effects while maintaining or increasing health benefits. This paper presents an approach to examining the relationship between contraceptive steroid levels at day 21 of the pill cycle and bleeding patterns. Baseline values and I-hour slopes had been found previously to provide a reasonable correlation with lipid levels and to be proportional to steroid dosages. 8 , 9 Yet, for the preparations used in this study, no significant relationship between bleeding patterns and baseline or I-hour slopes of contraceptives steroids was apparent. In agreement with other authors, there is a large intersubject variation in the serum levels of both ethinyl estradiol and norethindrone. 10. 14. 15 Steroids differ in absorption, distribution, first-pass effect, and elimination. The picture is further complicated by interpatient and interpopulation variations in plasma kinetics of contraceptive steroids. The results of this study suggest a need for further investigation of the pharmacokinetics of contraceptive steroids and the occurrence of undesired side effects. An attempt at controlling for other factors such as sex hormone-binding globulin capacity might shed more light on the relationship between steroid levels and bleeding patterns. We showed that differences in biologic responses associated with pill use cannot be explained solely on the basis of baseline values and I-hour slopes when determined at day 21 of the pill cycle.
1745
More detailed measurements over 24 hours or measurements at other times during the pill cycle may be necessary to further explore this relationship. REFERENCES 1. Wheble AM, Street P, Wheble SM. Contraception: failure in practice. Br J Fam Plann 1987;13:40-5. 2. De Lia JE, Emery MG. Clinical pharmacology and common minor side effects of oral contraceptives. Clin Obstet GynecoI1981;24:879-92. 3. Neel EU, Litt IF, Jay MS. Side effects and compliance with low- and conventional-dose oral contraceptives among adolescents. J Adolesc Health Care 1987;8:327-9. 4. Edelman DA, Kothenbeutel R, Levinski MJ, et al. An assessment of the side effects of switching from one oral contraceptive to another. Int J Gynaecol Obstet 1976; 14: 397-400. 5. Edgren RA, Nelson JH, Gordon RT, Keifer WS. Bleeding patterns with low-dose, monophasic oral contraceptives. Contraception 1989;40:285-97. 6. Trussell J, Kost K. Contraceptive failure in the United States: a critical review of the literature. Stud Fam Plann 1987;18:237-83. 7. Talwar PP, Berger GS. A prospective, randomized study of oral contraceptives: the effect of study design on reported rates of symptoms. Contraception 1979;20:329-37. 8. Burkman RT, Zacur HA, Kimball AW, et al. Oral contraceptives and lipids and lipoproteins. I. Variations in the mean levels by oral contraceptive type. Contraception 1989;40:553-61. 9. Burkman RT, Zacur HA, Kimball AW, et al. Oral contraceptives and lipids and lipoproteins. II. Relationship to plasma steroid levels and outlier status. Contraception 1989;40:676-89. 10. Zacur HA, Linkins S, Chang V, Smith B, Kimball AW, Burkman R. Ethinyl estradiol and norethindrone radioimmunoassay following Sephadex LH-20 column chromatography. Clin Chern Acta 1991;204:209-15. 11, Snowden R, Christian B, eds. Patterns and perceptions of menstruation. London: Croom Hel, 1983. 12. Belsey EM. The association between vaginal bleeding patterns and reason for discontinuation of contraceptive use. Contraception 1988;38:207-25. 13. Hill GA, Wheeler JM. Incidence of breakthrough bleeding during oral contraceptive therapy. J Reprod Med 1991; 36:334-9. 14. Goldzieher JW. Pharmacology of contraceptive steroids: a brief review. AM J OBSTET GYNECOL 1989;160:1260-4. 5. Back DJ, Breckenridge AM, Crawford FE, et al. Kinetics of norethindrone in women. I. Radioimmunoassay and concentrations during multiple dosing. Clin Pharmacol Ther 1978;24:439-47.
Discussion DR. RUDI ANSBACHER, Ann Arbor, Michigan. Although birth control pills were introduced for clinical use 32 years ago, there is still a paucity of information concerning how the steroids contained therein are metabolized or how side effects of these medications are correlated with the steroid content. This study, which emphasizes bleeding patterns of patients on three different formulations of oral contraceptives, sheds little light on what causes the differences from one patient to another. The authors state that there was no correlation between baseline or I-hour slopes of contraceptive steroids and intermenstrual bleeding. However, this is based on one blood sample obtained just before ingestion of the pill on day 21 of
