Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC Immune checkpoint inhibitors (ICIs) have transformed cancer therapy but are associated with a range of immune-related adverse events (irAEs). Case series, pharmacovigilance data and population-based studies have suggested a possible increase in fracture risk among ICI users, leading to the proposal of fractures as a potential novel irAE. However, evidence from randomised controlled trials (RCTs) specifically addressing this issue has been scarce and inconsistent, with guidelines from ASCO and ESMO not currently recognising fractures as a direct consequence of ICI use.WHAT THIS STUDY ADDS This systematic review and meta-analysis of 30 RCTs including 19 407 patients with solid cancers provides the first comprehensive synthesis of fracture risk in ICI monotherapy. The findings show a trend towards an 18% higher incidence of clinical non-pathological fractures among ICI users compared with non-ICI controls, with a more pronounced 47% increase when compared against chemotherapy alone. While these estimates did not reach statistical significance due to the low number of events and wide CIs, the results highlight possible fracture risk signals in specific subgroups such as patients without metastatic disease or those with prior chemotherapy exposure.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings underscore the need for prospective monitoring of skeletal health in patients treated with ICIs and call for systematic inclusion of fractures as predefined safety outcomes in future oncology trials. For clinical practice, awareness of potential fracture risk may guide proactive bone health management, particularly in higher-risk subgroups. From a policy perspective, the study highlights the importance of standardised adverse event reporting across RCTs to better capture uncommon but clinically significant treatment-related complications.Introduction Immune checkpoint inhibitors (ICI) have revolutionised oncological treatment and are currently considered the standard of care for several malignancies, being used by about 45% of patients with cancer in the USA. 1 ICI disrupt regulatory pathways of the immune system enhancing the activity of cytotoxic T-cells to target and eliminate cancer cells.2 However, immune-related adverse events (irAEs) occur frequently, are mainly related to the mechanism of action of ICI and may affect practically every organ system of the body.3 4 Case studies,5 6 analysis of pharmacovigilance data from the US Food and Drug Administration Adverse Event Reporting System5 and observational cohort studies from Canada7 and the USA8 have reported an early increase in the incidence of fractures in patients treated with ICIs leading to proposals to consider fractures a novel irAE of ICI.7 9–11 In the Canadian study of 1600 patients, fracture rates increased from 11.3 per 1000 patient years to 27.3 the first year of ICI treatment and declined to 17.6 in the second year. In the US cohort of 3137 patients with melanoma, major osteoporotic fractures rose from 1.3% to 1.8% in both the first and second year of ICI treatment.8 In sharp contrast, however, a study using VigiBase, which is the WHO’s global pharmacovigilance database and has accumulated data on approximately 36.6 million patients and 92.6 million adverse drug reactions, showed that the use of ICI does not increase the frequency of osteoporosis or fractures.12 To date, guidelines of the diagnosis and management of irAEs in patients treated with ICI from American Society of Clinical Oncology (ASCO)13 and European Society of Medical Oncology (ESMO)14 mention fractures only in relation to use of glucocorticoids given for alleviation of symptoms related to irAEs. Clarification, however, of the relationship between ICI treatment and bone fragility apart from its obvious clinical significance, is important for the management of patients with fractures. Compared with other serious adverse effects, fractures do not require discontinuation of ongoing treatments because these can be effectively treated or even prevented in patients at risk. Moreover, the bisphosphonate zoledronate, apart from its well-documented antifracture efficacy,15 was also shown to significantly reduce the risk of cancers in postmenopausal women ≥65 years,16 17 an action that may be related to the reported reduction of the production of cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) by this agent.18 To address the potential association between ICI use and fractures, we conducted a meta-analysis of the incidence of non-pathological clinical fractures using aggregated data reported in randomised controlled trials (RCTs) that