Development and validation of an algorithm to estimate incidence and prevalence of breast cancer patients with distant metastasis in Northern Ireland: A population-based study

doi:10.21203/rs.3.rs-3866487/v1

Introduction: The lack of systematic collection and reporting of breast cancer with distant metastasis (BCDM) means incidence and prevalence is largely unknown.

Methods: Patients with primary breast cancer (BC) (Stage I-III and “Stage not known”) and de novo patients (Stage IV at diagnosis) were identified from the Northern Ireland Cancer Registry and linked to hospital inpatient (PAS) and death records (GRO) over the period 1993 to 2020. An expert rules-based algorithm was developed to identify BCDM cases. This was validated using a dataset of 1,028 BC patients diagnosed in 2009 and followed up until 2017 for BCDM (184 BCDM cases). Sensitivity, specificity, positive and negative predictive values (PPV, NPV) were calculated, and incident and prevalent cases reported.

Results: Validation demonstrated high sensitivity (95.1%), specificity (99.2%), PPV (96.2%) and NPV (98.9%).

Between 2009–2020 incident cases were stable, averaging 254 per year, while the prevalent population increased from 587 to 911 (55% increase), suggesting improved survivorship.

In 2020 the highest proportion of patients living with BCDM were existing cases from previous years (71.2%), with 66.6% having progressive BCDM and 33% de novo BCDM.

Conclusion: The novel algorithm accurately identified BCDM cases which will help planning, care and support for BCDM patients.

Algorithm

breast cancer

distant metastasis

secondary

progression

prevalence

incidence

Breast cancer (BC) incidence has increased over recent decades, due to a number of factors including an aging population, development and extension of population-based screening programmes, greater public awareness of BC symptoms as well as various risk factors. Hormone related risk factors include later age of first childbirth, reduced breastfeeding periods and hormone replacement therapy usage. Other risk factors include elevated Body Mass Index (BMI) and alcohol consumption(1–8).

Increasing incidence, coupled with decreasing age-standardized mortality rates, has led to increased BC prevalence and greater numbers of women surviving well beyond their primary BC diagnosis in Northern Ireland (NI), the UK and worldwide (3,8,9). Consequently, there are increasing numbers of people diagnosed with Stage I-III BC who are at risk of subsequent disease progression, which can range from locoregional (LR) recurrence to distant metastasis (DM) (10–13).

BC with DM has poor survival outcomes, with less than 1 in 5 women with de novo Stage IV disease in NI surviving 5 years beyond diagnosis (8).

The growing number of BC survivors has implications for healthcare policy, planning and service delivery in both clinical and charitable sectors. There are rising costs associated with treatment, surveillance and follow-up of primary BC patients and, for those with disease progression to DM, additional treatments, surveillance and palliative care services.(14,15). One example of how rising demand is being managed in the primary breast cancer setting relates to the introduction of Self-Directed Aftercare (SDA) which encourages BC patients to self-manage their 5-year follow-up(16).

Cancer progression events, however, are not routinely collected by population-based cancer registries (PBCR) despite key indicators, such as the number and proportion of breast cancer patients with distant metastasis (BCDM) and disease-free survival periods, becoming increasingly important (17). Hospitals do not systematically collect data on BCDM and consequently the number of patients with BCDM is largely unknown (18,19).

Most previous studies measuring progression to DM across BC patient populations have used healthcare datasets, with PBCR studies rare and concentrated in a small number of countries (Belgium, Canada, Denmark, and USA) (11,20–27). In these studies, expert-driven rule-based or model-based algorithms have been developed but few studies have used population-based data and validated their algorithm against a gold standard dataset of known confirmed patients with BCDM (21,22,24,27,28).

BCDM studies based in the USA often use data from healthcare providers for development and validation of algorithms. Factors such as the lack of population-level coverage, patients moving in and out of catchment areas, Medicare claims data only applying to patients aged above 65 years and the use of hospital coding rather than standardised, internationally agreed coding used by PBCRs, all limit the reproducibility, generalisability, and ability to be externally verified for these studies (21,27,29–34).

This study aims to:

Develop a rule-based algorithm to identify BCDM cases by linking Northern Ireland Cancer Registry (NICR) population-based data with key NI health service datasets.

Test and validate the algorithm against a manually collected gold standard dataset of confirmed BCDM patients (35).

Develop estimates of BCDM incidence and prevalence in NI.

Report key demographic and clinical characteristics of BCDM patients.

