Steroidal contraceptives: effect on bone fractures in women

Laureen M Lopez; David A Grimes; Kenneth F Schulz; Kathryn M Curtis; Mario Chen

doi:10.1002/14651858.CD006033.pub5

. 2014 Jun 24;2014(6):CD006033. doi: 10.1002/14651858.CD006033.pub5

Steroidal contraceptives: effect on bone fractures in women

Laureen M Lopez ^1,^✉, David A Grimes ², Kenneth F Schulz ³, Kathryn M Curtis ⁴, Mario Chen ⁵

Editor: Cochrane Fertility Regulation Group

PMCID: PMC11127753 PMID: 24960023

Abstract

Background

Steroidal contraceptive use has been associated with changes in bone mineral density in women. Whether such changes increase the risk of fractures later in life is not clear. Osteoporosis is a major public health concern. Age‐related decline in bone mass increases the risk of fracture, especially of the spine, hip, and wrist. Concern about bone health influences the recommendation and use of these effective contraceptives globally.

Objectives

Our aim was to evaluate the effect of using hormonal contraceptives before menopause on the risk of fracture in women.

Search methods

Through April 2014, we searched for studies of fracture or bone health and hormonal contraceptives in MEDLINE, POPLINE, CENTRAL, EMBASE, and LILACS, as well as ClinicalTrials.gov and ICTRP. We examined reference lists of relevant articles for other trials. For the initial review, we wrote to investigators to find additional trials.

Selection criteria

Randomized controlled trials (RCTs) were considered if they examined fractures, bone mineral density (BMD), or bone turnover markers in women with hormonal contraceptive use prior to menopause. Eligible interventions included comparisons of a hormonal contraceptive with a placebo or with another hormonal contraceptive that differed in terms of drug, dosage, or regimen. They also included providing a supplement to one group.

Data collection and analysis

We assessed all titles and abstracts identified through the literature searches. Mean differences were computed using the inverse variance approach. For dichotomous outcomes, the Mantel‐Haenszel odds ratio (OR) was calculated. Both included the 95% confidence interval (CI) and used a fixed‐effect model. Due to differing interventions, no trials could be combined for meta‐analysis. We applied principles from GRADE to assess the evidence quality and address confidence in the effect estimates. In addition, a sensitivity analysis included trials that provided sufficient data for this review and evidence of at least moderate quality.

Main results

We found 19 RCTs that met our eligibility criteria. Eleven trials compared different combined oral contraceptives (COCs) or regimens of COCs; five examined an injectable versus another injectable, implant, or IUD; two studied implants, and one compared the transdermal patch versus the vaginal ring. No trial had fracture as an outcome. BMD was measured in 17 studies and 12 trials assessed biochemical markers of bone turnover. Depot medroxyprogesterone acetate (DMPA) was associated with decreased bone mineral density (BMD). The placebo‐controlled trials showed BMD increases for DMPA plus estrogen supplement and decreases for DMPA plus placebo supplement. COCs did not appear to negatively affect BMD, and some formulations had more positive effects than others. However, no COC trial was placebo‐controlled. Where studies showed differences between groups in bone turnover markers, the results were generally consistent with those for BMD. For implants, the single‐rod etonogestrel group showed a greater BMD decrease versus the two‐rod levonorgestrel group but results were not consistent across all implant comparisons.

The sensitivity analysis included 11 trials providing evidence of moderate or high quality. Four trials involving DMPA showed some positive effects of an estrogen supplement on BMD, a negative effect of DMPA‐subcutaneous on lumbar spine BMD, and a negative effect of DMPA on a bone formation marker. Of the three COC trials, one had a BMD decrease for the group with gestodene plus EE 15 μg. Another indicated less bone resorption in the group with gestodene plus EE 30 μg versus EE 20 μg.

Authors' conclusions

Whether steroidal contraceptives influence fracture risk cannot be determined from existing information. The evidence quality was considered moderate overall, largely due to the trials of DMPA, implants, and the patch versus ring. The COC evidence varied in quality but was low overall. Many trials had small numbers of participants and some had large losses. Health care providers and women should consider the costs and benefits of these effective contraceptives. For example, injectable contraceptives and implants provide effective, long‐term birth control yet do not involve a daily regimen. Progestin‐only contraceptives are considered appropriate for women who should avoid estrogen due to medical conditions.

Keywords: Female; Humans; Bone Density; Bone Density/drug effects; Bone Remodeling; Bone Remodeling/drug effects; Contraceptives, Oral, Hormonal; Contraceptives, Oral, Hormonal/adverse effects; Contraceptives, Oral, Hormonal/pharmacology; Estrogens; Estrogens/pharmacology; Fractures, Bone; Fractures, Bone/chemically induced; Medroxyprogesterone Acetate; Medroxyprogesterone Acetate/adverse effects; Medroxyprogesterone Acetate/pharmacology; Premenopause; Progestins; Progestins/pharmacology; Randomized Controlled Trials as Topic

Plain language summary

Hormonal contraceptives and bone health in women

Hormonal contraceptives have been related to bone changes in women. Whether such changes lead to more bone fractures later in life is not clear. However, bone health is a major public health concern. Bone density declines with age, and the change increases the risk of fracture. Due to concern about bone health, health care providers may not suggest hormonal contraceptives and women may not want to use them.

Through April 2014, we did computer searches for studies of birth control methods containing hormones and risk of fractures. Outcomes could also be bone mineral density or markers of bone changes. Birth control pills included types with both estrogen and progestin. Also included were implants and injectables with only progestin. We wrote to researchers to find other trials. We included randomized trials in any language that had at least three treatment cycles. The studies had to compare two types of birth control or one type of birth control or a supplement with a placebo or 'dummy' method.

We found 19 trials. Fifteen studies compared one birth control method with another hormone method. Two trials used a placebo or 'dummy.' One compared a hormone method to a method without hormones. None had fractures as an outcome and most looked at bone density. Birth control methods with both estrogen and progestin did not appear to affect bone health. However, 'depo,' which is injected and has only progestin, was related to lower bone density. The two depo trials with placebos showed increased bone density when some estrogen was given to women on depo. Bone density decreased in women who got a 'dummy' with the depo. Whether this decrease is important to the woman's health is not known. For implants, an etonogestrel implant with one rod showed a greater decrease in bone density than a two‐rod levonorgestrel implant. However, other implants studied did not show the same pattern.

The studies had data of moderate quality. Whether hormonal contraceptives affect fracture risk cannot be judged from current information. These contraceptive methods work well for birth control. Health‐care providers and women should think about the costs and benefits. For instance, injectable use can occur without a partner's knowledge, and is simpler than taking pills every day. Also, progestin‐only methods are suggested for some women with health problems who should avoid estrogen.

Background

Description of the condition

Steroidal contraceptives, particularly injectable contraceptives and combined oral contraceptives (COCs), have been associated with changes in bone mineral density in women. Whether such changes increase the risk of fractures later in life is not clear. However, osteoporosis is a major public health concern. Age‐related decline in bone mass increases the risk of fracture, especially of the spine, hip, and wrist (Howe 2011; Rachner 2011). The costs of osteoporosis‐related fractures can be substantial for the individual due to disability and to society for health and social care (Howe 2011). Concern about bone health influences the recommendation and use of these effective contraceptives globally.

Skeletal fragility results from suboptimal formation of bone mass and strength, as well as excess bone resorption (NIH 2000; Raisz 2005). Bone loss during contraceptive use may be temporary like that which occurs during pregnancy or breastfeeding (Gourlay 2004; ACOG 2008). Risk of future fractures after contraceptive use depends on whether the bone mass is restored.

Description of the intervention

Depot medroxyprogesterone acetate (DMPA)

DMPA is an effective contraceptive and the most widely‐used injectable (Bartz 2011). Data from developing countries showed median failure rates of 2.4% for injectables versus 10.3% for condoms and 6.5% for pills (Cleland 2004). First‐year failure rates for DMPA in the USA have been estimated at 0.2% for perfect use and 6% for typical use (Trussell 2011). If injectable use were limited due to concerns about effects on bone health, women might switch to less effective methods or use nothing, which could lead to increased pregnancy rates.

Of the injectable contraceptives, DMPA has attracted the most attention regarding bone health. DMPA may reduce bone mineral density (BMD), which is a potential concern for younger women who have not yet achieved peak bone mass. Early research indicated more bone loss among women who used DMPA before 20 years of age and those who used it for longer periods (Cundy 1998; Scholes 1999). More recently, two case‐control studies reported increased fracture risk for longer current use of DMPA (Vestergaard 2006; Meier 2010), although past users had little evidence of increased risk (Meier 2010).

In the US, the Food and Drug Administration requires a warning on DMPA labeling (FDA 2004; FDA 2011). It refers to BMD loss among DMPA users, especially younger women. The warning is based on limited evidence and may limit long‐term use (Kaunitz 2011). Major health organizations have recommended not restricting DMPA use among women 18 to 45 years old (WHO 2006; ACOG 2008; Guilbert 2009). In guidance about medical eligibility criteria for contraceptive use, DMPA is category 1 (no restriction) for women aged 18 to 45 years (CDC 2010; WHO 2009). For women outside that age range, DMPA is category 2, meaning the advantages generally outweigh the theoretical or proven risks.

Oral contraceptives (OCs)

OCs are the most commonly used reversible method in more developed countries (UN 2011). Failure rates for oral contraceptives in the USA (combined and progestin‐only) are estimated at 0.3% for perfect use and 9% for typical use in the first year (Trussell 2011).

Few associations have been noted between OC use and fracture risk in observational studies (Lopez 2012). A cohort study found OC ever‐users had increased risk for all fractures (Cooper 1993). However, a case‐control study, with later data from a subset, reported no association except for those with 10 years or more since use (Memon 2011). Another case‐control study reported increased risk, but only for those who had 10 or more prescriptions (Meier 2010). A cohort study of postmenopausal women found no increased fracture risk for OC use after excluding women with prior fracture (Barad 2005). Two other studies found little evidence of association between OC use and fracture risk. A cohort study noted increased risk for subgroups, such as those with longer use or specific intervals since use (Vessey 1998). A case‐control study reported increased risk for any fracture only among young women with less than average use (Vestergaard 2006).