compared the efficacy and safety of ICI monotherapy with those of placebo, no treatment or chemotherapy.Methods This meta-analysis conformed to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines. The review protocol is registered in the PROSPERO database (CRD42023448831). The literature search was conducted in PubMed (MEDLINE), Embase and the Cochrane Central Register of Controlled Trials from inception until December 2024. The search strategy incorporated both Medical Subject Headings and Emtree terms, along with free-text keywords, to identify relevant studies on ICI, specifically those targeting CTLA-4, programmed cell death protein 1 (PD-1) and programmed death-ligand 1 (PD-L1), together with filters for RCTs. We restricted our analysis to ipilimumab, nivolumab, pembrolizumab, atezolizumab, avelumab, durvalumab, cemiplimab and tremelimumab because these agents represent approved ICIs with established efficacy across a wide range of malignancies. Other investigational or less commonly used ICIs were excluded due to limited clinical use and insufficient RCT evidence. This approach allowed us to focus on clinically relevant agents and provide results that are directly applicable to current oncology practice. Only articles published in English were considered eligible for inclusion.All RCTs that included adult (>18 years of age) patients with solid cancers treated with ICI as monotherapy versus placebo/no treatment/chemotherapy were considered eligible for analysis. Exclusion criteria were as follows: (1) case reports, (2) case series, (3) abstracts and data from unpublished studies, (4) narrative reviews, (5) systematic reviews and meta-analysis, (6) observational studies, (7) RCTs that included combined ICI treatment and no inclusion of ICI monotherapy drug arm, (8) RCTs including patients with haematological malignancies and (9) RCTs that did not include a non-ICI control arm.The primary outcome of the meta-analysis was the incidence of clinical fractures in patients with solid cancer treated with ICI as monotherapy compared with placebo/no treatment/chemotherapy.Sources of information included journal articles, published trial protocols and/or ClinicalTrials.gov registration. In case of missing information, we communicated with the corresponding authors or the sponsor via email.We noticed that eligible clinical trials did not report the number of fragility fractures as adverse events in their full text publications or online supplemental material. To overcome this issue, we searched ClinicalTrials.gov for individual relevant adverse event data using each unique RCT registration number. Since individual patient data were not available, the study was based on aggregated data reported in the included RCTs.SP110.1136/bmjonc-2025-000868.supp1Supplementary data Fracture data were extracted from the ‘Adverse Events’ section of ClinicalTrials.gov and categorised using the predefined terminology available on the platform (eg, vertebral, spinal compression, hip, femur, femoral neck, upper limb, lower limb, pathological, osteoporotic). These terms were subsequently grouped into broader categories: (1) vertebral fractures, (2) non-vertebral fractures, (3) hip fractures, (4) pathological fractures (typically referring to fractures at sites of bone metastases) and (5) osteoporotic fractures (used when no specific site was indicated).Skeletal-related adverse events (ie, metastatic lytic lesions that could lead to fractures defined as pathological fractures) were also excluded from our analysis, as our focus was on assessing fragility-related fractures. Due to the overall low incidence of fragility fractures across studies, all anatomical fracture types were ultimately aggregated in the final analysis to improve statistical power and interpretability.Risk of bias analysis was performed according to Cochrane Risk of Bias tool V.2.0.We also conducted prespecified subgroup analyses to explore whether the effect of ICI therapy on fracture risk varied by:(1) M-cancer stage (metastatic vs non-metastatic disease); (2) prior chemotherapy exposure (drug-naïve patients vs those previously treated with any chemotherapy regimen); (3) ICI drug class (stratified by anti-PD-1, anti-PD-L1 or anti-CTLA-4 therapy); (4) treatment duration; (5) age; (6) sex; and (7) cancer type—stratified by tumour site, where data permitted.The primary outcome measure was ORs with corresponding 95% CIs. Meta-analysis was performed using both fixed and random effects models, while we assessed heterogeneity using the Q test and the I2 metric. No asymmetry was observed in the funnel plot, and similar results were obtained by the Egger’s test (p=0.934), indicating that