2014 Annual Meeting
The annual Cohort Consortium meeting, sponsored by EGRP and the Division of Cancer Epidemiology and Genetics (DCEG), was held on December 10-12, at the NCI Shady Grove Campus in Rockville, Maryland.
This summary reflects the open portion of the meeting conducted during December 10, 2014.
Session I: Welcome and Introductions
View Session I Details
Moderator: Dr. Deborah Winn
NCI Division of Cancer Control Cancer Control and Population Sciences (DCCPS)
Dr. Croyle provided an update on federal, NCI, and DCCPS activities relevant to the Cohort Consortium. He noted that NIH is expected to receive a small budget increase, but this will keep pace with the increasing costs of biomedical research. Some additional funds are expected to be available for Ebola and Alzheimer's research.
NCI has a new Deputy Director for Implementation Science, Dr. David Chambers. He participated in NCI's 7th Dissemination and Implementation Research Conference earlier in the week.
Dr. Lynn Penberthy is the new Associate Director of NCI's Surveillance Research Program (SRP). She led workshops in the fall to revisit gaps and incomplete aspects of SEER data and to develop ideas for improving SEER registry systems and data resources. Dr. Penberthy would be interested in receiving additional input from Cohort Consortium scientists regarding cancer surveillance needs and opportunities for expanding data linkages.
SRP is examining approaches for collecting more complete treatment data, including a study of oral anti-neoplastic treatment. NCI also is examining opportunities to obtain surveillance data from the private sector, for example, electronic medical record (EMR) data. NCI continues to examine challenges related to the use of Medicare claims data, including the problem of missing data for many variables. NCI is working with the American Society of Clinical Oncology (ASCO) on the Society's CancerLinQ™ project to expand cancer registry collection of biomarker and other information (e.g., residential history). Expanded collection of biomarker data might involve direct data feeds from laboratories and collaborations with private industry and data analytics groups. NCI also is interested in expanding tissue acquisition and repositories at cancer registries and is working on the development of a virtual tissue repository (VTR) linked to the Surveillance, Epidemiology, and End Results program (SEER). The VTR will be important for studies of rare tumors and could provide tissue information from population representative samples.
Dr. Croyle recognized the substantial scientific contribution of the Cohort Consortium. As an example, he mentioned the recent BPC3 paper published in the American Journal of Epidemiology on germline variance in breast cancer.
NCI Division of Cancer Epidemiology and Genetics (DCEG)
Dr. Stephen Chanock
Dr. Chanock described a recent NCI symposium, From Pedigrees to Populations, on the importance of epidemiology to all health research. During the symposium, NCI Director Harold Varmus said that epidemiology is connected to everything and "Almost all provocative questions begin with some epidemiological observation."
Dr. Chanock presented the Prevention Research Continuum, which covers etiology, prevention, and implementation activities. The Continuum provides guidance for the structuring of grants/studies. NCI's role in prevention research is foundational and focuses on etiologic studies. Observational studies can be as important as randomized prevention trials, which often may not be feasible or ethical.
NCI Cohort Consortium: Governance and Overview
Dr. Deborah Winn
Dr. Winn reviewed the governance of the Cohort Consortium. The Steering Committee is responsible for overall policy, management, and scientific direction of the Consortium. This Committee is comprised of principal investigators representing the diversity of the cohorts and three ex-officio NCI representatives. The Steering Committee conducts monthly teleconferences.
The Secretariat (now re-formed as the Steering Committee) developed Consortium bylaws and a governance structure over the past year with the goal of maximizing research; the changes were approved by the membership in late 2014 and are now being implemented. Key changes included an expanded Steering Committee, improved documentation of roles and responsibilities and the creation of the Chair-elect position. The Chair-elect replaces the Chair after a 1-year apprenticeship. In 2015, the Steering Committee Chair will be Dr. Elio Riboli and the Chair-elect will be Dr. Sue Gapstur.
The Consortium currently has 17 scientific and technical working groups. Four new scientific working groups were formed in 2014 on physical activity, biomarkers and breast cancer, early detection, and pooled analyses of risk factors for second cancers. One new technical working group on tissue collection and curation also was formed in 2014.
- View Session I Details
Session II: Science Reports and Panel on What's Next
View Session II Details
Moderator: Dr. Anthony Swerdlow
Overview of National Institutes of Health (NIH) and NCI Metabolomics Initiatives and Activities
Dr. Krista Zanetti
Dr. Zanetti provided an overview of metabolomic activities relevant to the Cohort Consortium. In 2012, the NIH Common Fund began to support a metabolomic research program with several components. NIH established six comprehensive metabolomic resource cores and two training programs to build infrastructure in this area. The NIH Common Fund Metabolomics program is reissuing the Collaborative Activities to Promote Metabolomics Research Administrative Supplement with a due date of February 13, 2015.
The Think Tank on the Use of Metabolomics in Population-Based Research was created to develop strategies for increased use of metabolomics in population-based research. The think tank included representatives from academic, nonprofit, government, and industry groups and focused on integrating data, resources, and infrastructure to support population-based studies of metabolomics.
A Trans-Agency Metabolomics and Epidemiology Working Group (MetEpi) recently was formed to promote strategies for establishing additional capacity for metabolomic analyses in population-based studies. The group involves members from the National Cancer Institute (NCI), National Heart, Lung, and Blood Institute (NHLBI), the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), and the Food and Drug Administration (FDA).