COCs may have little effect on BMD among healthy adult women. Prospective studies have indicated that ultra‐low dose COCs, containing 20 μg ethinyl estradiol, may affect bone development in young women (Cromer 2003). On the other hand, COCs with 30 to 40 μg ethinyl estradiol may have no negative effect and may even protect against bone loss, at least among women 30 years of age or more (Cromer 2003). Evidence from studies of varying designs indicates that BMD may be affected by COC use in adolescent and young women but not in adult premenopausal or postmenopausal women (Martins 2006; Herrmann 2010; Warholm 2012). However, COC use may have a negative effect on bone turnover markers, although the clinical significance of such change is unclear (Herrmann 2010).

Intrauterine device (IUD) or system (IUS)

For the levonorgestrel IUS, no mechanism is apparent that might affect bone health (Mansour 2012). However, a case‐control study reported reduced fracture risk for ever‐use of the hormonal IUD and longer use of that IUD (Vestergaard 2006).

Why it is important to do this review

Hormonal contraceptives are among the most effective and most widely‐used contraceptives. Concern about fractures may limit the use of these effective contraceptives. Women might switch to less effective methods or use nothing, potentially leading to increased rates of unintended pregnancy. The question about an association between steroidal contraceptives and fractures is important to examine systematically with the available evidence. Since our initial review in 2006, we also examined evidence of actual fracture risk in observational studies of hormonal contraceptives (Lopez 2012). In this update, we further examine the effect of using steroidal contraceptives before menopause on general bone health, based on evidence from randomized controlled trials.

Objectives

Our aim was to evaluate the effect of using hormonal contraceptives before menopause on the risk of fracture in women.

Methods

Criteria for considering studies for this review

Types of studies

We considered randomized controlled trials (RCTs) if they examined fractures, bone mineral density, or bone turnover in women who used hormonal contraceptives prior to menopause.

Studies were excluded if hormones were provided for treatment of a specific condition or if the study focused on women with a certain condition, such as endometriosis, polycystic ovary disease, or hirsutism. Also excluded were studies that provided hormone replacement therapy to postmenopausal women.

Types of participants

We included women in the identified trials who were randomly assigned to study groups.

Types of interventions

Interventions included comparisons of a hormonal contraceptive with a placebo or with another hormonal contraceptive that differed in terms of drug, dosage, or regimen. Interventions also included the provision of a supplement, for example, another hormone or a vitamin or mineral preparation, to one group.

We excluded interventions involving exercise, which appears to interact with hormonal contraceptives to affect bone health.

Types of outcome measures

Primary outcomes

The primary outcome was fractures occurring after baseline, particularly fractures of the spine, hip, and wrist.

Secondary outcomes

Bone mineral density, which could have been measured, e.g., at the femur, lumbar spine or whole body;
Biochemical markers of bone turnover (Vasikaran 2011a; Vasikaran 2011b), e.g.,
- bone formation ‐ serum osteocalcin, alkaline phosphatase, and type I procollagen;
- bone resorption ‐ serum calcium and C‐telopeptide; urinary pyridinoline and N‐telopeptides.

Search methods for identification of studies

Electronic searches

Through April 2014, we searched the computerized databases MEDLINE, POPLINE, Cochrane Central Register of Controlled Trials (CENTRAL), EMBASE, and LILACS for studies of fracture or bone health and hormonal contraceptives. In addition, we searched for recent clinical trials through ClinicalTrials.gov and the International Clinical Trials Registry Platform (ICTRP). The strategies are given in Appendix 1. Previous search strategies can be found in Appendix 2.

Searching other resources

We examined reference lists of relevant articles for other trials. For the initial review, we wrote to known investigators for information about other published or unpublished trials not discovered in our search.

Data collection and analysis

Selection of studies

We assessed for inclusion all titles and abstracts identified during the literature searches with no language limitation.

Data extraction and management

Two authors independently abstracted the data. Data were entered into RevMan, and a second author verified accuracy. Any discrepancies were resolved by discussion.

Assessment of risk of bias in included studies

Studies were examined for methodological quality, according to the principles recommended in Higgins 2011. Factors considered were study design, method for generating the randomization sequence, allocation concealment, blinding, and losses to follow up and early discontinuation. We also examined the methods used for assessing the outcomes.

Assessment of heterogeneity

None of the trials examined the same types of interventions. Therefore, we did not combine any trials in a meta‐analysis.

Data synthesis

For continuous variables, the mean difference (MD) was computed with 95% confidence interval (CI) using a fixed‐effect model. RevMan uses the inverse variance approach. For dichotomous outcomes, the Mantel‐Haenszel odds ratio (OR) with 95% CI was calculated using a fixed‐effect model.

We applied principles from GRADE to assess the evidence quality and address confidence in the effect estimates (Balshem 2011). When a meta‐analysis is not viable due to varied interventions, a 'Summary of findings' table is not feasible. Therefore, we did not conduct a formal GRADE assessment with an evidence profile and 'Summary of findings' table (Guyatt 2011).

For the 2011 update, we added an assessment of evidence quality using the GRADE approach (Higgins 2011). This assessment was based on the quality of evidence from the individual studies. In 2014, we refined the criteria used, based on our subsequent experience with other reviews. Evidence quality could be high, moderate, low, or very low. We considered the evidence from RCTs to be high quality initially, then downgraded for each of the following: a) randomization sequence generation and allocation concealment: no information on either, or one was inadequate; b) lack of blinding; c) follow up was 12 months or less for BMD measures only; d) losses were greater than 20% for the primary analysis.

Sensitivity analysis

In 2014, we added a sensitivity analysis. This included trials that provided sufficient data and evidence of moderate or high quality.

Results

Description of studies

Results of the search

The 2014 search produced 54 unduplicated citations from the main databases. In addition, we found seven unduplicated trials through ClinicalTrials.gov and ICTRP. Three new trials were added, including one that had been 'ongoing' in the previous update (Cibula 2012; Gai 2012; Sordal 2012). In addition new ongoing trial was added (Bonny 2013). An earlier ongoing trial is still awaiting classification due to lack of a report (Teva 2013).

We identified 19 randomized controlled trials that met the criteria for inclusion. Bone density was measured in 17 trials; the other two assessed biochemical markers of bone turnover (Paoletti 2000; Rad 2011). Twelve studies assessed BMD as well as biochemical markers. None had fracture as an outcome. Of the 17 trials that examined BMD, 14 measured the lumbar spine using dual‐energy X‐ray absorptiometry (DEXA), although the measurement site varied somewhat. The other three studies with BMD used computed tomography for the lumbar spine (Endrikat 2004), DEXA for the arm (Bahamondes 2006), and single photon absorptiometry for the arm (Naessen 1995).

Included studies

Most of the studies were 12 to 24 months in duration and two were 36 months long (Endrikat 2004; Kaunitz 2009). One trial was limited to six months (Naessen 1995). A crossover trial had the participants switch COCs at 9 months for a total duration of 18 months (Cibula 2012). Three studies focused on adolescents (Cibula 2012; Cromer 2005; Gai 2012).

The types and formulations of hormonal contraceptives varied. Eleven trials compared different COCs or regimens of COCs:

Two studied desogestrel‐containing COCs (Berenson 2001; Gai 2012).
Four examined levonorgestrel preparations as the investigational drug or the comparator (Endrikat 2004; Hartard 2006; Rad 2011; Sordal 2012).
Three examined gestodene preparations (Paoletti 2000; Nappi 2003; Cibula 2012).
Two examined drospirenone‐containing COCs (Nappi 2005; Gargano 2008).

In addition, five trials examined an injectable versus another injectable, implant, or IUD (Naessen 1995; Von Kesseru 2000; Cromer 2005; Cundy 2003; Kaunitz 2009), two compared two implants each (Di 1999; Bahamondes 2006), and one studied the transdermal patch versus the vaginal ring (Massaro 2010).

Risk of bias in included studies

Allocation

Study design and reporting varied in quality across these trials.

Randomization information was as follows:

Two trials had interactive voice‐response systems, based on computer‐generated random lists (Kaunitz 2009; Rad 2011).
Most used a random numbers table or a computer for random‐number sequence generation (Bahamondes 2006; Berenson 2001; Cibula 2012; Cromer 2005; Cundy 2003; Massaro 2010; Naessen 1995; Nappi 2003; Nappi 2005; Paoletti 2000; Von Kesseru 2000). Cromer 2005 mentioned block randomization techniques but did not specify the block size.
Gai 2012 reported randomized by 'drawing lots' without further explanation.
Five studies did not provide the method for sequence generation (Di 1999; Endrikat 2004; Gargano 2008; Hartard 2006; Sordal 2012).

Allocation concealment was unclear in many studies and not mentioned in others. As noted above, two had an interactive voice‐response system (Kaunitz 2009; Rad 2011). Bahamondes 2006 reported having sealed envelopes prepared at the WHO, and Cromer 2005 communicated that they used serially‐numbered opaque envelopes. Naessen 1995 used sealed envelopes, and Nappi 2005 reported the sequence was concealed until treatment was assigned. Two trials did not have any concealment and 11 trials had insufficient or no information.

Blinding

Some blinding was used in six trials: double‐blind (Cromer 2005; Cundy 2003; Endrikat 2004); investigators and providers (Kaunitz 2009); laboratory personnel (Massaro 2010); and participants (Berenson 2001).
Six trials were open‐label (Hartard 2006; Nappi 2005; Paoletti 2000; Rad 2011; Sordal 2012; Von Kesseru 2000).
Information on blinding was not available from seven studies (Bahamondes 2006; Cibula 2012; Di 1999; Gai 2012; Gargano 2008; Naessen 1995; Nappi 2003).

Incomplete outcome data

Losses were high in several trials, but largely due to method discontinuation or missing data. Losses greater than 20% threaten trial validity (Strauss 2005).

