MRI-identified multidimensional nodal features predict survival and concurrent chemotherapy benefit for stage II nasopharyngeal carcinoma
Article Category: Research Article
Pubblicato online: 13 dic 2022
Pagine: 479 - 487
Ricevuto: 10 set 2022
Accettato: 11 ott 2022
DOI: https://doi.org/10.2478/raon-2022-0047
© 2022 Yang Liu, Jianghu Zhang, Jingbo Wang, Runye Wu, Xiaodong Huang, Kai Wang, Yuan Qu, Xuesong Chen, Yexiong Li, Ye Zhang, Junlin Yi, published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.
Nasopharyngeal carcinoma (NPC) is a disease in head and neck that is especially prevalent in Southeast Asia, with an age-standardized incidence rate of five per 100,000 population.1 Given its complex anatomical location and high radiosensitivity, radiotherapy (RT) has been the mainstay of treatment. The predominant failure pattern of distant metastases highlights the importance of chemotherapy to improve survival outcomes in NPC patients.2, 3 Currently, the National Comprehensive Cancer Network (NCCN) guidelines recommend that concurrent chemoradiotherapy (CCRT) be the primary treatment for T1–2N1M0 patients and T2N0M0 patients with high-risk features.4 This is based on a prospective phase III study that showed that CCRT provided survival benefits over two-dimensional (2D) conventional RT alone for stage II NPC.5
With rapid technological advances, intensity-modulated radiotherapy (IMRT) has yielded excellent treatment outcomes for patients with stage II NPC, with 5-year overall survival (OS) and local control rates exceeding 90%.6 A recent phase II randomized clinical trial demonstrated that CCRT might be unnecessary for stage II patients treated with IMRT, as it failed to improve treatment efficacy while increasing the incidence of toxicity.7 A similar recommendation can be found in the latest European Society for Medical Oncology (ESMO) guidelines.8 Nonetheless, the survival outcomes remain poor for certain populations screened using adverse prognostic determinants (
American Joint Committee on Cancer (AJCC) N staging is consistently recognized as one of the most important determinants for prognosis prediction and treatment decision-making in stage II NPC. However, N1 disease represents a heterogeneous setting in which the size of the lymph node ranges from 1 to a 6-cm mass. Furthermore, the unresolved issue is the accuracy and consistency of N staging, which merely considers nodal size, laterality and location. With advances in imaging modalities, MRI imaging-based analyses have extracted diverse nodal features containing essential information closely related to NPC staging, treatment, and prognosis.12-16 As morphological lymph node features, high-level extranodal extension and central nodal necrosis have consistently been proven to be predictors of distant metastases.17-19 Furthermore, the burden-related lymph node features, such as volume and number of positive lymph nodes (PLNs), and number of positive lymph node regions, are proved prognostic factors for NPC progression.20-23 However, to the best of our knowledge, no comprehensive lymph node feature-based prognosis stratification and concurrent chemotherapy decision-making protocol has been reported for patients with stage II NPC.
In this study, we explored the prognostic value of multidimensional lymph node features in stage II NPC patients and established a lymph node feature-based nomogram. Furthermore, we examined the clinical validity of this model in guiding CCRT utility for patients with stage II NPC.
This study was approved by the Ethics Committee of our institute, and the requirement for informed consent was waived due to its observational nature. The present study was conducted in accordance with the guidelines from the Declaration of Helsinki.
Between January 2007 and December 2017, 242 stage II NPC patients were consecutively included (restaged according to the 8th AJCC staging system,). The eligibility criteria were as follows (Supplementary Figure 1): (1) histologically confirmed World Health Organization (WHO) type II–III NPC; (2) no evidence of distant metastases; (3) complete pretreatment evaluation and having complete baseline MRI scans of the nasopharynx and neck; and (4) treatment with IMRT or CCRT (cisplatin, cumulative dose ≥ 200 mg/m2).
