Background
According to Global Cancer Statistics 2020, there were >130,000 new cases of nasopharyngeal carcinoma (NPC) worldwide.1 Among these cases, approximately 4% to 10% are classified as de novo metastatic NPC (dmNPC).2–4 Unlike early and locoregionally advanced NPC, patients with dmNPC have limited treatment options and poorer survival outcomes. Recent trials have shown that adding PD-1 inhibitors to standard chemotherapy as a first-line treatment significantly improves progression-free survival (PFS) in patients with recurrent or metastatic NPC (RM-NPC).5–7 As a result, chemotherapy combined with immunotherapy (CT-IO) is recommended as the first-line treatment for RM-NPC by both NCCN and ESMO-EURACAN guidelines.8,9 However, therapeutic options and strategies for patients with dmNPC, a treatment-naïve condition, still significantly differ from those for posttreatment RM-NPC, particularly in managing local and regional lesions.10–12
Previous studies indicate that combining locoregional radiotherapy (LRRT) with palliative chemotherapy significantly enhances prognosis for patients with dmNPC compared with chemotherapy alone.11–15 Patients who underwent LRRT after chemotherapy exhibited a 2-year overall survival (OS) rate of 76.4%, whereas those receiving chemotherapy alone had an OS rate of 54.5%.15 However, these studies did not include immunotherapy, and only a few studies with limited sample sizes have assessed LRRT efficacy in patients with dmNPC receiving first-line CT-IO.16–18 Clinical consensus on this matter has not yet been reached. Additionally, the clinical features and treatment sensitivity of dmNPC vary, leading to different survival outcomes for patients receiving LRRT after systemic therapy.10,19,20 Given the complexity and heterogeneity of dmNPC, it remains unclear whether LRRT provides additional survival benefits after CT-IO and which patients would benefit most from additional LRRT in the era of immunotherapy.
This study used inverse probability of treatment weighting (IPTW) to investigate the efficacy of LRRT combined with CT-IO as first-line treatment for dmNPC. Additionally, we aimed to develop and validate a risk assessment tool to identify suitable patients who may benefit from additional LRRT following CT-IO for dmNPC.
Methods
Patient Selection
This study was approved by the Research Ethics Committee of Sun Yat-sen University Cancer Center (SYSUCC), with written informed consent waived (approval number: B2024-431-01). From May 2017 to June 2023, patients were included based on the following criteria: (1) previously untreated NPC with confirmed metastatic disease at initial diagnosis; (2) no history of other malignant tumors or synchronous malignancies; (3) received at least 2 cycles of platinum-based chemotherapy combined with anti–PD-1 immunotherapy as first-line treatment, regardless of LRRT; (4) presence of measurable lesions; and (5) availability of complete baseline and posttreatment clinical data for efficacy assessments. The authenticity of this study was verified by depositing the primary raw data on the Research Data Deposit public platform (www.researchdata.org.cn).
Patients underwent routine pretreatment evaluations, including detailed medical history, physical examinations, routine blood tests, serum biochemistry, plasma Epstein-Barr virus (EBV) DNA testing, and nasopharynx pathology. Imaging studies included abdominal ultrasonography, chest radiograph or CT, whole-body bone scan or PET/CT, and MRI of the nasopharynx and neck. Staging was based on the eighth edition of the AJCC/Union for International Cancer Control criteria. Experienced radiologists determined the number of metastatic sites and lesions based on imaging findings.
Treatments
All patients received first-line CT-IO. The palliative chemotherapy (PCT) regimens included GP (gemcitabine/platinum), TP (docetaxel/platinum), PF (platinum/5-fluorouracil), TPF (docetaxel/platinum/5-fluorouracil), and TPC (docetaxel/platinum/capecitabine). Because capecitabine is converted to fluorouracil in vivo, the TPF and TPC regimens were combined into the TPF regimen group. Anti–PD-1 antibodies used in the study included toripalimab, camrelizumab, sintilimab, tislelizumab, pembrolizumab, and nivolumab. CT-IO was administered every 3 weeks per cycle. A total of 367 patients received intensity-modulated radiotherapy (IMRT)–based LRRT following CT-IO. The radiation therapy techniques used have been previously described.21 A comprehensive summary of the treatment regimens is provided in Supplementary Table S1 (available online in the supplementary materials).
Follow-Up and Outcome
Radiologic evaluations (CT, MRI, or PET/CT) were selectively performed every 2 to 4 cycles during CT-IO and every 3 to 6 months thereafter to assess tumor response. Tumor responses were independently evaluated by 2 investigators using RECIST version 1.1.22 Plasma EBV DNA levels were monitored after every 1 to 2 treatment cycles. The primary endpoint of the study was PFS, defined as the time from treatment initiation to disease progression or death from any cause, whichever occurred first. After completing treatment, patients were followed up at least once every 6 months. Patients lost to follow-up at the time of their last contact were censored in the analysis. Adverse events (AEs) were graded at each follow-up visit according to CTCAE version 5.0 and the RTOG Late Radiation Morbidity Scoring Scheme.
