Table 3. Multivariate analyses of predicting pathological lymph nodes metastases: the final predictor for developing the nomogram.
| Variable | Logistic Regression | Nomogram | |||
|---|---|---|---|---|---|
| Training Group (N = 288) | |||||
| OR | 95% CI | p-value | AUCs | 95% CI | |
| MRI cT stage | |||||
| cT3-4 vs cT1-2 | 4.91 | [2.26,10.68] | 5.87 * 10−5 | ||
| MRI CRM involvement | |||||
| Involved vs clear | 3.07 | [1.28, 7.40] | 0.01 | ||
| Preoperative CEA | Training: 0.78 | [0.732, 0.837] | |||
| ≥ 5 ng/ul vs < 5 ng/ul | 1.62 | [0.88, 2.95] | 0.12 | Validation:0.71 | [0.619,0.801] |
| Tumor grade | |||||
| High grade vs low-medium grade | 4.91 | [0.89, 4.42] | 0.10 | ||
| LN size category | |||||
| largest LN size < 5 mm vs no LN detected | 2.05 | [0.95, 4.40] | 0.07 | ||
| largest LN size 5–10 mm vs no LN detected | 3.20 | [1.60, 6.42] | 1.04 * 10−3 | ||
| largest LN size > 10 mm vs no LN detected | 48.09 | [5.58, 414.34] | 4.24 * 10−4 | ||
Abbreviations: MRI, magnetic resonance image; CRM, circumferential resection margin; OR, odds ratio; CI, confidence interval; AUC, area under ROC curve; CEA, carcinoembryonic antigen; LN, lymph node.