there was no statistically significant evidence of small study effects in our data set (online supplemental figure 1)Patients and members of the public have not been involved in this research study.Full details of the search strategy, eligibility criteria, data extraction and methodology used are described in online supplemental material.Results A total of 46 RCTs were considered potentially eligible in the systematic review ( figure 1). The search process outlined in the PRISMA schema (figure 1) is described in detail in online supplemental material. In 16 RCTs, the incidence of fractures was not reported and we tried to contact the corresponding authors. The authors of seven RCTs responded but were not able to provide any information about the incidence of fractures; nine authors did not respond to our request. Subsequently, 16 RCTs were excluded from further analysis (online supplemental table 1).Figure 1PRISMA flow chart. *Eligible clinical trials that reported no information of clinical fractures on relevant publications, online supplemental material or protocol registration on ClinicalTrials.gov, even after personal communication with the corresponding authors via e-mail were excluded from further analysis. From: Page et al. 30 For more information, visit: http://www.prisma-statement.org/. ICI, immune checkpoint inhibitor; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.30 RCTs, involving 19 407 patients (13 022 males and 6385 females), were included (online supplemental table 2). Among these, 9331 patients comprised the non-ICI control group (3193 received placebo, 403 received no treatment and 5735 received chemotherapy), while 10 076 patients were treated with ICI monotherapy.These studies included the following solid cancer types: non-small cell lung cancer (n=9); small cell lung cancer (n=2); melanoma (cutaneous and advanced) (n=3); urothelial carcinoma (n=4); gastric/gastro-oesophageal cancer (n=5); triple-negative breast cancer (n=1); hepatocellular carcinoma (n=1); renal cell carcinoma (n=2); ovarian/fallopian/peritoneal cancer (n=1); cervical cancer (n=1); head and neck squamous cell carcinoma (n=1).121 incident fractures were ascertained; 68 in ICI users and 53 in non-ICI controls, and the follow-up period ranged from 7 to 78 months (median 19 months).Use of immune checkpoint inhibitor and fracture incidence Use of ICI was associated with an estimated 18% higher fracture incidence (OR 1.18) compared with non-ICI use, with a 95% CI of 0.82 to 1.70, not reaching statistical significance ( figure 2). In analyses restricted to studies with chemotherapy as the comparator (n=19), the effect estimate indicated higher odds of 47% in ICI users (OR 1.47), with the 95% CI ranging from a modest reduction of 7% to a considerable twofold increase (0.93 to 2.32) (figure 3; online supplemental figure 2a). When limited to RCTs with placebo as the comparator (n=10), the effect estimate was lower (OR 0.80; 95% CI 0.44 to 1.48) suggesting a 20% lower fracture incidence among ICI users (figure 3; online supplemental figure 2a), although imprecisely estimated with wide CIs.Figure 2Pooled ORs for fracture incidence related to ICI use compared with non-ICI users. Forest plot showing ORs with 95% CIs for fracture incidence in patients treated with ICIs compared with non-ICI controls across randomised controlled trials. The columns ‘Cases Doi’ and ‘Cases Control’ indicate the number of reported fracture events in the ICI-users and control groups, respectively. Each horizontal line represents the 95% CI for an individual study, with the square indicating the point estimate (size proportional to study weight) and the diamond indicating the overall pooled estimate using the Mantel-Haenszel method. The vertical line at OR=1.0 represents no difference in fracture risk. An OR greater than one suggests increased odds of fracture with ICI therapy, whereas an OR less than one suggests decreased odds. ICI, immune checkpoint inhibitor.Figure 3Subgroup analyses of fracture incidence associated with ICI users. Forest plot presenting ORs with 95% CIs for fracture incidence among ICI users compared with non-ICI controls, stratified by (1) comparator treatment (chemotherapy vs placebo), (2) M-cancer stage (non-metastatic vs metastatic disease), (3) prior exposure to chemotherapy (drug-naïve vs previous use of chemotherapeutic agents) and ICI drug-class (anti-CTLA-4, anti-PD-1, anti-PD-L1). Point estimates (squares) are proportional to study weight, and horizontal lines represent 95% CI. An OR greater than 1 indicates higher odds of fracture in ICI users relative to non-ICI controls, whereas an OR less than 1 indicates lower odds. CTLA-4, cytotoxic