Statistical Analysis
Comorbidity was defined as the presence of an additional concurrent medical condition, such as cerebrovascular disease, cardiovascular disease, chronic lung disease, chronic liver disease, chronic kidney disease, or diabetes. Plasma EBV DNA levels were categorized into variables based on previous studies.23,24 Patients were further stratified using C-reactive protein (CRP) levels ≥3 mg/L and lactate dehydrogenase (LDH) levels >250 U/L, as described in prior research.25,26
The median follow-up time was 28.9 months (95% CI, 27.3–30.5), calculated using the reverse Kaplan-Meier method. Statistical comparisons between groups were performed using the chi-square or Fisher exact test. To mitigate selection bias, inverse probability of treatment weighting (IPTW),27 a propensity score method, was used to balance differential distributions of factors between the LRRT and non-LRRT groups. A standardized mean difference (SMD) of ≤0.1 was considered ideal, whereas an SMD of ≤0.2 was deemed acceptable.
Kaplan-Meier curves were generated to compare PFS across treatment groups, both with and without IPTW adjustment. Given that LRRT is administered after completion of CT-IO, immortal time bias was a potential concern. To address this, we conducted a 6-month landmark analysis for PFS estimates.28 The Cox proportional hazards regression model was used to determine hazard ratios (HRs) and 95% confidence intervals.
To quantitatively predict prognosis, we developed a prognostic model. A total of 500 patients were randomly divided into a training cohort (n=350) and a validation cohort (n=150) at a 7:3 ratio. In the training cohort, univariable and multivariable Cox proportional hazards regression models identified prognostic factors for PFS. Variables with P<.1 in univariable analyses were included in the multivariable analyses. Model selection was guided by the Akaike Information Criterion (AIC) using a bidirectional stepwise multiple regression approach, iteratively adding and removing variables to identify the most significant predictors. Variables with P<.05 in the multivariable analyses were used to create a prognostic model, visually represented as a nomogram. The model’s performance was assessed using the C-index, time-dependent receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). Bootstrapping validation (1,000 bootstrap resamples) was performed to calculate a bias-corrected C-index. Each patient’s prognostic score was computed based on the developed model. The performance of the nomogram was assessed in the validation cohort, where Cox regression was conducted using the prognostic score as a covariate. Subsequently, the C-index, calibration plots, and calibration curves were generated from the regression analysis results.
Identifying patient subgroups with different treatment responses is crucial for tailoring individualized treatment strategies. To achieve this, we used a model-based trees approach29 to assess the personalized treatment impact based on nomogram scores, facilitating the identification of optimal candidates for LRRT. A 6-month conditional landmark analysis with IPTW adjustment was then performed, and Kaplan-Meier log-rank pairwise tests compared PFS among different risk groups with or without LRRT. All statistical analyses were performed using SPSS Statistics, version 26.0 (SPSS Inc.) and R version 4.3.3 (R Foundation for Statistical Computing), with visualizations created in GraphPad Prism version 8.3.0 (GraphPad Software, LLC). Two-tailed P values <.05 were considered statistically significant.
Results
Patient Characteristics and Outcomes
A total of 500 patients with dmNPC were included in the study between May 2017 and June 2023, with 367 receiving LRRT after first-line CT-IO and 133 not receiving LRRT. Median age was 46 years (IQR, 37–55), and 81.8% (n=409) were male. Most patients were diagnosed with T3–4 (n=457; 91.4%) and/or N2–3 (n=438; 87.6%) disease. Bone metastases was the most common metastatic site (n=354; 70.8%), followed by liver (n=177; 35.4%) and lung (n=121; 24.2%). Patient characteristics for the overall cohort are detailed in Supplementary Table S2. Median PFS (mPFS) was 29.4 months (95% CI, 26.3–45.2), OS was not reached (NR), and the overall response rate (ORR) was 89.4%. Baseline characteristics, categorized by treatment approach (with or without LRRT) before and after IPTW adjustment, are summarized in Table 1. Following IPTW adjustment, all variables achieved a SMD of <0.2, indicating balanced cohorts.
