Driving innovation through collaboration: development of clinical annotation datasets for brain cancer biobanking

Abstract

Background

A key component of cancer research is the availability of clinical samples with appropriately annotated clinical data. Biobanks facilitate research by collecting/storing various types of clinical samples for research. Brain Cancer Biobanking Australia (BCBA) was established to facilitate the networking of brain cancer biobanking operations Australia-wide. Maximizing biospecimen utility in a networked biobanking environment requires the standardization of procedures and data across different sites. The aim of this research was to scope and develop a recommended clinical annotation dataset both for pediatric and adult brain cancer biobanks.

Methods

A multidisciplinary working group consisting of members from the BCBA Consortium was established to develop clinical dataset recommendations for brain cancer biobanks. A literature search was undertaken to identify any published clinical dataset recommendations for brain cancer biobanks. An audit of data items collected and stored by BCBA member biobanks was also conducted to survey current clinical data collection practices across the BCBA network.

Results

BCBA has developed a clinical annotation dataset recommendation for pediatric and adult brain cancer biobanks. The clinical dataset recommendation has 5 clinical data categories: demographic, clinical and radiological diagnosis and surgery, neuropathological diagnosis, patient treatment, and patient follow-up. The data fields have been categorized into 1 of 3 tiers; essential, preferred, and comprehensive. This enables biobanks and researchers to prioritize appropriately where resources are limited.

Conclusion

This dataset can be used to guide the integration of data from multiple existing biobanks for research studies and for planning prospective brain cancer biobanking activities.

Brain cancers are relatively rare but usually fatal. Our understanding of inter- and intratumoral heterogeneity, and thus our ability to personalize therapies, is dependent on access to cohorts of patient samples. Despite increasing understanding of the molecular characteristics of subsets of primary brain cancers,¹ treatment largely remains “one size fits none” for the majority of adult primary brain cancers and many pediatric high-grade gliomas.

A key component of successful basic or translational cancer research is the availability of good-quality biological samples with appropriately annotated clinical data. Biobanks facilitate research by collecting and storing clinical samples, including tumor tissue, blood, and other samples, and are most informative for medical research when associated clinical data from consenting donors are linked to these samples.^2,3 Biobanks are becoming a vital part of translational cancer research, particularly with the increasing capacity to identify patient subsets that may benefit from targeted treatments. To make this “precision medicine” a reality, researchers require access to a large number of samples to enable the appropriate statistical power to answer research questions. Individual centers cannot necessarily biobank adequate numbers of samples for these studies,⁴ and researchers often cannot prospectively collect adequate sample numbers within a realistic time frame. This is particularly true for rare diseases such as primary brain cancers, for which a single biobank may collect only a small number of samples annually. Hence, networked biobanks are vital to facilitate research.

Brain Cancer Biobanking Australia (BCBA)⁵ was established by a brain cancer patient advocate in 2015 to facilitate the networking of brain cancer biobanking operations Australia-wide with the aim of providing researchers working in adult as well as pediatric brain cancer research access to the number, quality, and type of tissue samples and associated data they need to accelerate brain cancer research. BCBA is not a biobank itself, nor is it the custodian of any biospecimens or data. Rather, it acts as a “virtual biobank” and a portal for a network of 17 biobanks undertaking pediatric and/or adult brain cancer biobanking across Australia. BCBA has created an interactive online registry of the brain cancer biospecimens available for research within the network and facilitates researcher access to the tissue samples and associated clinical data via an online biospecimen search engine and application system.

To maximize the utility of biospecimens in a networked biobanking environment, procedures and data need to be consistent across different sites. Synoptic, prospective clinical annotation markedly improves the utility of biospecimens, by helping researchers to identify particular samples, and then be able to relate biological findings to clinical outcomes. Biospecimens without an essential set of clinical data items have limited use; therefore, it is necessary to have a minimum set of data items associated with the samples collected. The goal of this research was to undertake a scoping exercise to develop a recommended clinical annotation dataset for pediatric and adult brain cancer biobanks. Here, we present a recommended clinical dataset to be used as correlative clinical data in conjunction with brain tumor biobanking. The dataset focus is on primary CNS tumors of any histology, with the potential to include metastatic brain tumors if required.

Methods

A multidisciplinary working group consisting of members from the BCBA Consortium was established to develop clinical dataset recommendations for brain cancer biobanks that can be adapted by all biobanks regardless of their size or resource levels. The working group consisted of representatives from adult and pediatric neuro-oncology, neurosurgery, neuropathology, translational research, and patient advocacy. Because members of the working group were located across Australia, the initial working group meeting was held via teleconference, with follow-up discussions and communications conducted via email and face-to-face discussions at the annual BCBA Consortium meetings.