Study Population

Data on patients with primary BC (ICD 10 C50) diagnosed between 1993 and 2020 were abstracted from the population based NICR (36), which has full coverage of NI’s population of 1.9 million patients (37). To ensure patients who were at risk of developing a DM from a non-BC primary cancer were not included in the study the following criteria applied:

Patients were excluded if they had another primary cancer diagnosed before their primary BC.
Patients were excluded if they had another primary cancer diagnosed between their primary BC diagnosis and subsequent DM record.
Patients were included if their other primary cancer occurred after the DM, as were cases when the other primary was a non-melanoma skin cancer (NMSC; ICD10 C44).
Patients with pathologically verified morphology groupings that are not epithelial breast cancer were excluded e.g., sarcoma, carcinoids, neuro-endocrine tumours, and lymphoma.

The dataset contained 28,473 primary BC patients of which 1,245 patients are categorised as de novo BCDM (See Fig. 1). NICR has ethical approval (REC reference number:20/NI/0132) for collection of data used in this study.

Defining Breast Cancer Distant Metastasis

BCDM patients were categorised as being either de novo BCDM or progressive BCDM; the de novo BCDM category included patients whose primary BC was Stage IV at diagnosis, whilst patients in the progressive BCDM category were either initially diagnosed with LR disease (Stage I-III) or “Stage not known” and then developed DM. (“Stage not known” includes cases with insufficient clinical or pathology information on stage).

Administrative Dataset Linkage

For the period 1993–2020, patients with progressive BCDM were identified by linkage of NICR records with two datasets; the Patient Administration System (PAS) which has full coverage of all NI public hospital inpatient records and General Records Office (GRO) death registration datasets (38). Patients were considered to have progressive BCDM if they were recorded on either of the two linked datasets as having a secondary code (ICD 9 196–198 excluding 196.3, ICD10 C77-C79 excluding C77.3) following their BC diagnosis (39). The ICD 9 and 10 codes for axillary metastasis are 196.3 and C77.3 respectively. Axillary metastasis is included in the N category of TNM staging from the Union for International Cancer Control (UICC) and relates ipsilaterally to the primary cancer with LR involvement, and thus in most cases does not alone represent DM(39,40).

In addition, searching of the free text cause of death from GRO records (sections 1a, 1b, 1c and 2) was undertaken to identify BCDM cases. Of the search terms used to identify cases, “Metastatic breast cancer”, “metastatic breast carcinoma” and “carcinomatosis” were most commonly recorded, accounting for 48.2% of GRO text entries. In addition to electronic searching, manual scanning of GRO records was also used to identify cases, which was especially useful for unusual or less commonly used terms (e.g. “mets”) and to pick up those with incorrect spellings (Supplementary Table 1). Prior to 2000, NICR did not receive free text information from GRO death records.

Development of BCDM Algorithm

Administrative linkage and formulation of rules to identify BCDM were completed using R Core Team (2022) software. Rules were iteratively developed by an expert panel of NICR staff ensuring alignment of international PBCR rules regarding incident diagnosis of cancer(41). (See Supplementary Table 2) (Fig. 1).

Validation of BCDM Algorithm

The algorithm was validated using an existing Macmillan funded NICR research dataset, named “BCDM (Validation)”, consisting of BC patients diagnosed in 2009 who were followed up until 31/12/2017 for recurrence and progression. This validation dataset included 1,028 patients whose records were manually searched by experienced NICR staff by reviewing electronic software systems including regional radiology, pathology and Multidisciplinary Team Meeting (MDM) databases and oncology systems (35). 184 patients were found to either have de novo BCDM at diagnosis or to have developed progressive BCDM between 2009 and 2017.

The NICR + GRO + PAS BCDM test dataset, named “BCDM (Test)”, was restricted from the original dataset covering 1993 to 2020 to patients diagnosed in 2009 and followed up until 31/12/2017 to mirror the timeframes of BCDM (Validation), with patients who developed BCDM after 2017 excluded. Sensitivity, specificity, positive predictive values (PPV) and negative predictive values (NPV) analysis were determined by comparing the BCDM (Test) dataset with the BCDM (Validation) dataset (39).

Estimating incidence and prevalence

Following validation testing, the BCDM (Test) dataset was imported into STATA version 16 and the number of new (incident) BCDM cases per year by data source were calculated. The BCDM period prevalent population per year was also estimated by adding the number of living cases at previous year end to the new (incident) cases for that year. For example, the period prevalent population for 2020 was the total number of BCDM cases living on 31/12/2019 (regardless of when diagnosed) added to the new (incident) cases for 2020.

Patient and demographic characteristics of BCDM cases were examined to include BCDM classification type (progressive or de novo), sex, age at diagnosis of BCDM, Trust of residence (at primary BC diagnosis) and Northern Ireland Multiple Deprivation Measure 2017 (NIMDM2017) quintiles (at primary BC diagnosis) (42).

