Scolaris Content Display Scolaris Content Display

Base de datos Cochrane de Revisiones Sistemáticas Protocolo - Intervención

Pharmacological interventions for bone health in people with epilepsy

Declaraciones de intereses de los autores

Versión publicada: 10 enero 2024 Historial de versiones

https://doi.org/10.1002/14651858.CD014880
Contraer todo Desplegar todo

Abstract

Objectives

This is a protocol for a Cochrane Review (intervention). The objectives are as follows:

To assess the effectiveness of various pharmacological interventions (for treatment and for prevention) for bone health in people with epilepsy.

Background

People with epilepsy (PWE) are at increased risk of fractures compared to those without (relative risk 2.2, 95% confidence interval (CI) 1.9 to 2.5) (Vestergaard 2005), particularly fractures of the hip and spine (Shen 2014). Non‐traumatic fractures among PWE are associated with five times greater mortality compared to the general population (Whitney 2020). Epilepsy‐related fractures are a common cause of hospitalisation (Strezelczyk 2012) and contribute to significant economic costs (Cheng 2020; Verboket 2021). These fractures are also associated with psychosocial consequences, such as anxiety, depression, loss of independence and inability to meet obligations (Singaram 2019). Despite this significant burden, there is very little clarity on how to guide the management of bone health in PWE.

Fractures result from the interplay between the degree of impact and inherent bone strength. Among PWE, high impact fractures due to seizures account for only a third of all fractures (Zhao 2016). Furthermore, the risk of fractures is increased across all age groups and genders (Diemar 2019). Hence, it appears that poor bone health is an important determinant of fracture in PWE (Griepp 2021).

Description of the condition

Poor bone health results from changes in bone mineral content and microarchitecture that cause increased skeletal fragility, leading to fractures. Bone formation, bone resorption and bone turnover can all be adversely affected by epilepsy (Pack 2008). This relationship is only beginning to be understood, and multiple mechanisms are thought to be at play.

  1. Vitamin D. PWE are at risk of vitamin D deficiency (Fernandez 2018). The initial descriptions of bone diseases in epilepsy were in institutionalised individuals with osteomalacia, which occurs due to vitamin D deficiency in adults (Petty 2016). Special diets such as the ketogenic diet used in PWE are associated with vitamin D deficiency, and their metabolic effects, like acidosis, further adversely affect bone health (Bergqvist 2008).

  2. Secondary osteoporosis. Osteoporosis is defined as “a systemic disease characterised by low bone mass and architectural deterioration of bone tissue with a consequential increase in bone fragility and susceptibility to fracture” (NIH 2001). When these changes occur due to an underlying disease or drugs it is called secondary osteoporosis; anti‐seizure medications (ASMs) are a well recognised cause of secondary osteoporosis (Mirza 2015). ASM‐associated bone loss affects patients irrespective of age or gender (Griepp 2021; Pack 2008). In children and adolescents during critical phases, it may affect bone accrual and thereby lower peak bone mass (Meier 2011). ASMs are known to accelerate bone loss by many mechanisms (Arora 2016). The nature of ASMs may influence this bone loss, with early reports implicating enzyme‐inducing ASMs such as phenytoin (Petty 2016). Data on the effect of valproate, an enzyme inhibitor, on bone health are controversial. Newer drugs such as topiramate, eventrate and lamotrigine, especially in combination with valproate, have also been associated with reduced bone density (Meier 2011). The effect of duration of anti‐seizure medications on bone density is unclear, as some studies appear to show a reduction within two years while others show no change even over a 10‐year period (Griepp 2021). An interplay between genetics and ASM in determining the nature and severity of bone loss is also speculated (Dussault 2017). Blood levels of vitamin D, calcium, parathormone (PTH) and bone turnover markers are known to be altered by ASM intake, especially for those ASMs belonging to the first generation. Vitamin D metabolism is known to be affected at multiple levels by various ASMs (Dussault 2017). Altered sex hormone levels seen with some ASMs also contribute to bone loss (Arora 2016; Dussault 2017). A relationship may exist between exposure to ASMs (dose and number of ASMs used) and poor bone health (Beerhorst 2012; Griepp 2021).

  3. Others. Well‐known risk factors for poor bone health (e.g. smoking, comorbid illnesses, inadequate physical activity) are more common in PWE than those without epilepsy (Fernandez 2019). Age and gender also influence these relationships.