Patient Characteristics
| Variable | Overall Cohort | IPTW-Adjusted Cohort | ||||||
|---|---|---|---|---|---|---|---|---|
| LRRT Group (n=367) n (%) |
Non-LRRT Group (n=133) n (%) |
P Value | SMD | LRRT Group n |
Non-LRRT Group n |
P Value | SMD | |
| Age | .207 | 0.139 | .879 | 0.019 | ||||
| <45 y | 177 (48.2) | 55 (41.4) | 46.1 | 45.2 | ||||
| ≥45 y | 190 (51.8) | 78 (58.6) | 53.9 | 54.8 | ||||
| Sex | .260 | 0.125 | .909 | 0.014 | ||||
| Male | 305 (83.1) | 104 (78.2) | 82.4 | 81.9 | ||||
| Female | 62 (16.9) | 29 (21.8) | 17.6 | 18.1 | ||||
| Smoking | .353 | 0.104 | .746 | 0.04 | ||||
| No | 223 (60.8) | 74 (55.6) | 59.0 | 57.0 | ||||
| Yes | 144 (39.2) | 59 (44.4) | 41.0 | 43.0 | ||||
| Drinking | .260 | 0.125 | .482 | 0.087 | ||||
| No | 305 (83.1) | 104 (78.2) | 81.3 | 77.7 | ||||
| Yes | 62 (16.9) | 29 (21.8) | 18.7 | 22.3 | ||||
| Comorbidities | .887 | 0.027 | .888 | 0.017 | ||||
| No | 291 (79.3) | 104 (78.2) | 79.3 | 80.0 | ||||
| Yes | 76 (20.7) | 29 (21.8) | 20.7 | 20.0 | ||||
| Karnofsky performance score | .269 | 0.125 | .575 | 0.056 | ||||
| ≤80 | 28 (7.6) | 15 (11.3) | 8.9 | 7.3 | ||||
| >80 | 339 (92.4) | 118 (88.7) | 91.1 | 92.7 | ||||
| Body mass index | .052 | 0.285 | .88 | 0.106 | ||||
| <18.5 kg/m² | 26 (7.1) | 9 (6.8) | 7.0 | 7.1 | ||||
| 18.5–22.9 kg/m² | 151 (41.1) | 72 (54.1) | 44.9 | 42.0 | ||||
| 23.0–27.4 kg/m² | 157 (42.8) | 40 (30.1) | 39.3 | 39.0 | ||||
| ≥27.5 kg/m² | 33 (9.0) | 12 (9.0) | 8.9 | 12.0 | ||||
| Histologic classification | 1.000 | 0.074 | .552 | 0.063 | ||||
| WHO II | 1 (0.3) | 0 (0.0) | 0.2 | 0.0 | ||||
| WHO III | 366 (99.7) | 133 (100.0) | 99.8 | 100 | ||||
| Tumor category | .189 | 0.224 | .887 | 0.087 | ||||
| T1 | 4 (1.1) | 3 (2.3) | 1.3 | 1.3 | ||||
| T2 | 30 (8.2) | 6 (4.5) | 7.1 | 5.0 | ||||
| T3 | 180 (49.0) | 76 (57.1) | 51.5 | 52.3 | ||||
| T4 | 153 (41.7) | 48 (36.1) | 40.1 | 41.4 | ||||
| Node category | .066 | 0.284 | .881 | 0.096 | ||||
| N0 | 2 (0.5) | 1 (0.8) | 0.6 | 0.5 | ||||
| N1 | 50 (13.6) | 9 (6.8) | 11.5 | 8.7 | ||||
| N2 | 110 (30.0) | 33 (24.8) | 28.5 | 28.8 | ||||
| N3 | 205 (55.9) | 90 (67.7) | 59.4 | 62.0 | ||||
| Liver metastases | <.001 | 0.492 | .812 | 0.027 | ||||
| No | 260 (70.8) | 63 (47.4) | 63.8 | 65.1 | ||||
| Yes | 107 (29.2) | 70 (52.6) | 36.2 | 34.9 | ||||
| Lung metastases | .136 | 0.16 | .888 | 0.016 | ||||
| No | 285 (77.7) | 94 (70.7) | 75.9 | 76.6 | ||||
| Yes | 82 (22.3) | 39 (29.3) | 24.1 | 23.4 | ||||
| Bone metastases | .138 | 0.159 | .764 | 0.035 | ||||
| No | 100 (27.2) | 46 (34.6) | 29.2 | 27.6 | ||||
| Yes | 267 (72.8) | 87 (65.4) | 70.8 | 72.4 | ||||
| Number of metastatic sites | <.001 | 0.492 | .854 | 0.019 | ||||
| ≤2 | 333 (90.7) | 96 (72.2) | 85.6 | 85.0 | ||||
| >2 | 34 (9.3) | 37 (27.8) | 14.4 | 15.0 | ||||
| Number of metastatic lesions | <.001 | 0.628 | .908 | 0.015 | ||||
| ≤5 | 189 (51.5) | 30 (22.6) | 43.5 | 44.3 | ||||
| >5 | 178 (48.5) | 103 (77.4) | 56.5 | 55.7 | ||||
| Lactate dehydrogenase | .001 | 0.357 | .671 | 0.051 | ||||
| ≤250 U/L | 266 (72.5) | 74 (55.6) | 67.2 | 64.8 | ||||
| >250 U/L | 101 (27.5) | 59 (44.4) | 32.8 | 35.2 | ||||
| C-reactive protein | .076 | 0.192 | .712 | 0.046 | ||||
| <3 g/mL | 167 (45.5) | 48 (36.1) | 43.5 | 41.2 | ||||
| ≥3 g/mL | 200 (54.5) | 85 (63.9) | 56.5 | 58.8 | ||||
| Pretreatment EBV DNA | .052 | 0.207 | .943 | 0.009 | ||||
| <10,000 copies/mL | 217 (59.1) | 65 (48.9) | 55.3 | 54.8 | ||||
| ≥10,000 copies/mL | 150 (40.9) | 68 (51.1) | 44.7 | 45.2 | ||||
| Posttreatment EBV DNA | .050 | 0.206 | .881 | 0.018 | ||||
| Undetectable | 272 (74.1) | 86 (64.7) | 71.7 | 70.8 | ||||
| Detectable | 95 (25.9) | 47 (35.3) | 28.3 | 29.2 | ||||
| Cycle of first-line PCT | .027 | 0.246 | .793 | 0.035 | ||||
| <6 | 104 (28.3) | 24 (18.0) | 25.2 | 23.7 | ||||
| ≥6 | 263 (71.7) | 109 (82.0) | 74.8 | 76.3 | ||||
Abbreviations: EBV, Epstein-Barr virus; LRRT, locoregional radiotherapy; IPTW, inverse probability of treatment weighting; PCT, palliative chemotherapy; SMD, standardized mean difference.
