Materials and Methods
This cross-sectional study was done at seven centers in three regions of the developing world (Dhaka, Bangladesh; Campinas, Brazil; Shanghai, China; Cairo, Egypt; Mexico City, Mexico; Bangkok, Thailand; and Harare, Zimbabwe) from April 1994 to June 1997. It was approved by the Scientific and Ethical Review Group of the United Nations Development Programme (UNDP)/ United Nations Population Fund (UNFPA)/World Health Organization (WHO)/World Bank Special Programme of Research, Development and Research Training in Human Reproduction, and by ethical review committees at participating sites.
Women 30–34 years old were eligible for the study if they were not lactating or had not recently lactated; were not recently pregnant; had not had a hysterectomy or oophorectomy; did not use, currently or in the past, medications known to affect calcium metabolism; and did not have a chronic disease affecting calcium metabolism (Table 1, footnote).
Women attending family planning clinics in the seven centers who had at least 24 months of lifetime use of OCs, DMPA, or levonorgestrel implants were recruited consecutively, along with a comparison group of women (called never users of hormonal contraceptives) who had never used any hormonal contraceptive or had a lifetime exposure to hormonal contraceptives of less than 6 months. Women who had used two or more hormonal methods were assigned to the hormonal method most recently used for 2 or more years.
Of women who had never used hormonal contraceptives, 78% had never used any form of hormonal contraception. The mean duration of hormonal contraceptive use in the women classified as never users who had used hormonal contraceptives was 3 months (standard deviation [SD] 1.6 months) and the mean time since stopping use was 78 months (SD 50 months). Eighty-two percent (82%) of the assigned OC users used formulations containing 30 μg or more but less than 50 μg of estrogen, 15% used formulations containing 50 μg or more, and fewer than 1% used OCs containing less than 30 μg. The estrogen dose was unknown for 2% of the OC users. All DMPA users were on regimens of 150 mg every 3 months. At the time of this study, only the six-capsule subdermal levonorgestrel implant system was available. It releases 50–85 μg of levonorgestrel per day in the first 9 months of use and about 35 μg per day after 1 year.8
Bone mineral density (BMD) was measured at the distal radius, where trabecular bone predominates, and the midshaft of the ulna, where cortical bone predominates, in grams/centimeter2 (g/cm2) using single x-ray absorptiometry (SXA model DTX-100; Osteometer Meditech A/S Co., Rodovre, Denmark). Single x-ray absorptiometry was used because of its high reproducibility (coefficient of variation approximately 1%), its low cost, and ease of training.
The test/retest reliability of BMD measurements was assessed at each center during a pilot phase; additional training was offered at centers with low reliability. During the main study, ongoing assessment of test/retest reliability using the intraclass correlation coefficient9 showed values of 0.95 or more in four of the five centers with more than ten subjects. In that center, test-retest reliability for measures at the midshaft of the ulna was 0.90.
Information on other characteristics of subjects was collected in face-to-face interviews. Height and weight were measured using a standard protocol and stadiometers and scales provided by the WHO. Occupation was used as a surrogate for socioeconomic status because income levels cannot be compared across these diverse international settings. Occupation was classified into three groups according to the International Standard Classification of Occupations, Revised Edition 1968. High occupation included professional, administrative and management workers, and military officers. Medium occupation included clerical and sales workers. Low occupation included service workers, agricultural workers, and military workers other than officers. Subjects were classified according to both their own and their partner's occupation. Partners' occupation was used in the adjustment set because it was most strongly associated with BMD.
The characteristics of women in the study were quantified as mean ± SD, medians, or percentages, as appropriate. Differences in BMD across centers and the associations between BMD and selected characteristics of the study subjects and by contraceptive user group were assessed using analysis of variance techniques. Analysis of covariance was used to estimate BMD for each variable adjusting for center, for contraceptive user group, and for variables in an adjustment set, which consisted of variables for which the P value was less than .10 for BMD measured at both anatomic sites. For correlated variables, the one with highest partial R2 in a model that was also adjusted for centers was selected for the adjustment set. When examining a variable not in the adjustment set that was correlated with a variable in the adjustment set, the latter was dropped from the model.
