A bayesian approach to model the underlying predictors of early recurrence and postoperative death in patients with colorectal Cancer

Table 2 Predictors of non-terminal and terminal events utilizing Bayesian Independent AFT model with DPM baseline survival distribution

		Recurrence		Death without recurrence		Death after recurrence
		TR	95% CI	TR	95% CI	TR	95% CI
Age at Diagnosis (years)	Trend effect	0.956	0.942–0.970 *	0.996	0.906–1.135	0.999	0.961–1.037
Gender	Male	0.835	0.624–0.972 *	1.006	0.945–1.177	0.977	0.977–1.129
Metastasis to other sites	Yes	1.063	1.063–1.086 *	1.071	0.925–1.106	0.946	0.944–0.969*
Number of chemotherapies	Trend effect	1.069	0.989–1.069	1.019	0.909–1.147	1.045	1.045–1.045
Grade (differentiation level)	Well	Referent	---	---	---	---	---
	Moderate	0.974	0.877–1.062	1.059	0.950–1.078	0.996	0.947–1.001
	Poor	0.944	0.721–0.961 *	1.067	0.852–1.225	0.966	0.834–0.975*
Tumor size	Trend effect	0.890	0.855–0.960 *	1.002	0.972–1.038	1.039	1.000- 1.206*
PT-Stage	T2	Referent	---	---	---	---	---
	T3	1.048	0.927–1.050	1.018	0.994–1.083	1.066	0.975–1.233
	T4	1.006	0.914–1.068	1.095	0.978–1.095	1.077	0.992–1.162
	TX	NC	NC	NC	NC	NC	NC
PN-Stage	N0	Referent	---	---	---	---	---
	N1	1.033	1.033–1.077*	1.147	1.019–1.196 *	1.057	1.019–1.057
	N2	1.048	0.939–1.252	1.075	1.009–1.251 *	0.874	0.835–0.877
	NX	NC	NC	NC	NC	NC	NC

NC: not computable; CI: Credibility interval; TR: Time Ratio
Deviance information criterion (DIC = 1759), Logarithm of the pseudo marginal likelihood (LPML= -816)
The frailty component was significant in the multivariable model (Variance of frailties: 0.847, 95% CI (0.75 − 0.018))
Trend effect: The model considered the trend effect for ordinal categorical variables. The variables BMI category, cancer site, surgery, radiotherapy, chemotherapy, morphology, disease stage could not be entered in the model in the multivariable model (All P > 0.05)
*: P < 0.05

ISSN: 1471-2288