Description of the condition
Postmenopausal osteoporosis is a chronic metabolic disease related to the alternation of hormones, especially estrogen, during the perimenopausal or postmenopausal (or both) period in women. It is characterized by deterioration of bone mineral density (BMD), trabecula (reticular structure that forms the interior bone filled with bone marrow), cortical bone (compact stratum forming the outer shell of most bone), and extension of the medullary cavity (central cavity storing bone marrow). BMD is measured by dual-energy X-ray absorptiometry bone densitometer (DXA). For adult women, osteoporosis is defined as a value of BMD 2.5 standard deviations below the mean of young normal women (T-score ≤ -2.5) according to the World Health Organization (WHO) diagnostic criteria (Kanis 1994). The decreased BMD and microarchitectural changes will lead to fragility fracture, which is caused by mild to moderate force.
Postmenopausal osteoporosis-related fractures predominantly occur in vertebral, upper femur, or distal radius, and the prevalence of fractures positively correlates with menopausal years (Looker 1998; NAMS 2010). As the most serious complication of postmenopausal osteoporosis, fractures affect at least one-third of women aged 50 years and older. Besides, it has been estimated that lifetime risk for osteoporotic fracture is higher in women aged above 50 years than age-matched men (54% versus 5% to 6%) (Grossman 2001; Johnell 2005). Postmenopausal osteoporosis is a universal disease and subsequently results in heavy economic burden (Cummings 2002; Lewiecki 2004). Furthermore, the increase in life expectancy and elderly population may make the plight more serious. Therefore, the prevention and treatment of postmenopausal osteoporosis is urgent for the whole of society. Nonpharmacologic therapies of postmenopausal osteoporosis include optimizing nutrition, lifestyle changes such as doing more exercise, smoking cessation, avoiding excessive caffeine, and reducing alcohol consumption. Pharmacologic therapies include antiresorptive agents such as bisphosphonates, calcitonin, and hormone therapy, and anabolic drugs such as parathyroid hormone. The benefits of vitamin D with calcium for fracture prevention in postmenopausal osteoporosis have been demonstrated by Cochrane systematic reviews (Avenell 2009), and with bisphosphonates including alendronate, etidronate, and risedronate (Wells 2008a; Wells 2008b; Wells 2008c). Although fluoride increases BMD of lumbar spine, it does not have beneficial effects on reducing both vertebral and nonvertebral fractures (Haguenauer 2010).
Description of the intervention
Selective estrogen receptor modulators (SERMs) belong to a class of compounds that lack the steroid structure of estrogen, but have an alternative structure that binds to estrogen receptors (ER) (Riggs 2003). They have both estrogenic and antiestrogenic effects depending on the target tissue or gene. SERMs, which have been used clinically for several decades, are widely prescribed for estrogen-related diseases, such as breast cancer, ovulation induction, and hypogonadotropic hypogonadism (Shelly 2008a). Among pharmacologic therapies for postmenopausal osteoporosis, SERMs belong to antiresorptive agents, which focus on the relationship between estrogen efficiency and bone loss. Unlike hormone therapy, SERMs may avoid some side effects due to its selective effect (estrogenic and antiestrogenic effect) on different organs (Barrett-Connor 2002; Cummings 1999; Lippman 2006). As a large family with structurally various members, these compounds can be mainly classified into triphenylethylenes, benzothiophenes, naphthalenes, indoles, benzopyrans, and flavonoids.
Although triphenylethylenes were not originally designed for treatment of postmenopausal osteoporosis, clinical trials have demonstrated that they exert substantial positive effects on BMD in long-term management except for toremifene (Chi 2012; Moskovic 2012; Santen 2011). In contrast, administration of clomiphene and tamoxifen is limited by side effects such as abnormal vaginal bleeding, hot flashes, abdominal discomfort, headache, nausea, vomiting, or a combination of these. Furthermore, tamoxifen interacts with warfarin, rifampicin (rifampin), cholestyramine, and aromatase inhibitors (Morello 2003).
