Long-Term Evaluation and Normal Tissue Complication Probability (NTCP) Models for Predicting Radiation-Induced Optic Neuropathy after Intensity-Modulated Radiation Therapy (IMRT) for Nasopharyngeal Carcinoma: A Large Retrospective Study in China

Abstract

Purpose: To quantify the long-term evaluation of optic chiasma (OC) and/or optic nerve(s) (ONs) and to develop predictive models for radiation-induced optic neuropathy (RION) in nasopharyngeal carcinoma after intensity-modulated radiotherapy (IMRT).

Methods and materials: A total of 3,662 patients' OC/ONs with full visual acuity and dosimetry data between 2010 and 2015 were identified. Critical dosimetry predictors of RION were chosen by machine learning and penalized regression for survival. A nomogram containing dosimetry and clinical variables was generated for predicting RION-free survival.

Results: The median follow-up was 71.79 (2.63-120.9) months. Sixty-six eyes in 51 patients (1.39%) developed RION. Two patients were visual field deficient, and 49 patients had visual acuity of less than 0.1 (20/200). The median latency time was 36 (3-90) months. The 3-, 5-, and 8-year cumulative incidence of RION was 0.78%, 1.19%, and 1.97%, respectively. Dmax was the most critical dosimetry variable for RION (AUC: 0.9434, the optimal cutoff: 64.48 Gy). Patients with a Dmax ≥64.48 Gy had a significantly higher risk of RION (HR = 102.25; 95%CI, 24.86-420.59; P < 0.001). Age (>44 years) (HR = 2.234, 95% CI = 1.233-4.051, p = 0.008), advanced T stage (T3 vs. T1-2: HR = 7.516, 95% CI = 1.725-32.767, p=0.007; T4 vs. T1-2: HR = 37.189, 95% CI = 8.796-157.266, P < 0.001), and tumor infiltration/compression of the OC/ONs (HR = 4.572, 95% CI = 1.316-15.874, p=0.017) were significant clinical risk factors of RION. A nomogram comprising age, T stage, tumor infiltration/compression of the OC/ON, and Dmax significantly outperformed the model, with only Dmax predicting RION (C-index: 0.916 vs. 0.880, P < 0.001 in the training set; 0.899 vs. 0.874, P=0.038 in the test set). The nomogram-defined high-risk group had a worse 8-year RION-free survival.

Conclusions: In the IMRT era, Dmax <60 Gy is safe and represents an acceptable dose constraint for most NPC patients receiving IMRT. A reasonable trade-off for selected patients with unsatisfactory tumor coverage due to proximity to the optic apparatus would be Dmax <65 Gy. Caution should be exercised when treating elderly and advanced T-stage patients or those with tumor infiltration/compression of the OC/ON. Our nomogram shows strong efficacy in predicting RION.

Figure 1

Actual cumulative risk of RION over time (a) stratified by radiation dosage (b). RION-free survival probability curves of the low-risk and high-risk groups stratified by the nomogram in the training (c) and test sets (d). RION = radiation-induced optic neuropathy.

Figure 2

The time-dependent receiver operating characteristic (ROC) curve (a) and dose-effect curves (b) of Dmax for RION. Dmax had the highest AUC value at 8 years with 0.9434. The optimal cutoff of Dmax was 64.48 Gy (sensitivity = 0.955; specificity = 0.814). Solid and dashed lines indicate logistic regression and the 95% CI dose tolerance, respectively. RION = radiation-induced optic neuropathy; Dmax = maximum point dose; AUC = area under the receiver operating characteristic curve; CI = confidence intervals.

Figure 3

A nomogram for predicting 3-, 5-, and 8-year RION-free survival (a); calibration curves of the nomogram (b) in the training and (c) the test sets. RION = radiation-induced optic neuropathy; Dmax = maximum point dose.