Systematic review of the use of bone turnover markers for monitoring the response to osteoporosis treatment: the secondary prevention of fractures, and primary prevention of fractures in high-risk groups

Osteoporosis

Osteoporosis is a progressive systemic skeletal disease characterised by low bone mass and micro-architectural deterioration of bone tissue, with a consequent increase in bone fragility and susceptibility to fracture. 1,2

Bone turnover (remodelling)

Bone turnover is the process of resorption followed by replacement by new bone with little change in shape, and it occurs throughout a person’s life. Osteoclasts break down bone (bone resorption), releasing the minerals, resulting in a transfer of calcium from bone fluid to the blood. The osteoclast attaches to the osteon (layers of compact bone tissue surrounding a central canal), and secretes collagenase and other enzymes. Calcium, magnesium, phosphate and products of collagen are released into the extracellular fluid as the osteoclasts tunnel into the mineralised bone. Osteoblasts are mature bone cells responsible for bone formation and ossification. They produce the organic portion of the matrix of bone tissue, osteoid, which is composed mainly of type I collagen, and are responsible for mineralisation of the osteoid matrix. Ossification fixes circulating calcium in its mineral form, removing it from the bloodstream. Repeated stress, such as weight-bearing exercise or bone healing, results in the bone thickening at the points of high stress.

Remodelling in adults repairs micro-damage to bone and plays a role in the regulation of calcium homeostasis. An imbalance in the bone remodelling processes in adults is thought to impact on bone strength as a result of reductions in bone volume and mineralisation, loss of trabeculae, deterioration of trabecular connectivity, and the formation of resorption cavities and trabecular perforations. 3,4 Therefore, an increase in bone turnover where resorption exceeds formation is not only inversely correlated with bone mineral density (BMD), but may also alter bone architecture and porosity, increasing the risk of fracture beyond that due to reduced BMD, and can therefore be an independent predictor of fracture risk. 3–6

Diagnosis

Osteoporosis causes no symptoms until a bone is broken. As osteoporosis is associated with low bone density, bone density scanning [using dual-energy X-ray absorptiometry (DXA)] has become the most commonly used diagnostic technique. 2 There are accepted diagnostic criteria based on DXA: osteopenia (low bone mass) is present when the BMD is between 1 and 2.5 standard deviations below the mean value for young adults (BMD T-score of –1 to –2.5); osteoporosis is diagnosed when BMD is < 2.5 standard deviations below young adults’ (BMD T-score of < –2.5). 7

Risk of fracture

A reduction in BMD results in the thinning of the trabeculae and an increase in the fragility of the bones. 8 Therefore, people diagnosed with osteoporosis have an increased risk of suffering low trauma (fragility) fractures. When BMD is measured by DXA, a reduction of 1 standard deviation in BMD is reportedly associated with a 50–150% increase in the risk of osteoporotic fracture. 9 Increasing age is one of the major risk factors for osteoporosis; after 35 years of age bone loss increases gradually as part of the natural ageing process. 2 By 75 years of age, approximately half of the population will have osteoporosis. In addition, there is an increased risk of falling which increases the risk of fracture; one in two women and one in five men over the age of 50 in the UK will fracture a bone, mainly as a result of skeletal fragility. 2,10 The most common fractures in people with osteoporosis are of the wrists, hips and spinal bones; these are most common in older people, but younger people can sometimes be affected. 8,11

According to recent National Institute for Health and Care Excellence (NICE) guidance [clinical guideline (CG) 146], assessment of the risk of fragility fractures should be considered in:12

• all women aged 65 years and over and in all men aged 75 years and over

• in women aged under 65 years and in men aged under 75 years in the presence of risk factors, for example:

  • previous fragility fracture

  • current use or frequent recent use of oral or systemic glucocorticoids

  • history of falls

  • family history of hip fracture

  • other causes of secondary osteoporosis

  • low body mass index (BMI) (< 18.5 kg/m²)

  • smoking

  • alcohol intake of more than 14 units per week for women and more than 21 units per week for men.

An assessment tool for assessing fracture risk, FRAX^®, has been developed by the World Health Organization (WHO). 13 The factors taken into account are age, gender, weight, height, previous fracture, parental history of hip fracture, smoking status, the use of oral glucocorticoid steroids, a diagnosis of rheumatoid arthritis, the presence of a disorder strongly associated with osteoporosis and alcohol consumption, with or without BMD as determined using DXA. 12,14

Treatments for osteoporosis

Diet and exercise can be modified to improve a person’s fracture risk. Exercises considered best for people with osteoporosis are those that (1) are thought to have an effect on density and strength, such as weight-bearing exercises that cause force on the bones like jogging, stair climbing, walking briskly and resistance exercises, and (2) can reduce the risk of falling, such as balance training (e.g. tai chi), leg strengthening and flexibility training (e.g. yoga). Exercises that people with osteoporosis are advised to avoid are those that might increase the risk of falling, those that involve twisting the spine or bending from the waist, high-impact activities such as high-intensity aerobics or jumping and the use of excessive weight during resistance exercise. A diet containing foods rich in calcium and vitamin D (vitamin D is required for the absorption of calcium) or the use of calcium and vitamin D supplements can also improve bone strength.

The most common medical therapies for osteoporosis are bisphosphonate drugs. Bisphosphonates inhibit the activity of mature osteoclasts and reduce the rate of resorption. 4 The most commonly prescribed bisphosphonate is generic alendronate; other bisphosphonates include etidronate, risedronate (now available in generic form), ibandronate, and zoledronate. The recommended dose of alendronate is one 70-mg tablet per week, rather than 10 mg daily as originally prescribed, to reduce the incidence of gastrointestinal adverse effects and increase adherence. A strict technique must be adhered to when taking oral bisphosphonates to ensure satisfactory absorption. They must be taken on an empty stomach first thing in the morning, while remaining upright to prevent reflux, at least 30 minutes before the first food, drink or other medication of the day. The tablet should be taken with plain water only; other drinks (including mineral water), food and some medicines are likely to reduce the absorption of bisphosphonates. 15 Intravenously administered bisphosphonates are available; the recommended doses are 3 mg 3-monthly of ibandronate, or 5 mg annually of zoledronate. Pamidronate is not licensed for the treatment of osteoporosis but has been widely used off-licence at a dose of 30 mg quarterly.

Other medical therapies available include:

• raloxifene (Evista^®, Eli Lilly and Company Ltd): a selective oestrogen receptor modulator (SERM), which is a synthetic hormone that copies the effects of oestrogen on the bones

• strontium ranelate (Protelos^®, Servier Laboratories Ltd): a strontium(II) salt of ranelic acid, which is a dual-action bone agent that stimulates new bone growth and reduces bone loss

• teriparatide (Forsteo^®, Eli Lilly and Company Ltd): a recombinant form of parathyroid hormone (PTH 1–34) that helps regulate calcium levels and the activity of cells involved in bone formation

• denosumab (Prolia^®, Amgen Ltd): a monoclonal antibody that targets the RANK ligand

• hormone replacement therapy (HRT): a mix of hormones (oestrogens, progesterone or progestins, and sometimes testosterone) prescribed to post-menopausal women (natural or surgically induced) to reduce the symptoms caused by reduced circulating oestrogen and progesterone. The risk of development and progression of osteoporosis can therefore be reduced by the maintenance of oestrogen levels.

Burden of the disease on the NHS

Approximately 3 million people in the UK have osteoporosis, with about 20% of women aged 60–69 affected. There are thought to be about 230,000 osteoporotic fractures every year, with broken wrists, hips and spinal bones being the most common. Of the 60,000 people who suffer osteoporotic hip fractures each year, 15–20% are likely to die within a year from causes related to the fracture. 2

As stated in Diagnosis, above, there are a range of treatments available for osteoporosis, and the costs of these vary (pamidronate has not been costed as it is not licensed for use in osteoporosis):16

• Generic sodium alendronate: a 28-tablet pack of 10-mg tablets is £1.44 (approximately £19 annually); a four-tablet pack of 70 mg for once-weekly administration is £1.10 (approximately £14 annually). Fosamax^® (MSD) costs £23.12 for 28 10-mg tablets and £22.80 for four 70-mg once-weekly tablets.

• Generic sodium risedronate: a 28-tablet pack of 5-mg tablets is £17.99 (approximately £220 annually); a four-tablet pack of 35 mg for once-weekly administration is £19.12 (approximately £230 annually).

• Zoledronate: Zometa^® (Novartis) costs £174.17 for 4 mg in 5 ml, and Aclasta^® (Novartis) costs £253.38 for 5 mg in 100 ml – 5 mg administered once annually.

• Strontium ranelate (Protelos^®, Servier) costs £25.60 for 28 sachets each containing 2 g of granules daily (approximately £330 annually).

• Denosumab (Prolia^®, Amgen Ltd) costs £183.00 for 60 mg/ml in a 1-ml prefilled syringe – 60 mg administered 6-monthly (£366 annually).

• Raloxifene (Evista^®, Daiichi Sankyo) costs £17.06 for 28, and £59.59 for 84, 60-mg tablets – 60 mg daily (approximately £220 annually).

• Teriparatide (Forteo^®, Eli Lilly and Company Ltd) costs £271.88 for 250 µg/ml in a 3-ml pre-filled pen – 20 µg self-administered daily (approximately £3540 annually).

According to Hospital Episode Statistics (HES), in 2005–6 in England there were 5759 consultations and 4034 admissions (2368 emergency) for osteoporosis with a pathological fracture, and a further 8725 consultations and 8313 admissions (716 emergency) without pathological fracture. 17 For surgical interventions for fractures of the spine and hip (not only those associated with osteoporosis), there were 809 consultations and 667 admissions (353 emergency) for fixations of spinal fractures (approximately 26% in patients 60 years and older), and 46,812 consultations and 46,191 admissions (1611 emergency) for primary total prosthetic replacement of hip joint [depending on method used, approximately 50% (not using cement) to 85% (using cement) 60 years and older]. 17 Given the discrepancies in the numbers of hip replacements in the elderly and consultations of osteoporotic fractures, the incidence/consultation rate for osteoporosis may be underestimated. A recent report published by the Royal College of Physicians stated that only 32% (1933 out of 6083) of non-hip fracture and 67% (2324 out of 3484) of hip fracture patients had a clinical assessment for osteoporosis/fracture risk. 18 Osteoporosis reportedly costs the NHS and government £2.3B per year (£6M per day). 2

National Institute for Health and Care Excellence guidance

NICE has produced a number of technology appraisals (TAs) and CGs that have some relevance to this area. Three relevant TAs have been published: TA160 (Osteoporosis – primary prevention; postmenopausal women),19 TA161 (Osteoporosis – secondary prevention including strontium ranelate; postmenopausal women)20 and TA204 (Osteoporotic fractures – denosumab). 21

For the primary prevention of fractures, alendronate is recommended as the first-line treatment for most women at risk of fractures. Risedronate, etidronate and strontium ranelate are alternative treatments for post-menopausal women who cannot adhere to the required alendronate regimen, or those women with pre-specified combinations of T-score, age and number of independent clinical risk factors; strontium ranelate is not recommended as a first-line treatment for osteoporosis. Raloxifene is not a recommended treatment for the primary prevention of osteoporotic fragility fractures. 19 The recommendations for the secondary prevention of fractures are similar to those for primary prevention. The two differences are that (1) strontium ranelate can be used as a first-line treatment and (2) raloxifene is recommended as an alternative treatment for post-menopausal women who cannot adhere to alendronate, or in women with pre-specified combinations of T-score, age and number of independent clinical risk factors. 20 Denosumab has now also been added to the list of alternative second-line treatments for the primary or secondary prevention of fractures. 21

There are also four potentially relevant CGs available that deal with the management of independent risk factors for fracture: CG146 (Osteoporosis fragility fracture),12 CG21 (Falls: the assessment and prevention of falls in older people),22 CG59 (Osteoarthritis: the care and management of osteoarthritis in adults)23 and CG79 (Rheumatoid arthritis: the management of rheumatoid arthritis in adults). 24

This review will focus on patients being treated for osteoporosis with any of bisphosphonate, raloxifene, strontium ranelate, teriparatide or denosumab.

Bone turnover markers

Biochemical markers of bone turnover are used to monitor treatment response and may prove to be more useful than serial BMD measurements as they are non-invasive, relatively cheap compared with DXA, and there is an increased availability of auto-analysers in clinical chemistry laboratories.