A literature search using PubMed was conducted using key terms such as “brain cancer AND clinical data set for biobanks,” “brain cancer biobank,” “biobank AND datasets” in January 2016 to identify any published clinical dataset recommendations for brain cancer biobanks, and to ensure no duplication of effort nor conflict with any international or national recommendations. The working group was unable to identify any published brain cancer–specific clinical data recommendations for biobanks that could be adopted by BCBA for its biobank network.

An audit of data items collected and stored by BCBA member biobanks was also conducted to survey current clinical data collection practices across the BCBA network. The majority of BCBA biobanks responded to the request to provide their data items and dictionaries (Table 1).

Table 1.

List of Responding Brain Cancer Biobanking Australia Member Biobanks

Biobank Name	Brain Cancer Tissue Held
Australian Genomics and Clinical Outcomes of Glioma	Adult
Biospecimen Collections of Cancer Trial Co-operative Groups Clinical Trials, NHMRC Clinical Trials Centre	Adult
Brisbane Breast Bank	Adult (metastasis)
Children’s Cancer Centre Tissue Bank	Pediatric
Children’s Cancer Institute Tumour Bank	Pediatric
Department of Surgery, University of Melbourne, Royal Melbourne Hospital	Adult
Hunter Cancer Biobank	Adult
Kolling Institute Tumour Bank-NeuroEndocrine Biobank	Adult
Queensland Brain Tumour Bank	Adult
Queensland Children’s Tumour Bank	Pediatric
RPAH Neuropathology Tumour and Tissue Bank	Adult
South Australian Neurological Tumour Bank	Adult
Steve and Lynette Waugh Brain Tumour Bank	Adult
The Children’s Hospital at Westmead	Pediatric
Victorian Cancer Biobank	Adult

Abbreviations: NHMRC, National Health and Medical Research Council; RPAH, Royal Prince Alfred Hospital.

The Department of Surgery, University of Melbourne, Royal Melbourne Hospital, the Victorian Cancer Biobank, and the Hunter Cancer Biobank adopted the Australian Comprehensive Cancer Outcomes and Research Database (ACCORD) dataset for the CNS developed by the Royal Melbourne Hospital Neuro-Oncology Service through BioGrid.⁶ The ACCORD database is a web-based application for the collection and management of clinical information relating to cancer patients across a range of tumor streams, including the CNS. The working group also had access to the Queensland Oncology Online Central Nervous System Cancer form (Queensland Cancer Control Analysis Team, Queensland Government).

The data dictionaries were examined to identify the data items gathered and their classifications. The working group focused on defining and providing a structure and format for each data item. This was performed to limit free text where possible, and to ensure that data could be collected and stored in a standardized format.

The draft clinical dataset recommendations were compared with other biobank standards such as Biospecimen Reporting for Improved Quality (BRISQ)⁷and Minimum Information About Biobank Data Sharing (MIABIS 2.0)⁸ to ensure that the clinical data fields in these standards could be populated with the information gathered using the BCBA clinical dataset recommendations. BRISQ is a set of recommendations for reporting sample handling to ensure consistent and standardized information to strengthen communications and publications involving research studies using biospecimens.⁷ MIABIS 2.0 was developed by the Biobanking and BioMolecular Resources Research Infrastructure based on MIABIS.⁹ MIABIS 2.0 describes the content of biobanks, sample collections, and research studies at an aggregate level to enable data sharing within biobank networks.⁸

The draft clinical data recommendations were reviewed by biobank members, the BCBA Steering Committee, and Consortium Group to ensure that all necessary data items were included. Data items were categorized by consensus and discussion as essential, preferred or comprehensive to prioritize data points and hence resource use for biobanks, each of which has different funding models and data collection support. The draft dataset was circulated via email for review to the abovementioned groups. Feedback was further discussed and incorporated into the dataset, including level of prioritization of each data item, and changes were recirculated via email for further review by all groups in an iterative process until consensus was reached.

Results

Fifteen of the 17 member biobanks responded to the request for a list of the clinical data items collected by their biobank, with 11 of 15 providing detailed data dictionaries. The number and complexity of data items collected varied considerably and data item definitions and data item formats (eg, text, number, free text) were not always available within the lists provided by individual biobanks.

