Creating a data commons: The INternational Soft Tissue SaRcoma ConsorTium (INSTRuCT)

1.1 Soft tissue sarcoma

Soft tissue sarcoma (STS) comprises 7% of pediatric malignancies, with 40% of those representing rhabdomyosarcoma (RMS) and the remainder being classified as non-RMS STS (NRSTS). While 5-year overall survival (OS) for embryonal RMS is 66%–68%, outcomes for patients with alveolar RMS are less favorable, with a 5-year OS of 39%–50%.^{1, 2}

Compared to RMS, the study of NRSTS presents unique challenges. While more than half of pediatric STS are classified as NRSTS, this diverse group includes over 75 biologically unique sarcomas.³ Due to the rarity of each subtype of NRSTS, it has historically been studied out of necessity in “basket” trials where the varied diagnoses are treated uniformly, without an ability to take into account the diversity and unique biological behavior of each of these subtypes.⁴ Recent discoveries about the molecular underpinnings of STS subtypes have uncovered the molecular diversity within NRSTS, highlighting the biological distinctness of each of these tumor types.

Research has traditionally been conducted by cooperative groups, and discoveries have been limited by the paucity of patients with rare disease subtypes. Due to the rarity of these subtypes, international collaborative relationships have developed to pool data on rare subtypes.^5-8 Following this rich tradition of international collaboration, there was a desire to create a more comprehensive and reusable dataset and formally organized consortium for the study of pediatric STS. Pediatric STS experts engaged the Pediatric Cancer Data Commons (PCDC) team to assist in forming a consortium—and ultimately a data commons⁹—and developed the INternational Soft Tissue SaRcoma ConsorTium (INSTRuCT; Figure 1).

2.1 Beginnings

Under the auspices of the PCDC, INSTRuCT was established in 2017 to foster international research and collaboration focused on pediatric STS. INSTRuCT's membership is derived from North American and European cooperative groups (Children's Oncology Group [COG], European paediatric Soft tissue sarcoma Study Group [EpSSG], and Cooperative Weichteilsarkom Studiengruppe [CWS]) and includes data from prior studies sponsored by the International Society of Paediatric Oncology (SIOP) Malignant Mesenchymal Tumour Committee and Italian Association of Pediatric Hematology and Oncology (AIEOP) Soft Tissue Sarcoma Committee.

The foundational meetings took place as add-ons to established international research conferences. Six in-person meetings took place between October 2017 and October 2019 before transitioning to virtual meetings due to the COVID-19 pandemic. Each meeting was structured to include a plenary as well as breakout sessions focused on disease type (e.g., RMS vs. NRSTS) or clinical discipline. Initial clinical discipline work groups included pathology/biology and surgery, and later expanded to include radiation oncology and imaging. The early meetings established the structure of INSTRuCT and set overarching priorities, whereas subsequent meetings led to more granular discussions that established discipline-specific research questions and led to the development and approval of a standardized RMS data dictionary.

2.2 Establishing priorities

One founding charge was collection and aggregation of high-quality international clinical trial datasets into a data commons, which would increase access to cohorts of rare disease subtypes. Developing an international consensus for risk stratification in RMS was identified as the initial use for the data commons. Such an undertaking had recently been conducted by the International Neuroblastoma Risk Group (INRG), which similarly convened an international consortium of pediatric cancer researchers and combined international datasets in an effort to revise and standardize risk stratification for children with neuroblastoma.¹⁰ A member of the INRG Task Force was present at the founding INSTRuCT meeting to share background and insights learned in the process. Another priority was to focus on existing data from completed studies, rather than attempting to add new data to existing datasets.

Key clinical questions relating to RMS risk stratification included re-examination of primary tumor site designation as “favorable” or “unfavorable.” During the initial meetings, clinical trials from which data could be contributed were identified, and data elements to be contributed were selected so that work on a standardized data dictionary could begin. Priorities that emerged during subsequent meetings included identification and assessment of genetic markers, developing methods to link clinical data to genomic data, and improving outcomes in patients with relapsed and metastatic disease. A final aim was to develop international consensus on management of RMS and NRSTS, both to inform the conduct of international clinical trials but also to inform clinicians of “best practices.”

2.3 Clinical consensus building

One early task was to identify and standardize data elements to be included in a data dictionary to facilitate harmonization of datasets. The harmonization process highlighted similarities and differences between clinical trials and cooperative groups, including approaches to diagnosis and treatment. Data elements that were collected on case report forms were inventoried and analyzed, including how they were defined and collected. It quickly became apparent that consensus among international groups could facilitate retrospective harmonization of datasets from past clinical trials and—more importantly—could facilitate advances in clinical care, the conduct of future collaborative trials, and combination of new datasets. This prompted the development of four “consensus opinion” statements from the collaborative groups as the first publications to emerge from the consortium.^11-14 These consensus opinion papers incorporated review of the literature and clinical trials treatment protocols along with expert opinion to provide specific consensus recommendations to guide clinical care and the conduct of future clinical trials.

3.1 Governance structure

One early task was to establish a governance structure and define processes to be followed by the consortium. In particular, the specific roles and responsibilities of individuals and groups needed to be defined, and the relationships between the consortium, data contributors, and service providers needed to be detailed. Finally, policies were needed to outline the process for ingesting, requesting, and accessing data, in addition to dissemination of data.

3.2 Memorandum of understanding

One foundational document is the memorandum of understanding (MOU). Parties to the MOU include cooperative groups and institutions with data to contribute. In the case of INSTRuCT, using the INRG as a model, the cooperative groups formed the consortium on a foundation of trust, built on a long history of collaboration. The cooperative groups signed an MOU and created an Executive Committee.

