Model predicts lymph metastasis in oral cancer
MedWire News: Korean researchers report that a recently developed mathematical model can accurately predict the presence of lymph node metastasis in presurgical patients with oral cancer.
Determining the likelihood of such metastases in every preoperative oral cancer patient "may help in deciding the need for surgical lymph node dissection or additional preoperative treatment modalities that might improve survival," explain Ki-Yeol Kim and In-Ho Cha, from Yonsei University in Seoul.
Reporting in the journal Oral Oncology, the researchers created the model using a mathematical method known as principal component analysis (PCA).
They retrospectively assessed 90 patients with primary oral squamous cell carcinoma, who had already undergone surgical ablation and neck dissection, for four preoperative variables. These variables were: clinical tumor size (T), stage of the tumor, clinical nodal (N) stage, insulin-like growth factor II mRNA-binding protein 3 (IMP3), and cyclooxygenase-2 (COX-2).
All patients were randomly assigned to one of two groups: a training set (n=60) and a test set (n=30), and PCA was used among the training set to estimate how much impact each of the four variables had on the risk for the presence of lymph node metastasis.
In all, 31% of the whole cohort had dissected lymph nodes that were positive for cancer/tumor metastasis, with a similar rate among the training and test sets, at 68% and 70%, respectively.
When Kim and Cha closely analyzed the impact of each variable on the presence of oral cancer metastasis to lymph nodes, they found that clinical N stage had the strongest association with histological lymph node positivity.
Indeed, the majority (91.9%) of patients with clinical N stage 0 were histologically negative for lymph node metastases, whereas 53.6% of those with clinical N stage 2 were histologically positive for lymph node metastases.
Clinical T stage showed association with lymph node positivity, although it was not as strong a predictor of positivity as clinical N stage. IMP3 showed little association with lymph node status and COX-2 was not related to lymph node status at all.
Based on these findings, the researchers devised a predictive model using clinical N and T stage only. However, this model had a predictive accuracy of 53% - a figure that Kim and Cha believed could be improved. They therefore refined the model by including IMP3 and COX-2 as variables, and found that the accuracy of the model increased to 84%.
This finding demonstrates that these variables hold more predictive weight when used in combination compared with solitary use, says the pair.
Kim and Cha conclude that further study involving larger cohorts and different prediction models/clinical variables is required to validate the findings of the current study.
By Lauretta Ihonor