Raloxifene is the most well-known available benzothiophene and it is considered as second-line treatment for postmenopausal osteoporosis in most clinical guidelines (Hodgson 2003). Based on several clinical trials, especially the Multiple Outcomes of Raloxifene Evaluation (MORE) trial, raloxifene has a productive effect on bone loss and consequent osteoporosis-related fractures, and may also be beneficial for coronary heart disease (Barrett-Connor 2002; Kanis 2003; Siris 2002). However, its usage is limited by increased risk of venous thromboembolism for the first two years (Grady 2004). Besides, cholestyramine causes a 60% reduction of absorption and enterohepatic circulation of raloxifene, thus, cholestyramine or other anion exchange resins should not be coadministered with raloxifene (Morello 2003).
Like other SERMs, lasofoxifene (the naphthalene class) and available bazedoxifene (the indole class) are well absorbed orally and highly bound to plasma proteins. These two drugs are both approved for the prevention and treatment of postmenopausal osteoporosis.
The clinically available benzopyran ormeloxifene can be used as a weekly oral contraceptive agent. It is also effective in mastalgia and dysfunctional uterine bleeding, and shows antiresorptive activity in bone metabolism in vitro. Unfortunately, reported cases of urinary incontinence and uterovaginal prolapse limited its further development for chronic usage (Singh 2001).
Femarelle is a natural compound derived from soybean. It has been used for the treatment of menopausal symptoms and postmenopausal osteoporosis with no effect on thromboembolism and endometrial thickness (Nachtigall 2011; Yoles 2004).
Most of the above drugs have long serum half-lives (120 to 168 hours) except raloxifene (16 to 87 hours), and their doses ranges from 20 to 60 mg/day except for femarelle (644 mg/day) (Morello 2003; Yoles 2004). When considering SERMs prescription, liver function needs to be taken into account as most of them are metabolized in the liver and eliminated in the bile. Currently, new agents are under development for better efficacy, higher oral bioavailability, and fewer adverse effects/events.
How the intervention might work
Estrogen, a steroid hormone that dramatically fluctuates in the perimenopausal phase and decreases in the postmenopausal phase of women, has essential impact in regulating bone resorption and formation. Through body-widespread ERs, a type of a nuclear hormone receptor that belongs to the family of activated transcription factors (Bord 2001), estrogen can interact with special sequences of deoxyribonucleic acid (DNA) (Kumar 1988). In vitro studies have shown that estrogen could improve proliferation and early osteoblastic differentiation of mesenchymal stem cells (MSCs) (Hong 2011; Okazaki 2002), which have the ability of osteogenic differentiation, and can also act directly on osteoblasts by regulating bone nodule formation and matrix protein/cytokine production (Waters 2001). Different types of ER (α and β) may mediate different functions on different tissues, which forms the theoretical basis of developing SERMs. ER-α may play a more important role in increasing the expression of alkaline phosphatase and osteocalcin on the bone formation compared with ER-β (Bodine 1998). With high affinity for ERs, SERMs can alter receptor conformation, facilitate binding of transcription-related proteins of estrogen target genes, which further leads to tissue-dependent estrogen agonist responses in some target tissues such as bone, or antagonistic effects on uterine endometrium and breast tissue (Bryant 1999; Ke 1995; Lippman 2006). To date, clinically available SERMs have been used in many estrogen-related diseases including postmenopausal osteoporosis (Shelly 2008). For example, raloxifene shows high affinity for ERs on bone without significant binding to other major steroid hormone receptors. This will produce protective effect on the skeleton without stimulating an effect on the uterus and breast. Thus, theoretically, there are fewer concerns about any estrogen-like effect on endometrium or mammary tissue during raloxifene treatment (Bryant 1999).
