Depot medroxyprogesterone acetate (DMPA; 150 mg intramuscular or 104 mg subcutaneous injection), a long-acting reversible contraceptive used globally by women who require high contraceptive efficacy without increased thromboembolic risk,1,2 decreases bone mineral density (BMD)3,4 in adults in a reversible manner,5,6 with mean BMD declines of 5.4% (lumbar spine) and 5.2% (hip) after 5 years.7 In adolescents, mean decreases of 2.7% (spine), 4.1% (hip), and 3.9% (femoral neck) were reported, with spine BMD returning to pre-exposure levels 60 weeks after discontinuation.8 No evidence of reduced BMD was seen in postmenopausal women who had used DMPA during their reproductive years.9
In postmenopausal women, low BMD is associated with increased risk for fragility fractures (fractures occurring with little or no trauma attributable to severely compromised bone microarchitecture).10 Despite no established relationship between low BMD and fragility fractures in premenopausal women, concern that fracture risk could increase has led to limitations on DMPA use.11 Literature searches were completed using PubMed and Google Scholar for the years 1960–2012 using the search terms “Depo-Provera,” “Injectable medroxyprogesterone acetate,” “injectable MPA,” “injectable contraceptive,” “MP IM (IM MPA),” “DMPA or Depot MPA” and “fracture (fractures),” “fracture risk,” “osteoporosis,” “osteoporotic fracture,” “fragility fracture,” or “bone fracture.” No study estimating fracture risk before and after DMPA use has determined whether DMPA increases fracture risk. The present retrospective cohort study was conducted to address two questions: what is the association between DMPA use and fracture incidence? and does the association with fracture vary by level of DMPA exposure?
MATERIALS AND METHODS
Using data in the General Practice Research Database, this nonrandomized study compared fracture incidence in two cohorts of women, DMPA users and women using non-DMPA prescription contraceptives. The study used anonymized data and was deemed by the Research Triangle Institute Committee for the Protection of Human Subjects to be exempt, under United States law, from review by an Institutional Review Board.
The General Practice Research Database contains de-identified longitudinal patient records from more than 350 general practices in the United Kingdom (www.gprd.com) and has been validated for drug safety studies.12–14 The General Practice Research Database staff supplied electronic data for all women who met the following study criteria: acceptable patient status by General Practice Research Database data quality criteria; known year of birth; and at least one prescription contraceptive record, including DMPA, oral contraceptives, intrauterine device (IUD), cervical cap, or diaphragm, before age 50 years and before any record of bilateral oophorectomy dated between 1 January 1987 and 31 December 2005. For each eligible patient, we received all available General Practice Research Database records before and after the first recorded contraceptive prescription. These women constituted the full study cohort. For investigation of confounding, we restricted analysis to the subcohort of women who had at least 6 months of baseline data before contraceptive use.
For each patient, the index date was defined as the date of her first contraceptive prescription on or after the latest of 1 January 1987, the woman's practice registration date, or the date the practice was certified as having met General Practice Research Database data quality standards. Follow-up for each woman began on the index date and continued until the first of the practice's last contribution to General Practice Research Database, 31 December 2005, the woman's first recorded fracture, or her termination from the practice. Women who switched to DMPA after the index date first contributed observation time to the nonuser group (index date to first use of DMPA) and then to the DMPA user group; women who began DMPA at the index date contributed time only to the DMPA user group, regardless of other contraceptive usage.
Baseline medical history was used to assess the potential for possible confounding based on changes in incidence rate ratio estimates for DMPA exposure and fracture, with and without control of each potential confounder. We used standardization for the purpose of controlling confounding, based on distribution weights of person-years in the subcohort with at least 6 months of baseline history.
Each DMPA prescription generated 90 days of exposure, based on prescribing information.11 Periods of active exposure were defined by concatenating contiguous or overlapping prescription periods. Isolated DMPA episodes occurred when 90-day exposure periods did not overlap or flank the exposure period for another DMPA injection. Duplicate prescriptions (same date) were deleted and records occurring fewer than 63 days (fifth percentile for injection interval) after the last DMPA injection were disregarded.
A woman's period of active DMPA exposure extended from her first injection within an episode to the end of the 90-day exposure period of the last injection in the episode. Cumulative DMPA exposure for each woman was categorized as low (one to seven injections) or high (at least eight injections). We used three categories of DMPA follow-up time: “current exposure” was day 1 (DMPA administered) to day 90 (extended by 90 days for each new injection); “recent exposure” extended until 730 days after the last injection, based on an assumed BMD recovery time of 2 years (range 1 – 5 years, depending on bone site)5,6; and “past exposure” was all time 731 or more days after the last injection, unless there was a new prescription for DMPA.
