A Proposed Schema for ClassifyingHuman Research Biobanks

Peter H. Watson1,2 and Rebecca O. Barnes1

Human research biobanks have rapidly expanded in the past 20 years, in terms of both their complexity andutility. To date there exists no agreement upon classification schema for these biobanks. This is an importantissue to address for several reasons: to ensure that the diversity of biobanks is appreciated, to assist researchersin understanding what type of biobank they need access to, and to help institutions/funding bodies appreciatethe varying level of support required for different types of biobanks. To capture the degree of complexity,specialization, and diversity that exists among human research biobanks, we propose here a new classificationschema achieved using a conceptual classification approach. This schema is based on 4 functional biobank‘‘elements’’ (donor/participant, design, biospecimens, and brand), which we feel are most important to themajor stakeholder groups (public/participants, members of the biobank community, health care professionals/researcher users, sponsors/funders, and oversight bodies), and multiple intrinsic features or ‘‘subelements’’ (eg,the element ‘‘biospecimens’’ could be further classified based on preservation method into fixed, frozen, fresh,live, and desiccated). We further propose that the subelements relating to design (scale, accrual, data format, anddata content) and brand (user, leadership, and sponsor) should be specifically recognized by individual biobanksand included in their communications to the broad stakeholder audience.

Introduction

Human research biobanks are collections of humanbiospecimens (tissues, blood, and body fluids and their

derivatives) and associated data that are utilized for researchpurposes. These biospecimens and data are in some casesdesignated for research at the outset of the collection process,but often are collected for diagnostic procedures and laterrepurposed for research when residual material exists afterdiagnostic procedures have been completed (eg, clinical pa-thology archives). Human research biobanks are critical tothe current drive for personalized medicine as they providethe necessary biospecimens to fuel the majority of researchplatforms. There are already numerous examples of medicaladvances that have relied upon human research biospeci-mens.1,2 With the emphasis now on development of tools tofacilitate personalized medicine, the important role biobanksplay in medical research will only increase. This predictedtrend is supported by findings that over the past 2 decadesthere has been a significant expansion in biospecimen usageand, therefore, biobanking as an activity.3

The expansion of biobanking over the past 2 decades islargely due to evolution of research technologies and infor-matics capabilities. For example, the introduction of widelyavailable antibodies coupled with invention of the tissue

microarray enabled high-throughput screening of humantissues for suspected biomarkers.4 Technological advancessuch as these have meant that human biospecimens and datacan be technically approached as readily as cell lines andanimal models. This in turn has led to demand for biospe-cimens from the entire research spectrum, so that, whereasbiospecimens were once the domain solely of translationalresearch, they are now also critical for discovery and clinicalresearch.

Another factor underlying the expansion in human re-search biobanking is increased diversity of biobanks. Bio-banks have historically thought to be strictly comprised offormal entities within hospitals and research institutions,whereas it is now recognized that there are a variety of dif-ferent biobank types spanning a broad spectrum, from verysmall collections aimed at supporting a specific researchproject (mono-user biobanks), through collections associatedwith several research groups or clinical trials (oligo-userbiobanks), and finally to larger collection programs thatformally identify themselves as biobanks or repositories(poly-user biobanks). A good analogy to this spectrum ofbiobank types is that of ‘‘tools in the research workshop’’(Fig. 1). Biobanks belong to several general classes, butwithin these classes they have specialized aspects to theirdesigns, relating to the range of research questions they

1Tumour Tissue Repository, Deeley Research Centre, BC Cancer Agency, Victoria, British Columbia, Canada.2Department of Pathology and Laboratory Medicine, BC Cancer Agency and UBC, Vancouver, British Columbia, Canada.

BIOPRESERVATION AND BIOBANKINGVolume 9, Number 4, 2011ª Mary Ann Liebert, Inc.DOI: 10.1089/bio.2011.0020

327

intend to support. The range of questions in research reflectthe necessary incremental steps that lead research from basicdiscoveries to mature strategies, tools, and therapies thatimpact clinical care. Cancer research provides a good ex-ample of the progression of research questions from the basicto translational domain: what genes are different betweentumors; what cells express a specific gene within tumors;what effect does expression of a gene have on response totherapy?

