Radiation-induced late dysphagia after intensity-modulated radiotherapy in nasopharyngeal carcinoma patients: a dose-volume effect analysis

Dysphagia is a side effect of nasopharyngeal carcinoma chemo-radiotherapy (CRT) which greatly influences the quality of life of the patients. We analyzed late dysphagia in 134 patients with nasopharyngeal cancer undergoing radical radiotherapy (RT), and correlated these findings with dose–volume histogram (DVH) parameters of the swallowing organs at risk (SWOARs). DVH parameters of SWOARs were correlated with late dysphagia, and with RTOG/EORTC scale score and the M. D. Anderson dysphagia inventory (MDADI) score. The mean dose (Dmean) to the superior and inferior constrictor muscles (SCM and ICM) and age were associated with grade 2 late dysphagia. Receiver operating characteristic (ROC) curves showed that the threshold values for grade 2 late dysphagia were: Dmean to SCM ≥ 67 Gy, partial volume receiving specified dose of 60 Gy (V60) of SCM ≥ 95%, Dmean to ICM ≥ 47 Gy, and V50 of ICM ≥ 23%. The areas under the ROC curve were 0.681 (p = 0.02), 0.677 (p = 0.002), 0.71 (p < 0.001) and 0.726 (p < 0.001) respectively. Our study demonstrates a significant relationship between late dysphagia and the radiation doses delivered to the SCM and ICM. Our findings suggest that physicians should be cautious in reducing the RT dose to SWOARs in order to avoid severe dysphagia. Further prospective trials are necessary to recommend this as part of routine clinical practice.

Scientific RepoRTs | (2018) 8:16396 | DOI: 10.1038/s41598-018-34803-y dose-volume parameters and the observed incidence of late dysphagia in these patients. After a re-contouring of the SWOARs according to recently published guidelines, we determined the dose tolerated by the SWOARs that achieved the highest uncomplicated tumor control.

Materials and Methods
Patient population. A total of 158 patients with NPC who received IMRT at the First Affiliated Hospital of Guangxi Medical University from March 2013 to April 2014 were initially enrolled. The inclusion criteria were 1) treatment with curative IMRT at a dose delivered to the gross tumor volume (PGTVnx) of at least 66 Gy either alone or in combination with concomitant chemotherapy, 2) the availability of treatment plan record with DVH parameters, and 3) willingness to complete the stipulated questionnaires. The exclusion criteria were: 1) persistent/recurrent tumor, 2) distant metastasis, 3) previous radiotherapy for another head and neck tumor or with palliative intent, 4) Any RTOG/EORTC grade swallowing dysfunction before treatment and tumor invasion in oropharynx and hypopharynx, and 5) hoarseness, nasal regurgitation, lingual deviation and atrophy, coughing while drinking, unclear enunciation etc. The newly diagnosed NPC patients who suffered from dysphagia before treatment were still excluded so as to ensure that the observed dysphagia was only induced by radiation-related SWOARs dysfunction but not by lower cranial neuropathy. Based on these criteria, 24 patients were excluded: 3 were lost during the follow-up, 15 underwent a second treatment (re-irradiation and/or chemotherapy) due to either distant metastasis (n = 9) or loco-regional relapse (n = 6), and 6 were excluded because they had swallowing dysfunction before treatment. The remaining 134 patients received the questionnaires and provided informed consent. All methods were in accordance with the relevant guidelines and regulations, and were approved by the Ethical Review Committee of the First Affiliated Hospital of Guangxi Medical University. All clinical information of the participants is available and can be accessed.
Outline of swallowing structures. Based on published studies 9-12 , the SWOARs comprise of the following five muscles: the superior constrictor muscle (SCM), middle constrictor muscle (MCM) and inferior constrictor muscle (ICM) which are part of the pharyngeal constrictor muscle (PCM), the cricopharyngeus muscle (CPM), and the esophagus inlet muscle (EIM). The SCM, MCM and ICM form the posterior and lateral pharyngeal walls. The SCM extends from the caudal tip of the pterygoid plate to the lower edge of second cervical vertebra, MCM extends from the upper edge of third cervical vertebra to the lower edge of the hyoid bone, and ICM extends from below the lower edge of the hyoid bone to the lower edge of the arytenoid cartilage. CPM extended from below the lower edge of the arytenoid cartilage to the lower edge of the cricoid cartilage, and EIM consisted of the 1 cm of the muscular compartment of the esophagus inlet. Anatomical borders of each SWOAR are indicated in Table 1, and further delineated in Fig. 1.

