Safe and effective birth control is important to most young women, and over 90% of sexually active US women at risk of unintended pregnancy use contraception.1 Half of the 21 million US women who use reversible contraception choose oral contraceptives (OCs), and there are over half a million unintended pregnancies per year among these women.2–4 In spite of the high number of OC-associated pregnancies, biologic risk factors for OC failure largely remain unexamined.
One such factor of possible relevance is body weight. In a 1980 letter to the Medical Journal of Australia, Boden reported a cluster of low-dose OC failures among relatively heavy women (Boden DC. Unplanned pregnancies and the pill [letter]. Med J Aust 1980;1:391). Also in 1980, a small study by Stadel et al of ethinyl estradiol (EE) blood levels in OC users found that women in the lowest serum EE quartile had a nonsignificantly higher mean weight than those in the highest quartile (142.2 versus 133.5 lb), indicating that OC metabolism may be enhanced among heavy women.5 No more recent research on this topic is evident, although the lower-dose OCs currently marketed may have a greater impact on failure likelihood among heavy women. We conducted this analysis to investigate the relationship between a woman's weight and her risk of pregnancy while using OCs, using a population-based cohort of health maintenance organization enrollees.
MATERIALS AND METHODS
In this retrospective cohort analysis, we used data obtained in a case-control study of functional ovarian cysts and neoplasms conducted within Group Health Cooperative of Puget Sound (GHC) in western Washington State. This study was approved by the Fred Hutchinson Cancer Research Center and GHC Institutional Review Boards. All subjects in that study were female GHC enrollees aged 18–39 years, ascertained between January 1, 1990, and June 30, 1994. A total of 919 randomly selected potential control subjects (matched to case subjects on age and GHC clinic region) were approached for participation, of whom 755 (82.2%) completed an in-person interview and a dietary questionnaire after informed consent was obtained. We excluded potential controls from the ovarian cyst study if they had a bilateral oophorectomy or less than 6 months GHC enrollment before their reference date (dates assigned to controls to correspond to case subjects' diagnosis dates), or did not speak English. The 618 control subjects with any reported use of OCs for contraceptive reasons before the reference date provided the data for the analyses described in this report.
Study participants were interviewed in person by trained female interviewers who used a structured questionnaire that elicited a lifetime history of a variety of demographic, medical, and lifestyle factors, including contraceptive use. We used a life-events calendar and color photographs of all OC pills marketed in the United States to help women recall dates and specific brands of OCs used. For each reported pregnancy, we asked the participant if the pregnancy was confirmed (by serum or urine assay or by ultrasound examination), and what type of contraception she was using, if any, when she became pregnant. Height and body weight at reference date were obtained by participant self-report. We categorized weight by quartile, using the entire study group to define the quartiles.
The primary outcome of interest in these analyses was the number of pregnancies occurring while using OCs per 100 person-years of OC use, for women in each body weight quartile. Participants' entry into the cohort was the date of first OC use for contraception, and person-years of OC use included all subsequent periods of contraception-related OC use up until reference date. All analyses were performed using Stata Statistical Software Release 6.0 (Stata Corporation, College Station, TX) to calculate Cox proportional hazards regression,6 with failure defined as a confirmed pregnancy while using OCs, modeled as a function of time from entry into cohort. We plotted the Kaplan-Meier product-limit estimates of the survival function by weight quartile, to allow a visual comparison of OC failure likelihood. We then conducted Cox regression analyses to estimate the relative risk (RR) of OC failure associated with high body weight, allowing multiple failures per subject, and adjusting the standard errors for dependencies between failure times. We also conducted subanalyses limited to at most one failure per subject, with follow-up ending at first OC failure or at the date of last OC use for women with no failure, to investigate the possibility that individual susceptibilities could be partially responsible for any associations seen. We initially calculated RRs comparing each of the highest three body weight quartiles to the lowest quartile (reference group). Because the three lowest quartiles did not differ significantly and because we hypothesized a priori that there might be a threshold effect, to improve study power we then compared the highest body weight quartile with the three lowest quartiles combined (reference group). We tested for the interaction of body weight with smoking and race using the likelihood ratio test; no significant (P < .05) interaction was noted.
