Table 3.

Unadjusted and adjusted association between LN yield (≥ 10 vs. < 10) and recurrent cancer after surgical resection of T1 CRC

		Hazard ratio (95% CI)	P value	Maximum (97.5th percentile) IPWc	Post-IPW c-indexd	Complete case (%)e	Observed data points (%)f
Total cohort N = 1017	Unadjusted	0.27 (0.10–0.76)	0.01	–	–	100	100
	Adjusted for clinical factorsa	0.19 (0.06–0.60)	0.005	5.4 (4.4)	0.48	88	98
	Adjusted for clinical & histological factorsb	0.20 (0.06–0.67)	0.009	7.1 (4.4)	0.50	18	88
LN negative patients N = 933	Unadjusted	0.25 (0.08–0.83)	0.02	–	–	100	100
	Adjusted for clinical factorsa	0.21 (0.06–0.77)	0.02	5.6 (4.5)	0.48	88	98
	Adjusted for clinical & histological factorsb	0.23 (0.06–0.81)	0.02	7.1 (4.5)	0.49	17	87

CI confidence interval, CRC colorectal cancer, IPW inverse probability weighting, LN lymph node

^aAge (continuously), sex (male vs. female), tumor location (right colon vs. left colon vs. rectum), tumor size (continuously), tumor morphology (pedunculated vs. non-pedunculated) and lymph node metastasis (presence vs. absence) (the latter only in the total cohort, not in the analysis with LN-negative patients)

^bAge (continuously), sex (male vs. female), tumor location (right colon vs. left colon vs. rectum), tumor size (continuously), tumor morphology (pedunculated vs. non-pedunculated), invasion depth (deep vs. superficial submucosal invasion), lymphovascular invasion (presence vs. absence), differentiation grade (poor vs. moderate vs. good), and lymph node metastasis (presence vs. absence) (the latter only in the total cohort, not in the analysis with LN-negative patients)

^cThe maximum weight of a single patient used in the IPW adjusted analysis to obtain balance in potential confounders. This is a quality instrument to assess whether a single or a few cases influence the risk estimate excessively. As a rule of thumb this should be lower than 10% of the analyzed dataset (i.e., smaller than 100 and 90 for the total cohort and LN-negative patients, respectively)

^dThis is an estimate of the balance of confounders after adjusting by inverse probability weighting (0.50 complete balance, 1.00 complete unbalance)

^ePercentage of complete cases (i.e., cases with no imputed values for any of the evaluated variables of that analysis). Note: analysis was performed on the imputed dataset concerning all cases

^fPercentage of available data points before imputation. Note: analysis was performed on the imputed dataset concerning all cases