Executive Summary of the May 1-2, 2023 Precancer Atlas (PCA) Think Tank Meeting

Date Posted

On May 1-2, 2023, approximately 70 experts from academia, government, industry, and professional organizations globally met with the goal of developing a contemporary working definition or framework for the term ‘precancer.’  While there have been a number of working definitions proposed previously (e.g., 2006 National Cancer Institute Consensus Conference), there have since been substantial investments in research that have led to advancement of the understanding of some of the changes in cells and tissues prior to the development of cancer. Additionally, new methods and technologies have been developed for ascertaining and characterizing the transition to precancers. There is now an urgency to build upon this growing body of knowledge, to not only better understand the biology of precancers, but to quickly translate this knowledge into actionable cancer risk, prevention, and interception efforts.

Introduction and Charge

At the opening of the meeting, the National Cancer Institute (NCI) Division of Cancer Prevention Director stated that defining ‘precancer’ is perhaps the most important question for the field of cancer prevention. Defining precancer underpins the understanding of the malignant trajectory from early lesion to cancer, clinical outcomes, and opportunities to intervene. Further defining precancer will also stimulate additional research focused on the fundamental biological properties of these lesions. The importance of this question and far-reaching implications were evident in the charge given to participants in the meeting. The overall goal of the meeting was to propose a working definition or framework for the term precancer as well as recommendations for:

  • Studies undertaking comprehensive molecular and cellular characterization of precancers that could serve as models for the field
  • Building on ongoing efforts to characterize precancers, such as those led by the NCI Human Tumor Atlas Network (HTAN) supported through the Cancer Moonshot Initiative and the Precancer Atlas (PCA) initiative embedded within this, to develop effective prevention strategies
  • High-priority cancer types or at-risk populations with prospectively collected specimens in which meaningful studies can be conducted to move the field forward
  • Prevention- or early detection-focused studies with achievable milestones in 3-5 years

Individuals with diverse expertise were assembled to provide their perspectives on the definition of precancer given the implications for cancer research across all scientific disciplines (e.g., basic, epidemiology, translational, clinical, social, and behavioral), as noted by the charge above. Participants were asked to consider the definition or framework through the perspective of:

  • The molecular, cellular, immunologic basis of precancer, including the earliest changes detected and contributions from the immune and stromal cells in the surrounding microenvironment
  • The computational and bioinformatics challenges in data integration for precancer including data from genomic profiling, molecular and cellular characterization of the precancer and surrounding microenvironment, histopathology, spatial analyses
  • Pathology, including digital pathology, and the vast array of imaging techniques being utilized or new methods being developed to ascertain and characterize precancers in a moment in time and the changes that occur over time
  • The clinical and epidemiologic understanding of precancer and how the definition of precancer would influence future epidemiology and clinical studies as well as clinical management and decision-making, including the ability to predict the future trajectory of the lesion, changes in the surrounding tissue, and implications for the individual and their care
  • Integration of these molecular, pathohistological, and epidemiological data to predict clinical outcomes, both within an individual and across individuals within a population

Formal Presentations

Before meeting in subgroups to address the definition of precancer from various perspectives informed by methodologic expertise, brief talks focused on current progress in understanding precancer were presented from HTAN representatives leading its PCA initiative. Established in 2018, HTAN is leading the way in characterizing and integrating multiple types of data (e.g., genomics, histology, tumor microenvironment components and interactions, spatial analysis, and clinical data) to better understand the pre-malignancy to malignancy transition (HTAN Cell 2020) within selected cancer sites. Presentations at this meeting focused on lung cancer, sporadic and familial forms of colon cancer, breast cancer, and melanoma. More extensive information on these efforts can be found in a publication in Cancer Prevention Research. In addition, two plenary talks, one by Michael Stratton, Ph.D., entitled, “The Genomic Landscape of Premalignancy,” and another by Thea Tlsty, Ph.D., entitled, “Stromal Reprogramming and Cancer: Cause? Cure?,” helped to set the stage for further discussions in the breakout groups.

Summary of Breakout Group Discussions

Over two days and with these formal talks as background for discussion, the participants discussed the definition of precancer in four breakout groups aligned with the disciplinary perspectives noted earlier. Each breakout group then reported back to the full group of attendees with the goal of beginning to integrate these varied perspectives into a working definition or framework for precancer.

