- Email: [email protected]
The time frame of Epstein-Barr virus latent membrane protein-1 gene to disappear in nasopharyngeal swabs after initiation of primary radiotherapy is an independently significant prognostic factor predicting local control for patients with nasopharyngeal carcinoma
Int. J. Radiation Oncology Biol. Phys., Vol. 63, No. 5, pp. 1339 –1346, 2005 Copyright © 2005 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/05/$–see front matter
doi:10.1016/j.ijrobp.2005.05.051
CLINICAL INVESTIGATION
Head and Neck
THE TIME FRAME OF EPSTEIN-BARR VIRUS LATENT MEMBRANE PROTEIN-1 GENE TO DISAPPEAR IN NASOPHARYNGEAL SWABS AFTER INITIATION OF PRIMARY RADIOTHERAPY IS AN INDEPENDENTLY SIGNIFICANT PROGNOSTIC FACTOR PREDICTING LOCAL CONTROL FOR PATIENTS WITH NASOPHARYNGEAL CARCINOMA SHINN-YN LIN, M.D.,* KAI-PING CHANG, M.D.,†‡ MENG-SHU HSIEH, M.D., M.P.H.,§ SHIR-HWA UENG, M.D.,储 SHENG-PO HAO, M.D.,† CHEN-KAN TSENG, M.D.,* PING-CHING PAI, M.D.,* FU-TI CHANG, M.D.,* MING-HSUI TSAI, M.D.,¶ AND NGAN-MING TSANG, M.D., D.SC.* Departments of *Radiation Oncology, †Otolaryngology, §Family Medicine, 储Anatomic Pathology, and ¶Otolaryngology, Chang Gung Memorial Hospital, Linkou, Taiwan; and ‡Graduate Institute of Clinical Medical Sciences, Chang Gung University, Linkou, Taiwan Purpose: The presence of Epstein-Barr virus latent membrane protein-1 (LMP-1) gene in nasopharyngeal swabs indicates the presence of nasopharyngeal carcinoma (NPC) mucosal tumor cells. This study was undertaken to investigate whether the time taken for LMP-1 to disappear after initiation of primary radiotherapy (RT) was inversely associated with NPC local control. Methods and Materials: During July 1999 and October 2002, there were 127 nondisseminated NPC patients receiving serial examinations of nasopharyngeal swabbing with detection of LMP-1 during the RT course. The time for LMP-1 regression was defined as the number of days after initiation of RT for LMP-1 results to turn negative. The primary outcome was local control, which was represented by freedom from local recurrence. Results: The time for LMP-1 regression showed a statistically significant influence on NPC local control both univariately (p < 0.0001) and multivariately (p ⴝ 0.004). In multivariate analysis, the administration of chemotherapy conferred a significantly more favorable local control (p ⴝ 0.03). Advanced T status (> T2b), overall treatment time of external photon radiotherapy longer than 55 days, and older age showed trends toward being poor prognosticators. The time for LMP-1 regression was very heterogeneous. According to the quartiles of the time for LMP-1 regression, we defined the pattern of LMP-1 regression as late regression if it required 40 days or more. Kaplan-Meier plots indicated that the patients with late regression had a significantly worse local control than those with intermediate or early regression (p ⴝ 0.0129). Conclusion: Among the potential prognostic factors examined in this study, the time for LMP-1 regression was the most independently significant factor that was inversely associated with NPC local control. © 2005 Elsevier Inc. Nasopharyngeal carcinoma, Epstein-Barr virus, Latent membrane protein-1 gene, Radiotherapy, Local control.
INTRODUCTION
Because almost every NPC tumor cell carries the Epstein-Barr virus (EBV) genome (5), detection of EBV genomic deoxyribonucleic acid (DNA) such as latent membrane protein-1 gene (LMP-1) might indicate the presence of NPC (6 – 8). In recent years, polymerase chain reaction (PCR)– based detection of EBV genomic DNA has been employed as a powerful tool to detect the presence of NPC tumor cells with satisfactory sensitivity and excellent specificity (9).
Although radiotherapy is the cornerstone of treatment for nasopharyngeal carcinoma (NPC) and this tumor is much more radiocurable than other head-and-neck malignancies, the cumulative incidence of local recurrence after definitive radiotherapy ranges from 20% to 30% (1–3). Most local relapses after definitive radiotherapy occur within the first 2 years after primary treatment, with a median time of approximately 1.5 years (4).
Memorial Hospital (CMRPG33023, CMRPG33077-II) and the Taiwan National Scientific Council (NSC 89-2314-B-182A-089M08, NSC 93-2314-B-182A-128). Received Mar 4, 2005, and in revised form May 13, 2005. Accepted for publication May 15, 2005.
Reprint requests to: Ngan-Ming Tsang, M.D., D.Sc., Department of Radiation Oncology, Chang Gung Memorial Hospital, 5 Fu-Shin St., Kwei-Shan Hsiang, Taoyuan, Taiwan, R.O.C. Tel: (⫹886) 3-3281200 ext. 2600, 2645; Fax: (⫹886) 3-3280797; E-mail: [email protected]; [email protected] This study was supported in part by grants from the Chang Gung 1339
1340
I. J. Radiation Oncology
● Biology ● Physics
From the evidence demonstrated by prior studies in our laboratory, the presence of EBV LMP-1 gene in nasopharyngeal swab specimens indicates the presence of NPC tumor cells involving at least the nasopharyngeal mucosa; likewise, the disappearance of EBV LMP-1 implies the eradication of mucosal NPC tumor cells (9 –13). In one of our previous studies, it was found that patients with early LMP-1 disease remission during the course of radiotherapy (RT) seemed to have a better outcome in terms of local control. It was suggested that the responsiveness of LMP-1 to irradiation might reflect the aggressiveness and radiosensitivity of the NPC (10). However, it was not clear whether the above association could still exist after taking into consideration other well-established prognostic factors of NPC local control such as pretreatment T classification and the use of chemotherapy. Therefore, we assumed that the time for LMP-1 to disappear after initiation of primary RT represented the aggressiveness of NPC itself and indicated biologic response after treatment, which was highly associated with NPC local control. That is, hypothetically, the faster LMP-1 takes to disappear, the better NPC local control can be achieved after standard therapy. To investigate whether the time taken for LMP-1 to disappear after initiation of primary RT was significantly and inversely associated with NPC local control in a cohort of nondisseminated NPC patients after radical treatment, we performed serial examinations of nasopharyngeal swabbing with detection of EBV LMP-1 after initiation of primary RT. The time taken for LMP-1 to disappear after initiation of primary RT was estimated, and its prognostic value in predicting local control was comprehensively studied.
METHODS AND MATERIALS Study patients Between July 1999 and October 2002, a prospective cohort of NPC patients without distant metastasis at initial diagnosis was ascertained, consisting of 127 patients with NPC who were treated with definitive radiotherapy (cumulative dose of external beam radiotherapy ⬎68 Gy, with or without intracavitary brachytherapy) in the Department of Radiation Oncology in Linkou Chang Gung Memorial Hospital. Because it is well documented that patients with positive nasopharyngeal histology 12 weeks after completion of primary radical radiotherapy must have persistent disease and therefore early initiation of salvage therapy cannot be disputed (14), we have to emphasize that we excluded from our analysis the patients who had persistent presence of EBV LMP-1 in nasopharyngeal swabs even at 3 months after completion of primary RT. Evaluation of treatment response by detecting EBV LMP-1 had to be made within 3 months after completion of primary RT, and LMP-1 persistence indicating residual disease would not be declared until 3 months after completion of primary RT. Provided that the patient had at least two negative LMP-1 results, it was assumed that biologic remission of LMP-1 had been achieved irrespective of whether there was a suspicious lesion found by nasopharyngoscopic examination. In addition, each studied patient had to receive
Volume 63, Number 5, 2005
LMP-1 detection five or more times during the course of RT to prevent the possibility of sampling error. The staging system applied was the 1997 American Joint Committee on Cancer Cancer Staging (15), and histologic typing according to the World Health Organization (WHO) classification (16) was employed. Approximately one-half of the studied patients underwent magnetic resonance imaging (MRI) before primary treatment. If initial MRI was not done or unavailable, computed tomography (CT) imaging was an alternative when determining T classification. All patients in this cohort were treated in a consistent manner by the same well-experienced radiation oncologist. Conventional radiotherapy techniques without the component of intensity-modulated radiotherapy were employed. All recruited subjects underwent serial nasopharyngeal swabbing with PCR amplification to detect LMP-1 genome before or soon after initiation of RT, at least every 2 weeks during the RT course, upon completion of RT, and at a regular 2-month interval during followup. Each specimen of nasopharyngeal swabbing was obtained after the patient had signed the written informed consent.
PCR-based detection of the presence of EBV LMP-1 in nasopharyngeal swabs The nasopharyngeal swabbing techniques, tissue processing, laboratory methods, and PCR amplification have been described in detail and published previously (9, 10, 12). In brief, the patient’s nasopharyngeal wall was thoroughly brushed with a 15-cm long cotton stick, which had been prepared and immersed in phosphatebuffered solution. The specimen was then sent to the laboratory for DNA extraction and purification. We chose to amplify regions of the EBV LMP-1 gene (sense primer BN1, 5=-AGC GAC TCT GCT GGA AAT GAT-3=; antisense primer BN2, 5=-TGA TTA GCT AAG GCA TTC CCA-3=) for identification of viral DNA. For PCR amplification, oligonucleotide primers (sense, BN1; antisense, BN2) were used to detect the presence of LMP-1, which was a 316 base-pair PCR product in the extracted DNA.
