Prognostic factor analysis in patients with early-stage nasopharyngeal carcinoma in the USA

Nasopharyngeal carcinoma (NPC) is a malignant tumor primarily derived from the nasopharynx that is typically found in residents living in southern China, some Asian regions and North Africa [1]. According to a published report released by GLOBOCAN, the incidence of this cancer subtype is approximately 0.04–0.25% [2]. NPC is a highly fatal disorder that not only affects patient quality of life, but also induces a heavy burden on the medical system because the survival of patients with NPC 5 years after being diagnosed by detectable plasma Epstein–Barr virus DNA is approximately 49% [3]. NPC patients experience a worse survival outcome when the malignant tumor initiates remote metastases to other areas such as bone and lung [4]. Although diverse categories of treatments including chemotherapy, radiotherapy, immunotherapy and personalized treatments based on specific biomarkers have significantly increased NPC-related survival [5–8], some issues in clinical practice still remain unresolved. These include side effects of chemoradiotherapy such as nephrotoxicity and neurotoxicity, as well as uncertain outcomes and the high expense of the treatment methods mentioned above [9,10].

Similar to other types of cancer, NPC can be categorized into different stages according to the features of the neoplasm, the number of nodes and whether it metastasizes to regions that are adjacent to or distant from the primary site. Stage I and II carcinomas, which constitute the early stages of NPC, produce no remote metastases, unlike the advanced stages, according to the American Joint Committee on Cancer (AJCC) stage definitions [11]. Previous studies have shown that patients with early-stage NPC have a higher 5-year overall survival (OS) than those with an advanced stage of NPC [12]. Meanwhile, outcomes of children and adolescents with NPC are superior to those of adults in the advanced stage, as confirmed by diverse literature [13–15].

Even with the apparently high survival rates of early-stage NPC patients, few studies have emphasized the prognosis of these patients, despite the potentially lethal nature of the disease. Unfortunately, many patients die soon after the initial diagnosis of NPC [16], and a large number of patients with NPC acquire recurrent metastasis within a short time period [17]. To address the currently intractable issues existing in the NPC treatment process, we focused on the risk factors of patients with early-stage NPC and aimed to elucidate the prediction efficiency of our model. Consequently, the findings of our study may provide a better guide for clinical applications.

Patients & methods

Patients

The current study included 566 patients from 2004 to 2019 who were diagnosed with NPC at an early stage (I and II) and were registered in the Surveillance, Epidemiology and End Results (SEER) database, a publicly available source of data from the National Cancer Institute in the USA. The SEER database includes survival, clinicopathological and demographic data covering approximately one-third of the US population [18]. All participants were confirmed by positive histology (as opposed to clinical diagnosis only), direct visualization without microscopic confirmation, or positive microscopic confirmation; in some cases the method was not specified. Histology records of broad groupings were selected to classify the histological types, including codes 8010–8049 (epithelial neoplasms, not otherwise specified), 8050–8089 (squamous cell neoplasms), 8120–8139 (transitional cell papillomas and carcinomas), 8140–8389 (adenomas and adenocarcinomas), 8430–8439 (mucoepidermoid neoplasms), 8440–8499 (cystic, mucinous and serous neoplasms), 8500–8549 (ductal and lobular neoplasms) and 8560–8579 (complex epithelial neoplasms). A tumor (T) or node (N) stage was selected according to the Derived AJCC Stage Group, sixth edition (2004–2015). Regional lymph node (LN) surgery was performed using the field RX Summ–Scope Reg LN Sur (2003+). Early-stage NPC included stage I (T1N0) and stage II (T2a–2bN0, T1–2bN1) [1,19]. Early-stage NPC patients are those with the carcinoma confined to the nasopharynx (T1) or with tumor extending to the oropharynx/nasal cavity without parapharyngeal extension (T2a) or any tumor with parapharyngeal extension (T2b). Meanwhile, the lymph nodes of patients display no regional metastasis (N0) or have unilateral metastasis in LN(s), ≤6 cm in greatest dimension, above the supraclavicular fossa (N1). The inclusion criteria were: the tumor site was limited to the nasopharynx (site recode ICD-0-3/WHO 2008) and the tumor was in an early stage (AJCC Stage group 6th ed). The exclusion criteria comprised the following: unknown tumor stage, missing survival months, or unknown follow-up data.