Adjusted mean BMD was estimated according to the assigned hormonal contraceptive method and separately in women who were exclusive users of the method. These results are presented as adjusted mean differences between BMD in users of the contraceptive method and the comparison group of never users of hormonal contraceptives along with the 95% confidence interval (CI) for the differences in adjusted means.
We did analyses, restricted to exclusive users of each hormonal method, after categorizing women according to recency of hormonal contraceptive use (current or past use) and in categories of duration of use (less than 36 months, 37–48 months, and over 48 months).
A total of 2545 women were recruited to the study, but 71 women who were ineligible were excluded from this analysis. Table 1 gives the number of women in the analysis, categorized by contraceptive use and center. There were differences in most of the characteristics of women by hormonal contraceptive group (Table 2). This table also shows median duration of use for all users and for exclusive users of hormonal contraceptives.
Table 3 shows mean BMD at each anatomic site according to selected characteristics and by contraceptive user group, first adjusting only for center and then adjusting for center and the adjustment set. Bone mineral density was significantly (P < .001) associated with center.
After adjusting only for center, there were statistically significant (P < .05) associations of age, body mass index (BMI), years of lactation, and years since last lactation, with BMD measured at both the distal radius and the midshaft of the ulna. An association of parity with BMD was also observed, although at midshaft of ulna it was not significant. Bone mineral density was also significantly associated with occupation of the woman's partner. This association was complex, with both low and high occupation being associated with lower BMD. Higher BMD measured at the distal radius was also significantly (P < .05) associated with ever drinking coffee.
After adjustment only for center, there was a significant association of contraceptive method used most recently with BMD at both the distal radius and the midshaft of ulna. The significant association between BMD and centers persisted after adjustment for other variables in the table. After adjustment for center, assigned contraceptive method, and the variables in the adjustment set, the associations of age, BMI, occupation of the partner, coffee drinking, years of lactation, years since last lactation, and parity persisted for measurement of BMD at the distal radius. For measurement of BMD at the midshaft of ulna, after the same adjustment, only the associations between BMD and BMI, occupation of partner, and years of lactation persisted. After adjustment for center and the variables in the adjustment set, the significant associations of hormonal contraceptive method with BMD at the distal radius persisted.
Figure 1 shows the adjusted mean differences and 95% CIs for BMD between users of each hormonal contraceptive and never users of steroid hormone contraceptives according to the assigned hormonal contraceptive method and for exclusive users of each hormonal method. For women assigned to OCs and exclusive users of OCs, there were no differences in the adjusted means of BMD compared with never users of hormonal contraceptives. Users assigned to DMPA had significantly lower adjusted mean BMD at the distal radius. Exclusive users of DMPA had significantly lower adjusted mean BMD at both anatomic sites. In the analysis restricted to exclusive users of levonorgestrel implants, the adjusted mean BMD at the midshaft of the ulna was significantly lower for levonorgestrel implant users compared with never users of hormonal contraceptives. All of the adjusted mean differences in BMD between users of hormonal contraceptives and never users were of a magnitude that was less than one SD of the mean of never users of hormonal contraceptives.
Figure 2, which is restricted to exclusive users of each hormonal contraceptive method and never users, shows a consistent pattern in which use of hormonal contraceptives was significantly associated with BMD only in current users who used the method for a relatively short duration (24–36 months). For women who used OCs, adjusted mean BMD was higher in exclusive short-term current users than in never users of hormonal contraceptives (+0.017 g/cm2 for the distal radius; +0.016 g/cm2 for the midshaft of ulna). For those who used DMPA, the adjusted mean BMD was significantly lower in exclusive short-term current users than in never users of hormonal contraceptives (−0.023 g/cm2 for the distal radius; −0.014 g/cm2 for the midshaft of ulna). For women who used levonorgestrel implants, the adjusted mean BMD was significantly lower in exclusive short-term current users than in never users of hormonal contraceptives (−0.013 g/cm2 for the distal radius; −0.017 g/cm2 for the midshaft of ulna). For all three hormonal methods, there were no significant differences in BMD measured at either anatomic site between exclusive past users of the method and never users of hormonal contraceptives, nor were there differences between exclusive longer-term current users of any of the methods and never users of hormonal contraceptives. All observed associations pertain equally to BMD measured at the distal radius and the midshaft of ulna. All of these differences found in current short-term users of hormonal contraceptive were of a magnitude that was within 1 SD of the never users of hormonal contraceptives.