Description of the intervention

Many interventions for improved bone health exist and can be broadly classified as pharmacological and non‐pharmacological. Non‐pharmacological interventions are heterogenous (Coronado‐Zarco 2019) and may improve bone density (Howe 2011), but they are not the focus of this review.

Pharmacological interventions target various aspects of bone metabolism, such as bone formation, bone resorption and bone turnover (Pavone 2017). They are broadly classified into anti‐resorptive and anabolic agents. They improve the quality of bone and thereby prevent fractures. These interventions were primarily developed for, and studied in, primary osteoporosis. A network meta‐analysis reported that teriparatide was the most efficacious in preventing non‐vertebral fractures, while etidronate and denosumab both had good risk‐benefit ratios for fracture prevention in primary osteoporosis (Yang 2016). In secondary osteoporosis due to glucocorticoids, bisphosphonates (Allen 2016) and denosumab (Yanbeiy 2019) have been found to be effective in reducing vertebral fractures and preserving bone density. Similarly, these drugs may be effective in improving bone health in PWE. As these interventions have multiple methods of action, it is unclear if one or more will be more beneficial to bone health in PWE, given the complexity of bone disease in this group. Hence, we have included all possible pharmacological interventions in this review.

How the intervention might work

The pharmacological agents act by multiple mechanisms to ultimately reduce bone resorption (as with anti‐resorptive agents) or promote bone accrual (as with anabolic agents). The various agents are listed below with a summary of their mechanisms of action (Pavone 2017).

  1. Anti‐resorptive agents

    1. Vitamin D and calcium. These are both directly involved in bone metabolism at all stages. Optimum levels of these nutrients are required to ensure bone health. The daily recommended calcium intake is 1000 to 1200 mg/d in adults. Vitamin D can be given orally at 800 to 4000 IU daily or monthly at 60,000 IU, or as annual injections at 500,000 IU. To date, there is no conclusive evidence that supplementation reduces fractures, though observational studies seem to suggest that maintaining optimum levels of vitamin D and calcium is associated with reduced risk of fractures (Yao 2019).

    2. Bisphosphonates: These are the most commonly used anti‐resorptive agents. They inhibit osteoclast attachment to bone matrix, interfere with their differentiation and cause direct injury to the osteoclasts. This prevents bone resorption and improves bone strength. They can be grouped as nitrogen containing (alendronate, zoledronate, risedronate, etc.) or non‐nitrogen containing (etidronate). Different schedules, doses and modes of delivery are used for the different drugs. All these agents have been shown to increase bone mineral density (BMD) and prevent fractures – both vertebral and femoral – in people with primary osteoporosis (Zhou 2016).

    3. Hormonal. Sex steroids play an important role in reducing bone turnover and possibly in promoting bone accrual. Specifically, oestrogens inhibit bone resorption by inhibiting osteoclasts through the ERß receptor. Thus, post‐menopausal women with epilepsy are at increased risk of osteoporosis (Harden 2003). Oestrogens are given to women as hormone replacement therapy. The possible side effects of such therapy include venous thrombosis and cancer (Beral 2002). The selective oestrogen receptor modulators (SERMs) were developed to engage only the bone receptors of oestrogen. This enables this group of drugs to avoid the side effects of hormone replacement therapy while reducing bone resorption and increasing trabecular bone mass. SERMs have been shown to reduce vertebral fractures. However, their ability to reduce non‐vertebral fractures is not conclusive. Androgen replacement is indicated in osteoporosis only in cases of documented androgen deficiency (de Villiers 2018).

    4. Receptor activator of NF‐κB ligand (RANKL) inhibitors. The binding of RANKL to its receptors on pre‐osteoclasts is a key step in determining activation and differentiation of osteoclasts. RANKL inhibitors like denosumab block the binding of RANKL to RANK, causing inactivation of osteoclasts and their apoptosis. Denosumab has been shown to improve BMD and reduce fractures (Zhang 2020).

    5. Calcitonin. Unlike other agents in this category, calcitonin reduces bone resorption by its action on osteoblasts and not on osteoclasts. It is available as injectable, nasal and oral formulations. It has been shown to reduce bone turnover and increase BMD, though its role in fracture prevention is not well established (Munoz‐Torres 2004).