The definition and calculation of the time taken for LMP-1 to disappear after primary RT The independent variable that we intended to evaluate in this study was the time taken for LMP-1 to disappear after initiation of primary RT. For convenience and simplicity, we will call it the time for LMP-1 regression in the following text. In fact, it was not possible to identify the exact time when the EBV LMP-1 result had turned negative. We could only surmise that LMP-1 remission occurred between the date of the last positive LMP-1 result and the date of the first negative LMP-1 result. To be practical and realistic, we defined the date when LMP-1 first converted to be negative as a midpoint between the above two dates. Accordingly, the time for LMP-1 regression was calculated as the duration from the starting date of primary RT to the defined date when LMP-1 result converted to negative. Basically, the time for LMP-1 regression is a continuous variable in nature.
Statistical analysis The primary endpoint of interest in this study was NPC local recurrence. The time to develop NPC local recurrence was used as the outcomes measure. It is calculated as the duration from the date of the initiation of primary RT to the date of pathologic diagnosis of local failure, or to the date of censoring due to the following conditions: any causes of death, the occurrence of distant metastasis, and the date of last contact. The occurrence of distant
The time for LMP-1 to regress in NPC
metastasis was regarded as a censor in our study because distant metastasis from NPC or its related death was the most common competitive event for the development of NPC local relapse. Approximately one-fourth of all NPC patients would develop distant metastasis during follow-up. The majority of these NPC patients experienced distant metastasis within 2 years after completion of primary RT, with a median time of 17 months. Their median survival was only 10 months after distant metastasis (our unpublished data). A Cox proportional hazards regression model was performed to evaluate the impact of the time for LMP-1 regression and other potential prognostic factors on the local tumor control of NPC patients after definitive RT. The time for LMP-1 regression was treated as a continuous variable in the Cox proportional hazards regression model both in univariate and multivariate analyses. Local control rates, namely cumulative local recurrence-free probabilities, were calculated by the Kaplan-Meier method. A log–rank test was employed to compare the difference between the local control rates according to various potential prognostic factors. To evaluate the prognostic importance of the kinetic or pattern of LMP-1 regression, all studied patients were divided into three subgroups with different time frames of regression according to the time for LMP-1 regression. The principal exposure variable in our analysis, the time for LMP-1 regression, was then converted to a categorical and ordinal variable according to its quartiles. The related results and figures will be presented. The other potential variables of prognostic value in predicting NPC local control were also examined univariately or multivariately, including gender, age at initial diagnosis, primary tumor status (T2b-T4 vs. earlier than T2b), WHO histologic typing, the addition of intracavitary brachytherapy, whether there was an administration of chemotherapy, and overall treatment time of external photon radiotherapy. In general, a p value ⱕ0.05 was considered statistically significant.
RESULTS The median follow-up duration was 33.9 months, ranging from 6.4 to 58.5 months for all patients, and 36.3 months, ranging from 20.7 to 58.6 months, for survivors. The study cohort consisted of 94 men and 33 women. The age of primary NPC diagnosis ranged from 19.8 to 87.1 years old with a median age of 48.7 years. There were 45 patients (35.4%) whose primary tumor was staged as T2 at initial diagnosis. Thirty of these patients were further divided into T2a or T2b based on the status of parapharyngeal space extension if the reports of MRI or CT provided available information. According to WHO histologic classification, 48 cases (37.8%) were diagnosed as WHO type III, which denoted undifferentiated carcinoma; 72 cases (56.7%) were determined to be WHO type II, representing nonkeratinizing or poorly differentiated carcinoma; and only 3 patients’ primary tumors were interpreted as squamous cell carcinoma. In terms of treatment, there were 36 patients (28.4%) receiving at least one course of chemotherapy predominantly in a concurrent setting. Except for 4 patients, the majority of patients received intracavitary brachytherapy during the RT course. Overall treatment time of external photon radiotherapy, regardless of electron boost to neck
● S.-Y. LIN et al.
1341
Table 1. Demographics of the 127 investigated patients Patient characteristics Gender Male Female Age* Pathology–WHO typing I II III NE Primary tumor classification (T)† T1 T2 T2a ⫹ Unspecified T2b‡ T3 T4 Neck status (N)† N0–N1 N2–N3 Stage grouping† I II III IV Brachytherapy Yes No Chemotherapy Yes No Overall Tx time* The time taken for LMP-1 to disappear after initiation of primary RT* Pattern of LMP-1 regression Early (⬍16 days) Intermediate Late (40 days or more) The time taken for gross tumor to regress after initiation of primary RT (assessed by endoscope)*
No. of patients (%) 94 (74) 33 (26) Median
48.7 years of age
3 (2.36) 72 (56.69) 48 (37.80) 4 (3.15) 33 (25.98) 45 (35.43) 17 ⫹ 15 13 24 (18.90) 25 (19.69) 87 (68.5) 40 (31.5) 15 (11.8) 43 (33.9) 39 (30.7) 30 (23.6) 123 (96.85) 4 (3.15) 36 (28.35) 91 (71.65) Median 25% quartile Median 75% quartile 28 (22) 67 (52.8) 32 (25.2) Median
55 days 16 days 24.5 days 40 days
40 days
Abbreviations: WHO ⫽ World Health Organization; NE ⫽ not evaluable; Overall Tx time ⫽ overall treatment time of external photon therapy; AJCC ⫽ American Joint Committee on Cancer; RT ⫽ radiotherapy; LMP-1 ⫽ latent membrane protein-1. * It was regarded as a continuous variable. † The 1997 TNM staging system of AJCC was employed. ‡ T2b denotes parapharyngeal extension, which usually was evidenced by magnetic resonance imaging.
lymphatics, ranged from 46 days to 78 days, with a median of 55 days. The patient characteristics, disease-related factors, treatment characteristics, and the time for LMP-1 regression are shown in Table 1. By November 22, 2004, a total of 25 patients had developed local recurrence. The median time to develop NPC local failure after primary RT was 16.2 months, with a range between 7.3 months and 38.2 months.
1342
I. J. Radiation Oncology
● Biology ● Physics
The time for LMP-1 regression ranged from less than 1 week to 116.5 days after the initiation of primary RT (25% quartile, 16 days; median, 24.5 days; 75% quartile, 40 days). Here, according to the quartiles of the time for LMP-1 regression (16 and 40 days, respectively), we defined the kinetic of LMP-1 regression as being early regression if it required shorter than 16 days, late regression if it took 40 days or longer, and intermediate regression if LMP-1 disappeared between 16 and 40 days after starting RT. The local control rates according to the pattern of LMP-1 regression are presented in Fig. 1. It shows that the patients with delayed LMP-1 regression requiring 40 days or more after starting RT had a significantly worse local control than those with intermediate or early LMP-1 regression (log– rank test, p ⫽ 0.035). Their 2-year actuarial local failurefree rates were 67.41%, 84.15%, and 92.52%, respectively, according to the pattern of LMP-1 regression (Table 2). It was also observed that the difference in NPC local control existed between the patients with late regression and the remaining patients whose LMP-1 regression occurred within 40 days after initiation of primary RT (log–rank test, p ⫽ 0.0129). Furthermore, the cumulative local recurrence-free probabilities according to primary tumor status, WHO subtype, whether chemotherapy was administered, and overall treatment time of external photon radiotherapy were calculated. Figure 2 reveals that T2b disease or more advanced primary tumor extension conferred a significantly worse local control (log–rank test, p ⫽ 0.043). The patients receiving chemotherapy seemed to have a more favorable freedom from local failure than those who did not, but the difference failed to be statistically significant (p ⫽ 0.289). Likewise, the patients whose overall treatment time of external photon radiotherapy was more protracted than 55 days appeared to
Fig. 1. Freedom from local failure, according to the kinetic of latent membrane protein-1 regression. Early regression denotes that the time for latent membrane protein-1 to disappear takes less than 16 days after initiation of primary radiotherapy. Late regression indicates this specific time requires 40 days or more. Intermediate regression implies that remission of latent membrane protein-1 occurs between 16 and 40 days after starting radiotherapy.
Volume 63, Number 5, 2005
Table 2. The pattern of LMP-1 regression* according to the quartiles of the time for LMP-1 regression† Kinetic of LMP-1 regression
2-year local recurrencefree rate‡
HR (95% CI; [p value])
Early Intermediate Late
92.52% 84.15% 67.41%
1 1.709 (0.476–6.132; [p ⫽ 0.4]) 2.215 (0.96–5.111; [p ⫽ 0.06])
Abbreviations: HR ⫽ hazards ratio; 95% CI ⫽ 95% confidence interval; LMP-1 ⫽ latent membrane protein-1. * It was treated as a discrete variable. † It was treated as a continuous variable. ‡ Their Kaplan-Meier plots were compared by log–rank test. The p value was 0.035.
suffer a higher risk of local failure than those whose RT course could be completed within 8 weeks, but the difference did not reach statistical significance (p ⫽ 0.103). Table 3 shows unadjusted and adjusted hazard ratios for the time for LMP-1 regression and other various prognostic factors predicting NPC local recurrence. In the univariate analysis, longer time for LMP-1 regression and advanced T disease above T2a were statistically significant poor prognostic factors. All parameters concerned in the univariate analysis were also included in the multivariate analysis. They were the time for LMP-1 regression (treated as a continuous variable), age (treated as a continuous variable), gender (male/female), T status (T2b-T4/T1-T2a), N status (N2-3/N0-1), the use of chemotherapy (ever/none), brachytherapy (yes/no), and overall treatment time of external photon radiotherapy (⬎55 days/ⱕ55 days). The independently significant prognostic factors for NPC local recurrence identified by multivariate analysis were the time for LMP-1 regression and the use of chemotherapy. Advanced T status, overall treatment time of external photon radiotherapy longer than 55 days, and older age showed trends toward
Fig. 2. Freedom from local failure, according to extent of the primary tumor. Tumors of T1, T2a, and unspecified T2 diseases were grouped as early T status group. Tumors of T3 and T4 diseases were combined together with those of T2b classification to comprise the advanced T status group.
