Abstract
Human tissue biobanks are at the epicenter of clinical research, responsible for providing both clinical samples and annotated data. There is a need for large numbers of samples to provide statistical power to research studies, especially since treatment and diagnosis are becoming ever more personalized. A single biobank cannot provide sufficient numbers of samples to capture the full spectrum of any disease. Currently there is no infrastructure in the United Kingdom (UK) to integrate biobanks. Therefore the National Cancer Research Institute (NCRI) Confederation of Cancer Biobanks (CCB) Working Group 3 looked to establish a data standard to enable biobanks to communicate about the samples they hold and so facilitate the formation of an integrated national network of biobanks. The Working Group examined the existing data standards available to biobanks, such as the MIABIS standard, and compared these to the aims of the working group. The CCB-developed data standard has brought many improvements: (1) Where existing data standards have been developed, these have been incorporated, ensuring compatibility with other initiatives; (2) the standard was written with the expectation that it will be extended for specific disease areas, such as the Breast Cancer Campaign Tissue Bank (BCCTB) and the Strategic Tissue Repository Alliances Through Unified Methods (STRATUM) project; and (3) biobanks will be able to communicate about specific samples, as well as aggregated statistics.
The development of this data standard will allow all biobanks to integrate and share information about the samples they hold, facilitating the possibility of a national portal for researchers to find suitable samples for research. In addition, the data standard will allow other clinical services, such as disease registries, to communicate with biobanks in a standardized format allowing for greater cross-discipline data sharing.
Introduction
T
The National Cancer Research Institute (NCRI) Confederation of Cancer Biobanks (CCB) (http://www.ncri.org.uk/ccb/) is a consortium of organizations within the UK involved in the development, management, and use of biobank resources for cancer research. The CCB recently undertook a project to formulate a set of criteria against which a biobank can be assessed and accredited; these criteria cover the gamut of biobanking activities, which include consenting donors, risk management, and sample quality. From the initiation of the project, the aim was to devise data standards that would be applicable to all research biobanks, independent of any disease focus. The core long-term goal of the CCB is to create a catalogue and portal where all accredited biobanks are listed, and researchers can search for tissue samples across the accredited network safe in the knowledge that all samples meet a certain standard. Therefore, a data standard was created specifically detailing what information biobanks must be able to provide about the samples they hold; other criteria will be reported elsewhere.
In an attempt to improve the “discoverability” of sample collections, central directories of samples are proliferating. Ensuring that each biobank collects appropriate data has led to the development of data standards such as the MIABIS standard, 1 which determines the minimum amount of data that should be made available to a central directory in order to allow researchers to find the samples they require. The data standard presented here was drafted with some central core principles. The first was to ensure that this standard was compatible with other standards; where it overlapped with other standards, such as MIABIS, 1 the data terms present in those standards would be used rather than reinvented. The standard would not dictate the data terms to be used; instead it would ensure that the biobank supplies the current data definitions in use. While the data standard was devised by the CCB, it is intended to cover all biobanks and therefore has no disease focus; thus, it must facilitate extensions to be developed that ensure key disease-specific information can be added. Finally, the data standard must allow the description of every patient and individual sample aliquot to ensure that core information relating to the provenance of the sample can be found by the researcher, while maintaining patient confidentiality.
Materials and Methods
A working group (WG 3) consisting of CCB members from several biobanks was assembled and complemented with lay representation, researchers, and noncancer biobankers for the purpose of determining the requirements for a data standard. The existing data standards1,7,9 were first examined to ensure that should a standard that meets the requirements of the CCB already exist, it would be recommended rather than implementing a new standard. As no single data standard was found that met all the needs of the CCB, a new data standard was developed that would, where appropriate, adopt the existing standards wherever overlap was present. The focus of WG 3 was to develop a data standard that can be used in the development of a national catalogue of samples, irrespective of disease focus. In order to achieve this aim, the four key goals were to: 1) not repeat or conflict with similar work; 2) not focus on the definition of data terms; 3) offer a mechanism for extension; and 4) include patient and sample level data.
