National Benchmarking of Childhood Cancer Survival by Stage at Diagnosis in Italy (BENCHISTA-ITA): Study Protocol

Introduction

Survival for childhood cancers (CC) have significantly improved over recent decades, ranging from 78% to 81% in Europe. For many cancer types, 5-year survival now approaches 90%. However, progress has been limited for certain cancers, such as sarcomas.1 In Italy, the 5-year survival for childhood cancers reached 83% during the period 2010-2014.¹ Despite these improvements, documented variations in survival across European countries and Italian regions suggest that further progress is possible by addressing inequalities.^2,3
Several factors – such as socioeconomic status, geographic location, and national or regional clinical practices – can influence the timeliness of diagnosis and access to standard treatments for childhood cancer patients.^4,5 One of the most critical prognostic factors is the extent of tumour spread at diagnosis (stage). Childhood cancers differ from adult cancers in terms of biology and staging systems, which has historically made it difficult for population-based cancer registries (PBCRs) to collect standardized and complete staging data due to limited access to detailed clinical information.⁶
In 2014, an international group of clinicians and epidemiologists published the “Toronto Consensus Principles and Guidelines” (TG), which recommend specific staging systems for major childhood cancers.^7,8 The feasibility of applying these guidelines has been tested in Australia, in the Piedmont region (Italy), and in a European pilot study conducted through the Joint Action on Rare Cancers (JARC).^9-11
Risk stratification for children with cancer has not been widely reported in the literature. Accurate prognostic stratification is essential for interpreting and comparing survival across populations and time periods. Non-stage prognosticators (NSPs) for childhood cancers – such as molecular and genetic markers – are also important and are described in Gupta et al.,⁸ where their collection is strongly recommended for certain tumour types.
The overarching goal of this project is to better understand the reasons behind regional differences in childhood cancer survival in Italy and to identify areas for improvement. Since tumour stage is a key determinant of both prognosis and treatment intensity, the project aims to promote the adoption of the Toronto Guidelines by as many Italian cancer registries as possible for the most common solid paediatric tumours.
The Italian project “National Benchmarking of Population-Based Childhood Cancer Survival by Stage at Diagnosis” (BENCHISTA-ITA) is the national counterpart of the international BENCHISTA project (International Benchmarking of Childhood Cancer Survival by Tumour Stage).^12,13 Both projects share the following research questions:

• Are childhood cancers diagnosed at a more advanced stage in some regions or geographical groupings compared to others?
• Do survival rates by tumour stage vary among regions or geographical groupings?

To answer these questions, PBCRs will reconstruct tumour stage at diagnosis using the standardized Toronto Guidelines and collect follow-up data for at least three years to calculate overall survival (OS). Analysing stage at diagnosis within an international framework will help interpret survival differences across countries and regions.
An additional objective of BENCHISTA-ITA is to link its database with national clinical and hospital-based registries, including Mod.1.01 registry (the database managed by the Italian Association of Pediatric Hematology and Oncology – AIEOP)^14-16, the Italian Neuroblastoma Registry (RINB), and central nervous system (CNS) tumour registry. This linkage aims to enhance clinical data completeness and improve the quality of incidence and follow-up data. The project will also explore the phenomenon of treatment migration.¹⁷

Materials and methods

Selection of tumour type

The project aims to study stage distribution and survival for 9 solid paediatric cancers: medulloblastoma, astrocytoma, ependymoma, osteosarcoma, Ewing sarcoma, rhabdomyosarcoma, neuroblastoma, Wilms tumours, and retinoblastoma (Table S1, see online supplementary materials). These tumours were chosen based on one or more of the following considerations:
1. their generally good prognosis (retinoblastoma, nephroblastoma, localized neuroblastoma) and curability using standard treatment regimens;
2. the significant differences in outcomes already demonstrated between population-based survival studies;
3. low or no improvement in survival rates over a long period.
Together, these tumours represent more than half of all childhood solid tumours.¹ The updated European Network of Cancer Registries (ENCR) recommendation^18,19 should be followed to record the date of incidence used by the registry to define their cases meeting the inclusion criteria.