The time for LMP-1 to regress in NPC
● S.-Y. LIN et al.
1343
Table 3. Cox proportional hazards model analysis on time to develop local recurrence Univariate
Multivariate
Variable
p value
Crude HR (95% CI)
p value
Adjusted HR (95% CI)
The time of LMP-1 regression* T status (T1⬃2a vs. T2b and higher) Age* Sex (male vs. female) N status (N2 ⫹ N3 vs. N0 ⫹ N1) Chemotherapy (yes vs. no) Brachytherapy (yes vs. no) Overall Tx time (⬎ 55 days vs. shorter period)
⬍0.0001§ 0.0491‡ 0.0704† 0.8775 0.6792 0.294 0.9011 0.1085
1.028 (1.016–1.041) 2.327 (1.003–5.396) 1.032 (0.997–1.067) 0.934 (0.390–2.237) 1.194 (0.515–2.768) 0.592 (0.222–1.577) 0.881 (0.119–6.524) 1.904 (0.867–4.181)
0.0041§ 0.0554† 0.0968† 0.8959 0.2444 0.0316‡ 0.8732 0.0768†
1.021 (1.007–1.036) 2.66 (0.978–7.235) 1.03 (0.995–1.066) 0.938 (0.357–2.464) 1.767 (0.678–4.609) 0.291 (0.095–0.897) 1.183 (0.150–9.314) 2.091 (0.924–4.731)
Abbreviations: HR ⫽ hazards ratio; 95% CI ⫽ 95% confidence interval; Overall Tx time ⫽ overall treatment time of external photon therapy; LMP-1 ⫽ latent membrane protein-1. * Treated as a continuous variable. † p ⬍ 0.1; ‡ p ⬍ 0.05; § p ⬍ 0.01.
being prognostic factors. After adjusting for these five significant prognostic factors identified in the multivariate analysis, the time for LMP-1 regression still remained the most significant independent prognostic factor that was inversely associated with NPC local control (Table 4). To further clarify the inverse association between the time for LMP-1 regression and the time to develop local failure, we also performed subgroup analysis according to either T status or the kinetic of LMP-1 regression (Table 5). In consideration of the small number of patients and events in each T subcategory, the patients with primary diseases earlier than T2b were combined together to form the group with early T status, and the patients with T2b-T4 diseases were grouped as those with advanced T diseases. The prognostic value of the time for LMP-1 regression was very remarkable in the group of advanced T status (p ⬍ 0.0001, Cox regression analysis). However, the time for LMP-1 regression was not associated with NPC local control in the group of early T disease (p ⫽ 0.7371, Cox regression analysis). This observation is also expressed as KaplanMeier plots in Fig. 3, in which the patients with early LMP-1 regression were grouped together with those with intermediate LMP-1 regression. Table 4. Cox proportional hazards model analysis on time to develop local recurrence when adjusting for those five significant factors identified in the multivariate analysis Variable
p value
HR (95% CI)
The time of LMP-1 regression* T status (T1⬃2a vs. T2b and higher) Age* Chemotherapy (ever vs. none) Overall Tx time (⬎55 days vs. shorter period)
0.0072§ 0.0405‡
1.019 (1.005–1.034) 2.818 (1.046–7.594)
0.0618† 0.0478‡ 0.0897†
1.032 (0.998–1.066) 0.327 (0.108–0.989) 2.019 (0.897–4.546)
Abbreviations: HR ⫽ hazards ratio; 95% CI ⫽ 95% confidence interval; Overall Tx time ⫽ overall treatment time of external photon therapy; LMP-1 ⫽ latent membrane protein-1. * Treated as a continuous variable. † p ⬍ 0.1; ‡ p ⬍ 0.05; § p ⬍ 0.01.
The remaining part of Table 5 demonstrates that the prognostic value of the time for LMP-1 regression existed in the subgroup with late LMP-1 regression (p ⫽ 0.0016) and was still very remarkable after controlling for T status (p ⫽ 0.0074). Its prognostic value failed to reach statistical significance when the kinetic of LMP-1 regression was early or intermediate. These results further consolidated the findings exhibited in Fig. 1 and Table 2, in which the statistically significant difference in local recurrence-free rates exists primarily between the patients with delayed LMP-1 regression requiring 40 days or more and those whose regression occurred within 40 days after starting RT. DISCUSSION Numerous studies have evaluated the prognostic value of various parameters in NPC local control. The most wellestablished prognostic factors are primary tumor status (T classification) at initial diagnosis (1, 3, 17) and various factors comprising the T classification, such as parapharyngeal extension (18, 19), skull base invasion (20 –22), or cranial nerve involvement (3, 17, 20, 22). Other factors that have been reported include patient age (1, 17, 22), histologic type (3, 17), RT dose to the nasopharynx (2, 17, 23), overall treatment time of radiotherapy (24), treatment era (2, 25), the administration of chemotherapy (26 –28), the addition of intracavitary brachytherapy (29), and some clinical treatment response factors such as recovery of headache or cranial nerve palsy (21). To the best of our knowledge, no biologic factors associated with treatment response have been previously reported as prognosticators. In 1999, Kwong et al. (14) reported their study evaluating the prognostic significance of delayed histologic remission, which could be viewed as a pathologic response factor in NPC patients. However, they did not find any prognostic importance of delayed histologic remission. The discrepancy between the above Hong Kong study (histologic remission) and our study (the time for LMP-1 regression indicating biologic response) suggested that the tumor cells found in posttreatment biopsies might
1344
I. J. Radiation Oncology
● Biology ● Physics
Volume 63, Number 5, 2005
Table 5. Subgroup analysis on the prognostic value of the time for LMP-1 regression according to T status or the speed of LMP-1 regression Stratification T status Early T (earlier than T2b) Advanced T (T2b at least) The speed of LMP-1 regression Early regression (⬍16 days)
Exposure variable
p value
HR (95% CI)
The time for LMP-1 regression The time for LMP-1 regression
0.7371 ⬍0.0001*
0.991 (0.941–1.004) 1.03 (1.016–1.045)
The time for LMP-1 regression
0.2966 0.3125† 0.2966 0.8798 0.887† 0.872 0.0016* 0.0074† 0.137
Intermediate regression
T status‡ The time for LMP-1 regression
Late regression (ⱖ40 days)
T status‡ The time for LMP-1 regression T status‡
1.286 (0.802–2.063) 1.257 (0.806–1.960) 1.286 (0.802–2.063) 1.007 (0.923–1.099) 1.006 (0.922–1.098) 1.103 (0.336–3.164) 1.036 (1.013–1.058) 1.031 (1.008–1.054) 4.777 (0.608–37.524)
Abbreviations: HR ⫽ hazards ratio; 95% CI ⫽ 95% confidence interval. * p ⬍ 0.01. † Adjusted for T status; ‡ T2b-T4 vs. T1-T2a.
have lost their clonogenicity and reproductive potential if the detection of EBV LMP-1 at the same time revealed negative results. The time for LMP-1 regression could be regarded as an indicator of biologic response, which reflects both effectiveness of treatment and biologic aggressiveness of the NPC tumor itself. Provided that there is similar tumor burden or volume, it would be justifiable to suggest that the longer it takes for LMP-1 to disappear, the more aggressive the tumor is and the more intensive treatment the disease demands. In our analysis, the time taken for LMP-1 to disappear after initiation of primary RT was the most significant independent prognostic factor that was inversely associated with NPC local control. The time for LMP-1 regression did show a statistically significant influence on the time to develop local failure during the follow-up period. The time is so powerful as a prognosticator predicting NPC local control that it still retained its prognostic significance after controlling for other well-documented prognostic factors such as pretreatment T status, the use of chemotherapy, and overall treatment time of external photon therapy. In addition to the above findings specific to the purpose of this study, we had several other observations that supported the past related reports (1, 22, 24, 26, 27). For example, it was confirmed in our study that the administration of chemotherapy conferred a more favorable prognosis. Besides, the older the patient was, the higher the risk for NPC local failure. When the course of external photon radiotherapy was more protracted (i.e., longer than 8 weeks), NPC local control was also significantly poorer. Unlike some previous reports (29), our study failed to demonstrate the benefit in local control enhanced by the addition of intracavitary brachytherapy. In our study, it is clearly demonstrated that there is a statistically significant inverse association between the time for LMP-1 regression and the time to develop local failure after radical primary treatment. This association was very
remarkable in the patients with advanced T disease, but insignificant in the group with early T status. The prognostic importance of the time for LMP-1 regression existed only in those with advanced T status. No matter what the underlying reason was, it is evident that the patients with advanced T status certainly had tumors of heterogeneously biologic behaviors even though they had similar extent of primary tumors denoted by the same T classification. Analogously, the inverse association between the time for LMP-1 regression and the time to develop local failure is especially remarkable in the patients with late LMP-1 regression. Only for the patients with delayed LMP-1 regression, the time for LMP-1 regression itself could perform wonderfully as a prognostic factor associated with NPC local control. Although the time for gross tumor to regress assessed by direct nasopharyngoscope was not the very variable that we were interested in, we would like to share our sketchy observations about its correlation with the time for LMP-1 regression. In our preliminary analysis, the correlation of the two times was very heterogeneous in different patients. A large proportion of patients had their LMP-1 disappear earlier than the regression of their nasopharyngeal gross tumors as visualized by nasopharyngoscope. However, there were not a few patients who had persistent presence of LMP-1 for several weeks after their gross tumors had regressed completely without any mucosal irregularities. This disparity between biologic presence and nasopharyngoscopic detection of residual NPC tumors might have resulted from the loss of clonogenicity and reproductive potential in the residual tumors found by nasopharyngoscope. No matter how the correlation between both times was, we had attempted to incorporate the time for gross tumor to regress into the multivariate analysis. The prognostic value of the time for gross tumor to regress failed to be statistically significant, even without the confounding effect of the time for LMP-1 regression. On the contrary, the time for LMP-1 regression remained the most significant prognosti-
The time for LMP-1 to regress in NPC
Fig. 3. Subgroup analysis of the inverse association between the kinetic of latent membrane protein-1 regression and nasopharyngeal carcinoma local control according to T status. Panel (a) shows the subgroup of early T disease. Panel (b) shows the subgroup of advanced T disease. Gray lines indicate the patients with late latent membrane protein-1 regression who required 40 days or more.
cator of NPC local control even after the time for gross tumor to regress was incorporated into the multivariate analysis. In our study, local control of the NPC patients with parapharyngeal space extension (T2b) was the worst (data not shown), as poor as that of the group with T4 disease. The grave local control conferred by parapharyngeal space
● S.-Y. LIN et al.
