Copyright: ©Author(s) 2026.
World J Gastrointest Oncol. Jun 15, 2026; 18(6): 117851
Published online Jun 15, 2026. doi: 10.4251/wjgo.v18.i6.117851
Published online Jun 15, 2026. doi: 10.4251/wjgo.v18.i6.117851
Table 1 Clinical characteristic of esophageal patients, median (interquartile range)/mean (SD)
| Variables | Whole cohort (n = 206) | Training cohort (n = 144) | Validation cohort (n = 62) | P value |
| Age | 64.1 (7.5) | 64.5 (7.5) | 63.1 (7.4) | 0.218 |
| Gender | 0.570 | |||
| Female | 45 (21.8%) | 33 (22.9%) | 12 (19.4%) | |
| Male | 161 (78.2%) | 111 (77.1%) | 50 (80.6%) | |
| Body mass index | 23.1 (2.65) | 23.1 (2.7) | 23.1 (2.5) | 0.965 |
| Smoking and drinking | 0.267 | |||
| None | 86 (41.7%) | 60 (41.7%) | 26 (41.9%) | |
| Smoking alone | 40 (19.4%) | 29 (20.1%) | 11 (17.7%) | |
| Drinking alone | 13 (6.3%) | 6 (4.2%) | 7 (11.3%) | |
| Both | 67 (32.5%) | 49 (34%) | 18 (29%) | |
| Squamous cell carcinoma level | 1 (0.6, 1.4) | 1 (0.6, 1.4) | 1.1 (0.8, 1.3) | 0.520 |
| Cytokeratin 19 fragment antigen 21-1 level | 1.4 (0.9, 1.8) | 1.4 (0.9, 1.9) | 1.3 (0.8, 1.8) | 0.290 |
| Carcinoembryonic antigen level | 3.4 (2.6, 4.3) | 3.4 (2.6, 4.4) | 3.5 (2.6, 4.2) | 0.924 |
| Carbohydrate antigen 199 | 22.3 (14.9, 29.7) | 22.5 (14.8, 30) | 22.2 (16.8, 27.2) | 0.828 |
| Tumor location | 0.480 | |||
| Upper | 63 (30.6%) | 44 (30.6%) | 19 (30.6%) | |
| Middle | 130 (63.1%) | 89 (61.8%) | 41 (66.1%) | |
| Lower | 13 (6.3%) | 11 (7.6%) | 2 (3.2%) | |
| Tumor length | 3 (2, 4.5) | 3.4 (2, 4.5) | 3 (2, 4) | 0.082 |
| Computed tomography-reported nodal status | 0.407 | |||
| Negative | 148 (71.8%) | 101 (70.1%) | 47 (75.8%) | |
| Positive | 58 (28.2%) | 43 (29.9%) | 15 (24.2%) | |
| Tumor grade | 0.057 | |||
| G1 | 16 (7.8%) | 7 (4.9%) | 9 (14.5%) | |
| G2 | 150 (72.8%) | 109 (75.7%) | 41 (66.1%) | |
| G3 | 40 (19.4%) | 28 (19.4%) | 12 (19.4%) | |
| Depth of tumor invasion | 0.912 | |||
| cT1 | 47 (22.8%) | 33 (22.9%) | 14 (22.6%) | |
| cT2 | 58 (28.2%) | 40 (27.8%) | 18 (29%) | |
| cT3 | 91 (44.2%) | 63 (43.8%) | 28 (45.2%) | |
| cT4 | 10 (4.9%) | 8 (5.6%) | 2 (3.2%) | |
| Node stage | 0.506 | |||
| pN0 | 127 (61.7%) | 89 (61.8%) | 38 (61.3%) | |
| pN1 | 46 (22.3%) | 35 (24.3%) | 11 (17.7%) | |
| pN2 | 19 (9.2%) | 12 (8.3%) | 7 (11.3%) | |
| pN3 | 14 (6.8%) | 8 (5.6%) | 6 (9.7%) | |
| Lymph node metastasis status | 0.944 | |||
| Negative | 127 (61.7%) | 89 (61.8%) | 38 (61.3%) | |
| Positive | 79 (38.3%) | 55 (38.2%) | 24 (38.7%) |
Table 2 Performances of computed tomography to evaluate lymph node metastasis status in patients with esophageal cancer
| Statistic | Value (95%CI) | ||
| Whole cohort (n = 206) | Training set (n = 144) | Validation set (n = 62) | |
| Overall lymph node status evaluation | |||
| Sensitivity | 42.11% (30.86%-53.98%) | 42.59% (29.23%-56.79%) | 40.91% (20.71%-63.65%) |
