ETA2024 Poster Presentations Diagnosis of thyroid cancer-1 (10 abstracts)
1The First Affiliated Hospital of Sun Yat-Sen University, Clinical Trials Unit, Guangzhou, China; 2The First Affiliated Hospital of Sun Yat-Sen University, China
Objectives: Fine-needle aspiration biopsy (FNA) is the preferred preoperative pathological diagnostic method for thyroid nodules. Processing methods for FNA samples mainly include conventional smear (CS) and liquid-based preparation (LBP). There is still debate as to which method or their combination is better. This study aims to compare the diagnostic accuracy of the two methods and their combined usage.
Methods: We included thyroid cytopathology data from nine medical centers between 2013 and 2023. We identified a total of 35006 samples, of which 1049 samples used CS, 9677 samples used LBP, and 24280 samples used both CS and LBP. Among the samples utilizing both methods, 22472 had only a final cytological diagnosis based on two methods, while 1808 samples from prospective cohort had the diagnosis of CS, LBP and their combination, respectively. According to the Bethesda reporting system, we calculated the distribution of diagnostic categories for different methods and compared the diagnostic uncertainty (category III and IV). For category II, V, and VI samples with histological diagnosis, we compared the diagnostic accuracy of the different preparation methods. For the prospective cohort samples, we performed consistency analysis and compared the diagnostic performance of different preparation methods.
Results: Overall, the rate of diagnostic uncertainty (category III and IV) of CS was higher than that of LBP and combined CS and LBP (21.63% vs 14.01% vs 14.10%). 744 samples using CS, 4022 using LBP and 7270 using both methods under category II, V, or VI samples had definitive histological diagnosis. Their sensitivities (97.34% vs 98.84% vs 98.47%) and accuracies (95.83% vs 97.66% vs 95.82%) were comparable, while the specificity of LBP (79.34%) was higher than that of CS (60.00%) and their combination (67.58%). For the satisfactory samples (category II-VI) in the prospective cohort, the concordance rate between CS and LBP was 94.20%. The sensitivities (97.88% vs 96.79% vs 97.74%) and accuracies (96.97% vs 96.00% vs 96.89%) of three methods were comparable, while LBP showed higher specificity (75.00%) than CS (58.33%) and the combined methods (66.67%).
Overall data | Prospective cohort | |||||
CS n = 744 | LBP n = 4022 | CS&LBP n = 7270 | CS n = 531 | LBP n = 550 | CS&LBP n = 546 | |
Sensitivity | 97.34% | 98.84% | 98.47% | 97.88% | 96.79% | 97.74% |
Specificity | 60.00% | 79.34% | 67.58% | 58.33% | 75.00% | 66.67% |
Accuracy | 95.83% | 97.66% | 95.82% | 96.99% | 96.00% | 96.89% |
PPV | 98.30% | 98.68% | 97.01% | 99.03% | 99.03% | 99.05% |
NPV | 48.65% | 81.36% | 80.50% | 38.89% | 46.88% | 45.45% |
Conclusions: The diagnostic uncertainty rate of CS was higher than that of LBP and combined usage. The diagnostic accuracy of CS and LBP was comparable, and the combination of CS and LBP did not improve the diagnostic accuracy. Therefore, it is not suggested to combine CS and LBP in routine cytological diagnosis.