    6. Strontium ranelate. The mechanism of action is not clear. It probably modulates the calcium sensing receptors (CaSR), causing inhibition of osteoclast function and promotion of osteoblast differentiation. It has been shown to increase BMD, though its effect on fractures is less robust (Cianferotti 2013).

  2. Anabolic agents

    1. PTH analogues. Parathormone (PTH) is a key hormone in the regulation of bone metabolism. It activates osteoblasts and directly stimulates bone formation on active remodelling sites and on previously inactive bone surfaces. There is no accompanying increase in bone resorption. The PTH analogues are recombinant human PTH and promote bone formation by a similar mechanism. They have been shown to increase BMD values significantly, and reduce both vertebral and femoral fractures (Fan 2020).

Why it is important to do this review

Treatment of primary osteoporosis with anti‐resorptive and anabolic medications is effective in reducing the incidence of fractures (Yang 2016); it may also be cost‐effective (Li 2021). Some of these drugs have been found to be effective in secondary osteoporosis due to glucocorticoids (Allen 2016) and chronic kidney disease (Hara 2021). Little is known about these drugs in the management of poor bone health among PWE.

A Cochrane review on the effects of vitamins on seizure control reported on two studies of vitamin D supplementation (Ranganathan 2005). They noted that one of these studies showed a significantly increased bone mineral content among adult PWE using vitamin D supplementation (odds ratio 3.6, 95% CI 2.48 to 4.72). A more recent systematic review on vitamin D for bone health among adult PWE taking at least one ASM included four RCTs and five quasi‐experimental studies (Fernandez 2019). Six of these studies showed an improvement in BMD. This review included only PWE taking ASM. The review authors did not study fracture outcomes, and did not perform a meta‐analysis. Quality assessments were also inadequately reported. Other interventions such as bisphosphonates, which have been shown to be more effective in reducing fractures, have not been reviewed for use in PWE at all.

National guidance in the UK recommends the supplementation of vitamin D in PWE whilst acknowledging the lack of evidence (Cock 2015). Considering the burden of poor bone health in PWE and the lack of high‐quality evidence to guide its management, we are undertaking this review in order to critically appraise and summarise all the evidence pertaining to the various pharmacological interventions for bone health in PWE.

Objectives

To assess the effectiveness of various pharmacological interventions (for treatment and for prevention) for bone health in people with epilepsy.

Methods

Criteria for considering studies for this review

Types of studies

We will include randomised controlled trials (RCTs), including cluster‐randomised trials, and non‐randomised controlled cohort studies.

The interventions for improving bone health that we are evaluating in this review are the standard of care for primary osteoporosis, so comparisons between treated and non‐treated groups may not necessarily be done within an RCT. Some of the outcomes of interest, such as fractures or osteonecrosis, are long‐term outcomes requiring many years to manifest. Furthermore, the relationship between bone mineral density (a surrogate marker) and fractures (a patient‐important outcome), may be different among different PWE groups owing to differences in age, anti‐seizure medications etc. Thus, it is possible that investigators may use a non‐randomised cohort design to provide the intervention to all their study participants, to provide information on long‐term follow‐up, or to intervene in a highly selected PWE group. Hence, we believe these non‐randomised cohort studies may provide important complementary or sequential information that may not be available from RCTs (Schunemann 2013). We will not include cross‐over trials, as many of the interventions in the study may result in long‐lasting residual effects, so an effective wash‐out period before cross‐over of intervention may not be possible. Therefore, we will include only controlled cohort studies in this review, and we will analyse them separately from the RCTs. We will not limit inclusion of studies on the basis of their duration of follow‐up, publication status or methods of analysis.

Types of participants

We will include all studies which involved people with epilepsy (PWE) irrespective of age, gender, type of epilepsy, comorbidities, nature or duration of anti‐seizure therapy, or setting.

Types of interventions

In this review we will evaluate the following interventions.

Anti‐resorptive agents

  1. Vitamin D with or without calcium

  2. Hormonal therapy: selective oestrogen receptor modulators, oestrogen‐containing hormone replacement therapy, calcitonin therapy, tissue selective oestrogen complex, androgen therapy

  3. Bisphosphonate therapy

  4. RANKL inhibitors

  5. Strontium ranelate

Anabolic agents

  1. PTH analogues

For the RANKL inhibitors, strontium ranelate, and PTH analogues, only a single drug is available for each group, and we will pool the results from all the relevant studies for each group.