3.3 Executive Committee

Responsibilities of the Executive Committee described in the MOU included strategic planning; coordinating, appointing, and changing the data commons manager; amending the MOU as necessary; approving and managing membership; reviewing and approving requests to access data within the data commons; reviewing and approving contributions of data to the data commons; and approving of grant or funding applications submitted on behalf of, or which rely upon, the consortium.

The INSTRuCT Executive Committee is composed of a representative designated by each cooperative group to serve as co-chairs, two other representatives of each cooperative group (to be selected to ensure a balanced and complete representation of clinical disciplines and geography), two statisticians appointed by mutual agreement of the cooperative groups, and one representative of the PCDC, which serves as the data commons service provider. A consortium manager–with both program management and science expertise–plans, organizes, and facilitates meetings; handles project requests; and acts as a liaison between the consortium and the PCDC. The consortium manager also coordinates individual work groups and tracks action items to ensure goals of the consortium are met and coordination across the consortium.

The cooperative groups determined that the members of the Executive Committee will make decisions and handle disputes by mutual, unanimous agreement and consensus. These responsibilities and the decision-making framework bolstered trust among cooperative groups willing to bring their data into the PCDC, as they provided reassurance that the contributed data will be used responsibly and only for projects approved by the cooperative groups through their representation on the Executive Committee, which reviews project requests.

3.4 Work groups

The clinical discipline work groups are composed of members from each of the cooperative groups, with one member agreeing to serve as lead. To ensure communication with INSTRuCT leadership, Executive Committee members agreed to serve as a liaison to each work group. The Executive Committee drafted a charge to each work group. The Executive Committee opted to stagger the launch of the groups. NRSTS, pathology/biology, and surgery all convened for the first time in March 2018, followed by imaging in March 2020, and radiation oncology in October 2020.

In March 2021, the Executive Committee completed a review of the work group membership and progress. With 3 years of experience as guidance, the guidelines for work group composition were confirmed and term limits for the work group leads were adopted. For membership, the process used in 2018, whereby work group members are selected by the cooperative groups, was confirmed. As with the original arrangement, work group leads are selected by the work group members, and it was determined that the leads will be asked to serve 3 years, with the opportunity to extend another 3 years if they are willing and if the work group so decides, or the work group can name a new lead.

The statistical work group consists of statisticians, data managers, clinicians, and PCDC data and technical team representatives. The first charge was the creation of the harmonized RMS data dictionary. Additional responsibilities included ensuring the quality and ongoing growth of the data dictionary by working with the PCDC team to create quality control checks and identify data elements to be added to the data dictionary. Critical to the success of the commons is the partnership between the PCDC team and the cooperative group members with detailed understanding of the disease, the data, and the research questions that will be supported by the commons. This model has been adopted and improved in subsequent disease-oriented groups that have engaged with the PCDC, with the key elements of the multidisciplinary approach to this work remaining as a central principle.

3.5 Data contributor agreement

To meet regulatory requirements, the PCDC (through the University of Chicago) entered into data contributor agreements (DCAs) with each cooperative group or institution that intended to transfer data to the data commons. Of note, the University of Chicago, as the data commons service provider, enters into this agreement directly with the data contributor. INSTRuCT is not a legal entity and is neither directly contributing data nor directly receiving the data and is not a party to the agreement (Figure 2). The DCA provides the mechanism for the data contributor to outline what data are being transferred to the PCDC and any restrictions. The DCA also assures the contributor that the PCDC will release the data only to authorized users.

3.6 Data access and project request process

Once data have been combined into a data commons, policies and procedures are necessary to govern access to the data. Early efforts led to a cohort explorer tool that is freely available to any user who registers to use it. The cohort explorer tool was designed as a means to identify whether suitable cohorts of patient data were available to proceed with specific research projects to answer clinical questions.

Two mechanisms by which a request for access to line-level data can be initiated were devised: project requests can be initiated from within an existing INSTRuCT work group or by an independent investigator not as part of a work group.

3.7 Publication policy

The INSTRuCT Executive Committee quickly recognized the importance of a publication policy. After reviewing the INRG publication policy, the committee had a robust discussion to identify the key elements of the policy, inclusive of project team involvement, authorship, acknowledgements, and review process. Goals for authorship policies focused on ensuring appropriate representation from each contributing group and specialty (including statistical specialists), in addition to assignment of authorship and co-authorship. INSTRuCT has adopted the International Committee of Medical Journal Editors (ICMJE) authorship criteria for determining co-authors, as it is the de facto standard used within the medical literature.¹⁶ To support the training goals of INSTRuCT, full professors are asked to consider appointing an early-career investigator as the senior author, or co-senior authorship with a young investigator.

As with the project request process, the publication policy addresses projects undertaken by the INSTRuCT work groups (and therefore sponsored by the cooperative groups) and projects initiated by investigators (who may be members of one or more of the cooperative groups). The processes are the same for work group- and investigator-initiated projects, with the exception of author selection. For projects proposed by investigators, the INSTRuCT Executive Committee takes an active role in ensuring a complete and balanced project team that can address the proposed project hypothesis and use of the cooperative group data in INSTRuCT.

3.8 Data use agreement

To meet regulatory requirements, the PCDC (through the University of Chicago) enters into data use agreements (DUAs) with each investigator who will receive data pursuit to an approved project (Figure 2). As is the case with DCAs, the University of Chicago, as the data commons service provider, enters into this agreement directly with the data user and their institution. INSTRuCT is not a legal entity and is not directly providing data and therefore it is not a party to DUAs. The DUA provides the mechanism for the PCDC to detail the terms of use of the data being transferred from the PCDC. These agreements include any specific requirements from the DCA under which the data being provided were contributed.