The MetEpi Working Group has published several papers on opportunities and challenges of metabolomics in epidemiology research. The MetEpi Working Group also organized a symposium at Experimental Biology 2014 which resulted in a published summary of the session. The group has another symposium planned at the Metabolomics 2015 meeting in San Francisco. In 2014, the MetEpi Working Group organized the Think Tank on Metabolomics and Prospective Cohorts: How to Leverage Resources to further develop capacity for supporting metabolomic analyses in population-based studies across cohorts. Prospective cohorts were invited to participate in the think tank if they met the following criteria: 1) at least 100 participants with blood metabolomic data, 2) identified metabolites, 3) mass spectrometry platforms (preferred), 4) follow-up for disease, and 5) both cancer and non-cancer study outcomes. Three methods experts also were invited to join the think tank. Twenty-three disease cohorts, including 12 cancer cohorts and three established consortia currently participate in the think tank. The think tank is seeking additional cohorts that meet the criteria.
The Think Tank on Metabolomics and Prospective Cohorts also established an international metabolomic consortium and currently set priorities for that group. The Steering Committee is charged with developing the structure of the metabolomic consortium and establishing the mission, membership criteria, and policies.
Participants inquired about the importance of metabolomics to cancer research and epidemiology over the next 10 years. Dr. Zanetti noted that metabolomics is in the early stages of development. Many challenges need to be overcome to effectively use metabolomic technology, especially in larger cohort studies. Currently, NIH is working on building infrastructure and bringing together the right people to move the field forward.
Metabolites are important because they are stable (e.g., relative to proteins). It will be important to support methodological research on pre-analytic factors that affect metabolite levels (e.g., time from processing, freezing and thawing, type of specimen, impacts of anticoagulants, etc.). Currently, most metabolomic data come from blood samples. Urine and tissue samples need to be collected in future metabolomic studies.
Metabolomics in Consortia
Dr. Joshua Sampson
Dr. Sampson discussed ways that metabolomics might be used in the Cohort Consortium and epidemiology in general. The primary goals of metabolomic studies are to identify metabolites that either indicate the presence of disease or predict disease risk. Measurement of metabolites is advancing. The cost for analyzing a sample is decreasing, the number of biomarkers that can be examined in metabolite analyses is increasing, and reproducibility is fairly high with about half of measured metabolites having intraclass correlation coefficients (ICC) greater than 0.8. Two main platforms currently exist to measure metabolites at the mass level: nuclear magnetic resonance (NMR) spectroscopy and chromatography coupled with mass spectrometry. NMR is less expensive and non-destructive but less sensitive and able to examine fewer metabolites than chromatography/spectrometry.
It is possible to examine the association of metabolites with conditions and diseases across cohorts. For example, comparisons of metabolomic studies of BMI demonstrate similar effects of BMI on metabolites across cohort studies using different types of samples, populations, and protocols. These findings suggest that metabolomic studies can be conducted at the consortium level. A key advantage of the consortium is larger sample sizes/power to predict risk of disease for meta-analysis, replication, detailed analyses, and interpretation. Measurements are sensitive to study design and collection procedures, which mean that only meta-analyses, not pooled analyses, will be possible. Replication will be critical in this relatively new area of research. Detailed analyses will be important to examine, for example, intermediaries between known risk factors and cancer or normal metabolite levels for different ages. Studies will focus on different associations (e.g., gene, traits, behavior) and researchers will need to examine other studies to interpret their own results.
The collaboration inherent in consortia also is critical to metabolomic research. The available platforms measure different sets of metabolites, so collaboration on methods and possibly development of a common platform will be important. Standardized scales also need to be developed.
For most cancers, metabolites represent different pathways. Analyses across studies could help to correlate different cancers with different pathways.
ICC distribution depends on the metabolites measured and the population. Investigators should assess ICCs with each new population and platform. When stronger associations are found, sample sizes can be lower.
Participants suggested examining multiple repeat samples in metabolomic studies. They also inquired about the time period between collecting samples and disease diagnosis. Investigators interested in disease etiology or effects of the disease will need to examine samples closer to the time of diagnosis. Studies of long term causes of disease, however, would benefit from samples taken several years or decades before diagnosis. Initial studies probably should examine samples taken no more than 2 to 3 years before diagnosis to obtain stronger signals.
Systemic Metabolism and Pancreatic Adenocarcinoma
Dr. Brian Wolpin
Pancreatic cancer likely will surpass breast and colorectal cancer as the second leading cause of cancer death. Eighty percent of pancreatic cancers are found too late and no screening test for pancreatic cancer is available to the general population. Alterations in metabolites can be detected in patients with pancreatic ductal adenocarcinoma (PDAC) before diagnosis. These alterations need to be understood to predict PDAC and learn about its biology.
Dr. Wolpin conducted a prospective study using banked pre-diagnostic blood samples from four cohort studies and approximately 500 cases of PDAC. He compared metabolite profiles for these cases to those of matched controls. The metabolite profiling was performed using liquid chromatography mass spectrometry. Medical records were examined and lifestyle factors and other conditions considered. Many metabolites were removed from the analysis because they could not be reliably measured.
The study found that three branch chain amino acids (isoleucine, leucine, and valine) were elevated with early disease (2 to 5 years prior to diagnosis). Analyses revealed that the three BCAAs were independently associated with pancreatic cancer risk after controlling for BMI, diabetes, and physical activity (PA). Elevation in these BCAAs even predicted PDAC without an intermediate diagnosis of diabetes.
Dr. Wolpin also examined mouse models of pancreatic cancer: LSL- KRAS, Tp53, and Pdx1-Cre. The same BCAAs were elevated in the mice that went on to develop cancer. In addition, mice with elevated BCAAs had no other symptoms or diseases/conditions and diet was the same in all mice (amount and type). The mouse study further revealed that the levels of the BCAAs entering all of the mice through intestinal absorption were the same. The mice with elevated BCAA levels appeared to have the amino acids leaking from tissue, which would explain both the elevated BCAA levels and reduced skeletal muscle mass. Muscle wasting is associated with PDAC and may be explained by a tumor that absorbs certain BCAAs required for muscle tissue maintenance. This question requires further investigation.
