Association between primary-site surgery and overall survival in non-metastatic and metastatic colorectal cancer: a population-based study using the Surveillance, Epidemiology, and End Results database.

Background: Surgery is central to the management of localized colorectal cancer (CRC), whereas surgical decision-making in metastatic disease is more complex and highly individualized. This study aimed to investigate patterns of primary-site surgery recorded in the Surveillance, Epidemiology, and End Results (SEER) database, reasons for no surgery, patient and tumor characteristics, and observed overall survival (OS) in non-metastatic CRC, with a separate exploratory analysis of surgical patterns in metastatic disease. Methods: CRC tumor records diagnosed between 2006 and 2015 were identified according to prespecified eligibility criteria. Non-metastatic (M0) disease was the primary cohort. Descriptive analyses characterized recorded surgery, reasons for no surgery, baseline features, and survival; multivariable Cox regression, propensity-score matching (PSM), and sensitivity analyses estimated adjusted associations. Metastatic (M1) disease was analyzed separately as an exploratory cohort. Results: Among 14,130 M0 tumor records, five with unknown surgery status were excluded, leaving 13,807 with and 318 without recorded primary-site surgery. Surgery was not recommended in 213 of 318 no-surgery records (67.0%), and the method used to determine regional nodal status differed markedly by surgery status [absolute standardized mean difference (SMD), 2.158]. Median OS was 164 months with recorded surgery and 59 months without recorded surgery (log-rank P<0.001). In the primary Cox model, recorded surgery was associated with a lower average hazard of death over follow-up [hazard ratio (HR), 0.457; 95% confidence interval (CI): 0.397-0.527; P<0.001], but the proportional hazards assumption was not met. PSM yielded 317 pairs; the pair-stratified HR was 0.453 (95% CI: 0.344-0.598; P<0.001), and 60-month OS was 74.1% versus 50.2%. The M1 cohort comprised 694 tumor records. In the 2010-2015 subset (n=467), records without primary-site surgery more often had lung metastasis, bone metastasis, and at least two documented organ sites; four recorded surgical patterns were observed. Conclusions: In M0 disease, no recorded primary-site surgery was uncommon, usually reflected that surgery was not recommended, and was associated with substantially shorter observed survival. This association does not quantify the causal effect of surgery. Heterogeneous metastatic distributions and procedure patterns in M1 disease underscore the need for individualized clinical decision-making rather than routine primary-site surgery based on registry survival comparisons.