Assessment of bone-targeting agents use in patients with bone metastasis from breast, lung or prostate cancer using structured and unstructured electronic health records from a regional UK-based hospital
===========================================================================================================================================================================================================

* Anouchka Seesaghur
* Peter Egger
* Joshua Warden
* Ali Abbasi
* Bethany Levick
* Majid Riaz
* Peter McMahon
* Matthew Thompson
* Sue Cheeseman

## Objective

To assess use of bone-targeting agents (BTA) in patients with confirmed bone metastases (BM) from breast cancer (BC), non-small cell lung cancer (NSCLC) or prostate cancer (PC).

**Design** Retrospective cohort study.

**Setting** Regional hospital-based oncology database of approximately 2 million patients in England.

**Participants** Patients aged ≥18 years with a diagnosis of BC, NSCLC or PC as well as BM between 1 January 2007 and 31 December 2018, with follow-up to 30 June 2020 or death; BM diagnosis ascertained from recorded medical codes and unstructured data using natural language processing (NLP).

**Main outcomes measures** Initiation or non-initiation of BTA following BM diagnosis, time from BM diagnosis to BTA initiation, time from first to last BTA, time from last BTA to death.

**Results** This study included 559 BC, 894 NSCLC and 1013 PC with BM; median age (Q1–Q3) was 65 (52–76), 69 (62–77) and 75 (62–77) years, respectively. NLP identified BM diagnosis from unstructured data for 92% patients with BC, 92% patients with NSCLC and 95% patients with PC. Among patients with BC, NSCLC and PC with BM, 47%, 87% and 88% did not receive a BTA, and 53%, 13% and 12% received at least one BTA, starting a median 65 (27–167), 60 (28–162) and 610 (295–980) days after BM, respectively. Median (Q1–Q3) duration of BTA treatment was 481 (188–816), 89 (49–195) and 115 (53–193) days for patients with BC, NSCLC and PC. For those with a death record, median time from last BTA to death was 54 (26–109) for BC, 38 (17–98) for NSCLC and 112 (44–218) days for PC.

**Conclusion** In this study identifying BM diagnosis from both structured and unstructured data, a high proportion of patients did not receive a BTA. Unstructured data provide new insights on the real-world use of BTA.

*   Breast tumours
*   ONCOLOGY
*   Urological tumours
*   Respiratory tract tumours
*   Epidemiology

### STRENGTHS AND LIMITATIONS OF THIS STUDY

*   Our study uses both structured and unstructured patient medical history data to address the study aims.

*   The unstructured data is evaluated through natural language processing techniques.

*   Prescribing data originates from multiple data sources, and includes both inpatient and outpatient data.

*   This study relies on the quality and completeness of data collected from hospital records.

*   Insights from this study are limited to the routine practice in one regional area in the UK.

## Introduction

Bone is a frequent site of metastasis for breast cancer (BC), non-small cell lung cancer (NSCLC) and prostate cancer (PC), occurring in approximately 70% of patients with advanced BC,1 2 in 30–40% of all patients with NSCLC3 4 and in 80% of patients with advanced PC.5 6 Bone metastasis (BM) is a major cause of morbidity leading to severe pain, mobility difficulties and bone complications, also known as skeletal-related events.7–9 Bone-targeting agents (BTAs) reduce skeletal morbidity from metastatic bone disease and are used in patients with BMs across several tumour types. For most patients, whether symptomatic or not, clinical guidelines recommend starting a BTA as soon as bone metastases (BMs) are diagnosed.10–12

Records of BM depend on imaging practices in routine clinical practice. Imaging at baseline is performed to stage the patient and define the patient’s ongoing management. Throughout a patient’s disease journey, other imaging assessments may occur but repeat scans are not routinely performed unless clinically indicated. In electronic medical records (EMR), BM diagnoses are often not coded using medical codes,13 14 and may be captured in unstructured free text. Studies relying solely on BM diagnosis identified via structured data, may therefore, lead to an incomplete picture of the management of patients with cancer and BM.