1345
extension was not confounded by the use of chemotherapy. Namely, among all the patients receiving primary RT alone, those with T2b disease, like T4 patients, still suffered worse local control, as compared with the remaining patients. For T2b patients, a longer time for LMP-1 regression was the only independent prognostic factor predicting a poorer NPC local control, further reinforcing the prognostic importance and value of the time for LMP-1 regression. The poorest local control in T2b patients also reflected the clinical dilemma of irradiating and covering the parapharyngeal disease extension adequately without geometric miss by conventional radiotherapy techniques, while limiting the dose delivered to the adjacent critical normal organs such as the brainstem or spinal cord. After clarifying the prognostic significance of the biologic response represented by the time for LMP-1 regression in NPC local control, it is imperative to incorporate it in our clinical practice and decision. According to the results of this study, supposing that the time for LMP-1 regression calculated took 6 weeks (40 days approximately) after initiation of primary RT, it would be anticipated that a significantly shorter time to develop local failure, and a deemed poorer local control would be almost inevitable if the primary tumor status was T2b or more advanced. This justifies the initiation of boost or dose-escalation radiotherapy using stereotactic radiosurgery (SRS), which will result in enhanced local control and possibly overall survival. From 1997 onward, Goffinet et al. have reported their experience and results on radiosurgical boost for primary NPC patients (30 –33). SRS boost after external-beam radiotherapy provided excellent control in NPC patients with acceptable late toxicities. They even suggested that SRS boost should be considered for all NPC patients. Our study showed clearly that the NPC patients with locally advanced disease and delayed LMP-1 regression (i.e., approximately 6 weeks after initiation of primary RT) could be good and appropriate candidates for SRS boost treatment after external-beam radiotherapy. CONCLUSION Among the various prognostic factors predicting NPC local control analyzed in this study, the time for EBV LMP-1 to disappear after initiation of primary RT was the most statistically significant independent factor that was inversely associated with NPC local control.
REFERENCES 1. Yang T, Ng K, Wang H, et al. Prognostic factors of locoregionally recurrent nasopharyngeal carcinoma—a retrospective review of 182 cases. Am J Clin Oncol 1996;19:337–343. 2. Lee AW, Law S, Foo W, et al. Nasopharyngeal carcinoma: Local control by megavoltage irradiation. Br J Radiol 1993; 66:528 –536. 3. Sanguineti G, Geara FB, Garden AS, et al. Carcinoma of the nasopharynx treated by radiotherapy alone: Determinants of local and regional control. Int J Radiat Oncol Biol Phys 1997;37:985–996.
4. Chang JT, See LC, Liao CT, et al. Locally recurrent nasopharyngeal carcinoma. Radiother Oncol 2000;54:135–142. 5. Pathmanathan R, Prasad U, Sadler R, et al. Clonal proliferations of cells infected with Epstein-Barr virus in preinvasive lesions related to nasopharyngeal carcinoma. N Engl J Med 1995;333:693– 698. 6. Feinmesser R, Miyazaki I, Cheung R, et al. Diagnosis of nasopharyngeal carcinoma by DNA amplification of tissue obtained by fine-needle aspiration. N Engl J Med 1992;326:17–21. 7. Niedobitek G, Agathanggelou A, Nicholls JM. Epstein-Barr
1346
8.
9.
10.
11.
12.
13.
14. 15. 16.
17. 18. 19.
20.
I. J. Radiation Oncology
● Biology ● Physics
virus infection and the pathogenesis of nasopharyngeal carcinoma: Viral gene expression, tumour cell phenotype, and the role of the lymphoid stroma. Semin Cancer Biol 1996;7:165–174. Nicholls JM, Chua D, Chiu PM, et al. The detection of clinically occult nasopharyngeal carcinoma in patients following radiotherapy—an analysis of 69 patients. J Laryngol Otol 1996;110:496 – 499. Lin SY, Tsang NM, Kao SC, et al. Presence of Epstein-Barr virus latent membrane protein 1 gene in the nasopharyngeal swabs from patients with nasopharyngeal carcinoma. Head Neck 2001;23:194 –200. Tsang NM, Chuang CC, Tseng CK, et al. Presence of the latent membrane protein 1 gene in nasopharyngeal swabs from patients with mucosal recurrent nasopharyngeal carcinoma. Cancer 2003;98:2385–2392. Tsang NM, Chang KP, Lin SY, et al. Detection of EpsteinBarr virus-derived latent membrane protein-1 gene in various head and neck cancers: Is it specific for nasopharyngeal carcinoma? Laryngoscope 2003;113:1050 –1054. Hao SP, Tsang NM, Chang KP. Screening nasopharyngeal carcinoma by detection of the latent membrane protein 1 (LMP-1) gene with nasopharyngeal swabs. Cancer 2003;97: 1909 –1913. Hao SP, Tsang NM, Chang KP. Differentiation of recurrent nasopharyngeal carcinoma and skull base osteoradionecrosis by Epstein-Barr virus-derived latent membrane protein-1 gene. Laryngoscope 2001;111:650 – 652. Kwong DL, Nicholls J, Wei WI, et al. The time course of histologic remission after treatment of patients with nasopharyngeal carcinoma. Cancer 1999;85:1446 –1453. Fleming ID, Cooper JS, Henson DE. American Joint Committee on Cancer staging manual. 5th ed. Philadelphia: J.B. Lippincott; 1997. Shanmugaratnam K. Histological typing of tumours of the upper respiratory tract and ear. WHO international histological classification of tumors. 2nd ed. Berlin–New York: Springer-Verlag; 1991. Perez CA, Devineni VR, Marcial-Vega V, et al. Carcinoma of the nasopharynx: Factors affecting prognosis. Int J Radiat Oncol Biol Phys 1992;23:271–280. Heng DM, Wee J, Fong KW, et al. Prognostic factors in 677 patients in Singapore with nondisseminated nasopharyngeal carcinoma. Cancer 1999;86:1912–1920. Chua DT, Sham JS, Kwong DL, et al. Prognostic value of paranasopharyngeal extension of nasopharyngeal carcinoma: A significant factor in local control and distant metastasis. Cancer 1996;78:202–210. Sham J. Prognostic factors in nasopharyngeal carcinoma in-
Volume 63, Number 5, 2005
21.
22.
23. 24. 25.
26.
27.
28. 29. 30.
31.
32.
33.
vestigated by computer tomography—an analysis of 659 patients by Teo et al. Radiother Oncol 1992;25:216 –217. Lu TX, Mai WY, Teh BS, et al. Important prognostic factors in patients with skull base erosion from nasopharyngeal carcinoma after radiotherapy. Int J Radiat Oncol Biol Phys 2001; 51:589 –598. Teo P, Yu P, Lee WY, et al. Significant prognosticators after primary radiotherapy in 903 nondisseminated nasopharyngeal carcinoma evaluated by computer tomography. Int J Radiat Oncol Biol Phys 1996;36:291–304. Lee AW, Chan DK, Fowler JF, et al. Effect of time, dose and fractionation on local control of nasopharyngeal carcinoma. Radiother Oncol 1995;36:24 –31. Chang JT, See LC, Liao CT, et al. Early stage nasopharyngeal carcinoma: Radiotherapy dose and time factors in tumor control. Jpn J Clin Oncol 1998;28:207–213. Lee AW, Poon YF, Foo W, et al. Retrospective analysis of 5,037 patients with nasopharyngeal carcinoma treated during 1976 –1985: Overall survival and patterns of failure. Int J Radiat Oncol Biol Phys 1992;23:261–270. Lin JC, Jan JS, Hsu CY, et al. Phase III study of concurrent chemoradiotherapy versus radiotherapy alone for advanced nasopharyngeal carcinoma: Positive effect on overall and progression-free survival. J Clin Oncol 2003;21:631– 637. Al-Sarraf M, LeBlanc M, Giri PG, et al. Chemoradiotherapy versus radiotherapy in patients with advanced nasopharyngeal cancer: Phase III randomized Intergroup study 0099. J Clin Oncol 1998;16:1310 –1317. Ali H, al-Sarraf M. Chemotherapy in advanced nasopharyngeal cancer. Oncology (Huntingt) 2000;14:1223–1230. Chang JT, See LC, Tang SG, et al. The role of brachytherapy in early-stage nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 1996;36:1019 –1024. Le QT, Tate D, Koong A, et al. Improved local control with stereotactic radiosurgical boost in patients with nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 2003;56:1046 – 1054. Tate DJ, Adler JR Jr, Chang SD, et al. Stereotactic radiosurgical boost following radiotherapy in primary nasopharyngeal carcinoma: Impact on local control. Int J Radiat Oncol Biol Phys 1999;45:915–921. Chang SD, Tate DJ, Goffinet DR, et al. Treatment of nasopharyngeal carcinoma: Stereotactic radiosurgical boost following fractionated radiotherapy. Stereotact Funct Neurosurg 1999;73:64 – 67. Cmelak AJ, Cox RS, Adler JR, et al. Radiosurgery for skull base malignancies and nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 1997;37:997–1003.
