All patients included in this analysis were diagnosed with nasopharyngeal carcinoma, and were primary treated at our institution between May 2009 and March 2014. Patients were included if they met the following criteria: 1) age>16 and <80 years; 2) pathology confirmed nasopharyngeal carcinoma; 3) PET/CT scans performed four weeks prior to treatment; and 4) stage III and IV according to the 8th edition American Joint Committee on Cancer (AJCC) guidelines. We excluded patients who did not receive radiotherapy, patients with a history of other malignancies, and patients with a follow-up less than 5 years. This study was approved by the Ethical Committee at Fudan University Shanghai Cancer Center (FUSCC), and informed consent was obtained from all enrolled patients.
In our institution database of primary nasopharyngeal carcinoma patients, 171 patients were eligible for this study. Data included demographics, tumor characteristics, and treatment outcomes were retrospectively collected from the medical records. All patients were staged according to the 8th edition AJCC guidelines. Patients with metastatic disease at the time of initial diagnosis (M1 stage) were also included. Epstein-Barr virus (EBV) status was determined by testing plasma anti-EBV IgA antibodies using ELISA. EBV status was available for 62% of the patients.
Treatment and follow-up
All patients received Intensity-Modulated Radiation Therapy (IMRT) for a cumulative dose of 66 Gy (2.2 Gy/fraction/day) in 30 fractions for T1 and T2 disease or 70.4 Gy (2.2 Gy/fraction/day) in 32 fractions for T3 and T4 lesion. According to the tumor stage and other clinical characteristics, concomitant chemotherapy or targeted therapy was also performed. Induction chemotherapy was consisted of docetaxel 75 mg/m2 on day 1, cisplatin 75 mg/m2 on day 1, and 5-Fu 500 mg/m2/d continuously on day 1-5. With respect to concurrent chemoradiotherapy (CCRT), cisplatin 40 mg/m2 was used weekly during radiation. As for adjuvant chemotherapy, cisplatin 40 mg/m2 on day 1-3, and docetaxel 75 mg/m2 on day 4 after radiation. Cetuximab was used as a targeted drug with an initial dose of 400 mg/m2 followed by 250 mg/m2 weekly for the duration of radiotherapy. Individual treatment protocol was approved by the Nasopharyngeal Carcinoma multidisciplinary team in our institution after the consultation.
After completion of radiotherapy, physical examination, imaging examination, and nasopharyngoscopy were performed every 3 months in the first 2 years, then every 6 months in the third to fifth year and once a year thereafter. Local recurrence and distant metastasis were proven by pathologic evidence or radiologic evidence. We identified treatment response according to RECIST 1.1. The following endpoints were evaluated: progression-free survival (PFS) and loco-regional control (LRC). PFS was calculated from the first day of RT to the date of disease progression or was censored at the last follow-up date. LRC was measured from the first day of RT to the date of first recurrence in the primary tumor and/or lymph node.
PET/CT scanning procedure
18F-FDG was produced automatically by cyclotron (Siemens CTI RDS Eclips ST, Knoxville, Tennessee, USA) using Explora FDG4 module in our center. Radiochemical purity was over 95%. Before the 18F-FDG PET/CT, all patients were requested to fast at least 4 h. Venous blood glucose levels were maintained under 10 mmol/L. After injecting with 7.4 MBq/kg 18F-FDG, patients were kept lying comfortably in a quiet, dimly lit room for approximately 1 hour prior to scanning. Images were obtained on a Siemens biograph 16HR PET/CT scanner (Knoxville, Tennessee, USA). Transaxial intrinsic spatial resolution was 4.1 mm (full-width at half-maximum) in the center of the field of view. Data acquisition procedure was as follows: CT scanning was first performed, from the proximal thighs to head, with 120 kV, 80 ~ 250 mA, pitch 3.6, rotation time 0.5. Immediately after CT scanning, a PET emission scan that covered the identical transverse field of view was obtained. Acquisition time was 2 ~ 3 min per table position. PET image data sets were reconstructed iteratively using an ordered-subset expectation maximization iterative reconstruction (OSEM) by applying CT data for attenuation correction. Fusion images were reviewed and manipulated on a multimodality computer platform (Syngo, Siemens, Knoxville, Tennessee, USA). Two experienced radiologists analyzed and interpreted the images independently.
For quantitative analysis, maximum and mean of standardized uptake value (SUV) normalized to body weight were manually computed for primary tumor (SUVmax-T, SUVmean-T) and maximal neck lymph node (SUVmax-N, SUVmean-N) by drawing a region of interest (ROI). Meanwhile, metabolic tumor volume (MTV) was recorded at the absolute SUV threshold of 2.5 and the relative SUVmax threshold of 70%. Total lesion glucose (TLG) was calculated according to the formula: TLG = SUVmean × MTV. To evaluate intratumoral heterogeneity, heterogeneity index (HI) was obtained by dividing SUVmax by SUVmean for primary lesion and nodal disease.
The entire cohort was divided into a training cohort (n = 101) and a validation cohort (n = 70). The following parameters were assessed to identify predictors of recurrence: age, gender, EBV statue, histology, tumor staging, treatment, and PET parameters. Frequencies with percentages were used to describe categorical variables while medians with ranges were used for continuous characteristics. The differences of these parameters between these two cohorts were calculated. Mann-Whitney tests were used to compare the continuous variables, and Fisher's exact tests were used to compare the categorical data. The survival analyses were performed using the Kaplan-Meier method, and a two-sided log-rank test was used to compare groups.
The predictive model was constructed as suggested in the TRIPPOD statement. To develop a robust and well-calibrated nomogram predicting the risk of recurrence, a cox regression model was built using a training cohort of 101 patients and validated with a cohort of 70 patients. Firstly, a univariate Cox analysis was performed to assess relationships between risk factors and recurrence using the training cohort. The Harrell’s C-index was computed for the factor with significance of P < 0.05. Then, predictors were determined using the factors with significance of P < 0.1 and with highest C-index after univariate analyses, and the multivariate Cox regression model was developed with backward elimination. The Harrell’s C-index, the constant, and the standardized coefficient of the prognostic model were calculated. Lastly, based on this model, a nomogram was built to predict the individual conditional risk of 5-year recurrence.
To estimate the accuracy of the model, internal validation was performed by bootstrap algorithm, in which 1000 replications were constructed randomly, and the adjusted C-index and corresponding 95% confidence intervals were also computed. The calibration plot comparing the nomogram predicted versus observed probability was used to assess the accuracy. To test for generalizability, the developed nomogram derived from the training cohort was tested with the validation cohort. Three prognostic groups were created by categorizing the prognostic index computed from the model at the 55th and 89th percentiles using the X-tile software. These groups were called low-, intermediate- and high-risk groups. The same cut-off was applied in the validation cohort. All statistical tests were two-sided, and the p < 0.05 was considered statistically significant. All analyses were performed using SPSS (version 22.0; IBM Inc., New York, USA) and R version 3.5.3 (http://cran.r-project.org/mirrors.html).