In Von Kesseru 2000, loss of participants or missing data for BMD at 12 months was 48% and 79% in the two intervention groups. At 24 months, the figures were 70% and 84%.
Berenson 2001 losses were attributed to discontinuation or failure to obtain a bone scan within the required window: at 12 months, 62% and 68% for the two intervention groups; at 24 months, 71% and 54%.
In Endrikat 2004, loss was 52% at 36 months with 61% loss for bone data.
Cundy 2003 had a 29% loss due to early discontinuation.
In Cromer 2005, 24% withdrew by 12 months and 43% withdrew by 24 months. This does not include those without assessments due to early study closure.
Three trials had high overall losses: Kaunitz 2009 (39%); Rad 2011 (29%); Sordal 2012 (41%); losses to follow up were under 20%.

Effects of interventions

Progestin‐only methods

Six trials examined methods containing only the hormone progestin, including two trials of implants and four that examined the studied DMPA 150 mg.

Implants

Di 1999 examined the six‐capsule Norplant versus a similar domestic implant (manufactured in China). BMD at Ward's triangle was higher among Norplant users than domestic implant users at 12 months (mean difference (MD) 0.07; 95% CI 0.00 to 0.14, Analysis 1.4). Both types of implants had six capsules with the same amount of levonorgestrel. The groups did not differ significantly for BMD at the other locations, nor for serum and urinary measures.
In Bahamondes 2006, the implants studied were a single‐rod etonogestrel‐releasing implant and a two silicone rod levonorgestrel‐releasing implant. By 18 months, the etonogestrel‐implant group had a greater percent decrease in BMD at the midshaft ulna than the two‐rod levonorgestrel group (MD ‐0.39; 95% CI ‐0.56 to ‐0.22, Analysis 2.2) and at the distal radius (MD ‐1.00; 95% CI ‐1.09 to ‐0.91, Analysis 2.4). A secondary paper reported on BMD at 36 months, but the losses to follow up by that time were large and the groups were not significantly different for BMD at the distal radius.

1.4 — Comparison 1 Levonorgestrel‐releasing implants: Norplant versus Chinese implant, Outcome 4 Bone mineral density (Ward's triangle) at 12 months.

2.2 — Comparison 2 Etonogestrel‐releasing implant versus levonorgestrel‐releasing implant, Outcome 2 Percent change in bone mineral density (midshaft ulna) by 18 months.

2.4 — Comparison 2 Etonogestrel‐releasing implant versus levonorgestrel‐releasing implant, Outcome 4 Percent change in bone mineral density (distal radius) by 18 months.

Injectable DMPA 150 mg

Naessen 1995 randomized women to either DMPA 150 mg every 12 weeks or the levonorgestrel implant (Norplant). The DMPA group had a lower mean for alkaline phosphatase, a marker of bone formation, than the implant group at six months (MD ‐0.65; 95% CI ‐1.21 to ‐0.09, Analysis 3.1). The groups did not differ significantly for serum osteocalcin and calcium and for urinary hydroxyproline/creatinine. BMD data were shown in a figure rather than a table. By six months, BMD at the forearm reportedly increased in the levonorgestrel implant group (reported P = 0.006) and decreased insignificantly in the DMPA group. The group difference was reportedly significant at the proximal (reported P = 0.025) but not the distal forearm.

Two trials examined estrogen supplement versus a placebo for women on DMPA. All participants had an injection of DMPA 150 mg every 12 weeks.

- In Cromer 2005, one DMPA group received monthly injections of estradiol cypionate (E₂C) 5 mg whereas the other received the placebo supplement of 5 mL normal saline solution. Bone mineral apparent density (BMAD) was used to correct for variation in bone (see Characteristics of included studies). At 12 months, the groups with the estrogen supplement had increases while the placebo‐supplement group had decreases for spine BMD (MD 2.90; 95% CI 1.80 to 4.00, Analysis 4.1), spine BMAD (MD 2.70; 95% CI 1.60 to 3.80, Analysis 4.2), and femoral neck BMD (MD 3.20; 95% CI 1.36 to 5.04, Analysis 4.3). The groups were not significantly different for femoral neck BMAD (Analysis 4.4). At 24 months, the same trend was seen: spine BMD (MD 4.60; 95% CI 2.87 to 6.33, Analysis 4.5), spine BMAD (MD 4.90; 95% CI 3.11 to 6.69, Analysis 4.6), femoral neck BMD (MD 9.80; 95% CI 4.96 to 14.64, Analysis 4.7), and femoral neck BMAD (MD 7.10; 95% CI 0.50 to 13.70, Analysis 4.8). The trial was stopped early due to the differences reaching the predetermined significance level (P < 0.001).
- Cundy 2003 randomized DMPA users to daily intake of conjugated estrogens 62.5 μg or to a placebo supplement. BMD was measured at the lumbar spine, femoral neck, Ward's triangle, trochanter, and total body. For lumbar spine BMD, the group with the estrogen supplement had a small increase and the placebo‐supplement group had a small decrease by 12 months (MD 0.02; 95% CI 0.00 to 0.04, Analysis 5.1) and by 24 months (MD 0.04; 95% CI 0.02 to 0.06, Analysis 5.2). More than a fourth of the participants discontinued early. No significant changes were reportedly apparent in plasma calcium, phosphate, or alkaline phosphatase activity or in urinary N‐telopeptides/creatinine.

In Kaunitz 2009, intramuscular DMPA 150 mg/mL (DMPA‐IM) was compared with subcutaneous DMPA 104 mg/0.65 mL (DMPA‐SC). The groups did not differ significantly in the proportions with a 5% or greater decrease in total hip BMD at one, two, or three years. For lumbar spine BMD, more of the DMPA‐SC group had a 5% or greater decrease by year 3 (OR 2.11; 95% CI 1.00 to 4.45, Analysis 6.6). Losses due to discontinuation were high.

3.1 — Comparison 3 DMPA 150 mg versus levonorgestrel‐releasing implant, Outcome 1 Serum alkaline phosphatase at 6 months.

4.1 — Comparison 4 DMPA 150 mg + estradiol cypionate 5 mg versus DMPA 150 mg + placebo, Outcome 1 Percent change in bone mineral density (spine) by 12 months.

4.2 — Comparison 4 DMPA 150 mg + estradiol cypionate 5 mg versus DMPA 150 mg + placebo, Outcome 2 Percent change in bone mineral apparent density (spine) by 12 months.

4.3 — Comparison 4 DMPA 150 mg + estradiol cypionate 5 mg versus DMPA 150 mg + placebo, Outcome 3 Percent change in bone mineral density (femoral neck) by 12 months.

4.4 — Comparison 4 DMPA 150 mg + estradiol cypionate 5 mg versus DMPA 150 mg + placebo, Outcome 4 Percent change in bone mineral apparent density (femoral neck) by 12 months.

4.5 — Comparison 4 DMPA 150 mg + estradiol cypionate 5 mg versus DMPA 150 mg + placebo, Outcome 5 Percent change in bone mineral density (spine) by 24 months.

4.6 — Comparison 4 DMPA 150 mg + estradiol cypionate 5 mg versus DMPA 150 mg + placebo, Outcome 6 Percent change in bone mineral apparent density (spine) by 24 months.

4.7 — Comparison 4 DMPA 150 mg + estradiol cypionate 5 mg versus DMPA 150 mg + placebo, Outcome 7 Percent change in bone mineral density (femoral neck) by 24 months.

4.8 — Comparison 4 DMPA 150 mg + estradiol cypionate 5 mg versus DMPA 150 mg + placebo, Outcome 8 Percent change in bone mineral apparent density (femoral neck) by 24 months.

5.1 — Comparison 5 DMPA 150 mg + conjugated estrogens 625 µg versus DMPA 150 mg + placebo, Outcome 1 Change in bone mineral density (lumbar spine) by 12 months.

5.2 — Comparison 5 DMPA 150 mg + conjugated estrogens 625 µg versus DMPA 150 mg + placebo, Outcome 2 Change in bone mineral density (lumbar spine) by 24 months.

6.6 — Comparison 6 DMPA‐SC 104 mg versus DMPA‐IM 150 mg, Outcome 6 Decrease in lumbar spine BMD >= 5% from baseline (year 3).

Combination contraceptives

These 13 trials included 11 that compared combined oral contraceptives, as well as one of a combination injectable versus a non‐hormonal IUD and one of the transdermal patch versus the vaginal ring.

Oral contraceptives

Two compared desogestrel‐containing COCs versus other COCs:

Berenson 2001 randomized women to norethindrone 1 mg plus ethinyl estradiol (EE) 35 μg or to desogestrel 150 μg plus EE 30 μg. The norethindrone group had a significantly greater increase in BMD at the lumbar spine at 12 months than the desogestrel group (MD 1.83; 95% CI 0.42 to 3.24, Analysis 7.1). By 24 months, both groups had decreases from baseline but they were not significantly different. However, only about one‐third of the original participants remained at 12 months.
Gai 2012 also used desogestrel 150 μg plus EE 30 μg, but the comparison was cyproterone acetate (CPA) 2 mg plus EE 35 μg. The group with the desogestrel‐containing COC did not differ significantly from the CPA group for the BMD measures of lumbar spine or femoral neck at 12 or 24 months (Analysis 8.1 to Analysis 8.4).

7.1 — Comparison 7 Norethindrone 1 mg + EE 35 µg versus desogestrel 150 µg + EE 30 µg, Outcome 1 Percent change in bone mineral density (lumbar spine) by 12 months.

8.1 — Comparison 8 Desogestrel 150 µg + EE 30 µg versus cyproterone acetate 2 mg + EE 35 µg, Outcome 1 Change in BMD of lumbar spine by 12 months.

8.4 — Comparison 8 Desogestrel 150 µg + EE 30 µg versus cyproterone acetate 2 mg + EE 35 µg, Outcome 4 Change in BMD of femoral neck by 24 months.