Why it is important to do this review
Many published randomized controlled trials (RCTs) using SERMs showed beneficial effects on postmenopausal osteoporosis. The evidence supports the efficacy of raloxifene or bazedoxifene for the prevention of osteoporotic fractures. However, the results remain imprecise due to the small number of fracture events (Levis 2012; Murad 2012). In addition, published systematic reviews did not include other SERMs such as lasofoxifene, ormeloxifene, and femarelle. As clinicians do not have a global picture of all SERMs, which may further influence decision making, it is necessary to evaluate the efficacy and safety of all those drugs for the prevention and treatment of postmenopausal osteoporosis.
To estimate the clinical benefits and harms of all SERMs for postmenopausal osteoporosis in both prevention (women without postmenopausal osteoporosis) and treatment (postmenopausal osteoporosis with or without clinical history of previous osteoporotic fractures).
The principle objective of this review is based on the following questions:
- Is the use of SERMs effective and safe for the prevention and treatment of postmenopausal osteoporosis?
- Are SERMs more effective and safer than other approved drugs for prevention and treatment of postmenopausal osteoporosis?
- Which type of SERMs is the most effective or safest?
Criteria for considering studies for this review
Types of studies
We will only include RCTs in this review. We will exclude trials with inappropriate randomization strategies (e.g. participants were allocated by alternation, clinic record number, or date of birth) to avoid selection bias.
Types of participants
We will include postmenopausal women, surgically or naturally, with the following condition:
- those without osteoporosis;
- those with osteoporosis but without previous osteoporotic fractures;
- those with clinical history of previous osteoporotic fractures.
For naturally postmenopausal women, they should not be over 40 years of age and the duration of amenorrhea should be no less than 12 months. We will include women with history of fractures as long as the fractures had healed and were not malignant or pathologic fractures. We will exclude any women who have diseases of the thyroid, parathyroid, gastrointestinal tract, or liver, or take long-term medications such as corticosteroids, antiepileptic drugs, anticonvulsants, immunosuppressive agents for the possibility of secondary change on BMD. BMD will be measured by DXA and osteoporosis will be defined based on WHO criteria (T-score -2.5 or less) (Kanis 1994a). If the inclusion criteria were not detailed in the original paper, we will contact study authors for the information and perform separate analyses according to the three basic groups mentioned above (without osteoporosis, with osteoporosis without previous fractures, with osteoporosis and previous osteoporotic fractures). Cases will be excluded if available information for classification of above groups cannot be acquired.
Types of interventions
SERMs alone in any dosage and any form, or SERMs combined with calcium, vitamin D, or calcium plus vitamin D for prevention and treatment of postmenopausal osteoporosis.
The acceptable comparators may include:
- bisphosphonates (e.g. alendronate, etidronate);
- hormone therapy (estrogens, e.g. conjugated estrogen, 17b-estradiol; estrogen plus progesterone);
- parathyroid hormone;
If calcium, vitamin D, or calcium plus vitamin D are used as supplement therapy, this administration should be given to both intervention and comparison groups. The duration of intervention should be no less than 3 months.
Types of outcome measures
- Number of incident fractures with clinical significance (vertebrae, hip, and wrist).
- Quality of life.
- Total number of adverse effects.
- Number of withdrawals due to adverse events.
- Number of serious adverse events (death, life-threatening experience, hospitalization, disability, or incapacity).
- The total number of withdrawals.
- Mean change in BMD (lumbar spine, total hip, and femoral neck) measured by DXA.
We will include all trials that report at least one of the major or minor outcomes, and extracted all time points of measured outcomes in included studies. We will report all outcomes in the included trials, and at all time points reported. If the information of outcome measures of interest is insufficient, we will contact the corresponding authors of original article for further details.