Validation of algorithm

Validation testing of the algorithm’s BCDM estimates by comparing the BCDM (Test) dataset with the BCDM (Validation) dataset demonstrated high levels of accuracy, with high sensitivity (95.1%), specificity (99.2%), PPV (96.2%) and NPV (98.9%) found (Table 1) and overall accuracy of 98.4%. Of the 184 people identified as having BCDM, there were 7 false positive results, while of the 846 identified as not having BCDM there were 9 false negative results (Table 1).

Table 1

Performance results following validation of BCDM algorithm testing on BCDM (Test) dataset, against “known” 2009–2017 BCDM (Validation) dataset.
	TP	TN	FN	FP	Sensitivity, % (95% CI)	Specificity, % (95% CI)	PPV, % (95% CI)	NPV, % (95% CI)
Test Dataset n = 1,028	175	837	9	7	95.1% (93.8–96.4)	99.2% (98.6–99.7)	96.2% (95.0-97.3)	98.9% (98.3–99.6%)

Abbreviations: TP-true positives; TN-true negatives; FN- false negatives; FP-false positives; PPV-positive predictive value, NPV-negative predictive value, CI- Confidence Interval.

Estimation of incidence using BCDM algorithm

The number of new (incident) BCDM cases increased steadily between 1993 (n = 43) and 2009 (n = 253), remaining stable at an average of 254 new cases per year between 2009 and 2020 (Fig. 2). The stability of new BCDM case numbers from 2009 onwards is likely due to BC staging having reached high levels of completeness at this point, with “Stage not known” less than 10% after this time. Staging completeness improved from 45% in 1993 to 89% in 2008. There were notable increases in BCDM incidence in 1996 (n = 124), 2001 (n = 222), and 2006 (n = 230), which coincided with the diagnosis years which NICR conducted breast cancer audits and resulted in lower number of Stage ‘not-known’ and higher proportion of de novo (Stage IV) cases for these years (43). The absence of any increase in BCDM cases in 2012, also an audit year, is likely due to high completeness levels of staging were already established at this point (i.e. from 2009 onwards).

The lower incidence of BCDM between 1993 and 1999 is most likely due to the absence of primary BC records prior to 1993, alongside other factors such as the low proportion of breast cancers with stage assigned during this period and the lack of availability of free text information from death register certifications, which prevented free text mining to identify DM from GRO notifications.

Between 2000 and 2020 the algorithm’s free text-mining accounted for 77% of all progressive BCDM linked records (Fig. 1). From 2009 on, new BCDM case numbers have been stable, with an average incidence of 259 cases per year for the period 2016–2020. During this period, 31% of incident BCDM cases have been GRO only notifications, 16% PAS only notifications, 24% PAS + GRO notifications and 29% de novo stage IV notifications (Supplementary Fig. 1).

Prevalence of BCDM

Figure 3 shows the period prevalent population of BCDM per year, comprising BCDM cases newly diagnosed in that year and existing BCDM cases from previous years. BCDM incidence has remained relatively stable from 2009 to 2020, at 259 cases per annum. However, the number of existing cases has steadily increased over this period and consequently, the annual prevalent population has risen from 587 cases in 2009 to 911 cases in 2020 (55.2% increase). In 2009, existing cases accounted for 56.9% of the annual prevalent population, rising to 71.2% in 2020.

Table 2 shows that two-thirds (66.6%) of patients living with BCDM in 2020 had progressive BCDM and the remaining 33% were diagnosed with de novo BCDM.

Table 2

Patient characteristics of the Northern Ireland BCDM period prevalent population in 2020
		N (%)
Total BCDM population		911 (100%)
DM Type	De novo BCDM	304 (33.4%)
DM Type	Progressive BCDM	607 (66.6%)
Sex	Female	905 (99.3%)
Sex	Male	6 (0.7%)
Age at diagnosis of BCDM (Years)	< 40	52 (5.7%)
	40–59	387 (42.5%)
	60–79	354 (38.9%)
	80+	118 (13.0%
HSC Trust region of residence by postcode	Belfast HSCT	173 (19.0%)
	Northern HSCT	222 (24.4%)
	South-Eastern HSCT	181 (19.9%)
	Southern	183 (20.1%)
	Western	152 (16.7%)
Deprivation Quintile	Quintile 1 (Most Deprived)	169 (18.6%)
	Quintile 2	117 (19.4%)
	Quintile 3	170 (18.7%)
	Quintile 4	200 (22.0%)
	Quintile 5 (Least Deprived)	195 (21.4%)

More than half of patients with BCDM (52%) were over 60 years old when diagnosed.