To address these gaps in evidence on BM ascertainment, we used novel techniques to identify BMs in both structured medical code-based data, and unstructured free-text data from the hospital-based EMR database of the largest integrated regional cancer centre in the UK. This allowed us to identify a patient with comprehensive BM population to better understand the management of BM in patients with cancer. The current study aims to evaluate the real-world use and non-use of BTAs in patients with BC, NSCLC or PC with a BM diagnosis.

## Methods

### Outcomes measures

The main outcome measures were initiation or non-initiation of BTA following BM diagnosis, time from BM diagnosis to BTA initiation, time from first to last BTA and time from last BTA to death. Further details on BTAs used including extent of use were provided. Patient demographic and clinical characteristics as well as the treatment histories by tumour type, and by BTA use/non-use were also described.

### Data source

This hospital-based cohort study used EMR data from the REAL-Oncology database of England National Health Service (NHS) Leeds NHS Teaching Hospital Trust (LTHT). REAL-Oncology receives patient-level data directly from various clinical information systems, and each data source is linked at the patient-level via the patient’s unique identifier (figure 1).

![Figure 1](http://bmjopen.bmj.com/https://bmjopen.bmj.com/content/bmjopen/13/5/e069214/F1.medium.gif)

[Figure 1](http://bmjopen.bmj.com/content/13/5/e069214/F1)

Figure 1 
Leeds National Health Service Teaching Hospitals Trust (LTHT) data sources and linkages to create study data set. EMR, electronic medical records. MDT, multidisciplinary Team. JAC and EPRO are proprietary company names.

A two-phase approach was adopted to assess BTA use in patients with cancer and BM using secondary and tertiary care data. In Phase I, we applied novel techniques to identify patients with confirmed BM across all existing EMRs, whether structured or unstructured. In Phase II, we evaluated the use of BTAs within the identified study cohort. The study complied with the Hospital Trust’s Information Governance requirements. All data was fully anonymised and patients who had opted out of data sharing were removed from the study. Researchers do not work with identifiable data and work within a secure environment on a secure NHS network.

### Phase I: identification of BM diagnosis

Adult patients (aged ≥18 years at the date of primary cancer diagnosis) with a primary diagnosis of BC, PC and NSCLC (index date) were identified through International Classification of Diseases (ICD)-10 codes (Appendix A) (and additionally ICD Oncology 3rd edition (ICD-O-3) morphology codes for NSCLC, Appendix B) during the study period from 1 January 2007 to 31 December 2018. Patients with other primary malignancies prior to the index date or enrolled in a randomised controlled trial on BTA were excluded.

We included patients who had a BM either at their first diagnosis of primary cancer or developed BM at any time after initial primary cancer diagnosis. The BM diagnoses were identified via a BTA record, direct coding of BM and query of unstructured text from imaging, pathology and clinical summary reports using a natural language processing (NLP) approach. The NLP platform Interactive Information Extraction (I2E), that was developed by the company Linguamatics ([https://www.linguamatics.com/products/i2e](https://www.linguamatics.com/products/i2e)), was used to automate reviewing of unstructured text by looking for inbuilt and predefined keywords and phrases defined by clinical physicians with experience in diagnosing and treating patients with BM. A large percentage of the NLP-identified BM cases were manually checked by the data review team consisting of a senior physician and a data quality officer, and the information from this was used to improve the NLP query in a continuous feedback loop of checking and adjusting. Finally, all identified BM cases were manually reviewed by the data review team to provide final confirmation.

### Phase II: assessment of BTA use

#### Study population

From Phase I, all adult patients with BC, NSCLC or PC and a confirmed diagnosis of BM (identified from 1 January 2007 to 31 December 2018) were followed from BM diagnosis date to 31 June 2020 or death.