Head and neck T1- and T2-weighted MRI images (T1WI, T2WI) were obtained with a 1.5 or 3T scanner (GE Healthcare, Discovery MR, United States) axially, coronally, and sagittally. The axial slice thickness was 3 mm from the suprasellar cistern to the inferior margin of the sternal end of the clavicle.
The gross volumes of the primary tumor and nodes were identified as GTVnx and GTVnd, respectively. The diagnostic criteria of positive lymph nodes included: (1) any cervical lymph node with the shortest diameter > 10 mm, any retropharyngeal lymph node (RPLN) in the lateral group with the minimum diameter > 5 mm, and any visible RPLN in the median group; (2) central nodal necrosis or a contrast-enhanced rim; (3) extranodal extension; and (4) ≥ 3 contiguous converging lymph nodes, each with an minimum diameter > 8 mm.24 When more than two lymph nodes could not be distinguished from each other, they were counted as one. The maximal axial diameter of lymph nodes was measured on the largest plane of cross-sectional images. Extranodal extension was assessed on T1-weighted fat-suppressed contrast images and classified as follows (Figure 1): Grade 0, without extranodal extension; Grade 1, infiltrating the surrounding fat; Grade 2, matted nodes; and Grade 3, infiltrating the adjacent muscle, parotid gland, vessels and skin.13 Central nodal necrosis was identified as a focal area of high signal intensity on T2WI.19
Figure 1
Axial T1-weighted fat-suppressed contrast MRI images of grade 0–2 extranodal extension (ENE) in four stage II nasopharyngeal carcinoma patients.
All 242 patients completed whole-course simultaneous-integrated boost (SIB) IMRT. Doses of 69.96 Gray (Gy)/33 fractions were prescribed for GTVnx and GTVnd. High- and low-risk regions of the clinical target volume (CTV) received a prophylactic dose of 60.06 Gy/33 fractions and 50.96 Gy/28 fractions, respectively. The concurrent chemotherapy plan included a weekly (40 mg/m2) or three-weekly (80–100 mg/m2) cisplatin regimen.
The primary endpoint was OS (date of treatment start, to date of documented death from any cause or last follow-up). The secondary endpoints were disease-specific survival (DSS, date of treatment start to date of documented death from NPC-related cause or last follow-up), regional recurrence-free survival (RRFS, date of treatment start to date of the first observation of regional recurrence) and distant metastasis-free survival (DMFS, date of treatment start to date of the first observation of distant metastases).
Patients were followed up at least every three months for the first two years, every six months for the third to fifth years, and then yearly. Evaluations included complete physical examination, fiberoptic nasopharyngoscopy, MRI of the nasopharynx and neck, chest X-ray/CT, and abdomen sonography/CT with or without bone scans. Other additional investigations (
The Kaplan–Meier method was performed to estimate survival rates, and log-rank test was used to examine the significance of differences. Youden indexes derived from the time-dependent receiver operating characteristic (ROC) analysis were utilized to dichotomize continuous variables into categorical variables and determine the optimal cutoffs. Cox proportional hazards regression with backward selection was performed in multivariable analysis (MVA) and to calculate hazard ratios (HRs). To assess the effect of the continuous variables on OS, an additive Cox model was used to generate pointwise estimates of HR curves by using the “smoothHR” package.25 Based on the risk factors identified by MVA, a nomogram was developed. The calibration curves, Harrell concordance index (C-index), area under the curve (AUC) and decision curve analysis (DCA) were used to evaluate the performance of nomogram. Each patient was stratified into low- or high-risk groups based on the final sum of the nomogram scores. R (version 4.1.0) and SPSS (version 26.0) were adopted for the statistical analysis. All the tests were two-sided, and significance was defined as
The general characteristics of the 242 stage II patients are summarized in Table 1 (median age 50-years old, range 18–76; 173 men, male-to-female ratio 2.5:1). Overall, 158 (65.3%) patients received IMRT, and 84 (34.7%) received CCRT.