Prospective metabolite profiling should be used to examine cachexia and para-neoplastic diabetes in humans. Both of these conditions are unique to pancreatic cancer.
Some studies have demonstrated that the oral microbiome differs in pancreatic cancer patients. The microbiome could reflect levels of BCAAs.
Investigators were not able to examine glucose levels in the study because of the way samples were processed. They did examine hemoglobin A1c and insulin levels, which correlated with BCAA levels.
Immune Markers and the Role of Inflammation in Cancer
Dr. Danielle Carrick
Most epidemiologic inflammation studies are limited by the small number of analytes. Multiplex assays allow for the study of a larger number of markers using a small number of samples. In June 2014, NCI convened a workshop, Using Immune Marker Panels to Uncover the Role of Inflammation in Cancer, to share findings and lessons from cancer association studies using these methods, discuss the challenges with using multiplex immune marker panels in cancer epidemiology studies, and generate recommendations for moving the field forward. All speakers in this Cohort Consortium meeting's inflammation session participated in the workshop. Participants in the workshop included epidemiologists, immunologists, biostatisticians, and laboratory scientists. Sessions covered prospective cohort studies of circulating inflammatory marker panels and specific diseases/cancer types, biologic interpretations (from immunologists) of studies from the first session, state of the science for laboratory assays, and statistical issues.
Key recommendations from the workshop were to 1) encourage interdisciplinary research, 2) continue development/refinement of multiplex inflammation assays and incorporation of new platforms in epidemiology studies, 3) coordinate studies to control for multiple comparisons with adequate power, and 4) design discovery and replication studies, preferably measuring circulatory inflammatory marker levels over time prior to diagnosis.
Conceptual and Inferential Considerations in the Design and Analysis of Circulating Inflammatory Markers
Dr. Elizabeth Platz
Dr. Platz discussed design and statistical considerations when using multiplex assays in studying the association of circulating inflammatory markers with other immune markers and cancer outcomes. Her team's research in this area has identified many potential problems and possible solutions. A longer version of her talk is available at http://epi.grants.cancer.gov/events/immune-marker-panels/.
A key problem identified by Dr. Platz's team was intra-individual variability in immune marker concentrations over time. Most cancer cohort studies with stored blood only collect a single specimen. Exposures and infections cause immune markers to change over time. ICCs between 0.6 and 0.7 still can be obtained, which is consistent with other studies with different populations, number of repeated samples, and time intervals between samples. Nevertheless, these ICCs are not ideal for prediction. Solutions might involve excluding observations with above normal concentrations, which likely are due to transient infections. This approach would require determination of cut points and is difficult to do with large panels. ICC also can be estimated from other studies over time, but this approach depends upon many assumptions and often leads to high confidence intervals. Another possible solution is to select larger samples to allow for an attenuated relative risk (RR). Cohorts with repeated specimens also could be located, but this approach is expensive. Relevant exposure windows also may be missed given that samples usually come from middle aged and older people.
Another identified problem was unknown correlation between circulating immune markers and target organ inflammation and immune cell profile. Few studies have been conducted to date examining both tissue inflammation and circulating immune marker concentrations. The few studies that have been conducted show no correlation between circulating inflammatory markers in plasma and inflammation of bone or colon tissue in healthy people of normal weight. There is a need to collect more tissue from healthy people to determine the correlation between organ inflammation and circulating immune markers.
Reverse causation presents another problem in studies of inflammation biomarkers. Measures of cytokines might be confounded by the presence of cancer. It is difficult to determine, however, how long before diagnosis a cancer is/is not present. This problem even affects prospective studies.
Other issues include:
- Sampling of controls. Decisions about eligibility criteria for controls must be equally applied to cases to avoid bias.
- Cytokine SNP studies. SNP interactions with the environment (e.g., infectious agents, obesity, etc.) should be considered. The influence of a SNP on the gene product may not be manifest unless an elicitor of an immune response is present.
Circulating Immune Markers and Subsequent Lung Cancer Risk
Dr. Meredith Shiels
Several studies indicate that inflammation likely has an important role in the etiology of lung cancer. Dr. Shiels discussed two independent case-control studies nested in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) of the association between circulating inflammation markers and prospective lung cancer risk. These studies were able to measure multiple markers simultaneously using panels of multiplexed Luminex bead-based assays.
Replication and discovery studies were performed to reduce the markers examined to those with significant association with lung cancer risk. Ultimately, four immune markers (C-reactive protein [CRP], serum amyloid A [SAA], soluble tumor necrosis factor II [STNFRII], and CXCL9/MIG) were found to be significantly associated with lung cancer risk. Pooled analyses revealed similar results for current and former smokers and for lung adenocarcinoma and lung squamous cell carcinoma. CRP, SAA, and CXCL9/MIG levels were associated with lung cancer risk more than 6 years before lung cancer diagnosis.
These findings suggest that inter-individual variability in inflammatory response is associated with lung cancer risk. The studies matched cases and controls closely for for smoking, so this behavior is unlikely to account for the associations. Three of the four markers are elevated in people with inflammatory lung disease, so it is possible that underlying lung damage may account for some of the associations. Though three of the four markers can be produced by tumor cells, suggesting that associations may be induced by the presence of an undetected tumor, two of these markers were associated with lung cancer risk more than 6 years prior to diagnosis. In addition, these markers were not associated with later stage disease at diagnosis as would be expected if associations were due to markers produced by the tumors.