Efficacy of LRRT in the Overall Cohort
In the unadjusted Kaplan-Meier curves (Figure 1A), patients who received LRRT had a significantly longer mPFS compared with those who did not (NR vs 13.3 months; P<.001). Similarly, both the IPTW-adjusted (NR vs 15.1 months; P<.001; Figure 1B) and the 6-month conditional landmark IPTW-adjusted Kaplan-Meier curves (NR vs 21.5 months; P<.001; Figure 1C) showed a significantly longer mPFS in the LRRT group. In the univariate Cox proportional hazards regression analysis, LRRT was significantly associated with improved PFS (crude HR, 0.295 [95% CI, 0.227–0.385]; P<.001) (Figure 1A). The IPTW-adjusted (weighted HR, 0.359 [95% CI, 0.257–0.501]; P<.001; Figure 1B) and 6-month conditional landmark IPTW-adjusted (weighted HR, 0.477 [95% CI, 0.335–0.681]; P<.001; Figure 1C) Cox proportional hazards regression analyses revealed similar results.
(A) Unadjusted, (B) IPTW-adjusted,a and (C) 6-month conditional landmark IPTW-adjusteda Kaplan-Meier analyses of PFS for patients with dmNPC in the LRRT versus non-LRRT groups.
Abbreviations: dmNPC, de novo metastatic nasopharyngeal carcinoma; HR, hazard radio; IPTW, inverse probability of treatment weighting; LRRT, locoregional radiotherapy; PFS, progression-free survival.
aData are weighted proportions and not absolute numbers.
Citation: Journal of the National Comprehensive Cancer Network 23, 4; 10.6004/jnccn.2024.7086
Adverse Events
Adverse events (AEs) were reported in all patients across treatment groups (100%), with no grade 5 AEs observed (Supplementary Table S3). The most common AEs (≥50%) were anemia (93.7% in the LRRT group vs 94.0% in the non-LRRT group), leukopenia (81.2% vs 79.7%), neutropenia (76.6% vs 75.9%), nausea (69.8% vs 68.4%), hyponatremia (66.5% vs 57.1%), and vomiting (53.7% vs 51.1%). AEs that occurred significantly more frequently in the LRRT group included weight loss (50.1% vs 14.3%), increased creatinine (37.1% vs 21.8%), hypokalemia (35.1% vs 21.8%), and hypothyroidism (28.3% vs 19.5%) (all P<.05). No significant differences were observed between the groups in terms of grade ≥3 AEs. Radiotherapy-specific AEs included 23 (6.3%) cases of acute grade 3 dermatitis, 70 (19.1%) cases of grade 3 mucositis, and 2 (0.5%) cases of grade 3 xerostomia. For late AEs, 12 (3.3%) cases of grade 3 hearing loss and 3 (0.8%) cases of grade 3 neck skin fibrosis were observed.
In the LRRT group (n=367), patients receiving concurrent platinum-based chemotherapy (PBC) experienced significantly higher rates of grade ≥3 leucopenia, anemia, thrombocytopenia, nausea, and vomiting compared with those in the nonconcurrent PBC group (all P<.05; Supplementary Table S4). Among the 139 patients receiving concurrent PBC, 17 completed only 1 cycle (11 discontinued due to hematologic toxicity, 6 due to gastrointestinal toxicity), 105 completed 2 cycles (with 5 requiring dose reductions due to hematologic toxicity), and 17 completed 3 cycles (with 3 requiring dose reductions due to hematologic, gastrointestinal, and nephrotoxic effects, respectively).
Univariate and Multivariate Cox Regression Analysis
Baseline characteristics did not show significant differences between patients in the training and validation cohorts (all P>.05; Supplementary Table S2). Univariate Cox regression analyses were performed for PFS in the training cohort. As detailed in Supplementary Table S5, liver metastases, lung metastases, number of metastatic sites, number of metastatic lesions, LDH level, CRP level, pretreatment EBV DNA level, and posttreatment EBV DNA level were all showed significantly associated with PFS (all P<.05; Supplementary Table S5). Subsequently, variables with P<.1 were included in the multivariate Cox regression analysis. As illustrated in Table 2, liver metastases (yes vs no; HR, 1.432 [95% CI, 1.019–2.012]; P=.039), number of metastatic lesions (>5 vs ≤5; HR, 1.770 [95% CI, 1.232–2.542]; P=.002), LDH level (>250 vs ≤ 250 U/L; HR, 1.766 [95% CI, 1.274–2.449]; P<.001), and posttreatment EBV DNA level (detectable vs undetectable; HR, 3.600 [95% CI, 2.600–4.986]; P<.001) were identified as independent factors for PFS in patients with dmNPC.