#### BTA treatment

BTA treatment was determined through patients EMRs linked to the hospital pharmacy dispensing database JAC covering both inpatient and outpatient prescriptions, and the treatment prescribing database ChemoCare. We reported three phases of medication adherence (initiation, implementation and persistence) as recommended by the European Society for Patient Adherence, COMpliance, and Persistence.15 BTAs included two different classes of anti-resorptive agents: bisphosphonates (both intravenous and oral) and the Receptor Activator of Nuclear factor Kappa-B Ligand (RANKL) inhibitor denosumab.

#### Statistical analysis

Primary cancers, BM cases and BTA use, including type of BTA and switches between BTAs, were reported as counts and percentages. Patient characteristics were reported as percentages for categorical variables and medians (Q1–Q3) for continuous variables. The Kaplan-Meier method was applied to analyse time-to-event data of BTA records, such as time to first BTA, duration of BTA and time from last BTA to death. Counts of <6 were marked as such in all results to protect patient privacy. The SAS V.9.4 (SAS, Cary, North Carolina, USA) and R V.3.216 was used for all data management and statistical analyses.

#### Patient and public involvement

Patients were not involved in this study.

## Results

### Phase I: identification of BM diagnosis

In Phase I, we identified a total of 6142 patients with BC, 5202 patients with NSCLC and 5382 patients with PC primary cancer. Table 1 shows a summary of the different approaches and corresponding results for identifying BM diagnoses. Each of these approaches were reviewed to ascertain confirmation of a BM diagnosis: direct identification by NLP, identification by proxy based on a record of BTA treatment, identification by proxy based on a record of spinal cord compression and direct identification via diagnosis codes in structured EMR.

View this table:
[Table 1](http://bmjopen.bmj.com/content/13/5/e069214/T1)

Table 1 
Attrition table of study patient population in phase I

Table 1 shows the numbers and percentages of the three different methods of BM identification: NLP of unstructured data, evidence of spinal cord compression and BM in coded EMR fields. For BC, 573 patients were identified, with 527 (92%) via NLP-based querying of unstructured data. For NSCLC the total was 899, with 829 (92%) from unstructured data. For PC the total was 1017 and the results for unstructured data were 963 (95%). Further clinical expert review of all resulting cases detected additional false positives and yielded a final study cohort for BC, NSCLC and PC: 559 (9% of all primary cancer cases), 894 (17%) and 1013 (19%) patients with BM, respectively.

### Phase II: assessment of BTA use

Table 2 shows the patient demographic and clinical characteristics as well as the treatment histories of the final study cohort stratified by tumour type, and by BTA use/non-use. BTA initiation, implementation and persistence are shown in figure 2, followed by further details of the two most frequent BTAs in table 3.

View this table:
[Table 2](http://bmjopen.bmj.com/content/13/5/e069214/T2)

Table 2 
Patient characteristics of patient with final BM cohort

![Figure 2](http://bmjopen.bmj.com/https://bmjopen.bmj.com/content/bmjopen/13/5/e069214/F2.medium.gif)

[Figure 2](http://bmjopen.bmj.com/content/13/5/e069214/F2)

Figure 2 
BTA adherence: initiation, implementation and persistence. BM, bone metastasis; BTA, bone-targeting agent; NSCLC, non-small cell lung cancer.* includes patients who had BTA before their BM diagnosis** number of patients with a duration of at least one day

View this table:
[Table 3](http://bmjopen.bmj.com/content/13/5/e069214/T3)

Table 3 
BTA administration in patients with BM across the three cancers

#### Breast cancer

Among 559 patients with BC and BM, 47% (n=265) did not have a BTA prescription and 53% (n=294) received at least one BTA prescription, starting a median (Q1–Q3) of 65 (27–167) days from their BM diagnosis date (inclusive) to their first BTA initiation date (excludes nine patients with a BTA before BM). Median (Q1–Q3) duration of BTA therapy from first to last BTA record was 481 (188–816) days and median (Q1–Q3) time from last BTA to death was 54 (26–109) days (figure 2). Most patients (86%, n=254) received only one type of BTA. Table 3 provides details of two specific BTAs of different classes that were administered, the RANKL inhibitor denosumab (n=56, 19.1%) and the bisphosphonate zoledronic acid (n=229, 77.9%), both with the most frequent cycle duration of 28 days. During the follow-up period, a total of 52 switches were observed between BTAs. Of those, switches between denosumab and zoledronic acid were the most frequent: 30% (17/56) of all denosumab administrations ended in a switch to zoledronic acid, within a median (Q1–Q3) time to switch of 32 (28–57) days, and 5% (11/229) of all zoledronic acid administrations ended in a switch to denosumab, within a median (Q1–Q3) time of 78 (35–216) days. Patients with BTAs had a numerically higher percentage of oestrogen receptor status positive, progesterone receptor status positive, human epidermal growth factor receptor status compared with patients without a BTA (table 2).