Most BCBA member biobanks are unable to prospectively collect comprehensive data items because of resource constraints, and instead rely on information collected retrospectively by clinical departments from databases or paper medical records. Some biobanks stored only minimal information such as donor and consent details, and patient diagnosis, but were able to access medical records or hospital databases to collect data for research projects if required, and if human research ethics committee approvals permitted. The most comprehensive datasets were collected and stored by biobanks that were addressing clinical or epidemiological questions rather than molecular or laboratory research. The ACCORD dataset was also comprehensive as it is used to collect patient cancer information for clinical management.⁶

Examining the data items provided by BCBA biobanks, we identified 5 clinical data categories for brain cancer that reflect the clinical care from diagnosis to treatment and follow-up (Table 2): demographic, clinical/radiological diagnosis and surgery, neuropathological diagnosis, patient treatment, and patient follow-up. Patient information consists of donor-identifiable information for biobank internal use only, and patient demographic, lifestyle, and family cancer history details. Clinical and radiological diagnosis and surgery information includes clinical information on when and how the patient was diagnosed, the type of surgery, and any prior history of brain cancer. Neuropathological diagnosis data items relate to the pathological diagnosis details, including genomic and immunohistochemistry test results. Molecular characteristics now form part of the definition of (particular subsets of) CNS tumors in the most recent WHO classification.¹⁰ Treatment encompasses information about radiotherapy and chemotherapy, second- and higher-order surgery, and participation in clinical trials. The follow-up data items focus on information about brain cancer progression and clinical outcome and survival. Our clinical dataset recommendation is a tiered system to ensure that data collection is feasible and not so onerous that it becomes prohibitive. Accordingly, the data elements in each group were classified into 1 of 3 tiers: essential, preferred, and comprehensive (Supplementary Table 1).

Table 2.

Data Item Categories

Essential	Preferred	Comprehensive
Patient information	Patient information	Patient information
Personal details	Personal details	Personal details
(Last name, first name, UR number/MRN, name of hospital, date of birth, age at diagnosis, sex)	(Middle initial, post code, consultant name)	(Ethnicity)
		Lifestyle factors
		(Alcohol consumption, drinks per day, smoking status, past smoker, pack-year smoking history)
Biobank and consent details		Family cancer history
(Tissue bank consent, date of consent, consent version, biobank unique ID)		(Family history of brain tumors, family member with brain tumors, family member brain tumor type)
Clinical or radiological diagnosis and surgery	Clinical or radiological diagnosis and surgery	Clinical or radiological diagnosis and surgery
Diagnosis	Diagnosis	Diagnosis
(Date of diagnosis of CNS cancer, primary site of cancer, diagnosis basis, date of radiology, type of radiology)	(Radiological diagnosis)	(Radiological data)
Surgery	Surgery	Surgery
(Date of surgery, type of surgery)	(Extent of surgery)	(Preoperative ECOG)
Clinical syndromes	Clinical history	Clinical history
(Any germline syndromes relevant to brain cancer, type of germline syndrome)	(Past history of CNS cancer, type of previous CNS cancer, date of previous cancer)	(Treatment for previous cancer)
Neuropathological diagnosis	Neuropathological diagnosis	Neuropathological diagnosis
Tumor details	Tumor details	Tumor details
(Pathology report date, site of tumor, laterality of tumor, histological tumor type, metastasis primary site, WHO tumor grade, other information that maybe relevant) Clinical diagnosis (if pathological diagnosis is not available) (Clinical diagnosis, basis of clinical diagnosis)	(Tumor subtype, markers [1p19q analysis, MGMT methylation, Ki-67 analysis, IDH1 analysis, p53 analysis])	(Tumor size, markers [BRAF-KIAA1549 analysis, H3F3A K27 analysis, H3F3A G34 analysis, FGFR1 analysis, ATRX, BRAF, IDH1 codon 132, IDH2 codon 172, EGFR]; marker test)
Treatment information	Treatment information	Treatment information
Treatment	Treatment	Radiation therapy
(Did the patient undergo any treatment?)	(Availability of dates of treatment, treatment preoperatively or postoperatively)	(Start date, end date, dose, fractions, fields, stereotactic radiosurgery, completed treatment, if no, reason?)
	Chemotherapy	Chemotherapy
	(Concurrent chemotherapy + RT)	(Start date, end date, drugs used, dose reduction, completed, if no reason, toxicity)
	Clinical trials	Clinical trials
	(Clinical trial enrollment, name of trial)	(Was the trial randomized? Record arm of trial if unblinded)
Follow-up information	Follow-up information	Follow-up information
Follow-up	Follow-up
(Date, status, date of death)	(Evidence of recurrence, progression, stable disease, cause of death, other relevant comments)

Abbreviations: ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; IDH, isocitrate dehydrogenase; MGMT, O[6]-methylguanine-DNA methyltransferase; MRN, medical record number; RT, radiation therapy; UR, unit record.

