Prognostic variables for temporal lobe injury after intensity modulated‐radiotherapy of nasopharyngeal carcinoma

Abstract To determine predictive factors for temporal lobe injury (TLI) in nasopharyngeal carcinoma patient (NPC) treated with intensity‐modulated radiation therapy (IMRT). A total of 695 NPC cases treated with IMRT were retrospectively analyzed. TLI was diagnosed on MRI images. Volume‐dose histograms for 870 evaluable temporal lobes were analyzed, and the predictive factors for the occurrence of TLI was evaluated. Receiver operating characteristic curve (ROC) and Logistic regression analysis was used to determine volume‐dose parameters that predict temporal lobe injury (TLI). Univariate and multivariate analysis were used to analyze the predictive factors for TLI. The radiation dose‐tolerance model of temporal lobe was calculated by logistic dose‐response model. The median follow‐up time was 73 months. A total of 8.5% patients were diagnosed with TLI. Among all the volume‐dose parameters, logistic regression model showed D2cc (the dose Gray delivered to 2 cubic centimeter volume) was an only independent predictive factor. Multivariate analysis showed D2cc of temporal lobe, fraction size of prescription, T stage, and chemotherapy were the independent predictive factors for TLI. Logistic dose‐response model has indicated the TD 5/5 and TD 50/5 of D2cc are 60.3 Gy and 76.9 Gy, respectively. D2cc of temporal lobe, fraction size of prescription, T stage, and chemotherapy were the possible independent predictive factors for TLI after IMRT of NPC. Biologic effective doses (TD 5/5 and TD 50/5) of D2cc are considered to prevent TLI.


Introduction
Nasopharyngeal carcinoma (NPC) is a type of disease with distinct epidemiology, anatomy, and histopathology characteristics. It often occurred in China and southeast Asia, with the incidence rates in a range of 15 and 50 per 100,000 in southern provinces of China [1]. Radiotherapy (RT) with or without chemotherapy is a mainstay in the treatment of NPC [2]. Given the proximity of temporal lobes to NPC, temporal lobe injury (TLI) is one of serious late complications that affects memory [3], neurocognitive functions [3], physical functions [4], emotion [4], language [3], and the quality of life in survivors [4]. In an era of 2D radiotherapy, the reported rate of TLI ranges from 3-40% [5][6][7]. Recent advancement of dose-delivering technologies such as intensity-modulated radiation treatment (IMRT) allows normal adjacent structures including temporal lobes to be spared from high-dose radiation, which indeed led to reduction in incidence of radiation-induced temporal lobe injury that varied from 4.6 to 16% [8][9][10].
In 1991, Emami first described the dose constraint of the brain. The dose ≥60 Gy delivered to 1/3 of whole brain volume was found to be predictive for ≥5%

ORIGINAL RESEARCH
Prognostic variables for temporal lobe injury after intensity modulated-radiotherapy of nasopharyngeal carcinoma incidence at 5 years (TD 5/5) [11]. Quantitative Analysis of Normal Tissue Effects in the Clinic (QUANTEC) reported the relationship between the radiation dose and incidence of TLI. Based on 2D radiotherapy, for fractionated RT with a fraction size of <2.5 Gy, an incidence of radiation injury of 5 and 10% is predicted to occur at a biologically effective dose of 120 Gy (range, 100-140) and 150 Gy (range, 140-170), respectively. For twice-daily fractionation, a steep increase in toxicity appears to occur when the biologically effective dose is >80 Gy. For daily large fraction sizes (≥2.5 Gy), the incidence and severity of toxicity is unpredictable [12]. Lee previously reported that incidence rate of TLI was estimated to be 5% at 10 years when temporal lobe received 64 Gy (equivalent dose) based on 2D radiation therapy [6]. Of important notice, is that these data were created from conventional radiotherapy, and some heterogeneous factors such as different target volumes, endpoints, sample sizes, and irradiated brain regions would affect the clinical outcome.
However, in the modern era of IMRT, the reports about the TLI induced by IMRT or chemotherapy-IMRT were relatively limited. Su and coworkers reported that a total of 4.6% patients were diagnosed with TLI among 870 NPC at a median follow-up of 40 months. Of interest, TLI was not observed in T1-2 patients, the incidence rates were 3.1 and 13.4% in T3 and T4 patients, respectively. The Dmax and D1cc in injured temporal lobes (TLs) are greater than that in normal TLs. The 5-year incidence of TLI in patients with Dmax 64-68 Gy or D1cc 52-58 Gy is reported to be <5% based on this relatively short follow-up [13].
Apparently, previously recommended dose constraints for radiation-induced brain injury from QUANTEC are not adequately applied to IMRT of NPC. The dose constraints for TLI are urgently needed for NPC patients treated with IMRT. In this report, we conducted a retrospective analysis of 695 NPC treated with IMRT to determine predictive factors for TLI.