#### NSCLC

Among the 894 patients with NSCLC and BM, 87% (n=777) did not receive a BTA prescription and 13% (n=117) received at least one BTA prescription, starting a median (Q1–Q3) of 60 (28–162) days from their BM diagnosis date (inclusive) to their BTA initiation date (excludes eight patients with a BTA before BM). Median (Q1–Q3) duration of BTA therapy from first to last BTA record was 89 (49–195) days and median (Q1–Q3) time from last BTA to death was 38 (16–98) days (figure 2). A total of 12 patients with NSCLC received denosumab and 93 patients received zoledronic acid (table 3), both with the most frequent cycle duration of 28 days. The median number of administrations per patient was 2 (Q1–Q3: 1–11) for denosumab, and 1 (Q1–Q3: 1–3) for zoledronic acid. A total of 114 (97%) patients received only one type of BTA and <6 switches occurred between BTAs. Patients with BTAs had a numerically higher percentage of estimated glomerular filtration rate, anaplastic lymphoma kinase and programmed cell death receptor ligand-1 mutation data missing as well as a higher percentage of a history of overall radiotherapy (RT) or systemic anticancer therapy (SACT) or surgery compared with patients without a BTA (40.2% (=100%–59.8%) vs 31.5% (=100%–68.5%)) (table 2).

#### Prostate cancer

Among the 1013 patients with PC and BM, 88% (n=892) did not receive a BTA prescription and 12% (n=121) received at least one BTA prescription, starting a median (Q1–Q3) of 611 (295–980) days from their BM diagnosis date (inclusive) to their BTA initiation date (excludes one patient with a BTA before BM). Median (Q1–Q3) duration of BTA therapy from first to last BTA record was 115 (53–193) days and median (Q1–Q3) time from last BTA to death was 112 (44–218) days (figure 2). There were no patients on denosumab while 113 patients received zoledronic acid (table 3), with the most frequent cycle duration of 28 days. The median number of administrations per patient was 2 (Q1–Q3: 1–4) for zoledronic acid. Patients with PC BTA only had a record of one unique BTA with no switching recorded. Patients with BTA prescriptions had a numerically higher percentage of history of overall RT or SACT or surgery compared with patients without a BTA prescription (36.4% (=100%–63.6%) vs 17.1% (=100%–72.9%)) (table 2).

## Discussion

### Use of structured and unstructured data to identify patients with BM within LTHT

In this study, over 90% of all BM cases were identified through NLP-based querying of unstructured data. Healthcare professionals typically record BM detected during different diagnostic procedures in both structured and unstructured formats. Restricting the analysis to structured medical codes would have significantly underestimated the occurrence of BM in the three cancer cohorts in this hospital-based setting. Hence, use of NLP greatly enhanced the efficiency of the identification of BM cases from multiple unstructured data sources. The need for clinical review to eliminate false positive cases shows that further refinement of NLP models is still required.