Variability in bone turnover markers

Several factors can impact on the bone turnover marker levels, causing variability across samples, which can reduce repeatability and comparability, both within patients and between patients. These include specimen collection and storage;25,31–35 differences between analytical methods used;32,34 temporal variations (diurnal, menstrual, seasonal);25,31–35 diet and fasting;36 patient characteristics (age, gender or ethnicity);25,31,33,35 concomitant medication other than osteoporosis medications [HRT, anabolic agents, glucocorticoids, anticonvulsants, gonadotropin-releasing hormone (GnRH) antagonists or oral contraception];25,31 and comorbid conditions (renal impairment, liver disease, diabetes, thyroid disease, osteomalacia, systematic inflammatory diseases, degenerative joint disease, conditions causing immobility, or eating disorders). 25,31,33,35

Intrapatient variability for serum markers is lower than for urinary markers. 34 Some tests are more accurate when monitoring the response to specific treatments (e.g. CTX with bisphosphonates). Some tests have the advantage of not requiring the patient to fast prior to sampling (e.g. P1NP), or are less affected by diurnal variations (P1NP and BALP), and/or have lower overall intraindividual variability (BALP) than other bone turnover markers. 37 Each of these tests also has disadvantages: CTX has a large circadian rhythm, and therefore repeat sampling must be done at the same time of day, fasting is required prior to sampling, and the marker requires freezing soon after sampling as it can be unstable; BALP is affected by cross-reactivity with the liver form of alkaline phosphatase, limiting its use in patients with liver disease; and P1NP has a higher cost compared with other bone turnover markers. 37 Given the advantages that CTX, P1NP and BALP offer, and the availability of NTX, these are the bone turnover markers that will be investigated in the current review.

Use of bone turnover markers

The use of bone turnover markers varies greatly across the UK, in terms of both the test used and the frequency of its measurement. Several factors will need to be considered when choosing the bone turnover marker to be used, not least the availability of the assay methods. Bone turnover markers have a number of potential uses, including:6,37

1. predicting bone loss

2. identifying people at risk of primary or secondary osteoporosis and fracture

3. predicting treatment response prior to commencement

4. monitoring the response to osteoporosis treatment; identifying non-responders, which will include those not adhering with osteoporosis treatment (including patients not taking the medication or not following the instructions for administration)

5. identifying oversuppression of bone turnover in patient on long-term osteoporosis therapy

6. monitoring of people who have been on long-term treatment, or shown signs of oversuppression, and are taking a ‘treatment holiday’.

The main focus of this systematic review will be role 4: monitoring the response and non-response to osteoporosis therapy (and change in fracture risk).

Monitoring response to treatments for osteoporosis

There is currently no standard practice for the monitoring of patients receiving treatment for osteoporosis. The options include the use of repeated DXA, repeated measures of bone turnover markers, clinical review, or a combination of these. The use of DXA to monitor the response to osteoporosis treatment has limitations. Firstly, detectable changes in bone density due to treatment can take up to 2 years to become apparent;38 therefore, the identification of non-responders to treatment is delayed. Secondly, there is limited access to the technology and the test is relatively expensive (average £72 per scan). Thirdly, there is evidence that there is limited value in the regular monitoring of BMD in patients on bisphosphonate therapy. 39,40

As stated earlier, the relationship between bone turnover and bone density and architecture means that the rate of bone turnover may be an independent predictor of fracture risk;3–6 this can be measured using one or more of the bone turnover markers listed above. However, it is still unclear whether or not changes in bone turnover detected by bone turnover markers are reliable surrogate measures for improved bone density and architecture, and consequently accurate predictors of future fracture risk. Two studies have suggested that bone turnover markers can have independent predictive value in assessment of fracture risk. 41 If biochemical markers of bone turnover are reliable indicators of future fracture risk, their use may prove advantageous compared with serial BMD measurements, as not only are they non-invasive, relatively cheap compared with DXA, and the availability of auto-analysers in clinical chemistry laboratories is increasing, but a response to treatment can be detected much earlier than with DXA.

Changes in bone turnover rates have been detected in post-menopausal women within as early as 2 weeks after starting HRT,42 although the peak accuracy of changes in bone turnover markers to predict fracture risk in response to osteoporosis treatment may be later than this, between 3 and 12 months after initiating treatment, depending on the treatment and bone turnover marker used. 43–46 The ability to identify non-responders early within the treatment can be beneficial for patients by allowing early changes in management strategy if deemed necessary. The definition of treatment success varies depending upon the baseline risk of the patient being treated; in some patients a reduction in bone turnover would be considered a treatment success, but in others success may be a stabilisation of bone turnover. For all patients a continued increase in bone turnover rates would be considered a treatment failure. The definitions used throughout this project will reflect clinical practice and be based upon evidence for least clinical significant change.

There is a complex association between changes in bone turnover and fracture risk that is influenced by the treatment–bone turnover marker combination; the observed change in bone turnover markers will depend upon the treatment being administered. In studies of raloxifene, risedronate, alendronate and zoledronic acid, bone turnover markers have been reported as explaining between 28% and 77% of fracture risk reduction. 47

Bisphosphonates are antiresorptive therapies, and therefore they reduce the rate of bone resorption. Bone resorption is closely coupled to bone formation; consequently, there is usually a subsequent reduction in the rate of bone formation. This results in a transient uncoupling of bone turnover, which leads to a small increase in BMD. This increase in BMD may account in part for the decrease in fracture risk, but the reduction in bone turnover may independently improve bone strength by improving bone architecture and porosity. 48 Both raloxifene and denosumab reduce bone resorption, and therefore act as antiresorptive therapies; decreases in both bone resorption markers and subsequently bone formation markers should be observed in treatment responders as with bisphosphonates.

Teriparatide causes a small, transient increase in serum calcium, mainly due to the stimulation of tubular reabsorption of calcium from the proximal kidney tubules and increased calcium absorption from the bowel, but in a small part by increasing bone resorption (hence chronically elevated PTH can deplete bone). However, intermittent administration of PTH (i.e. daily injections of teriparatide) activates osteoblasts more than osteoclasts, stimulating new bone formation and increasing BMD. Therefore, a positive response in bone formation markers would be expected in treatment responders, with a subsequent increase in bone resorption markers due to the coupling of the processes, the opposite response to that seen with antiresorptive therapies.

Strontium ranelate increases new bone formation as well as reducing bone resorption and is classed as a dual-action bone agent. These effects are more modest than those seen with anabolic and antiresorptive treatments, with smaller positive changes in bone formation and negative changes in bone resorption markers, respectively. However, strontium ranelate appears to lead to persistent uncoupling of bone turnover.

The interpretation of changes in bone turnover markers is also influenced by the type of sample used: serum or urine. The intraindividual variability is greater for urine markers, giving serum markers a better signal to noise (S/N) ratio;34 the percentage change in a urinary biomarker needed to indicate a treatment response (least significant change) is greater than that required for a serum biomarker.

Identification of studies

The screening of titles and abstracts was conducted by two independent reviewers. All potentially relevant studies were retrieved where available, and two independent reviewers applied the inclusion criteria to the full papers. Disagreements were resolved by team discussion. Where consensus could not be reached at the title and abstract stage, the full paper was ordered. Inclusion was not restricted by language or date of publication. Abstracts were included if no additional information was available and there were sufficient outcome data to extract.

Inclusion and exclusion criteria

Data extraction strategy

Clinical data extraction was conducted by one reviewer using a standardised data extraction form and checked by a second reviewer. Discrepancies were resolved by discussion, with involvement of a third reviewer when necessary. Attempts were made to contact authors for missing data. Data from multiple publications of the same study were extracted and reported as a single study. Where applicable and available, extraction included data on study details (e.g. study/EndNote identifier, author, year, country, setting, number of participants and duration of follow-up), patient characteristics (e.g. age, gender, duration of osteoporosis, risk group, concomitant renal/liver disease; baseline bone turnover marker levels and BMD), details of intervention (serum or urine; sample collection details; pre-sampling preparations/restrictions; sample storage details; assay used; adjustments for creatinine excretion for urinary markers; delay between sample collection and assay; single/serial measures; intra- and interassay coefficients of variation; value for least significant change), study quality, and reported outcomes as specified above.

Economic data extraction was planned on the comparators, study population, main analytic approaches (e.g. patient-level analysis/decision-analytic modelling), primary outcome specified for the economic analysis, details of adjustment for QoL, direct costs (medical and non-medical) and productivity costs, estimates of incremental cost-effectiveness and approaches to quantifying decision uncertainty (e.g. deterministic/probabilistic sensitivity analysis).

Critical appraisal strategy

The quality of the individual studies was assessed at the study level by one reviewer and independently checked by a second; disagreements were resolved by consensus. The quality of included studies was assessed using relevant criteria suitable for the study design selected from standard checklists for RCTs, observational studies and economic evalutaions;62,66–68 topic-specific quality issues were incorporated where necessary (see Appendix 2 for details and guidance for completion).

Methods of data synthesis

Key study characteristics, patient outcomes and study quality were summarised in a narrative and tables. Meta-analyses suitable to the clinical data extracted were planned to estimate a summary measure of effect when sufficient numbers of comparable studies were available for an outcome. Given the substantial heterogeneity across the studies, this was not possible. It was also not possible to investigate the potential sources of heterogeneity that were specified in the protocol, as insufficient numbers of studies similar in other population, intervention and methodological characteristics were identified. The analyses planned were:

• investigation of potential subgroups of interest where sufficient data are available; for example, post-menopausal women (overall and for specific age ranges if data are available), elderly, skeletal site (hip, spine or wrist), and glucocorticoid-induced osteoporosis

• sensitivity analyses conducted, where appropriate, to investigate potential sources of heterogeneity such as study quality, and differences in sample acquisition, storage and assay methods.

A narrative synthesis of the methods and results of the cost-effectiveness studies was planned.

Quantity and overall quality of research available

As a result of the electronic and hand searching, 4002 papers were identified for initial screening. Of these, 437 were retrieved as full papers; 35 were reviews that underwent screening of their bibliographies. 5,6,11,15,34,47,50,69–96 Sixteen of the full papers were published in languages other than English. 37,52,74,76,79,81,95,97–105 Nineteen studies had duplicate publications; where this was the case14,40,42,43,56,103,104,106–144 one paper was allocated as the primary publication (full paper if the duplicate was an abstract or letter; the published paper if the duplicate comprised data from the manufacturer’s online trials database, the most recent publication, or the paper published in English) and used as the citation for the study throughout the review;14,40,42,43,56,106,131–140,142,143 relevant data were extracted from all publications where applicable. After full-paper screening, 42 studies (across 70 publications) met the inclusion criteria for the review of clinical effectiveness; the flow of studies through the review is given in Figure 1 (some of the studies were excluded for more than one reason). A list of excluded studies with the reasons for exclusion is given in Appendix 3.

FIGURE 1.

Flow of studies through the review.

The majority of the studies were conducted primarily, or entirely, in post-menopausal women; only four studies reported including men. 14,99,145,146 Where reported, the mean age ranged from 56.1 to 73.9 years. The definition used for the diagnosis of osteoporosis differed across studies; the definition was not provided in nine studies. 106,133,135,143,145,147–150 Most of the included studies were small, with the number of participants ranging from 22 to 3105 where reported (two studies did not report the number recruited40,151); 21 studies had fewer than 100 participants. 41,43,44,58,99,106,135,136,145,147,150,152–160 A summary of study characteristics is given in Table 1; full data extraction tables are provided in Appendix 4.

Quantity and quality of the included non-randomised studies

Of the 37 non-randomised studies, 21 were cohorts derived from the treatment arms of RCTs that compared a treatment regimen with either placebo or an alternative treatment. 14,40–43,132,137,139,140,146,147,151–153,156,159,161–165 The cohorts were derived either directly from reports of the RCT42,43,132,139,140,147,151–153,156,159,161 or from a paper reporting a post hoc analysis of the RCT40,146,162–164 or from a post hoc analysis of a subgroup of patients from the RCT. 14,41,137,165 These derived cohorts were assessed for quality as cohort studies rather than RCTs, as this is the manner in which they were used in the review.