doi:10.1016/j.ijrobp.2005.05.051
CLINICAL INVESTIGATION
Head and Neck
THE TIME FRAME OF EPSTEIN-BARR VIRUS LATENT MEMBRANE PROTEIN-1 GENE TO DISAPPEAR IN NASOPHARYNGEAL SWABS AFTER INITIATION OF PRIMARY RADIOTHERAPY IS AN INDEPENDENTLY SIGNIFICANT PROGNOSTIC FACTOR PREDICTING LOCAL CONTROL FOR PATIENTS WITH NASOPHARYNGEAL CARCINOMA SHINN-YN LIN, M.D.,* KAI-PING CHANG, M.D.,†‡ MENG-SHU HSIEH, M.D., M.P.H.,§ SHIR-HWA UENG, M.D.,储 SHENG-PO HAO, M.D.,† CHEN-KAN TSENG, M.D.,* PING-CHING PAI, M.D.,* FU-TI CHANG, M.D.,* MING-HSUI TSAI, M.D.,¶ AND NGAN-MING TSANG, M.D., D.SC.* Departments of *Radiation Oncology, †Otolaryngology, §Family Medicine, 储Anatomic Pathology, and ¶Otolaryngology, Chang Gung Memorial Hospital, Linkou, Taiwan; and ‡Graduate Institute of Clinical Medical Sciences, Chang Gung University, Linkou, Taiwan Purpose: The presence of Epstein-Barr virus latent membrane protein-1 (LMP-1) gene in nasopharyngeal swabs indicates the presence of nasopharyngeal carcinoma (NPC) mucosal tumor cells. This study was undertaken to investigate whether the time taken for LMP-1 to disappear after initiation of primary radiotherapy (RT) was inversely associated with NPC local control. Methods and Materials: During July 1999 and October 2002, there were 127 nondisseminated NPC patients receiving serial examinations of nasopharyngeal swabbing with detection of LMP-1 during the RT course. The time for LMP-1 regression was defined as the number of days after initiation of RT for LMP-1 results to turn negative. The primary outcome was local control, which was represented by freedom from local recurrence. Results: The time for LMP-1 regression showed a statistically significant influence on NPC local control both univariately (p < 0.0001) and multivariately (p ⴝ 0.004). In multivariate analysis, the administration of chemotherapy conferred a significantly more favorable local control (p ⴝ 0.03). Advanced T status (> T2b), overall treatment time of external photon radiotherapy longer than 55 days, and older age showed trends toward being poor prognosticators. The time for LMP-1 regression was very heterogeneous. According to the quartiles of the time for LMP-1 regression, we defined the pattern of LMP-1 regression as late regression if it required 40 days or more. Kaplan-Meier plots indicated that the patients with late regression had a significantly worse local control than those with intermediate or early regression (p ⴝ 0.0129). Conclusion: Among the potential prognostic factors examined in this study, the time for LMP-1 regression was the most independently significant factor that was inversely associated with NPC local control. © 2005 Elsevier Inc. Nasopharyngeal carcinoma, Epstein-Barr virus, Latent membrane protein-1 gene, Radiotherapy, Local control.
INTRODUCTION
Because almost every NPC tumor cell carries the Epstein-Barr virus (EBV) genome (5), detection of EBV genomic deoxyribonucleic acid (DNA) such as latent membrane protein-1 gene (LMP-1) might indicate the presence of NPC (6 – 8). In recent years, polymerase chain reaction (PCR)– based detection of EBV genomic DNA has been employed as a powerful tool to detect the presence of NPC tumor cells with satisfactory sensitivity and excellent specificity (9).
Although radiotherapy is the cornerstone of treatment for nasopharyngeal carcinoma (NPC) and this tumor is much more radiocurable than other head-and-neck malignancies, the cumulative incidence of local recurrence after definitive radiotherapy ranges from 20% to 30% (1–3). Most local relapses after definitive radiotherapy occur within the first 2 years after primary treatment, with a median time of approximately 1.5 years (4).
Memorial Hospital (CMRPG33023, CMRPG33077-II) and the Taiwan National Scientific Council (NSC 89-2314-B-182A-089M08, NSC 93-2314-B-182A-128). Received Mar 4, 2005, and in revised form May 13, 2005. Accepted for publication May 15, 2005.
Reprint requests to: Ngan-Ming Tsang, M.D., D.Sc., Department of Radiation Oncology, Chang Gung Memorial Hospital, 5 Fu-Shin St., Kwei-Shan Hsiang, Taoyuan, Taiwan, R.O.C. Tel: (⫹886) 3-3281200 ext. 2600, 2645; Fax: (⫹886) 3-3280797; E-mail: [email protected]; [email protected] This study was supported in part by grants from the Chang Gung 1339
1340
I. J. Radiation Oncology
● Biology ● Physics
From the evidence demonstrated by prior studies in our laboratory, the presence of EBV LMP-1 gene in nasopharyngeal swab specimens indicates the presence of NPC tumor cells involving at least the nasopharyngeal mucosa; likewise, the disappearance of EBV LMP-1 implies the eradication of mucosal NPC tumor cells (9 –13). In one of our previous studies, it was found that patients with early LMP-1 disease remission during the course of radiotherapy (RT) seemed to have a better outcome in terms of local control. It was suggested that the responsiveness of LMP-1 to irradiation might reflect the aggressiveness and radiosensitivity of the NPC (10). However, it was not clear whether the above association could still exist after taking into consideration other well-established prognostic factors of NPC local control such as pretreatment T classification and the use of chemotherapy. Therefore, we assumed that the time for LMP-1 to disappear after initiation of primary RT represented the aggressiveness of NPC itself and indicated biologic response after treatment, which was highly associated with NPC local control. That is, hypothetically, the faster LMP-1 takes to disappear, the better NPC local control can be achieved after standard therapy. To investigate whether the time taken for LMP-1 to disappear after initiation of primary RT was significantly and inversely associated with NPC local control in a cohort of nondisseminated NPC patients after radical treatment, we performed serial examinations of nasopharyngeal swabbing with detection of EBV LMP-1 after initiation of primary RT. The time taken for LMP-1 to disappear after initiation of primary RT was estimated, and its prognostic value in predicting local control was comprehensively studied.
METHODS AND MATERIALS Study patients Between July 1999 and October 2002, a prospective cohort of NPC patients without distant metastasis at initial diagnosis was ascertained, consisting of 127 patients with NPC who were treated with definitive radiotherapy (cumulative dose of external beam radiotherapy ⬎68 Gy, with or without intracavitary brachytherapy) in the Department of Radiation Oncology in Linkou Chang Gung Memorial Hospital. Because it is well documented that patients with positive nasopharyngeal histology 12 weeks after completion of primary radical radiotherapy must have persistent disease and therefore early initiation of salvage therapy cannot be disputed (14), we have to emphasize that we excluded from our analysis the patients who had persistent presence of EBV LMP-1 in nasopharyngeal swabs even at 3 months after completion of primary RT. Evaluation of treatment response by detecting EBV LMP-1 had to be made within 3 months after completion of primary RT, and LMP-1 persistence indicating residual disease would not be declared until 3 months after completion of primary RT. Provided that the patient had at least two negative LMP-1 results, it was assumed that biologic remission of LMP-1 had been achieved irrespective of whether there was a suspicious lesion found by nasopharyngoscopic examination. In addition, each studied patient had to receive
Volume 63, Number 5, 2005
LMP-1 detection five or more times during the course of RT to prevent the possibility of sampling error. The staging system applied was the 1997 American Joint Committee on Cancer Cancer Staging (15), and histologic typing according to the World Health Organization (WHO) classification (16) was employed. Approximately one-half of the studied patients underwent magnetic resonance imaging (MRI) before primary treatment. If initial MRI was not done or unavailable, computed tomography (CT) imaging was an alternative when determining T classification. All patients in this cohort were treated in a consistent manner by the same well-experienced radiation oncologist. Conventional radiotherapy techniques without the component of intensity-modulated radiotherapy were employed. All recruited subjects underwent serial nasopharyngeal swabbing with PCR amplification to detect LMP-1 genome before or soon after initiation of RT, at least every 2 weeks during the RT course, upon completion of RT, and at a regular 2-month interval during followup. Each specimen of nasopharyngeal swabbing was obtained after the patient had signed the written informed consent.
PCR-based detection of the presence of EBV LMP-1 in nasopharyngeal swabs The nasopharyngeal swabbing techniques, tissue processing, laboratory methods, and PCR amplification have been described in detail and published previously (9, 10, 12). In brief, the patient’s nasopharyngeal wall was thoroughly brushed with a 15-cm long cotton stick, which had been prepared and immersed in phosphatebuffered solution. The specimen was then sent to the laboratory for DNA extraction and purification. We chose to amplify regions of the EBV LMP-1 gene (sense primer BN1, 5=-AGC GAC TCT GCT GGA AAT GAT-3=; antisense primer BN2, 5=-TGA TTA GCT AAG GCA TTC CCA-3=) for identification of viral DNA. For PCR amplification, oligonucleotide primers (sense, BN1; antisense, BN2) were used to detect the presence of LMP-1, which was a 316 base-pair PCR product in the extracted DNA.
The definition and calculation of the time taken for LMP-1 to disappear after primary RT The independent variable that we intended to evaluate in this study was the time taken for LMP-1 to disappear after initiation of primary RT. For convenience and simplicity, we will call it the time for LMP-1 regression in the following text. In fact, it was not possible to identify the exact time when the EBV LMP-1 result had turned negative. We could only surmise that LMP-1 remission occurred between the date of the last positive LMP-1 result and the date of the first negative LMP-1 result. To be practical and realistic, we defined the date when LMP-1 first converted to be negative as a midpoint between the above two dates. Accordingly, the time for LMP-1 regression was calculated as the duration from the starting date of primary RT to the defined date when LMP-1 result converted to negative. Basically, the time for LMP-1 regression is a continuous variable in nature.