### BTA usage in patients with BC and BM

A European multicountry study (von Moos *et al*17) found that 88% of patients with BC with BM received BTA treatment, while 53% of patients with BC with BM received BTA treatment in our study. There are key differences between the studies. The von Moos *et al* study collected data in a cross-sectional survey of physicians that were treating patients with BM who were actively receiving treatment for their cancer. In contrast, 57% of all patients with BM in our study did not have a record of SACT, even though some of it may be a result of prescribing recorded outside the available data systems. A prospective study using a German tumour registry (Schröder *et al*18) reported a BTA treatment of 89% in patients with BC with BM with a median time to treatment from BM diagnosis of 3 weeks. Data collection was prospective and focused on an anticancer treated cohort, including outpatient treatment data. In contrast, our study obtained treatment data retrospectively from potentially incomplete hospital treatment databases and did not include treatment outside the hospital. Furthermore, there may be genuine differences in the use of BTAs in patients with cancer with BM between the UK and other European countries. Determinants of BTA prescribing in patients with cancer with BM were evaluated in several studies.17 19 20 Findings from von Moos *et al*17 indicate that some physicians base their BTA treatment decisions not only on clinical guidelines but also consider patients’ Eastern Cooperative Oncology Group (ECOG) performance score, disease burden and the presence of other sites of metastatic disease. For example, a patient with an ECOG performance score of 0–2 is considered fit enough to receive BTA treatment, but in the presence of extensive liver disease and low burden of bone disease, may not routinely receive a BTA. Our study showed a numerical difference in ECOG scores between patients with BTA and without BTA: 94% of patients with a BTA and 81% without a BTA had an ECOG score of 0–2. These findings suggest that BTA treatment is determined on a case-by-case basis within this setting and is not solely reliant on BTA guidelines.

### BTA usage in patients with NSCLC and BM

Diel *et al*21 investigated 242 patients with lung cancer with a diagnosis of BM and who received at least one BTA treatment in Germany from 2011 to 2015. Of these patients, 15% received denosumab and 63% zoledronic acid, while our study observed 10% of patients with NSCLC BM receiving denosumab and 80% zoledronic acid. The probability of patients still on denosumab after 6 months was 87% in Diel *et al*, compared with 13% in our study. The 2014 European Society for Medical Oncology bone health guidelines, which cover some of the German study time period recommend zoledronic acid or denosumab in patients with a life expectancy of greater than 3 months.10 The patients with NSCLC within our study had a median follow-up time of 87 (Q1−Q3: 37−205) days.

The low proportion of patients receiving BTA within the current study is likely due to the poor prognosis of these patients. Overall, survival data published by LTHT on patients with advanced non-squamous NSCLC (stage IIIB–IV) showed patients had a median survival of 4.1 months between 2007 and 2012 and 5.0 months between 2013 and 2017.22 The ECOG score further reflects the burden of disease in this population: 67% of patients with BTA with a score of 0–2 and 49% in patients without BTA.

### BTA usage in patients with PC and BM

The European multicountry study (von Moos, *et al*17 included an evaluation of patients with castrate-resistant prostate cancer (CRPC) and reported that 77% of patients with CRPC received at least one BTA. The von Moos *et al* study included patients who were actively receiving anticancer therapy. In our study, over 71% were diagnosed at stage IV and had BM at the time of PC diagnosis and 72% had no record of any other treatment such as SACT or surgery. A US-based study using claims and commercial databases, (Hernandez, *et al*7) identified BTA use in 52% of patients with PC with BM in 2012. Median time to first BTA was 35 and 37 days, respectively, for the claims and commercial databases. In our study we observed that 12% of patients with BM received a BTA, with a median time to first BTA of 610 days. While the National Institute for Health and Care Excellence guidelines do not recommend denosumab for patients with PC with BM in the UK,12 it is approved for use in patients with PC in the USA.23

## Strengths and limitations

A key strength of the current study is the use of extensive unstructured data from multiple EMR sources and application of NLP techniques to identify patients with BM. Leveraging unstructured data is especially important because BMs are likely identified at different diagnostic investigations and reported in different medical records. Access to multiple data sources and linkage within the LTHT database and the application of NLP methods enabled a more comprehensive account of the patient’s medical record data. Our findings show that the vast majority of BM cases would have been missed without evidence from unstructured medical record data, as BMs are typically not recorded through structured medical codes in this particular setting. In addition, the availability of both inpatient and outpatient prescribing data from multiple data sources is a strength of the study.