Significant Prognostic Factors for PFS in the Training Cohort
| Variable | HR (95% CI) | P Value |
|---|---|---|
| Liver metastases | ||
| No | Ref | |
| Yes | 1.432 (1.019–2.012) | .039 |
| Number of metastatic lesions | ||
| ≤5 | Ref | |
| >5 | 1.770 (1.232–2.542) | .002 |
| Lactate dehydrogenase | ||
| ≤250 U/L | Ref | |
| >250 U/L | 1.766 (1.274–2.449) | <.001 |
| Posttreatment EBV DNA | ||
| Undetectable | Ref | |
| Detectable | 3.600 (2.600–4.986) | <.001 |
Abbreviations: EBV, Epstein-Barr virus; HR, hazard ratio; PFS, progression-free survival.
Establishment and Validation of a Novel Prognostic Model
A nomogram was developed using the 4 prognostic factors identified for predicting 1-, 2-, and 3-year PFS in patients with dmNPC from the training cohort (Figure 2A). The model demonstrated good discrimination with a C-index of 0.721 (95% CI, 0.681–0.761). To evaluate its performance robustness, internal validation was performed using a 1000 bootstrap resampling method, yielding a bias-corrected C-index of 0.715. Time-dependent ROC curves showed AUC values of 0.788, 0.774, and 0.742 for 1-, 2-, and 3-year PFS, respectively (Figure 2B). The calibration plot indicated close alignment with the 45-degree ideal line, suggesting good calibration of predicted probabilities with observed outcomes in the training cohort (Figure 2C). Additionally, DCA revealed that the nomogram offered a better net benefit than treating all or none across a range of threshold probabilities in the training cohort (Figure 2D).
(A) Nomogram-based prognostic model for predicting 1-, 2-, and 3-year PFS in patients with dmNPC. Time-dependent ROC curves for predicting the 1-, 2-, and 3-year PFS rate in the (B) training and (E) validation cohorts. Calibration plots at 1, 2, and 3 years in the (C) training and (F) validation cohorts. Decision curve analyses at 1, 2, and 3 years in the (D) training and (G) validation cohorts.
Abbreviations: AUC, area under the curve; dmNPC, de novo metastatic nasopharyngeal carcinoma; EBV, Epstein-Barr virus; LDH, lactate dehydrogenase; LRRT, locoregional radiation therapy; PFS, progression-free survival; ROC, receiver operating characteristic.
Citation: Journal of the National Comprehensive Cancer Network 23, 4; 10.6004/jnccn.2024.7086
In the validation cohort, the C-index was 0.752 (95% CI, 0.698–0.806), indicating strong discriminatory ability of the nomogram. The AUC values for predicting 1-, 2-, and 3-year PFS were 0.776, 0.797, and 0.731, respectively (Figure 2E). As observed in the training cohort, the calibration curve demonstrated a strong correspondence between nomogram predictions and actual patient outcomes in the validation set (Figure 2F). DCA also confirmed the clinical utility of the nomogram in the validation cohort (Figure 2G).
Identification of Best-Fit LRRT Candidates
To identify optimal patient subgroups for additional LRRT and facilitate personalized treatment strategies, a model-based tree was constructed using prognostic scores derived from the nomogram in the training cohort (Figure 3A). The results indicated that a maximum depth of 3 was optimal. The tree showed that patients with a prognostic score of ≤73 derived more significant survival benefits from LRRT. Using this threshold, patients were stratified into low-risk and high-risk groups. Kaplan-Meier curves from the 6-month landmark time point demonstrated a notable PFS advantage in low-risk patients who received LRRT after CT-IO compared with those who did not (mPFS, NR vs 21.5 months; P<.001; Figure 3B). Conversely, no such benefit was observed in high-risk patients (mPFS, 18.9 vs 13.3 months; P=.161; Figure 3B). Similar results were observed when further validating the stratified treatment effects in an independent validation cohort (Figure 3C) and the overall cohort (Figure 3D).
(A) The stratified model-based trees classified patients into different treatment effects subgroups based on the prognostic score derived from nomogram. Six-month conditional landmark Kaplan-Meier analyses of PFS for low-risk (nomogram score ≤73) and high-risk (score >73) patients receiving LRRT following CT-IO versus those not receiving LRRT in patients with dmNPC in the (B) training, (C) validation, and (D) overall cohorts.
Abbreviations: CT-IO, chemotherapy combined with immunotherapy; dmNPC, de novo metastatic nasopharyngeal carcinoma; LRRT, locoregional radiotherapy; mPFS, median progression-free survival; NA, not available; NR, not reached.
Citation: Journal of the National Comprehensive Cancer Network 23, 4; 10.6004/jnccn.2024.7086
Discussion
NPC is characterized by high PD-L1 expression and an abundant presence of tumor infiltrating lymphocytes,30–32 highlighting its immunogenic nature and suitability for immune checkpoint blockade therapies. Currently, CT-IO is recommended as a first-line treatment option for R-M NPC in the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Head and Neck Cancers.9 In the specific subgroup of patients with dmNPC, previous studies have confirmed that adding LRRT after chemotherapy can significantly improve PFS.11–14 However, whether LRRT offers additional benefits following first-line CT-IO and which patient subgroups might benefit most from this approach remain unclear. In this study, we used IPTW-adjusted landmark analysis to assess the efficacy of LRRT after CT-IO and developed and validated a prognostic tool to identify ideal candidates for additional LRRT, potentially enhancing its clinical utility.