The Need for Biobank Classification

There currently exists no accepted classification system forhuman research biobanks, largely because of the way thatthe discipline of biobanking has evolved since its origin.Research biobanks originally began from the usage of re-sidual clinical samples. However, because this was a sec-ondary, unanticipated use of these materials, the researchcommunity eventually became aware of the limitations onwhat clinical laboratories could offer. This resulted in a ra-pid, unregulated proliferation of independent research bio-banks established to serve specific research interests. Thisnew research-based approach to biobanking has significantlyexpanded to embrace a range of specialized componentsincluding frameworks (privacy and security), equipment(processing, annotation, and storage), operating procedures(biospecimen accrual, processing, annotation, storage, re-lease, distribution, and tracking), clinical informatics (pa-thology, treatment, and outcome data), database structures(donor consent, inventory management tools, and querytools), policies (priorities and access processes), economicmodels (funding sources, user fees, and intellectual proper-ty), governance models (for strategy and operations), andpersonnel with specialized roles and training. This has meantthat biobanking, which was once a limited activity within

clinical pathology, has now evolved into a sophisticateddiscipline. However, this evolution has largely occurredoutside of clinical departments in an uncoordinated way.This has created major constraints in quality and capacity,compounded by the diversity of biobank design in the ab-sence of an accepted system for describing or classifyingbiobanks.

This absence of a biobank classification system has severalconsequences that are detrimental to the discipline. First, itmakes the diversity of biobanks hard to appreciate, and lackof appreciation is a major factor contributing to the chal-lenges faced by all biobanks in securing adequate resourcesand funding. Although it has been established that long-terminstitutional support is a requirement to ensure biobanksustainability,5,6 many institution decision-makers fail torecognize the importance, complexity, and operational needsof biobanks within their organizations. Similarly, many sci-entists believe that effective biobanking amounts to no morethan a freezer and a fraction of a technician’s time to fill thefreezer. Although this type of biobank does exist, there existsthe remainder of the biobank spectrum that encompassesincreasing levels of operational complexity and associatedcost. There will undoubtedly continue to be a need for themore basic types of biobanks [eg, formalin-fixed paraffin-embedded (FFPE), tissues collected without standardizedprotocols and held within pathology archives],3 but many ofthe newer research technologies, such as genomic analyses,require fresh biospecimens processed in a highly standard-ized way and that are annotated with extensive clinicaldata.7 If institutions, funding bodies, and the scientificcommunity fail to recognize and support the latter type ofbiobank, the biospecimens needed for research toward per-sonalized medicine will not exist.

The absence of a classification system may also soon im-pact research users by hampering the ability to interpret and

FIG. 1. The various types ofhuman research biobanks re-sult in a spectrum of biobanktypes: mono-user, oligo-user,and poly-user biobanks. AColor version of this figure isavailable in the online articleat www.liebertpub.com/bio

328 WATSON AND BARNES

reproduce research results. Until recently, it has been suffi-cient to report data on a biomarker with only limited infor-mation about the biospecimens used and their source.However, as the recent focus on ‘‘biospecimen research’’ hasalready proven, knowledge of the preanalytical variablesassociated with biospecimens is integral to downstream in-terpretation of research results.8,9 As the importance of pre-analytical variables has become appreciated, the biobankcommunity has also realized that these need to be docu-mented and this is reflected by several new initiatives. Forexample, the ‘‘Standard PREanalytical Code’’ (SPREC) rec-ommendations provide a framework to capture the keypreanalytical factors impacting biospecimens.10,11 It is only amatter of time until documentation of preanalytical variablesbecomes a requirement for research publications. The re-cently published ‘‘Biospecimen Reporting for ImprovedStudy Quality’’ (BRISQ) recommendations provide a set ofreporting parameters that should be included in biospecimenresearch publications, including preanalytical variables andother aspects of biospecimen-related research, such as clini-cal characteristics of participants and aspects of the accrualmechanism.11 Related to this, it will also soon be importantto define the biobank source to enable increased experimentreproducibility, as shown by a recent initiative to create a‘‘BioResource Impact Factor’’ (BRIF) to create a quantitativemeans to document the use and quality of research arisingfrom individual biobanks.12 These 3 important initiatives,SPREC, BRISQ, and BRIF, would all work toward increas-ing communication between biobankers and researchers toenable better interpretation of research results and increasedbiospecimen traceability and experimental reproducibility.The purpose of this article is to make the argument fordevelopment of a similar tool with a different focus, intendedto help communicate the different and diverse aspects ofbiobanks, particularly to the broader biobank stakeholderaudience (ie, groups other than the biobankers andresearchers).

