The general characteristics of the study patients (
n = 75) are summarized in Table
1. The median follow-up time in the entire cohort was 59.5 months (range: 2.6–140.9 months) and 76.0 months (range: 31.4–140.9 months) in surviving patients. There was no difference in the follow-up time for training and validation cohorts (median: 55.17 months, range: 2.6–140.9 months vs. 74.6 months, range 7.4–127.9 months,
P = 0.304). The most common primary tumor site was the oral cavity (
n = 22, 29.3%), followed by the oropharynx (
n = 21, 28.0%), nasal cavity/paranasal sinus (
n = 17, 22.7%), nasopharynx (
n = 8, 10.7%), hypopharynx (
n = 3, 4.0%), larynx (
n = 3, 4.0%), and ear (
n = 1, 1.3%). The histological diagnoses were as follows: ACC (
n = 35, 46.7%), MEC (
n = 25, 33.3%), adenocarcinoma (
n = 14, 18.7%), and acinic cell carcinoma (
n = 1, 1.3%). Sixty patients (80.0%) were considered to have tumors with high-risk histology according to the WHO classification. Surgery was the primary treatment modality in 45 study participants (60.0%). The median RT doses delivered to patients who received surgical and non-surgical treatment were 66 Gy (
n = 38, range: 54 − 76 Gy, 97.4% of cases ≥ 60 Gy) and 72 Gy (
n = 30, range: 66 − 80 Gy, all cases ≥ 60 Gy), respectively. The hospital identification number did not show any significant association with either T-, N-, and AJCC stages or survival endpoints (OS and RFS). In this extended study, we enrolled MSGC including all kinds of histology and more favorable overall survival (OS) was noted as compared with our previous paper that included only high-risk histology cases (5-year OS: 68.8% vs. 60.4%). Twenty-seven patients (36.0%) died during the study period, and the causes of death were as follows: MSGC,
n = 22; severe infection,
n = 1; hypopharyngeal cancer,
n = 1; pancreatic cancer,
n = 1; cerebrovascular disease,
n = 1; and traffic accident,
n = 1. Disease recurrences were observed in 31 patients (41.3%), with 13, 11, and 7 cases showing locoregional, distant, and concomitant locoregional plus distant recurrences, respectively. No difference in cancer death was noted in the training and validation groups (15 (33.3%) vs. 7 (23.3%) events, Fisher’s exact test,
P = 0.441). Nineteen (42.2%) and eight (26.7%) patients were dead in the training and validation cohorts, respectively (Fisher’s exact test,
P = 0.222), with eighteen (41.3%) and thirteen (40.0%) recurrent events noted in the corresponding groups (
P = 0.814). The training and validation groups did not differ in terms of clinical parameters and there were no differences in propensity scores (median: 0.32 versus 0.29, respectively,
P = 0.540).
Table 1
General characteristics of the study patients
Age: median [range], years | 52 [20-81] | 53 [22-78] | 51 [20-81] | 0.240 |
Sex | Female | 40 (53.3) | 25 (55.6) | 15 (50.0) | 0.646 |
Male | 35 (46.7) | 20 (44.4) | 15 (50.0) | |
Smoking | Yes | 21 (28.0) | 12 (26.7) | 9 (30.0) | 0.797 |
No | 54 (72.0) | 33 (73.3) | 21 (70.0) | |
Performance | ECOG 0 − 1 | 72 (96.0) | 43 (95.6) | 29 (96.7) | 1.000 |
ECOG 2 | 3 (4.0) | 2 (4.4) | 1 (3.3) | |
Treatment | Surgery | 45 (60.0) | 30 (66.7) | 15 (50.0) | 0.160 |
Non-surgery | 30 (40.0) | 15 (33.3) | 15 (50.0) | |
WHO histology | High-risk | 60 (80.0) | 38 (84.4) | 22 (73.3) | 0.255 |
Low-risk | 15 (20.0) | 7 (15.6) | 8 (26.7) | |
T-stage | T1 − T2 | 31 (41.3) | 16 (35.6) | 15 (50.0) | 0.239 |
T3 − T4 | 44 (58.7) | 29 (64.4) | 15 (50.0) | |
N-stage | N0 − N2b | 71 (94.7) | 42 (93.3) | 29 (96.7) | 0.646 |
N2c − N3 | 4 (5.3) | 3 (6.7) | 1 (3.7) | |
AJCC stage | I − II | 23 (30.7) | 12 (26.7) | 11 (36.7) | 0.445 |
III − IV | 52 (69.3) | 33 (73.3) | 19 (63.3) | |
Dead events | | 27 (36.0) | 19 (42.2) | 8 (26.7) | 0.222 |
Relapse events | | 31 (41.3) | 18 (40.0) | 13 (43.3) | 0.814 |
The median time to disease progression after treatment was 15.1 months (range: 2.4–69.1 months). Kaplan-Meier analysis revealed that patients with advanced T-stages, AJCC stages, or who were smokers had worse OS and RFS. Surgery and WHO high-risk histology were found to have an adverse impact on RFS, but not on OS. Patients with ACC tended to have a better OS (
P = 0.077) but a similar RFS (
P = 0.957) compared with other histology types. Among patients who were treated with surgery, positive margins, perineural invasion, and lymphatic invasion were identified in 24 (53.3%), 17 (37.8%), and 1 (2.2%) cases, respectively. The median nomogram score for OS [
13] was 132 (range: 0-254). The median nomogram score for RFS [
14] was 2.76 (range: 1.0-4.0). Notably, patients with advanced N-stage (N2c/N3,
n = 2/2) or poor PS (ECOG 2,
n = 3) had 5-year OS and RFS rates of 0%.