Prevalence of osteopenia and incidence of fragility fracture in adult patients with gastrointestinal cancer: protocol for a systematic review and meta-analysis

Abstract

Introduction

Osteopenia, defined as a bone mineral density (BMD) T-score between −1.0 and −2.5, is a precursor to osteoporosis and fragility fracture which results in substantial morbidity. Patients with gastrointestinal (GI) cancer are particularly vulnerable due to disease-specific mechanisms of bone loss and treatment-related toxicities, which compound the risk of osteoporotic fractures and adverse oncological outcomes. Despite this, the global prevalence of osteopenia and incidence of osteoporotic fracture in GI cancer populations remains poorly quantified, limiting opportunities for targeted surveillance, preventive strategies and policy development.

Objective

This systematic review and meta-analysis will estimate the global prevalence of osteopenia and global incidence of osteoporotic fracture in cohorts of adults with non-metastatic GI cancer and explore variation across treatment type, cancer site, geographical region and diagnostic modality.

Methods and analysis

This protocol follows the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) guidance. We will conduct a comprehensive search of MEDLINE, Embase, CINAHL, PubMed, Web of Science, Cochrane Central and ClinicalTrials.gov from 1994 to 2026. Eligible studies will include observational or interventional cohorts reporting prevalence of osteopenia, assessed using dual-energy X-ray absorptiometry (DXA) or CT scan, or incidence of fragility fracture, as determined by imaging findings, administrative codes or clinical history in cohorts of adult non-metastatic cancer populations. Five reviewers will independently and in pairs screen citations and full-text articles, extract data, and assess study quality using the Joanna Briggs Institute checklist for prevalence studies for each outcome. The primary outcome is the global pooled prevalence of osteopenia (diagnosed using DXA), calculated using a random-effect generalised linear mixed model. Secondary outcomes are the global pooled incidence rate of fragility fracture, and the pooled fragility fracture incidence rate ratio (IRR) comparing patients with GI cancer with the general population. All estimates will be presented with 95% confidence intervals. Heterogeneity will be assessed with I², τ² and prediction intervals for each outcome. Sources of heterogeneity in osteopenia prevalence will be explored through subgroup analyses by primary cancer location, treatment exposure, and geographical region. Sensitivity analyses, including an analysis of the pooled prevalence of osteopenia diagnosed using CT, and limited to studies deemed low risk of bias, will be performed. Publication bias will be evaluated using Doi plots supplemented with Luis Furuya-Kanamori indices. We will assess certainty in the evidence-base in duplicate by using the Grading of Recommendations Assessment, Development and Evaluation methodology for systematic reviews involving prognosis.

Ethics and dissemination

This study is a secondary analysis of aggregate, de-identified data extracted from previously published studies. As such, no new data were collected from human participants, and formal ethics board approval is not required. All aggregated data and analytical code used to complete this study will be made available on request to the corresponding author. The findings from this study will be disseminated through publication in a peer-reviewed scientific journal and presentation at national and international conferences.

PROSPERO registration number

CRD420261286673.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Piloted screening procedures, standardised data extraction and structured risk of bias assessment increase inter-rater reliability and reproducibility.

• Generalised linear mixed models will be used to directly model binomial data and generate robust pooled prevalence estimates.

• Variation in diagnostic definitions of osteopenia and fragility fracture may introduce misclassification bias.

• Generalisability is limited to the populations and geographical regions represented by the studies that will be included in this review.

Introduction

Rationale

Osteopenia and osteoporosis are disorders of bone remodelling that result in increased bone fragility and risk of fragility fracture, resulting from reduced bone formation and increased bone resorption.¹ These microarchitectural changes in the bone alter its mechanical properties and predispose to fracture.¹ The WHO first defined osteopenia in 1994 as a bone mineral density (BMD) 1.0 to 2.5 SDs below the mean BMD of a healthy young adult female (T-score between −1.0 and −2.5) as measured by dual-energy X-ray absorptiometry (DXA).¹ A T-score of ≤−2.5 or presence of fragility fracture is diagnostic of osteoporosis.¹ The global prevalence of osteopenia has been estimated to be 40%, which is double that of osteoporosis.² Although fragility fracture risk rises progressively with lower BMD, the absolute number of fractures is higher among individuals with osteopenia, reflecting the larger population affected compared with those who meet criteria for osteoporosis.³ Women with hip fractures related to low BMD experience 10–20% increased mortality compared with the general population.⁴ In 2019, 438 000 deaths and 16.6 million disability-adjusted life years were attributable to low BMD.⁵