This large multinational study confirms that there are differences in BMD between women living in Africa (Harare), Asia (Bangkok, Dhaka, Shanghai), and Latin America (Campinas, Mexico City). These differences were independent of BMI, socioeconomic status (measured using the occupation of the partner), and reproductive variables. Bone mineral density was highest in Harare, a setting with a high proportion of African women, and lowest in Dhaka, Shanghai, and Bangkok, which are settings with almost exclusively Asian women. The data should be interpreted with the knowledge that there are important differences in early childhood nutrition, exercise, and diet that were not taken into account by the study. We did not attempt to assess dietary intake of calcium because of the unavailability of a reliable and validated instrument to measure calcium intake from foods consumed in the large variety of settings in the study.
The study confirms the well-known association of BMD with BMI,1–3 which pertained even in this relatively thin (mean body mass index 24.8 kg/m2) study population. The percentages of women who smoked or drank alcohol were very low (14% for smoking and 5% for alcohol); heavy smoking and heavy alcohol use were infrequent. Low levels of exposure to smoking and alcohol could account for the lack of association of these factors with BMD in this study.
Several studies suggested that decreases in BMD during lactation are reversible after lactation ceases.10–14 When adjusted only for center, BMD was lower in women who lactated recently, but this association was no longer evident after adjustment for duration of lactation and other variables in the adjustment set. When adjusted only for center, women with a long (more than 5 years) duration of lactation had lower BMD than women with shorter duration of lactation. After adjusting for recency of lactation and other variables, BMD remained lower in women with a long duration of lactation. We interpret these findings as evidence supporting the notion that decreases in BMD during lactation are reversible up to total duration of lactation of 5 years. The findings suggest the possibility that total duration of lactation more than 5 years might result in persistent changes in BMD. Prior studies of this topic are contradictory.14
The interpretation of the data on BMD in relation to use of hormonal contraceptives is not straightforward because our study was cross sectional. We found evidence that the effect of hormonal contraceptives could be related to duration of use and time since last use, and we based our main conclusions about the effect of hormonal contraception on BMD on exclusive users of each hormonal contraceptive method after stratifying by duration of hormonal contraceptive use and time since last use. These analyses suggest that DMPA and levonorgestrel implants decrease BMD during early periods of current use but that the effect of DMPA and levonorgestrel implant use on BMD is reversible. The data suggest that OCs increase BMD during early periods of current use and also that BMD returns to the never users level during longer periods of continued use. Our conclusions apply to use of hormonal contraceptives in the middle of the reproductive age span and must be interpreted with the knowledge that median durations of use were 3–5 years. The suggestion that the effects of hormonal contraception on BMD are reversible could be substantiated only in longitudinal studies of relatively long duration.
Prior research on BMD and OC use is contradictory and confusing.15–26 Mehta15 identified seven studies that showed an increase in BMD in users of OCs and six studies that showed no effect of OCs on BMD. Differences in the size of the studies, the enrollment of postmenopausal women,20–22 the durations of OC use, and proportions of current OC users might explain the discrepancies between prior studies. Previous cross-sectional studies have not examined BMD in relation to recency and duration of OC use in a way that allows direct comparison with our results.
Prior research on BMD and DMPA use is limited and inconsistent. In a cross-sectional study by Cundy et al5 of 30 current users of DMPA aged 25–51 years and two comparison groups, one with premenopausal women and one perimenopausal women, DMPA users had BMD that was intermediate between the premenopausal and perimenopausal groups. In a study of 75 long-term (more than 3 years) DMPA users of reproductive age and an age-matched comparison group of nonusers of DMPA, Virutamasen et al27 found no statistically significant difference in radiographic measures of bone density between DMPA users and the comparison group. Cromer et al6 reported that among adolescents who used DMPA, vertebral bone density decreased from baseline density by 1.5% and 3.1% during 1 and 2 years of use, respectively, whereas for nonusers of hormonal contraceptives it increased by 2.9% and 9.5%, respectively. Gbolade et al7 found that DMPA users with more than 1 year of amenorrhea had overall bone density of the lumbar spine that was significantly lower (−0.33 g/cm2) than the population mean for women aged 20–59 years, but the decrease was not significant and was smaller in magnitude in the femoral neck.