Developing a Classification Schema

Classification is the process whereby ideas and entities, inthis case biobanks, are recognized, differentiated, and un-derstood. Determination of the overall entity and the cate-gories requires that the common purpose is defined at eachlevel. ‘‘Prototype,’’ ‘‘classical,’’ or ‘‘conceptual’’ approaches toconsider the components and elements that can be catego-rized within and between categories have been defined.13

The prototype approach involves recognition of generalprototypes to create the basic stems or concepts, and then thedefinition of grades of categories related to these. In theclassical approach, biobanks would be defined by their dif-ferences into discrete classes. In the conceptual approach,biobanks would be defined by looking first for inherentcommonalities to create groups.

Human research biobanks have existed in ‘‘prototype’’forms for more than 50 years. The initial prototype wasperhaps the pathology archive, restricted to FFPE materials.The prototypes of other forms of biobanks then evolved inresponse to the demands for increased numbers of biospe-cimens and non-FFPE biospecimen formats. These additionalbiobank types became recognized by the 1980s through theprototypical designations (eg, blood biobanks and fresh-frozen tissue biobanks) (Table 1). Second-generation versionsof these prototypes emerged in the early 2000s with im-proved oversight, access, and more diverse operational de-signs to address biospecimen needs in terms of format andquality, through standardized processes and annotation.8

Third-generation versions that incorporate knowledge frombiospecimen science and better defined preanalytical vari-ables are also now emerging.

The experience gathered and the examples provided bythese new generations of biobanks provide the elements toconsider a new approach to classification. Currently, bio-banks are most often described by a classical classificationschema based on the type of research they intend to support:(1) population study biobank, (2) basic research biobank, (3)translational study biobank, (4) clinical trial biobank, and (5)pathology archive biobank (Table 2). An example of theapplication of this type of classification method is theBiobanking and Biomolecular Resources Research Infra-structure (BBMRI) division of its catalog of European bio-banks into population-based biobanks and disease-basedbiobanks.14 As argued above, the increasing complexity andspecialization within the discipline of biobanking and theneed for recognition of the diversity of biobanks mean thatclassification of biobanks by other approaches would bemore useful for the many stakeholder groups other thanresearchers.

As a starting point, we can define the root class, ‘‘bio-bank,’’ stemming from the collection of human researchbiospecimens and associated data. All such biobanks havemany shared elements around the activities of accrual, pro-cessing, storage, and release and the rules, laws, and stan-dards for operations that involve external oversight bodies.

Table 1. A Prototypical Biobank Classification Schema

Classification Categories Characteristics

By prototype Pathology archive Collection consisting of clinical FFPE materials.These were the earliest forms of human biobanks.

Blood biobank Collection consisting of blood and its derivatives.These biobanks began following the advent of clinicalblood tests and clinical drug trials.

Fresh-frozen tissue biobank Collection consisting of fresh-frozen tissue. These biobanksbegan in association with the introduction of clinicalbiomarkers (eg, estrogen receptor testing) and increasedwith the advent of newer molecular biology-basedresearch methodologies (eg, polymerase chain reaction).

FFPE, formalin-fixed paraffin-embedded.

CLASSIFICATION OF HUMAN RESEARCH BIOBANKS 329

There are also important shared elements across many bio-banks around connecting with participants, standardizationof quality, oversight of biospecimens, and linkage to clinicaland outcomes data. These elements provide a rich set offeatures to serve as a basis for classification by a new classicalor conceptual approach. However, the assumption of clas-sical classification is that all major classes exist already andthat all classes are mutually exclusive. As the recent emer-gence of live cell biobanking suggests, biobanks are stillevolving in step with the science that they support, and sothe conceptual approach is more appropriate.