Although gastrointestinal (GI) cancers are responsible for one in four cancer diagnoses globally,⁶ advances in cancer screening, diagnostics and therapeutics have substantially improved survival across a wide range of GI malignancies.⁷,¹² More than 18 million cancer survivors are estimated to be living in the USA in 2025.⁸ This number is expected to surpass 22 million by 2035.⁸ With increasing survivorship, there is growing recognition of the complex sequelae that persist after completion of cancer-directed therapy.¹³ Among these, long-term health consequences such as treatment-related comorbidities are of particular concern, with implications for quality of life, long-term follow-up and healthcare resource utilisation.¹⁴,²⁰

Bone demineralisation, osteopenia and predisposition to fracture are important late effects of cancer and its treatment.²¹ In addition to skeletal metastases, several systemic therapies, including hormone therapy, chemotherapy, radiotherapy, corticosteroids and tyrosine kinase inhibitors, can disrupt normal bone remodelling, thereby predisposing patients to cancer treatment-induced bone loss.^{14 22} Proposed mechanisms include direct dysregulation of osteoblast and osteoclast differentiation, activity and apoptosis, as well as indirect pathways such as treatment-related renal dysfunction or marrow adipocyte differentiation.¹⁴ Surgical resection of GI malignancy has also been associated with decreases in BMD.²³ Retrospective studies have shown that up to 60% of patients with GI cancers have osteopenia around the time of cancer treatment.^{16 24 25} Furthermore, loss of BMD can be rapid following the initiation of cancer treatment, with reduced BMD seen within 1–3 years of cancer treatment.¹⁵²⁶,²⁸

These cancer-specific risk factors frequently coexist with established risk factors for BMD loss, including advanced age, low body mass index, postmenopausal status, physical inactivity, reduced mobility, smoking, alcohol use and insufficient calcium and vitamin D intake, which markedly increase the risk of osteopenia and osteoporosis.^{14 29} In turn, osteopenia substantially increases the risk of fragility fractures and adversely affects quality of life.^{17 18} In addition, impaired bone integrity creates a permissive microenvironment that may facilitate the development and progression of bone metastases.¹⁴ Given these consequences, cancer treatment-related bone loss has even been associated with increased mortality and worse prognosis.^{19 20 24 25 30}

Early diagnosis of cancer-related bone loss is critical to enable timely initiation of both lifestyle and pharmacological interventions to prevent fracture.^{31 32} Evidence-based lifestyle strategies to support bone health include regular physical activity incorporating muscle-strengthening and balance exercises, smoking cessation, moderation or avoidance of alcohol and caffeine and adequate intake of calcium and vitamin D.^{14 29} Pharmacological interventions, particularly antiresorptive agents such as bisphosphonates, should be considered in individuals with documented low BMD or in those with additional risk factors for BMD loss.³³ Bisphosphonate therapy has been shown to preserve BMD, reduce fracture risk and improve disease-free survival and recurrence rates among postmenopausal women with breast cancer.^{34 35} Comparable benefits, including reduced fracture incidence and improvements in BMD, have also been demonstrated with denosumab in patients with breast and prostate cancer.^{36 37}

Diagnosis of bone loss typically involves assessment of BMD using DXA at the hip, lumbar spine or femoral neck. Current guidelines recommend routine BMD evaluation in populations at high risk, such as patients with breast or prostate cancer, although such assessments are not yet standard practice across all malignancies despite recognition of this unmet need,²⁹³⁸,⁴⁰ and risk models have not been validated in patients with cancer.⁴⁰

Despite this growing body of evidence, the burden of osteopenia and fragility fracture in patients with GI cancers remains poorly defined compared with hormonally driven malignancies such as breast or prostate cancer.³¹ Given the high prevalence of GI cancers worldwide and the frequency with which their treatments affect nutritional status, metabolism, and overall bone health, there is a critical need for focused research in this population.⁶ Better characterisation of bone health, risk factors, and long-term outcomes in GI cancer survivors will prompt future research investigating causal associations of various aspects of cancer care to BMD loss, inform surveillance strategies, guide preventive and therapeutic interventions, and ultimately improve survivorship care.