Molecular, Cellular, and Immunologic Basis of Precancer (Breakout Group 1)

The molecular, cellular, and immunologic basis of precancer breakout group (Breakout Group 1) focused on interactions within and between epithelial cells and stromal cells, and contributions of extracellular matrix components. Precancer studies have largely focused on the epithelial cells, however, the precancer resides in a tissue that responds to the local environment. In conditions such as inflammation or fibrosis, there are changes in the environment that regulate stromal:epithelial cell interactions and expression of extracellular matrix proteins. These changes to the stroma occur before changes to the epithelial cells and can exert a dominant, pro-tumorigenic effect when interacting with epithelial cells. These changes can also be dynamic and reversible, providing the opportunity to revert precancers by developing interventions that could disrupt the pro-tumorigenic cues from the stroma and extracellular matrix.

Given the importance of stromal interactions in cancer progression and that these changes can be detected early, Breakout Group 1 called for additional studies of the local precancer microenvironment (PME), as well as how systemic events such as chronic inflammation shapes the PME. The PME is distinct from the tumor microenvironment (TME) and should be examined with the same intensity as studies currently focused on the TME.

Much of the excitement and discussion in Breakout Group 1 centered on the precancer microenvironment (PME), contextualizing this with a focus on both the tissue in which the precancer is found and the individual overall, and the information yet to be gleaned from this understudied area. Discussants also recognized that while there are organ-specific contexts for the PME, there are likely to be unifying principles, especially in the context of inflammation, that could have relevance to multiple cancers. This would provide the opportunity to advance understanding of precancers across many cancer types at one time, rather than focusing on only a single cancer at a time. Long-term goals included determining how alterations in the PME and other early events could be used as surrogate markers for endpoints in early phase clinical trials.

Pathology, Digital Pathology, and Imaging (Breakout Group 2)

Breakout Group 2 proposed an initial definition in which the precancer was an aberration from normal that predisposes to malignancy, but the aberration could be based on changes that occur beyond histology. For example, biological alterations could be used to characterize the lesion but there would need to be a clear connection between the biological alterations and a cancer outcome to determine why the lesion was designated a precancer. The group reiterated that the biology alone without a connection to outcomes was not sufficient to drive the definition and highlighted there are many data points that could be incorporated into a definition. Breakout Group 2 suggested creating a risk prediction model utilizing all available data for a given precancer and assigning value or weighting to the different data points.

This led into a discussion about the contributions of pathology, digital pathology, and AI with there being both excitement and caution about these technologies. Ideally, AI and digital pathology would provide the pathologist with more tools and add value to the information they provide. However, the availability of these technologies is not uniformly distributed and activities such as the storing of images can be costly. There are also challenges to obtaining fresh or frozen precancer specimens since most lesions are small and excised for evaluation, leaving limited or no tissue for additional studies, especially serially over time. Another challenge is lack of funding for non-directed biorepositories for storing samples.

Breakout Group 2 also discussed the contributions of the microenvironment and how to assess this in a high-throughput way that would be complementary to other approaches (e.g., analyzing the same tissues used for H&E staining). The challenges with examining the microenvironment include defining when during progression is the most critical time to assess these changes, how to assess this in small, heterogeneous samples, prior knowledge of epithelial/stromal cell interactions for a given precancer, and whether the lesion is being excised or monitored.

Genomic Landscape Analysis of Precancer: Challenges in Data Integration (Breakout Group 3)

Breakout Group 3 focused on the genomic landscape analyses and data integration began by highlighting how little is known about the precancer state. Characterizing driver mutations, enhancing epigenomic profiling, increasing understanding of biological intrinsic factors as well as extrinsic factors involved in the progression of cancer, and how these are aligned spatially and temporally are all aspects where additional data are needed and must be integrated to provide the most comprehensive view of precancer. These integrated data could then define hallmarks of precancer.

Data across these different data types and studies will need to be standardized. For technologies such as whole genome sequencing (WGS), there are defined standards accepted and implemented by the community. For emerging technologies such as single-cell and imaging-based assays, it may be necessary to define a key set of parameters that should always be measured so these are collected systematically in all studies. There will also be a need to develop reference standards. For example, the importance of the microenvironment has been highlighted repeatedly, yet there are no standard assays or references for assessing the microenvironment and comparing these data across studies.

