BackgroundWith the increasing use of CT screening for lung cancer and the advances in imaging and surgical techniques, as well as the improved understanding of the relationship between imaging features and pathology/prognosis of lung cancer, there is a need to reevaluate and tailor the surgical approach for treating early-stage lung cancer. The decision between lobectomy and sublobar resection depends on various factors, and surgical decision-making process is multifaceted, involving patient characteristics, clinical factors, surgeon factors, and hospital settings. Recent studies have shown that sublobar resection may be a viable alternative to lobectomy. Precision medicine aims to identify the most effective treatments and tailor the treatment strategy to individual patients, which will not only improve quality of care and long-term outcomes but also reduce unnecessary workups and treatments. ObjectiveThis dissertation research focuses on understanding the complex surgical decision-making process for early-stage lung cancer treatment, particularly in the context of sublobar resection versus lobectomy, as well as how one could improve the process by introducing objectivity with considerations of potential heterogeneous treatment effects in subgroups of early-stage lung cancer patients. This dissertation is divided into three objectives: 1) to examine the underlying latent constructs of tumor aggressiveness and surgical invasiveness, 2) to explore the factors involved in the decision between lobectomy and sublobar resection, including tumor aggressiveness, patient and surgeon characteristics, and nodule features, and 3) to explore how treatment effects on lung cancer-specific survival may vary between different subgroups of early-stage lung cancer patients.Methods In the first part of the analysis, using data from a prospective cohort of 578 patients who enrolled in the Initiative for Early Lung Cancer Research on Treatment (IELCART) project and underwent surgery for clinical stage IA NSCLC at the Mount Sinai Health System (MSHS) between 2016-2022, the study utilized exploratory and confirmatory factor analyses to examine the underlying latent constructs of tumor aggressiveness and surgical invasiveness, and a multiple-indicator, multiple-cause model to elucidate the relationship between tumor aggressiveness and surgical invasiveness, and how this relationship is modified by other patient and surgeon characteristics and nodule features. In the last part of our dissertation research, using data from 746 participants enrolled in the International Early Lung Cancer Action Program (I-ELCAP) between 1992-2022 who underwent surgical resection for first primary clinical stage IA (≤3cm) non-small cell lung cancer, we explored how treatment effects on lung cancer-specific survival may vary between different subgroups of early-stage lung cancer patients using machine learning based approaches under the counterfactual framework.ResultsUsing exploratory factor analysis and confirmatory factor analysis on data from IELCART-MSHS, this research identified two one-factor latent structures with acceptable fit: one characterizing tumor aggressiveness and another characterizing surgical invasiveness (CFI=0.99, TLI=0.98, RMSEA=0.032, SRMR=0.038). Using a MIMIC model, we found that tumor aggressiveness was positively associated with surgical invasiveness, but that this relationship was modified by patient and surgeon characteristics, as well as by other nodule features. Younger patients (β=-0.214, p=0.017), those with lower BMI (β =-0.181, p=0.049), those with tumors in the right lung (β=0.21, p=0.022), and those treated by male surgeons (β =0.274, p=0.019) were more likely to receive more invasive surgery. In the last part of our dissertation research, we deployed machine learning based approaches and revealed notable treatment effects heterogeneity of sublobar resection vs. lobectomy on lung cancer-specific survival, which varied based on sex, smoking status, comorbidity CT evidence of emphysema, and nodule consistency. Among males, there was no difference in median lung cancer-specific survival between sublobar resection and lobectomy (1.0, 95% CI: 0.99, 1.02), while females had longer median survival with sublobar resection (1.14, 95% CI: 1.13, 1.15). Our study also found that sublobar resection was favored for females, regardless of the presence of CT evidence of emphysema, and that former and never smokers benefited more from sublobar resection than current smokers. The benefit of sublobar resection was also found to be greater for solid tumors (1.24, 95% CI: 1.21, 1.26) than part-solid tumors (1.14, 95% CI: 1.07, 1.22) in female former/never smokers without CT evidence of emphysema. In male patients with self-reported history of diabetes or cardiovascular diseases, the comparison of lung cancer-specific survival between sublobar resection and lobectomy yielded inconclusive results. Lobectomy provided better outcomes in male patients who were current smokers, had no history of cardiovascular disease, but had CT evidence of emphysema (0.93, 95% CI: 0.87-0.99), while sublobar resection resulted in longer median lung cancer survival than lobectomy in male patients who were former or never smokers, had no history of cardiovascular disease or diabetes, and did not show CT evidence of emphysema (1.10, 95% CI: 1.07, 1.13), particularly in those with part-solid tumors (1.04, 95% CI: 1.03, 3.14).ConclusionsThis dissertation research reveals the complexity of surgical decision-making in early-stage lung cancer treatment. Our research not only provides the first examination of tumor aggressiveness as a latent factor involved in decision making network, but also provides important information regarding the magnitude and extent of factors involved in surgical decision making for early-stage lung cancer treatment. Our study provides valuable insights into the potential benefits of sublobar resection in specific patient populations, such as female patients, those with specific comorbidities, smoking status and nodule consistency on CT. Further research is needed to confirm these findings. Overall, the findings underscore the importance of personalized medicine, allowing for targeting of interventions to subgroup of patients where the treatment is more likely to be beneficial and improve treatment efficiency.