Development, Application, and Evaluation of Prediction Models for Cancer Risk and Prognosis (R21) | Apply Online

Frequently Asked Questions (FAQs)

What is PA-07-022?

PA-07-022 is a National Institutes of Health (NIH) National Cancer Institute (NCI) funding opportunity titled "Development, Application, and Evaluation of Prediction Models for Cancer Risk and Prognosis (R21)." It supports exploratory research focused on improving prediction models that estimate cancer risk (the likelihood of developing cancer) and prognosis (the likely course or outcome once cancer is present).

What is the main goal of this funding opportunity?

The main goal is to stimulate innovative, model-focused research that produces practical, testable cancer risk and prognosis prediction models. The emphasis is on creating, improving, applying, comparing, validating, and evaluating the real-world performance and utility of prediction models, rather than conducting basic discovery research.

What type of projects does NCI want to fund under this FOA?

NCI is looking for exploratory R21 projects that are grounded in real data and have a clear plan to evaluate model performance. Projects may refine established models using existing datasets, develop entirely new models, or conduct rigorous validation to ensure models work reliably across different populations and settings.

What does "prediction models for cancer risk and prognosis" mean in this context?

In this FOA, prediction models are tools or methods that combine one or more inputs (for example, clinical variables, demographic factors, pathology data, environmental exposures, outcomes, and genetic information) to estimate either (a) a persons chance of developing cancer (risk) or (b) a patients likely disease course or outcome after cancer is diagnosed (prognosis).

What grant mechanism is used for PA-07-022?

This opportunity uses the NIH R21 mechanism, which typically supports early-stage, proof-of-concept research, method development, and other high-impact ideas that are still exploratory but supported by data and a concrete evaluation strategy.

Who is the intended audience for this FOA?

NCI explicitly calls out clinicians, epidemiologists, geneticists, statisticians, and translational researchers. The FOA encourages teams that combine clinical insight with strong quantitative and translational methods.

What are the top scientific priorities emphasized by NCI for these prediction models?

The FOA highlights two major challenges as priorities: (1) integrating diverse data types that may live in different systems or formats, and (2) performing adequate validation so models are tested in independent populations and not only fit the data they were built on.

What kinds of data integration are encouraged?

The FOA encourages models that combine heterogeneous data, such as clinical and demographic variables, pathology findings, environmental and epidemiologic exposures, outcomes data, and genetic information. These data may come from multiple data warehouses or smaller "data marts."

What does NCI mean by "adequate validation" of a prediction model?

Adequate validation means demonstrating that the model performs reliably beyond the dataset used to develop it. The FOA specifically emphasizes testing models in separate, independent populations, which helps show that performance generalizes across different groups and settings.

What performance measures does NCI expect applicants to consider?

The FOA points to standard predictive performance metrics that should be established through validation, including sensitivity, specificity, positive predictive value, and negative predictive value. This signals an expectation of careful attention to real-world predictive accuracy, not just statistical significance.

Are projects that focus on discovering new biomarkers responsive to this FOA?

No. The FOA explicitly states that applications primarily focused on identifying or characterizing prognostic or diagnostic biomarkers are not considered responsive. The central contribution must be the prediction model work itself (development, application, comparison, validation, or utility evaluation), with any biomarkers serving as model inputs rather than the main endpoint.

Can biomarker information be included at all?

Biomarker or genetic information may be included as inputs to a prediction model, particularly as part of integrating diverse data types. However, the primary focus must remain on the modeling and validation framework and on demonstrating predictive performance and reliability.

What stages of cancer research does this FOA cover?

This FOA spans multiple cancer-related activity areas reflected in the CFDA numbers listed in the announcement (93.393, 93.394, 93.395, 93.399), which collectively cover prevention, detection/diagnosis, treatment, and control research. The common theme is improving prediction models relevant to these areas.

Is cost sharing or matching required?

No. The FOA states that it does not require cost sharing or matching.

Who is eligible to apply?

Eligibility is broad and includes nonprofit organizations (with and without 501(c)(3) status), small businesses, for-profit organizations (other than small businesses), private and public institutions of higher education, and other eligible entities as described in the full announcement.

Are foreign institutions eligible to apply?

Yes. Foreign institutions are explicitly eligible, which can be important for teams working with international cohorts or conducting validation in non-US populations.

Is this a time-limited opportunity, and what were the key dates?

The FOA was posted on November 1, 2006, used multiple receipt dates (recurring submission cycles rather than a single deadline), and was archived on February 2, 2010. That means it was active during that period and is no longer active after being archived.

What does it mean that the FOA used multiple receipt dates?

Multiple receipt dates means applications could be submitted across recurring cycles rather than one single deadline. The FOA was structured to accept submissions over time while it remained active.

What is the practical impact NCI is aiming for with these projects?

NCI is aiming for better-performing and better-validated cancer risk and prognosis prediction models, especially models that can integrate heterogeneous data and demonstrate consistent accuracy across more than one population. The expected impact is more trustworthy decision support and research tools for clinical and public health contexts.

What kinds of outcomes or deliverables fit the spirit of this FOA?

Based on the description, deliverables aligned with the FOA include prediction models with clear evaluation plans, evidence of validation in independent populations, and established predictive performance metrics (such as sensitivity and specificity). The focus is on models that are practical and testable, with demonstrated reliability across settings.

Is this FOA intended for basic discovery research?

No. The FOA emphasizes practical prediction tools and model evaluation rather than basic discovery efforts. It prioritizes work that strengthens prediction tools used by researchers, clinicians, and potentially the public.