Objectives

This systematic review and meta-analysis has multiple objectives. The primary objective is to estimate the global pooled prevalence of osteopenia in cohorts of adults with non-metastatic GI cancer. Subgroup analyses will explore variations in osteopenia prevalence across treatment type, cancer site, and geographical region. Secondary objectives are to estimate the pooled incidence rate of fragility fracture (a clinically significant consequence of osteoporosis) in this population, and the pooled fracture IRR comparing patients with GI cancer to the general population.

Methods

This protocol was written in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P).⁴¹ The completed checklist is available in online supplemental appendix 1. All authors have a background in related fields. Due to the expected scope of extracted data, the outcomes of this systematic review and meta-analysis may be published across several manuscripts. This systematic review has been registered on PROSPERO (registration number: CRD420261286673).

Eligibility criteria

We will include all relevant clinical trials and observational studies published after 1994, which is when the diagnostic criteria for osteopenia using DXA were established by the WHO.⁴² Studies conducted on human subjects published in all languages will be considered. Eligible publications must present original data.

Included studies will involve the target population of patients 18 years of age or older with any GI cancer diagnosis, being cared for in the inpatient or outpatient setting, at any point following cancer diagnosis. GI cancer includes oesophageal, gastric, intestinal, colorectal, appendiceal, anal, hepatobiliary and pancreatic. Studies will be excluded if more than 50% of the cohort has metastatic or stage IV cancer. The conditions of interest are (1) the prevalence of osteopenia, diagnosed using DXA or CT, with threshold values defined by authors of included studies, and (2) the incidence of fragility fracture, defined as fractures occurring after minimal to no trauma (eg, fall from standing height). In view of anticipated variability in the operational definition of fragility fracture among included studies, we will accept author-defined criteria consistent with low-trauma mechanisms, as determined by imaging findings, administrative coding or clinical documentation. We will include the largest cohort if overlapping cohorts are identified. Articles reporting fractures due to high-energy trauma and pathological fractures from metastatic cancer will be excluded. Studies that report only the incidence of osteopenia or the prevalence of fragility fractures will be excluded, since prevalence and incidence outcomes reflect distinct epidemiological measures and cannot be quantitatively synthesised within the same meta-analysis. Unpublished data, conference abstracts and grey literature will be excluded. Studies reporting modelled estimates will be excluded given that these are not independently observed data from primary study populations.

Information sources

On 2 October 2025, electronic bibliographic databases Ovid MEDLINE, Ovid Embase, CINAHL, Cochrane Central, PubMed, Web of Science and ClinicalTrials.gov were searched from 1994 to 2025 using a comprehensive search strategy developed in collaboration with an information specialist. An updated search conducted across all databases in 2026 prior to final manuscript submission will be performed to ensure search currency. Eligible studies will also be extracted from reference lists of the included studies. Two attempts will be made to contact corresponding authors for relevant data when not reported or extractable, and for clarification when required.

Search strategy

The comprehensive search strategy used to obtain relevant articles on the database Ovid MEDLINE is provided in online supplemental appendix 2. This search was then adapted for use in other databases. For all database searches, subject headings and key words included gastrointestinal cancer, metabolic bone disease, bone density, osteopenia, bone demineralization and fragility fracture.

Study records

Data management

The studies obtained from the comprehensive literature search have been imported into Covidence (Veritas Health Innovation, Melbourne, Australia) for consolidation and deduplication. This software will be used as a platform to streamline key methodological steps such as title/abstract screening and full-text review and enhance collaboration among the two independent reviewers. Extracted data will be stored in spreadsheets, with regular backups. A detailed audit trail of decisions made during screening, data extraction and analysis will be maintained. On completion of the review, all data will be archived in accordance with institutional requirements.