Inclusion criteria

Cases to be included are all children aged <15 years with the relevant histological codes (medulloblastoma, neuroblastoma, Wilms tumour, retinoblastoma, ependymoma) (Table S1). For astrocytoma, osteosarcoma, Ewing’s sarcoma, and rhabdomyosarcoma cases aged between 15 and 19 will also be included. For ependymoma, astrocytoma and retinoblastoma, non-malignant tumours will also be collected (Table S1). Cases must be diagnosed between 01.01.2013 and 31.12.2017, with one-year incidence back and forth considered for a registry to have a minimum of three consecutive years’ worth of cases, provided 3 years follow-up is ensured. Cases with incomplete staging information due to limited data should still be included and not excluded. All the CRs accredited by the Italian Association of Cancer Registries (AIRTUM) at the beginning of the project (2021) have been invited to contribute (16 regions). The expected number of cases estimated by geographical area for the three-year period of this study is shown in table 1.

The number of cases is primarily estimated using data from the EUROCARE-6 project (2005-2013). Some of them also participated, for a different period, in the JARC pilot study.¹¹
Cases ascertained only by death certificate (DCO), number of cases diagnosed by cytology or histology (microscopically verified), and those with unspecified morphology and topography codes (NOS) will be considered as data quality indicators for the completeness and accuracy of population-level registration of the nine diagnostic groups. The number of cases lost to follow-up and censored before the end date of follow-up will be used to assess the follow-up data quality.

Toronto Guidelines

Participant CRs will assign tumour stage at diagnosis using the TG supported by the detailed guidance based on the Australian experience produced as a study document.¹¹ The Australian guidelines document for the application of the TG was translated into Italian for clarity and made available through the AIRTUM website.²⁰ A tool to facilitate the reconstruction of stage called “CanStaging+” is available online.²¹ The TG includes a two-tiered staging system: Tier 2 is more detailed and intended for use in high-resource settings.^7,12 The full details of the Tier 1 and Tier 2 staging criteria for each tumour type are available elsewhere.²² All CRs are asked to provide Tier 2 stage if they can access the clinical details; otherwise, Tier 1 will be acceptable for the primary endpoints of assessing the proportions of localized vs metastatic tumour at diagnosis. Registries apply the TG for tumour staging to their existing data, using their usual data sources such as hospital discharge administrative files, pathological reports, and clinical records (and contribution of clinical registries with linkage). It was also requested to collect information regarding the clinical data sources for staging and the examinations performed (e.g., CT scan of chest versus X-ray of chest). Toronto stage is defined as the extent of disease at the time of diagnosis and is based on evidence acquired before treatment, with two exceptions: Wilms tumours staged according to the followed protocol (SIOP or COG)⁷ and tumours where investigations to exclude metastases were conducted shortly after surgery but before systemic therapy.

Quality and standardization of implementing Toronto Staging

To ensure the quality and standardization of implementing the Toronto staging guidelines, a comprehensive training programme has been proposed in line with the actions taken in the international project described by Lopez et al.²³ This programme includes training workshop sessions to introduce the Toronto Guidelines, followed by the exchange of fabricated cases to evaluate the consistency of TG application across different registries.
The training model was first implemented in a workshop held in Palermo in February 2020.²⁴ To further reinforce standardization, an additional online event was conducted in spring 2023. Training materials for 6 cancer types are available online; materials for rhabdomyosarcoma, astrocytoma, and retinoblastoma are available upon request from the corresponding author. These sessions involved the participation of cancer registry personnel and paediatric oncologists.

Variables to be collected

Variables to be collected by CRs for the nine tumours are compulsory or optional (table 2).

Compulsory variables

The major clinical variables are stage at diagnosis and examinations performed/used for staging. Compulsory demographic variables include year of birth, age at diagnosis, and basis of diagnosis and better described in Botta et al.¹² The structure of the record is presented in table 2. Follow-up data (life status and time in days from diagnosis to death or last follow-up) must be ensured up to at least three years from diagnosis;