Four trials used preparations containing levonorgestrel, either as the investigational drug or the comparator:

Endrikat 2004 compared levonorgestrel 100 μg plus 20 EE μg versus levonorgestrel 150 μg plus EE 30 μg. The two groups did not differ significantly in their slight decreases in BMD at 36 months (Analysis 9.1 to Analysis 9.4). Serum alkaline phosphatase increased and N‐telopeptides decreased, but change did not differ significantly between the groups. More than half of the participants were lost to follow up.
Hartard 2006 examined levonorgestrel 100 μg plus 20 EE μg versus desogestrel 150 μg plus EE 20 μg. By 12 months, the desogestrel group lost more areal BMD at the lumbar spine than the levonorgestrel group, but the difference was small (MD 1.41; 95% CI ‐0.11 to 2.93, Analysis 10.1). The desogestrel group had a greater decrease in serum alkaline phosphatase (MD 15.31; 95% CI 3.91 to 26.71, Analysis 10.3). The groups did not differ significantly in change in areal BMD at the femoral neck or in serum osteocalcin or C‐telopeptides (Analysis 10.2; Analysis 10.4; Analysis 10.5).
Rad 2011 compared a continuous regimen of levonorgestrel 90 μg plus EE 20 μg versus a cyclic regimen of levonorgestrel 100 μg plus EE 20 μg. The report provided standard errors and did not contain cell sizes, so we could not analyze any data. Reportedly, changes in osteocalcin and C‐telopeptides were not significantly different between the groups by cycle 13.
For Sordal 2012, the COC of interest was nomegestrol 2.5 mg plus [17ß] estradiol 1.5 mg (NOMAC‐E2), and the comparison was levonorgestrel 150 μg plus EE 30 μg. By cycle 26, the groups did not differ significantly for change in z‐score of the lumbar spine (Analysis 11.1) or femoral neck (Analysis 11.2).

9.1 — Comparison 9 Levonorgestrel 100 µg + EE 20 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 1 Bone mineral density (lumbar spine) at 12 months.

9.4 — Comparison 9 Levonorgestrel 100 µg + EE 20 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 4 Percent change in bone mineral density (lumbar spine) by 36 months.

10.1 — Comparison 10 Levonorgestrel 100 µg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 1 Percent change in areal bone mineral density (lumbar spine) by 12 months.

10.3 — Comparison 10 Levonorgestrel 100 µg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 3 Percent change in serum bone‐specific alkaline phosphatase by 12 months.

10.2 — Comparison 10 Levonorgestrel 100 µg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 2 Percent change in areal bone mineral density (femoral neck) by 12 months.

10.4 — Comparison 10 Levonorgestrel 100 µg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 4 Percent change in serum osteocalcin by 12 months.

10.5 — Comparison 10 Levonorgestrel 100 µg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 5 Percent change in serum cross‐linked telopeptides by 12 months.

11.1 — Comparison 11 Nomegestrol 2.5 mg + estradiol 1.5 mg versus levonorgestrel 150 µg + EE 30 µg, Outcome 1 Change in z‐score of lumbar spine after cycle 26.

11.2 — Comparison 11 Nomegestrol 2.5 mg + estradiol 1.5 mg versus levonorgestrel 150 µg + EE 30 µg, Outcome 2 Change in z‐score of femoral neck after cycle 26.

Gestodene‐containing COCs were the focus of three trials:

Paoletti 2000 randomized women to gestodene 75 μg plus EE 20 μg or gestodene 75 μg plus EE 30 μg. At 12 months, urinary deoxypyridinoline was lower in the EE 30 μg group than in the EE 20 μg group (MD 1.20; 95% CI 0.37 to 2.03, Analysis 12.2). The study groups were not significantly different for serum osteocalcin and urinary pyridinoline.
Nappi 2003 studied gestodene 75 μg plus EE 20 μg versus gestodene 60 μg plus EE 15 μg. The results were presented in figures without absolute values. The investigators reported no significant difference at 12 months between or within the groups in BMD at the lumbar spine or in serum osteocalcin. The study groups reportedly had significant declines in urinary pyridinoline and deoxypyridinoline by 6 and 12 months (reported P < 0.05), but the groups did not differ significantly.
The crossover study of Cibula 2012 compared gestodene 75 μg plus EE 30 μg versus gestodene 60 μg plus EE 15 μg. The participants were switched to the other formulation at nine months; study duration was 18 months. Measures included BMD of the lumbar spine, femur, and distal radius, as well as serum type I procollagen (PINP) and type I collagen cross‐linked C‐telopeptide (ßCTX1). The report included results of the full analysis of variance model (ANOVA) for lumbar BMD, PINP, and ßCTX1 (Table 1). Reportedly, dose was significantly associated with change in lumbar BMD (reported F‐ratio = 4.6; reported P value = 0.037). The COC containing EE 30 μg showed an increase while the COC containing EE 15 μg showed a decrease. Dose was also reportedly associated with a difference in PINP (reported F‐ratio = 8.3; reported P value = 0.005), but the text and figure were inconsistent regarding the direction of change. For ßCTX1, no significant difference was reported.

12.2 — Comparison 12 Gestodene 75 µg + EE 20 µg versus gestodene 75 µg + EE 30 µg, Outcome 2 Urinary deoxypyridinoline at 12 months.

3. Outcomes by 18 months (Cibula 2012).

Outcome at 18 months (crossover at 9 months)¹	Variable²	Reported  F‐ratio	Reported  P value
Change in lumbar spine BMD (g/cm²)	dose	4.6	0.037
Change in propeptide of type I procollagen (μg/l)	dose	8.3	0.005
Change in cross‐linked telopeptide (μg/l)	dose	0.7	0.397

Study	Comparison groups	Bone mineral  density:  # differences/  # measures¹	Biochemical  markers:  # differences/  # measures
Progestin‐only contraceptives
Implants
Di 1999	levonorgestrel 6‐rod (standard) vs levonorgestrel 6‐rod (domestic)	1/4	0/4
Bahamondes 2006	etonogestrel 1‐rod vs levonorgestrel 2‐rod	2/5	‐‐‐
Injectable DMPA
Naessen 1995	DMPA vs levonorgestrel 6‐rod implant	insufficient data	1/4
Cromer 2005	DMPA + E₂C vs DMPA + placebo	7/8	‐‐‐
Cundy 2003	DMPA + conjugated estrogens 62.5 μg vs DMPA + placebo	2/10	‐‐‐
Kaunitz 2009	DMPA subcutaneous vs DMPA intramuscular	1/6	‐‐‐
Combination contraceptives
Oral contraceptives
Berenson 2001	norethindrone 1 mg + EE 35 μg vs desogestrel 150 μg + EE 30 μg	1/2	‐‐‐
Gai 2012	desogestrel 150 μg + EE 30 μg vs cyproterone acetate 2 mg + EE 35 μg	0/4	‐‐‐
Endrikat 2004	levonorgestrel 100 μg + EE 20 μg vs levonorgestrel 150 μg + EE 30 μg	0/4	0/2
Hartard 2006	levonorgestrel 100 μg + EE 20 μg vs desogestrel 150 μg + EE 20 μg	0/2	1/3
Rad 2011	levonorgestrel 90 μg + EE 20 μg (continuous) vs levonorgestrel 100 μg + EE 20 μg (cyclic)	‐‐‐	insufficient data
Sordal 2012	nomegestrol 2.5 mg + 17ß estradiol 1.5 mg vs levonorgestrel 150 μg + EE 30 μg	0/2	‐‐‐
Paoletti 2000	gestodene 75 μg + EE 20 μg vs gestodene 75 μg + EE 30 μg	‐‐‐	1/3
Nappi 2003	gestodene 75 μg + EE 20 μg vs gestodene 60 μg + EE 15 μg	insufficient data	insufficient data
Cibula 2012	gestodene 75 μg + EE 30 μg vs gestodene 60 μg + EE 15 μg	1/1	insufficient data
Nappi 2005	drospirenone 3 mg + EE 30 μg vs gestodene 75 μg + EE 30 μg	0/1	insufficient data
Gargano 2008	drospirenone 3 mg + EE 30 vs drospirenone 3 mg + EE 20	0/1	insufficient data
Injectable
Von Kesseru 2000	norethisterone enanthate + E₂V vs Nova‐T IUD	0/2	‐‐‐
Patch versus ring
Massaro 2010	transdermal patch vs vaginal ring	0/1	0/3

Study	Comparison groups	Inadequate randomization  and allocation  concealment	No blinding	Follow up  <= 12 months  (only BMD)	Losses  > 20%	Quality of  evidence¹
Injectable (versus implant, injectable, or IUD)
Naessen 1995	DMPA vs LNG 6‐rod implant	‐‐‐	unclear	‐‐‐	‐‐‐	high
Cromer 2005	DMPA + E₂C vs DMPA + placebo	‐‐‐	‐‐‐	‐‐‐	‐1	moderate
Cundy 2003	DMPA + estrogen vs DMPA + placebo	‐‐‐	‐‐‐	‐‐‐	‐1	moderate
Kaunitz 2009	DMPA‐subcutaneous vs DMPA‐intramuscular	‐‐‐	‐‐‐	‐‐‐	‐1	moderate
Evidence quality from 4 DMPA trials						moderate
Implants
Di 1999	levonorgestrel 6‐rod (standard) vs levonorgestrel 6‐rod (domestic)	‐1	unclear	‐‐‐	‐‐‐	moderate
Bahamondes 2006	etonogestrel 1‐rod vs levonorgestrel 2‐rod	‐‐‐	unclear	‐‐‐	‐‐‐	high
Evidence quality from 2 implant trials						moderate to high
Combined oral contraceptives
Berenson 2001	norethindrone 1 mg + EE 35 μg vs desogestrel 150 μg + EE 30 μg	‐1	‐‐‐	‐1	‐1	very low
Gai 2012	desogestrel 150 μg + EE 30 μg vs cyproterone acetate 2 mg + EE 35 μg	‐1	unclear	‐‐‐	‐‐‐	moderate
Endrikat 2004	levonorgestrel 100 μg + EE 20 μg vs levonorgestrel 150 μg + EE 30 μg	‐1	‐‐‐	‐‐‐	‐1	low
Hartard 2006	levonorgestrel 100 μg + EE 20 μg vs desogestrel 150 μg + EE 20 μg	‐1	‐1	‐‐‐	‐‐‐	low
Rad 2011	levonorgestrel 90 μg + EE 20 μg (continuous) vs levonorgestrel 100 μg + EE 20 μg (cyclic)	‐‐‐	‐1	‐‐‐	‐1	low
Sordal 2012	nomegestrol 2.5 mg + 17ß estradiol 1.5 mg vs levonorgestrel 150 μg + EE 30 μg	‐1	‐1	‐‐‐	‐1	very low
Paoletti 2000	gestodene 75 μg + EE 20 μg vs gestodene 75 μg + EE 30 μg	‐‐‐	‐1	‐‐‐	‐‐‐	moderate
Nappi 2003	gestodene 75 μg + EE 20 μg vs gestodene 60 μg + EE 15 μg	‐‐‐	unclear	‐‐‐	‐‐‐	high
Cibula 2012	gestodene 75 μg + EE 30 μg vs gestodene 60 μg + EE 15 μg	‐‐‐	unclear	‐‐‐	‐‐‐	high
Nappi 2005	drospirenone 3 mg + EE 30 μg vs gestodene 75 μg + EE 30 μg	‐‐‐	‐1	‐1	‐‐‐	low
Gargano 2008	drospirenone 3 mg + EE 30 vs drospirenone 3 mg + EE 20	‐1	unclear	‐1	‐‐‐	low
Evidence quality from 11 COC studies						low
Injectable or patch versus ring
Von Kesseru 2000	norethisterone enanthate + E₂V vs Nova‐T IUD	‐‐‐	‐‐‐	‐‐‐	‐1	moderate
Massaro 2010	transdermal patch vs vaginal ring	‐‐‐	‐‐‐	‐‐‐	‐‐‐	high
Evidence quality from 2 trials of injectable or patch versus ring						moderate to high
*Overall evidence quality from 19 trials*						*moderate*