In multivariable models, the following time-dependent variables were evaluated for possible confounding effects: age (continuous and 5-year categories), smoking status (never, current, former), gravidity (0, 1, 2 or more), parity (0, 1, 2 or more), menstrual cycle regularity (10-day variation or less, 10-day variation or more), and time from end of OC episode to reference date weight measurement (less than 5 years, 5 to less than 10 years, 10 years or more). The time-independent variables, race (white, black, other), religion (none, Catholic, non-Catholic Christian, non-Christian), and age at reference date (continuous) were similarly examined. We included potential confounding variables in multivariable models if they changed the RR for OC failure associated with quartile of body weight by 10% or more.7
We conducted one set of subanalyses stratified by OC estrogen dose, allocating OC dose-specific exposure time for the various types of OCs a subject may have used. We classified as low dose all monophasic OCs with less than 50 μg of EE or less than 80 μg of mestranol, and as high dose all monophasic OCs with 50 μg or more of EE or 80 μg or more of mestranol. Because we wanted to determine if even a lower estrogen dose affected risk of OC failure, we conducted one analysis among users of very low-dose monophasic OCs, which we defined as less than 35 μg of EE.
To assess whether women with high body weight might have lower compliance, and therefore, higher failure rates with any user-dependent contraceptive method than other women, in one analysis we calculated barrier contraceptive failure rates by body weight and RRs associated with weight quartile, using the methods described above.
Women in quartile 1 (less than 56.5 kg) were more likely to be Asian, and women in quartiles 3 and 4 (62.5 to less than 70.5 kg and 70.5 kg or more, respectively) were more likely to be black (Table 1). At entry into the cohort (first OC use for contraception), smoking status also differed by quartile: women in quartile 4 were more likely to be smokers. Women in quartile 4 also were more likely to have had two or more pregnancies at the time of entry into the cohort. Study subjects were followed for 12.5 years on average, with a mean of 12.0 years of follow-up for women in the lowest weight quartile, 12.6 years for quartile 2 women, 12.6 years for quartile 3 women, and 13.2 years for women in the highest quartile. Mean ages at the conclusion of follow-up (reference date) from the lowest to highest weight quartile were 31.4 years, 31.4 years, 31.6 years, and 32.3 years, respectively. Over half of the study subjects had had two or more pregnancies by the end of follow-up, with little variation by weight quartile (quartile 1: 51.5%, quartile 2: 50.0%, quartile 3: 52.5%, quartile 4: 52.6%). Mean number of months of OC use during follow-up did not differ by weight quartile (quartile 1: mean 57.3, standard deviation [SD] 46.9, quartile 2: mean 54.7, SD 46.2, quartile 3: mean 53.2, SD 45.2, quartile 4: mean 54.4, SD 47.9).
There were 106 reported confirmed pregnancies in the cohort during 2821.8 person-years of OC use (3.8 per 100 person-years of exposure). The graph of the Kaplan-Meier survivor function indicated there were differences in OC failure likelihood by body weight quartile (Figure 1). In a multivariable model controlling for parity, we found a significantly elevated risk of reported pregnancy while using OCs (RR 1.6, 95% confidence interval [CI] 1.1, 2.4) associated with membership in the highest body weight quartile, compared with the other three quartiles combined (Table 2). In the subanalysis limited to one failure per subject shown in Table 3, we found that the association between the highest weight quartile and OC failure persisted (RR 1.7, 95% CI 1.1, 2.9).
In a set of subanalyses stratified by estrogen dose of monophasic OCs, weight-associated risk of OC failure was particularly elevated among women using very low-dose OCs (RR 4.5, 95% CI 1.4, 14.4) or low-dose OCs (RR 2.6, 95% CI 1.2, 5.9), after adjustment for parity, race, religion, and menstrual cycle regularity (Table 4). High-dose OC use was not associated with an increased risk of failure (RR 1.2, 95% CI 0.4, 3.5). The parity-adjusted risk of reported pregnancy while using barrier contraception was not significantly different among women in the highest body weight quartile than among other women (RR 1.2, 95% CI 0.8, 1.8).