Nevertheless, the study has some limitations due to the capture and documentation of inpatient BTA prescribing information. The comprehensive hospital drug dispensing data (JAC) is only available for the last 5 years. Although BTA treatment is also included in the oncology treatment database ChemoCare, BTA treatment is not always recorded within ChemoCare, especially for patients who receive a BTA during an inpatient admission. Hence, medications that were not prescribed using ChemoCare, including hormone therapy, and that were prescribed more than 5 years ago (not included in JAC), are not captured in this study. However, an assessment of BTA prescribing before and after the period of JAC availability showed only a marginal difference in BTA prescribing between the two periods. In addition, insights from this study are limited to the routine practice in the UK and reflect existing restrictions in reimbursement and access to BTA therapy within the country.

## Conclusion

To our knowledge, this is the first study that retrospectively identified patients with BM using both structured and unstructured data within England to characterise BTA use in clinical practice. Applying NLP to unstructured data should be considered as a useful additional strategy to identify BM and ascertain cases which would have been missed if only structured data were used. This study provided a different picture to existing literature on BTA use in Europe and the USA, highlighting the underuse of BTA treatment within patients with metastatic bone disease from BC, NSCLC or PC. These findings point to a complex decision-making process to prescribe bone protection therapy to patients with cancer. Further work is warranted to better understand individual patient medical need and treatment benefit, including repeating this work in other data sources to assess the benefit of using unstructured data.

### Supplementary data

[[bmjopen-2022-069214supp001.pdf]](pending:yes)

## Data availability statement

Data are not publicly available.

## Ethics statements

### Patient consent for publication

Not applicable.

## Acknowledgments

The authors acknowledge the contribution of Nina Snelling3 to the research work.

## Footnotes

*   Contributors AS, AA, PM and SC made substantial contributions to the design of the work. BL, MR, MT and PE contributed to the analysis of the data. AS, AA, PE and SC contributed to the interpretation of data. JW drafted the work. AS and PE made substantial contributions to substantively revise the manuscript. All authors reviewed the manuscript, provided final approval of the version to be published and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. SC is the guarantor.

*   Funding This study was funded by Amgen Ltd. The grant number is N/A.

*   Competing interests I have read the journal’s policy and the authors of this manuscript have the following competing interests. AS was an employee and equity holder in Amgen during the conduct of the study. PE, BL and MT were employed with IQVIA during the conduct of the study. JW had no conflict of interest to declare. AA reported contract work with Amgen. MR was employed with IQVIA and had an honorary contract with Leeds Teaching Hospitals Trust (LTHT) to access the data to produce analysis. PM worked for IQVIA during the initial development of the manuscript, and the analysis time period. SC’s part was funded by IQVIA.

*   Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

*   Provenance and peer review Not commissioned; externally peer reviewed.

*   Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

[http://creativecommons.org/licenses/by-nc/4.0/](http://creativecommons.org/licenses/by-nc/4.0/)
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: [http://creativecommons.org/licenses/by-nc/4.0/](http://creativecommons.org/licenses/by-nc/4.0/).

## References

1.  Pulido C, Vendrell I, Ferreira AR, et al. Bone metastasis risk factors in breast cancer. Ecancermedicalscience 2017;11:715. [doi:10.3332/ecancer.2017.715](http://dx.doi.org/10.3332/ecancer.2017.715)
    
    

2.  Xiong Z, Deng G, Huang X, et al. Bone metastasis pattern in initial metastatic breast cancer: a population-based study. Cancer Manag Res 2018;10:287–95. [doi:10.2147/CMAR.S155524](http://dx.doi.org/10.2147/CMAR.S155524)
    
    [CrossRef](http://bmjopen.bmj.com/lookup/external-ref?access_num=10.2147/CMAR.S155524&link_type=DOI) 

3.  Santini D, Barni S, Intagliata S, et al. Natural history of non-small-cell lung cancer with bone metastases. Sci Rep 2015;5. [doi:10.1038/srep18670](http://dx.doi.org/10.1038/srep18670)
    