Search methods for identification of studies
We will search the following electronic databases regardless of language or publications status:
- the Cochrane Central Register of Controlled Trials (CENTRAL) (current issue);
- MEDLINE (on Ovid, 1966 to present);
- EMBASE (1974 to present);
- Chinese Biomedical Literature Database on web (CBMweb, 1978 to present);
- VIP database (1989 to present);
- Wanfang Digital Periodicals of the Chinese Medical Association (CMA) database (1977 to present);
- biomedical content of Chinese National Knowledge Infrastructure (CNKI, 1979 to present), which includes China Academic Journal Network Publishing Database (CAJD), China Doctoral Dissertations Full-text Database (CDFD), China Master's Theses Full-text Database (CMFD), China Proceedings of Conference Full-text Database (CPCD), China Core Newspapers Full-text Database (CCND), Chinese Book Full-text Database (CBFD), and China Yearbook Full-text Database (CYBD).
Medical Subject Headings (MeSH) terms including osteoporosis, menopause, and selective estrogen receptor modulators will be adapted. Free-text words will also be introduced to complement the search strategy (Appendix 1).
Searching other resources
We will search for ongoing and registered trials in:
- Current Controlled Trials (www.controlled-trials.com/), 2004 to present;
- ClinicalTrials.gov, a service of the US national Institutes of Health (clinicaltrials.gov/ct2/home), 1997 to present;
- The WHO International Trials Registry Platform search portal (www.who.int/trialsearch/Default.aspx), 2007 to present;
- The Chinese Clinical Trial Register (www.chictr.org/), 2005 to present.
We will search for abstracts from the following databases of conferences:
- Gateway (gateway.nlm.nih.gov/gw/Cmd), 1993 to present;
- Medscape (www.medscape.com/resource/osteoporosis), 1994 to present;
- American Society for Bone and Mineral Research (www.asbmr.org/Default.aspx), 1974 to present;
- World Congress on Controversies, Debates & Consensus in Bone, Muscle & Joint Diseases (www.congressmed.com/bmjd/), 2012 to present;
- International Menopause Society (www.imsociety.org/), 2001 to present;
- The International Society of Gynecological Endocrinology (www.gynecologicalendocrinology.org/), 1998 to present;
- European Society of Gynecology (seg-web.org/), 2004 to present.
We will search for unpublished data from grey literature through:
- Centers for Disease Control and Prevention (www.cdc.gov);
- Health Canada (hc-sc.gc.ca);
- contacting experts, pharmaceutical companies, and authors of included studies.
We will also manually search the reference lists of all the RCTs and reviews obtained for eligible studies. Trials will be selected only if they meet all the inclusion criteria listed above.
Data collection and analysis
Selection of studies
WL and JZ will assess the eligibility of studies independently according to the criteria mentioned above irrespective of journal, authors, institution, and the magnitude and direction of the results. These two review authors will first examine each title from search results, and then identify potentially eligible studies through abstracts. Full text will be obtained for further eligibility evaluation of articles screened by abstract.
Data extraction and management
Two review authors (WL and JZ) will individually extract all valuable information from eligible studies by using a data collection form designed by the methodologist (TW) in our group based on the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Extracted items will contain general information of articles (title, author, author affiliation, institution, and source of this article); methods of study design (study design, unit of allocation, unit of analysis, power calculation, method of randomization, allocation concealment, and blinding); baseline of population (place, inclusion criteria, exclusion criteria, age, sex, ethnicity, work status, diagnosis of disease, number of participants recruited, number of participants who met inclusion criteria, number of participants randomized and number of participants followed up); details of interventions (route of administration, dose, duration, and control groups); outcomes (who carried out the measurement, what was measured at baseline and at follow-up, how was it measured, and how the tool was validated); analysis (statistical technique, intention-to-treat analysis or not, adjustment technique for confounding or not, number or percentage of followed up from each group); and outcome measures as described in 'Types of outcome measures' and 'Measures of treatment effect'. We will contact the corresponding authors of original articles to request missing details to identify eligibility and qualification of studies further. We will resolve any discrepancy in study selection or data extraction by discussion or arbitrated by a third review author (LX) if disagreements persist.