The draft data standard was also reviewed by other biobanks with active online catalogues, as well as by national projects examining the development of such catalogues to ensure the proposed standard had compatibility with these initiatives, and work was undertaken to provide a proof of concept catalogue system.
Results
As the membership of CCB is broad and open to any institution or researcher that collects cancer samples, the confederation includes a wide array of ‘biobank’ types. As such, it may not be possible for all custodians of samples to provide the same level of information as the more established biobanks; even within one biobank there may be some collections that have available varying levels of data. The data elements that must be supplied by all biobanks are provided in Table 1. The data required to meet the minimum standard is described in Table 2, and the data required to meet the best practice standard are described in Table 3. The main difference between the two standards is that the minimum standard includes brief information about the biobank and aggregated information about the sample, and the best practice standard includes information at the individual patient and individual sample levels.
Meeting the key goals
Not repeat or conflict with similar work
The minimum data standard is conceptually very similar to that documented in the MIABIS standard, 1 in particular in describing the biobank (Table 1a) and a collection of samples (Table 2a). Similarly, the best practice standard (Table 3) is more akin to the data standard used by the Breast Cancer Campaign Tissue Bank (BCCTB) (https://breastcancertissuebank.org/about-tissue-bank.php). Therefore, to achieve the first goal, the CCB data standard has, where appropriate, adopted the relevant terms from these other standards as indicated in the ‘Source’ column of Tables 1–3. The main difference from the MIABIS standard 1 is the inclusion of Table 2b (Sample Data). This addition provides information on different types of aggregated samples. For example, it is conceivable that one collection under the custody of a biobank may contain both cancerous and noncancerous materials or tissues, and fluid samples. In this scenario, there would be one entry in the Collection table (Table 2a) and two in the Sample Data table (Table 2b), one for the cancerous and one for the noncancerous samples.
Not focus on the definition of data terms
Focusing on the structure and format of the data terms within the standard rather than defining the specific terms to be used ensured this goal was achieved. The key focus of the standard was to seek agreement on the data that should be collected and how this should be structured. Therefore, the terms to be used by a biobank for the diagnoses as well as the terms to represent the organs from which the samples originate must be supplied by the biobank itself, represented in Tables 1b and 1c, respectively. This means that the biobanks included in the catalogue can continue to use their existing databases and ontologies. Some data may not be available from all biobanks; for example, it may not be possible for every biobank in the network to supply the time that the blood flow to the sample was cut-off prior to collection. It is inconceivable, within a network of biobanks, that all biobanks will always be able to provide the same level of information. Therefore, for every term within the data set, it is possible for the bank to mark that data as Unknown/Unavailable/Inappropriate.
Offer a mechanism for extension
A core aim of the CCB data standard was to ensure that it provides all core terminology used to allow any biobank within the network to communicate about the patients and samples under its custodianship, while providing sufficient areas where the standard could be extended. To facilitate this, the core data standard should not be changed by any disease specific area. However, there are fields named ‘Disease specific data’ (e.g., Table 3a) where the standard can be extended to allow either additional tables or fields to be placed at those levels.
The BCCTB was a leading partner in the development of the data standard and the core structure has been utilized within this breast cancer specific biobank. As well as demonstrating the ability of the standard to be extended, it also demonstrates the appropriate use of the patient and sample information. The CCB data standard was created with the belief that it could be used beyond cancer to describe samples. To test the feasibility of this belief, the CCB data standard was used within the STRATUM project (http://www.stratumbiobanking.org/data.html), which created a data standard for cataloguing respiratory disease samples. The core dataset of the CCB project was retained and extended where appropriate. While the STRATUM project is yet to be implemented, it demonstrates that conceptually the CCB data standard could be applied to a wider biobanking setting.