In this study, women in the highest quartile of body weight (70.5 kg or more) had a 60% higher risk of OC failure than women of lower weight. Further, the increased risk of OC failure associated with weight was higher for women using very low-dose or low-dose OCs. Boden speculated that such relationships might exist in a letter about a cluster of OC failures among Australian women weighing over 60 kg, but did not conduct a formal analysis of the question (Boden DC. Unplanned pregnancies and the pill [letter]. Med J Aust 1980;1:391). A recent report of a transdermal contraceptive patch trial had similar findings; the five women who became pregnant while using the patch had a mean weight of 74.2 kg, compared with a mean weight of 64.4 kg for the entire study group.8 Our results perhaps most closely parallel those of prior investigations by Gu et al of the effectiveness of contraceptive implants in a trial of 10,718 Chinese women.9,10 These authors found that pregnancy rates correlated significantly with weight, and the cumulative 7-year failure rate among women weighing 70 kg or more was over five times that of women weighing less than 50 kg. Although contraceptive implants contain only progestin, and OCs contain both estrogen and progestin, we speculate that similar mechanisms might explain potential associations between high body weight and contraceptive failure.
One possible explanation for OC failure among hormonal contraceptive users of high weight is that an overall enhanced metabolic rate leads to more rapid drug metabolism and subsequent insufficient serum progestin or estrogen levels for good contraceptive efficacy. Absolute weight is highly correlated with resting metabolic rate,11–13 and two basal metabolic rate equations that use weight alone have been found to correctly predict basal metabolic rate in OC users.14 Because weight is more strongly associated with metabolism than is body mass index (BMI),11 and because prior research in this area has been based on weight (Boden DC. Unplanned pregnancies and the pill [letter]. Med J Aust 1980;1:391),5,9,10 we defined body weight as the independent variable in this cohort study. We did conduct exploratory analyses recategorizing the cohort according to BMI quartile, however, and found that the likelihood of OC failure associated with the highest BMI quartile was similar to that associated with the highest body weight quartile (parity-adjusted RR 1.65). Additionally, although the number of women in each weight-BMI subcategory was too small to obtain conclusive results, there was some indication that within each weight category, higher BMI was associated with higher OC failure likelihood (data not shown). Because contraceptive steroids are lipids and can be absorbed in fat, it is plausible that, among women with high fat levels, decreased circulating steroid levels may contribute to the decreased effectiveness of OCs.
Not all women remembered the brand and dose of the OCs they had used in the past, limiting the number of person-years of exposure available for the dose-specific subanalyses. The overall recall of OC use in interviews using a photo album or life-event calendar prompt has generally been found to be excellent, but recall of specific brands may be less accurate.15–17 Although we could not confirm the accuracy of our participants' recall of OC use and brands, if misclassification was present, our results would only be an underestimate of the true risk of OC failure associated with body weight. One advantage of our inquiries about historical OC use was that we were able to obtain information about higher-dose OCs than are currently in common use, leading to our observation that the impact of high body weight was greater among women using monophasic OCs with less than 50 μg of EE than among those using monophasic OCs with 50 μg or more of EE. Although based on small numbers, the higher elevation of weight-related risk we found among users of very low estrogen dose OCs supports the possibility that the associations seen were not spurious.
A limitation of this analysis, which used data originally collected for another purpose, is that we did not have knowledge of women's weight throughout their reproductive lives, and medical records were not available to us. Our weight assessment was by self-report at the reference date of the ovarian cyst study, at the end of OC follow-up for these analyses. Subjects' reference dates were on average 76.5 months after the last OC use and did not vary significantly by weight quartile. It may be hypothesized that women who have experienced a full-term pregnancy have higher subsequent body weight on average than nulliparous women, and that this circumstance might be partially responsible for the observed associations. In one analysis limited to parous women, we found similar patterns of association between OC failure and high body weight (RR 2.1, 95% CI 1.4, 3.2), reassuring us that this is not a likely explanation for our findings. The accuracy of body weight reported by women participating in medical research has occasionally been questioned. Several researchers have investigated the validity of self-reported weight and height, and the majority have found that only small errors result.18–26 In general, under-reporting of high weights and over-reporting of low weights has been noted, which may lead to an underestimate of the prevalence of obesity but should have little impact on quartile-based weight analyses. If misclassification of body weight quartile was present, our results would be an underestimate of the real risk of OC failure associated with high body weight.