Assessment of risk of bias in included studies
Two review authors (WL and LX) will independently carry out an assessment of methodologic quality using The Cochrane Collaboration's 'Risk of bias tool'. Criteria for risk of bias assessment include selection bias, performance bias, detection bias, attrition bias, reporting bias, and other potential sources of bias (Higgins 2011). Each feature of the trials will be explicitly classified as 'low risk of bias', 'high risk of bias', or 'unclear risk of bias' (either lack of information or uncertainty over the potential for bias).
We will consider co-intervention and baseline imbalance as other potential sources of bias.
- low risk: details of co-intervention or baseline were equal in both intervention and comparison groups in the trial, no co-intervention in the trial (type, administration, or doses for co-intervention; height, weight, age, history of fractures for baseline);
- high risk: co-intervention or baseline was different between intervention and comparison groups;
- unclear: details of co-intervention or baseline was not described precisely and could not acquire details from the author of original trial.
The assessment results of 'Risk of bias' will be listed in summary figures. TW and JZ will check out the assessment and TW will be consulted to reach a consensus if any disagreements exist.
Measures of treatment effect
For dichotomous variables, we will extract the number of events and total sample size in each group to calculate the risk ratio (RR). For continuous data, we will extract means and standard deviations (SD) to calculate mean difference (MD). We will calculate standardized mean differences (SMDs) when outcomes are measured by different scales.
Unit of analysis issues
Unit of analysis issues such as cluster-randomized trial, cross-over designed RCTs, multiple intervention groups, and repeated measurement will be taken into account. The unit of analysis is individual rather than groups of individuals. If any cross-over designed RCTs are eligible for inclusion, we will extract and analyze data using the generic inverse-variance method following the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will perform meta-analysis of the results separate from other study designs. Repeated measurement will be dealt with in separate meta-analyses according to the different time points of measurement.
Dealing with missing data
We will contact authors of original trials to obtain relevant missing details. We will perform intention-to-treat analyses, including all subjects who were originally randomly assigned, if possible. Furthermore, we will calculate percentage of loss to follow-up and discuss the potential influence of missing data.
Assessment of heterogeneity
Included original studies may contain many types of heterogeneity identified as clinical heterogeneity (e.g. diversity in baseline characteristics of participants, interventions, outcomes), methodologic heterogeneity (e.g. variability in study design, risk of bias), and statistical heterogeneity, consequence caused by clinical or methodologic diversity. We will used the I
Assessment of reporting biases
We will investigate reporting bias from different aspects (e.g. publication, language, and outcome reporting). We will use funnel plots to evaluate potential publication bias as long as there are sufficient numbers of eligible RCTs (Egger 1997). We will report the possibility for selective reporting of outcomes if the outcome is found to be asymmetric. As funnel plots and publication bias are not necessarily related, we will not place excessive emphasis on it (Lau 2006).
We will perform meta-analyses under the guidance of The Cochrane Collaboration using Review Manager 5 (RevMan 2012). Where numbers or SD are not provided in included trials, we will calculate these data from confidence intervals (CI), standard errors (SE), exact P values, or other reported statistics. We will express the precision of both dichotomous and continuous data using corresponding 95% CI. We will consider the following interventions and comparators as clinically appropriate to combine in a meta-analysis. SERMs versus comparisons (placebo, calcium, vitamin D, calcium plus vitamin D, other SERMs, bisphosphonates, fluoride, parathyroid hormone, calcitonin, hormone therapy). For example, we will analyze comparisons of SERMs versus placebo and SERMs versus bisphosphonates separately. We will pool appropriate data into meta-analyses using a random-effects model.
Subgroup analysis and investigation of heterogeneity
We will perform subgroup analysis for outcomes based on clinic heterogeneity if we obtain enough data. We may carry out subgroup analysis based on the following:
- type of SERMs;
- co-interventions: comparison drugs with or without calcium or vitamin D or both;
- dosage: low or high dosage of SERMs;
- duration of studies: duration two years or less versus duration longer than two years;
- primary versus secondary prevention: comparison drugs in prevention versus treatment.