Construction of the nomogram

Independent factors regarding prognosis of 3- and 5-year cancer-specific survival (CSS) of NPC patients at an early stage were evaluated using Cox regression models. After determination by univariate or multivariate analyses, variables with a p-value < 0.05 were included. The nomogram of early-stage NPC patients based on the above results was constructed using R software (http://www.r-project.org/).

Nomogram validation & calibration

The evaluation of the precision of the model created by the methods described above was determined by assessing the area under the curve, concordance index and receiver operating characteristic curves [20]. After correcting for bias, the observed and predicted survival rates were compared using a calibration curve.

Statistical methods

To assess 3- and 5-year NPC-related survival, a Kaplan–Meier curve was established. All data were analyzed using SPSS v. 25.0 (IBM Corp., NY, USA). Statistical significance was set at p < 0.05.

Results

Patient characteristics

The characteristics of patients included in this study are shown in Table 1. The mean age of patients included in this research was 56.48 ± 13.45 years. Male patients with stage I NPC accounted for 50% of the entire cohort. This trend was also observed in the stage II group, in which the number of male patients accounted for 20.3% of the total cohort. The largest proportion of all patients were White, accounting for 33.4% (stage I) and 15.4% (stage II). The most common subtype of early-stage NPC histology was squamous carcinoma, which accounted for 63.2% of the tumors (stage I: 45.4%; stage II: 17.8%). In comparison with unknown primary site data (68.7%), patients with only one primary site and multiple primary sites accounted for 27.4 and 3.9% of the participants, respectively. The grade III (41.9%) and IV stages (41.0%) were more common than grade I (4.0%) and II stages (13.0%). All T2 stage patients were in the stage I group (40.1%). This trend was also applicable to N1 stage patients, who accounted for 52.5% of the entire cohort. A larger proportion of carcinomas (74.7%) were derived from another laterality. The proportion of tumors smaller than 3 cm was approximately 33.9%, that of tumors larger than 3 cm was 28.6%, and that of tumors of unknown size was approximately 37.5%. Surgery was performed to dissect the lymph nodes in 88.9% of the total patients. The 3- and 5-year CSS rates were 0.852 and 0.804, respectively. The 3- and 5-year OS rates were 0.805 and 0.737, respectively.

Survival circumstances & prognostic graph

The OS of the 566 patients was calculated from 176 patients who died due to reasons other than NPC. A total of 121 patients who died of early-stage NPC were selected to determine the CSS (Table 2). Univariate analysis showed that age (> 60 years) was an independent factor for OS in patients with early-stage NPC (p < 0.001). When compared with patients aged <50 years, the relationship of survival to the above age range was also applicable to CSS (p < 0.001). Sex acted as an independent factor for OS (p = 0.008) but not CSS (p = 0.097). Asian and Pacific Islander patients showed a lower hazard ratio (HR; p < 0.001). In addition to the aforementioned variables, race, tumor grade and tumor size were independent factors for both OS and CSS. The HR for grade IV patients was 0.415 (p = 0.011) for OS and 0.342 (p = 0.006) for CSS. Tumor size, when available, showed higher survival in both the OS (p = 0.003) and CSS (p = 0.004) groups for those with tumors ≤3 cm. The histological type, primary site, tumor, node stage and regional LN surgery did not act as independent factors for OS and CSS.

All of the above factors that were significant for CSS were included in the multivariable analysis. Significant differences were found for age, race and tumor size, which were applicable for both OS and CSS. The HR for OS for patients aged 60–69 years was 1.938 (p = 0.006), for age 70–79 years was 2.435 (p = 0.001) and for age >80 years was 6.380 (p < 0.001). For CSS, only the HR values of patients aged 70–79 and >80 years showed significance. When compared with White people, the HR values of Asian and Pacific Islanders in both OS and CSS groups were lower. A tumor size ≤3 cm resulted in an HR of 0.527 (p = 0.001) for OS and 0.465 (p = 0.001) for CSS. In contrast, we found no significant difference related to sex for CSS, although a significant difference was found in OS for females (HR: 0.659; p = 0.024). Tumor grade did not show a significant difference for either OS (p = 0.390) or CSS (p = 0.261; Table 3).