Characteristics of 242 patients with stage II nasopharyngeal carcinoma
| Characteristics | No. (%) |
|---|---|
| Age, median (range) | 50 (18–76) |
| < 50 | 120 (49.6) |
| ≥ 50 | 122 (50.4) |
| Sex | |
| Male | 173 (71.5) |
| Female | 69 (28.5) |
| AJCC 8th T stage | |
| T1 | 104 (43.0) |
| T2 | 138 (57.0) |
| AJCC 8th N stage | |
| N0 | 30 (12.4) |
| N1 | 212 (87.6) |
| AJCC 8th subgroup | |
| T1N1M0 | 104 (43.0) |
| T2N0M0 | 31 (12.8) |
| T2N1M0 | 107 (44.2) |
| GTVnx volume (cm3), median (range) | 19.1 (2.1–74.0) |
| < 13.7 | 70 (28.9) |
| ≥ 13.7 | 172 (71.1) |
| GTVnd volume (cm3), median (range) | 8.7 (0–71.0) |
| < 29.1 | 212 (87.6) |
| ≥ 29.1 | 30 (12.4) |
| Lateral of RPLNs | |
| None | 90 (37.2) |
| Unilateral | 113 (46.7) |
| Bilateral | 39 (16.1) |
| LN located in level III | |
| Yes | 195 (80.6) |
| No | 47 (19.4) |
| LN size (cm) | 1.7 (0–6.0) |
| MAD < 2.2 | 152 (62.8) |
| MAD ≥ 2.2 | 90 (37.2) |
| Number of positive LN | 2 (0– 9) |
| 0 | 30 (12.4) |
| 1 | 69 (28.5) |
| 2 | 60 (24.8) |
| 3 | 51 (21.1) |
| 4 | 19 (7.9) |
| ≥ 5 | 13 (5.4) |
| †ENE grade | |
| Grade 0 | 153 (63.2) |
| Grade 1,2 | 52 (21.5) |
| Grade 3 | 37 (15.3) |
| CNN | |
| No | 179 (74.0) |
| Yes | 63 (26.0) |
| Treatment | |
| IMRT | 158 (65.3) |
| CCRT | 84 (34.7) |
† ENE Grade 0 = none; Grade 1 = LN Infiltrating surrounding fat; Grade 2 = matted nodes; and Grade 3 = LN infiltrating adjacent muscle and (or) vessels and(or) skin
AJCC = American Joint Committee on Cancer staging system; CCRT = concurrent chemoradiotherapy; CNN = central nodal necrosis; ENE = extranodal extension; GTVnx/GTVnd = gross target volume of nasopharyngeal tumor/lymph nodes; IMRT = intensity-modulated radiotherapy; LN = lymph node; MAD = maximum diameter; RPLN = retropharyngeal lymph node
With a median follow-up of 10.3 years (range: 0.6– 17.5 years), the 5-year and 10-year OS, DSS, RRFS and DMFS rates of the entire cohort were 91.9% and 79.1%, 93.6% and 86.7%, 96.5% and 93.6%, and 92.3% and 89.1%, respectively. At the last follow-up, 42 (17.4%) patients had died, and 45 (18.6%) experienced treatment failure. The patterns of treatment failure and causes of death are shown in Supplementary Table 1.
Compared with IMRT, patients receiving CCRT had significantly higher rates of Grade 1–2 gastrointestinal reactions, ear toxicities and severe (Grade 3–4) leukopenia. No significant differences were observed in terms of the incidence or severity of late toxicities between the two groups (Supplementary Table 2).