The Elements of a New ConceptualClassification Schema

We propose that biobanks are classified by a conceptualapproach based on 4 functional elements—donor/participant,design, biospecimens, and brand—and multiple subelementsas outlined in Fig. 2. Each element will be discussed.

Donor/participant

This relates to the participant category, status, group, anddisease. The focus (category) can be on healthy control pop-ulations or diseased test subjects/participants. The test sub-jects may be alive or deceased (status), may belong torecognizable groups such as pediatric or adult (group), andmay be healthy or at risk for (preclinical state) or exhibitdisease (eg, noncancer, benign, or cancer).

Design

This relates to the accrual plan, scale, and the data format anddata content. Accrual can be prospective or retrospective, andthe source can be direct from the participant or indirect via amedical procedure, surgery, or autopsy, and in prospectivebiobanks the biospecimens and data can be collected once orat multiple time points. The collection can also involve par-

ticipants attending or deceased at a single health institutionor multiple clinical centers. The scale of collection can varyfrom very small collections in single laboratories related tosingle studies, to moderate- and large-sized collections re-lated to single and multiple research studies or collections.Scale is of course a relative term, but a recent review ofbiospecimen cohort size used in cancer research publicationssuggests that the average size has risen from 50 to 200 overthe past 2 decades.3 Our own experience within the tumorbiobanking field is that many current single studies and basicresearch laboratories host collections of < 200 cases, whereasformal biobanks designed to support multiple studies withcohorts selected using several criteria from the biobank stocktypically host > 1000 cases. On this basis we would proposeto define small/moderate/large as < 200, 200–1000, and> 1000, circa 2011. The data format within the database cancomprise linked or unlinked elements to the participant andbiospecimen (linked/identifiable elements can be coded oruncoded vs. irreversibly unlinked/anonymized biospeci-mens and data). The data content also vary in extent of an-notation and can include elements that relate to participant,biospecimen, disease, treatment, and outcome. With un-linked designs, data collection typically occurs once and inassociation with biospecimen accrual, whereas with linkeddesigns collection can occur before, during, or long afterbiospecimen accrual.

Biospecimens

This relates to the type, target, and method of preservation ofbiospecimens. The type of biospecimen includes blood, tissue,fluid, or cells; the target might be healthy or diseased targettissues; and preservation method might be formalin or otherfixation, freezing at 4�C, - 20�C, or - 190�C, or the collectionmight be focused on fresh materials (usually to obtain cell orfluid fractions), live cell preservation (usually throughfreezing), or desiccation.

Table 2. A Classical Biobank Classification Schema Based on Type of Research a Biobank Intends to Support

Classification Categories Characteristics

By type of researchthey support

Population study Focus of collection is on donors representing a defined population(normal, at risk, or with disease). Biospecimen focus is oftenblood/DNA samples. Usually corresponds to ‘‘large’’ sizeand ‘‘oligo’’ or ‘‘poly’’ user categories.

Researchproject

Focus of collection is variable (donor and/or patients with diseasetype). Biospecimen focus is variable and often involves tissueselected to represent normal and/or diseased tissue. Accrualthrough retrospective or prospective collection. Usually correspondsto ‘‘small’’ size and ‘‘mono’’ user categories.

Translationalstudy

Focus of collection is on patients representing specific disease typesand categories of disease accrued through either prospectiveor retrospective collection. Biospecimen focus is typically FFPEand/or blood samples with treatment and outcomes data.Usually corresponds to ‘‘moderate’’ size and ‘‘oligo’’ user categories.

Clinical trials Focus of collection is on patients with disease representing typesand enrolled prospectively into clinical trials to test new therapeuticapproaches. Biospecimen focus is mostly FFPE and/or blood sampleslinked to the trial treatment and outcomes data. Usually correspondsto ‘‘large’’ size and ‘‘poly’’ user categories.