For bisphosphonates, hormonal therapy, and vitamin D and calcium, there are many drugs administered through several routes, and at different doses and dosing frequencies (e.g. weekly oral 70 mg alendronate versus 5 mg zoledronic acid intravenously annually). However, there is evidence of a class‐effect for bisphophonates (Drake 2008), and studies of vitamin D have found no significant differences between high and low doses (Burt 2019). Oestrogen therapy at various doses or through different routes has a similar beneficial effect on the bone (Levin 2018). Hence, we believe there is sufficient similarity within each group to pool the data within that group, irrespective of individual drug, dose, duration or mode of delivery.

We will compare all interventions to placebo, usual treatment or no treatment. For interventions other than vitamin D, we will consider the use of vitamin D with or without calcium as the usual treatment. We will exclude comparison studies of active treatment groups.

Types of outcome measures

We will consider fractures (outcome of benefit of direct importance), quality of life (outcome of benefit of direct importance), adverse events (outcome of harm) to be the critical outcomes in this review. The BMD (surrogate marker of benefit), bone turnover markers and bone biopsy and functional measures are outcomes of benefit of the intervention, which we will study as secondary outcomes. We will use the last reported outcome for this review. All other important and relevant outcomes, like growth in children, will be summarised narratively as reported in the studies.

Primary outcomes

  1. Fractures. Fractures result in morbidity, increased risk of subsequent fractures, and increased mortality, especially among elderly people. There is also a significant economic cost for the fractures. Thus, they are an important outcome for studies of bone health and are recommended as an outcome by the Core Outcome Measures for Effectiveness Trials (COMET) initiative for osteoporosis (COMET 2017). For this review we will consider all fractures, irrespective of how they are diagnosed. Different fractures have different implications with respect to morbidity and mortality (Nazrun 2014). Hence, if there are sufficient data available, we will perform a subgroup analysis of vertebral and non‐vertebral fractures.

  2. Adverse effects of the interventions. We will evaluate osteonecrosis of the jaw (a condition associated with bisphosphonate therapy where there is necrosis of the jaw bone), atypical femoral fracture, any gastro‐oesophageal disorder, any musculoskeletal disorders, fever, hypersensitivity reactions, cellulitis, venous thromboembolism, stroke, oedema, hot flushes, acute kidney injury (AKI), urinary tract infections (UTI), sepsis, and any other complication that may occur.

  3. Quality of life among PWE can be measured using general health‐related quality of life measures like the 36‐item Short‐form (SF‐36) or 5‐dimension EuroQol questionnaire (EQ‐5D), or instruments specific to epilepsy, for example the Quality of Life in Epilepsy (QOLIE). To understand the influence of bone health on quality of life, osteoporosis specific instruments like the Osteoporosis Assessment Questionnaire (OPAQ), Osteoporosis‐Targeted Quality of Life Survey Instrument (OPTQoL), Quality of Life Questionnaire of the European Foundation for Osteoporosis (QUALEFFO), or the Osteoporosis Quality of Life Questionnaire (OQLQ) may be used (Madureira 2012). We will only include the results in a quantitative summary if two or more trials use the same instrument. If several scales are reported, we will individually summarise the results without assigning priority to any instrument.

Secondary outcomes

  1. Bone mineral density (BMD), measured by dual energy X‐ray absorptiometry (DXA), is a commonly reported surrogate outcome and is strongly related to consequent fracture risk (Bouxsein 2019). For this review we will include BMD measured at hip, lumbar spine, distal forearm and body less head. We will include studies reporting on BMD at any other site in the review but not include their data in the meta‐analysis. Bone mineral density can also be measured using quantitative ultrasound, quantitative computed tomography (CT) and magnetic resonance imaging (MRI). However, the metrics and sites of measurement for each of these techniques are different from those used for BMD measured by DXA (Nishiyama 2018). We will include studies using any of these techniques in the review but will summarise them separately if adequate data are available.