The ROC-determined optimal cutoffs for GTVnx volume, GTVnd volume and lymph node size to predict OS were 13.7 cm3, 29.1 cm3and 2.2 cm, respectively. Smooth HR trend showed that the log HR of overall death increased linearly with an increasing number of PLNs from 1 to 4 and reached a plateau from 5 and above (Supplementary Figure 2). Therefore, the number of PLNs was categorized into 6 groups (
Figure 2
Multidimensional nodal features of stage II nasopharyngeal carcinoma significantly associated with survival outcomes in univariate or multivariate analysis. Axial T2-weighted fat-suppressed image of
MAD = maximal axial diameter
Univariate analysis showed that age ≥ 50 years, GTVnd ≥ 29.1 cm3, lymph nodes located in level III (Figure 2A), lymph node size ≥ 2.2 cm (Figure 2B), PLN number (Figure 2C), and positive central nodal necrosis (Figure 2D) were associated with poor OS. MVA demonstrated that age, location and number of PLNs were independent determinants for OS (Supplementary Table 3). T stage was also included in the nomogram for its importance in clinical decision-making (Figure 3A). The calibration curves manifested excellent consistency in terms of the 5- and 10-year OS (Figure 3B–C).
Figure 3
Nomogram and calibration plots of 5- and 10-year overall survival (OS). Number of positive lymph nodes was as continuous variable.
The C-index of the nomogram in predicting OS was 0.726 (95% CI: 0.638–0.813) in comparison with 0.537 (95% CI: 0.498–0.608) for TNM stage (
Figure 4
Comparison of
The linear correlation between the nomogram score and the hazard of overall death is shown in Supplementary Figure 3. Based on cutoffs for the total score, patients were stratified into low-risk (score: 0–134) and high-risk groups (score: >134). There were no significant differences in any endpoint among the T1N1M0, T2N0M0, and T2N1M0 subgroups (Figure 5A–D). The high-risk group had significantly worse OS, DSS, RRFS or DMFS rates than low-risk group (Figure 5E–H).
Figure 5
Survival curves of the
Based on the stratified groups derived from the nomogram, we further analyzed the clinical efficacy of CCRT in two cohorts with different prognoses. Although a trend that CCRT improves OS was observed in the whole population, it did not reach a significant difference (Figure 6A). Nevertheless, high-risk patients could benefit from CCRT compared to IMRT (5-year OS 89.9%
Figure 6
Impact of concurrent
Figure 7
The 5-year and 10-year overall survival (OS) rates and hazard ratio (HR) between intensity-modulated radiotherapy (IMRT) group and concurrent chemoradiotherapy (CCRT) group according to
In this study, we comprehensively evaluated the prognostic value of multidimensional nodal features in patients with stage II NPC and further established a lymph node feature-based nomogram. The nomogram demonstrated superior predictive performance for OS and outperformed the current AJCC TNM staging. In addition, the nomogram-derived prognostic stratification contributes to identifying high-risk patients who are more likely to benefit from CCRT. To the best of our knowledge, this current study is the first to incorporate various nodal features into a nomogram to predict the prognosis of stage II NPC and to further guide the management of concurrent chemotherapy.