    

4.  D’Antonio C, Passaro A, Gori B, et al. Bone and brain metastasis in lung cancer: recent advances in therapeutic strategies. Ther Adv Med Oncol 2014;6:101–14. [doi:10.1177/1758834014521110](http://dx.doi.org/10.1177/1758834014521110)
    
    [CrossRef](http://bmjopen.bmj.com/lookup/external-ref?access_num=10.1177/1758834014521110&link_type=DOI) 
    
    [PubMed](http://bmjopen.bmj.com/lookup/external-ref?access_num=24790650&link_type=MED&atom=%2Fbmjopen%2F13%2F5%2Fe069214.atom) 

5.  Liu D, Kuai Y, Zhu R, et al. n.d. Prognosis of prostate cancer and bone metastasis pattern of patients: a SEER-based study and a local hospital based study from China. Sci Rep;10. [doi:10.1038/s41598-020-64073-6](http://dx.doi.org/10.1038/s41598-020-64073-6)
    
    

6.  Macedo F, Ladeira K, Pinho F, et al. Bone metastases: an overview. Oncol Rev 2017;11:321. [doi:10.4081/oncol.2017.321](http://dx.doi.org/10.4081/oncol.2017.321)
    
    

7.  Hernandez RK, Adhia A, Wade SW, et al. Prevalence of bone metastases and bone-targeting agent use among solid tumor patients in the United States. Clin Epidemiol 2015;7:335–45. [doi:10.2147/CLEP.S85496](http://dx.doi.org/10.2147/CLEP.S85496)
    
    [PubMed](http://bmjopen.bmj.com/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fbmjopen%2F13%2F5%2Fe069214.atom) 

8.  Asdahl PH, Sundbøll J, Adelborg K, et al. Cardiovascular events in cancer patients with bone metastases-A Danish population-based cohort study of 23,113 patients. Cancer Med 2021;10:4885–95. [doi:10.1002/cam4.4027](http://dx.doi.org/10.1002/cam4.4027)
    
    

9.  Hjelholt TJ, Rasmussen TB, Seesaghur A, et al. Risk of infections and mortality in Danish patients with cancer diagnosed with bone metastases: a population-based cohort study. BMJ Open 2021;11:e049831. [doi:10.1136/bmjopen-2021-049831](http://dx.doi.org/10.1136/bmjopen-2021-049831)
    
    

10. Coleman R, Body JJ, Aapro M, et al. Bone health in cancer patients: ESMO clinical practice guidelines. Annals of Oncology 2014;25:iii124–37. [doi:10.1093/annonc/mdu103](http://dx.doi.org/10.1093/annonc/mdu103)
    
    [CrossRef](http://bmjopen.bmj.com/lookup/external-ref?access_num=10.1093/annonc/mdu103&link_type=DOI) 
    
    [PubMed](http://bmjopen.bmj.com/lookup/external-ref?access_num=24782453&link_type=MED&atom=%2Fbmjopen%2F13%2F5%2Fe069214.atom) 

11. Coleman R, Hadji P, Body J-J, et al. Bone health in cancer: ESMO clinical practice guidelines. Annals of Oncology 2020;31:1650–63. [doi:10.1016/j.annonc.2020.07.019](http://dx.doi.org/10.1016/j.annonc.2020.07.019)
    
    [CrossRef](http://bmjopen.bmj.com/lookup/external-ref?access_num=10.1016/j.annonc.2020.07.019&link_type=DOI) 
    
    [PubMed](http://bmjopen.bmj.com/lookup/external-ref?access_num=32801018&link_type=MED&atom=%2Fbmjopen%2F13%2F5%2Fe069214.atom) 

12. Overview | denosumab for the prevention of skeletal-related events in adults with bone metastases from solid tumours | guidance | (NICE). 2012.
    