Statistical analysis The primary endpoint of interest in this study was NPC local recurrence. The time to develop NPC local recurrence was used as the outcomes measure. It is calculated as the duration from the date of the initiation of primary RT to the date of pathologic diagnosis of local failure, or to the date of censoring due to the following conditions: any causes of death, the occurrence of distant metastasis, and the date of last contact. The occurrence of distant
The time for LMP-1 to regress in NPC
metastasis was regarded as a censor in our study because distant metastasis from NPC or its related death was the most common competitive event for the development of NPC local relapse. Approximately one-fourth of all NPC patients would develop distant metastasis during follow-up. The majority of these NPC patients experienced distant metastasis within 2 years after completion of primary RT, with a median time of 17 months. Their median survival was only 10 months after distant metastasis (our unpublished data). A Cox proportional hazards regression model was performed to evaluate the impact of the time for LMP-1 regression and other potential prognostic factors on the local tumor control of NPC patients after definitive RT. The time for LMP-1 regression was treated as a continuous variable in the Cox proportional hazards regression model both in univariate and multivariate analyses. Local control rates, namely cumulative local recurrence-free probabilities, were calculated by the Kaplan-Meier method. A log–rank test was employed to compare the difference between the local control rates according to various potential prognostic factors. To evaluate the prognostic importance of the kinetic or pattern of LMP-1 regression, all studied patients were divided into three subgroups with different time frames of regression according to the time for LMP-1 regression. The principal exposure variable in our analysis, the time for LMP-1 regression, was then converted to a categorical and ordinal variable according to its quartiles. The related results and figures will be presented. The other potential variables of prognostic value in predicting NPC local control were also examined univariately or multivariately, including gender, age at initial diagnosis, primary tumor status (T2b-T4 vs. earlier than T2b), WHO histologic typing, the addition of intracavitary brachytherapy, whether there was an administration of chemotherapy, and overall treatment time of external photon radiotherapy. In general, a p value ⱕ0.05 was considered statistically significant.
RESULTS The median follow-up duration was 33.9 months, ranging from 6.4 to 58.5 months for all patients, and 36.3 months, ranging from 20.7 to 58.6 months, for survivors. The study cohort consisted of 94 men and 33 women. The age of primary NPC diagnosis ranged from 19.8 to 87.1 years old with a median age of 48.7 years. There were 45 patients (35.4%) whose primary tumor was staged as T2 at initial diagnosis. Thirty of these patients were further divided into T2a or T2b based on the status of parapharyngeal space extension if the reports of MRI or CT provided available information. According to WHO histologic classification, 48 cases (37.8%) were diagnosed as WHO type III, which denoted undifferentiated carcinoma; 72 cases (56.7%) were determined to be WHO type II, representing nonkeratinizing or poorly differentiated carcinoma; and only 3 patients’ primary tumors were interpreted as squamous cell carcinoma. In terms of treatment, there were 36 patients (28.4%) receiving at least one course of chemotherapy predominantly in a concurrent setting. Except for 4 patients, the majority of patients received intracavitary brachytherapy during the RT course. Overall treatment time of external photon radiotherapy, regardless of electron boost to neck
● S.-Y. LIN et al.
1341
Table 1. Demographics of the 127 investigated patients Patient characteristics Gender Male Female Age* Pathology–WHO typing I II III NE Primary tumor classification (T)† T1 T2 T2a ⫹ Unspecified T2b‡ T3 T4 Neck status (N)† N0–N1 N2–N3 Stage grouping† I II III IV Brachytherapy Yes No Chemotherapy Yes No Overall Tx time* The time taken for LMP-1 to disappear after initiation of primary RT* Pattern of LMP-1 regression Early (⬍16 days) Intermediate Late (40 days or more) The time taken for gross tumor to regress after initiation of primary RT (assessed by endoscope)*
No. of patients (%) 94 (74) 33 (26) Median
48.7 years of age
3 (2.36) 72 (56.69) 48 (37.80) 4 (3.15) 33 (25.98) 45 (35.43) 17 ⫹ 15 13 24 (18.90) 25 (19.69) 87 (68.5) 40 (31.5) 15 (11.8) 43 (33.9) 39 (30.7) 30 (23.6) 123 (96.85) 4 (3.15) 36 (28.35) 91 (71.65) Median 25% quartile Median 75% quartile 28 (22) 67 (52.8) 32 (25.2) Median
55 days 16 days 24.5 days 40 days
40 days
Abbreviations: WHO ⫽ World Health Organization; NE ⫽ not evaluable; Overall Tx time ⫽ overall treatment time of external photon therapy; AJCC ⫽ American Joint Committee on Cancer; RT ⫽ radiotherapy; LMP-1 ⫽ latent membrane protein-1. * It was regarded as a continuous variable. † The 1997 TNM staging system of AJCC was employed. ‡ T2b denotes parapharyngeal extension, which usually was evidenced by magnetic resonance imaging.
lymphatics, ranged from 46 days to 78 days, with a median of 55 days. The patient characteristics, disease-related factors, treatment characteristics, and the time for LMP-1 regression are shown in Table 1. By November 22, 2004, a total of 25 patients had developed local recurrence. The median time to develop NPC local failure after primary RT was 16.2 months, with a range between 7.3 months and 38.2 months.
1342
I. J. Radiation Oncology
● Biology ● Physics
The time for LMP-1 regression ranged from less than 1 week to 116.5 days after the initiation of primary RT (25% quartile, 16 days; median, 24.5 days; 75% quartile, 40 days). Here, according to the quartiles of the time for LMP-1 regression (16 and 40 days, respectively), we defined the kinetic of LMP-1 regression as being early regression if it required shorter than 16 days, late regression if it took 40 days or longer, and intermediate regression if LMP-1 disappeared between 16 and 40 days after starting RT. The local control rates according to the pattern of LMP-1 regression are presented in Fig. 1. It shows that the patients with delayed LMP-1 regression requiring 40 days or more after starting RT had a significantly worse local control than those with intermediate or early LMP-1 regression (log– rank test, p ⫽ 0.035). Their 2-year actuarial local failurefree rates were 67.41%, 84.15%, and 92.52%, respectively, according to the pattern of LMP-1 regression (Table 2). It was also observed that the difference in NPC local control existed between the patients with late regression and the remaining patients whose LMP-1 regression occurred within 40 days after initiation of primary RT (log–rank test, p ⫽ 0.0129). Furthermore, the cumulative local recurrence-free probabilities according to primary tumor status, WHO subtype, whether chemotherapy was administered, and overall treatment time of external photon radiotherapy were calculated. Figure 2 reveals that T2b disease or more advanced primary tumor extension conferred a significantly worse local control (log–rank test, p ⫽ 0.043). The patients receiving chemotherapy seemed to have a more favorable freedom from local failure than those who did not, but the difference failed to be statistically significant (p ⫽ 0.289). Likewise, the patients whose overall treatment time of external photon radiotherapy was more protracted than 55 days appeared to
Fig. 1. Freedom from local failure, according to the kinetic of latent membrane protein-1 regression. Early regression denotes that the time for latent membrane protein-1 to disappear takes less than 16 days after initiation of primary radiotherapy. Late regression indicates this specific time requires 40 days or more. Intermediate regression implies that remission of latent membrane protein-1 occurs between 16 and 40 days after starting radiotherapy.
Volume 63, Number 5, 2005
Table 2. The pattern of LMP-1 regression* according to the quartiles of the time for LMP-1 regression† Kinetic of LMP-1 regression
2-year local recurrencefree rate‡
HR (95% CI; [p value])
Early Intermediate Late
92.52% 84.15% 67.41%
1 1.709 (0.476–6.132; [p ⫽ 0.4]) 2.215 (0.96–5.111; [p ⫽ 0.06])
Abbreviations: HR ⫽ hazards ratio; 95% CI ⫽ 95% confidence interval; LMP-1 ⫽ latent membrane protein-1. * It was treated as a discrete variable. † It was treated as a continuous variable. ‡ Their Kaplan-Meier plots were compared by log–rank test. The p value was 0.035.
suffer a higher risk of local failure than those whose RT course could be completed within 8 weeks, but the difference did not reach statistical significance (p ⫽ 0.103). Table 3 shows unadjusted and adjusted hazard ratios for the time for LMP-1 regression and other various prognostic factors predicting NPC local recurrence. In the univariate analysis, longer time for LMP-1 regression and advanced T disease above T2a were statistically significant poor prognostic factors. All parameters concerned in the univariate analysis were also included in the multivariate analysis. They were the time for LMP-1 regression (treated as a continuous variable), age (treated as a continuous variable), gender (male/female), T status (T2b-T4/T1-T2a), N status (N2-3/N0-1), the use of chemotherapy (ever/none), brachytherapy (yes/no), and overall treatment time of external photon radiotherapy (⬎55 days/ⱕ55 days). The independently significant prognostic factors for NPC local recurrence identified by multivariate analysis were the time for LMP-1 regression and the use of chemotherapy. Advanced T status, overall treatment time of external photon radiotherapy longer than 55 days, and older age showed trends toward
Fig. 2. Freedom from local failure, according to extent of the primary tumor. Tumors of T1, T2a, and unspecified T2 diseases were grouped as early T status group. Tumors of T3 and T4 diseases were combined together with those of T2b classification to comprise the advanced T status group.
The time for LMP-1 to regress in NPC
● S.-Y. LIN et al.
