PROJECT SUMMARY/ABSTRACT Even the largest lung cancer screening trial to date, the National Lung Cancer Screening Trial (NLST) is insufficient on its own to answer questions about the effectiveness of lung cancer screening in diverse clinical settings. Findings from the NLST may not generalize to new populations due to the highly selected trial population, higher levels of adherence, and screening strategies that do not reflect current practice. Claims datasets, screening registries, and prospective studies offer opportunities for evaluating lung cancer screening strategies in real-world populations, but current methods that attempt to integrate this evidence are insufficient. Standard meta-analyses and simulation models use information from multiple trials and studies, but produce estimates that do not have a clear causal interpretation for any target population. The proposed project will bridge the gap between the available data and policy relevant questions by combining diverse information and developing new methods to estimate screening strategies that have a causal interpretation in nationally representative target populations. Our long-term objective is to reduce lung cancer mortality by combining multi-source data to learn about optimal lung cancer screening strategies in real-world populations. We will take steps toward this objective by achieving the following specific aims: (1) Use information from diverse sources to transport the effect of lung cancer screening strategies from trials to nationally representative target populations; (2) Identify individuals in nationally representative target populations who are most likely to benefit from lung cancer screening, including individuals traditionally underrepresented in trials, by jointly evaluating effect heterogeneity over multiple characteristics; (3) Model adherence patterns to lung cancer screening strategies in real-world settings and use these models to assess the effects of lung cancer screening strategies under different levels of adherence in nationally representative target populations; (4) Estimate in nationally representative target populations the comparative effectiveness of lung cancer screening strategies when the strategies have not been directly compared in the same trial but have been evaluated in different trials against a common comparator; (5) Use information from diverse sources as inputs in a simulation model to compare screening strategies that differ from those used in trials, in nationally representative target populations. This project will provide new insights on the comparative effectiveness of lung cancer screening strategies. Training in cancer prevention and statistical skills will launch my career as an independent researcher in data science and epidemiology that develops causal, statistical, and simulation methods to produce evidence that will lead to improved decision-making for cancer control strategies.

PROJECT NARRATIVE The proposed research on evidence synthesis via cancer-focused data integration is relevant to public health because it is a critical tool for evaluating lung cancer screening strategies and ultimately reducing cancer mortality. As lung cancer screening is implemented in more diverse settings, the applicability of trial findings to nationally representative populations is challenged. This work aligns with NCI’s priority to increase health equity across the cancer continuum because as a result of applying the proposed methods, we will produce high-quality evidence of lung cancer screening strategies for populations traditionally underrepresented in trials.