We also examined fracture incidence over time in DMPA users by calculating time-specific incidence rates since start of use. A Savitzky-Golay cubic polynomial smoothing filter with a 25-point window15 was used to fit a trend line. To examine fracture rates according to anatomic site, fracture events were coded using the Read and Oxford Medical Information System dictionary. Multiple fractures on the same date were counted as one event for incidence of “any fracture.” SAS 9 was used for basic analyses and Episheet (http://krothman.hostbyet2.com/Episheet.xls) was applied for standardizing rates and calculating confidence interval (CI) estimates of the standardized rates. For incidence rate ratios of zero, we used Stata software 9 to estimate the upper bound of a 95% exact CI.
We analyzed fracture incidence before and after the start of contraceptive use for women who used DMPA at some point and performed a similar analysis for nonusers of DMPA with respect to use of their qualifying non-DMPA contraceptive. This approach controlled for unmeasured variables that may differ between women but are generally constant over time within a woman, such as behavioral or lifestyle variables that are unavailable in the General Practice Research Database and therefore cannot be directly assessed but that could potentially affect the observed incidence of fractures.
We divided the subcohort of women who had 6 or more months of baseline history into ever-users and nonusers of DMPA, based on the choice of contraceptive after the index date, and calculated time to first fracture during the 6-month baseline period. Observation time within the baseline period was less than 6 months in those women who reported a fracture during the baseline period, and exactly 6 months otherwise. The number of person-years after starting contraceptive use was calculated beginning with the index date for nonusers or, for DMPA users, beginning with the date of first DMPA use and ending with study termination. If a DMPA user did not start DMPA on the study index date, the person-years of nonuse that occurred between the study index date and the start of DMPA use were not included in any before calculations or after calculations. Fractures during the baseline period were counted as “any” or “none.” Individual fractures were identified by the corresponding Read and Oxford Medical Information System code and were grouped into categories (axial, appendicular, or miscellaneous fractures) that may correlate with fracture etiology.10
The General Practice Research Database provided data for 330,684 women aged 15–50 years at the index date; 18,289 (5.5%) did not meet the eligibility criteria. The full study cohort included 312,395 women, and the subcohort with at least 6 months of baseline history included 166,637 (53%) women. In the full cohort, average follow-up was 5.9 years for DMPA users and 5.4 years for nonusers. The most common reason that follow-up terminated was reaching the 31 December 2005 end-of-study date (58.2%); other reasons were that the woman stopped contributing data to the practice, including by transfer or death (34.7%), an incident fracture (3.8%), or the practice stopped contributing data to General Practice Research Database (3.4%).
In the full cohort, 11,822 fractures occurred during 1,722,356 person-years of follow-up. Crude fracture incidence at any skeletal site during follow-up was 6.4 per 1,000 person-years for nonusers (8,887 fractures in 1,395,041 person-years) and 9.0 per 1,000 person-years for DMPA users (2,935 fractures in 327,315 person-years); the crude incidence rate ratio after contraception started was 1.41 (95% CI 1.35–1.47). Table 1 shows age-specific fracture rates in the full cohort.
In addition to the crude overall rates, fracture rate also was analyzed after being compared with results before DMPA or other contraception was initiated in the subgroup of participants with fracture data available before the index date; according to DMPA exposure (less than eight injections compared with eight or more injections) in the full cohort of DMPA users, and according to the anatomic location of the fracture in the full cohort of all participants. These analyses were performed after the magnitude of confounding had been assessed. Therefore, we present both the crude results and results standardized for measured confounders, as described.
To examine the potential for confounding of the results by pre-existing characteristics of the women in the study, the relevant demographic and medical history variables for which records were available in the General Practice Research Database were individually assessed as potential confounders.