Assessment of dysphagia.
All patients were followed-up to assess late dysphagia using the RTOG/EORTC scale 13 . Patient-reported clinical swallowing function was also assessed by the M.D. Anderson dysphagia inventory (MDADI) scoring, which consists of 20 questions with global, emotional, functional and physical subscales. In the MDADI questionnaire, a higher score is equated to a better quality of life and less swallowing trouble 14 .
Statistical analyses. Age, duration of follow up and dosimetric variables (D mean and V n ) were analyzed. Data conforming to normal distribution were compared using t-test, and data with non-normal distribution using the Wilcoxon Rank-Sum test. The χ 2 or Fisher's test were used to compare the grade 0-1 and grade 2 late dysphagia groups.
The clinical and dosimetric variables were assessed in the univariate analysis, and were incorporated into a binary logistic regression model to assess their independent contribution. Prior to the multivariate analysis, a correlation matrix was produced to identify those potential prognostic factors with high correlations, in particular between the DVH parameters. Based on Pearson correlation coefficients P ≥ 0.70 or variance inflation factor (VIF) ≥ 10 between the candidate prognostic factors, only one variable was selected and incorporated into the model to avoid multicollinearity, which may have negatively affected the generality of the model. The dosimetric variables that resulted from the independent factors of multivariate analysis and their corresponding significant DVH parameters in the univariate analysis were used for receiver operating characteristic (ROC) curve analysis, in order to calculate their threshold values of late dysphagia and assess their diagnostic capability. The dosimetric variables, which showed a correlation in the univariate analysis, were subjected to principal component analysis (PCA). Pearson correlation coefficients (r) were used to analyze the association between MDADI results and all demographic parameters, dosimetric variables and the degree of dysphagia (RTOG/EORTC scale). Statistical analysis was performed using the STATISTICA 22.0 software. A p-value of ≤ 0.05 was considered statistically significant. The figures were generated using GRAPHPAD 5.0. Table 2 summarized the clinical characteristics of 134 patients diagnosed with NPC. The median age of the patients was 44 years (range 18-71 years). During the median follow-up of 34 months (range 25-44 months), 71 patients (53%) were reportedly suffering from late dysphagia. Late dysphagia, according to the RTOG/EORTC scale, was scored as grade (G) 0 in 63 (47%) patients, G1 in 37 (28%) patients and G2 in 34 patients (25%). No cases of G3-4 toxicity were found. We then divided the patients into Group 1 (grade 0-1) and Group 2 (grade 2) based on the severity scores of dysphagia, and found that age and gender were related to the grade level of late dysphagia in the univariate analysis (p < 0.05).

Patents characteristics and late toxicity.
Association between the dosimetric parameters of SWOARs and late dysphagia. The correlation between D mean , V 50 , and V 60 to the affected constrictor muscles and the severity of the late dysphagia is shown in Table 3. The respective D mean to the PCM, SCM, MCM and ICM were each associated with G2 late dysphagia (p ≤ 0.001). Similarly, the respective V 60 of the SCM and MCM, and V 50 of the MCM and ICM were correlated with G2 late dysphagia (p ≤ 0.01). Multivariate analysis by forward elimination of insignificant explanatory variables was performed to adjust for various factors. Due to the high correlation between the D mean and the V 50 /V 60 of the constrictor muscles (Pearson coefficient > 0.8, p < 0.001), we used the D means to SCM and ICM for multivariate analysis, with age and gender as co-variants. Multivariate analysis showed that age (OR 1.050, 95%CI 1.005-1.098, p = 0.031), D mean to SCM (OR 1.170, 95%CI 1.018-1.344, p = 0.027) and D mean to ICM (OR 1.251, 95%CI 1.074-1.457, p = 0.004) were the independent predictors (Table 4). Finally, we evaluated the dose tolerance of the grade 2 late dysphagia using ROC curves in terms of the above significant independent variables. Due to the strong correlation between the V 50 /V 60 and D mean , V 50 of ICM and V 60 of SCM were also evaluated using ROC curve analysis. The significant dosimetric parameters and cut-off points of the ROC curve analysis are shown in Table 5. D mean to SCM ≥ 67 Gy, D mean to ICM ≥ 47 Gy, V 60 of SCM ≥ 95% and V 50 of ICM ≥ 23% were the threshold values of grade 2 late dysphagia ( Supplementary Figures 1-4). The areas under the ROC curves, as showed in Fig. 2, were 0.681 for D mean to SCM (sensitivity 0.647, specificity 0.65, p = 0.002) and 0.71 for D mean to ICM (sensitivity 0.676, specificity 0.67, p < 0.001). In the same way, the areas under the ROC curves were 0.677 for V 60 of SCM (sensitivity 0.735, specificity 0.6, p = 0.002) and 0.726 for V 50 of ICM (sensitivity 0.765, specificity 0.63, p < 0.001). In the principal component analysis, KMO and Bartlett tests showed correlations between the included variables (p < 0.001). The scree plot suggested three principle components as the optimal number which could explain 75.3% of the variation. Subsequently, a rotating element matrix was used to identify the dosimetric variables that had different attributes (Table 6). D mean to ICM, V 50 of ICM and V 60 of ICM had the most distinguishable contributions to the first principal component. Furthermore, the reliability and validity were examined by MDADI as showed in Table 7. The CPM and EIM were non-significant factors in relation to the grade 2 late dysphagia.