1343
Table 3. Cox proportional hazards model analysis on time to develop local recurrence Univariate
Multivariate
Variable
p value
Crude HR (95% CI)
p value
Adjusted HR (95% CI)
The time of LMP-1 regression* T status (T1⬃2a vs. T2b and higher) Age* Sex (male vs. female) N status (N2 ⫹ N3 vs. N0 ⫹ N1) Chemotherapy (yes vs. no) Brachytherapy (yes vs. no) Overall Tx time (⬎ 55 days vs. shorter period)
⬍0.0001§ 0.0491‡ 0.0704† 0.8775 0.6792 0.294 0.9011 0.1085
1.028 (1.016–1.041) 2.327 (1.003–5.396) 1.032 (0.997–1.067) 0.934 (0.390–2.237) 1.194 (0.515–2.768) 0.592 (0.222–1.577) 0.881 (0.119–6.524) 1.904 (0.867–4.181)
0.0041§ 0.0554† 0.0968† 0.8959 0.2444 0.0316‡ 0.8732 0.0768†
1.021 (1.007–1.036) 2.66 (0.978–7.235) 1.03 (0.995–1.066) 0.938 (0.357–2.464) 1.767 (0.678–4.609) 0.291 (0.095–0.897) 1.183 (0.150–9.314) 2.091 (0.924–4.731)
Abbreviations: HR ⫽ hazards ratio; 95% CI ⫽ 95% confidence interval; Overall Tx time ⫽ overall treatment time of external photon therapy; LMP-1 ⫽ latent membrane protein-1. * Treated as a continuous variable. † p ⬍ 0.1; ‡ p ⬍ 0.05; § p ⬍ 0.01.
being prognostic factors. After adjusting for these five significant prognostic factors identified in the multivariate analysis, the time for LMP-1 regression still remained the most significant independent prognostic factor that was inversely associated with NPC local control (Table 4). To further clarify the inverse association between the time for LMP-1 regression and the time to develop local failure, we also performed subgroup analysis according to either T status or the kinetic of LMP-1 regression (Table 5). In consideration of the small number of patients and events in each T subcategory, the patients with primary diseases earlier than T2b were combined together to form the group with early T status, and the patients with T2b-T4 diseases were grouped as those with advanced T diseases. The prognostic value of the time for LMP-1 regression was very remarkable in the group of advanced T status (p ⬍ 0.0001, Cox regression analysis). However, the time for LMP-1 regression was not associated with NPC local control in the group of early T disease (p ⫽ 0.7371, Cox regression analysis). This observation is also expressed as KaplanMeier plots in Fig. 3, in which the patients with early LMP-1 regression were grouped together with those with intermediate LMP-1 regression. Table 4. Cox proportional hazards model analysis on time to develop local recurrence when adjusting for those five significant factors identified in the multivariate analysis Variable
p value
HR (95% CI)
The time of LMP-1 regression* T status (T1⬃2a vs. T2b and higher) Age* Chemotherapy (ever vs. none) Overall Tx time (⬎55 days vs. shorter period)
0.0072§ 0.0405‡
1.019 (1.005–1.034) 2.818 (1.046–7.594)
0.0618† 0.0478‡ 0.0897†
1.032 (0.998–1.066) 0.327 (0.108–0.989) 2.019 (0.897–4.546)
Abbreviations: HR ⫽ hazards ratio; 95% CI ⫽ 95% confidence interval; Overall Tx time ⫽ overall treatment time of external photon therapy; LMP-1 ⫽ latent membrane protein-1. * Treated as a continuous variable. † p ⬍ 0.1; ‡ p ⬍ 0.05; § p ⬍ 0.01.
The remaining part of Table 5 demonstrates that the prognostic value of the time for LMP-1 regression existed in the subgroup with late LMP-1 regression (p ⫽ 0.0016) and was still very remarkable after controlling for T status (p ⫽ 0.0074). Its prognostic value failed to reach statistical significance when the kinetic of LMP-1 regression was early or intermediate. These results further consolidated the findings exhibited in Fig. 1 and Table 2, in which the statistically significant difference in local recurrence-free rates exists primarily between the patients with delayed LMP-1 regression requiring 40 days or more and those whose regression occurred within 40 days after starting RT. DISCUSSION Numerous studies have evaluated the prognostic value of various parameters in NPC local control. The most wellestablished prognostic factors are primary tumor status (T classification) at initial diagnosis (1, 3, 17) and various factors comprising the T classification, such as parapharyngeal extension (18, 19), skull base invasion (20 –22), or cranial nerve involvement (3, 17, 20, 22). Other factors that have been reported include patient age (1, 17, 22), histologic type (3, 17), RT dose to the nasopharynx (2, 17, 23), overall treatment time of radiotherapy (24), treatment era (2, 25), the administration of chemotherapy (26 –28), the addition of intracavitary brachytherapy (29), and some clinical treatment response factors such as recovery of headache or cranial nerve palsy (21). To the best of our knowledge, no biologic factors associated with treatment response have been previously reported as prognosticators. In 1999, Kwong et al. (14) reported their study evaluating the prognostic significance of delayed histologic remission, which could be viewed as a pathologic response factor in NPC patients. However, they did not find any prognostic importance of delayed histologic remission. The discrepancy between the above Hong Kong study (histologic remission) and our study (the time for LMP-1 regression indicating biologic response) suggested that the tumor cells found in posttreatment biopsies might
1344
I. J. Radiation Oncology
● Biology ● Physics
Volume 63, Number 5, 2005
Table 5. Subgroup analysis on the prognostic value of the time for LMP-1 regression according to T status or the speed of LMP-1 regression Stratification T status Early T (earlier than T2b) Advanced T (T2b at least) The speed of LMP-1 regression Early regression (⬍16 days)
Exposure variable
p value
HR (95% CI)
The time for LMP-1 regression The time for LMP-1 regression
0.7371 ⬍0.0001*
0.991 (0.941–1.004) 1.03 (1.016–1.045)
The time for LMP-1 regression
0.2966 0.3125† 0.2966 0.8798 0.887† 0.872 0.0016* 0.0074† 0.137
Intermediate regression
T status‡ The time for LMP-1 regression
Late regression (ⱖ40 days)
T status‡ The time for LMP-1 regression T status‡
1.286 (0.802–2.063) 1.257 (0.806–1.960) 1.286 (0.802–2.063) 1.007 (0.923–1.099) 1.006 (0.922–1.098) 1.103 (0.336–3.164) 1.036 (1.013–1.058) 1.031 (1.008–1.054) 4.777 (0.608–37.524)
Abbreviations: HR ⫽ hazards ratio; 95% CI ⫽ 95% confidence interval. * p ⬍ 0.01. † Adjusted for T status; ‡ T2b-T4 vs. T1-T2a.
have lost their clonogenicity and reproductive potential if the detection of EBV LMP-1 at the same time revealed negative results. The time for LMP-1 regression could be regarded as an indicator of biologic response, which reflects both effectiveness of treatment and biologic aggressiveness of the NPC tumor itself. Provided that there is similar tumor burden or volume, it would be justifiable to suggest that the longer it takes for LMP-1 to disappear, the more aggressive the tumor is and the more intensive treatment the disease demands. In our analysis, the time taken for LMP-1 to disappear after initiation of primary RT was the most significant independent prognostic factor that was inversely associated with NPC local control. The time for LMP-1 regression did show a statistically significant influence on the time to develop local failure during the follow-up period. The time is so powerful as a prognosticator predicting NPC local control that it still retained its prognostic significance after controlling for other well-documented prognostic factors such as pretreatment T status, the use of chemotherapy, and overall treatment time of external photon therapy. In addition to the above findings specific to the purpose of this study, we had several other observations that supported the past related reports (1, 22, 24, 26, 27). For example, it was confirmed in our study that the administration of chemotherapy conferred a more favorable prognosis. Besides, the older the patient was, the higher the risk for NPC local failure. When the course of external photon radiotherapy was more protracted (i.e., longer than 8 weeks), NPC local control was also significantly poorer. Unlike some previous reports (29), our study failed to demonstrate the benefit in local control enhanced by the addition of intracavitary brachytherapy. In our study, it is clearly demonstrated that there is a statistically significant inverse association between the time for LMP-1 regression and the time to develop local failure after radical primary treatment. This association was very
remarkable in the patients with advanced T disease, but insignificant in the group with early T status. The prognostic importance of the time for LMP-1 regression existed only in those with advanced T status. No matter what the underlying reason was, it is evident that the patients with advanced T status certainly had tumors of heterogeneously biologic behaviors even though they had similar extent of primary tumors denoted by the same T classification. Analogously, the inverse association between the time for LMP-1 regression and the time to develop local failure is especially remarkable in the patients with late LMP-1 regression. Only for the patients with delayed LMP-1 regression, the time for LMP-1 regression itself could perform wonderfully as a prognostic factor associated with NPC local control. Although the time for gross tumor to regress assessed by direct nasopharyngoscope was not the very variable that we were interested in, we would like to share our sketchy observations about its correlation with the time for LMP-1 regression. In our preliminary analysis, the correlation of the two times was very heterogeneous in different patients. A large proportion of patients had their LMP-1 disappear earlier than the regression of their nasopharyngeal gross tumors as visualized by nasopharyngoscope. However, there were not a few patients who had persistent presence of LMP-1 for several weeks after their gross tumors had regressed completely without any mucosal irregularities. This disparity between biologic presence and nasopharyngoscopic detection of residual NPC tumors might have resulted from the loss of clonogenicity and reproductive potential in the residual tumors found by nasopharyngoscope. No matter how the correlation between both times was, we had attempted to incorporate the time for gross tumor to regress into the multivariate analysis. The prognostic value of the time for gross tumor to regress failed to be statistically significant, even without the confounding effect of the time for LMP-1 regression. On the contrary, the time for LMP-1 regression remained the most significant prognosti-
The time for LMP-1 to regress in NPC
Fig. 3. Subgroup analysis of the inverse association between the kinetic of latent membrane protein-1 regression and nasopharyngeal carcinoma local control according to T status. Panel (a) shows the subgroup of early T disease. Panel (b) shows the subgroup of advanced T disease. Gray lines indicate the patients with late latent membrane protein-1 regression who required 40 days or more.
cator of NPC local control even after the time for gross tumor to regress was incorporated into the multivariate analysis. In our study, local control of the NPC patients with parapharyngeal space extension (T2b) was the worst (data not shown), as poor as that of the group with T4 disease. The grave local control conferred by parapharyngeal space
● S.-Y. LIN et al.