The remaining 16 studies were 14 uncontrolled cohort studies58,99,106,131,135,136,142,145,150,154,155,157,158,160 and two controlled cohort studies; one compared two groups with different prior treatment regimens,38 and the other treated one group of women with HRT and the other with alendronate, although the groups seem to have been established after recruitment. 44 Single cohorts were derived from both controlled cohort studies. In the only truly controlled cohort study, the recruitment of the control group was considered appropriate and the groups comparable at baseline. 38 Of the 16 studies that were designed as cohort studies, only four appeared to use consecutive recruitment (though this was usually inferred, not explicitly stated),38,142,154,160 one was not consecutive,136 and recruitment was unclear in 11. 44,58,99,106,131,135,145,150,155,157,158

Across the 37 non-randomised studies, 16 recruited a representative population,38,40,58,99,131,132,137,140,151,155,157,158,160,163–165 17 did not,14,41–44,106,136,139,142,146,152–154,156,159,161,162 and in the remaining four, it was unclear whether or not the population was representative. 135,145,147,150 Only nine studies provided sufficient intervention details to allow repetition. 38,41,58,153,155,157,158,160,163 Ten studies had no loss to follow-up for the analyses they conducted (some were post hoc analyses of the specific group of patients with the required outcome measures). 41,44,58,106,137,154,157,159,164,165 The controlled cohort study used a ‘modified’ ITT analysis; seven patients were excluded from the ITT population, and last observation carried forward was used for the imputation of missing data for those in the modified ITT population. 38 The reasons were given for losses to follow-up in a further 11 studies. 14,42,131,136,139,142,146,153,156,158,160

Confounders were clearly identified and described in five studies. 137,151,154,155,161 Although six studies adjusted for confounding factors in multiple regression analyses,137,139,151,155,161,163 only two adjusted for all the confounders considered important by the review authors (age, gender, prior fracture, baseline BMD, and BMI). 151,155 Twenty-seven of the studies had a minimum of 1-year follow-up in all patients. 14,38,41,42,44,58,99,106,131,132,135–137,139,142,145,150–152,155,157–161,163,165 The assessment of the reporting of adverse events of bone turnover markers was not assessed for these studies as their focus was not the assessment of the effectiveness or safety of bone turnover markers. A summary of the results of the quality assessment is given in Figure 2.

FIGURE 2.

Summary of the quality of the non-randomised studies.

Given the limitations of the non-randomised studies included, all were considered to be at a high or unclear risk of bias, and therefore the overall quality of each of the studies was considered to be low. The low quality assigned to the cohorts derived from RCTs is not necessarily a reflection of the quality of the original RCT; this classification was primarily driven by the post hoc nature of the selection of patients in the paper reporting the outcomes of interest for this review, which can introduce bias. The full results of the quality assessment and the guidance used for its completion, including the criteria on which the overall quality was primarily based upon, are given in Appendix 2.

Assessment of clinical effectiveness

Five RCTs56,133,143,148,149 and one post hoc analysis of a subgroup of patients41 from a previous RCT166 evaluated the effectiveness of feedback of bone turnover marker results on QoL and/or adherence, compliance, or persistence in post-menopausal women. These terms were defined variably across the studies, particularly the proportion of medication taken which was the definition used across the studies for adherence, compliance and persistence. We decided to standardise the definitions by adopting the system used by Delmas et al. :56 persistence is the time until discontinuation of medication, compliance is the proportion of medication taken, and adherence is a combination of persistence and compliance. None of the RCTs reported on the impact of bone turnover marker monitoring on treatment management.

Assessment of test accuracy

Most of the evidence available evaluating the accuracy of bone turnover markers was in the form of correlations between changes in bone turnover marker levels and BMD measured using DXA. The studies were extremely heterogeneous precluding the use of any meta-analytical models; the data were therefore presented in tables for each type of treatment (bisphosphonates, teriparatide, raloxifene, strontium ranelate and denosumab) with a brief narrative. Even if the data were not heterogeneous, the usefulness of these correlation data to inform the accuracy of bone turnover marker tests for identifying patients who remain at risk of fracture is limited; this is discussed in more detail in Limitations of the available evidence.

Twenty-seven studies reported the results of correlation analyses,14,38,40,42–44,58,106,131,135,136,140,145,146,150,152–155,157–163,165 eight reported the results of multiple regression analyses,132,137,139,151,153,155,161,163 and four reported both. 153,155,161,163 The r-values reported were derived using Pearson’s correlation in seven studies44,58,152,157,160,162,163 and Spearman’s rank correlation in 11;14,43,106,135,140,146,153,158,159,161,165 the method was not reported in nine studies. 38,40,42,131,136,145,150,154,155 Five studies reported predictive accuracy results in terms of sensitivity,140,156 results of receiver operating characteristic (ROC) analyses,140,156,159,161 or reported sufficient data to produce 2 × 2 tables of test performance. 163 One uncontrolled cohort study reported the difference in bone turnover marker measurements between those with and without a fracture,142 and another between those who did and did not have a response in BMD to treatment. 162

Assessment of test reliability and reproducibility

Four studies reported S/N ratio for a bone turnover marker in those being treated with either etidronate,156 teriparatide42 or raloxifene. 147,164 Each study calculated the S/N ratio differently, making comparisons across studies difficult (Table 17). Within-study comparisons show that sP1NP at 2 weeks had a lower S/N ratio than sCTX at 2 weeks but a higher S/N ratio at 25 weeks,147 and sP1NP had a greater S/N ratio than sBALP when measured at 6 months. 42 Blumsohn et al. 42 reported the intraclass correlation coefficients measured at two time points between 3 and 14 days apart, which were 0.988 for sBALP and 0.983 for sP1NP. Clowes et al. 147 reported the intraindividual coefficients of variation (CVs), which were 13.2 for sP1NP and 22.9 for sCTX.

Key findings

The included RCTs evaluated the impact of the feedback of results of bone turnover marker tests on QoL and/or adherence, compliance and/or persistence with bisphosphonate therapy. Although this is an important area of research, the usefulness of these data is limited because of the treatment regimens administered and the high baseline compliance and persistence rates; these issues are discussed further in Limitations of the available evidence, below. There was some evidence that the feedback of bone turnover marker results, and the message given, impacted on persistence with treatment; a positive result seems to encourage persistence and a negative result to discourage it. 56 This may not be the response expected to the feedback of such results in clinical practice. It may be expected that a message that treatment was not working could encourage adherence rather than discourage it, assuming that the potential risk of fracture is highlighted to the patient. It is worth noting that the RCT evaluating this, firstly, was conducted in older women aged 65–80 years, and may not be reflective of the response of younger post-menopausal women or of those with secondary osteoporosis, and, secondly, used daily risedronate which is a regimen that is not commonly used in current clinical practice.

Two of the included RCTs also reported on the QoL using the OPPS questionnaire; these studies reported small improvements for those patients receiving feedback in the overall, feeling informed and satisfaction scores133,148 and the confidence score,148 compared with those who did not receive feedback. Although this is based on only two studies, one received questionnaires from 563 of the 596 participants recruited, and the other study recruited 585 participants although it was not clear how many of these returned questionnaires. A study identified during the scoping review of modelling methods explored the impact of monitoring antiresorptive treatments using a bone turnover marker on quality-adjusted life-years (QALYs). 167 The decision tree (up to 3 months) and Markov model (3 months to 5 years) was based on a 60-year-old woman with post-menopausal osteoporosis with a total hip T-score of –3, no concurrent disease, and second-generation bisphosphonate therapy for 5 years. The comparators were no monitoring beyond a simple short-term follow-up to rule out adverse reactions, and monitoring with a serum bone resorption marker after 3 months of treatment. This modelling study showed small improvements in QALYs when adherence to treatment was assumed to be the same in both monitoring strategies, with the improvement being greater if monitoring improved adherence. 167

The majority of the evidence for the review of clinical effectiveness was based around the predictive accuracy of bone turnover marker tests in the form of correlations with changes in BMD; the limitations of these data are discussed in Limitations of the available evidence, below. Given the extreme heterogeneity across the studies, it was impossible to determine trends for any particular treatment–bone turnover marker monitoring combination. Studies correlating changes in bone turnover with either biopsy results or fracture outcomes were uncommon, with two studies using biopsy results as the comparator150,152 and seven using fracture. 40,42,132,137,139,151,161 The use of biopsy in all patients in a study is unlikely to be considered ethical because of the invasive nature of the procedure and the risk of complications. In addition, there are several limitations to the use of fracture outcomes as the comparator: the outcome is relatively rare, resulting in the need for larger studies; there needs to be a longer duration of follow-up in order to detect the outcome; and attrition is likely to be a problem over the time period required. These limitations may be the reason that DXA is most commonly used for the comparator. Data from the studies that used either biopsy or fracture were heterogeneous in terms of patient population and treatment regimen, and method of data analysis. The results of these studies were inconsistent and there were insufficient data to determine trends for any particular treatment–bone turnover marker monitoring combination.

There was no evidence available to inform the question as to the clinical effectiveness of treatment monitoring including a bone turnover marker over and above monitoring regimens where a bone turnover marker was not used. Further, the evidence available relating to the predictive accuracy, reliability and reproducibility was heterogeneous and of low quality, precluding the ability to draw any conclusions regarding the choice of bone turnover marker for use in monitoring in routine clinical practice.

Strengths and limitations of the review

The systematic review was based on an extensive search with well-designed search strategies. Abstracts were included when there were sufficient data to be extracted; authors of all potentially included abstracts were contacted in an attempt to identify full publications and/or obtain unpublished data. In addition, no study was excluded based on date of conduct or language of publication. Studies were selected using inclusion criteria defined a priori. We included RCTs of any size in the review of the clinical effectiveness of bone turnover marker monitoring. However, given (1) the large number of non-randomised studies and randomised studies from which cohorts were derived, from which data on test accuracy, reliability and reproducibility could have been extracted, (2) the apparent poor quality of this evidence base, and (3) the time constraints on the project, an additional criterion not specified in the protocol was applied. We excluded non-randomised studies and derived cohorts that included fewer than 20 osteoporotic patients, receiving one of the treatments of interest, in all the analyses of outcomes relevant to this review. This cut-off was used to distinguish between a case series and a cohort study, excluding from the review the lowest levels of evidence. This is the only change made to the protocol. Although the studies were too heterogeneous to pool using statistical meta-analytical models, we conducted a narrative synthesis in an attempt to summarise the evidence available.

Limitations of the available evidence

As with all systematic reviews, the reliability and generalisability of the results are governed by the quality and quantity of the evidence available. Although we included 42 studies, all were considered to be at a high or unclear risk of bias, and therefore of low quality. Furthermore, no RCTs were identified that addressed one of the primary aims of the review: to evaluate the clinical effectiveness of monitoring regimens that included at least one bone turnover marker, over and above monitoring regimens where a bone turnover marker was not used. Where RCTs were conducted, these reported primarily on compliance. 56,133,143,148,149

Adherence to treatment (in terms of both compliance and persistence), particularly with oral bisphosphonates, is a considerable problem in the management of patients with osteoporosis. 32,168–171 Oral bisphosphonates are associated with gastrointestinal adverse events,15,32,49–51,172 and require adherence to specific instructions for administration to maximise absorption and bioavailability that many patients find inconvenient. 15,32,53 Non-adherence to treatment regimens will result in a non- (or inadequate) response to treatment and a continued higher risk of fracture. 173,174 A recent systematic review of 27 observational studies, most of which were retrospective database analyses, reported a rate of fracture ranging from 6% to 38% in patients who were non-compliant, and 5% to 19% in patients who were non-persistent with osteoporosis medications. 175 The meta-analysis conducted included 12 of the studies and indicated a statistically significant increased risk of fracture of approximately 30% in patients who were non-compliant (OR 1.29, 95% CI 1.22 to 1.38), and 30% to 40% in patients who were non-persistent (OR 1.40, 95% CI 1.29 to 1.52) with treatment. 175 Most of the patients in the review received bisphosphonates; however, raloxifene, strontium ranelate, calcitonin, HRT, vitamin D and calcium were included in the regimens administered. 175

In the five RCTs that met the inclusion criteria for this review,56,133,143,148,149 a very high proportion of patients were adherent to medication; this is unlikely to be representative of clinical practice. Why there was such a high rate of compliance in the trials is unclear. It could be due to the use of a once-monthly dosing regimen used in most of the trials rather than daily or weekly dosing, as less frequent dosing is generally preferred by patients and can result in increased adherence. 32,173,176 Other potential reasons are: the short duration of follow-up used in most of the studies; that patients giving consent to be part of the trial are more likely to be compliant; and/or the increased attention and resulting patient awareness that being part of a trial may result in. It is, therefore, unclear how the feedback of bone turnover marker results would impact on the population seen in generally clinical practice, which is likely to be less compliant and/or persistent.

