Mixture cure model for estimating short-term and long-term colorectal cancer survival

We used mixture cure mode to separately investigate the risk factors for long-term and short-term survival of colorectal cancer patients.

Colorectal cancer (CRC) is the second most common cancer worldwide. In cancer studies, patients' survival is the most important indicator of patients' status. Classical methods in analyzing the survival data usually apply Cox proportional hazard regression.

The study was performed on 1121 patients diagnosed with colorectal cancer. Mixture cure model with Weibull distribution and logit link function was fitted to data.

Odds of long-term survival for rectum cancer patients were lower than for colon cancer patients (OR=0.29(0.09, 0.9)). Also, patients with the advanced stage of the disease had lower odds of long-term survival compared to early-stage patients (OR=0.24(0.06, 0.86)).In the short-term, the hazard of death for people with normal BMI was lower than the underweight group (HR=0.4(0.21, 0.76)). The short-term hazard of death for rectum cancer was about half of the short-term hazard for colon cancer (HR=0.49(0.29, 0.81)). Further, people with moderately (HR=2.11(1.26, 3.55)) and poorly (HR=4.04(2.03, 8.03)) differentiated tumor grade had a higher short-term hazard of death compared to people with well-differentiated grade.

Predictive variables of colorectal cancer survival showed different effects in short- and long -terms. Site topography was a prognosis for both long-term and short-term survival; BMI and tumor grade were short-term predictors of survival while stage was a long-term predictor of survival.