Additional analyses showed that the markers did not substantially improve the predictive value of a risk- based model. Circulating levels of these four markers, therefore, are not likely to be useful for lung cancer risk stratification. The markers, however, provide evidence for the role of inflammation in the etiology of lung cancer. Studies of predictors of lung cancer in never smokers are needed. The Cohort Consortium might be able to provide sufficient samples of never smokers for this type of study.
Circulating Inflammatory and Immune Markers and Risk of Non-Hodgkin's Lymphoma (NHL)
Dr. Mark Purdue
Dr. Purdue and colleagues examined the relationship between circulating inflammatory and immunologic markers on NHL risk in two nested case-control studies, PLCO and the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC). The nested PLCO study demonstrated an association between the immune markers soluble CD27 and 30 and NHL, a finding that had previously been observed in HIV-positive populations. The study also found an association with sTNFR1 and NHL risk.
Because of the small number of analytes examined in the first PLCO study, a second nested study was conducted to more broadly examine immunologic and inflammation pathways for NHL using Luminex bead-based assays. Four analytes, BCA-1, sTNFRII, and sVEGFR2, were independently associated with NHL after adjustment for other analytes.
Using the ATBC cohort, the investigators examined the same analytes identified in the PLCO studies in cases diagnosed long after blood collection. In this study sample, sCD23, sCD27, sCD30, and sTNFR2 were associated with NHL risk. The associations with sCD23 and sCD30 persisted for cases diagnosed as much as 15 to 20 years after blood collection. Analyses of markers associated with NHL subtypes found that sCD23 was strongly associated with chronic lymphocytic leukemia (CLL), although this finding may be explained by reverse causation. Associations with sCD23 and sCD30 observed over many years of follow-up for diffuse large B-cell lymphoma, an aggressive malignancy, are more consistent with an etiologic relationship.
These findings generally suggest that B-cell stimulation is important in the development of NHL. Investigators are now exploring pooling data across studies to better understand the associations of certain analytes with NHL subtypes. They also plan to explore the association of markers with suspected NHL risk factors and the underlying biologic mechanisms for this association.
Inflammation and Hepatobiliary Cancers
Dr. Jill Koshiol
Gall bladder cancer provides a good model for how inflammation contributes to carcinogenesis.
Dr. Koshiol measured immune markers in serum and bile in the Shanghai Biliary Tract Cancer case-control study. Several immune markers were substantially elevated in both serum and bile from gall bladder cancer cases compared to gall stone cases. Similar results were found in an analysis of immune markers in gallstone and gall bladder cancer patients in Chile.
These studies were cross-sectional so the immune marker levels could be related to disease effects. Cohort data are needed to determine the ability of immune markers to predict disease risk. The need for consortia is especially apparent with a rare cancer. Dr. Koshiol would like to use Cohort Consortium data to evaluate the role of immune markers in the development of gall bladder and liver cancer, which also is strongly linked to inflammation.
The EPIC study also found that sCD23 was associated with CLL many years prior to diagnosis. They could predict occurrence of the disease up to at least 7 years prior to diagnosis.
Participants asked about the benefit of examining a diversity of markers. They noted that different associations have been reported and few predictive markers have been identified in final analyses. Participants wondered about the correlation patterns between the various markers, the different aspects of immunity reflected by the markers, and whether they are complementary to each other. Dr. Shiels noted that the correlation between all markers in her studies was not strong but, among the four markers identified as most strongly correlated to lung cancer risk, CRP and SAA were more strongly correlated. Dr. Purdue noted that there was a nominal degree of correlation between key analytes identified in his study of NHL. When top findings were adjusted simultaneously, there was an attenuation of the magnitude of the odds ratio (OR). This finding suggests that an underlying process is being captured. His study, however, also found strong associations that remained independent upon adjustment for other analytes. These strong associations point to a single etiologic process that is likely to be important in studying disease subtypes such as CLL. In spite of correlations, important signals are generated in analyses of individual analytes.
Participants noted that the immune marker studies are important to guide future tissue studies. Investigators need to work with biologists to conduct both human tissue and animal studies of immune markers. Investigators interested in disease etiology and prevention need to know the usual state of analytes before the disease process begins, which indicates the need for healthy tissue. It is difficult, however, to obtain healthy tissue. Some tissue--for example, tissue from gall stone patients or ovarian tissue from women who have had a hysterectomy—is available from cancer-free patients. But tissue usually is obtained from people who have some medical condition. Tissue studies can be conducted in healthy animal models.
For multiplex panels, investigators need to consider whether the samples they plan to analyze are fit for purpose (i.e., how they are collected and stored and whether assays work). For example, panels from the Nurses' Health Study did not work for an immune marker analyses. Cytokine panels appear to be sensitive to multiple factors. Many investigators have had problems with low detectability. Dr. Shiels' group recently piloted a high-sensitivity cytokine panel that captures important interleukins. They plan to use this panel for future studies of lung cancer.
Participants inquired about approaches that might be used in the future to identify causal markers of various cancers. Dr. Purdue indicated that the markers identified in his study were not necessarily causal but captured underlying processes that drive lymphoma development. When a reproducible association between disease and a marker is found, investigators should follow up by examining biologic mechanisms that are driving that association. Dr. Carrick noted that the workshop also agreed that when designing future studies, it will be important focus on specific groups of markers for analysis and examine pathways rather than individual inflammatory markers.