Our primary finding demonstrated that adding LRRT after CT-IO significantly improved PFS in patients with dmNPC (HR, 0.295; P<.001). Notably, this benefit remained significant (HR, 0.477; P<.001) after adjusting for stabilized IPTW to control for confounders and conducting conditional landmark analysis to reduce the risk of selection and immortal time bias. This finding underscores the continued importance of LRRT in patients with dmNPC in the era of immunotherapy, consistent with findings from previous studies in the chemotherapy eras.11–14 Although some prior studies have suggested that adding LRRT after CT-IO prolongs survival in dmNPC,16–18 they were limited by small sample sizes and lacked landmark analysis, potentially leading to survival bias and an overestimation of LRRT’s efficacy.16–18 Given that the treatment regimen involving CT-IO followed by LRRT spans approximately 6 months, with LRRT administered only after CT-IO completion, conducting a 6-month landmark analyses is crucial to mitigate the risk of immortal time bias and prevent overestimation of LRRT’s effect on PFS. Therefore, our study benefits from a larger sample size (n=500) and employing more reasonable analytical methods compared with previous literature.
Why do patients with dmNPC derive additional benefits from LRRT following CT-IO? LRRT effectively controls the primary tumor and regional lymph nodes, reducing tumor burden, preventing further metastases, and alleviating symptoms. Additionally, based on the “long-tail” effect of immunotherapy,33–35 in combination with immunotherapy, LRRT can induce a synergistic mechanism known as the abscopal effect.36,37 This effect occurs when local irradiation triggers systemic immune responses, potentially leading to regression of distant metastases. Furthermore, LRRT enhances immunotherapy efficacy by promoting the release and presentation of tumor antigens,38 modifying the tumor microenvironment to increase immune cell infiltration, and reducing immunosuppressive cells.39,40 These factors collectively enhance the observed benefits in this patient population, highlighting that LRRT should be considered a valuable component of the therapeutic strategy for dmNPC.
However, patients with dmNPC exhibit considerable heterogeneity,10,41 making it challenging to establish consensus on the optimal candidates for LRRT following first-line CT-IO. Another important finding of our study was the development of a precise tool to identify patients with dmNPC most likely to benefit from LRRT. Our model incorporated key clinical factors associated with PFS, including number of metastatic lesions, presence of liver metastases, and LDH levels.17,42 Additionally, we identified clearance of EBV DNA levels after CT-IO as a crucial independent prognostic indicator in dmNPC. Previous research has consistently shown that dynamic changes in plasma EBV DNA copy numbers are closely linked to treatment responses to chemotherapy,43 radiotherapy,44 and immunotherapy.16,45,46 In the POLARIS-02 study,45 a rapid decrease in plasma EBV DNA copy number was significantly associated with a survival benefit from treatment with toripalimab monotherapy. Similarly, the CAPTAIN-1st trial46 found that early clearance of plasma EBV DNA was linked to an extended PFS in baseline-positive patients treated with a combination of camrelizumab and chemotherapy, compared with those who remained EBV DNA–positive. Our findings align with these observations, emphasizing that EBV DNA clearance after CT-IO predicts PFS in patients with dmNPC.
Our prognostic model was developed based on these identified factors that accurately predict individual outcomes and guide the selection of candidates for LRRT. Tree-based risk stratification using prognostic scores revealed that low-risk patients who received LRRT following CT-IO had significantly longer mPFS compared with those who did not received LRRT. In contrast, survival outcomes for low-risk patients who did not receive LRRT were comparable to those of high-risk patients, in whom no survival benefit from LRRT was observed. These findings underscore the importance of tailoring treatment for patients with dmNPC based on their risk profile. Ensuring that low-risk patients receive LRRT is crucial, whereas for high-risk patients—who may not derive additional benefit from LRRT—extreme caution should be exercised, and the exploration of more effective treatment strategies, such as local therapy to metastatic sites and targeted therapies, becomes a critical clinical priority.
Despite the promising results, the study has several limitations. First, as a retrospective study, it inherently carries the risk of selection bias, even with IPTW adjustment. Second, the lack of external validation across multiple institutions limits its generalizability. Third, the study had an intermediate-term follow-up period, with a median of 28.9 months. Moreover, given that the study population originated from an EBV-endemic region, it is essential to acknowledge that tumor characteristics may differ in nonendemic regions, warranting further investigation. Finally, although model-based trees offer valuable insights into subgroup treatment effects, they represent a step toward personalized medicine rather than individualized treatment recommendations. Future research should focus on developing methods to estimate individual treatment effects and optimize treatment strategies.
Conclusions
After addressing potential selection biases, our study demonstrated that LRRT following CT-IO improved PFS in patients with dmNPC. However, the survival benefit of LRRT could be heterogeneous. Our novel prognostic model categorized patients with dmNPC into low- and high-risk groups, showing that only low-risk patients demonstrated a significant improvement in PFS, whereas high-risk patients did not derive a clear benefit. To validate these results, a prospective clinical trial is essential.
Acknowledgments
We would like to thank the patients, clinical staff, data manager, and other support staff of Sun Yat-sen University Cancer Center.
References
- 1.↑
Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2021;71:209–249.
- 2.↑
Lee AWM, Ng WT, Chan LK, et al. The strength/weakness of the AJCC/UICC staging system (7th edition) for nasopharyngeal cancer and suggestions for future improvement. Oral Oncol 2012;48:1007–1013.