In order to determine whether the conclusions are robust for decision making, we will compare analyses including all studies with analyses restricted to studies at low risk of bias, to assess how results are affected by high or unclear risk of bias. In addition, we will also compare the fixed-effect model with the random-effects model. If imputed SDs are used in a meta-analysis, we will also compare all including trials with those excluding imputed statistical data.
'Summary of findings' table
To improve the accessibility of the review, we will complete a 'Summary of findings' table to present key information. The table will include all grading of the evidence according to the Grading of Recommendation Assessment Development and Evaluation (GRADE) approach (Schünemann 2011).
The 'Summary of findings' table will include the following outcomes:
- number of incident fractures - vertebral;
- number of incident fractures - hip;
- number of incident fractures - wrist;
- quality of life;
- total adverse events;
- withdrawals due to adverse events;
- serious adverse events.
Grading of the quality of evidence
We will use the GRADE approach to rank the strength of the overall quality and applicability of included studies according to the recommendation of the Cochrane Musculoskeletal Group (CMSG). We will grade the quality of evidence into four categories by using GRADEpro: high (further research is very unlikely to change our confidence in the estimate of effect), moderate (further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate), low (further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate), and very low (any estimate of effect is very uncertain) (Atkins 2004).
Many thanks to Shan Dan (Ph.D. Department of Obstetrics and Gynecology of the Second West China Hospital of Sichuan University) for giving advice on the revision of the protocol. We thank the editorial team of the Cochrane Musculoskeletal Group for help on preparing this protocol. We also thank the National Natural Science Foundation of China (Dr. Liangzhi Xu, Principal Investigator, Project No. 81070464) and Research Fund for the Doctoral Program of Higher Education, China (Dr. Liangzhi Xu, Principal Investigator, Project No. 20100181110007) for database support.
Appendix 1. Search strategy
We will use the following search strategy to search MEDLINE Ovid (1946 to present) and adapted for other databases:
4 bone density
5 bone densit$.mp
6 exp "bone and bones"
7 bone loss$.mp
8 bone mass$.mp
9 bone mineral densit$.mp
10 bone mineral content$.mp
11 bone age.mp
12 bone defect$.mp
13 bone deminerali?ation.mp
14 bone mineral$.mp
15 bone strength.mp
17 deminerali?ed bone.mp
19 exp Selective Estrogen Receptor Modulators
22 exp Clomiphene
25 exp Raloxifene
26 exp Tamoxifen
27 exp Toremifene
28 Tetrahydronaphthalenes/ or lasofoxifene.mp
30 Thiophenes/ or arzoxifene.mp
40 exp Estrogen Replacement Therapy
41 exp Hormone Replacement Therapy
42 or/19-41 (54687)
43 randomized controlled trial.pt
44 controlled clinical trial.pt
47 clinical trials as topic.sh
51 exp animals/ not humans.sh
52 50 not 51
53 18 and 42 and 52
Contributions of authors
Drafted the protocol: Wenjuan Li.
Revised the protocol: Liangzhi Xu, Xin Pan.
Search for relevant trials: Wenjuan Li, Jing Zhang.
Obtain copies of trials: Wenjuan Li, Liulin Tang.
Selection of trials: Wenjuan Li, Jing Zhang, Liangzhi Xu.
Extract data from trials: Wenjuan Li, Jing Zhang.
Enter data into RevMan: Xin Pan.
Carry out the analyses: Wenjuan Li, Wu Taixiang, Jing Zhang, Liangzhi Xu.
Interpret the analyses: Liangzhi Xu, Wu Taixiang.
Draft the final review: Wenjuan Li with contributions from all.
Update the review: Wenjuan Li, Jing Zhang, Liangzhi Xu.
Declarations of interest
Sources of support
- West China Second University Hospital, Sichuan University, China.Electronic database and specialist support
- Chinese Cochrane Center, West China Hospital, Sichuan University, China.Specialist support and evidence-based medicine (EBM) training
- No sources of support supplied