In the subcohort of women who had at least 6 months of baseline history available in the General Practice Research Database (n=166,367), fracture incidence after contraception started was similar to that in the full cohort. The crude fracture incidence rate was 6.6 per 1,000 person-years for nonuse of DMPA (4,939 cases in 744,242 person-years) and 9.1 per 1,000 person-years during DMPA use (1,574 cases in 173,713 person-years). The crude incidence rate ratio for any fracture in the subcohort was 1.37 (95% CI 1.29–1.45) and the crude incidence rate difference was 2.42 per 1,000 person-years (95% CI 1.94–2.91 per 1,000 person-years). In this subcohort, the age-standardized fracture rate per 1,000 person-years was 6.6 for nonusers and 9.3 for DMPA users, giving a standardized incidence rate ratio of 1.40 (95% CI 1.32–1.49) and a standardized incidence rate difference of 2.6 per 1,000 person-years (95% CI 2.13–3.17 per 1,000 person-years). These results are close to the crude results, indicating little confounding by age within the age range of the study population.
Table 2 shows the crude associations for each baseline variable with DMPA use and fracture risk, along with the incidence rate ratio standardized for age and potential confounders. All results were very close to the incidence rate ratio obtained when controlling only for age, indicating that none of the other potential confounders for which data were available in the General Practice Research Database contributed any material confounding to the overall incidence rate ratio for DMPA and fractures.
For the analyses of fracture incidence according to DMPA exposure and fracture site, the lack of confounding by measured risk factors for fracture prompted us to expand the study population to the full cohort of 312,395 women, including the 146,028 who had less than 6 months of medical history available before their index date, on the assumption that confounding as defined similarly would not be material for these additional women.
Table 3 shows that before contraception was started, the crude incidence rate ratio for fractures for women who later became DMPA users was 1.28 (95% CI 1.07–1.53) compared with women who never used DMPA. Further, neither of these two cohorts exhibited a meaningful increase in fracture rate after starting contraception. The incidence rate ratio for fracture in DMPA users (after and before) was 1.08 (95% CI 0.92–1.26). In these analyses, 72,190 person-years of “nonuse” by DMPA users who started DMPA after the index date were excluded.
Fracture history during the 6-month baseline period immediately preceding first contraceptive use was not identified as an important confounder in the age-standardized analyses because the follow-up data from the 585 women with a baseline fracture history did not materially influence the overall results. That is, the overall incidence rate ratio for fracture was similar with and without these relatively few women included in the analysis (Table 2). Nonetheless, both before and after contraception started, DMPA users had a greater fracture incidence than nonusers (Table 3).
Table 4 shows the effects of DMPA exposure by recency of use and cumulative number of injections. For current DMPA users compared with nonusers, crude excess fracture risk was greater in low-exposure users (incidence rate ratio 1.91, 95% CI 1.76–2.06) than in high-exposure users (incidence rate ratio 1.09, 95% CI 0.98–1.21) whose fracture risk was not different from that of nonusers. Age-standardized rates gave similar results. For recent and past DMPA use, fracture risk was similar between low-exposure and high-exposure users.
Figure 1 shows fracture incidence at each time point (1/10th year) since starting DMPA for current and recent users of DMPA, with a reference line indicating the nonuser incidence rate. The data points are statistically stable through approximately 1,500 days but are based on few events per time point after 1,500 days. Overall, the graph indicates that current and recent users do have a higher fracture rate than nonusers, but it is noteworthy that there was no important time trend with continuing use. In particular, the difference in risk between DMPA users and nonusers was present at the start of use, showed no discernible induction period during which BMD could have declined before fracture incidence increased in DMPA users, and did not meaningfully change up to day 1,500. The graph excludes periods of use for a woman that occurred after a gap of more than 2 years since her previous period of use.
Compared with nonuse, DMPA users had more codes for appendicular and miscellaneous (fingers, toes, face, skull, multiple trauma, and unspecified) fractures (Table 5); however, there was no excess risk for axial fracture codes (hip, pelvis, and symptomatic or clinical vertebral fractures; incidence rate ratio 0.95, 95% CI 0.74–1.23).
When used consistently and correctly, DMPA has contraceptive efficacy (0.22 failures per 100 person-years) comparable to IUDs and implants (0.27 failures);1 nevertheless, the possibility of an adverse effect on fractures has been a long-standing concern. However, the results of this study suggest that DMPA does not increase risk for fracture.
Hypoestrogenemia lowers BMD and, as a consequence, fracture risk is increased in postmenopausal women whose bone structures have been compromised because of large BMD declines (more than 20%) early in menopause;16,17 however, no similar BMD–fracture relationship has been demonstrated for reproductive-aged women. Therefore, as Grimes and Schulz18 have suggested, BMD may be an inappropriate surrogate endpoint in this population.