The latest NCCN guidelines recommend CCRT for stage II NPC, which is based on a phase III study from the conventional 2D-RT era that showed significant improvements in 5-year OS and progression-free survival (PFS) with CCRT over RT alone.4 The survival benefit of CCRT might be due to its radiosensitizing effect and the fact that it compensated for the dosimetric deficiencies of the conventional RT technique. In the IMRT era, the role of CCRT in stage II NPC has not been absolutely defined given the paucity of data from phase III trials. The majority of studies revealed that the addition of concurrent chemotherapy did not significantly improve survival but increased the prevalence of acute toxic reactions in patients with stage II NPC.7, 26 Nonetheless, Luo
Lymph node status is one of the most important determinants associated with prognosis and treatment decisions.15, 16 The determination of AJCC N1 classification depends primarily on the location and size of the ipsilateral cervical lymph nodes. Nonetheless, in recent years, alternative nodal parameters, such as the number of PLNs, extranodal extension and central nodal necrosis status, have been increasingly reported to better assess the profile of nodal burden.12-16 The number of MRI-positive lymph nodes has been reported to be a predominant independent prognostic factor for survival in NPC patients and is a better proxy for the cumulative effect of lymph nodes than AJCC N staging.22 Additionally, quantitative metastatic lymph node regions were shown to be superior to N classification in terms of prognosis in NPC.23 In addition, high-grade extranodal extension was reported as an evaluable predictor that could assist in the selection of stage II patients with a high risk of distant metastases.13,17,18 Another critical lymph node feature was central nodal necrosis, which was proven to be an independent negative prognostic factor in patients with NPC.12, 16 In the current study, by applying an additive Cox model based on P-penalized, we observed the impact of consecutive PLN on survival: the hazard of overall death increased linearly with increasing PLNs from 1 to 4, eventually reaching a plateau at PLNs of 5 and above. Based on this tendency, PLN surpassed the traditional AJCC N stage and emerged as an independent variable in the MVA model. Likewise, lymph node located in level III was found to be an important prognostic factor that outperformed N stage, which could be explained by the fact that lymph node often metastasizes in an orderly fashion from upper to lower level lymph nodes, and patients with level III lymph node involvement tend to have an increased number of positive lymph nodes. Lymph node size, GTVnd volume, extranodal extension and central nodal necrosis were no longer independent prognostic factors for OS, probably because they were surrogates for the number of PLNs or LNs located in level III and were included when all served as positive lymph node criteria in multivariable Cox analysis.
As per the results of the multivariate analysis, we finally included age, T stage, number of PLNs and location (lymph node located in level III) in the nomogram for OS. Although nomograms have been established in various manners in stage II NPC20-23, the notable hallmark of our nomogram is that it is the first to incorporate lymph node status, while in the previous nomogram, the effect of lymph node burden was merely assessed by N staging and lacked more detailed categorization.19 In terms of the discriminative performance of this nomogram, its C-index, AUC and DCA are chemoradiotherapy (CCRT) within
significantly better than AJCC TNM staging and are comparable to previous nomograms for stage II NPC.20-23
Despite the growing evidence supporting the prognostic value of lymph node features in NPC, very little stratification based on lymph node features has been incorporated into treatment decisions from a clinical management perspective. To our knowledge, only one previously published study of nomograms in NPC analyzed the ability of the lymph node features-based nomogram model to guide personalized induction chemotherapy combined with CCRT management in stage II–IVA patients.16 In the current study, we implemented the time-dependent ROC method to identify two subgroups according to the nomogram scores, which was more reasonable and discriminatory than arbitrary groupings based on medians or quartiles. We then found that CCRT was associated with improved survival outcomes in the high-risk group, with reductions in the overall death hazards of approximately 70%. These results underline that this nomogram may be valid enough to be used as a practical tool for clinicians to select stage II patients for concurrent chemotherapy.
This study has several limitations. First, the plasma EBV-DNA levels were not included in the survival analysis. Only five patients in this cohort had EBV-DNA over 500 copies/ml, which is consistent with previous studies reporting relatively low EBV-DNA copies in stage II patients29, 30, suggesting that their prognostic value may not be as critical as in stage III/IV patients. Another possible reason for the relatively low EBV-DNA in this cohort is the continued lack of standardized EBV-DNA testing31, 32; therefore, we sought to re-examine the prognostic value of pretreatment EBV-DNA for stage II NPC after standardization of testing methods. Second, although the sample size of the present study was the largest reported for nodal feature prognostication in stage II NPC to date, our data were obtained from a single cancer center. A larger sample size cohort for external independent validation is warranted to further assess the prognostic and predictive value of this lymph node feature-based nomogram.
In summary, this study established a nodal features-based nomogram with excellent performance, which provided statistically significantly superior discrimination to TNM staging in predicting the OS in patients with stage II NPC. Furthermore, the nomogram-based stratification effectively identified patients who were more likely to gain survival benefit from CCRT. It offers a useful tool for providing patient counseling and clinical assessments. To generalize the clinical utility of this nomogram, external validation should be considered.