[Clinical staging of nasopharyngeal carcinoma based on MRI: suggestions for improving the Chinese '92 staging system].

Background & Objective: Since the introduction of the Chinese '92 staging system of nasopharyngeal carcinoma (NPC), the diagnostic technology and therapeutic modality for NPC have been advanced obviously over the past 15 years. This study was to evaluate the staging parameters for NPC in a large cohort based on modern diagnostic and therapeutic modality to provide suggestions for improving the Chinese '92 staging system.
Methods: Between Jan. 2003 and Dec. 2004, 924 consecutive patients with newly diagnosed, nondisseminated biopsy-proven NPC, treated at Cancer Center of Sun Yat-sen University, were enrolled. All patients received magnetic resonance imaging (MRI) scan of the neck and nasopharynx before treatment. Induction was applied to evaluate the correlations among different T parameters. Cox regression model was used to investigate the prognostic values of different N parameters. According to the principle of the staging system, the indices of hazard consistency, hazard discrimination, prognostic value, and distribution were used to evaluate the proposed staging system.
Results: According to the Chinese '92 T classification, all cases of involvement of the prevertebra muscle, soft palate, pterygopalatine fossa, and orbit were incorporated with erosion of other parameters that belonged to the same or more advanced subgroup; 282 (91.3%) of the 309 cases of carotid sheath involvement were incorporated with erosion of other T3 parameters; 64 (85.3%) of the 75 cases of single anterior or posterior group of cranial nerve involvement were incorporated with erosion of other T4 parameters. The hazard ratios (HR) of local relapse for T3 stage with carotid sheath involvement [HR=1.635, 95% confidence interval (CI): 0.987-2.764] and T2 stage (HR=1.524, 95% CI: 0.910-2.368) were similar. The hazard ratios of local relapse for T3 stage with single site of skull base erosion (HR=3.567, 95% CI: 1.398-11.278), T3 stage excluding single site of skull base erosion (HR=3.891, 95% CI: 1.449-10.449), and T4 stage with involvement of the sphenoid sinus solely (HR=3.613, 95% CI: 1.437-11.854) were similar. The hazard ratios of local relapse for T3 stage with involvement of either anterior or posterior cranial nerves solely (HR=5.849, 95% CI: 2.069-14.500) and T4 stage excluding involvement of the sphenoid sinus (HR=6.618, 95% CI: 2.499-17.525) were similar. Multivariate analysis showed that lymph node metastasis level and laterality were independent predictors for distant metastasis. Therefore, according to the principle of concise, the parameters, including involvement of the prevertebra muslce, soft palate, pterygopalatine fossa and orbit, were deleted. According to the principle of hazard consistency, the involvement of the parapharyngeal spaces, including prestyloid space and poststyloid space, were defined as T2 stage, the involvement of the skull base, including pterygoid process, were defined as T3 stage, the involvement of the sphenoid sinus were defined as T3 stage, and the involvement of the cranial nerves were defined as T4 stage. N staging was optimized by incorporating level and laterality as staging criteria.
Conclusion: Based on MRI, the proposed T classification, N classification and clinical staging of NPC are reasonable according to the principle of hazard consistency, hazard discrimination, prognostic value and distribution, and should be recommended for clinical use.