The essential data items consist of information that can be collected by most or all biobanks. It was recognized by the working group that most biobanks collect clinical data around the time the donor undergoes surgery. This is the time that the biobank is in contact with the donor and has access to clinical information. Therefore, it was agreed that essential data items will mostly consist of items that would be required for almost all research studies and would be easily accessed by the biobank. Very few biobanks have the resources or patient access to enable comprehensive surveys at the time of diagnosis. The essential data items include donor-identifiable information, such as name and date of birth and consent details, under the demographic information category. These data items are used by biobanks only to identify the donor and are not for distribution to researchers.

Under the category of essential, clinical and radiological diagnosis information, date of diagnosis, basis for diagnosis, date of surgery, type of surgery, primary site of the tumor, and hereditary tumor syndromes are included, as these data fields would be required by a researcher to ascertain the timing and basis of diagnosis. Under neuropathological diagnosis information, essential data fields included pathology report date, tumor site, laterality of tumor and histological tumor type, tumor grade, clinical diagnosis, and neuropathological diagnosis according to WHO. The working group recommended that the histological tumor type follow the 2016 WHO Classification of Tumors of the Central Nervous System, to ensure alignment with international guidelines.¹⁰ Under the treatment information category, whether the patient underwent treatment was the only essential data field included. Under the follow-up information category, the essential data fields were date of follow-up and status, as these would provide information about patient survival, which would be required by most research projects. We would recommend that a contemporaneous survival update be requested whenever a biobank specimen is used for translational research in which survival is an outcome measure, and that survival information be entered in real time as obtained, whenever possible.

The preferred data category aimed to include data fields that would add further value to research projects. These data fields include prior history of CNS cancers, radiological diagnostic and imaging data, the timing of radiotherapy or chemotherapy, clinical trial participation, disease progression dates, and cause of death. These data items would allow research projects to examine disease progression and response to treatment.

The comprehensive data category consists of almost all the clinical information collected during the course of clinical management that may be required by research studies, particularly precision medicine research. Data fields under the comprehensive category include lifestyle information such as alcohol consumption and smoking status, and family history of brain cancer. This category also includes VASARI (Visually AcceSAble Rembrandt Images) feature set information; ¹¹ DICOM (Digital Imaging and Communications in Medicine) images of MRI, CT, and PET scans; brain tumor biomarker testing results such as ATRX mutation, MGMT (O[6]-methylguanine-DNA methyltransferase) promoter methylation status, and EGFR (epidermal growth factor receptor) amplification; and detailed treatment protocol information such as type and duration of treatment. Where relevant, imaging should follow standardized protocols.¹²

Discussion

To the best of our knowledge, there are no published minimum datasets for the clinical annotation of biobanked brain cancer specimens. These recommendations focus on those clinical fields required when collecting biospecimens for unspecified future research, and provide an opportunity to standardize the clinical data collected by the biobanks, facilitating the future construction of a large, harmonized prospective cohort. The tiered categorization of fields enables researchers to prioritize appropriately where resources are limited. The BCBA minimum dataset project fields can now be used to guide the integration of data from multiple existing biobanks for research studies, and plan prospective brain cancer biobanking activities. These recommendations can also be used to develop a distributed network biobank database incorporating the biospecimen data items recommended by the Biobank Certification Program,¹³ BRISQ,⁷ and the MIABIS recommendations^4,8,9 to collate the data previously collected by the biobanks for research use.

One challenge for biobank datasets may be remaining up to date with current recommendations. In 2018, the International Collaboration on Cancer Reporting developed a dataset for the pathology reporting of Tumors of the Central Nervous System (CNS).¹⁴ To remain current with best practice in diagnosis, BCBA will review these biobank clinical dataset guidelines annually to ensure that our guidelines are in line with international neuropathology reporting guidelines for CNS tumors. We will also review and refine these guidelines and data items regularly to align with international standards as required. We recommend that any database includes provision for collection of additional fields to enable the database to remain current with new biomarkers or treatment fields. We also recognize that there may be data items required by certain research projects that are not included in these guidelines, and cohort collections for specified research may still need to develop specific clinical data specifications to meet individual project goals.