Optional variables

This study also aims to evaluate the availability of additional relevant information from CRs regarding 9 tumour types and NSPs, primary treatment modalities, relapse/recurrence/progression, and cause of death. This data will be used for descriptive analyses, quality, and completeness across participating registries, aiding in data quality assurance and survival interpretation by stage. Specific NSPs (histology and cytogenetics) are requested for astrocytoma, medulloblastoma, rhabdomyosarcoma, neuroblastoma, and Wilms tumour, following the latest Toronto guidelines update.⁹ These include N-myc status for neuroblastoma, H3K27M mutations and Braf status for astrocytoma, WNT or SHH for medulloblastoma, PAX3-FKHR and PAX7-FKHR for rhabdomyosarcoma, and anaplasia for Wilms tumour. Additional features for astrocytoma and medulloblastoma include resection extent, anatomical location, and histological subtype, recorded using ICD-O-3. Individual treatment data (surgery, chemotherapy, radiotherapy) is collected. If first-line therapy is unclear, all treatments within the first 12 months post-diagnosis are included. Data on relapse/recurrence/progression within 3 years of follow-up is requested to assess feasibility. Cause of death categorized by tumour, toxicity, comorbidity, or other causes is also collected to explore survival rate differences due to treatment toxicity.

Statistical analysis

The formal statistical power to detect differences in stage distribution and survival rates between regions is necessarily limited by the number of incident cases for each of the 9 tumour types per region over the recent period for which cancer registries can provide tumour stage. Therefore, analyses of stage distribution and survival rates for each tumour type per region/group of regions will be descriptive, with 95% confidence intervals reported.
Approximately 1,400 cases are estimated to be collected (table 1). As the study is population-based, these are the largest numbers available and are not biased in ways that might affect institutional or clinical trial series.
Geographical areas were defined to maximize the sample sizes so Italian regions will be grouped according to the three conventional Italian geographical areas: North, Centre, and South-Islands. From the expected case numbers by registry (table 1), we have approximately 60% power to detect a 15% difference in lower stages (localized, loco-regional) versus more advanced stage (metastatic) between two regional groupings where there are 80-100 cases of each tumour type in each group. For group sizes with 100-130 cases, the power would be 70%. 
Overall survival for each tumour type will be analysed for all patients and broken down by appropriate tumour stage, using the standard Kaplan-Meier method reported with 95% confidence intervals. Survival differences among regional groupings and the corresponding contribution by variations in stage distribution will be studied by Cox model, including stage and other relevant prognostic variables (non-stage prognosticators) and confounders (age and sex) as covariates. To describe the proportion of cases treated and diagnosed in the same region and the movements among hospitals, the project will analyse patient pathways.

Linkage

This project will also test the feasibility of creating a linkage system developed with open-source R-based software to link population-based cancer registries and clinical databases for epidemiological surveillance, clinical network, and outcomes research.
To enable linkage, the completeness of identifiable variables will be evaluated in both data sources. Variables likely to be used for linkage include: sex, year of birth, residence, year of incidence, age in months, hospital of diagnosis, and vital status. These variables will help broadly identify patients across datasets.
Each record in the BENCIHSTA DB will be compared against all records in the clinical database through a two-step process: deterministic and probabilistic linkage.
Deterministic linkage involves matching records only when all selected variables have identical values. Successfully matched records are considered stable and excluded from further processing.
Probabilistic linkage is then applied to the remaining unmatched records. Here, a similarity score is calculated between each record in the two databases. If multiple source records exceed a predefined similarity threshold, they are placed in a temporary candidate set. The most appropriate match is then selected manually, based on expert judgment, contextual knowledge, and the most plausible outcome. The similarity score is a weighted function of the character-level similarity across the selected variables, as described in Contiero et al.25 Variable weights will be empirically determined through a pilot phase using data from the Campania and Piedmont PBCRs and the RINB.

Ethical approval and GDPR compliance

Ethical approval for the project has been given by the Ethical Committee of the Fondazione IRCCS “Istituto Nazionale dei Tumori” (INT) during the e-session held on 25th May 2021. The format of the data items to be collected has been agreed with the registries to be in the maximally de-identified format that minimises any risks to data privacy, in compliance with GDPR. INT will act as Data Controller for the project, with responsibilities as defined by article 26, GDPR and with the legal basis for data processing under article 6, GDPR 2018 being “Public interest” (clause (e)) and article 9 (special category) clause (j) – “research”.  