Selection process

A pilot using 50 citations will be performed prior to beginning title/abstract screening to ensure high inter-rater reliability, which will be assessed using Gwet’s Agreement Coefficient 1 (AC1). Observed agreement that is either substantial or almost perfect⁴³ will be considered sufficient to proceed. Five reviewers working in pairs will independently screen all studies at both the title/abstract stage and, for records meeting inclusion criteria, at the full-text stage. For full-text screening, an initial pilot of five articles will be conducted independently by all reviewers in pairs, with a target of 100% agreement prior to proceeding with the remaining full-texts. Citations selected at the title/abstract stage by either reviewer (SB, SM, KYI, NB, GY) will be selected for full text review, to be performed by the same reviewers. Any discrepancies during full-text review will first be resolved through consensus; if agreement cannot be reached, a third reviewer (NC) will adjudicate.

Data collection process

Data extracted will include information pertaining to the conditions of interest, context, population and planned subgroup analyses using predetermined forms, which will first be pilot tested across reviewers using three studies with a target of 100% agreement prior to proceeding with full data extraction. To minimise errors in the extraction of prevalence and incidence data, standardised data extraction forms with explicit field-level definitions will be used for all outcome variables. Any discrepancy between a study’s reported measure and its underlying design will be flagged, adjudicated, and documented, and the authors will be contacted to clarify measures reported.

Data items

Data will be extracted for the following domains wherever reported:

• Study variables: publication status, author, year published, study design, recruitment period, recruitment procedure, source of funding, conflicts of interest, number of participants.

• Population variables: country/region, average age, proportion of female participants.

• Context variables: primary site of cancer, proportion of stage IV patients, treatment exposure (eg, receipt of chemotherapy, radiation therapy or surgery prior to BMD measurement), average follow-up time, average time from cancer diagnosis to BMD measurement and/or fracture diagnosis, average time from last cancer treatment to BMD measurement and/or fracture diagnosis, occurrence of subsequent fractures, identification cancer recurrence alongside fracture diagnosis.

• Outcome variables and related definitions: osteopenia prevalence, method of ascertaining osteopenia diagnosis (ie, DXA, CT), BMD thresholds used for case definitions of osteopenia, interventions received to counteract reduced BMD (eg, physical activity regimen, bisphosphonate or denosumab use, calcium and/or vitamin D supplementation), definition of fragility fracture, fracture site(s) measured, method of diagnosis of fracture (imaging, self report, administrative codes), whether the fractures were symptomatic or incidental, symptoms reported (if any), fracture rate per 1000 person-years for all groups, IRR when cancer populations are compared with control populations within studies.

Outcomes and prioritisation

The primary analysis and main outcome of interest is the global pooled prevalence of osteopenia. Secondary analyses will explore the additional outcomes of global pooled incidence of fragility fracture and the pooled IRR of fragility fracture comparing patients with GI cancer and the general population.

Risk of bias in individual studies

The Joanna Briggs Institute Critical Appraisal Checklist for Studies Reporting Prevalence Data can be used to estimate risk of bias (RoB) for studies reporting prevalence and incidence,⁴⁴ and will be used to estimate RoB in each eligible study for each outcome of interest⁴⁴ (online supplemental appendix 3). This nine-item checklist addresses study sample frame, sampling method, sample size, subject and setting description, validity and reliability of methods used for diagnosis, statistical analysis, and response rate.

Two independent reviewers will assess the RoB of each eligible study in duplicate. Studies not meeting four or more quality criteria (indicated by four or more ‘No’ responses, as assessed by both independent reviewers) will be considered high RoB and excluded in a sensitivity analysis to investigate the impact of study quality on the final results (see Data synthesis). Disagreements in RoB assessment will be resolved through consensus, with independent third-party adjudication if consensus cannot be reached. Results will be displayed in a traffic light plot.

