From April 2007 to December 2017, 1,714 eligible patients were included in this study. The median age was 64 years old (range: 60-85 years), with a sex (M/F) ratio of 3.16:1.00. Non-keratinizing undifferentiated NPC (WHO type III) accounted for the majority of cases (97.8%). 38.6% of patients had pre-treatment plasma EBV DNA levels greater than 4,000 copies /mL. The proportion of patients classified as stage I, II, III and IV was 4.5%, 14.1%, 46.9% and 34.5%, respectively. There were 835 patients with an ACE-27 score of 0, 741 patients with a score of 1 and 138 patients a score of 2-3. Most patients underwent IC+CCRT (26.8%) or CCRT (36.9%), 10.0% underwent IC+RT, and 26.3% were treated with RT alone. The baseline characteristics are shown in Table 1.
The median follow-up time was 59.3 months (range, 0.39-170.09 months), 118 (6.9%) patients developed locoregional recurrence, 189 (11.0%) had distant metastasis and 33 (1.9%) had both. 463 (27.0%) patients died, the causes of death were tumor recurrence in 287 (16.7%) patients, complications after radiotherapy in 22 (1.3%) patients, non-NPC-related causes in 149 (8.7%) patients, and unknown cause in 5 (0.3%) patients. The non-NPC-related causes included 19 (1.1%) patients who died from other cancers, 57 (3.3%) patients from other chronic diseases, 6 (0.4%) patients from accidents, and 67 (3.9%) patients from natural death. The 5-year OS, DSS, DMFS and LRRFS for the whole cohort were 77.9%, 83.6%, 86.7% and 91.3%, respectively.
RPA-generated risk stratification
The RT alone dataset (n = 451) was randomly divided into a training set (60%) and a testing set (40%) for the explorative construction and validation of the RPA models for risk stratification. Table 2 shows the detailed clinicopathological characteristics for the RT alone training set among elderly NPC patients. The ROC curve analysis for pre-treatment plasma EBV DNA indicated that the TNM staging for elderly NPC patients could be redefined with a cut-off value of 4,000 copies/mL (Supplementary Figure S1). After adjustment in the multivariate analysis, plasma EBV DNA, LDH, and T stage were verified to have significant effects on OS (P <0.001, P = 0.001, P = 0.007, respectively) (Table 2). Risk stratification was performed using all validated predictors in the RPA. Automaticrpart algorithms were run to realize the modification of branches. Plasma EBV DNA and T stage remained in the final model while non-essential factors were removed.
310 elderly NPC patients treated with RT alone were classified into low-risk (n = 131; pre-treatment plasma EBV DNA ≤ 4,000 copies/mL & T1-2), intermediate-risk (n = 108; pre-treatment plasma EBV DNA ≤ 4,000 copies/mL & T3-4) and high-risk groups (n = 71; pre-treatment plasma EBV DNA > 4,000 copies/mL & any T) (Figure 1A). Survival curves showed significant discrimination in OS among the groups. The corresponding 5-year OS rates were 88.39%, 75.17% and 57.81%, respectively (P < 0.001; Supplementary Figure S2). As such, the low-risk group was considered the good-prognosis group, while the intermediate-risk and high-risk group were considered the poor-prognosis group. OS in the good-prognosis group was significantly higher than in the poor-prognosis group in the training set (HR = 0.309, 95% CI = 0.184-0.517; P < 0.001; Figure 1B). A similar situation was observed in the testing set (HR = 0.276, 95% CI = 0.113-0.670; P = 0.002; Figure 1C).
Selection of chemotherapy and chemotherapy regimen beneficiaries
In the whole elderly NPC patient dataset, chemotherapy was administered to 73.7% (1,263/1,714) of the patients. In the good-prognosis group, 180 patients received radiation therapy alone and 136 patients received chemoradiotherapy (CRT). The 5-year OS, DSS, DMFS and LRRFS of the CRT group vs. RT alone group were 88.1% vs. 88.3% (P = 0.570), 91.2% vs. 92.0% (P = 0.380), 90.9% vs. 95.3% (P = 0.047) and 96.0% vs. 94.9% (P = 0.580) (survival curves are shown in Supplementary Figure S3).
In the poor-prognosis group, chemotherapy was administered to 80.6% (1,127 /1,398) of the patients. The basic information is shown in supplementary Table E1. Patients who received RCT in the poor-prognosis group had a significantly improved OS than those who received RT alone, (HR = 0.70, 95% CI = 0.55-0.88; P = 0.003) and the high-risk subgroup (HR = 0.58, 95% CI = 0.43-0.78; P < 0.001) but not in the intermediate-risk subgroup (HR = 0.80, 95% CI = 0.54-1.17; P = 0.252). In the high-risk subgroup, CRT was validated to have significant survival benefits compared with RT alone for OS (P < 0.001), DSS (P = 0.012) and DMFS (P = 0.019), but not for LRRFS (P = 0.74) (see Supplementary Figure S4).