The Northern Health and Social Care (HSC) Trust had the largest population of patients with BCDM during 2020 with an estimated 222 patients (24.4% of total).

A higher proportion of prevalent BCDM cases in 2020 resided in less deprived areas (43.4% from quintiles 4 and 5) compared to those in more deprived areas (38.0% from quintiles 1 and 2).

The lack of systematic collection and reporting of BCDM means the number of patients living with BCDM is largely unknown, with data and intelligence lacking to effectively develop policies, plan and deliver Health and Social Care (HSC) services, including services within the charitable sector, for these patients (44,45).

We developed an algorithm to detect BCDM using datasets from population-based cancer registry (NICR), inpatient hospital administrative system (PAS) and death registration (GRO).

The rules-based algorithm detected patients with BCDM across NI with high levels of sensitivity (95.1%), specificity (99.2%), PPV (96.2%) and NPV (98.9%) (Table 1) and compared favourably to other algorithms estimating BCDM incidence utilising PBCR data (21,22,24,27,28).

To date there have been few studies that have assessed the incidence and prevalence of BCDM in the UK. We estimate that in 2020 there were 911 patients living with BCDM in NI, of which one third were de novo BCDM and two thirds progressive BCDM (See Table 2).

A study by Yip et al (2015) used mortality data to determine the number of people living with BCDM in the UK per annum. This study estimated there was 35,000 BC survivors in the UK, comprising those with DM and not in their last year of life (65.7%) and those in their last year of life (34.3%).

Clements et al (2012) also developed a model to estimate BCDM prevalence based on mortality and incidence data, with their model predicting approximately 3 to 4 prevalent BCDM cases for every BC death (46). Applying this to the NI population, where the average annual BC deaths for 2016–2020 was 314, would equate to between 942–1256 prevalent cases annually. The annual prevalent population we estimated, of 911 in 2020, was slightly below the lower end of this range, but importantly we have excluded cases with another primary cancer either before the primary BC diagnosis or between the primary BC diagnosis and the progressive BCDM (6.3% of the total BC population).

In a recent study which used secondary and tertiary care records from the English NHS Hospital Episode Statistics (HES) database to directly identify BCDM cases, Palmieri et al reported a prevalent population in England of 57,215 patients, and an incident population of 7,580, for the financial year 2020/21. However, this study included C77.3 (axillary node) cases within their definition of BCDM, did not exclude cases with other primary cancers and did not use a validation dataset. One of the strengths of our study is the exclusion of C77.3 cases from our definition of BCDM. We estimate that had C77.3 cases been included, total incidence would be over-estimated by 75% with specificity decreasing to 77.1%, and PPV decreasing to 48.7% (Supplementary Tables 2 and 3) (47).

As outlined above, previous studies have developed estimates of BCDM using a range of methods, including rule-based algorithms and machine learning models, and have used a range of datasets, which largely reflect their availability and suitability in the study’s region or country. Datasets that have been used to develop estimates including for example, hospital-based and administrative datasets. Our study is unique in that it used population-based cancer registry (NICR) data and linked this to inpatient hospital administrative system (PAS) and death record (GRO) datasets. The NICR process of staging for primary BC patients involves manual review by NICR staff of each patient’s electronic records, including pathology and radiology, with high levels of staging completeness in recent years, which supports confidence in the accuracy of de novo BCDM patient data in this study (48). We used a novel method of free-text mining of GRO death records for DM descriptors to identify patients with BCDM. Free text mining of death records between 2000–2020 yielded 77% of BCDM cases identified from GRO death notifications, with only 23% derived from GRO ICD10 secondary codes (See Fig. 1), highlighting the importance the contribution this technique makes in BCDM identification and estimation. Another study strength was the large validation dataset used, with 1,028 primary BC patients manually followed up for a period of 8 years leading to the identification of 184 persons developing DM. Very few previous studies used similarly sized validation datasets (22), with the majority validating against substantively smaller datasets (24,27,28,34)

Limitations

Our algorithm found stable numbers of newly diagnosed BCDM cases per year from 2009–2020, but lower numbers in the years before 2009. However, prior to 2009 GRO and NICR datasets were not as complete, with this especially true in the years before 2000. Prior to 2009 there were lower levels of staging completeness in the NICR dataset, whist between 1993 and 1999, the GRO death database did not allow for free-text mining as NICR did not receive cause of death text information at this time (See Fig. 2). There may as a result be some patients who developed BCDM in the period 1993–2008 and were still alive in 2020 who were not identified which may underestimate the period prevalent population in 2020.