Establishment of the nomogram & certification of the prognostic model

Regarding the independent factors for early-stage NPC, we included age, race and tumor size to construct a nomogram, as shown in Figure 1. As depicted in Figure 2, the prediction of 3- and 5-year CSS by the area under the curve was 0.753 (Figure 2A) and 0.715 (Figure 2B), respectively. A good correlation between the nomogram predictions and the actual observations for 3- and 5-year CSS was shown by the calibration curves (Figure 2C & D).

Discussion

NPC is a troublesome disease with a potentially high death rate. Proper management by ear, nose and throat doctors is urgently needed for patients. Previous research has focused on clarifying the risk factors for metastasis in NPC [21]. However, little emphasis has been placed on independent risk factors for patients with early-stage NPC. In this study, we found that age, race and tumor size were significant factors for predicting the survival of patients with stage I and II NPC.

Specifically, we found that age was an independent factor for patients with early-stage NPC. This finding is in accordance with previous research indicating that an increase in the age of patients with NPC caused HR values calculated by multivariate Cox analysis to increase significantly [22,23]. The above phenomenon was mostly attributed to the unsatisfactory outcome of therapeutic methods due to intolerance of frequent and high-intensity radiochemotherapy [23]. In addition, the high occurrence of cardiovascular/cerebrovascular disease and lung and colorectal cancer among older patients (compared with younger patients) may result in a higher death rate [24]. Prior research has indicated that Black people (who were categorized into another race subgroup in our research) were more likely to die from NPC than White people [25]. The higher death risk for Black patients with early-stage NPC may be due to unavailable or inadequate healthcare resulting in delayed seeking of medical intervention, influencing tumor staging at presentation [26]. Tumor size had a significant effect on death rate. Patients with tumors <3 cm had a lower risk of death than those with tumors >3 cm. This result is consistent with a previous study demonstrating that primary tumor size was an indicator predicting the CSS of NPC patients and affecting the practice of clinical treatments [27].

As shown in Table 3, sex did not predict CSS, although it was a significant predictor of OS in patients with early-stage NPC. Males had a higher risk of both early-stage NPC and NPC with bone metastases [28]. In addition, the predictive ability of tumor grade was lost for CSS and OS when it was included in multivariate Cox analysis. This is opposed to previous studies that showed no significant difference in both univariate and multivariate analyses [29]. With respect to race and NPC, we found that Asian and Pacific Islanders had a lower risk of death than White patients. This result is at odds with previous NPC research on race [30]. One possible reason for better outcomes in Asians would be the histological type of NPC: more of these patients have undifferentiated carcinoma, which has a better prognosis than keratinized squamous carcinomas [31]. This discrepancy may also exist because our research mainly focused on carcinoma derived from the nasopharynx rather than tumors derived from the entire head and neck (termed ‘oropharyngeal cancer’). This subtype of malignant tumor was found to be associated with a lower mortality in Asians than in Pacific Islanders, influencing the CSS, in different studies [32,33].

Our study had some limitations that should not be overlooked. First, the retrospective study design based on the SEER database incurred a possibility of selection bias. Second, the current research did not consider the impact of specific chemotherapy regimens, cycles of chemotherapy and radiotherapy dose, which are essential factors influencing outcomes and survival in patients with early-stage NPC. Additionally, a considerable proportion of patients with head and neck cancer have reduced performance status [34–38]; our study could not account for the effect of performance status which is a major prognostic factor associated with patient age. Third, some other prognostic factors, such as living habits (smoking and alcohol) and Epstein–Barr virus DNA biomarkers, were not available in the SEER database. Fourth, the regional limitations should not be discounted because the incidence of NPC is regionally unbalanced. Therefore, the results from US patients might not be successfully extrapolated globally. Fifth, the tumor size is an important factor for the diagnosis, prognosis and treatment of NPC. However, due to the limited patient numbers included in this study, enough subjects with data of tumor size from SEER*Stat software were not acquired. This should be further investigated in the future study.

Conclusion

The current study proposed some potential predictors for patients with early-stage NPC and constructed a nomogram to predict the 3- and 5-year CSS. These findings may provide guidance for ear, nose and throat clinicians to deliver better medication and treatments for patients with NPC.