Bias in our results would also be possible if heavier women over-reported the occurrence of pregnancy, perhaps because of an increased likelihood of irregular or prolonged menstrual cycles that were misinterpreted as spontaneous abortions. We think this is an unlikely possibility for two reasons: 1) all the relevant pregnancies were reported during periods in which participants were using OCs and were therefore not likely to have irregular cycles, and 2) all OC-related pregnancies were confirmed (by positive pregnancy test or ultrasound examination) according to the participants.
In pharmaceutical company-sponsored clinical trials, OC method failure is usually reported as around 0.1–0.3 pregnancies per 100 person-years of use.27–29 In contrast, typical failure rates among US users range from 3 to 8 per 100 person-years,2,4,30–34 leading some to speculate that most OC failure is due to women's failure to take OCs exactly as prescribed. We had no knowledge about the consistency of OC pill-taking behavior among our cohort, and one could hypothesize that less conscientious contraceptive use by heavy women might account for their increased likelihood of OC failure. In an attempt to address this issue of compliance, we considered in one set of analyses the risk of barrier contraceptive failure by body weight quartile, and found no association with the highest quartile. The lack of an association between barrier contraceptive failure and high body weight suggests that method failure rather than lack of compliance may be the primary mechanism for OC-associated failure among our heavy subjects.
The steroid hormone content of combination OCs has decreased dramatically over time, from 150 μg of EE in the 1960s to more recent formulations with an average dose of less than 35 μg of EE.35 Suppression of endogenous estrogen production in OC users may depend on estrogen dosage.29 Consequently, low-dose OCs may be approaching threshold levels of efficacy and good cycle control.36 Women's health care providers are encouraged by contraceptive experts to use the lowest OC dose possible, to decrease the possibility of adverse effects and increase the likelihood of compliance.28,29 These recommendations do not usually take into consideration possible metabolic variations among OC users. Our findings of increased risk of OC failure associated with high body weight, if replicated in larger studies with prospective objective documentation of both weight and periconceptional contraceptive use, suggest that body habitus may affect metabolism sufficiently to compromise contraceptive effectiveness in this subset of OC users. Consideration of a woman's weight may be an important element of OC prescription.
1. Abma J, Chandra A, Mosher W, Peterson L, Piccinino L. Fertility, family planning and women's health: New data from the 1995 National Survey of Family Growth. Vital Health Statistics 23. Washington, DC: National Center for Health Statistics, 1997.
2. Forrest JD, Singh S. Public-sector savings resulting from expenditures for contraceptive services. Fam Plann Perspect 1990;22:6–15.
3. Rosenberg MJ, Waugh MS, Long S. Unintended pregnancies and use, misuse and discontinuation of oral contraceptives. J Reprod Med 1995;40:355–60.
4. Fu H, Darroch JE, Haas T, Ranjit N. Contraceptive failure rates: New estimates from the 1995 National Survey of Family Growth. Fam Plann Perspect 1999;31:56–63.
5. Stadel BV, Sternthal PM, Schlesselman JJ, Douglas MG, Hall WD, Kaul L, et al. Variation of ethinyl estradiol blood levels among healthy women using oral contraceptives. Fertil Steril 1980;33:257–60.
6. Cox DR. Regression models and life tables. J R Stat Soc Serv B 1972;34:187–202.
7. Maldonado G, Greenland S. Simulation study of confounder selection strategies. Am J Epidemiol 1993;138:923–36.
8. Audet M-C, Moreau M, Koltun WD, Waldbaum AS, Shangold G, Fisher AC, et al. Evaluation of contraceptive efficacy and cycle control of a transdermal contraceptive patch vs. an oral contraceptive. JAMA 2001;285:2347–54.
9. Gu S-J, Du M-K, Zhang L-D, Liu Y-L, Wang S-H, Sivin I. A 5-year evaluation of NORPLANT contraceptive implants in China. Obstet Gynecol 1994;83:673–8.
10. Gu S-J, Sivin I, Du M, Zhang PD, Ying L, Meng F, et al. Effectiveness of Norplant implants through seven years: A large-scale study in China. Contraception 1995;52:99–103.
11. Owen OE, Kavle E, Owen RS, Polansky M, Caprio S, Mozzoli MA, et al. A reappraisal of caloric requirements in healthy women. Am J Clin Nutr 1986;44:1–19.