Pathologyarchive

Focus of collection is on cases treated by surgical approaches.Biospecimen focus is mostly FFPE samples. Usually correspondsto ‘‘large’’ size and ‘‘poly’’ user categories.

330 WATSON AND BARNES

Brand

This relates to the leadership, sponsors/custodians, and the userbase associated with the biobank. Biobanks are created by indi-viduals, often expert researchers and groups, and institutions.This occurs with support of dedicated sponsors, and as in othernonresearch fields, the biobank often becomes commonlyknown by these brand labels as these have clear value incommunication and recognition to key sponsors, stakeholders,and communities of users. The term ‘‘user’’ refers to the in-tended spectrum of users. This can be a single user (mono-user),several users such as in a research group (oligo-user), or thebiobank can be intended for multiple users (poly-user). The‘‘user base’’ categories define many of the design features.

Considerations in the Applicationof a Conceptual Classification Schema

The elements and subelements outlined earlier provide acommon terminology for building a classification schema forbiobanks. We have recently applied the above schema, ele-ments, and subelements as a basis for formulating a plan forcertification of Canadian tumor biobanks. Inherent to thiscertification plan we needed to functionally classify biobanks

to refine the different education and information needsaround standardization of biospecimen quality. However,this is a relatively narrow context (ie, to guide developmentof a detailed program plan for tumor biobanks and mostlyfrom the biobank stakeholder perspective) in which to testapplication of this schema. To be more widely applicable, anew biobank classification should be simple, flexible, andgeneral. Classification should also serve as a communicationtool with the various biobank stakeholders and thereforemust relate to the need to communicate the elements andsubelements that are most important and of most interest toall the major stakeholder groups. We propose an example ofone approach describing how to apply this schema to anindividual biobank to stimulate broader discussion.

Example of How to Apply a ConceptualClassification Schema

To apply this conceptual classification schema to our ownbiobank, the British Columbia Cancer Agency’s Tumour Tis-sue Repository (BCCA-TTR), we created a matrix (Fig. 3) todisplay the different biobank stakeholders and elements. Thestakeholders we considered were comprised of 5 groups: (1)public/participants, (2) members of the biobank community,

FIG. 2. Proposed conceptual classification schema for biobanks based on 4 elements—donor/participant, design, biospe-cimens, and brand—and multiple subelements. A Color version of this figure is available in the online article at www.liebertpub.com/bio

CLASSIFICATION OF HUMAN RESEARCH BIOBANKS 331

(3) health care professionals/researcher users, (4) sponsors/funders, and (5) oversight bodies. All elements and theirsubelements were then organized into rows and the subele-ments that we considered to represent the principal concernsfor each of these stakeholder groups were identified with an‘‘X.’’ This enabled a score to be calculated for each subelement,equal to the number of stakeholders for which each subele-ment was important (0–5). Using this approach, we identifiedthe most important common elements for classification of ourbiobank to be brand (user, leadership, and sponsor) and de-sign (scale, accrual, data format, and data content).

Conclusion

The adoption of a more refined classification system isimportant for the future of human research biobanks. Giventhe wide spectrum of biobank stakeholders, a single classi-fication schema may not serve all biobanks optimally, but itmay be valuable and possible to agree upon the elements of aprimary classification schema and secondary terms forcommon use. We propose that a conceptual approach based

on shared functional elements should be the basis for suchclassification. Unlike other recent suggestions to classifybiospecimens for research users (SPREC),10 standardizebiospecimen-related research reporting (BRISQ),11 and re-porting of biobank impact factors (BRIF),12 the approachpresented here would allow for detailed biobank classifica-tion with common terminology that is of most value foroperational communication between biobanks and theirnonbiobank stakeholders. However, to be adopted, a classi-fication method must be simple, broadly applicable, and useelements that are the least likely to change over the life of thebiobank. Therefore, we conclude that brand and its subele-ments should be the core basis for classification to commu-nicate with all stakeholders and that design, biospecimen,and donor elements should be selectively used as secondaryclassifiers. As an example, our own biobank name (BCCancer Agency’s Tumour Tissue Repository) incorporatesbrand and disease subelements, but for consistency with thisproposed schema and to more broadly communicate acrossour stakeholders, we might consider a future modification tothe title ‘‘BC Cancer Agency Poly-user Tumour Repository’’