Most of the evidence available relating to the accuracy of the bone turnover markers was in the form of correlations with BMD, measured using DXA. Some of the correlations, but not all, were statistically significant. The fact that a correlation coefficient is statistically significant is not indicative of a strong association between the two variables. The likelihood of the results of a correlation analysis achieving statistical significance is influenced by study sample size; small samples are unlikely to have sufficient power to detect statistically significant results. Indeed, the results showed non-significant but strong correlations in small studies, and significant but weak correlations in larger studies; the majority of the studies where a strong correlation was detected were small, thus lacking the power to detect significant results. Although pooling these studies to derive a summary estimate could be a solution to increase the statistical power, the high level of clinical heterogeneity across the studies precluded the use of meta-analytical techniques. In addition, a non-significant result reflects only that there is no linear association (which is what the correlation coefficient evaluates) and not that there is no association at all.

These correlation data may be further limited in their usefulness, as BMD is only one factor that impacts on bone strength and therefore on fracture risk,177 and the association between bone turnover markers and BMD may vary depending on the population and on the menopausal status in women. 70 There are three factors that determine bone strength: density, architecture and porosity,177 and therefore increases in BMD explain only a limited proportion of the reduction in fracture risk. 91,178 Bone turnover impacts on all the three aspects of bone strength,177 and so unlike BMD, biomarkers used to measure bone turnover reflect changes in overall bone strength, and hence potentially fracture risk. 91,177 It has been suggested that it is reductions in activation frequencies and reductions in bone turnover resulting in thickening of critical trabeculae, reductions in perforation of trabecular plates and promotion of bone mineralisation, rather than increases in BMD, that reduce a patient’s risk of fracture with antiresorptive therapy. 48,179 In studies of raloxifene, risedronate, alendronate and zoledronic acid, bone turnover markers reportedly explained between 28% and 77% of the fracture risk reduction, compared with changes in BMD explaining up to 28% with the same agents. 47 Therefore, even if there is no significant correlation between changes in bone turnover markers and increases in BMD, it cannot be assumed that there is no correlation with fracture risk. BMD is a poor surrogate on which to assess the accuracy of bone turnover markers, and any presumption that bone turnover markers are not effective in identifying patients who are not responsive to treatment and still at risk of fracture based on these data would be unwise. 70,177,178

The lack of value of the correlation data was compounded by the extreme heterogeneity across the studies in terms of patient populations, the definitions used to diagnose osteoporosis, treatment regimens, bone turnover marker tests used and their timing, and the DXA sites used and the timing of DXA scans; no two studies reported the results for the same combination of these factors, precluding the use of any meta-analytical models. In addition, where reported, there was heterogeneity across studies in terms of obtaining and handling of samples. For the studies evaluating serum bone turnover markers, patients were not fasted in one study using sBALP,142 storage temperature ranged from –20 °C140,157,158,160 to –80 °C,58,163 and the assay method used also varied (most noticeably for sBALP, which was measured using chemiluminescence,42,43 enzyme immunoassay,142,154,156,159,165 ELISA,40,58,160 immunocapture,157 IRMA,14,132,139,140,151,152,157,161,164 or an ostase assay where the variant was not reported146). For studies evaluating urinary bone turnover markers, where reported, collection was the first morning void151 or second morning void,44,56,137,142,152,158,161 storage temperature ranged from –20 °C44,137,140 to –70 °C,154,158 and the assay method used varied, with chemiluminescence10,162 and ELISA44,56,99,131,140,142,152,154,155,158,161 being used in studies that reported the assay; two of the studies using uNTX did not report whether or not the results were corrected for creatinine. 56,136 These limitations in the data prevented us from drawing any strong conclusions regarding the relationship between change in bone turnover markers and BMD, or fracture, from the results of these correlation analyses.

Correlations between changes in bone turnover markers and biopsy results and, more importantly, fracture outcomes70 may be more useful than correlations between changes in bone turnover markers and BMD. These two outcomes do not solely measure changes in bone density, but incorporate bone porosity and microarchitecture. In addition, an assessment of the association with fracture incidence is a direct assessment of the association with the event the bone turnover markers are attempting to predict at the time of testing. The only study to report correlations to both BMD and fracture incidence showed that post-menopausal women over 55 years of age receiving teriparatide reported statistically significant correlations between changes in sP1NP at 1 month and BMD at 24 months, but not fracture. 42 However, the correlations with BMD were weak, and the correlation coefficients for the association with fracture were not reported; therefore, the strength of the association with fracture is unclear. Whether this would be true of other population–bone turnover marker–treatment combinations, or if sP1NP was measured after more than 1 month of treatment, is uncertain.

The assessment of the relationship between changes in bone turnover markers and biopsy or fracture outcomes would be further improved if confounding factors were adjusted for in a multivariate regression analysis. Even if significant correlations are identified between bone turnover markers and these other measures for fracture risk, it is unclear whether or not the use of correlation statistics fully explains the relationship. If these variables were to be incorporated into a multiple regression analysis where other important predictive variables are also included, it is possible that these other variables, or combination of variables, may be stronger predictors of fracture risk than bone turnover markers. In addition, where there is a non-significant association between change in bone turnover markers and fracture risk, the association may change and become significant when other predictive factors are included in a multivariate regression analysis, owing to a synergistic effect of combined variables. These apparently non-significant associations may therefore be important influences within a multivariate regression analysis. Therefore, evaluations of the association between changes in bone turnover markers and subsequent fracture risk outcomes should incorporate confounding factors. In order to assess the accuracy of bone turnover markers in terms of their ability to identify patients who remain at risk of fracture, research needs to be conducted to investigate their independent predictive value of fracture risk.

Data from studies conducting multiple regression analyses gave some indications that a significant association between the changes in bone turnover markers and the incidence of hip or vertebral fractures in patients being treated with bisphosphonates was observed. However, although we have indicated that data from multiple regression analyses that adjust for important confounding factors are of more value than data from correlation analyses, the limited use of these analyses, and the heterogeneity across the studies that did conduct them, limits the usefulness of these data.

As a consequence of the limitations of the data discussed, there is currently insufficient high-quality, consistent evidence available to draw any firm conclusions on the ability of changes in bone turnover markers to identify patients not responding to treatment, or to predict future fracture risk. There are substantial gaps in the clinical evidence base, particularly in terms of the impact of bone turnover marker monitoring on treatment management decisions, and the independent predictive value of bone turnover markers for future fracture. Further research is required. However, given the large number of possible combinations of patient population–treatment–monitoring regimens, decision-analytic modelling will be an essential component of that research if we aim to inform efficient decision-making. Therefore, suggestions for future research are discussed in Chapter 6, Suggested research priorities so that these can incorporate the needs of the assessments of both clinical effectiveness and cost-effectiveness.

Comparison with previous systematic reviews

One review included RCTs that compared antiresorptive treatment regimens with placebo in post-menopausal women, reported associations between changes in a bone turnover marker (or BMD) and the incidence of fracture. 75 The review was based on 18 trials involving 26,494 women, which amounted to 69,369 woman-years. Poisson regression analyses were used to investigate associations between the difference in percentage change in bone turnover marker between the treatment and placebo groups and the relative risk of fracture, weighted by woman-years. As with our review, the trials were heterogeneous in terms of the bone turnover markers used, treatment regimens, patient population and duration of follow-up over which fracture was assessed. Antiresorptive therapy in general resulted in a decrease in resorption bone turnover marker, and a concomitant decrease in the rate of bone formation turnover markers. Therefore, a decrease in either type of marker is seen as an indication of a reduction in bone turnover. Overall, regression coefficients were 0.0067 [standard error (SE) 0.0034; range of change in bone turnover marker compared with placebo across studies +1% to –70%] for changes in resorption bone turnover marker and 0.0134 (0.0051; range of change in bone turnover marker compared with placebo across studies +7% to –56%) for formation markers. There was a 40% decrease in fracture risk for those treatments that decrease resorption markers by 70% compared with placebo and a 44% decrease in fracture risk for those treatments that decrease formation markers by 50% compared with placebo. Changes in bone turnover marker were significantly correlated with changes in BMD (p ≤ 0.002); regression coefficient (R²) = 0.58 for resorption markers and 0.41 for formation markers. 75

It is unclear how comparable the results of this review are to our review, for several reasons. Firstly, trials that evaluated calcitonin and alendronate combined with oestrogen were included in the analyses; these are treatments not being considered in our review. Calcitonin is now authorised only for short-term use in Paget’s disease and acute bone loss due to sudden immobilisation and hypercalcaemia caused by cancer, and not for treatment of osteoporosis,180 and oestrogen is a HRT used to treat and prevent a range of post-menopausal symptoms which could lead to an overestimation of the effectiveness of alendronate. Secondly, data for resorption markers (urinary deoxypyridinoline, CTX, NTX and urinary hydroxyproline) were combined, as were data for formation markers (serum osteocalcin and sBALP); the specific bone turnover markers used in each study were not reported in the publication, and therefore it is unclear how many of the studies used the bone turnover markers being evaluated in our review. Thirdly, the analyses were restricted to changes in bone turnover markers at 12 months (three studies used data from 6 months where 12-month data were not available); our review was particularly interested in the changes at 3 (and secondarily at 1 and 6) months, as it is the early detection of treatment non-responders that makes bone turnover marker monitoring a potentially useful strategy. In addition, sensitivity analyses conducted by the review authors showed that the overall statistically significant association between resorption and bone formation turnover markers, and the risk of fracture was lost when the largest trial that administered raloxifene was removed from the analysis (p = 0.09), and for resorption bone turnover markers when any one of three alendronate trials were removed, two of which were the second and third largest trials. Therefore, it is uncertain whether there are specific population–treatment–bone turnover marker combinations that produce significant associations with fracture, or whether this is a sample size issue and that most studies were underpowered to detect the association.

A second review that addressed a broader question, including studies of patients with and without osteoporosis, conducted a separate analysis investigating the association between bone turnover markers and fracture outcomes in osteoporotic patients receiving antiresorptive therapy. 47 This section of the review included seven publications; one of these reported only on the results of serum osteocalcin, which is not being evaluated in our review. All six of the relevant studies included in the review by Vasikaran et al. 47 were included in our review. 40,132,137,139,151,161 The same data (results from regression analyses) were extracted for both reviews from one study. 151 However, the data extracted from the other studies varied between the two reviews, making comparisons of the results of the two reviews more difficult. Vasikaran et al. extracted the percentage treatment effect explained (TEE) for two of the studies,132,137 an outcome not considered in our review; we extracted regression data from both of these studies,132,137 whereas Vasikaran et al. extracted regression data from only one. 132 Predictive data in the form of AUCs were extracted by both reviews from one study; we also extracted regression data. 161 Both reviews extracted regression data from a further study; we extracted data only for sBALP, as this is the only bone turnover marker that had results reported for the subgroup of patients with osteoporosis. The analyses of the other bone turnover markers included patients with osteopenia who were not being considered in our review. 139 For the final study included in the review by Vasikaran et al. , we extracted data on the correlations between bone turnover markers (sP1NP and sCTX) and fracture outcomes, whereas Vasikaran et al. 40 extracted the relative risk/odds ratio of fracture for the treatment compared with placebo in the subgroup of patients who had been tested using sP1NP. In addition, our review included results from regression analyses from three studies,154,155,163 and other predictive data from three studies,142,156,163 that were not included in the review by Vasikaran et al. The review by Vasikaran et al. did not consider results from correlation analyses;47 we included data on correlations between bone turnover markers and fracture from two studies. 40,42

Vasikaran et al. concluded that the evidence available relating to the association between bone turnover marker changes and fracture risk reduction is promising, but further studies are needed to address sample handling, the timing of bone turnover marker testing, and the statistical methods used; an assessment of whether or not the final bone turnover marker level is a guide to fracture risk was suggested. Vasikaran et al. also included an assessment of the impact of bone turnover monitoring on adherence to osteoporosis medication. Vasikaran et al. included two studies in this assessment, one of which was included in our review56 and the other excluded as it was conducted on patients with osteopenia rather than osteoporosis. 181 We included a further four RCTs that evaluated the impact of feedback of bone turnover marker results in patients with osteoporosis. 133,143,148,149 Vasikaran et al. 47 concluded that both of the studies included in their review showed that the feedback of results from a positive result encouraged adherence, and a negative result discouraged it. As already mentioned, this response may be surprising; however, the two studies were conducted in populations that were not representative of a general osteoporotic population or of a population of osteoporotic post-menopausal women (one conducted in older women aged 65–80 years56 and the other in patients who had not yet developed osteoporosis181). Therefore, it is unclear whether or not this response would be representative of the osteoporotic population seen in clinical practice. In addition, the inclusion of additional RCTs in our review casts doubt on the impact of feedback of bone turnover marker test results on compliance and adherence. However, as stated above in Limitations of the available evidence, these studies are unlikely to be reflective of clinical practice.

