Skip to main content
. 2020 Nov 19;3(11):e2025881. doi: 10.1001/jamanetworkopen.2020.25881

Table 2. Model Accuracy and Survival Predictions for Treatment According to Model Recommendations.

Model C index OS, median (IQR), months HR (95% CI)a P value HR, IPTW (95% CI)a P value
Patients receiving recommended treatment Patients not receiving recommended treatment
RTOG 95-01, ≥2 lymph nodes NA 111.8 (102.1-118.9) 98.1 (91.8-108.9) 0.96 (0.89-1.05) .38 0.89 (0.81-0.98) .02
EORTC 22931, T3-4 disease, except T3N0 larynx N2-3, LVI, deep nodes with oral/oropharynx cancer NA 111.3 (105.4-117.6) 95.3 (86.9-108.2) 0.93 (0.86-1.01) .07 0.90 (0.82-0.99) .03
DeepSurv 0.693 (0.675-0.711) 118.1 (111.5-126.5) 90.6 (79.8-98.1) 0.79 (0.72-0.85) <.001 0.76 (0.69-0.84) <.001
N-MTLR 0.691 (0.673-0.709) 116.4 (109.7-123.3) 93.5 (85.9-101.1) 0.83 (0.77-0.90) <.001 0.80 (0.72-0.88) <.001
RSF 0.695 (0.676-0.713) 111.4 (101.1-120.3) 99.5 (91.7-110.1) 0.90 (0.83-0.98) .01 0.96 (0.87-1.06) .41

Abbreviations: EORTC, European Organisation for Research and Treatment of Cancer; HR, hazard ratio; IPTW, inverse probability of treatment weighting; IQR, interquartile range; LVI, lymphovascular invasion; N-MTLR, neural multitask logistic regression; OS, overall survival; RSF, random survival forest; RTOG, Radiation Therapy Oncology Group.

a

HRs are given for the patients who received a recommended treatment, compared with those who did not. Results are compared with decision rules derived from the RTOG and EORTC trials, wherein patients were recommended to receive chemoradiation if they met any of the intermediate risk criteria specified by the trials and radiation alone if not.