Patient selection
A total of 695 NPC patients treated with definitive IMRT in our institution were retrospectively reviewed and analyzed to determine predictive factors for TLI from June 2004-June 2009. All patients had a pathology-proven NPC and had a local disease or locoregional disease (stage I-IVa + b) according to the UICC 7th TNM staging system, prior to definitive IMRT. Recurrence patients were excluded from our study. The basic characteristics such as age, gender, hypertension, diabetes, blood lipid, chemotherapy, target therapy and et al. were recorded carefully in Table 1.
The study was approved by the ethics committee of our institution.

Radiotherapy protocol and temporal lobe contour
Radiation planning was designed and optimized using the CORVUS 3.4-4.2 inverse treatment planning system. The IMRT plan was implemented through dynamic intensitymodulated coplanar arc irradiation using a multileaf 8 Gy/f), in 30-33 fractions. All the patients received radiation to the lymph node drainage areas in the lower neck using 60 Co split-field techniques or 6 MV X-ray split-beam techniques with a prescription dose of 46-50 Gy [14]. X-ray was used to evaluate the setup errors before treatment delivery. The patients were treated with one fraction daily over 5 days per week. The dose limits for each normal organ were followed according to the Radiation Therapy Oncology Group protocol 0225 (RTOG0225). The whole process of IMRT was carried out according to an institutional treatment protocol previously described [14]. The temporal lobe (TL) delineated in the original treatment plan were reviewed, and was found that original TL contours did not include the areas that overlapped the tumor target volumes such as CTV and PTV, due to an inherent limitation of the Corvus system. For purpose of this study, we re-delineated the entire TL and obtained the following dosimetric parameters: Dmax, Dmean, D0.1cc (the dose Gray to 0.1cc of the TL volume), D0.5cc, D1cc, D2cc, D3cc, D5cc, D10cc, D15cc, and D20cc.

Chemotherapy
Of the 695 patients, 163 patients received radiotherapy alone, while 532 received concurrent chemoradiotherapy with cisplatin 80 mg/m 2 every 3 weeks for 2-3 cycles. Among 532 patients, 86 patients received 2-3 cycles of neoadjuvant chemotherapy combined with concurrent chemoradiotherapy. 47 patients received concurrent chemoradiotherapy combined with 1 to 2 cycle of adjuvant chemotherapy. The neoadjuvant chemotherapy regimen was TPF, and the adjuvant chemotherapy regimen was cisplatin.