Discussion

This study was encouraged by the strong participation of registries in the pilot study, funded by the JARC.¹¹ The seven Italian PBCRs contributing data on neuroblastoma and Wilms tumours demonstrated high completeness of staging according to the Toronto Guidelines. Currently, approximately 80% of the Italian population of children and adolescents is covered by PBCRs.²⁶ Another important aspect is the contribution of Italy to an international project funded by Children with Cancer UK,^12,13 where Italian registries participated alongside 27 European and 4 non-European countries. A further strength was the close involvement of paediatric oncologists in the development of the TG⁷ and in the implementation of the population-level study.
Compared to the international study, this project includes: the linkage with the clinical registry to improve that completeness of clinical variables; the collection of the information regarding the hospitals to track patients’ pathways; three additional solid tumour types, including two CNS tumours (astrocytoma and ependymoma). For these tumours, stage at diagnosis is less relevant than morphological diagnosis, biological features, tumour site, and grading. We also observed that the incomplete collection of benign and borderline lesions hinders accurate geographical comparisons and complicates pathological diagnosis, which is often confirmed only a short time after initial diagnosis.²⁷ Retinoblastoma was included due to the importance of early diagnosis and centralised treatment in enabling salvage therapy.
To improve standardization across registries, three types of training courses were conducted, with high participation from registry staff. Course content was available in both Italian and English, accessible online, and supported by a help desk and a quality assurance document.
Childhood cancers are rare diseases, and thus it should be feasible to collect complete clinical information, which should be routinely included in cancer registration. A questionnaire on data sources (clinical and administrative) used to collect stage at diagnosis was distributed to registries. Most PBCRs participating in BENCHISTA also obtained additional data from external sources and had access to imaging results for staging, although they rarely had the opportunity to consult clinicians in ambiguous cases. In this study, registrars were also asked to document the diagnostic examinations used to determine the stage at diagnosis. This information is essential to assess the intensity of staging procedures and the precision of stage assignment. Understanding how exhaustive the diagnostic work-up was in excluding distant metastases will help interpret stage migration and improve geographical and temporal comparisons.28 Advances in diagnostic procedures and awareness of them may influence the proportion of advanced-stage tumours.
Most participating PBCRs are general registries without age limits, and thus include children and adolescents who may not be treated in paediatric oncology centres or be part of the AIEOP network. When linked with PBCRs, the AIEOP registry (Mod. 1.01)14,15 revealed that a proportion of children and adolescents with cancer – especially adolescents – were not known to paediatric oncology centres. For this reason, we requested the collection of information on hospitals of diagnosis and primary treatment. We also strengthened collaboration with AIEOP by linking three hospital-based registries: neuroblastoma, CNS tumours, and Mod. 1.01. The latter was established in 1970 to evaluate the paediatric hospital network and facilitate patient enrolment in clinical studies. This linkage will improve the completeness of clinical variables and enable more efficient case collection. The linkage protocol and results will be described in a dedicated paper.
One of the added values of the project is that both the AIRTUM and AIEOP are partners in the project.

Limitation of the study

The study faces some limitations due to the rarity of these cancers. Having a small number of cases in some categories, after stratification by type of childhood cancer and stage, leads to low statistical power, which reduces the ability to detect a real effect if one exists. It also results in wider confidence intervals, indicating greater uncertainty in the estimates. This is partly due to the study period spanning limited three/four years of incidence. Another limitation of this study is the high number of registries involved in the study, sometimes more than one per region. To address the lack of standardisation in stage data collection, we invested in targeted training initiatives. Additionally, centralising data quality and analyses is expected to enhance consistency across registries. However, a further limitation may arise from data access restrictions, which could limit the availability of clinical details from certain registries. These challenges are largely due to varying interpretations of GDPR and privacy regulations.
In conclusion, this study aims to enhance the interpretation of geographical differences in childhood cancer survival within Italy and across Europe – leveraging participation in the BENCHISTA International project – for the most common paediatric solid tumours, and to disentangle the respective contributions of stage at diagnosis and treatment. We recognize that many factors influence delayed diagnosis (e.g., cultural factors, paediatric healthcare planning, distance to specialized centres) and access to standard treatment (e.g., timely and accurate diagnosis, availability of treatment protocols, sociocultural barriers). We encourage PBCRs to routinely collect these clinical variables to strengthen connections with the clinical community and inform policy decisions.
Communicating the results of this study in collaboration with patient and family associations may help drive the changes needed to improve childhood cancer outcomes.