Previous database studies of DMPA and fracture applied different types of cohorts and age ranges, making comparison with the current study's cohort of contraceptive users potentially difficult; however, the age-specific and gender-specific fracture rates reported by Van Staa et al16 for women up to age 65 years appear similar to the rates found in the current study for unexposed women. A case-control study of fracture and DMPA use found adjusted odds ratios for DMPA use between 1.18 (95% CI 0.93–1.49; one to two cumulative prescriptions) and 1.54 (95% CI 1.33–1.78; 10 or more prescriptions).19 Although fractures likely to be trauma-related (leg, foot, hand) were associated with DMPA in that study, spine and hip fractures (potentially related to low BMD) were not.19 A second case-control study of fractures and use of DMPA or IUD reported adjusted odds ratios of 1.44 (95% CI 1.01–2.06) for DMPA use and 0.75 (95% CI 0.64–0.87) for IUD use, but the authors acknowledged that unassessed confounders may have affected the results.20 A prospective, nonrandomized study examined risk factors for stress fracture among female army recruits.21 Smoking and excessive alcohol intake were associated with an increased incidence in fractures, but contraceptive use showed no important association, except for DMPA use in a subcohort of non-Hispanic white women (incidence rate ratio 1.71, 95% CI 1.01–2.90).21 No studies, however, have examined fracture incidence before DMPA use.
There were several notable findings in this study. Although DMPA users had higher fracture risk than nonusers, risk did not increase after starting DMPA but was similarly present before DMPA was initiated. This comparison controlled for confounders not measurable in the General Practice Research Database but that probably remained constant over time within a woman. Risks for hip and vertebral fractures are particularly sensitive to BMD changes and routinely serve as primary endpoints in randomized osteoporosis trials, whereas finger, toe, face, and skull fractures typically are not analyzed in these trials22 because their incidence does not correlate with BMD changes because these fractures probably result from trauma.10,17 Therefore, although few if any true fragility fractures would have been expected in this study, because of the young mean age of the women, we found no difference in risk for axial fractures (hip, symptomatic vertebral, or pelvic fractures), along with the largest incidence rate ratio being finger, toe, face, or skull fractures. This is consistent with the overall higher fracture incidence in DMPA users being unrelated to a decrease in BMD, but possibly it is attributable to differential risks for traumatic injury. Neither the time trend for fractures after DMPA started nor the trend with cumulative DMPA exposure fits a pattern that is consistent with a causal effect of DMPA. Fracture rates were greater in DMPA users immediately on beginning contraceptive use as well as before beginning. Furthermore, longer-term DMPA users had a lower fracture risk than shorter-term users, which would be unexpected if the higher fracture risk among DMPA users was related to an adverse effect of DMPA on bone.
A limitation of our study is the possibility of uncontrolled confounding by factors for which we had no data (eg, education level, participation in sports or other risk-taking behaviors of a physical nature, and attitudes toward contraception) or for which we had incomplete data (eg, body mass index and smoking history). Confounders also may have been misclassified, through inaccuracy in recording medical history or through missing data, and thereby reduced the ability to control for differences in such factors between the exposure groups.
Some studies have reported important baseline differences between women self-selecting DMPA and women choosing other prescription contraceptive methods, including lower socioeconomic status and educational attainment,1 as well as differences in risk factors for falls, fractures, and osteoporosis.20 In one prospective study of the effects of DMPA on BMD, important differences precluded a planned postbaseline comparison between the DMPA user and nonuser cohorts.23 Although the potential confounders measurable in the current study did not have confounding effects, the results suggest that unassessed factors, possibly related to behavior, lifestyle, or unreported medical history, differed between cohorts and could be the underlying cause of the higher fracture risk in the DMPA cohort. Reporting bias also may have affected the results. Visits every 3 months by current DMPA users may have enabled more accurate fracture reporting, whereas nonusers or past users generally visit less often.
In conclusion, the overall findings of this study demonstrated that DMPA use was associated with greater fracture risk compared with nonuse, both before and after DMPA use began. Although reversible decreases in BMD are well-documented with DMPA, the current findings do not suggest that these BMD decreases had any clinically important effect on fracture risk during study follow-up (mean was approximately 6 years). The association between DMPA use and higher fracture risk does not appear to be the result of a causal effect of DMPA but may represent inherent differences in fracture risk between women who elected to use DMPA and women who used other prescription contraceptives.
1. Winner B, Peipert JF, Zhao Q, Buckel C, Madden T, Allsworth JE, et al.. Effectiveness of long-acting reversible contraception. N Engl J Med 2012;366:1998–2007.