    

13. Jensen AØ. n.d. Validity of the recorded International classification of diseases, 10th edition diagnoses codes of bone metastases and skeletal-related events in breast and prostate cancer patients in the Danish national Registry of patients. CLEP;1:101. [doi:10.2147/CLEP.S5446](http://dx.doi.org/10.2147/CLEP.S5446)
    
    

14. Liede A, Hernandez RK, Roth M, et al. Validation of international classification of diseases coding for bone metastases in electronic health records using technology-enabled abstraction. Clin Epidemiol 2015;7:441–8. [doi:10.2147/CLEP.S92209](http://dx.doi.org/10.2147/CLEP.S92209)
    
    [PubMed](http://bmjopen.bmj.com/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fbmjopen%2F13%2F5%2Fe069214.atom) 

15. De Geest S, Zullig LL, Dunbar-Jacob J, et al. ESPACOMP medication adherence reporting guideline (emerge). Ann Intern Med 2018;169:30–5. [doi:10.7326/M18-0543](http://dx.doi.org/10.7326/M18-0543)
    
    [PubMed](http://bmjopen.bmj.com/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fbmjopen%2F13%2F5%2Fe069214.atom) 

16. Team RC. A language and environment for statistical computing. R foundation for statistical computing. Vienna, Austria:; 2022. Available: [https://www.R-project.org/](https://www.R-project.org/)
    
    

17. von Moos R, Lewis K, Massey L, et al. Initiation of bone-targeted agents in patients with bone metastases and breast or castrate-resistant prostate cancer actively treated in routine clinical practice in Europe. Bone 2022;154. [doi:10.1016/j.bone.2021.116243](http://dx.doi.org/10.1016/j.bone.2021.116243)
    
    

18. Schröder J, Fietz T, Köhler A, et al. Treatment and pattern of bone metastases in 1094 patients with advanced breast cancer-results from the prospective German tumour Registry breast cancer cohort study. Eur J Cancer 2017;79:139–48. [doi:10.1016/j.ejca.2017.03.031](http://dx.doi.org/10.1016/j.ejca.2017.03.031)
    
    

19. Butler AM, Cetin K, Hernandez RK, et al. Treatment dynamics of bone-targeting agents among men with bone metastases from prostate cancer in the United States. Pharmacoepidemiol Drug Saf 2018;27:229–38. [doi:10.1002/pds.4360](http://dx.doi.org/10.1002/pds.4360)
    
    [PubMed](http://bmjopen.bmj.com/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fbmjopen%2F13%2F5%2Fe069214.atom) 

20. McGrath LJ, Overman RA, Reams D, et al. Use of bone-modifying agents among breast cancer patients with bone metastasis: evidence from oncology practices in the US. Clin Epidemiol 2018;10:1349–58. [doi:10.2147/CLEP.S175063](http://dx.doi.org/10.2147/CLEP.S175063)
    
    

21. Diel I, Ansorge S, Hohmann D, et al. Real-World use of denosumab and bisphosphonates in patients with solid tumours and bone metastases in Germany. Support Care Cancer 2020;28:5223–33. [doi:10.1007/s00520-020-05357-5](http://dx.doi.org/10.1007/s00520-020-05357-5)
    
    [PubMed](http://bmjopen.bmj.com/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fbmjopen%2F13%2F5%2Fe069214.atom) 

22. Snee MCS, Thompson M. Temporal trends in treatment and overall survival among patients with incident NSCLC in the UK: a REAL-oncology database analysis from the I-O optimise initiative. presented at: 9th European lung cancer Congress (ELCC) [Geneva, Switzerland]. 2019. Available: [https://www.leedsth.nhs.uk/assets/ca7effdffd/28.-Snee\_REAL-ONC\_44P\_ELCC2019\_FINAL-Poster\_4Apr19.pdf](https://www.leedsth.nhs.uk/assets/ca7effdffd/28.-Snee\_REAL-ONC\_44P_ELCC2019_FINAL-Poster_4Apr19.pdf)
    
    

23. FDA Approves Denosumab for the Treatment of Bone Loss in Patients With Prostate or Breast Cancer. Cancer Network: FDA, 2011.