FIG. 3. Matrix of biobank elements and subelements grouped into areas of interest for the proposed stakeholder groups. Asan example of application of this matrix, those elements and subelements applicable to the BC Cancer Agency’s TumourTissue Repository (BCCA-TTR) have been recorded (‘‘X’’). The column on the far right shows the score for each subelement,equal to the number of stakeholders for which each subelement was important (0–5). A Color version of this figure isavailable in the online article at www.liebertpub.com/bio

332 WATSON AND BARNES

and to adopt the secondary terms ‘‘prospective’’ and ‘‘large’’biobank when appropriate for the target audience.

Acknowledgments

This work was conducted as part of the BC CancerAgency’s Tumour Tissue Repository Program (supported bythe BC Cancer Foundation), the BC BioLibrary (supported bya grant from the Michael Smith Foundation for Health Re-search), and the Canadian Tumour Repository Network(supported by a grant from the Institute of Cancer Research,Canadian Institutes of Health Research).

Author Contributions

Both authors contributed to the ideas presented here andto the development and preparation of the manuscript. Bothauthors read and approved the final manuscript.

Disclosure Statement

There are no conflicts of interest to disclose.

References

1. Zwick E, Wallasch C, Ullrich A. HER2/neu: a target for breastcancer therapy. Breast Dis 2000;11:7–18.

2. Topol EJ, Murray SS, Frazer KA. The genomics gold rush. JAMA2007;298:218–221.

3. Hughes SE, Barnes RO, Watson PH. Biospecimen use in cancerresearch over two decades. Biopreserv Biobank 2010;8:89–97.

4. Kononen J, Bubendorf L, Kallioniemi A, et al. Tissue microarraysfor high-throughput molecular profiling of tumor specimens.Nat Med Jul 1998;4:844–847.

5. Hewitt RE. Biobanking: the foundation of personalized medi-cine. Curr Opin Oncol 2011;23:112–119.

6. Riegman PH, Morente MM, Betsou F, et al. Biobanking for betterhealthcare. Mol Oncol 2008;2:213–222.

7. Ginsburg GS, Burke TW, Febbo P. Centralized biorepositories forgenetic and genomic research. JAMA 2008;299:1359–1361.

8. Barnes RO, Parisien M, Murphy LC, Watson PH. Influence ofevolution in tumor biobanking on the interpretation of transla-tional research. Cancer Epidemiol Biomarkers Prev 2008;17:3344–3350.

9. Barnes MG, Grom AA, Griffin TA, et al. Gene expression profilesfrom peripheral blood mononuclear cells are sensitive to shortprocessing delays. Biopreserv Biobank 2010;8:153–162.

10. Betsou F, Lehmann S, Ashton G, et al. Standard preanalyticalcoding for biospecimens: defining the sample PREanalyticalcode. Cancer Epidemiol Biomarkers Prev 2010;19:1004–1011.

11. Moore HM, Kelly AB, Jewell SD, et al. Biospecimen reporting forimproved study quality (BRISQ). Cancer Cytopathol 2011;119:92–102.

12. Cambon-Thomsen LM. BRIF: bio-resource impact factor. AnInternational Working Group towards An Operational Index toPromote and Recognize Sharing of Biological Samples and As-sociated Data. Biopreserv Biobank 2011;9:98.

13. http://en.wikipedia.org/wiki/Categorization.14. Riegman PH, de Jong BW, Llombart-Bosch A. The Organization

of European Cancer Institute Pathobiology Working Group andits support of European biobanking infrastructures for transla-tional cancer research. Cancer Epidemiol Biomarkers Prev2010;19:923–926.

Address correspondence to:Peter H. Watson

Director, Tumour Tissue RepositoryDeeley Research Centre

BC Cancer Agency2410 Lee Avenue

Victoria, BCCanada V8R 6V5

E-mail: pwatson@bccancer.bc.ca

Received 14 April, 2011/Accepted 18 May, 2011

CLASSIFICATION OF HUMAN RESEARCH BIOBANKS 333