2. Centers for Disease Control and Prevention. US medical eligibility criteria for contraceptive use, 2010: adapted from the World Health Organization Medical Eligibility Criteria for Contraceptive Use. 4th Edition. Morb Mortal Wkly Rep MMWR 2010;59(RR-4):1–86.
3. Cromer BA, McArdle BJ, Mahan JD, Zibners L, Naumovski Z. A prospective comparison of bone density in adolescent girls receiving depot medroxyprogesterone acetate (Depo-Provera), levonorgestrel (Norplant), or oral contraceptives. J Pediatr 1996;129:671–6.
4. Westhoff C. Bone mineral density and DMPA. J Reprod Med 2002;47(Suppl 9):795–9.
5. Kaunitz AM, Arias R, McClung M. Bone density recovery after depot medroxyprogesterone acetate injectable contraception use. Contraception 2008;77:67–76.
6. Scholes D, LaCroix AZ, Ichikawa LE, Barlow WE, Ott SM. Injectable hormone contraception and bone density: results from a prospective study. Epidemiology 2002;13:581–7.
7. Kaunitz AM, Miller PD, Rice VM, Ross D, McClung MR. Bone mineral density in women aged 25-35 years receiving depot medroxyprogesterone acetate: recovery following discontinuation. Contraception 2006;74:90–9.
8. Harel Z, Johnson CC, Gold MA, Cromer B, Peterson E, Burkman R, et al.. Recovery of bone mineral density in adolescents following the use of depot medroxyprogesterone acetate contraceptive injections. Contraception 2010;81:281–91.
9. Walker-Orr BJ, Evans MC, Ames RW, Clearwater JM, Cundy T, Reid R. The effect of past use of the injectable contraceptive depot medroxyprogesterone acetate on bone mineral density in normal post-menopausal women. Clin Endocrinol (Oxf) 1998;49:615–8.
10. Stone KL, Seeley DG, Lui LY, Cauley JA, Ensrud K, Browner WS, et al.. BMD at multiple sites and risk of fracture of multiple types: long-term results from the Study of Osteoporotic Fractures. J Bone Miner Res 2003;18:1947–54.
11. British National Formulary No.62. London (UK): BMJ Group and RPS Publishing; 2011.
12. Jick SS, Kaye JA, Vasilakis-Scaramozza C, Garcia Rodríguez LA, Ruigómez A, Meier CR, et al.. Validity of the general practice research database. Pharmacotherapy 2003;23:686–9.
13. Van Staa TPA. The use of a large pharmacoepidemiological database to study exposure to oral corticosteroids and risk of fractures: validation of study population and results. Pharmacoepidemiol Drug Saf 2000:9:359–66.
14. Khan NF, Harrison SE, Rose PW. Validity of diagnostic coding within the General Practice Research Database: a systematic review. Br J Gen Pract 2010;60:e128–36.
15. Savitsky A, Golay MJE. Smoothing and differentiation of data by simplified least squares procedures. Anal Chem 1964;36:1627–39.
16. Van Staa TP, Dennison EM, Leufkens HGM, Cooper C. Epidemiology of fractures in England and Wales. Bone 2001;29:517–22.
17. Kanis JA. Diagnosis of osteoporosis and assessment of fracture risk. Lancet 2002;359:1929–36.
18. Grimes DA, Schulz KF. Surrogate end points in clinical research: hazardous to your health. Obstet Gynecol 2005;105:1114–8.
19. Meier CR, Brauchli YB, Jick SS, Kraenzlin ME, Meier CR. Use of DMPA and fracture risk. J Clin Endocrin Metab 2010;95:4909–16.
20. Vestergaard P, Rejnmark L, Mosekilde L. The effects of DMPA and intrauterine device use on fracture risk in Danish women. Contraception 2008;78:459–64.
21. Lappe JM, Stegman MR, Recker RR. The impact of lifestyle factors on stress fractures in female army recruits. Osteoporos Int 2001;12:35–42.
22. Cummings SR, Ensrud K, Delmas PD, LaCroix AZ, Vukicevic S, Reid DM, et al.. Lasofoxifene in postmenopausal women with osteoporosis. N Engl J Med 2010;362:686–96.
23. Johnson CC, Burkman RT, Gold MA, Brown RT, Harel Z, Bruner A, et al.. Longitudinal study of depot medroxyprogesterone acetate (Depo-Provera) effects on bone health: study design, population characteristics and baseline bone mineral density. Contraception 2008;77:239–48.