Progress in Molecular and Histopathological Subdivision of Tumors and Relevance to Cohort Epidemiology
Drs. Lorelei Mucci and Mia Gaudet
Dr. Gaudet provided an overview of the new Cohort Consortium Tissue Working Group. The purpose of this Working Group is to integrate tissue biomarkers with epidemiologic studies. Group goals are to facilitate the sharing of lessons learned from tissue collection efforts with those interested in collecting tissue, develop best practices and guidance for collecting tumor blocks, and create a forum for discussing new technologies and assays and new collaborative research projects. Tissue is important for identifying causal factors, cancer subtypes, and new biomarkers.
Dr. Mucci discussed a survey distributed to Cohort Consortium investigators in November 2013 asking who collected or was interested in collecting tissue. Forty-two investigators responded and 15 reported that they were collecting tissue for at least one disease site, mostly for breast cancer (11 cohorts). Eight cohorts were collecting colorectal cancer tissue, seven collected ovarian cancer tissue, and five were collecting prostate and lung cancer tissue. The Working Group plans to conduct an expanded survey about the number of tissue blocks available and whether tissue microarrays (TMAs) have been constructed.
The Working Group also plans to educate through quarterly webinars. One webinar already has been conducted on collecting tumor tissue in cohort studies (available online). Participants were encouraged to provide input for the new Working Group by emailing Drs. Gaudet, Mucci, or Carrick.
Overview of Tissue Collection in Epidemiologic Studies
Dr. Peter Campbell
Dr. Campbell discussed the process of collecting tissue from participants in the CPS-II study. He is the PI of the colorectal cancer portion of the CPS-II Tissue Repository.
For the CPS-II study, investigators verified cancer diagnoses through patient medical records and cancer registries. They had to reabstract hospital contact information and accession numbers from the medical records. Once consent was obtained from patients, tissues were sent to pathologists for review before processing. Key milestones in the process of obtaining tissue were IRB approval to accrue diseased tissues directly from hospitals and obtaining consent. Almost 70 percent of patients contacted (or spouses in the case of deceased patients) gave their consent.
In the United States, most hospitals store tissues for 10 years. In the CPS-II study, investigators found that that were able to obtain tissue for 90 percent of patients if hospitals were contacted within 10 years of diagnosis. This proportion declined to less than 30 percent if more than 10 years had passed since diagnosis.
Additional challenges encountered in tissue collection include:
- Refusals from teaching and other hospitals that do not participate in external research.
- Reluctance of some hospitals to send tissue blocks. They only will send slides, which are more subject to loss or degradation of antigens and often incur additional costs for handling.
- Rising cost of tissue materials.
- Increasing clinical use of neoadjuvant therapies.
Research Opportunities for Using Tissues in Epidemiologic Studies
Dr. Rulla Tamimi
Dr. Tamimi discussed the characterization of tumor heterogeneity. Molecular testing of tumors can provide important insights into disease mechanisms and subtypes.
Many platforms currently exist for measuring somatic alterations in DNA and RNA. Formalin fixed paraffin embedded (FFPE) can be used for assays but older samples might present some difficulties given variability in processing and degraded DNA/RNA. These assays need to be scalable and cost effective. Tissue microarrays (TMAs) reduce workload because one block contains representative samples from a large number of participants, which reduces workload.
Morphologic context of molecular alteration is critically important. C-path technology is available to rapidly capture information on histology and morphology.
Dr. Tamimi discussed the various concepts, platforms, and assays for cancer molecular pathology. When choosing the right assay, investigators must consider the most relevant analyte (DNA/RNA/Protein), desired breadth of the analysis, desired sensitivity, and in situ versus ex situ. The location/source is known for in situ samples obtained directly from the tumor. Ex situ samples are easier to obtain, but may mix different tissue types and lack subcellular localization.
SNPs for the whole genome can be examined using either DNA or RNA, but DNA is most stable. Dr. Tamimi discussed the relative benefits of assays for measuring DNA alterations. Polymerase chain reaction assays (PCR) are cheaper, more sensitive, useful for studying genes with hotspot mutations, but are not comprehensive. Sequencing is less sensitive but more comprehensive to capture the mutational landscape of a tumor.
Several new assays are available for measuring RNA alterations, each with their own advantages and disadvantages. Technologies for examining proteins in FFPE tissues are in the early stages of development, but immunohistochemistry allows for the study of single proteins. Protein computational image analysis has improved objectivity and reproducibility, is scalable (a pathologist is not needed to individually score), and can extract hundreds of quantitative images from each image.
Participants asked about best practices for preserving material that investigators have extracted or collected for future work (beyond assays they already planned to run at the time of tissue collection). In particular, they wanted to know whether it is possible to store tissue sections for immunohistochemical work as well as extracted RNA. For immunohistochemistry, antigenicity in tissue sections decreases rapidly with time, so investigators try to send all tissue sections to laboratories as soon as they are sectioned. For DNA and RNA, no studies have compared the quality of downstream assays and the effects of freezing and thawing and processing after an extended period of time. Work needs to be done to examine factors that affect the quality of extracted DNA and RNA.
For CPS-II, about a quarter of the tissue received came in the form of pre-cut slides. The investigators were advised to dip slides in paraffin to preserve antigenicity and prevent oxidation. They also store tissues at very cold temperatures. Prospective collection increases the likelihood that investigators will receive unstained slides rather than tissue blocks.
Most investigators conducting tissue studies focus on cancers that offer sufficient numbers of cases. The problem is that less common cancers will not be studied in the context of cohorts. The Cohort Consortium may need a planned effort to collect tumor blocks for rarer cancers. One goal of the Tissue Working Group is to bring together investigators interested in rare cancers and document cohorts that are collecting tissues from these sites. Few incentives exist, however, for collecting tissue from rare tumors.