MRI protocol
MRI brain was performed using a 1.5-Tesla system (Signa CV/i; General Electric Healthcare, Chalfont St. Giles, United Kingdom), To diagnose TLI, our neuroradiologists examine the area from the suprasellar cistern to the inferior margin of the sternal end of the clavicle using a head-and-neck combined coil. T1-weighted fast spin-echo images in the axial, coronal, and sagittal planes (repetition time, 500-600 msec; echo time, 10-20 msec), and T2-weighted fast spin-echo MRI in the axial plane (repetition time, 4000-6000 msec; echo time, 95-110 msec) were obtained before injection of contrast material. After intravenous injection of gadopentetate dimeglumine (0.1 mmol/kg body weight Gd-DTPA, Magnevist; Bayer-Schering, Berlin, Germany), spin-echo T1-weighted axial and sagittal sequences and spin-echo T1-weighted fat-suppressed coronal sequences were performed sequentially, using similar parameters to before injection. The section thickness was 5 mm with a 1 mm inter-slice gap for the axial plane, and 6 mm with a 1 mm inter-slice gap for the coronal and sagittal planes. All brain MRI images were centrally reviewed by two neuroradiologists, Each MR was first independently reviewed. For those with any disagreement, a joint review was then performed with a consensus report as the final report for this study analysis. Diagnostic criteria for TLI were as follows: (1) white matter lesions, defined as areas of finger-like lesions of increased signal intensity on T2weighted images; (2) contrast-enhanced lesions, defined as lesions with or without necrosis on post-contrast T1weighted images with heterogeneous signal abnormalities on T2-weighted images; (3) cysts, round or oval welldefined lesions of very high signal intensity on T2-weighted images with a thin or imperceptible wall as previously reported [15]. Recurrence or metastasis involved in TL was excluded from this study analysis.

Follow-up
The duration of follow-up was calculated from the completion of IMRT to either the day of death or the day of last examination. All the patients were follow-up every 3 months in the first years, every 6 months in the second year and every 12 months in the following years. Nasopharyngeal mirror, endoscopic examination, detailed physical examination and MRI were performed during follow-up. Follow-up MRI of the nasopharynx and/or neck was performed every 6-12 months, or whenever tumor recurrence was suspected or neurologic signs or symptoms occurred. The latency of TLI was measured from the day of IMRT completion to the date of MRI diagnosis.
Statistical analysis SPSS 19.0 was used for statistical analysis. Actuarial survival rates were calculated using the Kaplan-Meier method. Student's t test was used for comparing the dose between TLI and non-TLI group. Receiver operating characteristic curve (ROC) was used for screening volume-dose parameters to predict TLI. Logistic multiple stepwise regression was used for predicting the TLI rate M. Feng et al. Radiation-Induced Temporal Lobe Injury of NPC from the physical and biological equivalent dose of the 11 volume-dose parameters. For prognostic factor of TLI, long-rank test was used for univariate analyses, and Cox proportional hazard model was used for multivariate analyses. TLI. The response variable was defined by classifying each temporal lobe depending on whether changes on MRI images. Dose and volume response curves were calculated with the nonlinear regression model using the logistic dose-response model as below [16]: In the equation, b stands for the regression coefficient, X 1 , X 2 …X m stand for the value of different volume-dose parameters. Both physical dose and biological equivalent dose (BED) of temporal lobe were established. BED was calculated by linear quadratic model The criterion for statistical significance was set at α = 0.05, and P values were determined from two-sided tests.

Dosimetric analysis of temporal lobes
In our study, the longest latency of TLI is 55 months, so the 59 patients with TLI all occurred within 60 months after IMRT. Of the 59 patients with TLI, 22 were (1) diagnosed with bilateral TLI, and these 44 injured temporal lobes were included in this study. The other 37 patients were unilateral TLI, and 26 of the 37 patients who developed unilateral TLI within 60 months after IMRT were followed up for at least 60 months, so these 37 injured temporal lobes were included, and the uninjured temporal lobes in these 26 patients were considered as normal lobes for dose analysis. The 377 of 636 patients without TLI were followed up for at least 60 months and underwent MR imaging examinations at least once during the first 60 months of follow-up to exclude TLI. Therefore, the actual incidence of TLI at 5 years is analyzed in 436 patients. Eighty-one injured temporal lobes and 780 normal temporal lobes were enrolled in dosimetric analysis.

Univariate and multivariate analysis of TLI
For the univariate analysis, T stage, chemotherapy, diabetes, physical dose of temporal lobe (D2cc), biological dose of temporal lobe (D2cc) and fraction size of prescription was the substantial predictive factors. For the multivariate analysis, only T stage, chemotherapy, D2cc of temporal lobe, and fraction size of prescription were independent risk factors for TLI (Table 4).

