(P-623) Generating Hypotheses About Never-smoking Lung Cancer Risk Factors Using Large Primary Care Databases

Background: Lung cancer in never smokers (LCINS) is the sixth most common cause of cancer death, but its etiology remains unclear

Objectives: To identify medical conditions or lifestyle factors associated with LCINS using a data-driven approach to generate hypotheses.

Methods: We used the UK’s Clinical Practice Research Datalink (CPRD; covering ~8% of the UK) to perform a population-based nested case-control study. Cases were first primary LCINS (N=1701) diagnosed between 1988-2018 in patients aged 30-89. Never-smoking controls (N=16963) were 10:1 frequency-matched on age, sex, clinical practice, year of diagnosis, and length of registration. Never smoking status was ascertained through clinical records. The study population had no evidence of prior cancer and ≥1 year of practice registration between study entry and 1 year prior to selection/diagnosis. Medical conditions and lifestyle factors were identified through diagnostic Read codes, and grouped into 89 categories (e.g., hypertension, angina) and 26 unifying themes (e.g., cardiovascular disease [CVD]) when ≥15 NSLC cases comprised the category. Exposure was assessed in the 1-10 and 10-32 years prior to the index date (date of lung cancer diagnosis or control selection). Conditional logistic regression and logistic regression with hierarchical random effects were used to estimate the association (odds ratios [OR], 95% CI) between each theme or category and LCINS. A multiple testing correction was applied for associations lacking a priori evidence.

Results: Study participants were 55% female and were registered in their practice for a median of 15 years (IQR: 6,28) prior to the index date. In the 1-10 years prior to selection, among all themes, infections/inflammation (1.2, 1.1-1.4), CVD (1.3, 1.1-1.4), and anemia (1.5, 1.2-1.9) were associated with higher risks of LCINS. Within the infections/inflammation theme, 3 novel conditions including autoimmune disease (1.3, 1.1-1.5) were associated with LCINS. Five known risk factors within the infections/inflammation theme including chronic obstructive pulmonary disease ([COPD] 2.9, 2.2-3.9) were also identified. In the 10-32 years prior to selection, COPD (4.3, 2.2-8.6), respiratory infections (1.3, 1.0-1.7), and deep vein thrombosis (2.3, 1.5-3.4) were associated with higher risks, and a healthy diet (0.8, 0.6-1.0), a lower risk, of LCINS.

Conclusions: We agnostically identified established and novel risk factors for lung cancer in never smokers, which will be validated in external data. Large clinical databases may be useful in generating hypotheses about poorly understood diseases to drive more targeted epidemiological research.