SEER is trying to expand the VTR, which could provide more rare tumor samples. In addition, the NIH Common Fund program, GTEx, has tissue samples available from healthy people upon request (no cost involved).
In response to a participant's question about standardization within the Cohort Consortium, Dr. Campbell indicated that he had discussed possibly submitting an application for harmonizing the KRAS markers for colon cancer.
Markers of Early Detection Working Group Report
Dr. Mia Gaudet
Dr. Carrick reviewed the aims and conclusions of the 2013 DCCPS- and DCP-sponsored Early Detection Cancer Research Workshop to facilitate collaboration in biomarker research. The workshop examined the feasibility of biospecimen sharing within the existing structure and using existing cohorts.
Dr. Gaudet and others met the previous day to discuss ideas for streamlining access to biospecimens in cohort studies for research on early markers of detection. This early group promoted the establishment of a small working group to meet with NCI leadership, clinical and laboratory scientists, and cohort investigators to further discuss efforts around this research topic. Some ideas included expanding the Cohort Consortium database to include additional information about biospecimens, preferably by year before diagnosis, universal policies for accessing biospecimens for discovery and validation in cohort studies, and modifying the Cohort Consortium proposal to create unified guidance for discovery and evaluation of biospecimen use across cohorts. Another idea was to have a common review panel for these proposals to provide recommendations that allow each cohort to determine whether to opt in/out of the project. Lastly, the group discussed the values of educational workshops to bring together basic scientists and epidemiologists to familiarize each other with their vocabulary, the value of biospecimens and other data from cohort studies, and important considerations and methodological issues associated with use of biospecimens and other data in cohort studies.
- View Session II Details
Session III: Workshop and Committee Reports
View Session III Details
Moderator: Dr. Wei Zheng
Overview of EGRP Cohort-Related Activities
Dr. Daniela Seminara
In December 2013, DCCPS held a meeting on Transforming Epidemiology for 21st Century Medicine and Public Health. Recommendations for transforming the practice of epidemiology included:
- Increasing access to data, metadata, and specimens to foster collaboration, ensure reproducibility and replication, and accelerate translation to population health impact.
- Supporting the harmonization of existing epidemiologic data and the creation of study repositories.
- Supporting processes for registration of new studies, data access and sharing, and collaborative analyses.
Recommendations for expanding cohort studies across the lifetime and including multiple health outcomes included:
- Mapping and registering existing cohort studies worldwide.
- Engaging with stakeholders and field leaders to discuss the concept of a national cohort for multiple health-related outcomes.
Meeting participants also recommended optimizing the use of resources for epidemiologic studies by encouraging the leveraging of existing resources.
Dr. Seminara emphasized the importance of transparency, accessibility, reproducibility, and translation in Cohort Consortium research. Current priorities for cohort-related activities include the building and solidifying of infrastructure, establishing best practices and guidelines for research across cohorts and for multilevel epidemiology studies. Guidelines also are needed to facilitate data sharing across cohorts.
Cancer Epidemiology Descriptive Cohort Database (CEDCD)
Dr. Amy Kennedy
NCI is in the advanced stages of assembling a descriptive database for cancer cohorts (CEDCD).
The CEDCD will be a web-based searchable database that includes cohort profiles and contact information; study design and eligibility criteria; enrollment and biospecimen numbers; cancer outcomes and deaths; policies, protocols, and questionnaires; publications and funded research projects; and links to cohort sites and related research projects. Initially, the database will include all cohorts focusing on cancer as a primary outcome, which includes the Cohort Consortium and other cohorts funded by EGRP. The CEDCD is expected to serve as a model for a worldwide cohort registration project.
Developers plan to send prefilled forms to PIs for completion at the end of 2014. Once forms are received, data will be cleaned and uploaded. Launch of the live website is scheduled for spring of 2015. Online updates of the website are expected to occur annually.
The home page will provide information on the purpose of the database, a description of its contents and the process for collecting data, and guidance for using the database. Users will be able to search on several criteria, compare information, and export data to Excel.
The CEDCD will increase transparency, thereby reducing bias, improving reliability, facilitating the planning of collaborations, the efficient use of resources, the identification of research priorities, and the performance of peer review. Challenges include the changing nature of longitudinal protocols and the burden to the investigators in providing timely updates.
Participants asked whether the CEDCD site would link to dbGaP. The home page will have links to other publicly available related sources.
Participants asked that expanding the database as information on tissue and early markers of detection expands be considered. Working groups would like these types of data to be searchable on a database that either is publicly accessible, available only to the members of the Consortium, or with a combination of public and private components. Working groups should share with developers the type of information they would like to include in the database in the future so that the CEDCD is designed to accommodate that information.
There has been discussion about a VTR that would include information on cases. This concept could be expanded in a way that facilitates access to and requesting of biospecimens.
Report on Recommendations of NCI's October 2014 Workshop on Novel Approaches and Challenges to Data Harmonization
Dr. Gabriel Lai
Data harmonization is crucial to conducting "big science" through pooled and meta-analyses. In addition, data harmonization is particularly important as the breadth and scope of collaborative studies increase; the research questions become more complex, requiring integration of multilevel datasets, and limited resources increase the need for efficiency and maximum use of existing resources.
Advantages include increased power, larger sample sizes, pooling, improved generalizability of results, increased validity of comparative research, and opportunities for multiple secondary analyses and interdisciplinary research.
NCI's Workshop on Novel Approaches and Challenges to Data Harmonization reviewed best practices for multilevel data harmonization for application to multilevel epidemiology research, including new tools and approaches. Topics discussed included innovative approaches to the harmonization of epidemiologic risk factors, clinical and outcomes data, biomarkers, genomics, and analytic issues related to data harmonization.