Discussion
Temporal lobe injury (TLI) is one of the most serious late effects after the definitive radiotherapy of NPC. Results from this study and other few studies [9,13,17] have demonstrated that incidence rates of TLI in patients treated with modern IMRT are indeed significantly reduced, in comparison with previous reports in the era of conventional radiotherapy [11,12]. Similar to the previous report [13], we observed TLI in patients treated with T3-T4    NPC, but not in patients with T1-2 NPC. However, the incidence of TLI was 8.49% in patients with T3-4 in our study at a median follow-up of 73 months, which was higher than 6.7% observed in a previous study at a median follow-up of 40 months [13]. This difference could be related to a relatively short follow-up time in the previous study [13] (median follow-up of 40 months vs. 73 months in this study), and in this study, we noticed a median time to diagnose TLI was 38 months. The above difference might explain a high incidence rate of TLI in patients with NPC in this study.
The precise mechanism that causes TLI remains unknown, but TLI is likely to be associated with the volume and dose of TL irradiated. Su found that IMRT with rV40 <10% or aV40 <5cc, and a Dmax < 68 Gy or D1cc < 58 Gy for the TL was less likely to develop TLI [13]. Among several volume-dose and volume-dose parameters, Sun reported only D0.5cc was predictive for TLI (D0.5cc < 69 Gy) [18]. In this study, we noticed the area under cures (AUC) of D2cc was the predictable factor for TLI among all the dosimetric parameters. Based on the dose-response analysis (Figure 3), we noticed the close correlation between incidence rate of TLI and D2cc of temporal lobe, which indicates the importance of AUC of D2cc to predict TLI. The D2cc < 60.31 Gy for TD5/5 and 76.85 Gy for TD50/5 thresholds should be considered as the constraint dose of TL in the IMRT of NPC. However, we are still not able to define specific TLI areas that are associated with radiation dose, even though we carefully review the CT dosimetry and MRI.
Radiation-induced injury of temporal lobe is a highly complex and multifactorial process. Radiation tolerance may vary depending on patient-and tumor-specific characteristics, as well as treatment modifications. Our multivariate analysis showed T stage, chemotherapy, D2cc of temporal lobe and fraction size of prescription were the independent risk factors for TLI. Similarly, Zeng reported the T category, chemotherapy and Dmax of temporal lobe were the significant factors affecting the risk of temporal lobe injury [18]. Furthermore, Lee [11] reported the 10-year actuarial incidence of TLN was 4.6% for patients irradiated to 60.0 Gy with 2.5 Gy per fraction, and up to 18.6% for those irradiated to 50.4 Gy with 4.2 Gy per fraction. In a prospective trial, 25 patients are treated by IMRT for a total of 70 Gy with 2.34 Gy per fraction to GTV, and TLI becomes one of the common late complications with a high incidence at 12% [19]. Our results showed that fraction size of prescription (>2 Gy) was the independent risk factors for TLI, which was consistent with the previous studies. Besides the volume-dose parameters of radiation, the age, chemotherapy, diabetes, and hypertension were reportedly predictive factors for TLI [20,21]. Results of our study confirmed that chemotherapy was indeed the prognostic factor for TLI, but others (diabetes, hypertension, alcohol and smoke, blood lipid, and target therapy) were not found to be independent risk factors for TLI in this study. In this study, 76.5% patients received chemotherapy.
Shortcoming of this study is retrospective nature. The follow-up time might not be enough for NPC, even though, to the best of our knowledge, this study has the longest median follow-up time (73 months) among the literature reports on IMRT of NPC. In addition, current MRI diagnosis of TLI is not completely defined, even though we have assigned two neuroradiologists to independently review each MRI. The NTCP model established in this study warrants further validation in a separate external validation cohort, and the D2cc e constraints based on this NTCP model should be treated with caution before use.
In a conclusion, results from this study have suggested that radiation dose of temporal lobe, T stage, fraction size of prescription, and chemotherapy were the independent predictive factors for TLI. D2cc was an important parameter for TLI. TD5/5 and TD50/5 of D2cc are 60.3 Gy and 76.9 Gy, these biological effective dose thresholds should be considered in IMRT planning of NPC.