Information on BMD in users of levonorgestrel implants is very limited. Naessen et al28 found an increase in forearm bone density in 11 premenopausal women who were followed up longitudinally in a randomized trial comparing DMPA with levonorgestrel implants. Cromer et al6 observed no adverse effect of levonorgestrel implants on vertebral bone density in their follow-up study of adolescents. An effect of levonorgestrel implants in decreasing BMD is not expected because the 19-nor-progestogen released by the levonorgestrel implant is widely considered to have a beneficial effect on bone density.
Levonorgestrel implant use can suppress ovarian function and production of endogenous estrogen. This suppression might be greater in women with small body mass.29 We examined the possibility that the effect of levonorgestrel implants on BMD during early short-term use might be limited to women who were thin and who might have greater suppression of endogenous estrogen when using levonorgestrel implants. The analysis was inconclusive because there were too few heavy women who were current short-term users of levonorgestrel implants. Future studies should determine whether body mass might modify an effect of levonorgestrel implants on BMD. Information on menstrual patterns during implant use was not collected. Analysis of BMD in relation to such patterns might also be informative.
Bone density is the result of a dynamic process in which bone is constantly being formed and reabsorbed. The hormonal mechanisms that govern bone formation and reabsorption are complex. It is plausible that there are homeostatic mechanisms that maintain BMD despite hormonal stimulation.
The cross-sectional nature of the study is a limitation. The age range of subjects was narrow and the use of hormonal contraceptives might have a different effect on BMD at older or younger ages. Bone mineral density was measured at the distal radius and midshaft of ulna using single x-ray absorptiometry. Findings cannot be generalized to other clinically important skeletal sites, such as the femoral neck and the spine, where fracture risk is of most concern.
A 1992 WHO study group established that BMD within one SD of the young adult reference mean is normal; a low bone mass (osteopenia) was defined as more than one SD below the reference mean but less than 2.5 SD below this value; and osteoporosis was BMD 2.5 SD or more below the young adult mean.30 All of the reduced values of BMD associated with hormonal contraceptive use we observed in this study were within one SD of the mean of never users. Thus, whatever transient small changes in BMD might occur early after initiation of use of some hormonal contraceptives are small and might not be clinically important.
1. Cummings SR, Kelsey JL, Nevitt MC, O'Dowd KJ. Epidemiology of osteoporosis and osteoporotic fractures. Epidemiol Rev 1985;7:178–208.
2. Sowers MR, Galuska DA. Epidemiology of bone mass in premenopausal women. Epidemiol Rev 1993;15:374–98.
3. Stevenson JC, Lees B, Devenport M, Cust MP, Ganger KF. Determinants of bone density in normal women: Risk factors for future osteoporosis. BMJ 1989;298:924–8.
4. DeCherney A. Bone-sparing properties of oral contraceptives. Am J Obstet Gynecol 1996;174:15–20.
5. Cundy T, Evans M, Roberts H, Wattie D, Ames R, Reid IR. Bone density in women receiving depot medroxyprogesterone acetate for contraception. BMJ 1991;303:13–6.
6. Cromer BA, Blair JM, Mahan JD, Zibner SL, Naumovski Z. A prospective comparison of bone density in adolescent girls receiving depot medroxyprogesterone acetate (Depo-Provera), levonorgestrel (Levonorgestrel implant), or oral contraceptives. J Pediatr 1996;129:671–6.
7. Gbolade B, Ellis S, Murby B, Randall S, Kirkman R. Bone density in long term users of depot medroxyprogesterone acetate. Br J Obstet Gynaecol 1998;105:790–4.
8. Fraser IS, Tiitinen A, Affandi B, Brache V, Croxatto HB, Diaz S, et al. Levonorgestrel implant consensus statement and background review. Contraception 1998;57:1–9.