A third review evaluated the use of bone turnover markers for the monitoring of osteoporosis therapy in post-menopausal women. 6 This review included 48 studies, most of which were excluded from our review because they evaluated serum osteocalcin, tartrate-resistant acid phosphatase 5b (TRAPc5b) or P1CP; recruited patients who did not have osteoporosis or mixed populations where results for those with osteoporosis were not reported separately; made comparisons between baseline bone turnover marker test results and BMD or fracture outcomes only; had fewer than 20 treated patients in their analyses; or received an osteoporotic treatment that was not one of those being considered in this review. Of the 48 studies included in Funck-Brentano et al. ,6 nine assessed the correlation between changes in bone turnover markers and either BMD or fracture risk in at least 20 patients with osteoporosis receiving one of the treatments being considered in this review, and were therefore included in our review. 132,139,140,151,157,160,161,163,165 We included a further 17 studies that reported correlations between bone turnover markers and BMD or fracture risk in patients with osteoporosis receiving one of the treatment regimens of interest,14,38,40,42–44,58,99,106,131,135,145,146,153,155,159,160 two that reported correlations with results from biopsy,150,152 and one that reported correlations with vertebral strength index. 136 Funk-Brentano et al. 6 did not include data from regression analyses or other types of predictive data. The review by Funk-Brentano et al. concluded that short-term changes in bone turnover markers were significantly correlated with BMD variation, but there was no evidence that they predict benefit on fracture risk at the individual level. A high proportion of the non-significant correlations were in those studies that also met the inclusion criteria for our review and, therefore, if based solely on those nine studies, the conclusions drawn are unlikely to have been so strong.

Country

The country setting for four studies was North America:167,182–184 three in the USA167,182,183 and one in Canada. 184 The remaining eight papers were conducted for western European countries. 185–192 Five studies were conducted by the same authors in Belgium,188–192 (these are included separately because there were slight variations in the methods used), one was undertaken in the UK,187 one in Sweden,185 and one in the Netherlands and the UK. 186

Type of model

Nine studies were based on individual patient-level Markov (or state transition) microsimulation models. 182,185–192 Five of these were by the same group of Belgian investigators. 188–192 Two studies used a combined decision tree and Markov modelling approach,167,184 and one used a Markov cohort model alone. 183

Study objective

The objectives of the studies were to estimate the cost-effectiveness of different treatments for osteoporosis;186–188,190 to model adherence or persistence to treatments of osteoporosis;182,183,185,189,191,192 to model the cost-effectiveness of a multifaceted intervention to improve osteoporotic care by encouraging patients to come forward and receive treatment after fracture of the wrist;184 and to evaluate different follow-up regimes for antiresorptive treatments for post-menopausal women with osteoporosis. 167 The latter assessed only health benefits, measured in QALYs, and did not include costs.

Adherence definition

Nine studies used the term ‘adherence’. 167,185–192 Ten studies used the term ‘compliance’. 167,182,185–192 Nine studies used the term ‘persistence’. 183–185,187–192

In Chapter 3 (see Assessment of clinical effectiveness) we adopted standard definitions for adherence, compliance and persistence owing to variation in their use. Persistence is the time until discontinuation of medication, compliance is the proportion of medication taken, and adherence is a combination of persistence and compliance. The remaining sections will use these definitions, rather than those specified by the study authors, to ensure consistency. However, where the terms used differ from these definitions, this will be highlighted.

How compliance, persistence and adherence were modelled

How treatment management was modelled

Only one study incorporated a change of treatment into their model structure. 167 The model allowed for switching from a bisphosphonate to a second-line treatment, such as teriparatide, if results of a bone turnover marker test during follow-up led to the conclusion that compliance or response to treatment was inadequate. The authors appeared to have assumed that bone turnover markers were able to correctly identify responders and non-responders to treatment, and so test accuracy data were not included in their model.

Summary of approaches for modelling adherence and treatment change

Only one study modelled treatment change as a result of bone turnover marker monitoring during treatment, and this study did not include test accuracy data. The other studies only modelled aspects of adherence to treatment; compliance was the most commonly modelled variable, with persistence and compliance being distinguished in some studies.

Compliance was considered a binary variable in most studies that modelled compliance. That would allow a relative risk of compliance to be utilised in a model of monitoring strategies that gave test feedback to patients to encourage adherence. The cut-off point for determining compliance/non-compliance was 80% in several studies, chosen to be consistent with that reported in the literature, although it is not clear how that cut-off point was derived. It is likely that there will be more evidence on the risk of fracture for compliant and non-compliant patients defined with that cut-off point than for other cut-off points. Potential confounders should be accounted for when estimating the relative risk of fracture between compliant and non-compliant patients.

Real-life persistence rates were estimated, where possible, based on observational data from databases, and hazard rates could be estimated if survival models fit the data. Six studies modelled compliance, non-compliance and persistence separately, and therefore incorporated the different aspects of adherence.

Including an estimate of primary non-adherence where patients are prescribed treatment is a useful approach if there are a significant proportion of patients for whom that applies. These primary non-adherent patients would not be captured in a survival model.

Current practice

Based on discussion with a clinical advisor the timings of GP visits and DXA tests in current practice are presented in Figure 3. In this monitoring strategy decisions on treatment change as a result of a poor DXA test result occur at either 2 or 3 years, depending on whether or not a fracture occurred within the first 2 years. Decisions may also be influenced by the occurrence of side effects and patient statement on compliance to treatment: if a patient discontinues treatment because of side effects, then a different class of treatment may be prescribed. If there is a poor DXA test result and the patient claims to be compliant to treatment then a new class of treatment would be prescribed; however, if the patient admits to non-compliance then an alternative dose or mode of delivery of the same treatment may be appropriate, where one is available.

FIGURE 3.

Current and feasible monitoring strategies for response to treatment. a, GP visit: enquire about patient adverse events and adherence. The numbers in circles mark different stages of the monitoring strategy. Tx, treatment; +ve, positive; –ve, negative.

Alternative monitoring strategy using bone turnover marker tests

We are not aware of any published guidelines on the use of bone turnover markers in the UK for monitoring response to treatment for osteoporosis which could be used to inform alternative monitoring strategies incorporating bone turnover markers. Therefore, based on discussion with a clinical advisor, a feasible monitoring strategy using bone turnover marker tests was constructed and is presented in Figure 3. The purpose of this figure is to allow the presentation and discussion around the type of data that might be required to evaluate similar strategies.

The numbers in circles indicate four different stages to the strategy. In this hypothetical strategy, patients are started on treatment at stage 1. At stage 2, after 3–6 months, depending on the type of test and treatment, a first bone turnover marker test is done. The results are fed back to the patients in order to improve, or encourage continued, adherence. At stage 3, a second bone turnover marker test is done only for patients who received a negative result on the first test. This would be done at 6–12 months depending on the type of test and treatment. Using a predefined decision rule after a second negative test result, a treatment change would be recommended. The clinician may recommend a new treatment class to everyone with two negative test results. Alternatively, if the clinician has assessed patient adherence, he or she may recommend a new treatment class only to adherent patients and a different mode of delivery and/or frequency of the same treatment class to non-adherent patients where one is available.

Given the lack of evidence, it has not been possible to determine clinically relevant strategies for evaluation. The timing of the bone turnover marker tests may vary by the type of test used and the type of treatment administered that the test is monitoring a response to. The optimal timing of a bone turnover marker test needs to be determined by, for example, studies that evaluate the S/N ratio at different time points; there may be a trade-off between a greater S/N ratio and delay to response monitoring. While some S/N ratio data were identified in the review, these were inadequate to address the issue of optimal timing and combination of treatment and tests.

Different cost-effectiveness analysis methods

Based of the review of modelling methods, the two most common models employed to address this type of decision question are:

1. individual patient-level Markov simulation

2. decision tree and Markov model combined.

Both of these modelling methods require the combination of evidence on the risk of fracture given different risk factors, the effect of feedback on response rates and bone turnover marker test accuracy data.

Modelling data requirements for hypothetical strategy

Test accuracy

When diagnostic or prognostic tests are included in a decision-analytic model, test accuracy needs to be considered and the clinical outcomes for patients with correct (true-positive and true-negative) and incorrect (false-positive and false-negative) test results need to be incorporated. 193 Test accuracy will be an important factor in any decision model used to investigate the current decision problem as test errors may result in an incorrect treatment pathway being followed. A non-responder incorrectly identified as a responder (false-positive) would likely remain on their current treatment, and would not benefit from a change in treatment class. Conversely, a responder who was incorrectly classified as a non-responder (false-negative) may be prescribed a different drug unnecessarily. Where patients have true-positive (a responder identified as a responder) or true-negative (a non-responder identified as a non-responder) results, they will benefit from appropriate treatment management choices based on the correct test results.

For this decision problem, test accuracy will need to be determined for each of the tests within the pathway; in the example, test accuracy for bone turnover markers would be needed for stages 2 and 3, and test accuracy for DXA at stage 4. The major limitation for establishing the accuracy of bone turnover marker tests in identifying treatment non-responders is the lack of a gold standard.

There are a range of reference standards available, but each has its limitations. Bone biopsy could be seen as the ideal reference standard. However, given the invasiveness of the test and the high risk of complications it would be considered unethical to conduct studies where all patients would undergo biopsy to confirm treatment response. The next most complete reference standard would be the use of BMD with clinical follow-up to determine fracture incidence to supplement the BMD results. Where BMD indicated a response to treatment sufficient enough to reduce a patient’s fracture risk but the patient went on to have a fragility fracture, the patient could be reclassified as a non-responder. This would mean that if the bone turnover marker was negative in this patient, it would be reclassified from a false-negative to a true-negative. Whether a patient who did not have a response in BMD but did not go on to have a fragility fracture within a certain time frame could be reclassified from a non-responder to a responder is less certain.

Less perfect reference standards are the incidence of fracture alone and BMD alone, the latter being the poorest available reference standard. The lack of suitability of BMD as a surrogate for fracture risk against which we can judge the accuracy of bone turnover tests is discussed in Chapter 3 (see Limitations of the available evidence). An alternative method for estimating the test error rates for a test at a particular time point would be to use the intraindividual variation data that is used to calculate S/N ratios, but these estimates may be inferior to test accuracy data with an appropriate comparator. 194 When establishing the accuracy of DXA, the only reference standards are bone biopsy and fracture incidence, and the limitations of these have already been discussed.

Once the proportion of true-positives, true-negatives, false-positives and false-negatives had been established, fracture risk for each of these groups would need to be estimated.

Measures of adherence

As discussed previously, adherence has two aspects: compliance and persistence – both of these are important causes of non-response to osteoporosis treatment. Clearly, adherence to treatment will affect a patient’s response to treatment, at all stages of the treatment pathway. Therefore, any attempt to model adherence would apply to each stage in Figure 3. Several studies modelling the effectiveness of treatment have modelled compliance and persistence. Our decision problem involves the ability of tests to identify treatment responders and non-responders rather than to establish treatment effectiveness. For patients identified by a test as a non-responder (bone turnover markers at stage 2 and 3; DXA at stage 4), treatment non-response could have a number of causes, including: non-compliance; non-persistence; an underlying, untreated cause of the osteoporosis; an inability to absorb the drug; and/or test error. The issue of test accuracy has been addressed in earlier in this chapter (see Test accuracy).

Including measures of adherence in the model would be of use only if there would be different treatment pathways for adherent and non-adherent non-responders, and/or there was evidence that feedback of bone turnover marker test results increased adherence. It is possible that there would be different treatment pathways for adherent and non-adherent non-responders for those who were treated with first-line oral bisphosphonates, as true non-responders are unlikely to respond to a different dosing frequency or mode of administration, but patients who are non-compliant with a daily or weekly oral dose may become compliant with a monthly oral dose or intravenous administration. Therefore, the importance of modelling adherence may differ depending on the choice of first-line treatment. Adherence rates may improve by providing feedback from bone turnover marker tests to the patients, but there is no strong evidence for this effect.