The core focus of this work was to facilitate the creation of a national catalogue of available samples. Although this has not yet been completed, the Edinburgh and Dundee Experimental Cancer Medicine Centres (ECMC) (http://www.ecmcnetwork.org.uk/network-centres/edinburgh/) have adopted the data standard when developing a system to allow researchers to discover what samples are available across these two independent biobanks, one based in Dundee and the other in Edinburgh. Therefore, even though both biobanks use different terminology for their samples and two independent database systems, the data standard was able to provide a mechanism for the databases of both biobanks to communicate about the samples available at each site.
Include patient and sample level data
The ECMC trial of the data standard showed that each ECMC site could upload an anonymized version of their data pertaining to the patient (Table 3a). Some data relating to the patient, such as the patient diagnosis (Table 3b), may change over time so this is separated from the main Patient table to allow multiple entries to be attributed to the Patient. In a similar fashion, there are some properties linked to the sample that are time sensitive, such as the age of the patient at the time the sample was taken, any diagnosis at the time the sample was taken, or the consent conditions for the sample. The Sample Group (Table 3c) is used to provide information relevant to samples that were all collected on the same day, which allows the time sensitive information (diagnosis, age, consent conditions) to be attributed to all the grouped samples rather than having to link them individually. The Tissue Sample (Table 3e) and Fluid Sample (Table 3f) tables represent the individual aliquots and their properties, including some key quality information. The two were separated as there are clearly additional fields for tissue samples, such as the type of organ and the location of that organ (contained within the Solid Specimen Table 3d) that are not relevant to a fluid sample such as blood. Conversely, there are properties that a blood sample will require, such as volume, that are not applicable to tissue samples. Again, where the field is found in another data standard, such as the Storage Temperature, the definition from that standard has been used. The reason for requesting this level of detail is to ensure that researchers can find combinations of samples that may not have been predefined within a collection of samples using the Collection Table. In addition, the researcher can search for samples based on key quality control parameters.
Discussion
The CCB data standard provides a mechanism for building on the work of the previous MIABIS standard, 1 enabling researchers to find suitable available samples based on the individual characteristics of patients. The data standard is designed to empower the researcher with all the information available to help them decide which samples are appropriate for their research. An alternative option would have been to ask every biobank in the UK to provide only the data that was known to be available for all biobanks. This approach of going to the lowest common denominator actively undermines the potential benefit of any national registry, as key quality metrics are not made available to the researcher, especially in a climate where many standards and journal guidance are asking for such information.7–9 In the same light, the data standard should not exclude collections by placing too high a burden on entry, yet should include collections that do not have the same level of detail. Although these collections may not be appropriate for use in all scenarios, they should still be visible to the researcher as they may still be of some use. Therefore, the CCB data standard provides a mechanism to ensure that all sample collections can be accounted for while introducing a level of quality metrics at the individual sample level.
The CCB data standard avoids detailing the exact terms that each site must use to comply with the standard. Instead, the data standard focuses on agreement on the data to be collected and how it should be structured. This approach does introduce a concern for implementation of the data standard as differences in the meaning of terms will have to be mapped within the central registry. However, the adoption of the data standard by both the ECMC and BCCTB demonstrates that this challenge is technically possible to overcome.
The CCB provides a network of biobanks that are seeking to implement the data standard as part of a harmonization project in which any biobank seeking to be accredited must meet certain pre-defined standards. As such, and in combination with this data standard, the CCB will be able to provide a national registry of samples in the UK, and so provide a one-stop portal for researchers to source the most suitable samples available for their research within a framework of accredited standards.
Footnotes
Acknowledgments
We wish to thank members of Working Group 3 from the NCRI CCB Harmonization Project (Joint Chairs: Ian Forgie and Philip Quinlan; WG Members: Anne Carter, Bill Greehalf, Elwyn Shing, Gita Mistry, Helen Bulbeck, James Flanagan, John Brinsley, Kwok Pang, Mairead MacKenzie, Martin Groves, and Stuart Griffiths); the STRATUM Working Group, Carol Dawson and Paul Mitchell from the ECMC Edinburgh Centre and Breast Cancer Campaign.
Author Disclosure Statement
No competing financial interests exist.