In addition to obtaining the data, other challenges include collecting the right samples and creating appropriate resources for sharing data. It is a continuum from normal tissue to precancer to cancer, therefore, normal and precancer samples will be a heterogeneous collection reflecting this continuum. In addition to defining precancer, there may also need to be standardized definitions of what normal is, where there is an accepted range or reference for comparison. Data from normal samples may be generated from a variety of different research initiatives, including ongoing research studies or through the newly launched NIH Common Fund program, Somatic Mosaicism across Human Tissues (SMaHT), and could be leveraged to understand early changes in the transition from normal to precancers. Across these efforts data will need to be standardized, shared, and integrated with other research activities.

Other major themes from Breakout Group 3 included the need for better models and technologies to further understand precancers. Multiple discussants highlighted the importance of being able to cluster multi-omics data into neighborhoods or communities that can quantified and possibly characterized as hallmarks of cancer, as well as developing technologies to assess precancers in three-dimensional space rather than only in a section on a slide. Understanding physical distances and spatial alignment is a high priority. Furthermore, the spatial and temporal components of precancer development in a human are not likely to be represented in current mouse models. By beginning to understand these elements, new integration models could be developed that would more faithfully replicate precancer development in humans.

Clinical and Epidemiologic Basis of Precancer (Breakout Group 4)

Breakout Group 4 were charged with focusing on prevention-related clinical and epidemiological aspects of defining precancer. They started with the important question of who the audience would be for this definition. The group reached the consensus that the primary audience would be the research community developing novel approaches to characterizing precancers and understanding the trajectory of these lesions in the progression to cancer. The shared goal is clinical intervention at the earliest possible point. Therefore, this definition also needs to be clinically relevant and understandable to patients and clinicians.

For epidemiology and clinical studies, the definition must be practical and incorporate a connection to progression to cancer. The definition must also include methods for ascertaining and characterizing lesions that meet this definition to facilitate integration into epidemiology and clinical studies. Therefore, the starting point for informing clinical and epidemiology studies is pathology with morphological changes necessary to identify the lesions, accompanied by other alterations (e.g., molecular, metabolic, microenvironment changes). Along with identifying the lesions, sufficient information needs to be collected to discern risk for progression and design an effective intervention to disrupt progression. There is also a need to balance the desire for longitudinal studies with large tissue banks with small, well-defined cohorts for understanding the etiology of precancers and subsequent outcomes such as progression.

Breakout Group 4 found it difficult to develop a succinct definition but was clear that there needed to be 1) a detectable morphological change that could be accompanied by other alterations; 2) enough information not just to identify a precancer but to predict risk of the lesion transitioning to cancer; 3) a minimum set of criteria to characterize precancers which could be differ between different types of cancer (e.g., hematologic and solid tumors). In addition, the group advocated for the definition to be adaptive since methods for ascertaining precancers are changing, resulting in epidemiology studies changing as new tools and methods are developed. Ultimately, a single definition may not be possible but rather a framework for considering what meets the threshold of a precancers may be more feasible.

Conclusions and summary

In the final session, strong support for developing an updated definition of precancer that incorporates recent knowledge of the fundamental biology of precancers and the potential to translate this into prevention strategies was reiterated. This definition could take several forms ranging from proposing a generalizable framework that could be applied to precancers to developing a statistical definition that would quantify and weight different types of data. These definitions or frameworks could differ across various types of cancer and would provide a basis for future studies aimed at understanding the precancer to cancer trajectory. Furthermore, terminology for this definition will need to be carefully selected as this needs to be understood by clinicians and patients, as well as provide a connection to subsequent cancer outcomes.

When developing future studies using or further refining this definition of precancer, there was discussion regarding which cancer types to prioritize. Different parameters for prioritization were suggested, such as: number of cases for a given cancer, availability of samples, lethality, screening programs for defined cancers, epidemiology data supporting determination of a precancer lesion, and known relationship to future outcomes. Many potential cancer types were nominated for prioritization in future studies. The overall group also raised the possibility of prioritizing based on a thematic approach focused on risk factors (e.g., smoking, obesity, chronic inflammation, genetic predisposition) to address multiple cancers.

New technologies and resources will also be needed to comprehensively study precancers. This includes identifying studies that are ongoing or will be starting soon in which prospective samples could be collected (e.g., COMET: A Study for Low-Risk DCIS; the Tomosynthesis Mammographic Imaging Screening Trial (TMIST); and the UK Our Future Health), as well as opportunities to access samples from prior studies. Ideally, both prospective and retrospective samples will be analyzed using multiple technologies and their resulting data types (e.g., whole genome or whole exome sequencing, epigenomics, metabolomics, microenvironment alterations, single cell data, spatial analyses) will provide a connection to long-term clinically relevant outcomes of importance to patients and providers. With the potential to collect a variety of different types of data, platforms and methods will need to be developed where a core set of assays and reference standards can be included and the data appropriately integrated. This will require strong, centralized data coordination and mechanisms for making the data broadly accessible. These platforms will also need to be flexible as new technologies and new types of data become available. Additional emerging areas included prioritizing spatial analyses and understanding how to incorporate data from multi-cancer detection (MCD) assays, as these become more prevalent, into a dynamic definition of precancer.