9. Fleiss J. The design and analysis of clinical experiments. New York: John Wiley & Sons, Inc., 1986.
10. Caird LE, Reid-Thomas V, Hannan WJ, Gow S, Glasier AF. Oral progestogen-only contraception may protect against loss of bone mass in breast-feeding women. Clin Endocrinol 1994;41:739–45.
11. Drinkwater BL, Chesnut CH. Bone density changes during pregnancy and lactation in active women: A longitudinal study. Bone Miner 1991;14:153–60.
12. Kalkwarf HJ, Specker BL. Bone mineral loss during lactation and recovery after weaning. Obstet Gynecol 1995;86:26–32.
13. Laskey MA, Prentice A. Effect of pregnancy on recovery of lactational bone loss. Lancet 1997;349:1518–9.
14. Sowers MF, Corton G, Shapiro B, Jannausch ML, Crutchfield M, Smith ML, et al. Changes in bone density with lactation. JAMA 1993;269:3130–5.
15. Mehta S. Bone loss, contraception and lactation. Acta Obstet Gynecol Scand 1993;72:148–56.
16. Enzelsberger H, Metka M, Heytmanek G, Schurz B, Kurz C, Kusztrich M. Influence of oral contraceptive use on bone density in climacteric women. Maturitas 1988;9:375–8.
17. Fortney JA, Feldblum PJ, Talmage RV, Zhang J, Godwin SE. Bone mineral density and history of oral contraceptive use. J Reprod Med 1994;39:105–9.
18. Gambacciani M, Spinetti A, Taponeco F, Cappagli B, Piaggesi L, Fioretti P. Longitudinal evaluation of perimenopausal vertebral bone loss: Effects of a low-dose oral contraceptive on BMD and metabolism. Obstet Gynecol 1994;83:392–6.
19. Garnero P, Sornay-Rendu E, Delmas PD. Decreased bone turnover in oral contraceptive users. Bone 1995;16:499–503.
20. Goldsmith NF, Johnston JO. Bone mineral: Effects of oral contraceptives, pregnancy, and lactation. J Bone Joint Surg 1975;57:657–68.
21. Hreshchyshyn MM, Hopkins A, Zylstra S, Anbar M. Associations of parity, breast-feeding, and birth control pills with lumbar spine and femoral neck bone densities. Am J Obstet Gynecol 1988;159:318–22.
22. Kleerekoper M, Brienza RS, Schultz LR, Johnson CC, and the Henry Ford Hospital Osteoporosis Cooperative Research Group. Oral contraceptive use may protect against low bone mass. Arch Intern Med 1991;151:1971–6.
23. Lindsay R, Tohme J, Kanders B. The effect of oral contraceptive use on vertebral bone mass in pre- and post-menopausal women. Contraception 1986;34:333–40.
24. Lloyd T, Buchanan JR, Ursino GR, Myers C, Woodward G, Halbert DR. Long-term oral contraceptive use does not affect trabecular bone density. Am J Obstet Gynecol 1989;160:402–4.
25. Mazess RB, Barden HS. Bone density in premenopausal women: Effects of age, dietary intake, physical activity, smoking, and birth-control pills. Am J Clin Nutr 1991;53:132–42.
26. Recker RR, Davies M, Hinders SM, Heaney RP, Stegman MR, Kimmel DB. Bone gain in young adult women. JAMA 1992;268:2403–8.
27. Virutamasen P, Wangsuphachart S, Reinprayoon D, Kriengsinyot R, Leepipatpaiboon S, Gua C. Trabecular bone in long-term depot-medroxyprogesterone acetate users. Asia-Oceania J Obstet Gynaecol 1994;20:269–74.
28. Naessen T, Olsson S, Gudmundson J. Differential effects on bone density of progestogen-only methods for contraception in pre-menopausal women. Contraception 1995;52:35–9.
29. Brache V, Alvarez-Sanchez F, Faundes A, Tejada AS, Cochon L. Ovarian endocrine function through five years of continuous treatment with levonorgestrel implant subdermal implants. Contraception 1990;41:169–77.
30. World Health Organization. Assessment of fracture risk and its application to screening for postmenopausal osteoporosis. WHO Technical Report Series 843. World Health Organization, Geneva, 1994.