T-lymphocyte-associated protein 4; ICI, immune checkpoint inhibitor; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1.Sensitivity analyses Meta-regression analysis was performed to investigate the effect of (1) M-cancer stage, (2) previous exposure to chemotherapeutics and (3) the drug class of ICI used, in the incidence of fractures ( figure 3, online supplemental figure 2b–d).The highest OR was observed in patients without metastatic disease, where ICI users demonstrated nearly a twofold higher fracture incidence compared with non-ICI controls (OR 1.97; 95% CI 0.95 to 4.08). ICI users with prior chemotherapy exposure also had a higher incidence of fractures (OR 1.44; 95% CI 0.74 to 2.79). All other subgroup comparisons indicated little difference in fracture incidence between ICI users and non-ICI controls with wide confidence intervals (figure 3, online supplemental figure 2).Further analyses classified by treatment duration, age, sex and cancer type that were initially considered could not be performed due to limited variability and reporting. Treatment duration was broadly similar across studies, most patients were within a comparable adult age range (55–70 years), fracture outcomes were not reported separately by sex and cancer type was predominantly solid tumours without adequate representation of distinct subgroups.Risk of bias assessment All studies were randomised as per protocol. Most of them (20/30) had low risk of bias, eight studies raised some concerns: four in terms of the randomisation process, three for the allocation concealment and one study for the selection of reported results. Two RCTs showed high risk of bias, both in terms of the randomisation process. As the collection of adverse event information on fractures data was solely extracted from the Clinical Trials database, we did not consider this as a bias for selective outcome reporting. Details of the risk of bias are summarised in online supplemental figure 3. Sensitivity analyses by omitting the two high risk of bias trials also did not alter the results in the pooled fracture risk of ICI monotherapy versus non-ICI controls (online supplemental figure 4).Discussion Our systematic review and meta-analysis of 30 RCTs, comprising 19 407 patients with solid cancers, demonstrated a very low frequency of fractures, 0.7% in the 10 076 patients treated with ICI monotherapy and 0.6% in the 9331 patients who received placebo or chemotherapy, resulting in wide CIs for the estimate. The OR indicated an 18% increased fracture incidence in ICI users compared with non-ICI controls, although with wide CIs, not reaching statistical significance. When chemotherapy was used as the comparator, the estimated effect size suggested a 47% higher fracture incidence in ICI users. By contrast, comparisons with placebo suggested a lower fracture incidence in ICI users but again with wide CIs.In meta-regression analysis based on M-cancer stage or previous chemotherapy exposure, the highest OR was observed in ICI users without metastatic disease, who demonstrated a nearly twofold higher fracture incidence compared with non-ICI controls, and in ICI users with prior chemotherapy exposure that showed a 44% higher fracture incidence compared with non-ICI controls. Taken together, our findings show a higher fracture incidence in ICI users, particularly in certain clinical subgroups, although with considerable uncertainty probably due to the low number of fractures reported.Previous observational studies7 8 have reported higher frequencies of fractures within the first year after the initiation of ICI treatment (1.8% in both studies), including broader cancer types7 or treatment regimens (ie, combination of different types of ICI).7 8 By contrast, our analysis focused specifically on ICI monotherapy in RCTs with solid types of cancer only, which may inherently involve different risk profiles. Moreover, a recent report from the WHO VigiBase database, including 108 512 ICI users—the largest cohort reported to date—showed an even lower frequency of fractures (0.34%).12 These variations across studies indicate that the ascertainment and the reporting of fractures in the context of ICI therapy may be influenced by differences in study design, patient selection or data collection methods which could ultimately lead to under-reporting.The potential molecular mechanisms involved in ICI-related disruption of bone metabolism and structure19 are not yet fully elucidated. ICIs are known to stimulate the release of proinflammatory cytokines from activated T-cells and increase the production of RANKL, leading to enhanced osteoclast formation and activity.20 Consistent with this