The ideal method to model adherence would be to include adherence from the start of treatment. A primary non-adherence estimate could be modelled, which is a percentage of patients who are recommended treatment but do not start treatment. This could be estimated from the number of prescriptions that are not filled; however, this would underestimate the value, as some patients may have a prescription filled and not commence treatment. To include measures of adherence in subsequent stages of a model, the proportions of non-compliant non-responders and compliant non-responders would need to be established. There are limitations with the methods available for determining these estimates. It is likely that the most common means of identifying these data would be based on patient reporting, which is prone to bias. Measures such as the MPR rate are also problematic, as patients may have their prescriptions filled but not take all of the medication prescribed. In the case of bisphosphonates, patients may take their tablets but not comply with the strict regimen required, adding a further complication to the establishment of a rate of compliance. Persistence could be estimated by reviewing the attrition rates from RCTs evaluating treatment effectiveness; this would likely underestimate the rate. Such a review could be supplemented with data from observational studies that use sources such as GP databases. 183 Any estimate of compliance, persistence or overall adherence would be highly uncertain, making it extremely important that it is incorporated into the modelling in a manner which allows appropriate sensitivity analysis to be undertaken.

In addition to the adherence data, the fracture risk for these population groups would also need to be determined. If it can be assumed that the relative risk of fracture given treatment versus no treatment is based on a compliant population, then the relative risk of fracture given good compliance versus poor compliance enables the calculation of the risk of fracture for poorly compliant patients to treatment. There would be an interaction between compliance and response. The optimal cut-off point is uncertain and needs to be established prior to modelling.

One aspect of adherence deserves separate consideration, namely discontinuation rates due to treatment-related adverse events; this is a particular issue for bisphosphonates. Gastrointestinal upsets are common with oral bisphosphonate therapy, but an additional, serious, side effect would have to be incorporated in the model; osteonecrosis of the jaw (ONJ). This is a side effect specific to the use of bisphosphonates, and can lead to the need for costly dental surgery. Although this is a rare adverse event, there is some evidence that the risk of developing ONJ is much higher with intravenous bisphosphonates than with oral preparations. 195 The rate of ONJ may be investigated initially by a review of long-term RCTs that evaluate the clinical effectiveness and safety of bisphosphonate therapy; the review could supplement these data with the available prospective observational and retrospective studies.

Summary of modelling approaches and available evidence

The key part of a cost-effectiveness analysis of bone turnover marker tests for monitoring response to treatment for osteoporosis is accounting for test accuracy, the prognostic outcomes for true- and false-positive and negative test outcomes, and the effect of bone turnover marker feedback on patient adherence to treatment. This affects both who benefits from bone turnover marker feedback and who benefits from treatment change. These data were either absent completely in the evidence identified in this review, were insufficient given the different tests and treatments, or were applicable to populations with unrealistic adherence rates for clinical practice.

Compliance and persistence are commonly modelled separately. The effect of test feedback on adherence is often reported as the effect on compliance or persistence. Given the relationship between the two and bone turnover marker levels, this does require assumptions to simplify the model.

This section has focused on those gaps in the evidence that would be essential to any future decision-analytic model but may be difficult to establish estimates for. Other variables, such as estimates of treatment effectiveness and data on utilities or other relevant QoL outcomes, would be required. Information would also be required on resource use, including the cost of tests, GP/clinic visits, treatment costs, and the costs associated with treating serious side effects such as ONJ.

Strengths

We conducted a rigorous systematic review which addressed clear research questions using pre-defined inclusion criteria. Comprehensive literature searches were conducted to locate all relevant published and unpublished studies without any language restrictions, thereby minimising both publication and language biases. Efforts were also made to identify additional studies by hand-searching the reference lists of relevant publications. We are therefore confident that we have been able to include all the relevant studies in the evaluation of the clinical effectiveness, predictive accuracy, reliability and reproducibility, and cost-effectiveness of bone turnover markers for monitoring treatment response that met our inclusion criteria. Each stage of the review was conducted in duplicate, reducing the risk of error and bias. Owing to the high degree of clinical heterogeneity across the studies in terms of patient populations, treatment regimens and duration of follow-up, the clinical data were appropriately synthesised using a narrative approach.

As we were unable to conduct a cost-effectiveness analysis, we conducted additional assessments of the economic models used to address similar decision problems in order to inform future decision-analytic modelling that may be undertaken to address the decision problem investigated in this review.

Limitations

The main limitations of the review of clinical effectiveness were the lack of data on the effectiveness of monitoring regimens and the poor quality of the data that were available; all of the included studies were judged as low quality. This lack of robust data relating to the comparative predictive values of the different bone turnover marker tests for monitoring the response to a specific osteoporosis treatment precludes the possibility of determining which bone turnover marker, if any, is superior in terms of its ability to identify treatment non-responders and predict future fracture risk when used for monitoring osteoporosis treatment.

Despite the moderate amount of correlation data identified for the evaluation of the relationship between changes in bone turnover markers and BMD, it remains unclear whether or not these findings can also be utilised as an indicator of the association with the outcome of fracture risk. As discussed in Chapter 3 (see Limitations of the available evidence), BMD is a poor surrogate for fracture risk. 47,70,177,178 Treatment-induced changes in BMD account for a limited proportion of the observed reduction in fracture risk;91,178 therefore, using BMD as a surrogate for assessing the accuracy of bone turnover markers for identifying patients on treatment who remain at risk of fracture is inappropriate.

Although the incidence of fracture is a more robust outcome measure for fracture risk,70 there was a paucity of data correlating the changes in bone turnover marker levels to this outcome. There was also a paucity of studies that adjusted the associations between changes in bone turnover markers and either changes in BMD, fracture, or other measures such as spinal strength indices or results of biopsy, for confounders in multiple regression analyses. As discussed in Chapter 3, where strong associations are identified between bone turnover markers and fracture or other measures in correlation analyses, there is no evidence that these would produce significant associations when other important confounding variables are included in regression analyses. Alternatively, where there is a non-significant correlation between change in bone turnover markers and one of these outcomes, the association may change and become significant when other predictive factors are included in a multivariate regression analysis. Either way, assessing the association between changes in bone turnover markers and any outcome in isolation, without adjusting for confounders within a multivariate regression analysis, is unlikely to reflect the true association between these variables within a patient; it was impossible to draw any conclusion as to whether or not these bone turnover markers were able to identify treatment non-responders and predict fracture risk independently of BMD measurements in patients receiving osteoporosis treatment.

It should be noted that the results from the studies utilising correlation and regression analyses were inconsistent. This may be due to the considerable clinical heterogeneity across the included studies in terms of the definitions used to identify those with osteoporosis for inclusion in the studies, patient populations recruited, the treatment regimens administered, and the type and timing of the tests being evaluated. Most of the included studies had small sample sizes, resulting in low statistical power to detect a significant association.

The analysis of test reliability and reproducibility in women being treated for osteoporosis was limited; very few studies reported these data. Test reliability and reproducibility is most commonly measured in either healthy individuals or control subjects who are not receiving treatment. Although this provides baseline intraindividual and interindividual CVs and a S/N ratio, it does not inform us as to how these tests perform in women receiving treatment. Whether receiving treatment would increase or decrease these measures of variability is unknown.

The lack of any published decision-analytic models investigating the decision problem being addressed in this review, and the lack of evidence on the effectiveness of monitoring treatment response using bone turnover markers, meant that the cost-effectiveness of different monitoring strategies could not be investigated at this time. In order to construct such a model, the large gaps in the current evidence base will need to be filled. We identified these gaps and the data required for a future cost effectiveness analysis of different monitoring strategies and discussed how these data could be obtained in Chapter 4 (see Economic model). The uncertainties that remain and research priorities are highlighted in the section below and in Chapter 6 (see Suggested research priorities), respectively.

Suggested research priorities

In order to determine whether or not bone turnover marker monitoring improves treatment management decisions and ultimately impacts on patient outcomes in terms of reduced incidence of fracture, RCTs are required. The predictive accuracy of bone turnover markers for future fracture outcomes in patients receiving osteoporosis treatment could be investigated using prospective, long-term observational studies with large sample sizes. However, as the nature of bone turnover marker response is determined by the mechanism of action of the drug, any future research needs to identify the most appropriate treatment–test combinations in order to identify whether or not the predictive accuracy of a particular bone turnover marker can be maximised to aid treatment management decisions. All future studies should adopt a standardised definition of osteoporosis, such as the WHO criteria. 7

There are potentially a large number of patient population–treatment–test combinations; therefore, conducting RCTs or even larger observational studies to establish the effectiveness for all of these combinations would not be feasible. Therefore, it is likely that identifying the most promising combinations would be beneficial in order to ensure that the most promising are evaluated in the more costly and time-consuming studies such as RCTs. This would include not only identifying which bone turnover test best identifies non-responders to specific treatments, but also the optimal timing of these tests. This may feasibly be achieved through the use of a patient registry, where relevant pre-specified standardised data would be collected and trends both in the use of different tests and in outcomes related to test–treatment combinations could be identified. However, without more widespread use of these tests in clinical practice, the usefulness of such a registry would be questionable. If a registry is not established, a survey of the current use of bone turnover markers may be useful. An alternative to establishing a registry to identify promising test–treatment combinations might be to undertake smaller, less costly studies that identify those treatment–bone turnover marker combinations that have the lowest inter- and intraindividual patient variability (and therefore a higher S/N ratio). These smaller feasibility studies could be used to help to identify the most promising combinations for future more costly research. Such studies would be of use only when conducted in the context of establishing inter- and intraindividual patient variability in bone turnover markers and would not be useful if the decision was made to either establish a patient registry, or if there was considered to be sufficient experience within the clinical setting already available to identify treatment-test combinations that could be evaluated in effectiveness studies.

To further limit the number of RCTs and other costly and time-consuming research, there are some areas of uncertainty that could be classified as low priority. These could be investigated initially within a decision-analytic framework, once sufficient evidence becomes available on monitoring effectiveness and the predictive value of bone turnover markers. By using this strategy, those areas of uncertainty that are key drivers of cost-effectiveness can be identified. Further research can then focus on investigating the impact of those areas of uncertainty that most influence the cost-effectiveness of the monitoring regimens, rather than being conducted to inform those estimates and assumptions to which the cost-effectiveness analysis is robust.

We consider that the research priority is to identify the most promising treatment–test combinations. This can be achieved either by conducting small variability studies or, if more widespread use is feasible, by initiating a patient registry to collect data. The former would be quicker, easier and less costly, but the quality of the data would be poorer and likely to be collected in small selected populations. This would mean that the results may not reflect the broader population seen in clinical practice, and the choices made as to which treatment–test combinations to evaluate in a RCT may be inappropriate. Once the most promising treatment–test combinations have been identified, well-designed RCTs can be conducted to evaluate the clinical effectiveness of those monitoring regimens; this would include measuring outcomes such as the proportion of non-responders, adherence rates, treatment management decisions, and fracture outcome. Data from these RCTs along with other sources can then be included in a decision-analytic model in order to investigate cost-effectiveness.