This study presents estimates of BCDM in NI which, in the absence of routine systematic collection of BCDM incidence in NI (45), depend on clinical coding of DM by HSC Trusts and which are not linked to the original primary BC diagnosis. However, as virtually all DM cases in NI are diagnosed in a secondary or tertiary care hospital within a universal healthcare system, this recording is likely to have high levels of completeness and accuracy in recent years. The timeliness of BCDM recording by HSC staff was also not known, and delays in recording BCDM by HSC staff may lead to an underestimate of more recent cases. However, as this effect will likely be distributed across the annual prevalent figures reported here, its impact is likely to be small. Further information on diagnostic and treatment pathway data for these patients may help improve the timeliness of BCDM diagnoses.

Another possible limitation is that patients with another primary cancer prior to their primary BC, or with another primary cancer prior to their BCDM record, were excluded. This represented 6.3% of all LR (Stage I-III) and “Stage not known” patients (See Fig. 1). Manual note review for this cohort may help ascertain if the metastatic event was associated the primary breast, rather than the other primary, cancer, but this level of detail was not available for this study.

The algorithm developed detected BCDM with high levels of sensitivity, specificity, PPV and NPV.

Annual incidence of BCDM increased steadily between 1993 and 2009, from 43 to 253, although case numbers in earlier years may have been underestimated due to primary BC cases not being available before 1993, low levels of staging completeness and poorer quality data. From 2009 onwards BCDM incidence has remained relatively stable at about 250 new cases per year.

The annual period prevalent population increased substantially between 2009 and 2020 from 587 to 911, an increase of 55.2%, suggesting improved BCDM survival profiles over this time. In 2020, the highest proportion of patients living with BCDM were existing cases from previous years (71.2%), while two-thirds of patients (66.6%) had progressive BCDM compared to one third being de novo (Stage IV at diagnosis).

Increasing survivorship of patients with BCDM has implications for health and social care services policy, planning and service delivery, including the charitable sector. The novel algorithm presented here will help policy development and the planning and delivery of care and support for BCDM patients and their carers. Further work is required to examine survival trends of patients with progressive DM and also to explore diagnostic, treatment and management pathways for this important group of patients.

Ethics approval and consent to participate

NICR has ethical approval for collection of the data used in this study - Northern Ireland Research Ethics Committee (NIREC) Reference Number:20/NI/0132)

Competing Interests

The authors have no conflicts of interest to declare that are relevant to the content of this article.

Funding

This study was supported by Cancer Focus Northern Ireland (Charity no.101307) but they were not involved in the study design, collection, analysis, interpretation of data, the writing of this article, or the decision to submit for publication.

Author Contribution

All authors contributed to the conception or design of the work. AA, DB, DD, DF, GS, HM, JK and SH contributed to the formulation and development of the study’s algorithms and validation plans. Merging of the PAS and GRO datasets to NICR BC dataset were completed by AA with algorithm rules applied under supervision from SH. SH completed the analysis and prepared the initial manuscript draft. All authors reviewed and edited the manuscript with AMcB reviewing and editing from the patient’s perspective. Supervision of the project was undertaken by DB.

Acknowledgements

This work would not have been possible without NICR staff (Cancer intelligence officers, IT, administrative and management staff) whose knowledge, skill and dedication allowed data to be assimilated and analysed. This work uses data provided by patients and collected by HSCNI as part of their care and support, and for which we are very grateful. The NICR is funded by the Public Health Agency of Northern Ireland. Special thanks to Ann McBrien and Julie Ann Lillis who advocated for incidence and prevalence statistics for patients with BCDM in NI to aid improvements in this area. Julie Ann has sadly passed away at the time of publication of this report. We would like to dedicate this publication to her.

Data availability

The datasets used in the study are not available publicly due to the need to maintain patient confidentiality. However, study data can be accessed via a secure facility within the NI Cancer Registry. The processes for accessing the data are available from the NICR contactable via email: [email protected].