  2. Bone turnover markers. Procollagen type I N terminal propeptide (P1NP), a bone formation biomarker, and cross‐linked telopeptide of type 1 collagen (βCTX‐1), a bone resorption biomarker, are accepted by the UK National Osteoporosis Guideline Group (NOGG), the National Osteoporosis Foundation in the US, the International Osteoporosis foundation (IOF) and The European Society for Clinical and Economic Aspects of Osteoporosis, Osteoarthritis and Musculoskeletal Diseases (ESCEO) as being the most useful biomarkers, particularly for monitoring bisphosphonate therapy and predicting fracture risk (Kuo 2017). Hence, we will consider these two biomarkers for this review. We will also evaluate total alkaline phosphatase as it is a widely available biomarker, though not as specific. We will report all other biomarkers as presented in the studies.

  3. Bone biopsy and histomorphometry. Some studies of bone fragility may use structural and kinetic histomorphometry variables on uncalcified transilial bone biopsy specimens to evaluate therapies. We will also consider such studies in this review. For bone biopsy, we will present a summary of the indices measured, together with their findings, but not include them in the quantitative synthesis.

  4. Functional outcomes. Multiple scales are available that evaluate various measures, such as task‐oriented activities, timed mobility tests, balance activities, gait patterns and muscle strength. Some have been incorporated into batteries, such as functional fitness and functional autonomy batteries (Varahra 2018). We will also report any such measures in this review if presented by the included studies.

Search methods for identification of studies

Electronic searches

We will search the following databases.

  • Cochrane Register of Studies (CRS Web), using the search strategy shown in Appendix 1

  • MEDLINE (Ovid) 1946 onwards, using the search strategy shown in Appendix 2

  • SCOPUS 1823 onwards

CRS Web includes randomised, or quasi‐randomised, controlled trials from PubMed, Embase, ClinicalTrials.gov, the World Health Organization International Clinical Trials Registry Platform (ICTRP), the Cochrane Central Register of Controlled Trials (CENTRAL), and the Specialised Registers of Cochrane Review Groups, including Epilepsy. We will not limit the searches by time, publication status or duration, but we will restrict them to English language studies.

Searching other resources

In addition, we will search the following sources.

  1. Reference lists of review articles, included studies and clinical practice guidelines

  2. Citations of the articles for related letters or errata

Data collection and analysis

Selection of studies

We will combine all citations and abstracts obtained through searches and remove any duplicates. Two groups of review authors (RA & DM, PPN & JSP) will independently screen and assess the entire list of citations, abstracts and full articles. Within each group, each author will review about half the citations. The initial screening (stage 1) will exclude articles whose titles or abstracts are clearly irrelevant and will also remove duplicates. The remaining citations and abstracts, including doubtful citations or abstracts, will be included in the next stage of screening, where we will retrieve the full text of all the articles selected in stage 1. We will collate multiple reports of the same study so that each study is the unit of interest rather than each report. Each group of review authors will independently evaluate these and any other articles identified as potentially relevant for inclusion in the review. At this stage, we will include any articles which both groups of review authors agree to include and exclude those that both groups agree to exclude. In case of conflict, review authors will attempt to reach consensus through discussion. If this is not possible, the senior review author of the team (SN) will adjudicate. We will document this process in sufficient detail to complete a PRISMA flow diagram and characteristics of included and excluded studies tables.

Data extraction and management

We will develop a data extraction form to collect the following data from each of the selected studies.

  • Study details (such as author, citation, identifier, location, length of follow‐up, sponsorship details)

  • Participant details (e.g. inclusion and exclusion criteria for participants, demographic details, type of epilepsy, anti‐seizure medications and duration)

  • Details of the intervention and comparator (drug, dose, route, duration)

  • Outcome details (outcomes, including the type and site of fractures and how they are measured, adverse events, outcomes reported)

  • Details for risk of bias assessment (e.g. study design; study quality, including method of randomisation, sampling procedures, recruitment, attrition; and statistical methods)

For the evaluation of cohort studies, we will also extract the following information.

  • Age, sex, type of epilepsy, anti‐seizure medications (nature, number and duration), previous fractures, post‐menopausal state and setting of the study (community or institutional)

We will develop this form a priori and pilot it on at least one study. Four authors, working in pairs (RA & DM, PPN & JSP), will independently extract these data from the selected studies. In the event of missing information, we will try to contact the authors for clarification. In case of conflict, we will attempt to reach consensus through discussion; if this fails, a senior reviewer (SN) will adjudicate. We will enter the extracted data into RevMan Web for analysis (RevMan Web 2023).