1746 Saleh et al.
the cycle (approximately 24 hours after ingestion of the previous pill) and a second blood sample taken 1 hour after ingestion of the pill on the same day. Helton et al. I. 2 in the mid-1970s were the first to note the marked variability in an individual's metabolism of the active hormones, especially the estrogenic component, contained in oral contraceptives. In 1989 Goldzieher" emphasized that the blood levels from a given oral dose of ethinyl estradiol and norethindrone vary widely from patient to patient because of the partial inactivation of the steroids on their first pass through the enterohepatic circulation. Therefore wide ranges in bioavailability can occur in an individual and between individuals. In a 1992 review article Zacur and Stewart' summarized the information currently known about the mechanism involved in the metabolism of the oral contraceptive pill steroids and the effects that other drugs and medications may have on subsequent steroid levels. However, there are still many gaps in our knowledge. Ansbacher' in 1991 stated that the use of a single day test period for comparing low-dose contraceptive agents could not be extrapolated to the entire 28-day cycle, because the individual patient variation in the absorption of the active ingredients and the latter's subsequent metabolic breakdown cannot be assured. In addition, mean patient data were not sufficient to ensure similar bioavailability of a product in a particular patient. In the study we are reviewing today the bioavailability data are not given for individual patients but for the entire group, and these data are expressed as means. Were there differences noted in individual patients, were some patients outliers, and if there were outliers, how did these affect the mean data for the entire group? Can one really accept data from hormonal assays that have a between-assay variation of 18.7% for ethinyl estradiol and 12.5% for norethindrone? Correlation of hormonal levels with the status of the endometrium would enable us to properly adjust dosage to correct the bothersome symptom of intermenstrual bleeding in individual patients. Admittedly, this would be most costly, but without so doing we will be left with the intermenstrual bleeding patterns described in this study. REFERENCES 1. Helton ED, Williams MC, Goldzieher ]W. Human urinary and liver conjugates of 17 a-ethynylestradiol. Steroids 1976; 27:851-67. 2. Helton ED, Simmons R, Meltz ML, Goldzieher ]W. Variability in conjugates of ethynylestradiol produced by in vitro incubation of human liver tissue. Contraception 1977;16: 256-60. 3. Goldzieher ]W. Pharmacology of contraceptive steroids: a briefreview. AM] OBSTET GYNECOL 1989;160:1260-4. 4. Zacur HA, Stewart D. New concepts in oral contraceptive pill use. Curr Opin Obstet Gynecol 1992;4:365-71. 5. Ansbacher R. Interchangeability oflow-dose oral contraceptives. Are current bioequivalent testing measures adequate to ensure therapeutic equivalency? Contraception 1991;43: 139-47.
June 1993 Am J Obstet Gynecol
DR. FRANK W. LING, Memphis, Tennessee. Complications of oral contraceptives, such as breakthrough bleeding, are commonly referred to as only "minor risks" or "side effects." Nevertheless, it is these annoyances that are the primary reason for noncompliance. Minimizing or eliminating these side effects could help patients more effectively use the contraceptive products available. Twenty percent is a commonly quoted incidence of breakthrough bleeding for oral contraceptive users, with higher rates occurring with low-dose preparations. The optimal combination oral contraceptive agent will, for a given individual, maximize contraception while minimizing untoward side effects. In a study originally designed to evaluate the relationship between oral contraceptives and changes in lipoprotein and lipid levels, the authors report bleeding patterns of 192 women on three different pills. The authors also attempted to correlate these biologic responses with the bioavailability of plasma steroids. Both goals of the study have been at least partially achieved, because the authors have successfully described bleeding patterns in these three groups and determined that there is no correlation of either baseline value or I-hour slopes of ethinyl estradiol and norethindrone at day 21 of the cycle with the amount or type of bleeding observed. Several questions regarding definitions and methods do remain, however. For example, the authors state that women on higher dosages of either steroid had intermenstrual bleeding in continuation with the withdrawal period. This seems to contradict Fig. 5, in which the number of withdrawal bleeding days per cycle for OC 1 and OC2 did not exceed 5 days during any of the nine cycles. For a woman to have intermenstrual bleeding continuous with withdrawal bleeding, according to the authors' definitions, withdrawal bleeding would have to have occurred through day 28 and into day 1 of the subsequent cycle. Should this day 1 bleeding really be considered "intermenstrual"? Clarification of this point would be helpful. Also, because there was no attempt to quantitate the amount of bleeding on any given day, could the authors speculate on their observations regarding this very crucial issue? Because patients will often accept a small amount of spotting but will not accept a large amount of bleeding, could this lack of quantification of bleeding seriously affect the interpretation and clinical use of these data? Also, because the number of intermenstrual bleeding days was fewer than 3 days per month over the 9-month period, even for the OC3 users, was this felt to be a significant problem by the subjects when interviewed? It was further stated that there was no evidence that patients were lost to follow-up or stopped participation because of an inability to tolerate their bleeding pattern. It was interesting to note that the other side effects such as amenorrhea, dysmenorrhea, nausea, headache, depressed mood, breast tenderness, and acne had frequencies ranging from 0.5% to 44.3%. Were these the reasons cited for discontinuing the pill, and if so, are there similar frequencies of these side effects in the