Search terms

1. (P1NP or PINP).ti,ab. (593)

2. (procollagen adj3 propeptide).ti,ab. (429)

3. (procollagen adj3 peptide).ti,ab. (557)

4. (collagen adj3 propeptide).ti,ab. (191)

5. (BSAP or BALP or BAP).ti,ab. and (bone or bones or biomarker$ or biological marker$).af. (1090)

6. bone specific alkaline phosphatase$.ti,ab. (1196)

7. bone alkaline phosphatase$.ti,ab. (1153)

8. bone source alkaline phosphatase$.ti,ab. (1)

9. (CTX or NTX).ti,ab. and (bone or bones or biomarker$ or biological marker$).af. (2032)

10. crosslaps.ti,ab. (296)

11. (telopeptide$ adj3 collagen).ti,ab. (668)

12. (n-telopeptide$ adj3 collagen).ti,ab. (214)

13. (c-telopeptide$ adj3 collagen).ti,ab. (184)

14. bone turnover marker$.ti,ab. (1333)

15. bone metabolic marker$.ti,ab. (190)

16. Biological Markers/ and exp "Bone and Bones"/ (3520)

17. ((biochemical marker$ or biomarker$ or biological marker$) adj2 bone$).ti,ab. (2261)

18. bone marker$.ti,ab. (1437)

19. or/1-18 (10656)

Line 19 captures bone turnover marker terms

1. 20. exp osteoporosis/ (39,832)

2. 21. osteoporo$.ti,ab. (43,304)

3. 22. 20 or 21 (55,772)

Line 22 captures osteoporosis terms

1. 23. diphosphonates/ or alendronate/ or clodronic acid/ or etidronic acid/ (15,156)

2. 24. (bisphosphonate$ or diphosphonate$).af. (17,117)

3. 25. (alendronate or alendronic acid or fosamax or actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan or osteofar or osteoform or osteoral or osteotrat or recalfe or terost or aldrox or fosval or holadren or leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos or fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosal).af. (4270)

4. 26. (clodronate or clodronic acid or bonefos or loron or clodron or lodronat or ascredar or ostac or clastoban or lytos or clasteon or climaclod or clodeosten or clody or difosfonal or dolkin or moticlod or niklod or ossiten or osteonorm or osteostab or mebonat).af. (1827)

5. 27. (etidronate or etidronic acid or EHDP or didronel or difosfen or detidron or osteodidronel or bonemass or osteotop or didronate or diphos or anfozan or biotredine or dralen or etidron or etiplus or feminoflex or maxibral or oflocin or osfo or ostedron or osteodrug or osteoton or ostogene or somaflex or sterodome or sviroxit or tilferan or dronate-os or etidrate or osteum).af. (7341)

6. 28. (ibandronic acid or ibandronate or boniva or bondronat or bonviva or bandrobon or bonat).af. (766)

7. 29. (pamidronate or pamidronic acid or aredia or ADP sodium or aminomux or pamdosa or pandrat or pamisol or pamitor or osteopam or aredronet or pamidria or pamifos or pamidran or linoten or xinsidona).af. (2470)

8. 30. (risedronate or risedronic acid or actonel or ductonar or ribastamin or ridron or risedon or optinate or risofos or benet or acrel).af. (2039)

9. 31. (zoledronic acid or zoledronate or zometa or zomera or aclasta or reclast or zoldria).af. (2319)

10. 32. (tiludronic acid or tiludronate or skelid).af. (139)

11. 33. (neridronic acid or neridronate or nerixia).af. (63)

12. 34. (olpadronic acid or olpadronate).af. (73)

13. 35. (cimadronic acid or cimadronate).af. (92)

14. 36. (piridronic acid or piridronate).af. (0)

15. 37. (icandronic acid or icandronate or bisphonal).af. (1)

16. 38. (minodronic acid or minodronate).af. (57)

17. 39. (raloxifene or biofem or ciclotran or evista or ketidin or loxifen or raxeto or optruma or bonmax or estroact or ralista or celvista).af. (2870)

18. 40. Raloxifene/ (2063)

19. 41. (strontium ranelate or protelos).af. (421)

20. 42. Teriparatide/ (1167)

21. 43. (denosumab or prolia or xgeva).af. (504)

22. 44. (teriparatide or forteo or forsteo).ti,ab. (587)

23. 45. (treatment$ or treat or treated or treats).ti,ab. (3,156,022)

24. 46. dt.fs. (1,504,349)

25. 47. or/23-46 (3,906,922)

Line 47 captures intervention terms

1. 48. 19 and 22 and 47 (1879)

Line 48 combines bone turnover marker, osteoporosis and intervention terms

1. 49. exp animals/ not humans/ (3,667,503)

2. 50. 48 not 49 (1733)

Line 50 excludes animal-only studies

Search terms

S48 S21 and S24 and S47 (155)

S47 S25 or S26 or S27 or S28 or S29 or S30 or S31 or S32 or S33 or S34 or S35 or S36 or S37 or S38 or S39 or S40 or S41 or S42 or S43 or S44 or S45 or S46 (330,229)

S46 treatment* or treat or treated or treats (327,921)

S45 teriparatide or forteo or forsteo (102)

S44 denosumab or prolia or xgeva (108)

S43 "strontium ranelate" or protelos (41)

S42 (MH "Raloxifene") (441)

S41 raloxifene or biofem or ciclotran or evista or ketidin or loxifen or raxeto or optruma or bonmax or estroact or ralista or celvista (636)

S40 "minodronic acid" or minodronate (1)

S39 "icandronic acid" or icandronate or bisphonal (0)

S38 "piridronic acid" or piridronate (0)

S37 "cimadronic acid" or cimadronate (0)

S36 "olpadronic acid" or olpadronate (3)

S35 "neridronic acid" or neridronate or nerixia (4)

S34 "tiludronic acid" or tiludronate or skelid (4)

S33 "zoledronic acid" or zoledronate or zometa or zomera or aclasta or reclast or zoldria (367)

S32 risedronate or "risedronic acid" or actonel or ductonar or ribastamin or ridron or risedon or optinate or risofos or benet or acrel (238)

S31 pamidronate or "pamidronic acid" or aredia or "ADP sodium" or aminomux or pamdosa or pandrat or pamisol or pamitor or osteopam or aredronet or pamidria or pamifos or pamidran or linoten or xinsidona (223)

S30 "ibandronic acid" or ibandronate or boniva or bondronat or bonviva or bandrobon or bonat (109)

S29 etidronate or "etidronic acid" or EHDP or didronel or difosfen or detidron or osteodidronel or bonemass or osteotop or didronate or diphos or anfozan or biotredine or dralen or etidron or etiplus or feminoflex or maxibral or oflocin or osfo or ostedron or osteodrug or osteoton or ostogene or somaflex or sterodome or sviroxit or tilferan or dronate-os or etidrate or osteum (172)

S28 clodronate or "clodronic acid" or bonefos or loron or clodron or lodronat or ascredar or ostac or clastoban or lytos or clasteon or climaclod or clodeosten or clody or difosfonal or dolkin or moticlod or niklod or ossiten or osteonorm or osteostab or mebonat (85)

S27 alendronate or "alendronic acid" or fosamax or actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan or osteofar or osteoform or osteoral or osteotrat or recalfe or terost or aldrox or fosval or holadren or leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos or fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosalalendronate or "alendronic acid" or fosamax or actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan or osteofar or osteoform or osteoral or osteotrat or recalfe or terost or aldrox or fosval or holadren or leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos or fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosal (765)

S26 bisphosphonate* or diphosphonate* (3182)

S25 (MH "Diphosphonates") or (MH "alendronate") (3160)

S24 S22 or S23 (10,599)

S23 osteoporo* (10,505)

S22 (MH "Osteoporosis+") (8995)

S21 S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 or S10 or S11 or S12 or S13 or S14 or S15 or S16 or S17 or S20 (772)

S20 S18 and S19 (164)

S19 (MH "Bone and Bones") (2891)

S18 (MH "Biological Markers") (12,859)

S17 "bone marker*" (108)

S16 ("biochemical marker*" or biomarker* or "biological marker*") n2 bone* (224)

S15 "bone metabolic marker*" (4)

S14 "bone turnover marker*" (129)

S13 c-telopeptide* n3 collagen (44)

S12 n-telopeptide* n3 collagen (42)

S11 telopeptide* n3 collagen (168)

S10 crosslaps (16)

S9 (CTX or NTX) and (bone or bones or biomarker* or biological marker*) (143)

S8 "bone source alkaline phosphatase*" (0)

S7 "bone alkaline phosphatase*" (57)

S6 "bone specific alkaline phosphatase*" (101)

S5 (BSAP or BALP or BAP) and (bone or bones or biomarker* or biological marker*) (61)

S4 collagen n3 propeptide (37)

S3 procollagen n3 peptide (18)

S2 procollagen n3 propeptide (45)

S1 P1NP or PINP (25)

Search terms

osteoporosis AND ("bone markers" OR "bone turnover markers")

Search terms

#1 P1NP or PINP (133)

#2 procollagen near/3 propeptide (68)

#3 procollagen near/3 peptide (140)

#4 collagen near/3 propeptide (26)

#5 (BSAP or BALP or BAP) and (bone or bones or biomarker* or biological marker*) (189)

#6 "bone specific alkaline phosphatase*" (385)

#7 "bone alkaline phosphatase*" (189)

#8 "bone source alkaline phosphatase*" (0)

#9 (CTX or NTX) and (bone or bones or biomarker* or biological marker*) (464)

#10 crosslaps (55)

#11 telopeptide* near/3 collagen (152)

#12 "n-telopeptide*" near/3 collagen (37)

#13 "c-telopeptide*" near/3 collagen (45)

#14 "bone turnover marker*" (23)

#15 "bone metabolic marker*" (3)

#16 ("biochemical marker*" or biomarker* or "biological marker*") near/2 bone* (108)

#17 "bone marker*" (55)

#18 MeSH descriptor Biological Markers, this term only (6199)

#19 MeSH descriptor Bone and Bones, this term only (1156)

#20 (#18 AND #19) (258)

#21 (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #20) (1370)

#22 MeSH descriptor Osteoporosis explode all trees (2750)

#23 osteoporo* (4841)

#24 (#22 OR #23) (4841)

#25 MeSH descriptor Diphosphonates, this term only (738)

#26 MeSH descriptor Alendronate, this term only (498)

#27 MeSH descriptor Clodronic Acid, this term only (166)

#28 MeSH descriptor Etidronic Acid, this term only (370)

#29 bisphosphonate* or diphosphonate* (1336)

#30 alendronate or "alendronic acid" or fosamax or actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan or osteofar or osteoform or osteoral (775)

#31 osteotrat or recalfe or terost or aldrox or fosval or holadren or leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos or fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosalalendronate or "alendronic acid" or fosamax (107)

#32 actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan (51)

#33 (osteofar or osteoform or osteoral or osteotrat or recalfe or terost or aldrox or fosval or holadren) (2)

#34 leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos (0)

#35 fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosal (22)

#36 clodronate or "clodronic acid" or bonefos or loron or clodron or lodronat or ascredar or ostac or clastoban or lytos or clasteon or climaclod or clodeosten or clody or difosfonal or dolkin or moticlod or niklod or ossiten or osteonorm or osteostab or mebonat (290)

#37 etidronate or "etidronic acid" or EHDP or didronel or difosfen or detidron or osteodidronel or bonemass or osteotop or didronate or diphos or anfozan or biotredine or dralen or etidron or etiplus or feminoflex or maxibral or oflocin or osfo or ostedron or osteodrug or osteoton or ostogene or somaflex or sterodome or sviroxit or tilferan or dronate-os or etidrate or osteum (535)

#38 "ibandronic acid" or ibandronate or boniva or bondronat or bonviva or bandrobon or bonat (177)

#39 pamidronate or "pamidronic acid" or aredia or "ADP sodium" or aminomux or pamdosa or pandrat or pamisol or pamitor or osteopam or aredronet or pamidria or pamifos or pamidran or linoten or xinsidona (422)

#40 risedronate or "risedronic acid" or actonel or ductonar or ribastamin or ridron or risedon or optinate or risofos or benet or acrel (350)

#41 "zoledronic acid" or zoledronate or zometa or zomera or aclasta or reclast or zoldria (373)

#42 "tiludronic acid" or tiludronate or skelid (29)

#43 "neridronic acid" or neridronate or nerixia (23)

#44 "olpadronic acid" or olpadronate (12)

#45 "cimadronic acid" or cimadronate (0)

#46 "piridronic acid" or piridronate (0)

#47 "icandronic acid" or icandronate or bisphonal (0)

#48 "minodronic acid" or minodronate (2)

#49 raloxifene or biofem or ciclotran or evista or ketidin or loxifen or raxeto or optruma or bonmax or estroact or ralista or celvista (578)

#50 MeSH descriptor Raloxifene, this term only (373)

#51 "strontium ranelate" or protelos (59)

#52 denosumab or prolia or xgeva (64)

#53 MeSH descriptor Teriparatide, this term only (126)

#54 treatment* or treat or treated or treats (330,609)

#55 (#25 OR #26 OR #27 OR #28 OR #29 OR #30 OR #31 OR #32 OR #33 OR #34 OR #35 OR #36 OR #37 OR #38 OR #39 OR #40) (2500)

#56 (#41 OR #42 OR #43 OR #44 OR #45 OR #46 OR #47 OR #48 OR #49 OR #50 OR #51 OR #52 OR #53 OR #54) (330,912)

#57 (#55 OR #56) (331,447)

#58 (#21 AND #24 AND #57) (536)

Search terms

# 45 #44 AND #19 AND #18

# 44 #43 OR #42 OR #41 OR #40 OR #39 OR #38 OR #37 OR #36 OR #35 OR #34 OR #33 OR #32 OR #31 OR #30 OR #29 OR #28 OR #27 OR #26 OR #25 OR #24 OR #23 OR #22 OR #21 OR #20

# 43 Topic=(treatment* or treat or treated or treats)

# 42 Topic=(denosumab or prolia or xgeva)

# 41 Topic=("strontium ranelate" or protelos)

# 40 Topic=(raloxifene or biofem or ciclotran or evista or ketidin or loxifen or raxeto or optruma or bonmax or estroact or ralista or celvista)