Next steps

This meeting provided two days of robust discussion reviewing current understanding of precancers and opportunities for advancing this research through developing a framework for defining precancer that incorporates multiple research perspectives. This proposed framework will outline critical elements, e.g., histopathology, cellular architecture, and spatial-genomic images, that must be included in defining precancer along with supporting elements, e.g., molecular, cellular, genomic and tumor microenvironment characterization, to strengthen evidence in support of the consequential precancer. As noted, no single mutation or other factor is essential or deterministic in isolation – but each contributes to risk. The overall objective will be to define the hallmarks of precancer in a manner that is quantifiable and distinguishes normal versus precancer for a given set of variables. These hallmarks will include the precancer environment and the exposome that are necessary but not sufficient to determine phenotypically normal but possibly at-risk lesions from precancer on а trajectory to becoming cancer. This dynamic, temporal component is crucial. Factors such as germline variation, somatic mutations, local stromal and matrix environment, immune environment/fitness/senescence, exposures, and their phenotypic consequences (e.g., inflammation), are likely to raise the risk of cells giving rise to a consequential precancer.

The discussions at this meeting revealed several common themes (see Summary Table) and served as a starting point for developing a more comprehensive definition or framework for precancer. The group agreed to continue these discussions and committed to producing a framework that can be communicated to the cancer research community within the next year.

Meeting summary prepared by Jessica M. Faupel-Badger, Ph.D., M.P.H., Planning Officer, NCI Division of Cancer Prevention with additional input from PCA Think Tank Editorial Team.

Summary Table – Common themes and starting point for developing a definition of precancer.

Beginning Framework for Defining Precancer: Common Themes Across Breakout Groups and General Discussion

Considerations for a developing a definition:

  • Audience for precancer definition: Goal is to support research to advance understanding of the precancer state and progression to cancer. Researchers are the primary audience, but the definition must also be clinically relevant and understandable to patients and clinicians.
  • Definition vs. Framework: The definition of precancer could vary across different cancers (e.g., solid vs. hematologic malignancies; sporadic cancers vs. cancers with a hereditary component), requiring cancer-specific definitions. Alternatively, a more general framework highlighting common fundamental features, characteristics, and properties (i.e., hallmarks) of precancers could be created.
  • Definition needs to be dynamic and adaptable: The definition of precancer should not be limited by current technology or methodology. New tools and methods for ascertaining and characterizing lesions are being developed, which will change epidemiology, clinical, and basic science study design.
  • Definition linked to outcomes: Need to know trajectory to cancer to inform risk prediction and develop effective interventions.

Necessary components of a definition:

  • Histology as a starting point: Must be able to ascertain and characterize the lesion for epidemiology and clinical studies, as well as for studies examining fundamental biology.
  • Microenvironment as a key contributor: Alterations in the microenvironment occur before there is detectable cancer and contribute to cancer progression. These alterations could include changes in immune cell populations, stromal cell communication, and expression of extracellular matrix proteins. The definition of precancer should incorporate changes in the precancer microenvironment (PME).
  • Integrating multi-omic data: Characterization of metabolic changes, alterations in microenvironment, and spatial components can be quantified and included in the definition of precancer and prediction of risk. All information should be collected, and the data weighted for risk-prediction models.
  • Incorporating timing and transitions: Precancers arise from normal tissues and transition to cancers, across varying amounts of time and levels of risk. Definitions of precancer should incorporate these transitions (e.g., normal to precancer as well as precancer to cancer) and the timing of progression to enable prevention and interception studies.