Conflicts of interest: none declared.

Funding: Children with Cancer UK (Grant References: 20-329) and Associazione Italiana per la Ricerca sul Cancro (AIRC) Italy (Grant reference: IG 2020 - ID 24933).

References

1. Botta L, Gatta G, Capocaccia R, et al. Long-term survival and cure fraction estimates for childhood cancer in Europe (EUROCARE-6): results from a population-based study. Lancet Oncol 2022;23(12):1525-36. doi: 10.1016/S1470-2045(22)00637-4
2. AIRTUM Working Group; CCM; AIEOP Working Group. Italian cancer figures, report 2012: Cancer in children and adolescents. Epidemiol Prev 2013;37 (1) Suppl 1:1-225.
3. Gatta G, Botta L, Rossi S, et al. Childhood cancer survival in Europe 1999-2007: results of EUROCARE-5 – a population-based study. Lancet Oncol 2014;15(1):35-47. doi: 10.1016/S1470-2045(13)70548-5
4. Erdmann F, Feychting M, Mogensen H, Schmiegelow K, Zeeb H. Social Inequalities Along the Childhood Cancer Continuum: An Overview of Evidence and a Conceptual Framework to Identify Underlying Mechanisms and Pathways. Front Public Health 2019;7:84. doi: 10.3389/fpubh.2019.00084
5. Dang-Tan T, Franco EL. Diagnosis delays in childhood cancer: a review. Cancer 2007;110(4):703-13. doi: 10.1002/cncr.22849
6. Mullen CJR, Barr RD, Franco EL. Timeliness of diagnosis and treatment: the challenge of childhood cancers. Br J Cancer 2021;125(12):1612-20. doi: 10.1038/s41416-021-01533-4. Eratum in: Br J Cancer 2021;125(8):1178. doi: 10.1038/s41416-021-01548-x
7. Gupta S, Aitken JF, Bartels U, et al. Paediatric cancer stage in population-based cancer registries: the Toronto consensus principles and guidelines. Lancet Oncol 2016;17(4):e163-72. doi: 10.1016/S1470-2045(15)00539-2
8. Gupta S, Aitken J, Bartels U, et al. Development of paediatric non-stage prognosticator guidelines for population-based cancer registries and updates to the 2014 Toronto Paediatric Cancer Stage Guidelines. Lancet Oncol 2020;21(9):e444-51. doi: 10.1016/S1470-2045(20)30320-X
9. Aitken JF, Youlden DR, Moore AS, et al. Assessing the feasibility and validity of the Toronto Childhood Cancer Stage Guidelines: a population-based registry study. Lancet Child Adolesc Health 2018;2(3):173-79. doi: 10.1016/S2352-4642(18)30023-3
10. Sacerdote C, Mosso ML, Alessi D, et al. An application of the Toronto Childhood Cancer Stage Guidelines in three population-based cancer registries: The case of central nervous tumors. Pediatr Blood Cancer 2020;67(6):e28303. doi: 10.1002/pbc.28303
11. Gatta G, Botta L, Capocaccia R, Cañete A, Pritchard-Jones K; JARC Pilot Study Toronto Guidelines Working Group. Staging childhood cancers in Europe: Application of the Toronto stage principles for neuroblastoma and Wilms tumour. The JARC pilot study. Pediatr Blood Cancer 2021;68(9):e29020. doi: 10.1002/pbc.29020
12. Botta L, Gatta G, Didonè F, Lopez Cortes A, Pritchard-Jones K; BENCHISTA Project Working Group. International benchmarking of childhood cancer survival by stage at diagnosis: The BENCHISTA project protocol. PLoS One 2022;17(11):e0276997. doi: 10.1371/journal.pone.0276997
13. The BENCHISTA Project. Available from: https://www.ucl.ac.uk/population-health-sciences/child-health/research/developmental-biology-and-cancer/benchista-project (last accessed: 28.01.2025).
14. Pession A, Rondelli R, Haupt R, et al. Sistema di rilevazione dei casi di tumore maligno in età pediatrica in Italia su base ospedaliera. Italian J Pediatr 2000;26:333-41.