- 3.↑
Pan JJ, Ng WT, Zong JF, et al. Proposal for the 8th edition of the AJCC/UICC staging system for nasopharyngeal cancer in the era of intensity-modulated radiotherapy. Cancer 2016;122:546–558.
- 5.↑
Mai HQ, Chen QY, Chen D, et al. Toripalimab or placebo plus chemotherapy as first-line treatment in advanced nasopharyngeal carcinoma: a multicenter randomized phase 3 trial. Nat Med 2021;27:1536–1543.
- 6.↑
Yang Y, Pan J, Wang H, et al. Tislelizumab plus chemotherapy as first-line treatment for recurrent or metastatic nasopharyngeal cancer: a multicenter phase 3 trial (RATIONALE-309). Cancer Cell 2023;41:1061–1072.e4.
- 7.↑
Chan ATC, Lee VHF, Hong RL, et al. Pembrolizumab monotherapy versus chemotherapy in platinum-pretreated, recurrent or metastatic nasopharyngeal cancer (KEYNOTE-122): an open-label, randomized, phase III trial. Ann Oncol 2023;34:251–261.
- 8.↑
Bossi P, Chan AT, Even C, Machiels JP. ESMO-EURACAN clinical practice guideline update for nasopharyngeal carcinoma: adjuvant therapy and first-line treatment of recurrent/metastatic disease. Ann Oncol 2023;34:247–250.
- 9.↑
Pfister DG, Spencer S, Adkins D, et al. NCCN Clinical Practice Guidelines in Oncology: Head and Neck Cancers. Version 4.2024. Accessed May 1, 2024. To view the most recent version, visit https://www.nccn.org/
- 10.↑
Zou X, You R, Liu H, et al. Establishment and validation of M1 stage subdivisions for de novo metastatic nasopharyngeal carcinoma to better predict prognosis and guide treatment. Eur J Cancer 2017;77:117–126.
- 11.↑
Rusthoven CG, Lanning RM, Jones BL, et al. Metastatic nasopharyngeal carcinoma: patterns of care and survival for patients receiving chemotherapy with and without local radiotherapy. Radiother Oncol 2017;124:139–146.
- 12.↑
Zeng L, Tian YM, Huang Y, et al. Retrospective analysis of 234 nasopharyngeal carcinoma patients with distant metastasis at initial diagnosis: therapeutic approaches and prognostic factors. PloS One 2014;9:e108070.
- 13.↑
Hu J, Kong L, Gao J, et al. Use of radiation therapy in metastatic nasopharyngeal cancer improves survival: a SEER analysis. Sci Rep 2017;7:721.
- 14.↑
Hu SX, He XH, Dong M, et al. Systemic chemotherapy followed by locoregional definitive intensity-modulated radiation therapy yields prolonged survival in nasopharyngeal carcinoma patients with distant metastasis at initial diagnosis. Med Oncol 2015;32:224.
- 15.↑
You R, Liu YP, Huang PY, et al. Efficacy and safety of locoregional radiotherapy with chemotherapy vs chemotherapy alone in de novo metastatic nasopharyngeal carcinoma: a multicenter phase 3 randomized clinical trial. JAMA Oncol 2020;6:1345–1352.
- 16.↑
Hu YJ, Lu TZ, Zhang H, et al. Locoregional radiotherapy improves survival outcomes in de novo metastatic nasopharyngeal carcinoma treated with chemoimmunotherapy. ESMO Open 2023;8:101629.
- 17.↑
Chen Y, Chen C, Peng H, et al. Risk-adapted locoregional radiotherapy strategies based on a prognostic nomogram for de novo metastatic nasopharyngeal carcinoma patients treated with chemoimmunotherapy. Sci Rep 2024;14:3950.
- 18.↑
Liu ZQ, Zhao YN, Wu YS, et al. Immunochemotherapy alone or immunochemotherapy plus subsequent locoregional radiotherapy in de novo metastatic nasopharyngeal carcinoma. Oral Oncol 2023;147:106583.
- 19.↑
Shen L, Dong J, Li S, et al. M1 stage subdivision and treatment outcome of patients with bone-only metastasis of nasopharyngeal carcinoma. Oncologist 2015;20:291–298.
- 20.↑
Sun XS, Liang YJ, Liu SL, et al. Subdivision of nasopharyngeal carcinoma patients with bone-only metastasis at diagnosis for prediction of survival and treatment guidance. Cancer Res Treat 2019;51:1259–1268.
- 21.↑
Sun XS, Liu SL, Luo MJ, et al. The association between the development of radiation therapy, image technology, and chemotherapy, and the survival of patients with nasopharyngeal carcinoma: a cohort study from 1990 to 2012. Int J Radiat Oncol Biol Phys 2019;105:581–590.
- 22.↑
Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45:228–247.
- 23.↑
Chen SY, Duan XT, Li HF, et al. Efficacy of sequential chemoradiotherapy combined with toripalimab in de novo metastatic nasopharyngeal carcinoma: a phase II trial. Cell Rep Med 2023;4:101279.
- 24.↑
You R, Liu YP, Lin M, et al. Relationship of circulating tumor cells and Epstein-Barr virus DNA to progression-free survival and overall survival in metastatic nasopharyngeal carcinoma patients. Int J Cancer 2019;145:2873–2883.