| Specificity | 80% (72.08%-86.50%) | 77.78% (67.79%-85.87%) | 85% (70.16%-94.29%) |
| PPV | 55.17% (44.38%-65.50%) | 53.49% (41.21%-65.36%) | 60% (38.06%-78.55%) |
| NPV | 70.27% (65.70%-74.47%) | 69.31% (63.64%-74.45%) | 72.34% (64.34%-79.13%) |
| Accuracy | 66.02% (59.11%-72.46%) | 64.58% (56.19%-72.37%) | 69.35% (56.35%-80.44%) |
| Thoracic lymph node status evaluation | |||
| Sensitivity | 47.92% (33.29%-62.81%) | 51.35% (34.40%-68.08%) | 36.36% (10.93%-69.21%) |
| Specificity | 86.08% (79.68%-91.06%) | 85.05% (76.86%-91.20%) | 88.24% (76.13%-95.56%) |
| PPV | 51.11% (39.11%-62.99%) | 54.29% (40.66%-67.30%) | 40% (18.39%-66.35%) |
| NPV | 84.47% (80.46%-87.79%) | 83.49% (78.25%-87.66%) | 86.54% (80.26%-91.04%) |
| Accuracy | 77.18% (70.84%-82.73%) | 76.39% (68.60%-83.06%) | 79.03% (66.82%-88.34%) |
| Abdominal lymph node status evaluation | |||
| Sensitivity | 28.89% (16.37%-44.31%) | 26.67% (12.28%-45.89%) | 33.33% (11.82%-61.62%) |
| Specificity | 95.03% (90.44%-97.83%) | 93.86% (87.76%-97.50%) | 97.87% (88.71%-99.95%) |
| PPV | 61.9% (41.80%-78.62%) | 53.33% (31.05%-74.36%) | 83.33% (38.76%-97.53%) |
| NPV | 82.7% (79.82%-85.25%) | 82.95% (79.59%-85.85%) | 82.14% (76.24%-86.83%) |
| Accuracy | 80.58% (74.51%-85.75%) | 79.86% (72.37%-86.08%) | 82.26% (70.47%-90.80%) |
Table 3 Comparison of two machine-learning models with computed tomography in evaluating lymph node metastasis in patients with esophageal cancer
| Statistic | Training set | Validation set | ||
| Value | 95%CI | Value | 95%CI | |
| Computed tomography predicted | ||||
| Sensitivity | 42.6% | 29.25%-56.8% | 40.9% | 20.7%-63.7% |
| Specificity | 77.8% | 67.8%-85.8% | 85.0% | 70.2%-94.3% |
| PPV | 53.5% | 41.2%-65.4% | 60.0% | 38.1%-78.6% |
| NPV | 69.3% | 63.6%-74.4% | 72.3% | 64.3%-79.1% |
| Accuracy | 64.6% | 56.2%-72.4% | 69.4% | 56.4%-80.4% |
| Model-1 predicted | ||||
| Sensitivity | 85.45% | 73.34%-93.50% | 66.67% | 44.68%-84.37% |
| Specificity | 88.76% | 80.31%-94.48% | 92.11% | 78.62%-98.34% |
| PPV | 82.46% | 72.18%-89.49% | 84.21% | 63.45%-94.25% |
| NPV | 90.80% | 83.82%-94.95% | 81.40% | 71.15%-88.59% |
| Accuracy | 87.50% | 80.97%-92.42% | 82.26% | 70.47%-90.80% |
| Model-2 predicted | ||||
| Sensitivity | 89.09% | 77.75%-95.89% | 75.00% | 53.29%-90.23% |
| Specificity | 85.39% | 76.32%-91.99% | 89.47% | 75.20%-97.06% |
| PPV | 79.03% | 69.34%-86.27% | 81.82% | 63.38%-92.12% |
| NPV | 92.68% | 85.56%-96.44% | 85.00% | 73.75%-91.95% |
| Accuracy | 86.81% | 80.16%-91.87% | 83.87% | 72.33%-91.98% |
- Citation: Gu RT, Li X, Cheng W, Wang XW, Jin H, Liu T. Machine-learning models integrating preoperative clinical factors and circulating tumor DNA features predict lymph node metastasis in esophageal carcinoma. World J Gastrointest Oncol 2026; 18(6): 117851
- URL: https://www.wjgnet.com/1948-5204/full/v18/i6/117851.htm
- DOI: https://dx.doi.org/10.4251/wjgo.v18.i6.117851