15. Pession A, Rondelli R. The Italian hospital-based registry of pediatric cancer run by AIEOP. Epidemiol Prev 2008;32(2):102-5.
16. Didonè F, Botta L, Tittarelli A, et al. Enhancing quality of information for Italian Childhood Cancer data: the BENCHISTA-AIRC Project. In: GRELL 2024. Abstract accepted as oral presentation presented according to the conference sessions. Page 33. Available from: https://www.grell-network.org/_files/ugd/f45023_20d01cb1411544e685bf8aeee44775c0.pdf
17. Pession A, Quarello P, Zecca M, et al. Survival rates and extra-regional migration patterns of children and adolescents with cancer in Italy: The 30-year experience of the Italian Association of Pediatric Hematology and Oncology (AIEOP) with the Italian hospital-based registry of pediatric cancer (Mod. 1.01). Int J Cancer 2024;155(10):1741-50. doi: 10.1002/ijc.35074
18. European Network of Cancer Registries (ENCR). Available from: https://encr.eu/sites/default/files/pdf/incideng.pdf (last accessed: 28.01.2025).
19. Recommendations for a Standard Dataset for the European Network of Cancer Registries. Available from: https://www.encr.eu/sites/default/files/pdf/recommendations.pdf (last accessed: 28.01.2025).
20. Stadiazione dei tumori maligni infantili per i registri tumori di popolazione. Italian translation of the Toronto Guodelines. Available from: https://www.registri-tumori.it/cms/sites/default/files/pubblicazioni/Toronto%20Childhood%20Cancer%20Stage%20%28Italian%29.pdf (last accessed: 27.08.2025)
21. CanStaging. Available from: https://www.canstaging.org/ (last accessed: 28.01.2025).
22. Aitken JF, Youlden DR, Moore AS, et al. Childhood Cancer Staging for Population Registries According to the Toronto Childhood Cancer Stage Guidelines. Brisbane (Australia): Cancer Council Queensland and Cancer Australia; 2017. Available from: https://cancerqld.blob.core.windows.net/content/docs/childhood-cancer-staging-for-population-registries.pdf (last accessed: 05.02.2024).
23. Lopez-Cortes A, Didonè F, Botta L, et al. Cancer data quality and harmonization in Europe: the experience of the BENCHISTA Project – international benchmarking of childhood cancer survival by stage. Front Oncol 2023;13:1232451. doi: 10.3389/fonc.2023.1232451. Erratum in: Front Oncol 2024;14:1397101. doi: 10.3389/fonc.2024.1397101
24. Corso: La stadiazione dei tumori infantili secondo le Linee Guida di Toronto. Available from: https://www.registri-tumori.it/cms/eventi/corso-la-stadiazione-dei-tumori-infantili-secondo-le-linee-guida-di-toronto (last accessed: 28.01.2025).
25. Contiero P, Tittarelli A, Tagliabue G, et al. The EpiLink record linkage software: presentation and results of linkage test on cancer registry files. Methods Inf Med 2005;44(1):66-71.
26. Mappa dei registry di popolazione. Available from: https://www.registri-tumori.it/cms/pagine/mappa-dei-registri-di-popolazione (last accessed: 28.01.2025).
27. Gatta G, Peris-Bonet R, Visser O, et al. Geographical variability in survival of European children with central nervous system tumours. Eur J Cancer 2017;82:137-48. doi: 10.1016/j.ejca.2017.05.028
28. Botta L, Didonè F, Lopez-Cortes A, et al. International benchmarking of stage at diagnosis for six childhood solid tumours (the BENCHISTA project): a population-based, retrospective cohort study. Lancet Child Adolesc Health 2025;9(2):89-99. doi: 10.1016/S2352-4642(24)00302-X. Erratum in: Lancet Child Adolesc Health 2025;9(3):e9. doi: 10.1016/S2352-4642(25)00037-9. Erratum in: Lancet Child Adolesc Health 2025;9(5):e13. doi: 10.1016/S2352-4642(25)00096-3.