Study¹	Comparison groups	Fracture	Bone mineral  density²	Biochemical  markers²
Injectable DMPA
Naessen 1995	DMPA vs levonorgestrel 6‐rod implant	‐‐‐	‐‐‐	1/4
Cromer 2005	DMPA + E₂C vs DMPA + placebo supplement	‐‐‐	7/8	‐‐‐
Cundy 2003	DMPA + estrogen vs DMPA + placebo supplement	‐‐‐	2/10	‐‐‐
Kaunitz 2009	DMPA subcutaneous vs DMPA intramuscular	‐‐‐	1/6	‐‐‐
Implants
Di 1999	levonorgestrel 6‐rod (standard) vs levonorgestrel 6‐rod (domestic)	‐‐‐	1/4	0/4
Bahamondes 2006	etonogestrel 1‐rod vs levonorgestrel 2‐rod	‐‐‐	2/5	‐‐‐
Combined oral contraceptives
Gai 2012	desogestrel 150 μg + EE 30 μg vs cyproterone acetate 2 mg + EE 35 μg	‐‐‐	0/4	‐‐‐
Paoletti 2000	gestodene 75 μg + EE 20 μg vs gestodene 75 μg + EE 30 μg	‐‐‐	‐‐‐	1/3
Cibula 2012	gestodene 75 μg + EE 30 μg vs gestodene 60 μg + EE 15 μg	‐‐‐	1/1	‐‐‐
Combination injectable or patch versus ring
Von Kesseru 2000	norethisterone enanthate + E₂V vs Nova‐T IUD	‐‐‐	0/2	‐‐‐
Massaro 2010	transdermal patch vs vaginal ring	‐‐‐	0/1	0/3

Date	Event	Description
7 May 2014	New search has been performed	Search updated.
25 March 2014	New citation required but conclusions have not changed	Three new trials included (Cibula 2012; Gai 2012; Sordal 2012); one ongoing trial added (Bonny 2013); one study excluded (Berenson 2012).
27 January 2014	Amended	Criteria revised for assessing evidence quality (Data synthesis; Table 19).  Sensitivity analysis added (Table 20).

Date	Event	Description
8 June 2011	New search has been performed	Searches were updated for MEDLINE, POPLINE, and CENTRAL
21 April 2011	New citation required but conclusions have not changed	Three new trials were included (Kaunitz 2009; Massaro 2010; Rad 2011). Other studies were excluded or added as Ongoing studies.
29 March 2011	New search has been performed	Searches were updated
18 December 2008	New citation required but conclusions have not changed	Added: two new trials (Gargano 2008; Hartard 2006), follow‐up data from earlier trial (Bahamondes 2006), and an ongoing trial (Schering‐Plough 2011a). An additional trial was excluded (Teichmann 2009a).
24 November 2008	New search has been performed	Searches updated in Nov and Dec 2008
15 April 2008	Amended	Converted to new review format.
28 June 2006	New citation required and conclusions have changed	Substantive amendment

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density (L2 to L4) at 12 months	1	60	Mean Difference (IV, Fixed, 95% CI)	0.04 [‐0.02, 0.10]
2 Bone mineral density (femoral neck) at 12 months	1	58	Mean Difference (IV, Fixed, 95% CI)	0.05 [‐0.01, 0.11]
3 Bone mineral density (trochanter) at 12 months	1	58	Mean Difference (IV, Fixed, 95% CI)	0.03 [‐0.02, 0.08]
4 Bone mineral density (Ward's triangle) at 12 months	1	58	Mean Difference (IV, Fixed, 95% CI)	0.07 [0.00, 0.14]
5 Serum osteocalcin at 12 months	1	60	Mean Difference (IV, Fixed, 95% CI)	0.16 [‐0.58, 0.90]
6 Serum alkaline phosphatase at 12 months	1	60	Mean Difference (IV, Fixed, 95% CI)	1.94 [‐13.29, 17.17]
7 Urine hydroxyproline/creatinine at 12 months	1	60	Mean Difference (IV, Fixed, 95% CI)	1.70 [‐2.86, 6.26]
8 Urine calcium/creatinine at 12 months	1	60	Mean Difference (IV, Fixed, 95% CI)	‐20.12 [‐44.01, 3.77]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density (midshaft ulna) at 18 months	1	111	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.02, 0.02]
2 Percent change in bone mineral density (midshaft ulna) by 18 months	1	111	Mean Difference (IV, Fixed, 95% CI)	‐0.39 [‐0.56, ‐0.22]
3 Bone mineral density (distal radius) at 18 months	1	111	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.02, 0.02]
4 Percent change in bone mineral density (distal radius) by 18 months	1	111	Mean Difference (IV, Fixed, 95% CI)	‐1.0 [‐1.09, ‐0.91]
5 Bone mineral density (distal radius) at 36 months	1	75	Mean Difference (IV, Fixed, 95% CI)	‐0.01 [‐0.03, 0.02]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Serum alkaline phosphatase at 6 months	1	19	Mean Difference (IV, Fixed, 95% CI)	‐0.65 [‐1.21, ‐0.09]
2 Serum osteocalcin at 6 months	1	19	Mean Difference (IV, Fixed, 95% CI)	‐1.44 [‐3.39, 0.51]
3 Serum calcium at 6 months	1	19	Mean Difference (IV, Fixed, 95% CI)	0.02 [‐0.05, 0.09]
4 Urinary hydroxyproline/creatinine at 6 months	1	19	Mean Difference (IV, Fixed, 95% CI)	7.30 [‐14.52, 29.12]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Percent change in bone mineral density (spine) by 12 months	1	69	Mean Difference (IV, Fixed, 95% CI)	2.90 [1.80, 4.00]
2 Percent change in bone mineral apparent density (spine) by 12 months	1	69	Mean Difference (IV, Fixed, 95% CI)	2.7 [1.60, 3.80]
3 Percent change in bone mineral density (femoral neck) by 12 months	1	69	Mean Difference (IV, Fixed, 95% CI)	3.2 [1.36, 5.04]
4 Percent change in bone mineral apparent density (femoral neck) by 12 months	1	69	Mean Difference (IV, Fixed, 95% CI)	1.2 [‐2.01, 4.41]
5 Percent change in bone mineral density (spine) by 24 months	1	34	Mean Difference (IV, Fixed, 95% CI)	4.6 [2.87, 6.33]
6 Percent change in bone mineral apparent density (spine) by 24 months	1	34	Mean Difference (IV, Fixed, 95% CI)	4.9 [3.11, 6.69]
7 Percent change in bone mineral density (femoral neck) by 24 months	1	34	Mean Difference (IV, Fixed, 95% CI)	9.8 [4.96, 14.64]
8 Percent change in bone mineral apparent density (femoral neck) by 24 months	1	34	Mean Difference (IV, Fixed, 95% CI)	7.10 [0.50, 13.70]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Change in bone mineral density (lumbar spine) by 12 months	1	27	Mean Difference (IV, Fixed, 95% CI)	0.02 [0.00, 0.04]
2 Change in bone mineral density (lumbar spine) by 24 months	1	26	Mean Difference (IV, Fixed, 95% CI)	0.04 [0.02, 0.06]
3 Change in bone mineral density (Ward's triangle) by 12 months	1	26	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.03, 0.03]
4 Change in bone mineral density (Ward's triangle) by 24 months	1	25	Mean Difference (IV, Fixed, 95% CI)	0.04 [‐0.00, 0.08]
5 Change in bone mineral density (trochanter) by 12 months	1	26	Mean Difference (IV, Fixed, 95% CI)	0.03 [‐0.01, 0.07]
6 Change in bone mineral density (trochanter) by 24 months	1	25	Mean Difference (IV, Fixed, 95% CI)	0.02 [‐0.01, 0.05]
7 Change in bone mineral density (femoral neck) by 12 months	1	27	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.04, 0.04]
8 Change in bone mineral density (femoral neck) by 24 months	1	26	Mean Difference (IV, Fixed, 95% CI)	0.02 [‐0.02, 0.06]
9 Change in bone mineral density (total body) by12 months	1	27	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.01, 0.01]
10 Change in bone mineral density (total body) by 24 months	1	26	Mean Difference (IV, Fixed, 95% CI)	0.01 [‐0.01, 0.03]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Decrease in total hip BMD >= 5% from baseline (year 1)	1	328	Odds Ratio (M‐H, Fixed, 95% CI)	0.60 [0.29, 1.26]
2 Decrease in total hip BMD >= 5% from baseline (year 2)	1	207	Odds Ratio (M‐H, Fixed, 95% CI)	0.78 [0.43, 1.40]
3 Decrease in total hip BMD >= 5% from baseline (year 3)	1	117	Odds Ratio (M‐H, Fixed, 95% CI)	1.16 [0.56, 2.41]
4 Decrease in lumbar spine BMD >= 5% from baseline (year 1)	1	328	Odds Ratio (M‐H, Fixed, 95% CI)	0.79 [0.47, 1.32]
5 Decrease in lumbar spine BMD >= 5% from baseline (year 2)	1	208	Odds Ratio (M‐H, Fixed, 95% CI)	0.58 [0.34, 1.01]
6 Decrease in lumbar spine BMD >= 5% from baseline (year 3)	1	115	Odds Ratio (M‐H, Fixed, 95% CI)	2.11 [1.00, 4.45]