Best practices for data harmonization across the Cohort Consortium include available and readily accessible data, comprehensive meta-data documentation, and data compiled and classified in a compatible way (and necessary resources for doing this).
Key ongoing challenges to data harmonization include the lack of coordinating centers and the extra time and resources required for coordinating centers as well as the large effort expended for each project even with many cohorts participating in the same project. The last challenge might be reduced by defining and standardizing data-sharing policies. Other challenges include inconsistent interpretation of data dictionaries, inconsistent definition of phenotypes due to changing target variables, irreconcilable differences in exposure assessment between group, and data-sharing practices/policies.
Workshop recommendations included:
- Develop a collaborative framework/infrastructure, including authorship guidelines.
- Increase cooperation of multiple PIs and programmers.
- Facilitate linkage of existing databases.
- Increase interdisciplinary input.
- Create a Data Coordinating Center to ensure rigor of data, promote communication, document processes (useful for standardization), and manage and distribute data.
- Establish a Data Advisory Committee.
- Prioritize the use of standard measures and meta-data.
- Develop an initial general data transfer agreement.
- Establish a public repository for algorithms/coding.
- Consider potential next steps following the completion of a study (e.g., translation, replication).
- Create an ongoing forum for exchange of information and assessment of current practices.
- Provide funding and training support (e.g., junior investigators, new skills for experienced investigators such as combining/pooling data).
A Data Coordinating Center is needed for a Consortium, but some large cohort studies also might need their own data coordinating centers. Participants emphasized the need to begin by using existing coordination structures.
IRBs can complicate the data harmonization process. Requirements across countries will be a problem. Some of these issues will be discussed in the summary paper to be submitted to Cancer Epidemiology, Biomarkers & Prevention (CEBP).
Report on Recommendations of NCI's October 2014 Workshop on Data Sharing in the Cohort Consortium
Dr. Joanne Elena
The Workshop on Data Sharing in the Cohort Consortium was conducted to identify opportunities and resolve barriers to facilitate controlled access to epidemiologic data from observational studies.
The goals of the workshop were to identify:
- Challenges and possible solutions to sharing epidemiologic data.
- Key elements and potential epidemiologic data sharing model(s).
- Resources needed to facilitate increased epidemiologic data sharing.
The Data Sharing Planning Committee is seeking ideas from cohort members on ways to facilitate more efficient data sharing across cohorts.
The Workshop identified 42 challenges that can be broadly collapsed into the following categories:
- Resources (e.g., funding/programmer time).
- Data complexity.
- Maintaining quality of data/methods.
- Lack of standardization.
- Career concerns and recognition of PIs. In consortia, many authors appear on publications, which may impact the careers of investigators under the current academic advancement system.
- Technical, storage, and infrastructure needs. Technology can be used to make the data sharing process more efficient.
Recommended solutions to these challenges focused on resources. Suggestions included real time data collection, efficient use of programmers, posting effective practice documents online, and a centralized IRB for multi-site projects. Ideas are requested for what individual cohorts can do and what NCI can do. NCI is continuing the discussion of ongoing data sharing policies and standards. Participants were encouraged to approach NCI with challenges and ideas to facilitate easier data sharing.
With regard to the recommendation for a multi-site IRB, NCI is seeking comment on a proposal for a single-site IRB. Participants should examine and comment on that proposal.
Participants wanted clarification about what audience NCI would encourage members to share their data with. For these purposes, the conversation is confined to the research community. Dr. Elena acknowledged the strong commitment to data sharing that the Cohort Consortium members have shown using established processes. It will be useful to measure data sharing activities among Cohort Consortium members and discuss ways to reward these activities. An analysis of publications might be one way to evaluate data sharing activities.
Update on the NCI Funding Opportunity for Cohort Studies
Dr. Joanne Elena
A U01 has been released on Core Infrastructure and Methodological Research for Cancer Epidemiology Cohorts. This program announcement (PAR) focuses on establishing, maintaining, and upgrading new or existing infrastructure for cancer cohort studies. For example, this mechanism could be used to expand recruitment, support high-quality biospecimen banks, or link to medical records.
The PAR does not support hypothesis-generated research. Rather, it is designed to support infrastructure, including but not limited to expanding recruitment and retention, biospecimen collection and biobanks enhancement, and molecular characterization.
NCI would like to know how to help cohorts develop a more efficient infrastructure. Informing NCI about the tasks that require the most time and cost would allow NCI to better support the cohorts.
The PAR has recently been re-released (http://grants.nih.gov/grants/guide/pa-files/PAR-15-104.html).
Dr. Elena clarified that with the infrastructure component for risk, cohorts of more than 10,000 participants (and survivor cohorts greater than 2,000 survivor per cancer site) are required to be funded under this U01. Cohort studies with fewer than the required numbers can fund their infrastructure through an R01.
Participants asked about using this mechanism to establish coordinating centers for this U01. NCI must release a specific call for infrastructure development, such as this U01, to support the development of a coordinating center.
Cancer Registries and Data Linkages: Challenges and Opportunities
Cancer Registry of Norway
Dr. Giske Ursin
Dr. Ursin described the experience of the Cancer Registry of Norway. National cancer registries have existed in Nordic countries for longer than in other countries. Norway implemented compulsory cancer reporting in 1952.
Cancer registration in Norway incorporates pathology reports, death certificates, information from radiotherapy locations, hospital patient administrative systems, and clinical reports. Cancer data are available through the main cancer database, clinical cancer registries, screening programs, and other databases. The registry has access to complete (not estimated) population denominator data. NORDCAN, an online cancer database includes data from Norway, Sweden, Denmark, Faroe Islands, Finland, and Iceland. Data from NORDCAN can be used to show differences in cancer rates across countries.
