Predictive Model and Precaution for Oral Mucositis During Chemo-Radiotherapy in Nasopharyngeal Carcinoma Patients

Abstract

Purpose: To explore risk factors for severe acute oral mucositis of nasopharyngeal carcinoma (NPC) patients receiving chemo-radiotherapy, build predictive models and determine preventive measures.

Methods and materials: Two hundred and seventy NPC patients receiving radical chemo-radiotherapy were included. Oral mucosa structure was contoured by oral cavity contour (OCC) and mucosa surface contour (MSC) methods. Oral mucositis during treatment was prospectively evaluated and divided into severe mucositis group (grade ≥ 3) and non-severe mucositis group (grade < 3) according to RTOG Acute Reaction Scoring System. Nineteen clinical features and nineteen dosimetric parameters were included in analysis, least absolute shrinkage and selection operator (LASSO) logistic regression model was used to construct a risk score (RS) system.

Results: Two predictive models were built based on the two delineation methods. MSC based model is more simplified one, it includes body mass index (BMI) classification before radiation, retropharyngeal lymph node (RLN) area irradiation status and MSC V55%, RS = -1.480 + (0.021 × BMI classification before RT) + (0.126 × RLN irradiation) + (0.052 × MSC V55%). The cut-off of MSC based RS is -1.011, with an area under curve (AUC) of 0.737 (95%CI: 0.672-0.801), a specificity of 0.595 and a sensitivity of 0.786. OCC based model involved more variables, RS= -4.805+ (0.152 × BMI classification before RT) + (0.080 × RT Technique) + (0.097 × Concurrent Nimotuzumab) + (0.163 × RLN irradiation) + (0.028 × OCC V15%) + (0.120 × OCC V60%). The cut-off of OCC based RS is -0.950, with an AUC of 0.767 (95%CI: 0.702-0.831), a specificity of 0.602 and a sensitivity of 0.819. Analysis in testing set shown higher AUC of MSC based model than that of OCC based model (AUC: 0.782 vs 0.553). Analysis in entire set shown AUC in these two method-based models were close (AUC: 0.744 vs 0.717).

Conclusion: We constructed two risk score predictive models for severe oral mucositis based on clinical features and dosimetric parameters of nasopharyngeal carcinoma patients receiving chemo-radiotherapy. These models might help to discriminate high risk population in clinical practice that susceptible to severe oral mucositis and individualize treatment plan to prevent it.

Keywords: nasopharyngeal carcinoma; dosimetric parameter; preventive measures; radiation-induced oral mucositis; radiotherapy.

Copyright © 2020 Li, Li, Jin, Lin, Fang, Yang, Shen, Chen, Zhang, Chen, Chen and Chen.

Figures

Figure 1

(A) Box plots of dose-volume objectives distributions. Mean values are indicated by the horizontal lines within boxes and median values are represented by the diamonds. Severe oral mucositis group (grade ≥ 3) and non-severe oral mucositis group (grade = 1, 2) data were drawn as red forward diagonal and blue backward diagonal boxes respectively. Note: *Statistically significant at p=0.05 level. (B) Area under curve (AUC) of all the dose-volume objectives under both oral cavity contour (OCC) (red solid line) and mucosa surface contour (MSC) (blue dashed line) methods. Each AUC is acquired from the ROC curve of each objective. Most dose-volume objectives under MSC method show better performance (higher AUC) than OCC method in terms of predicting severe oral mucositis.

Figure 2

(A, F) LASSO coefficient profiles of 19 clinical features and 19 MSC and OCC based dosimetric parameters; (B, G) Ten-fold cross-validation for tuning parameter selection in MSC and OCC based LASSO model. (C–E, H–J) ROC curve for MSC and OCC based model: (C, H) Training group, (D, I) Testing group, (E, J) Entire group. The point on the curve is cutoff value for RS and the following bracket contains specificity and sensitivity. Abbreviation: RS, risk score; OCC, oral cavity contouring; MSC, mucosa surface contouring; AUC, area under curve; CI, confidence interval.

Figure 3

Patients distribution of V10% and V15% by using TOMO and IMRT. (A) MSC V10%, (B) MSC V15%, (C) OCC V10%, (D) OCC V15%. Abbreviations: OCC, oral cavity contouring; MSC, mucosa surface contouring.

Figure 4

Computed tomography (CT) scan of a nasopharyngeal carcinoma patient with mucosal surface contours (MSC) (up, area filled with blue) and oral cavity contours (OCC) (down, area filled with brown). MSC involves the mucosal surface while OCC encompass more solid tissue, like tongue, maxillary bone, etc. The green line and pink line are isodose curve of 50Gy and 55Gy respectively. Abbreviations: OCC, oral cavity contouring; MSC, mucosa surface contouring.