Data synthesis

We will perform a meta-analysis to determine the pooled global osteopenia prevalence. The analysis will include data from all eligible studies that include information regarding prevalence of osteopenia diagnosed using DXA. We will use a random-effects generalised linear mixed model (GLMM) based on logit transformed proportions. Τ² will be estimated using a maximum-likelihood approach. GLMMs better account for within-study variance and the binomial structure of prevalence data, especially if there are rare total event counts or small sample sizes.⁴⁵,⁴⁹ This analysis will result in a pooled proportion with a 95% confidence interval (CI). All results will be visually represented in Forest plots.

The pooled fragility fracture incidence rate will be calculated as a secondary outcome. Eligible cohorts will have a fragility fracture incidence rate (total events divided by person time) extracted. Incidence rates will be scaled per 1000 person-years. For this analysis we will use a GLMM for count data (random-effects mixed-Poisson regression), which has been shown to be more powerful across a range of conditions including recurrent, rare or zero events that are non-normally distributed, incomplete follow-up, and the inclusion of a small number of studies.^{50 51} This will result in a pooled incidence rate with a 95% CI.

We will also calculate the pooled IRR comparing patients with GI cancer with the general population, if sufficient studies are identified. For each study, fracture counts and corresponding person-time will be extracted for both the GI cancer cohort and the comparator population. The pooled IRR will be estimated using a GLMM similar to methods described for pooled fragility fracture incidence, with log person-time included as an offset. This model will yield a pooled IRR with 95% CI.^{50 51}

For all analyses described above, statistical heterogeneity will be quantified with I² and τ².^{52 53} We will also report a 95% prediction interval to estimate the dispersion of the true estimates in the GI cancer population.^{54 55} The multiplier for the prediction interval will be based on the t-distribution or the standard normal distribution, depending on the number of studies and heterogeneity. Direction and magnitude of heterogeneity will be used to assess inconsistency when evaluating the overall strength of the body of evidence (see Confidence in cumulative evidence), acknowledging that I² is generally higher with proportions meta-analyses.^{56 57}

Potential sources of heterogeneity in the outcome of osteopenia prevalence will be explored using the following pre-specified subgroup analyses:

1. Treatment exposure (hypothesis: higher prevalence following cancer treatment).

2. Primary site of cancer (hypothesis: higher prevalence in gastric cancer).

3. Geographic region (hypothesis: higher prevalence in Asia-Pacific regions vs European/North American regions).

We will modify the Instrument to assess the Credibility of Effect Modification Analyses criteria for randomised controlled trials to assess the credibility of any subgroup effects.⁵⁸

The following sensitivity analyses will be conducted provided there is sufficient volume (at least three studies per analysis):

1. Analysis limited to studies deemed low RoB.

2. Analysis limited to studies using Hounsfield units on CT to diagnose osteopenia.

All analyses will be performed in R⁵⁹ using the metafor package.⁶⁰

If quantitative synthesis is not feasible, results will be summarised by conducting a structured narrative synthesis following published guidelines.⁶¹

Meta-bias(es)

Publication bias will be assessed for each key outcome visually by the creation of Doi plots, and quantitatively supplemented using the Luis Furuya-Kanamori (LFK) index with the a priori concern for positive publication bias (studies with high prevalence of osteopenia or fracture are more likely to be published).⁶²,⁶⁴ The LFK index quantifies the degree of asymmetry in the Doi plot, with more extreme asymmetry suggesting publication bias. Values between −1 and +1 indicate no asymmetry, whereas values between −2 and −1 or +1 and +2 indicate minor asymmetry, and values more extreme than ±2 indicate major asymmetry. No other formal tests for publication bias will be performed.^{53 63}

Confidence in cumulative evidence

For each key outcome, the strength of the body of evidence contained within each study will be assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework, which incorporates assessment of risk of bias, imprecision, inconsistency, indirectness and publication bias into an overall assessment of certainty of evidence.⁶⁵ Specific guidance on the use of GRADE in systematic reviews of prevalence and incidence does not exist; however, GRADE for systematic reviews involving prognosis is readily adaptable to questions of prevalence and incidence.^{66 67} A key principle is that for prevalence and incidence reviews, observational studies start as high-certainty and can be downgraded for risk of bias, imprecision, indirectness, inconsistency or evidence of publication bias.⁶⁶ Overall quality and therefore confidence in the evidence will be rated as either high, moderate, low, or very low, and represented in a GRADE summary of findings table and/or evidence profile.

