[Prediction of papillary thyroid cancer recurrence according to preoperative data].

Abstract

Objective: To create a formalized method for predicting papillary thyroid cancer recurrence after hemithyroidectomy based on preoperative data.

Material and methods: At this stage of the study, we selected 101 patients with papillary thyroid cancer who underwent surgical treatment in 2017-2023. Recurrence was observed in in 47 patients. Fifty-four patients had no recurrence within 5 years after surgical treatment, i.e. these patients underwent surgery in 2017-2018. To find prediction rules, we used original classification method based on searching for subsets of variables and piecewise linear rules separating classes in pairs with subsequent voting of such rules to make a decision.

Results: The exam was carried out using a training sample (101 cases) and sliding control method (10 tests on 10 random cases). On the training sample, sensitivity of predictive algorithm was 91%, specificity 78% and error rate 13%. The aggregated result of 10 trials using sliding control method revealed sensitivity of predictive algorithm 86%, specificity 75% and error rate 15%. This result is close to overall sample and confirms the effectiveness of this method for predicting recurrence.

Conclusion: The pilot experiments revealed the patterns in data for potential prediction of recurrence based on preoperative indicators. Further study of this problem may be valuable for decision-making and adjustments in the management of patients with papillary thyroid cancer.