External Attendees

Adewole Shomari Adamson, M.D.
MPP Dell Medical School at The University of Texas at Austin

Robert W. Allan, M.D.
U.S. Department of Veteran Affairs - North Florida / South Georgia Veterans Health System

Manisha Bahl, M.D., M.P.H.
FSBI Massachusetts General Hospital

Esteban Braggio, Ph.D.
Mayo Clinic Arizona

Russell Ray Broaddus, M.D., Ph.D.
University of North Carolina School of Medicine

Alexandra S. Brown, M.D., FASCP
American Society for Clinical Pathology

Joshua Campbell, Ph.D.
BU-BMC Cancer Center

Andrew Tan Chan, M.D., M.P.H.
Massachusetts General Hospital

Robert Coffey, M.D.
Vanderbilt University

Tim Coorens, Ph.D.
Broad Institute of MIT and Harvard

David Crosby, Ph.D.
Cancer Research UK

William Dahut, M.D.
American Cancer Society (ACS)

Jeremy Davis, M.D.
National Cancer Institute (NCI)

Li Ding, Ph.D.
McDonnell Genome Institute

Rebecca Clare Fitzgerald, OBE FMedSci
University of Cambridge

Irene Ghobrial, M.D.
Dana-Farber Cancer Institute

Julie Gralow, M.D., FACP, FASCO
American Society of Clinical Oncology

Ernest Hawk, M.D., M.P.H.
The University of Texas MD Anderson Cancer Center

Jeffrey Hooke, M.D., FCAP
Uniformed Services University of the Health Sciences

Yujin Hoshida, M.D., Ph.D., AGAF, FAASLD
University of Texas Southwestern Medical Center

Shelley Hwang, M.D., M.P.H.
Duke University Comprehensive Cancer Center

Catriona Jamieson, M.D., Ph.D.
Sanford Stem Cell Institute

Neil Kay, M.D.
Mayo Clinic

Al Kovatich, Ph.D.
Murtha Cancer Center Research Program (MCCRP)

Ken Lau, Ph.D.
Vanderbilt University School of Medicine

Mark Lawler, MRCPath, FRCPath, Ph.D.
Queens University Belfast

Scott Lippman, M.D.
Moores Cancer Center at UC San Diego Health

Karen Lu, M.D.
The University of Texas MD Anderson Cancer Center

Anirban Maitra, M.B.B.S.
The University of Texas MD Anderson Cancer Center

Angelo M De Marzo, M.D., Ph.D.
Johns Hopkins University School of Medicine

Daniel Thomas Merrick, M.D.
University of Colorado, Denver/Anschutz Medical Campus Department of Pathology

Janette Merrill, MS CHES
American Society of Clinical Oncology (ASCO)

Ajit Johnson Nirmal, Ph.D.
Brigham and Women’s Hospital

Timothy Rebbeck, Ph.D.
Dana-Farber Cancer Institute

Pierre Saintigny, M.D., Ph.D.
Centre Léon Bérard

Michael Paul Snyder, Ph.D.
Stanford Medicine

Avrum Spira, M.D., M.Sc.
Johnson & Johnson

Michael Rudolf Stratton, Ph.D.
Wellcome Sanger Institute

Thea Tlsty, Ph.D.
University of California, San Francisco

Kenneth Tsai, M.D., Ph.D.
H. Lee Moffitt Cancer Center & Research Institute

Ignacio Ivan Wistuba, M.D.
The University of Texas MD Anderson Cancer Center Life Science Plaza

Laura Wood, M.D., Ph.D.
Johns Hopkins University School of Medicine

Matthew Young, Ph.D.
National Cancer Institute

NCI Attendees

Jessica M. Faupel-Badger, Ph.D., M.P.H.
Philip E. Castle, Ph.D., M.P.H.
Curtis C. Harris, M.D.
Shannon Hughes, Ph.D.
Sharmistha Ghosh-Janjigian, Ph.D.
Erika M. Kim, Ph.D.
Victor Kipnis, Ph.D.
Tracy Lively, Ph.D.
Felicia Evans Long, M.B.A.
Guillermo Marquez, Ph.D.
Richard Mazurchuk, Ph.D.
Lori Minasian, M.D., F.A.C.P.
Susan Orr
Howard L. Parnes, M.D.
Christos Patriotis, Ph.D., M.Sc.
Angie Petruzzelli
Jo Ann Rinaudo, Ph.D.
Vikrant Sahasrabuddhe, M.B.B.S., M.P.H., Dr.P.H.
Shizuko Sei, M.D.
Sudhir Srivastava, Ph.D., M.P.H., M.S.
Eva Szabo, M.D.
Asad Umar, D.V.M., Ph.D.
Juan Miguel Villanueva, B.S.
Paul Wagner, Ph.D.
Wendy Wang, Ph.D., M.Sc.
Nicolas Wentzensen, M.D., Ph.D., M.S.