Methods in regression analysis in surgical oncology research-best practice guidelines

Using real working examples, we provide strategies and address challenges in linear and logistic regression to demonstrate best practice guidelines and pitfalls of regression modeling in surgical oncology research.

To demonstrate our best practices, we reviewed patients who underwent tissue expander breast reconstruction between 2019 and 2021. We assessed predictive factors that affect BREAST-Q Physical Well-Being of the Chest (PWB-C) scores at 2 weeks with linear regression modeling and overall complications and malrotation with logistic regression modeling. Model fit and performance were assessed.

The 1986 patients were included in the analysis. In linear regression, age [β = 0.18 (95% CI: 0.09, 0.28); p < 0.001], single marital status [β = 2.6 (0.31, 5.0); p = 0.026], and prepectoral pocket dissection [β = 4.6 (2.7, 6.5); p < 0.001] were significantly associated with PWB-C at 2 weeks. For logistic regression, BMI [OR = 1.06 (95% CI: 1.04, 1.08); p < 0.001], age [OR = 1.02 (1.01, 1.03); p = 0.002], bilateral reconstruction [OR = 1.39 (1.09, 1.79); p = 0.009], and prepectoral dissection [OR = 1.53 (1.21, 1.94); p < 0.001] were associated with increased likelihood of a complication.

We provide focused directives for successful application of regression techniques in surgical oncology research. We encourage researchers to select variables with clinical judgment, confirm appropriate model fitting, and consider clinical plausibility for interpretation when utilizing regression models in their research.