Arnold M, Morgan E, Rumgay H, Mafra A, Singh D, Laversanne M, et al. Current and future burden of breast cancer: Global statistics for 2020 and 2040. The Breast [Internet]. 2022 [cited 2023 Aug 16];66:15–23. Available from: https://doi.org/10.1016/j.breast.2022.08.010
Renehan AG, Tyson M, Egger M, Heller RF, Zwahlen M. Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. www.thelancet.com [Internet]. [cited 2023 Aug 23];569. Available from: www.thelancet.com
Ahmad AS, Ormiston-Smith N, Sasieni PD. Trends in the lifetime risk of developing cancer in Great Britain: comparison of risk for those born from 1930 to 1960. Br J Cancer [Internet]. 2015 [cited 2023 Aug 23]; Available from: www.bjcancer.com
Beral V, Bull D, Doll R, Peto R, Reeves G. Breast cancer and breastfeeding: collaborative reanalysis of individual data from 47 epidemiological studies in 30 countries, including 50 302 women with breast cancer and 96 973 women without the disease. The Lancet. 2002 Jul 20;360(9328):187–95.
Aicr, WCRF. Diet, Nutrition, Physical Activity and Cancer: a Global Perspective A summary of the Third Expert Report. [cited 2023 Aug 23]; Available from: http://gco.iarc.fr/today
Guthmuller S, Carrieri V, Wübker A. Effects of organized screening program on breast cancer screening, incidence, and mortality in Europe. J Health Econ [Internet]. 2023 Aug 20 [cited 2023 Aug 23];102803. Available from: https://linkinghub.elsevier.com/retrieve/pii/S0167629623000802
Rumgay H, Shield K, Charvat H, Ferrari P, Sornpaisarn B, Obot I, et al. Global burden of cancer in 2020 attributable to alcohol consumption: a population-based study. Lancet Oncol [Internet]. 2021 Aug 1 [cited 2023 Aug 23];22(8):1071–80. Available from: https://pubmed.ncbi.nlm.nih.gov/34270924/
NICR. Female Breast Cancer 1993-2021, NI Cancer Registry 2023 Report [Internet]. Belfast; 2023 Oct [cited 2023 Oct 26]. Available from: https://www.qub.ac.uk/research-centres/nicr/FileStore/OfficialStats2021/Reports/female-breast-cancer-official-statistics-report.pdf
Breast cancer mortality statistics | Cancer Research UK [Internet]. [cited 2023 Oct 31]. Available from: https://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/breast-cancer/mortality
Mariotto AB, Etzioni R, Hurlbert M, Penberthy L, Mayer M. Estimation of the Number of Women Living with Metastatic Breast Cancer in the United States. [cited 2023 Aug 15]; Available from: www.seer.cancer.gov/seerstat
Holleczek B, Stegmaier C, Radosa JC, Solomayer EF, Brenner H. Risk of loco-regional recurrence and distant metastases of patients with invasive breast cancer up to ten years after diagnosis - Results from a registry-based study from Germany. BMC Cancer [Internet]. 2019 May 30 [cited 2023 Aug 16];19(1):1–14. Available from: https://bmccancer.biomedcentral.com/articles/10.1186/s12885-019-5710-5
Lord S (Sally) J, Daniels B, Kiely BE, O’Connell DL, Beith J, Pearson S, et al. Long term risk of distant metastasis in women with non-metastatic breast cancer and survival after metastasis detection: a population-based linked health records study. Medical Journal of Australia. 2022 Oct 17;217(8):402–9.
Geurts YM, Witteveen • A, Bretveld • R, Poortmans • P M, Sonke • G S, Strobbe • L J A, et al. Patterns and predictors of first and subsequent recurrence in women with early breast cancer. Breast Cancer Res Treat. 165.
Gogate A, Wheeler SB, Reeder-Hayes KE, Ekwueme DU, Fairley TL, Drier S, et al. Projecting the Prevalence and Costs of Metastatic Breast Cancer From 2015 through 2030. [cited 2023 Aug 15]; Available from: https://academic.oup.com/jncics/article/5/4/pkab063/6320071
Sun L, Cromwell D, Dodwell D, Horgan K, Ruth Gannon M, Medina J, et al. Economic Evaluation Costs of Early Invasive Breast Cancer in England Using National Patient-Level Data. [cited 2023 Oct 25]; Available from: www.elsevier.com/locate/jval
Anderson H, Ong J, Sloan S, Mallon P, McIntosh S, Irwin G. TH1.1 Self Directed Aftercare is a Safe, Efficient and well tolerated way to follow up Breast Cancer Patients. British Journal of Surgery [Internet]. 2022 Aug 9 [cited 2023 Aug 16];109(Supplement_5). Available from: https://dx.doi.org/10.1093/bjs/znac248.181