Discussion
Although IMRT with concomitant chemotherapy for NPC has a high rate of local control, radiation-induced late dysphagia is usually debilitating to patients. Furthermore, late dysphagia in NPC patients treated with IMRT is poorly understood. It is essential to know the dose tolerance for dysphagia in order to predict the safety of treatment plans. The aim of this retrospective study was to improve the understanding of late dysphagia and thus optimize IMRT for NPC. The effect of the radiation dose and volume on the healthy tissues is a major concern in radiotherapy. Patterson et al. evaluated the swallowing function of 18 NPC patients using a patient-reported oral function score with fiberoptic endoscopic examination; 5 patients suffered from moderate late dysphagia but no severe cases were observed 4 . Another study by Xu et al. showed that 50% of NPC patients treated with IMRT had swallowing dysfunction 15 . Our results, which are consistent with the previous studies, showed that 53% of the patients experienced late dysphagia and grade 2 late dysphagia occurred in 25% of the patients. Therefore, treatment-related late toxicity may worsen patients' quality of life.
Some recent studies have assessed the relationship between the dose-volume parameters of swallowing structures and late dysphagia in patients undergoing treatment for head and neck tumors 9,16,17 . Deantonio et al. reported that >50 Gy D mean to SCM and MCM correlated with grade 2-3 late dysphagia, and D mean to MCM is the only significant predictor of late grade dysphagia 18 . In the prospective study of Feng et al., patients with adequate aspiration received >60 Gy D mean to PCM 19 . Furthermore, studies have also analyzed the incidence     IMRT is widely used for patients with NPC, and delivers a high radiation dose to tumors while maintaining a safe dose for normal tissues surrounding the tumor. This technique also exhibits excellent tumor coverage. Compared to three-dimensional conformational radiation therapy, IMRT in head and neck reduces adverse effects such as dysphagia and thus improves quality of life (QOL) 26,27 . Currently, despite expert recommendations to spare a portion of the SWOARs in order to reduce dysphagia, the dose constraint to PCM is unclear 26,28 . Studies have shown different results, possibly due to methodological differences and the ambiguous contouring    of the SWOARs. Limiting the radiation dose to the crucial SWOARs is expected to decrease the incidence and severity of radiation-induced dysphagia with IMRT. A study used the new technique of swallowing sparing IMRT (SW-IMRT), and reduced the doses to the SWOARs based on the following criteria listed in order of priority: (1) minimizing the mean dose to the SCM, (2) minimizing the mean dose to the MCM, (3) minimizing the mean dose to the supraglottic larynx, and (4) minimizing the proportion of the EIM receiving ≥60 Gy (EIM V 60 ). Compared to the standard IMRT (ST-IMRT), SW-IMRT reduced the mean dose to the various SWOARs, along with a 9% mean reduction (3%-20%) in predicted physician-rated RTOG/EORTC grade 2-4 swallowing dysfunction 27 . The results of our study must be viewed cautiously because of several limitations. Due to the retrospective approach, we used the previously described dose tolerance in clinical practice in a prospective manner till they were validated. The second limitation is regarding the method of delineation of the swallowing structures that, although performed by an experienced radiation oncologist, are not fully standardized and could thus result in a bias. Other limitations are the small sample size, lack of anatomical examinations such as endoscopy, MRI, and barium meal test.

Conclusion
We found a significant relationship between late dysphagia and the radiation dose delivered to superior and inferior pharyngeal constrictor muscles during radiotherapy of NPC. The D mean to SCM ≥ 67 Gy, V 60 of SCM ≥ 95%, D mean to ICM ≥ 47 Gy, and V 50 of ICM ≥ 23% correlated with grade 2 late dysphagia. The optimal modality to assess late dysphagia and the most appropriate dose limits of constrictor muscles are still open issues and should be further investigated. A prospective study is needed to validate the findings of the present study, and to determine whether the dosimetric benefits of this treatment strategy translate into better clinical outcomes.