Assessment of risk of bias in included studies

Pairs of review authors (RA & DM, PPN & JSP) will independently assess the risk of bias for each study using the Cochrane risk of bias 2 (RoB 2) tool, as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2022b).

We will assess the risk of bias in results for fractures, adverse events, quality of life and BMD, according to the following domains.

  1. Bias arising from the randomisation process

  2. Bias due to deviations from intended interventions: we will consider the effect of adherence

  3. Bias due to missing outcome data

  4. Bias in measurement of the outcome

  5. Bias in selection of the reported result

Each pair (RA & DM, PPN & JSP) will independently evaluate these domains using the RoB 2 excel tool. In the event of missing information, we will try to contact the authors for clarification. In case of conflict, we will attempt to reach consensus through discussion; if this fails, a senior review author (SN) will adjudicate. We will use the signalling questions in the RoB 2 tool (Higgins 2022b), and rate each domain as 'low risk of bias', 'some concerns' or 'high risk of bias'. We are interested in deviations from adhering to the intervention. As the effects of the interventions are likely to accrue over time, we believe this is a more important effect than assignment. We will assess the outcomes at the last reported time point in the study. We will summarise the risk of bias judgements across different studies for each of the domains listed for each outcome. The overall risk of bias for the result is the least favourable assessment across the domains.

For cross‐over or cluster‐randomised trials, we will use the recommended variation of RoB 2 (Higgins 2022a).

For non‐randomised studies, we will use the ROBINS-I tool to assess the risk of bias for the same outcomes (Sterne 2020). This tool evaluates the following domains: confounding, selection bias (including attrition), information bias (classification of intervention, deviation from interventions and outcome measurement), and reporting bias at pre‐intervention, intervention and post‐intervention time frames. We will consider the following factors to be important confounders: age, sex, type of epilepsy, anti‐seizure medications (nature, duration and number), previous fractures, post‐menopausal state and setting of the study (community or institutional). Important co‐interventions to be considered are the use of vitamin D with or without calcium supplements. On the basis of the ROBINS‐I tool, we will classify studies into low, moderate, serious or critical risk of bias.

We will present the risk of bias summary on every forest plot, use it to perform sensitivity analysis, and incorporate it into the GRADE assessment.

Measures of treatment effect

For dichotomous outcomes, we will express effect measures as the risk ratio (RR) or hazard ratio (HR) with 95% confidence intervals (CI). Effect measures of outcomes measured on continuous scales will be expressed as the mean difference (MD) with 95% CI. If standardisation is possible, then we will use standardised mean differences (SMD). If studies include a mixture of change‐from‐baseline and final value scores, we will use the (unstandardised) mean difference method. We will interpret SMD based on Cohen’s effect size estimates, with < 0.2 being a trivial effect, 0.2 to < 0.5 being a small effect, 0.5 to < 0.8 a moderate effect and > 0.8 a large effect (Tellez 2015). If standardisation is not possible, we will present the outcomes as reported. For studies that have not reported outcomes completely, we will attempt to contact the study authors to obtain more information. If this is not successful, we will only report the results in the narrative synthesis.

Unit of analysis issues

The unit of analysis is the individual participant. In the case of cluster‐randomised trials, we will carefully examine if the unit of analysis is adequately addressed during the reporting of the study results. If not, we will attempt to reanalyse or recalculate the standard error by adjusting for the design effect, as per the guidelines in the Cochrane Handbook (Higgins 2022a). In the event of outcomes being reported at multiple time points, we will use the last time point measured. For studies evaluating multiple interventions at the same time, we will attempt to combine groups and make a single pair‐wise comparison. If this is not possible, we will divide the common control arm equally between the comparisons (Higgins 2022a).

Dealing with missing data

We will contact the trial authors in the event of missing data. When data are not presented in the format required, we will perform simple imputations, such as mean from median, etc (Deeks 2022). We will not attempt to perform any more complex imputations.

Assessment of heterogeneity

To assess statistical heterogeneity, we will use the I2 statistic and a Chi‐square test. We will interpret the I2 values as follows.

  • 0% to 40%: might not be important

  • 30% to 60%: may represent moderate heterogeneity

  • 50% to 90%: may represent substantial heterogeneity

  • 75% to 100%: considerable heterogeneity.