Methods	Randomized controlled trial; enrollment from Aug 2003 to Jul 2004.  Sample size calculation based on assuming a BMD change of 0.014 g/cm² by 18 months; 48 women per group would provide 80% power to detect difference of 0.008 (presumably between groups).
Participants	111 women, 19 to 43 years, requesting implant for contraception. Women were a subset of a larger study by UNDP/UNFPA/World Bank/WHO.  Inclusion criteria: not pregnant or lactating within 12 months of enrollment.  Exclusion criteria: women with chronic diseases, such as diabetes mellitus, chronic renal failure, hyper/hypothyroidism, hyper/hypoparathyroidism, hepatitis, cancer or pituitary disease. Also excluded were women who used calcium or Vitamin D supplements, anticonvulsants, any corticosteroids, thiazide diuretics or drugs for thyroid disease.
Interventions	1) Single‐rod etonogestrel‐releasing implant (Implanon, NV Organon, Oss, Netherlands) (N=56) versus 2) Two silicone rod levonorgestrel‐releasing implant (Jadelle; Schering Oy, Turku, Finland) (N=55).  Insertions performed in first 5 days of cycle; no wash‐out period for those with previous contraception. [9 women had amenorrhea at insertion due to DMPA use.] Duration: 18 months
Outcomes	Bone mineral density (BMD) at midshaft of ulna (mainly cortical bone) and at distal radius (mainly trabecular bone). Measures done in non‐dominant arm with DEXA. Measures taken at baseline and 18 months.  Later report (Monteiro‐Dantas, 2007) provided 36‐month data for BMD at distal radius (also ultra‐distal radius, not available in earlier report and not used in this review).
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomization system
Allocation concealment (selection bias)	Low risk	Allocation concealment with sealed envelopes prepared at WHO
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not mentioned
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Losses: 18 months, all participants included in analysis (primary paper); 36 months (secondary paper), loss was 32% (36% Implanon and 29% Jadelle).

Methods	Randomized controlled trial in USA; recruitment from May 1996 to Jan 1999.
Participants	179 women, 18 to 33 years, who had undergone baseline bone scan as part of larger contraceptive study.  Inclusion criteria: due to funding (US Department of Defense), women had to meet criteria for entry into armed forces (high school graduate or equivalency diploma, no felony arrests, within 36% ideal body weight for height, free of medical conditions or physical disabilities that would affect completion of military training).  Exclusion criteria: currently pregnant or breastfeeding, had injection for contraception in past 6 months, took OCs in past month, or had medical contraindication to hormonal contraception.
Interventions	1) Ethinyl estradiol 35 μg plus norethindrone 1 mg OC (Ortho Novum 1/35; Ortho Pharmaceutical Corporation, New Brunswick, NJ) (N=87) versus 2) Ethinyl estradiol 30 μg plus desogestrel 150 μg OC (Mircette; Organon Corporation, Oss, Netherlands) (N=92).  Data were reported for 28 and 35 women at 12 months.
Outcomes	Bone mineral density of lumbar spine (L1 to L4) with DEXA at baseline and at 12 months. Baseline scans performed within 2 months of initiation; follow‐up scans were done 10 to 14 months after baseline.   Berenson 2004 reported on scans at 24 months for those who obtained a second scan.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random numbers table used
Allocation concealment (selection bias)	High risk	No allocation concealment, according to communication with investigator.
Blinding (performance bias and detection bias)   All outcomes	Low risk	Providers not blinded to treatment; package labeling was removed for participants. Pills were referred to as "red" or "green."
Incomplete outcome data (attrition bias)   All outcomes	High risk	Losses due to discontinuation of contraceptive method or failure to have bone scan (at all or within required window): at 12 months, 68% for norethindrone‐containing pill and 62% for desogestrel‐containing pill; at 24 months, 71% and 54%, respectively.

Methods	Randomized, crossover trial; investigators from Czech Republic. No a priori sample size calculation. Reportedly had post hoc power analysis; no detail provided.
Participants	56 adolescent females who requested hormonal contraception at adolescent gynecology unit. Inclusion criteria: age 15 to 19.5 years, BMI 20 to 27 kg/m², regular menstrual cycle. Exclusion criteria: recent or past COC use; systemic or chronic diseases including endocrine disorders; using medication that could influence bone metabolism or reliability of hormonal contraception (e.g. corticosteroids, antiepileptics, thyroid hormones); drug use or smoking > ten cigarettes per day; daily dietary calcium intake < population average (< 600 mg/day); endurance physical activity; immobilization or invalidity; COC contraindication.
Interventions	1) Gestodene 75 μg plus EE 30 μg versus 2) Gestodene 60 μg plus EE 15 μg Groups switched COCs after 9 months of use; total study duration was 18 months. No 'washout' period between study periods, reportedly for ethical reasons, i.e., high risk of pregnancy among adolescents.
Outcomes	BMD at lumbar spine L1 to L4, total proximal femur, femoral neck and distal radius, total body mineral content. Reported median and quartiles for percent changes from baseline. Serum propeptide of type I procollagen (PINP, bone formation marker).  Serum type I collagen cross‐linked C‐telopeptide (ßCTX1, bone resorption marker). Percent changes in biochemical outcomes shown in figures without absolute values.
Notes	Control group of nonusers did not want hormonal contraception (not included here). Unable to obtain further data from investigator regarding outcomes, e.g., means and SD.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random numbers (STATA software).
Allocation concealment (selection bias)	Unclear risk	no information
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	no information
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Loss by 9 months: group 1, EE 15 μg 3/28 = 10.7%; group 2, EE 30 μg 3/28 = 10.7%; Loss by 18 months (after crossover): group 1, EE 30 μg 6/28 = 21.4%; group 2, 5/28 = 17.9%; total loss 19.6%.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Percent change in bone mineral density (lumbar spine) by 12 months	1	63	Mean Difference (IV, Fixed, 95% CI)	1.83 [0.42, 3.24]
2 Percent change in bone mineral density (lumbar spine) by 24 months	1	67	Mean Difference (IV, Fixed, 95% CI)	0.84 [‐1.24, 2.92]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Change in BMD of lumbar spine by 12 months	1	277	Mean Difference (IV, Fixed, 95% CI)	‐0.00 [‐0.03, 0.02]
2 Change in BMD of lumbar spine by 24 months	1	261	Mean Difference (IV, Fixed, 95% CI)	‐0.01 [‐0.03, 0.02]
3 Change in BMD of femoral neck by 12 months	1	277	Mean Difference (IV, Fixed, 95% CI)	‐0.00 [‐0.02, 0.02]
4 Change in BMD of femoral neck by 24 months	1	261	Mean Difference (IV, Fixed, 95% CI)	‐0.01 [‐0.03, 0.01]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density (lumbar spine) at 12 months	1	39	Mean Difference (IV, Fixed, 95% CI)	‐2.10 [‐17.84, 13.64]
2 Bone mineral density (lumbar spine) at 24 months	1	39	Mean Difference (IV, Fixed, 95% CI)	‐1.60 [‐18.29, 15.09]
3 Bone mineral density (lumbar spine) at 36 months	1	39	Mean Difference (IV, Fixed, 95% CI)	‐3.53 [‐17.27, 10.21]
4 Percent change in bone mineral density (lumbar spine) by 36 months	1	39	Mean Difference (IV, Fixed, 95% CI)	0.40 [‐3.00, 3.80]
5 Percent change in serum alkaline phosphatase (36 months)	1	39	Mean Difference (IV, Fixed, 95% CI)	‐57.80 [‐160.03, 44.43]
6 Percent change in cross‐linked N‐telopeptides (36 months)	1	39	Mean Difference (IV, Fixed, 95% CI)	‐7.70 [‐31.90, 16.50]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Percent change in areal bone mineral density (lumbar spine) by 12 months	1	42	Mean Difference (IV, Fixed, 95% CI)	1.41 [‐0.11, 2.93]
2 Percent change in areal bone mineral density (femoral neck) by 12 months	1	42	Mean Difference (IV, Fixed, 95% CI)	0.08 [‐2.42, 2.58]
3 Percent change in serum bone‐specific alkaline phosphatase by 12 months	1	42	Mean Difference (IV, Fixed, 95% CI)	15.31 [3.91, 26.71]
4 Percent change in serum osteocalcin by 12 months	1	42	Mean Difference (IV, Fixed, 95% CI)	7.71 [‐2.49, 17.91]
5 Percent change in serum cross‐linked telopeptides by 12 months	1	42	Mean Difference (IV, Fixed, 95% CI)	16.39 [‐8.41, 41.19]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Change in z‐score of lumbar spine after cycle 26	1	76	Mean Difference (IV, Fixed, 95% CI)	‐0.10 [‐0.22, 0.01]
2 Change in z‐score of femoral neck after cycle 26	1	76	Mean Difference (IV, Fixed, 95% CI)	‐0.05 [‐0.16, 0.06]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Urinary pyridinoline at 12 months	1	20	Mean Difference (IV, Fixed, 95% CI)	0.20 [‐1.65, 2.05]
2 Urinary deoxypyridinoline at 12 months	1	20	Mean Difference (IV, Fixed, 95% CI)	1.20 [0.37, 2.03]
3 Serum osteocalcin at 12 months	1	20	Mean Difference (IV, Fixed, 95% CI)	1.60 [‐1.87, 5.07]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Lumbar spine BMD at 12 months	1	40	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.09, 0.09]
2 Urinary pyridinoline at 12 months	1	35	Mean Difference (IV, Fixed, 95% CI)	0.20 [‐0.65, 1.05]
3 Urinary deoxypyridinoline at 12 months	1	35	Mean Difference (IV, Fixed, 95% CI)	0.10 [‐0.24, 0.44]
4 Serum osteocalcin at 12 months	1	35	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.26, 0.26]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Percent change in bone mineral density (lumbar spine) by 12 months	1	21	Mean Difference (IV, Fixed, 95% CI)	‐0.7 [‐3.08, 1.68]
2 Percent change in bone mineral density (lumbar spine) by 24 months	1	14	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐2.74, 2.74]