Volume 168, Number 6, Part 1 Am J Obstet Gynecol
patients who did complete the study? If these side effects were not the reason the patients dropped from the study, I would be interested in knowing why they did, I would be particularly interested in knowing which side effects had an incidence of 26.5% and 44.3%. In Fig. 1, in comparing intermenstrual bleeding rates in each cycle among the three preparations, the authors state that the ninth cycle data were discarded because of limited numbers. Because the study described 192 patients who completed 9 months, this loss of data in the final month of the study is confusing. I noted that the number of days that elapsed between the last active pill day and the onset of withdrawal bleeding was analyzed in only those cycles in which there had not been intermenstrual bleeding. Why was this specific exclusion used? Because intermenstrual bleeding could include even a small amount of spotting early in the cycle, the authors should explain what they feel the relationship might be between intermenstrual bleeding early in the cycle with withdrawal bleeding at the end of the cycle, which would warrant exclusion of these cycles. The authors point out that their incidence of intermenstrual bleeding appears above the range usually reported for monophasic pills and may be caused by their failure to exclude women with a previous history of menstrual irregularities. Because a previous history of menstrual irregularity was not controlled for, were other factors which might affect menstrual disturbance or oral contraceptive metabolism similarly not controlled? A final consideration is whether there were assurances as to the timing of ingestion of the pill on a daily basis. Because the authors excluded cycles in which a pill was missed, should a similar adjustment have been made if a pill had been taken at a markedly different time of the day? DR. SALEH (Closing). Although much of the clinical information presented in this study relative to pill dosage and vaginal bleeding are known by most clinicians, a significant goal was to attempt to correlate clinical behavior of birth control pills and steroid bioavailability. We felt this was a unique opportunity to study the subject in depth by means of a randomized clinical trial. We hoped to relate bleeding patterns to contraceptive sex steroid levels for the time period and samples available to us. Unfortunately, as discussed, we were unable to demonstrate a significant relationship. As pointed out by Dr. Ansbacher, I agree that multiple samples drawn at different time intervals would have been helpful. However, because of cost restraints, the two time periods were chosen. They were initially
Saleh et al.
1747
chosen on the basis of a pilot study including 58 patients followed with multiple samples over a 24-hour period. The I-hour slope was found to be a reliable indicator of steroid metabolic behavior when lipid levels were concerned. We know that ethinyl estradiol and norethindrone tend to peak within a few hours of oral administration and then progressively decrease afterward. We are aware that ethinyl estradiol and norethindrone concentrations vary markedly over a 24-hour period, and this may complicate analyses of steroid bioavailability and biologic responses. Bleeding may be related to the amount of variation rather than hormone level. Relative to bleeding in patients lost to follow-up, the main reason for patients discontinuing the study was relocation out of the Baltimore area. In those patients, the incidence of bleeding was 44% and was not statistically different from the study group. Relative to the pattern of bleeding in the withdrawal period, the number of bleeding days within the withdrawal period, or days 21 to 28, was higher for the higher dosage preparation only during the first 3 months of the study. Those who had the highest endometrial stability during the first part of the cycle tended to bleed more during the withdrawal period, whereas those on the lowest steroid dosage preparation bled more during the intermenstrual period, or days 1 to 21. In one graph patients were excluded from the ninth month. The intermenstrual bleeding rates were omitted because the last visit often occurred in the middle of the ninth month, making calculation of a rate for that month impossible. What were the factors in the patients' history that may have influenced the outcome? Patients selected for the study were between 18 and 35 years old, had regular menstrual cycles lasting between 25 and 35 days, and did not have any contraindication to birth control pills or previous endocrinologic or metabolic disorders. Body weights were within 115% of the ideal body weight. Patients did not breast-feed, become pregnant, or take steroid or birth control pills within 6 months of the study. Why were higher rates of bleeding found in this study? As stated, women with previous menstrual irregularities were excluded from the study. Comparison between different studies is difficult because of the difference in study design and because the definition of intermenstrual bleeding is different. We used a liberal definition, including all bleeding irrespective of the amount, whereas other authors have used bleeding enough to stain a pad or significant enough to be a problem to the patient.