This will be further influenced by the magnitude and direction of treatment effects and the strength of evidence for heterogeneity (e.g. the P value from the Chi2 test, or a CI for I2) (Deeks 2022). If significant statistical heterogeneity exists (substantial or considerable heterogeneity as per guidance in the Cochrane Handbook (Deeks 2022)), we will perform a random‐effects meta‐analysis. We will explore the reasons for such heterogeneity and discuss the same.

Assessment of reporting biases

When an outcome is unavailable for synthesis from a given study, we will use the Outcome Reporting Bias In Trials (ORBIT) system for classifying the reasons for the missing results (Page 2022). In the light of this information, we will consider the bias introduced in the synthesis and will report the studies with missing results in the forest plot (Page 2022).

If more than 10 studies are included in the review, we will construct a funnel plot to assess for reporting bias. To investigate the presence of small study bias in a possible meta‐analysis, we will check whether a random‐effects model estimate of the intervention effect is more beneficial than a fixed‐effect model estimate of intervention effect.

Data synthesis

We will perform meta‐analyses only when such an approach is meaningful; that is, if the interventions, participants and outcome measures are sufficiently similar for pooling of effect measures. Two authors (RA and SN) will assess the similarity and document the reasons. We will use inverse variance weighted methods.

The interventions bisphosphonates, vitamin D/calcium and hormonal therapy have some inherent variability, such as in the dose, drug, mode of delivery etc., and we anticipate some variability in the effects. Thus, we will use a random‐effects meta analysis model for these interventions (Deeks 2022). We will explore possible subgroup analysis if there are adequate data.

We will use RevMan Web to analyse the data and perform separate analyses for the different study designs (randomised and non‐randomised studies) and different intervention categories. We will perform inverse variance weighted fixed effect meta‐analysis for the RANKL inhibitors, strontium ranelate and PTH analogues, unless there is statistical heterogeneity as described above. We will present these results in tabular forms as well as forest plots.

Subgroup analysis and investigation of heterogeneity

We will perform the following subgroup analysis if there are sufficient data.

  1. By age group: paediatric (< 18 years), adult (18 to 65 years) and elderly (> 65 years)

  2. By nature of antiepileptic therapy, as enzyme inducing versus non‐enzyme inducing

  3. By duration of trials (≤ 6 months versus > 6 months)

  4. By gender

  5. By duration of anti‐seizure medication

  6. By nature of fracture, as vertebral and non‐vertebral fractures.

Based on the data obtained, we may consider further subgroup analyses.

We anticipate considerable clinical and methodological heterogeneity. Patient‐level, interventional‐level and outcome‐level factors are likely to be different between studies (Gagnier 2012). We will extract key covariates at all three levels. We will evaluate clinical heterogeneity narratively, and discuss the implications of such differences in the review. Methodological heterogeneity is also likely, and we will assess this and consider any implications as part of the risk of bias assessment of included studies. We will perform sensitivity analysis to assess the impact of bias in studies and will also discuss it narratively.

Sensitivity analysis

We will perform sensitivity analyses for primary outcomes by excluding studies judged to be at high overall risk of bias.

Summary of findings and assessment of the certainty of the evidence

We will present the findings of the review in a summary of findings table for the following outcomes: fracture, quality of life, adverse events and BMD. The comparator used will be placebo, usual treatment or no treatment. For bisphosphonates, we will consider the use of vitamin D/calcium to be the usual treatment. The magnitude of effect of the intervention will be presented as absolute and relative measures for dichotomous outcomes (fracture and adverse events), and as mean differences or percentage change from baseline for continuous measures (quality of life and BMD). For each of these outcomes, this table will provide information regarding the effect size, quality of evidence and the number of participants and studies for each analysis. We will report on vitamin D/calcium and bisphosponates because these are the most widely used interventions. We will report on these outcomes at six months' follow‐up.

The risk of bias within a trial (methodological quality), the directness of evidence, heterogeneity, precision of estimates and risk of publication bias are all considered when deciding the quality of evidence by the GRADE approach. The overall risk of bias as determined by the ROB 2 tool will feed into the GRADE assessment. The quality of evidence is then graded as very low, low, moderate or high using the GRADE approach (Schunemann 2022). This quality refers to the extent to which we can be confident that the estimates from the synthesis of studies are close to the true effect.