# 39 Topic=("minodronic acid" or minodronate)

# 38 Topic=("icandronic acid" or icandronate or bisphonal)

# 37 Topic=("piridronic acid" or piridronate)

# 36 Topic=("cimadronic acid" or cimadronate)

# 35 Topic=("olpadronic acid" or olpadronate)

# 34 Topic=("neridronic acid" or neridronate or nerixia)

# 33 Topic=("tiludronic acid" or tiludronate or skelid)

# 32 Topic=("zoledronic acid" or zoledronate or zometa or zomera or aclasta or reclast or zoldria)

# 31 Topic=(risedronate or "risedronic acid" or actonel or ductonar or ribastamin or ridron or risedon or optinate or risofos or benet or acrel)

# 30 Topic=(pamidronate or "pamidronic acid" or aredia or "ADP sodium" or aminomux or pamdosa or pandrat or pamisol or pamitor or osteopam or aredronet or pamidria or pamifos or pamidran or linoten or xinsidona)

# 29 Topic=("ibandronic acid" or ibandronate or boniva or bondronat or bonviva or bandrobon or bonat)

# 28 Topic=(etidronate or "etidronic acid" or EHDP or didronel or difosfen or detidron or osteodidronel or bonemass or osteotop or didronate or diphos or anfozan or biotredine or dralen or etidron or etiplus or feminoflex or maxibral or oflocin or osfo or ostedron or osteodrug or osteoton or ostogene or somaflex or sterodome or sviroxit or tilferan or dronate-os or etidrate or osteum)

# 27 Topic=(clodronate or "clodronic acid" or bonefos or loron or clodron or lodronat or ascredar or ostac or clastoban or lytos or clasteon or climaclod or clodeosten or clody or difosfonal or dolkin or moticlod or niklod or ossiten or osteonorm or osteostab or mebonat)

# 26 Topic=(fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosal)

# 25 Topic=(leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos)

# 24 Topic=(osteofar or osteoform or osteoral or osteotrat or recalfe or terost or aldrox or fosval or holadren)

# 23 Topic=(actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan)

# 22 Topic=(osteotrat or recalfe or terost or aldrox or fosval or holadren or leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos or fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosalalendronate or "alendronic acid" or fosamax)

# 21 Topic=(alendronate or "alendronic acid" or fosamax or actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan or osteofar or osteoform or osteoral)

# 20 Topic=(bisphosphonate* or diphosphonate*)

# 19 Topic=(osteoporo*)

# 18 #17 OR #16 OR #15 OR #14 OR #13 OR #12 OR #11 OR #10 OR #9 OR #8 OR #7 OR #6 OR #5 OR #4 OR #3 OR #2 OR #1

# 17 Topic=("bone marker*")

# 16 Topic=(("biochemical marker*" or biomarker* or "biological marker*") near/2 bone*)

# 15 Topic=("bone metabolic marker*")

# 14 Topic=("bone turnover marker*")

# 13 Topic=("c-telopeptide*" near/3 collagen)

# 12 Topic=("n-telopeptide*" near/3 collagen)

# 11 Topic=(telopeptide* near/3 collagen)

# 10 Topic=(crosslaps)

# 9 Topic=((CTX or NTX) and (bone or bones or biomarker* or biological marker*))

# 8 Topic=("bone source alkaline phosphatase*")

# 7 Topic=("bone alkaline phosphatase*")

# 6 Topic=("bone specific alkaline phosphatase*")

# 5 Topic=((BSAP or BALP or BAP) and (bone or bones or biomarker* or biological marker*))

# 4 Topic=(collagen near/3 propeptide)

# 3 Topic=(procollagen near/3 peptide)

# 2 Topic=(procollagen near/3 propeptide)

# 1 Topic=(P1NP or PINP)

Limits: Lemmatization – OFF

Search terms

osteoporosis AND ("bone markers" OR "bone turnover markers")

Search terms

1. (P1NP or PINP).ti,ab. (0)

2. (procollagen adj3 propeptide).ti,ab. (0)

3. (procollagen adj3 peptide).ti,ab. (0)

4. (collagen adj3 propeptide).ti,ab. (0)

5. (BSAP or BALP or BAP).ti,ab. and (bone or bones or biomarker$ or biological marker$).af. (0)

6. bone specific alkaline phosphatase$.ti,ab. (0)

7. bone alkaline phosphatase$.ti,ab. (0)

8. bone source alkaline phosphatase$.ti,ab. (0)

9. (CTX or NTX).ti,ab. and (bone or bones or biomarker$ or biological marker$).af. (0)

10. crosslaps.ti,ab. (0)

11. (telopeptide$ adj3 collagen).ti,ab. (0)

12. (n-telopeptide$ adj3 collagen).ti,ab. (0)

13. (c-telopeptide$ adj3 collagen).ti,ab. (0)

14. bone turnover marker$.ti,ab. (0)

15. bone metabolic marker$.ti,ab. (0)

16. ((biochemical marker$ or biomarker$ or biological marker$) adj2 bone$).ti,ab. (0)

17. bone marker$.ti,ab. (0)

18. or/1-17 (0)

19. osteoporo$.ti,ab. (24)

20. (bisphosphonate$ or diphosphonate$).af. (2)

21. (alendronate or alendronic acid or fosamax or actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan or osteofar or osteoform or osteoral or osteotrat or recalfe or terost or aldrox or fosval or holadren or leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos or fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosal).af. (5)

22. (clodronate or clodronic acid or bonefos or loron or clodron or lodronat or ascredar or ostac or clastoban or lytos or clasteon or climaclod or clodeosten or clody or difosfonal or dolkin or moticlod or niklod or ossiten or osteonorm or osteostab or mebonat).af. (0)

23. (etidronate or etidronic acid or EHDP or didronel or difosfen or detidron or osteodidronel or bonemass or osteotop or didronate or diphos or anfozan or biotredine or dralen or etidron or etiplus or feminoflex or maxibral or oflocin or osfo or ostedron or osteodrug or osteoton or ostogene or somaflex or sterodome or sviroxit or tilferan or dronate-os or etidrate or osteum).af. (0)

24. (ibandronic acid or ibandronate or boniva or bondronat or bonviva or bandrobon or bonat).af. (0)

25. (pamidronate or pamidronic acid or aredia or ADP sodium or aminomux or pamdosa or pandrat or pamisol or pamitor or osteopam or aredronet or pamidria or pamifos or pamidran or linoten or xinsidona).af. (0)

26. (risedronate or risedronic acid or actonel or ductonar or ribastamin or ridron or risedon or optinate or risofos or benet or acrel).af. (7)

27. (zoledronic acid or zoledronate or zometa or zomera or aclasta or reclast or zoldria).af. (1)

28. (tiludronic acid or tiludronate or skelid).af. (1)

29. (neridronic acid or neridronate or nerixia).af. (0)

30. (olpadronic acid or olpadronate).af. (0)

31. (cimadronic acid or cimadronate).af. (0)

32. (piridronic acid or piridronate).af. (0)

33. (icandronic acid or icandronate or bisphonal).af. (0)

34. (minodronic acid or minodronate).af. (0)

35. (raloxifene or biofem or ciclotran or evista or ketidin or loxifen or raxeto or optruma or bonmax or estroact or ralista or celvista).af. (2)

36. (strontium ranelate or protelos).af. (0)

37. (teriparatide or forteo or forsteo).ti,ab. (0)

38. (denosumab or prolia or xgeva).af. (0)

39. (treatment$ or treat or treated or treats).ti,ab. (15,352)

40. or/20-39 (15363)

41. 18 and 19 and 40 (0)

Search terms

1. (P1NP or PINP).ti,ab. (908)

2. (procollagen adj3 propeptide).ti,ab. (511)

3. (procollagen adj3 peptide).ti,ab. (663)

4. (collagen adj3 propeptide).ti,ab. (225)

5. (BSAP or BALP or BAP).ti,ab. and (bone or bones or biomarker$ or biological marker$).af. (1440)

6. bone specific alkaline phosphatase$.ti,ab. (1388)

7. bone alkaline phosphatase$.ti,ab. (1421)

8. bone source alkaline phosphatase$.ti,ab. (1)

9. (CTX or NTX).ti,ab. and (bone or bones or biomarker$ or biological marker$).af. (2846)

10. crosslaps.ti,ab. (429)

11. (telopeptide$ adj3 collagen).ti,ab. (839)

12. (n-telopeptide$ adj3 collagen).ti,ab. (251)

13. (c-telopeptide$ adj3 collagen).ti,ab. (228)

14. bone turnover marker$.ti,ab. (1780)

15. bone metabolic marker$.ti,ab. (232)

16. (marker/ or biochemical marker/ or biological marker/ or disease marker/) and (bone/ or bone turnover/) (3965)

17. ((biochemical marker$ or biomarker$ or biological marker$) adj2 bone$).ti,ab. (2782)

18. bone marker$.ti,ab. (1800)

19. or/1-18 (12,593)

20. exp osteoporosis/ (75,240)

21. osteoporo$.ti,ab. (58,711)

22. 20 or 21 (88,169)

23. bisphosphonic acid derivative/ or alendronic acid/ or clodronic acid/ or etidronic acid/ (29,528)

24. (bisphosphonate$ or diphosphonate$).af. (16,503)

25. (alendronate or alendronic acid or fosamax or actimax or alenato or arendal or berlex or brek or elandur or findeclin or lafedam or lendronal or marvil or maxtral or oseotenk or osteofene or osteonate or phostarac or regenesis or silidral or tilios or bonalen or cleveron or endronax or minusorb or norvic or osdron or ossomax or ostenan or osteofar or osteoform or osteoral or osteotrat or recalfe or terost or aldrox or fosval or holadren or leodrin or oseotal or osteofem or osteosan or alefos or fosalen or ostalert or bifosa or osteofos or fosalan or maxibone or adronat or alendros or dronal or genalen or onclast or drovitan or landrolen or sinfract or aldronac or aliot or denfos or fixopan or genalmen or osteodur or porosal).af. (14,288)

26. (clodronate or clodronic acid or bonefos or loron or clodron or lodronat or ascredar or ostac or clastoban or lytos or clasteon or climaclod or clodeosten or clody or difosfonal or dolkin or moticlod or niklod or ossiten or osteonorm or osteostab or mebonat).af. (4623)

27. (etidronate or etidronic acid or EHDP or didronel or difosfen or detidron or osteodidronel or bonemass or osteotop or didronate or diphos or anfozan or biotredine or dralen or etidron or etiplus or feminoflex or maxibral or oflocin or osfo or ostedron or osteodrug or osteoton or ostogene or somaflex or sterodome or sviroxit or tilferan or dronate-os or etidrate or osteum).af. (7748)

28. (ibandronic acid or ibandronate or boniva or bondronat or bonviva or bandrobon or bonat).af. (3253)

29. (pamidronate or pamidronic acid or aredia or ADP sodium or aminomux or pamdosa or pandrat or pamisol or pamitor or osteopam or aredronet or pamidria or pamifos or pamidran or linoten or xinsidona).af. (7680)

30. (risedronate or risedronic acid or actonel or ductonar or ribastamin or ridron or risedon or optinate or risofos or benet or acrel).af. (6164)

31. (zoledronic acid or zoledronate or zometa or zomera or aclasta or reclast or zoldria).af. (7257)

32. (tiludronic acid or tiludronate or skelid).af. (744)

33. (neridronic acid or neridronate or nerixia).af. (266)

34. (olpadronic acid or olpadronate).af. (242)

35. (cimadronic acid or cimadronate).af. (9)

36. (piridronic acid or piridronate).af. (0)

37. (icandronic acid or icandronate or bisphonal).af. (2)

38. (minodronic acid or minodronate).af. (182)

39. (raloxifene or biofem or ciclotran or evista or ketidin or loxifen or raxeto or optruma or bonmax or estroact or ralista or celvista).af. (8201)

40. raloxifene/ (7977)

41. (strontium ranelate or protelos).af. (1325)

42. "parathyroid hormone[1-34]"/ (3339)

43. (teriparatide or forteo or forsteo).ti,ab. (913)

44. (denosumab or prolia or xgeva).af. (1737)

45. (treatment$ or treat or treated or treats).ti,ab. (3,950,972)

46. dt.fs. (2,648,356)

47. or/23-46 (5,432,565)

48. 19 and 22 and 47 (2670)

49. exp animal/ or exp nonhuman/ (5,469,091)

50. exp human/ (13070114)

51. 49 not (49 and 50) (4,361,154)

52. 48 not 5 (2495)