1345
extension was not confounded by the use of chemotherapy. Namely, among all the patients receiving primary RT alone, those with T2b disease, like T4 patients, still suffered worse local control, as compared with the remaining patients. For T2b patients, a longer time for LMP-1 regression was the only independent prognostic factor predicting a poorer NPC local control, further reinforcing the prognostic importance and value of the time for LMP-1 regression. The poorest local control in T2b patients also reflected the clinical dilemma of irradiating and covering the parapharyngeal disease extension adequately without geometric miss by conventional radiotherapy techniques, while limiting the dose delivered to the adjacent critical normal organs such as the brainstem or spinal cord. After clarifying the prognostic significance of the biologic response represented by the time for LMP-1 regression in NPC local control, it is imperative to incorporate it in our clinical practice and decision. According to the results of this study, supposing that the time for LMP-1 regression calculated took 6 weeks (40 days approximately) after initiation of primary RT, it would be anticipated that a significantly shorter time to develop local failure, and a deemed poorer local control would be almost inevitable if the primary tumor status was T2b or more advanced. This justifies the initiation of boost or dose-escalation radiotherapy using stereotactic radiosurgery (SRS), which will result in enhanced local control and possibly overall survival. From 1997 onward, Goffinet et al. have reported their experience and results on radiosurgical boost for primary NPC patients (30 –33). SRS boost after external-beam radiotherapy provided excellent control in NPC patients with acceptable late toxicities. They even suggested that SRS boost should be considered for all NPC patients. Our study showed clearly that the NPC patients with locally advanced disease and delayed LMP-1 regression (i.e., approximately 6 weeks after initiation of primary RT) could be good and appropriate candidates for SRS boost treatment after external-beam radiotherapy. CONCLUSION Among the various prognostic factors predicting NPC local control analyzed in this study, the time for EBV LMP-1 to disappear after initiation of primary RT was the most statistically significant independent factor that was inversely associated with NPC local control.
REFERENCES 1. Yang T, Ng K, Wang H, et al. Prognostic factors of locoregionally recurrent nasopharyngeal carcinoma—a retrospective review of 182 cases. Am J Clin Oncol 1996;19:337–343. 2. Lee AW, Law S, Foo W, et al. Nasopharyngeal carcinoma: Local control by megavoltage irradiation. Br J Radiol 1993; 66:528 –536. 3. Sanguineti G, Geara FB, Garden AS, et al. Carcinoma of the nasopharynx treated by radiotherapy alone: Determinants of local and regional control. Int J Radiat Oncol Biol Phys 1997;37:985–996.
4. Chang JT, See LC, Liao CT, et al. Locally recurrent nasopharyngeal carcinoma. Radiother Oncol 2000;54:135–142. 5. Pathmanathan R, Prasad U, Sadler R, et al. Clonal proliferations of cells infected with Epstein-Barr virus in preinvasive lesions related to nasopharyngeal carcinoma. N Engl J Med 1995;333:693– 698. 6. Feinmesser R, Miyazaki I, Cheung R, et al. Diagnosis of nasopharyngeal carcinoma by DNA amplification of tissue obtained by fine-needle aspiration. N Engl J Med 1992;326:17–21. 7. Niedobitek G, Agathanggelou A, Nicholls JM. Epstein-Barr
1346
8.
9.
10.
11.
12.
13.
14. 15. 16.
17. 18. 19.
20.
I. J. Radiation Oncology
● Biology ● Physics
virus infection and the pathogenesis of nasopharyngeal carcinoma: Viral gene expression, tumour cell phenotype, and the role of the lymphoid stroma. Semin Cancer Biol 1996;7:165–174. Nicholls JM, Chua D, Chiu PM, et al. The detection of clinically occult nasopharyngeal carcinoma in patients following radiotherapy—an analysis of 69 patients. J Laryngol Otol 1996;110:496 – 499. Lin SY, Tsang NM, Kao SC, et al. Presence of Epstein-Barr virus latent membrane protein 1 gene in the nasopharyngeal swabs from patients with nasopharyngeal carcinoma. Head Neck 2001;23:194 –200. Tsang NM, Chuang CC, Tseng CK, et al. Presence of the latent membrane protein 1 gene in nasopharyngeal swabs from patients with mucosal recurrent nasopharyngeal carcinoma. Cancer 2003;98:2385–2392. Tsang NM, Chang KP, Lin SY, et al. Detection of EpsteinBarr virus-derived latent membrane protein-1 gene in various head and neck cancers: Is it specific for nasopharyngeal carcinoma? Laryngoscope 2003;113:1050 –1054. Hao SP, Tsang NM, Chang KP. Screening nasopharyngeal carcinoma by detection of the latent membrane protein 1 (LMP-1) gene with nasopharyngeal swabs. Cancer 2003;97: 1909 –1913. Hao SP, Tsang NM, Chang KP. Differentiation of recurrent nasopharyngeal carcinoma and skull base osteoradionecrosis by Epstein-Barr virus-derived latent membrane protein-1 gene. Laryngoscope 2001;111:650 – 652. Kwong DL, Nicholls J, Wei WI, et al. The time course of histologic remission after treatment of patients with nasopharyngeal carcinoma. Cancer 1999;85:1446 –1453. Fleming ID, Cooper JS, Henson DE. American Joint Committee on Cancer staging manual. 5th ed. Philadelphia: J.B. Lippincott; 1997. Shanmugaratnam K. Histological typing of tumours of the upper respiratory tract and ear. WHO international histological classification of tumors. 2nd ed. Berlin–New York: Springer-Verlag; 1991. Perez CA, Devineni VR, Marcial-Vega V, et al. Carcinoma of the nasopharynx: Factors affecting prognosis. Int J Radiat Oncol Biol Phys 1992;23:271–280. Heng DM, Wee J, Fong KW, et al. Prognostic factors in 677 patients in Singapore with nondisseminated nasopharyngeal carcinoma. Cancer 1999;86:1912–1920. Chua DT, Sham JS, Kwong DL, et al. Prognostic value of paranasopharyngeal extension of nasopharyngeal carcinoma: A significant factor in local control and distant metastasis. Cancer 1996;78:202–210. Sham J. Prognostic factors in nasopharyngeal carcinoma in-
Volume 63, Number 5, 2005
21.
22.
23. 24. 25.
26.
27.
28. 29. 30.
31.
32.
33.
vestigated by computer tomography—an analysis of 659 patients by Teo et al. Radiother Oncol 1992;25:216 –217. Lu TX, Mai WY, Teh BS, et al. Important prognostic factors in patients with skull base erosion from nasopharyngeal carcinoma after radiotherapy. Int J Radiat Oncol Biol Phys 2001; 51:589 –598. Teo P, Yu P, Lee WY, et al. Significant prognosticators after primary radiotherapy in 903 nondisseminated nasopharyngeal carcinoma evaluated by computer tomography. Int J Radiat Oncol Biol Phys 1996;36:291–304. Lee AW, Chan DK, Fowler JF, et al. Effect of time, dose and fractionation on local control of nasopharyngeal carcinoma. Radiother Oncol 1995;36:24 –31. Chang JT, See LC, Liao CT, et al. Early stage nasopharyngeal carcinoma: Radiotherapy dose and time factors in tumor control. Jpn J Clin Oncol 1998;28:207–213. Lee AW, Poon YF, Foo W, et al. Retrospective analysis of 5,037 patients with nasopharyngeal carcinoma treated during 1976 –1985: Overall survival and patterns of failure. Int J Radiat Oncol Biol Phys 1992;23:261–270. Lin JC, Jan JS, Hsu CY, et al. Phase III study of concurrent chemoradiotherapy versus radiotherapy alone for advanced nasopharyngeal carcinoma: Positive effect on overall and progression-free survival. J Clin Oncol 2003;21:631– 637. Al-Sarraf M, LeBlanc M, Giri PG, et al. Chemoradiotherapy versus radiotherapy in patients with advanced nasopharyngeal cancer: Phase III randomized Intergroup study 0099. J Clin Oncol 1998;16:1310 –1317. Ali H, al-Sarraf M. Chemotherapy in advanced nasopharyngeal cancer. Oncology (Huntingt) 2000;14:1223–1230. Chang JT, See LC, Tang SG, et al. The role of brachytherapy in early-stage nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 1996;36:1019 –1024. Le QT, Tate D, Koong A, et al. Improved local control with stereotactic radiosurgical boost in patients with nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 2003;56:1046 – 1054. Tate DJ, Adler JR Jr, Chang SD, et al. Stereotactic radiosurgical boost following radiotherapy in primary nasopharyngeal carcinoma: Impact on local control. Int J Radiat Oncol Biol Phys 1999;45:915–921. Chang SD, Tate DJ, Goffinet DR, et al. Treatment of nasopharyngeal carcinoma: Stereotactic radiosurgical boost following fractionated radiotherapy. Stereotact Funct Neurosurg 1999;73:64 – 67. Cmelak AJ, Cox RS, Adler JR, et al. Radiosurgery for skull base malignancies and nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 1997;37:997–1003.