Discussion
HPSCC is one of the most common head and neck malignant tumours. Advanced HPSCC often results in death with local recurrence and distant metastasis within five years after treatment. Despite medical advances, the 5-year survival rate remains at approximately 40%(1). Some studies suggested that 80% of patients with HPSCC who underwent surgery at the initial diagnosis had pathology suggestive of neck metastases upon examination(15, 16). The local lymph node metastasis can increase the recurrence and mortality of HNSCC. Hence, the metastatic load of lymph nodes is one of the most important prognostic indicators of HNSCC(17-19).
The AJCC/ UICC tumor, node, metastasis (TNM) staging system is widely used for HPSCC, in which pN is described for the size, location, laterality, number, and extra-envelope invasion (ENE) of positive lymph nodes. However, it is vulnerable to the total number of surgically removed lymph nodes, for which we need improvement on the predictive risk stratification model(20). LNR, the ratio of MLN to HLN, is a surrogate mathematical marker affected by several factors, including the extent of cervical lymphatic dissection, the surgeon’s cervical dissection technique, and the quality of pathologic evaluation(21). Multiple studies have validated the value of LNR in various tumors, including HPSCC, for predicting prognosis. However, a major limitation of LNR is that there is no corresponding discriminatory power for the ratio of 0 and 1. The predictive value of LODDS has been validated in various tumors (10, 11, 13, 14). LODDS distinguishes patients without positive lymph nodes by adding 0.5 to both positive and negative lymph nodes. Its predictive value in stage III and IV HPSCC has not been tested yet.
Several clinical studies have verified the predictive value of LNR in HPSCC. Yu and colleagues conducted a study involving 279 patients with HPSCC, where the subjects received pretreatment (radiotherapy/chemotherapy) before neck dissection. The multifactorial analysis showed that LNR was an independent predictor of prognosis(22). Another study of 81 patients with HPSCC suggested a better prognosis for pN1 with LNR < 0.1 and pN2 patients(23). A meta-analysis of stage III and IV LSCC and HPSCC suggested that LNR was a better prognostic indicator than pN(6). Our study has selected 166 patients with HPSCC who underwent primary lesion resection with cervical lymph node dissection at initial diagnosis, with or without adjuvant therapy (radiotherapy/chemotherapy) after surgery. We set a cut-off value of 0.11 for LNR based on the patient’s DFS. Patients with LNR ≥0.11 had significantly poorer OS and DFS values (P<0.05).
Patients with LNR values of 0 and 1 can be further risk stratified by LODDS, which has been shown to have a better predictive value for tumor prognosis than pN and LNR in oral squamous cell carcinoma(13), esophageal cancer(11), breast cancer(14), gastric cancer(12), bladder cancer(10), and colorectal cancer(9). However, studies in LSCC(24) and cervical squamous carcinoma (25) suggested LODDS was predictive of progression but did not provide any improved predictive performance over LNR, which is coherent with our findings. In our study, LODDS was associated with both OS and DFS, and a LODD ≥0.91 was associated with poorer DFS (P<0.001) but not with OS. We believe this is caused by the dataset only having 5 patients with an LNR value of 1 and 12 patients having an LNR value of 0, which does not allow display the true advantage of LODDS over LNR.
In our study, for the first time, we selected patients with stage III-IV HPSCC and verified the predictive value of LODDS. We also evaluated the predictive value of pN, LNR, and LODDS for advanced HPSCC and found that pN was a stronger predictor than LNR and LODDS for OS, and LNR was stronger than pN and LODDS for DFS. In the multivariate analysis, LNR showed a greater predictive value for DFS than pT, pStage, surgical margins, SIRI, and LODDS. Meanwhile, LNR, pT, and surgical margins all had independent predictive values for OS. In contrast to other studies, our study suggested that HLN ≥15 did not improve DFS. These contradictive findings may be related to the fact that our research came from a single center and was retrospective. Our study is also limited by a small clinical sample size with potential selection and recall bias. We expect multi-center, prospective studies to be available in the future.