Methods	Randomized controlled trial in the USA; 4 health clinics in large metropolitan setting Enrollment from May 2000 to Dec 2002. No sample size calculation provided
Participants	123 adolescent females. Inclusion criteria: adolescent girls, age 12 to 18 years, who were seeking contraception and selected DMPA.  Exclusion criteria: use of DMPA, pregnancy or abortion in past 6 months; use of OCs in past 3 months, chronic medical condition or treatment that may affect bone, or need for confidentiality in contraception.
Interventions	All participants: DMPA as 150 mg deep intramuscular injection every 12 weeks. 1) Monthly intramuscular injections of 5 mg estradiol cypionate (supplement) (N=65) versus 2) 5 mL normal saline solution (placebo) (N=58). Study duration: 24 months
Outcomes	Bone mineral density obtained at L1 to L4 lumbar vertebrae, total hip (left), femoral neck, trochanter, and Ward's triangle; DEXA was used.  Bone mineral apparent density (BMAD) was used to correct for variation in bone: BMAD = BMC(L1 to L4) / Ap^3/2, where BMC = scanned bone mineral content and Ap = projected area; BMAD = BMC(femoral neck) / Ap²(femoral neck). Outcome means were adjusted for baseline body weight and bone mineral density.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Stratified by recruitment site. 'Blocked randomization techniques.'
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Low risk	Double blinded: study subjects, technicians conducting the dual energy x‐ray absorptiometry scans, and clinicians providing health care to the participants Nurse providing injections was not blinded.
Incomplete outcome data (attrition bias)   All outcomes	High risk	The trial was stopped early due to interim analysis showing the differences between the groups reached the pre‐determined significance level of < 0.001. Estrogen supplementation was offered to participants who had the placebo; exercise and diet counseling was offered to all active participants. Many participants had not had 12 or 24 months of observation (N=24 and N=33, respectively). In addition, 30 young women withdrew by 12 months (30/123 = 24%) and 53 withdrew by 24 months (53/123 = 43%).

Methods	Randomized controlled trial in Auckland, New Zealand.  Analysis by intention to treat.
Participants	38 women recruited from family planning clinics.  Inclusion criteria: age <= 45 years, long‐term users (>= 2 years) of DMPA, areal BMD (from lumbar spine) <= 1.20 g/cm2 (young adult average).  Exclusion criteria: known metabolic bone disease, taking drugs (other than DMPA) that can affect bone density, FSH > 20 U/liter.
Interventions	All participants: DMPA 150 mg injection every 12 weeks. 1) Conjugated estrogens 62.5 μg (Premarin; Wyeth‐Ayerst, Collegeville, PA) (N=19) versus 2) Placebo (N=19), taken orally each day for 2 years. Duration: 24 months
Outcomes	Areal BMD by DEXA at lumbar spine (L2 to L4), femur, and total body; lumbar spine BMD was the primary outcome.  Results were reported for plasma calcium, phosphate, total alkaline phosphatase activity; fasting urine N‐telopeptide/creatinine ratio.
Notes	Author provided data for all BMD measures (article had graphs) as well alkaline phosphatase (AP) and N‐telopeptide NT. However, AP and NT had missing data at baseline, and the distributions appeared to be skewed, so those data were not analyzed in this review.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	According to the investigator, randomization was balanced by blocks of 10; within a block, subject was "allocated a treatment number." Random numbers above the median were assigned to treatment and those below were assigned to placebo.
Allocation concealment (selection bias)	Low risk	According to communication with the investigator, allocation concealment was accomplished with serially‐numbered opaque envelopes.
Blinding (performance bias and detection bias)   All outcomes	Low risk	Double blind
Incomplete outcome data (attrition bias)   All outcomes	High risk	27 women completed 18 months or more (29% loss to early discontinuation from study)

PERMALINK

Steroidal contraceptives: effect on bone fractures in women

Laureen M Lopez

David A Grimes

Kenneth F Schulz

Kathryn M Curtis

Mario Chen

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Background

Description of the condition

Description of the intervention

Depot medroxyprogesterone acetate (DMPA)

Oral contraceptives (OCs)

Intrauterine device (IUD) or system (IUS)

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Assessment of heterogeneity

Data synthesis

Sensitivity analysis

Results

Description of studies

Results of the search

Included studies

Risk of bias in included studies

Allocation

Blinding

Incomplete outcome data

Effects of interventions

Progestin‐only methods

Implants

1.4. Analysis.

2.2. Analysis.

2.4. Analysis.

Injectable DMPA 150 mg

3.1. Analysis.

4.1. Analysis.

4.2. Analysis.

4.3. Analysis.

4.4. Analysis.

4.5. Analysis.

4.6. Analysis.

4.7. Analysis.

4.8. Analysis.

5.1. Analysis.

5.2. Analysis.

6.6. Analysis.

Combination contraceptives

Oral contraceptives

7.1. Analysis.

8.1. Analysis.

8.4. Analysis.

9.1. Analysis.

9.4. Analysis.

10.1. Analysis.

10.3. Analysis.

10.2. Analysis.

10.4. Analysis.

Methods	Randomized controlled trial in Beijing.
Participants	61 women, aged 25 to 40 years, who sought contraception counseling at Ob‐Gyn Hospital.  Inclusion criteria: regularly menstruating, no OC use in past 3 months or Norplant removed >=6 months prior, no medications or diseases known to interfere with bone metabolism.
Interventions	1) Norplant implant (Leiras Pharmaceuticals, Turku, Finland) (N=30) versus 2) Domestic implant (Capsulae Levonorgestreli Silasticus; Yalujiang Medical Factory, Dandong, Laioning Province, China) (N=31). Both types had 6 capsules, each with 36 mg levonorgestrel for a maximum 216 mg levonorgestrel. Study duration 1 year
Outcomes	BMD of lumbar spine (L2 to L4) and proximal femur via DEXA.  Serum osteocalcin (BGP) and alkaline phosphatase (ALP); urine hydroxyproline/creatinine and urine calcium/creatinine.
Notes	Attempts to reach the investigator were unsuccessful
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	randomly divided into two groups
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	No information
Incomplete outcome data (attrition bias)   All outcomes	Low risk	One implant removed at subject's request and data excluded from analysis.

Methods	Randomized controlled trial in Germany  Analysis was per protocol.
Participants	100 healthy women at one center.  Inclusion criteria: current users of COCs for > 2 years, but no EE >= 50 μg for more than 6 months; started COCs after 16th birthday.  Exclusion criteria: contraindications for COC use; smoking >15 cigarettes for women < 30 years and any smoking for women > 30 years; use of DMPA in previous 6 months, use of other sex hormones during treatment, coexisting diseases (unspecified), diagnostically unclassified genital bleeding, and history of migraine with menstruation.
Interventions	1) 20 μg ethinyl estradiol (EE) plus 100 μg levonorgestrel OC (20/100; Schering AG, Berlin, Germany) versus 2) 30 μg ethinyl estradiol with 150 μg levonorgestrel OC (30/150; Schering AG). First tablet taken on first day of withdrawal bleeding with no wash‐out period. 36 consecutive 28‐day cycles. Study duration was 3 years
Outcomes	Trabecular BMD of lumbar spine (L1 to L3) using computed tomography;  alkaline phosphatase and N‐telopeptides.
Notes	Attempts to reach the investigator were unsuccessful
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Low risk	Double blind
Incomplete outcome data (attrition bias)   All outcomes	High risk	88 took at least one dose; authors' analysis only included the 48 who provided data at baseline and cycle 36 (52% loss). Loss to follow up was 15% overall (15/100). Bone density data provided for the 39 with data at all time points (61% loss).

Methods	Randomized trial; unspecified location (protocols approved in Shandong, China). No sample size calculation provided.
Participants	300 women from 16 to 18 years old, attending family planning clinics and requesting birth control. Inclusion criteria: regular menses, non‐hormonal contraception, no breastfeeding or delivery for at least 6 months, no pregnancy; did not take calcium, vitamin D, or bone‐affecting medication. Exclusion criteria: chronic disease, such as diabetes mellitus, renal dysfunction, thyroid and parathyroid diseases, hepatitis or pituitary diseases.
Interventions	1) Desogestrel (DSG) 150 μg plus EE 30 μg versus 2) Cyproterone acetate (CPA) 2 mg plus EE 35 μg Duration: 2 years
Outcomes	BMD measured at lumbar spine (L 2 to L4) and femoral neck by DEXA.
Notes	A third group was not randomized; women did not want to use hormonal method.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	By 'drawing lots'
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	No information
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Loss by 12 months: DSG+EE 12/150 = 8%; CPA+EE 11/150 = 7.3%. Loss by 24 months: DSG+EE 23/150 = 15.3%; CPA+EE 16/150 = 10.7%.

Methods	Randomized trial in Italy No sample size calculation provided.
Participants	44 women, 21 to 34 years old, presenting at clinic. Inclusion criteria: age of menarche 12 to 14 years, ovulation during pretreatment cycle, BMI > unclear (report had error in lower cutoff of '2') < 25, normal menstrual cycles, normal diet without high or low caloric intake.  Exclusion criteria: confirmed or suspected pregnancy, pregnancy or breastfeeding in past year, liver disease, vascular or metabolic disorder, disorder of bone metabolism, treatment with drugs that affect bone metabolism or drugs that interfere with contraceptive steroids, other contraindication for COCs.
Interventions	1) Drospirenone 3 mg plus EE 30 μg versus 2) Drospirenone 3 mg plus EE 20 μg Duration of 12 months
Outcomes	BMD of lumbar spine (L1 to L4) by DEXA.  Bone turnover markers: serum and urinary calcium, urinary pyridinoline and deoxypyridinoline (resorption); serum osteocalcin (formation). Results were presented in figures, without absolute values, for urinary pyridinoline and deoxypyridinoline and for serum osteocalcin.
Notes	'Controls' were not randomized, and therefore were excluded from analysis in this review; they did not request contraception. Unable to obtain from investigator further information on methodology or additional data.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	No information
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Losses: 7% (3/44) included 1 excluded for protocol violation (EE 30 μg group) and 2 early discontinuations (EE 20 μg group).