Clinical cancer registries have detailed data on diagnosis and treatment, as well as some follow-up data. In Norway, eight clinical cancer registries currently have national status and receive governmental funding; the prostate and colorectal registries are the oldest at more than 10 years old. Clinical cancer registries with data on breast, lung, melanoma, lymphoma/CLL, and gynecologic and children's cancers are newer.
All the Nordic countries have screening programs on breast and cervical cancer. Dates and outcomes of every screening from 1995 onwards are registered in Norway. Colorectal cancer screening programs exist in Denmark and in some parts of other Nordic countries, including a pilot project in Norway.
Norway has cause of death, employment, education, salary, and country of birth data at the Statistics Norway registry (special rules apply regarding access to these data). Other registries include a medical birth registry, vaccination registry, prescription registry, and national patient registry with discharge records and information on outpatient diagnosis, including comorbid conditions.
Challenges in using data from Nordic cancer registries include ethical approval, restrictions on some types of data (e.g., socioeconomic), lack of a common IRB across Nordic countries, differences in the way clinical data are recorded between countries, and the complexity of screening data (read official documentation). Dr. Ursin recommended involving someone from the local cancer registry when attempting to use more detailed cancer registry data, and in particular when using screening data, from a Nordic country. Investigators also should carefully review the websites for the cancer registries. The Association of Nordic Cancer Registries can be a useful resource. Tabular data are widely accessible from the Nordic registries. For investigators who need more detail, there may be a cost associated with obtaining such data, in order to cover staff time spent on a project.
Participants asked about efforts to link cancer registry information to tumor blocks in Nordic countries. In Norway, registries have specimen numbers. Tumor blocks are kept at the hospitals and are never destroyed. Investigators must obtain consent from patients to use tumor blocks, although the ethical committees will in some instances waive that criteria. The registries have patient addresses but obtaining consent is a lot of work.
Cancer Registration in the United States
Drs. Meir Stampfer and Debbie Winn
In the United States, there are cancer registries in all 50 states. NCI's SEER program covers 28 percent of the U.S. population and the Center for Disease Control and Prevention's (CDC's) National Program of Cancer Registries (NPCR) covers the rest of the United States (and some Pacific Islands). Annual reports on cancer trends are developed every year in a collaborative effort between the North American Association of Central Cancer Registries (NAACCR), NCI, and CDC.
Currently there is no single source of cancer registration information in the United States. The only way to follow up with a cohort to determine who was diagnosed with cancer is through interviewing study subjects or contacting cancer registries one by one. Death certification occurs in states and laws and regulations vary. Investigators also need to work with IRBs in every state where they are seeking cancer registration data.
NCI is working on a virtual cancer pooling project to make cancer registration data available through a Linkage Coordination Center. The concept is being pilot tested at the University of Southern California, which is linked to multiple cancer registries through the Adventist Health Study. The pilot phase will examine whether counts of incident cancer cases in each participating registry can be obtained through this linkage approach. NCI plans to meet with the FDA and other organizations in February to expand this project. NCI is seeking a third party to perform the linkage in the near future.
The Social Security Death Master File, which provided rapid information on fact of death, is now less useful for this purpose because it is missing many deaths. National Death Index (NDI) information is still available. NDI information has a high level of completeness and is relatively timely.
A meeting at the National Academy of Sciences was held in September to discuss researchers' access to death records. There is widespread interest in this issue. As an initial step, getting more rapid ascertainment of fact of death information would be very useful to researchers.
- View Session III Details
Session IV: Summary and Open Discussions
View Session IV Details
Moderator: Dr. Bill Blot
Participants discussed the highlights of the posters that presented details of 12 different studies. Two of the posters concerned studies of vitamin D in relation to cancer risk. Evidence suggests that increased vitamin D intake lowers the risk for some kinds of cancer (particularly colorectal cancer in women), but not others (including breast cancer, as shown in one poster). People with darker skin tend to have lower circulating levels of vitamin D, which might account for some disparities in the incidence of colorectal cancer between Blacks and Whites. Prospective studies of vitamin D levels show that higher levels are associated with reduced all-cause mortality and this effect is stronger in Whites, with the association persisting after adjustment for BMI and physical activity levels.
Participants noted that the posters demonstrated the particular advantage of the Cohort Consortium in enabling studies to obtain adequate numbers of samples of serum to examine pre-diagnostic analyte levels for rare cancers. They recommended using the posters as examples of success in applications for funding.
Participants discussed the advantages of collecting tumor tissue routinely, selectively, or not at all. Some cohorts have or are planning to request funding to routinely collect tumor tissue. Study sections might reject routine tumor tissue collection for rare cancers because of the small number of cases. Applications therefore should emphasize plans to pool tissue data. Applicants also should emphasize that data will be useful broadly, not just for the proposed study. In the case of special initiatives (e.g., for collection of tumor tissue for rarer cancers) NCI staff can influence study sections prior to review.
In 2015, NCI plans to conduct regular meetings to discuss signature studies. Participants should send suggestions to Nonye Harvey or Debbie Winn for signature studies that they think should receive priority.
- View Session IV Details
Twitter Hashtag and Live Tweets from the Meeting
On Thursday, December 11, NCI staff tweeted key points and highlights from the main meeting using the meeting hashtag, #NCICohortConsortium. An archive of all the #NCICohortConsortium tweets can be viewed at the Healthcare Hashtag Project.
If you have questions about the meeting, contact Nonye Harvey, M.P.H. at the National Cancer Institute.