Warren JL, Yabroff KR. Challenges and Opportunities in Measuring Cancer Recurrence in the United States. JNCI J Natl Cancer Inst [Internet]. 2015 [cited 2023 Aug 22];107(8):134. Available from: https://academic.oup.com/jnci/article/107/8/djv134/944858
Cancer Strategy Steering Group. A Cancer Strategy for Northern Ireland 2022-2032 [Internet]. Belfast; 2022 Mar [cited 2023 Oct 25]. Available from: https://www.health-ni.gov.uk/sites/default/files/publications/health/doh-cancer-strategy-march-2022.pdf
Smith A, Durston V, Mills S. Long term risk of distant metastasis in women with non-metastatic breast cancer and survival after metastasis detection: a population-based linked health records study. Vol. 218, Medical Journal of Australia. 2023.
Valachis A, Carlqvist P, Ma Y, Szilcz M, Freilich J, Vertuani S, et al. ARTICLE Overall survival of patients with metastatic breast cancer in Sweden: a nationwide study. [cited 2023 Sep 18]; Available from: https://doi.org/10.1038/s41416-022-01845-z
Izci H, Tambuyzer T, Tuand K, Depoorter V, Laenen A, Wildiers H, et al. A Systematic Review of Estimating Breast Cancer Recurrence at the Population Level With Administrative Data. [cited 2023 Aug 24]; Available from: https://academic.oup.com/jnci/article/112/10/979/5817318
Izci H, Macq G, Tambuyzer T, De Schutter H, Wildiers H, Duhoux FP, et al. Machine Learning Algorithm to Estimate Distant Breast Cancer Recurrence at the Population Level with Administrative Data. Clin Epidemiol [Internet]. 2023 [cited 2023 Aug 24];15:559–68. Available from: https://pubmed.ncbi.nlm.nih.gov/37180565/
Habbous S, Barisic A, Homenauth E, Kandasamy S, Forster K, Eisen A, et al. Estimating the incidence of breast cancer recurrence using administrative data. 2023 [cited 2023 Aug 22];198:509–22. Available from: https://doi.org/10.1007/s10549-022-06812-z
Cronin-Fenton D, Kjaersgaard A, Mette Nørgaard •, Amelio J, Liede • Alexander, Rohini •, et al. Breast cancer recurrence, bone metastases, and visceral metastases in women with stage II and III breast cancer in Denmark. [cited 2023 Aug 22]; Available from: https://doi.org/10.1007/s10549-017-4510-3
Colzani E, Johansson AL V, Liljegren A, Foukakis T, Clements M, Adolfsson J, et al. Time-dependent risk of developing distant metastasis in breast cancer patients according to treatment, age and tumour characteristics. 2014 [cited 2023 Aug 22]; Available from: www.bjcancer.com
Lord S (Sally) J, Daniels B, Kiely BE, O’Connell DL, Beith J, Pearson S, et al. Long term risk of distant metastasis in women with non-metastatic breast cancer and survival after metastasis detection: a population-based linked health records study. Medical Journal of Australia. 2022;217(8).
Ling AY, Kurian AW, Caswell-Jin JL, Sledge GW, Shah NH, Tamang SR, et al. Using natural language processing to construct a metastatic breast cancer cohort from linked cancer registry and electronic medical records data. [cited 2023 Aug 28]; Available from: https://github.com/stamang/CLEVER
Jung H, Lu M, Quan ML, Cheung WY, Kong S, Lupichuk S, et al. New method for determining breast cancer recurrence-free survival using routinely collected real-world health data. BMC Cancer [Internet]. 2022 Dec 1 [cited 2023 Aug 24];22(1):1–10. Available from: https://bmccancer.biomedcentral.com/articles/10.1186/s12885-022-09333-6
Ritzwoller DP, Hassett MJ, Uno H, Cronin AM, Carroll NM, Hornbrook MC, et al. Development, Validation, and Dissemination of a Breast Cancer Recurrence Detection and Timing Informatics Algorithm. 2017 [cited 2023 Aug 24]; Available from: https://academic.oup.com/jnci/article/110/3/273/4349662
Sathiakumar N, Delzell E, Yun H, Jooste R, Godby K, Falkson C, et al. Accuracy of Medicare Claim-based Algorithm to Detect Breast, Prostate, or Lung Cancer Bone Metastases. Med Care [Internet]. 2017 [cited 2023 Aug 24];55(12):e144–9. Available from: https://pubmed.ncbi.nlm.nih.gov/29135778/
Chubak J, Onega T, Zhu W, Buist DSM, Hubbard RA. An electronic health record-based algorithm to ascertain the date of second breast cancer events. Med Care [Internet]. 2017 [cited 2023 Aug 24];55(12):e81–7. Available from: /pmc/articles/PMC4592686/
Whyte JL, Engel-Nitz NM, Teitelbaum A, Gomez Rey G, Kallich JD. An Evaluation of Algorithms for Identifying Metastatic Breast, Lung, or Colorectal Cancer in Administrative Claims Data. Med Care [Internet]. 2015 Jun 29 [cited 2023 Aug 24];53(7):e49–57. Available from: https://pubmed.ncbi.nlm.nih.gov/23524464/