concept, reduced bone mineral density was reported in PD-1 and PD-L1 KO mice.21 However, in vitro and in vivo studies have challenged these findings. PD-1 inhibitors reduced the number and activity of human osteoclasts in vitro,22 while reduced number of osteoclasts and mild osteopetrosis, characterised by increased volume of cancellous bone attributed to decreased bone resorption, had been found in PD-1-deficient mice.23 Similarly, in other experiments, PD-1-deficient mice were protected against bone destruction induced by femoral inoculation of Lewis lung cancer cells, while intravenous administration of PD-1 antibody to wild type mice inhibited osteoclastogenesis.24 These findings raised even the question of whether PD-1 inhibitors might be a potential treatment of osteoporosis.10 Further support was recently provided by a report of significant osteogenesis by human osteoblasts obtained from patients on exposure to proinflammatory cytokines, as those released during treatment with ICIs.25 Adding further complexity, a recent in vivo study demonstrated a dual effect of genetic and pharmacological inhibition of PD-L1 on bone, revealing both protective and deleterious impacts on skeletal integrity, and identifying advanced age and female sex as risk factors for bone loss.26 27 The main strength of our study is the comprehensive review of the literature, by identifying the presence of fractures not only in the articles of included RCTs but also in their published results in ClinicalTrials.gov site. Limitations of our study may include the lack of individual patient data regarding risk factors related to bone fragility (eg, prior fracture history, osteoporosis treatment and glucocorticoid use) and the absence of bone health assessments in most RCTs. Second, while the time frame of the included studies (ranged from 7 to 78 months, with a median of 19 months) is consistent with the follow-up durations of large oncology trials and is comparable to that of observational studies reporting increased fracture risk may still be insufficient to fully capture the long-term skeletal effects of ICI. Third, the types of solid tumours included in our meta-analysis varied considerably and comprised malignancies with differing propensities for bone metastasis, precluding the evaluation of fracture incidence according to tumour type. Fourth, the exclusion of a significant proportion of eligible trials (16 out of 46) due to missing fracture data may introduce selection bias, while the reliance on ClinicalTrials.gov adverse event data may affect data consistency. Lastly, we should also consider general limitations inherent to meta-analyses, including challenges of synthesising heterogeneous patient populations and treatment regimens due to variability in the design of each trial. Nevertheless, fractures are a clinically objective outcome that is unlikely to be entirely overlooked, even in the presence of variable adverse event reporting standards.In conclusion, our meta-analysis provides the first comprehensive synthesis of available RCT data of the relationship between ICI use and fracture incidence in patients with malignant diseases, showing a trend towards an increase in fracture risk among ICI users with solid cancers. Our results, however, also highlight the current gaps in fracture reporting across RCTs with ICI users and underscore the urgent need for standardised and transparent adverse event reporting to improve the accuracy of safety assessments.Similar methodological approaches have been used to investigate other uncommon but clinically significant adverse events related to cancer immunotherapy. Recent studies have examined the incidence of non-infectious uveitis following ICI use and in patients receiving BRAF inhibitors, providing insights into how rare treatment-related complications can be identified and quantified using large-scale cohort and registry data.28 29 Future perspectives Our findings highlight important gaps in current evidence regarding the skeletal effects of ICI. Future randomised trials should prospectively include fractures as predefined safety outcomes and even ensure longer follow-up durations to capture delayed skeletal events. Detailed data on bone health—including prior fracture history, bone mineral density and use of antiosteoporotic or glucocorticoid therapy—should be systematically collected. Stratification by tumour type and presence of bone metastases is also essential, given that certain cancers are more prone to osteolytic lesions. Addressing these factors will improve our understanding of fracture risk in patients treated with ICI and guide appropriate bone health monitoring in clinical practice.