Patient characteristics
A total of 992 cases of RLS (524 males and 398 females) were enrolled in our study. We first detected the prognostic factors for all the participants through univariate Cox regression analysis (Table 1). Our data indicated that age, marital status, histological type, AJCC stage, N stage, M stage, surgery, lymph node dissection, radiation, chemotherapy, and distant organ metastasis (liver, lung, bone, and brain) were prognostic factors for RLS (P<0.05). A multivariate analysis was then performed (Table 2). The results showed that age, histological type, AJCC stage, surgery, and chemotherapy were prognostic factors for RLS (P<0.05). Therefore, as an independent prognostic factor for RLS, age deserves further exploration.
The age of 70 years was also identified through X-tile software as the optimal cut-off value (Figure 1), which divided the patients with RLS into two age groups. A total of 619 patients with RLS diagnosed older than 70 years old were identified as the elderly group, and 303 patients were classified in the young group (<70 years old). Then, the characteristics of the elderly and young groups were compared, and the data showed that patients with RLS in the elderly group were significantly different from those in the young group in many aspects, such as marital status, surgery, radiation, and chemotherapy (P<0.05, Table 3). Of the 992 patients, mortality occurred in 268 patients (27.0%) at the follow-up deadline. By comparison, the elderly patients with RLS had a higher overall mortality rate (41.58%) compared to that in the young patients with RLS (22.94%) at the end of follow-up (P<0.001).
Prognostic factors of the elderly patients with RLS
Through univariate Cox regression, our results revealed that age, histological type, AJCC stage, N stage, M stage, surgery, radiation, chemotherapy, liver metastasis, and lung metastasis showed significant effects on overall survival of the patients above 70 years old (P<0.05, Table 4). On the other hand, unlike the elderly group, the prognostic factors for the young group were sex, histological type, AJCC stage, N stage, M stage, distant metastasis (liver, lung, and bone metastasis), surgery, lymph node dissection, and chemotherapy (P<0.05, Table 4). These results further revealed the difference between the two age groups.
A subsequent multivariate Cox regression analysis was performed based on the univariate Cox regression analysis, and four independent prognostic predictors (age, histological type, AJCC stage, and surgery) for the elderly group were identified. The data demonstrated that age was an unfavorable prognostic factor (HR=1.05, 95% CI=1.01-1.09). There were significant differences in prognosis between different histological types. Compared to the dedifferentiated type, the mixed type has a worse prognosis. Moreover, higher grades of AJCC stage also indicated worse overall survival. Compared to stage I, higher grades, such as stage II (HR=2.08, 95% CI=1.02-4.22), stage III (2.32, 1.31-4.10) or stage IV (3.44, 1.64-7.21), were associated with worse OS. In regard to treatment, surgery also showed a significantly positive impact on overall survival (0.31, 0.18-0.54). All the details are shown in Table 5. Such results are also different from the young patients, whose independent prognostic factors did not include age (Table 6).
The correlations of the three categorical variables among four independent predictors for the elderly group (histological type, AJCC stage, and surgery) with overall survival obtained by the log-rank test were also demonstrated through the survival curve (Figure 2). Then, a nomogram of the elderly population was constructed based on the four independent prognostic factors.
Nomogram Construction
According to the multivariate models, nomograms that combined all five independent prognostic factors were developed to predict overall survival at 1, 3 and 5 years for the elderly group (Figure 3).
Validation and Calibration of the Nomogram
The C-indexes of the 1-, 3- and 5-year survival nomograms were 0.737 (95% CI 0.692-0.782), 0.737 (0.692-0.782) and 0.7367 (0.692-0.782) for OS in the elderly group, respectively. The discrimination ability of the nomogram for OS was evaluated by comparing the AJCC7 stage. The corresponding C-index values of the AJCC7 stage were 0.631 (95% CI 0.576-0.685), 0.6306 (0.576-0.685) and 0.6306 (0.576-0.685). Therefore, the C-index of the nomogram was significantly higher than that of the AJCC7 stage (P<0.001), especially at 3 and 5 years, reflecting the better overall discrimination ability of the nomogram.
Then, the discrimination ability of the nomogram at 1, 3, and 5 years was examined, and the corresponding AUCs were 0.749, 0.804 and 0.810, respectively. However, the AUCs were 0.679, 0.710 and 0.646 for the AJCC7 stage (Figure 4), which were all lower than those of the nomogram (P<0.001). Therefore, the newly constructed model had a better discrimination ability than that of the AJCC7 stage.
We further verified the accuracy of the nomogram using NRI and IDI in elderly individuals. The AJCC7 stage was set as the old model. The NRIs for OS at 1, 3, and 5 years were 0.371 (95% CI 0.169-0.498), 0.364 (0.235-0.519), and 0.553 (0.167-0.846), respectively. The IDIs for OS at 1, 3, and 5 years were 0.072, 0.138, and 0.211 (all P<0.001), respectively. Hence, the nomogram had better accuracy than the AJCC7 staging system for predicting OS at 1, 3, and 5 years.
Calibration curves demonstrated acceptable agreement between the predicted death rate and observed values of overall survival (Figure 5). Thus, we believe the nomogram model had a high accuracy for the prediction of OS at 1, 3 and 5 years for elderly patients with RLS.