For chemotherapy regimens, 38.1% (429 /1,127) of patients received IC+CCRT, 48.1% (542 /1,127) received CCRT and 13.8% (156 /1,127) received IC+RT in the poor-prognosis group. Patients who received IC+CCRT and CCRT had significantly improved OS than those who received RT alone (IC+CCRT vs. RT alone: HR = 0.64, 95% CI = 0.48-0.85; P = 0.002; CCRT vs. RT alone: HR = 0.70, 95% CI = 0.54-0.91; P = 0.008) but not in the IC+RT group (IC+RT vs. RT alone: HR = 0.83, 95% CI = 0.59-1.17; P = 0.306) (survival curves are shown in Figure 2A). Discrimination of survival benefits were not evident compared with RT alone for DSS (IC+CCRT vs. RT alone: P = 0.148; CCRT vs. RT alone: P = 0.184; IC+RT vs. RT alone: P = 0.574), DMFS (IC+CCRT vs. RT alone: P = 0.389; CCRT vs. RT alone: P = 0.355; IC+RT vs. RT alone: P = 0.706) or LRRFS (IC+CCRT vs. RT alone: P = 0.910; CCRT vs. RT alone: P = 0.736; IC+RT vs. RT alone: P = 0.515) (survival curves are shown in Figure 2B-D). More detailed information of patient chemotherapy regimens in the poor-prognosis group are listed in Supplementary Table E2.
Subgroup analysis in the poor-prognosis group
To clarify whether other clinical factors impacted the prognosis of elderly NPC patients in the poor-prognosis group, univariate analysis was performed with OS, DSS, DMFS and LRRFS. Only the ACE-27 score was significant in predicting OS, DSS, DMFS and LRRFS (all P < 0.05). However, age and smoking status were significant in predicting OS (age ≥70 years vs. 60-64 years, P > 0.001; age 65-69 years vs. 60-64 years, P = 0.001; smoking vs. non-smoking: P = 0.022), as shown in Table 3.
In the poor-prognosis group, there were 591 patients with an ACE-27 score of 0, 482 patients with a score of 1 and 54 patients with a score of 2-3 in the CRT group and 117 patients with a score of 0, 101 patients with a score of 1 and 53 patients a score of 2-3 score in the RT alone group, respectively. The 5-year OS for CRT vs. RT-alone with ACE-27 scores of 0, 1 and 2 were 76.0% vs. 70.0% (P = 0.014), 80.5% vs. 68.2% (P = 0.150) and 58.5% vs. 62.2% (P = 0.490), respectively (survival curves are shown in Figure 3). In the intermediate-risk group, the 5-year OS for CRT vs. RT-alone with ACE-27 scores of 0, 1 and 2 were 82.9% vs. 74.7% (P = 0.058), 86.2% vs. 81.7% (P = 0.580) and 66.1% vs. 73.1% (P = 0.240), respectively (survival curves are shown in Supplementary Figure S5). In the high-risk group, the 5-year OS for CRT vs. RT-alone with ACE-27 scores of 0, 1 and 2 were 70.0% vs. 59.6% (P = 0.014), 73.3% vs. 54.2% (P = 0.037) and 48.5% vs. 47.7% (P = 0.660), respectively (survival curves are shown in Supplementary Figure S6).
In the poor-prognosis group, there were 648 patients aged 60-64 years old, 358 patients were 65-70 years old and 121 patients aged ≥70 years old in the CRT group and 58 patients aged 60-64 years old, 96 patients aged 65-70 years old and 117 patients aged ≥70 years old in the RT alone group, respectively. The 5-year OS for CRT vs. RT-alone with 60-64, 65-70 and ≥70 years old were 80.9% vs. 75.9% (P = 0.068), 73.3% vs. 63.4% (P = 0.270) and 64.8% vs. 67.1% (P = 0.820), respectively (survival curves are shown in Figure 4). In the intermediate-risk group, the 5-year OS for CRT vs. the RT-alone with 60-64, 65-70 and ≥70 years old were 85.9% vs. 89.7% (P = 0.840), 79.7% vs. 75.1% (P = 0.630) and 81.5% vs. 70.5% (P = 0.088), respectively (survival curves are shown in Supplementary Figure S7). In the high-risk group, the 5-year OS for CRT vs. RT-alone with 60-64, 65-70 and ≥70 years olds were 75.9% vs. 47.4% (P༜0.001), 66.1% vs. 51.5% (P = 0.130) and 46.6% vs. 62.2% (P = 0.120), respectively (survival curves are shown in Supplementary Figure S8).
In the poor-prognosis group, there were 487 smokers and 640 non-smokers in the CRT group and 86 smokers and 185 non-smokers in the RT alone group, respectively. The 5-year OS for CRT vs. RT-alone with smokers and non-smokers were 73.2% vs. 67.0% (P = 0.091) and 80.1% vs. 68.4% (P = 0.004), respectively (survival curves are shown in Supplementary Figure S9). In the intermediate-risk group, the 5-year OS for CRT vs. RT-alone with smoking and non-smoking were 79.1% vs. 77.9% (P = 0.840) and 87.2% vs. 76.4% (P = 0.099), respectively (survival curves are shown in supplementary Figure S10.). In the high-risk group, the 5-year OS for CRT vs. RT-alone with smokers and non-smokers were 67.0% vs. 50.5% (P = 0.009) and 72.8% vs. 57.5% (P = 0.004), respectively (survival curves are shown in Supplementary Figure S11.)