A total of 1035 patients with NM-GBA were included and randomized into a training (n=727) and validation (n=308) cohort. The cumulative survival in the entire cohort was 73.5% at 1 year, 44.2% at 3 years and 33.9% at 5 years. The baseline characteristics of the patients included in the training and validation sets were balanced (Table 1).
The C-index, LR, AIC and BIC were used to compare the predictive abilities of stage N, LNR and LODDS. The LODDS system showed the highest C index and LR test as well as the lowest AIC and BIC, indicating that the superior performance of LODDS (C index: 0.648, LR test: 133.4, AIC: 5863.572, BIC: 5872.009) on LNR status (C-index: 0.637, LR test: 128, AIC: 5868.957, BIC: 5877.395) and N (C-index: 0.622, LR test : 104.1, AIC: 5892.882, BIC: 5901.309) in terms of prediction of OS for NM-GBA. To determine the optimal model, we performed stepwise forward regression with the LODDS and ten other significant modules. Finally, age, sex, chemotherapy, stage, grade, LODDS and height were included. Univariate Cox regression analysis revealed that 9 variables (age, gender, stage, level, LODDS, height, race, T, and marital status) were significantly associated with OS in the training cohort. In multivariate Cox regression analysis, age, gender, chemotherapy, stage, grade, LODDS, and height remained independent prognostic factors (Table 2).
Construction and validation of the nomogram
The new prognostic model for NM-GBA based on these variables was constructed. A nomogram displaying the predictor variables and the corresponding point scales has been presented in Fig. 2. The nomogram estimated a patient’s probability of survival based on a total score calculated by adding zero to 100 points for each individual predictor. Most of the patients in the current study had total risk points ranging from 137 to 384.
The C index of the nomogram in the training and validation cohorts was 0.730 (0.708–0.752) and 0.746 (0.715–0.777), respectively. The time-dependent AUC was >0.7 for the prediction of OS at 1, 3 and 5 years in the training and validation cohorts (training cohort: AUC over 1 year = 0.802 (0.766-0.838), 3 year AUC = 0.803 (0.770–0.835), 5 year AUC = 0.794 (0.756–0.832), validation cohort: 1 year AUC = 0.784 (0.749–0.819), 3 year AUC = 0.784 (0.751–0.817 ), 5-year AUC = 0.786 (0.751–0.821)), indicating favorable discrimination of the predictive model. The calibration plots demonstrated good agreement between the events predicted by the nomogram and observed (Fig. 3a–f). The high calibration and discrimination performance of the nomogram was confirmed in the validation cohort.
Clinical Value of the Nomogram in Comparison to the AJCC 8th Edition TNM Staging System
To assess the accuracy of the change in risk classification, we calculated the NRI, IDI, C-index, and AUC from the AJCC 8th edition nomogram and TNM classification system. In the nomogram test in the training cohort, the NRI for OS at 1, 3 and 5 years was 0.648 (95% CI = 0.532-0.862), 0.625 (95% CI = 0.480-0.757) and 0.589 (95% CI = 0.480-0.757). % CI=0.434–0.739), IDI for OS at 1, 3 and 5 years was 0.073 (95% CI=0.058–0.088, P
These results indicated that the prognostic performance of the newly constructed model was superior to that of the traditional AJCC TNM staging system. DCA showed that the nomogram could better predict OS at 1 year, 3 years and 5 years in patients with NM-GBA. Compared to the AJCC 8th Edition system, the nomogram added more net benefits to almost all threshold probabilities in the training and validation cohorts (Fig. 3g–l).
Clinical risk stratification of patients with NM-GBA based on nomogram score
We finally stratified the risk of patients in the training and validation cohorts based on the total score calculated by the nomogram. Patients can be divided into four groups: nomo 1 (total score