12. Owen OE. Resting metabolic requirements of men and women. Mayo Clin Proc 1988;63:503–10.
13. Ravussin E, Burnand B, Schutz Y, Jequier E. Twenty-four-hour energy expenditure and resting metabolic rate in obese, moderately obese, and control subjects. Am J Clin Nutr 1982;35:566–73.
14. Piers LS, Diffey B, Soares MJ, Frandsen SL, McCormack LM, Lutshini MJ, et al. The validity of predicting the basal metabolic rate of young Australian men and women. Eur J Clin Nutr 1997;51:333–7.
15. Norell SE, Boethius G, Persson K. Oral contraceptive use: Interview data versus pharmacy records. Int J Epidemiol 1998;27:1033–7.
16. Hunter DJ, Manson JE, Colditz GA, Chasan-Taber L, Troy L, Stampfer MJ, et al. Reproducibility of oral contraceptive histories and validity of hormone composition reported in a cohort of US women. Contraception 1997; 56:373–8.
17. Coulter A, Vessey M, McPherson K. The ability of women to recall their oral contraceptive histories. Contraception 1986;33:127–37.
18. Bolton-Smith C, Woodward M, Tunstall-Pedoe H, Morrison C. Accuracy of the estimated prevalence of obesity from self reported height and weight in an adult Scottish population. J Epidemiol Community Health 2000;54:143–8.
19. Roberts RJ. Can self-reported data accurately describe the prevalence of overweight? Public Health 1995;109:275–84.
20. Kuskowska-Wolk A, Bergstrom R, Bostrom G. Relationship between questionnaire data and medical records of height, weight and body mass index. Int J Obes Relat Metab Disord 1992;16:1–9.
21. Rowland ML. Self-reported weight and height. Am J Clin Nutr 1990;52:1125–33.
22. Le Marchand L, Yoshizawa CN, Nomura AMY. Validation of body size information on driver's licenses. Am J Epidemiol 1988;128:874–7.
23. Stewart AW, Jackson RT, Ford MA, Beaglehole R. Underestimation of relative weight by use of self-reported height and weight. Am J Epidemiol 1987;125:122–6.
24. Stewart AL. The reliability and validity of self-reported weight and height. J Chronic Dis 1982;35:295–309.
25. Palta M, Prineas RJ, Berman R, Hannan P. Comparison of self-reported and measured height and weight. Am J Epidemiol 1982;115:223–30.
26. Stunkard AJ, Albaum JM. The accuracy of self-reported weights. Am J Clin Nutr 1981;34:1593–9.
27. Trussell J, Kost K. Contraceptive failure in the United States: A critical review of the literature. Stud Fam Plann 1987;18:237–83.
28. Hillard PJA. Oral contraception noncompliance: The extent of the problem. Adv Contracept 1992;8(Suppl.):13–20.
29. Hatcher RA, Trussell J, Stewart F, Cates W, Stewart GK, Guest F, et al. Contraceptive technology. 16th ed. (revised). New York: Irvington Publishers; 1994.
30. Jones EF, Forrest JD. Contraceptive failure rates based on the 1988 NSFG. Fam Plann Perspect 1992;24:12–9.
31. Kost K, Forrest JD, Harlap S. Comparing the health risks and benefits of contraceptive choices. Fam Plann Perspect 1991;23:54–61.
32. Jones EF, Forrest JD. Contraceptive failure in the United States: Revised estimates from the 1982 National Survey of Family Growth. Fam Plann Perspect 1989;21:103–9.
33. Grady WR, Hayward MD, Yagi J. Contraceptive failure in the United States: Estimates from the 1982 National Survey of Family Growth. Fam Plann Perspect 1986;18:200–9.
34. Grady WR, Hirsch MB, Keen N, Vaughan B. Contraceptive failure and continuation among married women in the United States, 1970–75. Stud Fam Plann 1983;14:9–19.
35. Gerstman BB, Gross TP, Kennedy DL, Bennett RC, Tomita DK, Stadel BV. Trends in the content and use of oral contraceptives in the United States, 1964–88. Am J Public Health 1991;81:90–8.
36. Goldzieher JW. Pharmacology of contraceptive steroids: A brief review. Am J Obstet Gynecol 1989;160:1260–4.