Discussion

Osteopenia and fragility fracture are associated with considerable morbidity, and these risks are further amplified in patients with cancer, who not only experience accelerated BMD loss through cancer-specific mechanisms but also face worse oncological outcomes.^{19 20 24 25 30} Given the rising global burden of osteopenia, and the heightened clinical consequences of low BMD and resulting fragility fracture in cancer survivor populations,³¹ establishing reliable estimates of the global burden of osteopenia and fragility fracture is both timely and essential.

The findings of this study will have several important clinical, research and policy implications. From a clinical perspective, establishing reliable estimates of the global prevalence of osteopenia and incidence of fracture in patients with GI cancer will underscore the need for routine assessment of bone health in oncology and survivorship settings, particularly if low BMD and fracture are diagnosed incidentally. By quantifying the current burden of osteopenia and fracture, our results will help clinicians to appreciate the risk of these conditions, tailor surveillance strategies, and consider preventive interventions such as nutritional supplementation, exercise programmes or pharmacological therapy.³¹ Moreover, our findings will facilitate shared decision-making and patient counselling, ensuring that patients are appropriately informed about the potential impact of cancer treatment on long-term bone health.

Importantly, in the GI cancer population, fractures may arise from distinct and clinically important aetiologies that require differentiation, including osteoporotic fragility fractures⁶⁸ and pathological fractures secondary to metastatic or primary bone disease.⁶⁹ Distinguishing between these entities is essential for clinical decision-making and guiding further treatment. Therefore, accurate fracture screening and classification is a fundamental component of clinical care in patients with GI cancer.

At the health systems level, these findings will inform planning and resource allocation. Prevalence and incidence data can guide the development of cost-effective models for screening, fracture prevention and multidisciplinary survivorship care. For example, estimates of population burden may justify referral to fracture clinics and integration of bone health assessments into oncology follow-up care pathways. Furthermore, by highlighting the scale of the problem globally, our study may encourage guideline committees to consider a question related to systematic bone health surveillance for GI cancer survivors. Importantly, identifying geographical and demographic differences in prevalence will shed light on inequities in access to bone healthcare, particularly in low- and middle-income countries and among marginalised groups, thereby supporting more equitable policy development.

In addition, the results of this study will provide a critical framework for advancing research. The anticipated heterogeneity across studies identified in this review will further highlight the need for standardised definitions, measurement methods, and reporting practices in bone health research. By drawing attention to these anticipated methodological gaps, this work may catalyse efforts to harmonise research practices and promote the development of consensus definitions and outcomes, particularly with respect to the CT-based diagnosis of osteopenia. Prevalence estimates will serve as baseline data for clinical trials testing interventions to mitigate BMD loss and prevent fragility fractures in GI cancer survivors, including the use of bisphosphonates, which have been shown to reduce treatment-related BMD loss and fracture risk in select cancer populations⁷⁰ and in osteopenic populations.⁷¹,⁷³

This systematic review and meta-analysis will be limited by the anticipated heterogeneity of the included studies. Variability in the definitions and diagnostic methods used to identify the key outcomes of osteopenia and fragility fractures is expected to introduce measurement inconsistency. Our findings will only be generalisable to the populations and healthcare settings included in the analysis.

Patient and public involvement

This protocol was informed throughout its design by patient and public involvement through collaboration with representatives from MyGutFeeling - Stomach Cancer Foundation of Canada, a key advocacy group for gastric cancer survivors. Patients and patient-advocates were first engaged during the formulation of the research question. Patient partners provided feedback helping ensure the analyses remained aligned with survivorship priorities and could generate patient-relevant guidance for long-term care. Patient partners will not be involved in data extraction or analysis, but will play an essential role in dissemination. With their input, we plan to co-develop patient-friendly summaries and educational materials to share results with the broader gastric and GI cancer communities.

Amendments: Any important amendments to this protocol will be documented in detail, along with the date, rationale and description of the change. Deviations will be updated in the PROSPERO record and will be clearly described and justified in the final published review.

Data availability statement

Data are available upon reasonable request.