Because most tumors involve changes in various systems during their development and progression. Therefore, the detection indicators in peripheral blood that can reflect the general condition have very important diagnostic and prognostic value for neoplastic diseases, such as circulating tumor DNA and circulating tumor cells 17,18. However, PTC is mainly concentrated in the local thyroid gland due to its low malignancy, and indicators currently considered to be directly related to the lesion, such as thyroglobulin, which is highly heterogeneous in different population studies and can only be used to indicate the condition of postoperative recurrence, but not for preoperative prediction19,20. Therefore, the preoperative severity of PTC is rarely effectively assessed by biomarkers or other parameters in peripheral blood.
Our study included common peripheral blood tests preoperatively, which have not been clearly demonstrated to have a clear association with the degree of progression of PTC, such as metastasis and invasion. Some studies have reached some preliminary conclusions by retrospective analysis, especially by constructing clinical prediction models together with ultrasound imaging indicators, or supervised learning models based on machine learning algorithms. After combining the results of preoperative thyroid antibodies or a serological marker, the predictive efficacy of ultrasound for LNM can be further improved21,22. However, the limitation of these studies lies in the relatively limited sample size on the one hand, and the very complex composition of peripheral blood on the other hand, and the combination of single or several indicators may not be able to measure the potential relationship between peripheral blood and thyroid tumors more comprehensively. In addition, researchers tend to pay more attention to the overall predictive power of diagnostic models in diagnostic tests, especially when performing supervised machine learning. The contribution of each predictor to the model is not particularly interpretable23. For example, image-omics studies rely more on the power of the improved algorithm itself for disease prediction than on the clinical significance of each independent variant6.
In this study, we included a large number of multidimensional peripheral blood tests. PTC patients were mainly classified according to these peripheral blood parameters by means of unsupervised clustering. Because pathologic conditions of whether PTC developed metastasis and invasion were not included in the classification process, and unsupervised algorithms only emphasized differences between data rather than convergencing24. Therefore, when our results indicate that patients after clustering do have differences in PTC invasion and metastasis between different clusters, it is even more likely to illustrate the potential association of these indicators with the progression of PTC when combined. Among them, laboratory tests in peripheral blood such as blood counts, biochemical parameters, thyroid function, and coagulation function all contribute significantly to the clustering results of high-risk populations for PTC metastasis and invasion.
Among the biochemical parameters, the most relevant ones were ALT and AST, which mainly reflected the patient 's liver function. Present results suggest that nonalcoholic fatty liver disease is associated with a risk of thyroid cancer and that this risk increases with increasing ALT levels, in line with our findings 25. However, in another study, the increased ratio of AST and ALT was found to be elevated in patients with thyroid cancer26.Different studies have shown the complexity of the two transaminases for the development and progression of PTC. Overall, the primary etiology of nonalcoholic fatty liver disease is closely related to metabolism, and previous studies have shown that the metabolomics of thyroid cancer also has a unique mechanism of change27, reflecting the complexity of the process of tumor development, therefore, changes in AST and ALT may also indirectly reflect the potential burden of thyroid cancer development on liver metabolism. Blood count parameters are also very complex for metastasis and invasion of PTC, and usually progression of cancer leads to the development of anemia, so for most cancers, low hemoglobin is a risk factor for progression28.This is contrary to our findings. We suggest that this phenomenon may occur in association with angiogenesis during tumor metastasis and invasion, as well as causing a compensatory increase in systemic red blood cells by a hypoxic state in the local microcirculation29. Neutrophils are considered protective against invasion and metastasis of PTC in our study, which could reflect that elevated levels of neutrophils are more beneficial in inhibiting cancer progression in the absence of co-infection, as also confirmed by related studies30, 31. However, the increase in total leukocyte count still predicts a poor prognosis in PTC, indicating that inflammatory cells in peripheral blood may still reflect a pro-tumor trend as a whole. In addition, thrombin time and FT3 levels were both significant adverse factors for PTC progression. What is known is that adverse effects on coagulation function emerge as a chronic inflammatory state that adversely impacts tumor prognosis, which has been demonstrated in other studies32. Although hyperthyroidism itself is a protective factor for the development of thyroid cancer, low TSH is not conducive to the growth of thyroid cancer cells. However, regardless of TSH, a cohort study has similarly shown that high levels of thyroid hormones are also associated with thyroid cancer incidence33. We suggest that increased thyroid hormone synthesis may also reflect active physiological function of the thyroid gland and also predict active cell division and proliferation, so it exists as a risk factor for adverse pathological features of PTC in our study.
To explore biomarkers that are more biologically meaningful and intuitive for PTC. Methods for high-throughput sequencing have also been used in different studies. Although the samples examined in these studies were mainly postoperative pathological tissues of the thyroid gland, or thyroid tissue obtained by fine needle aspiration cytology before surgery, rather than peripheral blood. However, sequencing results can also be reflected in indicators in peripheral blood by the association of the gene co-expression level behind them, and novel sequencing-based biomarkers, such as non-coding RNAs 34,35, have also been developed for peripheral blood of PTC patients. In line with the conclusions of our study, these similar markers, such as MiR-221 and MiR-146, were also detected to be specific in patients with chronic liver injury36, inflammatory disease37, and thrombus-related disease38. Thus, our study also provides a rationale for exploring potential common pathogenic mechanisms between PTC and other diseases. Certainly, because this study was a retrospective study, bias in cohort selection remains. To further demonstrate the reliability of our findings, more rigorous prospective studies are still needed to validate our results in the future.