- 25.↑
Korevaar JC, van Manen JG, Dekker FW, et al. Effect of an increase in C-reactive protein level during a hemodialysis session on mortality. J Am Soc Nephrol 2004;15:2916–2922.
- 26.↑
Kastritis E, Morel P, Duhamel A, et al. A revised international prognostic score system for Waldenström’s macroglobulinemia. Leukemia 2019;33:2654–2661.
- 27.↑
Chesnaye NC, Stel VS, Tripepi G, et al. An introduction to inverse probability of treatment weighting in observational research. Clin Kidney J 2022;15:14–20.
- 28.↑
Li WZ, Lv SH, Liu GY, et al. Development of a prognostic model to identify the suitable definitive radiation therapy candidates in de novo metastatic nasopharyngeal carcinoma: a real-world study. Int J Radiat Oncol Biol Phys 2021;109:120–130.
- 29.↑
Seibold H, Zeileis A, Hothorn T. Model-based recursive partitioning for subgroup analyses. Int J Biostat 2016;12:45–63.
- 30.↑
Chang AMV, Chiosea SI, Altman A, et al. Programmed death-ligand 1 expression, microsatellite instability, Epstein-Barr virus, and human papillomavirus in nasopharyngeal carcinomas of patients from the Philippines. Head Neck Pathol 2017;11:203–211.
- 31.↑
Ono T, Azuma K, Kawahara A, et al. Prognostic stratification of patients with nasopharyngeal carcinoma based on tumor immune microenvironment. Head Neck 2018;40:2007–2019.
- 32.↑
Wang YQ, Chen YP, Zhang Y, et al. Prognostic significance of tumor- infiltrating lymphocytes in nondisseminated nasopharyngeal carcinoma: a large-scale cohort study. Int J Cancer 2018;142:2558–2566.
- 33.↑
Cali Daylan AE, Halmos B. Long-term benefit of immunotherapy in metastatic non-small cell lung cancer: the tale of the tail. Transl Lung Cancer Res 2023;12:1636–1642.
- 34.↑
Chen H, Li Z, Qiu L, et al. Personalized neoantigen vaccine combined with PD-1 blockade increases CD8 (+) tissue-resident memory T-cell infiltration in preclinical hepatocellular carcinoma models. J Immunother Cancer 2022;10:e004389.
- 35.↑
Zhao J, Dong Y, Bai H, et al. Multi-omics indicators of long-term survival benefits after immune checkpoint inhibitor therapy. Cell Rep Methods 2023;3:100596.
- 36.↑
Zhang Z, Liu X, Chen D, Yu J. Radiotherapy combined with immunotherapy: the dawn of cancer treatment. Signal Transduct Target Ther 2022;7:258.
- 37.↑
Ngwa W, Irabor OC, Schoenfeld JD, et al. Using immunotherapy to boost the abscopal effect. Nat Rev Cancer 2018;18:313–322.
- 38.↑
Fucikova J, Kepp O, Kasikova L, et al. Detection of immunogenic cell death and its relevance for cancer therapy. Cell Death Dis 2020;11:1013.
- 39.↑
Liu S, Wang W, Hu S, et al. Radiotherapy remodels the tumor microenvironment for enhancing immunotherapeutic sensitivity. Cell Death Dis 2023;14:679.
- 40.↑
Wang Y, Liu ZG, Yuan H, et al. The reciprocity between radiotherapy and cancer immunotherapy. Clin Cancer Res 2019;25:1709–1717.
- 41.↑
Chan SK, Lin C, Huang SH, et al. Refining TNM-8 M1 categories with anatomic subgroups for previously untreated de novo metastatic nasopharyngeal carcinoma. Oral Oncol 2022;126:105736.
- 42.↑
Sun XS, Liang YJ, Liu SL, et al. Establishment and validation of a nomogram for predicting survival in patients with de novo metastatic nasopharyngeal carcinoma. Oral Oncol 2019;94:73–79.
- 43.↑
Chan ATC, Hui EP, Ngan RKC, et al. Analysis of plasma Epstein-Barr virus DNA in nasopharyngeal cancer after chemoradiation to identify high-risk patients for adjuvant chemotherapy: a randomized controlled trial. J Clin Oncol 2018;36:JCO2018777847.
- 44.↑
Wang WY, Twu CW, Chen HH, et al. Plasma EBV DNA clearance rate as a novel prognostic marker for metastatic/recurrent nasopharyngeal carcinoma. Clin Cancer Res 2010;16:1016–1024.
- 45.↑
Wang FH, Wei XL, Feng J, et al. Efficacy, safety, and correlative biomarkers of toripalimab in previously treated recurrent or metastatic nasopharyngeal carcinoma: a phase II clinical trial (POLARIS-02). J Clin Oncol 2021;39:704–712.
- 46.↑
Yang Y, Qu S, Li J, et al. Camrelizumab versus placebo in combination with gemcitabine and cisplatin as first-line treatment for recurrent or metastatic nasopharyngeal carcinoma (CAPTAIN-1st): a multicentre, randomised, double-blind, phase 3 trial. Lancet Oncol 2021;22:1162–1174.