This activity coincides with a number of international initiatives to improve biobanking standards and biospecimen quality for research. The Canadian Office of Biobank Education and the Canadian Tissue Repository Network have established a biobank certification program to help improve the quality of biobanking practices and raise standards through the provision of education modules and document templates for polices and standard operating procedures.¹³ The biobank certification program helps biobanks improve their operations and meet world biobanking standards. Other developments, such as the Biorepository Accreditation Program established by the College of American Pathologists and the Biospecimen Proficiency Testing Program established for biobank accreditation by the Integrated Biobank of Luxembourg under the International Society for Biological and Environmental Repositories, aim to improve biospecimen quality.^13,15 Furthermore, there have been developments such as BRISQ for reporting sample handling by biobanks⁷ and the previously mentioned dataset recommendations MIABIS 2.0, with variations of this also developed to help biobank networks collate, promote, and supply fit-for-purpose samples to researchers.^4,8,9 These activities complement our reported dataset.

Although this manuscript focuses on the development of a clinical biobank dataset, we recognize that there are substantial practical implementation and funding issues that affect the ability of any team to collect clinical data. Sample handling and storage are very important aspects of biobanking, with recent developments internationally aiming to improve biobanking standards and biospecimen quality for research. This is beyond the scope of the current manuscript; however, we also plan to develop standard operating procedures for sample collection, processing, and storage, informed by the literature and evolving international standards. Integrating data collection within an electronic health record or automated extraction of such information is ideal to reduce additional clinician burden, obviate the need to fund data collection, and reduce errors and disruption to workflow. Furthermore, the process of developing this dataset did not incorporate a recommendation on frequency of data updates, with the exception of contemporaneous survival data, as this may be resource constrained in some environments. Another limitation of this dataset is that it is relevant to the current management of brain cancer at the time of development; clearly, as pathology and management evolve, any dataset must be able to incorporate key changes through a process of planned revision, including movement between categories. Further updates to this dataset will be available on the BCBA website.⁵

A future goal includes integrating a platform for registry clinical trials, to rapidly and cost effectively evaluate common clinical questions and extend the current evidence base for effective care. Furthermore, this database may also be appropriate as a platform for a clinical quality registry using health outcome data to understand the appropriateness and effectiveness of care. To achieve this, a separate structured process to identify key quality indicators would be required and some fields that are currently considered preferred or comprehensive in the context of translational research and biobanking may be critical fields when considering quality of care.

In conclusion, this process, informed by the literature, current guidelines, and multidisciplinary expert consensus, has developed a harmonized research-ready dataset that will enable the most relevant clinical data to be prioritized for collection by brain cancer clinical biobanks for linkage to collected specimens or use in independent clinical research.

Funding

BCBA was supported by grants from Cancer Council NSW, Roche Australia, Robert Connor Dawes Foundation, and The Isabella and Marcus Foundation through the Australian Communities Foundation for this work.

BCBA Working Group Members

Craig Gedye, Bryan Day, Manuel Graeber, Robyn Leonard, Kerrie McDonald, Anna K. Nowak (Working Group Chair), and Mythily Sachchithananthan (BCBA Senior Project Coordinator).

BCBA Consortium Members

A/Prof Michael Back, Prof Michael Besser, Prof Andrew Boyd, A/Prof Michael Buckland, Prof Jennifer Byrne, Dr Lawrence Cher, Dr Raymond Cook, Dr Bryan Day, Dr Jerry Day, Dr Andrew Davidson, Ms Lisa Devereux, Dr Mark Dexter, Dr Roy Donnelly, A/Prof Kate Drummond, Dr Lisa Eckstein, A/Prof Hui Gan, Ms Therese Garrick, A/Prof Craig Gedye, Dr Nick Gottardo, Prof Manuel Graeber, Prof Clive Harper, Dr Rosalind Jeffree, Prof Terrance Johns, A/Prof Mustafa Khasraw, Dr Ganessan Kichenadasse, Dr Eng-Siew Koh, Dr Wendy Lipworth, Ms Robyn Leonard, Dr Louise Ludlow, A/Prof Kerrie McDonald, Dr Andrew Moore, Dr Lenka Munoz, Dr Najmun Nahar, Prof Anna Nowak, Dr Sarah Olson, Dr Rebecca Ormsby, Dr Jonathon Parkinson, Ms Audrey Partanen, Ms Emma Raymond, Prof Roger Reddel, A/Prof Peter Robbins, Prof Mark Rosenthal, Dr Jodi Saunus, Dr Brindha Shivalingam, Prof John Simes, Dr Brett Stringer, Ms Heather Thorne, A/Prof Claire Vajdic, Dr Winny Varikatt, Prof David Walker, Dr Helen Wheeler, Prof Deborah White, Dr Sonia Yip, and A/Prof David Ziegler.

Conflict of interest statement. None declared.