Methods	Randomized trial in Munich, Germany No mention of sample size calculation or power.
Participants	52 women, 18 to 24 years old, recruited via advertisements and mass mailings.  Inclusion criteria: discontinued OCs for at least 2 months; spontaneous menstrual cycle during run‐in period.  Exclusion criteria: BMI > 30, history of smoking (>20 cigarettes/day), alcohol consumption (>20 g/day), exercise > 1 hour/week in past, age of menarche > 15 years, long‐term use of OC (> 50% of time since menarche), previous menstrual disorders, past or current pregnancy, hypertension (> 140/90 mmHg), abnormal values for clinical chemistry, current or past medication with drugs that influence bone metabolism, use of contraceptive implants in past 6 months, diseases that affect bone (except well‐controlled Type I diabetes mellitus or well‐controlled hypothyroidism)
Interventions	1) Levonorgestrel 100 μg plus EE 20 μg (N=24) versus 2) Desogestrel 150 μg plus EE 20 μg (N=28); Duration of 13 cycles
Outcomes	BMD (spine L2 to L4 and femoral neck) by DXA. Serum bone‐specific alkaline phosphatase; intact osteocalcin; serum C‐terminal cross‐linked telopeptides.
Notes	'Controls' were not randomized, and therefore were excluded from the analysis in this review; they did not choose hormonal contraception.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	High risk	Open
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Losses due to discontinuation (e.g., adverse events, personal reasons): 19%; desogestrel group, 21%; levonorgestrel group, 17%.

Methods	Randomized controlled trial at 36 sites in USA, 9 sites in Canada, and 3 sites in Brazil; conducted from 2001 to 2004. Sample size of 250 for each group was based on 80% power to detect 2% difference in bone mineral density at 2 years.
Participants	535 women aged 18 to 35 years, sexually active, desiring long‐term contraception. Inclusion criteria: regular menstruation in past 3 months, negative urine pregnancy test, and willingness to rely upon DMPA‐SC or DMPA‐IM for contraception for at least 2 years. Exclusion criteria: use of oral contraceptives, implant, or hormonal IUD in past 2 months or DMPA‐IM in past 10 months (contraceptive patches and rings were not in use at the time of study enrollment); lumbar spine or femur BMD T‐score of less than −1.0, history of pathologic or compression fracture; abnormal cervical cytology; undiagnosed abnormal genital bleeding; known or suspected pregnancy; history of breast cancer, thrombotic event, hepatic or renal disease, alcoholism or other drug abuse (in past 5 years); uncontrolled hypertension, active hepatic or renal disease, type 1 diabetes, or poorly controlled type 2 diabetes; and taking anticancer agent aminoglutethimide. Calcium, multivitamins, and other mineral supplements were not required nor prohibited if part of the participants' normal daily regimen.
Interventions	1) DMPA 104 mg/0.65 mL by subcutaneous injection every 3 months (DMPA‐SC) versus 2) DMPA 150 mg/mL by intramuscular injection every 3 months (DMPA‐IM) Duration: 2 years; secondary objectives included assessing efficacy, safety, and tolerability over 3 years
Outcomes	Primary: percent change (baseline to 2 years) in BMD at total hip and lumbar spine; dual X‐ray absorptiometry used. Efficacy endpoint was pregnancy (urine test). Adverse events recorded throughout study.
Notes	Study was sponsored by Pfizer. Three investigators were not compensated, investigators reportedly retained full editorial control over content of paper; other 2 were Pfizer employees. Report did not specify where the analysis was conducted or by whom.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Interactive voice‐response system based on computer‐generated random list
Allocation concealment (selection bias)	Low risk	Interactive voice‐response system
Blinding (performance bias and detection bias)   All outcomes	Low risk	Investigators and evaluators at each site were blinded. Independent person received the study syringes, maintained the randomization code, and administered the study drug.
Incomplete outcome data (attrition bias)   All outcomes	High risk	Participant flow chart provided. Overall losses by 1 year: 39% (38% DMPA‐SC, 40% DMPA‐IM). Loss to follow up by 2 years (1 year not available): 19% overall; 16% DMPA‐SC, 22% DMPA‐IM Efficacy analysis included all participants who received at least one dose of study drug and made at least one visit after receiving study drug.

Methods	Randomized controlled trial at university‐based family planning clinic in Italy during 2008. No mention of sample size calculation.
Participants	40 healthy women, 23 to 34 years old, requesting contraception. Inclusion criteria: age of menarche 12 to 14 years, demonstrable ovulation during pretreatment cycle, body mass index (BMI) > 20 and < 22 kg/m², normal menstrual cycles and normal diet. Exclusion criteria: confirmed or suspected pregnancy, pregnancy or breastfeeding in past year, liver disease, vascular or metabolic disorder, disorder of bone metabolism (Paget disease, hyperparathyroidism, renal osteodystrophy) and treatment with drugs that affect bone metabolism (bisphosphonates, sodium fluoride, calcitonin, estroprogestins or anabolic steroids, corticosteroids, calcium or vitamin D, phosphate, thiazide diuretics) or drugs that interfere with contraceptive steroids (barbiturates, antiepileptics, rifampicin, griseofulvin), contraindications for the use of hormonal contraceptives
Interventions	1) Contraceptive patch delivering norelgestromin 150 μg plus EE 20 μg daily versus 2) Vaginal ring releasing etonogestrel 120 μg plus EE 15 μg daily Duration: 12 months
Outcomes	Bone mineral density by DEXA of lumbar spine (L1 to L4) Bone formation: serum osteocalcin; bone resorption: urinary pyridinoline and deoxypyridinoline.
Notes	Results for bone formation and resorption were presented in figures. On request, the investigator provided means, standard deviations, and Ns for analysis. Study was supported by institutional funds from investigators' university department
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated list
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Low risk	Blood and urine samples analyzed in laboratory blinded to treatment group; absorptiometry performed by blinded assessor
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Participant flow chart provided. Losses due to drop‐outs were 20% in each group; reasons for discontinuation were personal, irregular bleeding, and reaction to patch. No losses to follow up mentioned.

Methods	Randomized trial in Uppsala, Sweden.
Participants	22 women, 20 to 45 years, seeking contraceptive advice at family planning clinic, University Hospital.
Interventions	1) Levonorgestrel implant (Norplant) versus 2) DMPA 150 mg every 12 weeks Study duration: 6 months
Outcomes	Bone density in distal forearm by single photon absorptiometry. Bone density results shown in figure; no absolute numbers for variance. Serum calcium, alkaline phosphatase, and osteocalcin; urinary hydroxyproline/creatinine.
Notes	Investigator was unable to retrieve the BMD means and SD; publication was > 10 years old.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	According to the investigator, randomization procedure was a 'computer‐based randomization.'
Allocation concealment (selection bias)	Low risk	Sealed envelopes were opened after inclusion of each subject.
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	No information; interventions differed (injection versus implant)
Incomplete outcome data (attrition bias)   All outcomes	Low risk	19/22 women completed the 6‐month follow up; bone density measured in 18 (18% loss).

Methods	Randomized trial
Participants	40 healthy women desiring contraception.  Inclusion criteria: 22 to 34 years old, age of menarche 12 to 14 years, ovulation in pretreatment cycle, BMI > 20 and < 22, normal menstrual cycles, normal diet without high or low calories.  Exclusion criteria: pregnancy or breastfeeding in past year, liver disease, vascular or metabolic disorder, bone disease or disorder, smoking >10 cigarettes/day, migraine with aura, drugs that affect bone metabolism or with steroidal contraceptives, hysterectomy or oophorectomy, other contraindications for COCs.
Interventions	1) Ethinyl estradiol 20 μg plus gestodene 75 μg OC (FEDRA, Schering; Milan, Italy) (N=20) versus 2) Ethinyl estradiol 15 μg plus gestodene 60 μg OC (ARIANNA, Schering; Milan, Italy) (N=20) Study duration 1 year
Outcomes	BMD by DEXA at posterior‐anterior lumbar spine (L1 to L4)  Serum osteocalcin, urinary pyridinoline and deoxypyridinoline
Notes	Results shown in figures without absolute values. Unable to obtain further information from the investigator.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated list
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	No information
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Loss: 8%; 3/40 women dropped out or missed the follow‐up visit

Methods	Randomized controlled trial in Naples, Italy
Participants	48 healthy women in family planning clinic.  Inclusion criteria: 22 to 34 years old, age of menarche 12 to 14 years, ovulation in pretreatment cycle, normal menstrual cycles, no abnormal dietary requirements.  Exclusion criteria: pregnancy or breastfeeding in past year, liver disease, vascular or metabolic disorder, bone disease or disorder, smoke >= 10 cigarettes/day, migraine with aura, drugs that affect bone metabolism, drugs that interfere with steroidal contraceptives, hysterectomy or oophorectomy, other contraindications for COCs.
Interventions	1) Ethinyl estradiol 30 μg and drospirenone 3 mg OC (Yasmin; Schering, Milan, Italy) (N=24) versus 2) Ethinyl estradiol 30 μg and gestodene 75 μg OC (Ginoden; Schering) (N=24) Study duration 1 year
Outcomes	BMD by DEXA at posterior‐anterior lumbar spine (L1 to L4).  Serum and urinary calcium, serum osteocalcin, urinary pyridinoline and deoxypyridinoline. Results for biochemical measures shown in figures without absolute numbers.
Notes	Unable to obtain further information from investigator.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomization sequence
Allocation concealment (selection bias)	Low risk	Sequence was 'concealed both to researchers and patients until treatments were assigned.'
Blinding (performance bias and detection bias)   All outcomes	High risk	No blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Loss: 6%; 3/48 women dropped out.