Nordstrom BL, Whyte JL, Stolar M, Mercaldi C, Kallich JD. Identification of metastatic cancer in claims data. Pharmacoepidemiol Drug Saf. 2012 May;21(SUPPL.2):21–8.
Xu Y, Kong S, Cheung WY, Bouchard-Fortier A, Dort JC, Quan H, et al. Development and validation of case-finding algorithms for recurrence of breast cancer using routinely collected administrative data. [cited 2023 Aug 24]; Available from: https://doi.org/10.1186/s12885-019-5432-8
Cairnduff V, Dwyer L, Lardner S, Burns colm. 2019 NAACCR/IACR Abstract Program: Investigating the Characteristics of Women With Breast Cancer Recurrence In Northern Ireland (NI). In: Investigating the Characteristics of Women With Breast Cancer Recurrence In Northern Ireland (NI) [Internet]. Vancouver: The North American Association of Central Cancer Registries (NAACCR) and the International Association of Cancer Registries.; 2019 [cited 2023 Aug 18]. p. 167–167. Available from: http://www.iacr.com.fr/images/AnnualMeetings/Abstracts/2019Vancouver_Book-of-Abstracts.pdf
ICD-10 Version:2016 [Internet]. [cited 2023 Oct 5]. Available from: https://icd.who.int/browse10/2016/en#/II
Census | Northern Ireland Statistics and Research Agency [Internet]. [cited 2023 Sep 26]. Available from: https://www.nisra.gov.uk/statistics/census/2021-census
Search GRONI’s online records | nidirect [Internet]. [cited 2023 Aug 24]. Available from: https://www.nidirect.gov.uk/articles/search-gronis-online-records
International Classification of Diseases for Oncology, 3rd Edition (ICD-O-3) [Internet]. [cited 2023 Aug 18]. Available from: https://www.who.int/standards/classifications/other-classifications/international-classification-of-diseases-for-oncology
Ervik M, Napolitano G, Hawkins ST, Yeung K, Fox C, Jeyaraj A, et al. Canstaging+ [Internet]. 2021 [cited 2023 Aug 18]. Available from: https://www.canstaging.org/site/breast_c_v8
ENCR Recommendations | European Network of Cancer Registries [Internet]. [cited 2023 Sep 5]. Available from: https://encr.eu/ENCR-Recommendations
Ijpelaar J, Power T, Green B. Northern Ireland Multiple Deprivation Measures 2017. Journal of the Statistical and Social Inquiry Society of Ireland [Internet]. [cited 2023 Oct 5]; Available from: https://www.nisra.gov.uk/publications/northern-ireland-multiple-deprivation-measure-2017-steering-group
Cairnduff Victoria, Fitzpatrick D, Gavin A. Monitoring care of female breast cancer patients in Northern Ireland diagnosed 2012 with comparisons 1996, 2001 and 2006. Belfast; 2015.
Smith A, Durston V, Mills S. Long term risk of distant metastasis in women with non-metastatic breast cancer and survival after metastasis detection: a population-based linked health records study. Vol. 218, Medical Journal of Australia. 2023.
Cancer Strategy Steering Group. A Cancer Strategy for Northern Ireland 2022-2032 [Internet]. Belfast; 2022 Mar [cited 2023 Oct 25]. Available from: https://www.health-ni.gov.uk/sites/default/files/publications/health/doh-cancer-strategy-march-2022.pdf
Clements MS, Roder DM, Yu XQ, Egger S, O’Connell DL. Estimating prevalence of distant metastatic breast cancer: A means of filling a data gap. Cancer Causes and Control [Internet]. 2012 Oct 1 [cited 2023 Aug 15];23(10):1625–34. Available from: https://link.springer.com/article/10.1007/s10552-012-0040-9
Palmieri C, Owide J, Fryer K. Estimated Prevalence of Metastatic Breast Cancer in England, 2016-2021. JAMA Netw Open [Internet]. 2022 Dec 1 [cited 2023 Aug 15];5(12):e2248069–e2248069. Available from: https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2799862
UKIACR. United Kingdom and Ireland Association of Cancer Registries (UKIACR) Performance Indicators 2020 report on 2018 data [Internet]. 2020 [cited 2023 Oct 25]. Available from: https://www.ukiacr.org/sites/ukiacr/files/file-uploads/miscellaneous/UKIACR%20PIs%20%202020%20commentary%20on%202018%20data%202.0.pdf

No competing interests reported.

SupplementaryInformation.docx

Development and validation of an algorithm to estimate incidence and prevalence of breast cancer patients with distant metastasis in Northern Ireland: A population-based study

Status:

Version 1

Abstract

Figures

Introduction

Methods